Modern Ion Beam Materials Analyisis

Handbook of Modern
Ion Beam Materials Analysis

Second Edition
Handbook of Modern
Ion Beam Materials Analysis
Second Edition
EDITORS:
Y ongqiang Wang and Michael Nastasi
IMIRlsl
Materials Research Society
Warrendale, Pennsylvania
Single article reprints from this publication are available through

University Microfilms Inc., 300 North Zeeb Road, Ann Arbor, Michigan 48106
CODEN: MRSPDH
Copyright 2009 by Materials Research Society.
All rights reserved.
This book has been registered with Copyright Clearance Center, Inc. For further information, please
contact the Copyright Clearance Center, Salem, Massachusetts.
Published by:
Materials Research Society
506 Keystone Drive
Warrendale, PA 15086
Telephone (724) 779-3003
Fax (724) 779-8313
Web site: http://www.mrs.orgl
Library of Congress Cataloging-in-Publication Data
Handbook of modem ion beam materials analysis I editors, Yongqiang Wang and Michael A Nastasi. -2nded.
p.cm.
Includes bibliographical references and index.
ISBN 978-1-60511-215-2 (alk. paper)
1. Ion bombardrnent--Handbooks, manuals, etc. 2. Solids-Effect of radiation onHandbooks, manuals, etc. 3. Materials-Analysis-Handbooks, manuals, etc. I. Wang,
Yongqiang. II. Nastasi, Michael Anthony
QC702. 7.B65H36 2009
620.1'127--dc22
2009045714
Manufactured in the United States of America
Disclaimer of Liability
Neither the editors, authors, contributors, their affiliated institutions, nor the
Materials Research Society makes any warranty or representation, express or
implied, including the warranties or merchantability and fitness for a particular
purpose, or assumes any legal liability or responsibility for the accuracy,
completeness, or usefulness of any information, apparatus, product, or process
disclosed, or represents that its use would not infringe privately owned rights.
CONTENTS
ACKNOWLEDGEMENTS ...................................................................................... iv
CHAPTERS
1.
Introduction ........................................................................................................ 1
2.
Energy Loss and Energy Straggling ................................................................. 5
3.
The Interaction of Charged Particles with Nuclei ......................................... 27
4.
Backscattering Spectrometry .......................................................................... 43
5.
Elastic Recoil Detection: ERD ........................................................................ 81
6.
Nuclear Reaction Analysis: Particle-Particle Reactions ............................. 125
7.
P.rticle-Induced G.mm. Emission: PICE ................................................. 147
8.
Nucle.r Reactions for Hydrogen Analysis ................................................... 175
9.
Hydrogen Analysis by Proton .. Proton Scattering ......................................... 187
10.
Ch.rged p.rticle Activation Analysis .......................................................... 207
11. Particle-Induced X-Ray Emission: PTXE ...................................................... 231

12. Ion Channeling ..................................~ ............................................................ 247
13. Nuclear Microprobe Analysis ......................................................................... 285
14. Data Analysis Software for Ion Beam Analysis ............................................... 305
15. Pitfall'. in Ion Beam Analysis ........................................................................ 347
16. Instrumentation .nd L.bor.tory Pr.ctice ................................................... 385
17.
Radiological Safety ......................................................................................... 425
iii
Acknowledgements
The development of this Handbook wat; enthusiastiacally
supported by our immediate supervisors, Anna Zurek and Toni
Taylor. It would not exist without the support of the Department
of Energy and in particular the Materials Sciences and
Engineering Division, Office of Basic Energy Sciences. We also
Analysis for providing a superb foundation for this second

edition-in fact, some of the authors from that first 'edition also
repeat for this one. We thank the following individuals who have
helped us to review some of the materials and/or have made
valuable suggestions for the new book: 1. Tesmer, W.N.
Lennard, H. Shen, G.A. Glass, B.L. Doyle, W.A. Lanford, L
Usov, R. Greco, and J. Zhang. We again thank the many
contributors who have either written Dew chapters/appendices or
made significant updates to the old chapters/appendices for this
book. Their names and institutions are listed below,
want to thank our families for their constant support during the
development of this Handbook.
We arc grateful.to the Materials Research Society (MRS) for
undertaking the formidable task of publishing this handbook.
The Director of Publications and Marketing, Gail Oarc, and her
staff offered valuable assistance in guiding the book to
Los Alamos, New Mexico, USA, July 2009
completion.
We gratefully acknowledge the editors and authors of the first
Y.Q. Wang
0/ Modem Ion Beam Materials
M. Nastasi
edition of the Handbook
Boston University
Soumendra N, Basu
CNRS-CERl, Orleans, France
Gilbert Blondiaux
Jean-Luc Debrun
Institute 0/ PhYSics and Power
Engineering, Russia
Alex Gurbich
Instituto Tecno16gico e Nuclear,
Sacavem, Portugal
Centra de Fisica Nuclear da
Universidade de Lisboa
Nuno Barradas
Los Alamos National Laboratory

Michael Bozoian
Aaron Dick
Lynn Foster
Michelle Lee
Carl J. Maggiore
Michael Nastasi
Joseph R. Tesmer
Yongqiang Wang
Paul Scherrer institUle and ETN Zurich

Switzerland
Max Dobeli
Research Institute/or Technical Physics
and Materials Science (AfFAJ. Budapest.
Hungary
Gabor Battistig
SAPIR, Institut des NanoSciences de
Paris and Universite de Marie et Pierre
Curie, Paris, France
Ian Vickridge
Sandia National Laboraton'es
J, Charles Barbour
Gyorgy Vizkelethy
State University o/New York at Albany
William A Lanford
Technical Research Centre of Finland
Juha-Pekka Hirvonen
Texas A&M University
Lin Sbao
Universitat der Bundeswehr Munchen,

Germany
A. Bergmaier
Gunther Dollinger
Patrick Reichart
University of Arizona
Rjchard P. Cox
John A Leavitt
Laurence C. McIntyre, Jr,
University ojHelsinki
Reijo Lappalainen
Jyrlci Raisanen
era Rauhala
University a/Guelph, Canada

J. lain Campbell
Stephanne Taylor
University a/North Carolina

Max L. Swanson
University ofSurrey, UK
Chris Jenycs
McMaster University
John A. Davies
Paolo Rossi
University 0/ Western Ontan'o

William N, Lennard
Jan V. Mitchell
Middle Tennessee State University

Martha R. Weller
United Stales Naval Academy

James Ziegler
University o/Wisconsin
Paul M. DeLuca, Jr,
The UniversityandINFN, Padua, Italy'
Vanderbilt University
Robert A. Weller
National University a/Singapore

Mark Breeze
iv
CHAPTER
D
INTRODUCTION
Y. Q. Wang and M. Nastasi
Los Alamos National Laboratory, Los Alamos, New Mexico, USA
The second edition of the Handbook of Modern Ion Beam

Materials Analysis is largely based on the framework of the first
edition, often refered to as the "Black Book", edited by J.R.
Tesmer and M. Nastasi (Tesmer and Nastasi, 1995). The Btack
Book was a direct outgrowth of the widely used (but now out-of
print) ion Beam Handhook/or Materials Analysis, often referred
to as the "Green Bookn , edited by 1.W. Mayer and E. Rimini
(Mayer and Rimini, 1977). The Green Book is primarily a
compilation of data and, to the novice, not very user-friendly_
The intent of writing a new handbook was to expand and update
the Green Book, as well as to provide a different, more
organized style of presentation. Emphasis is placed not only on
the existing and new data, but also on practical examples of how
analysis techniques are applied to common problems.
Background materials are also presented where appropriate. The
discussions are aimed at an audience with an education
equivalent to that of a graduate student in materials-science- and
engineering-related fields.
The first edition of the Handbook of Modern Ion Beam
Materials Analysis was produced in 1995 and bas been popularly
used by many students, ion beam researchers, and accelerator
laboratories throughout the world ever since. Over the past
decade, steady advancements have occurred in many ion-beamanalysis-related fields, including ion stopping, data reduction
software, nuclear microprobe development, and new ion beam
techniques for materials characterization. According to the
o~e search engine SearchPlus, well over 20,000 publications
smce the Black Book's publication have reported the use of one
or more ion beam analysis techniques. New advancements in ion
beam analysis have inspired us to produce the second edition of
the Handbook of Modern Ion Beam Materials Analysis. As well,
the ion beam analysis community has responded that the
omission of the particleinduced Xray emission (PIXE)
technique from the first edition was a mistake that should be
corrected in the new edition of the handbook.
In addition to the revision of many of the original chapters in

the first edition, four new chapters and corresponding
appendices have been added to the second edition. These new
chapters are: Hydrogen Analysis by Proton-Proton Scattering
(Chapter 9), Particle-Induced X-ray Emission: PIXE (Chapter
11), Nuclear Microprobe Analysis (Chapter 13), and Data
Analysis Software for Ion Beam Analysis (Chapter 14). At the
same time, two of the chapters in the first book have been kept
largely unchanged: Backscattering Spectrometry (Chapter 4) and
Charged Particle Activation Analysis (Chapter 10).
Because of significantly increased content, the second edition
of the bandbook is divided into two volumes. Volume 1, which
c~ntain~ Chapters 1 through 17, presents fundamental principles,
diSCUSSIons, and examples. Volume 2 contains Appendices 1
through 21, which are the compilations of relevant data. A
v~teran of ion beam analysis might well be able to get along
Wlth Volume 2, whereas the lessexperienced reader will find the
principles, discussions, and examples in Volume 1 to be of
interest.
The ion beam analysis (lBA) po~ion (first volume) of the
handbook is a collection of materials analysis techniques that
utilize energetic ion beams as probes, including Rutherford
backscattering spectrometry (RBS), elastic recoil detecion
(ERD), nuclear reaction analysis (NRA.), particle-induced X-ray
emission (PIXE), and ion channeling. Over the years,
backscattering spectrometry has been extended to include the
non-Rutherford region for enhanced measurements of light
elements, termed elastic backscattering spectrometry (EBS) or
nuclear resonance scattering (NRS). Elastic recoil detection also
goes by several. other names, such as elastic recoil detection
anal~i5 (ERDA), elastic recoil spectrometry (ERS), and forward
recoll spectrometry (FRES). Analogous to PIXE, particleinduced gamma emission (pIGE) has become a distinctive
branc~ of nuclear reaction analysis. Unlike spectrometry
techmques such as RBS and ERD where mass-energy ambiguity
Chapter 1
is unavoidable, PIXE and PIGE are spectroscopy techniques that

use characteristic X-rays from atomic structurc or y-rays from
nuclear structwe for postive element identification.
induced charge (IBlC) microscopy. Thus, a new chapter

(Chapter 13) is introduced to discuss basic principles of highenergy nuclear microprobes and their applications in materials
characterization.
The fundamental physics of the interactions bctween incident

ions and individual atoms (and their constituent particles) in the
solid provides the underlying science for ion beam analysis.
These interparticle interactions depend on many parameters
including ion velocity and energy, ion and atom size, atomic
number and mass, and distance of closest approach between the
ions and atoms in the solid. The ion-atom interactions can be
grouped as "major" and "minor" interactions depending on their
effects on the energy and trajectory of the incident ion (Bird and
Williams, 1989). Major interactions include large-anglc
scattering, atomic displacement (vacancies and replacement),
sputtering, inner-shell ionization. and nuclear reactions.
Emission products from the major interactions are used to derive
infoID"Jation on material composition and structure and provide
the basis for elemental ion beam analysis. Minor interactions
control the slowing of the ion (energy loss) and perturb its
trajectory (multiple scattering) and, more importantly, bring to
the methods of lEA a unique capability for deriving depth
information without physically removing/sputtering the surface
target material, thus providing the basis for ion beam depth
profiling. Details about the interactions between incident ions
and target electrons and the associated energy loss process are
presented in Chapter 2, and the physics of the interactions
between charged particles and target nuclei is presented in
Chapter 3. Specific ion beam analysis techniques pertaining to
charged-partic1e-target-nucleus interactions are discussed in
Although computer programs to analyze data from RBS and

NRA date back to the early 1970s, the importance of data
analysis software was not fully realized until the mid-1980s
when the need for fast, nondestructive, and quantitative analysis
of more complicated thin-film structures and dopant diffusion
profiles was quiekly recognized in the rapidly growing
semiconductor industries. Currently, there are over a dozen
software packages available to aid routine ion beam analysis,
which includes converting energy spectra into appropriate depth
profiles, Simulating experimental conditions, and educating or
training new ion beam analysis users. Details of the underlying
physics, differences in mathematical algorithms, and
intercomparisons of these data analysis software packages are
included as a new chapter (Chapter 14) in this new edition.
Is IBA a nondestructive technique? The answer obviously
depends on the definition of the term "nondestructive". For
example, the leakage current of a silicon p-n junction might
increase Significantly after exposure to the irradiation associated
with ion beam analysis, but its dopant profiles might be hardly
changed. In general, from the macroscopic and compositional
point of view, IBA is a nondestructive technique. From the
microscopic point of view, ion-bearn-induced damage on the
target always exists. However, the most damage introduced by
megaelectronvolt ions in lEA is in the region where the ions
come to a halt, normally several microns below the surface. So
in surface and thin-film analysis, IBA is generally considered as
a nondestructive teChnique. However, caution must be taken
when organic compounds, polymers, insulators, and singlecrystallinc targcts are analyzed. Element loss, sample charging,
and unwanted channeling behavior could introduce additional
uncertainties when these types of targets are subjected to the
analyzing ion beam. Specific pitfalls in ion beam analysis are
discussed in Chapter 15.
Chapters 4 through 10. The basic principles of particle-induced

X-ray emission (PIXE) and its comparisons to energy-dispersive
spectroscopy (EDS) using an electron probe and X-ray
fluorescence (XRF) spectroscopy using an X-ray probe are
presented in Chapter 11.
In addition to element/composition analysis and depth

profiling, ion beam analysis also provides a powerful means to
measure radiation damage and defect information in singlecrystalline targets. This is accomplished by combining one of the
Ion beam analysis relies on 1-5 MY ion accelerators to

produce well-characterized (in terms of mass, energy, charge
state. spot size) ion beams as analyzing probes. Standard vacuum
chambers, nuclear insttumentation, and dctection electronics are
used to perform the measurements and to acquire ion beam data.
IBA instrumentation and laboratory practice are discussed in
elemental analysis techniques (e.g., RBS, NRA, PIXE) with the

ion-confinement characteristics (channeling) uniquely exhibited
by highiy ordered singlecrystalline lattice structures. In fact, ion

channeling has become an imperative tool for determining the
lattice location of simple defects such as interstitials before they
agglomerate into large defects or loops for electron microscopy
observations. Jon channeling is discussed in Chapter 12.
Chapter 16.
Finally, we urge that caution be observed when using beam
energies exceeding the Coulomb barriers of the target elements
or of the particles of the beam. Under these conditions, nuclear
reactions, whether wanted or unwanted, will occur, and there can
be associated radiation hazards from both prompt and induced
radiation. The extent of these hazards depends on the beam, beam
Over the past two decades, rapid development in accelerator

rrllcrobeam technologies and the associated new characterization
techniques uniquely related to nano- and micro-sized ion beams
have expanded our traditional view and application of ion beam
analysis in materials science. ,These include not only three~
dimerisional elemental mapping through fl.-RES and twodimensional trace-element measurements through ll-PIXE, but
also new ion beam microscopy techniques such as ion beam
current, and target (sample) material, as well as the facility layout

and shielding. Always consult your local health physics
professional before undertaking measurements involving nuclear
Introduction
reactions, including those involving elastic resonances. More on

radiological safety is presented in Chapter 17.
REFERENCES
In the analysis universe, there are many surface and thin-film

analysis techniques; so many, in fact, that most of us cannot
remember or even recognize their acronyms. When should one
choose or not choose IDA? It really depends on what one wants
to know about his or her sample. For ex.ample, megaelectronvolt
IBA techniques can obtain composition information independent
of the chemical bonding (electronvolt range) as compared with
X-ray photoelectron spectroscopy (XPS) or Auger electron
spectroscopy (ABS). This is considered to be a weakness of IBA
in that no chemical information about the sample can be
obtained. At the same time, this could also be a strength, given
that this insensitivity to chemical effects allows quantitative
elemental analysis. In fact, RBS is considered the most
quantitative analytical technique in surface and thin-film
analysis. In the end, all analytical techniques have their pros and
cons, and IBA should be used in conjunction with or in
complement to other analytical techniques to be more effective
for modern materials cbaracterization.
Analysis, Academic Press, Sydney, Australia.

Mayer, J.W., and Riroini, E. (eds.) (1977), Ion Beam Handbook
for Materials Analysis, Academic Press, New York.
Bird, J.R., and Williams, J.S. (1989), Ion Beams/or Materials
Tesmer, J.R., and Nastasi, M. (eds.) (1995), lIandbook 0/

Modern Ion Beam Matenais Analysis, Materials Research
Society, Pittsburgh, PA.
CHAPTER
ENERGY
Loss
AND ENERGY STRAGGLING

E. Rauhala
University of Helsinki, Helsinki, Finland
J. F. Ziegler
United States Naval Academy, Annapolis, MD, USA
CONTENTS
2.1INTRODUCTION .................................................................................................................. 7
2.1.1 Basic concepts and deimitions:.................................................................................. 8
2.1.2 Units .......................................................:.................................;................................ 9
2.1.3 Compilations and programs .................................................................................... 10
2.2 THEORY AND MODELS OF ENERGY LOSS .............................................................. 10
2.2.1 Basic physics ..............................................................................................................10
2.2.2 The scaling rule .........................................................................................................11
2.2.2.1 Effective charge .................................................................................................. 12
2.2.3 Computer calculations of stopping powers ........................................................... 12
2.2.3.1 Comparisons of stopping programs to data .................................................... 13
2.2.4 Special interests for stopping data in silicon ......................................................... 14
2.2.4.1 Comparisons of helium stopping data in Si among differeut sources .......... 14
2.2.4.2 Comparisons of helium stopping data in Si:
crystalline versus amorphous ........................................................................................ 15
2.2.5 Elemental targets versus mixtures and compounds ............................................. 16
2.2.5.1 Bragg's rule ......................................................................................................... 16
Chapter 2
2.2.5.2 Accuracy, deviations, and corrections to Bragg's rule ................................... 17

2.2.6 Stopping cross section aud the depth scale ............................................................ 17
2.2.6.1 Thin targets .................:....................................................................................... 18
2.2.6.2 Thick targets ....................................................................................................... 18
2.2.7 Isotopes....................................................................................................................... 18
2.3 STRAGGLING...................................................................................................................... 19
2.3.1 Models ........................................................................................................................ 19
2.3.1.1 Bohr's theory ....................................................................................................... 19
2.3.1.2 Corrections to Bohr's theory, other models .................................................... 19
2.3.13 Straggling in mixtures and compounds ........................................................... 20
23.1.4 Additivity of energy loss iluctuations ............................................................... 21
2.3.2 Comparisons to experimental data ......................................................................... 21
2.4 RANGE .................................................................................................................................. 21
2.4.1 Range calculations and simulations........................................................................ 21
2.4.2 Ranges in compounds ............................................................................................... 21
2.4.3 Isotopic effects ........................................................................................................... 21
REFERENCES
........................................................................................................................ 22
Chapter 2
The experimental data on ion energy loss are overwhelming.

illustrates the stopping cross sections and a
2.1. INTRODUCTION
Figme 2.1
Knowledge of the slowing down of ions in traversing matter is

of fimdamental importance in the techniques of materials analysis
using beams of charged atomic particles. Depth perception in these
semiempirica1 fit for just one element, aluminum, for ions Be

through U. The data comprise 148 data sets with 931 data points.
The total database of this Internet-based compilation (SRlM,
Stopping and Range of Ions in Matter) comprises more than
techniques follows directly from the energy lost by the probing

particles, and the energy loss affects both qu.antitative and
25,000 data points. Figure 2.2 shows a comparison between

experimentaL results and the fitted curve for the same data. As
compositional analysis.
The physics of energy loss phenomena is very complex,

involving many kinds of interactions among the projectile ion.
target nuclei, and target electrons. Because of their significant
importance in many fields of physics, these phenomena have been
the subject of intense studies since the beginning of the 20th
century.
observed, most data points fall within 5-10% of the curve (from
http://www.SRlM.org).
...
00
...
.......
OIl
S
tJ
600
I.
I:
.,
>~
iI
---
lUI 400
....~
...0
(/l
S 200
:3
....~
S
::I
...<
10 2
10 3
Energy ( keV /amu )

FIG. 2.1. Stopping cross sections and a semiempirical fit for aluminum, for ions Be through U. (from http://www.SRlM.org). (Multiply
value in units of keV/amu by ion mass in amu for ion energy in keY; e.g., 103 keV/amu for 4He corresponds to a 4He ion energy of
4000 keY.)
Chapter 2
t'
QJ
.c:
r,..
'+'
1.2
QJ
....S
s..
QJ
1.0
~
~
r:1
I I
....~
.,
..
0.8
10%
+'
CJ)
10 1
10 3
_.-
10 4
Ion Energy ( keV/amu )
FIG. 2.2. Comparison between data and fit for the data ofFig. 2.1 (from http://www.SRIM.arg).
The theoretical treatment of energy loss phenomena has been
reviewed by Bohr (1948), Wbaling (1958), Fana (1963), Jackson

(1962, 1975), Bichsel (1972), Inokuti (1971), Sigmund (1975),
AhJen (1980), Linmark and Ziegler (1980), Ziegler (1977, 1980),
KUIrulkhov and Komarav (1981), Ziegler et 0/. (1985),
incident
International Commission on Radiation Units & Measurements

(ICRU) (1993, 2005), Sigmund (2004, 2006), Ziegler 0/ 0/. (2008),
~..
and others.
.: .. t"
The experimental methods have been reviewed and discussed

by, for example, Northcliffe (1963), Betz (1972), Chu (1979),
Bauer (1987), Mertens (1987), Powers (1989), ICRU (1993,
2005), Geissel et 0/. (2002), and Ziegler e/ 0/. (2008).
1rnn:Snitttea .
~artkI~
E:.i:AF!
:Zi
Zj
,MI
MJ
Energy loss in the relativistic energy region bas been reviewed

by Ablen (1980) and Lindbard and Sorensen (19%), among others.
FIG. 2.3. Particles of atomic number Z] and mass M] penetrating
through a material layer comp:>sed of atoms with atomic number
~ and mass M2 . The layer has a thickness of tix = t, corresponding
to an areal density of Nt. The initial energy of particles is Eo; the
energy lost by the particles is LlE.
Energy loss, .6.E
llli = average kinetic energy loss of point-like atomic particles
traversing matter.
Units: [Llli] ~ eV, keV, MeV, etc.
2.1.1. Basic concepts and defmitions
The definitions of the concepts related to the energy loss of

atomic particles traversing matter, such as energy loss, specific
energy loss, stopping, stopping cross section, stopping power and

stopping force, vary in the literature. Which definition applies in a
particular case can be established from dimensional considerations.
We adopt the following notation and definitions (see Fig. 2.3):
Energy Loss and Energy Straggling
E15 also sometimes called stopping power or linear energy transfer

(LEI).
Energy loss is a characteristic of the particle (e.g., "energy loss of

He ions", "energy lost by the ion beam")
Stopping cross section is a characteristic of the medium (e.g.,

"stopping cross section of silicon for He ions").
Stopping power, stopping force, dEJdx

dEldx ~ S:!J.EI!Jx -> dEldx, when!Jx -> 0, where t.E is the energy
loss upon traversing the distance m.
2.1.2. Units
Units: [dEldx] = MeVlmm, eV/_ etc.

dEldx might also be called specific energy loss.
Atomic mass units are used throughout this chapter. M J ;;;;: Ml[amu]
and M, ~ M,[amu] thus stand for the masse, of the ion and target
medium, respectively, in atomic mass units. For a compound or
mixture, take M2 to mean the average atomic mass; for example,
Stopping power is a characteristic of the medium (e.g., "stopping

power of silicon for He ions", "stopping power for He ions in
silicon").
for quartz M..

,
Stopping cross section,
Sl02
= (MSi
+ 2lVlo)/3
The mass density of the
p[glcm3]. Vo
medium is
is the Bohr velocity for atomic electrons.
Unit conversions relevant to energy loss calculations are presented
in Table 2.1.
,= (l1N)(dEldx) or ,= (l/p)(dEldx).
Units: [] = eVJ(1015 atoms/cm2) orkeV/(mglcm2), etc., as N is the
volume density (e.g., in atoms/em3) and p is the mass density (e.g.,
in g/cm'J.
Table 2.1. Unit conversions for energy, velocity, thickness, areal density, stopping power, and stopping cross section.
Multiply units
by
for units
Example
MeV
---
MeV/amu
4 MeV 'He - 1 MeV/amu
vIvo
(MeV/amuJ,n
M,[amu]
0.1581
1.389 x 10'
1.661 x 1O'M,[ amu]
(MeV/amu)'"
mI,
v/v, - I - 0.025 MeV/amu H

6
2 MeV4He~ VHe - 9.82 X 10 mls
om
IOta atomsfcm2 for Au -170 nm
om
100 )!g/cm' for C - 285 urn
lOJ~ atomslem2
100 llg/cm2 for Au - 305 x lOl~

1
atoms/em
MeV/(mg/cm')
100 eV cm~/10 ~ atoms for A1 20 3 ~

2.95 MeVcm'/mg
[M, = (2M" + 3Mo)/5; MAl = 2698,
Mo= 16.001
keV/)!m
30 eV cm2lI015 atoms forSi

-150 keV/~
1015 atoms/em2
p[g/cm']
10
)!g/cm'
p[g/em']
10'
Jlg/cm
1.66IM,[amu]
2
eV em 1I 0 1:'i atoms
1.66 I M,[amu]
eV cm21101:'i atoms
10'p[g/cm'J
1.66 I M,[amu]
Chapter 2
2.1.3. Compilations and programs

Since 1995, the year of the publication of the first edition of
this handbook, computer programs to perform calculations of
energy loss phenomena have become readily and widely available.

The inclusion of printed tables of stopping powers in this chapter is
thus no longer reasonable. They have all been omitted in this
edition. Plots of stopping powers as data sources in the appendices
have also been excluded.
The number of experimental studies on ion energy loss is

overwhelming. These data on ion stopping and ranges are coUected
in a number of printed compilations. In addition, there are Web
sites where compilations and summaries of data and theory, as 'Well
as programs for stopping calculations, are available. Unless stated
otherwise, an data and calculations refer to energy loss in
amorphous material. For energy loss in crystalline material. see
Section 2.2.4.2 below.
Printed compilations:
Whaling (1958)
Littmark and Ziegler (1980)
Northcliffe and Schilling (1970)
Ziegler(1977,1980)
Johnson and Gibbons (1970)
Janni (1982)
Bichsel (1972)
Ziegler et al. (1985)
Ziegler and Chu (1974)
Huben et al. (1990)
Brice (1975)
Berger and Paul (1995)
Gibbons et al. (1975)
[CRU (1993)
Winterbon (1975)
Paul and Schinner (2003)
Andersen and Ziegler (1977)
Ziegler (2004)
Huben I al. (1980)
ICRU (2005)
Ziegler e/ al. (2008)
Data compilations on Web sites:
MSTAR (II. Paul and A. Schinner): http://www.exphys.uni[w.ac.at/stopping!
SRlM (J.F. Ziegler): http://www.srim.org!
SRIM (J.F. Ziegler): bttpJIwww.srim.org!
Stopping Power for Light Ions (II. Paul): http://www.exphys.unilinz.ac.atlstopping!
the
A short desaiption of
programs above is given in Section
A3.1 of Appendix 3. Parameters of some oftbe compilations,
tables, and programs are listed in Section A3.2 of Appendix 3.
Computer software for stopping calculations:

ASTAR and PSTARCM.J. Berger, J.S. Coursey, M.A. Zucker,
and J. Chang): http://www.exphys.uni-lw.ac.atlstopping!
2.2. THEORY AND MODELS OF ENERGY LOSS
ATJMA (II. Geissel, C. Scheidenberger, P. Maizacher, J.

Kunzendorf, H. Weick): httpJlwww-linux.gsi.del-weickiatimai
2.2.1. Basic physics

..
The relative importance of the interaction processes between
the ion and the target medium depends mostly on the ion velocity
and on the charges of the ion and target atoms. The energy loss of
ions is' usually considered in three velocity regimes, namely, low,
intermediate, and high velocities. The classification is based on the
CASP (p.L Grande and G. Schiwietz):

http://www.hmi.de/peoplelschiwietz!casp.html
GEANT4 (Geant4 collaboration):
http://geant4.web.cem.chlgeant4/
10
Energy Loss and Energy Stragglir1g
ion velocity in comparison to the orbital velocity of the atomlc

electrons (see Fig. 2.4).
introduced (Northcliffe and Schilling 1970). According to t::l:ris

suggestion, as a consequence ofEq. (2.1), the effective chargel'Ll.so
relates the energy loss of different ions by scaling: The ratio of "the
energy loss of the heavy ions to that of the light ions is obt2.ine!i.. by
the square of the effective charge.
At ion velocities v that are significantly lower than the Bohr

velocity of the atomic electrons Ve, the ion carries its electrons and
tends to neutralize by electron capture. At very low velocities,
elastic collisions with the target nuclei, that is, nuclear energy loss,
dominates. As the ion velocity is increased, the nuclear energy loss
diminishes as lIE. Simultaneously, electronic energy loss (Le.,
inelastic collisions with the atomic electrons) soon becomes the
main interaction. The total energy loss is taken as the sum of the
nuclear and electronic contributions.
For most applications of ion beam analysis, nuclear st~ing

is small, For example. above 200 kcV/amu, the nuclear stOf.P'ing
contribution is typically below 1% of the electronic stOfPing
contribution.
Many different theoretical approaches to treat the vwi GUS
aspects of the slowing dovro of ions in matter exist For treatises On
theory, the reader is referred to the literature. Major theoretical
advances accomplished recently include the works of SigIll:und
(1982), Oddershede and Sabin (1983), KUrth and co-workers
(1981), Brandt and Kitagawa (1982), Ziegler et al. (1~85),
Lindhard and Sorensen (1996), Sigmund and Schianer (2000,
2002a, 2002b, 2oo2c). Grande and Schiwietz (2002), M'YJ'lf"d et
al. (2001,2002), Arista (2002), and Sigmund (2006).
In the low-ian-velocity range from v :::;' 0.1 Vo to v z Z]2IJ ve, the

electronic energy loss is approximately proportional to velocity or
Eln. Ion energy loss in the low-ion-velocity region was first treated
by Bohr (1913, 1915, 1948), Lindhard and co-workers (1953,
1963), and Firsov (1959).
At high velocities v Vo, the charge state of the ion
increases, and finally, the ion becomes fully stripped of its
electrons. At a given velocity, the energy loss is proportional to the
square of the ion charge. The theoretical framework for the energy
loss of high-velocity ions originates from the works ofBobr (1913,
1915, 1948), Bethe (1930, 1932), Bethe and Heitler (1934), and
Bloch (1933a, 1933b), The Bethe-Bloch formula has the general
form
dE
dx
(,,(E)
NZ, Z,e ) f ,
M,
2.2.2. The scaling rule

Based on the analysis of ex.perimental energy loss data and
theoretical considerations, semiempmcal fOImulations of the
effective charge and the energy loss for all kinds of projectile ions
in all elemental targets have been derived. Although the effective
charge concept and the scaling approach can be very useft.J.l in
predicting stopping powers, it has been argued (e.g., SignlUnd,
2004,2006) that these concepts are not well based theoretically.
(2.1)
Writing the stopping cross sections from Eq. (2.1) for two
different projectiles a and b having the same velocity v in a
medium ~. we obtain
where e is the electron charge and f(ElM 1) is a function

depending only on the target, not on the type of projectile.
10 g.
:~
~
" . , 1 "",m
.-., .
:l!
.....r-' i
t , '.
..
,)
."
Lao
'J
0.001
,I
(2.2)
j
I
Here, y is the fractional effective charge, defined according to
,,
I
equation:
,, " ,
Z",rr(V,Z,)e = r2,e.
I,
~
J..
."1
\ II
,I
I,
,
,,I
1.
".,Dll
!ntt",,;d.
[.~
,.,
(EI.h) I :MiN
"
(2.3)
The stopping cross section of a heavy ion 011) is then

obtained from the stopping of protons (H) at the same velocity
(assuming 1H ~ 1) as
1000
(2.4)
FIG 2.4 Regimes of heavy ion stopping illustrated by 0 in

aluminum (from Sigmund, 2004; data compiled by H. Paul (see
http://www.exphys.uni-l.in.z.ac.at.stopping).
This is the scaling rule of stopping cross sectioDS. At the same

velocity, we have
To extrapolate the theoretical treatment from the high-velocity

region to the intermediate-velocity region where the ion is only
partially stripped, concept of effective charge has been
(2.5)
11
Chapter 2
The higher the ion energy, the closer y approaches unity.

When vIvo = 6.325 (E[MeV/amu])II2 I, the ion is assumed to be
fully stripped and y ~ 1.
Quick estimates of heavy ion stopping in elements can be
calculated from the proton stopping values if one assumes 1 equal
to unity. Depending on ion velocities v Vo, this procedure is
usually accurate to better than 20%.
EXAMPLE 2.1. ApproximationYHc "" 1.
Assuming YHc = 1 leads to the follO\~ing mnemonic: 4He ions at 2
MeV lose energy 4 times more rapidly than lH ions at 0.5 MeV.
This rule is accurate to about 3%.
EXAMPLE 2.4. Effective charge ofLi ions.

Ziegler (1980) gives a parameterization for Li ions (Z, ~ 3) as
Yu =A\I-exp[ -(B+C)]1
A =1 +( 0.007 +5xlO~' Z,)exp{-[7.6-ln(E u[keV/amuj)
J'}
B ~ 0.7138 +0.002797Eu [keV/amu1

C =1.348 x 10"' (EdkeWarnu])' .
This parameterization can be used to calculate 'Li ion stopping in
EXAMPLE 2.2. Approximation YLi = 1. Calculate the approximate

'Li ion stopping cross sections in carbon at 2, 5, and 10 MeV.
As Mu ~ 7M H, we have EH ~ EJ.!7 "285, 715, and 1430 keV at
the same velocities. Taking Yu = I and Zl]YLi2 = 9, we can write
Eq. (2.4) as eti2, 5,10 MeV) ~ 9,,"(285, 715, and 1430 keV).
The program PSTAR gives <H(285, 715, and 1430 keV)" 0.489,
0.282, and 0.177 MeV cm'/mg, and thus, ed2. 5. and 10 MeV)"
4.40,2.54, and 1.60 MeV cm'/mg.
carbon at, for example, 2, 5, and 10 MeV.

Eli ~ 2, 5, and 10 MeV correspond to -285, 715, and 1430
kcV/arou (- 3.4, 5.3, and 7.6v/vo). Eq. (2.7) yields Yu (2, 5, and 10
Me~ 0.80, 0.97, and 1.00.
From Eq. (2.5) and EXAMPLE 2.2, we then find <uC2, 5, and 10
MeV) ~ 0.4899(0.80)', 0282 9(0.97)', and 0.1779(1.00)' MeV
cm'/mg ~ 2.81,2.38, and 1.60 MeV cm'/mg. These values can be
compared to those obtained in EXAMPLE 2.2.
2.2.2.1. Effective charge

For the heavy-ion fractional effective charge y, different ZIand Z2-dependent formulations have been proposed in the
EXAMPLE 2.5. Effective charge for heavy

(Z, > 3) ions.
literature, including those of Northcliffe and SChilling (1970),

Forster e/ af. (1976), Anthony and Lanford (1982), and Hubert 0/
ai. (1989, 1990). Semiempirical parameterization:s based on an
extensive amount of experimental data were presented by Ziegler
(1980), Ziegler e/ af. (1985), and Hubert el af. (1990).
Ziegler (1980) gives 1 ror heavy ion scaling (Z, > 3) from proton
stopping in the energy range 200 keV/amu < E < 22 MeV/amu as
y,. = 1- exp( -A) [1.034 -0.177 exp (-{).08114Z HI )]

A =B+0.0378 sin(rrB/2)
(2.8)
EXAMPLE 2.3. Effective charge of He ions.

For 'He ions, Ziegler el of. (1985) and Ziegler and Manoyan
(1988) presented the simple formula
More
recent
Z,-
and
Z,-dependent
effective-charge
parameterization fonnulae for 2.5-75 MeV/amu heavy irn;tS, with

'He used as me reference particle, are found in Ziegler et af. (1985)
and Hubert et al. (1990). These parameterizations require more
extensive calculatioru; than those in EX.A1v1PLES 2.3-2.5.
(2.6)
2.2.3. ComputeI'" calculations of stopping powers

Computer programs for the calculation of stopping powers of
all elemental matter, any mixture, and many compound materials
for any ion over a wide range of energies can be downloaded from
or used on the Internet Most programs rely on the semiempirica1
approach and the scaling and effective charge concepts; a few are
more theoretically oriented. Usually, both the electronic and
nuclear contributions of stopping can be calculated separately.
where FlM, is in keV/amu (M, ~ 4.0026) and the fitting constant

values are '" ~ 0.2865, 0.1266, -{).001429, 0.02402. -{).01135. and
0.001475 for i ~ 0-5, respectively.
According to Eq. (2.6), the mnemonic in EXAMPLE 2.1 is

accurate to 3%, the factor ZHe)H/ being actually 3.89 instead of 4.
At E H ~ 0.5, 1.0, 1.5, and 3.0 MeV, the factors are 2.88, 3.46,
3.75, and 3.99, respectively.
12
Energy loss and Energy Straggling
At the moment, at least the following programs are fre_ely

available (sec Section AJ.I in Appendix 3 for details, tenns, and
conditions for use): AST AR, PST AR, ATIMA, CASP, and
GEANT4, MSTAR, and SRlM. Many of the programs offer
siJnple online interfaces for quick calculations of stopping and
graphs in elemental matter. SRIM has a module (see Section AJ.J
in Appendix 3 for details) that can be incorporated into any
Figure 2.5 shom stopping cross-section data for 4He iom in

aluminum from the SRIM database and a semiempirical ClLVe
fitted by SRlM, together with a plot of the deviation from the fit
(Ziegler ef ai., 2008). This is an example of a typical <ase
relevant to ~on beam analysis where large amounts of data enst.
The scatter between the bulk. of the data is between 5% and 1C~.
computer program as a subroutine.
Figure 2.6 shows experimental data for 12C ions in alumin..1IIl

and curves from several different programs/tabulations. Again.." a
large scatter within the data can be observed. In addition, at the
stopping maximum around 5 MeV, the curves show deviatbns
between the different programs on the order of 10%.
2.2.3.1. Comparisons of stopping programs to data

Data compilations, tables, and programs show a large scatter
of data. The scatter depends on ions, target materials, and energies.
Two examples, He and C ions in aluminum, are illustrated next.
.
'"
......
60
N~
I
>
40
.!i"
g:
.s
'"
.."
20
0.8
f-+--+-I-I----f--f.-f,H-
+--H-I-I--~ft~;.'j~t11
'"
10~
10.t.
( keV/amu )
Ion
FIG: ~.5. Measured stopping power data for 4He ions in aluminum from 43 data sets and 421 data points for 'He ions (top) and
deVlatton of the data from the semiempirical fit (bonom) (from hnp:llwww.SRIM.org). (Multiply units oHeV/amu by 4 for He energy
mkeV.)
\3
Chapter 2
Considering the energy range more relevant from the point of

view of ion beam analysis, namely, 0.1 MeV/amu < ElA < 10
MeVlamu, we find again tha~ in general, the standard deviations of
normalized differences between the data and calculations are
similar for the various programs. The semiempirical approaches
show a slightly better overall performance, with standard
deviations on the order of 2-8%. All programs do better for higher
energies: above 1 MeV/amu, the standard deviation is usually
below 5%.
We can conclude that, based on the comparisons described
above, uncertainties on the order of several percent are to be
expected in stopping power calculations.
2.2.4. Special interests for stopping data in silicon
Because of its enormous technological' interest, silicon is a
special case in applications of ion beams to materials analysis. The
stopping of Si as a target material, especially for He ions, is of
great interest to beam applications.
FIG. 2.6. Measured stopping power data for He ions in

aluminum from many different sources. Eaeh symbol
corresponds to a data point, with the different letters
corresponding to different publications. The various curves
correspond to different fitted semiempirical curves from the
literature. (Multiply units of Energy per nucleon by 12 for 12e
energy
In
MeV.)
(From
http://www.exphys.unilinz.ac.at/stopping.)
Unless otherwise specified, the data and calculations usually

present energy loss of ions in amorphous silicon. However, data
and ca1culations have also been published for energy loss of
various ions channeled along axial and planar directions (see
Section 2.2.4.2).
Comparisons of programs, data compilations, and tables can

be found for sevemhoDS and materials (ICRU, 2005; Paul, 2006,
2007, 2009a; Paul and Schinner, 20030, 2005, 2006), Section
A3.2 of Appendix 3 provides a comparison of parameters for
tables and programs from ICRU Report 73 (2005).
2.2.4.1. Comparisons of helium stopping data in Si among

different sources
Sev~ral recent publications have discussed the stopping of
silicon for He ions, for example, Bianconi er al. (2000) and
Barradas e/ al. (2002) (and references therein). The surface yield
of a backscattering spectrum is a measure of the stopping power: it
is inversely proportional to the stopping cross section on the
inward and outward paths of the ions in the target material. Table
2.2 presents results of an intemationaJ collaboration (Bianconi et
al.. 2000) to define the surface yield of a Rutherford backscattering
spectrum ~) for ~e ions incident on amorphized silicon. In the
collaboration, the measurements of five independent laboratories
were in agreement to within 2%. The table also compares
calculations by semiempirical fits.
Tables of statistical comparisons of data and program

calculations by Paul and coworkers are provided in Section A3.4
of AppendJx 3. Two of the programs included, PASS (Sigmund
and Schinner, 2002e; Sigmund ef al., 2003) and CASP are based
aD theoretical treatments, whereas the others are semiernpirica1,
most of them being based on the effective charge concept.
As observed from the tables in Section AJ.4 of Appendix 3,

the standard deviation cr (defined in Section A3.4) between data
and computer calculations (tabulated values) is on the order of 213% for all programs and tables listed. This applies to all ions and
all stopping media studied in a wide energy range ofEiA = 0.001100 MeV/amu.
Table 2.3 shows a comparison of measured and calculated

electronic stopping powers of amorphous Si for 4He ions.
Concluding from these particular data sets, the calculations
reproduce the experimental data within about 3-4%.
14
Table 2.2. Experimental surface RBS yield (Ho) for Si as a function of energy and calculations using different semiempirical fits (rrGm
Bianconi et aI., 2000. RUMP refers to http://www.genptot.comi, SRlM2000 to ~ttp:l/www.srim.orgl and KKKNS refers to Kome et
al., (1998a)).
Ho counts/(fLC msr keY)

1.0
MeV
1.5
MeV
2.0
MeV
2.5
MeV
3.0
Mev
Experiment
58.5
28.8
18.3
12.7
9.47
RUMP
57.2
27.7
17.1
12.0
9.06
SR1M2000
57.6
28.1
17.3
12.1
9.09
KKKNS
60.5
29.5
18.3
12.8
9.69
Table 2.3. Stopping cross sections from several recent publications compared with serniempirical fits (from Barradas et al., (2002),
with the stopping cross sections from AST AR and SRIM-2008 added. "This work" refers to Barradas et al., (2002), Lulli to Lulli et
aI., (2000), Hoshino to Hoshino el 01., (2000), ZBL85 to Ziegler el 01.. (1985), KKKNS to Konac el aI., (1998a)). The measurelllent:;
of Barradas et ai. (2002) agree with those of Bianconi et al. (2000) within about 3%, and the calculations also reproduce the data to
within about 3%,
Sa ~eV/IO]5 at/emIl
Eolli'!e~
0.5
I.!
1.5
1.6
2.1
2.6
2.5
This work
70.9 3.9
62.10 41
60.43 39
54.20 34
52.75 32
47.5\ 37
46.32 40
42.0 33
4L0763
37.6573
This 'Norka
68.1 1.4
62.31::1:: 31
60:69 25
54.33 15
52.81L7
47.33 24
46.08 27
41.51 15
40_62 36
37.0644
68.76
61.15
59.35
-52.88
51.46
46.48
45.39
41.55
40.7
37.66
Lulli et al. [19]

Hosblno eta!. [43]
55.3 1.2
41.59
4504 1.0
50.3 1.1
ZBLS5 [23J
71.30
63.4
61.5
54.9
53.5
4&.4
47.3
43.3
42.4
39.20
KKKNS [7J
67.41
59.95
58.19
51.84
50.45
45.57
44.5
40.74
39.9
36.92
ASTAR
66.20
6004
59
53.5
52.3
47.8
43.1
39.30
61.2
59.5
53.2
51.8
45.9
42
38.10
SRIM-200&
Section AJ.5 in Appendix 3 shows figures ofhdium ion energy loss data in silicon and fits to these data from semiempirical calculations.
The mechanism of energy loss along a channeled trajectory

is complicated and not fully understood, The channeled ions
traverse in a region of reduced electron density; thus, in the
electronic stopping regime, ions in a channeled trajectory
experience a reduced energy loss as compared to the random
trajectory.
2.2.4.2. Comparisons of belium stopping data in Si:

crystalline versus amotphous
The energy loss of ions traversing along an axial or planar
direction in a crystalline material, a "channeling direction",
differs from the energy loss of ions traversing an amorphous
materiaL A direction that is neither an axial nor planar direction
in a crystal is often referred to as a "random" direction.
IS
Chapter 2
The reduction of the energy loss or stopping power is

defined as the ratio of the energy loss or stopping power for an
ion traversing along a specific axial or planar direction in a
crystalline material to that for an ion traversing an amorphous
sample of the same material. This ratio is thus less than 1.
Experimental data on the energy loss in channeled

directions can be found in the literature. For the specific case of
He in silicon, there is a large variation in stopping values. Table
2.4 presents data for 'He along the (100), (110) and (Ill)
channels in silicon. The published data are given as a function of
energy or as a function of angle of incidence of the ion.
Table 2.4. Ratio of the energy Joss for "He ions in the axial direction to the energy loss in amorphous Si. The values have been read from
published graphs.
Energr~eV)
Sto~~ing
0.96
1.0
1.0
1.0
1.0
Axis
(100)
(100)
(100)
(100)
(100)
0.78
0.80
0.84'
0.88
0.89
Azevedo e/ al. (1998)

Sbao el al. (2006)
Bernardi et aI. (2006)
Yamamoto el al. (1999)
dos Santos e/ al. (1995)
2.0
2.0
2.0
2.0
2.0
2.0
(100)
(100)
(100)
(100)
(100)
(100)
0.71
0.90'
0.85'
0.74
0.64
0.76
ShaD el al. (2006)

Greco el al. (2007)
Bemardi e/ al. (2006)
Yamamoto el al. (1999)
Azevedo., al. (1998)
dos Santos el al. (1995)
(110)
0.54
0.57
0.53
0.52
0.48
Shao e/ al. (2006)

Azevedo el al. (2001)
Shao et al. (2006)
Sbao el al. (2006)
Azevedo el al. (2001)
0.84
0.72
0.71
Shao e/ al. (2006)

Azevedo ./ al. (2001)
Shao el al. (2006l
1.0
1.2
1.5
2.0
2.0
1.0
1.2
1.5
(I 10)
(I 10)
. (110)
(llO)
(111)
(111)
~lll~
ratio
Reference
Absolute stopping data given; ratio calculated assuming KKKNS stopping (Konac et at, 1998a, 1998b).
>I<
According to the data in Table 2.4, the reductions of the

stopping power for 1.0 and 2.0 MeV 'He in (100) Si are on the
order of 0.85 ( 0.05) and 0.75 ( 0.1), respectively.
2.2.5. Elemental targets versus mixtures and compounds

2.2.5.1. Bragg's rule
A simple linear additivity rule of energy loss in compounds

can be used on the assumption that the interaction processes
between ions and component target clements are independent of
the surrounding target atoms. Consider a compound or mixture
A",B,. Bragg's rule (Bragg and Kleeman, 1905) for the stopping
cross section AB [eV cm2/atom] for the compound can be written
Monte Carlo (Posselt, 1994; Yamamoto et al., 1999; Lulli et

al., 2002; Nowicki el al., 2005, and references therein) and
molecular dynamics (peltola et aI., 2004) computer programs have

been used in connection with stopping calculations of channeling
phenomena.
as
16
state efleets have been discussed by Ziegler et al. (2008) <and

solid-gas differences by Paul (2009b).
(2.9)
where m + n is normalized to unity. Note that the appropriate units
should be eV cm2/atom, as this equation docs not hold for stopping
power in other tmits such as keV/!Jlll
Figure 2.7 displays both the chemical comp01md effect and

the physical phase effect for H ions in water. This is an exau-::q,le
where Bragg's rule fails. Hydrogen undergoes a large s~
cbange in compounds; typically, its stopping changes by up to ~ 0%
in compounds. The figure shows the stopping of water (vapor and
ice) for H ions, calculated both using Bragg's rule and witb the
chemical compound corrections.
EXAMPLE 2.6. 2.0 MeV 4He ion stopping in silicon oxide Si0 2
SRlM-2006 gives ,Si(2.0 MeV) ~ 46.88 eV cm'I10 1l atoms and
sO(2.0 MeV) = 38.36 eV cm2/101S atoms. For Si02, we then have
SiO
a = (lssi + 2(P)13 = 4l.02 eV cm2/101S atoms.
2.2.6. Stopping cross section and the depth scale

The stopping power, dE/dx, or the stopping cross seetin."" E,
can be used to relate the energy lost by the probing ion.s, .6.E, too the
thickness penetrated in the target material. In principle, the
calculations involve solving integrals of the form
2.2.5.2. Accuracy, deviations, and corrections to Bragg's rule

Bragg's rule assumes the interaction between the ion and the
atom to be independent of the environment. The chemical and
physical state of the medium is, however, observed to have an
effect on the energy loss. The deviations from Bragg's rule
predictions are most pronounced, on the order of 10-20%, around
the stopping maximum for light organic gases and for solid
compounds containing heavier constituents, such as oxides and
nitrides.
D,E=
Eo
J(dE/dx)dx
o
or
x= J(dEldxtdE.
(2.10)
The former gives ~ when the depth dependence of stopping

is assumed, whereas the latter gives the depth scale x for a given
energy intervalLlE ~ E - E,.
To allow for the effects of chemical state, Ziegler and

Manoyan (1988) developed the "cores and bonds" (CAB) model
originally based on the works of Kreu~ and co-workers (1988 and
earlier references therein), Chan et 01. (1978), Sabin et al. (1985),
and Oddershede and Sabin (1989). The model assumes the energy
loss of ions in a compound to be due to two contributions: the
effect of the cores, that is, the closed electron shells of atoms, and
the effect of the chemical bonds. such as H-C or C=C bonds.
Hydrogen Stopping in Water- Vapor and liqUid.
The model works best for cases where the physical state of the
target has little effect and for target; similar to those in the original
experiments (see Ziegler and Manoyan, 1988). The largest
differences between the CAB calculations and simple Bragg's
additivity rule predictions are found around the stopping
maximwn. The differences reduce with increasing energy,
disappearing for ion velocities above lOvo. The average accuracy
of the calculation is better than 2% when compared to data from
severnl hydrocarbon targets.
[e
,Iii!.
aond ~tt:(oPQr) == O.!4
\"'~ ~rfi(~ortdl "" (loa?

o,Lo-_-'--'-LJ-'..Ll....IO..O---"--'-LJ-L...J.I...OOUO
Hydrogen ENERGY (keV jamul

FIG. 2.7. Chemical compound and physical phase effects for H
ions in vrater. The curves include both COrrettions. (From
http://www.srim.org/SRIMICompounds.htm.)
Physical state effects, that is, the dependence of stopping on

the physical state of the medium, have also been observed. In
general, most experimental data indicate <20% larger stopping in
vapor than in the solid phase for light ions (Thwaites, 1985). For
heavy ions, the opposite has been observed, the stopping being
<20% lower in gases than in solids (e.g., Herault et aI., 1991, and
references therein). Experiments on both chemical and physical
state effects have been reviewed by Thwaitc, (1985, 1987) and
Bauer (1990). The stopping of light and heavy ions in compounds
and the material dependence of stopping power have been studied
on the basis of the binary theory of electronic stopping by Sigmund
er ol. (2003) aod Shanna 0/ 01. (2004). Most recently, physical.
Depending on the target thiekness and the stopping variation
in the ion energy region in question, dEldx or E can be evaluated

either at the ion swface energy Eo or at some mean energy, for
example, En ~ (E + E,)I2. For 'He ions around 2 MeV, these
methods usually differ by a few percent for the first few hlilldred
nanometers of target For high accuracy, numerical ,integration of
the integrals in Eq. (2.10) should be used for thick layers and near
the stopping maximum
17
Chapter 2
The same units for thickness should be used as found in

stopping; for example, use areal density Nt [10" atowem'] if the
stopping cross section is given in eV cm2/10 15 atoms.
EXAMPLE 2.7. ProIOn depth scale in carbon. What is the 2.0 MeV
proton energy lost in a carbon target for depths of (a) 1000 nm and (b)
20~?
2.2.6.1. Thin targets
The
swface
energy
approximation
dE
AE = dx (Eo) &.: or 8E = E(Eo)Nl This
also be used to obtain the mean energy E... =
which
mean
energy
approximation,
yields
approximation
can
Eo -
AEI2 for the

then
gives
dE
AE= dx (E av )&':or8E=E..)Nl
2.2.6.2. Thick targets
Numerical integration of Eqs. (2.10) proceeds as follows:
Estimate hE (or Llx or Nt) as above. Divide the target into n slabs.
Use the surface energy approximation in each of these slabs; that
is, calculate the energy lost in the ith slab, .6.Ei. and the total ~ in
the target ao;
t.E =
(e;_,)(Nt);
(2.11)
i:: t.E;,
i-=l
...
where
dE .......-.,
dx. (Ei_d
and
~CE.t-I) are the stopping cross section and
energy loss, respectively, evaluated at the energy of the ion at the (i

- I)th ,lab.
18
-'--- ----._------_.
(a) The surface energy approximation gives dE = (2.866 eV cmll10lS

atoms( (17.6)( lOll atomslcm 2):::: 50 keV.
(b) We calculate ~E :::: 1000 keY by the swface energy approximation
as above. This implies E.. = F.-8E/2 = 1500 keY. Assuming 1500
2
15
keY) =3.506 eV cm /l0 atoms, the mean energy approximation thus
2
yields AE = (3.506 eY cm /lO B atoms) x (353 x lO"atomslcm2) ::::
1235 keY.
Numerical integration over six slabs results in AE = 1250 keV: Each

slab thickness corresponds to (Nt)i = (353/6) x lOIS atomslcm1. This
gives the energy loss in the flISt slab as .6.EI = ECEo)(Nt}. = (2.866 eV
cm'1l0" atoms) x [(353/6) x 10" atoms/em'] = 168.5 keY.
The energy at the end of the first slab is then EI = Eo _ !lEI = 2000 _
168.5 keY = 1832 keY. Energy Joss ill the second slab is evaluated at
this energy: dE, =E,)(Nt); = (3.051 'V cm'lI 0" .Ioms) x [(353/6) x
2
lOIS atorns/em ] = 179.5 keV. We get E2 = EI - .6.El = 1832 -179.5
keY = 1652 keY at the end of the second slab. Continuing the
procedure, the energy losses in the remaining four slabs are obtained
as dE, (i = 3-6) = 193.0, 210.3, 233.7. and 268.1 keY. Summing up,
the total proton energy loss is liE = 1:.1..& (i = 1--6) = 1253 keY.
dE
t.E I =-(E
dx ,- ,)&.:.I or
t.E; =
The thicknesses COrrespond to areal densities (see Table 2.1 for unit
conversions) of 17.6 x lOll and 353 )( 10'8 C atomslcm2, respectively.
2.2.7. Isotopes
The electronic stopping for a given ion X with an atomic
nwnber Z, at an energy EJA is the same regardless of ion mass
number A. The stopping of 2 MeV 2H ions is thus equal to the
stopping of 1 MeV IH ions.
For isotopic compositions in the target medium differing from

the natural composition, the stopping [MeV cm2/~g] is to be
approximated by scaling with the average mass ratio as follows:
stopping(mediwn[M!,,z,]) ~ stopping (mediUlll [M",z,D '(M.J
M!v)'
flE,X)
where M ,v is the average mass of the medium with the
natural isotopic composition.
Eo
2.3. STRAGGLING
When a beam of charged particles penetrates matter, the
slowing is accompanied by a spreading of the energy distribution
of the particles. This phenomenon is called energy straggling. It is
due to statistical fluctuations in the number of collision processes.
In materials analysis with ion bearns, straggling broadens the
measured 'energy distributions and resonances and impairs depth
and mass resolutiuns (Fig. 2.8).
X=R
2.3.1. Models
2.3.1.1. Bohr's theory

When the energy transferred to target electrons in the
individual collisions is small compared to the width of the energy
loss distribution, the distribution is close to a Gaussian distribution.
The condition for a Gaussian distribution can be formulated
(Besenbacher et aI., 1980) as
Nt[atoms/em']2: 2x1020
~(E[MeV/amU])'
Z,
E-
FIG. 2.8. The evolution of the energy disttibution as a :function of

depth as the ion traverses through a medium. The energy SPread
due to straggling increases with depth until me lowest-energy ions
reach the energy of maxirnwn stopping. Below this energy level,
the energy loss of the ions decreases with decreasing energy
leading to the "energy bunching" of the straggling distnbUtion.
(From Wilken and Fritz, 1976.)
(2.12)
ZI
2.3.1.2. Corrections to Bohr's theory, other models

Bohr's treatment has been improved by many authors to
extend its applicability to lower velocities of light ions (Z\ <:::: ~).
Lindhard and Scharff(195J) proposed a simple correction for ion
velocities below E[keV/amuj ~ 75/Z,
For thick targets, where the energy loss during penetration

exceeds 25%, the Gaussian approximation fails.
In the limit of high ion velocity, the energy loss is dominated
by electronic excitations. In this region, straggling is almost
independent of projectile velocity. In the Gaussian distribution
regime, the following simple expression was derived by Bohr
(1948)
n,'[keV'l =0.26Z/ Z,Nt[l 01 atoms/ern'],
Q' _ {O.5L(X),
_
Qn'
(2.13)
1,
for E[keV/amu1< 75Z,

for E[keV/amu1~ 75Z,
(2.14)
L(x) = 1.36Xl12 -O.016x'l2
where fiB 2 is the Bohr value for the variance (standard deviation)
of the average energy loss fluctuation (full width at half-maximum
heigbt (FWHM) ~ 2.355 rlj.
E[keV/amu]
(2.15)
25Z,
Bonderup and Hvelplund (1971), Chu (1976), Besenbacher el al.

(1980), Sigm1.md and Fu (1982); and Sigmund and Haagerup
(1986) presented more refined and extensive calculations. In the
non-Gaussian regions and for heavy ions, energy straggling has
been described in Livingston and Bethe (1937), Landau (1944),
Symon (1948), Vavilov (1957), Schulek I al. (1966), Tschalilr
(1968), Bichsel and Saxon (1975), and Sigmund and Schinner
(2002a,). For heavy ions, the charge exchange, in addition to the
target nonunifonnity effects, becomes increasingly important. The
EXAMPLE 2.8. From Eq. (2.13), we obtain (within 4% acCUI1lcy)

for 'He ioos: rl.'[keV'j " Z,Nt[IO iS atoms/em'j. This relation is
helpful for quick estimates of"He ion Bohr straggling.
19
Chapter 2
c~arge exchange effects have been treated, for example, in Sofield

(1990). No appropriate theory for slow heavy ions in light targets
exists.
~
Yang ef al. (1991),Yang and MacDonald (1993) and Yang

(1994), completed a survey of the theo-existing H, He, and heavyion straggling data and developed a fitting function for the Chu
'
(1976) model. Using the effective charge and scaling approach for
energy straggling and the Chu model and considering correlation
effects and charge change effects, they obtained an empirical
furmula fitting the data for heavier ion straggling.
Figures 2.9 illustrate He, Li, C, 0, CI, and I ion struggling in
carbon, silicon, germanium, tin and gold as calculated from the
..
empirical formula by Yang ef al. (1991), normalized to Bohr

struggling.
'.r'
EXAMPLE 2.9. Straggling of 5.0 MeV helium ions in gold.

Bohr straggling as calculated from Eq. (2.13) is n.'lNt ~ 82 keY'
2
1S
em II0 atoms. In a gold layer of 10 18 atomslcm2 (about 170 run),
this corresponds to a spreading in energy of ns ::: 9 ke V. Figure
2.9 gives om." 0.8 for He ions in gold at 5.0 MeY. A straggling
contribution of about 7 keY can thus be expected to the 5,0 MeV
helium beam energy resolution after transmission through a 170nmgoldfoil.
2.3.1.3. Straggling in mixtures and compounds
A linear additivity approach for treating compounds

(mixtures) si.milar to that for energy loss was proposed by Chu
(1976) for energy straggling according to the Bohr model.
AmBo (m + n = 1) with an atom
density N.... [atoms/cm'j and the atomic densities NA and N.
Consider a compound (mixture)
Assume straggling values of oA and OS for the elements in a layer

of thickness t Straggling nAB in the compound layer A",B, can
then be obtained from the expression
NABt
NAt
AB
Noting that mN
N.t
FlG. 2.9. Straggling of He, Li, C, 0, CL and I ion<; in caJbon, silicon,

gennaoium, tin, and gold as calculated from the empirical formula of
Yang ef al. (1991), relative to Bohr straggling. In the figure at left, the
order of the curves for He and Li ions is the same as that for C and 0
ions.
EXAMPLE 2.10. Bohr straggling of 'He ions in 10" atoms/em'
ofSi02.
We now have A.nBo = SilOlJ NS,it = 0.33 x NSiOz t = 0.33
atoms/cm and Not = 0.67 x NsiO::!
(2.16)
t=
10 18
0.67 x 10 18 atomslcm2.
Eq. (2.13) for Bohr straggling then gives
NA and nNAB = Ne this equation reduces to
(nAB)' =(OA)' +(06)'.
..... -nervY (M_V)
(nAB)' m(nA)' n(nB)'
---
..... _-----------
(2.17)
(n~)' ~ 0.33 x 0.26 x

4.80 keY'.
Z,'z, keY' ~ 0.33
x 0.26 x 4x14 keY'~
(n~)'~ 0.67 x 0.26 x

5.57 keY'.
Z,'z, keY'~ 0.67 x 0.26 x 4x 8 keV'~
From Eq. (2.17), we obtain
(n~o,), ~ (n~;)' +(n~)' ~ (4.80 + 5.57) key2~ 10.4 keY'

=> n~i02 =3.22keY.
20
Energy loss and Energy Straggling
2.3.1.4. Additivity of energy loss fluctuations

The different Gaussian contributions to energy loss fluctuation
can be added in quadrature. The energy resolution in a spectrwn,
for example, arises from several different contributions: the system
resolution nOET , energy straggling nSfRJ beam energy profile
OSEAM, and so on. These contributions give the tOlal variance of
the energy loss fluctuation
(2.18)
The same applies to contributions to energy loss fluctuation
originating from multilayered targets or from target thickness
variations. In the former case, the total variance of energy loss
fluctuation is !hOT2 = n/ + Q B2 + ... , where A and B stand for
successive layers in a medium penetrated by the ions. A
distribution in target thickness with a variance fit increases the
energy loss fluctuation by nl!.t2 = (dEJdx)\6ti (Besenbacher et aI.,
1980).
2.3.2. Comparisons to experimental data
Energy spreading in experimental spectra often results mostly
from contributions other than energy straggling. Nonuniformity of
targets (rough surfaces and interfaces) and correlation effects in
energy loss (Besenbacher et al., 1980; Yang et al., 1991) in solid
media tend to yield fluctuations in energy loss that are often much
larger than energy straggling. These effects are expected to apply
to routine ion beam analysis.
Qualitative agreement for light ion experimental data with the
predictions of Bohr (1948), Lindhard and Scharlf (1953),
Bonderup and Hvelplund (1971), and Chu (1976) is generally
observed, although data not corrected for the nommifonnity and
correlation effects can, in some cruses, exceed the straggling
predictions by hundreds of percent
2.4. RANGE
The definitions of the range parameters vary in the literatw'e.
high-energy region (above -I MeV/amu) by assuming S,(E')'"

become increasingly inaccurate.
2.4.1. Range calculations and simulations

In range theory, the range distnbution calculation is regcrCl..ed
as a transport problem describing the motion of the ions dur:i..::ng
their slomng to zero energy. The calculations can be brGlc:::ily
classified into analytic methods and simulation methods.
The analytic approach has been developed in Lindhard (I al.
(1963), Brice (1970), Sanders (1968), Winterbon el al. (1 97 0),
Sigmund (1969), and Littrnark and Ziegler (1980), as well as it "the
PRAL (projected Range Algorithm) code by Bieroack (198 ]),
Biersack and Ziegler (1982) and Ziegler el al. (1985). The l'R...AL
program, based on transport theory and the -stopping fonnalism
developed by Ziegler et al. (1985), is included in the SR..:rM
program package (Ziegler el al., 2008). Multilayered targets = o t
be treated by the transport theoretical calculations.
Simple universal expressions for ion projected ranges ~ in
the low-energy region, where nuclear stopping dominates, \;JI."rere
proposed by Winterbon el al. (1970) and Kalbitzer and Oet2Il1ann
(1980) and later refined, for example, by Gupta and Bhattacharya
(1984), Izsack el al. (1986), and Kido ., al. (1986).
The simulation methods are based on molecular dyr.atnics
(MO) or binary collision approximation (BCA) calculations. The
BCA simulation method involving Monte Carlo calculations has
developed into many well-established computer codes such as
SRlM (Biersack and Haggrruut, 1980; Ziegler et al., 1985;
Ziegler, 2004; Biersack et aI., 1991), CRYSTAL TRIM (POSselt,
1994), EVOLVE (Rousch ., al., 1981), PlBER (Adesida and
Karapiperis, 1982), TRIDYN (Moller and Eckatein, 1984), TCIS
(Cui and Li, 1985),
(Chou and Ghoniem, 1987), and
MARLOVIE (Robinson and Torrens, 1974)_ The simulation
approach can yield ranges and range distributions for multilaYered
targets over a wide range of Ion energies
mops
Usually, projected range (or penetration depth) is defined as the

mean depth from the target swface at which the ion comes to a
halt. Range along the path (or total path length) is the total mean
distance of the ion traversed along the trajectory. Because ion
propagation is not rectilinear, the former is smaller than the latter_
By range straggling, we mean the width of the ion range
distribution. The range distribution is further characterized by
higher moments, such as skewness and kurtosis_
2.4.2. Ranges in compounds

Accurate treatment of ion ranges in compoundc; requires
extensive calculations. The range can be approximated by
averaging the ranges for the ion in each element of the compound,
weighted by the relative composition
A,.,B,: RAE ~mRA + nRB, with m+nnomtaliz~d to unity. (2.20)

2.4.3. Isotopic effects
Approximations exist for treating isotope effectS. The range
The total path length can be defined by the integral

E
R(E)=
dE
IN[S,(E~+S,(E~]'
o.
For energies below the stopping maximum. these calculaic:::..:ns
for different isotopes of ions ~ X at energies E/A can be scaled

(2.19)
with the mass number ratio: Rangetz;X)
~ Range(~XHA.'/A).
The range of2 MeV 2H ions is thus twice the range of 1 MeV lH
where both the electronic and the nuclear contrloutions to the

stopping pow~r are considered. Estimates of the mean range along
the path can be calculated from this integral [Eq. (2.19)] in the
ions.
21
Chapter 2
For isotopic compositions in the target medium differing from

the natwal composition, range can be approximated from the
Bichsel, H., and Saxon, R.P. (1975), Phys. Rev. All, 1286.
relation Range(mcdium[M!.,z,lJ = Range(medium[M~,Z,lH
Biersack, J.P., and Haggmark, L.G. (1980), N"eI. lns!rum.

Methods 174, 257.
M!vJM:av), where Mav is the average mass of the medium with the
natural isotopic composition.
Biersack, J.P. (1981), Nuel. i/'lStrum. Methods 182/183, 199.

Biemack, J.P. (1982), Z. Phys.. A305, 95.
Biersack, J.P., and Ziegler, J.F. (1982), In Ion Implantation
Techniques (Ryssel, B., and Glawischnig, R, eds.),
Springer, Berlin, p. 157.
REFERENCES:
Adesida, L., and Karapiperis, L. (1982), Radiat. Elf. 61, 223.
AhIen, S.P. (I 980). Rev. Mod Phys. 52, 121.
Andersen, H.H., and Ziegler, J.F. (1977), Hydrogen Stopping
Powers and Ranges in All Elements, Pergamon, New

York.
Anthony, J.M., and Lanford, W.A. (1982), Phys. Rev. Al5, 1868.
Arista, N.R. (2002), Nuel. Ins!rom. Methods B195, 91.
Azevedo, G. de M., Behar, M., Dias, J.F., Grande, P.L., dos

Santos, J.H.R., Stoll, R., Klatt, Chr., and Kalbitzer, S.
(I 998), Nuel. ins!rum. Methods B13~138, 132.
Azevedo, G. de M., Dias, J.F., Behar. M., Grande, P.L., and dos
Santos, IH.R., (2001), Nuc!. Instrum. Methods B174,
407.
BillTIldas, N.P., Jeynes, C., Webb, R.P., and Wendler, E. (2002),
Nucl. Ins!rum. Methods 8194,15.
Bauer, P. (1987), Nud. instrum. Methods B27, 301.
Bauer, P. (! 990), Nuel. ins!rum. Methods B45, 673.
Berger, MJ., and Paul, H. (1995), "Stopping Powers, Ranges and
Straggling", in Atomic and Molecular Data for
Radiotherapy and Radiation Research, report IAEATECDOC-799, lAEA, Vienna, Austria, chap. 7, pp. 415545.
- ._.. Bernardi, F., Araujo, L.L., Behar, M., and Dias, J.F. (2006), Nuel.
Ins!rom. Methods B249, 69.
Besenbach.r, F., Andersen, J.U., and Bonderup, E. (I980), Nuel.
I/'IS!rum. Methods 168, I.
Bethe, H.A. (1930), Ann. Phys. 5, 325.
Bethe, H.A. (1932), Z Phys. 76, 293.
Bethe, H.A., and Heider, W. (1934), Proc. R. Soc. London A146,
83.
Betz, H.D. (1972). Rev. Mod Phys. 44, 465 ..
Biersock, J.P., Steinbauer, E., and Bauer, P. (1991), Nuel.lnstrum.

Methods B61, 77.
Bloch, F. (1933a), Ann. Phys. 16,287.
Bloch, F. (1933b),
z. Phys. 81, 363.
Bohr, N. (1913), Philos. Mag. 25, 10.

Bohr, N. (1915), Philos. Mag. 30, 581.
Bohr, N. (1948), Mot. Fys. Medd K Dan. Vidensk. Selslr. 18 (8),
24 (19).
Bonderup, E., and Hvelplwul, P. (1971), Phys. Rev. A4, 562.
Bragg, W.H., and Kleeman, R. (1905), Philos. Mag. 10,5318.
Brandt, W., and Kitagawa, M. (1982), Phys. Rev. B25, 5631.
Brice, D.K. (1970), Rodiat. Elf. 6, 77.
Brice, nK (1975), ''High Incident Ion Energies", vol. I of Ion

Imp/antation
Range
and
Energy
Deposition
Distributions. Plenum Press, New York.
Chan, E.K.L., Powers, D., Lodhi, AS., and Brown, R.B. (1978),
Appl Phys. 49, 2346.
Chou, P.S., and Ghottiem, M. (1987), Nuel. Instrum. Methods
B28,175.
Chu, W.K. (1976),Phys. Rev. A13, 2057.
Chu, W.K. (1979), in Accelerators in Atomic Physics (Richard, P.,
cd.), Academic Press, New York.
Cui, F.-Z., and Li, H.-D. (1985), Nuel. I/'IStrum. Methods B718,
650.
dos Santos, I.H.R. Grande, P.L., Boudinov, H., Behar, M., Stoll,
R., Klan, Chr., KaIbitzer, S. (1995), Nucl Instrom.
Methods B106, 51.
Fano, U. (1963), Annu. Rev. Nuel. Sci. 13, I.
Fixsov, O.B. (1959), Zh. Elcsp. Teor. Fiz. 36, 1517. See also:
Firsov, O.B. (1957), Zh. Eksp. Teor. Fiz. 32, 1464;
(1957), Zh. Elcsp. Tear. Fiz. 33, 696; (1958), Zh. Eksp.
Teor. Fiz. 34, 447.
Bianconi, M., Abel, F., Banks, J.e., Climent Font, A, Cohen, C.,
Doyle, B.L., Lotti, R., Lulli, G., Nipoti, R., Viclaidge, I.,
Walsh, D., and Wendler, E. (2000), Nucl. Instrum.
Methods BI61-163, 293.
Forster, J.S., Ward, D., Andrews, H.R., Bail, G.c., CostO, G.J.,
Davies, W.G., and Mitchell, LV. (1976), Nuel. I/'IStrum.
Methods 136, 349.
Bichsel. H. (1972), American Institute. 0/ Physics Han.dbook, 3rd

Ed. (Gray, D.E., ed.), McGraw-Hill, New York, pp. 8142-8-189.
22

Kumakhoy. M.A., and Kornarov, F.F. (1981), Energy Loss and J~n
Ranges in Solids, Gordon and Breach, New York.
Geissel, H., Weick, H., Scheidenberger, C., Bimbot, R., and

Gardes, D. (2002), Nucl. Instrnm. Methods B195, 3.
KiIrth, E., and Wedell, R. (1981), Phys. Lell. 86A, 54. See also:
Kiirth, E., and Wedell, R (1983) Phys. Status &>lidi
Gibbons, J.F., Johnson, W.S., and Mylroie, S.W. (1975), Projected
Range Statistics: Semiconductors and Related Materials,

2nd Ed., Halsted Pre;;s, Stroudsbury, FA.
B116, 585. KfirJi, E., WedeU, R., Semrad, D., and Bauer,
P. (1985) Phys. Status Solidi B127, 633.
Grande, P.L., and Schwietz, G. (2002), Nucl. instrnm. Methods

BI95,55.
Landau, L. (1944), 1. Phys. USSR 8,201.
Grcco, R., Luce, A, Wang, Y., and Shao, L. (2007), Nud.

insrrum. Methods B261, 538.
Lin, H.H., Li, L.W., and Norbeck, E. (1986), Nucl. Ins/rum.

Methods B17, 91.
Gupta, S.K., and Bhattacharya, P.K (1984), Phys. Rev. B29, 2449.
Lindbard, J., and Scharff, M. (1953), Mot. Fys. Mood. K Dan.

Videnuk Se/sk 27 (15).
Herau1~ J., Bimho!, R., Gauvin, H., Kubica, B., Anne, R., Bastin,
Lindbard, J., Scharff, M., and ScbWtt, H.E. (1963), Mat. Fys.
Medd. K Dan. Vidensk Se/sk. 33 (14). See also:
Lindbard, J. (1954), Mat. Fys. Medd. K Dan. Vidensk
Selsk 28, 8. Lindbard, 1., and Winther, A. (1964), Mat.
Fys. Medd. K Dan. Videnslt. Selsk 34, 4. Lindhard, J.,
Nielsen, v., and Schartl; M. (1968), Mat. Fys. Medd. K
Dan. Vidensk. Selsk. 26, to. LindhaTd, J., Nielsen, V.,
Scharff, M., and Thomsen, P.v. (1968), Mat. Fys. Medd.
K Dan. Vidensk. Se/sk 33, 10.
G., and Hubert, F. (1991), Nucl. Instrnm. Methods B61,

156.
Hoshino, Y., Okazawa, T., Nishii, T., Nishimura, T., Kido, Y.,
(2000) Nud. instrnm. Methods BI71, 409.
Hubert, F., FIeill)', A, Bimho!, R, and Gardes, D. (1980), Ann.
Phys. (France) 5S, 3.
Hubert, F., Bimbo!, R, and Gauvin, H. (1989), Nue/. Instrum.
Methods B36, 357.
.
Hubert, F., Bimbot, R, and Gauvin, H. (1990), At. Data Nucl.
Dala Tables 46, I.
Lindbard, 1, and Serensen, AH. (1996), Phys. Rev. A53, 2443.

Lirtmark, U., and Ziegler, I.F. (1980), Handbook of Range
Distributions for Energetic Ions in All Elements,
ICRU (1993), Stopping Powers and Ranges for Protons and

Alpha Particles, ICRU Report 49, ICRU, Bethesda, MD.
Pergamon, New York.
ICRU (2005), Stopping Powers of Ions Heavier than Helium,
Livingston, M.S., and Bethe, HA (1937), Rev. Mod. Phys. 9,245.
ICRU Report 73, Oxford University Press, New York; 1.

ICRU 5 (1), I.
Lulli, G., Albertazzi, E., Bianconi, M., Bentini, G.G.,
Nipoti~ R.,
Lotti, R. (2000), Nucl. Instnlm. Methods B170, 1.
Inokuti, M. (1971) Rev. Mod. Phys. 43,297.
Lulli, G., A1bertazzi, E., Bentini, G.G., Bianconi, M., Lotti, R
Izsack, K., Benhold, J., and Kalbitzer, S. (1986), Nud. inutrum.

Methods B15, 34.
(2002), Nucl. Instrnm. Methods B193, 103.
Jackson, 1.D. (1962), Classical Electrodynamics, Wiley, New
M.ynard, G., Zwicknage~ G., Deutsch, C, and Katsonis,K.

(2001), Phys. Rev. A63, 1.
YOtk, chap. 13.

Jackson, J.D. (1975), Classical Electrodynamics, 2nd Ed., Wiley,
Maynard, G., Sarrnzin, M., Ratsonis, K., and Dimitriou, K. (2002),

Nucl. Inutrnm. Methods B193, 20.
New York, chap. 13.
Mertens, P. (1987), Nuel. Instrum. Methods B27, 315.
Ianni, J.F. (1982), Atomic Data and Nuclear Data Tables 27, 147.
Moller, W., and Eckstein, W. (1984), Nud. instrum. Methods B2,

814.
Johnson, W.S., and Gibbons, I.F. (1970), Projected Range
Sraristics in Semiconductors, Stanford University Press,
Nonhcliffc, L.C. (1963),Annu. Rev. Nucl. Sci. 13, 67.
Stanford, CA.
Nonhcliffe, L.C., and Schilling, R.F. (1970), Nucl. Data Tobles 7,

233.
Kalbitzer, S., and Oetzmann, H. (1980), Radiat. Eff. 47, 57.
Kido, Y., Konomi, I., Kakeno, M., Yamada, K.., Dolurui.e, K, and
Kawamoto, J. (1986), Nud. Instrum. Methods B15, 42.
Nowick4 L, Turos, A., Ratajczak, R., Stonert, A., and Garrido, F.

(2005), Nucl. Insrrum. Methods B240, 277.
Kanae, G., Kalbizer, S., Klatt, C, Niemann, D., and Stoll, R.

(1998a), Nue/. instrum. Methods B138, 159.
Oddershede, J., and Sabin, J.R (1983), Phys. Rev. Len. 51, 1332.
See also: Oddershede, I, and Sabin, l.R. (1984), At. Data
Nucl. Data Tables 31, 275. Oddershede, 1., and Sabin,
I.R. (1987), Phys. Rev. A35, 3283. Sabin, l.R., and
Oddernhede, J. (1982), Phys. Rev. 26,3209. Sabin, J.R.,
and Oddershede, J. (1984), Phys. Rev. A29, 1757. Sabin,
Konae, G., Klatt, C, and Kalbizer, S. (1998b), Nucl. inutrum.

Methods B146, 106.
Kreutz, R, Neuwirth, W., and Pietsch, W. (1988), Phys. Rev. A.22,
2598.
23
Chapler 2
J.R., Oddershede, J., and Sigmund, P. (1985), Nud.

Inslrum. Methods B12, 80. Sabin, I.R., and Oddersbede,
Sigmund, P. (1975), ''Energy loss of charged particles in solids", in
Radiation Damage Processes in Materials (Du Puy.

CHS., ed), Noornoff, Leiden, The Netherlands, chap. I.
I. (1987), Nucl. Inslrum. Methods B24125, 339. Sabin,

I.R., and Oddershede, I. (1985), Int. J. Quantum. Chern.
Symp. 19, 733.
Sigmund, P. (1982), Phys. Rev. Al6, 2497. See also: Sigmund, P.,
and Winterbon, I<. (1985), Nucl. Instrum. Methods B12,
1.
Oddersbcdc, J., and Sabin, I.R. (1989), Nud. Instrum. Methods

842,7.
Sigmund, P. (2004), Stopping of Heavy Ions. A Theoretical

Approach, Springer Tracts in Modem Physics Vol. 204,
Paul, H., and Schinner, A (2003a), Nud. Jnslrum. Methods B209,

252.
Springer, Berlin.
Paul, H, and Schinner, A. (2003b), At. Data Nucl. Data Tables 85,
377.
Sigmund, P., and Fu, D.-I. (1982), Phys. Rev. AlS, 1450.
Sigmund, P., and Hangemp.U. (1986), Phys. Rev. A34, 892.
Paul, H, and Sehinner, A. (2005), Nucl. Instrum. Methods B227,

461.
Sigmund, P., and Schinner, A. (2000), Eur. Phys. J. D12, 425.

Sigmund, P., and Schinner, A. (2002a), Eur. Phys. J. D23, 20 I.
Paul, H. (2006), Nud. Instrom. Methods B247, 166.
Sigmund, P., and Schinner, A. (2002b), Nuel. lnslnim. Methods

BI93,49.
Paul, H., and Schinner, A. (2006), Nucl. Ins/rom Methods B249,

1.
Sigmund, P., and Schinner, A. (2002c), Nuel. Instrum. Methods

BI95,64.
Paul, H. (2007), Nud. Instrum. Methods B261, 1176.

Paul, H (2009a), "Some New Results on Stopping Power for Fast
Ion",
Sigmund, P., Fettouhi, A., and Schinner, A. (2003), Nuel. Inslrnm.

Me/hods B209, 19.
in Application of Accelerators in Research and
Industry: Twentielh International Conftrence. AlP
Sigmund, P. (2006), Panicle Penetration and Radiation Effects.

General Aspects and Stopping 0/ Swift Point Particles,
Springer-Verlag, Berlin.
COIlference Proceedings, American Institute of Physics,

New York, vol. 1099,pp. 251-254.
Paul, H. (2009b),Nuci.instrum. Methods B267, 9.
Sofield, C.J. (1990), Nuel. Instrum. Methods B45, 648 and

references therein.
Peltola, I., Nordlund, 1<., and Keinonen, I. (2004), Nud. lnstrum.

Methods B217, 25.
Steward, P. (1968), PhD Thesis, University of California,
Possel!, M. (1994), Radiat. Eff. Defects Solids 130/131, 87.
Berkeley, CA; Lawrence Radiation Laboratory Report

UCRL-18127.
Powen>, D. (1989), Nucl. Instrum. Methods B40/41, 324.
Symon, I<.R. (1948), PhD Thesis, Harvard University, Cambridge,
Robinson, M.T., and Torrens, I.M. (1974), Phys. Rev. B9,5008.
MA.
Rausch, M.L, Andreadis, T.D., and Goktepe, O.F. (1981), Radiat.

Eff. 55, 119.
Thwaites, D1. (1985),Nuc/. Inslnim. Methods BU, 84.

Thwaites, D.l. (1987), Nucl.lnslnim. Methods B27, 293.
Sabin, lR., Oddershede, I., and Sigmund, P. (1985), Nucl.

Jnstrum. Methods B12, 80.
Tschaliir, C. (1968), Nud. Jns/nim. Methods 64, 237. See also:

Tschallir, C (1967), PhD Thesis, University of Southern
California, Los Angeles, CA. Tschallir, C (1968), Nuel.
lmtrum. Methods 61, 141.
Sanders, I.B. (1968), Can. J. Phys. 46, 455.

Shao, L., Wang, Y.Q., Nastasi, M., and Mayer, I.W. (2006), Nuc!.
lnslnim Methods B249, 51.
Vavilov, P.V. (1957), Sov. Phys. JETP 5,749.
Sharma, A., Fettouhi, A., Schinner, A., and Sigmund, P. (2004),

NucI. Instrum. Methods B218, 19.
Wbaling, W. (1958), 'The energy loss of cbarged particles in

matter", in Handbuch der Physik (Fliigge, S., cd),
Springer, Berlin,Bd. XXXIV, pp. 193-217.
Schulek, P., Golovin, B.M., Kulyulcina, L.A, Medved, S.V., and

Pavlovich, P. (1966), SOY. J. Nuc/. Phys. 4, 996.
Wilken, B., and Fritz, T.A. (1976), Nud. Instrum. Methods 138,
331.
Sigmund, P. (1969), Phys. Rev. 184,383.
24
WinterbOll, K.B., Sigmund, P., and Sanders, I.B. (1970), Mat. Fys.
MeM K Dan. Vidensk. Se/sk. 37 (14).
Winterbon, K.B. (1975), "Low Incident Ion Energies", vol. 2 of
Ion Implantation Range and Energy Deposition
Distributions, Plenum Press, New York.
Yamamoto, Y., Ikeda, A., Yoneda, T., Kajiyama, K., and Kido, Y.
(1999), Nucl. Instrum. Methods B153, 10.
Yang, Q., (1994), Phys.'Rev. M9, 1089.
Yang, Q., O'Connor, D.J., and Wang, Z. (1991), .'Iue/. Ill.'Itrum
Methods B61, 149.
Yang, Q., and MacDonald, R.J., (1993), .'Iuel. Iostrum. Methods
B83,303.
Ziegler, 1.F., and Chu, W.K. (1974), At. Dolo .'Iuel. Dala Tables
13,463.
Ziegler, IF. (1977), Helium Stopping Powers and Ranges ir-.:2 All
Elemental Matter, Pergamon, New York.
Ziegler, J.F. (1980), Handbook of Stopping Cross-Sections i"
Energetic Ions in All Elements, Pergamon, New Y011:::.....
Ziegler, 1.F., Biersack, J.P., and Littmark, U. (1985), The Slop~ing
and Range of Ions in Solids, Pergamon, New yo~ vol.
1.
Ziegler, 1.F., and Manoyan, I.M. (1988), Nue/. Jnsrrum. M,".hods
B35,215.
Ziegler, 1.F. (2004), Nucl. Instrum. Methods B219-220, 102.
Ziegler, J.F., Biersack, J.P., and Ziegler, M.D. (2008), The
Stopping and Range of Ions in Matter, LuLu Press,
Morrisville, .
NC.
See:
http://www.srim.org/SRIM%20Book.hlrn.
25
CHAPTER
THE INTERACTION OF CHARGED

PARTICLES WITH NUCLEI
A. F. Gurbich
Institute of Physics and Power Engineering, Obninsk, Russia
Contributor: Y. Q. Wang
CONTENTS
3.1 INTRODUCTION ............................................................................................................ 28
3.2 FORMAL THEORY OF ELASTIC SCATTERlNG ...................................................28
3.3 MECHANISMS OF THE PROJECTILE-NUCLEUS INTERACTION ..................32
3.4 PHENOMENOLOGY ..................................................................................................... 35
3.5 EVALUATION OF CROSS SECTIONS ......................................................................37
3.6 KINEMATICS OF NUCLEAR REACTIONS .............................................................37
3.7 INTERNET -BASED NUCLEAR PHYSICS RESOURCES RELEVANT TO rnA .39
3.8 EXAMPLES
3.8.1 EXAMPLE 3.1. Identification of peaks in the d + 27AI spectrum ................... 40
3.8.2 EXAMPLE 3.2. Resonances in the '7AI(p,y)'8 Si reaction .................................40
9
3.8.3 EXAMPLE 3.3. Resonances in the nuclear reaction IHC F,ay)"O and
19F(p,ay)160........................................................................................................... 41
3.8.4 EXAMPLE 3.4. Kinematic transformations for the 'He + d --+ p + 4He +
18.35 MeV reaction ............................................................................................. .41
3.8.5 EXAMPLE 3.5. Conversions between 3Ue(d,p)4Hc and 'U('He,p)'He ...........42
REFERENCES ................................................................................................................... .42
27
Chapler 3
Nucleons are held together inside nuclei by nuclear forces.

These forces are strong attractive forces acting only at short
distances. They exhibit the property of saturation,. and as a result,
nuclear forces are attributed exchange character. Nuclear forces
depend on spin, not on electric charge, and they are not central
forces. The nature of the nuclear forces has not yet been fully
clarified.
3.1 INTRODUCTION
Pioneering achievements in ion beam analysis (lBA) were
made by nuclear physicists who developed a promising
application of existing instruments and knowledge to materials
investigations. Currently, IBA is a routine technique in materials
analysis, and few people using it in analytical work have even an
undergraduate education in nuclear physics. Nevertheless,
except for Rutherford backseattering spectrometry (RBS) and
particle-induced X-ray emission (PIXE) spectroscopy, all other
IBA methods are based on nuclear interactions, so at least a
basic knowledge of the underlying physics is needed to
competently employ these methods in practice.
The short range of strong nuclear forces leads to a strict

demarcation between the regions where only long-range
Coulomb forces or only nuclear forces operate, as the latter are
at least 100 times greater than Coulomb forces at shan distances
of about 1 fin.
The fundamental results in nuclear pbysics relevant to IBA

were .obtained in the 1960s, and now, this field is regarded as
"mature". The mechan.lsms of projectile-Ducleus interaction are
known, and appropriate models have been developed. However,
it appears that, beyond Rutherford scattering, nuclear physics
theory is unable to calculate a priori the cross section of such
interactions with an accwacy acceptable for IBA Moreover,
even such an important quantity for rnA as the energy at which
the cross section begins to deviate from the Rutherford value
cannot be reliab[y predicted (Gurbich, 2004).
Assuming that a nucleus is a uniformly charged spbere of

radius R, the dependence of the electrostatic potential energy on
the distance r for the projectile-nucleus system is
Zze'
Vc(r)=
~'( r')
- - 3--
2R
R'
~R
for
for
r.:s:; R
(3.2)
where Z and z are the charge numbers of the nucleus and

projectile, respectively. For the attractive forces, the potential
energy is negative, and the nuclear potential can be represented
as a potential well that is about 40-50 MeV deep. For the
combined Coulomb and [Juclear potential, the transition from
repulsion to attraction proceeds in a narrow region in the vicinity
of the boundary of the nucleus (see Fig. 3.1).
Although nuclear physics theory cannot provide sufficiently

accurate cross-section data when the calculations are based
simply on fIrst principles, a particular cross section can generally
be reliably represented by adjusting model parameters. Thus,
theory provides a powerful tool for data evaluation based on
critical analysis of the available experimental results within
theoretical constraints. The evaluation of nuclear data is a
routine procedure in preparing neutron data for different
applications. The same approach can be successfully applied to
charged-particle cross sections.
10
T .. ---... ----
The nuclear physics concepts and models relevant to IBA

are discussed in this chapter. This chapter begins with a
>QJ
consideration of the formal quantum mechanical theory of

scattering, followed by a review of the mechanisms of
projectile-nucleus interaction. Models for dealing with differen1
processes are then introduced, and their application to crosssection evaluation is presented. Some basic reaction kinematics
is also included, but comprehensive discussions on reaction
kinematics are deferred to Appendix 4. In addition, the sources
of nu~lear data are indicated. Finally. a few examples are given
at the end of the chapter to show the relevance of these nuclear
concepts and models to real IBA applications.
15
r, fm
-10
-30
40
FIG. 3.1. The combined Coulomb and nucIear
potential of a nucleus. The classical turning
points for a panicle with kinetic energy T are
located at rl and r2.
3.2 FORMAL THEORY OF ELASTIC SCAlTERING

An atomic nucleus is a strongly bound system of nucIeons
located in a small domain with a typical size of
R,,(l.!-1.5)A w fm (1 fm= [0-13 em),
(3.1)
For a charged projectile to Teach the range of action of

nuclear forces, it should have some kinetic energy T sufficient to
overcome the Coulomb potential bamer of height
where A is the mass number.
28
The Interaction at Charged Particles with Nuclei
Zze'
=-
For a single immovable nucleus placed in the path af2L.:l ion

beam of intensity equal to N particles per square centimeter- in 1
s, the number of the lons scattered in the angle interval from.. 8 to
e + de is dN = 21tbdbN~ where b and db are derived fioICl.. Eq.
(3.5). The relative number of particles scattered into a given
solid angle is equal to the area of the ring with radii b and b -+ db
surrounding the scattering center
(3.3)
which is on the order of 1 MeV even in the interaction of singly

charged particles with the lightest nuclei.
According to quantum mechanics, a particle with kinetic
energy below the potential energy barrier height can penetrate
through the barrier, with the transparency of the barrier being
given by the fannula
dN
da= - = 27tbdb
N
(L6)
This is the differential cross section that is expressed ::for a

target containing n nuclei per unit area by the weU-kumVD
Rutherford formula. For the center-of-mass frame of referen.ce, it
is written as
(3.4)
where j.l = Mml(M + m) is the reduced mass and rl and r2 are the
classical twning points (see Fig. 3.1).
da = n(Zze
Thus, although the electric charge of atomic nuclei hinders

the initiation of nuclear reactions with low-energy charged
particles, such reactions are still feasible at energies below the
potential barrier.
dQ
(1. 7)
mv' 4sin4!!.
2
The conversion of this equation into a more practical farm
in the laboratory frame of reference is discussed in Appendix 4
(EXAMPLE A4.1).
If a projectile-nucleus interaction is conditioned solely by

long-range electric forces, then the differential cross section of
an elastic scattering event is derived from the conservation of the
energy and angular momentum (see Fig. 3.2 for the scheme of
the scattering).
At short distances on the order of the range of the nllclear

forces, simultaneous knowledge of the momentum and the
impact parameter of the scattering particle is impossible
according to Heisenberg's uncertainty principle. Consequently,
in principle, nuclear forces cannot be incorporated into any
. theory in which the trajectory of the particle is assumed to be
determined. Thus, classical physics is inapplicable, and quantum
mechanics must be used.
'
iJ
.. i<'
+~ ____ ~-l\-=~~~~ _________ .
+Ze
To explain the basics of this approach., nuclear scattering is

considered frrst for the simplest case of a spinless projectile with
no charge. According to quantum mechanics, the state of a
particle is described by the wave function'll, which is obtained
as a solution of the wave equation. For the case of elastic
scattering of spinless nonidentical particles, the wave equation
takes the form of the Schr6dinger equation with a spherically
symmetric potential V(r)
2m
~'V+h'(E-V)\jI=O
FIG 3.2. Scheme of projectile scattering by
the electric field of the nucleus.
(3.8)
where E is the particle energy.

Prior to scattering, the wave function 'V far a particle with a
given momentum p has the fonn of a plane wave:
As long as the interaction law (Le., the dependence of the

force on the distance) is known. it is possible to find .the
dependence of the scattering angle on the impact parameter b,
which is expressed in the nonrelativistic case by the relation'
where k is a propagation vector
tan e = 2Zze'
mV'b'
k=-=-
'V = e. ik:,
(3.5)
P
tz
29
(3.9)
);.
(3.10)
Chapter 3
Here,
A.:::: IJ2n, where J.... is the de Broglie wavelength.
This function is a solution of Eq. (3.8) in the case ofV(r) ~

0, that is, an equation of the form
2m
tI'
L'l1jI + -EIjI =
(3.11)
that is normalized to correspond to the flux density being equal

to the projectile velocity.
In the course of scattering, the plane wave interacts with the

field of the nucleus VCr), giving rise to a spherical wave
divergent from the center of the interaction. This wave has the
form
;kr
1(9)~
(3.12)
Thus, the last stage of the scattering process (after

scattering) is depicted by a superposition of the two waves, the
plane wave and the spherical wave, as
FIG. 3.3. Illustration oftbe definition
of the differential c:;ross section.

or
(3.16)
(3.13)
Here, 9 is the scattering angle, ftS) is the amplitude of the

divergent wave, and the liT factor represents the decreasing of
the flux in inverse proportionality to the square of the distance.
By definition, the differential cross section, do, is equal to

the fraction dN IN of projectiles scattered into the given solid
angIe dfl. Assuming a unity density of panicles in the primary
beam, the flux density is N = V[S-l cm-21, where v is the particle
velocity. The number of particles dN traversing the surface
element dS per unit time is determined by the probability of
- . - finding particles in the elementary volume dV ~ vi' sin e de dq>
(see Fig. 3.3), and the probability density is given by the square
of the modulus of the scattered wave function:
dN= 1(9)=-;10/' vr'sin9d9d<p
(3.14)
r'v
LA,P'(cos9)Rk/(r)
,.,
(3.17)
where ~(r) is a radial wave function and P,(cos 9) is a Legendre

polynomial [p, ~ I, P, ~ cos 6, P, ~ (3cos e - 1)12, etc.].
Far from tIle center of scattering (at large distances r), the
radial function for each component I can be represented in the
form of two partial spherical waves, one of which is convergent
lU-1"i)] and the other of which is divergent [e{ kr-I~)].
For the initial stage depicted by a plane wave, the two

waves have equal amplitudes and
(3.18)
Thus, the plane wave expressed through an expansion over
Legendre polynomials has the form
1/(9)I'r vdQ =1/(9)1' dQ
1jI=
[e -{
Taking into account the facts that the velocity does not
change in elastic scattering and that sin ede dcp = dn, it is
apparent that the differential cross section is expressed through
the square of the modulus of the amplitude of the scattered
wave:
dN
do=
N
Thus, the angular distribution of the scattered particles is

defined by the f{9) function. For quantitative analysis of the
elastic scattering, Eqs. (3.8) and (3.11) are considered in
spherical coordinates. The general solution of these equations
has the form
(3.15)
ih
~(2/+1)i'
[{a-,~)'-e +-I~)l
.
~(cos9) e
'
e = 1..;,
,...
30
21kr
(3.19)
The Interaction of Charged Partides with Nuclei
Here, each of the spherical waves c:orresponds to particles

moving with a given orbital momentum I and is characterized by
the angular distribution P,(cos 8) (see Fig. 3.4).
V(r), 'l'(r)
olk
,
------j'
..... -.
..,'/
FIG. 3.4. Angular distributions for

the fust three partial waves.
FIG. 3.5. Fonnation of the wave phase shift.
In the process of scattering, an additional divergent

spherical wave arises. As a result, the ratio between the
convergent and divergent waves changes. The change of the
ratio can be formally taken into account by a coefficient at the
divergent wave
To summarize, the differential cross section for eLastic

scattering is calculated from Eq, (3.16), with the scattering
amplitude being expressed through phase shifts 01 according to
Eq. (3.24). The phase shifts for partial waves are calculated by
solving the Schrodinger equation [Eg. (3,8)] with an assumed
potential V(r). Thi.s equation is split into angular and radial parts.
The asymptotic general solution for the radial equation is
(3.20)
In case of the elastic scattering, the fluxes for the convergent

and divergent waves should be equal to each other for each value
of 1. This means that !SI!2 = I, so the factor S[ can be written as
(3.25)
The phose shifts o[ are defined by the boundary conditions
and are functions of k and 1but do not depend on the scattering
angle.
(3.21)
where 0, is called the phase shift.
If the projectile is charged. it interacts with the comb ined
Coulomb and nuclear fields of the target nucleus. The relation
Physically, the phase shift is explained by the difference in

the wave velocity in the presence of the nuclear force field and
outside the nucleus, as illustrated in Fig. 3.5.
for the scattering ~plitude is then

(3.26)
The partial wave after scattering then has the form
Rk/(r) ~ e{ kr- 1
i+
25r
-e -{ kr-/i)
where fc(6) and
0'1
are the amplitude and phase shift, respectively,
of the Coulomb scattering.
(3.22)
The S, values defined by Eq, (3.21) can be considered as

elements of some diagonal matrix. which is called the scattering
matrix.ln the case of pure elastic scattering, the phase shifts ~ are
real numbers. However, they become complex if inelastic
scattering is also present in the scattering process. This
corresponds to decreasing of the amplitude of the divergent Waves
(i.e., IS,I < 1).
The solution of Eq. (3.8) for the final stage of scattering is
(3.23)
In the case that a projectile has a nonzero spin, the entire

scheme described so far remains valid. However, the equ~tions
become more complicated because the radial wave equation splits
into (2s + 1) equations. Suppose the projectiles are protons with
spin Yl. Then, the spin of the, bombarding particle can be
The relation between the scattering amplitude and phases

can be shown to be
(3.24)
31
Chapter 3
combined with angular momentum 1 in two ways, resulting in total

angular momentum values ofj = I Y:z..
The proton elastic scattering differential cross section is
obtained in this case by solving the Schr5dinger equation for
partial waves as do/elf! =
1A(9)1' + IBee)I'. with the scatlering
amplitudes A(9) and B(9) defined by the relations
A(e) =fc(9)+~ I[(/+IW +IS,- -(21 +1)]exp(2icr,)P'(cos9);

21k ''''
According to a general result of quantum mechanics, the

energy of particles moving in a finite domain of space can assume
anJy discrete values. Consequently, all nuclei have discrete energy
levels, as shown schematically in Fig. 3.6, and so, for light nuclei,
the cross section for this mechanism has a resonance structure. For
heavy nuclei, the level density is high. and the levels overlap at
relatively low excitation energy, where the level width r becomes
greater than the distance D benveen levels. This results in a
smoother dependence of the cross section on energy. In the
intermediate case, when r - D, the cross section exhibits a
strongly fluctuating structure (so-called Ericson fluctuations;
Ericson and Mayer-Kuckuk, 1966). Because of the long lifetime

and the uncertainty relation (written in energy-time coordinates as
6E . L\.t 2: Ii). the widths of the compound nucleus levels are
(3.27) and (3.28)
where fdO) is the amplitude of Coulomb scattering, crl represents
the Coulomb phase shifts, p[(cos 6) represents the associated
Legendre polynomials,
st
and s~
rather small, as are the 'Widths of the resonances obsen'ed in the

cross section. Because of the interference bet\.veen potential and
resonance scattering, the excitation function has a typical Structure
with resonances pictured as dips and bumps rather than as Breit-
Wigner functions (Fig.3.7).
are scattering matrix
elements for different spin orientations, ~d k is the wave
number.
The only unknown~ in the cross-section relations are the
phase shifts 51. and as soon as they are determined, the cross
section can be -calculated. The difference between classical and
quantum mechanical results for the scattering process is striking:
In the classical approach, a particle having zero impact parameter
scatters straight back toward its point of origin, whereas the
angular distribution for the corresponding partial wave with zero
angular momentwn is isotropic (see Fig. 3.4).
E,
E,
..
JD
:=:=:=--:-~='\
'\
",t)
E,
E,
3.3 MECHANISMS OF THE PROJECTILE-NUCLEUS

INTERACTION
The mechanism of scattering considered so far is called direct
or potential scatlering because it proceeds through the direct
interaction of a single bombarding particle with a potential well
representing a nucleus. Nuclear interactions at low energies can
also proceed in two stages through the mechanism of a compound
nucleus. The first stage of the interaction is the absorption of the
bombarding panicle by the target nucleus and the production of an
intennediate, or compound, nucleus. The compound nucleus is
Light
Heavy
nucleus
nucleus
FIG. 3.6. Diagrams of the energy levels for light and
heavy nuclei and the corresponding cross section
behavior.
always highly excited because the absoroed particle brings both its
Even though the excited levels overlap for the compound
nucleus formed in the interaction of protons with a mediumweight or heavy nucleus, there are still some strong sharp
resonances in the cross sections. These isobaric analogue
resonances (lARs) are caused by the population in the
compound nucleus of so-called isobaric analogue states.
Except for electrical forces, the n-n, n-p, and p-p interactions
are similar. Therefore, the energy levels of isobaric (equal-A)
nuclei are relatively insensitive to the interchange of a protoD
and a neutron. The isobaric analogue states in the (Z+I,N)
kinetic energy and the bond energy of the absorbed nucleons into
the nucleus produced. The second stage is the decay of the
compound nucleus with the emission of some particle. The
original particle can always be emitted, in whicb case the original
nucleus is re-formed. A typical lifetime for a compound nucleus is
14
_10- 5, which is very long compared -with the time of direct
interaction defined as the time (10-23_10-21 s) needed for the
bombarding particle to pass through the region occupied by the
potential well of the nucleus.
32
The Interaction of Charged Particles with Nuclei
nucleus will have the same properties as the energy states in

the (Z,N+l) nucleus, but will be shifted to higher energy by an
amount equal to
(3.29)
because of the additional Coulomb energy, SEc, associated
with the extra proton, less the neutron-proton mass difference
(Fig. 3.8).
1000
750
500
~
250
Potential Scattering
C 750
0
500
"'"
250
0
750
500
250
0
'"'"
Ie
()
J"
Breit-Wigner Resonances
Z,N+l
~"
FIG. 3.8. Scheme showing isobaric analogue

states.
Total
r~
0.5
1.0
1.4
2.0
1.5
Energy, MeV
1.2
FIG. 3.7. Interference of the potential and resonance

scattering for 12C(p,Po)12c. Note that the resonance
dramatically affects the cross section at energies much
below the Coulomb barrier.
1.1
13
0.9
"C
0.8
~-
0.
I'1A
1'1
1'11'1
natcr(p,po)
0.7
0.6
B'ab =141
0.5 L -_ _ _
2.65
In addition to elastic scattering. which is characterized by

the conservation of kinetic energy and the same types of
particles in the entrance and exit channels, different types of
nuclear reactions can take place. The possibility for some
reaction to occur is governed by the conservation laws. These
are the conservation of the number of nucleons, the electric
charge, the energy, the angular momentum, and the parity.
Application of the conservation of energy gives the Q value of
the reaction
j
i
o~ 1.0 F--zf'-'::"""'--':""=';""~4..4~""
464
In most cases, it is because of lARs that the elastic

scattering cross section departs from the Rutherford value for
medium and heavy nuclei at sub-Coulomb energies. If the
resonance structure of the compound nucleus is known, these
resonances can be taken into account in calculations of the cross
section (Fig. 3.9).
Q~ (M + M,)c' - (M, + M,)c' ,
Wylie and Zamboni, 1971

--Theory
1.3
~~-
2.70
_ _-~-~--~
2.75
2.80
Energy, MeV
FIG. 3.9. Isobaric analogue resonances in proton
elastic scattering from chromium (Nud. Instr. and
Meth. B217 (2004) 183).
respectively. The precise values for nuclear masses can be
found in Audi et al. (2003). Q-value calculators based on these
data
are
available
on
the
Internet
at
http://www.nndc.bnLgoviqcalc/ and http://nucleardata.nuclear.
Iu. se/database/masses/ .
(3.30)
where Mj, M2, M3, and M4 are the masses of the projectile,
target nucleus, outgoing particle, and residual nucleus,
Obviously, for elastic scattering, the Q value is zero. If Q

is negative, the reaction is endoergic with a threshold and can
33
Chapter 3
proceed only when the projectile energy exceeds this threshold.

For positive Q, the reaction is exoergic and can take place in
principle at any energy. If the energy of the projectile is
transferred to one nucleon or to a small group of nucleons, the
reaction is direct; alternatively, if the energy of the projectile is
transferred to all of the nucleons, the reaction proceeds through
the compound nucleus mechanism. When a particle or a
gamma ray is emitted from the compound nucleus, the residual
nucleus can stay both in the ground state and in an excited state
provided that the projectile has enough energy.
For non-elastic processes induced by protons, the most
important ones for IBA are inelastic scanering, (p,o.), (p,y),
(p, ay), and (p,p'y) reactions. Inelastic scattering is possible
when the projectile energy in the center-of-mass 'system
exceeds the excitation energy of the first level in the target
nucleus. Additional peaks are then observed at energies lower
than the elastic peak energy in backscattering spectra. For
inelastic proton scattering groups, the outgoing proton energies
are calculated using the general inelastic collision kinematic
relations of Appendix 4 but with the Q values replaced by the
level energies of the target nucleus taken with a negative sign.
E, ,....,..:.:M:;,,- Eo +Q,
M,+M 2
where
(3.31)
Eo is the initial proton energy; Q is the reaction energy;
and M1 and M2 are the projectile and target nucleus masses,

respectively. The energy for the emitted primary y-rays is
calculated as
E(i) =E _E(i)
1
x
level'
(3.32)
where Elevel is the energy of the excited state in the residual

nucleus. In addition to the primary gamma quanta, a cascade of
y-rays corresponding to the transitions between the levels of the
residual nucleus is also emitted. Their energies are calculated as
the differences between the respective level energies. The
branching ratios and other detailed information on levels and
gammas can be found in the literature and in the electronic
database at http://..........,..-y-nds:iaea.orglnudat2/. Information on
nuclear structures is also available from different sources
(Nuclear Data Sheets; Ajzenberg-Selove compilation for A = 320 in many issues of Nuclear Physics and at
http://www.tunl.duke.eduINuclDatal; End' and Van der Leon
compilation for A = 21-44 in several issues of Nuclear Physics;
Firestone e.t aI., 1996) and from the Evaluated Nuclear Structure
Data File (ENSDF) database at http://www.nndc.bnl.gov/ensdfl.
It should be noted that the data presented in various sources
might be slightly different because each of the data files is
obtained as a result of the analysis of many sets of experimental
data that are not necessarily consistent and the evaluation
depends on the approach employed.
(p,a) reactions are mainly exoergic (Q - 1-3 MeV).

However, because of the high Coulomb barrier for a particles,
the cross section is not large, and for (p,n) reactions that lead
to excited states in the residual nucleus, the cross section for
low-energy protons is, as a rule, negligible.
The (p,p,),) reaction (see the scheme of the reaction in Fig.

3.11) is possible only if the projectile energy exceeds the first
energy level in the target nucleus. This type of reaction is
actually inelastic scanering accompanied by y-emission from the
residual nucleus. Reactions such as (p,ay) and (p,ny) proceed in
a manner similar ro that shoVID in Fig. 3.11, except that another
particle instead of proton is emined from the compound nucleus.
r
.'.
For reactions induced by deuterons, three mechanisms

contribute to the total amplitude of the process: stripping,
resonance, and statistical. Stripping is a direct process that
proceeds as follo-ws: Because of electrical forces, a deuteron is
oriented in the course of the interaction in such a way that the
proton is farther from the nucleus than is the neutron. Because
the deuteron binding energy is small (-2.2 MeV), the neutron
can be absorbed by the nucleus, while the proton keeps moving
in a fon.vard direction. The statistical mechanism corresponds to
the case when the compound nucleus levels completely overlap.
It is assumed that the pro~ctile transfers its energy to all of the
nucleons of the nucleus. As a result, the nucleus becomes
''heated'' to some temperature. it exists in the excited state for a
relatively long time until energy is concentrated onto one of the
nucleons (or clusters), which "evaporates" from the nucleus [see
details of the theory in Vogt (1968), for example].
A,Z
FIG. 3.10. Scheme of a (p,y)

reaction.
In a (p,y) reaction, y-rays of different energies are emitted
corresponding to transitions on different levels of the residual
nucleus (Fig. 3.10). Because, in addition to its kinetic energy, the
proton brings to the compound nucleus its binding energy
(which is, on average, ~8 MeV), the compound nucleus becomes
highly excited. The excitation energy Ex is
34
The Interact/on of Charged Particles with Nuclei
3.4 PHENOMENOLOGY
The interaction of a projectile with an atomic nucleus Car::::::l, in
principle, be descnbed theoretically as the sum of eleme .:tary
interactions between an incoming particle and each of the
nucleons comprising the nucleus. However, even assuming that
the potential of the scattering of ODe nucleon by the ~ther
nucleon were known (which is not actually the case), such a
description would encounter the insoluble many-body problex:n.
Therefore, it is practical to apply a phenomenological

approach that consists of considering the projectile interac tion
with the nucleus as a whole, where the nucleus is represented by
an appropriate potential. The potential parameters are found by
fitting theoretical calculations to the available expe-rimental
data. To make this approach more physical, the potential shape
is derived from the known features of the nucleon-nue leon
interaction and from distributions of matter and charge ill the
nucleus.
A,Z
As stated in Section 3.2, knowledge of the phase shifts

completely determines the elastic scattering cross section. There
are different pheno-rnenological approaches to deriving phase
shifts from experimental data. For example, phase shifts can be
found by directly fitting the cross-section relations (phase-shift
analysis); by solving the SchrOdinger equation for partial waves,
with the parameters of the nuclear potential being free; and
within the framework of the R-matrix theory (Lane and Thomas,
1958) by adjusting compound nucleus ievel parameters and
boundary conditions.
FIG. 3.11. Scheme ofa (p,p'y) reaction.

Because the binding energy for deuterium is only -1
MeV/nucleon whereas it is -8 MeV/nucleon for other nuclei, the
binding energy of a deuteron in a compound nucleus appears to
be about 14 MeV. Consequently, the excitation energy of a
compound nucleus formed by the absorption of an accelerated
deuteron is high, and so, various reactions are possible and are
highlyexoergic.
When alpha particles are used as projectiles, the gammaemission mechanism resembles that for protons. Because alpha
particles are strongly bound systems, the reactions with nucleons
in the exit channel are endoergic. Thus, the compound nucleus
produced as a result of the capture of a low-energy alpha particle
can decay mainly by relaxing back to the elastic channel or by
emitting gamma rays. The threshold for the (a,p) reaction is
usually greater than 1 MeV, and because of the Coulomb barrier,
the cross section for the (a,p) reaction is, as a rule, small.
Different mechanisms contribute to the alpha elastic scattering
cross section beyond the energy ~egion where it follows the
Rutherford law. These are direct (shape elastic) scattering,
compound elastic scattering, resonance scattering of a djfferent
origin. and exchange processes consisting of the exchange of
nucleons between the alpha particle and the target nucleus in the
course of scattering.
In the case when the nucleus is represented by means of a

potential well, only a few parameters with slight (if any)
dependence on energy are used. The potential is real if elastic
scattering is the only channel and complex if absorption take
place (optical model; see Hodgson, 1963). In the general case,
the potential is of the form
Finally, it is worth noting that, for any reaction except for

elastic and inelastic scattering, the residual nucleus might be
radioactive. Every radioactive isotope bas its 0\\10 unique
characteristics: a half-life and the energy of the decay products.
Thus, provided that the half-life is detennined, the decay
products are known, and their energy is measured, the atomic
number and atomic mass can be identified for the isotope. This
is used in charged particle activation analysis (see C~apter 10).
Uso
(3.3 3)
where U c is the Coulomb potential defined by Eq. (3.2)
U R(r) = -V.I. (r)

U 1 (r) = 4a,WD
d/,(r)
dr
(3.35)
--
=( ~)' V
! d/"
d
I.s
/R(r)=[l+exp(r:~,
Jr
mnc
R , =rAlfl
,
35
(3.34)
so
(3.36)
(337)
(3.38)
Chapter 3
The potential terms represent, in sequence, the real central

volume potential of depth V It. the imaginary central surface
potential of depth WD (volume absorption is negligible at low
energies), and the surface spin-orbit potential of depth Vso; flr)
is the Saxon-Woods form factor; Rx is the half-value radius; Rx
is the diffuseness parameter; A is the target mass number; mit is
the It-meson mass; c is the velocity of tight; and I and s denote
the angular momentum and spin operators, respectively.
:a
~
400
160
(p,PO)"O ,",/ V.=58 MeV, Wo=O

"
V =57 MeV, W =0
9=1700
'
,,
200
s:
~exP(2il.:)[e'p(-2~:)+exP(2i~.)
where J..~
+ iJl~
ir. 1
E, - E--jr
2
l'
,,'. '.
, '
<J)
<J)
<J)
...- ..... -.\: ....
Ol
The cross section calculated with this model is smooth, with .

rather broad resonances observed only at energies such that the
conditions for standing waves to form in the nucleus potential
well are fulfilled. Narrow resonances, which commonly
influence the cross section for light and medium-heavy nuclei,
are not reproduced within the framework of the potential model.
When few and sparse resonances are present in the excitation
function, these resonances can be taken into account by adding
resonance terms to the diagonal elements of the scattering
matrix. Then, for protons as an example, an element of the
scattering matrix has the fOrnl
,/
'\
E 300
c
-;
100
.'
V =57 MeV, Wo= 0.5
Me~-o-
0-
R
o~~~~~~~~~~~~~~
3.0
3.5
4.0
4.5
5.0
5.5
6.0
Energy, MeV
FIG. 3.12. Sensitivity of single-particle resonances to the
potential parameters (from Nuc!. {mfr. and Meth. 8261
(2007) 4ll1) ..
unrealistic for lower energies. Several attempts have been made to
develop a global set for the low-energy region. However for a
close reproduction of the measured angular distribution or
excitation function, the parameters of the model should be fitted to
experimental data specific to that energy range.
(3.39)
is the offresonance nuclear phase shift and the
The optical model is also often used for the case of elastic
scattering of complex particles such as alpha particles. There are
quantities Eo, r, and rp are the energy, total width, and partial
elastic width, respectively. The subscript I is the relative angular
momentum of the proton and the target in units of 11. The plus
sign refers to the case when J = I + Vz, and the minus sign
corresponds to the case when J = 1 - Y2. The quantity tf1, is the
resonance phase shift. An example of the corresponding
calculations is presented in Fig. 3.9 of Section 3.2.
well-known difficulties in specifying the alpha optical potential at

low energy. These are the continuous and discrete ambiguities and
the limitations related to the applicability of the optical model near

and below the Coulomb barrier. In addition, the anozmlous largeangle scattering and quasi-molecular resonances obseIYed in alpha
elastic scattering are phenomena that are outside the framework of
the usual optical model.
In most cases of Iow-energy scattering of charged particles

on light nuclei, the contribution of the reaction channels is
negligible, so the imaginary potential is close to zero. In the
region of single-particle resonances, the calculated cross section
is extremely sensitive to the potential parameters, as illustrated
in Fig. 3.12.
Various modifications of the standard optical model have

been made to improve its adequacy. Among them are using form
factors more complex than Saxon-Woods and splitting the
potential well depth according to the partial wa..'e I-number.
R-matrix theory is the most adequate approach when narrow

resonances dominate the cross section (Fig. 3.13). This implies
that the interaction proceeds through the compound nucleus
mechanism
It should be noted that, at Jow energy, the optical model is not

applicable in the case ofligbt nuclei because a basic assumption of
the model, namely, the averaging over the compound nucleus
levels, is not met.
The R-matrix theory matches wave functions outside the

range of the nuclear forces and internal wave functions at the
nucleus boundary, with the internal function being expanded
in terms of eigenfunctions of the radial equation with
appropriate boundary conditions to achieve stationary states.
It is essential that the eigenfunctions correspond to actual
states of the compound nucleus. If a wave function and its
derivative are known at the boundary of the nucleus, the wave
function can be found eveI)where outside the nucleus. Thus,
Generally. the results obtained using the optical model for the
scattering of nucleons from mediwn-weight and heavy nuclei are
quite reliable in the energy range of about 10-100 MeV. However,
at energies in the vicinity of the Coulomb barrier and in the subCoulomb region, some anomalies have been found in the
systematic dependence of the optical potential parameters on mass
number and energy. Therefore, the results obtained with global
sets acquired through optimization based on a wide collection of
experimental data measured at higher energies appear to be
36

300
2,.,
not be possible to judge whether the difference caIl.. be

regarded as real or whether it is caused by experimental er-:ror.
"AI(p,p/,AI
8=170'
~
~
:0'"
.s
~"
""6
"0
160(u,uO)"O
2000
-'"
~
100
Ll
'"
1000
E,,,7.6 MeV
1500
1000
-Theory
John et ai, e,.m.=169.2'
C"
1200
1400
1600
.800
2000
Energy (keY)
"0
""6
Hunt et ai,
eom =163.8'
500
"0
o~~~~~~~--~
FIG. 3.13. R-mattix calculation of the cross section in the case

of a strong resonance strucrure (Nucl. Jnstr. and Meih. B190
100
12{)
140
160
160
eom.
(2002) 237). Experimental points are taken from mANDL.
FlG. 3.14. Comparison of the results of crosssection mea.c;ureInents

within thc framework of a theoretical appf02.ch. Experimental points are
taken from [BANDL.
the S matrix elements [Eq. (3.21)] become expressed through

the level parameters of the compound nucleus.
The evaluation procedure consists of the following Steps:

First, a search of the literature and of nuclear databases is
made to compile relevant experimental data. Then, data from
different sources are compared, and the reported experimental
conditions and errors assigned to the data are examined.
Based on this comparison, the apparently reliable
experimental points are critically selected. Free parameters of
the theoretical model, which involve appropriate physics for
the given scattering process, are then fitted within the limits
of reasonable physical constraints. The model calculations are
finally used to produce the optimal theoretical diff:rential
cross section. Thus, data measured under dIfferent
experimental conditions at different scattering angles become
incorporated into the framework of a unified theoretical
approach, and ex.citation functions for analytical purpo~es can
be calculated for any scattering angle, with relIability
exceeding that of any individual measurement. An example of
the evaluation results is shown in Fig. 3.15. In order to meet
the needs of the IBA community, some elastic scattering
cross sections were evaluated, and the Sigma Calc Web site
was developed (http://www-nds.iaea.orgisigmacalci) to
provide convenient and reliable smooth curves for dcr(E)/rln
at any angle.
3.5 EVALUATION OF CROSS SECTIONS

Except for pure Rutherford scattering, exact knowledge
of the cross section cannot be derived from any experiment or
calculation. These data can be estimated only with some
limited degree of confidence. The main source of the data is,
of course, experiment. Many differential cross sections
relevant to IBA have been measured, and most of the data
llave been compiled in databases [mainly in the Ion Beam
Analysis Nuclear Data Library (IBANDL) at httpJiwwwnds.iaea.orgfibandlf and Experimental Nuclear Reaction Data
(EXFOR) at httpJiwww-nds.iaea.orgiexforij. However, there
are several reasons why these data should be evaluated prior
to their use. When the compiled cross-sections are compared
(e.g. using the !BANDL overlay graphics) the comparison
often reveals discrepancies which are far beyond quoted
experimental errors. In SOUle cases, the cross section has a
fine structure with a typical width between 0.1 and 10 keY
that is often measured in insufficient detail (see an example
in Fig. 3.13). Moreover, because of the dependence of the
cross section on the scattering angle, the available data can be
employed only in the case of a scattering geometry very close
to the geometry used in the cross-section measurements.
Theory gives a unique instrument to resolve the problem of
comparing data acquired in different experiments. An
example is given in Fig. 3: 14, where two cross sections
measured at slightly different angles are shown. The
difference in angIe is 5.4, whereas the cross sections differ
by a factor of -2.4. Without theoretical calculations, it would
3.6 KINEMATICS OF NUCLEAR REACTIONS

The laboratory frame of reference is the frame where the
detector is situated. In rnA experiments, the target nucleus is
always in rest in this frame. For a projectile of mass MI moving
along the x axis toward a target nucleus of mass M 2 , the point
with coordinate Xc defined as
37
200
Chapl.r 3
150
100
energy associated with the movement of the center of mass and

the kinetic energy associated with the movement of the particles
relative to the center of mass. For a projectile with kinetic
energy Eo (in the laboratory frame), thi.s means that
50
0
~
"-
(/)
......
.0
50
'6
'tJ
(3.42)
100
'-"
"
E=(M,+M,)V; E
,
2
+",
150
elab=150
where En:. is the kinetic energy of the colliding particles in their

relative motion in the eM system. The substitution of Eq. (3.41)
into Eq. (3.42) gives the relation between Ere, and Eo
M, E
M,+M, '
E~,
300
250
(3.43)
which is used, for example, to find the correspondence between

the resonance energy in the laboratory system and that in the
center-of-mass system. It also follows from Eq. (3.43) that an
endoergic reaction is possible only for beam energies
200
150
100
50
elab =110
(3.44)
0
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
It is useful to have explicit fonnulas for the angle

transformation between the eM and laboratory frames that
include the projectile energy Eo in the laboratory frame only
Energy (MeV)
tane=
FIG. 3.15. Evaluation of the 'cross section for proton

elastic scattering from oxygen (NucJ. Instr. and MethB 129 (1997) 311). Experimental points reported in
different works are taken from ffiANDL.
sin eo
y, + cos 8,
(3.45)
cos9, =[cos'e(I-Y,'Sin'e)]'12 -y,sin'e,
(3.46)
where
(3.40)
where Xl and X2 are the projectile and target coordinates,
respectively, is the center of mass (eM) of the system
comprising these two particles. This point moves in the
laboratory reference frame with the velocity
Y,= ( M,M, "' (I+M,+M,

)
M,M,
y, = M,M,
M,M,
(3.41)
M,
)-'12 ,
Eo
)'12(1+ M,M,+M, EoQ J-"'
(3.47)
(3.48)
with indices i = 3 and i = 4 standing for the outgoing particle and

the residual nucleus, respectively.
where VI is the projectile velocity.

The center-of-mass reference frame is defined as the frame
with its origin fixed at the point Xc in Eq. (3.40). The target
nucleus is obviously not at rest in the eM reference frame;

however, its advantage is that the sum of the momenta of the
colliding particles in the eM frame is zero both before and after
the collision.
According to Konig's theorem, the kinetic energy of a
The total cross sections are invariant for the reference-frame

transformations. However, the differential cross sections doldO.
depends on the frame. The relation between the differential cross
sections expressed in the CM and laboratory frames is derived
from the equality of the numbers of particles emitted in the
corresponding solid angles in the two frames:
(3.49)
system consisting of a projectile and a target is the kinetic
38
or
dO"
do- sin e de d", = do- sin e de dq> ,

d!l
d!lc
c
C
c
( 1-y,' sin'
8)
112
<In [( 1-y,'
.
sin' e )112 +y, cose ]'
(3.50)
The azimuth angles <p and tpc arc equal, and for polar angles,
(3.5::2)
,
where y, are defined by Eq'. (3.47) and (3.48).
the transformation relations are given by Eqs. (3.45) and (3.46).
More transformation formulas for different particular Cases
Then, for the transformation of the differential cross

section, the following relations can be derived:
are listed in Appendix 4.
3.7 INTERNET-BASED NUCLEAR PHYSICS

RESOURCES RELEVANT TO IBA
(3.51)
Presently, practically all of the infonnation on properties of

atomic nuclei and nuclear reactions can be retrieved from
databases via the Internet. The Internet resources mentioned
throughout the text of this chapter and employed in the eXlUl'"lples
are summarized in Table 3.1.
Table 3.1. Summary of Internet-based nuclear physics resources relevant to rnA
Resource address
Description
http://www.nndc.bnl.gov/qcaIC/
Q-value Calculator (QCalc)
http://ntJ<leardata.nuclear.I~.se/database/masses/
Nuclear structure and decay data: NuBase database with

Q-value calculator
http://www.nndc.bnl.gov/ensdfJ
Evaluated Nuclear Sttucture Data File (ENSDF)
http://www.nndc.bnl.gov/nudat2/
NuDat 2.5: Search and plot nuclear structure and decay data
http://www.nndc.bnl.gov/masses/mass.mas03
Atomic Mass Adjustment 2003
http://www.tunl.duke.edulNuciDatai
Energy Levels of Light Nuclei, A = 3-20
httpl/www-nds.iaea.orglibandll
Ion Beam Analysis Nuclear Data Library (!BANDL)
http://www-ndsjaea.orglexfor/
Experimental Nuclear Reaction Data (EXFOR)
http:((www~nds.iaea.org!sigm.acalc/
SigmaCalc: Evaluated differential cross sections for rnA
39
Chapter 3
3.8 EXAMPLES
3.8.1 EXAMPLE 3.1.ldentificatioD of peaks in d + l7AI
spectra
3.8.2 EXAMPLE 3.2. Resonances in the 27Al(P;r)28Si reaction
Figure 3.16 shows two spectra obtained from deuterons
Figure 3.17 shows the yield data for the 27 A1(P,y)"Si reaction.
impinging on an 27Al foil.

Both the "AI(d,a)"Mg and "AI(d,p)"AI reactions have positive
reaction energies. The corresponding Q values were calculated at

http://nucleardata.nuclear.1u.seJdatabaselrnassesl.
The energies of low-lying states in the residual nuclei 25 Mg and

26A1 are taken from http://www.nndc.bnl.gov/ensdf7. Energies of
the outgoing particles are calculated according to the equations
of Appendix 4.
For this reaction, Q = 115&4.934 keY (taken from

http://www.nndc,bnJ.gov/qcalc/). The excitation energy of the
28Si compound nucleus can be calculated according to Eq.
(3.31). The level energies for 23Si in the interval corresponding
to
the
proton
energy
range
are
available
at
http://www.nndc.bnl.gov/nudat2/. Proton resonance energies can
then be calculated as
E'
'"
~ (E'
-Q) MSi ~ (E' -11584.934) 27.976927

MAl'
. 26.981539'
where the masses MAl and MSi for 27Al and 28 Si , respectively,
were taken from Audi e( ai., 2003 (available online at
http://www.nndc.bnl.gov/m assesimass.mas03).
Eo:' keY
12741
11 400
'2
:0
FIG. 3.16: Spoctnt of alpha particles measured with a ''!bin''

detector (top) and protons measured with a filtering foil 0[1 the
detector (bottom) induced by deuterons impinging on 27AI foil
[Nuc/. Instr. and Meth. B266 (2008) 3535].
12Q'J~
13128
IU21
3.8.3 EXAMPLE 3.3. Resonances in the nuclear reactions IHe~,a1)160 and 1!1F(p,o.y)160
Figure 3.18 shows the energy-level diagrams of different nuclei for the case of fluorine detection, where 19p(p,ay)16 0 has been lLsed
most freqlJently (see Chapter 7). Its reverse reaction, lH(I9p,O:y) Hi O, has commonly been used for hydrogen analysis (see Chapter- 8).
Resonance parameters for IHe9FJa.y)2~e were measured for the interaction between fluorine and hydrogen far the praton-indll..ced
at
reaction 19F(p,ay)2~e. The parameters are listed in Table 2.29 of Tilley et al., 1998 (also available
http://www.tunl.duke.edulNuclDataJ). A strong resonance is observed in this reaction at a proton energy of 872.11 keY. In the ceo..terof-mass frame of reference, the resonance energy is [see Eq. (3.43)]
Eres '_ EH
em
lab
MF
H+
For fluorine as the projectile, the resonance energy in the laboratory frame of reference is
It is worth noting that, if mass TIlimbers are used in the calculations instead of atomic masses, the error in the resonance energy is as
large as 130 keV.
The resonance width of 4.30 keY is given in Table 2.29 of Tilley et al. (1998) in the center-of-mass system. It is transfonned into the
laboratory frame in a similar manner:
4.30 1.007825+18.998403
1.007825
85.36 keY.
The resonance cross section for gamma production should be identical for 1~(p, ay) lliO at the resonance energy ofEp= 872.11 keY
and for IHe~, cry)160 at the resonance energy ofEl~F= 16.44 MeV.
position of the detector of recoiled alpha particles. The angle
trimsformation is made with help of Eq. (3.46). For registering
alpha particles at 130, the calculations give 138.5 e in the eM
system, with the values for the particle masses and for Qof the
reaction
being
taken
from
http://nucleardata.nuclear.
lu.se/databaselmassesl. Because the velocities of the outgoing
proton and alpha particle are aligned in opposite directions in the
eM frame of reference, the angle Scm for the outgoing proton is
180 0 _ 138.5 0 = 41.5 0 This value has to be converted into the
laboratory system, which gives Slab = 40.1 e. The cross section
0,(9) ~ 66.8 mb!sr for the 'He(d,p)'He reaction at this angle
(40.1) is taken from IBANDUEXFOR. Finally, the Crosssection 04($) = 47.7 mb/sr for detecting alpha particles at the
laboratory angle ~ ~ 130' is computed according to Eq. (A4.20)
of Appendix 4.
3.8.4 EXAMPLE 3.4. Kinematic transformations for the JHe
+ d"""" P + "He + 18.35 MeV reaction

This reaction can be applied in different ways using either
3He nuclei or deuterons as projectiles and registering either
protons or alpha particles. Some information on the differential
cross sections for the interaction between 3He nuclei and
deuterons is available in the databases EXFOR and IBANDL
(see Table 3.1 for database addresses). For the given detector
position at 130 0 in the laboratory frame of reference, the cross
section for detecting protons with a 420 keY deuterium beam is
56.5 mb!sr. Using Eq. (A4.14) (see Appendix 4), one can obtain
the proton particle energy of 13,94 MeV. Using the same
equation with M3 and M" exchanged in the quantities A-D, one
can obtain the alpha-particle energy of 2.82 MeV at 1300
However, there is no information on the cross section in the
case wben the recoil "He is registered. The problem is to find the
cros:s section for detecting alpha particles at this angle. To make
the transformation for the cross section, flrst, it is necessary ta
find the angle for outgoing protons that corresponds to the
41
Chapter 3
"Ne
".,,1115
"l'
Ep
IkeVJ
SI'Oble
"Il
Stable!
Sf'obl!
3_8.5 EXAMPLE 3.5. Conversions between 'He(d,p)'H. and
'H('He,p)'He
!. . .
Another problem is to establish the correspondence between the
~.020
('~ .w.
"Ne
'He(d,p)'He and 'H('He,p)'He reactions. Thc cross-section

maximum is reached in the 'He(d,p)'He reaction at about 420
keV. Following the procedure applied in EXAMPLE 3.3 it is easy
to find that the cross-stx:tion maximum in the lHeHe,p)"He

reaction occurs at
S'Jl
484
,3100
22'
. alJlCl"'i1 _
r-
12.81<
8,'15
E'H, = Ed M'll< = 420 3.016029 629 keV .

""
"" Md
2_014101
~
6.06
In the eM frame of reference, the cross sections for the direct and
reverse reactions at any given energy are the same. The
transfonnation of the 3He{d,p)"He cross section into the CM
system is performed according to Eq. (A4.18) (see Appendix 4),
with the quantities A, B, C, and D being calculated assuming M I =
Md and M2 = M JHC ' To determine the result for the 2HeHe,ptHe
121
4969
14 248
'--
11.612
19FIP.aJ"O
1 .
19Ff p,n}"Nt
"Flp.yJ"N,
reaction, the cross section obtained in the eM system for 3He(d,p)"

is reverted into the laboratory frame using the same equations but
assuming M] = M JHc and M2 = Md. Tbese calculations are rather

'--
FIG. 3.18. Energy-level diagrams of different nuclei for the case of

fluorine analysis. A plot of cross section versus proton energy is
included for 19p(p,cry)160 resonance reactions. Use 'Nith permission
from Chapter 4 of Ion Beam for Materials Analysis, edited by 1.R
Bird and 1.S. Williams, Academic Press, Australia, 1989.
tedious and are performed much more easily using a spreadsheet

program such as Excel. As the end results, the calculations show
that, when protons are registered at a 130 0 laboratory angle and
the cross section for the JHe(d,ptHe reaction is assumed to be
56.5 mb/sr at a resonance energy of Ed - 420 keV (taken from the
EXFOR
database,
bttp://W\>w-nds.iaeaorg!exfor/),
the
2H(~e,ptHe reaction gives 55.6 mb/sr, for a resonance energy of
E lHC = 629 keY. The difference between these two differential
cross sections is small only because the reactions have a very high
Q value (18.35 MeV).
REFERENCES
Audi, G., Wapstra, A.H., and Thibault, C. (2003), Nuel. Phys.
A728,337.
Hodgson, P.E. (1963), The Optical Model o[Elastic Scallering,

Clarendon Press, Oxford, UK.
Ericson, T., and Mayer-KucktJ.k, T. (1966), Annu. Rev. Nucl. Sci.
Lane, A.M., and Thomas, R.G. (1958), Rev. Mod. Phys. 30, 257.
16,183.
Tilley, D.R., Cbeves, C.M., Kelley, J.H., Raman, S., and Weller,
H.R. (1998), Nuel. Phys. A636, 247.
Firestone, R.B., Shirley, V.S., Baglin, CM., Zipkin, J., and Chu,
S.Y.F. (1996), Table o[Isotopes, 8th Ed., Wiley-Interscience,
Berkeley, CA..
Vogt, E.W. (1968), ''The Statistical Theory of Nuclear
Reactions'\ in Advances in Nuclear Physics (Baranger, M., and

Vogt, E.W., cds.), Plenum Publishers, New York, vol. 1.
Gurbich, AF. (2004), Nucl.lnstrum. Methods H217, 183.
42
CHAPTER
Backscattering Spectrometry
J. A. Leavitt and L. C. Mcintyre, Jr.
University of Arizona, Tucson, Arizona, USA
M.R. Weller
Middle Tennessee State University, Murfreesboro, Tennessee, USA
CONTENTS
4.1 INTRODUCTION ......................................................................................................... .45
4.2 FUNDAMENTALS ........................................................................................................ .46
4.2.1 The kinematic factor and mass resolution ........................................................... 46
4.2.2 Elastic scattering cross sections ............................................................................ 46
4.2.2.1 Definition ofthe differential cross section ................................................... 47
4.2.2.2 Rutherford cross sections .............................................................................. 47
4.2.2.3 Non-Rutherford cross sections .....................................................................47
4.2.3 Experimental geometry ...... ..................................................................................49
4.2.4 Effects of energy loss of ions in solids ..................................................................49
4.2.4.1 Defmitions ................................................................................................. ,.....49
4.2.4.2 Depth scale ......................................................................................................50
4.2.4.3 Depth resolution .............................................................................................51
4.2.4.4 Surface spectrum beigbt. ...............................................................................52
4.2.4.5 Peak widtbs for tbin fIlms .............................................................................52
4.2.4.6 Mean energy in tbin fIlms ............................................................................. 52
4.3 SAMPLE ANALySIS ......................................................................................................53
4.3.1 Tbin-fIlm analysis ................................................................................................ .54
4.3.1.1 The peak integration method ........................................................................ 54
Multilayer films ..................................................................................................... 57
Use of elastic scattering resonance....................................................................... .58
43
Chapter 4
Overlapping peaks ................................................................................................. 58

4.3.1.2 Stoichiometry by surface spectrum beights .................................................59
4.3.1.3 Peak width methods ....................................................................................... 60
4.3.2 Surface layers of bulk materials ......................................................................... 64
4.3.2.1 Stoichiometry by surface spectrum heights ................................................. 64
4.3.2.2 A deficiency method ....................................................................................... 65
4.3.2.3 Depth prof'iling ............................................................................................... 67
4.3.3 Numerical computations .....................................................................................70
4.3.3.1 Areal density calculations ............................................................................. 70
4.3.3.2 Computer simulation ..................................................................................... 70
4.4 EXPERIMENTAL CONSIDERATIONS .............. ,..............................:....................... 70
4.4.1 Design factors ....................................................................................................... 70
4.4.1.1 Sensitivity ......................................................................................................~.70
4.4.1.2 Mass resolution .............................................................................................. 70
4.4.1.3 Depth resolution ............................................................................................. 70
4.4.2 Ion beams and ion energies ................................................................................. 71
4.4.2.1 Comparison of beams and energies .............................................................. 71
4.4.2.2 Medium.energy backscattering .................................................................... 71
4.4.3 Cautions ................................................................................................................ 74
4.4.3.1 Sample eharging ............................................................................................. 74
4.4.3.2 Channeling ...................................................................................................... 75
4.4.3.3 Extraneous peaks ........................................................................................... 75
. __,,_ .. _.4.4.3.4 Radiation exposure ........................................................................................ 75
4.4.4 Examples ............................................................................................................... 75
4.4.4.1 Depth prof'lle: 1.8 MeV 'He ........................................................................... 76
4.4.4.2 Mass resolution: 15 MeV 16 0 ........................................................................ 76
4.4.4.3 Depth resolution: 25 MeV 3'CI ..................................................................... 77
4.4.4.4 Beam modification: 250 keV He.................................................................. 78
------'--'--4.4.4.5 Depth prof'lle with elastic resonance: 3 MeV 'He ....................................... 78
REFEREN CES .................................................................................................................... 79
'.4
:1
.
.1
21
t
3
:~
"
~,.
.j,
oj
.s;ti.,
.'.~
-?
.f,!
.~
44
.~
.@
Here, Ni is the atomic density (atoms per unit volume) oftb.<:: ith
element, and t is the physical film thickness. If the scattering is
Rutherford (pure Coulomb) scattering, then 0;(E.9) c:tr::l. be
calculated from
4.1 INTRODUCTION
Backscattering spectrometry Using ion beams with energies in
the megaelectronvolt range has been used extensively for
accurate determination of stoichiometry, elemental areal density,
and impurity distributions in thin films. Measurement of the
number and energy distribution of ions backscattered from
atoms in the near-surface region of solid materials allows
identification of the atomic masses and determination of the
distribution of target elements as a function of depth below the
surface.
(zze')'
0'1l(E,9)= ~
Application of the technique to thin-film analysis is illustrated

in Fig. 4.1 for the ideal case of a two-element thin film of
uniform composition on a low-mass substrate. Analysis ions
scatter elastically from target atoms with energy characteristic of
the mass of the struck particle; they also lose energy passing into
and out of the film material. Energy analysis of the backscattered
ions by the detection system yields the backscattering spectrum
displayed in the lower portion of Fig. 4.1 in the form of counts
per channel versus channel number. The channel number is
normally linearly related to the backscartered ion energy, E 1
Appearing in the spectrum is a nearly flat -topped "peak." for
each element present in the film. The peak widths are caused by
the energy loss of the analysis ions in the film material.
energies (E~ , E~) of the high-energy sides of the peaks into the
expression
Ml sin 9(
Ml' -M,"
sin)9"
(4-A)
where ~ and ~ are the atomic numbers of the incident and target
ions. respectively. This equation is given in cgs units. A u.seful
number in evaluating this equation is e2 =:: 1.44 X 10-13 MeV- em.
For very thin films, E, the energy of the analysis ion immediately
before scattering can be taken as Eu. For thicker films, the Q1ean
energy of the analysis ions in the :film should be used for E (see
Section 4.2.4.6).
Q ions al Eo
The film elen:tents can be identified by insertion of measured
I
,I
4[(M/-M l1Sin 1aY'1 +M1COSeJ
Detector
(4.1)
2000
to calculate the kinematic factor K for the ith element. Eo is the

incident-ion laboratory kinetic energy. In turn, the kinematic
factor, K, is given by
E~ E~ Eo
III
j j
1000
(4.2)
(.)
where 9 is the laboratory angle through which the incident ion is

scattered and M I and M z arc the masses of the incident and
target particles, respectively. Because the parameters M}, Eo, and
9 are usually known, M2 can be determined, and the target
element can be identified.
500
10DO
Channel number
FIG. 4.1. Basic backscattering spectrometry. Experimental
geometry (top). Backscattering spectrum (bottom) for a hvoelement compound (AmBJ film of uniform composition on a
low-mass substrate.
The areal density. (Nt)i. in atoms per unit area, can be

determined for the ith element from knowledge of the detector
solid angle, .0, the integrated peak count Ai for Q incident ions,
and the measured or calculated cross section alCE,9) using the
expression
The average stoichiometric ratio for the compound film

(A.,BJ can be calculated from Eq. (4.3) to be
n _ NB
AB 0" A (E, 9)
m - NA ~ AA . O"B(E,e)
(4.3)
45
(4.5)
Chapter 4
Note that this ratio depends only on the ratio of measured

integrated peak counts AalAA and knowledge of the crosssection ratio GAlas. The hard-to-measure quantities Q and n
bave cancelled,
Conversion of an areal density, (Nt)j, into a physical film
thickness, t, requires knowledge of the film density, PAB. The
relevant atomic densities, N~ and N~, can be calculated
from
N oIJI ;mpASNO
- - andN AS ; npABN.
-_
A
M~
B
M~
(4.6)
and then used in
(4.7)
to calculate t. Here, No is Avogadro's number, and MAB = mMA +
nlvts is the molecular weight of compound AmBn.
measured stoichiometric ratios by reducing backscattering peak

overlap, as well as to improve mass resolution and sensitivity to
light elements. The cross sections are non-Rutherford for light
target elements. The higher-energy heavy ions are used to
improve mass and depth resolution and enhance the sensitivity to
medium and heavy elements. The lower-energy ions are used to
improve mass and depth resolution and sensitivity for all
elements, In addition, detector developments, particularly for
time-of-flight and electrostatic analyzers, have made the use of
the lower-energy ions feasible.
It is the aim of this chapter to present basic equations and

tabular and graphical data needed to analyze backscattering
spectra. A few examples illustrating how this infonnation is used
for the analyses arc included. For a more extensive discussion. of
the principles of backscattering spectrometry, see Chu el al.
(1978). Although this cbapter is not a review of recent advances
in the field) it concludes with a succinct discussion of
experimental considerations that might be of use to the analyst in
choosing the experimental setup that will produce optimum data.
4.2 FUNDAMENTALS
Typicaiuncertainties in the results are =1:::3% for areal densities

and a few tenths of I % for average stoichiometric ratios. The
actual uncertainty in the physical film thickness is usually
unknown because the film densities are usuaUy unknown.
The
preceding discussion indicates the power of

backscattering spectrometry for analysis of a simple film (see
Section 4.3.1.1 for further experimental details). Historically, the
majority of backscattering analyses have been performed using
4He analysis ions with energies in the 1-2 MeV region. The
reasons for this are that (1) the available accelerators produced
beams with these energies; (2) data for the energy loss of 4He in
the elements were better known than data for for other ions; (3)
the silicon surface barrier detector energy resolution for 4He is
about 15 keY; and most importantly, (4) the backscattering cross
sections for 4He incident on all elements more massive than Be
are nearly Rutherford in this energy region.
The principal strengths of Rutherford backscattering (RES)
with 4He ions are as follows: (1) It is an absolute method that
does not require the use of standards; meaningful uncertainties
can usually be assigned to the results. (2) It is quick and easy,
with a typical data acquisition time of about 10 min. (3) It is
frequently nondestructive. (4) It can be used for depth profiling
(with 10-30-nm depth reSOlution), The techniquets principal
weakness is that it is not good for trace-element analysis. It has
moderate sensitivity (_10-4) to heavy elements in or on light
matrixes but very poor sensitivity ( .... 10-1) to light elements in or
on heavy matrixes.
4.2.1 The kinematic factor and mass resolution

Equation (4.2), the expression for the kinematic factor K,
results from application of conservation of energy and
momentum to the two-body collision between isolated particles
of masses M t and M 2 Incident beam particle 1 (at laboratory
kinetic energy Eo) is scattered "With final laboratory kinetic
energy EJ through laboratory angle e by target particle 2
(initially at rest in the laboratory). The binding energy of particle
2 in the target is neglected. The kinematic factor is independent
of the nature of the force between the particles (as long as energy
is conserved). Tables of kinematic factors, K, are included in
Appendix 5 for the scattering of lH and 4rIe; the kinematic
factors for other ions can be calculated using Eq. (4.2). Average
kinematic factors, K., are also listed. where the average mass Ml
for natural isotopic abundance has been used in the caIculation
ofK [Eq. (4.2).
For fixed e, the energy separation, 8Et. for beam particles
scattered by target particles ofmass difference AM2 is [from Eq.
(4.1)]
dE, =E.(=JdEM,.
(4.8)
If L1EJ is set equal to BE. the minimum energy separation that

can be experimentally resolved, then 5M2, the mass resolution of
the system is
More recent years have seen increasing use of higher-energy
ligh' ions ('H, 'He). higher-energy heavy iollS ("0, "CI), and
even lower-energy ions (He. 12C) for backscattering analyses,
1iM2 - EO ( dK )
<1M2
The higher-energy light ions arc used to improve the accuracy of
46
(4.9)
1i
,,
'.
Figure 4.2 contains info~tion that can be used to estimate 3M2

for a given experimental situation if the overall energy
resolution, oE, is known. The quantity oE contains contributions
from detector resolution, straggling, beam energy spread, and
various geometric effects (O'Connor and Chunyu, 1989).
(9 E) ~
es,
(.2..J
dQ(E) _1_
Nt
Q nee)
(4. I(J)
where Nt is the number of target atoms per urit area

perpendicular to the beam and [dQ(E)]iQ is the fraction. of
incident particles scattered into the small solid angle 0(8)
centered at deflection angle 6. If dQ(E) is replaced by A and
subscript i is added, Eq. (4.3) results.
The mass resolution at the sample surface is usually

detennined primarily by the detector resolution; straggling
dominates for layers deep in the sample. For fIxed BEfEo, Fig.
4.2 indicates that SM2 improves with increasing analysis beam
mass. This is somewhat deceptive because oE frequently
depends on the analysis beam mass. For instance, if a typical
surface barrier detector is used with 5 MeV 4He and 12C analysis
beams, then oM? at M2 = 100 u is actually somewhat better (2.2
u versus 3.2 u) for 4He than for 12C because the detector
resolution is ~15 keY for 4He and -50 keY for l2C at 5 MeV
(Leavitt et al., 1988). There are, of course, iIl5tances where the
detector resolution improves with increasing beam mass; such is
the case for the time-of-flight detector.
4.2.2.2 Rutherford cross sections

If the force between the incident nucleus (Ml,Zle,E) and the
target nucleus (M2,Z2c, initially at rest) is assumed to be the
Coulomb force, Fl2 = (ZI~e2/?ri, then use of the definition in
the preceding section results in Eq. (4.4), the expression for the
Rutherford cross section in the laboratory system Numerical
values of laboratory Rutherford cross sections for beams of IH,
4He, 7Li, i2C, and 14Si at 1 MeV incident energy are given in
Appendix. 6 for several backscattering angles. An accurate
approximation for large backscattering angles and (MliIvh) 1 is
Mass resolution conversion
(Chu etal., 1978)
2000~~~~~~~~~~~
cr.(E,e)~002073(Z~~, J[sin~m-2(~:J]
1800
1600
1400
~1200
with E in MeV and
~ 1000
in b/sT [\ b (bam) = 10- cml 1.

24
4.2.2.3 Non-Rutherford cross sections

Experimental measurements indicate that actual cross
sections depart from the Rutherford values at both high and low
energies for all proj~ctile-target pairs. The low-energy
departures are caused by partial screening of the nuclear cbarges
by the electron shells surrounding both nuclei. Results of several
investigations (L'Ecuyer et aI., 1979; Hautala and Luornajarvi,
1980; Andersen et al., 1980; MacDonald et al., 1983; and
Wenzel and Whaling, 1952) indicate that these low-energy
corrections are given with adequate accuracy by (L'Ecuyer et al.,
1979)
~ 800
600
400
20~~~~~~~::====~:J
o
CiR
[4.11)
20 40 60 80 100120140160
M2
FIG. 4.2. Plots of(dK/dMlfl versus target mass Mz for several
analysis beams (M 1 = ion mass). Units ofMI and Mz are u. The
alaR
laboratory backscattering angle is 1800 The plots can be used to
=I
0049Z Z"
.
I
1
(4.12)
E~
estimate the mass resolution 6M2 ~ (6ElEo) (dKldMX' ifSEIE"

is specified.
or by (Wenzel and Whaling, 1952)
alaR =1
O.0326Z Z'"
I
(4.13)
Eo<
4.2.2 Elastic scattering cross sections

4.2.2.1 Definition ofthe differential cross section
The average differential cross section, cr(e,E), for scattering
of beam particles of incident energy E by target particles in a
thin film is defined by
for light-ion analysis beams with megaelectfonvolt energies.

Here, ECM is the center-of-mass kinetic energy in keY. In
practice, replacing ECM by ELab produces negligible error. A table
47
Chapter 4
of low-energy corrections is given in Appendix 6. Algorithms

for rapid computations of cross sections for medium-energy (SO1000 ke V) backscattering are discussed by Mendenhall and
Weller (I99Ia).
ratio can be adversely affected if the cross sections of interest are

non-Rutherford at the higher ion energy. The analysis-ion energy
should usually be chosen to be as high as possible to take
advantage ofthe reduction in peak: overlap, but with a value such
that the relevant cross sections have accurately measured values
that do not vary widely in the region just below the incident ion
energy. If the ratio of measured-to-Rutherford cross section
varies slowly with energy, the non-Rutherford effect can easily
be included in calculations using Eqs. (4.3) and (4.5) by simply
dividing the values of ~ by the non-Rutherford enhancement
factors (a/oR)j at the mean energy of the projectile in the film
and proceeding with the calculation ~ if the' cross sections wer~
Rutherford. Because non-Rutherford cross sections for light
The high-energy departures of the cross sections from

Rutherford behavior are caused by the presence of short-range
nuclear forces. Some measurements and calculations (Bozoian et
al., 1990; Bozoian, 1991a and 1991b; and Hubbard er al., 1991)
regarding the onset of these high-energy departures are
summarized in Fig. 4.3 for lH, "He, and 7Li analysis beams. The
straight lines in Fig. 4.3 represent rough boundaries separating
the region of Rutherford behavior (below the line) from the
region of non-Ruthenord behavior (above the line). Equations
resulting from the least-squares fits to the points in Fig. 4.3 are
For 'H:
E:
z (0.12 O.OI)Z, - (0.5 0.1)
E: ~ (0.25 0.0 I)Z, + (0.4 0.2)

For'Li: E: ~ (0.3300.005)Z, -(1.40.1)
For 'He:
where
16
E:
(4.14)
7U
14
12
is the laboratory projectile kinetic energy (in
MeV) at which the backscattering cross section (for 160 0 < SLab
< 180) deviates from Rutherford values by 4% for a target atom
of atomic number ~. Note that cross sections for backscattering
of lH at 1 MeV are non-Rutherford for Z2 :5 15, and cross
.. sections for 'He at 2 MeV are Rutherford for Z,:> 6. At present,
no practical method exists for rapid and accurate calculation of
these high-energy non-Rutherford cross sections; they must be
measured.
10
~
>
'He
~8
a:.Q
zj
W
Advantages associated with the use of non~Rutherford cross

sections such as improved accuracy in the determination of
stoichiometric ratios and increased sensitivity for the detection
of light elements in heavy-element matrixes frequently justify
the additional work required to make the measurements.
According to Eg. (4.5), the average stoichiometric ratio for two
film elements depends only on the ratio of integrated peak
counts, ApjAA, and the cross section ratio, (SA/aB' Thesc
stoichiometric ratios can be determined as accurately as a few
tenths of a percent by acquisition of sufficient data if the
backscattering peaks are well separated and if the cross-section
ratios are accurately known. However, use of analysis-ion
energies such that the cross sections are Rutherford (and
therefore accurately knovvn) frequently produces backscattering
peaks that overlap. The uncertainties in the peak count ratios
resulting from deconvolution/simulation techniques often
severely limit the accuracies of the stoichiometric ratios,
particularly in cases of nonuniform film composition.
4
2
0
10
20
30
40
50
Target atomiC number (Z2)

FIG. 4.3. Laboratory projectile energies, E Lab, at which
backscattering cross sections (for 160 < ~ < 180) deviate
from the Rutherford values by 4% versus target atomic number,
~, for iH, 4He, and 7Li projectiles. The straight lines are leastsquares fits to the data points [Eq. (4.14)]. The 'H and 'He points
are experimental (see Bozoian, 1991b, for references). The 7Li
daia were obtained from optical model calculations (Bozo ian,
199Ia).
Use of higher analysis-ion energies usually results in desired

reduction of peak overlap because the energy loss of the analysis
ions in matter decreases with increasing energy in the energy
range normally used. However, the accuracy of cite cross-section
48
Backscattering Spectromehy
the perpendicular distance x below the sample surface at wb..ich a

backscattering event has taken place are as follows:
target elements are frequently many times the Rutherford value
whereas corresponding cross sections for the heavier target

elements can remain Rutherford, increased relative sensitivity
[or light-element detection results. For example, the relative

sensitivity for detection of C in thin films on Si substrates is
enhanced relative to Rutherford by a factor of7 for 3.8 MeV 4He
d.
and by a factor of more than 100 for 4.27 MeV 4He. In another
d.
instance, 4He ions with energies of 8.1-9.1 MeV have been used
in superconducting
films by a factor of -25 over Rutherford (Martin et al., 1988;
Barbour el al., 1988).
to enhance the sensitivity for detection of 0
x
casal
x
cosS,
d.
d_
(IBM)
cos9 l
(4.15)
x
(Cornell)
cos 91 , cos 9 2
4.2.4 Effects of energy Joss of ions in solids

4.2.4.1 Definitions
Many of the features of a backscattering speCtnlItl are
determined by the energy loss of the analysis beam ions as they
traverse the sample material. Consequently, quantitative
knowledge oftms energy loss is a key element in understanding
a backscattering spectrum. This section offers only a brief
sWilinary of the relevant quantities and relations relating to
energy loss, but a full exposition is given in Chapter 2.
Strong, narrow, isolated resonances in the non-Rutherford

cross sections can be used for depth profiling oflight elements in
Of on heavy matrixes'. Examples are the 3.04 MeV resonance in
the 4He- 160 cross section (Cameron, 1953), the 2.525 MeV
resonance in the IH_9Be cross section (Mozer, 1956; Leavitt et
al., 1994), and the 4.26 MeV resonance in the 4He_ 12C cross
section (Bittner and Moffat, 1954).
There is considerable information in the literature regarding
measured non-Rutherford cross sections for IH and "He
projectiles. Most of these data were acquired during the 19505
and 19605 in connection with studies of nudear-Ievel structure.
The data have usually been presented in graphical form only, as
differential CfOSS sections in the center-of-mass system versus
projectile energy in the laboratory system. An unpublished
compilation (Jarjis, 1979) gives some tables of numerical values
prqduced by use of a computer digitizer. A few more recent
reports contain tabular as well as graphical data on the measured
cross sections. Information on non-Rutherford cross section for
I H and 4He analysis beams is presented in Appendix 7.
The stopping power of a material for a particular ion is
usually defined as the energy loss per distance traveled in the

material, denoted as dEldx. This quantity depends on the ion and
the material traversed, as well as the energy of the ion. TyPical
units of the stopping power are eV/A or eV/nm.
Another quantity. the stopping cross section
4.2.3 Experimental geometry

Two experimental arrangements in common use are referred
to as the mM and Cornell geometries. For both geometries, the
incident beam is horizontal, and the sample surface is vertical.
For the IBM geometry, the scattered beam (directed at the
detector), the incident beam, and the sample normal are all in the
same horizontal plane. In the Cornell geometry, the detector is
directly below the incident beam; the inCldent beam and the
scattered beam are in a vertical plane. In both geometries, the
angle between the sample normal and the incident beam is 8 1,
and the angle between the sample normal and the scattered beam
is 6 2, The "tilt" axis is a vertical axis through the beam spot on the
sample sUlface, so the tilt angle is OJ (see Fig. 4.1) in both cases.
The relation between the scattering angle 8 and 6 1 and 6 2 is
for
the
IBM
geometry
(4.16)
(atoms/cm3).
The reader should be

where N is the atomic denSity
aware that the literature is not consistent in these definitions~ and
a common unit of "stopping power" used in nuclear physics is
eV/(mr/cm2), where the material '<thickness" is given in units of
mgfcm (actually an areal density) (Northcliffe and Sthilling,
1970).
Values of ion stopping cross sections in all elements are
available from an extensive study based on semiempirical fitting
of experimental data (Ziegler el al., 1985). More up-la-date
stopping cross section data can be found in SRlM 2009
(www.SRIM.org). The variation of stopping cross section for '1fe
ions Mth energy is similar for all elements, showing a broad
maximum below about J MeV. The decrease in energy loss of
"He ions with increasing energy above about 1 MeV is
responsible for the fact that the elemental peak widths in a
backscattering spectrum are narrower at higher incident energies
in this energy range.
and
cose 2 = cos(1t-e) cos 81 for the Cornell geometry. In the ffiM

geometry, for a given tilt angle a], the angle 62 depends on the
direction ofS I ; that is, it depends on whether the sample normal is
rotated toward or away from the direction of the scattered beam.
The relations between the inward and outward path lengths and
An approximation, called the Bragg rule (Bragg and Kleeman,

1905), is commonly used to calculate stopping cross sections of
ions in compounds or mixtures of different elements. This
49
L....
is defined as
traversed. This quantity is independent of the volume density of

the material. Typical units are 10- 15 eV cm2 The relation between
these two quantities is given by
dE
'-=NE
dx
9=1t-19, 92 1
E:,
the energy loss per atom per unit area (areal density) of material
Chapter 4
approximation simply assumes that each target atom acts

independently in the energy loss process and ignores any effects of
chemical bonding in the material. lbis rule can be expressed for a
compmmd AmB, by
Usin2 the second eoualitv in Eo. (4.18), we fmd
N~203 eAl + Nif20 3 eO
2 3
(:!:)Al 0 =
=4.7xI0 22 x44 X10-15
(4.17)
+ 7.1 X 1022 x 35 x 10-15
which gives the compound stopping cross section in terms of

energy loss per molecule per unit area traversed. The
corresponding stopping power is given by
=46xIO
( -dxdE)~ =N~ ~ =N "~, +N~,
--46 eV
A
(4.18)
dE
aE=J-dx
dx
and N:a are the atomic densities of A and B, respectively, in the

compound (where the m and n subscripts on A and B are
suppressed in this notation). (See EXAMPLE 4.1.) Deviations
from the Bragg rule have been reported, and a discussion of the
effect of chemical binding on stopping powers is given in Ziegler
of thick targets, a computer is usually used to calculate the
EXAMPLE 4.1. Calculate the stopping cross section and stopping

power of2 MeV 4He'i- in AI 1 0 3 using the Bragg rule. Use tabulated
values of stopping cross sections (see Appendix 3, Section 3.1), fAI
=44 x 1O-ls eV cm2 andEo =35 x 10- 15 eV cm' in Eq. (4.17) to fmd
energy of ions at a depth below the sample surface by numerical

integration of Eq. (4.19), with the target material divided into
thin slabs and a parameterization used to calculate the stopping
power at successive ion energies. Notice that the calculation of
ion energy loss requires knowledge of the composition of the
sample, which might be the object of the analysis and therefore
unknown. An iterative procedure is often required in such cases,
with an assumed composition mied to calculate an initial
approximation of energy loss, which is, in turn, used to calculate
improved values of the composition (see Section4.3.1.1).
(2)( 44+ 3 x35):x lO-15
= 193:x la-IS eV cm~.

To fmd (dEldx)"'4 O, ,we calculate the molecular density NAi,o, as
g
= pN~
4--)(6)(10
11
molecules
=_-,cm~_'____....!!mlSo!!.l_
102~
The range of ion beams in materials and the phenomena of

energy straggling are also of importance in understanding a
mol
= 2.5xlO
3
12
backscattering spectrum (see Chapter 2). Range elements A and

B are used in the stopping cross section factor. The relations for
element A are given by the equations infonnation can be used to
estimate the maximum analysis depth possible with a given
beam at a given incident energy. Recall, however, that the beam
must scatter and re-emerge from the sample to be useful in
backscattering analysis. A rough criterion for accessible depth
proposed by Chu e/ al. (I 978) is that the energy of the scattered
particle at the detector should be greater than one-fourth of the
incident energy.
All 0 J molecules
em'
The stopping power is obtained from Eq. (4.18) as
0
dE)Al2 3
( dx
(4.19)
In cases of thin targets, the stopping power can often be taken as

constant, with tbe necessary quantities evaluated at the incident
energy (surface-energy approximation) or at the mean energy of
the beam in the target (mean-energy approximation). In the case
and Manoyan (1988) (also see Chapter 2, Section 2.2.5.2).
NAl.O,
em
The calculation of energy loss, AE, of ions traversing a solid

involves the integration of the stopping power as ill
where NAB is the molecular density (moleculesicm3) and N~
0
8"'4 , =
seV
=N Al 20 3 e Al203
=2.35 X 1022 x 193 X 10-15
=46 eV
An altemate method, using an atomic basis, involves calculating the
4.2.4.2 Depth scale
atomic densities of each element in the molecule:

N~lO] = 2 x 2_35 x lOzz = 4.7 x IOn AI atoms/em]
To utilize the depth profiling capabilities of backscattering, it

is necessary to relate the energy of the scattered particle to the
depth in the sample where the scattering occwred. This
relationship depends on the energy loss of the analysis ion
traversing the sample, the kinematic factor for the scattering, and
and
N~201 =3x2.35xlOl.l =7.1xl0 I2 0atoms/cml.
50
(surface-energy approximation) or the mean beam energy in the

sample (mean-energy approximation) is satisfactory.
the orientation of the sample normal relative to both the incident

beam and the detector direction. We denote the difference in
energy at the detector of a particle scattered at the surface and a
particle scattered at a depth x, measured perpendicular to the
sample surface, as liE. This quantity is given as a function of x
by
LIE = [S]x
4.2.4.3 Depth resolution

The considerations in the previous section lead to an
expression for the depth resolution in backscattering analysis.
Using Eq. (4.20), we see that the minimum detectable depth
difference, Sx, is related to the minimum detectable scattered
particle energy difference, 6, by
(4.20)
where [S] is called the energy loss factor and is defined by
1+ (dE)
-[ (dE)
dx
dx
[5]= K -
in cos 9 1
1]
liE
out cos 9 2
[SI
(4.21)
Sources of energy spread include'detector resolution, energy

spread in the incident beam; straggling, and kinematic effects.
Because energy straggling increases as the ion beam tray- erses
the sample, the depth resolution degrades with depth in the
sample. A common practice is to quote depth resolution at the
surface, which need not include a straggling contribution. A
convenient approximation is to assume that all sources of energy
spread are Gaussian and to add them in quadrature.
The depth resolution can be improved by increasing [S]. This is
usually done by tilting the sample normal relative to the
incoming beam (i.e., increasing 9 1 andlor 9 2). The effect is to
increase the path length required to reach a given depth
(measured perpendicular to the surface) in the sample. This
increases the energy difference of the scattered particle for a
given depth difference. [t should be noted that the use of large
tilt angles introduces additional sources of energy broadening
and requires that tbe sample surface be reasonably flat. Several
studies concerned with optimizing depth resolution by target
tilting can be found in the literature (\Villiams and Moller, 1978;
O'Conner and Chunyu, 1989; and Boerma et al., 1990).
The kinematic factor K and the angles 9, and O2 are defined in

Sections 4.2.1 and 4.2.3, respectively. and the stopping powers
are those for the analysis ion on the inward and outward paths.
The corresponding relation involving the areal density and
stopping cross sections is given by
LIE = []Nx
(4.22)
for a single-element sample. The quantity [e] is called the

stopping cross section factor and is defined by
[,] =
(Kl:. _1_
cosS
H.
_1_)
(4.23)
cosS 2
For multi-element samples, we see that the depth-energy

relation depends on the struck particle; therefore, separate
relations must be calculated for each element in the sample. In
the case of compound AmBn, N becomes the molecular density
NAB (molecules/crn1 and corresponding values of K and Eout that
apply for scattering from elements A and B are used in the
stopping cross section factor. The relations for element A are
given by the equations
EXAMPLE 4.2. Calculate the depth-scattered ion energy

differences for 2 MeV ~e+ in A1 20 3 We again consider the case al
= 0 and !h = 10. Here, we must calculate separate differences for
AI and 0, The Al and 0 stopping cross section factors are ca1culated
using Eq. (4.25). The K factors for "He on Al and b are 0.5525 and
0.3625, respectively, resulting in energies after scattering (at the
surface) of 0.5525 x 2.0 = 1.105 MeV for Al and 0.3625 x 2.0 =
0.725 MeV for O. We evaluate the elemental stopping cross
(4.24)
[]:-' =(K,~_1_H::..._1_)
cos
9(
(4.26)
8x=-.
--
(4.25)
cos9l
AllO J
Similar relations can be written for element B. The lower index

refers to the scattering element, and the upper index refers to the
stopping material. A surface energy approximation is often used
for stopping cross section factors. In this appoximation. Ein is
evaluated at the incident energy Eo, and E. oul is evaluated at
energy KEo. The resulting stopping cross section factor is
usually written as ['0]. (See EXAMPLE 4.2.)
sections involved in out,Al and

surface-energy approximation:
~203 = 2x ef;; + 3 x e~ =2 x
A110)
out,O
44
at these energies using the
x 10-1!;
+ 3 x:35 x 10-15 =193 xlO- lS eV on 1
~t2 =
x E~.A1 +
x ~t,Al =
x 51
15
X 10-
+ 3x 46 X 10-15 =240 x 10-15 eVan 2
Because both the energy loss factor and the stopping cross
section factor require the evaluation of energy-dependent
parameters, an integral over the relevant energies should be
perfonned. In many cases, use of the incident beam energy
e~t.~J
= 2 x ~t.o + 3 X e~t.o =2 x 54 X 10-15

+:3 x 48 X 10-15
>=I
252 X 10- 15 eVan2
We can now calculate the stopping cross section factors
51
Chapter 4
AI,O, _
[ EO'<'1
]
-
"'1,0,
Ajt.:;.
_1_
cosS)
where [&o~ and [eo]~ are the compound stopping cross

section factors as defined in Eq. (4.25) using the surface-energy
"', _l_
+E-.,...J
cosS]
approximation.
[.co, =0.5525xI93xlO-" + 240xI0- 1'

xl.015 = 350xI0- 1
The stoichiometry of a multi-element sample can be

calculated by comparing the surface heights of the elemental
peaks (see Section 4.3.1.2 and EXAMPLE 4,3).
eVem'
and
yJO
[610"
"0
1
A10
1
=K s'''''
--+8 " - _
0 ill
cosS
..... 0 cosO,
4.2.4.5 Peak widths for thin nIms

The energy width of elemental peaks in single- or multielement samples. where the peaks are at least partially resolved
and wider than the system energy resolution, can be calculated
from Eq. (4.22) or (4.24) by replacing x by the film thickness l
Conversely. the elemental areal densities can be calculated from
experimentally determined peak widths using these equations
and a knowledge of the stopping cross sections. This procedure
is described in Section 4.3.1.3.
[.]~,o, = 0.3625 x 193 x 10-
+ 252 X 10-1
xl.015 = 326x10-1S eVem'
Using the molecular density

moleculeslem 3, we find
AIZO)
... =( ev)
A
.... (76.6-ev)
A XX
Of =[, j "N "x

'"
82.3- xx
'"
Of o =['j"N"x=
.0
where .c.EAl and .6.Eo are the energies of ions scattered from AI and
0, respectively, relative to ions scattered at the surface and x is the
depth in the target where the scattering took place.
l)
Then calculate the mean energy of the ions in the film. EO).
using
-(1)
AESEA
=EO---m2
(4.30)
This result represents a first-order correction. The procedure

should be iterated until E changes by Jess than a spetified
percentage between successive iterations. For the second
iteration, the values of (Nt)~J) should be calculated using Eq.
(4.27)
(4.3) with E = E(1); then .6Ef~1 and E(2) are calculated using
(4.31)
The surface heights of the two elemental peaks in the compound

AmBn are given by
"A<EoJ nQm&/([ol~ cos 8t)
HB.o ="BOlo) QQn&/([oJj}B cos 8t )
(4.29)
i~l
where (; is the energy width per channel and [Eo] is the surfaceenergy approximation of the stopping cross section factor
[defined in Eq. (4.23)] evaluated at the incident energy. The
remaining symbols are those defined in Eq. (4.3). This equation
is obtained from Eq. (4.3), where the areal density (Nt)
contributing to one channel of the spectrum at the surface is
&1[EoJ [see Eq. (4.22)].
HA,O =
the energy loss of the ions passing through the film, AE~nEA,
using
AE~EA = ti(Eo)(Nt)~EA
4.2.4.4 Surface spectrum beight

An important characteristic of a backscattering spectrum is
the height (counts/channel) of the front edge of an elemental
peak corresponding to scattering from the top surface of the
sample. We consider here cases where the sample produces
elemental peaks that are much wider than the energy resolution
of the detection system. For a singleelement target, the surface
height is given by
Ho ~ ,,(Eo)OQ&I([Eo)COS
4.2.4.6 Mean energy in thin films

To calculate the mean energy of the analysis ions in a thin
film containing r elements, use Eq. (4.3) with E = Eo to calculate
(Nt):EA in the surface-energy approximation (SEA). Calculate
(4.32)
(4.28)
52
Most of the data for the sample analyses descrihed in this

section were acquired with the standard backscattering setup
described in Leavitt (1987). Ion beams from a vertical, singleended 5.5 MY Yan de Graaff accelerator were deflected through.
90 by a bending magnet into a horizontal collimating beamlio.e
that preceded the target chamber. A 25 mm2 strrface barrier
detector, placed about 150 mm upstream from the target holder at
an angle 10 below the beam (Cornell geometry), subtended.
solid angle n ~ 0.78 msr at the targeL Hence, the backscattering
angle is e = .170. Detector pulses due to the backscattered.
analysis ions were preamplified, shaped and amplified, and
sorted by a pulse height analyzer (FHA), The resuit, in the form
of counts/channel versus channel number, constituted the
bacbcattering spectrum of the target. These data were sent to a.
PC for disk storage, integration of peaks, plotting, data analysis.
and so on. Some typical operating conditions and parameter
ranges for this system are given in Table 4.1.
EXAMPLE 4.3. Calculate surface heights for 2 MeV \Ie+ on

AhO}. We assume the following experimental parameters for
this calculation:
n = 10-3 sr
I> ~ 1 keYI channel
Q ~ 6.24 x lOll incident particles (10 fie charge)
61 ~ 0, 6, ~ 10 (scattering angle ~ 170)
The Rutherford cross sections for Ai and 0 are listed in
Appendix 6 as 0.2128 x 10-24 cm2/sr and 0.0741 x 10-24 cml/sr,
respectively. The compound stopping cross section factors are
those found in the depth scale example (EXAMPLE 4.2):
(O]~,OJ =350~1O-15 eV cm 2
(SO]~,O, =326xlO-15 eVcm 2
Using Eq. (4.28), we find

H
= crf!Q25
(O]~,O,
A1,0
Table 4.1. Typical experimental operating conditions and

parameter ranges used during acquisition of backscattering
spectra described in Section 4.3.
24
0.2128 X 10- x 10-' x 6.24x lOll x2 x 10'

350xlO
~
15
Experimental parameter
76 cts.
_ crf!Q31i
0,0 -
eo
Analysis ion energy

Beam cross section
Beam current
Integrated charge
lAI,O,
0
24
0,0741 x 10- dO-' x 6.24 X lOll x 3 X 10'

1
Detector energy resolution for 4He ions
326 X 10 '
~ 43 cis (or particle:;.lchannel).
Data acquisition time
Vacuum
PumE!-down time
Several iterations might be required if the energy loss in the lihn

is an appreciable fraction of Eo. The final (Nt)~!) values are
calculated using Eq, (4.3) with E ~ E<O, where E'" is the final
mean energy obtained from the iteration process. If the scattering is
Rutherford and the film is not too thick, then
(Nt)ff) ~
1
-m
E
( Eo
Values
M.V
mmxmm
1.05.0
nA
10200
~C
5-100
k.V
min
Torr
min
1.5 x 1.5
15
5-10
2x10-'
15
The beam energy (or the bending-magnet field) was calibrated

using three (n,y) resonances in "Mg (Endt and van der LeWl, 1967)
and (a,a) resonances in ''N (Herring, 1958) and "0 (Hausser et ai"
1972), which covered the 4He beam-energy range 2437.4 keY to
5058 keY, See Appendix 19 for additional calibration points. The
beam energy was known to within 5 keY and was stable to less
than 1 keY; the beam-energy spread was less than 0.5 keY.
)2 (Nt)SEA
1
Units
(4.33)
4.3 SAMPLE ANALYSIS

This section starts with a very brief description of apparatus
and operating conditions in a "typical" backscattering
laboratory. The specific laboratory that is described was chosen
for reasons of familiarity and concreteness. Procedures for
determining certain necessary experimental quantities, such as
analysis beam energy and the energy/channel of the pulse height
analyzer, are also briefly described. Examples of actual data
analysis follow. Several of these analyses are explicitly done "by
hand" to clearly illustrate the power and simplicity of the
technique. The section concludes with a brief discussion of
computer-assisted data analysis, which is, of COUT~e, current! y
used for most analyses.
The conversion of PHA channel number, n, to backscattered
ion energy, EJ, was accomplished by least-squares fitting

(4,34)
to backscattering peak data (n;, E:) from a very thin film
containing To, Nb, AI, and 0 on a C substrate. The value of E: for
the ith film element was calculated with Eq. (4.1) using known values
of K; and Eo. The value of Do the peak channel for the ith element,
was read from the spectrum (see Appendix C of Chu et (JJ.,
1978). Thus, values of energy/channel, &, and energy intercept,
E', were determined for the amplifier gain used.
53
Chapter 4
The number of analysis ions, Q. incident on the target was

calculated from the total charge, Qt, deposited in the insulated
large! chamber during the run; Q ~ Q'!e, where e = 1.602 x 10-19
C' (Coulombs). It was assumed that the incident ions bore a
cbarge of +e. (See Chapter 15 for possible corrections.) The
cbarge Q' was divided by the dead-time mtio (DTR) to account
for the fact that the PHA did not accept pulses during a portion
of the time while Q' was being collected. The DTR factor was
usually taken as the ratio of the "real" time to "live" time (both
provide element identification. elemental areal densities, the

stochiometric ratio, and a thickness estimate for this film.
Table 4.2. Quantities whose values are usually known prior to
acquisition ofbackscattering data for a particular sample.
Symbol
supplied by the PHA); this factor was usuaUy :s 1.02 [or the
analyses discussed. in the follOwing subsections. Other methods
for determining the dead-time correction factor are discussed in
Chapter 15.
Daily measurement of the areal density ofa secondary standard

Ta film that had been calibrated with a Bi-implanted-in-Si RBS
standard (Eshbach, 1983) provided another correction factor, C B;.
This factor was related to the efficiency of charge collection and
the value of solid angle used. Its value was that required to give
the correct areal density for the Ta standard. Typical CBi values
were near 1.00, with an uncertainty of D.03; this uncertainty of
-3% included the uncertainties in charge collection and solid
angle measurement, as well as the uncertainty (2.5%) in the Bi
RBS standard itself.
&, E'
1:;(9)
A number of apparatus parameters and measured or

calculated quantities were- usually considered known prior to the
acquisition of data for a particular sample; a summuy of these is
given in Table 4.2. Certain other quantities peculiar to a
particular run on a particular sample are listed in Table 4.3.
cr;(E,e)
.;(E)
4.3.1 Tbin-rum anaJysis
Identity (mass and atomic number) of the analysis

ions
Incident laboratory kineti<: energy of the analysis
beam ions
Laboratory angle through which the analysis ion is
scattered
Solid angle subtended by the detector at the target
Angles between the semple normal and the incident
and backscattered beams, respectively
Energy/channel and energy intercept, respectively,
parameters connecting the backscattered energy, EI
ofan analysis ion with the PHA channel number, n,
byE 1 =n&+E'
Correction factor (related to efficiency of charge

collection and solid angle measurement) that gives
correct (Nt)Bi value for Bi RES standard
Kinematic factor for target element i and
backscattering angle e
Cross section for scattering of anaJysis ions of
laboratory energy, E, through angle 9, for the ith
target element
Stopping cross section for analysis ions of
laboratory energy E in the ith target element
Table 4.3. Quantities whose values are determined by a

particular run on a particular sample.
4.3.1.1 The peak integration method

If the backscattering peaks are well separated so the
integrated peak counts, ~, can be accurately determined from
the spect.:rum, then the peak integration method can be applied in
the simple direct manner discussed in Section 4.1. A slightly
modified version ofEq. (4.3)
., _ Ai cos 81C SI exDTR
(Nt)
(4.35)
Q'no'R (E,e)(~)
OR
Quantity
should be used for the calculation of (Nt)i. the areal density of

the ith element in the film. The symbols are defined in Tables
4.2 and 4.3. The ratio, (0/0'010 is the DODRutherford correction
mctor discussed in Section 4.2.2.3. Note that, in instances where
the integrated charge and solid angle are Dot well knO"'ll, the
quantity CBv(Q'!l) can be obtained from the substrate (ifthere is
one), assuming that the stopping cross section factor and
scattering cross section are knO'WD for the substrate.
Symbol
Quantity
Q'
Integrated cbarge deposited on the sample during

themo
DTR
Dead time ratio for the PHA during the run
ni
Channel number (at half-maximum) of the highenergy edge oftbe signal due to scattering from
the itb target element at the sample swfa.ce
nj"
Channel number (at half-maximum) of the lowenergy edge of the backscattering peak due to the
ith target element
Hi.o
Spectrum height (counts/channel) of the signal

due to scattering from the ith target element at
the sample surface
Integrated eounts in the peak due to scattering
from the ith element (in a thin film)
Energy width of the peak due to scattering from
the ith target element (in a thin film)
The spectrum shown in Fig. 4.1 is actually that of a Cd-Fe
film on a Si substrate; the analysis beam ions were 3776 keV
~c. EXAMPLE 4.4 uses data taken from this spectrum to
54
EXAMPLE 4.4. This example is an application of thc peak
integration method of analysis of the two-element thin film

whose spectrum is shown in Fig. 4.1 .
A priori acquisition parameters (symbols defined in Table 4.2):
Eo-3776keV
9=170'
81 =0
8 = 100
2
n - 0.78 Imr
CB , = (0.99 0.03)
= (3.742 0.005) keVlchannel
E' - (8 3) keY
K..(170') - 0.7520
Koa(170') = 0.9039 (from Appendix 5)
E:
=n,' + E' =(757 1)(3.7420.OO5) +(8 3)

.=(28416)keV
E: - n,'+E' = (9101)(3.742O.OO5)+(83)
= (3413 7) keY
K =
E~ = (284IH) =O.752O.OO2
Eo
(37765)
K = E; = (34137) ;;O.904O.002
... Eo (37765)
Therefore, elements A and Bare Gd and Fe, respectively. (Note

that element A could also be Tb, hecause K", = 0.9048.)
- 0.2469 X 10-24 ~
sr
OCd (E ,170') _
R
Calculation of elemental areal densities, (Ntk

First, the values of At are calculated from the integrated counts
in thc regions of interest. In this instance, a constant background,
detennined by the counts per channel just above the particular
peak, is subtracted.
21.53 X 10-24
(3.776)2
2
-1.510 X lO-z, ~
sr
AFe - 64957 - 79 (128) _ (64475 261) cts
(hom Appendix 6)
21
(0.049)(2)(26)'/3
(....)
ACd
0.998
3776
In this case, the background correction is almost negligible; the
calculated uncertainties in the values of Ai are statistical

Next, the areal densities in the Surface-energy approximation,
(Nt)r", are calculated using Eq. (4.35) ",th E ~ Eo:
-1- .:.:(0",.04=9)",(2:e,).:.:(64..:;),-4_/3. _ O. 993
OR Cd
3776
[from Eq. (4.12)J

E" (3776 keY) ~ 5l.4 x 10- 1' eV cm'
," (3676 key) ~ 52.2 x 10-1' eVem'
E"" (3776 keY) ~ 86.3 x 10-1> eV cm'
Ef1rl (3676 keY) ~ 87.5 x Wi' eV cm'
(from TRIM1985)
(Nt\~.c (644i5 261)(LOO8)(O.99 to03Xl602x 10-19 ) at:om;,/a:n1

'Fe
=(2.68 0.08) x IOU!
(757 I)
n'=(660 I)
H~ ~ (1020 20) ct.\
atom,/c:m 2
Then, the mean energy of the 4He ion in the film, .,(1), is
calculated (to first order) using Eq. (4.29) for the first-order energy
loss, <'>E~EA ,of the ions in the film and Eq. (4.30)35 follows:
Parameter values directly associated with the spectrum shown
Q'~20.QJ j.lC
(20.01 x lO"'"6XO.nl x 10OJ )(O.2469 x l(j=Z")(o.998)
(Nt)~ =(0.7090.1121) x 1018 atoms/anI .
in Fig. 4.1 (symbols defined in Table 4.3):
n.-
=103978 - :~ (130) =(103823 323) cts
DTR-l.008
n, - (91O I)
nA~(812
<'>Ei:' =.Fe (Eo)(Nt)~;" +.Cd (Eo)(Nt)~
I)
H~ ~ (640 20) cts
=(51.4 X 10-15 )(2.68 X lOIS)
Integrated counts in spectral regions of interest (initial and final

channel numbers are listed):
channels (789-918) ~ 103978 cts; (920-960) ~ 49 cts
channels (640-767) ~ 64957 cts; (768-788) -79 cts
+ (86.3 X 10-15 )(0.709 x 1018 ) eV

-199 keY
SEA
ji(l)=E o -
Element identification [wing Eqs. (4.34) and (4.I)J,
55
<'>Em
2
=3776_ 199 -3676keV.
Chapter 4
The areal densities, (NtW). including the first-order correction for

energy loss in the film are [adjusting the Rutherford cross section
in Eq. (4.35) to the mean energy in the sample, using the inverse
square dependence on energy]
electron-shell corrections, are believed to be Rutherford at the

analysis energy used. Of course, an uncertainty assigned to the
cross-section ratio can be easily included in the calculation (see
Chapter 16).
The value of the physical film thickness, ~ could be calculated
using Eqs. (4.6) and (4.7) if the density of the film were lmown. It is
not, in this instance. However, it is customary to produce an
(Nt)(t) = (3676)2 (Nt)SEA = 2.54 x 1018 atoms/em2

. Fe
3776
F.
estimate of the film thickness in such cases using the following
(Nt)~ = 0.672 x 1018 atOIns/cm 2
procedure: The actual Gd-Fe mixture is replaced by an elemental
Results of an additional iteration of this procedure using Eqs.

(4.31), (4.32), and (4.33) (notethal e" and gOd are evaluated at (1'
are
bilayer with the same areal densities for Gd and Fe. The
elemental layers are assumed to have elemental bulk densities. The
thickness of this replacement film serves as the estimate of the
physical thickness of the original film. The values of the elemental
bulk densities are N Fe = 8.44 x lOll atoms/em) and NOd ""' 3.02 x
10" atomslcm'(Appendix I). Hence, from Eq. (4.7), we have
~.:! = (52.2 x 10-15 )(2.54 x 1018 )

+(87.5 x 10-15 )(0.672 x 1018 ) eV
2.55 XlO
191
=368lkeV
tGd
(Ntj<2) = (3681)2 (Nt)5EA
Fe
3776
1018
3.02 X 10 22
an =223nm
525 DID .
Summary of the analysis rcsuJts for the two-element film of

Fig. 4.1: The stoichiometric ratio is
Additional iterations produce no further change in the (Nt)j

values. so these are the final values of tbe energy loss in the film,
the mean analysis-ion energy in the film, and the elemental areal
densities glven by the mean-energy approximation.
NFe =3.780.02
NGd .
so the stoichiometry of the film is Gd(O.209:<1.oonFe(O.79U:O.OOI)' The

areal densities are
The average stoichiometric ratio for this film can be calculated

using Eq. (4.5):
(Nt)" ~ (2.55 0.08) x 10" atoms/em'
a~(Eo, 170). (a~)Gd

a~(Eo, 170)
Note that no uncertainty is assigned to this final thickness

estimate. In addition, the judicious analyst should indicate that
this estimate might not accurately represent the actual physical
thickness of the film.
1018 atoms/cm2
(Nt)g; = (0.674 0.021) xl0 18 atoms/em 2
NGd - AGd
0.674
tCdF. -
Fe
(2.55 0.08)
NFe _ APe .
em = 302 DID
8.44 x loT'
=191keV
E(2) =3776 _
l8
(Nt)"" ~ (0.674 0.021) x 10" atoms/em' .
(..Q...)
erR Fe
The film thickness is about 0.5 J.UIl. Note that the uncertainty in
the ratio NFeiNOd is about 0.5%, whereas the (Nt); values are
uncertain by about 3%. This is a typical result and reflects the:
cancellation of the hard-to-measure quantities Q and n from the
ratio.
_ (64475 261) . 21..S,l. 0.993

- (103823 323) 3.521 0.998
=3.780.02.
The peak integration method can be applied in this simple and

direct manner to the analysis of any single-layer film whose
backscattcring peaks do not overlap. It is not necessary that film
composition be uniform as a function of depth. The resulting
stoichiometric ratios will, of course, be the average values for
the film. For very thin films, the backscattering peaks will not
If the "molecular" formula for the film is written as 0dmFeu.. then

0.001. The quoted uncertainties
are statistical, that is, due to the uncertainty in the ~ ratio. The
cross-section ratio bas been regarded as exact in this calculation.
The cross sections for He-Fe and He-Gd, ex~pt for the small
m =0.209 0.001 and n =0.791
56
..
have the nearly flat portions on top but will be sharp peaks; in
this iru;tance, the channel nwnbers of the peak centers should be
used in Eqs. (4.1) and (4.34) ror element identification (see
Appendix C in Chu.' al., 1978).
In the case of the spectrum shown in Fig. 4.4, the peak

integration method can be only partially apptied. The
stoichiometry and areal densities of the Al20lAr) layers can be
completely determined. However, the NiFe layer stoichiomeTry
cannot be determined from these _data because the Ni and Fe
peaks overlap. For this spectrum, the Ni and Fe edges are
separated by less than 10 channels. The surface barrier detector
Mullilayer films
The peak. integration method can also be applied to the
analysis of multilayer films. The spectrum of a trilayer film on a
energy resolution (oE == 15 keY) is about 8 channels, and in

addition, both Fe and Ni have several isotopes. The result is that
this particular NiFe film is probably too Urick for accurate
determination of its stoichiometry by 4He ion beam analysis with
surface barrier detectors. Films afNiFe as thick as 25 run can be
analyzed with 5 MeV 'He (Leavitt el al., 1985). In such cases,
thinner "witness" films, made by the same deposition
procedures, are sometimes used to provide the stoichiometric
ratio, NfJ'NNi. The integrated peak counts, Ar~ and Atr., can then be
calculated from [see Eq. (4.3)J
C substrate is shown in Fig. 4.4. The peaks associated with a

particular layer can be identified by noting peak locations in
spectra taken at different tilt angles (8,). The high-energy edges
of the peaks associated with the surface layer have the same
locations in these spectra. Buriedlayer peak locations shift
toward lower energies as the tilt angle is increased because the
effective overlayer thicknesses increase with tilt angle as (cos
elr]. The locations of peaks associated with a particular buried
layer all shift by about the same number of channels as the tilt
angle is varied. Two backscattering spectra, one taken with 8 I ~
0 and the other with 8 1 between 45 0 and 600 , are usually
sufficient for assignment of the peaks to appropriate layers.
Once the peaks have been assigned to layers, the peak
integration method can be applied to successive layers, starting
with the surface layer.
(4.36)
and the measured integrated ANife peak counts (with A = Ni, B =
Fc), thus providing the Ai values for use with the peak
integration method. As indicated, the analysis of multilayer films
that produce spectra with non-overlapping peaks can be carried
out by hand, but the procedure is tedious and bears a strong risk
of numerical error in the calculations. Computer assistance with
The analysis-ion energy incident on a particular layer is, of

course, the energy Eo minus the energy lost by the ions
traversing the layers that cover the layer being analyzed. The
values of (Nt); for a particular layer should be ca'k:ulated in the
mean-energy approximation before proceeding to the next layer.
these calculations is strongly recommended.
FIG. 4.4. Tbe 1.9 MeV 4He backscattering spectrum of a trilayer film on a carbon substrate. The backscattering signals from the three
layers are clearly separated. The Ni and Fe peaks are not resolved.
57
Chapter 4
Use ofelastic scattering resonance

Figure 4.5 shows a case where a strong non-Rutherford
resonance (Mozer, 1956; Leavitt etal., 1994; see Appendix 7) in
9
the 1H- Be cross section causes sufficient enhancement of. the
Be signal that the peak integration method can be used to
. determine the areal density of the Be layer. The net integrated
count in the Be peak in this spectrum is All< = (2353 '" 172) cts;
the large uncertainty is caused by subtraction of the substantial
signal due to the AbO] substrate. The backscattering spectrum
taken with 1.9 MeV 4He shows DO sign of a Be peale. At the
2525 keV proton analysis energy used, the cross sections for Be,
0, and AI are all non-Rutherford; only the IH-Ge scanering is
Rutherford. The 1H-~e cross-section enhancement factor for
thin targets (Leavitt el at., 1994) is (cr/cr.) = (60 '" 3).
4000
2.525 MeV 'H
CI)
<:
AJ,O,
C'
I~FAk-
::J
0
(.)
i:se
Substitutio~ of these values of

and (olcr0Be. along with the
appropriate values of Q', n, a 6\, DTR, and CBi> into Eq.
(4.35) gives (Nt),. = (313 29) x 10" Be atoms/em'. Ifbulkdensity Be is assumed (Appendix 1), then the estimate of the Be
film thickness is tee = 26 nm. When a nanow resonance and a
thick target are used, the average cross section over the range of
ion energies in the target must be known. This, of course, varies
with the target thickness. The principal source of uncertainties in
the final results is the uncertainty in the measured value of the
cross section. It is clear that considerable work remains to be
done concerning measurement of nonRutherford cross sections.
Non-Rutherford resonance cross-section data are compiled in
Appendix 7.
=101
9u.a
=170
Be
2000
G.
I
",.
I
I
I
I
I
.
......... ......_......
500
420
Channel number
440
FIG. 4.5. The 2.525 MeV lH backscattering spectrum of a bilayer film on a sapphire substrate. The fact that the
]H-~e cross section
is about 60 times the Rutherford value at this energy allows the Be backscattering signal to be observed. section . .It is clear that
considerable work remains to be done concerning measurement of non-Rutherford cross sections. Non-Rutherford resonance crosssection data are compiled in Appendix 7.
Overlapping peaks
Figure 4.6 shows the spectrum of a thin film with
backscattering peaks that overlap. The film consist.. of 50
bilayers of Fe-Mo with a total estimated thickness of about 360
om. The individual layers are so thin that the peaks from
peak integration method that utilizes peak heights and widths

can be used for analysis of films of uniform composition. This
method is descnbed in Section 4.3.1.3 and applied to the data of
Fig. 4,6.
individual layers are Dot resolved by the detection system (liE ==

15 keV, .1E~ == 5 keV). Therefo~, the film can be treated as
a compound film of FeMo, and the appearance of the spectrum
indicates that this "mixture" is unifonn throughout the film. The
peak integration method cannot be used for analysis of these
data in the simple, direct manner described in EXAMPLE 4.4
because the peaks overlap. However, a modified version of the
One procedure for analyzing data of the type shown in Fig.

4.6 is to use computer fitting to separate the contributions that
the different elements make to the total counts in the overlapping
peaks. Peaks of specified shape are least-squares-fitted to the
experimental data; the results of the fit provide the ~ values
with appropriate uncertainties. For a uniform film that is not too
thin. a specified peak shape represented by two half-gaussians
and a trapezoid works well (McIntyre et al., 1987). Figure 4.7
58
. ',"I'
t...
,<
.~
shows the result of application of this fitting procedure to the

data of Fig. 4.6. The total integrated count is AFeMo = 564380;
the fitting procedure divides the counts as
~ 181300 2700
and AMa = 383100 3300, Insertion of these values, along with
appropriate values of Q', DTR, eBb and so on, into Eq. (4.35)
yields (Nt)" ~ (142 5) x 10 16 Fe atoms/em', (Nt)",o ~ (114 4)
at the higher analysis energy. The 4He-O cross section is not

Rutherford at the higher energy (see the resonance near chn:mel
130), but analysis of the substrate is not of interest in this case.
Aoalysis of the data shown in Fig. 4.8 yields (Nt)P.~ (142 4) x
lOl6pe atoms/cm2, (Nt)MQ = (115 3) x 1016 Mo atoms/cm2, and
A,.
N,oiNMo ~
16
Of course, the simplest procedure (and often the quickest)

for analysis of the film of Fig. 4.6 would be to produce a
spectrum rith separated peaks so that the peak integration
method could be applied in the simple manner described in
EXAMPLE 4.4. Sueh a spectrum (Fig. 4.8) with separated peaks
was produced simply by increasing the ~c beam energy to 3.8
Mev. The 4 He-Fe and 4 He- Mo cross sections are still Rutherford
1.9 MeV 4He
4000
(1.236 0.006). Comparison of these results with
those produced by computer-assisted separation of the Fig. 4.6

peak data indicates that the (Nt)t values and their uncertaioties
are about the same, but the accuracy of the average
stoichiometric ratio. NF~'Mo. is substantially improved by USe of
the higher-energy data. This comparison not only points (Jut a
deflillte advantage to the use of higherMenergy analysis beams,
but also instills confidence in the reliability of the compll.terassisted peak separation procedure.
10 Mo atoms/em', and NfoiNMO - (1.24 0.02).
Q' = 25!IC
&LAS
= 170"
1/1
C
o'"
U2000
500
1000
Channel number
FIG. 4.6. The 1.9 MeV 4He backscattering spectrum ofa 360 nm Fe-Mo multilayer (50 bilayers) film on a sapphire substrate. The
backscattering peaks for Fe and Mo overlap.
M
4.3.1.2 Stoichiometry by surface spectrum heights

The stoichlometry of thin films whose composition as a
function of depth is uniform can be obtained from the spectrum
heights of the edges corresponding to scattering from the surface
atoms. The stoichiometric ratio for the two-element film. A..,B... is
where the explicit expression for [E.];,' [from Eqs. (425) and
(4.17)]
IE lAB ';'m[KAeACEo) + EA(KAEo)]

oA
cose l
cose,
[from Eq. (4.28)1
m _ NA
H1~o "a [eol1B

HB.O "A [Eola
---~-,;Jf"-'-xB
NB
(4.37)
(4.38)
59
L ..
Chapter 4
The symbols are defined in Tables 4.2 and 4.3. A similar

expression for [o]~ can be obtained by' exchanging A and B
and m and n.
Note that the ratio [eo]:' /[eoCI contains the ratio min, which
is, of course, the object oftbe calculation.
3000
2500
2000
-8
III
. 1:
~
1500
.',
- ....
~.~
,,,,;:::;;.,,,,
~
.,
Mo
1000
','
Fe
500
o
460
560
660
760
Channel
860
960
'.\
FIG. 4.7. The least-squares-fit of specified peak shapes (two half gaussians and a trapezoid) to the data (dots) of Fig. 4.6 provides
integrated peak. counts, Apt and A Mo for use with the peak integration analysis procedure. This fitting procedure is applicable to films
that bave uniform composition.
This (t:o] ratio seldom departs from 1.00 by more than a few
percent. An itemtive procedure can be used to obtain the min
ratio. For the first iteration, min is calculated from appropriate
H:, H~,
(fA.
as that of thin films providing that the composition of the
material is uniform as a function of depth. (See EXAMPLE 4.5.)
and O'B values using Eg. (4.37) with the [eo] ratio
4.3.1.3 Peak width methods

Data of the type shown in Fig. 4.6' can also be analyzed using
the -peak width method, Calculation of the areal density by this
method does not require knowledge of either Q', the charge
collected, or n, the detector solid angle. The method is usually
not as: accurate as the peak integration method, it but can be used
in cases where the effiCiency of charge collection is suspect.
taken as 1.00. This value of min is then used in Eq. (4.38) to

calculate the [eol mtio, which, in turn, is used in Eq. (4.37) to
obtain a more accurate value of min. The procedure is iterated
until the min ratio ceases to change. Note that this technique can
be used to determine the stOichiometry of bulk material as well
60
,
:/.
:1.8 MeV 4He
4000
Q'
= 50 JJC
NF.
NMo =1.236O.OO6
~::;,
o
2000
o
AI
I..----_-..J(
o
500
1000
Channel number
FIG. 4.8. The 3.8 MeV 4He backscattering spectrum of the film whose 1.9 MeV spectrum is shown in Fig. 4.6. The backscattering
peaks for Fe and Mo are separated.
For the case of the two-element compound film AmBIl' Eq.

(4.24), with x = 1, yields the following expressions for areal
densities (Nt)A and (Nt).:
and peak integration methods. Note that combination ofEqs. (4.36),

(437), and (4.40) yields
AAB
(4.42)
(4.39)
where AEA and AEB are the energy widths (full width at half-
which can be used to obtain the integrated peak counts necessary

for use of the peak: integration method. In essence, once the film
stoichiometry is obtained from the data with Eq. (4.42), Eq.
(4.36) can be used to find the contributions of the individual
elements to the combined peak for use in the peak integration
method.
maximum, F\VHM) of the peaks. The quantities [,];' and [s]:'

can be approximated using Eq. (4.38) with Eo replaced by Ii, the
mean energy of the analysis ions in ~e film. Note that factors in
Eq. (4.39) such as m/[E]:B depend on the stoichiometric ratio
min. This ratio can be evaluated using a combination of stepheight and peak-width data. From Eq. (4.39) or (4.24), we have
In summary, a comparison ofrcsults of the various methods

for analyzing the FelMo multilayer of Figs. 4.6, 4.7, and 4.8 is
provided in Table 4.4. Clearly, the procedure that produces the
results of highest accuracy is use of the peak integration method
on data with separated peaks. It is particularly important that the
peaks be separated if the film composition varies with depth.
However~ it is frequently not possible to produce spectra whose
peaks are separated. If the film composition is uniform, then
computer-assisted separation of overlapping peaks frequently
produces areal densities that have accuracies comparable to
those obtained in the preceding examples. However, the
accuracy of the stoichiometric ratio is Usually degraded by a
factor of 2 or 3 compared to that discussed here. The use of the
peak mdth method (EXAMPLE 4.6) produces results of inferior
[1:1" _ABA
[Elii' - ABB
(4.40)
Hence, from Eq. (4.37)
~=&= H~o. au. AEA
n
NB H3 CJA AB.
---_______ .______
(4.41)
For films ofuniforrn composition, a useful analysis alternative that
avoids much of the labor associated with the peak width method
as applied in EXAMPLE 4.6 is a combination of the peak width
61
Chapter 4
accuracy because this method depends on the accuracy of

stopping cross sections (estimated to be about 5%). However, it
is a method that can be carried out by hand for the case of
overlapping peaks. Further, it can be used as a crude (5%) check
on the accuracy of charge collection and solid angle
measurement. If the peak width method is applied in its pure
fonn [Eq. (4.39)], it can be used to check for the presence of
accidental channeling. The combination peak v.idth and peak.
integration method (EXAMPLE 4.7) produces results of
accuracy comparable to the peak.
where (8 3) etsfehanoel is the background on the high-energy

side of the FeMo siJn1al.
N
Calculation of the stoichiometric ratio, ~;
N.,
For the first iteration, use of Eq. (4.37), with the ['0] ratio taken
as 1,QO, provides
NFe _m. (7004O) 9.251 0.992

N Mo - -; (1520 40) . 3.521 . 0.996
EXAMPLE 4.5. This example illustrates the determination of

film stoichiometry from measured values of spectral step heights
(for the uniform film of Fig. 4.6).
Then, use of Eq. (4.38) and the interpolated stopping cross

sections given in the first step, with min = 1.21 to calculate the
[eo] ratio, gives
A priori acquisition parameters (the symbols are defmed in

Table 4.2) are (the analysis beam is 4Hc)
9 = 170'
8,=10'
Eo= 1892 keY

a,~o
n = 0.78 rnsr .
K,,(I 70') = 0.7520
(from Appendix 5)
K" x Eo ~ 1423 keY
~ = (1.821 0.004) keV/chanoel
cr~"(EI'1700)J =3.521
( "~'(E,,170')
'" 9.251
(~J
OR
=0.996
Fe
1.21 08.
(1.21)
C El = (1.05 0.03)
"",(170) = 0.8471
"", x Eo ~ 1603 keY

E' ~ (10 3) keY
l""o
[EO]~~
(from Appendix 6)
(~J
erR
=0-'lM~of.,-
+[(0.8471)(93.0)+
[(1.21)[(07520)(72.0)+
.
=0.992
Mo
[(0. 847J)(72.0) + ;~:s] 1

:'~;S]
;:~~] :
+[(0.7520)(93.0)+_1_02_J
0.985
=(1.03O.OS) .
[from Eq. (4.12)]
From TRIM1985 interpoJationprovides the values
FinallY, from Eq. (4.37), the stoichiometric ratio,

NFdNMois
""(1892 keY) = 72.0 x 10-" eVem';

.M(1892 keY)=93.0x 1O-"eV em'
N pe
N Mo
""CI603 keY) = 76.5xI0-" eVern';
(1.210.OS)
(103 0.05)
(117010)
.
Relevant parameter values directly associated -with the
because further iteration of this procedure produces no change in

NFJNMo' Note that this value is in agreement with that (1.24
0.02) obtained by use of the peak integration method with the
computer-assisted peak separation shovm in Fig. 4.7 and also
that (1.236 0.006) obtained using 3.8 Mev He ions shown in
Fig. 4.8. The use of step heights to obtain stoichiometric ratios
usually produces results of rather limited accuracy.
spectrum shown in Fig. 4.6 (symbols defined in Table 4.3):

DTR~ 1.002
= (700 40) ets
H~~ ~(l52040) cts
EXAMPLE 4.6. This example illustrates the determination of

elemental areal densities by the peak width method for a unifonn
''''(1603 keY)~ 98.7xlO-" eVem'

.'"(1423 keY) = 79.1xlO-" eVern';
e Mo (1423 keY)::: I02xlO- IS eV cm2
H;:;
two-element film (for the data shown in Fig. 4.6).
Q'=20.01 ~C
nF" = (777 I)
n,; ~ (549 I)
TIM, ~ (878 I)
OM,' = (643 I)
AF,M, ~ 564380 - (8 . 3)(371) = 561412 1343 ets
The a priori acquisition parameters and parameter values

directly associated with the spectrum shown in Fig. 4.6 are
explicitly listed in EXAMPLE 4.5. Symbols are defined in
Tables 4.2 and 4.3.
Calculation of peak energy widths, 6.EFe and AE Mo
62
.~
1lEf~ - (n fe - n' Fe)ti
(Nt):' ~ (1.23 0.06) x 1018 atomslem'
~ [(777 I) -(549 1)](1.821 0.004)
because further iteration of this procedure results in no further

change in the (Nt) values.
=(4153) keV.
AE"" =(4283) keV.
Table 4.4. Comparison ot results of analyses of the FeJlv10 film (of

Figs. 4.6, 4.7, and 4.8) by various anatysis methods.
Calculation of the stoichiometric ratio [Eq. (4.41)]:
~=
NMo
(70040) (9.251) (0.992) (4153)

(15200 t 40) . (3.521)' (0.996)' (4283)
=(l.l70.08).
Calculation of areal densities (Nt)Fe and (Nt)Mo:
In the surface-energy approximation. use ofEqs. (4.38) and
(4.39) yields
6E Ee
(Nt)SEA
Fe
[Eo Jj:eMo
e /m
(415 3)
(278 14)
1013
X
=(1.47 0.07) X 1018 atoms/cm 2
(Nt)5EA
Mo
.:IEMo
(E lFe.YoIn
o Mo
(428 3) lO'S
(347 17) x
NF.
Anatysls method
(NtlF,
(Nt)!to
Peak Integration method with

separated peak data. Fig. 4.8
142:4
115.3 1.236 0.006
Peak integraUon methOd, with

computer-assisted separation
of over1apping Fe and Mo
peaks, Fig. 4.7.
142 5
fl44 1.24.0.02
Peak wid1h method (with

spectrum heights).
Example 4.6, Fig. 4.6.
1447
1236 1.17 0.08
Combination of peak width

aryd peak integration methods,
136 7
116::t: 5
1.17:t. 0.08
1.17 0.10
NMo
Example 4.7, Ag. 4.6.
Stoil:::hiometry by spectrum
4.5, Fig. 4.6
hej~ht, Exam~le
0.26 0.06) x 1018 atoms/em 2
-,Nt! values are in lJnits 0110 16 atomslcm2.
To first order, the energy loss of the analysis ions in the film is
[Eq. (4.29)]
.:IE~EA =(1.49) (72.0) + (1.23) (93.0) =224 keV
EXAMPLE 4.7. This example illustrates use of a combination

of the peak width and peak integration methods to obtain areal
densities in a two-element thin film of uniform composition
(Fig. 4.6).
.~
so the mean analysis-ion energy is [Eq. (4.30)J, to first order,

E,l, =1892 - (224/2) = 1780 keV. This
With the data from Fig 4.6 listed in Example 4.5, Eq. (4.42)
yields
mean energy, (1), should be used in the evaluation of the
[Eo];' and [Eo t,' faclors by Eq. (4.38) to obtain more accurate
AFe
values of (Nt)" and (Nt)M' through use of Eq. (4.39). It is, of

. course, necessary to interpolate additional values of eFe and EMQ
for the relevant energies from data -given in TRIM:! 985 The
results are
[e(E(1) )]~O
m
1+
Me
(H~~8
)( AEMo )
HFeMo
AEF
e
FerO.
561412 1343
- 1+ (152040).
C4283)
(288 14) x 10-15 eV crn 2
(700 40)
(415 3)
=(173000 7500) cts

A Mo =(388000 7500) cts
eMO
[ (-W)t
AFeMo
_ (348 17) X 10-15 eVem2
The areal densities in the mean-energy appro;Umation are [Eq.

(4.39)]
Upon substitution of these values into Eq. (4.35). along with

appropriate values of the other required parameters (listed in
Example 4.5), the areal densities in the surface-energy
approximation are obtained a5
(Nt);:'" ~ (1.44 0.07) x 1018 atoms/em'
63
Chapter 4
The symbols are defined in Tables 4.2 and 4.3. Note that the
ratio [E,l~n I[e,l:n in the generalized form of Eq. (4.37)
(Nt)~~ = (1.52 0.08) x 1018 atoms/cm2

(Nt)~
=(1.31 0.05) x 1018 atoms/ cm2 .
contains' the ratios min, ... , mlp, which are, of course, the objects
of the calculation. As before, these [~] ratios seldom depart
from 1.00 by more than a few percent. An iterative procedure
can be used to obtain the min, ... , mlp ratios. For the first
iteration, min is calculated from Eq. (4.37) with the ['o) ratio
taken as 1.00. These ratios can then be substituted into Eq. (4.43)
to calc':!late the [so] ratios, which are then used in Eq. (4.37) to
obtain more accurate values of min, and so on.
Two additional iterations. utilizing Eqs. (4.29) through (4.33),

yield the areal densities in the mean-energy approximation as
(Nt)~
= (1.36 0.07) x 1018 atoms/ cm2
(Nt)~
=(1.16 0.05) X 1018 atoms/ cm2
The backscanering spectrum of a ceramic glass sample is

exhibited in Fig. 4.9. Cross sections for 4He on all indicated
elements are nearly Rutherford except for small corrections [Eq.
(4.12) due to electron shells. Substitution of H;::n step-height
width method. Its principal advantage is thaI it can easily be

carried out by hand, particularly if peak integration calculations
have been computerized to perfonn the mean-energy

approximation. The determination of stoichiometry from spectral
values (obtained from the SPectrum) into Eq. (437) with the ['o)
ratio taken as 1.00 yields the stoichiometry listed in Table 4.5.
Atomic fractions can be obtained from the stoichiometric ratios
provided by Eq. (4.37). For example, the 0 atomic fraction, F o,
is givcn by
step-heights (EXA1vIPLE 4.5) produces results of only modest

accuracy, the calculation can be done by hand. This method is
also applicable to bulk materials of uniform composition. It is

worth noting, again, that all of the preceding analysis methods
(except the first) apply only to the case of films whose
composition does not vary with depth.
(4.44)
4.3.2 Surface layers of bulk materials

4.3.2.1 Stoichiometry by surface spectrum heights
The stoichiometry of materials whose composition does not
vary with depth can be detennined from measured surface
spectrum heights and use of generalizations of Eqs. (4.37) and
(4.38). If elements denoted by Am. Bn ... D~ are present in the
material, terms such as H~ and [o]~ in Eqs. (4.37) and
The 0 atomic percent is 100Fo . Note that the uncertainties in the

atomic percentages are rather large. The principal source of
uncertainty is the difficulty in reading precise values of H~.o
from the graph (see Table 4.5). In view of these large
uncertainties, calculation of the [f:o] mtios using Eq. (4.43) is not
worthwhile in this instance.
(4.38) should be replaced by the terms H;::n and [,l~-'>,
where,
for
example,
Table 4.5. Summary of analysis results obtained from the spectrum
of the ~ramic glass sample shown in Fig. 4.9.
[e lAB... n,; m [KAeA(Eo) + eA(KAEo)].

o A......
COS 6
cos 6
1
(4.43)
dR (1 MeV, 170')
(x to-" .m'lsr)
Atomic
perce'nt
Element
(counts)
0
Na
SI
365:s: 30
0.2965
63.8:s: 1.6
125 t5
0.5993
10.8,. 1.0
410::1:: 15
0.9905
1.861
21.5:t.1.3
707
Zn
175 6
Cd
64
III':B .. D
1,0
9,.,
4.701
12.09
2.0 O2
1.9O.1
0.039 0.004
J
,J
:?
,,
,I
I
I
1.9 MeV 4 He+
9 LAe = 1700
Q'=50pC
~NEl1,S~K~Cdo.04
2000
Sc
'0"
(J
1000
Zn
HSl,oI
0
j Cd
HZn,o
1000
500
Channel number
FIG. 4.9. The 1.9 MeV 4He backscanerlng spectrum ofa ceramic glass. The indicated stoichiometry was determined from the step
heights. For uncertainties in the subscripts (atomic percentages), sec Table 4.5.
4,3.2.2 A deficiency method
A deficiency method can be used to analyze near-surface
two-element materials even when the steps for both elements are
not ,,;sible on the spectrum, provided that the identity of the
lighter element is knovro. The spectrum for the he-dvier pure
element is compared with that for the two-element material. The
deficiency in the heavy-element signal caused by the presence of
the light element is noted and used to obtain the material
stoichiometry at the sample surface. Application of this method
~ _ N~Ox _ [~o]~ (H~.O

- N~Ox - [EO]a,
H~~x-
1)
(4.45)
to obtain the near-surface stoichiometric ratio. Equation (4.45) is
obtained from Eqs. (4.27), (4.28), and (438). with

KAEB (Eo)
cos 91
to a case of oxidation of tungsten is illustrated in Fig. 4.10. The

spectra of pure W and WO x samples are compared. The numbers
of analysis ions incident on the two samples are the same, of
course. The signal from 0 in the WO x spectrum (edge near
channel 370, not sho\VD.) is not visible on top of the W signal. If it
is assumed that the deficiency in the W signal above channel 800 is
caused by the presence of 0 in WOX! then the surface stoichiometry
+ EB (KAE o )
cos 92
(4.46)
and x = Dim, A = W, and B = O. Interpolated values of the stopping

cross sections f.w and 0 (TRIM1985) must be used in the
calculation ofx. The resulting near-surface stoichiometry is W03 ;
the 6% tmcertainty is due to uncertainties in the step heights and
energy loss data. From the appearance of the spectrum, it is
apparent that the WO) layer extends to a depth corresponding to
about channel 890. The W0 3 film thickness, t, can be estimated
by inserting the integrated counts, A w, between channels 890 and
can be calculated by inserting measured W step heights, H:,ll and

H:~' , from Fig. 4.10 into the expression
960 into Eq. (4.3) to calculate (Nt). and then using Eq. (4.7) to
estimate t. Of course, bulk density is assumed for the W0 3 film;
consequently, no uncertainty is quoted for the thickness
estimate.
65
Chapter 4
Q'=5j.C
::
-'"
2000
________
I:
=3.G4~O.17
/"reW
------------L ,
--~
:s
(.)
I
I
1000
\
I
I
'" T
W.lnWOx
I HWOx
t---'43oA
800
900
Channel number
of
--J \
W,o
1000
FIG. 4.10. Portions of 1.9 MeV 4He backscattering spectra

a pure tungsten sample (dashed line) and an oxidized tungsten sample
(solid line). The spectra were obtained under identical experimental conditions (same integrated charge, BLob, etc.). Analysis by the
deficiency method (see Section 4.3.2.2) produced the WO J stoichiometry and film thickness shown.
66
5000 r---~------~
4
4000
characteristic of many samples, determination of sample

stoichiometry" as a function of depth is of considerable interest.
In this section, two examples of depth profiling of the impurity
o in two Mo/Si multilayer samples are discussed. The samples,
46A and 46B. are identical multilayers consisting of 11 layea:-s of
M 0 HiO A each) aod 10 layers of Si (-100 A each) on Si and C
substrates, respectively. Rutherford backscattering spectra of
these samples are shown in Figs. 4.11(.) and 4.12(0). The
individual layers are not resolved, so the films can be regarded
as mixtures of Mo aod Si. For Sample 46B, the impurity 0
signal is clearly separated from the C, Si, aod Mo signals in the
spectrum. so the 0 atomic fraction can be depth-profiled usiog a
generalized version of Eq. (4.28) to convert observed 0 peak
spectral heights into concentration ratios at appropriate depths.
For Sample 46A, the 0 signal is essentially unobservable in the
Rutherford backscattering spectrum [Fig. 4.12(a)], so the 0
depth profile cannot be obtained from these data. In fact, SaITlple
46B was intended to witness the amount of 0 in Sample 46A.
The purpose of the measurements (described later in this section)
utilizing the strong non-Rutherford resonance in the ~He-O cross
section to compare 0 depth profiles in Samples 46A and 468
was to investigate the validity afthis witness procedure.
a)1.9MeV He o.,sopC ~=17rr
MoISt
~ 3000
170 nm
MO'\.
t.,
"""""',----
02000
SI
1000
0
~1
200
1'\
400
i.
600
800 1000
Channel
(a)
An approximate 0 depth profile for Sample 468 can be quickly
obtained by the following procedure: The peak integration method

(Section 4.3.1.1) can be used to analyze the data of Fig. 4.11. The
results are listed in Tahle 4.6; note that the total film thickness is
estimated to be -170 TIm. Of course, the Mo and Si are assumed to
be distributed in pure elemental layers of bulk density, and the
r---------------~60
200
. b)
50
-li
40
IJJ
100
50
s~~e
'00
'00
30~
..
o
::~::
20.2
/_
.....
.:'.':... ,.
.~,::~,...:.............. \... .
presence of 0 is ignored in making this thickness estimate. ~Use

the average 0 atomic percent for the tihn is 15% and the
concentration appears to vary linearly 'Nith depth, the spec1:n..un height (53 cts) near the center of the quadrilateral sketched on the 0
peak data [Fig. 4.11 (b)] cOITeSJlonw to 15 at % O. This establishes
the vertical scale as 3.5 counts per 0 atomic percentage given that
spectrum height is proportional to concentration for low irr..purity
concentrations. If it is assumed that the two peaks near channels 370
and 320 [Fig. 4.1 1(b)] are doe to 0 on the front and back runaces of
the film, then the horizontal scale is 3.4 om per channel. The 0
atomic percent varies nearly linearly from about 9% at the substrate
interface to about 200/C at the film surface, as indicated in Fig.
4.11(b).
8.
Depth (hm)
-e
c
150
:,:.. .
10
,
~.,
OL---........:;:..c.....:=:.::Jt.:.~............~h;:.lO
200
250
300
350
400
Channel
(b)
Table4.6. Results of a peak integratiOn analySis of the RBS data

lor Sample 468 shown in Fig. 4.11.
FIG. 4.11. (a) The 1.9 MeV 'He Rutherford backscattering

spectrum of Sample 46B, a 170 nm MoiSi multilayer film on a C
substrate. (b) An expanded view of the 0 peak region used for
depth profiling 0, with the 0 atomic percent as a function of
depth below the surface indicated.
Element
Mo
4.3.2.3 Depth profiling

All detailed examples of sample analysis presented 50 far
involve films or materials whose compositions do not vary with
depth below the surface. Because variable composition is
Si
0
67
Areal density
(x 1015 atoms/Cm2)
percent
406 19
531 .25
161 .12
37.0.0.4
48.4 0.4
14.7.0.5
Atomic
t
(nm)
63
106
1
Chapter4
The strong non-Rutherford resonance in the 4H e-O cross

section at a 'He energy of 3.04 MeV (Cameron, 1953) can be
used to depth profile a in Sample 46A. The resonance is narrow
(-10 keY wide), strong (a/crR " 23 at the maximum), and
moderately lsolated. For 4 He energies slightly greater than 3.04
MeV, the 0 signal from the film region occupied by the
resonance is strongly enhanced compared to the Si signa1. Figure
4.12(b) shows the strong 0 peak due to 0 at the surface of
Sample 46A. The net integrated counts in the 0 peak were
obtained by subtracting the large Si background. Backscattering
spectra for both Samples 46A and 46B were taken at several ~e
energies above and below 3.04 MeV, and the integrated counts
in the 0 peak: were obtained in each case. The net 0 peak counts
as a function of incident "He energy are plotted in Fig. 4.13 for
both Samples 46A and 46B.
The vertical scale, in
atomic percent, was detennined
be 20% (as was determined from the RBS spectrum of sample

468 as described previously). The horizontal scale, in terms of
nanometers, was determined from the calculated 57 ke V energy
loss in Sample 46B (using appropriate stopping cross sections
with areal densities given in Table 4.6) and the estimated 170nm thickness. Thus, the energy-to-deptb conversion is 3.0
nmlkeV. The resulting horizontal depth scale is shown at the
bottom of Fig. 4.13. Inspection of Fig. 4.13 indicates that there is
considerably more 0 in Sample 46B than in Sample 46A. Most
of the 0 in Sample 46A is very near the surface, and Sample
46B does not senre as a good witncss for Sample 46A as far as 0
is concerned.
5000.--------------------.
a) 1.9 MeV 4He Q' =50 JC 9t.Ae = 170
5000 ,----------------------,
b) 3.035 MeV"He Q';:: 50 pO &tAB =170
4000
terms of 0
by requiring that the 0 atomic percent at the Sample 468 surface
.sA
~3ooo
'"
o
U 2000
1000
1000
i
_. _ 0 ~=_'="~......,.:'7_'__'_'_':"~~-'--"-_:_J
-- -- .._- .- -_.- 0 . -- 200 400
600 800 1000
SI
~_ ~
--~
200
400
600
Channel
Channel
(a)
(b)
800 1000
FIG. 4.12. (.) The 1.9 MeV 'He Rutherford backscattering spectrum of Sample 46A, a 170-nm MoiSi multilayer film on a Si substrate.
(b) The 3.035 MeV 'He backscattering spectrum of Sample 46A. The enhanced o peak in (b) is due to the strong non-Rutherford
resonance in the 4He-O cross section at the .(He laboratory energy of 3.04 MeV.
68
Depth profile of Oxygen in MolSl multllayers
-onSIC46A)
____ OnC(46B)
"" 10,000
.S:
.l!l
<::
20
--He-O
x-section
8.
c
10 o'!
.!:!
::r
o><
5000
o L~,---,_..,......:~==::::;:::",~~-,-_-,-~--.J 0 ~
3100
3080
3060
3040
3020
He beam energy (keV)
3000
E'
2000
1000
Depth below surface (A)

FIG. 4.13. Oxygen depth profiles for Samples 46 A and 468 obtained from the data shown in Figs. 4.11 and 4.12.
Energy (MeV)
.0.5
3000
1.0
1.5
2500
-""..
0
(.)
2000
1500
1000
500
0
0
200
400
600
800
1000
Channel
FIG. 4.14. A RTJ1\..fP simulation (smooth line) of the spectrum shown in Fig. 4.6 (rough line). The experimental parameters given in
EXAMPLE 4.5 and the Fe and Mo areal densities given in tile first row of Table 4.4 were input to the program
69
Chapter 4
4.3.3 Numerical calculations
discussed briefly. For a more detailed discussion of detector ';,

properties, see Chapter 16.
4.3.3.1 Areal density calculations

A basic calculation in backscattering analysis involves the
use of Eq. (4.35) in determining elemental areal densities (Nt
values) from peak areas and experimental parameters. This is the
peak. integration method described in Section 4.3.1.1. Because of
energy loss of the analysis ions in the sample, scat1ering takes
place over a range of energies. Because this calculation involves
the scattering cross section, which is a function of energy, an
integration should be performed. One approach is to divide the
target into thin layers for the numerical integration.
The primary factors t4at must be taken into consideration in

the design of a backscattering measurement system are the
sensitivity, mass resolution, depth of analysis, and depth
resolution required to obtain the desired information. For a given
measurement. the achievable limit for each of these quantities
depends on the choice for a number of ex.perimental parameters
such as the ion beam and the ion energy. The type of detector
used for analysis also influences the ultimate quality oftbe data
obtained. The final data generally result from a complex
interplay of effects arising from each decision.
In the mean-energy approximation. an iterative procedure is
used as described in Section 4.2.4.6 and EXAMPLE 4.4. This

procedure is usually carried out with a microcomputer. The
algorithm used is essentially that demonstrated in EXAMPLE
4.4; however, the calculations often involve samples containing
several layers, each with several elements. A parameterization of
stopping cross sections, as described in Section 4.2.4.1, is an
essential ingredient of programs using this procedure, because
the scattering cross section in Eq. (4.35) must be evaluated at the
mean energy of the beam in a particular layer. Another important
function of programs calculating areal densities from peak areas
is the propagation of uncertainties in the experimental quantities
to obtain reasonable uncertainties in the final elemental area
densities and stoichiometric ratios.
4.4.1.1 Sensitivity
The most obvious factor affecting the senSitivity of a given
back.scattering measurement is the scattering cross section, which
clearly varies with the mass, charge, and energy of the incoming
ion. However, the ultimate sensitivity is also influenced by the
level of counts from other sources (for example, pulse pileup or
heavyatom, thick-target scattering) underlying the region of
interest An additional consideration is the practical limit on the
time available to do the experiment. If a different beam provides
a higher cross section but is o'n.ly available at (or limited to) a
much lower beam current, the achievable sensitivity might be
less than that obtainable using ions with a smaller scattering cross
section. Acceptable beam currents might be limited by detector
considerations. Finally, sensitivity is limited by beam effects on
the target such as beam heating and sputtering, which can
destroy fragile targets.
4...3.3.2 Computer simulation

Another approach to analyzing backscattering spectra with
overlapping peaks is by computer simulation involving the
calculation of the energy sPe(:trum of backscattered particles
using a specified target composition and experimental
parameters. The procedure is to alter the target composition until
the calculated and measured energy spectra are closely matched.
This can be done by manually changing the target composition
for each iteration or by using a leastsquares fitting procedure to
find the target composition that best fits the measured spectrum.
From the early study of this approach by Ziegler et alA (1976) to
more advanced computer codes such as RUMP (Doolittle, 1985,
1986, 1990), SlMNRA (Mayer, 1997,2007), and Data Furnace
(Barradas, 1999, 2004, 2007), various ion beam analysis
software are now available. Readers are referred to Chapter 14
for more information on computer simulations. As a simple
example, a RUMP simulation of the spectrum of Fig. 4.6 is
shown in Fig. 4.14. Information about this program can be
obtained from http://www.genp1otcomlindex.hlm.
4.4.1.2
'.1
,
~.
Mass resolution
The achievable mass resolution for a given measurement

depends on factors involving the ion beam, its energy, the target
composition, and properties of the detector. The actual separation
in energy, .6.E, for scattering from atoms having adjacent values of
M, is given by Eq. (4.8) and depends on the ma:;ses of both the
incoming ions and the target atoms, as well as the ion energy.
The ability to distinguish betwcen adjacent values of M 2
however, depends also on the overall energy resolution 8E of the
experiment [Eq. (4.9)]. The overall energy resolution BE depends
on the detector energy resolution, energy straggling of ions
within the target, and the energy spread of the incoming ion beam.
Other experimental parameters such as the size of the beam spot
and the solid angle of the detector also influence the energy
resolution.
4.4 EXPERIMENTAL CONSIDERATIONS

4.4.1 Design factors
Choices about ion beams, ion energies, and detectors for
backscattering must be made in the context of the measurement
pl3.Wled. In the following sections, we discuss some common
choices for ion beams and energies, considering the advantages
and disadvantages of each. Detector considerations are also
4.4.13 Depth resolution

When an experiment requires depth profiling of a sample, one
must consider both the accessible depth and the depth resolution
when choosing an experimental technique. Both are influenced
by the same factOlS. Clearly. if an ion loses energy rapidly, the
surface thickness producing energy-resolved particles will be
1,
.J,
70
Backscatterlng Spectrometry
smaller. The energy loss factor, [8], depends on the incoming ion
greatly enhanced for 3.05 MeV 4He scattering, where the elastic
cross section is resonant. For scattering with heavier ions,. the
mass resolution. improves, with optimum results obtained for a
22 MeV 12C analysis beam, where the high-energy edges
resulting from scattering from the two i~otopes of eu are dearly
separated. Resolution worsens for higher beam masses, Depth
resolution for high-mass target atoms also improves with
increasing beam mass, with the 12C spectrum distinctly
indicating an increase in Y concentration with depth.
and its energy, the target composition, and thc orientation of the
target. The achievable depth resolution ox for a measurement [Eq.
(4.26)] is also proportional to SE, so Sx dcpends on the detector
resolution, the energy stability of the beam, and energy
straggling in the target, as well as the parameters that determine
[8].
4.4.2 Ion beams and ion energies
Originally, backscattering analyses were performed almost
exclusively using protons or alpha particles with energies in the
Rutherford scattering regime for most target elements. Although
this analytical technique is convenient, it does not provide the mass
resolution, sensitivity, or depth resolution required for some
measurements. Light ion beams of both higher and lower
energies are currently being used in specialized applications that
require high sensitivity and bigh depth resolution. Heavy ion
beams have also been shown to offer enhanced results over
conventional techniques for some applications. The main features
of the various techniques are compared in Table 4.7.
4.4.2.2 Medium-eRergy backscattering

Considerable work has also been done using lower-energy
ions for backscattering analysis. This technique, called mediumenergy backscattering (M:EBS), commonly employs analysis
ions ranging from 'He to 12C with typical incident energies
between 100 ke V and 500 keV. Because of the relatively poor
energy resolution of surface barrier detectors, other analysis
systems such as time-of-flight (TOF) detectors (Mendenhall and

Weller, 1989, 1990a, 1990b, 1990c) are used to detect the
scattered ions.
One of the obvious advantages to MEBS is the larger elastic
scattering cross section. It should be noted, bowever, ,that
screening corrections to the cross section are generally necessary
in this energy range. As for any technique, the sensitivity of
MEBS is limited by background levels. Thus, MEB S is
relatively insensitive to light target constituents such as ox:ygen
in the presence of heavier target constituents. However, MEBS
with TOF detection can provide improved sensitivity for heavier
species because there is no background arising from pulse pileup ~
as in a surface barrier detector. (The background arising from
high count rates in a TOF detector is flat and distributed
uniformly across the entire spectrum, permitting easier and less
ambiguous subtraction.)
4.4.2.1 Comparison of beams and energies

In 1991, several backscattering techniques were compared for
the analysis of a high-temperature superconducting film,
y IBa2Cu3~ _~ deposited on an AbO) substrate (Rauhala et aI.,
1991). Figure 4.15 shows backscattering spectra obtained using
2.5 Me V protons and alpha particles of varying energies. Figure
4.16 shows similar spectra obtained using more massive analysis
ions. In each case, the analysis ions were scattered into a surface
barrier detector at a scattering angle of 1700 . The incident and
exit angles were both 50 with respect to the surface normal.
From Fig. 4.15, it should be noted that, whereas none of the light
ions provide good mass resolution for the more massive species
in the target, the sensitivity for oxygen detection in the ,target is
71
Table 4.1, A comparison of some of the main fealures of somo common backacaHering techniques,
Conventional
backscetterlng
High energy. light Ion
High energy. heavy Ion
Medium energy
bBCk8canartng
backSclittering
backscatterlng
BaamJenargyl
detectors
"He with 0,5 MeV < E < 2.5 MeV,

lH with 0.5 MeV < E < 1.0 MeV.
Surlace barrier detectors commanly used.
4He WIth E > 2.5 MeV,

with E > 1
MeV. Sunaco barrior detectorssre commonlyuaed.
Any beam more masalve than ~He.

Common beams are 12C, ,aO,1aF, end
35CI. Baam snorgles ,..5 MeV, Surfaoo
banler and IIme-oI.ff1ghl daleetof8aTe
I)'plcal.
IIn'llon beam may be used. Common

beamssro"He, 7U, II!C. nma-ol-nlght
detectors are moat uaelul.
Cr08s section
Rutherlord e~cept for Isolated

resonanceslnvoMng low mass lar'
get atoms; negligible screening,
Ruthorford 'or higher ma33SS. NonRUlherford lor some IIghlerlarget specles. Exp9rfmental/y determined cross
sections must ~9 used. Negligible
screening.
Beam energies Bre chosen 10 give

Rutherford cross sections, Screening
corrections requIred parUcularly for
heavy target epElcl8s.
Rutherford, bUt screening corrections

requIred
SenslUvlty
Umlled for low larger masses due

tounderlylng spectrum from heavy
subslrate or matrix apeelas. AlSO
may be limited by detector effects
such as pulse pile-up.
Enhanced sensitivity I() low mass specles due to r~lo.tlve enhancemant In

cross seellon relative 10 that tor hIgher
mass species.
Increased beam energy cancele potontlsl Incr(la&e In aensltlvll)' arising

from hlghetZ,.lesssenslUvelorllghter
species Ihan conventional baCkseatlerlng due 10 decrease In CfOS!! aactlon
relallvato heavier species. ForsurfElce
barrier detectors, etfects of pulse pileup on speclrum are reduced.
Umlted eensUMty 10 low mao species

due to undertyln~ spectra Irom heavier
substrata or matrix species. Tlm&-olIl1ght analyslsrnlnlmlzes ellectsol mndomcolncldence background. Ullimate
ssnslUvlly Umlted by spultarfng oltarget.
Opllmum forlowertarget masses,
Mass resolution Improv8sllnearl'l wilh

Incrl3i1slng beam energy.
Mass separation Improves with Increa8~

Ing beam mass. However, for surface
barrier detectors. degradation 01 dalector resolution with Ion mass may
result III over-all dlKlrel1S0 In resolution.
Mass separation Is smaller than lor

other techniques, bul energy resolulion oltlm9-0I..nlQht delector ylaldscomparable ma" rasolutlon. Resolution
msy bellmlled by energy stability 01 Ion
lH
;j
Mass
reaolutlon
poor for heavter ~peclas. LImited

by Inlrinsle detectrJr resolution
be"",.
b",,",,,,,
Depth
praflllng
RelatlvelysmalllSJ rasullaln large

profiling depth, bt.rt limited reeolulion. Roaolutlonlsllmlted at higher
energle9 by IS) ana ellOwer aner[lIes by Increased e,"er~y
strag[lilng.
For broad resonances. IImltatlons are

31milar to those of conventlonsl back
scattering. For narrow resonancs6.
depth of profile Is limited by proIClml1y of
other resonances while depth n!!IIllu_
lion Is Ilmiled by wfo'lh of resonance.
Increased IS} relaUve to light lona leads

10 ehallower profiling depths 8Ildsuperior poIenUal depth resolution. Depth
re80lulion may bo limited by detector
resolullon.
Lorge [SI and low besm energy reliulls

In very shallOw analysis deplh. Tlmeot-nlght depth resolUllon5 of order 1 nm
exceed thai available with other techn!quOl). Reoolutlon may ba limited by
energy &tabUny of beam
other
consIderations
All target Specl8S can bs anslyzed

simultaneously wUh an approprlale deteelor. Inslgnlllcant radletlon damage tor txlth detector and
targel surface. Large acceleralor
not required.
Deplh prom9s for all lergat species

may bo obtained simultaneously II resanancas ere broad, but narrow resonances requIre a series 01 measure
mania. Nuclear reaction products may
complicate spec1rum. Aadlntlon effecta
on delet:tor& and targetsurfacos unlm
portant. Requires a lBlger accelerator
than conventional baCkaCetleling.
An targelspecJea heavier than the beam

ioo lnay be analyzed simultaneously.
Dalect()n; Gnd lorgets eubJ8Ct to slgnlOcsnt radiation damage. Pulse height
delects In surfaC8 bamer deteelQra can
complicate analySis ot spectra. Production 01 high onergy, noavy Ions requiros II large eccelerator.
All target spocioa may be snaly2.ed

almultanaously. RedUCed beam heatIng of targets permits analyals 01 fmglle
targets. Sputtering may damage tar~
gets. parlleutartyforheav1erlona. Analy
sts may be performed with amaltlow
energy ecceleratQnJ liuch as Ihose used
lor Ion Implantation.
,_~,,"....<,,~,_.",__ ,__ .,"...~_~>,_~,,~... '."",. ,. ,.' ,. .."
<,' .
',.':'!~
V1Ba2Cu3Or.,.IA1203
4.3 MeV 4He
~
I: 5
:I
~4
~3
:e2
CO
-1
"
~ 0 E - - - - L - -_ _' - _ _ _'-~_L..LY
200
400
600
800
Channel number
FIG. 4.15. Comparison ofbackscattering spectra obtained at various beam energies for !H and 4He analysis ofY IBa2CU~Orx. Taken
from Rauhala et at. (1991).
The sensitivity of this technique is also influenced by beam

effects on the target. In particular, sputtering of the target is an
important consideration in this energy range and often is the
limiting factor in sensitivity of this technique. As a pmctical
matter, the achievable sensitivity for any technique might be
limited by the time available for the analysis and the total
integrated charge which can be deposited in the target during
that time. The lower energy ion beam used for MEBS results in a
lower level of target heating as well as a higher cross section,
thus permitting use of a higher beam current and lower total
analysis doses. In some instances, this combination makes it
possible to analyze fragile targets which are destroyed by higher
Y,S""Cu,O,.xlAJ2O,
28 MeV 2BsJ
energy beams.
200
400
600
The mass resolution obtained with MEBS is comparable to

that of conventional backscattering, but only because the
detectors used with this technique have a much better energy
resolution than do surface barrier detectors. Figure 4.17 shows a
MEBS spectrum obtained using a time-of-flight detector with
450 keY 12C as the analysis ion. The target consisted of a
800
Channel number
FIG. 4.16. Comparison ofbackscattering spectra for heavy-ion

analysis ofYIS.,eu,o,.,. Taken from Raubal. et 0/. (1991).
73
Chapter 4
.
')
atomslcm 2),
scattered from the same target also at 150 into a swface barrier
detector having a resolution of 15 keY. The natures of the
spectra are different in that the former is a plot of yield versus
flight time whereas the latter shows yield versus energy, but it is
clear that comparable mass resolutions are obtained for the two
cases.
0
sputter..deposited film containing eu (2.7 x IOu

Ag
(10" atoms/em'), and Au (3 x 10" atoms/em') on an Al
substrate. For this measurement, the beam current was 10 nA,
the total integrated cbarge was 10 Ile, and the scattering angle
was 150 0 The beam was incid~t along the surface normal.
Figure 4.1& shows a simulated spectrum for 2 MeV 4He
,'i
.:;
450 keV 12C+
500
Cu
400
300
;'
~ ..
200
Au
100
Ag
'.
"
200
100
300
400
Flight tIme (ns)

F]G. 4.17. Time-of-flight spectrum forbackscattering analysis using 450 keY
containing Au, Ag, and Cu on an AI substrate.
12C
analysis ions. Target was a sputter-deposited film
stability of the incident beam, and special attention might need

to be given to this aspect of accelerator performance.
MEBS is particularly useful technique for obtaining depth

profiles at the surface of samples. Stopping powers are relatively
high at the energies used, and the excellent detector resolution of
time-of-flight systems allows composition variations to be
detected with a resolution of a few nanometers. Figure 4.19
(Mendenhall and Weller, 1991b) shows. comparison between
two spectra obtained using a 270 keV 7Li analysis beam. The:
shift of 2 keV in the Ti edge for a TiSh film relative to that for
Ti metal indicates the presence of a layer of SiOz only 3.5 om
thick. The accessible depth of analysis is much smaller than that
available for higher-energy beams. however.
4.4.3 Cautions
As with any experimental procedure, there are a number of
possible sources of problems both in setting up the experiment
and in analyzing the result". Many of these problems are
discussed in some detail in Chapter 15, but we mention a few
briefly here as well.
4.4.3.1 Sample cbarging

Many of the possible sources of problems in backscattering
analysis arise from details of the experimental setup. For
insulating targets, sample charging can alter the incident energy
of the analysis beam by several kiloelectronvolts (significant for
MEBS analysis) and also can lead to inacCUIacies in current
integration when the sample discharges. Another s~urce of
potential error in current integration is inadequate cbarge
collection. Both of these difficulties, together with possible
preventive measures, are discussed in Chapter 15.
Finally, accelerator requirements for this technique are

considerably different from those of higber-energy analysis
methods. In particular, the beam energy used for MEBS
measurements is available from the same types of low-energy
accelerators used for ion implantation and for other work in
materials science. ]t should be noted. however, that the energy
resolution of this technique is strongly influenced by the energy
74
~.:
Energy (MeV)
1.3
1.4
1.5
1.6
1.7
1.8
10
rl,r----.,--~~I-----,_,----._,----~I
0.8
:E
CIl
>.
0.6 -
"C
1!iii
E
0.4
o
t-
Ag
0.2 -
0.0
700
750
800
11
850
Channel
FIG. 4.18. Simulated energy spectrum forbackscattering analysis using 2 MeV 'H
was th e_~M~.
Ii F 4 17
Au
900
950
1000
1-"
. h .. -I::
-"
.
e ana YSIS IOns Wit a s"-tJ.ace bamer detector. Target
4.4.3.2 Channeling
Experimental parameters such as the beam energy and the
target orientation can lead to spectra that are difficult to
interpret. For crystalline targets, channeling (Chapter 12) can
occur if the beam is incident normal to the target surface,
resulting in surface peaks and reduced subsurface yields in the
backscattering spectrum. Although channeling can also occur for
other angles depending on the material. it is possible to orient the
target in a non-channeling position relative to the beam. Rotation
of the target during bombardment is also used to avoid
channeling.
4.4~.4 Radiation exposure

~~lly, we note that consideration should be given to the
ra~latJ.on levels associated with any backscattering technique.
HIgh-energy beams can present SeriDUS radiation hazards and
even bc.ams with kiloelectronvolt energies generate X-rays.
Appropnate precautions against radiation exposure should be
taken (see Chapter 17).
4.4.3.3 Extraneous peaks
4.4.4 Examples
Backscattcring spectra can also be complicated by processes

other than elastic backscattering that occur simultaneously in the
target Inelastically scattered analysis ions and nuclear reaction
The selection of detector, ion beam, and ion energy depends

the specific requirements of a given measurement. Some
examples are described in this section.
products can both generate

backscattering spectrum.
OD
75
extraneous
peaks
in
the
Chapter 4
3000
-'"
8000
6000
TI
2000
::>
0
4000
2 keV
1000
2000
o~-L------
______~~~__-A~O
100
200
Energy (keV)
FIG. 4.19. Backscattering specmun obtained using 270 keY 7Lt ions with a TOF analyzer. The superposed spectra correspond to a
titanium foil target and a titanium silicide film. Taken from MendenhaU and Weller (1991 b).
4.4.4.1 Depth prome: 1.8 MeV 'He
angular distributions of material sputtered from a Ag-Au alloy.

Figure 4.21 shows a spectrum similar to those obtained in that
work. This spectrum was obtained using 15 MeV 160 ions
combined with a sUrface barrier detector. The peaks identified as
Ag and Au correspond to areal densities of (3.63 0.10) x 10 13
2
atomslcm and (3.73 0.19) x 10 12 atoms/cro2 , respectively.
Although the 'Widths associated with each peak are large, each
element is clearly resolved. Note particularly that, even with the
relatively poor resolUTIon associated with surface barrier
detection of heavy ions, a eu eontaminant peak is clearly
resolved from a surface peak. of Fe. Also, there is no background
in the region of interest arising from scattering from the AI foil
substrate. (The low level-background arises -from bulk Fe
contamination of the foil.) The absence of any pulse pileup in
the vicinity of the Au and Ag peaks permitted the analysis to be
performed with beam currents as high as 200 nA. It should be
noted that the pulse height defect associated with the detection
of heavy ions by surface barrier detectors would have made it
difficult to identify these elements had they been unknown
species. In that case, it would have been appropriate to use a
different type of 4etector.
Conventional backscattering using megaelectronvolt alpha

particles detected by a surface barrier detector is the most
convenient and appropriate analysis technique for many
measurements. Figure 4.20 shows the results of a typical series
of measurements performed using this technique. In this
experiment, Yu et al. (1991) used alpha backscattering to study
the formation of TrSi 3 layers by high-dose iridium ion
implantation into silicon. For this application, good mass
resolution was not necessary because the identity of the target
constituents was already knovm. However, depth resolution was
important because depth profiles of the implanted iridium and of
the silicon concentration were needed. By orienting the target
and detector as shown in the figure, a depth resolution of
approximately 5 nm was obtained. This was sufficient to
demonstrate clearly the effects of increasingly greater Ir
implantation doses on both the Ir depth profile and the Si
distribution.
4.4.4.2 Mass resolution: 15 MeV 160
Heavy-ion backscattering has been used for a number of
applications where improved mass resolution for heavy clements
was needed. Weller (1983) used this technique to measure
..
76
-~-
Channel #
FIG. 4.20. RES spectra of iridium-implanted silicon for three different implant doses, obtained using 1.8 MeV alpha backscattering and
surface barrier detectors (Yu ef al., 1991) .
. --200
..
~.~
..
Ag
100
Cu
Au
10
12
Energy (MeV)
FIG. 4.21. RBS spectrum of sputter-deposited Ag and Au on an aluminum foil substrate, obtained by scatering 15 MeV 16 0 ions into a
surface barrier detector. Taken from Weller el al. (1990).
obtained by scattering into (a) a surface barrier detector and (b) a
TOF detector. 'Whereas little information about the layered
structure can be obtained from the surface barrier spectrum, the
TOF spectrum clearly resolves the first eight layers of the
structure. Energy dispersion, which worsens progressively with
depth into the target, limits the analyzable depth in this case.
4.4.4.3 Depth resolution: 25 MeV lSCt

Heavy-ion backscattering bas been combined with TOF
analysis to obtain good depth andlor mass resolution. Dobell et
al. (1991) used 25 MeV "CI to profile a sample consisting of 10
layers,8 nm thick, of lnO.ISG:l,J.82As alternating with 20-nm-thick
layers ofGaAs. Figure 4.22 shows the In portion of the spectrum
77
.."l
Chapter 4
100
a}RDS
80
~
:::>
ion-beam modification of materials because the same accelerator

and sometimes even the same ion beam can be used for both
modification and analysis.
60
40
~o~----------
__
~.-
____
20
0
400
300
200
100
Depth (nm)
15
b)TaFRBS
~
:::>
10
ot.....~~_-J..._~-J...~.;..-.L.JJ........
300
200
100
Depth (nm)
FIG. 4.22. RBS spectra obtained from a target consisting of 10
layers, 8 nm thick, of lno.18Gau.82As alternating with 20-nm-thick
layers of GaAs. Only In region of spectra shown for 25 MeV
3:5Cl ions scattered into (a) a surface barrier detector and (b) a
TOF detector. Taken from Dobeli et at. (1991)
4.4.4.4 Beam modification: 250 keY 4He

Medium-energy backscattering analysis is most useful for the
analysis of very thin films and for measurement of depth profiles
very near the surface of materials. Mendenhall and Weller

(1990b, 1990c) used this technique to investigate the destruction
of CaF2 and MgF2 films by ion bombardment, irradiating targets
with a relatively high current of 250 keV 4He. The beam current
was periodically reduced to a much smaller value for analysis so
that the dose dependence of the target composition was
measwed. Results sho'NIl in Fig. 4.23 for CaF2, converted from a
TOF spectrum to an energy spectrum, clearly indicate the loss of
F from the sample for increasing irradiation doses. As seen "With
this app1ication, MESS can be particularly useful in the study of
FIG. 4.23. Energy spectra showing relative concentrations of Ca

and F in a CaF1 film before and after inadiation with 250 ke V
'He. The CaF, film was analyzed with 250 keV'He using. TOF
detector. The TOF spectrum was then converted to the more
familiar energy spectrum. Taken from Mendenhall and Weller
(1990b).
4.4.45 Depth profile with elastic resonance; J MeV "He

Non-Rutherford elastic scattering is often used to obtain
concentrations of low-Z target constituents that are difficult to
analyze when the Cross section is Rutherford Vizkelethy et aJ.
(1990) demonstrated the use of this technique for profiling oxygen
in buried layers. Figures 4.24 and 4.25 sWw spectra obtained by
scattering alpha particles with energies of 3.054 MeV and 3.254
MeV, respectively. from a target consisting of a 20-nm SiOl layer
sandwiched between two 300-DID layers of Cu deposited on a
graphite substrate. In Fig. 4.24, the backscattering yield from a
surfitce oxide layer ichannel 210) is enhanced by the 3.04 MeV
resonance in the I OCa,a)160 elastic scattering cross section,
whereas the much smaller yield from the buried oxide layer
(channel 150) results from the smaller off-resonance cross section.
In Fig. 4.25, however, scattering from the buried layer is resonant,
yielding a more accurate estimate of the oxygen concentration at
that location.
78
'
Backscattering Spectrometrt
"Application of Accelerators in Research and IndtLstry:

Proceedings of the Eighteenth International Conference on the
Application of Accelerators in Research and Industry (C~
Energy (MeV)
1,s
1,0
~5
~''''''''~~~'''~t
2004)", McDaniel, F,D" Doyle, BL, Vizkelethy, G"

Johnson, RM_, aod Sisterson, 1M, (eds.) (2005), Nucl.
instrum, Methods B241 (1-4), 1-1038.
"Application of Accelerato" in Research and Industry Proceedings
of the Nineteenth International Conference on The Application

of Accelerators in -Research and Industry, Nineteenth
International Conference on The Application of Accelerators in
t.
Research and Industry", McDaniel, F,D" Doyle, B.1" BenItzhak, L, Hamm, RW" Johnson, RM_, Sisterson, J.M"
200
"'"
Channel
Slaughter, D.R., Thevutbasan, S., Vizkelethy, G., Wang, Y.,

and Wells, D.P. (eds.) (2007), Nucl. Instrum, Methods B261
soo
.00
(1-2), l-l224,
"Application of Accelerators in Research and Industry:
Twentieth International Conference", McDaniel, F.D.,
FIG. 4.24. Backscattering spectrum of a target consisting of a

20-nm Si0 2 layer sandwiched between two 300-nm layers of eu
deposited on a graphite substrate. The incident beam was 3.054
MeV alpha particles for which the 160(a,a) 160 elastic cross
section is resonant. Taken from Vizkelethy et at. (1990).
0_'
i4
~
Ii'
Doyle, B,L. (OOs,)(2009), AlP Con! Proc, 1099, 1-1030_

Barbour, J.e., Doyle, B.L, and Myers, S.M, (1988), Phys, R<", B38,
7005,
Barradas. N,P.. Jeynes, e., Jenkin, M" and Marriott, P.K.
(1999), Thin Solid Films 343-344, 31,
Energy (MeV)
1,0
t.s
2.S
2.0
--
~.
,
.-
'"
>=
-;\~
Barredas, N,P, (2007), Nucl. Instrum. Methods B261, 418.

Bittner, J,W" and Moffat, RD, (1954), Phy', Rev, 96,374,
'"\
--- - -
Barradas, N,P, (2004), Nuc!. Instrum, Methods B225, 318_
Boerrna, D,O., Labohm, F" aod Reinders, J.A (1990), Nue{,

Instrum, Methods B50, 291.
Bozoian, M_, Hubbard, K,M" and Nastasi, M, (1990), Nucl.

- - [nstrum, MethaciYB51, 311.
II
Bozoian, M, (1991 a), Nucl. I",/rum, Methods B56---S7, 740,

Bozman, M, (I 99 Ib), Nue!. [nstrum_ Methods B58, 127,
Bragg, W H., and Kleeman, R (I905), Philos, Mag, 10, 5318_
Cameron, J.R. (1953),Phys, Rev, 90, 839,

Chu, W-K., Mayer, J,W" and Nicolet, M-A, (1978),
Backscatten'ng Spectrometry, Academic Press, New York.
100
200
300
400
500
Channel
Dobeli, M"
Haubert, p.e, Livi, RP., Spicklemire, S.J.,

Weathers, D,L., and Tombrello, T.A (1991), Nue/. In,trum_
Methods B56---57, 764,
FIG, 4.25. Analysis of the same target used in Fig. 4.24 with a
3.254 MeV alpha beam. Taken from Vizkelethy et ai, (1990),
Doolittle, L.R, (1985), Nuc!. [nstrum, Methods B9, 344,
It is not possible in an article of this length to describe or
provide an example of every variation of backscattering analysis.

Two sources for references on current backscanering techniques
and applications are the proceedings: of two biannual international
conferences. Application of Accelerotms in Reseorch and Industry
(2005,2007,2009) and Ion Beam Analysis (2004, 2006, 2008),
Doolittle, LR_ (I 986), Nucl. [nstrum, Methods Bt5, 227,

. DOOlittle, L.R (I990), in High Energy and Heavy Ion Beams in
Materials Analysis (Tesmer, J.R, Maggiore, C.)., Nastasi,
M" Barbour, J,C., and Mayer, j,W" eds,), Materials
Research Society,Pinsburgh, PA, p. 175,
End!, P,M., and van der Leun, C. (1967), N.ci, Phys_ AIOS, l.
Eshbach, H,t, (1983), The Bi-implanted-in-Si RBS standard was
supplied by H.L Eshbach, Central Bureau of Nuclear
REFERENCES
HR., Besenbaeher, E, Loftager, Po, and Moller, W,
(1980), Phys, Rev, A21, 1891.
Andersen.
Measurements. Geel, Belgium.
79
Chapter 4
Hausser, 0., Ferguson, AI., McDonald, A.B., Sz6ghy, 1M.,

Alexander, T.K., and Disdier, D.L. (1972), Nucl. Phys.
AI79,465.
Hautala, M., and LUQrnajarvi, M. (1980), Radiat. Elf 45,159.
Herring, D.F. (1958), Phys. Rev. 112, 1217.
Hubbard, KM., Tesmer, J.R., Nastasi, M., and Bozoian, M.

(1991), Nud. Instrom. Methods B58, 121.
"Ion Beam Analysis 16: Proceedings of the 16th International

Conference on Ion Beam Analysis", Vizkelethy, G.,
McDaniel, F.D., Tbcvuthasan, S., Tesmer, J.R. (eds.)
(2004), Nud. Instrum. Methods B219-220, 1-1084.

Co;nference on Ion Beam An3Iysis", Boerma, D.O.,
Climent-Font, A., Respaldiza, M.A., Ager, F.J., Garcia-
Mendenhall, M.H., and Weller, R.A (1990a), Nud. InstnJm.

Methods B47, 193.
Mendenhal~ MR., and Weiler, RA. (1990b), Nucl.lnstnJm.

Methods B51, 400.
Mendenhall, M.H., and Weller, RA. (1990c), Appl. Phys. Lett. 57,
1712.
Mendenhall, M.H., and Weller, R.A. (I 99la), Nucl. InstnJm.
Methods B58, II.
Mendenhall, M.H., and Weller, R.A (199Ib), Nucl. Instrum.
Methods BS9--liO, 120.
Mozer, F.W. (I956),Phys. Rev. 104, 1386.
Northcliffe, L. C, and Schilling, R.F. (1970), Nucl. Data Tables
A7,233.
Lopez, G. (eds.) (2006), Nuc!. Instrom. Methods B249, 1942.
O'Connor, D.J., and Chunyu, T. (1989), Nucl. Instrum. Methods

B36, 178.

Conference on Ion Beam Analysis", Pathak, A.p., Avasthi,
Petrov, I., Braun, M., Fried, T., and S.therolom (1983), J. Appl.
Phys. 54, 1358.
D.K., Dev, B.N. (eds.) (2008), Nuc!. Instrom. Methods B266

(8) 1149-1932.
Jarjis, R.A. (1979), Nuclear Cross Section Data for Surface

Analysis (Department of Physics, University of
Manchester), unpublished.
Leavitt, l.A., Stoss, P., Edelman, c.R., Davis, R.E., Gutierrez,

S., Jubb, N.J., and Reith, T.M. (1985), N"cl. Instrum.
Methods BIO-II, 719.
Leavitt, J.A. (1987), Nucl. Instrom. Methods B24-25, 717.
Leavitt, J.A., MclntyTe Jr., L.C., Stoss, P., Ashbaugh, M.D.,
DezfouIY-AIjomandy, B., Hineeli, M.F., and Van Zijll, G.
(1988), Nucl. Instrom. Methods B35, 333.
Leavitt, J.A., McIntyTe Jr., L.C., Champlin, R.S., Stoner Jr., J.O.,
Lin, Z., Ashbaugh, M.D., Co., RP., and Frank, J.D. (1994),
Nucl. instrom. Methods B85, 37.
L'Ecuyer, I., Davies, 1.A., and Matsunami, N. (1979), Nucl.

Jnstrum. Methods 160,337.
Raubala, E. (1987), J. Appl. Phys. 62,2140.

Rauhala, E., Saarilahti, J., and Nath, N. (1991), Nud. Jnstnim.
Methods B61, 83.
TRIM 1985, by Ziegler, J.F., Biersack, J.P., and Littmark, U.
"The Stopping and Range of Ions in Solids", yol. 1 of The

0/ Ions in Matter, Pergamon Press,
Stopping and Ranges

New York.
Vizkelethy, G., Revesz, P., Li, J., Matienzo, L.J., Emmi, F., and
Mayer, J.W. (1990), in High Energy and Heavy Ion Beams
in Materials Analysis (Tesmer. 1.R. Maggiore, C.J.,
Nastasi, M., Barbour, 1.C., and Mayer, J.W., eds.), Materials
Research Society, Pittsburgh, PA, p. 235.
Weller, M.R. (1983), Nucl. Jnstrum. Methods 212, 419.
Weller, M.R, Mendenhall, M.H., Haubert, P.C., D6beli, M.,
and, Tombrello, T.A. (1990), in High Energy and Heavy Ion
Beams in Materials Analysis (Tesmer, J.R, Maggiore, C.J.,
Nastasi, M. Barbour, J.C . and Mayer, J.W., eds.), Materials

Research Society, Pittsburgh, PA, p. 139.
Macdonald, J.R.', Davies, J.A., Jackman, T.E., and Feldman, L.c.

(I 983),./. Appl. Phys. 54, 1800.
Wenzel, W.A., and WbaIing, W. (1952), Phys. Rev. 87,499.
Martin, J.e, Nastasi, M., Tesmer, I.R., and Maggiore, C.J.
Williams, J.S., and MOiler, W. (1978), Nuc!. Instrom. Methods

157. 213.
(1988), Appl Phys. Lett. 52,2177.

Mayer, M. (1997), SIMNRA User's Guide, Technical Report
IPP91113, Max-Planck-Institut fiir Plasmaphysik, Garching,
Germany.
Yu, K.M., Katz, B., Wu, I.C., and Brown, I.G. (1991), Nuc!.
Instrom. Methods B58, 27.
Ziegler, J.F., Lever, RF., and Hirvonen, J.K. (1976), in Ion
Beam Surface Layer Analysis (Meyer, 0., Linker, G., and
Kappler, F., eds.), Plenum Press, New York, p. 457.
Ziegler, J.F., Biersock, J.P., and Littmark, U. (1985), "ne
Mayer, M. (2007), SIMNRA User's Guide, Max-Planck-Institut

fUr
Plasmapbysik,
Garching,
Germany,
http://www.rzg.mpg.de!%7EmamJ.
McIntyre Jr., L.C., Ashbaugh, M.D., and Leavitt, J.A. (1987),
Mater. Res. Soc. Symp. Proc. 93,40 I.
Stopping and Range of Ions in Solids", vol. 1 of The

Stopping and Ranges of Ions in Malter, Pergamon Press,
New York.
Mendenhall, M.H., and Weller, R.A. (1989), Nucl.lnstrum.

Methods B40-41, 1239.
Ziegler, I.F., and Manoyan, J.M. (1988), N"d. Instrom. Methods

B35,215.
80
ELASTIC RECOIL
DETECTION: ERD
G. Dollinger and A. Bergmaier
Universitat der Bundcswehr Munchen, Neubiberg, Germany
CONTENTS
5.1 INTRODUCTION ....................................................................................................83
5.2 FUNDAMENTALS ...................................................................................................84
5.2.1 Kinematics ........................................................................................................84
5.2.2 ERD cross sections ........................................................................................ 84
- -- --
5.3 METHODOLOGY .....................~.:;;.: .........:...........::....:.::::::........................ 89

5.3.1 ERD using a range foil .................................................................................... 90
5.3.2 AE-E analysis ....................................................................... 93
5.3.3 Time of fligbt versus energy (TOF-E) metbod .......................................... 97
5.3.4 Magnetic and electrostatic spectrograpbs ................................................... 100
5.3.5 Transmission ERD for tbin-foil analysis ............................................ 101
5.3.6 Coincident ERD or Elastic Recoil Coincideuce Spectrometry (ERCS) ... 102
5.4 QUANTITATIVE ERD: DATA EVALUATION ......................................... 103
5.4.1 Thin layer analysis ........................................................................ 103
5.4.2 Thick-layer analyses ...................................................................................... 103
5.4.2.1 Simulation-fitting approach ................................................................. 104
5.4.2.2 Direct transformation scbeme (slab analysis) ........................ 105
5.5 DEPTH RESOLUTION IN ERD ANALySIS ............................................ 105
5.5.1 Beam contribution to energy spread ............................................................ 107
5.5.2 Doppler effect ............................................................................................... 107
5.5.3 Energy straggling ............................................................................ 108
5.5.4 Multiple scattcring effects ................................................................ 109
5.5.5 Detector resolution ......................................................................................... 110
5.5.6 Correction of kinematic sbift ............................................................ 111
81
-,
- --- ._--
Chapter 5
5.5.7 Deconvolution of ERD spectra ..................................................................... 112

5.6 ACCURACY AND SENSITIVITY IN ERD ANALYSIS ................................... 114
5.6.1 Accuracy ......................................................................................................... 114
5.6.2 Sensitivity: Trace element analysis ........................................................ 115
5.6.2.1 Number of detectable counts ................................................................ .116
5.6.2.2 Limits through background .................................................................. 116
5.6.2.3 Plural scattering effects ......................................................................... 116
5.6.2.4 Secondary nuclear reactions ................................................................. 117
5.6.3 Irradiation damage ........................................................................ 118
REFEREN CES ............................................................................................................... 121
82
Elastic Recoil Detection: ERD
5.1 INTRODUCTION
ERD (elastic recoil detection) or ERDA (elastic recoil
detection analysis) (Ecuyer et al., 1976; Doyle and Peerey, 1979;
Arnold Bik and Habraken. 1993; Assmann et aI., 1996),
sometimes called PRES (forward recoil spectrometry) (Mills et
aI., 1984), uses elastic scattering as the main interaction between
a smft projectile and a target atom for elemental and depth
profile analysis as it does Rutherford backscattering
spectrometry (RES) [Fig. 5. I (a)]. The main difference from RBS
(Chapter 4) is the detection of the recoil atom and not the
scattered projectile. The yield of recoil ions of a certain element
is directly proportional to the elemental content in thc layer that
is investigated. Matrix effects of surrounding atoms have only
minor effects on the scattering cross section because the
scattering cross section is, as in RBS, dominated by the large
interaction potentials between the atoms at short distances
(Coulomb or strong interaction potentials). Thus, ERD has the
potential to perform quantitative elemental analysis. The only
influence of the matrix on the scattering yield arises through
cbanneling effects, thus allOwing for structural analysis (Chapter
The main advantage of ERD compared to RBS is that ERD

solves the ambiguity problem inherent in RES depth profiling
where mass information and depth information are both ~ed in
one energy spectrum of the scattered projectiles. Because the
re(;oil atoms of the target are analyzed by ERD, their nuclear
charge andlor mass can be independently determined in addition
to their total energy. Thus, the determination of the elemetlt and
its depth of ongin follows event by event. This especially holds
true for light-element analysis (that is, Zj < 10, where Zi is the
atomic number of a target atom), where RES has a low
sensitivity because of cross-section scaling by Z? and where
RBS suffers from a high background produced from projectiles
that are scattered deeper in the sample at lower energies from
heavy matrix atoms. As a consequence, ERD call deliver
quantitative depth profile analysis especially of light elements
with a sensitivity in the part-per-million range or even better.
High-resolution setups are able to analyze light elements with a
depth resolution better than 1 nrn at slightly reduced sensitivity.
In particular, hydrogen, which is not accessible by RES, can
easily be profiled by ERD (Doyle and Peercy, 1979). The
potential of ERD to provide quantitative analyses of light
elements Vlith good senSitivity and depth resolution makes it one
of the most suitable tools for elemental depth profiling.
12).
I'"
-1
ERD: recoil ions
i, EI
"',
i ~
The drawback of ERD generally is the necessity of a more

sophisticated experimental setup compared to RES, for which a
small-scale electrostatic particle accelerator and a single silicon
projectile ions
detector already offer very good performance. In ERD, recoil
atoms are scattered to scattering angles <p, with Cjl < 900 in the
laboratory frame. These forward scattering directions mean low
. _ angles of incidence for projectiles and low angles of exit for
recoil ions during analysis of thick targets [Fig. 5.1(b)], which
complicate beam focusing and adjustment of the target and
RBS:
detector angles. In addition, the large number of projectiles
backscattered projectiles
(a)
scattered in the direction of the ERD detector have to be
separated from the recoil ions. Energetic heavy-ion beams are
required to get the full potential of ERD in a baekground.free
ERD: recoil ions for analysis:
analysis of all light and medium-heavy elements because the
scattering angle is limited when the heavy ions are heavier than
the matrix atoms (Stoquert et ai., 1989; Assmann er ai., 1996).
However, in the case of hydrogen analysis, a simple detector
configuration is sufficient, for example, a passive absorber foil in
front
of a silicon detector to suppress scattered projectiles and
(b)
heavy matrix recoil ions from reaching the detector (Section 5.3)
(Doyle and Peerey, 1979). Thus, even He beams can he utilized
Fig. 5.1. (a) Principle ofERD as a counterpart ofbackscattering
for hydrogen analysis by ERD. The general ERD setup requires
geometry. In the analysis of thin foils, both the scattered
larger accelerators than are usually used for dedicated RBS and
projectile and the recoil ion scattered into the forward direction
particle-induced X-ray emission (FIXE) systems in order to
from the sample can, in principle, be analyzed. The involved
apply heavy-ion beams with specific energies of 0.2-2
scattering angles are given by kinematics. (b) Principle of
MeV/nucleon, for example, there are several ERD experiments
analyzing a thick sample by ERD. Because recoil ions are
running with 1()"'35 MeV chlorine, 4()"'350 MeV gold, or 4()'"
scattered to angles <p smaller than 90 0 , the recoil ions have to be
170 MeV krypton or iodine beams. This also means enhanced
analyzed under a shallow scattering geometries with small
effort for the detector equipment to acquire elemental
incident angles a and exit angles 13.
-,
.r=,
.,
' -... '
83
,1
Chapter 5
~E/Ej, when developed into a Taylor series around the mean

scattering angle <p, reads as
information (nuclear cbarge and/or mass) separated from the

depth information, as the latter is obtained from the energy
measurement. The more advanced detectors consist of either
energy loss ven;us energy (t.E-E) (Stoquert el 0/., 1989;
Gebauer el al., 1990; Assmann el a/., 1994; Bergmaier ef aI.,
1998) or time of flight versus energy (fOF-E) detector
telescopes (Groleau el al., 1983; Thomas el a/., 1983; Whitlow
et al. 1987; Ami et aI., 1992) or even magnetic (Gossett, 1986;
Dollinger el al., 1992) or electrostatic (Enders el aI., 1992)
spectrographs. These technical requirements have limited the
widespread usc of ERD compared to RBS, despite the obvious
advantages of the former in many applications. However,
increasing numbers of ion beam laboratories are setting up ERD
as a standard tool for elemental analysis. Especially the use of
TOF detectors and magnetic or electrostatic spectrographs
allows ERD a.nalysis with submegaelectronvoltlnucleon heavyion beams that are already available in many ion beam
laboratories.
Because the scattering angle of common ERD - analysis is

between 30" and 45", the first-order contribution dominates the
kinematic effect and is much larger than in common RBS
experiments.
Scattering angles of (Jl :::::: 45" are normally used for ERD
experiments in order to obtain sufficient recoil-ion energies for
accurate elemental identification and energy measurement and in
order to avoid excessive spreads in kinematic energy (i.e., tan ~
::; 1, Eq. 5.2). As long as the projectiles are lighter or not much
heavier than the atoms of the layer to be analyzed, the projectiles
will also be scattered into the detector. An elegant method to
circumvent this problem is to use projectiles that are much
heavier than the main elements in the matrix. The projectiles do
not scatter imo the recoil-ion detector if the scattering angle rp is
chosen to be larger than 'Pm:u, which is given by kinematics as
5.2 FUNDAMENTALS
In the following sections,

relationships for ERD.
we summarize the basic
5.2.1 Kinematics
In an ERD experiment, recoil ions from binary elastic
collisions are analyzed (Fig. 5.1). As a good ~pproximation. the
recoil ions are at rest before scattering, and they are emitted at
angles <p with cp < 90" in the laboratory frame because of twobody kinematics. A projectile of energy Ep transfers the amount
of energy Ei to a recoil atom of type i in the sample during a
binary collision, where
rp~
.M
= arCSlllM.
(5.3)
5.2.2 ERD cross sections

As long as strong interactions do not contribute
significantly, the cross section dcrJdQ is generally given by a
more or less screened Rutherford cross section, which is given in
the laboratory frame by
(5.1)
do,
(d.Q1... -K
Tbe parameter ~ is the kinematic factor ofERD. and M" and Mi
denote the masses of the projectiles and the recoil atom,
respectively. Deviations from Eq. (5.1) occur as a result of the
thermal and zero-point motion of the target atoms before
scattering, which arc treated in terms of a ''Doppler effecf' in
Section 5.5.2. Other deviations can be introduced by some
inelastic processes in the electronic shells. In general, inelastic
nuclear scattering has to be avoided in order to prevent the
ambiguity problems in elemental separation and depth scales that
arise in RBS analysis.
( 'J'(
Z,Z,e
4m:,E,
M, + M;
M,
)'
COS' '1"
(5.4)
where z" and ~ are the atomic numbers of the projectiles and the
scattered target atom of type i, respectively.
K::; 1 is a mUltiplicative factor that is introduced to correct

the pure Rutherford cross section in the case that screening from
electrons is not negligible. As long as screening is moderate
(e.g., K > 0.9), the factor K can be accurately approximated as
(AudelSen ef ai, 1980; Dollinger er al., 1996)
A change in scattering angle, drp. makes a change, BE;. in the

recoil-ion energy, E;, according to Eq. 5.1. This change is called
the kinematic effect. The relative change in recoil-ion energy.
84
1+
K = [ 1+ v.
I,Lab
E,
scattering cross section enhanced by almost a factor- of 4

compared to those of other light elements because Z/A; = 1
in the case of hydrogen compared to Z/Ai ::::: 1/2 for the other
V;,ub)'
2E,
elements.
(5.4.)
(2EpCOS~J
V.LLab
If the ions overcome the Coulomb barrier, strong nuclear

interactions dominate the cross sections, and Eq. (5.4) no longer
']'
holds true. A "safe" specifiC projectile energy E,lm, for- pure

Rutherford l':TQSS section is given when F-p is smaller tban a
threshold energy E Thres where
using Vl,l.ab in eV given by
M +M'x48.73xZ Z.Z
~ 213 +z211
V.1,Lab ='
M.
pip
,.
Mp
JZ,Z,e'
ETb= = ( 1+- - - - .
M, 41t8oRo
(S.4b)
(:5.6)
The cross sections can deviate by less than about I % from the
Rutherford values when a safe minimum distance of closest
The accuracy to which the correction factor K can be

approximated through Eq. (5.4a) for K < 0.9 is unclear. As a
simple approximation of Eq. (5.4), one can use the atomic mass
numbers Ap and Ai of the projectiles and the recoil ions,
respectively, and insert the projectile energy in units of MeV into
approach
of
Ro
=[1.25( A,u' + A,"') + 5]:fin
is
used
(Wakefield et of, 1970) where fin is femtometer or lxtO 13 m. At

larger projectile energies, one must rely on experimen tal or
theoretical values for the elastic scattering cross sections. In the
case of heavy projectiles, one can use optical model calculations,
the equation
as demonstrated by Pusa el af. (2003).
=2.074XIO-Mcm'[~J'(A,+A'l'_3 .(5.5)
(dcr,)
d.Q u'
E,[MeVj
lA,
cos ~
The specific energies, Epi'lVIp, where Rutherford cross

sections can safely be applied for screening factor values of K >
0.99 and K> 0.90 are shown in Fig. 5.2.

Equation (5.5) has a maximum deviation from the exact
Rutherford cross section [Eq. (5.4) at K ~ 11 of 1.5% when
analyzing hydrogen by heavy ions.
Several specific characteristics of the ERD Rutherford cross

sections as given by Eq. (5.4) can be identified:
;..
---- ---- -----
. -;.,:..~.-;:.. ...... ---
---
The cross section increases with scattering angle cp

according to a lIe cos 3 q dependence. This increase in cross
section is accompanied by a decrease in recoil-ion energy ~
that is proportional to cos' ~ [Eq. (5.1)].
~ning factor K < 9o~
In the case of projectiles that are much heavier than the

recoil ions (Mp M j ), the cross section scales as
2
cr ex: Z, (M" + MJ2 ex: Z:
20
using the same projectile
40
prnjecflte alOmi~.numher-
FIG. 5.2. Limits of specific energies for projectiles where

Rutherford cross section') with low screening corrections (K >
0.9 or K > 0.99) can be applied as a function of the nuclear
charge of the projectiles. In addition, the limits from which the
scattering cross section becomes non~Rutherford through nuclear
interactions [Eq. (5.6)] are also plotted.
energies or as cr ex: Zp when the ion velocity, and thus the

specific energy EplMp> is kept constant.
The cross section is nearly constant for all recoil ions that
are much lighter than the projectile (Mi Mp), and thus,
similar sensitivities can be obtained in terms of possible
count rates. The only exception is hydrogen, which has a
85
Chapter 5
Table 5.1. References for elastic cross-sectioll data for helium ERD on protons. pQ-le,p)He, in the laboratory system.
Angle (laboratory system) (deg)
60.00
Energy range (MeV)

3.00-4.50
Reference
Bogdanovic Radovic and Benka
55.00
2.50-4.50
50.00
2.50-4.50

(ZOO I)
Bogdanovic Radovic and Senka
45.00
2.50-4.50
(2001)
(2001)
41.50
41.50
40.00
2.43-2.97
2.44-2.98
2.50-4.50
31.30
31.30
30.00
1.95-3.05
1.95-3.05
2.50-4.50
30.00
30.00
30.00
30.00
21.50
21.50
20.00
20.00
1.0()-2.50
5,03-5.96
9.5()-11.6
1.0()-12.5
2.0()-2.92
10.00
10-40
2.0()-Z.92
1.0()-2.5
1.51-1.95
1.0()-2.50
0.6()-5.00

(2001)
Nagata ef aJ.. 1985
Yamaguchi et al., 1983
Bogdanovic Radovic and Benka,
2001
Yamaguchi ef al. 1983
Nagata ef 01., 1985
Bogdanovic Radovic and Benka,
2001
Quillet ef al., 1993
Wetteland ef 01., 1998
Browning et at., 2004
Keay and Ingram, 2004
Yamaguchi et al., 1983

Nagata ef aI., 1985
Quillet el al., 1993
Nagata ef aI., 1985
Quillet el 01., 1993
Kim el al., 1999
.~
2004). This fit is especially useful in the. resonance region at

about 8 MeV helium ion energy where cross sections are
enhanced by up to a factor 100 compared to Rutherford cross
sections. which has a positive effect when considering beam
damage effects (see Section 5.6.3). The data can also be easily
When light ions are used as projectiles, the Coulomb barrier
is easily overcome already at moderate energies, especially in the

widespread case of analyzing hydrogen isotopes by helium or
lithium ERD. A collection of references for experimental data of
He-proton (p(He,p)He) elastic cross sections is given in Table
5.1. Most of the data are also collected on the Web site of the
Ion Beam Analysis Nuclear Data Library (IRANDL; bttp:llwwwnds.iaea,orglibandl/). There is also a tbeory for elastic p(He,p)He
scattering over a large energy range, where only a few
parameters are fit to existing experimental data (Pusa et al.,
retrieved from the SigmaCa1c Web site (bttp:llwwwnds.iaea.orglsigmacalcl#). These values might be the best choice
for evaluating helium ERD for hydrogen analYSis. A compilation
of data is shown in Fig. 5.3 for some angles over a wide range of
energies.
~;
1,
86
I~_=~oor"
-;j~,250 '.
,:~.
" ~'1~99,
~'-"
"' ...
5
6
7
'He energy [ MeV I
I ql': 30
10
.. -.
- - ........
5
6
'He energy [ MeV I
1.=35'
10
"
eo 400
..
'
'200
'"
is
,,
.... .. .... .......

;.;
"
600
tl
is
.' .. "'.
~E
800
:5
250
S
6
7
"'He energy [MeV]
....... 9
'0
10
.,
200
.. -.. ..
3
-. ..
_10 _
'He ene'rgy [ MeV I
FIG. 5.3. Elastic scattering cross sections, He-p, as calculated using SigrnaCaic (http://www-nds.iaea.orglsigmacalcl#) (solid line)
compared to the Rutherford cross section (dashed line) at various scattering angles.
Table 5.2. References for elastic cross-section data for helium ERD on deuterium, d(He,d)He, in the laboratory system.
41.50
40.00
35.00
31.30
30.00
30.00
30.00
25.00
21.50
20.00
20.00
10.00
Energy range (Me V)

2.15-3.05
0.98-2.87
0.98-2.87
1.96-3.05
0.96-2.93
1.00-2.50
9.00-11.0
0.96-2.93
2.00-2.92
1.00-2.5
0.98-2.87
1.00-2.50
87
Reference
Nagata et al., 1985
Kellock and Baglin, 1993

Nagata ef aI., 1985
Kellock and Bag1in, 1993
Quillet ef al., 1993
Browning et al., 2004
Nagata ef al., 1985
Quillet ef af., 1993
Q:nllet ef al., 1993
Chapter 5
The differential cross section dO' is calculated in units of ~~

d.nl'
millibams per steradian (mb/sr) when the energy is in
megaelectronvolts and parameters a, b, c, and d are used as given '\
~:
in Table 5.3 for a set of scattering angles <p. Only a few data are
available for non-Rutherford cross sections for lithium ERD on 3'
hydrogen (see Table 5.4), but none are available for lithium
ERD on deuterium.
In the case of analyzing deuteriwn by helium ERD, no

theory is available to describe the cross sections. Therefore, one
can use experimental data for the differential cross section in the
laboratory frame where they are available [scc Table 5.2 or the
mANDL Web site (http://www-nds.iaea,orglibandIJ)J, or one
can use the following fit function introduc<d by Kellock and
Baglin (1993) for the nonresonane< reg;on (see Fig, 5.4):
"
(5,6.)
Table 5.3. Fit parameters for experimental scattering cross-section data (laboratory frame) for d(He,d)He according to Eq. (5.6a)
(Kellockand Baglin, 1993):
(:)Ub
= exp ( aE+ b+i+ :'
'P (deg)
a (1IMeV)
20
25
30
35
-0.143
0.451
0,133
0,097
b
6,043
2,849
4.383
4.511
c (MeV)
-0,898
4,738
2,196
2,588
d (MeV')
1,476
-U03
-0,042
-0.278
Table 5.4. References for elastic cross-section data for 1Li ERD on protons p(,U,p) 1Li in the laboratory system.
45.00
30.00
Energy range (Me V)
Reference
2.28-5.70
2.28-5.70
Siketic, 2005
Siketic, 2005
- -..:::"-.:.....- -----
88
Elastic Recoil Detection: ERO
...
"
. cO ij~Ki-'
o ... ~
:.:....::-:.~ft
~~In:.
-',.,;.~
:,."flibt:n:
"'-"c;'_.".,.;;'",~",~...;.
'0
..,......-..
~,.:.;.-
--".'
.'
~~~7C-""",",,,,,,~.,J;;1 ~o
3,0 '.
'4fteene.!'U'i"(v.eV)
1"'!
1.S .
.:L.O
. ~~G~~Yl
.'
E;IG.S.17.
,'-'
..
FIO.:S,lS..
-_.
o
o ... lcr"
-PatyDa!II.IIni
... ,....
- ~.
~ .'':'~'-''''-'-'~ .:.
....
i.ii.
1.0
..
'a . '," ; .
:ss-
.... F&n-bd
",
-.~-.-~....
>v
...
.,/
Cl:!:Do~._~_
. . N . . . . . . . .
'L'!A~~--~"~~~~2D7"----~..
~----~
..
.~,~.~~--,~.~~~..
"==----"
..~--~..
"He enerwJMeV)
"HrI energ)' (MeV)
flG. 5.19.
,flG.520.
._.w-
-~ftl
.~--~
'';.0
FIG. 5.4. Experimental elastic scattering data and an empirical fit according to elastic scattering cross sections He-d compared to
Rutherford cross section (dashed line) according to Eq. (5,6.) and Rutherford crosssection data (Kellock and B.glin, 1993),
An additional prerequisite for any detector, however, is to
5.3 METHODOLOGY
The aim of an ERD experiment is to perfonn elemental
analysis by detecting the recoil ions in forward directions. There
are two main tasks to be solved: The recoil ions have to be
identified in terms of their nuclear charge or mass, and their total
energy has to be measured in order to obtain the depth
information.
handle the huge number of projectiles scattered into the direction

of the recoil-ion detector. The possibility of using heavy
projectiles to avoid scattering of the projectiles into the recoilion detector has already been mentioned (Section 5.2.1): If
projectiles much heavier than the elements in the sample are
used, no projectil~s are scattered to angles largeI than qluw. [Eq.
89
Chapter 5
(5.3)] because of two-body elastic scattering kinematics. Thus,

the detector can be placed at scattering angles cp > 'Pmu to avoid
scattered projectiles entering the detector.
If the scattering angle cannot be chosen large enough
because of the lack of sufficiently heavy enough projectiles or
because of the need for small scattering angles, the scattered
projectiles have to be bandIed in some way. The number of
projectiles scattered into the direction of the detector (NRBs) is
much larger than the number of recoil ions (NERD) for normal
ERD scattering geometries where the scattering angle cp is
smaller than 45. The ratio NrudNERD is given by the ratio of the
corresponding scattering cross sections. Considering Rutherford
cross sections for botb scattering projectiles at a matrix element
of nuclear charge Z2 and mass M2 and recoil ions of type i of
interest for ERD analysis, one obtains
(5.7)
To obtain a rough estimate of scattered projectiles compared to

recoil ions, we assume the relations Mp ~ M2 ~ Mi and M j ;::: 2Zj ,
which implies
N
-.!!l!.. " cot'
cp.
from both the ERD and RBS signals. In this case, it is desired to
accept the scattered projectiles in the ERD spectrum. However,
the huge nwnber of projectiles scattered from heavy matrix
atoms should be prevented from reaching the detector in order to
attain a sufficient count rate of small elemental fractions at low
background when sensitive analysis. of light elements is the aim
of an ERD experiment.
In the foHowing sections, we summarize the various
methods used to identify the nuclear charge or mass of the recoil
ions independently from depth analYSis.
5.3.1 ERD using a range foU

A range foil in front of the recoilion detector can fulfil two
tL<ks (Doyle and Peercy, 1979): (1) It can prevent the buge
background of scattered projectiles from reaching the detector if
it is thick enough to stop them. (2) If the range foil is thick
enough, it can also stop the recoiled heavy matrix atoms and
keep them from reaching the detector. This is important because
the heavy matrix atoms are not the aim of the analysis but -can
otherwise dominate the ERD spectrum.
The foil can be partially replaced by gas-filled volume
with thin entrance and exit foils. Such a setup allows a uniform
adjustment of the effective thickness of the stopper medium by
adjusting the pressure inside the volume in order to obtain
optimum thickness as a range medium.
(5.8)
NERD
The simplest arrangement for perfonning ERD analysis on a

thick sample with some type of elemental recognition is to use a
single detector, such as a silicon detector, with a sufficiently
thick range mediwn in front of it at forward detection angles
(e.g., 30-45) (Fig. 5.5a) or even at 0 when analyzing thin foils.
The projectiles must be much heavier than the target elements
considered for analysis, because then, the range of the heavier
projectiles is less than the range of the lighter and thus faster
recoil ions. In this way, the projectiles and also the heavier recoil
ions are completely kept from reaching the detector. An energy
spectrum is collected only for all recoil ions penetrating the
range foiL When such a range foil is used, it is possible to
perform hydrogen analysis already when using helium ion
beams, for example, at 2-4 MeV. Thus, standard RBS setups can
easily be transformed to hydrogen ERD setups by adding an
energy detector equipped with a range foil in the forward
direction and using shallow angles of incidence of the ion beam
onto the sample.
This ratio quickly increases much larger than 1 with decreasing

scattering angle q> < 45. For the analysis of hydrogen by helium
ERn, in which case the projectiles and the analyzed elements are
much lighter than the matrix (Mi M2 and Mp M2), one
obtains
NRBS
NERD
--=
M, )' (1 + coscp)' cot , cpo

M, +M,
cos<p
(5.9)
In this case, the relative number of projectiles scattered into the

ERD detector is even larger than for heavy-ion ERD_ Thus, a
huge number of scattered projectiles have to be counted and
identified in addition to the desired recoil-ion yield, or one has to
make precautions to keep the scattered projectiles from entering
tbe detector. Higher energies of the He projectiles help to reduce
the number of forwardscattered projectiles entering the detector
because the scatterin cross section for the ions on the matrix
atoms decreases as c but the ERD cross section on hydrogen is
already enhanced over the Rutherford cross section and might
even increase because of the strong interaction_
Figures 5.5(b) and 5.5(c) show the sample setup for ERD
using 2.8 MeV He ions at a scattering angle of 30 and the
resulting energy spectrum. The sample was a 400-nm
polystyrene layer on top of deuterate<! poly(methyl methacrylate)
(PMMA) tha~ in tum, was on 400 nm of poly(vinyl chloride)
(PVC). The recoil ions were transmitted through a l2_1.J.!m
aluminized Mylar foiL Both the deuterium and the hydrogen
recoils appear in one energy spectrum_ The depth profile
Sometimes, it can be useful to analyze the scattered yield of

projectiles from heavy elements as an RES signal simultaneously
with the yield of recoil ions from the light elements in order to
obtain full elemental information in one single measurement
90
calculated by slab analysis (Section 5.4.2.2) is shown in Fig.

5.5(d). It shows a depth resolution of about 50 nm, which is
mainly influenced by kinematic broadening through the solid
angle of detectioll (Section j.5 .6) and by energy straggling OC the

recoil ions in the range foil.
2.8 MeV He+
Recoils
Hanc;l~
400nm PS
D-(PS-FiMMA)
400 nmpVC
(bl
Energy (MeV)
Depth (rim)
1.2
0.6
0.8
1.0
6000 . - - - , - - - - , - - - - - , - - - - - . - - - - ,
5000
1000
800
600
400200
6~-r~~rr~~rT~~rT"~~~,",&
N-
d)
"Q
5"
"b
c)
4000
x 4
E
~
c
-;;-3
3000
11
3 .:
c
2000
'"-- -I
...........,
,,,
1000
~
<:2
B
C
~-
2 c
~
U
0 ,
I!
, .,J'"
.0
80
100
120
140
":160
180
100
Channel
80
60
......... " ' ,
...
40
Depth (Vgrnlcnf)
20
FIG. 5.5. (a) Principle of a range-foiVsolid-state detector combination for the analysis of light elements by ERD. The projectiles have
to be much heavier than the element of interest, so that the projectiles scattered from any type of matrix element will be stopped in the
range foil. If thick enough, the range foil transmits only the light recoil elements. Such an arrangement is sometimes sufficient for the
identification of light recoil ions. This method is widely used for analyzing hydrogen in thin films. (b) Layered sample that is analysed
for hydrogen isotopes by utilising 2.8 MeV He-beam and range foil (!2~m aluminised mylar) method (Green, 1994). (c) Energy
spectrum of hydrogen and deuterium recoil ions from the sample of (b) collected at a scattering angle cp = 30 and an angle of
incidence a = 15 Q The detector covered a solid angle of detection Ml = 3.6 msr. The H and D recoil signals are almost completely
separated with their approximate sUlface peak positions as labelled (Green, 1994). (d) Concentration depth profIles for H and 0
determined from the energy spectrum of (c). The concentration and depth scales are shown in densiry-exclusive units along the left and
bottom axes, respectively. The right and top axes were determined by assuming the density of the sample is 1.04 g1cnr (Barbour,
1995).
The conditions for ERD analysis were a 4 MeV He beam, a 20"un-thick A1 range foil in front ofa silicon detector, an incidence
angle of Il ~ 20, and a scattering angle of ~ ~ 40 [Fig. 5 .6(all
The sensitivity in hydrogen detection can be down to the

atomic parts per mllion (at-ppm) range. Figure 5.6(a) shows five
ERD hydrogen spectra as obtained from a bulk copper sample.
91
Chapter 5
(Tripatbi e. al., 2004). No ion other than hydrogen recoils were

able to penetrate the range foil. The untreated surface usually
contains some 10 15_10 16 atoms/cm2 of hydrogen as long as the
surface is not well prepared under ultrahigh-vacuum (UHV)
conditions, which makes a background in the bulk through plural
scattering (see Section 5.6.2.3). After subsequent sputtering
under UHV. the surface hydrogen coverage is reduced, and thus,
this surface treatment reduces the background in the bulk to the
sub-part-per-million level. The spectra show the sensitivity that
can be achieved when using a hydrogen-less range foil (here
aluminium). Plural scattering of ions stopped in the range foil
can produce additional background \\o'hen using hydrogencontaining range foil such as the commonly used Mylar foils
(Tripatbi el aI., 2004).
a)"" r-~.,-,-~~~_~~~~~~
A better depth resolution than from He ERD can be

achieved by the range-foil ERD technique when_ heavier
projectiles are used because the heavier projectiles allow the
range-foil thickness to be reduced. In' addition, other light
elements such as Li, Be, B, C, N, 0, and F become accessible.
An example is presented in Fig. S.6(b) for the analysis of a 300run-thick Si j N4 layer on top of a silicon substrate by 28 MeV
silicon ions at a scattering angle of q> = 300 with use of a 12-,...m
Mylar range foil in front of a silicon detector (Barbour and
Doyle, 1995). The foil transmits only the light recoil ions from
the sample, The energy spectrum shows hydrogen signals from
the Si3N4 layer added to the counts fTOm nitrogen. Depth
resolution for hydrogen has improved in this case, as one can
estimate from the slopes in the energy spectra, but it is still
limited as a result of the energy-loss straggling in the range foil.
b)
. (1,) Before 'Spullerir,g

. cBj ,15 rmlutes. ~PtJttenn,"
". 11!11 C) 3O'mlfwl$.sl:llltierlhg
, (O).4S.mim.iles$i:iutterlng
. :El BO irin'uieS *,"firIn~
liLllJ._e--1 X1 ,
.,_'~~~~~~~~~~lL~.=_.~~~_~O
;Z
-:iQa
.' ~
, om',)
.....
......
.f:ne:I'UY ~V)
Energy(....eV}
1000
r----'i-----i4- - - - . . : ; ._ _--.,
200
The drawback of the simple range-foil detector arrangement

in heavy-ion ERD. however, is that all recoil atoms reaching the
detector are collccted in one energy spectrum. The various recoil
atoms show up at different energies in the spectrum because of
oL---,----,r----r~~~~~
a
100
400
500
their different kinematic factors and their respective energy

losses in the range foil [Eq. (5.1)J. Thick samples spread the
energy spectrum of each element, leading to overlapping mass
and depth infonnation and thus to a similar ambiguity problem
as in RBS analysis. As a consequence. the measurement
sensitivity for elements in overlapping spectra is limited.
FIG. 5.6. (a) Hydrogen profiles frOID a bulk copper sample

using range-foil analysis and a 3 MeV He beam. Hydrogen
contamination is present only at the copper surface. The peak
width reflects depth resolution. Sputtering reduces the hydrogen
contamination with time. A bulk sensitivity of less than 1 ppm is
achieved at low surface contamination. (b) Energy spectrum of
recoil ions collected from an approx.imately 300-om-thick silicon
nitride layer. A 28 MeV Si beam was used for the analysis with
<p = 300 and a. = 25:>. The hydrogen is present as a minor
constituent of the film remaining after atmospheric-pressure
chemical vapor deposition at 800 c. The hydrogen signal
appears on the nitrogen background.
How does one choose the optimum range foil? The

thickness depends on the range-foil material and the types of
ions that should be stopped or transmitted. In standard cases, the
foil should be thick enough to keep the projectiles out of the
recoil detector. The sum of the foil plus the dead layer of the
detector should be equal to the range of the scattered projectiles
plus about 3 times the range of straggling in order to keep most
of the projectiles from reaching the detector. Good estimates can
be obtained from range calculations using the SRIM code
(htrp://www.srim.org/;Zieglere.al. 1985). Ranges in Mylar,
aluminium, and CF4 gas are plotted in Fig. 5.7 for various ions as
functions of their energy. Sometimes, thinner foils can be
acceptable when projectiles are sufficiently slowed do'WIl to be
sorted out of the spectrum. Foils that are too thick reduce the
transmitted recoil-ion energies and increase energy straggling.

Regarding the effects of energy straggling and small-angle
scattering, the range foil should consist of atoms that are as light
as possible, such as beryllium or carbon materials. The simplest
foils are some type of plastic material such as Mylar or Kapton
foils, whereas beryllium is seldom used because of its toxicity.
92
However, hydrocarbon materials have the disadvantage that their

light ions, especially hydrogen, can be scattered by secondary
reactions from the range foil into the ERn detector and produce
of llli and E=. All light and medium-heavy elements, including

hydrogen, are well-separated and identified.
background for the hydrogen analysis (see Section 5.6.2.3)

(Tripatbi el al., 2004). Thus, hydrogen-free materials are a better
directly from the energy losses of the various elements at a given

energy (Fig. 5.9). The separation works efficiently even for
various isotopes of the same element when the total energy is
beyond the stopping-loss maximum. In that case, the stopping
force Sj scales as
The potential of elemental separation can he estimated
choice for sensitive hydrogen analysis. Because hydrogen-free
carbon materials are difficult to prepare, aluminium foils can

offer a good compromise, but quality (uniform thickness,
absence of pinholes. etc.) has to be chetked carefully. The foils
should be amorphous or nanocrystalline in order to avoid
s. 0:: Z2M In Ei
channeling effects in the range foil.
E[ .
(5.1 0)
Heavier recoil ions can have recoil energies such that the
energy loss is less than the energy-loss maximum, in which case
the scaling of Eq. (5.10) no longer holds true. Elemental
separation from the energy-loss signal is hardly possible at low
ion energies (Fig. 5.9) when the energy-loss differences for the
various elements are smaller than the 6E resolution. The
important exceptions are hydrogen isotopes,. which can be well
separated far below the energy-loss maximum.
5.3.211E-E analysis
The full potential of ERD analysis is obtained when the
elemental analysis is performed independently from the depth
analysis. In addition to the energy of the recoil ions, one needs a
second independent signal to separate the elemental information
from the depth information. A quite simple solution is to use a
.6E-E serup. where an energy-loss signal is gained from the
recoil ions using a thin transmission detector in coincidence with
the signal of a second detector that is placed directly behind it in
order to measure the residual energy (Stoquert et al., 1989;
Gebauer et aI., 1990; Assmann et al., 1994; Bergmaier el al.,
1998). Adding the energy-loss and residual-energy signals gives
the total energy of the detected ions. If the .6.E signal is plotted as
a function of the total energy, the different elements are
separated from each other because of their different energy-loss
signals for a given total energy. An example of a dE-E
measurement using a detector where the 8E part consists of a gas
ionization chamber and the residual energy, Erell, is measured by
a silicon detector is shown in Fig. 5.8. The analyzed sample was
an approximately 200-nm-tbick Al.Gal_. .N layer deposited on
AhO). It was analyzed using a t70 MeV 1271 beam, a scattering
angle of q> = 40.2, and an incident angle of a = 15, The energy
loss, tt.E, is plotted versus the total energy, Bill" which is the sum
There are several ways to- set up 8E-E detectors. A

straightforward solution is to combine rn,'o solid-state detectors,
one behind the other, where the ftrst detector is a thin (Le., 5-10
/lm) silicon detector (Yu and Gustafsson, 1986; Arnold Eik el
al., 1992; Wielunski el aI., 1997) [Fig. 5.10(a)J. Although these
detectors are simple to handle, they have disadvantages. The first
disadvantage is the low energy resolution of the AE detector
because of the large noise level of such thin solid-state detectors.
A second disadvantage is the sensitivity of these detectors to
radiation damage, wbich is a severe problem because a large
nUIilber of low-energy recoil and scattered projectiles are
stopped by this detector whereas only a few particles reach the
residual energy detector. An additional range foil limits both
elemental separation and dcpth resolution as a result of energy
reduction and straggling
93
"-
""
~]
.,
Chapter 5
~:
. 10000
'1000
1 oil
, 1
'cij
a
o
,f
c
.Q
0,1
aluminum
O,Ol,~~~,,",,~~=~~~=~~~~
0,1
10
100
1000
ion energy [MeV J
0,0~,L"1~~"4~~~10;-'---U.~100!;;--'~-';1~OOO
ion energy [MeV I
.!
,5
"
94
. "
.~.~-
.'
.80
'"''
"':>'"
~' .. '; ...'.'~'~"-~,,

'ob'1~~~~.~,~~~~~,~a~~~~~'00~~'~"'~'~~1ono
:4'. ..
20
:'
lpri ene
"
,'~-~o
. -- c
~~a
t-ii""""""
Etot [MeV]
'.:.~~ .
............ ,........ .
FIG. 5.8. ilE-E spectrum of a 200-rrun-thic:k Al,;Gal_xN layer on

an AlzOJ substrate obtained using a 170 MeV 1271 beam and a
hybrid detector at cp = 41.5 0 where the ilE part was measured
using a gas ionization chamber tilled with CF 4 gas and the
residual energy was measured using a silicon detector
(Bergmaier er a/.,1998).
......
"~"
....,' ..... 'Ii

'H
Mylar (H.C",O,)
1,1
Better perfonnance has been demonstrated using ionization-
- 10
chamber-t}pe gas detectors [Fig, 5,10(b)] where the anode is

divided into two consecutive stripes. The electrons in the
detector are allowed to drift transversally to the impinging recoil
..........
100
10
Ior1
.....
allergy
.......
1000
[MeVI
--. ~
ions in order to obtain a .dE and a residual energy signal. A
Frisch grid confignration (Frisch, 1944; Assmann er ai., 1994)
makes the energy signal independent of the ion position where

the ions enter the detector. Although the energy resolution for
protons and He ions is still' slightly worse for ionization
chambers than for silicon detectors, ionization chambers offer
reasonable energy resolution even for such light ions, and it is up
to a factor of 0,5 better for heavy ions (Kottler er ai"~ 2006),
Thus, for ions heavier than lithium, gas ionization chambers
even outperform silicon detectors in terms of providing better
energy resolution. In addition, these detectors do not suffer from
irradiation damage if the detector gas is routinely changed.
Moreover, the cathode-t~-anode signal ratio can easily be used
to obtain additional position information wlth a resolution of
better than 1 mm. This allows for the acCurate measurement of
.' ,..,,~.
0,
... ......
'
". .
....,
1'0.0
....
CF.(gas)
'Ii ';"-'...~
He ..
. (p=tol3l:10'Pa..T"'293K}
";,
ogo)J,,.~---"-~=~~~~~lka-c....-'"--'''''jJ'OO"--~+-'-~U;;,aao
. ion enerii/ I MeV j
FIG, 5.9, Stopping forces of (a) aluminium, (b) Mylar fail

(H gC100 4), and (c) CF4 gas for various ions as a function of ion
energy as calculated by SRIM-2008,04 (http://www.srim.org/).
95
Chapter 5
energy loss
"
AE
ions
transmission silicon detector
(5-10 Jll11)
silicon
stop detector
40mm
top view
~~~~sl:TJ
dx=O-7rnm
~gasout
cathode
./1
impinging - - ' /
~"s
. e("ltfan~ foils
position sensitive
Si PIN diode
I~
.... ,~ri.~ ~~d. ,
.AE.1.
. .6.E2 .:." '.;.':;
ai"iode
gas In
FIG. 5.10. Various t.E-E detector setups for elemental analysis in ERD analysis. The ions enter from the left, deposit a fraction of
their energy in the t section, and deposit the residual cnergy in the second section of the detector. (a) Schematic for an solid-state
detectors for .1.E and residual energy. (b) Electrode configuration of an ~E-E ionization chamber (Assmann et at., 1994). The Frish
grid makes the anode signal independent of the position of the ion track. The cathode is structured in a backgammon geometry in order
to obtain additional position information. (c) Hybrid gas (fOf .1.E) and solid-state (fOf residual energy) ERD detectof. The hybrid
detector has larger dynamics in energy and type of particle analysis. The anode is divided into two stripes to obtain twil:e as much AE
information for redundant elemcntal information.
the recoil-ion scattering angle at large dctcctor acceptance in

order to correct for the kinematic shift (Section 5.5.6). Thus,
depth resolution is maintained when large solid angles of
detection (e.g., several millisteradians) are used (see Section
5.5.6). In addition, two-dimensional position resolution can be
introduced when, for example, the cathode is structured in a
configuration has been used to perform, ERD blocking

experiments (Nolte et al., 1998). The disadvantage of such
detectors is that they are not commercially available but rather
have to be set up by each group itself, although consttucrion
schemes are available from the pioneering groups. Another
disadvantage is that the length of the detectors is limited. As a
backgammon configuration (Assmann et al., 1994), Such a
96
consequence, it is normally not possible to simultaneously

identify hydrogen recoil atoms when analyzing heavier elements.
To circumvent the problem of analyzing recoils with too
large of a range of variation in pure ionization-chamber-type
detectors, hybrid detectors consisting of one or several .6..E gas
ionization chambers and a silicon detector for residual energy
measurements offer a good compromise in order to stop all
recoil ions ofintcrcst [Fig. 5.10(c)J (Bergmaier et al., 1998). In
the hybrid case, the residual energy solid-state detector can be
operated as a position-sensitive detector in order to correct the
kinematic shift or to perform blocking experiments. Solid-state
residual energy detectors suffer from irradiation damage but not
as much a:s solid-state 6.E detectors do. A regular change of the
detector after some beam time might provide acceptable
performance with respect to the advantages it gives over pure
ionization-chamber setups.
~
MCP2
'1Op 12
... ,
500
"0
4dO
.I:
U
;:
300
.2'
Li:
0
A general limit of the fiE-E technique is the minimum

energy required to separate the various recoil ions through a
specifiC energy loss. AJthough separation of hydrogen isotopes is
possible even at very low ion energies, other detectors have to be
used to obtain elemental or mass information of the recoil ions at
lower energies, as such information is obtained only when using
projectiles at or even below 1 MeV/nucleon.
III
200
i=
100
5.3.3 Time of flight versus energy (TOF-E) method

A second widely used method for recoil-ion identification is
to analyze the mass of the recoil ions by measuring their energy
Ei and velocity Vi indep~d~ntly:
i )'
M = 2E2i =2E(TOF
L
I
Vi
(5.11)
Whereas the energy measurement can be perfonned by either a

solid-state detector (Groleau et al., 1983; Thomas et aI., 1983;
Whitlow el al., 1987; Arai et ai., 1992) or an ionization chamber
(Kottler et al., 2006), the velocity of the ions is measured by the
time of flight of the ions (TOF~ through a defined flight path L
(Fig. 5.11). This method can especially be used at low recoil-ion
energies where the 6B-E technique is no longer useful, although
some TOF-E setups are used even at higher energies (Bohne et
aI., 1998).
100
12 MeV
200
300
400
:;00
Recoil Energy I ch
- FIG. 5. \ \. Principle of TOF-E methods: (a). Start-stop
configuration for TOF measurement that allows bigh TOF
resolution. Secondary electrons that are generated in thin foils
(usually carbon) are used to start and stop signals at the
beginning and end of a flight path L. The energy of the ions is
measured in an independent energy detector. (b) Example of a
TOF-E ERD analysis, showing the potential of separating the
masses of all light and heavy elements in the sample to obtain
background-free elemental spectra. Depth profiles can be
generated from energy spectra that are separately projected for
each resolved mass.
Although the principle seems to be relati vel y simple, the
TOF-E method requires sophisticated detectors to achieve its
full potential. The first problem to be solved is to measure the
97
---
----------- ----Chaptet 5
time of flight of the ions accurately. Because mass resolution
oM, depends on both TOF, resolution O(TOFi) and Ii; resolution
when the pulse width is small enough and fast stop detectors are
used or when long flight paths are utilized.
SEil
&., =
M,
(0,)'
+2.(OIOF,)'
E,
TOF,
5.12
one must obtain a relative rOF resolution better than

6(TOF i)rrOF j < 0.5% and an energy resolution of slightly less in
order to obtain a mass resolution better than 11100. Recoil ions
of 1 MeV/nucleon have a velocity of about 1 nslcm. Thus, a time
resolution of 250 ps is necessary at a flight path of L = 50 em.
Higher recoil-ion energies or a desired better relative mass
resolution requires better time resolution or longer flight paths.

Thus, large-area timing detectors and energy detectors are
necessary to cover several milllsteradians of solid angle of

detection.
To date, two methods have been used to measure TOF. The
first is to llse a pulsed beam and a solid-state energy detector
simultaneously to create a fast timing signal. The timing signal is
used as a start pulse for timing electronics [e.g., time-to-analog
converter (TAe) or time-to-digital converter (TDC)], and a fast
timing signal correlated to the beam pulses, for example,
deduced from chopper or buncher resonators used for beam
pulsing pwposes is used to stop it. TOF can be deduced from
this time difference At as TOF = 10 - 6t for an arbitrary constant
to The problem is to obtain a pulsed beam with high quality (no
background beam between pulses and a subnanosecond time
spread) and to achieve sufficient ii.ming resolution from the
large-area energy detector. Pulsed beams from cyclotrons as well
as chopped and bunched beams from electrostatic accelerators
are used. Typically, the overall time resolution achieved by a
pulsed-beam system is no better than 0.5 TIS (Bohne et aI., 1998).
This is sufficient to separate hydrogen isotopes and the light
recoil ions, but it has limited. sensitivity and mass resolution for
heavier elements. On the other hand, in principle, there seems to
be no limit to obtaining better perfonnancc using pulsed beams
The other widely used method to obtain TOF data is based

on separate start and stop detectors that each consists of a thin
foil that creates a secondary electron pulse when an ion is
transmitted through the foil [Fig. 5.1 L(a)]. The secondary
electrons are accelerated by a grid close to the foil and led to a
secondary-electron detector, which is, in most cases, a
muLtichannel plate (Mep) detector_ An example for such a TOFE analysis is shown in Fig. 5.11 (b).
To avoid large time spreads due to fast electrons that stem
from forward-scattered electrons at the exit side of the foil, one
can use the electrons emitted from the entrance side. There, the
electron energy spectrum is close to a Gaussian distribution with
a mean energy of about SeV and a similar energy spread, and
additional energy tails are small. The number of electrons
emitted depends on the foil material, espo::ially its electron
affinity and bowing of the electron conduction band at the
surface, the foil thickness, and the electronic stopping loss of the
ions. The electron yield is normally greater than unity for all iom:
except high-energy hydrogen ions.
Several configurations are used to direct the secondary
electrons from the beam direction onto the channel plate (Fig.
5.12):
(a) The secondary electrons are guided in a homogeneous

magnetic field and deflected 1800 onto the MCP [Fig.
5.12(a)] (Renfordt et af., 1981; Zebelmann et af., 19~7).
(b) The secondary electrons are reflected by an electrostatic
mirror onto the MCP [Fig. 5.12(b)] (Busch et af., 1980).
(c) The thin foil traversed by the recoil ions is tilted by a certain
angie. Thus, the electrons can be accelerated
perpendicularly from the foil, reaching the MCP outside the
beam directioll after some drift length without deflection
[Fig. 5.12(c)] (Groleau et of., 1983)_
98
'Jrid 1kV
,,/<>:.~;bon f~iiOV
,.' ..... :
ion~s~~~~~~_;'f'~~'~~~'
grid "2kV.. " ", 0
9ri~ {~V'
.sii:fe ~I~'~'
. . .. : '
'
' . '
HatH:um cyclotrori
____~____~
.'
. ~~~ 'MCI'> in 1KV
time detector
~~\
""
\1
anode
4-5 kV
2-3 MCI'
detectors
god
2kV
carbon foil
OkV
FIG. 5.12. Various configwations to deflect secondary electrons out of the ion path at constant electron flight paths to the electron
detector for fast timing: (a) 180 0 magnetic field deflector, (b) electrostatic mirror configuration consisting of electrostatic grids, (c)
tilted configuration with 00 electron detection.
Method (b) is probably the most widely used, although the

most compact is method (c). A disadvantage of method (c) is
that the effective foil thickness increases because of the tilt.
However, tilting the foil has the advantage of intrinsically
correcting, to first order. for the kinematic shift of the time of
flight. The situation is plotted in Fig. 5.13. The kinematic energy
shift given by Eq. (5.2) transfers into a shift in the TOF when the
flight path L is constant for all scattering angles:
.1TOF
TOF
- - = tan <p.1<p + higher orders .
angle cp relative to the beam direction (Fig. 5.13) [Dollinger

2009]. However, the kinematic shift in the energy-dispersive
detector remains and can be corrected only through an additional
measurement of scattering angle (Bohne et aI., 1998).
In most cases, thin carbon foils (1-10 Jlg/cm2, meaning a
linear thicknessof 5-50 run) are used as secondary-electron
emitters because they can be produced thinnest with respect to
energy loss, energy-loss straggling, and angular scattering
effects. The thinnest give minimized energy straggling and are
used for high-resolution work (Liechtenstein et al., 1997).
Especially for hydrogen analysis. a problem remains because of
low and possibly varying detection efficiency that results from a
low secondary-electron yield. The electron yield increases with
roi I thickness until saturating at several micrograms per square
centimeter (Koschar et al.,. 1989). The thickness where
(5.13)
The scattering angle varies because of the solid angle of

detection and the beam spot size on the target. Both effects are
corrected to first order when both foils are tilted by the scattering
99
Chapter 5
saturation occurs depends on the recoil-ion velocity. The average

secondary--electron emission from carbon is less than unity for
protons of about 1 MeV. The total efficiency is limited to about
60010 even when detectors with single-eJectron sensitivity are used.
Use of thin-film coatings such as MgO (Gujratbi and Bultena,
1992) or LiP (Bohne et ai., 1998) or other salts provides some
improvements, but the stability of the secondary-electron yields
from these layers remains a problem. In addition, increasing the
thickness of the foils leads to larger energy straggling and sma1langle scattering. Detection efficiencies for heavier recoils,
however, are very close to unity already when using pure carbon
foils because of the fact that the electron yield (scaling with the
electronic energy loss) is much larger than unity for the heavier
recoil ions.
In most cases, the relative TOF resolution 6(TOF;)fTOFi is

much better than the relative energy resolution of the energydispersive detector. Values of o(TOFi)ITOFi < 0.5% can be
'achieVed using moderate flight paths (-O.5m) and good timing
resolution (o(TOFi) ;; 200 ps]. When the recoil-ion mass has
been identified, the time-of- flight information can be used to
calculate the recoil-ion energy where the timing resolution
transforms into a relative energy resolution
oE; = 2 o(TOP)
E;
TOF;
(5.14)
Thus, depth profiles calculated from TOF spectra show

improved. depth resolution compared to standard solid-state
detectors.
53.4 Magnetic and electrostatic spectrographs

.. Any magoetic spectrograph [Fig. 5.14(a)] (Gossett, 1986; .
Bergmaier et ai., 1996; Kimura and Nakajima, 2001) allows for
some recoil-ion identification, because the radius fi of a recoil
ion of charge qi is given for a homogeneous magnetic field B by
M:v
.j2M;E;
r = - ' - ' = -'----'--'.'
q;B
q;B'
(5.15)
If the energy E j of the recoil ions is determined in addition to

their position in the focal plane of a magnetic spectrograph, then
identification of recoil mass Mi becomes possible, ex.cept for
some ambiguity due to the various charge states qi of the recoil
ions. Sophisticated focal-plane detectors have been constructed
that measure the position and scattering angle of the recoil ions
by two position-sensitive resistive-wire proportional counters.
Several redundant measurements of 6E and total energy E are
made by using a segmented ionization chamber, thus allowing
for redundant particle identification from elemental number and
mass analysis [Fig. 5.14(b)].
Any energy-dispersive detector in rOF-E, AE-E, or even

magnetic analysis can be replaced by a transverse electric field,
as is used in electrostatic spectrographs [Fig. 5.14(c)] (Enders et
af., 1992; Schiettekatte et ai., 1996). Using the electrical or
magnetic analysis in front of a stopper foil allows analysis of the
energy or momentum of the recoil ions with high precision
before the particles are transmitted through the thick range foil
and elemental and mass identification is performed. The
achievement of much better energy resolution (a total relative
energy resolution of 7 x 104 has been reported for a real ERD
analysis (Dollinger et ai., 1992, 2004; Dollinger, 1993) and thus
high. deptb resolution is indeed the main reason to accept the
FIG. 5.13. TOF detector setup where the kinematic effect is

corrected to a fIrst-order approximation by tilting the start and
stop detectors by a scattering angle of rp with respect to the
recoil-ion direction.
100
large additional experimental effort when using electrostatic or

magnetic spectrographs. In addition, there are cases in which the
use of redundant particle identification helps to obtain a high
sensitivity (part-per-million or even better sensitivi~) for

elemental analysis where simpler ERD setups might suffe::r from
background (Bergmaier el ai., 1996).
1m
. focal plane
ion
ho,~nv'
focal plane detector
High Hesolution
8RD-Spe-cLrQmeter
1m
FIG. 5.14. (a) Q3D magnetic spectrograph used for particle identification and high resolution depth profile analysis (Dollinger el aI.,
1992). (b) Electrostatic spectrometer for high-resolution ERD analysis (Enders et aI., 1992).
5.3.5 Transmission ERD for thin-foil analysis

The most simple case of ERD can be used when analyzing
thin foils where the energy losses of the projectiles and recoil
ions are small. The foil can be analyzed in transmission
geometry using a single detector at a certain scattering angle, for
example, at Il' = 30, without any range foil. In that case, the
elements are identified just by their energy through kinematics
[Eq, (5.1)] (Assmann and Maier, 1995). Usc of projectiles much
heavier than the matrix atoms keeps the scattered projectiles out
of the detector, and the energy scale can directiy be calibrated
into mass scale because of the monotonic and (approximately)
linear energy-mass relationship. Ambiguities with scattered

projectiles Vlill arise only when lighter projectiles are used, but
such difficulties can be handled when the stoichiometry- of the
target is simple. A passive foil in front of the detector and a welladjusted scattering geometry might be suitable for separating
ambiguity problems for certain cases through different energy
losses and kinematics of the projectiles and the recoil ions.
Analysis of thick foils produces overlapping profIles of different
elements, which destroy the potential of ERD for the elementselective detection of recoil ions, and the more sophisticated
detection schemes as-described above should be used instead.
101
Chapter 5
5.3.6 Coincident ERD

Spectrometry (ERCS)
or
Elastic
Recoil
Coincident
Kinematic coincidence is an elegant method for

independently obtaining mass and depth information when
analyzing thick foils. It requires the simultaneous detection of
both the scattered projectile and the recoil ion from a two-body
reaction. At least one of the particles has to be transmitted
through the sample, and the sample has to be thin enough for this
pwpose. Three different detection geometries can be utilized
(Fig. 5.15; Hofi;llss ef 01., 1991) in which the projectile, the
recoil ion, or both are transmitted through the target. To make
the method suitable for thicker targets, high-energy light ions
(i.e., proton or helium beams) are preferred because of their low
stopping cross sections. The laboratory-frame scattering angles S
and <fl of the scattered projectiles and the recoil ions,
respectively, are directly related as long as elastic scattering is
considered:
(scatiered
10115\
M,
(b)
(5.16)
sin(2<p) cot 3 + cos(2<p)
IM
deiectcH 2
(reeDlled afGl'I'lsl
. ;:\y::
ion beam
. ,-,~~-, .~~---
I\{I
By measuring or selecting the two scattering angles, it is possible

to determine the mass M; of the recoil ion from the known
projectile mass Mp. Alternatively, the energy relationship
between the two particles of a scattering interaction can be used
to determine particle masses or the scattering angle, In the case
of elastic scattering with negligible energy loss, one obtains the
following relationship using the kinematic factors Is. of scattered
projectiles and k; of'l'coil ions of type i [Eq. (5.1)J:
del~Clof
(scaitered Ions)
'
E, -E= = E, -EM, =k, -k, =2.k, _I =(~M,Mj '<05,,,,)-1

E/ -rEus
Ep
I'
+M/
..
".
(5.17)
".
.. !: :. -t .
If substantial energy loss is involved, the energies of the

projectile (E,), scattered projectile (E RBS ), and recoil ion (E;)
depend on the depth, x, where scattering takes place. Because
the fractions of energy loss are different for the various particles
and angles involved, a consequence is that the ratio (Ei _
ERBS)/(EERD + ERBs) cbanges with depth. However, the ratio
remains an unequivocal function of scattering angle fP and
masses Mj and Mp that can be calculated or calibrated in detail
(Hofsllss ef al., 1990, 1991). A special case is the analysis of
hydrogen by proton-proton scattering, where projectile and
target nucleus are identical particles. This case has most
potential for hydrogen analysis especially at ion microprobes
because it causes the lowest damage potential (Reichart et al.,
2004). The special features of proton-proton sca~ering are
discussed in detail in Chapter 9.
102
. ~ f:'I"
.-
'.:
,...
",
!,
..
:... ,
' .. ,: .~.~'~.
,:
FIG. 5.15. Setups for coincident ERDA: (a) Both scattered

particles are transmitted through the sample. The detectors sit on
opposite sides of the sample. (b) Both scattered particles are
transmitted through the sample. When the particles have
different masses, they are scattered to different angles. Thus, m'o
circular detectors can be used to cover the maximum possible
solid angle of detection. (c) Tilted foil method. Only one particle
from the two coincidentally detected particles (normally the
lighter one) is transmitted through the sample. Thus, heavier
atoms can be detected as in standard ERD analysis in relatively
thick, free-standing samples.
The main advantage of using wincidel1t ERD is that light

projectiles can be used at high projectile energies where nuclear
reactions can take place. Because of the kinematic coincidence,
even nuclei produced by nuclear reactions can be sorted out.
This allows the use of light.ion beams, given that the scattering
cross sections are enhanced compared to Rutherford cross
sections, and thus offers a reduced damage potential (see Section
(5.19 )
5.6.3).
If all relevant elements contributing to the layer or foil are
5.4 QUANTITATIVE ERD; DATA EVALUATION
5.4.1 Thin-layer analysis

There is a straightforward way to obtain the elemental
concentration of a thin layer either because the recoil ions can be
directly assigned to the layer of interest (as is often the case) or
because a thin foil is analyzed. The number of detected ions Yi
scattered from element i is given by the following equation when
Np projectiles hit the target:
N. do
N do
Y =N - ' E.-' Ml=N n . - E . -i L'lQ.
I
PAidD,
piAlclQ
measured
Note that the concentrations can be determined without the

knowledge of the number of incident projectiles (beam charge)
and without the exact detector geometry, in particular, withQut
the solid angle of the detector. The accuracy of the elemental
fractions DI is given by counting statistics, knowledge of "the
detection efficiencies Cj for the various elements, ao.d the
accuracy of the cross sections do/d!!. For thinlayer analysis,
the accuracy can additionally be limited by elemental
contributions from neighboring layers that produce backgroLlnd
in the layer of interest. The accuracy afthe elemental fractions is
discussed in more detail in Section 5.6.1.
N j atoms are within the beam'
area A on the target (with the areal density NlA). Thus, Il; ~ N/N
is the atomic fraction of element i, da/dn is the differential
recoil scattering cross section at scattering angle <p, and .Ml is
the solid angle of the detector. .Ej is the efficiency for the
detection of element i in the recoil-ion detector including dead
The values of interest are the concentrations nj of all

elements i in the layer. The ratio of the concentrations of two
elements can be calculated directly from the associated yields
and differential cross sections as
= 1), the concentration ni can be calculated as
(5.20)
(5.1S)
time effects.
k=l
It is assumed that Ni atoms of type i out of m different elements

or isotopes and a total of N =
(Ln
. 5.4.2 Thick-layer analyses

The energy spectra of aU recoil atoms together contain the
elemental depth profiles. To extrnct the depth profiles from the
energy spectra, one needs the energy depth relationship for each
element and/or isotope. During scattering at depth x, a portion ki
(kinematic factor) [Eg. (5.1)] of the projectile energy E,(x) is
transferred to the recoil atom: E;(x) ~ k;E,(x). For a recoil atom
of type i scattered at a certain depth x underneath the surface, the
detected energy of the recoil E~(x) is written as

m [
. sex')]
sex')] dx' E:"'ex) = k, { E, - ,f!
njex')-'-:dx'}- 'm
JL[ njex')~
o j=1
sma.
()
The first term calculates the recoil energy directly after the
scattering event by using the projectile energy reduced by the
energy loss of the projectile in the sample up to thickne.o;s x and
the kinematic factor k j of Eq. (5.1). The second term de:scribes
the energy los5 of the recoil ion on its way back to the surface,
j:1
smi3
fs~ex')dx'.
(5.21)
and the third term is the contribution of the dead layer or range
in front of the detector. The stopping force,
foil
S~(x) = dE~ Idx , indicates the energy loss per path length dx
in the passive layer (denoted as layer 0). The calculation uses
Bragg"s rule that the stopping force in a layer at depth x' is the
103
~,
Chapter 5
sum of the stopping forces Spix~ and S~(x~ of the pure

elements j for the projectiles and the recoil ions, respectively,
weighted by the corresponding elemental fractions n;(x} The
angles a and ~ are the angles between the surface and the beam
direction and recoil-atom direction, respectively [Fig. 5.l(b)].
Note that these angles are defined to result in shallow angles, SO
that the sines can be approximated by their arguments for rough
estimates.
A,(x) 8(E:"')
n;(x)
YJEf"'(x)]
N N dO"; "'Oe.
A 'dO
'
..
..
(5.24)
AA
Equations (5.21) and (5.24) are the master equations to be

solved for the concentration profiles Di(X:)'
The concentration Dj(X) of element i at depth x is obtained

from
n;(x) =
= Y,[Ef"'(x)] PWA.~
.6.e.
N --"'
dcr;
dO
Transforming the linear thickness into an "area] density",

[xp(x)A,]/[A,(x)], of each layer, one can' calculate the depth
profiles on the areal depth scale (i.e., atoms/em' scale). Only by
knowledge of the density profiles, that is, from an independent
density measurement, can one obtain the linear depth scale of the
depth profiles. On the other hand, if linear dimensions of a layer
of interest are known from elsewhere, one can evaluate the
density profile of the layer by ERD analysis.
(5.22)
The yield Y, [Ef'" (x)] issummcd over an energy bin lie; of the
recoil energy spectrum of species i where E~(x) corresponds
The set of equations given by Eqs. (21) and (24) only

implicitly contain the concentration profiles, because the recoil-
to the recoil energy at thickness x. The" energy bin ne; at energy

. SIab of tholckness Ox at depth x,
E iD" corresponds to a certam
ion energy, E~et J depends on Dj(X) and vice versa. The solution
which is given by the effective stopping force aE~1;1 lax, which
of this set of equations can be obtained in several ways, as

described in the following subsections.
is the change in the detected energy for a change in thickness Ox

and is obtained as a derivative ofEq. (5.21). The total nwnber of
atoms per unit area, N/A, is given by
N =
"'e;
p(x)A,
8(E"") ,
A,(x)--'-
5.4.2.1 Simulation-fitting approach

The Simulation-fitting approach involves the following
steps: Assume concentration profiles for the m relevant elements
in the sample. Calculate the m energy spectra including a11
knowledge on physical and experimental parameters, and change
the assumed profile until an optimum fit to the data is obtained.
This procedllre can be done with most of the programs
developed for ioa beam analysis spectrum evaluation using an
analytical or Monte Carlo-type simulation approach. Those
programs known to the authors are listed in Table 5.5. One
should be aware that m individual spectra have to be generated
for the m elements in the sample and compared to the measured
spectra, making fitting by hand a complex task. Automatic
versions of data fitting can be used, but they still have to be
carefully adjusted for reliable data to be obtained.
(5.23)
Ox
where p(x) is the density of the layer at depth x, A,(x) is the

average atomic number (gImol) in the layer, and A, is
Avogadro's number (A" ~ 6.629 x 10" atoms/mol). Thus, the
denominator on the right-hand side of Eq. (5.23) is the effective
stopping cross section for the detected recoil ions given in
standard units (i.e., eV em'), Combining Eqs. (5.22) and (5.23),
one obtains the fraction Dj(X) of element i at depth x as
,.
104
Table 5.S. Programs available for data evaluation and simulation.
Nam<: of program
GrSA
TyPe of analysis
Analytical simulation
RUMP
RBX
NDF (Data Furnace)
SIMNRA
MCERD
Monte Carlo simulation
CORTEO
KONZERD
SRIM

Direct transformation (slab analysis)
Transport of ions in matter tables and
DEPTH
5.4.2.2 Direct transformation scheme (slab analysis)

Because a complete set of m energy spectra i.s measured for
the m elements, the RBS ambiguity of elemental concentration
and depth profile is avoided. This allows for straightforward data
evaluation by ,olving Eqs. (5.21) and (5.24) iteratively. To do
so, one divides the sample into slabs of unifonn thickness Ax
(i.e., same number of atomstcm1). Assuming arbitrary values for
the m elemental concentrations nj(xl) in the first (surface) layer
(depth Xl), one can determine the fust energy bin li.ej for each
element from Eq. (5.21). Thus, the number of detected recoil
ions of each type i, Yi(E~), is determined for this first layer,
and the concentration can be calculated according to Eg. (5.24).
These results give new starting values to calculate the elemental
concentrations n;(xJ) of the first layer a second time. This
proccdure nonnally converges quickly to self-consistent values
for the ni(x\), The same procedure is perfonned fOT the
subsequent slabs xJ, where the integration in Eq. (5.21) is
replaced by a sum over the slabs in front of the actual layer Xj,
assuming a constant elemental fraction across the slab thickness
6x. Thus, the concentration profiles ni(x) ue calculated directly.
Under the condition
..
I , D;"'I
Reference(s)
Raubala, 1984; Saarilahti and Rauhala,
1992
Dootittie, 1985,1986
K6tai, 1994, 1997
Szilagyi et al., 1995; Szilagyi, 2000
Barradas'et al., 1997, 1998; Jeynes et
01,,2003
Mayer, 1999,2002
Sajavaara et al., 2000; Arstila et al.,
2001
Schiettekatte, 2008
Bergmaier et at., 1995
http://www.srim.org/;Zieglerelal.
1985
The direct transformation scheme has been used by several

ERD groups because it often gives the most direct information
from the measured spectra in terms of the desired depth profiles
without detailed physics simulations. The most widespread
program for this direct transformation with various variants is
the program KONZERD (Bergmaier et al., 1995), which is
. based on the program SQUIEKY (Borgesen e/ a!., 1982) that
was originally developed for RES analysis. The program is
based on a roam algorithm that is also able to calculate the
statistical error for the calculated concentration profiles from the
count statistics. An example of such a direct transformation is
shown in Fig. 5.16, where the elemental profiles of an AlGaN
sample on AhO] substrate have been extracted from the raw data
of Fig. 5.8. An error bar is also shown, as it was calculated
. ~- directly from the statistical error in the energy bins belonging to
this ,lab.
The direct transformation approach delivers depth profiles
that contain the full experimental resolution. Thus, the profiles

are not subject to overinterpretation as can be the case in the
simUlation-fitting
approach.
However,
unfolding
the
instrumental function,. for example, by Bayesian methods, is
hardl y possible, and one must, for instance, rely on the limited
depth resolution given by the experimental conditions.
the set of equations is
overdetermined if the product of NpMl is fixed by measured

values. Therefore, NpM'l is normally left as a fitting parameter
that is used as a control parameter to determine whether its value
. remains constant over the entire depth and whether it coincides
with the measured value within the given errOrs. If all relevant
elements of a sample are measured but one, the mC8S\UCd
product NpMl can be used instead of the missing energy profile
to calculate the concentration profiles of all elements including
the element of the missing energy spectrum.
5,5 DEPTH RESOLUTION IN ERD ANALYSIS

The deceleration of the projectiles on their way into the
sample and of the recoil ions on their way out of the sample
converts into an energy-depth relationship that is evaluated in
terms of elemental depth profiles. The depth resolution, bx. is
therefore directly related to the total energy resolution, aEi. of
the experiment and the shift, dE~' [Eq. (5.2I)J, of the detected
energy of recoil ions of type i obtained for a layer of thickness
dx
105
Chapter 5
Ox=
BE
Oct
dE, Idx
(5.25)
Considering a Gaussian energy distribution obtained from

an infinitesimally thin layer, the energy width aEi will be taken
as the full width at balf-maximum (fwhm) of the energy
distribution, making two adjacent layers at a distance 8x
distinguishable, where the fwhm is related to the standard
deviation of the mean,
by fuhm ~ -/8()n 2)&. In absolute
cr,
values, the energy shift dE~ Idx for a certain layer thickness dx
varies with each type of recoil ion. A better overview is obtained

when using relative va1ues for energy resolution and stopping
forces. Close to the surface, in the absence of a range foil the

depth resolution ax is given by (Dollinger et al., 1992).
10
20
30
40
depth [10" at! em ']

~~~
Ox
,_~
:~::,,::,,,::,::,:c"!:'::::::~'
(dE./dx)
E.
+ (dE;ldx)
sin a
E;
sinJ3
(5.26)
10"
106
tiro
J= 10.2
EiC
Q)
E
Q)
:a;
____ _
_ _ _ _-I
Al
As can be seen from this equation, the depth resolution depends

on the sum of the relative energy losses of the projectiles and
recoil ions divided by the sines of the angle of incidence and exit
[Fig. 5.1(b)], respectively. Although the absolute values of the
energy losses of heavy projectiles are much larger than those of
the ligbt recoil ions, the relative energy losses do not differ that
much, and thus, the recoil ions can contribute substantially to the
total relative energy loss. The use of shallow incident angle a
and exit angle ~ relative to the sample surface, rather than the
angles to the surface normal, makes it easy to estimate the
. _. influence of changes in these angles on the energy resolution
from Eq. (5.26) because the sine functions can be approximated
by their arguments in the case of small a and ~. It can also be
recognized from Eq. (5.26) that a shallow angle of Incidence a.
helps to enhance the contribution of the energy loss of the
projectiles and that the contribution of the recoil ions to the
relative energy loss can be kept low for J3 > a.. A relative energy
loss of about 0.1% is obtained at about 0.5 MeV/nucleon for
heavy ions and shallow angles of incidence of a < 50 for single
atomic layers (e.g., 10 15 atoms/cm2). Thus, even monolayer
depth resolution comes into reach. Although shallow exit angles
P increase the total energy loss, it is normally not belpful to
make ~ too small because differences in path length become
large for large opening angles of the detector. ~ :; 2a offers a
good compromise considering all aspects.
"'~':-
Ga
to
c
........
\
\
10'3 \ \
.0
I
I
I
I
J
\'-~-~"f--""'''-:
1o '--,-'-...,.-.o-,---_;>--'-----:--;--~~
a
10 '.
20
30
.40
depth [10'1 at/em' ]

FIG. 5.16. Elemental depth profiles as obtained from the raw
data of Fig. j,8 for an AixGal_x)N layer 00 an AhO) substrate
using the direct transformation routine KONZERD (Bergmaier
el al., 1995): (a) linear plot, (b) logarithmic plot also sbowiog
the minor element (0) of the layer (Dollinger e/ a1., 2004).
~!
1;
-';
';1
.:,;,1
1
The spread of the relative energy loss is the sum of several
contributions. Assuming statistically independent Gaussian-type
contributions to a first approximation, the total energy spread
SE/Ei for recoil ions of type i is the quadratic sum of the 1\vhm
of each contribution
A
:~i
:~
iJ
.,,
.~
:~
Elastic Recoil Deteclion: ERD
I
(5.27)
The first three contributions are due to the experimental
which can be cut from any intense ion beam using a lrighresolution analysis magnet and its slit system.
semp and do not change with analysis depth. They consist of the
energy spread introduced through the phase-space volume being
occupied by the beam, the intrinsic detector resolution, and a
Doppler-effect contribution, respectively. In addition, there are
The
to the relative energy sp.l"ead,

from the beam's angular spread, d'l'bao" is
dominated by a kinematic effect that is given by Eq. (5.2). "Note
that a beam divergence of d<p = I rnrad already causes a CJ.2%
contribution
(oE.,IE,~,
two depth-dependent contributions that are due to energy

straggling and small-angle scattering effects~ respectively.
energy spread through kinematic effects at a scattering angle of

<p = 45. Because kinematic effects decrease for small scattc::ring
angles {Eq. (5.2)], larger beam divergences can bc accepted at
reduced scattering angles. The bearn's angular spread also makes
a depth-dependent contribLltion because of its effect on path
length. This contribution has to be linearly subtracted from the
contribution of the pure kinematic effect in order to obtaiIl the
total {8EpfEp~Ip' However, the other depth-dependent
contributions nonnally dominate when the beam divergence is
well-controlled, and the path length contribution from. the
angular spread of the beam can usually be neglected.
All effects are included in many simulation packages (see
Table 5.5 for a list of available programs). A more detailed

examination of the effects can be pcrfonned using Monte Carlo
codes, where detailed energy distributions '2.5 well as nonGaussian contributions are treated in detail (Table 5.5). Surface
and interface roughness or other lateral structures in the sample
affect the energy profiles and thmi increase the measured energy
spread. The influence of roughness on the depth profiles has
been included in some of the newer evaluation programs. The
problem is that different structures lead to similar energy profiles
that cannot easily be separated from energy spread effects
according to Eq. (5.27). Careful treatment of all contributions
must be implemented to ensure correct interpretation of the data
and to obtain additional information on roughness.
5.5.2 Doppler effed

The kinematics given by Eq. (5.1) is correct for a recoil ion
at rest in the laboratory frame before scattering. Because 0 f the
thennal and zero-point motion of the atoms in the solid state, the
condition of an atom at rest before scattering is not ex.actly
We next describe the main effects on energy resolution, and
thus depth resolution, according to Eq. (5.27).
fulfilled. Thus, the energies calculated from Eq. (5.1) have to be

corrected. Because of similarities with the acoustic case, where
the velocity of the emitting source bas to be considered, this
5.5.1 Beam contribution to energy spread

The relative energy spread, (oEpI'Ep)'xam2 , relevant for the
total energy spread of an ERn experiment consists of two
contributions: the relative energy spread of the beam as
effect is also called the Doppler effect in the case of particle

scattering. Although the ground-state and ex.citation energies
~m of the atoms before scattering are below 1 eV, the energy
spread introduced by the Doppler effect is up to a factor of 10-3
of the analY-.led energies. In the case of isotropic motion of the
atoms before scattering at temperature T, the contribution of the
Doppler effect to the energy spread for Ep Ei,exe and E,, kT
delivered from the accelerator, (8E,lEp)"......, aod the energy

spread introduced by the angular spread of the beam,
)EpIE~~. Because the two contributions are, in general,
statistically indepehdent, one obtains
is calculated as (Kruse and Carstanjen, 1994; Kruse, 1995):
(~-~)E,a,COth(E"",).
( BE,)
E. dop =J8ii2
cas m
E. EP
kT
(5.28)
In standard cases, the energy spread, (BEy'Ep)beatn,M?, of

electrostatic accelerators is in the range from 1 x 10""" to 1 X 10-3
The energy spread of cyclotron beams is normally worse, but in
most cases, it is sufficient to perform standard ERD. Only highresolution ERD requires energy spreads on the order of2 x 10-4,
't'
(5.29)
The Doppler effect makes a significant contribution to the

energy-loss spread at low recoil energies E i. Zero-point motion
and excitation energies CEi,ac) are the largest and Ei is smallest in
analyses of hydrogen isotopes because they are the lightest
atoms. Ei,ae far ex.ceeds thermal energies at room temperature
for hydrogen atoms [E",,, kT ~ cath (E"" IkT) od 1
107
'.',
..
... - ...---------~-'-'---'-------
Chapler 5
Hydrogen bound in water molecules has a large ~ac value (~e

= 0.46 eV), which is not much reduced for deuterium (Eac =
0.35 eV). When hydrogen is chemisorbed to metal surfaces, Eo;c
is reduced, for example, to about 0.10 eV and 0.07 eV for
hydrogen and deuterium, respectively. For instance, when
analyzing cbemisorbed hydrogen with 0.2 MeV/nucleon heavy
ions such as Ne, CI. or Au ions, the energy spread from the
Doppler effect is calculated to be (oE;IE,)dop ~ 1.9 x 10" for
hydrogen bound to water, which is a significant and limiting
contribution to the total energy spread in high-resolution
hydrogen analysis. For heavier target atoms. for which Ei,CJ;~
becomes much smaller than kT, Eq. (5.29) can be approximated
as
OEi)
( E.
= _.J8_ln_2
dOP
COscp
J... __l )kT,
( Ei
BE(X)]
[
...,
E,
and the contribution of the Doppler effect to energy spread

becomes only a minor fraction.
The anisotropic case of the Doppler effect is described in
Kruse and Carstanjen (1994) and Kruse (1995), where the
excitation energy is considered to be much larger in one
direction than in the other dimensions (one-dimensional case).
5,5.3 Energy straggling

Whereas the energy and angular spread of the beam and the
detector resolution can be controlled by the experimentalist,
energy straggling cannot be avoided and is, in fact. one of the
main limits for depth resolution in depth profiling. [BE(x)lE]sln!
can be approximated by Bohr straggling (Bohr, 1948) at high
energies and by additional correction factors for ion energies
around or below the stopping maximum
(5.30)
=_.J8_ln_2 _e_' 47tZ, N x[W,'CZ"E,)Z,' + W;'(Z"E)Z,'],

V
sma.
smj3
E 41tEo
where ~ is the average nuclear charge of the target and NN is

the atomic density in the target The factors Wp(l"E,) and
WO<l"E;) are factors th,t correct the Bohr straggling to the nue
energy-loss scattering values where the main considerations are
charge-state fluctuations (which enlarge straggling) and
screening effects (which reduce energy straggling) for 110t
completely stripped ions at low energies. The factors cannot
currently be calculated accurately for heavy ions at low ion
energies, and semiempirica1 values such as Yang's correction
mctors (Yang and O'Connor, 1991) can be used,. In the case of
thick targets and changing elemental content in the layer. the
contribution of each individual layer has to be quadratically
added for the projectiles on their way into and for the recoil ions
on their way out of the target. The additional energy spread of a
passive layer (e.g., absorber foil or dead layer of the detector)
must also be quadratically added. We refer to Chapter 2 for a
detailed discussion of energy straggling.
(5.31)
for a certain layer thickness .6x decreases when the incident

angle a is reduced, and thus, depth resolution increases. Shallow
angles of incidence, therefore, give the best depth resolution as
long as the roughness or waviness of the sample and path-Ien~
effects of small-angle scattering do not deteriorate the depth
resolution.
Optimum depth resolution is obtained at about the energyloss maximum, where the ratio of energy spread to energy loss is
a minimum.. This is because the energy spread remains nearly
constant over a large range of ion energies. The ratio of th~
cnergy spread to the maximum energy Joss decreases slightly
with decreasing ion mass. Thus, lighter ions at their energy-loss
maximum are preferable to heavy ions in terms of depth
resolutio~ but the differences are small for different types of
projectiles. To better control the effects of beam damage,
however. projectile energy can be reduced below the energy-loss
maximum. causing a slight loss in depth resolution but a gain in
2
the nwnber of counts because of the -lIP" increase in the
scattering cross section. However, small-angle scattering effects
(see next section) have a greater influence on depth resolution at
low ion energies and thus set a lower limit for the energy. Ion
energies of less than 0.2 MeV/nUcleon might, therefore. be
unfavorable for ERD work.
Especially for heavy-ioD beams at or below the stopping

force maximum, the accuracy of available straggling data is
limited. For example, energy straggling of 60 MeV Ni ions in
thin carbon foils shows 30% less straggling than calculated from
Y aug's fonnul. (Schmelmer e1 al., 1998). Because straggling,
(oElE)_, scales only with the square root of the path length, the
ratio [(oElE),.,.J/[IIEI""j of straggling to the energy shift IIElllx
108
,l
~.
.,l,
:~
1
I
"{
{
~
!'
angle scattering affects both the incident projectiles and the

recoil ions in a statistically independent manner, the total effect
is the sum of the kinematic and path-length effects of both .
particles.
5.5.4 Multiple scattering effects

Multiple scanering effects influence depth resolution in two
ways (Fig. 5.17). The first effect is kinematic, as the original
direction.s of the projectiles and recoil ions are changed. Thus,
the scattering angle for an ERD scattering event also changes.
The angular spread, d<p, is transfonned into an energy spread,
(5FJEJ.", through scattering kinematics according to Eq. (5.2).
A second contribution, (SE/EJpath, comes from the lateral spread
introduced by multiple scattering when the path length of the
ions between the sample surface and a certain depth varies.
]n ERD experiments, scattering angles q> between 300 and

0
45 are commonly used, where a strong first-order contribution
to the energy spread from small-angle scattering d<p occurs
according to Eq. (5.2). Its influence can be reduced if smaller
scattering angles arc used (e.g., q> = 100 or q> == 15), as is often
done in high-resolution ERD work because of the reduced firstorder contribution (Dollinger et ai., 1998a). For thin-foil
analysis, even the 0 0 configuration can be chosen where firstorder contributions are zero. A second possibility to reduce
multiple scattering effects, in general, is to enhance projectile
energy, because small-angle scattering 'Nidth scales as lIE
2
whereas plural scattering scales according to l1E .
recpiUons :
Angular distributions from small-angle scattering can be

calculated accurately, and thus, so can energy distributions
(Sigmund and Winterbon, 1974). The full widths at halfmaximum of the angular distributions of the incident projectiles,
dlp p , and the emitted recoil ions, dlpj, can be approximated by
FIG. 5.17. Sketch of the kinematic and path-length contributions

to energy spread through small-angle scattering. Energy is
reduced through kinematic effects when the angle of incidence a
is increased to Ql and thus the scattering angle <p is enlarged.
This energy reduction can be partially compensated through an
accordingly shorter path length. Similar compensation of the
kinematic and path-length effect occurs when the angle of
incidence is reduced to Q2 and thus the scattering angle qJ is
reduced and the path length is increased.
(pasztielai.,1991)
41toE p a
Small-angle scattering has - large cross sections. These-multiple scattering events lead to a continuous angular
distribution close to a Gaussian angular distribution and reduce
the depth resolution already below the flfSt atomic layers of the
analyzed sample. larger scattering angles occur a second time
after the ERD scattering event with only a small probability.
Such plural larger-angle scattering leads to tails in the energy
and thus in the depth profiles. Here, we discuss mainly multiple
scattering and its influence on depth resolution. Plural scattering
limits sensitivity and is discussed in Section 5.6.2.3. Analytical
simulation codes already include multiple scattering effects, as
well as plural scattering to some extent, whereas the full effects
can be calculated accurately only by Monte Carlo-type codes. A
compilation of available simulation codes is listed in Table 5.5.
{"a'
[_1_ xp(x)A,
sinet A,(x)
d<p:~ C 2Z;Z,e' {"a' [NX

.
In Eq. (5.32), Z, ~
In;Z;
]}D
(5.32)
xp(x)A,
sin ~ A, (x)
' _ 41toE;a
]}D
represents the average value of the
atomic numbers in the target. The constant C = 0.5 and the

exponent D = 0.62 give reasonable values for the parameter
range relevant for ERD analysis; a is a screening radius for the
interaction potential given by
O.885a o
a
(
Z,
2/3
+Z, 213 )112
with the Bohr radius ao = 0.529
The two contributions, from kinematics and from path

length, have much more influence on ERD spectra than in
con.ventional RBS analysis where the kinematic shifts at
scattering angles close to lSOc arc: small and changes in path
length are also small for perpendicular incidence on the target.
However, shallow incidence angles produce similar path-length
effects for ERD and RBS depth profile analysis. Because small-
10
x 10- m.
Thus, the fust-order contributions of the amall-angle

scattering distribution of the projectiles and recoil ions to the
encrgy spread as calculated from Eq. (5.2) is
109
L
~ C 2Z Z,e'
d<p
,.
Chapter 5
(E~E)
tio,p ""-
2Z,Z,,' [ , I
Ep tan'P C 4m:oE pa
1tEl
<P(X)A.]D
sino. Al(X)
(533)
(534)
The projectile contribution normally dominates the small-angle

scattering effect compared to the contribution of the recoil ions
because ZlEj and l/sin p are smaller than ZIE, and l/sin a in
standard ERD experiments.
spread is dominated by the energy-loss spread contribution,

which has a large uncertainty in iu. value. The partial
compensation of the smallangle scattering contributions, that is,
the kinematic and path-length effects, becomes visible in the
more or less constant contribution of the small-angle scattering
effects to the total energy spread at an intermediate depth range.
The squareroot behavior can be recognized at small depths,
whereas the XL62 behavior dominates at large depths. The
calculation shows that it becomes possible to resolve single
graphite layers close to the surface when the relative detector
resolution is much better than 10-3.
-'tcltalertlltOY~
-. erta'VIlos:i straggrll'lg (scaIed~
1.'
--- rnutllple Scal!emg (1KOIl1on)

__ ~ I!f'II!'r0" BpnlIId oCthe be~m
_
be3m~'
-
1.' ~
~60 MeV ~I ~ carPo~
The path-length effect can be calculated from the lateral

spread when its geometrical effect is approximated in first order:
o:e
~'1O"'
',... 'encrw loss stf&gg:il"l!l

.~-
mulliple~[prcftttiIe)
dE.
= 0.56
( 'E)
E JIIIb.i E,
dX' _X_cot a
Ei sin~
C2Z;Z,e' ["" _1_ XP(X)A.]'.

4JU:oEi8
(5.36)
sin~ Al(x)
0.'
and dominates the energy spread at large depths
when shallow angles of incidence a are used. Becawe of the

also makes long tails in the energy spectra. Thus, one should
to dominate the energy spread.
The kinematic and geometric effects from small-angle
scattering cannot be quadratically added because they depend
statistically on each other. The situation can be discussed with
respect to Fig. 5.17. If an ion is multiply scattered Sllch that the
ERn scattering event requires a larger angle than the usual
scattering angle fp, the recoil-ion energy is reduced. Commonly,
this situation is accompanied by a shortened path length Wltil the
depth ~ is reached, and thus, energy loss becomes smaller and
recoil-ion energy larger. As a consequence, the kinematic cffect
and the path-length effect can compensate for each other in some
way. Because of the statistical nature of the small-angle
scattering effects, there is never complete but only partial
compensation.
All contributions to energy resolution and thus depth
resolution are analytically treated and approximately added in
the right manner in the simulation program DEPTH (Szilagyi e/
al., 1995). As an example, the results are plotted in Fig. 5.18 for
the analysis of a pure graphite sample using 60 MeV l27y
,projectiles at a. = 4 0 , fp = 10. In this case, the total energy
110
0'OO~~~~~~~~'~5~~,,~,;~~~~;:~:::
dePth (1014 at/cmt , n'ml
long tails in the lateral distributions from plural scattering, this
avoid measurement in the regime where path-length effects begin
15,
,. g.
Whereas both the kinematic effect and energy-loss straggling

X1.62
,. i
1.0 ~
0.'
scale with the square root of the thickness x, the path length
scales with
0.
FIG. 5.18. Calculation of all contributions to the energy spread

(left ordinate) and thus depth resolution (right ordinate)
according to Eq. (5.27) for the analysis of a pure graphite sample
by 60 MeV 1211 ions. Depth scale and depth resolution are given
in natural
which can also be read on a nanometer scale
when assuming a density of p = 2 g1cm3 which is close to the
density of perfect graphite (p = 2.28 g/cm'). The calculation was
performed using the simulation program DEPTH (Szih\gyi el al.,
uruts,
1995).
5.5.5 Detector resolution
In most cases, the detector resolution is the limiting
contribution to depth resolution close to the surface, where
energy-loss straggling and small-angle scattering effects are
small (see Fig. 5.18). Energy-dispersive detectors such as silicon
detectors and ionization chambers are limited in energy
resolution because of electronic noise and the statistics of charge
creation. Although silicon detectors can reach an energy
resolution of close to 10 keY for 1 MeV protons, the energy
resolution decreases for heavier ions as a result of energy-loss
straggling in the dead layer of the detectors; inhomogeneities of
the dead layer; and as sometimes postulated, by plasma density
effects. The relative energy resolution, SEIEt, has been shown to
5.5.6 Correction of kinematic shift
remain nearly constant (close to 1%) for all ions analyzed in an

ERD experiment using solid-state detectors at optimized
conditions. Ionization chambers, as often used in ERD analysis
because of the possibility of easily dividing the detector into
several ~ stripes, offer less energy resolution for protons but
remain more constant in energy resolution with increasing
nuclear cbarge of the recoil ions. It has already been shown that
use of ultrathin, homogeneous Si)N4 entrance windows for the
gas detector often gives energy resolution that is comparable to,
and sometimes even better than, that of solid-state particle
detectors, with the additional advantage of baving nearly no
sensitivity to irradiation damage (Kottler et at., 2006). Dividing
the detector into several tJ.E stripes decrcases the energy
resolution because of the additional electronic noise
contributions from the additional amplifiers needed. All together,
as already stated above, energy-dispersive detectors give total
energy resolutions of not much better than 1% at best, which
leads to a depth resolution close to the surfaee of between 5 and
50 run depending on the energy and, thus, the relative energy
loss of the projectiles, as well as on the angle of incidence a.
Using the range-foil ERD method reduces the energy resolution
even further and therefore reduces the depth resolution
significantly because of the large energy straggling of the reCl}il
ions in the several-micromctcr-thick range foil.
A major problem for energy and thus depth resolU""1ion

comes from the angular acceptance of the detector. It adds an
additional contnbution due to a kinematic effect to the dete etor
energy resolution. Large solid angles of detection are require d to
minimize irradiation damage effects in the target during an-BRD
experiment (see Section 5.6.3) and to achieve a suffic::ient
counting rate. The variatio~ dcpv, due to the opening angle of the
detector in the scattering plane produces kinematic broadening
(8E/E;)n according to Eq (5.2). For a desired rnaxirnum
contribution of (oE/EJ1I to the total energy spread, one r::nust
limit the opening angle dcpn of the detector considering the li:near
approximation ofEq. (5.2) for a non-0 setup.
60 MeV "'I'" beam
002-layers of graphite
Improved energy resolution is obtained using time-of-flight

(TOF) detection, especially for ion energies in the range of 0.2
MeV/nucleon, where the energy resolution of energy-dispersive
detector is limited (see Section 5.3.3) (Giangrandi el ai., 2007).
Because the energy resolution from TOF measurements is better
than that from energy-dispersive detectors, one can normally use
the TOF infOlm.ation to produce the depth information in TOF-E
detector configurations after particle identification. The limit in
TOF measurement is given by the energy and angular spread
introduced by the start detector in start-stop setups. It can be
circumvented in the pulsed-beam configuration, but in this case,
high-quality pulsed beam andlor long flight paths must be used.
(a)
Much better energy resolution is obtained using magnetic or

electrostatic spectrographs. The intrinsic energy resolution can
be as good as 2 x 10 (Uifller et at., 1973), and a total energy
resolution of? x 10-4 (Dollinger, 1993) has been reported in a
real ERD experiment that contained all additional contributions
to the energy spread according to Eq. (5.27). This energy
resolution allows for the resolution of single atomic layers from
a pure graphite sample (Fig. 5.19) (Dollinger et aI., 1998b).
Beyond several atomic layers, the energy spread increases as a
result of depth-dependent straggling effects, and monolayer
depth resolution is lost (see also Fig. 5.18).
o
(b)
0.5
hlllE [%]
FIG. 5.19. (a) Schematic drawing of graphite analyzed b) highresolution ERD. (b) Experimental data for the ERD analysis of
graphite by 60 MeV 1271 ions. Several single atomic layers an be
resolved, as shown by the deconvolution model (dasht:d lines)
(Dollinger e/ aI., 1998b).
III
'.p
Chapter 5
corresponding energy spectra and, therefore, depth profiles. The

thin-foil spectra can be used to calibrate thc correction for thickfoil analysis.
(5.37)
A special case of high-resolution kinematic correction has
been applied using a Q3D magnetic spectrograph, where the
kinematic shift was eliminated up to the third-order contribution
[Eq. (5.2)] by using additional multipole fields (quadrupole,
bexapole, and octupole) for beam deflection (Dollinger, 1993).
The kinematic shift contribution (BE/EJJ. for an acceptance

angle of dQJ.L perpendicular to the scattering plane has only a
quadratic and higher-order dependence. Thus, the maximum

angle dtp.l for an acceptable contribution (5E/c;h to the energy
shift is calculated as
Because of path-length differences for the recoil ions, the

angle-energy relationship cbanges with analyzed depth from that
obtained for pure scattering kinematics in the analysis of thick
targets.
The change depends on the contribution of the recoil-ion
(5.38)
energy loss to the total energy loss and changes with the
analyzed recoil ion and with the ERD measurement geometry,
that is, the angles of incidence a. and exit p. Again, as in the case
The same condition as in Eq. (5.38) is obtained in the case of a" - . of small-angle scattering, the effect of path length decreases the
0 setup for both angular directions where the flrst-order
kinematic energy shift. The correction for this depth-dependent
kinematic effects are zero. A cUlVed slit following a circle with
behavior is similar to that for the kinematic shift itself. It can be
constant scattering angle <p and an aperture dcpl enlarges the
controlled by an angle-energy plot for any recoil species of a
usable solid angle of detection in the case of q> 0. However, it
multilayer target (Dollinger, 1993; Neumaier el at., 2001).
is inconvenient to use the required axial detector geometry, where,
in addition, one has to account for a change of exit angle ~.
5.5.7 Deconvolution of ERD spectra
*"
Although all contributions to energy-loss straggling [Eq.

(5.27)] can be optimized by choosing appropriate beam
conditions, detector resolution, and ERD geometry, the result is
a limited depth resolution.
The only valuable enlargement of the solid angle of

detection is possible when the scattering angle is mea:sured and
when this information is used to correct for the kinematic shift.

This approach has been used in position-resolving detector
setups, for instance, in pure ionization-chamber approacbes
(Assmann e/ al., 1994),' as well as hybrid gas-solid-state
detectors [Fig. 5.10(.)] (Bergmaier e/ ai., 1998) or TOF-E
The only possibility to drive depth resolution beyond the

limit as obtained from Eqs. (5.26) and (5.27) is to measure the
elemental profiles with good statistics, which allows one to
deconvolute the measured spectra from the contributions to the
energy spread. The accuracy of a deconvolution, however.
depends on a perfect knowledge of the experimental parameters,
especially the accurate geometry and energy calibration of the
spectra and, most critically, knowledge of the total energy spread
at each energy for every analyzed element. Because the
theoretical values of energy-loss spread, which dominates the
total energy spread in standard ERD experiments, have large
uncertainties, there is only limited use of deconvolution schemes
for improving depth profile analysis regarding depth resolution.
setups (Bohne et af., 1998). The kinematic correction using the

position-sensitive detector of Fig. S.lO(c) is shown in Fig. 5.20
for the analysis of a thin foil target Figure 5.20(a) shows
measured data plotting the position (calibrated in scattering
angle) versus energy without kinematic correction. The data
show the expected decrease of energy with increasing scattering
angle according to Eq. (5.1). Figure 5.20(b) was obtained after
the energy for all scattering angles had been corrected to that of
the mean (bere, <p = 38") by multiplying each measured energy
by a factor of (cos~ 38)/(cos2 <pl. The correction yields vertical
lines in the angle~nergy relationship and minimuni widths in the
112
".
:...J-
'/
"
37
,~ '37;5
'g>
"
'l'
een .", '."
:':C
"
.~
.. 5. .. ~
In
. _.;
. 39
'9.51"-
39.5
.0
d)
soo
200
u
m
150
400
-c
>=
:- 300
>-
100
200
AI
50
.AD
100
o :-O'"'-~f.'0"""~-:2'='0~'"'":3~O"-'-"-..
f:;-.J...-:50::'""oL,:."o""...J
0a
10
20
30
40
JAl
50
60
.-.-.. -."recoil energy', MeV 1-----------.-.:...-.----~, recoiLenergy .LMeY 1.. ___.______
-FIG. 5.20. (a) Plot of scattering angle versus recoil-ion energy measured from a 20-nm-thick ~~bon foil cov.ered by~a thiii'~of aAl 0
.,.--
~.
,-
--
2 3- -
film. The energy spread from energy loss is negligible, The angular information is obtained from a position.sensitive silicon detector,
The kinematic energy shift is visible for all elements, resulting in broad peaks in energy projection. (b) Softwarecorrected spectrum
with a constant energy-angle relationship and enhanced energy resolution.
of 1 run assuming a density of 2 glcm3 The figure shows the

original data and the most probable profile that represents the
data with the constraint of minimum informational content.
Figure 5.21(a) shows the evaluation of the data using a
calculated resolution function AoEl'TH from the DEPTH code
(Szilagyi el aI., 1995), whereas Fig. 5.21(b) shows the same
evaluation but lL'iing an experimental resolution function ~ as
obtained from analyzing a thin multilayer sample. In addition, an
To avoid overevaluated profiles. one can use Bayesian~type

data analyses within the framework of maximum entropy
approaches and additional regulations such as adaptive kernels
(Fischer el aI., 1997; von der Linden, 1995). Applied to ERD

data, the evaluation methods search for a depth distribution that
is consistent with the measured data under the knowledge of
instrumental and physical constraints as discussed in Section 5.5
and that has the minimum informational content avoiding
artificial structures in the depth profiles. An example of that
approacb is shown in Fig. 5.21, where the rzrofile of 13C atoms
implanted at low energies in an amorphous 2C layer is shown as
measured by high-resolution ERD using 40 Me V Au ions and a
error band (hatched area) that represents the I" band of aU

possible profiles that would also be in line with the measured
data from the given statistics is plotted. The two reconstructions
differ from each other because of differences in the response
functions. However, the bimodal distribution that was subject of
this investigation is reproduced in both approaches,
Q3D magnetic spectrograph (Neumaier ., al., 2001). Note the

depth scale, where 1 )( 1016 atomslcm1 corresponds to a thickness
113
..
Chapter 5
5.6 ACCURACY AND SENSITMTYIN ERD ANALYSIS

ERD is a method for depth profile analysis that allows
quantitative and sensitive elemental analyses. This is due to the
dOminating two-body scattering reactions and of obtaining a
clear separated signal for each element, making quantification of
the elemental content easy. Here, we discuss in more detail the
5.6.1 Accuracy
The relative statistical error, 6.n/nj, of the elemental fraction of
element i present in a thin layer is obtained from the
uncertainties, t.. YIc:o of all elements k present in the layer as can
be deduoed from Eq. (5.19):
accuracy of the elemental conccntration and depth scale and the

sensitivity that can be obtained in ERD analysis.
An.
I
- ' =n, Y,
. ~ata
(a)
data
(b)
,.,
Y0 only when the yields Y.... can be integrated without any

ambiguity with other elements, when no background is present,
and when there is no overlap with neighboring layers. In the
latter case, errors of background subtraction for various reasons
have to be considered carefully in obtaining correct estimates for
llY,. In addition, Eq. (539) holds true only when the scattering
cross sections dav'dO and the efficiencies St: are known much
better than the statistical error of the elemental counts. Relative
Rutherford cross sections are nonnally known much better than
I % as long as the contributions from screening and strong
interactions are kept small. Nuclear reaction cross sections can
be much more uncertain. Additional uncertainty can arise as a
result of irradiation damage effects (see Section 5.6.3) when the
elemental content changes significantly during analysis. In
amorphous materials, a relative statistical error doVJt1. to 0.1 %
has been achieved in the analysis of thin foils or thin layers when
the elemental signals from the layer are well-separated. This
accuracy requires a number of counts per element that exceeds
several millions and good control of all of the systematic errors.
..
0'
]'
AY.
----L[dcr.
E
,=. dQ
The uncertainty is given by statistics (.6.Y k ~.JY: for large
l222ev
II
dQ
'm
(5.39)
c
c
8
dcr. ;)
(I-n,. " )AY; +(n,-,
MAP reconstruction
with confidence interval
with respect 10 A...

2
In thick-layer analysis, there is additional uncertainty from

the lack of precise knowledge of the stopping cross sections that
enters into the accuracy of determining the elemental fractions
througb Eq. (5.24), as well as the depth scales througb Eq.
(5.21). When the contribution of the stopping cross sections for
the recoils can be kept small, thc total influence of stopping on
the accuracy of the elemental fractions remains small, because
then, the energy bins ei where the yields Yi arc determined all
scale proportionately. However, the ratio of the energy loss of
the recoil ions to the projectiles changes from ooe element to
another when the recoil ions carry a substantial fraction of the
effective energy loss through the target. Thus, an additional error
in the elemental fractions can be introduced
depth [10"allcm' J
FIG. 5.21. J3e profiles of 222 eV Be implanted in pure
amorphous 12C analyzed by40 MeV Au at '" ~ 15 using a Q3D
magnetic spectrograph (Neumaier et aI., 2001). The spectra are
deconvoluted using the deptb-dependent resolution function and
sbow 10 error bands: (a) Calculated using the DEPTH code
(Szilagyi et aI., 1995). (b) Obtained using an experimental
resolution fimction as extracted from analysing a lJefl2e
multilayer. This resolution function assumes worse energy
resolution than assumed in (a), resulting in sharper profiles than
in (a). Both profiles show a type of bimodal structure in the
range profiles.
With respect to the depth scale accuracy, aH uncertainties of

the involved stopping cross sections scale linearly with the frac
114
.0
hydrogen analysis with MeV He-ERD are discussed in [Wang
tion of their contribution to the effective stopping cross section.

Bragg's rule for adding elemental stopping cross sections to
obtain those of mixed targets can hold to better than 1%
accuracy for ion beam..c:: in the MeVlnucl. range.
2004].
5.6.2 Sensitivity: Trace element analysis
The elemental sensitivity is influenced by several factors.
The scattering angle tp and the angles of incidence a. and

exit ~ contribute similar uncertainties. The angles have to be
controlled to 1 mrad (0.06) accuracy to obtain highly accurate
ERD analysis. Another problem can arise in cl)'Stalline materials
where channeling or blocking effects can occur, thus changing
the yields dramatically. The effects are small when the incident
angle is kept far from the main crystal axes or planes. However,
some effeets might still arise from high-index crystal axes or
planes when incident beams with low divergence are used for
high-resolution work. Sample rotation with fixed scattering
geometry can help to reduce such effects. An overall relative
accuracy of better 1% in elemental concentration can be
obtained when all effects are well-controlled.
First, sensitivity is determined by the number of counts One

can obtain for a certain element.
Second, the backgrowtd level in the region of interest in "the
detector signals limits sensitivity. This background can nave
different origins, such as limited elemental separation, dark
counts, scattered atoms from somewhere other than tb.e a-Yea
or depth of interest in the target, \Wong signals in "the
detector (Le., pileup effects), plural or multiple scattenng
effects, secondary reactions, and others.
The potential for accurate elemental determination in a

thick, layered sample is demonstrated by the ERD analysis of an
AIGaN layer on an AhO] substrate. The raw data are shown in
Fig. 5.8, and the elemental profiles are in Fig. 5.16. The aim of
this measurement was thc detcrmination of the Al-to-Ga ratio x
in the Al~Gal_JlN top layer. The data analysis problem was that
the AIlO) substrate also contained AI that had to be separated
from the A1 content in the AlJlGai-:l:.N layer. Because stopping
cross sections for heavy ions are not known accurately enough to
achieve the desired high accuracy, it is essential to characterize
the aluminium slope at the interface' in order to extract an
integrated number of AI counts belonging to the top layer only.The high accuracy obtained can be demonstrated when the data
are compared to the AI fractions detennined from X-ray
diffraction (XRD) (Fig. 5.22), where the change of the lattice
parameter is converted into the AI fraction using Vegard's law
between pure AlN and pure GaN. [Vegard's law assumes a
linear relationship between the Ai fraction and the change in
lattice parameter and holds fairly well for this mixed
semiconductor (Angerer et 01., 1997).] In fact, these ERD
measurements eonfmned Vegard's law by comparison with the
XRD data. In addition, the small scatter of the data ar01md the
linear relationship shows agreement with a 0.5% relative
accuracy.
Third are effects of irradiation damage, which ultimately

limit the number of accessible counts, because any analysis
using energetic ion beams induces changes in the
stoichiometry by displacement collisions, leading to iQnbeam-induced diffusion and sputtering.
These influences are discussed in more detail in the follOWing
sections.
1.0
0.8
-~
~
0.6
.a'"
~"
Q,4
<l
:;;!
02
1.0
It is worthy pointing out that when MeV He-ERD is used in

d<r.
hydrogen analYSis, the scattering cross sections dl is usually

non-Rutherford [See Section 2.2]. To increase the analysis
accuracy, a known hydrogen standard is often used in the
measurement. Two ~es of hydrogen standards are commonly
used: mo'NIl atoms/cm (kapton et al.) and known atoms/cm 2 (Himplanted Si etc.). Their advantages and disadvantages in
FIG. 5.22. Plot of AI fraction in AlxGal-;,N layers as measured

by ERD in comparison to that extracted by XRD. Assuming
Vegard's law would lead to the full line representing a direct
proportional relationship. The fit to the data (dashed line) shows
only minor deviations (.Angerer et al., 199~).
115
".+
Chapter 5
5.6.2.1 Number of detectable counts

If a large number of scattered projectiles are detected in the
recoil-ion detector, the number of detectable recoils might be
limited by the limited ERn count rate accepted by the detector
and by a limited beam time. If the projectiles are kinematically
kept out of the detector or stopped in front by a range foil, the
minimum elemental fraction is given by the fraction of the count
rates of a trace element relative to the matrix elements divided
by the attributed scattering cross sections. As long as pure
Rutherford cross sections are involved and the projectile is much
heavier than the recoil ion (Mp}) Mj), the scattering cross section
is nearly the same for all target elements [Eq. (5.4)]. The sole
exception is hydrogen atoms (Mi = I), for which the cross
section is nearly four times larger [Eq. (5.4)J. Thus, Ibe
advantage in ERD of obtaining the elemental concentrations
directly from the relative count rates of the various elements
becomes a disadvantage in view of sensitivity when the count
rate and beam time, and thus the total number of dctected counts,
are limited. For a minimum yield Yj = 10 for a trace element of
atomic fraction ni, one has to count about 10/0j counts of the
1
matrix atoms. That means that a count of 10 recoil ions makes a
detection limit of about 1 at.-ppm. To obtain bcttecsensitivities,
it might be recommended that the recoils of the main matrix
elements be prevented from reaching the detector in order to
count only the desired trace element. This might be done by
adjusting a stopper foil in front of the detector. As a result, the
recoil ions from the matrix atoms would be stopped. but the light
trace elements would pass. This method is applicable onJy for
trace elements that are much lighter than the matrix elements,
because then the ranges of light recoil ions are much larger than
the range of beavy recoil ions. Magnetic or electrostatic
separators -or spectrographs are also applicable to reduce the
count rate of matrix atoms. These separators or spectrographs
can be combined with some range foil, but because of multiple
charge states emined from the matrix atoms at all ener~es up to
the maximum energy, complete selection is often not possible. In
those cases, the count-rate capability is one serious limit for trace
element sensitivity. For pure hydrogen analysis, a range foil in
front of the detector or a magnetic deflection with some much
thinner foil in front is sllfficient to obtain spectra that arc
completely free of scattered projectiles and matrix recoils in the
detector. In that case, there is no sensitivity limit through countrate problem>.
5.6.2.2 Limits through background

There are several reasons for background in the detector that
can mask the true events of a trace element The number NB of
background counts directly limits sensitivity. One can subtract
the background from the signal if the background is stable over
time and if it can be calibrated or calculated. In this case, the
uncertainty of background subtraction is at least
might depend on the elemental content of the sample, the count

rate, detailed beam and target geometry, and so on. In this case,
the background is in part unpredictable, and the minimum
sensitivity is given by an effective background value that lies
between
,IN; and 2N
B.
Another background problem might stem from parasitic

reactions on the target holder or on the beam entrance slits if the
beam is not weU-collimated. that is, if there is a beam halo
hitting nontarget material containing the element of interest.
These effects can be avoided by using slits situated in front of
the last bending magnet and by using additional antiscattering
slits behind the magoet that cut off any eventual beam halo.
5.6.2.3 Plural scattering effects
ERD analysis relies
two-body large-angle scattering
between the .projectiles and the recoil atoms and . a
. quasicontinuous electronic energy loss in the sample. Secondaty
scattering events create problems in obtaining correct
concentrations or elemental depth profiles. Whereas multiple
small-angle scattering mainly influences depth resolution
thrOllgh more or less Gaussian angular distributions, plural
scattering influences sensitivity through long tails, producing
backgroWld in the relevant spectra. The tails occur as
background in layers close to another layer that contains a larger
amount of the element of interest. In general, the problem is
much more serious for ERD analysis than for RBS analysis
because of the shallow angles of incidence and exit with respect
to the target surface. Small secondary scattering angles are
sufficient to significantly change the ion direction with respect to
the target normal. Even an inversion of the ion beam direction
can easily occ:ur. Such particles can penetrate a layer of the
sample a second time and thus have a probability to scatter the
elements at lower energies. Although steeper angles can be
produced, making a tail at larger energies and thus in the
dircction of smaller depths with respect to a layer (Carstanjen,
1998), plural scattering develops mainly low-energy tails and
thus tails into deeper layers. The effects have been investigated
by several researchers. A detailed experiment is presented in Fig.
5.6(a) where the hydrogen content ofa mecallic bulk sample with
surface hydrogen coverage, was analyzed by range-foil ERD
using a 4 MeV He beam [Fig. 5.6(a)J. Sub-pan-per-rnillion
sensitivity was possible only when the hydrogen surfaee peak
had been reduced by sputter treatment. The experiment showed
that the effect is more serious for heavy matrixes because plural
on
Z2
scattering scales by 2 . The integrated content of an element m

the low-energy tail varies over orders of magnitudes depending
on the projectile mass, its energy, the ERD geometry, and the
mean nuc1eBJ charge of the target The effect can be reduced by
enhancing the projectile velocity (energy per nucleon), because
the plural scattering cross section scales as (Z l fE)2 according to
Fa, which
gives the lower limit in sensitivity. However, the background
116
Rutherford cross sections. Steeper angles u and

reduce background from plural scattering effects.
P also help to
energy is sufficiendy below the Coulomb barrier for o.ucL ear

reactions. The He signals are assigned to 12C(llC,ai~e
reactions occuning as secondary reactions of carbon recoils. in
the diamond sample (Dollinger el al., 1996). It was found that
many more He ions were detected for a 130 MeV Ni beam than
for a 60 MeV Ni beam (Do!1inger et al., 1996). This effcct from
secondary reactions is an issue for heavy-ion ERD only.
Secondary scattering o~ recoil ions or projectiles in or in

front of the detector also produces background. For example, the
scattering of hydrogen by the large number of projectiles in a
hydrogen-containing stopper foil produces background in
hydrogen analysis, It can be significantly reduced by using
nominally hydrogen-free stopper foil such as aluminium
(Trip.thi et al., 2004), but surface coverage .nd impurity

concentrations of hydrogen in the stopper foil remain a problem
if no care is taken. Heavy-ion ERD also suffers from secondary
scattering effects in the detector (i.e., in the AE part of a AE-E
detector), where a secondary large scattering event of a recoil
ion will make a slower ion and an additional recoil, depositing
additional encrgy in the 6.E part. This higher 6.E is interpreted as
a recoil ion of larger Z value and is incorrectly assigned to
another element if, accidentally, a correct ~E relationship is met
for another element. Such a contribution is seen in Fig. 5.23. 1t
shows a 6.E-E spectrum (raw data) of recoil ions from a
diamond sample detected at 40 0 from 130 MeV Ni projectiles
(Dollinger et al., ] 996). The secondary reaction and the pilelJp
contribution make a large background for all elements bet\.veen
carbon and magnesium, as well as some background for the
heavier elements through triple accidental coincidence pileup. In
this case, the background amounts to 10-40 ppm for the analysis
of nitrogen or oxygen in carbon and is mainly due to the high
count rate in the AE part of the detector (-7 kHz in this case). On
the other hand, high count rates are necessary to obtain enough
data with limited beam time to obtain high sensitivity. Sensitivity
for elements lighter than the matrix atoms (carbon) or heavier__ ..
than about twice that amount is limited by the number of total
counts only. Thus, better than 1 at.-ppm sensitivitihas been' _.
reported for ERD count rates of 2 kHz at less than 3-h
FIG. 5.23. 6.E-Ercs spectrum of ERD analysis of a diamQnd

sample using a 130 MeV 5~i beam. In addition to the main
carbon signal, the spectrum shows some type of oxygen
contamination and a large background consisting of pulse pileup
mainly from the 6.E detector and some secondary reactions. In
particular, He ions are detected from IlC(nC.a)2~e secondarynuclear reactions, whereas thc corresponding Ne signals are
measurement time (Dollinger et al., 1996).
hidden in the plleup signals (Dollinger et al., 1996).
To reduce background, one can introduce redundant

measurement of the elemental information of the recoil ion. This
might be several 6.E measurements or a combination of several
types of elemental analYSis, such as the combination of TOF-E
with 6.E-E or a combination of a magnetic spectrograph with a
multistrip 6.E measurement. The latter was realized in an
experiment to detect nitrogen in carbon at sub-part-per-million
concentrations (Bergmaier et al., 1996).
r2$ldual energy
! a.u.]
Beams of heavy ions, such as CI, Cu, I, or Au, as used for

heavy ion ERD can have specific energies up to several
MeV{nucl. The energies are normally chosen safely below the
Coulomb barrier for reactions with any type of target atoms.
Light recoil atoms, however, scattered off these heavy ions
(especially hydrogen isotopes, carbon, nitrogen, and oxygen)
have nearly twice the velocity of the projectiles. Thus, they have
nearly four times the specific energy of thc projectile and,
because they have a much lower nuclear charge than the
projectiles, exceed the Coulomb barrier with nearly all elements
in the target. Thus, these recoil ions can induce secondary
nuclear reactions at a reaction cross section 0"22. which will
generatc new elements reaching the detector. Consequently, a
background of the produced elements will be detected, resulting
in an apparent concentration n y . This concentration of secondary
reactions can be calculatcd assuming a constant center-of-mass
5.6_2.4 Secondary nuclear reactions

Artificial signals can also be produced from secondary
nuclear reactions, producing elements that are not present in the
target sample. The effect is visible in Fig. 5.23 as well. The 6.EE plot shows signals for He although there is no He inside the
diamond layers. In addition, the energy of many' of the He ions
exceeds the energy from ERD kinematics. Direct nuclear
reactions of the 130 MeV Ni beam are not expected because the
117
,...
Chapter 5
(eM) reaction cross section
O'n
differently regarding ion-bearn-induced alterations of elemental

profiles, especially in terms of the effects of electronic
interactions of ions with solids.
for all scattering angles as
(Dollinger et al., 1996)
T= _ E"", -21n T~
= f
n
y
0"" COS
8n:(dEIdx)2~'.E.",
E"",
T~
E"",
1&t01
1
1~.oo
'1e-41
The factor f arises because of the transformation from the eM to

the laboratory system and is f Il::j 1.7 for a symmetric system.
,-
(dEldx)2-jo2.,m is the energy loss of the mean matrix atom
1~
within the matrix at the Coulomb barrier energy Elhll given in the
,.....
laboratory frame through Eq. (5.6) but with R. = I.3fm(A"n +

Ajll3). Tdetmio (T&r,mu) is the minimum (maximum) energy of
carbon recoils detected. Obviously, ny depends mainly on the
numerator of the second fraction in Eq. (5.40), which is
determined by the factor x:. = T ma/Eth22. A plot of the
denominator against the factor x is shown in Fig. 5.24. For
example. the analysis of pure carbon (graphite or diamond
samples) by 130 MeV (60 MeV) Ni ions exhibited an apparent
helium concentration of 3 x 10-4 (5 x 10-9) from '2Ce2C,cc.i~e
secondary nuclear reactions, in quite good agreement with Eq.
(5.14). The Ne ions from the same reactions were also detected
but were somewhat obscured by pulse pileup effects from carbon
recoil ions of the matrix. The contribution of secondary nuclear
reactions decreases quickly with increasing nuclear charge 22 of
the matrix atoms because the Coulomb barrier EIh.12 scales with
2
Z2 and the stopping force also scales close to Zz.2 at high recoilion energies whereas the nuclear reaction cross section 0'21 scales
only with A2213 for nonresonant nuclear reactions.
1&-OS
10-06
10
;x:
= T"",,' Ec,..
FIG. 5.24. Plot of the function f(x) ~ x -lIx-2Iog(x).
The most appropriate way to compare the various ion beam

analysis methods with respect to irradiation damage is to define
damage numbers that give the orders of magnitude of the various
effects. A common way to define these damage numbers is to
take them as the numbers of atoms that are displaced (Ddispl) or
ionized (Deice) before one of those atoms is detected (Reichart et
aI., 2002). These numbers allow for the comparison of
completely different IDA techniques or different ion beam and
detection conditions with respect to their irradiation damage
potential because it gives the number of displaced or ionized
atoms relative to the number of events detected for a certain
element.
5.6.3 Irradiation damage

Irradiation damage is generally one of the main limits of ion
beam analysis because any material is, to some extent, sensitive
to irradiation damage effects from energetic ion beams. At the
least, ion displacement collisions will lead to elemental
reorganization and atomic losses (i.e., sputtering). Also,
electronic interactions through ionization processes and thennal
spike or Coulomb explosion phenomena can lead to similar
effects ifthe material is sensitive, for ex:.ampie, organic materials
and several insulator materials. In view of the sensitivity and
accuracy of ion beam analysis, the only information of interest is
whether a desired measurement can be performed before the
elemental profiles have been significantly altered by ion
bombardment. the analysis of light elements is often more
critical than the analysis of heavier elements because several
: light species can be mobile if once activated by any process.
This is especially true for hydrogen isotopes. but is also an issue
for other species such as lithium, nitrogen, oxygen,
~ydrocaroons, and the noble gases. Various materials behave
The number of displaced atoms per detected event, DdispU is

given by the ratio of the total displacement cross section, Odispb
to the total detection croSs section, (do-JcIn)Ml. In the case of
ERD assuming a pure Rutherford detection cro~s section (Eq.
(5.4)], the damage number for displacement reactions, Ddispb is
approximately given by (Dollinger etal., 1996)
cr~1
1t
(dcr/ctn)Ml
Ml
E_ cos' "'(I. In E,,)

-,
E"",
(5.41)
2E....
with E,~ = [4M"M/(M, + Mu'JE, being the maximum energy

that can be transferred to the recoil atom of type i. E.:!ispI is the
minimum displacement energy that has to be transferred to an
lIS
are present in a solid volume-of 1 .u.m1. Thus, conventional EED

analysis is not possible without displacing each atom several
times. The technique can be pushed as far as possible by -the
following options: (1) Use ERD reactions with enhanced D<:mRutherford cross sections, (2) use reduced ion energies (and
thereby lower Ei,max) when Rutherford cross sectioos are
applicable, (3) enlarge the solid angle of detection, or (4) enla,-ge
the beam diameter. To obtain a sensitivity below 10 12 aticm:2 in
ERD of high depth resolution, it is recommended that the sam.ple
be scanned in order to enlarge the effective beam diameter up to
square-centimeter dimensions and that a solid angle of detection
in the range of at least several millisteradians be used.
atom of the solid to create a displaced atom. It 1S about three

times the binding energy of the atom, 50 E displ =: 25 eV can be
used as a rough estimate. For hydrogen,- Edi.!pl = 10 eV is a better
estimate. The cross section for atomic displacements, CTdispi, is
evaluated according to the Kmchin-Pease model (Kinchin and
Pease, 1955), which also accounts for secondary nuclear
collision cascades and uses Rutherford scattering cross sections,
as follows:
(5.42)
If elemental loss is induced by electronic interactions, the

s
6
varies in order of magnitude between 10 and 10 in
situation is even worse. The damage cross section due to
the case of standard ERD conditions (r.p < 45[>, E" betv.teen 0.2
ionizations can be approximated in the same way a.s that due to
MeV/nudeon and 2 MeV/nucleon), and the ratio nltJD. is
nuclear
collisions by replacing the displacement cross section,
4
normally around 10 3_10 for ERD experiments when large solid
O"dispb in Eqs. (5.41) and (5.42) by an ionization cross section,
angles of detection are used. It follows that displacement
aion. An estimate for Ginn can also be obtained from Eq. (5A2)
numbers, Ddispb are in the range of 109_10 11 displaced atoms per
when considering Rutherford scattering on electrons by setting ~
detected count. In the case that the ERD cross section is
~ 1 and M;;;' In, ~ Mpl1836 aod by using the appropriate values
enhanced by nonRutherford conoibutions, D disp1 is reduced by a
for Edispl ;::; 5 eV and Ec,max according to two-body kinematics
factor of crRU1~/crERD. Because, at worst, all of the displaced atoms
[Eq. (5.1)].
might be lost from the region of interest by diffusion once
activated (especially possible in the case of hydrogen isotopes),
The number of ionizations, DI.OI\' until one count is obtained
one should displace no more than 10% of the atomic species of
is about 1000 times the number of di.splaced atoms, DdispJ, for
interest from the sample in order to keep close to the original
high-energy beams Atoms (e g , hydrogen) can diffuse out of the
concentrations. On the other hand, one should detect a minimum
analysis area once activated by IOnizahon, although it has been
of 10 counts to obtain a meaningful value for the elemental
reported that double Iomzations are required to lose an hydrogen
concentration. Both requirements imply a minimum number of
atom (Maree et aL, 1996a), Heavy-ion beams have the potential
lOll_lOB atoms in the layer of interest being analrz~_~ ?y. ~__ to ,create jon. tracks _thrOl!gh collective effects such as thermal
2
ERD experiment. Using-an- illuinimi.ted-area of 1 mm , a
spikes (Toulemonde et aL, irioO) where'rnaierial islost m.Ore'-:'::---'
13
15
2
minimum elemental coverage of 10 . _10 at.lcm has to be
effectively than through'pure collision cascades (Toulemonde elpresent for an accurate analysis to be performed. These numbers
ai., 2003). A comparison of hydrogen loss induced by various
set a limit on the minimum concentration that can safely be
light- and heavy-ion beams used' for ERD analysis of various
19
2
analyzed. For a thick layer of 10 atomslcm , one has a safe
hydrocarbon materials is provided in Table 5.6. The calculated
damage cross sections O"displ and aiolJ are listed, together with
situation down to part-permillion concentrations, but thinner
layers ~ll be probl.ematic. ERD with .cl~se to ato~c.-layer dep~
experimentally determined release cross sections are!. Whereas
resolUTIon suffers III the end from hffilted S~[lSltLV1ty. E~ m
the measured release cross sections arel are close to a.fupl When
microbearn analysis, where the number of avallable atoms m the
protons are used as projectiles, they can approach aim in the case
layer of ions scales inversely with the area covered by the beam,
of heavy-ion beams.
is even more limited. Thus, for instance, only about lOll atoms
Ei,ma./Edispl
119
. '3Z
Chapter 5
Table 5.6. Hydrogen release cross sections, OJ'l:io derived from different experimental data compared with the calculated damage cross
sections for di'iplacements, O"di5J1' and ionizations, Gioo. according to Eq. (5.42). The threshold energy was set to Edisp ~ 10 eV for
displacements and EiOll ~ 5 eV for ionizations.
arel (
Experiment (reference)
experiment)
(mb)
O"disp
(mb) crimi (mb)
5 x 10 10
3 X IOu
7):' 1013
2 x 10 14
1011
2)(
1011
2 x to'
I x 107
1 )( 107
3)( 108
3)( 108
I x 10'
7 x 10'
3 X 1013
2 X 1010
3 X 1014
4 x 10 10
4)( lOll
2 x 10"
5 x lOll
1 x 10 10
3 x 10'
6 x 10'
6 X 10 12
5)( 1011
4)( 10 10
6 x 10'
17 MeV p on Mylar (Reichart et al., 2004)

2.55 MeV p on Mylar (Sj6land ef al., 1997)
3 MeV p on Mylar (Hacke el 01., 1993)
2 MeV He on porph)Tin (Maree et 01., 1996a)
2 MeV He on porphyrin (de Jong et aI., 1997)
4 MeV He on porphyrin (de Jong et aI., 1997)
7.6 MeV He on porphyrin (de Jong et aI., 1997)
10 MeV Si on porphyrin (Marie et 01., 1996b)
46 MeV Cu on porphyrin (Maree et of., 1996a)
60 MeV Ag on a-C:H (Mareeel al., 1996b)
7 MeVN on a-C:H (Baumann et 01., 1987)
3.5 MeV N on a-C:H (Baumann et 01.,1987)
1.6 MeV Nt on a-C:H (Baumann et al., 1987)
2.15 MeV AI on a-C:H (Baumann et 01., 1987)
87 MeV Br on SiON/Si (Walker et 01., 1998)
46 MeV Cu on a-Si:C:H (U1lersma el aL, 2000)
17 MeV F on diamond (Scllschop et al. 1980)
10 MeV Si on W/Au (Grambole et 01.,1999)
8 X 107
2)( 10'
5)( lO"
5 X 10"
2
1011
5 x lOll
7 x 10 12
j x
2 x 1011
2 X 10 13
3 x 10 10
3 l( 10 10
5 X 1011
5 x 1011
3 x 1011
2 X lO"
1 x 10 13
l( 1013.
1 x 1011
I x 10 14
4 x 10 10
4 x 10 10
3 x 10 9
2 x 1011
1013
7 x 1013
6 x 10 12
3 X 10 13
energy are used (Maree et aI., 1996a). This is because ERD

cross sections scale approximately as a a:: (z.,Mpffip)2 for ion
beams that are much heavier than the elements of interest and
because lighter-ion beams also show some electronic effects in
elemental release. In addition, the contribution of elemental loss
through electronic stopping is strongly reduced when heavy-ion
beams are used at moderate energies where the electronic
stopping is below the energy-loss maximum.
Obviously. protons have. by far. the lowest damage cross

sections. Thus. coincident detection of the protons from protonproton scattering is by far the best option for hydrogen analysis
with regard to minimizing damage potential. This is not only
because it entails the lowest ionization effects. but also because
the elastic scattering cross section is enhanced by a factor of
about JOO in high-energy proton-proton scattering because of
strong intemctions and because huge solid angles of detection
(e.g., >2 sr) can be used. In particular, microbeam analysis with
beam spots on the order of 1 JlID or less is possible only when
using proton-proton scattering. (See Chapter 9 for a detailed
discussion of proton-proton scattering.) The disa.dvantage of
proton-proton scattering that the depth resolution is poor and
that the scattered protons have to transmit the sample, thus
necessitating small sample thicknesses. Therefore, for high depth
resolution in the nanometer range, one must employ ERD
analysis using non-proton beams.
Because the elemental loss strongly depends on the type of

material and/or element investigated. one has to measure the
observed elemental concentration versus time of measurement in
order to exclude a Significant loss of the element during any type
of ERD experiment. An example of such a time-dependent
elemental fraction is shown in Fig. 5.25, where deuterium
content was measured in a deuterated amorphous carbon (a-C:p)
sample by 210 MeV I and by 80 MeV Au ions. At moderate loss
(in the case of 80 MeV Au ions), one can easily extrapolate the
concentration to zero time of measurement. An overly strong
loss of hydrogen (210 MeV 1701 ions in this case) with nonlinear
behavior masks the true concentration. and thus, the ultimate
limit of ERD analysis is reached.
At first, the use of heavy-ion beams for stlndard ERD

analysis seems to be worse than use of lighter ones because of
the huge damage cross sections. A comparison of the damage
nwnbers for various experiments, however, shows an even
slightly better sensitivity when hcavy projectiles of moderate
120
~&
.......
.
1 . :T
.Doi!--''-t1O.--'-~2l;;D~-l30.--'--40!n'-!50;;-'-'/'OO.-'-~7"'O--'-'.!80.--'-~80''..J'06
'fluence [1010 ions cm~2]
~.01
0.1
1~
100
fluence [10111 ions cm4 J
FIG. 5.25. Plot of the deuterium content of a-deute rated amorphous carbon (a-C:D) layer as a function of heavy-ion beam fluence using
80 MeV Au and 210 MeV I beams (Behrisch etal., 2000): (a) linear scale, (b) logarithmic scale offluenee.
Baumann, H., Rupp, Th_, Bcthge, K., Koidl, P_, and Wild, Ch_
(1987), Eur_ Mater. Res. Soc. Can! Proc. 17,343_
Behrisch, R., Mayer, M., Jacob, W., Assmann, W., Dollinger,
G_, Bergmaier, A., Kreissig, U., Friedrich, M_, Sun, G.Y.,
Hildebrandt, D., Akbi, M., Sclmeider, W., SchleuBner, D.,
REFERENCES
Andersen, H.H., Besenbacher, F., Loftager, P., and Moller, W.

(1980), Plrys. Rev. A21, 1891.
Angerer, R., Brunner, D., Freudenberg, F., Ambacher, 0.,

Stutzmann, M., Hopler, R., Metzger, T., Born, E., Dollinger,
G., Bergmaier, A, Karsch, S., and Komer, HJ. (1997),
Appl. Phys_ Lett_ 71, 1504.
Arai, E., Funaki, H_, Katayama, M" Oguri, Y, and Shimizu, K.
Knapp, W., and Edelmann, Ch. (2000), J. Nucl. '-Water. 281,
42.
Bergmaier, A, Dollinger, G_, and Frey, CM. (1995), Nuel_
Instrum, Methods B99, 488,
Bergmaier, A, Dollinger, G" Faestenruum, T" Frey, CM"
(1992), Nuel. Instrum. Methods B64, 296,
Arnold Bi!<, W,M" deLaat, CTAM" and Habraken, F,H,P,M, " ____ Ferguson, M" Guttier, H, (1996), Diamond Relat. Mater_ 5,(1992), Nuel. Instrum, Methods B64, 832,
995
.
.
.
. .". _ .
:
Arnold Bi!<, W,M" and Habraken,F,H,P,M_ (1993), Rep. Prog. - Bergmaier,-A_, Dollmgei, G" and'tiey, C,M. (1998),- Nuel. .
Instrum, Methods BI36-138, 638,
Plrys, 56, 859,
Arstila, K" Sajavaara, T" and Keinonen, j, (2001), Nucl.
Bogdanovic Radovic, I., and Benka, O. (2001), Nucl. Instrum,
Instrum, Methods B174, 163.
Methods B174, 25_
Assmann, W., Hartlmg, P., Huber, H., Staat, P., Steffens, H. t and
Bohne, W., Rohrich, J., and Roschert, G. (1998), Nuc1. Instrum.
Steinhausen, Ch, (1994), Nuel. Instrum_ Methods B89, 13 L
Methods B139, 219.
Assmann, W., and Maier, HJ. Nucl. Instrum. Methods A362
Bohr, N. (1948), Mat. Fys: Medd. K Dan. Vidensk. Selsk. 8, 18.
(1995) 143,
Bergesen, p" Behrisch, R" and Scherzer, RM,U, (1982), Appl,
Assmann, W., Davies, lA., Dollinger, G., Forster, J.S., Huber,
Phys. A 27, 183.
H" Reichelt, Th" and Siegele, R. (1996), Nuel, Instrum,
Browning, J,F" Langley, R.A., Doyle, B,L., Bank", J,C" and
Methods B1l8, 242_
Wampler, W,R. (2004), Nuel, instrum, Methods B219--220,
Barbour, J.e., and Doyle, B.L (1995), in Handbook of Modern
317.
ion Beam Materials Analysis (Tesmer, J.R., and Nastasi,
Busch, F., Pfeffer, W., Kohlmeyer, B., Scbiill, D., and Pihlhofer,
F_ (1980), Nuc/, Instrum, Methods 171, 7L
M" eds.), Materials Research Society, Pittsburgh, PA, p_

112,
Barrad.. , N,P" Jeynes, C, and Webb, R,P, (1997), App/' Phys,
Lett, 71291.
Barrad.. , N,P" Marriott, P,K" Jeynes, C, and Webb, R.P,
(1998), Nuel. Insrrum. Methods B136-138, 1157,
Carstaojen, H,D, (1998), Nucl. Instrum, Methods B136-138,

1183,
de Jong, M_P., Maas, AJ_H" van Ijzendoom, Ll., Klein, S,S"
and de Voigt, MJ,A, (1997), J. Appl, Plrys, 82, 1058.
Dollinger, G" Faestermann, 1., and Maier-Komor, p, (1992),
Nuel. Instrum, Methods B79, 422.
121
-e-
Chapter 5
Kim, C.S., Kim, S.K., and Choi, H.D. (1999), N"el. Instrum.
Methods B155, 229.
Kimura, K., and Nakajima, K. (2001), Appl. Phys. Leu. 78,
Dollinger, G. (1993), Nuel. Instrum. Methods B79, 513.

Dollinger, G., Boulouedninc, M., Bergmaier, A, Faestermann,
T., and Frey, C.M. (1996), Nucl. Instrum. Merhods B118,
291.
Dollinger, G., Frey, C.M., Bergmaier, A., Faestennann, T.
(1998.), Nucl. Instrum. Methods B136-138, 603.
Dollinger, G., Frey, C.M., Bergmaier, A., and Faestermann, T.
(1998b), Europhy,. Lett. 42,25.
Dollinger, G., Bergmaier, . A., Goergens, L., Neumaier, P.,
VandeNorSt, W., and Jakschik, S. (2004), Nuel. Instrum
Methods B219-220, 333-343.
Dollinger, G. (2009), to be published.
Doolittle, L.R. (1985), Nucl. Instrum. Methods B9, 344.
Doolittle, L.R (1986), Nuel. Instrum. Methods B15, 227.
Doyle, B.L., and Peerey, P.S. (1979), Appl. Phys. Lett. 34, 811.
Bcuyer, J.L., Bras,ard, C., Cardinal, c., Chabbal, J., Deschenes,
L., Labrie, J.P., Terreault, B., Martel, J.G., and St.-Jacques,
R (1976), 1. Appl. Phys. 47, 381.
Enders, Th., Rilli, M., and Carstanjen, H.D. (1992), Nuel.
Instrum. Methods B64, 817.
Fischer, R, Mayer, M., von def Linden, W., and Dose, V.
(1997), Phys. Rev. E55, 666.
Frisch, O. British Atomic Energy Report, BR-49, 1944.
Gebauer, B., Fink, D., Goppelt, P., Wilpert, M., and Wilpert, Th.
(1990), N"el. Instrum. Methods B50, 159.
Giangrandi, S., Arstila. K... Brijs, B., Sajavaara, T.) Vantonune,
A., and Vandervorst, W. (2007), Nuel. Instrum. Methods
B261,512.
Gossett, C.R (1986), Nuel. Instrom. Method, B15, 481.
Grambole, D., Herrmann, F., Behrisch, R., and Hauffe, W.
(1999), N"cI. Instrom. Methods B158, 647.
Green, P.F. (1994), private communication (Sandia National
Laboratories, Albuquerque, NM).
Groleau, R, Gujrathi, S.C., and Martin, l-P. (1983), Nuc!.
Instrom. Methods 218, 11.
Gujrathi, S.C., and Bultena, S. (1992), Nuel. Instrum. Methods
B64,789.
Hacke, M., Meijer, J., Stephan, A., Bukow, H.H., H6fert, M.,
and Rolfs, C. (1993), Z. Phy,. A346, 309.
Hofsa.ss, H.C., Parikh, N.R, Swanson, ML., and Chu, W.K..
(1990), Nuel. Instrom. Methods B45, 151.
Hofsass, H.C., Parikh, N.R, Swanson, M.L., and Chu, W.K.
(1991), Nuel. Instrum. Methods B58, 49.
Ion Beam Analysis Nuclear Data Library (IBANDL):
http://www-nds.iaea.orglibandV.
Jeynes, C., Barradas, N.P., Marriott, P.K.., Boudreault, G.,
Jenkin, M., Wendler, E., and Webb, RP. (2003), 1. Phys.
D: Appl. Phys. 36, R97.
Keay, J.C., and Ingram, D.C. (2004), Nuc!. Instrum. Methods
B222,690.
Kellock, A.J., and B.glin, J.E.E. (1993), Nuel. Instrum. Methods
B79,493.
1679.
Kinchin, H.G., and Pease, RS. (1955), Rep. Prog. Phys. 18, 1.
Koschar, P., Kroneberger, K, Clouvas, A., Burkhard, M.,
Meckbach, W., Heil, 0., Kemm1er. 1., Rothard, H.
Groeneveld, K.O., Schramm, R., and Betz, H.-D. (1989),
Phys. Rev. A40, 3632.
K6tai, E. (1994), Nuel. Instrum. Methods B85, 588.
Kotai, E. (1997),AlP Con! Proe. 392, 631.
Kottler, C., DobeIi, M., Glaus, F., and Suter, M. (2006), Nuel.
Kruse, 0., and Carstanjen, H.D. (1994), Nuc!. Instrum. Methods
t
B89,191.
Kruse, O. (1995), PhD thesis,'Universitllt Stuttgart Stuttgart,
Germany.
Liechtenstein, V.Kh., Ivkova, T.M., Olshanski, .0.,
Feigenbaum, L, Di Nardo, R., and DobeIi, M. (1997), Nuc!.
Instrum. Methods A397, 140.
L6filer, M., Scheerer. H.-I., and Vonach, H. (1973), Nuc!.
Instrum. Methods Ill, 1.
Maree, C.RM., Kleinpenning, A.. Vredenberg, AM., and
Habraken, F.H.P.M. (1996a), Nuel. Instrom. Methods B118,
301.
Maree, C.H.M., Vredenberg, A.M., and Habraken, F.H.P.M.
(1996b), Mater. Chem. Phys. 46, 198.
Mayer, M. (1999), AlP Con! Proe. 475, 54.
Mayer, M. (2002), Nuel. Instrum. Methods B194, 177.
Mills, PJ., Green, P .E., Pa1mstrom, C.l, Mayer, lW., and
Kramer, EJ. (1984), Appl. Phys. Lett. 45,958.
Nagata, S., Yamaguchi, S., Fujino, Y., Hori, Y., Sugiyama, N.,
and Kamada, K. (1985), Nuel. [nstrum. Methods B6, 533.
Neumaier, P., Dollinger, G., Bergmaier, A., Genchev, I.,
Goergens. L., Fischer, R., Ronning, c., and Hofs!ss, H.
(2001), Nuc!. Instrum. Methods B183, 48.
Nolte. H., Assmann, W., Huber, H., Karamian, S.A., and
Mie,kes, H.D. (1998), Nuel. [nstrum. Methods B13&-138,
587.
Faszti, F., Szilagyi, E., and K6tai, E. (1991), Nuc!. Instrom.
Methods B54, 507.
Pus., P., Raubala, E., Alanlm, T., and Raisnen, J. (2003), J.
Appl. Phys. 93, 6370.
Pusa, P., Rauh.la, E., Gurbich, A., and Nurmela, A. (2004),
Nuc!. Instrom. Methods B222, 686.
QuilicI, V., Abe~ F., and Schott, M. (1993), Nuel. Instrom.
Methods B83, 47.
Raubala, E. (1984),J. Appl. Phys. 56,3324.
Reichart; P., Dollinger, G., Datzmann, G., Hauptner, A., Korner,
H.-J., and Scbmebner, O. (2002), Nuc!. Instrum. Methods
B197, 134.
122
Reichart, P., Dollinger, G., Bergmaier, A., Datzmann, G.,

Toulemonde, M., Assmann, W., Trautmann, C., GrUner, F.,
Hauptner, A., Komer, H.-I., and Kriieken, R. (2004), Nuci.
Mieskes, H.D., Kueal, H., and Wang, Z.G. (2003), Nucl.
iostrum. Methods B219, 980
Instrum. Method, B212, 346.
Renfordt, R.E., Noll, H., and Sapotta, K (1981), Nucl. IllStrum.
Tripatbi, A., Kruse, 0., and Cantanjen, H.D. (2004), Nud.
Methods 185, 157.
instrum. Methods B219-220, 435.
Saarilahti, I., and Raahala, E. (1992), NucJ. Instrum. Methods
Ullersma, E.H.C.,UUersma, P., and Habraken, F.H.P.M. (2000),
B64,734.
Phys. Rev. B61, 10133.
Sajavaara, T., Arstila, K., Laakso, A., and Keinonen, I. (2000),
von der Linden, W. (1995), Appl. Phys. A60, 155.
Nuci. Instrum. Methods BI61-163, 235.
Wakefield, B., Naqib, I.M., Harper, R.P., Hall, I., and Christy,
Sehiettekatte, F., Chewier, A., Chewier, N., Plantier, A., Ross,
A. (1970), Phys. Lett. B31, 56.
G.G. (1996), Nucl. Instrum. Methods B118, 307.
Walker, S.R., Davies, 1.A., Foster, 1.S., Wallace. S.G., and
Schiettekatte, F. (2008), Nucl. Instrum. Methods B266, 1880.
Kockelkoren, A.C. (1998), Nuci. instrum. Methods B136Sehmeimer, 0., Dollinger, G., Frey, C.M., Bergmaier, A., and
138,707.
Karseh, S. (1998), Nucl. Instrum. Methods B145, 261.
Wang, Y.Q., (2004), Nue!. Instrum. Methods B219-220, 115.
Sellschop, I.P.F., Madiba, c.c.P., and Annegam, H.I. (1980),
Wetteland, C.I., Maggiore, C.J., Tesmer, I.R., He, X-M., and
Nuc/.Iostrum Methods 168, 529.
Lee, D-H. (1998), "Magic" Energies for Detecting Light
Sigmund, P., and Winterbon, K.B. (1974), Nue!. inslrum.
Elements with Resonant Alpha Particle Backscattering,
Methods 119, 541.
report LA-UR-98-4867, Los Alamos National Laboratory,
SigrnaCalc: http://WVIW-nds.iaea.orgisigmacalcl#.
Los Alamos, NM.
Siketic, Z., Bogdanovic Radovit, I., Muta, H., and JakSi6, M.
WhitlOW, H.I., Possnert, G., and Petersson, C.S. (1987), Nucl.
(2005), Nud. InSIr. and Melh. B229, 180
Sjoland, K.A., Kristiansson, P., Elfman, M., Malmqvist, K.G.,
Wielunski, M., Mayer, M., Behrisch, R., Roth, J., and Scherzer,
Pallon, I., Utui, R.J., and Yang, C. (1997), Nucl. Instrum.
B.M.V. (1997), N"d. Instrum. Methods B122, 113.
Methods B124, 639
Yamaguchi, S., Fujino, Y., Nagata, S., Kaneko, H., Hashimato,
Stopping and Range of Ions in Matter (SRlM):
K., and Hirabayashi, M. (1983), Nucl. instrum. Methods
http://www.srim.org/.
218,598.
Stoquert, J.P., Guillaume, G., Hage-Ali, M., Grob, J.1., Ganter,
Yang, Q., and O'Connor, D.l (1991), Nucl. Instrom. Methods
c., and Siffert, P. (1989), Nucl.lnstrum. Methods B44, 184.
B61,149.
Szilagyi, E., Paszti, F., and Amsel, G. (1995), Nucl. Instrum.
Yu, R., and Gustaf,son, T. (1986), Surf Sci. 177, L987.
Methods BlOO, 103.
Zebelmann, A.M., Meyer, W.G., Halbach, K., Poskanzer, A.M.,
Szilagyi. E. (2000), Nucl.lnstrum. Methods BI61-163, 37.
Sextro, R.G., Gabor, G., and Landis, D.A. (1977), Nuc/.
Thomas, lB., Fallavier, M.; Ramdane, D., Chevarier, N., '3;d. ------lnstrum Methods 141, 439. . .. --- ..
.
Chevarier, A. (1983), Nucl. Instrum. Me/hods 218,125.- Ziegler, I.F., Biersack, I.P., and Littmark,' U. (1985), The
Toulernonde, M., Dufour, C., Meftah, A., and Paumier, E.
Stopping and Range of Ions in Solids, Pergamon Press, New
(2000), Nucl.lnstrum. Methods BI66-167, 903.
York, vol. 1.
123
-_._-_._------
:,-"
CHAPTER
NUCLEAR REACTION ANALYSIS:

PARTICLE- PARTICLE REACTIONS
G. VizkeJethy
Sandia National Laboratories, Albuquerque, New Mexico, USA
Contributors: G. Battistig, I. Vickridge, M. Nastasi, J. R. Tesmer, and Y. Q. Wang
CONTENTS
6.1 GENERAL PRINCIPLES AND EQUIPMENT ..............................................................127
6.2 GEOMETRY AND METHODS FOR Fll..TERING UNWANTED PARTICLES ........128
6.2.1 Absorber foil technique ................................................................129 ~_. ___ .._
6.2.2
Thin detector technique .....................................................................................129
6.2.3
Electrostatic or magnetic deflection (also electrostatic detector and magnetic

spectrometer) technique .....................................................................................129
6.2.4
Time-of-flight (fOF) technique ...........................................................................129
6.2.5
Coincidence tecllnique ....................................................................................... 129
6.3 MEASUREMENT METHODS ......................................................................................130

6.3.1 Overall near-surface contents ............................................................................. 130
6.3.2 Nonresonant deptb profiling ............................................................................. 130
6.3.3 Resonant deptll profiling ................................................................................... 133
6.4 USE OF STANDARDS .................................................................................................. 134
6.5 USEFUL REACTIONS ................................................................................................... 136
6.5.1
Proton-induced reactions .................................................................................... 136

~i(p,a)'He and ~i(p,'He)"He, Q = 4.02 MeV ........................................... 136
6.5.1.2 'Li(p,a)'He, Q = 17.347 MeV ..................................................................... 136
6.5.1.3 "Be(p,a)~I, Qa = 2.125 MeV; 9Be(P,d)8Be(2a), Qd = 0.5592 MeV .............. 137
6.5.1.4 lOB(p,a)'Be, Q = 1.147 MeV ....................................................................... 137
6.5.1.1
125
Chapter 6
6.5.1.5 l1 B(p,a)8Be, Q = 8.582 MeV ...................................................................... 137

6.5.1.6 lSN(p,a)lZC, Q = 4.966 MeV ...................................................................... 137
6.5.1.7 180 (p,a/ sN, Q = 3.9804 MeV .................................................................... 137
6.5.1.8 "F(p,a)160 , Q = 8.1137 MeV ..................................................................... 138
6.5.2 Deuteron-induced reactions .............................................................................. 139
6.5.2.1 D(d,p)T, Q = 4.033 MeV ............................................................................ 139
6.5.2.2 'He(d,pj'He and 'He(d,aj'H, Q = 18.352 MeV ........................................ 139
6.5.2.3 'Li(d,aj'He, Q = 22.374 MeV .................................................................... 139
6.5.2.4 'Be(d,tj8Be, Q = 4.496 Me V ....................................................................... 139
6.5.2.5 'Be(d,aj'Li, Q = 7.153 MeV ....................................................................... 139
6.5.2.6 ,oB(d,aj8Be, Q = 17.818 MeV .................................................................... 139
6.5.2.7 12C(d,pjI3C, Q = 2.722 MeV ...................................................................... 139
6.5.2.8 13C(d,pj14C, Q = 5.951 MeV ...................................................................... 140
6.5.2.9 "N(d,pjISN, Q = 8.610 MeV ...................................................................... 140
6.5.2.10 14N(d,aj"C, Q = 13.574 MeV ..................................................................... 140
6.s.2.U 1SN(d,pj''N, Q = 0.266 MeV ...................................................................... 140
6.5.2.12 ISN(d,aj 13 C, Q = 7.687 MeV ....................................................................... 140
6.5.2.13 16 0 (d,pj 17 0, Q = 1.917 MeV ...................................................................... 140
6.5.2.14 16 0 (d,aj 14N, Q = 3.11 MeV ........................................................................ 140
6.5.2.15 180(d,pj"O, Q = 1.732 MeV ...................................................................... 141
6.5.2.16 l80(d,aj16N, Q = 4.247 MeV ...................................................................... 141
6.5.2.17 "F(d,pj"F, Q = 4.374 MeV ....................................................................... 141
17
6.5.2.18 "F(d,aj 0, Q = 10.031 MeV ..................................................................... 141
6.5.3 'He-induced reactions ........................................................................................... 141
6.5.3.1 Dc'He,p)'He and DeHe,aj'U, Q = 18.352 MeV ....................................... 141
6.5.3.2 'Li('Ue,pj 8Be, Q = 16.7863 MeV ............................................................... 141
6.5.3.3 , Be('H e,pjII B, Q = 10.3219 MeV .............................................................. 141
6.5.3.4 'BceUe,aj8Be, Q = 18.9124 MeV .............................................................. 141
6.5.3.5 ,oBe('Ue,pj 12 C, Q = 19.6931 MeV ........................................................;.... 142
6.5.3.6 l1 Bc'He.pj I3 C, Q = 13.1853 MeV; "B~'Ue,dj"C, Q = 0.4635 MeV ........ 142
6.5.3.7 12 q 'Ue,pj"N, Q = 4.7789 MeV; nq He,aj"C, Q = 1.8563 MeV ......... 142
6.5.3.8 160 ('Ue,pj 18F, Q = 2.0321 MeV; "O(,Ue,a)ISO , Q = 4.9139 MeV ........ 142
6.5.4 Alpha-induced reactions ....................................................................................... 142
6.5.4.1 ,oB(a,p)13C, Q = 4.0626 MeV; II B (a, pj 1C , Q = 0.7839 MeV ................. 142
6;5.4.2 1'N(a,p)17 0, Q = -1.1914 MeV ................................................................... 143
6.5.4.3 19F(a,p)22Ne, Q = 1.6746 MeV ................................................................... 143
REFERENCES ..................................................................................................................... 143
126
"
Nuclear Reaction Analysis: Particle Partide Reaclions
6.1 GENERAL PRlNCIPLESAND EQUIPMENT

Nuclear reaction analysis (NRA). like Rutherford
backscattering spectrometry (RBS), exploits the interactions of
charged particles with atoIPic nuclei for analysis. However, in
contrast to RES, where the particle-nucleus collision is elastic
and mediated by the Coulomb force, in NRA, probing beam
species and energies are chosen such that the incident particle
overcomes the Coulomb repulsion barrier between itself and the
target nucleus. The interaction is then dominated by the nuclear
forces and can only be adequately described by nuclear quantum
theory. As the incident particle first crOsses the "Coulomb
barrier" (see Appendix 8) and starts to feel the nuclear forces,
the cross section deviates from the Rutherford cross section,
giving rise to non-Rutherford elastic scattering (see Chapter 4
and Appendix 7).
electronics with elastically scattered primary beam, i"t: is

necessary to filter this large flux. Several methods for achie"",ing
this are described in this chapter; however, the most cOIU:non
way is to place a thin fihn (often Mylar) in front oftbedet"'ctor
so that only the energetic particles produced by exothermic
nuclear reactions can pass through. The energy stragg;ling
induced by such a filter is now the limit to the achievable dEpth
resolution. To maximize count rates, it thus becomes accept<lhle
to use larger solid angles for detection because the degradation
of depth resolution due to kinematic broadening and the po. ()TCr
intrinsic resolution of large-area detectors no longer dominate.
For RBS, a solid angle of a few millisteradians is usll.ally
used, whereas for NRA, solid angles of 50-100 msr are
typically employed.
It is useftll to use larger incident beam currents to compensate
for the lower cross sections.
Filtering techniques are necessary to eliminate the large flux
of elastically scattered primary beam.
Particles produced by the nuclear reactions can have !D uch
higher energies (up to 14 MeV) than those commonly
encountered in RBS, thus necessitating use of detectors with
large sensitive depths in these cases.
With smaller distances of closest approach during the

collision (decreasing nuclear charges, increasing energies),
internal nuclear energies come into play, so the interaction is no
longer elastic; mass is not conserved. This internal energy
available for the reaction is denoted by Q. The cross sections
now differ strongly from the Rutherford cross section, generally
being much smaller and varying in complicated ways with
incident particle energy and detection angle. For incident
energies used in ion beam analysis (0.1-5 MeV), the Coulomb
repulsion is most easily overcome when a low-Z projectile (Z =
1. 2) is incident on a low-Z target nucleus (Z = 1-15). With
some exceptions, nuclear reactions start to have significant cross
sections from about 500 keY for incident protons and deuterons
on targets up to about aluminium. NRA is complementary to
RBS in that it is most sensitive for detecting light elements on a
heavy substrate.
Because the nuclear reactions are isotope-specific, with

negligible values for higher-Z targets, we can also observe that
there is no natural background from the high-Z components
of the target, and
~~ offers many possibilities for stable isotopic tracing
studies (e.g., Ir'2H, 12If3C, I4II'N, 161180, etc.).
Usually; there is- no: inalyticai fonn of the nuclear cross
sections, so that reliable experimental design and data analysis
depends largely on the availability of measured cross sections in
the energy range and at angles of interest. Some of these data. are
listed in Appendix 11. General information about nUclear
reaction
cross
section
data
is
available
at
http://www.nndc.bnl.gov/ in the National Nuclear Data Center
(NNDC), Brookhaven National Laboratory, UptOD, NY. A more
useful comprehensive database of nuclear reaction cross sections
of specific interest for NRA is maintained by the International
Atomic Energy Agency [Ion Beam Analysis Nuclear Data
Library (lBANDL): www-nds.iaea.orglibandl]. lBANDL
incorporated thc SigmaBase and NRABase (archival copies are
available
at
bttp:llwww.rn1a.kfki.huisigmabasel
and
bttp:/lwww.mfa.kfki.buisigmabaseJprograrnsinrabase2.htmI. respectively) and now incorporates many more cross sections,
includIng many of those mitially compiled by Jarjis (1979). It is
continually updated, and the cross sections can be downloaded
in formals suitable for use with simulation programs and for
plotting.
An excellent discW;sion of the fund~ent~l physics involved

in NRA can be found in Feldman and Mayer (1986), and several
older but very useful review articles discuss NRA in detail
(Amsel and Lanford, 1984; Bird, 1980; Deconninck. 1978;
Mayer and Rimini, 1977; Ziegler. 1975). Detailed reviews of
rnA of the first 10 elements have been published in the
Proceedings of Ihe Third International Conference on Chemical
Analysis (1991). The basic concepts of relevant nuclear theory

are presented in Chapter 3 of this handbook, and the kinematics
of these inelastic collisions is presented in Appendix. 4. Many
developments and applications of NRA can be found in the
proceedings ofillA conlerences (1973-2009).
The equipment required for NRA is essentially the same as
that for RES: particle accelerator, scattering chamber, high or
ultrahigh vacuum, surface barrier detector, standard nuclear
electronics, and multich~el analyzer.
The main differences arise from the large difference in
magnitude between the Rutherford and nuclear reaction cross
sections. To avoid overloading the detector and acquisition
For a given incident particle interacting with a given

nucleus, several different nuclear reactions might be possible,
127
Chapter 6
and for a given nuclear reaction, some of the available energy,
Q, might remain as internal energy, leaving the product nucleus

in an excited state, We thus refer to the Q value of the reaction in
which the product nucleus ends up in the ground state as 00.
whereas Ql and Q2. refer to reactions in which the product
nucleus ends up in the first and second excited states, and so on.
The particles that are detected are denoted similarly; for
example, we refer to IO(d,Po)"O (Q, ~ 1.917 MeV) when the
17
0 atom ends up in the ground state, and 1'0(d,PI) 170 (QI ~
1.046 MeV) when the 170 atom ends up in the first nuclear
excited state. Nuclear level diagrams showing nuclear structure
and properties of nuclear levels can be found for the target
IIIlISses A ~ 3 (Ajzenberg-Selove, 1987a), A ~ 5-10 (AjzenbergSelove, 1988), A ~ II and 12 (Ajzenberg-Selove, 1990), A ~ 1315 (Ajzenberg-Selove, 1986a), A ~ 16 and 17 (AjzenbergSelove, 1986b), A ~ 18-20 (Ajzenberg-Selove, 1987b), and A ~
21-44 (Endt and van der Leun, 1978). Many of these data are
available on the TUNL Web site (httpllwww.tunI.duke
.edulnucldata), and the 00 values can readily be calculated from
the most up-to-date known masses of the nuclides using Qcalc
(http://www.nodc.bnl.gov/qcalc). One such example is shown in
Fig. 6.1, where the IIB(p,a)sBe reaction has frequently bcen
applied for analysis of trace amounts ofboran.
depending on the circumstances. The scattering geometry is the

same as that in RBS, except that, in NRA, some filtering
equipment is placed in front of the detector. Figure 6.2 shows a
typicall\'RA geometry with an absorber.
Emlx), E(Emlx),Q,8)
flO. 6.2. Typical scauering geometry used in NRA

experiments. Eo is the energy of the incident beam; Ein(x) is the
energy of the iuciden. particle at depth x; E[E;,(x),Q,8] is the
energy of the reaction product in the sample at a depth :x; Q is
the Q value of the rcaction (see Appendix 10); Eom(x) is the
energy of the reaction product that was produced at a depth x in
the sample, after traveling out of the sample but prior to passing
through the absorber foil; and E.r,,(x) the energy of the reaction
product after passing through the absorber foil. The angles a.in
and Clout are the angles ofthe incident beam and of the detector,
respectively, relative to the surface normal of the sample surface,
and E> is the traditional scattering angle, that is, the angle
between the incident and outcoming beams (8 in this chapter is
identical to 8 in Fig. 4.1 of Chapter 4.)
,~,.~ __-=;;-/F,\
3.
6.2
GEOMETRY
AND
METHODS
FOR
FILTERING UNWANTED PARTICLES
The geometry in NRA can be either backward or forward,
'"
The energy of the particles leaving the target is givCll by

"'(OO5a.... )
FIG. 6.1. Schematic illustration of proton-on-II B nuclear
E~,(x) = E[E" (x),Q,El]-
reaction channels (Moncoffre 1992).
S"",(E) <lx,
(6.1)
NRA spectra are sometimes difficult to interpret because

peaks of different particles (or the same particles with different
energies) can overlap. The choice of optimum experimental
parameters (incident beam energy, detection angle, absorber
thickness, or other filtering techniques) is an essential part of the
design of an NRA measurement In some cases, narrow
resonances exist [e.g . in 180(p,n)l~ that allow high-resolution
depth profiling in the same way as for the (p,y) resonances
treated in Chapter 7.
If(cos<<;.)
E;,(x)=E o -
S.,(E)dx,
(6.2)
where x is the dcpth at which the nuclear reaction occurs; Eo is

the energy of the incident beam; Q is the energy released by the
reaction; S(E) is the stopping power of the incident ion and the
outgoing reaction product; ain and <loul are the angles of the
incident beam and of the detector, respectively, relative to the
surface Donnal of the sample surface; E[Ein(x),Q,0] is the
energy of reaction product (see Chapter 3 and Appendix 4); and
128
Nuclear Reaction Analysis: Particle-Particle Reactions
e is the traditional scattering angle, that is, the angle between

the incident and outgoing beams (e in this chapter is identical to
in Fig. 4.1 of Chapter 4). For calculation of the stopping
powers, see Chapter 2 and Appendix 3. For Q-value
calculations, please refer to Appendix 4. Q values for deuteroninduced reactions can also be found in Appendix 10.
Five filtering methods can be used, as discussed in the

following sections.
6.2.1 Absorber foil technique
The simplest method to stop the scattered beam is to place an
absorber foil in front of the detector. Usually, Mylar is used for
the foil because it is pinhole-free, although aluminum foil is
sometimes used. The energy of the particles after passing
through the foil is
E,,,(x) = Eoo,{x)-
'j S,b,CE) dx.
(6.3 )
where X~bs and Sabs(E) are the thickness and the stopping power
of the absorber foil. respectively. In Appendix 10, the energies
of protons and alpha particles after having passed through
frequently used thicknesses of Mylar are tabulated for deuteroninduced reactions from 2H to 19F. Readers can use SRIM
deflection distance depends on the energy, mass, and chat:-ge of

the particle at a constant electrostatic or magnetic field, SQ that
the different particles are separated in space. This method gives
better depth resolution than use of an absorber, but because of its
complicated measurement setup, it is rarely used, and we will
not discuss it in detail here (Moller, 1978; MoUer et ol., ] 977;
Chaturvedi et al., 1990; Hirvonen and Luecke, 1978).
6.2.4 Time-of-flight (fOF) technique
This method is frequently used by nuclear physicists "When
they have to distinguish between a large number of different
particles that are generated simultaneously. The techniqlle is
based on the simultaneous measurement of the energy and
velocity of the detected particle. From these two values, the
mass of the particle can be calculated. The velocity is measured
by measuring the time elapsed between the detection () i the
particle in two sequential detectors that are placed at a fixed
distance from each other. This method gives the same depth
resolution as described in Section 6.2.2 for the thin detector
technique, but it requires sophisticated electronics. The full rOF
measurement also requires a multi-dimensional multichannel
analyzer, which can be found in many of today's ion beam
laboratories. A simplified version of TOF is obtained when the
measured velocity is not stored with the energy, but is used in a
coincidence system to reject signals from particles with
velocities below a certain threshold level (SokoloY et al., 1989).
Another disadvantage of this method is that, although the

(http://srim.org) to calculate more precise energy distrubutions
multichannel analyzer does not detect aU of the low-energy
for any kind of particles passing through various absorber foils.
particles, those particles reach the detector, and this huge
Because of the different stopping powers of alpha particles and
scattered-ion flux can damage or kill the detector. This techoique
is used in ERDA (elastic recoil detection analysis), where the
protons, the interfering a. peaks can usually be filtered out by the
appropriate ftlter thickness. The disadvantage of this r:nethod is
difference between the flux of the forward-scattered alphas and
that it results in poor depth resolution because. of large ~nergy . ~ the recoiled hydrogen isotopes is not so large. Additional details
straggling in the absorber foil.
concerning TOF can be found in Cbapters 5 and 16.
6.2.2 Thin detector technique
This technique is used when proton and a. peaks overlap and
the Ct spectrum contains more information than the proton
spectrum. The opposite of this situation is when the a. particles
are filtered out using absorber foil. Because of the difference in
the stopping powers between a particles and protom, filtering
can be achieved with a weIl-<:hosen detector thickness: whereas
the a. particles fully stop in the sensitive thickness of the
detector, the protons lose only a fraction of their energy. This
measurement can be made using either thin fully depleted dEldx
detectors, which are especially designed to measure energy loss
and are quite expensive, or low-resistivity detectors with low
. bias VOltage (Abel et al., 1990). A detailed study bas been
carried out by Arosel et al. (1992). This technique is usually
combined with the absorber foil technique.
6.2.5 Coincidence technique

The detection of bad.scattered particles can also be
suppressed by using a coincidence arrangemenL The two
reaction products are detected in different detectors in
coincidence (i.e., the event is registered only if both detectorn
give a signal). One of the detectors sbould be placed at an angle
that is unreachable for one of the reaction products (this angle
can be calculated from the kinematics). In this case, only the
other detector can detect the particle, which cannot reach the
first detector because of the coincidence requirement. The
residual background is due only to the accidental coincidences.
Applications of this technique are described in Pretorius and
Peisach (1978) and Wielunski and MBller (1990). This techoique
is limited to transmission geometries (thin foil) because of the
conservation of momentum. As in the TOF technique, all of the
particles (scattered and reaction products) hit the detector, which
means that the detector can be destroyed by the large ion current.
6.2.3 Electrostatic or magnetic deflection (also electrostatic

detector and magnetic spectrometer) technique
These techniques are based on the effects of the magnetic
and electrostatic fields on energetic charged particles. The
129
Chapter 6
the depth resolution. The depth scale can be calculated frOID Eqs.
(6.1), (6.2), and (6.3) hy calculating the detected energy of the
reaction product for different depth values. An alternate
tcchnique in obtaining the depth scale is to use an expression
similar to that used in ion hackscattering [see Eq. (4.22) in
Chopter 4]
6.3 MEASUREMENT METHODS

6.3.1 OveraU near-surface content!;
]f the cross section of a given nucleus changes slowly with
energy in the vicinity of Eo. the absolute number of nuclei per
square centimeter can be determined in thin layers independently
of the concentration profile and of the other components of the
target. Let Eo be the energy of the incident beam and .6E be the
energy loss in the thin layer to he measured. If <f(E) z cr(Eo) for
Eo > E > E - .6.., then the number of counts in the peak. area A
from the thin layer is
A = Q,Ocr(Eo)Nt ,
cos am
.
(6.5)
where N is the atomic density and []m is the nuclear reaction
stopping cross-section factor, defined as
(6.4)
(6.6)
where Qc is the collected charge, Q is the solid angle. and Nt is
the number of nuclei per square centimeter. If all of the
The reaction factor n, is taken from Eq. (6.8) (also see Table
6.1) and weights the energy loss of the incident particle in the
same fashion a.s the kinematic factor weights the backscattering
stopping cross-section factor (E]in in Eq. (4.23) of Chapter 4.
Values of ein and ~I can be calculated assuming either a
surface-energy or mean-energy approximation. When either the
standard or unknown contains more than one element, the Bragg
rule must he applied toEq. (6.6). (See EXAMPLE 6.2.)
quantities in Eq. (6.4) were known precisely, the absolute value
of the nuclei per square centimeter would be easy to calculate.
Unfortunately, usually, none of the above values (except the

number of counts) is known precisely enough. and the
measurement arrangemenl is changed between measurements.
(For issues on absolute current measurement problems, see
Chapter 15.) As the peak. area A depends only 00 the number of
nuclei per square centimeter in different matrixes, the
measurement can be compared to a well-known standard, that is,
Nt ~ (AlA,.,,,,,)Nt,.,.,,,,. (See EXAMPLE 6. I.) The accuracy of
the measurement is then detennined only by the statistics of the
measurement and the 'accuracy of the standard The problem of
standards and their use is discussed later in this chapter (Section
6.4). The thickness of these thin layers is on the order of
thousands of angstroms.
Although proper calculation of the depth resolution requires

convolution of the various energy spread functions, a bandy
approximation can be obtained by assuming that all energy
spread distributions are Gaussian, in which case the depth
resolution in depth x is given by
Table 6.1. Values of fitting parameters (IE and i3e: in the expression
E3 = ae: El +i3E forB = 165".
FItting
Q
Reaction
(MeV)
14N(d,o:)12C
13.579
'''c(d.p) 14C
5.947
range
~.
(MeV)
(MeV)
0.122:t 0.013
9.615:t 0.016
0.8 - 1.4
0.005
5.339 :t 0.006
0.6 - 2.0
0.571
:t
1SN(d,o.)13C
7.683
0.273 0.010
5.487:t 0.014
0.6 - 2.0
9Se(d,p)10Be
4.585
0,456
j:
0.006
3.923 :t O.OOS
0.6 2.0
'2c(d,p)''c
2.719
0.591
::!:
0.003
2.384 0.004
0.6 - 2.0
''O(d.~!'70
1.919
0.685 " 0.002
1.722 0.003
0.6 - 2.0
where ~ is the detector resolution; ~ is the geometrical

energy spread; .6.Ems is the energy spread due to multiple
scattering; LlEabs is the energy straggling in the absorber foil; crill
and O'QU! are the straggling for the incident ion and the reaction
product, respectively; and EJ and E2 are the energies of the
incident ion and the reaction products, respectively, at depth x.
Among these terms, the multiple .scattering distribution is the
one that deviates most from the Gaussian fonn. The only
difference from the RBS depth resolution is that the kinematic
factor k is replaced by oE I/oE2 . For some reactions, this
derivative is larger than unity, and the stopping power of the
incident ion is also large because of the usually low incident
energy. The depth resolution (if no absorber is used) can,
therefore, be very good near the surface.
6.3.2 Nonresonant depth profiling

As the thickness of the sample becomes larger, the peak of the
reaction product becomes broader. The peak shape' is the
convolution of the concentration profile with the cross section and
130
.~
Nuclear Reaction Analysis: Particle-Partide Reactions
For very light nuclei at backward angles, this derivative is

smaller than zero. Therefore, the energy of the reaction product
increases with increasing depth. Often encountered examples
wh3ere
situation can arise are D(d,p)T, DeHe,afUe, and
D( He,p) He. In such cases (and if the stopping power of the
reaction product decreases with increasing energy or changes
only very little), the contribution of the deep-lying nuclei
extends the spectrum toward higher energies. At a certain depth,
the depth resolution Ixcomes infmite, as the denominator
becomes zero. The program DEPTH has been developed as a
comprehensive computer code for calculating depth resolution,
including all of the above terms (Szilagyi eJ a/., 1995)
th!S
For a slowly changing cross section, the depth profIle can be

calculated in the same way as in the case of RBS, and the
Rutherford cross section should be replaced by the actual cross
section. Where the cross section changes significantly along the
measured thickness, the depth profiles can best be calculated by
using simulation programs to fit the measured spedrum
(SENRAS by Vizkelethy, 1990; SlMNRA by Mayer, 1999;
NDF by Barradas, 1997). A comprehensive intercomparison of
the main simulation programs is given in Barradas et al. (2007).
EXAMPLE 6.1. Consider a SiOiSi sample where the thickness
of the oxide layer is unknown. We want to measure this
thickness by measuring the oxygen content of the sample. The
best reaction for this measureLlent is 160(d,Pl)1iO. This reaction
bas a plateau below 900 keY down to 800 keY, as shown in fig. 6.3
(between the arrows).
In this example, we use a deuteron beam with Eo = 834 keY and

detect the protons at
= 135. To suppress the backscattered
deuterons, we use a 12-JilD-thick Mylar foil in front of the
detector (aU of the deuterons with energies below 900 keV are
stopped by this foil). This reaction produces two groups of
protons, Po (Q = 1.919 MeV) and p, (Q = 1.048 MeV). The
energies of the protons can be calculated from the general
kinematic factor
.
EJ '12 = B -+ (B' + C)'12 -.... (u EEI +"PE )"' '

where
(6.8)
The variables M" M2, M), M.., Eh El , and e in Eq. (6.8) are
defmed in Fig. A4.1 in Appendix 4. The second form ofE3 given
in Eq. (6.8) is a linear approximation of the ftrst form. The'
variables aE and ~E are specific to the reaction under study and
depend aD the detection angle 0. Values for aE. ~E. and Q are
2iven in Table 6.1 for some .specific reactions.
In this example, Mit M z, M 3 , and !>v4 are the masses of the
deuteron, 160, proton, and ]70, respectively. The energy E1 is
defined as being equal to Eo, and E) is the final proton energy,
which, in this example, is 2.352 and 1.556 MeV for Po and Ph
respectively. The energies measured after the protons have
passed through the Mylar foil can be calculated from the data in
Appendix 10. Although those energies were calculated at e =
150, the energy of the protons in our case differs only slightly
(energies in Appendix 10 are 2.36 and 1.58 MeV), so we can
assume that the protons will lose the same amount of energy.
Data from Appendix 10 show that pc loses 0.24 MeV in the
absorber foil and PI loses 0.35 MeV. Therefore, the energies of
the protons after they have passed through the absorber foil are
2.312 III1d 1.206 MeV.
Let us estimate the max.imum thickness of Si0 2 that can be
measured at this energy. The stopping power of deuterons in
Si0 2 changes slowly between 800 and 900 keY, so we can use
an average, which is 7.6 eV/A. The energy range that we can use
is approximately 30 keY, so the maximum thickness is
x ... =
S(E)
keV
7.6 ,VIA
~ 4000 A
In Fig. 6.4, the measured spectrum (Vizkelethy, 1990) can be

seen. There are two ex.tra peaks that require some explanation.
The peak at cbanoels (i.e., high energies) comes from the
12C(d,p)13C reaction. During ion bombardment in a typical rnA
vacuwn chamber (10-6 Torr), a hydrocarbon layer is always
deposited on the surface. The small broad peak between the Po
and Pl peaks comes from the silicon. The proton from D(d,p)T
reaction is not clearly seen in this measurement because its
~roton energy is almost the same as the proton energy from
'O(d,PO)"0. A later work done by (Jiang et ai., 2003) on a
similar sample at a scattering angle of 150 did detect protons
from the D(d,p)T reaction, where deuterium was implanted into
the sample as a result of the deuterium bombardment during the
analysis.
The next step is to calculate the total number of oxygen
atoms from Eq. (6.4). UsuallYI the absolute value of the cross
section and the solid angle of the detector are not known
precisely, so standards must be used. The best oxygen standard
is Ta20s made by anodic oxidation (Amsel 1978a). In this
example, we use an 80 V thick Ta20s standard in which the
oxygen content is 6.69 x 101i \60 atomslcm2 . We do the same
measurement with the standard and calculate the peak. integrals.
Because we collect the same charge in both cases, the ox.ygen
(M,M,E, )'12
'"
cos
M,+M,
and
~= 30
M,Q+E,(M, -M,)
M,+M,
131
-ct
Chapter 6
content is obtained by the comparison of the peak integrals [see
Eq. (6.13)]
Nt
:xo"
=N
"
----. "O(d,pol"'o
"O(d.P)"O.6-15O
'"o{doP,J"O
~=6.69xlOll~::4.54xlont
31875
t,..,o, YTa,O,
which corr~onds to 1000 AofSio,.

EXAMPLE 6.2. Consider the application of the 12C(d.Po)IlC
reaction in analyzing the amount of carbon in a diamondlikccarbon (DLC) coating that contains a substantial amount of
hydrogen. We will take am ~ 0" and """ - 15'. From Appendix 10,
we see that a relatively flat plateau e:tists in the region between 0.9
and 1.0 MeV for the 12C(d,Po)13C reaction. The proton energy
following a reaction between 12C and a 1 MeV deuteron at the
sample swiace is 3.0 MeV as calculated using Eq. (6.S) and Table
6.1, Values of em for the 1 MeV deuteron in carbon and hydrogen
are 7.19 x lOtS eV cm2/atom and 1.99 x lOIS eV cm2!atom,
respectively. Values of ~ for the 3.0 MeV proton in carbon and
hydrogen are 2.15 x IOU eV cm2!atom and 0.47 x 1015 eV
2
cm /atom, respectively. From Table 6.1, aE = 0.59. From these data,
we can calculate the nuclear reaction stopping cross-section factor
for each of the elements, carbon and hydrogen, in the sample for
the 12C(d,Po)IJC reaction as
[J~ ~(O.59X7.17+ ~:~~}W1S

=6.45xlO- 15
- -- - -- -
o ---.--------.---.----.------IncIcIe<It ham Ene'lll' (keVl
17
FIG. 6.3. The "O(d,p) 0 reaction for both Po and PI reaction

products. This reaction has a plateau below 900 keV down to SOO
keV (from Gurbich and Molodtsov, 2004).
~OO~~rr~~~"~~'T~~rr
,~~
t'o(d.P11170
2000
J!l
1500
5o
u 1000
(evem')
I-
.11
"o(d.PoInO
Catom
lISf(d.p)ZlSI
and
100
[cJ~ =(O.59xI.99+ ~:;~}W-IS

=1.22xlO- I '
500
200
300
~)Ue
400
500
600
Ch8Elnels
FIG. 6.4. The measured spectrum from 834 keY deuterons on a
(evem')
SiO,lSi sample (from Vizkelethy, 1990). There are two extra

peaks. which require some explanation. The peak at hi~h
Hatom
channels (i.e., high energies) comes from the 12c(d,p)1 C

reaction; during ion bombardment a hydrocarbon layer is always
deposited on the surface. The small broad peak between the po
and PI peaks enmes from the "Si(d,p)"Si reaction. The protons
from the D(d,p)T reaction are not identified here because they
have almost the same energy as those from 16 0 (d,Po) 170
reaction. A later work by Jiang e/ al. in 2003 did show a
deuterium peak because deuterium had been implanted into the
sample as a result of the deuterium bombardment.
Asswniog that the OLe coating is composed of 40 at. %

hydrogen and 60 at. % carbon, the Bragg rule gives
[E);.-H = O.4(e]~r + 0.6[];,
= 4.36xlO-" (ev em!)
atom
132
..
6.3.3 Resonant depth profiling

As Eq. (6.7) shows, the depth resolution ofNRA is generally
vel)' poor because of the absorber. The depth resolution can be
improved if resonances exist. The method is very similar to the
resonance profiling by (p,y) resonances described in Chapter 7.
The detector signal goes through a windowed single-channel
analyzer, where the energy window is set to include only the
reaction products from the resonance reaction, to a multiscaler.
The energy can be changed either manually or by an automatic
energy scanning system. An indicative depth scale is given by
Eo - E~scmanc~
S(E)/ cos u;. '
(6.9)
where ErnollDn~c is the resonance energy. The energy-loss integral

can be calculated by using the average energy approximation,
that is, E = (Eo - Ensonaru:~)/2, which is similar to the RBS average
energy approximation (Chapter 4), as the energy change is
usually small. Under the major simplifying assumption that all
contributions to the energy spread are Gaussian, the depth
resolution is given approximately by
(6.10)
whcrc
is the width of the resonance, as( is the energy
straggling of the incident particle, llli btam is the energy spread of

the incident beam (inc1uding the Doppler effect due to thermal
motion of the target atoms), and 6.Ems is the energy spread due to
small-angle multiple scattering. The resonance line shape is

generally of Breit-Wigner form, and when this is a significant
contribution to energy resolution, its width should not, strictly
speaking, be added in quadrature to the other contributions. The
beam energy straggling in the near-surface region is also far
from Gaussian, highly asymmetric, and rapidly varying with
depth. For such circumstances, the only way to calculate depth
resolution properly is to convolve the various contributions and
apply some criterion to decide when two separated deltafunction concentration profiles can be said to be resolved.
Nevertheless, Eq. (6.10) is useful as a frrst-Qrder idea of energy
resolution. As Eq. (6.10) shows, the depth resolution close to the
surface can be improved by tilting the sample, but this is at the
cost of degraded depth resolution deeper in the target due to the
increased energy straggling and the effects of small-angle
multiple scattering
extract the information from the spectrum, one should

deconvolve it with the depth-depcndcnt resolution function~ (See
EXAMPLE 6.3.) The method for the calcu1atioo of the'
concentration profIle in this case is the same as that presented for
(p,y) reactions in Chapter 7. When the resonance width and
beam energy spread are sufficiently narrow (on the order- of a
few times the average energy lost to an electron in a single ionatom collision), it is necessary to treat the individual ion-atom
collisions in a statistical way. The theory of narrow resonance
profiling (instrumentation function, straggling close to the
surface, simulation of resonance spectra, and so on) with. such
narrow resonances is discussed in detail in Maure! (1980).
Maurcl e/ al. (1982), and Vickridge (1990). Vickridge (J990)
also describes SPACES, a PC-based computer program tha.l uses
this statistical approach to calculate excitation curves
corr,esponding to arbitrary concentration profiles.
EXAMPLE 6.3. Here, we give an example in which the
ISO(p,a)15N 629 keY resonance was used to determine oxygen
exchange in high-I" superconductor material (Cheang Wong et
aI., 1992). First, let us estimate the depth resolution of the method
and calculate the depth scale using zero tilt aIlgle. The
composition of the sample was measured by RBS, Y1BaIU1.A.
Because the high Te: layer was relatively thin, we can use an average
stopping power of 10.87 eVIA, calculated using the Bragg rule
(see Chapter 2). Equation (6.9) shows the depth scale to be 1
keV/tO.S7 eV/A = 92 AlkeV. Because the resonance is 2.1 keY
wide, the depth resolution on the surface is approximately 200 A
(the energy spread was assumed negligible relative to the
resonance width). At first an 180 reference sample was measured
to calibrate the accelerator energy more precisely. The spectrum
lS sho'NO in Fig. 6.5. This resonance sits on a continuous
background; therefore, the spectrum does not go to zero even
when the resonance is not inside the ISQ-containing layer. This
explains the counts below the resonance energy.
The spectrum of the annealed high~T~ sample is shown in Fig.

6.6. A peak can be found !it the resonance energy that indicates
an enriched surface layer. To calculate the depth profile from the
spectrum, we must deconvolute the spectrum. The a. yield
(counts per channel) at beam energy Eo is given by

n
E;
j=i
Y(E) = K~>j~ f(Epxj>E) cr(E) dE
(6.11)
where K is a scaling factor; c; is the concentration of ISO in the jth

layer, cr(E) is the reaction cross section; and
The measured excitation curve represents the convolution of

the concentration profile and the energy spread of the detection
system, which is the same as the denominator in Eq. (6.10). To
(6.12)
133
."
Chapte'6
is the energy distribution of the incoming beam with initial

energy Et in depth Xj. with ECEi.x) the beam energy and aj the
straggling at depth J1} Several methods are available to carry out
the deconvolution, as described in Chapter 7. In this case, we
used the Levenberg-Marquardt method (press et al. 1986) to fit
the concentration proftle, X' = [Y;~
- Y(E)]' .
The best fit
was achieved by using a 150-A-thick surface layer enriched to

30%, followed by constant volume enrichment to 3.5%.
"
'"g 10
'ii
"E
,
g
Requirements for the measurement apparatus are as follows:

Current measurement must be reproducible. which
implies a bigh-stability current integrator, efficient
secondary-electron suppression, and very good vacuum
to avoid charge exchange after the beam analyzing
magnet (for He or heavier beams).
Energy calibration must be precise and reproducible,
Detection angle (if the detector is not in a permanent
position) must be reproducible.
Dead-time correction and pileup rejection must be
accurate,
I-
Proton energy [keY]
FIG. 6.5. The "O(p,a)l'N 629 keY resonance from a Ta,O,

reference sample. This resonance sits on a continuous
background, and therefore, the spectrum does not go to zero
outside the reaction energy,
These requirements are the same as those required by highprecision RES, so they are not discussed further here (see
Chapters 4 and 16). The frequently used standards are LiF, CaF,
(Dieumegard elal., 1980), NbN (Maurel and Amsel, 1983), TaN
(Amsel and David, 1969), and Ta,O, (Pringle,1972; Phillips and
Pringle, 1976; Amsel el 01., 1978" 1978b). The measurements
of the standard and the unknown sample should be performed
under the same conditions. namely, the same solid angle, same
incident energy, same absorber, and also the same uncertainty of
the current meastliement. In this case Cif the above require!l1ents
are ful.filled), the total amount of the measured nuclei in the
unknown sample is given by
..
.,. '.O(p. ~'''N on yaaCuO
,
1.0
'
t
,
'tj
Requirements for good standards are as follows:

The thickness of the layer should be small enough not
to cause significant change in the cross section.
The lateral uniformity over the beam area must be high.
The standard should be amorphous to avoid channelling
effects.
The targets should have long-term stability in air, in
vacuum, and under ion bombardment.
The preparation of the target should be highly
reproducible .
"
C)
6.4 USE OF STANDARDS

As mentioned previously, there are uncertainties in the
calculation of the absolute overall surface can lent of an element.
The best way to overcome these uncertainties is to use reference
targets. A detailed discussion of the preparation and use of
reference targets can be found in Arose! and Davies (1983).
O.S ~
A Q~f
Nt = Nt ----,~'A
Q
'-
u...~u....~1..L.~'..L.'~..L.~.L..~.L....~~~...
0.0
620
625
630
635
640
64S
Proton energy [keY]
650
"'
655
(6.13)
where A and Am are the peak areas in the spectra; Qc and Q~f are
the collected charge; and Nt and Ntm are the total amounts of the
measured nuclei for the unknown sample and the reference
standard, respectively. Equation' (6.13) assumes that the
thicknesses of both the reference sample and the unknown
sample are small enough to use an energyindependent cross
section and that their respective peaks can be fully integrated. If
FIG. 6.6. The l'O(p,a)l'N 629 keY resonance spectrum from a

hlgh-Tc sample following an 180 oxygen-exchange anneal. A
peak can be found at the resonance energy that indicates an
enriched surface layer,
134
the same amount of charge is collected in both cases, the Q~f IQ<;
Table 6.2. Cross sectIon ratios.
tem can be eliminated.

Isotope
Another way to calculate the absolute overall surface content

of an element is to IDeasW"e the cross section ratios of different
elements at a given energy. With this method, only one precise
reference target is needed. (See EXAMPLE 6.4.) lbe disadvantage
of the method is that one is very unlikely to find a wide energy
region over which the cross sections of different elements exhibit
plateaus. This limits the measurable thickness. For "0, "N, and
at 0 ~ ISO and at 972 keY, the values are given in Table 62
(Amsel and Davies, 1983). This table permits unknown standards
to be calibrated to a known standard (e.g., high-precision Ta20S
reference standard, which can be prepared easily by anodic
oxidation).
(d.p,)
UN
"c
In the case of the l80(p,a)15N resonance, the use of a .

standard is also beneficial. It helps to dctennine the precise
energy of the resonance and the energy spread of the beam. See
Chapters 7 and 8 for more about the use of standards with
resonances.
~XAMPLE 6.4. COrlsider a sample contaminated on the surface by
oxygen and carbon. Because it is quite difficult to prepare a carbon
standard, a Ta20S reference sample will be used. A good oxygen
standard is too thick for us to consider the oxygen cross section to
be con~tant at 972 keY, where the ratio O'( 12C)/cr I6 0) is known.
rherefore, we first measure the oxygen content of the sample with
the 16 0 (d,PI)I'O reaction at 850 keY. The cross section is quite
constant down to 800 keY. Comparing the counts with a Ta10s
reference standard from Eq. (6.13), we can calculate the oxygen
content. we then change the energy to 972 keY and measure the
oxygen again and the carbon with the 12C(d,PoiJc reaction. Because
we want to measure only surface contamination, we assume that
both O'C 12 C) and 0'(160) can be considered constanl We already
know the oxygen content from the 850 ke V measurement, and using
data from Table 6.2, c{L2C)/ae 60) = 1.91, we can easily calculate
the carbon content as
Cross sectIon ratio (c,,/col
Nuclear reaction
(d,Po)
0.032
(d.p,+P2)
0.078
(d,P4)
O.OSS
(d,ps)
0.45
(d,Pe)
0.057
(d.",,)
O.OOSO
(d,.a,)
0.119
(d,pp)
1.91
Consider the binary-clement unknOVlO sample composed of

elements A and B with the composition AlBy, where x and y are
the atomic fractions of A and B respectively (i.e., x + y = 1).
Because we are interested in the amount of element A in our
unknown. we choose a standard containing element A and
assume that the standard is composed of the elements A and C
with the composition AzC vr , where z+w-l. following the
development of Eq. (4.28), the surface heights of element A in
f
our two samples, in the unknown, H~G' and in the standard or

known, H~o' are given by
cr A (Eo)QuQul>u x
[l~ casCl"
(6.15)
and- _. __. _ - '
where O'ACEo) is the nuclear reaction cross section for element A;
(6.14)
[I: l::" and [I: l:c are the nuclear reaction stopping crosssection factors for element A in the unknown and the known,
respectively [as defined in Eq. (6.6)]; and the variables n, Q,
where Ntc and Nto are the absolute contents and Yc and Yo are the
counts of carbon and oxygen, respectively.
and fi; correspond to the detector solid angle, the number of

incident ions, and the energy width per channel, respectively.
We can rewrite the second part of Eq. (6.15) in terms of (JA(Eo)
and substitute into the first part of Eq. (6.15) to obtain
In many instances, the standard is a bulk specimen, and the

unknown sample can be either a bulk sample or a thin film. For
the situation where the standard and the unknown are both bulk,
a surface-energy approximation to the nuclear reaction analysis
can be used to obtain the composition at the surface.
H U [el
AB
x=~~Rz
Ac
HKA,
(6.16)
[el A.,
135
-1
Chapter 6
wbere R is given by
can be found in Appendix 11 and a more extensive list in

mANDL (www-nds.iaea.orglibandl), so only the general
features are mentioned here.
(6.17)
6.5.1 Proton-induced reactions
are
When the unknown and the knoYtn samples

examined under
identical conditions (the same geometry and detector, integrated
charge, and amplifier settings), R = L In the use ofEq. (6.16),
one must make a first guess of the composition of the unknown
so that an input value for [t]~ can be obtained. As new values
for x are calculated, the value of [E]~ must be changed using
the Bragg rule. Usually, several iterations are required before the
value of x calculated using Eq. (6.16) and the value of x used in
the calculation of
[]1:
converge.
For the conditions where the standard is a bulk specimen and

the unknown is a thin film, the surface height of the known must
be compared to the area of the unknown. Again, we assume a
known with composition AzCw and an unknown with
composition AxBy. The surface height from the known is given
by the second part of Eq. (6.15). Assuming that the unknown is
thin enough to have a constant cross section, the total yield, or
peak integral, is given by
cr A (E,)OuQu CN;:"t)
(6.18)
COSU io
where N~ is the atomic density of element A in the unknown,
t is the thickness. and N~t is the areal denSity. Rewriting the
The (p,a) reactions can be divided into two groups,

depending on their Q values. The flrst group has a relatively low
Q value
'Be, ''s, "AI), and the second group has a much
)ligher Q value CLi, I1B, "0, "F, "No, "P). In Figs. 6.7 and 6.3,
the energy dependence of the reaction products on the incident
energy can be seen at = 150. It is useful to choose an energy
with no overlap between two different reactions. For example,
Fig. 6.8 demonstrates that the simultaneous detection of 2~a
and "P by (p,<1) is impossible because the energies of the
emerging a. particles are the same for the two reactions almost
over the entire range.
fLi,
As noted previously, problems arise when the absorber foil

technique is used. This is the simplest filtering method, but it
cannot always be applied. Above a certain backscattered proton
energy, the absorber, which stops the backscattered. (or forwardscattered) protons, also stops the a. particles. In Fig. 6.9, we
sbow calculated values of the minimum a, d, and }He energies
above which these particles can penetrate a Mylar absorber but
the protons are stopped. These data were calculated using TRIM
or SRIM (see Chapter 2), with the absorber thickness chosen to
be the range of the proton beam and then the <1, d, and 'He
energies calculated as the values that gave the same range as that
of the proton beam. Usually, a slightly thicker absorber is used
to be sure to stop all of the protons. In the case of very light
substrate. this absorber thickness can be lower, depending on the
maximum energy of the scattered protons.
I
I
I
I
I
I
I
II
I
. The relationship between the proton energy and the n, d, and
~e energies is practically linear, with slopes of 1.35, 3.6, and
second part ofEq. (6.15) in terms of O'A(Eo), substituting into Eq.
4.15 for d., 3He, and a., respectively.- These slopes can provide a
good guideline to the appropriate energy in a given situation.
(6.18), and solving for N;:"t yields
6.5.1.1 'Li(p,<1)'He and 'Li(p,'He)'He, Q = 4.02 MeV

This reaction is not useful because the low Q value means
that only for low incident proton energies can a foil thickness be
found that will stop the elastically scattered protons but still let
through the emitted particles. At
= 150, the highest proton
energy for which a foil can be used is 0.370 MeV for detection
of Cl and 0.575 MeV for detection of 3He.
(6.19)
where
(6.20)
6.5.1.2 'Li(p,a)'He, Q = 17.347 MeV
This reaction is excellent for 'Li detection. The cross section
was measured at (3 = 90 between 1.4 MeV and 2.0 MeV and at
0= 1200 between 0.4 MeV and 2.5 MeV (Sweeney and Marion,
1969). The cross section increases very smoothly up to 3.26 mb/sT.
Detailed cross-section measurements were earned out by
6.5 USEFUL REACTIONS

In this section, we list some useful or potentially useful
reactions for light ions incident on elements from hydrogen to
flUOrine, along with their parameters. The detailed cross sections
136
"'
Cassagnou el al. (1962) for 0 ~ 85', 109', 119", 128', 138', 147",
15Y, and 164' between 1.5 MeV and 3.5 MeV and by Maurel (in
Mayer and Riminl, 1977) at = 150 in the energy range ofO.S2.0 MeV. The cross section has a broad maximum around 3
MeV. An example application of the reaction can be found in
Sagara et al. (1988). The (p,a.) reaction that can interfere is
1~(p,a.)160. but if one chooses the proton energy carefully, it can
be avoided.
105', and 1200 between 0.9 MeV and 1.3 MeV and at 0 = 1400
between 0.9 MeV and 3.0 MeV (Hagendom and Marion, 1957).
A broad resonance exists around 1 MeV, with a narrower one at
1.21 MeV. Cross sechoDS at ~ 120' and 0 ~ ISO' between 2.5
MeV and 5.0 MeV can be found in lausel-Hlisken and
Freiesleben (1977).lnterferenee with 180(p,a.)''N is possible.
6.5.1.3 'Be(p,a.)'Li, Q. ~ 2.125 MeV; 'Be(p,d)'Be(2a.), Q. ~

0.5592 MeV
Both reactions have a sufficiently low Q value that the use of
an absorber is limited to very low energies, The cross sections of
both reactions were measured at approximately 8 = 90 between
0.1 MeV and 0.7 MeV (Sierk and Tombrello, 1973). Both
reactions show broad resonances at approximately 330 keY, and
the (p,a) reaction cross section is almost isotropic up to 0.36
MeV. Above 450 keV, both cross sections vary slowly, so that
they can be used for overall surface content measurement (an
absorber cannot be used at this energy). The accessible depth is
limited by the presence in the spectrum of the deuterons
produced by the 9se{p,d)sSe(2a) reaction. For examples of
depth profiling using the '13e(p,a)6L i reaction, see Shaanan el al.
(1986) and Reichle el al. (1990).
6.5.1.4 "B(p,<lj'Be, Q ~ 1.147 MeV
The Q value is too low for this reaction to be very useful for
lOs measurement. Cross sections of both l~(p,a,iBe and
lDs{p,al)7Se were measured at 0 = 50 and 0 = 90 between 2
MeV and II MeV (Jenkin el al., 1964).
6.5.1.5 "B(p,a.)'Be, Q ~ 8.582 MeV
The cross section of the (P,ao) reaction is much smaller than .
that of the (P,C(.I) reaction, and so, the latter is us~lly used. The
cross section of the (P,o.l) reaction was measured at 9 = 85.4
and
~ 151.85' between 0.8 MeV and 6.0 MeV (S)TIlons and
Treacy, 1963). The reaction changes very slowly between 1
MeV and 2 MeV. For (P,ao), broad resonances can be found
around 2.0,2.5, and 3.8 MeV at 0 ~ 151.85', but" is very =11
even at the resonances as compared with the cross section of
(p,o.,).
At lower energies, (P,a.l) has a very broad resonance (f = 300

ke V) at 650 ke V with cr peaking at 90 mb/sr and a narrower
resonance (r ~ 6.2 keV) at 163 keV with (J ~ 5 mb/S! in the
range 0 ~ /40-160' (Ligeon and Bontemps, 1972).
Examples of applications can be found in Lu et al. (1989)
(high,mergy plateau) aod in Scanlon el al. (1986) and
Moncoffre el al. (1990) (Iow~ergy resonances).
6.5.1.6 "N(p,a.)"C, Q ~ 4.966 MeV
Instead of this reaction, the 1~(p,ay)12C reaction is usually
used where the y-ray is detected (see Chapters 7 and 8). The
cross section was measured at = 15, 25. 40, 60, 75, 90,
137
6.5.1.7 180 (p,a)''N, Q ~ 3.9804 MeV

This is probably the most frequently used (P,ex.) reaction. The
cross section was measured at 0 = 68,128, and 170." bc;:twecn
0.9 MeV and 2.0 MeV (Amsel, 1964); at ~ 165' between 0.5
MeV and 1.0 MeV (Amsel and Samuel, 1967); and at e
155 0
between 1.7 MeV and 1.775 MeV (Alkemade el al. 1988).
Overall surface content measurements can be performed at =
165 0 and ~ = 730 keY, where the cross section changes very
slowly. The cross section has several resonances that can be used
for depth profiling. Values for the width of the resonance, r, and
the cross section cr as a function of proton energy are given in
Table 6.3. The 1165 keY resonance is very narrow, but it is
weak and sits on a large continuum in the cross section. The
most easily used resonance is that at 629 ke V, for whi ch the
shape of the cross section has been measured very precisely
(Maurel, 1980). The 152 keV resonance has been studied
intensively (B~ttistig et al., 1991, 1992). A very :narrow
resonance width I 00 e V) was found, and despite its low cross
section, the fact that the cross section is isotropic in the Vicinity
of the resonance allows use of an annular detector with a large
solid angle. This resonance has been extensively used for
isotopic tracing studies of oxidat.ion mechanisms of Si and SiC
(see, e.g., Viekridge el al., 2002, and Vickridge el al., 2007).
Depth resolution on the order of 1 run is possible (Trimaille,
1994). ---
Table 6.3. "F(p,a)"O parameters:

Energy (9)
(MeV)
0.777(90)
0.777(150)
0.840(150)
0.843190)
1.347(150)
1.354190)
1.652(150)
1.713(90)
1.842(t50)
1.901(90)
cr
(keV)
(mb/sr)
7
10
32
17.5
36
36
90
72
122
0.075
0.17
0.22
0.13
3.2
1.05
1.4
2.9
32
1.6
Chapter 6
rable 6.4 190 Cp,al"N panunetero

Energy
r
(MeV)
(keV)
152
629
1165
1766
-"
(mb/sr)
0.05
2.1
0.05
60
4.5
135
6.5.1.8 19F(p,a)"O, Q ~ 8.1137 MeV

In this case, the y portion of the (P,Cly) reaction is used more
frequently (see Chapters 7 and 8). The a. portion of the cross
section was measured at 0 ~ ISO' and 90' between 0.5 MeV and
2.0 MeV (Dieumegard et of., 1980). For overall surface content
determination, E = 1260 ke V can be used because of the
stationary part that extends down to 1100 keY. At this energy
range, this reaction is absolutely background-free, and there are
Energy of protons [MeV]
FIG. 6.8. The energy of the reaction products (willI relatively

high Q values) for several (p,a.) reactions as a function of proton
no interfering reactions. Several resonances exist in this energy

range, and values for the width of the resonance,
.-'
r, and the cross
energy.
section cr as a function of proton energy are given in Table 6.4.

For resonance measurements, the 19F{p,ay/60 reaction is highly
recommended.
~
!. 15
3.0
"c"
J= 10
2.5
0
"'"
--""
_.- d
,;
2.0
a
~
..""
15
>
~1.o
05
Energy of protons [MeV]
0.0
2
3
Energy or proton [MeV]
FIG. 6.9. The minimurp
Ct,
d, or 3He energy, above which the a.,
d, or lHe particles can penetrate the Mylar absorber but protons

are stopped. These data were calculated using TRIM with the
thickness set equal to the range of the proton beam. The a, d,
FIG. 6.7: The energy of the reaction products (with relatively

low Q values) for several (p,a), (p,'He), and (P,d) reactions as a
function of proton energy.
and 3He energies are those that gave the same range as the
proton beam in the Mylar. The relationship between the proton
energy and the a., d, or 3He energy is linear, and the slopes are
1.35. 3.6, and 4.15 for d, 3He. and a., respectively.
138
.- -Energy of deuterons (MeV)
FIG. 6.10. The minimum detectable energy of p, JHe, and a.

particles as a function of the deuteron energy using a Mylar
absorber. The procedures used to calculate these energies are the
same as were used for Fig. 6.9. The relationship between the
deuteron energy and the energy of the reaction particle is linear.
and the slopes are 0.76, 2.65, and 3.05 for P, 'He, and a,
respectively.
6.5.2 Deuteron-induced reactions

Almost all of the light elements have deuteron-induced
reactions, most of which have positivc Q values. We do not
mention all of them here, but cite only those that are commonly
used. The reactions are mainly (d,p), but (d,a) and (d,'He)
reactions are also used. The compound nuclei in these reactions
usually have many accessible excited states and, therefore, have
many groups of emitted particles [e.g., 14N(d,pI-7)1~, 19F(d,pl(6)2~. etc.). The large number of emitted particle groups can
easily cause interference between different pairs of elements.
Here, we do not include the. reactionMproduct energy vemlS
deuteron energy curves for the deuteron-induced reaction as we
did for proton reactions, because researchers can calculate these
for their particular detection angle and elements. In Appendix 10
are listed the energies of the particles generated in (d,p) and
(d,a) reactions at 0 = 150 and Ed = 0.9 MeV and 1.2 MeV
(Amsel, 1991). Appendix 10 also contains the energies of the
particles aftcr they have passed through a Mylar absorber foil
whose thickness was chosen to stop the elastically scattered
deuterons, A diagram showing energies of the emitted protons at
o = 150 and Ed = 1.8 MeV from elements A = 2-31 can be
found in Viclaidge (1988).
6.5.2.1 D(d,p)T, Q = 4.033 MeV

This reaction is always present when a deuteron beam is used,
because of the reaction between the incoming beaD::! and
deuterium already stopped in the sample. Its inteIlSit),; thus
increases with incident dose, and in frequently used stalLdards,
the proton peak from this reaction can become significan .... (For
example, in 160 measurements, this peak appears betw~n the po
and Pt peaks.) A compilation of crossMsection data can be found
in Jannie and Seagrove (1957) .
6.5.2.2 'Hd,p)'He and 'He(d,a)'n, Q -18.352 MeV
See Section 6.5.3.1 on D('He.p)'He and D('He,a)'H
reactions.
6.5.2.3 'Li(d,a)'He, Q = 22.374 MeV
The cross section for this reaction was measured at = 1500
between 0.5 MeV and 2.0 MeV (Maure1 el al., 1981). It bas a
quasiplateau at approximately 0.65 MeV that can be used for
o,,'eraU surface content measurement.
6.5.2.4 'Be(d,t)'Be, Q = 4.496 MeV

The cross section for this reaction was measured at E) = 90
from 1.5 to 2.5 MeV (Biggerstaff <I al . 1962). Plateaus can be
found above 2 MeV.
6.5.2.5 'Be(d,a)'Li, Q 7.153 MeV

The cross sections of the (d,exo) and (d,Ctt) reactions were
measured at 0 = 90 and 0 = 165 (Biggerstaff el al., 1962).
The cross section of (d,ao) shows plateaus between 1.5 MeV and
2.5 MeV at 0 = 90 and between 1.0 MeV and 1.5 MeV at 0 =
165. -. - -_.
6.5.2.6 IOB(d,a)'Be, Q = 17.818 MeV
The cross section was measured at e = 156" between 1.0
MeV and 1.8 MeV for both (d,,,") and (d,n,) (Purser and
Wildenthal, 1963). Both cross sections have slowly changing
regions [1.0-1.8 MeV for (d,no) and 1.4-1.6 MeV for (d,a,)],
but the cross section of (d,al) is 10 times higher than that of
(d,,,,,) and Q, (14.878 MeV) is still high; therefore, the (d,n,)
reaction is recommended.
6.5.2.7 12c(d,p) 13 C, Q - 2.722 MeV
Detailed cross-section measurements were carried out at
different angles from 1.8 MeV to 3.5 MeV by McEllistrem et al.
(1956). The cross sections of (d,Po) and (d,p,) (Q, = --{).37 MeV)
were measured by Kasby el al. (1960) at 0 = 47.6', 80.5',
158.4', and 165 for (d,Po) and at 0 - 80.5 for (d,p,) in the
range of 0.7-2.0 MeV. The (d,Po) cross section has a plateau at
backward angles above 0.9 MeV. This plateau is usually used to
The minimum detectable energies of p, lHe, and Ct as

functions of the deuteron ellergy using Mylar absorber are
shown in Fig. 6.10. As in the case of protons, the relationship is
practically linear, with slopes of 0.76, 2.65, and 3.05 for p, 3Re,
and a, respectively.
139
Chapter 6
6.5.2.12 "N(d,a)"C, Q = 7.687 MeV

The cross sections of (d,,,,,) and (d,(1I) were measured al 0 =
90' from 1.2 MeV to 2.5 MeV (Mansour ef al. 1964). That
article also presents angular distributions for (d,no), (d,al), and
(d,a.Z+3) at various energies. A new measurement was made by
Sawicki ef al. (1986) at 0 = 150' in the 0.8-1.3 MeV energy
range. A more recenl measuremenl (Vickridge ef al., 1996)
showed significant divergence from these previous works, in
both energy and angular dependences; however. these cross
sections do have a slowly changing region between 1.0 MeV and
1.2 MeV. It was shown (Vickridge et ai., 1995) that it is possible
10 simultaneously depth profile "N and I'N via (d,,,) reactions.
measure overall carbon content in the surface layer. Thin oxygen

standards can be used to determine the absolute amount of
carbon, using the cross-section ratio of the (d,p) reactions on \60
and I'e (Amsel and Davies, 1983) as shown in Table 62.
6.5.2.8 13C(d,p)"C, Q =5.951 MeV
This cross section was measured at 0 = 20 and
135 in
the energy range of 0.6-3.0 MeV (Marion and Weber, 1956).
The cross section has a plateau around 1.1 MeV at = 165 that
can be used for overall surface content measurement. The peaks
from 12e and He are well separated because of the large
difference in the Q values. This has allowed isotopic tracing
studies in which Ile and BC are measured in a single
measurement (Colaux et 01., 2007)
e ;: : ;
6.5.2.9 "N(d,p)"N, Q = 8.610 MeV

The cross section for this reaction was measured at 9 = 150
from 0.6 MeV to 1.4 MeV for Po, p, + P,. p" p., and p, (Amsel
and David, 1969). At higher energies, the cross section was
measured at = 60, 90 a , 120, and 150 in the energy range of
2.6-5.0 MeV for Po-PIO (Koenig et 01., 1977). The croSs section
was aho measured at
= 150 from 0.5 MeV to 0.65 MeV for
p, (Berti and Drigo, 1982) and from 0.5 MeV to 2.0 MeV for po
(Simpson and Earwaker, 1984). There is a plateau in the PI + pz
cross section at = 150 0 between 1.05 MeV and 1.25 MeV. The
ratio 1'N(d,Po)I["Si(dPI)+"Si(d,p,)] was measured al 0 = ISO'
between 1.72 MeV and 1.82 MeV (Vickridge, 1988). This ralio
is almost constant and so can be used for high precision silicon
nitride stochiometry measurement. Because of the large Q value,
very thick detectors (~300-500 urn) should be used to stop the
r:rotons. Some of the froton groups can interfere with the
'C(d,p)13 C and l'O(d,p) 70 reactions. Also, thc peaks from the
different proton and Q groups can overlap in the spectrum. This
effect can be avoided by careful choice of the absorber foil and
the use of thin detectors to move the proton peaks to lower
energies in the spectrum.
65.2.13 l'O(d,p)17 0, Q = 1.917 MeV

This is the most frequently used reaction for 160 detection,
and probably the most widely used reaction in NRA. The cross
seclions for (d,Po) (Qo = 1.046 MeV) and (d,p,) (Q, = 1.046
MeV) were measured at <:) = 164.25 0 in the 0.7-2.2 MeV energy
range (Seiler ef al., 1963), at 0 = 10' and 87' from 0.8 MeV 10
2.0 MeV (Amsel, 1964), and at 0 = 135' from 0.2 MeV to 3.0
MeV [(d,PI) from 0.52 MeV] (Jarjis, 1979) and at 0= 150' from
0.7 to 1.8 MeV (Gurbich and Molodtsov, 2004). Because the
D(d,p)T and "O(d,p) 190 (if "O-enriched samples are
investigated) reactions can interfere with (d,po) peak. and because
the cross section of(d,PI) is larger than that of (d,po), the p, peak
is usually used for 160 content detennination. The cross section
of (d,Pl) was rneasmed with bigh precision at = 150, between
0.4 MeV and l.l MeV (Amsel and Samuel, 1967). In most
cases, the plateau between 0.8 MeV and 0.9 MeV is used. In the
case of materials having low-Z matrixes (particularly Al and Si),
the deuteron energy should be kept low enough that the effect of
the Coulomb barrier favors the cross section for the 160 (d.p)110
reaction. For very thin surface films containing 160, it might
nevertheless be advantageous to use the (d,Po) reaction, in which
= 900 Peaks from
case the cross section is higher ~t
11B(d,p)I'B and I'N(d,p)I'N also interfere with the PI p'eak.
Therefore, if the target contains a significant amount of 1 B or
I'N, the Po peak should again be used.
6.5.2.10 "N(d,a)"C, Q = 13.574 MeV

The cross sections of(d,~) and (d,al) were measured at =
90' from 1.0 MeV to 3.5 MeV (Ishimatsu ef al., 1962) and at 0
= 150' from 0.6 MeV to 1.4 MeV (Amsel and David, 1969).
Cross sections of (d.Cl(J) to (d,n8) were measured at = 60 0 , 90,
120, and 150 0 between 2.6 MeV and 5.0 MeV (Koenig et ai.,
1977). The (<1.(11) cross section shows a quasipIateau between
1.05 MeV and 1.4 MeV and is the preferred reaction for NRA.
Careful choice of the absorber foil and the use of thin detectors
can move interfering proton peaks from 14N(d,p)l~ to lower.
energies in the spectrum.
6.5.2.11 "N(d,p)''N, Q = 0.266 MeV

Becau..c;;e of the low Q value, this reaction is not useful.
140
6.5.2.14 "O(d,o.)"N, Q = 3.11 MeV

This reaction is used for high-depth-resolution 160 profiling.
The cross section was measured at 8 = 164.25 from 0.7 MeV to
2.2 MeV (Seiler ef al., 1963); at 0 = 90",135', and 165' in the
0.8-2.0 MeV energy range (Amsel, 1964); and at 0 = 145'
belween 0.75 MeV and 0.95 MeV (Turos ef al. 1973). Usually,
only a thin absorber foil or no absorber is used, because an
absorber that would stop the deuteron beam would also stop the
alpha particles. Therefore, this reaction cannot be used on high-Z
targets. where the energy of the backscattered deuterons is
almost equal to the beam energy. Because of the thin absorber
(or no absorber), the Po peak from the 160(d,Po)170 reaction can
interfere with the {l peak. To avoid this interference, one should
use a thin detector (-30 Jlm). This technique (depth resolution,

optimization, etc.) is discussed in detail by Turos ef al. (1973).
6.5.2.15 180 (d,p)"0, Q = 1.732 MeV
This reaction can interfere with the 160~d'PO)110 reaction if
the target contains a significant amount of sO. To resolve the
peaks from the different isotopes, one should use a highresolution detection technique. The explorable target thickness is
16
limited by the energy difference between the protons from 0
18
and 0.
6.5.2.16 "O(d,a)"N, Q =4.247 MeV
The cross sections of (d,o.IHd,~) were measured at
=
165' from 0.8 MeV to 2.0 MeV (Amsel, 1964). The a peaks
from
are well separated from the a peak from 160 . This
13
16
reaction is 'useful only for simultaneous
0 and 0
measurement; otherwise, the IgO(p,(l)t~ reaction should be
used becallse of its higher cross section.
tao
65.2.\7 "F(d,p)"F, Q = 4.374 MeV

This reaction is important because it can interfer~ with the
l'O(d;PI) 17 0 reaction. With thick targets, the peak of
19F(d,pl1+pl2i~ overlaps with the PI peak from to.
65.2.18 I'F(d,a)I'O, Q = 10.031 MeV
The cross section was measured at <:> = 150" between 0.8
MeV and 2.0 MeV for both (d,CL,) and (d,CLI) (from 1 MeV)
(Maurel et aI., 1981). The plateau in the (d,Ut) cross section
from 1.3 MeV to 1.6 MeV is used for I~ determination, with E.!
= 1.6 MeV. Because of the large difference in the Q values of
the (d,p) and (d,a) reactions, they do not interfere with each
other (at least with a. and al)' The a. signal from 'Li(d,CLl'He
and the almost continuous Ct spectrum from the 10s(d,a)3Be
reaction can overlap with the alphas from I~. The simultaneous
detection of light elements and the interference are discussed in
detail in Maurc1 et al. (1981). For data (energy levels, crosssection measurement, etc.) derived from the heavier elements,
see Eodt and van der Leun (1978) and IDANDL (wwwnds.iaea.orglibandl).
relationship between the energy of 3He and the energy of the

reaction products after they have passed through an absorber foil
that stops the primary beam.
6.5.3.1 n('He,p)'He and n(,He,I1)'H, Q = 18.352 MeV
The cross sections and angular distributions of the peotons
were measured by several groups in the early 1950s: He(d,:p)'He
at ~ 0 from 0.2 MeV to 1.6 MeV (Bonner ef al., 1952) and at
= 86' from 0.24 MeV to 0.94 MeV (Yarnell ef al., 1953) and
D('lle,p)'He at various angles in the range 0.1-1>.8 MeV (:Kunz,
1955). All three sets of measurements show almost DO angular
dependence; therefore, Appendix II includes total cross sections
instead of differential cross sections. The 3He(d,p),He cross
section was remeasured by Moiler and Besenbacher (1980). The
total cross section shows a broad resonance around 0.64 MeV
for D('lle,p)'He [400 keV for 'He(d,pl'He]. This reaction is
used for the detection of both deuterium and JHe detecting either
the protons at backward angles (Pronko and Pronko, 1974;
Pronko, 1974; Dieumegard ef aI., 1979; Qiu ef al., 1989; Payne
e/ al., 1989) or the alphas (Moller, 1978; Chatuverdi ef al., 1990;
Langley et al., 1974) at forward angles, or for detecting the two
types of particles in coincidence (Wielunski and MoUer, 1990).
The high-energy protons should be detected either by a very
thick (1500 ~m) surface barrier detector (Dieumegard ef al.,
1979; Payne et al., 1989) or by a plastic scintillator (Qiu et al.,
1989). The backscattered deuterium particles should be rejected
by an absorber or by electrostatic (Moller, 1978) or magnetic
(Chatuverdi et al., 1990) deflection, or a coincidence
measurement can be used (Wie1unski and Moller, 1990). It
should be noted that the reaction at backward angles has inverse
kinematics, that is, the lower the energy of the incident 3He, the
higher the energy of the protons and alphas. -
6 J
8 ~-
'_'"
6.5.3.2 b( He,p) Be, Q - 16.7863 MeV

The cross sections for eHc,po) and eHe,Pl) were measured at
o = 0' and ISO" between 1.0 MeV and 5 MeV (Schiffer ef al.,
1956).
65.3.3 'Be(,He,p)l1B, Q = 10.3219 MeV
The cross sections for eHe,po) and CHe,Pl) were measured at
0"" 0, 7, 30, 60, 90, 120, and 150" from 1.8 MeV to 5.1
MeV (Wolicki er al., 1959).
6.5.3 'He-Induced reactions

With the notable exception of the D('He,p)'He and
OeHe,o.)iH reactions, lHe-induced reactions have not been
nearly as extensively used as the (P,CL) and (d,p) reactions. We
can nevertheless also cite the use of 3He for detection of 12e
(Gossett, 1981) and 160 (Li er al., 1991). These reactions offer
us an alternative reaction to detect He and 160, mainly where
deuteron beams cannot be used. In Fig. 6.11, we show the
6.5.3.4 'Be('He,a)'Be, Q = 18.9124 MeV

Angular distributions and differential cross sections were
measured at 0 = 40, 90, 125, and 1500 in the range of 2.510.0 MeV (Bilwes ef al., 1978).
141
Chapter 6
6.5.3.5 "S(,He,p)"C, Q 19.6931 MeV

The cross sections for eHe,po) and eHe,p,) were measured at
e = 0' and 90' from 0.5 MeV to 5 MeV (Schiffer e/ al., 1956).
6.5.3.6 "B(,He,p)"C, Q = 13.1853 MeV; "B(,He,d)"C, Q =
10.4635 MeV
The cross sections of (,He,d.,), (,He,po), and (,He,p,+p,+p,)
were measured at e = 60, 90, and 163 0 between 3.0 MeV aDd
5.5 MeV (Holmgren e/ al., 1959).
:;G
!
-,
--- p
"< 4
"0:
..
E'
6.5.3.7 "C('He,p)"N, Q 4.7789 MeV; "C('He,a)"C, Q

1.8563 MeV
The cross sections of eHe,ao) and eRe,Prill) were
measured at 0 = 76 and 159.4 in the energy range of 2.0--5.5
MeV (Kuan et ai., 1964). More recently, the cross sections of
eHe,po-Pl) were measured at e = 90 from 2.1 to 2.4 MeV
(Tongelal., 1990).
<
w
---
2
3
Energy of 'He [MeV]
6.5.3.8 "O(,He,p)"F, Q = 2.0321 MeV; "O(,He,a) 15 0, Q =

4.9139 MeV
The angular dependence and the differential cross section at
e = 90' were measured from 2.1 Me V to 3 .1 MeV for (,He,a.)
and eRe,Po-P7) (Bromley et al., 1959). The cross section for

0.0 was measured at e = 70, 80 0 , 90, 100 0 , 120 0 , 1300, 140,
and ISO' in the energy range 00.2-6.2 MeV (Ott and Weller,
1972). The cross section around a broad resonance at 2.37 MeV

at 90' was remeasured with high precision by Abel et al. (1990)
and Lennard et af. (1989). The application of this reaction is
discussed in Abel et al. (1990).
Most of these cross sections contain many narrow resonances

that might allow high-resolution depth profiling, but in most
cases, the resonances are neither isolated nor Darrow enough for
practical application. These cross sections are not included in
measurements. For example, the resonances of the l~(CL,a)l~
3000
>'
..
!!'
_'4N(a.o:)''N
-
2500
- '4N(a,p)l'Q
-.~
"'si(a.pflp
,
,
,, ,,
, ,, ,
.!!.
1l
,
"~
,
-,,
,
0: 2000
1500
.{
;
"
~
a;
5003000
3500
4000
4500
5000
5500
5000
Incident (I energy (keV]
Appendix II.
The (a,p) reactions with negative Q values can interfere with
RBS measurements at high enough energies. This interference
could be impoI1ant in the case of elastic resonance
'4
FIG 6.11. The relationship between the energy of 3He and the
energy of the reaction products after they have passed through an
absorber foil that stops the primary beam.
6.5.4 Alpha-induced reactions

A few of the light elements have positive Q values (lOS, liB,
19F, ~a, 27A1 , 31 p, and 3S Cl) for a-induced reactions, which
would make it possible to detect protons under almost
backgrOtwd-free conditions; however, the cross sections are high
enough only at relatively high energies, usually above 2 MeV.
-.-
--_.- ..... ~.-:=-::::.~::: ...
FlG. 6.12. A comparison of the energies of the reaction products for

the 28Si(a,p)J1p, 1"N(a....p)170. and 14N(a.,a)14N reactions as a
function of incident a energy.
lO
6.5.4.l B(l1,p)"C, Q = 4.0626 MeV; "B(a,p)"C, Q = 0.7839

MeV
These two reactions wcre studied mainly for nuclear physics
(for details, see Dayras el al., 1976; Shire el al., 1953; Shire and
Edge, 1955; Wilson, 1975; Lee and Schiffer, 1959; Mani et al.,
reaction are above 3 MeV. and the peak of the protons coming
from the ''N(a,p)"0 reaction (Q = -1.1914 MeV) can overlap
with the peak of the elastically scattered a. particles in a certain
energy range.
142
1963). For boron detection, the IOB(a.,p)IJC reaction was used

around the 1.567 MeV resonance (Annigliato e/ at, 1978) and
around the 2.67 MeV resonance (Olivier et a/., 1976). The
lIB(a,p)I'C reaction was used at 2.67 MeV (Olivier el al., 1976).
These reactions were investigated by McIntyre ct a/. (1992).
Amsel, G., and Davies, J. (1983), Nuel. Instrum. Methods 218,

177.
Amsel, G., and Lanford, WA (1984), Annu. Rev. Nud. Pari.
Sci. 34, 435.
Amsel, G. (1991), private communication (Ecole Normale
Superieure, Paris, France).
6.5.4.2 I4N(a,p)"0, Q - -1.1914 MeV
Although this reaction has a negative Q value, it is worth

mentioning here. The cross section was measured by Herring e/
al. (1958) and Kashy eI al. (1958) and mOle receotlyby Lin eI al.
Amsel, G., Faszti, F., Szilagyi, E., and Gyulai, J. (1992), Nud
Instrum. Melhods B63, 421.
(1993). The "Si(a,p)"P reaction (Q =-1.916 MeV) can interfere
Armigliato, A., Bentini, G.G., Ruffini, G., Battaglin, G., Delia
with this reaction. To get an estimate for thc region of

interference, see Fig. 6.12, which shows the energies of the
protons from the 1~(a,p)170 and 28Si(o.,p)31p reactions and also
the energy of the elastically scattered a particles from 14N. An
example ofan application can be found in Doyle el a/. (1985).
Mea, G., and Drigo, A.V. (1978), Nuel. Instrum. Melhods

149,653.
Barradas. N. (1997) Appl. Phys. Lell. 71,291. Also see the IBA
DataFurnace
Homepage
at
http://www.ee.surrey.ac.uklibclndf.
6.5.4.3 "F(a,p)nNe, Q =1.6746 MeV
Barradas, N .P., Arstila, K., Battistig, M. Bianconi, M.,

Dytlewski, N., Jeynes, E. Kotai, E., Lulli, G., Mayer, M.,
Rauhala, E., Szilagyi, E., and Thompson, M. (2007), Nud.
Inslrum. Melhods B262, 281.
For details on this reaction, see Endt and van der Leun
(1978). The resonance around 2.443 MeV was used for depth
profiling by Smulders (1986) and Kuper et al. (1986). McIntyre
et al. (1989) studied the depth profiling capabilities of this
reaction.
Baskhin, S., and Richards, HT. (1951), Phys. Rev. 84, 1124.
Battistig, G., Amsel, G., and d'Artemare, E. (199l), Nucl.

REFERENCES
Abel, F., Amsel, G., d'Artcmarc, E., Ortega, C, Siejka, J., and
Vizkelethy, G. (1990), Nuellnstrum. Methods B45, 100.
Battistig, G., Amsel, G., and d'Artemare, E. (1992), Nuel.
Ajzenberg-Selove, F. (1986a), Nuel. Phys. A449, I.
Berti, M., and Drigo, A.V. (1982), Nuel. Instrum. Methods 201,
473.
Ajzenberg-Selove, F. (1986b), Nucl. Phys. A460, 1.

Ajzenberg-Selove, F. (1987a), Nucl.Phys. A474, 1.,
. __ . _.
Biggerstaff, J.A., Hood,R.F., Scott, H., and McEllistrem, M.T.

--,. (1962), Nud. Phys. 36, 631.
.-.-.
Ajzenberg-Selove, F. (1987b), Nucl. Phys. A475, 1.

Ajzenberg-Selove, F. (1988), Nuel. Phys. A490, 1.
Bilwes, B., Bilwes, R., Ferrero, J.L., and Garcia, A. (1978), J.

Phys. 39,805.
Ajzenberg-Selove, F. (1990), Nuel. Phys. A506, 1.
Bird, J.R. (1980), Nuel.lnstrum. Methods 168, 85.
Alkemade, P.F.A., Slap, CAM., Habraken, F.H.P.M., and van

der Weg, W.F. (1988), Nuellnstrum. Methods B35, 135.
Bird, J.R., Duerden, P., and Clapp, R.A. (1981), Nud. Instrum.
Melhods 191, 345.
Amsel, G. (1964), thesis, Ann. Phys. 9,297.
Bonner, T.W., Conner, J.P., and Lillie, A.B. (1952), Phys Rev.
88,473.
Amsel, G., and Sarouel, D. (1967), Anal. Chem. 39, 1689.
Bromley, DA, Kuhner, JA, and Almquist, E. (1959), Nuel

Phys. 13, 1.
Amsel, G., aodDavid, D. (1969), Rev. Phys. Appl. 4, 383.

Amsel, G., Nadai, J., Ortega, C., Rigo, S., and Siejka, J. (1978a),
Nuellnstrum. Melhods 149, 705.
Cassagnou, Y., Jerooymo, J.M.F .. Sadeghi, A., and Forsyth, P.D.

(1962), Nuel. phys. 33,449.
Amsel, G., Nadai, J.P., Ortega, C., and Siejka, J. (1978b), Nuel.
Instrum. Methods 149, 713.
Chater, R.J., Kilner, l.A., Reson, K..l., Robinson, A.K., and

Hemment, P.L.F. (1990), Nuel.Inslrum. Melhods 845, 110.
Arosel, G., d'Artemare, E., and Girard, E. (1983), Nucl. Instrum.

Melhods 205,5.
Chaturvedi, u.K., Steiner, U., Zak, 0., Krausch, G., Shatz, G.,
and Klein, J. (1990), Appl. Phys. Lett. 56, 1228.
143
-- -_. - - - - - - - - - -
..
"'\'1
Chapter 6
Cheang Wong, J.e., Ortega, C, Siejka. 1., Trimaille, L, Sacuto,
Jenkin, J.G., EaIwaker, L.G., and TiUerlon, E.W. (1964), Nucl.

Phys. 50, 517.
A., Balkanski, M., and Vizkelethy, G. (1992). Nucl.

lns/rum. Methods B64, 179.
liang, W.L., Shutthanandan, V, Thevutbasan S., McCreadYI
Colaw<, J.L., Thome, T., and Terwagne, G. (2007), Nucl.

lns/rum. Methods B254, 25.
D.E., and Weber, W.I. (2003), Nucl. /ns/rum. Methods

B207,453.
Dayras, RA., Switowski, Z.E., and Tombrello, A.T. (1976),

Nuel. Phys. A261, 365.
Kashy, E., Miller, PD., and Risser, J.R (1958), Phys. Rev. 112,
547.
.
Deconninck, G. (1978), Introduction

Chemistry, Elsevier, Amsterdam.
Radioanalytjcal
Kashy, E., Perry, RR, and Risser, J.R (1960), Phys. Rf:v. 117,
1289.
Dieumegard, D., Dubreui~ D., and Amsel, G. (1979), Nucl.

lns/rum. Methods 166, 431.
Koenig, W., Riccato, A., Stock, R., Cuzzocrea., P., Perillo, E.,
to
Sandolini, M., and Spadaccini, G. (1977), Nuova Cimento

A39,9.
Dieumegard, D., Maurel, B., and Amsel, G. (1980), Nucl.

lns/rum. Methods 168, 93.
Kuan, H-M., Bonner, T.w., and Risser, J.R. (1964), Nucl. Phys.
51,481.
Doyle, B.L., Follstaedt, D.M., Picraw<, S.T., Yost, F.G., Pope,

L.E., and Knapp, J.A. (1985), Nucllnstrum. Methods B7--8,
166.
Kunz, W.E. (1955), Phys. Rev. 97,456.

Kuper, F.G., De Hosson, lTh.M., and Verwey, J.F. (1986), J
Appl. Phys. 60, 985.
Elwyn, AJ., Holland, RE., Davids, C.N., Meyer-Shulzrneister,

L., Mooring, F.P., and Ray Jr., W. (1979), Phys. Rev. C20,
1984.
Langley, R.A., Picraux, S.T., and Vook, F.L. (1974), J. Nucl.

Mater. 53, 257.
Endt, P.M., and van der Leun, C. (1978), Nucl. Phys. A310, 1.
Lee Jr., L.L., and Schiffer, J.P. (1959), Phys. Rev. 115, 160.
Feldman, L.C., and Mayer, J.W. (1986) Fundamentals of
Surface and Thin Film Analysis, North-Holland, New York,
Lennard, W.N., Tong, S.Y., Mitchell, LV., and Massowni, G.R

(1989), Nucl. Instrom. Methods B43, 187.
Chapter 12.
L~
Gossett, C.R (1981), Nucl.lns/rum. Methods 191, 335.

Gurbieb, A., and Molodtsov, S. (2004),Nucl.lnslrum. Methods
B226,637.
J., Vizkelethy, G., Revesz, P., Mayer, J.W., Matienzo, L.J.,

Emmi, F., Ortega, C., and Siejka, J. (1991), Appl. Phys. Lett.
58, 1344.
Ligeon, E., and Bonternps, A. (1972), J. Radioanal. Chern. 12,
_Hagen~~rn:!.B~,andMarion, J.B. (1957), Phys. Rev. 108, 1015.
. 335.
Herring, D.F., Cbiba, R., Gasten, B.R, and Richards, R.T.

(1958),Phys. Rev. 112, 1210.
Lin, Z., McIntyre Jr., L.e., Leavitt, I.A., Ashbaugh, M.D., and
Hirvonen, J.K., and Luecke, W.R. (1978), Nuel. lns/rum.

Methods 149,295.
Lu, X, Yuan, Y., Zbeng, Z., Jiang, W., and Liu, J. (1989), Nucl.
lnstrum. Melhods B43, 565.
Holmgren, HD., Wolicki, E.A., and Johnston, RL. (1959),

Phys. Rev. 114, 1281.
Mani, G.5 .. Forsyth, P.D., and Perry, RR. (1963), Nucl. Phys.
44,625.
Isbimatsu, T., Nakashima, T., Kata, N., and Morita, S. (1962), J,
Mansour, N.A., Saad, H.R, Saleh, Z.A., Sayed, E.M.,

Zaloubovsky, 1.1., and Gontchar, V.1. (1964), Nucl. Phys.
65,433.
Cox, RP. (1993), Nucl.lns/rum. Methods B79, 498.
Phys. Soc. Jpn. 17, 1189.
Jarjis, R.A (1979), Internal Report, University of Manchester,
Manchester, U.K.
Marion, J.B., and Weber, G. (1956), Phys. Rev. 103,167.
Jarmie, N., and Seagrove, J. (1957), Charged porlicle cross

sections, Los Alamos Scientific Labomtory Report LA2014, Los Alamos National Laboratory, Los Alamos, NM.
Iausel-HUsken, S., and Freiesleben, H. (1977)",
363.
z.
Marion, lB., Weber, G., and Mozer, F.S. (1956), Phys. Rev.
104, 1402.
Maurel, B. (1980), thesis. Universire Paris VIT, Paris, France.
Phys. A283,
Maurel, B., Amsel, G., and Diewnegard, D. (1981), Nuc1.

lns/rum. Methods 191, 349),
144
. '~
Nllclear Reaction Analysis: Particle-Particle Reactions
Maurel, B., Arosel, G., and Nadai, J.P. (1982), Nucl. Instrum.
Methods 197, 1.
Pringle, J.P.S. (1972), J. Electrochem. Soc. 119,482.

Proceedings of IBA Conferences: 1973 Yorktown Heig/J.ts, NY
(Thin SaUd Films 19); 1975 KArlsruhe, Germany ~lenum,
New York); 1978 Washington, D.C. (Nucl. IF1Strum.
Methods 149); 1980 Aarhus, Denmark (Nucl. IMStrum.
Methods 168); 1981 Sydney, Australia (Nud. I""trum.
Methods 191); 1983 Tempe, p.;z. (Nud. Instrum. Methods
218); 1985 Berlin, Germany (Nud. Instrum. Methods BI5);
1988 Johannesburg, South Africa (Nucl. Instrum. kiethods
835); 1989 Kingston, Jamaica (Nud. Instrum. kiethods
B45); 1991 Eindhoven, The Netherlands (Nucl. I"Strum.
Methods B64); 1993 Balatonfiired, Hungary (Nucl. [:nstrum.
Methods B85); 1995 Phoenix, AZ (Nuc!. Instrum. kiethods
B118); 1997 Lisbon, Portugal (Nud. Instrum. Methods
B136-138); 1999 Dresden, Germany (Nucl. [:nstrum.
Methods BI61-163); 2001 Cairns, Australia (Nucl. [>!Strum.
Methods BI90); 2003 Albuquerque, NM (Nucl. [:nstrum.
Methods B219-220); 2005 Seville, Spain (Nucl. [mtrum.
Methods B249).
Maurel, 8., and Arosel, G. (1983), N"d. Instrum. Methods 218,

159.
Mayer, I.W., and Rimini, E. (eds.) (1977), Ion Beam Handbook,
Academic Press, New Yark.
Mayer, M. (1999), "SlMNRA, a Simulation Prograur for the
Analysis of NRA, RBS and ERDA", in Proceedings of the
15th International Conference on the Application of
Accelerators in Research and Industry, Duggan, 1.L., and
Morgan, I.L. (eds.), American Instihlte of Physics
Conference Proceedings 475, p. 541. See also the SIMNRA
Home Page: http://www.rzg.mpg.del-mami.
McEllistrem, M.T., Jones, K.W., Chiba, R., Douglas, RA.,
Herring, D.F., and Silverstein, E.A. (1956), Phys. Rev. 104,
1008.
Mcintyre Jr., L.C., Leavitt, l.A., Dezfouly-Aljomandy, B., and
Oder, J. (1988), Nucl. Instrum. Methods B35, 446.
Proceedings of the Third International Conference on Chemical

Analysis (1991), Nucl lnstrum Methods 866.
McIntyre Jr., L.e., Leavitt, J.A., Ashbaugh, M.D., DezfoulyArjomady, B., Lin, Z., Oder, J., Farrow, R.F.C., and Parkin,
S.S.P. (1989), N"cI. Scj. Technol. 1, 56.
Pronko, P. (1974), J. Nud. Mater. 53, 252.

Pronko, P.P., and Pronko, J.O. (1974), Phys. Rev. B9, 2870.
Mcintyre Jr., L.e., Leavitt, lA., Ashbaugh, M.D., Lin, Z., and
Stoner Jr., J.O. (1992), N"c!. Instnlm. Methods B66, 221.
Purser, K.H., and Wildenthal, B.H. (1963), Nud. Phys. 44, 22.
Moncoffre, N., Millard, N" laffrezic, H., and Toussct, J. (1990),

Nucl.Instrum. Methods 845, 81.
QCale (Q-value calculator) is

http://www.nndc.bnl.gov/qcalc.
available
onlme
at
Monco/Ire, N., Nucl.lnstrum. Methods (1992) B66. 126.. __ ._. Qiu, Q., Arai, E., Aratani, M, Yanokura, M., Nozaki, T., Obji,
Y., and Imura, R. (1989), Nud. Instrum. Methods B44, 179.
Moller, W., Hufschmidt, M, and Kamke, D. (1977), Nu~l.
, Reichle, R~ Berisch, R., and Reith, I. (1990),' NuC!. Instrum.
lnstrum. Methods 140, 157.
Methods B50, 68.
Moller, W. (1978), Nuc! Instrum. Methods 157, 223.
Sagara, A., Kamada, K., and Yamaguchi, S. (1988), Nucl.
Moiler, W.o and Besenbacher, F. (1980), Nucl. Instrum. Methods
168, 111.
Sawicki, 1.A., Davies, lA, and Jackman, T.E. (1986), Nucl.
Olivier, C, McMillan, J.W., and Pierce, T.B. (1976), J.
lnstrum. Methods 815, 530.
Radioanal. Chern. 31, 515.
Scanlon, P.J., Farell, G., Ridgway, M.C., and Valizadeh, R
Ott, W.R, and Weller, H.R. (1972), Nud. Phys. AI79, 625.
Payne, RS., Clough, A.S., Murphy, P., and Mills, P.J. (1989),
Nucl. Instrum. Methods B42, 130.
Schiffer, I.P., Bonner, T.W., Davis, R.H., and Prosser lr., F.W.
(1956), Phys. Rev. 104, 1064.
Phillips, D., and Pringle, I.P.S. (1976), Nud. Instrum. Methods

135,389.
Seiler, F., Jones, c.R., Anzick, W.J., Herring, D.F., and Jones,
K. W. (1963), Nucl. Phys. 45, 647.
Press, W.H., Fainnery, B.P., Teukolsky, S.A., and Vetterling,

W.T. (1986), Numerical Recipes, 1st Ed., Cambridge
University Press, Cambridge, U.K.
Shaanan, M., Kalish, R., and Richter, V. (1986), Nucl. Instrum.

Methods B16, 56.
Shire, E.S., Wormald, 1.R., Lindsay-Jones, G., Lunden, A., and

Stanley, A.G. (1953), Philos. Mag. 44, 1197.
Pretorius, R., and Peisach, M. (1978), N"cl. Inst. Methods 149,

69.
145
1,
Chapter 6
Shire, E, and Edge, RD. (1955), Philos. Mag. 46, 640.
Vickridge,I.C. (1990), thesis, Universite Paris VII, Paris,
Siejka, I., Nadai, P., and Amse1, G. (1971), J. Electrochem. Soc.

118,727.
Sierk, A.J., and Tombreilo, T.A. (1973), Nucl Phys. Al10, 341.
SigmaBase;
An
archival
site
is
available
online
at
http://www.mfa.ldki.hulsigmabasei.
France.
Vickridge, I.C. Trompetter, W.J., Browo, LW.M., and Patterson,
IE. (1995), Nucl. Instrum. Methods 899 454.
Vickridge, I.e., Trompetter, W.J., and Coote, G.E. (1996), N"cl.
lnstrum. Methods 8108367.
Vickridge, Le., Trimaille, l, Ganem, J' Rigo, S., Radtke, C.,
Smulders, P.J.M. (1986), Nucl.lnslrum. Methods 814,234.
Sokolov, J., Rafailovich, M.H., Jones. RA.L., and Kramer, E.l
Baumvo~ I.J.R, and Stewle, F.C. (2002). Phys. Rev. Lett.

89(24) 256102.
(1989), Appl. Phys. Letl. 54, 590.

SRlM, http://srim.org
Vickridge, I.C., 'Ganem, I., Hoshino, Y., and Trimaille, I. (2007),

J. Phys. D40, 6254.
Sweeney Jr., W.E., and Marion, I.B. (1969), Phys. Rev. 182,
1007.
Vizkelethy, G. (1990), Nucl. Instrum. Methods 845, I.

Wielunski, M., and MOiler, W. (1990), Nucl. Instrum. Methods
850,23.
Symons, G.D., and Treacy, P.B. (1963), N"cl. Phys. 46, 93.
Szilagyi, E., Paszti, F., and Amse1, G. (1995), Nue!. lnstrum.
Methods, 8100, 103.
'.~
Wilson, S.J. (1975),Phys. Rev. Cl1,107l.

Wolicki, E.A., Holingren, H.D., Iohnston, RL., and Geer Illsley,
E. (1959), Phys. Rev. 116, 1585
Tong, S.Y., Lennard, W.N., Alkemada, P.F.A., and Mitchell,

LV. (1990), Nucl.lnstrum. Methods 845, 91.
Trimaille,1. (1994), thesis, Universite Paris VII, Paris, France.
Yarnell, J.L., Lovberg, RH., and Stratton, W.R (1953), Phys.

Rev. 90, 292.
Turos, A., Wie1unski, L., and Batez. A. (1973), Nucl. Instrum.

Methods Ill, 605.
Ziegler, J.F. (1975), New Uses ofJon Accelerators, Plenum,

New York.
Vickridge,I.C. (1988), Nucl. Instrum. Methods 834, 470.
146
. -~
,[
I
CHAPTER
I
I
I
PARTICLE-INDUCED GAMMA
EMISSION: PIGE
J. Raisanen
Contributors: J.-P. Hirvonen and R. Lappalainen
CONTENTS
7.1 INTRODUCTION ......................................................................................................149
7.2 PARTICLE--GAMMA REACTIONS ........................................................................149
7.2.1 Deptb profiling with resonance reactions .......................................................... 149
7.2.1.1 Fundamentals .................................... ;..................... ,................................... 149
7.2.1.2 Choice of reaction ........................................................................................ 152 ..
7.2.1.3 Experimental considerations ...................................................................... 152
Accelerator ........................................................................................................... 152
Gamma-ray detection .......................................................................................... 153
Detector and beam energy calibration ................................................................153
7.2.1.4 Cautions ........................................................................................................................ 153

Factors affecting measurements ........................................................................ 153
Background and interferences ........................................................................... 154
Concentration and depth se3Ies ................................................................. 155

Standards ..................................................................................................... 156
Depth resolution and sensitivity ................................................................ 157
Supporting data analysis software ................................................................ 158
Depth profiling of the light elements ......................................................... 159
Hydrogen and helium ......................................................................................... 159
Lithium, beryllium, and boron ........................................................................... 159
7.2.1.5
7.2.1.6
7.2.1.7
7.2.1.8
7.2.1.9
Carbon .................................................................................................................. 159

Nitrogen ............................................................................................................... 160
Oxygen ................................................................................................................. 160
147
~--.
_.. _ . _ - - - - - - - - - - - - - - -
Chapter 7
Fluorine ................................................................................................................ 163

Neon ..................................................................................................................... 163
Sodium ................................................................................................................. 163
Magnesium .......................................................................................................... 163
Aluminum ............................................................................................................ 163
Silicon .................................................................................................................. 164
Heavy elements .................................................................................................... 164
7.2.2 Elemental analysis using particle-induced gamma-ray emission (PIGE) ......... 164
7.2.2.1 Thick-target gamma-ray yield ....................................................................164
7.2.2.2 Stopping power correction ......................................................................... 164
General guidelines .............................................................................................. 164
Use ofsUlndards ...............................................................................................:.. 165
7.2.2.3 Experimental considerations ...................................................................... l65
In-vacuum measurements................................................................................... 166
External-beam measurements ............................................................................ 166
7.2.2.4 Cautions ...................................................................................................... .166
Geometry.............................................................................................................. 166
Background and interferences ........................................................................... 166
Peak broadening .................................................................................................167
7.2.2.5 Detection sensitivity .................................................................................... 167
7.2.2.6 Elemental analysis of the light elements by PIGE ................................... 167
Proton-induced y-ray emission ........................................................................... 167
Deuteron-induced y-ray emission ....................................................................... 17 0
. Triton-induced r-ray emission ...............................................................;............ 170
a-particle-induced y-ray emission ...................................................................... 170
7Li-induced y-ray emission ................................................................................. 171
Heavy-ion-induced y-ray emission ..................................................................... 171
7.2.2.7 Advantages and limitations ofPIGE ......................................................... I71
.. ._____ MFE.RE_NSE~~................................................................................................................... .172
1
~'
:~
j
<
.~
..
.~
148
Particle-Induced Gamma Emission: PIGE
7.1 INTRODUCTION
The unique feature of any technique for nuclear reaction
analysis (NRA) is that it is sensitive only to specific isotopes.
This: of course, restricts its general applicability, but makes
NRA a powerful tool for those cases in which it can be utilized.
To understand the limitations and advantages of nuclear reaction
analysis with "(-ray emission and to enable the reader to decide
when these techniques are worth uSing, it is necessary to
compare this technique to other methods perfonned with
energetic ion beams.
(SIM:S). However, compared to SIMS and other sputlerio.g-based

prOfiling techniques, nuclear reaction allalysis with 1-ray
emission is nondestructive. Moreover, it is normally quite rapid,
and the number of samples to be measured" is no! a limiting
factor.
7.2PARTICLE-GAMMAREAcnONS
I
When a beam of charged particles, such as protons, heliwn
ions, or deuterons, bits the sample surface layer, nuclear
reactions are induced, and gamma radiation is emitted. Here, the
discussion is restricted to prompt gamma emission, meaning that
radiation is detected during irradiation, in contrast to activation
techniques where the rndiation is detected after ilTadiation.
Activation analysis, which has been used widely to measure bulk
concentrations of elements with mediwn or heavy atomic
weights, is discussed in detail in Chapter 10. Activation methods
have also been used for profiling by etching the sample surface,
layer by layer.
Nuclear reaction analysis with "(-ray emission is a less

frequently used ion beam analysis technique than Rutherford
backscattering (RBS), elastic recoil detection analysis (ERDA).
or particle-induced X-ray emission (PIXE), although its use has
increased significantly as a complementary method to other ion
beam methods. This is understandable because elements or
isotopes appear in RES, ERDA, and PIXE spectra in a simple,
systematic, and predictable way. This makes the interpretation
and analysis of the measurements quite straightforward. In
nuclear reactioos, even different isotopes of the same element
behave in different ways, and no systematics can be found. For
this reason, a researcher who wants to utilize nuclear reactions
has to know, isotope by isotope, the energies of resonances, the
energies of y-rays, and so OD. In many cases, knowledge of the
properties of the isotope involved is not sufficient; possible
interfering reactions in the sample matrix must bc known as
well. This results in extensive tables full of numerical data, as
can be seen in Appendix 12.
Wben a sample is bombarded with particles at a fix.ed energy,

the total gamma-ray yield from a certain reaction can be used to
deduce the average concentration of the corresponding element
in surface layers. With light ions at energies of a few
megaelectronvolts, "the maximum depth analyzed is typically
well below 50 1-JlD. When resonance reactions arc used for
profiling, the gamma-ray yield of a resonance reaction within a
proper gamma-ray energy range is measured as a function of
bombarding energy starting just below the resonance. The
profiled depth in these cases is typically limited to a few
microns. The use of particle-gamma reactions for the
RBS. ERDA. and PIXE are considered reasonably general
techniques that can also be used to probe the chemical __ measurement of depth profiles and elemental analysis of the
composition of completely unknown samples. In contrast,
light elements by PIGE spec~oscopy is discussed in Sections
nuclear reaction analysis js limited to the detection of the light
7.2.1 and 7.2.2.
elements only. Furthermore, only a few isotopes have the
potential for depth profiling. These can be found in the low7.2.1 Depth profiling with resonance reactions
7.2.1.1 Fundamentals
atomic-number section of the table of isotopes. Isotopes such as
!]C, l~, IS 0 ,19p, nNe, ~a, 24Mg , 26Mg, 27A l, and 30Si are
Most of the light nuclei (Z < 30) bave strong. sharp
resonances in the cross section a(E) of nuclear reactions induced
nonnally used. Hydrogen profiling employing inverse reactions
by light ions at low bombarding energies
MeV). Here, the
is also widely used and is discussed in Chapter 8. Special
discussion is limited to resonances leading to gamma-ray
expertise is required to try depth profiling of isotopes that are not
emission, and bombarding particles of mainly protons are
in the above list. In addition to depth profiling, 'Y-ray reactions
considered. Alpha-particle-induced reactions also show
can be used in elemental analysis in very much the same manner
resonance features, but they are far less frequently employed for
as X-rays in PIXE. The use of the charged-particle-induced
depth prOfiling because of the limited selection of sufficiently
ganuna-ray emission (pIGE or PIGME) method enables
strong and narrow resonances. (For hydrogen depth profiling
detection of the light elements with good sensitivity. The use of
with resonance reactions, see Chapter 8.) The nuclear reactions
different bombarding particles and energies offers varying
showing resonance features exploited in depth profiling are
sensitivity for different elements, which makes this technique
briefly introduced in the following discussion. In resonance
trickier to use.
capture rea<.:tions, the projectile is captured by the target nucleus
A to form a nucleus B in an excited state. The reaction can be
In summary, nuclear reaction analysis with 'Y-ray emission
written as
provides a quantitative and efficient technique in those cases in
which it can be utilized. Its unique fcature is a sensitivity for
isotopes resembling that of secondary ion mass spectrometry
(7.1)
149
Chapter 7
where B denotes the compound nucleus in an excited state,

which then decays to the ground state B through remission.
Typical examples of such reactions employed in light-element
depth profiling are (p,y) proton capture reactions. The excitation
curves are of the Breit-Wigner type (Lorentz function) in which
the excited states of the compound nucleus correspond to
reaction resonances. When no change of particles occurs but the
target remains in an excited state, the process is called inelastic
scattering. The reaction can be formally written as
a+ A -+ A + a'.- A + a'+y.
(7.2)
Typical inelastic scattering reactions involved with particlegamma analysis are (p,p'y) reactions. With light (Z ,;; 20) target
nuclei, the excitation curves show resonance features on top of a
continuum. With heavier target nuclei, the inelastic scattering
occurs well below the Coulomb barrier, and the mechanism is
Coulomb excitation. In such reactions, the cross sections
increase smoothly as a function of bombarding energy. In
rearrangement reactions, particles other than the bombarding
particles are emitted. Such a reaction can be written as
a + A -+ B + b --+ B + b + 1'.
(7.3)
Typical such reactions employed for depth profiling are (P,"l')

reactions, as rearrangement reactions also show resonance
features.
The use of resonance reactions for light-element depth
profiling is based on bombardment of the sample with an
energy-analyzed partic1e beam and detection of the induced y_
ray emission as shown in Fig. 7.1. If the energy of the beam
corresponds to the energy of a narrow resonance Er, resonance
reaction can take place only on th.e surface. A.t higher
bombarding energies, the particles must slow down in the
sample to the depth at which Er is reached, when the reaction is
possible. The yield detected at this bombarding energy, E.,
reflects the concentration at this depth. Heo.ce, the excitation
curve of the resonance reaction YCEb). that is, the yield of the
reaction versus the bombarding energy, gives an estimate of the
concentration distribution in the surface layer (see Eq. 7.5.). As
examples of tbin- and thick-target excitation cwves, the curves
for Al are provided in Fig. 7.2.
(E-E,)'+r'/4'
particle beam ~>-_....j._J L
E,>E,
FIG. 7.1. Principles of the resonance profiling method. The

sample is bombarded by particles having energy higher than the
resonance energy (or the same as the resonance energy, which
corresponds to surface layer analysis). Gamma rays originate
from reactions taking place at the depth where the bombarding
particle energy coincides with the resonance energy. The
gamma-ray yield is then measw-ed as a function of the
bombarding particle energy, whioh provides the data for
extracting the concentration profile.
~'r---~--~-------------------,
i
J
--
I
!
,-~-~:----
..
The cross section of an isolated resonance, for example, as

seen on the thin-target excitation curve of Fig. 7.2 is given by
the Breit-Wigner formula as
IT)'1. 'r,r >
where a and b refer to the incident and outgoing particles,

respectively; y is a statistical factor including spin numbers; ). is
the de Broglie wavelength 0..2- 1IEb2); r is the resonance width;
and ra and rb are partial widths. Resonance cross sections are
usually tabulated either as a cross section at the resonance energy
[a(E,) ~1t)'A'4r.r.,r' in bams (b), where 1 b~ W" crn') or as a
resonance strength [S ~ (21 + I)r,r.,r in electronvolts, where 1
is the spin quantum number of the resonance level]. Both of
these parameters are directly related to the measured gamma-ray
yield. If the energy resolution of the beam is smaller than the
resonanoe width (0) r /2.355), then the total gammaray yield
from the resonance is proportional to the maximum of the cross
section. On the other hand, if 0b ) r /2.355, the total yield is
proportional to the integral of the cross section and resonance
strength S.
..
FIG. 7.2. Tb.in- and thick-target excitation WIVes for

"Al(p,y)"Si. [Figure adapted from the data of Hunt and 10nes
(1953).]
(7.4)
150
In the case of a thick target, the reaction resonances are seen
Here, C is a constant for given detection conditior:l..S (for

example, the same geometry and collected charge). If the
resonance is narrow and beam energy resolution is good, the
functions g, f, and 0 are sharp, and therefore, the yiel<l curve
corresponds well to the actual distribution.
as abrupt increases in the gamma-ray yield (steps in the curve),

so the excitation curves are composed of successive plateaus. At
higher particle energies (around 2 MeV in case of protons), the
curve becomes monotonic, and the yield increases rapidly as a
function of energy. The resonances begin to overlap, a
continuum is reached, and the occurrence of isolated resonances
vanishes. Such behavior is typical for protons on light nuclei.
For heavier target nuclei, the Coulomb excitation becomes the
dominant process.
"
1.0
In thick ~target yield curves obtained with a narrow resonance,

one can clearly note an effect called the Lewis effect. This
results in a maximum just above the resonance energy, followed
by a shallow minimum, before reaching the constant value. In
Fig. 7.3 is displayed the experimental yield curve for the
27AI(p;y)28Si resonance at 992 keY, clearly showing the Lewis
effect. The charged particles lose energy in discrete steps, and
some or these steps can be larger than the natural width of the
resonance. It is possible that particles having initial energies well
above the resonance energy ER can jump over the resonance,
resulting in the observed minimum. Particles incident at ~ will
all have ror a finite time the correct energy to interact, resulting
in the noted maximum. The Lewis effect is noticeable only if the
resonance width and the experimental resolution are on the order
of 100 e V or less. The Lewis effect and its impact on nuclear
resonance depth profiling have been considered, for example, in
the work ofVickridge and Amsel (1996).
.
z'
:?:
z
v AI(p,y)
992 k.V
--1>.=0.00
--1>.=0.05
_. _. 1>.=0.10
.. "1>.=0.15
- '''1>.=0.20
0.5
500
1000
1SOD
EE, (eV)
Experimental yield curve (points) from the
27 A1(p,y) 28 Si resonance in solid aluminum. (Figure adApted from
the data ofVickridge and Amsel (1996).1
FIG.
7.3.
As an ex.ample of the effect of ex.perimental resolution, the

measured range profile of 100 keV "Ne ions implanted in
The sharpness of the resonance peaks/edge (Fig. 7.2) depends
vanadium is illustrated in Fig. 7.4 (represented by the points).
on the resonance width and energy spread of the ion beam. In
The solid line is the yield curve calculated Using Eq. (7.4) for the
practice, the experimental width of the resonance profile is ." assumed concentration profile described. by the dashed line. This
split-Gaussian concentration profile was obtained by a fitting
increased because of the natural width of the resonance, the
energy resolution of the beam, and the energy straggling of the
procedure comparing the calculated and measured yield curves.
In this case, the experimental resolution is determined mainly by
beam particles during the energy loss Et, - E,.. To deduce the
the cnergy straggling (function f). especially in the tail part of
precise shape of the actual concentration profile, N(x), the shape
of the resonance cross section, aCE); the energy distribution of
the distribution.
the beam, g(Et"E); and the energy straggling of particles at depth
The actual distribution N(x) can be deconvoluted from the
x, f(E,E',x), have to be taken into account
integral Eq. (7.5) using numerical methods. However, this
m
m
m
problem is ill~defined, and there are several mathematically
Y(Eb) = C Idx IdE IdE' N(x) g(Eb,E) f(E,E',x) cr(E')
correct solutions. The solution is rarely reasonable or correct,
unless physically reasonable additional boundary conditions,
o 0
0
such as positivity and smoothness of solution, are used in
(7.5)
solving the equation.
151
1
Chapter 7
Depth (nm)
100
200
300
5
~
c
'"....
'C
100 keY 22Ne+ __ V
Gi
's.
...>-,
?-
til
,,
----
-,
850
870
860
880
Proton energy (keV)

FI~. 7.4. G~a-ray yie!d curve for an implanted 100 keY 22Ne distribution in vanadiwn. The solid line is the yield calculated for the
split-Gaussian com:entrauOD profile (dashed line) obtained using a fitting procedure. The dotted line is the Monte Carlo simulation for
amorphous backing.(Lappalainen. 19&6).
7.2.1.2 Choice of reaction
used for hydrogen detection, because of the large resonance

cross section, easy availability of nitrogen beams, and good
depth resolution at the surface (see Chapter 8).
Among light-ion-induced resonance r~actions, those induced

by protons have been used mostly for profiling. In fact, most of
the light nuclei bave (p,y) or (P,II'() resonances at bombarding
energies below 2 MeV. Table A12.1 in Appendix 12 lists the
main useful resonances of this type for most of the light
elements. This table is based on the earlier compilation by
GoLicheff et al. (1972). with some useful resonances added for
. elements beavier than fluorine. Table A12.2 in Appendix 12 lists
the strongest (p,y) resonances by proton energy below 3 MeV.
The table is based on the tabulations of Butler (1959) and the
previous edition of the handbook, with updating of the data. For
practical depth profiling, an isolated, strong (S large) and narrow
(r small) resonance should be selected.
7.2.1.3 Experimental considerations

Instrumentation is discussed in detail in Chapter 16. Here, the
discussion is limited to some special features relevant for
profiling measurements. The necessary equipment for profiling
with narrow resonances is an accelerator and a gamma-ray
detector with a suitable data collection and analyzing setup.
Accelerator
The main requirements of the accelerator for profiling are
reasonable energy resolution and the possibility of changing
beam energy easily. A magnetic field is used to analyze the
energy of the beam and to separate particles of only one type
from the original beam eontaining several ions with different
charge states. The bombarding energy is generally increased
stepwise by adjustmg the magnetic field and terminal voltage.
However, more sophisticated fully automatic hysteresis-free
energy scanning systems (Amsel ei al., 1983; Meier and Richter,
1990) have been developed to increase the energy of the beam
while keeping the analyzing magnetic field constant and
selecting the energy by electrostatic deflection plates just before
and after the analyzing magnet. In profiling, the advantage of the
high energy resolution of the beam is restricted to very narrow
resonances 0.2 keV) an~ the outennost surface layer 50 nm),
In principle, (a.y), (a.ny), and (a,pr) resonances could offer

higher depth resolution than proton resonances, because of the
higher stopping power of He ions relative to H jons. However,
only a few wen-isolated, sufficiently strong (n;y) resonances
exist and are suitable for profiling. Some of these are collected
in Table A12.3 in Appendix 12.
The same resonances that have been used for profiling with
protons have been extensively utilized with inverse resonance
reactions to proftlc hydrogen. These resonances are listed in
Appendix 13. In particular, the 6.4 MeV 'H(''N,a1)''C
resonance with the detection of 4.43 MeV gamma rays is widely
152
Partide-Induced Gamma Emission: PIGE
as energy straggling due to the deceleration process starts to

dominate the total depth resolution deeper in the sample. High
energy resolution is necessary in some applications of narrow
resonances such as studies of the Lewis effect and Doppler
broadening of the resonance width because of the vibrational
motion of the detected atoms (Fujimoto, 1990; Mitchell and
Rolfs, 1991).
Doppler effect unless the gamma-my detector is at an angle of 90

with respect to the beam (Fig. 7.5), Because the Doppler shift is
typically less than 1% of the gamma-ray energy, the effect can only
be seen using a high-resolution detector such as an HPGe detector.
:;
10'
Gamma-ray detection
In particle-induced gamma emission measurements, the
radiation is generally detected either by a sodium iodide
scintillation NaJ(TI), bismuth germanate (EGO), or higb-purity
Ge (HPGe) detector. In all of these detectors, single- and doubleescape peaks are observed in addition to the full-energy peak
(photopeak) corresponding to the 'Y-ray energy. These escape
peaks are observed 0.511 MeV and 1.022 MeV below the
photopeak and are prominent and especially useful with highenergy gamma rays. The Nal scintillation counter is
conventionally used for profiling when high efficiency is needed
and energy resolution is not critical. This is normally the case
wilen Light elements are profiled in samples of high Z value
using well-isolated resonances. BGO detectors have higher
efficiency for a given size than NaI delectors and a larger
fraction of pulses in the full-energy peak. Furthermore, the BOO
detectors offer a bettcr signal-to-background ratio 3Ild a compact
setup, which are especially necessary when profiling with weak
resonances. HPGe detectors are generally used for the analysis
of total average concentrations of elements in the surface layer,
because these detectors have high enough energy resolution to
distinguish nearby gamma rays from competing reactions.
Before profiling, it is helpful to measure a gamma spectrum
from the sample with a Ge detector 10 identify .possible
interfering peaks from competing reactions or laboratory
background and to determine the suitable energy window
(region of interest) to sum the total yield. Improved efficiency
geometries have also been utilized leading to nearly 41t geometry
(piel et al., 1996).
:r7AI{p,,f'sI. 1m koV
~103
~
e
.:a 102
s:
10'
10'
0
10
12
Energy (MeV)
FIG. 7.5. Gamma-ray spectrum for the 992 keY "Al(p,y)"Si

resonance measured using a thin-film AI sample and a Ge
detector. Notice that each high-energy gamma ray gives three
peaks in the spectrum: the full-energy photopeak (PP), the single
escape (SE) peak, and the double escape (DE) peak. The
photopeak corresponds to the case when the full energy of the
gamma ray is absorbed in the detector. The escape peaks are
. _observed when on~ .or: both of the annihilation gamma rays (511
keV) escape from the detector.
.
--'.
-",
:::
:.... ,;..;....,---
The energy calibration and resolution of accelemtors is

usually determined utilizing the same narrow, strong resonances
that are used for profiling (see also Appendix 19). Tn the case of
protoo., the 991.90 0.04 keY 21 A1(P,r)13 Si resonance aod the 429.57
0.09 keV "N(p,cry)"C resonance are generally used. Tbe energy
distribution of the beam is assumed to be a G311ssian
Detedor and beam energy calibration

One standard procedure at the beginning of a set of
measurements is the calibration of the gamma-ray detection
system as well as the beam energy. The gamma-ray detection
system is typically calibrated for energy and efficiency using a
radio~ctive source for low-energy gamma rays and well-known
gamma-ray resonance transitions for high-energy gamma rays.
Suitable radioactive sources such as S7CO, mCs 6OCo 22~
l~~
207
10'1'
'
,
,
'Eu, Bi, and
I cover the range from 0.1 MeV to 3.0 MeV.
For higher gamma-ray energies, resonances such
the 992 keV
27 A1(P,'Y) 28 Si resonance and the 441 keY 7Li(p,y)8Bc resonance or
an Am-Be source can be used. Figure 7.5 shows a spectrum measured
for the 992 keV 27Al(p;yfSi resonance. The gamma rays that
correspond to the transitions from the states VYitb a very short
lifetime are emitted while the residual nuclei are Slowing down
in the sample. Therefore, these gamma peaks shift because of the
g(Eb,E)=[
(211:) cr b
]exP{-[(Eb-E)/crbll/2J, (7.6)
where Eb is the mean energy and crb is the standard deviation of

the energy distnoutioo. CJb can be measured, for example, by using
a resonance yield curve for a Urick sample, which will be treated
in Section 7.2.1.6.
as
7.2.1.4 Cautions
Factors affecting measurements
The main factors determining the efficiency of measurements
are the strength of the resonance, the beam current, and the
153
Chapter 7
gamma-ray detection efficiency. These factors can be expressed

in a simplified relation for the measured gamma-yield
y '" w(6)nItSIE"
The background of the measured gamma-ray yield curve is

mainly due to natural radioactivity in the surrounding materials
competing reactions caused by contaminants or othe:.
com:pon~ts of the sample, reactions in slits and apertures
collimatmg the beam, and noise in the electronics. Figure 7.6
shows a typical laboratory background measured with a Ge
de.te~tor (see also Section 7.2.2.4). The energy, intensity, and
ongm of the strongest peaks are indicated. The strongest peaks
(7.7)
v.:he~e e. is the efficiency of the detector, w(9) is the angular

dlstnbutIOn of the gamma radiation (9 is the angle between the
incident beam and the target---detector line), n is the solid angle,
I is the beam current, t is the measurement time, S is the
are due 10 2D'n (2614.6 keY) and <OJ( (1461 key), which are also
:esonance strength, and E( is the resonance energy. For a given

lSo~p.e, both the SIEr dependence and the amount of background
radiation favor the use of low bombarding energies.
often used for detector energy calibration. The natural

background occurs below 3 MeV (see Table A12.I8 in
. The beam current can usually be increased up to a few

~croamperes when the sample is cooled with water or liquid
rutrogen and a large beam spot is used. In some cases, it might
be necessary to move the beam spot during profiling because of
dete~o~tion of .the sample surface, for example, blistering and
e~fohatJOn at high fluences. The profiled element might also
mIgrate as a result of heating of the sample and radiationenhanced diffusion. Sometimes, the counting rate of the
detection system becomes a limiting factor with high beam
currents. When the high counting rate is due to very intense lowenergy ganuna rays
MeV) from reactions in the sample, lead
absorber plates can be used to decrease the intensity of the lowenergy gamma rays.
Because the effective solid angle of the detector varies

roughly as 1/(2, the sample-to-detector distance should be taken
in:o ~ccount in the measurement chamber design, detector
shl:ld.mg. and absorber attachment The intensity of gamma
radiation depends on e, the detection angle with respect to the
beam direction., as
w(O)= l+a,P,(cos O)+a.P.(cos 9)+higber-order terms
and gamma rays and on minimization' of the interferences, so

these factors are discussed next in more detail.
Appendix 12), and it can be significantly reduced by properly

shie~ding the detector and reducing the data collection time using
a higher beam current. Quite often, the desired yield curve

under~oes interference from resonances or continuous gammaray yIeld due to contaminants or other components of the
sample. Table A12.2 in Appendix 12 lists the strongest (p,y) and
(p,ay) resonances for proton energies below 3 MeV. For each
resonance, the table includes the energy of three gamma rays
that are of the greatest aid in identifying !be resonance, the total
cross section at the resonance peak in millibams (mb) [or the
integrated cross section in electronvolt bams (eV b)] or the
resonance strength in electronvolts (eV), and lhe full width at
half-maximum (fwhm) of the resonance in the laboratory system
of coordinates. In principle, all of thc resonances below and even
some broad resonances above the bombarding energy can cause
interfering gamma yield. The effect of contamination depends on
its distribution. If a contaminant is on the surface only. the
gamma yield due to the contaminant corresponds to resonances
n~ar ~e bombarding enc:rgy. On the other band, if the interfering
YIeld IS due to a contamInant uniformly distributed in the sample

or to a certain component of the sample, the interfering yield
corresponds to the thick-target gamma-ray yield, and steps in the
yield will OCCUI when the energy of a specific resonance of the
interfering isotope is achieved. Tables AI2.8a, A12.8b, and
(7.8)
Al2.ge in Appendix 12 list the gamma rays and yields obtained

from thick targets by proton bombardment. These tables aid in
the selection of a proper resonance for prOfiling, as well as in the
interpretation of possible interferences.
where P;(cos 9) is the Legendre polynomial of order i and

coeffi~ients a; .depend on the type of gamma transition. Usually,
the third term IS small, and a good average yield can be obtained
at the angles of 55 and 125, for which P2. -is:::: o.
The most common interfering resonances are those of

fluorine, nitrogen, and carbon. These elements have several
strong Tesonances. Some of those resonances are vet)' broad
(50-500 key), which makes them even more likely candidates
for interference, even if these elements are located in the surface
layer only. However, fortunately, the gamma-ray yield from the
resonance reactions of each of these elements corresponds to a
separate energy window (region of interest). The most prominent
gamma rays for fluorine resonances are in the energy range of
6.13-7.12 MeV, those for nitrogen resonances at 4.43 MeV, and
those for carbon resonances either at the low energies of 2-3
. Gen~ral1y, the detector efficiency decreases rapidly with

mcreaslOg gamma-ray energy. Therefore, low-energy gamma
ra?,s from a resonance reaction should be included in the energy
wmdow unless interfering gamma rays reduce the signal-tobackground ratio significantly.
Background and interferences

The reliability and accuracy of the measured depth profile
depend on the correct identifi.:ation of the interfering resonances
154
'}
changes significantly with increasing bombarding energy. In

these cases, the right background level and ttle correc,t
subtraction of interferences can be confinned by measuring a
few points of the distribution proflle using a Ge detector which
typically allows the characteristic gamma rays of the inctividual
resonances to be resolved. In some cases, it is possible to
determine the background and interferences by measuring the
yield curve from a similar sample but without the isotope of
interest or with a very thin layer containing this isotope. This is
often possible when ion-implanted profiles are measured.
Me V or around 9 Me V. The use of resonance tables and the

selection of the proper energy windows is explained with
specific examples in Section 7.2.1.9.
If the profile is limited to the surface layer, the gamma-ray

yield can be measured on both sides of the distribution, and
background subtraction can generally be done using a linear
approximation. In some cases, the right background level
(especially above the resonance) is not that obvious, for
example, when the measured profile extends beyond the
maximum analyzable depth or when the interfering gamma yield
104
lD
'"lD
~
~
."
III
N
\!!
"...m
103
"
"'"
~
I:
iO
lD "
t:!.g
W
<II N
::l
"
"C
Gi
>= 102
a-2O&n
101
b _ 214e1
c _228 Ac
d _ 212ei
,-.-~~-------
500
1500
1000
2000
2500
3000
Energy (keV)
FIG. 7.6. Typical laboratory background spectrum measured with an unshielded Ge detector. The origins of the strongest
peaks are indicated.
7.2.1.5 Concentration and depth scales

Let us assume that the background and possible interferences
have been subtracted from the measured yield curve. As
discussed in Section 7.2.1.1, the measured gamma-ray yield curve
gives a fairly good estimation of the actual concentration profile.
The gamma-ray yield versus bombarding energy curve can be
easily converted to a concentration versus depth profile. When
the gamma yield for element A from sample AmBo. (m + n = 1) is
compared to that of a reference standard containing a known
atomic fraction of element A, the difference in the stopping cross
sections of these two media must be taken into account. The
atomic fraction of the measured element in the sample can be
obtained from
(7.9)
where 5t represents the standard, A the measured element, and B

the rest of the sample atoms. Yst is the gamma-ray yield step
height of the resonance measured with the standard. The
stopping cross sections, e, of different atoms at the reSonance
energy can be calculated using, for example, the SRIM-2008
program (see Appendix 3). Stopping cross sections of
multicomponent materials are calculated using Bragg's rule,
EAmBn = mEA + nEB, unless experimental stopping cross sections
are available.
155
Chapter 7
The depth x required to slow down bombarding particles with

a mean energy Eb to an energy E' is given by
"'dE
= jS(E)'
(7.10)
where SeE) is the stopping power of the sample material. In

practice, the depths Xi corresponding to successively increasing
bombarding energies Ei can be obtained, for example, using the

simple approximation
x(E)= E, -E,
SeE,)
x(E) = x. +
,
,-I
2(E. -E. ,)
'
,-
SeE,) + S(E,_,) ,
(7.11)
i~2
Table 7.1. Example of the calculation of depth and concentration

scales for a NbN coating on steel.
Energy
(leoV)
Yield
(counts)
Depth
(nm)
Concentration
429.00
429.33
429.66
430.97
432.29
434.93
450.94
452.29
453.63
456.34
459.05
4B1.n
464.50
487.92
500
710
0
2.0
4.0
12.1
20.1
36.4
135.8
144.3
152.8
169.9
187.0
204.3
221.7
243.6
26.8
35.5
51.5
58.7
59.4
1190
1455
1485
1495
1510
1425
1200
945
185
25
10
0
(8l %)
59.7
60,0
57,9
51,8
43,9
11.1
1.61
0,65
where SeE;) is the stopping power of the sample material at

energy E i- This is calculated using Bragg's rule and taking into
account the atomic fraction m at this data point. Typically,' the .. 7.2.1.6 Standards
energy is in keY, S in keV/run, and depth in nm. However, if the
In analysis with' nuclear reactions, quantitative results are
density of the sample is not known, the depth is obtained in units
generally
obtained using standard samples containing the
of atomsicm 2 or nglcm 2,
element under consideration. Tbe standard is measured in the
same setup as the samples, and the yield is scaled with the total
Example 7.1. As an example, the simple formula in Eq. (7.11) is
collected charge of bombarding particles. In the case of
used to calculate concentration and depth scales for the measured
resonances, the resonance yield step height is detennined by
data given in Table 7.1. The nitrogen profile was measured for a
measuring the gamma-ray yield from the standard in the
NbN coating on steel using the 429 keY J~(p,(I"y)I2C resonance. A
bombarding energy range between E, - Sf aod E, +
There
thick TiN sample was used as the standard, and the measured yield
are several requirements for a good standard, such as
step was YIII = 1471. The standard was measured using the same
homogeneity, stability under beam heating, composition similar
settings as for the NbN samples. The stopping cross section values
to that of tbe samples, and not too many interfering resonances
needed in Eq. (7.9) are calculated at tbe resonance energy (429 keY)
or gamma-ray -yields. Some typical standards are listed in Table
and are E.N = 9.01. En = 17.64, and Es = Nb = 25.46 in units of
2
A12.4 in Appeodix 12. In many cases, high-purity elemental
eV/(lOls atJcm ). Therefore, the stopping cross section for TiN is
standards can be used, as well as stable, nonvolatile compounds,
l3.32 eV/(IOIS at./cm2). Then, for example, the concentration
such as oxides, carbides, nitrides, and carbonates. Standards can
corresponding to the first data point can be obtained using Eq. (7.6)
also be made by mixing high-purity powders in known ratios.
as
However, special care must be taken to ensure uniform mixing
and stability under irrn.diation; this limits the cboice to
0.50x500x25.46
anhydrous materials with a reasonably high melting point.
m ~ :-:-:::-:-:-::-::--'::-:'-'--=:-'-'-:"':--:--:----:-c:-c:Furthennore, surface roughness and even thin impurity films can
1471 x 13.32 + 0.5 x500 x (25.46 -9.01)
degrade depth resolution for a resonance step. In the case of
insuJators, it is necessary to use a thin cOnductive film, mask, or
or other neutralizing arrangement to prevent charging of the
grid
,,-(429 keY)
sample surface. High-purity heavy metals (Ta, Mo, etc.) are
suitable for this purpose.
sr.
~ 0.163
Measurement of a resonance yield cwve for a thick

homogeneous sample is typically used to deteIIIIine the exact
position of the resonance, the energy resolution at the surface,
and the calibration factor for conversion of measured yields 10
concentrations. Figure 7.7a illustrates this type of yield curve for
keVinm
Because the stopping power increases only slightly with the

energy, the depth corresponding to the second point is
X
a 992 keY
(429.33-429.00) keY =0.33/0.163 run =2.0 run

0.163 keY/run
27A1(P;y)
28
Si resonance measured from a bulk:
aluminum plate. The yield curve for a thjck sample is a

smoothed step function, with the resonance energy
corresponding to the point where the yield is one-half of the step
156
Particle-Induced Gamma Emission: P1GE
height. Figure 7.7b shows a Gaussian function and its integral.

These correspond to the energy distribution of the beam on the
surface and approximately the experimental yield curve for a
thick sample when the resonance is narrow compared to the
beam energy resolution (Fig. 7.7a). The energy interval
corresponding to the yield increase from 12% to 88% is the
experimental resolution (fwhm). The experimental beam
resolution was determined for the yield step in Fig. 7.la by

fitting the yield curve using the beam resolution as a 1itting
parameter. In this case, the simple procedure described onbove
would give the same result. However, when the resonance width
is larger, the thick-target yield curve is not an integrated
Gaussian because of the longer tail for the Breit-Wigner rc.TIllula
than for
Gaussian with the same value ofr.
the
0.'
05
..
0.'
02
990
Proton energy (keV)

992
994
996
0.'
998
1000
0.0
2.
*100
:(.
~
:
~A~.~I~
:!!
22
~:~~~::y
oL-~==='~'~'______~______~--1
OA
'"
1.
100
<a)
C.O
2M
Depth (nm)
(b)
0.'
...
18
(]
r."o..ubV
.,
1.0
.,
FIG. 7.7. (a) Experimental resonance yield step for the 992 keY 27 Al(P.y/8Si resonance. The sulid line is a fit obtained using the
Vavilov theory for straggling. (b) A Gaussian distribution and its integral.
of the analyzing magnet. The Doppler broadening follows a

7.2.1.7 Depth resolution and sensitivity
Gaussian law, and the width can be obtained from
The depth resolution is detennined by the energy resolution
of the beam, Db; Doppler broadening due to the vibration of the
target .atoms, flo; th~ reson~ce width, tr;'-and the energy-~- ~-=-,- ...... ~ (2M E
(7.13)
stragglmg of bombardmg particles as they penetrate the sample,
= 2.355
'b
,
n,,(x). The total depth resolution, d(x), can be obtained using
M,
the Gaussian approximation as
where M 1 and Ml are the mass numbers of the ion and target
nuclei, respectively; k is the Boltzmann constant; and T is the
temperature of the sample in kelvin. For a 1 MeV proton beam at
d(x) = n", (x)
(7.12)
room temperature, no is -100 eV. which is of the same order of
Sex)
magnitude as the widths of the narrowest proton resonances.
Because of the M11M2 factor, this term becomes more important
where n",(x) =[n,,' + no' + r.' + !1,.(X)')'I2.
when heavy ions are used to profile hydrogen. On the other
hand, no can be minimized by cooling the sample with liquid
This definition of resolution corresponds to the experimental . nitrogen.
width of an infinitely narrow distribution located at depth x,
where the full width at half-maximum of the energy distnbution
The energy straggling can be approJ<imated by the LindhardisOIIJ!'
Scharff model (Lindhard and Scharff. 1953) as
no
The energy resolution of the beam is typically in the range of

0.05-1.0 keY and can be improved, for example, by collimating
the beam and stabilizing the terminal voltage and magnetic field
-kT)V'
n",(x)= O.2942CZj ((Z,IM,)x[nm]r(gIcm']J"' [keY]

(7.14)
157
Chapter 7
with
measured yield curve using only Eq. (7.5). Extra conditions such
as smoothness, positivity, or boundary values are always needed.
fOflC<3
Because the detection sensitivity depends on several factors

experimental arrangement, gamma-my detection
efficiency, interfering gamma radiation, beam current,
measurement time, and substrate, the sensitivity is specific to a
given measurement. However, by taking into account the
resonance strength and the gamma-ray emitted, order-ofmagnitude estimates for the sensitivity of each resonance can be
calculated. Table AI2.S in Appendix 12 gives these estimates for
light elements. The estimates are for the total concentration of
element (unless a particular isotope is specified) and for the
strongest, narrow resonance of the element.
forl<~3
such as
L(l<) = 1.36K'12 -0.016K312
K = 4.0321 x 10-' Eb[keV]I(Z,M,)

where C is the correc:tion factor to Bohr's fonnulation; ZI and Ml
are the atomic and mass numbers, respectively, of bombarding
partides; and Z2 and M2 are the corresponding values for the
target. Although this approximation is supposed to be valid
typically at depths of a few hundred nanometers, it agrees fairly
well with experimental results even doser to the surface. See
Chapter 2 for additional discussion about energy straggling.
7.2.1.8 Supporting data analysis software

Several computer programs (Land et al., 1976; Maurel et ai.,
1982; Deconninck and van Oystaeyen 1983; Smulders. 1986;
Vickridgc and Amscl, 1990; Rickarcls, 1991; Landry and Schaaf.
2001) have been developed to extract an accurate depth profile
from the measured yield curve. Mostly applied methods use
either an iteration procedure starting from the measured curve as
the first approximation -or the fitting of the parameters of a
certain distribution function. The first approach is suitable for all
types of profiles. On the other hand, for example, ion-implanted
profiles can often be described with split-Gaussian, Pearson, or
spline functions, which are easy to fiL In both cases, a complete
profile can normally bc obtaincd in IC5s than 30 min.
Figure 7.8 illustrates the ueplh resolution of I MeV proton

and 2 MeV 4He+ beams in silicon and tantalum at. room
temperature as calculated using the approximation. given in Eq.
(7.14). nb and rr are assumed to be 0.1 keY for both beams. This
corresponds to profiling with very narrow resonances and good
beam resolution. The depth resolution with protons is 4 run and
I run at the surface in silicon and tantalum, respectively.
However, straggling dominates the resolution at larger depths,
even at depths of a few nanometers. Therefore, it is evident that
extreme beam energy resolution amI minimization of Doppler
broadening are necessary only in special surface studies with
narrow resonances, because the energy straggling of the beam
rapidly becomes the main factor affecting depth resolution
below the surface. The depth resolution with alpha-particle
resonance reactions is better mainly because of the relatively
large stopping power of "'He+ ions compared to protons.
Unfortunately, only a few reasonably strong and well-isolated
alpha resonanres exist (see Table A12.3 in Appendix 12)
The Wlfolding of the real profile from the experimental,

broadened gamma-ray yield curve is a typical deconvolution
problem. The main difference between various programs is the
calculation of beam energy straggling. The most accurate
method is the stochastic calculation of a discrete-energy-Ioss
proccss (Maurel et ai., 1982). This type of calculation is
necessary in high-resolution surface studies, for example, to
study the Lewis effect or the therlllal vibration of surface atoms
(Mitchell and Rolfs, 1991). When the proftle is measured with a
broad resonance (rr > 1 keY) or low beam energy resolution or
when the accurate profile shape at the outermost surface is not of
interest, straggling can be calculated accurately enough using the
Landau-Vavilov theory (Landau. 1944; V.vilov, 1957). The
values of these distributions can be obtained by interpolating
from tabulated values or by computer algorithms (Rotondi and
Montagna. 1990). When the resonance is hroad (r = 4 keV for I
Me V protons), the straggling distribution can be obtained using
a simple Gaussian approximation similar to that llsed for beam
resolution [Eq. (7.6)]
There are basically two ways to solve the deconvolution

problem: by least-squares fitting of a certain analytical function
describing the real profile or by iterative solution of thc
distribution using a suitable algorithm that takes into account
proper physical boundary conditions. These programs can easily
be used.to simulate the yield curve for a certain user-provided
concentration distribution. This type of user input is fairly easy.
However, it must be kept in mind that no matter what unfolding
method is used, a unique solution is never determined by the
Il8
600
r--------,--~---------,---.--~----,
500
51
E 400
c:
c:
o
,., 300
::s
Protons
T.
'0
Ul
CI)
a: 200
Helons
SI ././
/"
--
100
/"
Ta
Depth, nm
FIG. 7.8, Calculated depth resolution curves for 1 MeV protons and 2 MeV 4He+ ions.
--~.. ,.:....~
- ' --.-' ' - _-r_-"-"_' __ "-~_'''~_'-''~_ . ' . ~-~~-'. --------_
. _
Nevertheless. low-energy (P.y) resonances of Li (441 keV) and

liB (163 keY) have been used for profiling up to a few microns
with almost zero background because of the high energy of
gamma rays and the low bombarding energy (Toivanen et 01.,
1985). The 'Li(a,y)llB resonance at 950 keY offer> higher
resolution and has been utilized, for example, in the profiling of
lithium in silicon (Zinke-Allmang el aJ., 1986).
7.2.1.9 Depth profIling oflhe light elements

This section provides some typical profiling examples. The
best resonance choices, interferences, and some references are
given for each elemenl Additional data can be found in
Appendix 12.
Hydrogen and helium

Heavy-ion accelerators have been used with 7Li, I'N, and 1~
beams for profiling hydrogen and helIum (see Chapter 8). The
HCLi,y)B8e reaction offers sensitivity on the order of -1 ppm
and a profiling range up to several microns with a 3-7 MeY
beam (padawer el al. 1974; Adler and Schulte. 1978). 'Li beams
have also been used for helium profiling, using the 1.43 MeY
resonance of the 4HeCLi,y/ 1Be reaction (Scbulte, 1978).
Nitrogen and fluorine beams are used conventionally for highresolution hydrogen profiling in surface layers (Livingston et aI.,
2001; Kumar and Raju, 2004).
Carbon
The 1748 keY resonance of the iJC(p,y)i'N reaction is the
only narrow and strong resonance of carbon. However, it must

be kept in mind that, at such a high bombarding energy, most of
the light elements have several broad resonances, leading to high
background at low gamma-ray energies. and even in the region
of the high-energy primary gamma rays (9.17 MeV),
interferences are likely. For example, when detenninin,g 13C_
implanted profiles in silicides, the main interfering gamma-ray
yield originates from Si. This is demonstrated in Fig. 7.9, which
shows the gamma-ray spectra taken from DC-implanted T~ and a
pure Si target at the bombarding energy of 1750 keY. All Si
isotopes. 28Si, 29Si, and 30Si, have several resonances below 1750
keY. The energy range of gamma rays from each isotope is
Lithium, beryllium. and boron
These elements are generally profiled with nuclear reactions

by detection of charged particles. All (p;y) resonances are very
broad, limiting the resolution to a few tens of nanometers at best.
159
Chapter 7
shown. In this case, the energy window for the yield curve
measurement has to be restricted to include only the 9.17 MeV
gamma-ray photopeak because of the interfering gamma rays
from Si.
Nitrogen
Both 14N and 1~ isotopes can be profiled with several
resonances. Those used most often are the 429 and 898 keV
12
1'N(p,ay) C and 1059 keY I'N(P,r)"O resonances. The best
depth resolution can be obtained with alpha particles, using the
1531 keY 1"N(Ct,y)IBp resonance. However, generally, the 429
keY (p,ny) resonance offers the best sensitivity, a longer
profiling range, and good depth resolution. Because the use of
this resonance is one of the most popular applications, the
selection of a proper energy window is discussed briefly here.
At such a low bombarding energy, the main interferences are

generally due to fluorine, carbon, and 14N (Fig. 7.10). The use of
a Ge detector is often essential for identification of the source of
interference. The suitable energy window for nitrogen detection
is about 3.1-4.8 MeV, which is just above the natural
background (Fig. 7.6). Note that the energy window is selected
to take into account the poor energy resolution of the Nal
detector. Because the background due to fluorine includes the
same ganuna rays within a wide bombarding energy range,. the
ratio of gamma-ray yields within the energy ranges 5.0-6.2 MeV
and 3.1-4.8 MeV can be takcn as a constant This fact can be
exploited when subtracting the fluorine interference at various
bombarding energies. The same procedure can be used to
subtract the background due to 1'N resonances, as their gammaray energies fall mainly within the 5-7 MeV range. This
interference contribution becomes even more significant when
the bombarding energy is above 700 keY, which is typical, for
example, in the case of thick nitride coatings.
In thick nitride coatings, the main interfering yield comes
from ''N(P,r) and 1'N(P,r) reactions, along with a minor
interference from fluorine and carbon contamination. However,
these can be subtracted using various energy windows as
discussed above. The energy nmge 429-898 keY typically
aliows a profiling range up to -2.5 Jl1Il in nitride coatings.
Higher-energy resonances can be applied as well, although the
background then increases significantly. Foe example, the 1.65
MeV resonance can be used up to 6 JllIl, but overlapping with
the 1.21 and 1.98 MeV resonances must be taken into account
and subtracted.
Oxygen
Oxygen profiling is possible using the IBO(p,y) resonances
and enriched samples. All resonances have complicated decay
schemes, leading to many intense gamma-ray peaks. Table
A12.1 in Appendix 12 lists only the strongest, high-energy
gamma rays. All resonances decay through the low-energy
excited states of 19p, leading to 110 and 197 keV gamma rays
that might also be due to proton inelastic scattering from 19F.
The resonance at 1167 keY has been used for high-resolution
surface studies. Figure 7.11 shows 1BO profiles in 90010 180_
enriched TazOs layers (Maurel et aI., 1982). The prominent
surface peak is due to the Lewis effect, as discussed before
(Section 7.2.1.1, Figure 7.3).
10'
!
(a)
10'
Energy (MeV)
10'
(b)
10'
Energy (MeV)
10
FIG. 7.9. Measured gamma-ray spectra of (aJ a I'C-implanted Ta

sample and (b) a pure Si sample at the bombarding energy of
1750 keY.
Employing the 1~(p,(f:y)IZC reaction, detection limits up to a

few parts per million can be reached using a low-background
detection system and reasonably high beam currents.
160
300
.
~
19p(p,cq)'50
250 i-
"--'....
.,. 200
'2
::l
.e 150
~
>=
'2c(p,-r)13 N
C
"C
Qi
i-
.::l
100 -
50
J.~
2
(a)
tSN(p,ccr) 12c
1~(p,y)'SN
.
~
...,. . ,"
..dI.
I.A
.",.
I
I? ,T~Ir
~
8
Energy (MeV)
300
250
.__.. .,. 200
'E
::l
.e 150
-'"
(b)
"C
]i
> 100
N
50
Energy (MeV)
FIG. 7.10. Gamma-ray spectra for carbon, nitrogen, and fluorine measured using a 500 keV proton beam with (a) a Ge detector and
(b) a Nal detector.
161
Chapter 7
100
200
300
Channels
i7.A1 (P.~ 992 keY
Gz, =100 eV
Go = 101.5 eY
20
(~+D::.140eV)
r=105eV
XA\.o~ c
i/--
0,5,10,20,50,1 oo,2OCJA
..- ---
---
..I
~
(b)
,.
' ,.',..: ",
o L _ _ _ _~=-_=:.:._.._..___-_-_- _-L- _-_-_- =-.:-:.- :.-: ;-: ;~: ;-: ~:=-"_"~-:."': .~:.-, " ' =__
o
20
40
60
Channels
FIG. 7.11. (a) Excitation curves for 90% lBO_enricbed Ta20~ layers of increasing thickness X. The lines are best fits obtained using
stochastic simulations with the parameters given in the figure. (b) Calculated excitation curves for a thick aluminum sample covered
with an AI 20) layer of thickness X. The dotted lines represent the theoretical contributions of a 20-run-thick oxide layer and aD
aluminum sample covered by an absorber equivalent to that oxid.e layer (Maurel et a/., 1982).
"
162
,,
Fluon'ne
Fluorine is one of the most profiled elements, with (p,n'Y)
resonances and a typical interfering contaminant due to its many
strong and broad resonances. The most frequently employed
resonances are those at E, 340.5 and 872 keY. They mainly
decay emitting 6129 keY gamma rays, being typical for fluorine
contamination. Figure 7.12 illustrates the profiling of fluorine in
glass using the 340 keV resonance (Deconninck and van
Oystaeyen, 1983). The calibrated experimental resolution
function is plotted as a function of depth for the 340 keY

resonance and the closest higher-energy resonance of 484 ke V.
The background yield is very low, and profiles up to 1.5 lUll can
be measured without interferences. In this case, interferences
from the other fluorine resonances cannot be resolved using a Ge
detector, as all resonances decay through emission of the same
gamma rays. This is typical for all rearrangement-reaction
resonances.
(483.6 keV resonance)
0.3
>0
>.
"
iii
I:
Q)
0.2
.E
I:
:;::
1\1
'6
"''"
"
ffi
>
"
0.0
--=-ill
~
:>
"
0
"
>:!
Radiation emission
0.1
>
~
>
~
0.4
>
Glass surface
>
0.2
>
>
~
:=
'"
"
1\1
CC
0.0 I.L.......:"..L..-~c.......~~....lL...........l!...~d::............:::>.L......!.<:~....>..:_.......J
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1:6 1.8 2.0
Depth below surface (340 keV resonance)
(microns)
FIG. 7.12. Calculated yields initiated by the 340 keY resonance in fluoridated glass at different beam energies. The curve corresponds
to the experimental resolution. Interference from the next resonance at 483 ke V is also represented.
,...,,,,---,,,- -
r~~onanCe cPalt~~~-~; aI., 1983).
..
-"
In both cases, a Ge detector

should be applied, and because the resonances are rather weak,
high beam current is needed, making effective sample cooling
necessary.
Neon
In principle, the Urree stable neon isotopes can be profiled
with (P,'Y) resonances, but only the 2~e (natural abundance
90.51%) and 22Ne (9.22%) resonances are sufficiently Darrow
and well isolated. The 1169 keY '''Ne(p,y)21 Na resonance is
especially well isolated, but the relatively low-energy gamma
rays (3.54 MeV) involved are easily overlapped by interfering
gamma rays. However, the 22Ne(p,y)23 Na resonances with high-energy gamma rays can be used for high-accuracy profiling even
with light backings (Lappalamen, 1986).
Aluminum
Aluminum has only one stable isotope, "AI. which has at
least 22 (p,y) resonances below 1 MeV. They are all very narrow
< 200 eV), but none of them offers a wide energy range free
of other resonances both above and below the resonance. The
two strongest resonances are those at :p = 632 and 992 keY
given in Table AI2.! (Appendix 12). Both of these resonances
decay with primary high-energy gamma rays that can be well
separated from those due to other light elements such as fluorine,
nitrogen, and carbon. In tbe case of deep profiles (up to several
microns), interfering nearby 21A1(P.'Y) 28Si resonances can be
excluded using primary high-energy gamma rays and a Ge
detector. Also, the strong (S = 250 eV) and narrow (r < 200 eY)
resonance at 1684 keY of the 27Al(p,p''Y)21Al reaction has been
shown to provide fast and accurate depth profiling of aluminum
(Mateus e/ al., 2008).
cr
Sodium
Sodium distributions have been successfully profiled using

the lOll keY "Na(p,cry)"'Ne resonance (Delia Mea, 1986;
Trocellier ef al., 1982) and the 309 keY "Na(p,y)"Mg
resonance.
Magnesium
Magnesium can be profiled with the 1548 keY 2~g(p,yt' AI

resonance (Anttila el al., 1977) or the 2010 keY "Mg(p,p'y) 'Mg
163
Chapter 7
Silicon
Silicon has been profiled mainly using the 620 keY
JOSi(p,y}3Ip resonance. Profiles can be measured with high depth
resolution using a Nal detector and a 7.0--8.3 MeV energy
window. Several weak, low-energy resonances for profiling 29Si
have been employed, but doing so requires improved gamma-ray
detection efficiency (piel ef al., 1996).
Heavy elements
Elements heavier than silicon are seldom profiled with (P,1)
resonances. This is because these resonances are typically broad
and not well isolated and otCW at high bombarcting energies.
Phospborus and sulfur are the main exceptions, both having
narrow resonances for profiling. The (p,y) resonances of Ti, Cr,
Ni, and Fe can be used for profiling, but generally, it is
necessary to use a HPGe detector to measure primary
characteristic gamma rays.
7.2.2 Elemental analysis using particle-Induced gamma-ray
emission (PICE)
.
In principle, any ion could be used with proper bombarding
energy to induce gamma-ray emission. Protons and alpha
particles are mostly used, but some sparse measurements have
bcen made using light ions such as deuterons, tritons, 3 He , Li, B,
N, F, andCI.
The P1GE method is typically less sensitive than the similar

method utilizing X-ray detection (PIXE). However, the gammaray peaks are generally well isolated, and the energy is high
enough that correction for absorption is not necessary. In
addition, the high penetrability of gamma rays simplifies the
experimental arrangements and calculations. Generally, PIOE is
used with other complementary ion beam analysis methods such
as PIXE (partiCle-induced X-ray emission), BS (backscattering
spectrometrY), and NRA (nuclear reaction analysis with particle
detection).
The main improvements in the use of PIGE during the past
decade or so include large, high-resolution Ge detectors, external
beams, and the nuclear microprobe. External beams are now a
standard feature in PlOE and PIXE facilities. External beams
offer several advantages, such as the ability to measure organic
samples (even volatile) or large specimens (e.g., paintings); easy
sample preparation; effective cooling; and easy, reliable beam
current integration, even for insulators. Nuclear microprobes
(see Chapter 13) are mainly used with BS, PIXE. and other
methods that have high yields with low beam currents, but they
have occ~ionally been applied as well for analyses of light
elements using PIGE (Mosbab and DUJ1luci, 1997). PIGE has
also been employed in channeling studies of light-element
impurity-atom lattice-location determinations in crystalline
materials.
7.2.2.1 Thick-target gamma-ray yield

More or less systematic measurements of gamma-ray yield
versus bombarding energy under proton, deuteron, and alpba
bombardment have been made in several laboratories. Tables
AI2.Sa through AI2.13d in Appendix 12 list the absolute thicktarget gamma-ray yields for elemental analysis by protons,
deuterons, and alpha particles. Data for 7Li and heavier ions are
summarized in Tables A12.15, A12.16, and A12.17 (Appendix
12). The yields given in the tables are absolute gamma-ray yields
of the pure elemental samples and are given per solid angle
[steradian (sr) and total collected beam cbarge [microCoulomb
We)]. No error limits are given, however. The main source of
error is the stopping power. which can be as much as 20% in the
compound cases. On the other hand, these absolute yields are
seldom used for high-accuracy analysis because comparison
with a proper standard with a composition similar to that of the
sample is generally possible.
7.2.2.2 Stopping power correction

In the following treatment, thick targets (those with a
thickness exceeding the bombarding particle range) are assumed,
and absorption effects of the r-rays are not taken into account.
Also, straggling effects are ignored and smooth cross-section
curves without resonance structures are assumed. Thc gamrnaray yield from a thick target can be approximated as
E,
Y, =ru:f..; fa,(E)ISm(E) dE,
(7.15)
o
with i referring to the measured nuclide and m to the matrix. n is
the number of bombarding particles, E is the detection efficiency
including the solid angle, F" is the incident ion energy, fwi is the
weight fraction of nuclide i, 8m is the stopping power for the
matrix, and oJE) is the cross section for the specific reaction. In
addition, we assume thaI the nuclide to be measured is evenly
spread through the matrix.
General guidelines
The aim is to formulate an equation that enab1es estimation of
concentrations using thick-target gamma-ray yields of pure
elements. The main concern is finding a procedure that accounts
for the relationship between stopping power and gamma-ray yie1d
as a function of bombarding energy. The following procedure
follows the guidelines of Kenny ef al. (1980). A similar procedure
has also been presented, for example, by Ishii ef al. (1978a,b,c).
Using Eq. (7.15), the formulation for the thick-mget gammaray
yield for nuclide i in matrix m is obtained as
Y, =f,.;[Y,...,(Ep) s,(Ep)/Sm(Ep)l- fY"",(E)
~~~
dE
(7.16)
where Y PlIIl' is the pure-element yield under the same
experimental conditions, and SlE) is the stopping power for the
164
I
.1
pure element. The exact solution can be obtained numerically,

but unfortunately, the pure-eiement yields are generally not
accurately known as a function of bombarding energy, as would
be required for accurate analyses. Therefore, for practical
reasons, the use of simplified general guidelines has been
suggested:
= Y B Y;" SA (E) f
wB
yB
A
pure
S (E)
wA'
(7.19) ,
Example 7.2 illustrates the use of this procedure.
Example 7.2. In a PIOE measurement, peaks for bor<:>n and

magnesium were Doted in the spectrum.. The peak areas for the
429 keY, 975 keY, and 1014 keY gamma-ray peaks were 800
(7.17)
counts, 1000 counts, and 1500 counts, respcctively. A detection

angle of 55<1 was employed in the measurement, and the en.ergy of
the bombarding protons was 3.1 MeV. A cahbratioo standard set
was prepared for boron, and tile obtained boron concentration in
the sample was 10 ppm by weight. Using Eq. (7.19) and talang
advantage of the yield values (Y,=) of Table At2.8a in
Appendix 12, the following concentration values for Mg and Al
are obtained:
(7.18)
Here, the energy E1I1 is the energy at which the yield for the
pure-element target bas decreased to one-half of its initial value.
that is, Y 1Mt:(E1I2) = 1/2 Y pure{Ep). The difference in the yield
values calculated from Eq. (7.16) compared to those obtained
using the geoeral guidelines [Eqs. (7.17) and (7.18)] was studied
by Kenny ef 01. (1980). They found that the guidelines provided
values within 1.5% of the values obtained using Eq. (7.16).
IWlcts 72xlif N,<lOrr' 9.62MV,(rrgan-')10

IUlcts 12xlo' N,<lOr)-< 8.47MV,(rrgcrn-') JPIl""85JPI1
Use ofstandards
l.Wcts 72x]if N,U.cst 9.62 MV,(rrgcm')
Concentrations of elements distributed homogeneously in
c ~
]O~35JPI1
thick samples can be obtained by comparison to standards. Many
lillds 4.6x10' N,(j.Cs-t RISMV,(rrgan-')
.
multicomponent elemental standards exist for the analysis of
geological, biological, and medical samples [e.g., National
lrutitute of Standards and Technology (NISn, Institute for
The relevant stopping power values have been taken from SRlMReference Materials and Measurements (lRMM), International
2008. Note for possible peak overlapping.
Atomic Energy Agency (lAEA)J. In many cases, researchers can
prepare the proper standards themselves, by mixing a known -_ ..
amount of the element to be determined into a matrix similar to
. The general guideline has been applied successfully, Jor
the one being studied. For example, the detennination of carbon,
example, in the determination of Na, Mg, AL P, aDd CI in
nitrogen, and oxygeo can be carried out with organic compouods
geological samples (Olabanji et ai" 1996). It should be pointed
as standards. The known stoichiometry for C, N, and 0 of thc
out that, for accurate analysis, the relevant pure-element yields
compounds can be used directly. As a restriction preparing one's
should be measured using the appropriate geometrical setup (see
own standards, the seleoted organic compounds should not have
also Section 7.2.2.4). For an order-of-magnitude resul~ the
literature pure-element yield values can be applied.
very high concentrations of any single element (above about 40
wt %). This procedure is a consequence of the similarity of the
stopping powers for typical organic samples and the compounds
7.2.2.3 Experimental considerations
used as standards (Lapatto and Riiisiiaen, 1988).
Several variations of the experimental arrangements used for
PIGE analyses have been reported. The trend has been to
The stopping power values can be obtained, for example, using
construct multipurpose chambers for simultaneous PIGE, PIXE,
the SRlM-2008 program (www.SRlM.org). If the e~citation curve
and RBS measurements. When low concentrations are
is known as a :function of the bombarding energy, more accurate
determined,
minimal system background is essential. In choosing
calculation can be made using a simple computer program,
the construction and lining materials for the experimental setup,
although this is not generally necessary. The general guidelines
the thick-target gamma-ray yield tables of Appendix 12 can be
presented in Eqs. (7.17) and (7.18) can be employed, which, in
employed. It should be noted that the best materials of choice are
twn. simplifies the standardization procedure. In this case, the
not the same for PIXE and PIGE.
concentration of only one element (element A) in the unknown
_0
sample material must be detennined by use of a standard. The

concentrations for all other elements (elements B) can then be
obtained using the general guideline
_ _ _ _ _ _ _ _ __
In principle, there are two possibilities for experimental

arrangements: invacuum measurements and external-beam
measurements.
165
Chapter 7
In-vacuum measurements
In all experimental considerations, careful shielding and
lining are needed owing to particle scattering behavior. Tantalum
is commonly used for shielding from scattered particles. With
tantalum, low-energy gamma-ray lines occur, so better materials
are tin and lead. Tin is ideal when the proton energy is below
....2.5 MeV, and lead can be used at higher energies. Lead is often
used as an absorber in front of the detector to reduce the count
rate and improve the sensitivity. Copper is often used as a
construction material because it is a good conductor of heat. A
drawback of copper, however, is its contribution of spurious
peaks to the ganuna spectra.
External-beam measurements
External-beam PIGE can be applied to the analyses of
sensitive materials (e.g., paintings). Special attention should also
be paid to the exit-foil material. Polyimide (H"C"N,O,)
(Kaptan) is commonly used in external-beam PIXE
measurements, and it can also be employed for PIGE analyses.
The induced background is low but interference can result from
the C, N, and O. Also, exit-foil cooling might be required in some
cases. With organic exit foils and ions heavier than protons and
deuterons, a cautious approach should be adopted because of the
large energy losses involved, which can lead to rapid foil rupture.
In PIGE, high beam currents are often used, and therefore,
metallic exit foils, such as havar (a high-strength alloy), titanium,
or nicke~ are more coDvenient This is a.q important point to be
considered, especially when carbon, nitrogen, and oxygen are
analyzed The gamma-ray peaks originating from the exit foil can
be used to nonnalize the beam current of various measurements
(Raisanen, 1986). The gamma-my yields are generally
sufficiently high for such pwposes but not so high that they
provide a disruptive background signal In some cases, PIXE and
..... PIGE measurements are done simultaneously with the same
setup, although the best sensitivity is not then achieved for both
methods. A multipurpose external-beam PlXE-PIGE-RBS setup
is depicted in Fig. 7.13.
geometry, possible peak: broadening (see below) should also be

considered.. The effect of angular distribution on thick-target
yield for the 'Li(p,p'y) reaction has been studied by Aslam et al.
(2002).
7.2.2.4 Cautions
It should be noted that, when perfonning PIGE measurements
with ion energies close to the energies of strong resonances,
caution should be taken because a small change in the
bombarding energy can lead to significant changes in the yields.
Some additional points that should be kept in mind are discussed
in this section.
Geometry
Generally, one can expect thick-target gamma-ray yields to be
isotropic. However, when the yield is mainly due to strong
resonances, noticeable deviations from isotropic distributions can
be observed (see, for example, Kenny et aJ., 1980). In such cases,
the angular distributions should be taken into account if the
general guidelines arc employed. When fixing the measurement
166
FIG. 7.13. The exit foil set-up is positioned at I, the X-ray

detector is at 3, the particle detector is at 4, the HPGe detector is
at 5, the detector for beam current monitoring is at 6, and 7 is the
video camern. (Adapted from Calligaro et al., 2004.)
Background and interferences
Interfering background gamma rays can be categorized into
two groups, laboratory background and prompt background. Very
often intensity ratios of gamma-ray peaks must be compared in
order to accurately subtract the background contribution. The
laboratory background must be determined with the same
detector as used in the actual analy:les (see also Fig. 7.6). It
should be remembered that the total background can vary with
both laboratory and experimental setup. Significant background
can also originate from other sources, sucb as activated slits and
similar beamline equipment. This contnbution varies as a
function of time. Prompt background has two sources: the
detector and the experimental setup. The most important origin of
detector-induced background peaks is the neutrons produced in
the sample. In the case of 7-9 MeV protons, relative neutron
yields for several pure elements and compounds are given in
Table AI2.8d (Appendix 12). These fast neutrons induce
reactions with the Ge isotopes of the detector crystal and create a
background that is difficult to avoid without loss of detection
sensitivity. The neutron flux to the detector can be reduced
significantly with a thick neutron absorber, for example, lO-cmthick boron-doped paraffin, plastics, or water. The neutroninduced peaks are significantly broader than nonnal gamma-ray
peaks and are mainly in the low-energy region I MeV). Most
typical of these are the 596 and 691 keV peaks. Also, reactions
with Fe, AI, Cu, and B present in the detector eud cap occur, and
the resulting gamma-ray peaks are visible in the measured

spectra. The resulting interferences in the spectra depend on the
target material and the ion type and energy. It should be noted
that use of deuterons as' bombarding particles results in severe
neutron yields. Typical background gamma-ray lines are listed in
Table A12,18 (Appendix 12),
detection of the specific element. On the other hand, sensitivities

can be much worse if the sample contains significant amounts of
other elements with high gamma-ray yields or if the sample can
withstand only low beam currents. Using special experimental
arrangeml!nts, such as a pulsed particle beam for background
suppression, allows improved sensitivity to be obtained
(Gurbich, 1997), Also, the application of a so-called clover-GoBGO detector system for elemental analyses using PI GE has
proven advantageous (Elekes et ai., 1999), improving detection
limits by a factor of 2-3. This setup consists of a clover-type
composite HPGe detector surrounded by a BGO anti-Compton
shield. The detection sensitivity has been also improved by
measuring gamma-n coincidences (Ene el al., 2001), instead of
standard PIGE. The limiting factor, however, is the n.eed for
more complex equipment and increased measuring times.
Peak broadening
In some cases, the gamma-ray peaks are significantly broader
than the resolution of an HPGe detector. When gamma rays are
emitted while the residual nucleus is moving, a shift in the
ganuna-ray energy is observed. This broadening is mainly due to
the short lifetimes of de-excitation states and the high recoil of
the light nucleus (Doppler effect) or to a large resonance width.
The Doppler effect is especially significant in the case of light
target nuclei. The Doppler-shifted energy of gamma rays is
determined by the velocity of the recoiling nucleus relative to
the observer according to the expression (up to second order)
(7.20)
where is the angle between the beam direction and the sampledetector line, v is the velocity of the excited nucleus, c is the
velocity of light in a vacuum, and E~ is the energy of the
gamma ray if the nucleus decays at rest.
The recoil nucleus must decay during flight before stopping.
In the case of solid materials, the lifetime of the excited state
should be between 5 x 10- 15 and 2 x 10-12 S in order to cause
Doppler broadening. The shape of the Doppler-shifted gammaray peak depends on the lifetime of the excited state and the
stopping power of the retarding medium. In cases where the
products of the average stopping power and mean lifetime are
the same, similar line shapes are obtained. The effect is most
prominent at 0 and is not observable at 90. Peak broadening is
evident with protons for Ey values of 478 keY (Li), 3562 keY
(Be), 429 keV (B), 4429 keV (N), and 1634 keY (Na) and with
'He+ ions for E, values of 478 keY (Lil, 4439 keV(Be), 3684
keY (B), 1275 keY (F), and 1809 keY (Na). The peak
broadening is useful in identifying the origin of the gamma rays,
especially for elements having only one gamma-ray line.
7.2.2.6 Elemental analysis of the light elements by PIGE

Systematic thick-target gamma-ray yields have been tabulated
in the literature for protons, deuterons, tritons, alpha particles,
7Li, nC, 14N, 160, and J.5CI ions. The available references are
provided in Table 12.6. Also, the possible inclusion of an atlas
of relevant gamma-ray spectra is indicated when available. The
most useful charged-particle-induced reactions for the elemental
analysis of light elements are listed in the tables of Appendix 12.
In situations when different isotopes can be detected, the
relevant reactions are provided even if their yields are not high.
The yields for low-energy 300 keY) gamma rays tend to be
more inaccurate, e.g., due to self-absorption in the target
material. For these reasons, the very low energy lines are usually
ex.eluded from the tables. It should be noted that, when pureelement targets were not available, targets in compound form
. were applied. For such elements) the error limits arc higher
owing to the needed stopping power correction: In the following
subsections, various particle-induced gamma-ray yields are
slUVeyed from the point of View of elemental analyses. For
practical elemental analyses, it is often equally important to have
knowledge of the elements exhibiting low gamma-ray yields,
which are discussed at the end of each section.
Proton-induced y-ray emission

In most PIGE studies, protons have been used as bombarding
particles, The light elements are discussed keeping in mind their
analyses by the PIGE method. The relevant reactions can be
found in the tables of Appendix 12. Analysis of hydrogen is
presented in Chapter 8 of this handbook.
7.2.2.5 Detection sensitivity

The sensitivity of the PIGE method depends on several factors:
reaction cross section. beam current, irradiation time, detection
system, and sample matrix. Sensitivities achievable with various
bombarding particles are summarized in Tables A12.9, AI2.12,
A12,14, and Al2.17 (Appendix 12), These should be taken as
order-of-magnitude values only because the experimental
arrangements and bombarding energy are not optimized for the
Lithium has two stable isotopes 5Li (7.5%) and 7Li (92.5%).
'Li can be easily detected using the 478 keY y-ray line, which
is slightly Doppler-broadened. For 6Li analysis, no usable y-rayemitting reactions exist. When analyzing samples with high
lithium concentrations, one should keep in mind that neutrons
are easily emitted.
167
";- 'i
Chaptel7
Carbon has two stable isotopes, IlC (98.9%) and IlC (1.1 %).
Detection of carbon at low proton energies is Dot sensitive. A
good sensitivity for "c is obtained at higber bombarding
energies employing the 12C(P,p'r)12C reaction and the 4439 keY
line. At energies below the strong 5.370 MeV resonance, the
sensitivity is poor, and above about 8 Me VI the background
usually increases significantly. The optimum energy region is
thus between 7.5 MeV and 8.0 MeV. In some cases, nitrogen can
cause interference (sec below). l3e can be determined with
moderate sensitivity at low bombarding energies.
Nitrogen has two stable isotopes, ''N (99.63%) and "N (0.37%).
The most significant 1-ray line originating from "N at low
proton energies is the 4439 keV line due to the 1~(p,ar)12C
reaction. The Doppler-broadened peak in the spectrum can be
easily distinguished, but the obtained sensitivity is moderate.

The optimwn proton energy is close to 2 MeV (Hanninen and
Raisanen, 1984), and the obtained detection limit about 0.2
wl %. At higher proton energies the sensitivity for nitrof.en
analysis can be increased significantly using the 1"N(p,p',,{) ~
reaction and the 2313 ke V line. Because of the strong resonance
at 3.903 MeV, the optimum energy for this reaction is 4.0-4.5
MeV. At higher energies, the neutron yield increases, which
results in a higher background and poorer sensitivity.
Sodium has one stable isotope, 2~a.

A good sensitivity for sodium analysis is obtained by
detecting the 440 keY y-rays. The y-ray line at 1636 keY is
somewhat Doppler-broadened.
Magnesium has three stable isotopes, 24Mg (78.99%), 2SMg
(10.00%), and "Mg (11.01%).
Magnesium can be analyzed with a relatively good
sensitivity. At low proton energies, the most prominent lines are
at 390 keY, 585 keY, 975 keY (from "Mg), and 1369 keY (from'
24Mg). Unfortunately, the 585 keV line has interference due to
background from ~, which interferes in long measurements
and when the magnesium concentration is low. It should be
noted that the 1369 keV line also originates from aluminium. In
principle, with proton energies above 3 MeV, all magnesium
isotopes can be detected.
Aluminium has one stable isotope, 21Al.
Aluminium has several strong r-ray lines at 844 ke V, 1014
keV, 1369 keV, and 1779 keY. The 1369 keY line migbt have
interference from sample magnesium. The 1779 ke V line may
contain interference from sample silicon and phosphorus. Also
the 844 keY and 1014 keV lines which are most suitable for
aluminium analysis have interference from magnesium,
21
26 Mg (p;y) A1, but the magnesium yields, are fortunately rather
low,
Oxygen has three stable isotopes, "0 (99.76%), 17 0 (0.038%),

and "0 (0.20"10).
Unfortunately, in general, the analysis of oxygen by PIGE is
not sensitive. In principle, the various isotopes can be
determined, but not with good sensitivity. At low bombardmg
energies, 160 can be detected by the 495 keY y-rays resulting
from the 160(p;y)17F reaction. A significantly better sensitivity
for 160 detection is obtained at energies above 7 MeV using the
"O(p,p'y)"O reaction and the 6129 keY line. The optimum
energy region for this reaction is about 7.5-8.5 MeV. Possible
interference due to sample fluorine and the resulting y-rays from
the 19p(p,ar)160 reaction should be properly taken into account.
Silicon has three stable isotopes, UiSi (92.23%), 29Si (4.67%),

and lOSi (3.10%).
Silicon can be analyzed by PIGE with moderate sensitivity.
The "{-ray line at 1779 keY, from the 28Si~,p'y)28Sj reaction, has
interference resulting from the "Al(p,y) 'Si and " P(p,a1)"Si
reactions. The 1273 keV line is most useful for silicon analysis.
All silicon isotopes can be deterrriined employing >3 MeV
protons.
Phosphorus has one stable isotope, lip.
In the analysis of phosphorus, several interfering and
overlapping r-ray lines and reaction channels occur. The 1266
keY line, from the J'p(p,p'yi'p reactions, overlaps with lines
originating from silicon and sulfur, lOSi(p,y)llp and 34S(p,uy)3Ip,
respectively. The 2233 keV r-ray line does not have overlapping
and is safest to use. Unfortunately. it has a rather low yield.
Fluorine has only one stable isotope, 19p.

Fluorine detection can be carried out by using either the 6129
keY "{-rays from the 19p'(p,cry)160 reaction or the 110 keV and
197 keV "{-rays from the l~(p,p'y)l9J' reaction. The 110 keV and
197 keY y-!ays offer a better sensitivity (by a Ilictor of about 5)
(Shroy e/ aI" 1978), but the detector should face the front side of
the sample to avoid absorption effects. This is a disadvantage,
because for multielemental analysis, an absorber in front of the
detector is often needed to reduce the proportion of low-energy
r-rays. The detector solid angle is also smaller compared with
the situation where the detector is positioned immediately
behind the sample. The interference of the l80(P,r)1~ reaction in
fluorine analysis by the 19F(p,p'''{i9F reaction was studied by
Grambok el 0/, (2000).
Sulfor bas four stable isotopes, "5 (95.02%), "5 (0.75%), "s
(4.21%), and "5 (0.02%).
Sulfur can be detected employing the reaction 32S(p,p'r)12S
(E, = 2230 keV) at the level of 100 ppm (by weigbt) from thick
organic samples and at the 100 nglcm2 level from aerosol
samples (Raisanen and Lapatto, 1988). The optimum
bombarding energy is about 4.9 MeV. The reaction shows two
strong resonances at energies of 4.77 MeV and 5.10 MeV. At
168
Partide-Induced Gamma Emission: PIGE
low bombarding energies, the analysis of sulfur by PIGE is

impractical. In analyses of sulfur, care should be taken to
compensate for the possible overlapping "(-ray peaks in the
spectra. Such are possible from the elements of phosphorus
[2230 keY 3Ip(p,y)"S; 2233 keY llp~,p'y)llpJ, silicon [2233
keY lOSi(p,y)'IP; 2235 keY lOSi(P.p'y) Sil, and chlorine [2230
keY )~Cl(p,ny)32S]. Fortunately, these elements have more
dominating lines at other energies that can be used for checking
their presence in samples. For sulfur analysis, PIXE is more
sensitive, but because of overlapping X-ray peaks and absorption
effects, PIGE is often a useful complementary method to FIXE.
Manganese has ODe stable isotope, sSMn.

Manganese bas a significant line at 931 keY. .At the
bombarding energy of 5.5 MeV, the line at 1408 keY is also
useful for analytical purposes (Badica el al., 1996).
Iron has four stable isotopes, S4Fe (5,9%), s6pe (91.72~), s7Fe
(2.1%), and "Fe (0.28%).
.
Iron has a significant line at 847 lee V that can be used for
practical purposes. It is often found as a background peak in
PIGE spectra due to scattering from stainless steel materials.
Unfortunately, the 847 keY line often overlaps with the strong
844 keY line from aluminium.
Chlorine has two stable isotopes, "Cl (75.77%) and "Cl

(24.23%).
The strongest line from chlorine (at proton energies above 3
MeV) is the 1219 keY line, wbich is due to the "Cl(p,p'y)"CI
reaction ..The same 1'-rays also originate from the 34S{P;yiSCI
reaction. The 1763 keY line does not experience overlapping.
Cobalt has one stable isotope, 59CO.

Cobalt bas two distinct lines at 339 keY and 1332 keV. The
overlapping with nickel should be noted.
Nickel has five stable isotopes, "Ni (68.077%), "Ni (26.233%),

'INi (1.140%). 61N i (3.634%), and "Ni (0.926%).
The strongest y-ray line that is useful for nickel analysis is at
1332 keV. Possible overlapping with cobalt should be noted.
Potassium has three stable isotopes, 3~ (93.2581 %), ~

(0.0117%). and 41K (6.7302%).
The analysis of potassium is most feasible with 2168 keVyrays. In geneml, the PlOE method is not sensitive fqf potassium
analyses. It should be noted that neutrons are emitted from
potassium samples even at low proton energies.
Copper bas two stable isotopes, 6lCu (69.17%) and "Cu

(30.83%).
Copper can be best detected by the 992 keV r-ray line at low
proton energies and by the 962 keY line at high energies.
Calcium has six stable isotopes, "Ca (96.941%), "Ca (0.647%),

"Ca (0.135%). "Ca (2.086%), "Ca (0.004%), and "Ca
(0.187%).
The most intense y-ray line at low bombarding energies is at
373 keY. Otherwise, the same remarks as for potassium are
valid, including neutr~D emission at 10:-'" protop energies.
Zinc has five stable isotopes, 64Zn (48.6%), 66 Zn (27.9%), 67Zn

(4.1%), "Zn (18.8%), and 70Zn (0.6%).
Zinc has a prominent r-ray line at 360 keY. The various
isotopes can be determined with moderate sensitivity; see Table
A12.7 (Appendix 12).
Scandium has one stable isotope, 4~SC.
In analyses of thin targets the resonance structw'es of the

excitation curves should be taken into account. In Boni et al.
(1988), useful thin-target excitation curves for PIGE analysis of
Li, B, F, Mg, AI, Si, and P arc given. The data are provided for
proton energies from 2.2 MeV to 3.8 MeV at a measurement
angle of 90".
Several ,,(-ray lines originate from scandium. The presence of

scandium in ordinary samples is generally very rare, and its
analysis is not generally of significant importance.
Titanium has five stable isotopes, .46.yi (8.0%), 47Ti (7.3%), 4'Ii
(73.8%), "Ti (55%), and "1'i (5.4%).
In some applications (for example, in the analysis of painting
pigments), it is useful to determine the transition elements by
PIGE. Titanium can be analyzed using the 983 keY y-ray line.
To obtain quantitative infonnation on the sensitivity of the

method, typical detection limits obtained under practical
conditions (-30-min measurement) for biomedical and organic
samples are provided for selected elements in Table A12.9
(Appendix 12). Sensitivity is deflDed in this context as the
minimum detectable concentration. The minimum detectable
peak is asswned to be three times the square root of the
background at the fwhm of the peak. For Li, B. C, N, 0, and S.
the E,. values are the optimum proton energies for the analyses,
and the detection limits stated are the best values obtainable by
the technique. As can be noted, the optimum proton energies for
the analyses of all elements do not coincide. The detection limit
values depend strongly on the sample composition, espedally on
Vanadium has two stable isotopes, 50y (0.25%) and .sly

(99.75%).
Vanadium can be analyzed employing the "y-ray lines at 319
keY and 749 keY. The 319 keY y-ray line overlaps with a line
originating from zinc.
Chromium has four stable isotoPes, soCr (4.345%), Her

(83.79%), <lCr (9.50%), and "'Cr (2.365%).
Chromium has a prominent line at 379 keV. For the analysis
of the different isotopes, see Table A12.7 in Appendix 12.
169
1
Chapter 7
DTGE can also be used for S, el, and K analyses. The detection
limits obtainable depend on the sample and the background
the concentrations of the elements for which the method

detection sensitivity is good (e.g., Li, B, Na, and F).
caused by neutron-induced y-rays in the detector.
Low-yield elements:
The y-ray yields of the heavy elements are generally rather
low. Several such elements are employed as low-background
materials for construction purposes. With proton energies below
3 MeV, the best low-yield elements are tin (no l-ray lines) and
Low-yield elements:
The element, of Fe, Ni,
ell, Zn,
So, To, W, and Pb can be
recommended as low-yield backing or construction materials

when the deuteron energies are below 1.8 MeV (Kiss et ai.,
lead (disadvantage of lead is the intense Pb KX-ray yield;
1994). At energies above 3 MeV, only To, W, and Pb should be
otherwise, no y-ray lines). Tantalum is also a good candidate,
employed. The use of aluminium in measuring chambers (AI is

frequently used in setups constructed for simultaneous analyses
by PIXE) ,hould be avoided if deuterons are employed.
although the rather strong y-ray lines at 136 keV, 165 keV, and
302 keV increase the background stightly.
Triton-induced y-ray emission

Only one systematit;; study on triton-induced thick-target y-ray
yields can be found in the literature (Borderie and Barrandon,
1978). in that study, 2.0 MeV, 3.0 MeV, and 3.5 MeV tritons
were employed for selected elements. The best sensitivity values
faT relevant elements are collected in Table A12.l2 (App<:ndix
Deuteron-induced y-ray emission

Two systematic studies of deuteron-induced thick-target y-ray
yields have been presented in the literature (Kiss et al., 1994;
Elekes et al., 2000). The studied elements include Z = 3-20 in

the energy range of 0.7-3.4 MeV. In employing deuteroninduced y-ray emission (DIGE), the possibility of an associated
neutron hazard should be kept in mind. For example, this
demands careful beam alignment:
.
12).
Low-yield elements:
The best low-yield materials are the heavy elements; for
example, Mo, Pt, and Au can be recommended.
The most intense y-ray lines of the light elements following

1.8 MeV deuteron bombardment, along with the nuclear
reactions involved, are listed in Table AI2.l0 (Appendix 12)
The selected energy of 1.8 MeV is ideal for analyzing samples
containing high oxygen concentrations, as the ener@: is just
a-particle-induced y-ray emission

The available systematic studies involving ,,(-ray yields for a
particles are provided in Tables AI2.13a-<: (Appendix 12). The
light-element r-ray yields induced by 2.4 MeV a-particles are
provided by Lappalainen el 01. (1983). A selection of the highest
below the 1.828 MeV neutron threshold of the 'O(d,n)I7F

reaction. At this energy, clearly the best peak-ta-background
ratios are obtained
l-ray yields induced by 3.5 MeV a-particles is given by Borderie

The limits of detection for the transition elements in steel by
5 MeV deuterons were determined by Peisach el at. (1994). The
reactions yielding the best detection limits for the transition
elements are provided in Table A12.1! (Appendix 12). The
and Barrandon (1978). The binned thick-target gamma-ray

yields for Be, B, N, F, No, Mg, and AI inducad by 5.6 MeV, 6.3
MeV, 7.0 MeV, 7.7 MeV, 8.8 MeV, and 10 MeV alpha-particles
are provided by Heaton et al. (1995), In Giles and Peisach
(1979), the spectra obtained by 5 MeV a-particle bombardment
of the light elements are presented together with the relative
- values correspond to the collected charge of 1 mC.

The total cross section of the 14N(d,P'Y)I~ nuclear reaction
for analytical applications has been detennined in the energy
region of 50()'-1500 keV (van Bebber el 01., 1998). The gammaray production cross-section curves for the 6Li(d,py)'Li,
'Be(d,ny)''lJ, "B(d,py) 12B, "O(d,py) 170, and "F(d,py)"'F
reactions in the energy range 0[0.6-2 MeV were determined by
Sziki et al. (2006). As an example, the high-precision detection
16
Of 0 in high-critical-temperature (high-Te) superconductors by
1.8 MeV deuterons was presented by Vickridge el 01. (1994).
Oxygen was determined employing the 160 (d,py) 170 reaction
intensities and net counts/collected charge. The absolute gammaray yields for Ti through Zn (excluding V) induced by.5 and 10
MeV a-particles are given in Kocsonya et al. (2006).
It should be noted that. with a-particle energies below 5 MeV,
no ')'-ray lines originate from carbon. The elements that are easily
dete<:table by u-particle bombardment are Li, Be, B, and F.
When comparing the yields obtained by 2.4 MeV a-particles
with those obtained by 2.4 MeV protons, it can be concluded
that the a-particle-induced yields are higher only for beryllium.
and detecting the trompt 871 keV y-rays emitted from the first
For beIYllium, the sensitivity obtained by a-particle-induced rray emission (sometimes referred to as AlGE) is about 1000
excited state of I O. The obtained uncertainty for the ana1yses

was less than 1%.
time, better than that obtained by protons. AlOE could be of
specific interest in the anfllysis of Li, Be, B, and F from samples

that are rich with Na, Mg, or A1 (for which AlGE is not
sensitive). Such analyses ,by protons are not sensitive. Additional
advantages of the use of a-particles are the low neub'on yields
The DIGE method is sensitive for C, N, and 0 detennination

by 1.8 MeV deuterons. It is also rather sensitive for the detection
ofNa and Si. Ifproton energies above 4 MeV are not available,
170
and the simplicity of the gamma spectra. BecaUse of the short

lifetime of the de-exciting state and the high recoil of the
nucleus, a clear Doppler broadening occurs in several r-ray
lines. The broadening is useful in identifying the origin of peaks
in gamma spectra. Generally, for several elements, the r-ray
energies involved are rather low. This can cause problems with
absorption.
.
Low-yield elements:
Of the light elements, carbon can be used as a low-yield
material. Of the heavier elements from Co and Ni, no y-rays
originate. Also, the yields for Y, Nb, In, and Sn are sufficiently
low for these materials to be employed as construction materials.
7Li-induced y-ray emission
Energetic 1Li ions are sometimes used in RBS measurements
for substituting 4 He ions. Lithium-ion-induced y-ray emission
can be used in connection with other ion beam analytical
techniques for simultaneous detection of the light elements.
Gamma-ray yields induced by 12 MeV and 18 MeV 7Li ions are
provided by Raisanen (1990). Elemental analysis using 7Li ions
is surveyed in Table AI2.l5 (Appendix 12).
Although the y-ray yields increase with increasing particle
energy (from 12 MeV to 18 MeV), the background in the spectra
are also elevated. As a result, the minimum detectable amount of
an element is not necessarily lower for increased particle energy.
The exact optimum bombarding energy depends on the matrix
and the element to be determined. A limitation on the use of 1Li
ions is that Li, Be, and B cannot be detected, as no significant rray lines originate from these .e)ements.
Low-yield elements:
..... ,..........- .
Lead is a good low-yield material to be used with 7Li ions.
Because of its rather high neutron yield, carbon is not usable as a
construction material in experimental setups when 7Li ions are
employed. Tungsten, hafuium and tantahun are good backing
materials, for example, for nuclear physics experiments with 7Li
beams. No significant y-ray lines originate either from
zirconium.
Heavy-ion-induced y-ray emission
He, UN, and J60 ions: Backscattering spectrometry (BS) and
elastic recoil detection analysis (ERDA) can be carried out with
16
12e, I"N, and 0 ions. In such dete~ations. the simultaneous
we of the ion-induced y-ray emission technique can be
exploited. Absolute y-ray yields induced by 22 MeV and 28
MeV 12C ions, 28 MeV 14N ions, and 28 MeV and 33 MeV 160
ions for the light elements are available in the literature (Seppala
et aI., 1998). The bombarding energies are well above the
Coulomb barriers of the light elements (2 < 9), for which several
strong y-ray lines are reported. The highest yields obtained by all
ions are for beryllium. The yields induced by 12C, l"N, and 160
ions can be compared ,with the values obtained by other ions.
With 12C, '~. and 160 ions, the values are systematically higher
for beryllium, carbon, nitrogen, and oxygen than for 2.4 MeV
protons (equivalent energy/nucleon). The "(-ray yields induced
by 2.4 MeV a-particles are systematically lower for all elements
except for beryllium. It can be coI1cluded that the 12C, '""N, and
16
0 ion-induced yields are sufficiently high for p:ractical
detection of carbon, nitrogen, and oxy~en. The relevant reactions
and yields for analyses with 12e, ~, and 160 bea::m.s are
provided in Table AI2.l6 (Appendix 12).
Low-yield elements:
The yields are low for elements heavier than titanium., and all
such elements can be used as construction materials. Also, the
neutron yields induced by the heavy ions are low.
JJCI ions
The usability of 55 MeV 3sCI+7 ions for elemental analyses
was demonstrated by Borderie el al. (1979). For all other
elements except helium, lithium, and beryllium, 55 MeV is
below the Coulomb barrier. The method was found suitable also
for analyses of heavier elements in special cases. The y-ray
yields and detection limits obtainable from a niobium backing
are listed in Table A12.17 (Appendix 12). The detection limit
values refer to one hour of irradiation with a beam current of 0.5
!lA.
Low-yield elements:
As expected, heavy elements are the best low-yield materials.
In all cases, the reactions are Coulomb excitation, and the
involved y-ray energies are low.
7.2.2.7 Advantages and limitations of PIGE

An advantage of PIGE is the high penetrability of gamma
rays, thus diminishing matrix effects. The detector can be
positioned immediately behind the sample, providing a
maximum solid angle for improved sensitivity. The ex.perimental
setups are simple, and with external beams, the usability of the
technique can be enhanced. Volatile materials can be analyzed,
and sample preparation is simple. As a rule, very small amounts
of sample material are needed. Also, surface topography does
not significantly influence the determinations. The technique is
nondestructive and fast. Using PIGE in combination with other
ion beam techniques, nearly all elements can be detected
simultaneously. On the other hand, PIGE can be used only for
the analysis of selected isotopes, and its sensitivity for many
elements is only moderate. Because the optimal conditions, such
as the bombarding particles and their energy, depend on the
matrix and the isotope to be detected, no universal "best"
conditions and physical parameter choices can be provided.
Mateus et al. (2005) developed a code for the quantitative
analysis of light elements by PIGE. In this approach. standards
are not needed, as the known nuclear reaction excitation
functions are integrated. Naturally, this approach relies on
accwate excitation functions and stopping power corrections.
171
.,';
Chapter 7
REFERENCES
Adler, P.N., and Schulte, RL. (1978), Scripta Metall. 12,669.
Hunt, S.E., and Jones, W.M. (1953), Phys. Rev. 89, 1283.
Ishii, K., Sastri, C.S., Va11adon, M., Borderie, B., and Debrun,
lL. (1978a), Nuel.Instrum. Methods 153, 507.
Amsel, G., d'Artemare, E., and Girard, E. (1983), Nucl.Instrum.

Methods 205, 5.
Ishii, K., Valladon, M., Sastri, C.S., and Debrun, 1.L. (l978b),
Nucl.Instrum. Methods 153, 503.
Anttila, A., Bister, M., Fontell, A., and Winterbon, K.B. (1977),
Radiat. EfJ. 33, 13.
Aslam, Prestwich, W.V., and McNeill, F.E.
Radioanal. Nuel. Chern. 254, 533.
Ishii, K., ValladoD, M., and Debrun, J.L. (1978c), Nucl. Instrum.
Methods 150, 213.
(2002), J.
Kenny, M.J., Bird, J.R., and Clayton, E. (1980), Nucl. Instrum.

Methods 168, 115.
Badica, T., Besliu, C., Ene, A" Olariu, A" and Popescu, I.
(1996), Nuel.Instrum. Methods B111, 321.
Kiss, A.z., Biron, I., Calligaro, T., and Salomon, J. (1994), N.el.
Boni, c., Cereda, E., Marcazzan Braga, G.M., aod DeTomasi, V.

(1988), Nuel.Instrum. Methods 835, 80.
Kocsonya, A, Szokefalvi-Nagy, Z., Torri, A., Raubala, E., and

Rliisanen, 1. (2006), Nuel. Instrum. Methods 8251, 367.
Borderie, B., and Barrandon, J.N. (1978), Nuel. Instrum.

Methods 156,483.
Kumar, S., and Raju, V.S. (2004), Nuel. Instrum. Methods B226,
631.
Borderie. B., Barrandon. IN., Delaunay. B., and Basutcu, M.

(1979), Nucl. Instrum. Methods 163, 441.
Land, DJ., Simons, D.G., Brennan, J.G., and Brown, M.D.
Butler, l.W. (1959). NRL Report 5282, U.S. Naval Research
(1976), "Unfolding Tecliniques for the Determination of
Laboratory, Washington, DC.
Distribution Profiles from Resonance Reaction GrnmmaRay Yields", in Ion Beam Surface Layer Analysis, Meyer,
Calligaro, T., Dren, J.-C., Salomon, J. t and Walter, Ph. (2004),
0., Linker, G., and Klippeler, F. (eds.), Plenum, New York,
Nuel. Instrum. Methods 8226, 29.
yol.2, p. 851.
Deconninck, G., and van Oystaeyen, B. (1983), Nuel. Instrum.

Methods 218, 165.
Landau, L. (1944), J. Phys. USSR 8,201.

Landry, F., and Schaaf, P. (2001), Nuel. Instrum. Methods B179,
Delia Mea, G. (1986), Nuel. Instrum. Methods B15, 495.
Elekes, Z., Kiss,
262.
A.z., Gyilrky, Gy., SomoIjai, E., and Uzonyi, I.
Lapano, R., and Raisanen, 1. (1988), Phys. Med. BioI. 33, 75.
(1999), Nuel. Instrum. Methods 8158, 209.
Lappa1ainen, R., Anttila, A, and Rliisanen, 1 (1983), Nuel

Elekes, Z., Kiss, A.Z., Biron, I., Calligaro, T., and Salomon, 1.
(2000), Nucl. Instrum. Methods BI68, 305.
Lappalainen, R. (1986), Phys. Rev. B34, 3076.
Ene, A., Badica, T., Olariu, A., Popescu, LV., and Besliu, C.
Lindhard, J., and Scharff, M. (1953), Mat. Fys. Medd K Dan.

Vidensk. Selsk. 27 (15).
Fujimoto, F. (1990), Nuel. Instrum. Methods 845, 49.
Livingston, RA., Schweilzer, 1.S., Rolfs, C., Betker, H.W., and
Giles, I., and Peisacb, M,'(1979), J. Radioanal. Chern. 50, 307.
Kubsky, S. (2001),J. Mater. Res. 16,687.
Golicheff, I., Loeuillet, M., and Engelinann, Ch. (1972), I.

Radioanal. Chern. 12, 233.
Mateus, R., Jesus, A.P., and Ribeiro, J.P. (2005), Nuel. Instrum.
Grambok. D., Neelmeijer, C., Knoll, K.., and Herrmann, F.
Mateus, R., Fonseca, M., Jesus, A.P., Luis, H., and Ribeiro, J.P.
Methods 8229, 302.
(2000), Nucl. Instrum. Methods BI61-163, 269.
Gurbich, A.F. (1997), Nuel. Instrum. Methods B129, 429.
Maurel, B., Amsel, G., and Nadai, J.P. (1982), Nuel. Instrum.
Methods 197, I.
Hanninen, R., and Rliisanen, 1. (1984), Nuc!. instrum. Methods

B5,43.
Meier, J.H., and Richter, F. W. (1990), Nuel. Instrum. Methods

847,303.
Heaton, R.K.., Lee, H.W., Robertso, B.C., Norman, E.B., Lesko,

K.T., and Sur, B. (1995), Nuc!. Instrum. Methods A364,
Mitchell, G.E., and Rolfs, C. (1991), Nuel. Instrum. Methods

B56-57,473.
317.
172
Mosbah, M., and Duraud, I.-P. (1997), Nuc/. Instrum. Methods

8130,182.
Shroy, RE., Kraner, H.W., lones, K.W., lacobson, I.S., and

Helier, L.l. (1978), Nuel. Instrum. Methods 149, 313.
Olabanji, S.D., Haque, A.M.L, Zandolin, S., Ajayi, R T., Buoso,
Smulders, P.IM. (1986),Nucl. Illstrum. Methods B14, 234.
M.C., Ceccato, D., and Cherubini, R. (1996), Nucl. Instrum.

Me/hods 8109-110, 262.
SRIM2008 (2008), Ziegler, I.F., private communication (U.S.

Naval Academy, Aaoapolis, MIl) (www.srim.org):
Padawer, GM., Kamynowski, E.A., Stauber, M.e., and

D'Agostino, M.D. (1974), Geochim. Cosmochim. Acta,
Suppl. 5-2, 1919.
Szoo, G.A.., Simon, A, Szikszai, Z., Kertesz, Zs., and Dobos, E.
Palt,moo, R, Raisanen, I" Hautala, M., and Aattila, A. (1983),

Nucl. Instrum. Methods 218,785.
Toivanen, R.O., Hirvonen, l-P., and Lindroos, V.K. (1985),

N.el. Instrum. Methods B7-8, 200.
Peisach, M., Pineda, c.A., and Pillay, A.E. (1994), Nuc!.

Ins/rum. Methods B85, 142.
Tweenier, P., Nens, B" and Engelmann, Ch. (1982), Nuel.

Instrum. Methods 197,15.
Piel, N., Scbulte, W.H., Berheide, M., Becker, H.W., Borucki,
van Bebber. H.. Borucki, L., Farzin, K., Kiss, I\.Z., and Schulte,
(2006), Nncl. Instrum. Methods 8251, 343,
W.H. (1998), Nucl.Instrum. Methods 8136'-138, 72.
L., Grama, c., Mehrhoff, M., .and Rolfs, C. (1996), Nuc!.

Instrum. Methods B1l8, 186.
Vavilov, P,V. (1957), SOY. Phys. JETP 5, 749.
Riiisanen, I. (1986),Nucl.Instrum. Methods 817, 344.
Vickridge, I., and Amsel, G. (1990), Nuel. Instrum. Methods

B45,6.
Raisanen, 1, and Lapatto, R (1988), Nud. Instntm. Methods

B30, 90.
Raisanen, I. (1990), Nucl.Instrum. Methods A299, 387.
Vickridge, I.C., Talioo, j" and Preslaod, M. (1994), Nuel.

Instrum Methods 885, 95.
Rickards, I. (1991), Nuc!. IrlStrum. Methods B56'-57, 812.
Vickridge, I.e., and Arnsel, G. (1996), Nucl. Instrum. Methods

BI08,403.
Rotondi, A., and Montagna, P. (1990), Nud. I",trum. Methods

B47,215.
Zinke.AlImang, M., Kassler, V., and Kalbitzer, S. (1986), N.eI.

Schulte, R.L. (1978), Nncl.Instrum. Methods 149, 65.

Seppalii, A" RaISanen; I., Raubala, E., and Sz6k~fal;i::-NagY,Z.
(1998), Nud. Instrum. Methods B143, 233: ..."-'
.. ' .".,
173
NUCLEAR REACTIONS FOR

HYDROGEN ANALYSIS
w. A. Lanford
State University of New York at Albany, Albany, New York, USA
CONTENTS
8.1
INTRODUCTION .................................................................................................................... 176
8.2 REACTIONS USED TO PROBE FOR 'H ............................................................................ 176

8.3
RESONANT NUCLEAR REACTiONS ....................................... ,........................................ 176
8.4 EXPERIMENTAL DET AILS ................................................................................................. 177

8.4.1 Beam current integration ........... ;.......:... ;.........: ..............:.;.. :;.-......:......:.;............;........ 177 ... c.- co:. -..
8.4.2 Beam geometry ..........................................:............................:..............:...................... 177
.
8.4.3 H loss during analysis ................................................................................................. 178
8.4.4 Backgrounds ................................................................................................................ 178
8.4.5 Warning for those running N'+ beam from a single-ended machine ..................... 179
8.4.6 Warning for those running N beams from a tandem accelerator ........................... 179
8.5
ANALYSIS OF RESONANT NUCLEAR REACTION DATA ......................................... 179

8.5.1 Chamber calibration: Determination ofK ............................................................... 181
8.5.2 Energy loss values ....................................................................................................... 182
8.5.3 Depth resolution .......................................................................................................... 182
8.5.4 Off-resonance contributions ....................................................................................... 182
8.5.5 High-sensitivity measurements .................................................................................. 183
8.6 SELF-CONSISTENT ANALYSIS COMBINING NRA AND RBSDATA: TIlE NEED FOR
NRA DATA FOR PROPER RBS ANALYSIS ......................................................................... 184
8.7 SUMMARY AND CHECKLIST OF NUCLEAR REACTION
PROFILlt"IG PROCEDURES AND CONCERNS ................................................................ 185
REFERENCES
........................................................................................................................... 186
175
Chapter 8
8.1 INTRODUCTION
Megaelectronvolt ion beam analysis for hydrogen is
important for the following reasons:
Hydrogen is probably the most common elemental
contaminant, especially in thin~film materials.
The presence of hydrogen can have dramatic effects on the
electrical, mechanical, and chemical properties of some
materials.
Hydrogen is invisible in most modem analytical probes.
resolution, maximwn depth of profile, interference reactions,

and ease of data analysis. Another consideration is case
(expense) of producing the necessary heavy-ion beam.
Perhaps the first point to make is that essentially only
resonant nuclear reactions have been used to profile for IH. One
reason for this is that (as outlined in the next section), resonant
nuclear reactions provide a natural way to measure depth
profiles by measuring the nuclear reaction yield as a function of
beam energy. Depth information could be obtained using
nonresonant nuc1ear reactions by measuring the energy of the
cbarged particles produced in the nuclear reactions. However, in
many (most?) cases, the kinematics is very unfavorable if these
particles are detected at backward angles. It often turns out that
the energy of the detected particle (because of the large centerof-mass velocity) is low and almost independent of the depth at
which this particle is produced. If the detector is placed at
forward angles where particle energies are higher, one might just
as weU look directly at the elastically recoiling protons. This is a
useful approach that is referred to as elastic recoil detection
(ERD) and is discussed in Chapter 5.
In addition, hydrogen is chemicaUy versatile, having valences

of both +1 and -1; hence, it can react with most elements. For
example, in semiconductors, hydrogen can passify both
acceptors and donors. Further, hydrogen is many orders of
magnitude more mobile than other common contaminants. This
means that, if hydrogen is present in a system even at room
temperature, it has the mobility to reach critical regions in short
times.
Because of this siruation, rnegaelectronvoIt ion beam
techniques have been developed to detect hydrogen.. These
include nuclear reaction techniques (Leicb and Tombrello, 1973,
and many others) as discussed in this chapter, and elastic recoil
detection methods (Cohen et at, 1972; L'Ecuyer et 01., 1976;
and many others) as discussed in Chapter 5.
Of
The three resonant reactions that have been used far more
than other.; arc those induced by "N (Lanford el ul., 1976), IOF
(Leich and Tombrello, 1973), and 'Li (Adler ef 01. 1974).
Parameters for each of these reactions are given in Table A13.1
of Appendix 13 (Trocellier and Engelmann, 1986; Xiong er 01.,
1987; Barnes el 01., 1977). Each of these reactions has its own
advantages. For most applications, the I~ reaction has the
advantage of having the best combination of analytic
characteristics (depth resolution and sensitivity). The 19p
reaction bas the advantage that it can be conducted using natural
(as opposed to isotopically enriched) F in the accelerator ion
source.' The 1Lj reaction has the advantage of allowing profiling
to much greater depths in a sample than either of the other
reactions.
To put this field in some perspective, over the past 10 years

so, hundreds of papers have been publiShed in which
megaelectronvolt ion beams were used to measure H

concentration profiles in a vast variety of materials, including
electronic materials (amorphous Si, diamond, silicon nitride.
boron nitride, silicon dioxide, etc.), metals (diffusion of H in
metals, electromigration of H in metals, stress-induced transport
. of H in metals, and others), glasses (weathering of glasses.
obsidian hydration, solar wind implants on lunar glasses, and
many more), and superconductors.
In the following sections, the experimental details for using

the 1~ resonant nuclear reaction to determine H concentration
profiles are discusse4 in detail. Although there are some
differences, ill general, the same considerations apply in the use
of the I'F and 7Li reactions.
The purpose of this chapter is to provide practical handbook

infonnation that can be used to allow someone to quickly set up
a H profiling system and to analyze data taken with such a
system. This puzpose is to be contrasted with that of another
review written by the present author (Lanford, 1992), which
emphasizes the general considerations in hydrogen analysis but
does not present the useful tables or other practical information
given here.
8.3 RESONM'T NUCLEAR REACTIONS

Figure 8.1 shows a schematic representation of the 1~
resonant nuclear reaction profiling method. This method makes
use of the nuclear reaction:
8.2 REACTIONS USED TO PROBE FOR 'H

In principal, any proton-induced nuclear reaction can be used
"in reverse" to probe for IH by bombarding the sample to be
analyzed with a beam of the appropriate heavy ions and
measuring the number of nuclear reactions that occur with
hydrogen in the target. However, some reactions provide better
analytic cha.ractc.ristics than others, in tenus of sensitivity, depth
UN + IH -. 12C + 4He + gamma-ray.

This reaction has a large cross section at the resonance energy
(6.385 MeV I~ laboratory energy), with the cross section only a
few kiloelectronvolts away being four orders of magnitude
176
r
I
Nuclear Reactions for Hydrogen Analysis
smaller (Horn and Lanford, 1988) To use this reaction for

analysts, the sample to be analyzed IS bombarded with 15N with
an energy at or above the resonant energy. and the number of
characteristic gamma rays produced in the target is measured
""ith a scintillation detector. When the sample is bombarded with
1'N at the resonance energy, the gamma-ray yield is proportional
to the amount of H on the surface of the sample. When the
sample is bombarded with l~ above the resonance energy, there
are negligible reactions with surface H atoms because the energy
is above the resonance energy. However, as the 15N ions
penetrate the sample, they lose energy and rea.ch the resonance
energy at some depth. Now, the gamma-ray yield is proportional
to the amount of H at this depth. Hence, by measuring the
gamma-ray yield as a function of beam energy, the H
concentration as a function of depth is deteITIlined.
/~
"WindOii!
Beam
The chamber shown in Fig. 8.2 also has a filament to

neutralize cbarge on insulating samples. This filament is a lowvoltage, low-power light bulb with the glass broken off and the
filament coated to lower the work function of the tungsten. The
coating wed in this case is a commercial product made by
General Electric for this purpose. The filament is powered by a
battery and is electrically enclosed in the Faraday cage. Hence,
turning the filamenl on for insulating samples has no effect on.
the beam current integration system. This procedure allows
analysis of insulators or conductors with equal ease.
--75A
/ / /
8.4.1 Beam current integration

Because this technique relies on making an absolute
measurement of ganuna-ray yield per incident ion., it is
important to integrate beam current carefully. Most importantly,
this means suppressing any electrons that might enter or leave
the Faraday cage. In the case of the chamber shown in Fig. 8.2,
in addition to the -300 V electrostatic suppression (shown),
there is also (not shown) a permanent magnet mounted near the
insulated coupling in the beamline. This combination of electric
and magnetic suppression is very effective.
It is also important that the vacuum conditions he good
enough that no charge exchange occurs between the ion beam
and the residual gas'in the bearnline. Otherwise, the charge
carried to the Faraday cage by the t~ ions might not be the
same as that selected by the analyzing magnet, and the
relationship between the integrated beam current and the number
of incident ions would no longer be known.
Hydrogen
detection
resonance
bismuth germinate (EGO) crystalllneated about 2 em bebind the

sample (see Chapter 16). The following subsections discus's
some points that should be considered.
6.385 :t 0.005 MeV
ii;}? !
For a more general discussioll of beam current integration,

see Chapter 15, Pitfalls in Ion Beam Analysis.
8.4.2 Beam geometry

It is also important to make certain that the incident beam hits
the sample-and only the sample-once it enters the Faraday
cage. There can be no beam-defining slits near the target, or
hydrogen on these slits will cause reactions that will be detected
by the gamma-ray detector, thereby causing a false signal. In the
case of the setup shown in Fig. 8.2, the beam is viewed on a
scintillator (a glass microscope cover slide in one of the sample
positions) from the back (gamma-ray detector removed), and the
ion beam is focused and steered in such a way that the beam will
cleanly hit the sample to be analyzed when the sample is rotated
into the beam.
FIG. 8.1. A schematic representation of the l~ resonant nuclear

reaction profiling technique. The lSN bombards the sample and
reacts with H within a "resonance detection window" at a depth
x in the target Measuring the reaction yield as a function of
beam energy detennines the H concentration as a function of
depth.
SA EXPERIMENTAL DETAILS
Figure 8.2 shows a schematic of the experimental chamber
used for H profiling at the State University of New York
(SUNY) at Albany, NY. The 1'N beam enters from the right and
bombards the samples, which are mounted on the circumference
of a rotatable sample wheel. The gamma rays from the reaction
are detected by a scintillation detector [a NaI crystal or a
The chamber discussed in this section bas been in operation

at SUNY-Albany for more than 20 years, and any possible
changes in detection efficiency (or beam current integration
efficiency) have been monitored over this long period by repeat
177
Chapter 8
practice, this is not a serious problem. Although a small amount

of frost might form on the samples, it pwnps off in a few
minutes following evacuation.
runs on the same sample. Remarkably, although a smal1

deterioration in BGO detector resolution has been obscIVed, the
overall efficiency of this system has remained constant to within
a few percent over this long period.
In summary, bydrogen loss during analysis can be an

important issue, and whenever analysis 1S being conducted on a
new material, it is essential to test for hydrogen Joss. In a recent
analysis of SiCyHl' made by plasma deposition, an initial
measwement (with relatively intense beam) showed that -2030'10 of !be hydrogen could be lost during that analysis.
However, by limiting beam intensity and total beam dose
(consistent with good counting statistical accuracy), three repeat
measurements showed relative hydrogen contents of
1.000:0.998:0.973; thus, with care, hydrogen loss can be
rendered completely negligible for these materials
Rotatable
Photomultiplier
valve
There are also situations in which it might be important to

measure the hydrogen contents of samples when the samples are
held in a hydrogen gas. For example, some metals take up
hydrogen from the gas phase but will give off this hydrogen
when placed in a vacuwn. With differential pumping, it is
possible to carry out this analysis with the samples held in H2
gas at pressures up to a few Torr (Horn e/ al., 1987).
sample
wheel
Samples
FIG. 8.2. A schematic of the chamber used for 15N hydro:en

profiling at the State University of New York at Albany. The ~
beam bombards samples mounted on the circumference of the
sample wheel. A gamma-ray detector is mounted outside the
vacuum about 2 cm behind the samples. The chamber is a
carefully designed Faraday cage with no slits near the targets.
The filament provides electrons to stop insulating samples from
charging in the beam.
8.4.4 Backgrounds
The large scintillation detectors used to measure the gammaray yield for NRA are also sensitive to other sources of
radiation, most notably cosmic rays and natural radioisotopes
(1.46 MeV gamma rays from the decay of"'K and 2.61 MeV
gamma rays from the decay of
to 203Pb). In some
situations, background radiation can also be produced by the
accelerator. When measuring H at low levels, it is important to
measure background carefully. This can be d.one, for example,
by recording the gamma-ray counts from a "H-free" sample such
as a clean Si wafer that has a very low level of intrinsic
hydrogen.
mu
8.4.3 H loss during analysis

Some materials lose H during analysis. The loss can be very
rapid (e.g., CR,) or negligible (e.g., TiH,), depending on the
nature of the material. One of the great advantages of nuclear
reaction analysis (NRA) for H is that it is easy to repeat
measurements to see if the H content is changing during the
analysis. Because of the possibility ofhydrogen loss, it is useful
to be able to spread the analyzing beam over a large area on the
sample. The setup shown in Fig. 8.2, for example, has a beam
raster system that allows the beam (after it is focused) to be
rastered over a rectangular area up to 1 em or more on a side.
With such a system, it is possible to measure even very delicate
materials, such as CH2 , with negligible loss during a run.
In a few situations, background nuclear reactions between the

incident I~ ions and isotopes in the target can also occur. In
genem1, this is not expected to be a serious problem because the
15N beam is far below the Coulomb barrier for every element
except hydrogen. However, it is not below the Coulomb barrier
for deuteriwn, and a background reaction is observed for
deuterated samples. One reaction that occurs is D + I~ --. l~ +
H. The "N beta-decays (7.2-, balf-life) to 1'0 with
accompanying high-.mergy (6.13 MeV, 7.11 MeV, and 8.87
MeV) gamma rays. The rate of this reaction can be monitored
because it is not a resonance reaction, so it varies smoothly "With
beam energy. Hence, by measuring the background as a function
of energy from below the resonance energy to above it, the
background yield can be estimated. Also, this background
reaction produces some gamma rays of higher energy than the
l~ gamma rays. Hence, this background can be monitored by
measuring the number of gamma rays above 4.4 MeV. Care
For some materials. hydrogen can be lost simply when they

are placed in a vacuum. An example of this is hydrated glasses
that lose water if placed in vacuum at room temperature. For
such materials, freezing the samples to -30C or -40 C before
evacuation (and keeping them cold during analysis) can
effectively stop this hydrogen loss (Schnatter ef aI., 1988). In
such a procedure, one has to be cOllcemed that atmospheric
water will condense on the samples before evacuation. In
178
'f
should be taken to note that this background is not a prompt

gamma-ray emitter, but rallier, the gamma rays are delayed by
the beta decay of "N (7.2-s half-life).
The other background reaction that has been observed with a
ISN beam is with Li in the sample. Above about 7 MeV incident

energy, high-energy gamma rays are produced by reactions
between 1~ and Li in the target.
Detailed investigations of the reactions between 1~ and
deuterium and Li do not seem to be available in the literature.
Fortunately, neither deuteriwn nor Li is commonly present in
of plasma-deposited amorphous Si on a crystal Si substrate.

These plots show both raw data (i.e., gamma-ray coun1:s as a
function of beam energy) and fmal results (H concentration as a
function of depth). The convcrsion of the raw data to final
results involves multiplying the two raw data scales (iUC.es) by
constants, as outlined below.
The stopping power, dEJdx, for I~ in most materials is
essentially constant for beam energies betvreen 6.4 and lOMe V,
the energies of interest for NRA profiling. Consequently, the
conversion of the beam energy scale to a depth scale (x) is
accomplished simply with the relation
E - Eres
samples, so this is usually not a problem. However, one needs to

be aware of this possible problem and to take precautions when
these backgrounds might be present.

8.4.5 Warning for those running a Nl+ beamfrom a single-
ended machine
When running I5N 2+ beams from a single-ended accelerator
without a system in the terminal that selects the charge-to-mass
ratio of the particle being accelerated, there will likely also be a
one-quaner-energy "contaminant" beam. This beam comes from
the acceleration ofN 2+ ions that break up after acceleration, but
before the magnetic analysis system. This one-quarter-energy
beam and the full-energy beam follow identical trajectories in a
magnet field (they have the same value of MElq'). Separating
this beam from the full-energy beam requires electrostatic
analysis (e.g., deflecting the beam with electric steerers and
bringing it back with magnet steerers).
i
I
I
dE/dx
X=
(8.1)
where E is the beam energy and Eres is the resonance energy

(6385 MeV). In cases where dEldx is not constant. one would
have to integrate dE/dx (or use range differences) to relate tbe
energy above the resonance energy to depth within the sample.
The conversion of the yield data (Y) to hydrogen
concentration, n(x), is ahnost as easy. Fonnally, the gamma-ray
yield, Y, is given by
y =
Qn(x)s(x)dx
(8.2)
where Q is the number of incident ions, n is the hydrogen

concentration, and cr is the cross-section, By cbanging variables
to an integral over E, using dx = dE/(dEIdx), this becomes
Y =Q
8.4.6 Warning for those running N beams

from a tandem accelerator
Nitrogen does not form a stable negative ion. Hence, attempts
to obtain a N" beam from the ion source of a tandem accelerator
will not be successful. However, there are a number of suitable
negative-ion complexes, such as NH-, NH2-. and C"W, which are
easily available using direct-extraction negative-ion sources.
Which complex to utilize depends on several factors that, in
turn, depend on the characteristics of a particular accelerator,
such as which complex is most easily produced in its ion source.
However, in general, NIr is preferable to CN" because, after
stripping in the terminal, for a given terminal voltage, the N ions
from the fonner ba ve higher energy and hence strip to a higher
charge state. Any Coulomb explosion that might occur during
stripping is also less for NJr compared to CN". It should also be
noted that, when such complexes are extracted from ion sources,
other complexes with the same atomic mass might be present,
for example, 0- and NH2-. Hence, it might be important to
optimize ion source conditions by monitoring the analyzed beam
on target, not just the beam measured at the ion source.
[n(E) a(E)]/(dE! dx)dE
(803)
Note that, over the energies at which the cross section is

nonzero, n and dEJdx are constant and can be taken through the
integral sign, and a(E) is given by the Breit-Wiener fonnula
2
a(E)=
r
aO4
(E-E
i
res
+L
y_Qnlt/2aor
dE/dx
(8.4)
(8.5)
or
n(x) = K (dE/ dx) Y(x)
(8.6)
where K is a constant that reflects all of the cross section

parameters and, in practice, also the detector efficiency.
The gamma-ray yield, Y, is therefore converted to

concentration wing Eq. (8.6). The constant K is determined for
an experimental chamber by making a measurement on a sample
of known H content (see EXA.M:PLE 8.1), although, in principle,
it can also be determined from the known reaction parameters
8.5 ANALYSIS OF RESONANT NUCLEARREACllON DATA

Figure 8.3 shows NRA data for measurements of H in a
sample of Si implanted with 1016 HJcm2 at 40 keY, and a sample
179
Chapter 8
and an evaluation of the detector efficiency. It is important to

note. however, that K is independent of the material being
analyzed All of the matrix-dependent effects are in the dEldx
In summary, NRA profile data, such as those in Fig. 8.3. are

converted to H concentration as a function of depth using Eq.
(8.1) for the depth scale and Eq. (8.6) for the concentration scale.
term ofEq. (8.6).
Depth in silicon (11m)
0.0
0.1
0.2
0.3
0.4
0.5
8
300
10
.Q
.r
6 E
.!!
"
0
tJ
:..
-"
......
u
0....
200
!!!
tJJ
4 ~
U
c:
Counts -:- 50
c:
0
tJ
C,)
100
2
- ---- -- -
0
6.4
6.5
6.6
6.7
__
B_a:~~ro~:d
6.8 6.9
15N energy (MeV)
0.0
~ 1000
.!!
7.0
7.1
7.2
J:
Depth In microns
0.1
0.2
0.3
0_5
0.4
0.6
Amorphous silicon
~ 0.55 microns
8"
8"
500
1 - - - - - - - t..E =0.765 MeV
o~----~--
0.2
__
______
0.4
____
__
0.6
E-ERES (MeV)
FIG. 8.3. Typicall~ nuclear reaction profile data for (top) a Si sample implanted with I016H1cm 2 at 40 keY and (bottom) a thin film
of hydrogenated amorphous Si on a Ctystalline Si substrate. Both raw data (gamma-ray counts versus beam energy) and final results
(H concentration versus depth) are given. Note that the solid data point (bottom) was measured after the profile had been completed to
verify that H was not lost during analysis.
180
I
EXAMPLE 8.1. For the chamber shown in Fig. 8.2, using a
3-in.-diameter by 3-in.-long cylindrical BGO detector 2 em
behind the sample, with Y being counts per 1 ~C (2 J.lC of
1~2l of incident I~ particles, integrating all events in the
4.4 MeV gamma-ray full-energy peak and the 3.9 MeV
single-escape peak, the constant K is obtained as K = 0.45 X
10" ~mI(MeV cm'), or Eq. (8.6) becomes
n ~ 0.45 x 10" (dEJdx) x Y
(8.7)
EXAMPLE 8.2. Measurements on Cl-h

Measurement of Y for a variety of poiyethylene saQ'lples
gave 522 counts per 0.06 flC of IW 2+ beam. Note tha-t this
material can lose H very rapidly in the analysis beaIO. To
minimize this effect. the beam current was very low (a few
nanoamperes), and it was rastered over an area with
dimensions of several millimeters on a side. Note also that
different sources of "polyethylene" gave different count rates
(by as much as 5% from the average).
the data in Fig. 8.3 used a slightly different chamber with a

different detector, so the constant K for those data is
dE/dx for His 28.4 MeV/(mglcm2), and for carbon, it is 8.4

MeV/(mglcm2). See Table A13.2 in Appendix 13. The Bragg
rule for combining stopping powers gives the sto:pping
~ower of CH 2 as
different from that given here. Data using this setup are
discussed further in Section 8.5.1 on calibration procedures.
dEJdX CH, =(2x28.4+12 x 8.4)/(2+12)
where dEldx is in units of MeV/micron, Y is in units of
counts per 2 !le, and n is in units of H atorns/cm3 Note that
=11.26 MeV/(mglcm')
8.5.1 Chamber calibration: Determination of K
Although, in principle, the chamber constant K can be
determined from-the reaction cross section, resonance width, and
detector efficiency, it is dangerous to trust the literature for
absolute cross sections to the accuracy needed for analysis. The
usual procedure is to record H profile data from a sample of
known H content and determine K by comparing these data with
the known H content. There are two types of standard samples
that can be used: one bas a known amount of H per unit area
2
(cm ), and the other has a known amount of H per unit volume
(cm').
Assuming a density of I gfcm) for polyethylene implies that

the dE/dx value [in the units used for Eq. (8.6)J is 1.13
MeV/)J.IIl and that the number of C atoms per cubic
centimeter in polyethylene is 4.3 x 1022 at./cm).
(Note that, to convert dEldx from units ofMeVl(rnglcl11') to
MeV/!J1D.just multiply the former by the specific gravity in
glcm' and divide by 10.)
The usual way to make samples. of known Hlcrn? is by ion ...

implantation. The calibration of the chamber shown in Fig. 8.2
was established in this way. This procedure has thl! advantage of
allowing the determination of K for a chamber with an accuracy
limited only by the precision with which the experiment can be
carried out, that is, the precision of the implantation and the
precision of the H profile measurement. Determination of K by
using samples of known Wern 3 relies on knowing the value of
dE/dx for the sample. Because dEldx is generally known only to
within 5-10%, this limits the accuracy with which K can be
detennined by this procedure to 5-10%, at best.
Profiles of implant data, such as those shown in Fig. 8.3, can
be used to determine K. In this case, Eq.. (8.6) is integrated over
the yield curve, giving nt (atoms/cm') = KN(E) dE.
181
Using this value of dE/dx, along with the measured Y (522

counts/0.06 "C = 17,400 countsl2 ~C), in Eq. (8.7) gives the
hydrogen content ofo ~ 8.84 x 1022 H atom/em3, or the ratio
of Hie atoULS as Hie = 8.84/4.3 = 2.05, in good agreement
with the ex.pected value of2.
Note that, in the <::alculation of the HIe atom ratio, the
assumed density of the polyethylene cancels out. This
cancellation can be confmned by the reader by carrying out
the calculation for the Hie ratio assuming a different density .
for polyethylene. (For example, assuming that the density is
2 glcm' doubles the dE/dx value in units ofMeV/flII\, which
doubles the deduced absolute H concentration. Doubling the
assumed density also doubles the deduced absolute C
content. Hence, taking the ratio betvleen the two, the
assumed densIty cancels.)
Chapter 8
clearly important to determine this contribution. For the case of

the l~ reaction, this off-resonance contribution has been
carefully measured (at least for beam energies below 8 MeV)
(Hom aod Lanford, 1988). The counts from the off-resonance
cross section can be determined by integrating the area under the
surface-layer peak (high hydrogen). The off-resonance
conbibution is then
Probably the more common way to determine K is to record

data for samples of known H/cm3 . There are a large variety of
possible materials, including plastics and metal hydrides, that
could be used (see Table A13.3 in Appendix 13). To illustrate
this procedwe for the chamber shown in Fig. 8.2, the use of
polyethylene as a "standard" is presented in EXAMPLE 8.2.
This is not the procedure actually used to establish K for this
chamber, but is used as a check of the ion-implantation-based
calibration and as an easy way to verify that nothing in the
analysis chamber has changed over time.
Y'ff~' = (1.28
(8.8)
Hr')!
where I is the integrated area (in keV counts) of the surface layer
peak.
It .is perhaps worth emphasizing that, unlike some procedures

that require standards for each class of materials to be analyzed,
for nuclear reaction analysis; one has to establish the chamber
constant K only ouce. Given K. using the procedure outlined
above, any material for which dEJdx is knO\VD can be analyzed.
2.0
8.5.2 Energy loss values

Table AI3.2 in Appendix 13 provides a list of elements with
their densities (specific gravities) and I~, 1~, and 'Li stopping
powers at their resonance energy. These stopping powers are
from the program STOP written by James Ziegler of the U.S.
Naval Academy at Annapolis, MD (Ziegler et aI., 1985).
f1: Lorentz
t2: G3uss
1.5
8.5.3 Depth resolutioll
There exists some confusion about the depth resolution of
these methods in the literature. Commonly, the near-surface
depth resolution was considered to be given simply by the width
of the resonance divided by dEldx. Deeper into the sample, the
depth resolution would be dominated by straggling of the incjdent
beam. For the ''1' case, the resonaoce is so narrow (1.8 keV) that
the motion of H bonded in a solid will shift the reaction on and
off resonance' (Zinke-Allmang e/ al., 1985). As a result, the
reaction behaves as if the cross section were not Breit-Wigner
(Lorentzian) but rathcr Gaussian, with the width
.. -determined by how the H is bonded (in the range of5-15 keV). For
the I~ reaction, this Doppler effect dominates the near-surface
depth resolution.
1.0
FlG. 8.4. Plot of the width that results from the convolution of a
Gaussian with a Lorentzian (Breit-Wigner), as well as the result
for a Gaussian convoluted 'with a Gaussian (Arnsel and Maure~
1983). The results are expressed as the ratio of the total (T) width
to the width of one of the functions. See EXAMPLE 8.3.
In general, the evaluation of the depth resolution of resonant

nuclear reaction profiling requires the convolution of a Gaussian
width (from beam eoergy _ggling and Doppler effects) with a
Breit-Wigner (Lorentziau) width (from the resonance cross
section). This has to be done numerically. III fact, Arnsel and
Maw-el (1983) already did this and presented the general result in
graphical fonn (Fig. 8.4).
EXAl\1PLE 8.3. Convolution of a Gaussian and a Lorentzian.

As an example, we evaluate the depth resolution (fwhm)
of a resonant nuclear reaction (Breit-Wigner width of 25
keV) at a depth in a target where the straggling is 20 keV (a
Gaussian distribution with a fwhm of 20 keV). In the
notation of Fig. 8.4, r]trl = 20125 = 0.8. From Fig. 8.4,
reading the GaussianlLorentzian curve at r l ff2 = 0.8 gives
rTlfl = 1.45 or a total width of 1.45 x 25 = 36 keY. The
depth resolution is then given by this energy width divided
by the value of dEldx at the resonance energy.
8.5.4 Off-resonance contributions

When profiling through a surface layer with a high H content
into a material with a low H content, the off-resonance cross
section interacting with H in the surface layer can contribute to
the count rate. This is shown schematically in Fig. 8.5. It is
"
.~.
..','
;~
182
I
EXAMPLE S.4. Evaluation of the Off-Resonance Yield.
Figure 8.6 shows the raw data for a hydrogen profile of a
sample that consisted of plasma-deposited silicon nitride
(very high hydrogen content). on aluminwn (low hydrogen
content), on a silicon substrate. The surface nitride gave a
very high count rate (about 5500 counts/4.8 ~C). The
(a)
question was whether the low count rate observed in the

aluminum (energies 6.7 MeV and 7.3 MeV) was due to real
hydrogen in the aluminum or to the off-resonance yield -with
hydrogen in the surface nitride. Integration of counts in the
~ Hydrogen-bearing layer
surface nitride layer needs to be done carefully. which is not

possible for the data plotted in the logarithmic plot of Fig.
8.6. Roughly, one can see in this figure that there are about
5500 counts from the resonance energy (6385 keV) to about
6550 keY. Careful integration of .these data gives I = 0.96 x
10' keV counts. Using Eq. (8.8), the off-resonance yield
should be
(b)
Y,ff= ~ (1.28 x 10-') x (0.96 x 10') ~ 12.3 counts.

This is indeed about the count rate observed in the aluminum
layer, and the ~onclusion is that there is no measurable
hydrogen in this layer. The large peak at the back edge of the
aluminum (around 7.5 MeV) is real and clearly not due to
off-resonance reactions. Rather, this peak results from the
accumulation of hydrogen at the AlISi interface during I
sample annealing.
FIG. 8.5. A schematic representation of profiling through a

hydrogen-bearing surface layer into a low-H region. In situation
(b), the off-resonance cross section can interact with the Ii in the
surface layer, thereby contributing to the count rate. The counts
from the process are (to a good approximation) just proportional
to the integrated amount of H in the surface layer times the otfresonance cross section. See EXAMPLE 8.4.
10000 , - - - - - - - - - - - - - - ,
.
8.S.S High-sensitivity measurements
The ('N profiling technique outlined above gives a sensitivity
on the order of hundreds of parts per million (atomic), with
1000
.
.
.
beam currents on the order of SO nA and run times on the order
.....
- ..
ofa few minutes. To measure sarilples with lower H content, it is ......... ~
necessary to (1) increase the beam current, or (2) decrease the
100
background, or (3) increase the detector efficiency, or (4) use a
reaction with a larger cross section. All of these are done. The
increase in beam current depends on the accelerator available
10
and how much beam a sample will take (with cooling and beam
rastering). The decrease in gamma-ray detector background is
1~~~~~~~~~
accomplished with material and electronic shielding of the
6.3 6.5 6.7 6.9 7.1 7.3 7.5 7.7 7.9
gamma-ray detector to reduce the number of cosmic-rdY events
Beam energy (MeV)
recorded (Damjantschitsch et o/.. 1983; Kuhn et 0/.. 1990; Horn
FIG. 8.6. Raw data for a hydrogen profile ofa sample consisting
and Lanford, 1990). New well-type BGO detectors provide a
of a thin layer of plasma silicon nitride on aluminum 00 a silicon
high detector efficiency compared to the commonly used 3-in. x
substrate. The very high-countrate surface peak (6.385-6.55
3-in. cylindrical detector (Kuhn et a/., 1991). Going to the 13.35
MeV) corresponds to the H in the silicon nitride. The low-countMeV 1~ resonance increases the gamma-ray yield by an order
rate region (6.7-7.3 MeV) corresponds to the alwninurn layer,
of magnitude (at a sacrifice in depth resolution and more
and the peak at 7.5 MeV corresponds to the Al/Si interface. The
complicated off-resonance contributions). Hence, there are many
approximately 10 counts in the Al region are attributed to the
ways to increase the sensitivity of this method, and all of these
off-resonance reaction with the surface H layer. See EXAMPLE
approaches have been explored. However, experience indicates
8.4.
that many research problems in this field can be effectively
studied without such changes.
---.,'-~.--.
!
1
183
_-_._.
Chapter 8
8.6 SELF-CONSISTENT ANALYSIS COMBINING NRA

AND RBS DATA: THE NEED FOR NRA DATA FOR
PROPER RBS ANALYSIS
changes in the H coment cannot explain such a large change in

the aSi ratio. This intui~ve answer is wrong.
Energy (MeV)
Because RBS does not see a signal from hydrogen,

iDteIpretation of RES spectra usually ignores the effects of
hydrogen. Although this assumption. is fme in many cases, it is
not always appropriate. Indeed, for the class of materials (widely
used in microelectronics) made by chemical deposition from
silane, methane, and, ammonia, among other gases, ignoring the
presences of hydrogen can result in serious misinterpretation of
RBS data. This potential problem is illustrated in this section. In
addition, the self-consistent procedure for combining NRA data
for hydrogen with conventional RBS is demonstrated in detail.
1500
'"
Figure 8.7 shows 2 MeV RBS spectra recorded from samples

of SiC)fto These spectra were recorded from an as-deposited
film and from a sample from the same deposition after
annealing. The goal of the ion beam analysis is to determine y
and z, that is, the C and H contents of the film before and after
annealing.
Because the RBS signal from C is measw-ed on the large and
(because of the effects of accidental channeling) llncertain
"background" due to scattering from the Si substrate, it is
difficult to determine the C signal accurately. Hence, a better
way to conduct the analysis is to measure the absolute Si signal
and determine the C content "by subtraction", that is, fit the
recorded RBS spectra absolutely by varying Ibe ClSi ratio (y).
Such an analysis procedure requires that the absolute solid angle
of the detection system is known and that the chamber has a
precise beam current monitoring system. In practice, what is
normally done is to measure RBS spectra on the unknown
relative to the RES spectra of similar samples of precisely
known composition. In the present case, these known samples
were a thermally grown SiO~ film and a plasma-deposited
amorphous Silio.19 fibn. Bolb of these standard samples gave the
same solid angle to -1 %.
1000
soo
'\Si
o .
"
100
200
300
Channel
,00
'00
600
Fig. 8.7. 2 MeV RES spectra from an as-deposited film of

SiCyHz and from an annealed film from the same deposi.tion. The
Si-substrate step is located near 450 channel. Note the !Dcreased
Si step height for the annealed sample, consistent 'with the los5
ofH, and possiblyC, during the anneal.
Given this zeroth-order C content, the NRA hydrogen data

can be analyzed to obtain a frrst-order WSi ratio for these films.
(This analysis is shown in more detail in EXAMPLE 8.5.). The
results of tills analysis are HlSi - 1.18 and 0.33, respectively.
This makes it clear that the zeroth-order assumption ofHlSi =.

2 for the initial RBS analysis was far from correct With the
first-order values for HlSi for these films, the RUMP fitting of
the RBS spectra can be repeated to obtain better values for c/Si.
The re,ults ore ClSi - 0.89 and 0.88, respectively. Note that the
qualitative interpretation of the effect of annealing is very
different from that made before including the NRA hydrogen
data. Now, the O'Si content is seen to be the same after
annealing, suggesting that only hydrogen is emitted from the
sample when annealed.
Confronted with the two spectra shown in Fig. 8.7 t it is easy

to determine the CJSi ratio for each sample simply by using
RUMP (Doolittle, 1985) to fit Ibe two spectra. In order to do
this, some assumption needs to be made about the H content of
the two films. If the hydrogen content is ignored (Le., assumed
to be zero), the C/Si ratio deduced from fitting the spectra will
be in error. In the present case, the H content of each film was
initially assumed to be SiC~H2 (a reasonable guess for the H
content of such films).
Repeating the cycle of analysis using these new values for
CJSi to re-analyze the NRA data and using the new H/Si values
to determine new c/Si values results in final compositions of
SiC, O.sHL14 and SiCl.ooHlJ.3.4
The results of fitting these spectra under this assumption are

c/Si ratios of 0.65 and 0.41 for the as-deposited and annealed
samples, respectively. Note that, on face value, these results
suggest that the C content of these films decreases with
annealing, perhaps through the emission of methane from the
sample. Without a calculation, intuition might suggest that any
Note that the uncertainty in the Reontent has a mueh greater

effect on the inferred C content than the reverse. The HlSi ratio
changes only slightly with better information about the ClSi
ratio, whereas the deduced O'Si ratio cbanges dramatically 'Nlth
better infonnation on the H content.
184
Nuclear Reactions far Hydrogen Analysis
Example 8.5. Combining NRA and RBS.

First, it should be noted that, when expressing the hydrogen
content as an atomic ratio, the density ofllie film being analyzed
is not needed. Hence, in the analysis example presented here, the
H/Si atomic ratio is detennined without making any assumption
about fihn density.
Repeating the RUW fit of the RES data and then the re-analys!s
of the NRA data gives new values for y and z as follows:
zeroth-order first-order
second-order
y,z
y, z
y,z
as-<iepnsited
0.65,2.00
0.88, U8
1.05, 1.24
annealed
0.89,0.33
0.41,2.00
1.00.0.34
For simplicity, consider SiC;Jf/: to be one molecular unit. The
Comments
It is worth noting that the effect of hydrogen is large in this
case because RES spectra were being analyzed making use of
only the Si RES signal to determine the C/Si ratio, that is,
determining the carbon content by subtraction. If the RBS
spectra were analyzed by measuring the RES C signal relative to
the RBS Si signal, the effect ofR would be negligible. However,
accurate measurement of the RBS C signal is difficult because of
the large and uncertain background from the Si substrate.
gram molecular weight of this unit is

gmwt = 28 x 1 + 12x Y + Ix z,
with the numbers 28, 12, and I being the approximate atomic
masses ofSi, C, andH, respectively.
Using the elemental stopping powers given in Appendix 13 and
the Bragg rule for combining stopping powers, the stopping
power of this material is
dEldx
= (28
An alternative approach would be to deposit the films on a

light substrale (e.g., carbon) to avoid the difficulty of measuring
the C RBS signal on a difficult background. Unfortunately, this
is often not practical.
x 6.24 + Yx 12x 8.45 + z x Ix 28.4)lgmwt
where the stopping powers for 6.4 MeV nitrogen in Si, C, and H
are 6.24, 8.45, and 28.4 MeVI(mglcm'), respectively.
The simpler case where the non-hydrogen composition of the

film is known (i.e., y = c/Si in this case) also requires selfconsistent analysis similar to that illustrated here. In this case, an
initial guess for the hydrogen content needs to be made to
calculate the stopping power, which is used to analyze the NRA
data for hydrogen, which is, in tum, used to calculate stopping
power, and so on. Experience shows that this analysis cycle
generally converges in one cycle.
The stopping power in ~eV/Jlm is just p{dEIdx)/IO, where p is

the film density in g/cm J .
The density of molecules (or Si atoms) for the film is p x (6.02 x
lOlJ)/gmwt, where 6.02 x 1023 is Avogadro's number.
Making use of these results and Eq. (8.6) gives the HlSi alomic
ratio as
~
~
<-
K[(dE/dx)p/IO]Y
Si
p(6.02xlO")/gmwt
Z=-
-'-.
8.7 SUMMARY AND CHECKLIST OF NUCLEAR

PROFlLH,G c PROCEDURES
AND
REACTION
CONCERNS
.. P
Given the availability of suitable energy beams of l~, 19F, or
'u and a gamma-ray detector (e.g., 3-in. x 3-in. Nal or BOO),

NR<\' for H profiles requires only the ability to bombard the
K(dEldx)(gmwt)Y
sample cleanly with the analyzing beam, to integrate the incident

beam effectively, and to measure the resultant gamma rays with
a fixed efficiency. Probably the two most common difficulties
encountered revolve around the questions of whether the beam is
hitting the sample and only the sample, and whether the H is
stable in the sample during analysis. The first of these can be
managed by taking some care in designing the beam geometry
(no slits near the target but ability to monitor beam position,
beam size, and any possible beam halos). The second pitfall can
be managed by taking care to rrllnimize the beam intensity on the
target by rastcring thc bcatTl, by running the minimum total
number of ions required to get good statistical accuracy, and by
cooling the target.
IOx(6.02xlO")
where K is the chamber calibration constant (0.45 x 10 19 for
Albany), dE/dx. is the stopping power of the film in
MeV/(mglcm'J, gmwt = gram molecular weight of our
molecular unit, and Y is the NRA reaction yield (in counts12 J-lC
for SUNY -Albany).
In the present example, using the carbon contents obtained by

fitting the RBS spectra using RUMP (i.e., y = 0.65 and 0041,
respectively, for purposes of an initial calculation of stopping
power), the assumed hydrogen content of HlSi = z = 2, and the
NRA counts recorded for these samples (Y ."., 5280 counts/2 I-lC
and 1618 counts!2 ~C, respectively) gives hydrogen contents of
z =HlSi = l.l8 and 0.33, respectively.
Analysis of the data is done by using Eqs. (8.1) and (8.6),
with the chamber calibration determined by running one or more

samples of known H content.
185
Chapter 8
REFERENCES
Adler, P.N., Kamykowski, E.A., and Padawer, a.M. (1974), in
Hydrogen in Metal, American Society of Metals, Materiais Park,
OH,p.623.
Amsel, G., and Maurel, B. (1983), Nud. Instrum. Methods
218,183.
Barnes, C.A., Overley, lC., Switzkowski, Z.E., and Tombrello,
T.A. (I 977), Appl. Phys. Lett. 31, 239.
Cohen, B.L., Fink, C.L., and Degnan, I.H. (1972), J. Appl Phys.
43,19.
Damjantschitsch, H., Weiser, M., Heusser, G. Kalbitzer, S., and

..._ Mannsperger, H. (1983), Nucl. lnstrum. Methods 218, 129.
Doolittle, L.K (1985), Nucl. lnstrum. Methods B9, 344.
Horn, K.M., Lanford, W.A., Rodbell, K., and Ficalora, P.
(1987), Nucl. lnstrum. Methods B26, 559.
Horn, K.M., and Lanford, W.A. (1988), N.d. lnstrum. Methods

B29,609.
Kuhn, D., Sturm, H. and Rauch, K (1991), in 10th International

Conference on Ian Beam Analysis (1-5 July, 1991. Eindhoven.
The Netherlands) Book ojAbstracts, p. 73.
Lanford, W.A., Trautvetter, H,P., Zieger, J.F., and Keller, J.
(1976), Appl. Phys. Lett. 28, 566.
Lanford, W. A. (1992), Nucl. lnstrum. Methods in Phys. Res.
B66,65.
L'Ecuyer, J" Brassand, C., Cardinal. C., Chabbal, l, Deschenes,
L., Labrie, J.P., Terreault, R, Martel, J.G., and St. Jacques, R. J.
Appl. Phys. (1976), 47, 381.
Leich, D.A., and Tombrello, T.A. (1973), Nucllnstrum. Methods
108,67.
Schnatter, KH., Doremus, KH., and Lanford, W.A. (1988), J.
Non-Cryst. Solids 102, II.
.
Tracellier, P., and Engelmann, Ch. (1986), J. Radioanal. N"d.
Chem. 100,117.
845,256.
Xiong, F., Rauch, F., Sm, C., Zhou,' Z.,Livi, RP.; and
Tombrello, T.A. (1987), Nud. lnstrum. Methods 'Phys. Res.
B27,432.
Khabibullaev, P.K., and

Skorodumov, a.G. (1989),
Determination of Hydrogen in Materia/s, Nuclear Physics
Methods, Sprioger-Verlag, Berlin, Germany.
Zigeler, I.F., Biersack, J.P., and Littmark, U. (1985), The

Stopping and Range of Ions in Solids, Pergamon Press, Oxford,
UK.
Kuhn, D., Rauch, F., and Baumann, H. (1990), N.d. lnstrum.

Methods 845, 253.
Zinke-Allmang, M., Kalbitzer, S., and Weiser, M. (1985). Z.

Phys. A230, 697.
Horn, KM., and Lanford, W.A. (1990), Nud lnstrum. Methods
'1
"
,',
186
CHAPTER
II
HYDROGEN ANALYSIS BY PROTONPROTON SCATTERING
P. Reichart and G. Dollinger
Universitiit der Bundeswehr Miinchen, Neubiberg, Germany
CONTENTS
9.1 INTRODUCTION .......................................................................................................... 189
9.1.1 Basics of the proton-proton scattering technique ............................................. 189
9.1.1.1 Kinematics .................................................................................................... 189
9.1.1.2 Coincidence filtering .................................................................................... 190
9.1.1.3 Energy analysis ............................................................................................. 191
9.1.1.4 Timing ........................................................................................................... 191
9.1.2 Quantification .....................................................................................................192
9.1.2.1 Loss due to angular spread .......................................................: ...:::.:..:. .:.......192
9.2 SENSITIVITY LIMIT .................................................................................................. 194
9.2.1 True events versus accidental coincidences ..................................................... 194
9.2.2 Irradiation damage ............................................................................................ 195
9.2.3 Hydrogen microscopy using a proton microprobe ......................................... 197
9.3 DEPTH PROFILING .................................................................................................... 198
9.3.1 Depth scale .......................................................................................................... 198
9.3.2 Depth resolution ................................................................................................. 199
9.3.2.1 Correlated-path-Iength effect, IiE,,.th .......................................................... 199
9.3.2.2 Small-anglc-scattering-path effect, SEa ..................................................... 199
9.3.2.3 Energy-difference method ........................................................................... 200
9.3.2.4 Sum of contributions ....................................................................................200
9.4 EXPERIMENTAL REQUIREMENTS ....................................................................... 200
9.4.1 Sample thickness ................................................................................................ 200
9.4.2 Detector thickness .............................................................................................. 202
187
Chapter 9
9.5 ADDITIONAL REMARKS .................. ..................................................................... 203

9.5.1 Relativistic corrections to sum of scattering angles ........................................ 203
9.5.2 Quantum effects due to identical particle scattering ...................................... 203
9.5.2.1 Mott scattering ............................................................................................. 204
9.5.3 Relativistic transformations .............................................................................. 204
REFERENCES
.................................................................................................................. 205
188
Hydrogen Analysis by Proton-Proton Scattering
9.1 INTRODUCTION
This chapter presents a brief collection of the basic
principles, advantages, limits, and technical requirements of
pratall-jJrotan (Pp) scattering far hydragen analysis. The
principle of proton-proton scattering is illustrated in Fig. 9.1.
Two protons are scattered elastically from a reaction between an
incident proton of kinetic energy Eo and a hydrogen atom in the
target and are detected in coincidence in transmission geometry.
The benefit of coincident detection is that it filters out elastic ppscattering events from all other reactions, in particular from
nuclear reactions, giving a background-free hydrogen analysis.
Because the absolute kinetic energy minimwn (in the laboratory
system) for pion production is 286 MeV, this scattering can be
regarded as essentially elastic even up to energies of about 350
MeV (Bethe, 1999).
and the energy of each scattered proton is
=K,E. =(cas' ale. and

E. =K.E. =(cos' ~)E. =(sin' alE.
E,
where Etn is the energy of the projectile before scatt<:ring. For (

sum of the energies of the two scattered protons, ~e obtain
E,+E.=E..
which is the energy of the protons before scatteri:ng. Note tJ
we use Eo for the incident beam energy before entering the tarj
but Ein for the incident energy before scattering.
The striking advantage of proton-proton scattering is that,

with only a little effort-two detectors and some coincidence
electronics---one can achieve a high sensitivity in every ion
beam laboratory with MeV protons. This tcchnique was
discavered by Cahen et al. (1972), wha already demanstrated
the detection of 10 at.ppm (atoms per 106 total atoms) hydrogen
in iron using 17 MeV protons. Lower proton energies were
applied by Paduschek and Eichinger (1980, 1981) and later by
Martinsson and Kristiansson (1993). The enhanced sensitivity of
the technique for hydrogen analysis compared to elastic recoil
detection (ERD) was emphasized by Green and Dayle (1995).
This huge benefit for scanning proton microprobes :in terms of
(;oncurrent lateral/depth resolution and sensitivity was discussed
in detail with respect to the low damage potential by Reichart et
al. (2002). It was in 1996 and 1997 that the method was first
applied to three-dimensional bydrogen mapping and microscopy
(Dujrnic et al., 1996; Sjoland et al., 1997). With dedicated setups
at the Munich microprobe SNAKE [Superconducting Nanoscope
for Applied nuclear (Kern) physics Experiments], a detection
limit af even 0.08 at-ppm was achieved in 2004 in a 50-~
thick diamond sample (Reichart et al., 2004a, 2004b), together
with a lateral resolution of only 0.6 ~. The latter microprobe
uses high-energy protons up to 28 MeV, which allows the
analysis of the required freestanding samples up to a thickness of
several hundred microns. Thus, there are only a few limitations
in sample preparation, and proton-proton scattering is very
attractive for hydrogen analysis in cases where standard methods
fail in tenns of sensitivity, spatial resolution, and calibrated
quantification.
<-;J
sample
9.1.1 Basics oftbe proton-proton scattering technique

9.1.1.1 Kinematics
In the noorelativistic approximation (see Section 9.5.1 for
deviations), the protons have equal mass. The scattering angles e
and $ are then related to each other by
..
......
,.
. ...
FIG. 9.1. Hydragen analysis bypratan-protan scattering. T

possible detector setups are shown with the detector oriented
perpendicular and (b) parallel to the beam direction
transmiss~oD geometry.
(9.1)
189
Chapter 9
essential to achieve the highest sensitivity in microprobe

analysis.
9.1.1.2 Coincidence filtering

Time and angular coincidence filtering offers the possibility
to suppress all scattering events other than pp-scanering events
in order to obtain hydrogen counts without background. In total,
three conditions give a very tight coincidence window on true
pp-scattering events:
(1) time coincidence (nanoseconds),
(2) scattering plane (opposite detector elements), and

(3) angular sum (90).
As a qualitative example of 17 MeV proton-prown
scattering, Fig, 9.2 shows the coincidence filtering for the
analysis ofa Q,9-f..llll-thick Mylar foil. The spectra were recorded
using a special dctector that is divided into 16 sectors and 48
rings (see Fig. 9.7 in Section 9.2.1). In Fig. 9.2(.), the sum
energy of all incoming events is plotted as a function of the
scattering angle of each particle. The spectrum includes the true
pp-scattering events (I) and various single events from protons
scattered from various nuclear reactions (II-IV). In addition,
there is a massive background due to protons scattered
elastically from the C, N, and 0 atoms (marked with dashed
line). Because these protons are not stopped in the I-mm-thick
detector, they deposit only a fraction of their energy, which
varies with scattering angle, The spectrum in Fig. 9.2(b) shows
coincidence events that fulfill all three filter conditions,
specifically, (I) double hits within a 2-3-05 time windOW, where
(2) Lbe reaction point and the two detection points lie in one
plane (hits in opposite sectors) and (3) additionally the sum of
scattering angles is required to be 90~. In conclusion, one obtains
a nearly background-free spectrum of the true pp-scattered
events, with some artifacts/tails on the outer scattering angles
caused by detector edge effects.
Note that this dedicated detector setup allows the possibility
of using ail three filter conditions and, as shown in Fig. 9.2,
demonstrates the anguJar dependence of the background.
However, high sensitivity is already achieved by simple setups
with ODe or "more detector pairs that filter with conditions (1) and
(2). Condition (1) is, in fact, the most critical. Furthennore, one
can use the energy difference of the emerging protons, L\E = Ee ~,and write condition (3) as
lill- Eo - E, - Em(eos'
a - sin' 6) -
Em cos 26
FIG. 9.2. Example of proton energy spectra from a O.9-~-thick

Mylar foil and the power ofpp-coincidence filters: (a) All events
with multiplicity M :: I. ] marks the pp-scanering events with
the coincident protons adding to a total energy of 17 MeV
(incident proton energy). II-IV mark various bands of proton
events where the incident proton is inelastically scattered with
excitation of C. N. and 0 atoms" The elastically scattered
"single" protons are marked with the dashed line: they leave
only a fraction of their energy depending on the scattering angle
(17 MeV protons are not fully stopped in the detector chip). (b)
Double hits (M: = 2) within a 2-3-n5 coincidence window that hit
the detector in opposite sectors and result in an angular sum of
90~" Tails on the outer scattering angles remain as artifacts
because of detector edge effects: protons hitting near the edges
of the detector do not give the full energy signal.
(9.4)
Hence, the scattering angle is determined by the energy

difference, and one can use a correlation to the energy sum in
order to achieve some filtering. Specifically, the energy loss has
to be taken into accOlmt, and 8E has to be evaluated with Eo and
E; as defined in Eqs. (9.11) (Section 9.3.1). This technique

suggested by Sjoland of 01. (1997) is good for filtering with
smaller detectors and very useful for correction of the path
length effect (see Section 9.3.2). However, when using a large
solid angle of detection, an angular resolving detector is
190
25 MeV protons. It shows that the depth resolution is sufficient

to reveal a low bulk hydrogen content that would otherwise be
hldden under the surface-byrlrogen background.. This main
feature and the depth resolution limit of the method are
discussed in Section 9.3.2.
9.1.1.3 Energy analysis

For thick targets, tJ:le sum of energies in the proton-proton
event is related to the depth of the scattering event, and it is
possible to generate a depth profile of the hydrogen distribution
vvith this information. A.s an example, Fig. 9.3 shows a depth
profile ofa 130-1llD!bick quartz sample that was analyzed uswg
Sljm
21
of energy ~fcojncident protoils (li,111V)
'23
l2'
pa~ksuri'~c( ,
~lOd :H~atJcm'
"'24
front surface,
-10" n-.item'<
.buikregion
c.:i 0;' H-itJem'
li
i
depth
resolution
-8~m
depth resolution
FWHM-15~
higb
I"",' energy
background
eoergy
SG
-100
-50
appr"ox. depth scale (pm)
FIG. 9.3. Depth profile from 25 MeV proton-proton scattering of. quartz sample that was polished to a thickness of about
130 J1ID and analyzed without further surface preparation. Hence, surface peaks are dominant but can be sepamted from the
bulk bydrogen content of about 10 at-ppm.
9.1.1.4 Timing
The scattered protons have different flight times to the
detector if 9 +- 45. For scattering on a thick sample with
thickness d, the maximum differeoce in the time of flight for
scattering is observed at z = 0 (at the froot of the sample; see
Fig. 9.1). Assuming a constant stopping f<?rce along the paths,
one can calculate, in a Doorelativistic approximation, the time
difference, 6.1.1.(9), for the two protons hitting a perpendicular
detector plane at distance D [Fig. 9.1(a)) or the time difference,
t.tR(e), for the two protons hitting paralle.l detectors a with

separation distaDce of2R [Fig. 9.l(b)] as
COS S
(
Eo cos' 9 _ Sed )-'

cosS
.
., 9 Sid
- smS
E sm
(
o
sine
l!Ild
191
J-'
(9.5)
Chapter 9
that the approximation of constant stopping force is not valid for

very thick targets!
In Fig. 9,4(a), D.t (8) is plotted as an example for a typical
experiment (see figure caption). Time of flight differences 6t are
in the order of 1 DS even under steep angles. A parallel detector
setup [Fig. 9.1(b)) has a smaller time of flight difference as
shoVID. in Fig. 9.4(b) because the difference in the length of the
flight path is compensated, to first order, by kinematics (D.t = a
ford_D).
sin 9 Eo cos' a_ S,d )-'

cos
,
S,d
- cosS Eosin 8--.sma
-'
, (9.6)
respectively. with the stopping forces Se and S, for the protons

scattered to angles 8 and $ [see Eqs. (9.12), Section 9.3,1]. Note
. (a)__ ._ Detektor
:i
lBearn AXIs
D=29 rnrn'
3 MeV
d (Si) =
10)lm /
.5
2 .5
:
.
.
. ..
,; I.S
.E
.,
: 3 MeV
"
~OMeV
d=O
...
-",#'
... "
rnm .' .,.
3 MeV
! d (Si) = 10 J.lm
2.5
..,,
.
I.S
30 MeV
d=O
........
.;: O.S
,.,
:
"
....
..
-R=~O
IOO,...m,'
,
'
;;:'"
d (Si) = :
..
Detclctot nBeairi AXis
30 MeV
d (Si) ~ 100 J-lm
...
05
"~- ~...
.##
--SJ.5""~-~-6~0""--~65""-':""''':''~70.64'~5'~'~-';'--;;~;';~~0~'''''''''''~5SO::.:''-''-'',-'''-'''~'6O"~';;:'';'''::'':':'':';~~~::-'...;d::.=~~O.
''''5'''''',:=::_SLO""_,
4
a (lab Systern)
Ei (lab system)
FIG. 9.4. Difference in time offiigbt for coincident protons scattering from a Si sample (thickness d) at different proton energies, as
labeled. The results are shown ror setups mth (a) the detector perpendicular to the beam IIXlS at distance D 20 mm [ef. Fig. 9.1 (a)J
.,and (i_the detector parn.Jlel to the beam axis at radius R = 20 mm [ef. Fig. 9.l(b)J.
9.I.Z Quantification
(do/dO)" is nearly constant in the centerof-mass (eM) system
Proton-proton scattering gives the hydrogen content without
at most detection angles (i.e., laboratory angles of 30-60 0 )

because of the dominant S-wave nuclear scattering (see
Appendix 14). Note that this is not true in the laboratory system
because of the dO transformation. Figure 9.5 shows an excerpt
of the available data for relevant energies given in the laboratory
system. In Appendix J 4, the corresponding data are available in
tabulated form, along with a simple approximate fit for the data.
respect to bonding and chemical environment as long as

channelinglblocking effects can be neglected. The areal atomic
density, PH dx, of hydrogen in the sample is directly related to
the detected coincidence scattering events, NP!"
with the
(differential) pp-scattering cross section (dolctn)pp a:;
N..
with
cr,,= J(dcr)
ill
(9.7)
9.1.2.1 Loss due to angular spread
n" dQ "
Multiple or small-angle scattering causes angular spread of

the incoming and emerging protons that increases with the path
length of the scattered protons. Ibis causes a loss of true events
for the corresponding detectors that increases rapidly with
decreasing size of the detector elemeqt or apectwe. This problem
is briefly addressed in Cohen et al. (1972), who approximated
In Eq. (9.7), No is the nwnber of incident protons. and Epp

accounts for the coincidence detection efficiency. If two
opposite detectors are used for coincidence detection, then Dw is
tbe solid angle covered by only one detector because of the fact
that identical particles arc being scattered. At high energies,
192
,.
Hydrogen Analysis bY Proton-Proton Scattering
the root-mean-square angle of multiple scattering, oa, roughly

by
6a/deg"
~1.8Z1 ~
(9.8)
where ZJ is the atomic number of the target material and.6E/E is

the fraction of its total energy lost upon passing through the
target. This gives a spread of up to a few degrees for typical ppscattering experiments, especially in heavy-element target
materials, and it is a critical constraint for the detector setup. A
more detailed discussion can be found in Hofsass et al. (1991),
for example, or a theoretical approach in Szilagyi ef aJ. (1995).
However, a detailed simulation by Monte Carlo methods or
some other approach that can improve the accuracy of ppscattering experiments is still missing, but is under development
within the CORTED code from F. Schiet!ekatte (2008).
1$0
160
~eB111"""""
.. eI4.Ia~v
. e '"1eJ I.k\'
;..
.... e .. \ulobtv
'a2Ul~
.. e29AL.l4Iv
..... E. So.:UiI(lV
100 15" 20 2S" 30" 35" .a"' 4S"
'"- -&0' 65'
nr
1~--
SO"
Scattering angle 9...
FIG. 9.5. Elastic pp-scatlering cross-section data Oaboratory

system). Data points are connected for better visibility. Reported
relative errors are smaller than 1% for = 20, ... , 70. These
data and further data including those fcr inelastic scattering are
available from individual sources (Allred et al., 1952; Berdoz et
at., 1986; Blair el at., 1948; Burkig ef al., 1959; Dayton and
Schraok, 1956; Imai ef at., 1975; Jarmie ef at., 1970, 1971,
1974; Jannie and Jett, 1974; Jeong ef at., 1960; Johnston and
Young, 1959; Kikuchi ef at., 1960; Knecht ef al., 1959;
Panofsky and Fillmore, 1950; Slobodrian et at., 1968, 1976;
Wassmer and MuIuy, 1973; Worthingtoo et at., 1953; Yntema
and White, 1954) or from collections (Bystricky ef aI., 1980) or
databases such as EXFORJCSISRS (IAEA Nuclear Data
Services, 2009j NNDC Databases, 2009). Gray data points were
calculated from symmetrical angles. Original data with a fit
approximation and also relativistic transformation equations are
given in the Appendix 14. Note that the differential cross section
is symmetrical about 90 0 in the center-of-mass system for the
nonrelativistic approximation, but not in the laboratory system
because of the dO: transformation.
.
The loss due to lateral spreading can be empirically

corrected by a polynomial or exponeotial fit (Wegden ef at.,
2005; Komatsubara ef at., 2006), but note that the fit function is
strongly material-dependent and also depends on the geometry
of the scattering detector. Because accurate theoretical models
for these dependencies are not currently available, it is
recommended calibration targets with a similar thickness and
stoichiometry be used. Suitable hydrogen standards for
calibration are Mylar, Kapton, and t)pes of polycarbonate.
Calibration for materials with higher atomic number can be
achieved either by stacking these material between thin foils, as
was done for Fig. 9.6, or by using materials with known
hydrogen content. Implantation of hydmgen might also be a
solution.
193
.. E"1UI1$NeV
~200
As an example, Fig. 9.6 shows a pp-scattering spectrum of a

sandwich sample consisting of 60-J.llll Si sheets separated by 0.9fUIl-thick Mylar foils [Fig. 9.6(a)]. It was analyzed using a 25
MeV proton beam. The detector setup consists of seven
corresponding pairs of 40 mm x 8.5nun strips, eacb pair
covering a solid angle of detection of about 2x1'lpp = 2xlO mSr.
Figure 9.6(b) .shows a plot of the two energies measured with
two detectors in coincidence against each other. The- energies
show a quasilinear relationship that is somewhat bent at lower
total energies because of the path-length effect (Section 9.3.2).
In Fig. 9. 6( c), the same energy spectrum is translated into an
approximate depth profile from the total energy using Eq. (9.14).
The integrated peak content [Fig. 9.6(d)] sbows that the number
of detected coincidence events decreased at lower depths, even
though aU Mylar foils had the same thickness. This.decrease is
due to the increasing lateral spread with the longer path length of
the scattered protons through the SilMylar stack.
Chapter 9
.)
Si-. Wafer - ~ J.li:n each
,,
.,
direction
~.
Mylar Foii - 0.9 p.m each
xi@'
/~I
b)
2.0
e=
b
u
:
dded:or Ii
"
t
H:rdrn~1l Depth Profile
1$
u
'"'-. 'll
,,
f4 ..
<)
detector i .:
bcaJi,
. ..
. Salllplt Cr"css Stttfon.
.... ,.
1.5
"':
1.0
..
0.5
.1(1)
400
d)
EnellO' CorrelatIon
16
I.
>u
:;:
.E
~
""
I.
16
14
(ntegrnted Pe1!k Content

,~
\-'
" ',e" I.
'lle
A'
11
ID
.=
:t:,.
~.
,-j"
,1
L'
. ,it
-.
8
J
12
- [nugy.~t~~I~orn III ~eV'
:~:
~;r.
18
I.
12
12
';
'"
dO'
20
......... 2
'~~ -
I.
. --"~Oo
11)0
~OO
. '~App~~m2tt~ drplh % iu p"m"
300
.40~
FIG. 9.6. (a) Proton-proton scattering spectra of a sandwich sample that consists of five O.9-jJ.m-thick Mylar foils separated by.
hydrogen-free sheets of60-~m Si. The incoming proton energy is 25 MeV. (b) Correlation of the energies of coincident protons. The
slight bending appears at lower total energy as a result of path-length effects. (c) Depth profile of coincidence events with approximate
depth scale according to Eqs. (9.14) (Section 9.3.1). The depth resolution improves toward the back side ofllie sample because of the
decreasing energy spread and lower path-length effect. (d) Depth profile of integrated peak content that decreases toward the front side
ofllie sample (z "" 0) because of the increasing angular spread of the protons (true event loss).
From this calibration measurement, one obtains the

efficiency factor f"p for quantification in Eq. (9,7), which
depends on the depth z because of the lateral spreading effect.
The calibratioG also accounts for losses due to factors such as
jitter in the timing electronics and other detector effects, as well
as systematic erroTS in the beam current measurement. Countrate-dependent effects sucb as dead time and pileup should be
monitored and corrected for each measurement, of course.
194
9.2 SENSITIVITY LIMIT
9,2.1 True events versus occidental c(lincidences

As shown already by Cohen et al. (1972), even for a simple
detector geometry of two detector halves, in principle, it is
possible to detect a hydrogen fraction as low as 30 parts per
billion (Ppb) in common target materials in a reasonable
detection time. The sensitivity is limited by two factors: (1) the
irradiation qamage mainly caused by the incident proton beam as
discussed in Section 9.2.2 and (2) the true-ta-accidental

coincidence rate, that is, the signal-to-noise ratio (SNR).
Consider two corresponding detector elements with the total
count rate ns in each detector element. Following reasoning
similar to that of Cohen el at. (1972), one obtains the ratio of the
true pp-coincident count rate, TIpP' and the accidental count rate,
D.a.;, for small atomic hydrogen concentrations, CH. as
ncO"
SNR = -2!. = _,_H_--f!!.,

Dac
TJE
(9.9)
IDs cr p
where'[ is the length of the coincidence time window, opp is the

scattering cross section for proton-proton scattering as in Eq.
(9,1), and
O"pp =
r (dO"/ctn) P dD
hw
15
the sum of all proton
scattering cross sections ("singles") for scattering on the nuclei

of the matrix constituents within Opp (1.e., one-half of the total
solid angle of detection). In addition, T\E is the fraction of all
scattered events that fall in the energy window of true protonproton events (sum energy).
In principle, at a given arbitrarily tiny hydrogen

concentration CH, one can reduce the total count rate ns until a
satisfactory SNR is obtained according to Eq. (9.9). However,
that would reduce the hydrogen count rate as well, because the
coincidence and total rate are related by npp = cH(appfop)ns, and
measurement time would increase unacceptably. Hence, one
must set a desired minimum count rate of true events, (opp)min =
NmuITmca>, where at least N miu proton-proton counts must be
registered within a reasonable measurement time, T mea5, in order
This demonstrates that partper-million sensitivity is. easily

achieved using just two opposite detectors with standard
coincidence electronics. To achieve higher sensitivity ?Vith aD
acceptable detection time, particularly when using low bea.m
currents in microprobe setups, the only way is to increase the
number of detector elements, 2M. At the 17 MeV pp-scattering
setup at SNAKE, the annular detector shown in Fig. 9_7 with
384 corresponding detector pairs (48 x 16 rings and sec10rs) is
used, which, in principle, allows for the detection of 0~02 at.
ppm of hydrogen at SNR = 10 with the previously mentioned
values of the other parameters at a count rate of 1 cOlOClt/h. In
reality, multiple/angle scattering effects already playa crucial
role and reduce the effective number of usable elements, 50 that
a detection limit of "only" (CWmin = 0.08 at.ppm was achieved
(Reichart et al., 2004b) at a count rate of roughly 3 countslb and
a total measurement time of only 4 h. Note that the timing
resolution was 2 ns for that elCperiment, with a total count rate of
100 kHz in all detector elements together.
The angular spread as discussed in the preceding section

limits the minimum reasonable size of the detector elements.
Moreover, with increasing number of detector elemen~, the ratio
of inactive interstrip area relative to the active area increases.
This leads to tails in the energy spectra and negatively affects the
sensitivity. Detector misalignment becomes a critical issue as
weD, as illustrated in Fig. 9.8(a) for an annular detector and for
strip detectors in general. Also, the paths of the scattered protons
(particularly at steep angles) might hit more than one detector
element [Fig. 9.8(b)] in the case of multistrip semiconductor
detectors. These types of problems can be corrected using a
software analysis.
to obtain a reliable hydrogen signaL This gives
9.2.2 Irradiation damage

In all IBA methods, the incident beam causes irradiation
damage that will change the locations of a certain number of
atoms to be detected before enough data for good statistics are
collected. In this sense, irradiation damage limits the sensitivity,
(9.10)
as discussed in Section 5.6.3 in Chapter 5 of this handbook. As

with hydrogen analysis, in fact, such damage is reduced by a
factor of about 106 for the pp-scattering method compared to all
where the SNR is increased by a factor of M when M detector

pairs are used.
other rnA methods for hydrogen analysis. Thus, broadbeam

analysis of hydrogen in any sensitive material such as biological
or organic material and in particular microprobe analysis
becomes possible. Three main facts offer this enormous
advantage [a detailed treatment is given in Reichart et al.
(2002)]:
For all but very heavy elements, at an incident energy of 17

MeV, the scattering cross sections, oIl' are of the same order of
magnitude as opp (Cohen et aT., 1972). Thus, a typical minimum
sensitivity of CH = 0.3 at.-ppm can be obtained for SNR = 10,
Nmin = 10, Tmeas =10 h, '[ :::: 1 us, and TJE :::: 0.2 when using one
detector pair (M = I). The background equals the signal (SNR =
1) at a hydrogen concentration of CH = 0.1 at.-ppm for these
conditions. The total count rate in the detector will then be
around Os = (npp)mir!CH = 3 kHz, which is a reasonable count rate
for standard detectors and acquisition systems.
195
/
Chapter 9
boou!
(leclOt) .
(ring)
FdIG'd9_~d' Detector chip used at SNAKE. Micron

ope
Semiconductor Design S2 with 48 rings at the p-doped side and 16 sectors at the
D-
51 C.
example, at 20 MeV. it is enhanced by up to a factor of 500.

However, the damage cross section, which is dominated by
small-angle scattering events, remains Rutherford.
(3) The ~smis~ion geometIy and the sum-energy analysis of
elastIc scattenng events enables the use of huge solid angles
of detection, ~ > 1 sr, without being affected by kinematic
effects.
b)
FIG. 9.8. Effects of <a) detector misalignment and (b) the

problem of steep trajectories hitting neighboring detectors.
(I) The hydrogen release cross sections listed in Table 5.6 .in
Chapter 5 for various beam/target material combinations
indicate that the release cross sections of MeV proton beams

are much lower than those of heavy ion beams. The data
also show that the nuclear displacement cross section, 0disp.
which is proportional to the nuclear stopping power, is a
useful upper-limit approx.imation for hydrogen release in the
case of proton beams, whereas ionization effects play a
minor role.
(2) The detection cross section of proton-proton scattering is
strongly enhanced compared to the Rutherford scattering
cross section because of the strong interaction potential; for
196
,The meaning of the damage "numbers approximated by

ReIchart et aT. (2002) is best illustrated for the case of
microprobpe analysis: In the case of ERD, damage numbers of
Ddisp = 10 displaced atoms and Dioo. = 1012 ionized atoms before
one atom is detected would mean that, in an analyzed volume of
I'ta"
.
II
J.lDl Con lOJDg a maxImum of 10 hydrogen atoms (i.e. a
concentration of 100%), all of these atoms would be displaced
once and ionized 1000 times before 100 counts of hydrogen
e~cn~ w~re detected. In nuclear reaction analysis (NRA), the
SItuatIon IS somewhat better, but microprobe analysis would still
not be possible at all at part-per-million levels of hydrogen.
Proton-proton scattering is the only alternative with damage
4
numbers ofDdisp = 10 and Dm = 107, for example, for 20 MeV
proton-proton scattering. As pointed out earlier in this section,
the release of hydrogen by ionization is much less severe than
\or ~eavier ion beams, and the damage nwnber Ddisp is the upper
~mlt ~or hydrogen release. This gives the unique possibility to
mvestlgate ~ydrogen concentrations as low as 1 at.-ppm (equal
to about 10' H atoms/}.UTr at microscopic dimensions using a
proton microprobe (Fig. 9.9) and even lower for many solids,
such as diamond, where bydrogen is much more stable (Reichart
el al., 2004a).
together with the corresponding scan position, one can generate

a two-dimensional image of the hydrogen distribu.1ion
illustrated in Fig. 9.9 for a mayfly wing (Dollinger el a[_, 2003)
with an image that reveals its capillary structure. The generation
of the map took only about 15 min. and did not destroy the
organic material, nor did the hydrogen yield decrease with the
applied dose. The energy analysis of both protons was used to
obtain information on the third dimension of the hydrogen
distribution with micrometer resolution, hence provid.:ing the
unique p~ssibility fOT three-dimensional hydrogen micros copy.
as
9.2.3 Hydrogen microscopy using a proton microprobe

Because of its low damage potential, proton-proton scattering
is the only technique that can be used with nuclear microprobes
to perfonn sensitive hydrogen microscopy. PerpendiCUlar
incidence in the transmission geometry does not deteriorate the
focus of the beam. Hence, by scanning the focused beam
laterally across the sample and recording the coincidence events
sample
: -"
~f
2D~ilppitig
1
L-l!!!!!."->','
depth profile
.j
!Eo
L.j- __ _
FIG. 9.9. (a) Principle of three-dimensional hydrogen microscopy by proton-proton scattering using a nuclear microprobe. (b)
Example of hydrogen microscopy performed on a mayfly wing using 10 MeV protons.
Lateral spread can be kept low when high-energy protons
are used: Figure 9.10 shows an SRlM2008 simulation (Ziegler et

al., 2008) of3 MeV and 20 MeV protons penetrating a diamond
sample. With 3 MeV protons, the full width at half-maximum
(fWhm), ffit, of the lateral spread is ffit = I J.ll1l at a depth of 30
J.UD. For 20 MeV protons, the value is only 6x = 140 nm at this
depth and still only 310 run at a depth of 50 JllIl. Hence,

submicrometer resolution in the lateral direction is possible
using. high-energy protons and targets thinner than 50 J.Illl
(Reichart et al., 2004b). Heavier target elements, however,
enlarge the lateral spread.
Chapter 9
J MeVprptmis .
20 MeV protons
30
sample depth (11m)
50
sample depth (11m)
FIG. 9.10. SRIM2008 simulation (Ziegler el al., 2008) of (a) 3 MeV protons compared to (b) 20 MeV protons penetrating a diamond
sample.
9.3 DEPTH PROFILING

9.3.1 Depth scale
With respect to the energy loss of the protons on their path
througb the sample, the sum of the energies of the two protons
gives a depth scale of the scattering events. For a first-order
approximation, the stopping forces along the path of the incident

beam and each of the scattered protons can be taken as constant
E (45') ~ E, - 2./2S",d -( S, - 2./2S",)z or
z( 45') = E..m
- Eo + 2./2S",d
2./2S", - s,
(9.14)
(Fig. 9.11). The mean energies of the protons leaving the sample
at angles of 8 and ~ are then given by
E,(e)~(Eo
-Soz)cos
d-z
e-s,-cos 8
and
E.(e)~(Eo -Soz)sin'e-s. d-z
(9.11)
SID 8
with the depth of the scattering event Z, the incident energy

and the stopping forces So = S(o) and
S,(e)~S[(Eo-Soz)cos'eJ
and
S. (8) ~ S[(Eo - Soz)sin 2 OJ

The sum energy, E~um =
Eo,
(9.12)
Ee + Et> is calculated as
Sa
S. (d-z).
E_(e,z,d)~Eo-Soz- --+-.( cose
(9.13)
SlOG
1
,
An easy approtimation to first order for the depth scale for

scattering at 45" can be obtained by assum..i.ng Se_w = St-4s' ~
S(ErI2) = SI12 to be constant for all values of z, so that
FIG. 9.11. Kinematics ofproton-protoD scattering. Scattering on
a sample with thickness d. A proton with incident energy Eo and
stopping force So is scattered at depth z. The outgoing protons
with stopping power Sa and S~ are scattered at angles S' and $,
respectively, and leave the sample with energies Eo and 4
198
For protons in the MeV range down to energies of a few

hundred keV, the stopping force can be approximated quite
accurately by the Bethe-Bloch formula without relativistic
corrections
(.18)
with Zp = I, can be seen to depend on the square root of the

electron density of the sample matrix, Ile'= nTZT.
SeE) = _ dE = c, In(c,E) ,
dx
E
(9.15)
Contribution (4) is geometrical effects, bEg~(e,i,d). One

can see from Eq. (9.3) that, to first order, the kinematic effect
disappears. However, to second order, the analysis of thick
targets involves two additional effects, as discussed in the next
two subsections: the correlated-path-length effect, OEpailu and the
small-angle-scattering-path effect, 5Ea .
with the constant terms Cl and C2 in Sl units being given as

(9.16)
where e is the electronic charge in SI units, me is the electron

mass, Zp = 1 is the atomic number (of hydrogen), Mp = mp is the
projectile mass, ~ = DTZr =(NAIAT)PTZT is the electron density
of target matrix, and I : : : 1 L5Z r eV is the mean ionization
potential (e.g., from Berger et ai., 1993). Therefore, Eqs. (9.14)
can be calculated simply using S'I2 ~ 2ln(c,EoI2)So and So
c, In( c,Eo)/Eo.
With this approximation, one can calibrate the depth ::cale

from the sum energy of the coincident protons_ [Note thai the
offset of the scale depends on the total thickness of the sample,
but not the relative calibration.] Tbis is done in Fig. 9.6 as an
example for the sandwich sample consisting of five Mylar foils
separated by sheets of 60-f.UIl-thick Si. From this figure, one can
see that the depth resolution improves with positive depth z
because of the decreasing energy spread, as explained in the next
.subsection. The actual depth scale deviates because of the
constant-stopping-force approximation. For exact calibration of
the depth scale, Eqs. (9.13) and (9.14) have to be replaced by
numerical analysis (slab technique) with nonconstant stopping
forces.
9.3.2.1 Correlated-path-length effect, BEp, "

The uncertainty, 00, in the determination of the scattering
angle makes a contribution 5Epath to the total energy spread. This
is because the total energy loss of the protons changes with
scattering angle as a result of both different path lengths and
different stopping forces. Therefore, for an uncertainty of 89 in
the scattering angle, the contribution of the so-called correlatedpath-length effect is obtained as
(8 d,1'fJ\ _{ ~
s..., + S, S..., ] d-z
~""pm;z, ) - 1XIS6 ca;(e+l'fJ) sine sin(e+88) (
).
);1'
(9.19)
This equation can be expanded as a Taylor series to second order

in 50. within the Bethe-Bloch energy range, assuming the
logarithmic tenns to be constant for all involved energies {e.g.,
assuming In[m,,M,-'j'E(e,z)] = In(m,,M,-'r'E,)}, as follows:
Il'p.~{{: :Ja3t-l~~l ~~l)&~}(d-?)

(9.20)

Because of low stopping force of high-energy protons, the
depth resolution from proton-proton scattering is moderate,
being in the range of 1 f1IIl. From Eqs. (9.13) and (9.14), one can
see that the resolution of the total energy E~llm! limits the depth
resolution. The contributi~Ds from (1) the detector energy
resolution, oEdeh and (2) the beam energy spread, 15Ebcam , are
given by the setup. Contribution (3) is the energy loss straggling,
oEstragg(8,z,d), of the incoming and emerging protons, which
causes an energy spread as the quadratic sum of the
contributions of each proton path
BE ... (8,z,d)=C,_ (_I_+_._l_)(d_Z)+Z
"
cos8
sm8
This approximation gives about 20% higher values than an

exact calculation of Eq. (9.19) and can therefore be taken as a
good upper limit for bE,.. At = 45, the linear tenn of Eq.
(9.20) is zero. Thus, only second- and higher-order tenns
influence depth resolution. This changes when large solid angles
of detection are applied using outer scattering angles far from
45 for detection (Reichart et aI., 2002).
9.3.2.2 SmaH-angle-scattering-path effect, SEa

Even if be = 0, the angular spread of the protons give a
spread in the path length of each proton trajectory and, hence, a
spread in energy oEa(O,z,d) (Reichart ef a/., 2002), given by
(9.17)
.
d-z
8E.(8;z,d)=2
In Eq. (9.17), the factor c,n", given in SI units as
199
)[
sine
00;8
' + s,-.-, fu,-,,== )' ,(9.21)
(~-,-fu
~
rosS
=e
sm8
sine
Chapter 9
where 00. is the angular spread caused by small-angle scattering

along the path with length x (Sigmund and Winterbon, 1974).
9.3.2.3 Energy-difference method
In conclusion, the geometrical effects predominantly
deteriorate the depth resolution when the angular spread or
uncertainty, 08. grows large, especially when one uses a large
solid angle of detection (outer scattering angles). The detector
aperture otp, the small-angle scattering spread 00., and possibly a
beam divergence 013 ~ independent contributions to 09, so that
69 = ~6a.' + 6'1" + 6P' .

However, the contributions of a finite detector aperture, olp,
and the small-angle scattering spread, Sa., can be eliminated if
one uses the kinematic infonnation that is inherently available
from the energy difference of the two protons, dE = Ee - Et .
With this information, ODe can achieve a very accurale
determination of the scattering angle, as suggested by Sjoland e/
al. (1997), because llE is strongly dependent on the scattering
angle. The achievable angular accuracy is given by (Reichart et
al., 2002)
""r
(cose sine
ae"'="12E,sin28+3(d-Z)S(E,) sin'S + cos's
J,r~
(9.22)
.-- where Ez = Eo - SoZ and the overall energy resolution is given by

(9.23)
with the contributions from above as a function of 9, z. and d.
- -~-~ -wiiliiliis...;~~~~ch. an angular resolution of less than 10
mrad is possible even for scattering angles of around 60
(Reichart et a/., 2002). Note that the beam divergence at
microprobes is of the same order of magnitude and might
contribute dominantly as well.
93.2.4 Sum of contributions
Finally, the contributions to the energy spread can be
compared and added quadratically if all are assumed to be
independen~ so that SE~' ~2SE...' + S~' + SE.' + oE,...'.
Figures 9.12(a) and 9.12(b) show the depth resolution resulting
from the above relations for a diamond sample using 3 MeV and
20 MeV proton beams, respectively. For the outer scattering
angles (i.e., 35'/55') the geometrical path-length effects grow to
be a main limitation on the depth resolution for thick targets.
One can correct for the correlated-path-Iength effect by applying
a good detector angular resolution (thin dotted line) or by
detennining the angle using the energy-differeoce method (thin
solid line). The top thick lines give the overall resolutjon with
the latter angle determination method (solid) or with the angular
resolution of the detector fixcd at ocp = 10 mrad (dotted line).

The lower eliagrams [Figs. 9.12(c) and 9.l2(d)] show the
calculated overau depth resolutions at different scattering angles
for an iron sample (note the different scales compared to the
above diagrams).
As a result, a relative depth resolution of .5-10% of the
target thickness can be achieved for light target elements or
10-20% for heavier elements, depending on the maximum
scattering angle. This is sufficient and essential to separate
hydrogen surface peaks from bulk hydrogen that would
otherwise destroy the good sensitivity for hydrogen in the bulk
(Fig. 9.3).
9.4 EXPERIMENTAL REQUIREMENTS

As stated in Section 9.2, it is very easy to achieve part-permillion sensitivity with only two opposite detectors and a
standard coincidence readout system (Coben et al. 1972).
Hence, proton-proton scattering is possible with low effort at
every ion beam laboratory that is using MeV protons. Hydrogen
microscopy is possible using a proton microprobe, but in order
to achieve a reasonable imaging time and low irradiation
damage, a large solid angle of detection and a setup with
multiple coincidence detectors is essential (Reichart et a/. ,
2D04b). The most crucial restriction is due to transmission
geometry. The demands are a minimum proton energy with
respect to the sample thickness and also a minimum detector
thickness in order to stop the emerging protons. These points are
discussed in the following subsections .
9.4.1 Sample Thickness
The maximum range of the protons limits the sample
thickness in transmission geometry. Even more critical for
detection of the protons, their energies must be above the
threshold of the detector. Staying above a minimum outgoing
energy of about Emm = 100-200 keV is recommended because
small-angle or multiple scattering effects will have a critical
influence on coincidence analyses below that energy. Figurc
9.13 shows the maximum sample thickness where a steeply
scattered proton has an outgoing energy of Emin = 200 ke V at
scattering angles of 45, 50. 55, and 60 for various sample
materials.
For proton energies above 200 keY, the Bethe-Bloch formula
is a good approximation for the stopping force. Taking
logarithmic terms as constant, one can calculate analytically the
maximum sample thickness at a minimum outgoing energy of
Emmas
d~
cosS
.(E'
4e
E')
o cos
-min
c,21n(c,E, cos' 8)
(9.24)
with CI and C2: as in Eq. (9.15). The maximum sample thickness

calculated by this approximation is plotted for Si as the dotted
line in Fig. 9.l3. It gives lower values than the true maximum
thickness but, hence, a good threshold to be on the safe side.
200
Hydrogen Analysis bv Proton-Proton Scattering
;)
,.
.,
.: pfldIl::d~in\N:iI\gi!br~.oos"";wo round)
--::-:-'P3!II~AIO,..(lm#_hy~d!~Jb-ao".) -.
-:G,J
'_,..;.'~_~-::=:_~;;..,dt~_~~._
bijO
<. .. ':
.'iJ
'u
<1':.~~~,,",",,!~I'<""~pr_ilf:.. ~b:y~p.)
-'.;', row_ In-l.urabt_l"tSlIbtIIIIiIli...nd)
...... 1DiaJ.1l.. (Ba;kby.ICt";O"ei".m.c.J
".,. .
It.::vn~.spr=l~~.o,b=ndk~aP .. Q}
,Diamonii
,.,
E.:03W
. 8=450
.,
,, .
..
Diamond
Diamond
Eo =3MeV
8 = 35/55
E. = 2.0 MeV
e = 35/55
ma.ximum
thickness
".
..
'00)
maximum
tllicknCis
'"
d)
'l
Iton
. Iron
" 3 MeV
20 MeV
dS
"
FIG. 9.12. The different contributions for depth resolution/energy spread as labeled for a diamond sample (nT = 1.7 x lOll at./cmJ)
using (a) 3 MeV and (b) 20 MeV protons depending on the target thickness d. (c) and (d) show the overall depth resolution only for an
iron sample at different scattering angles (beam divergence Sf), 0 mrad; detector energy resolution per proton, 1%, with a minimum of
20 keV).
.
201
Chapter 9
.:,._-_., . .- .
.~';.';-;.~~'.:~ ";...,v\,,,, '.t" ,b
',.....:~u~
. .'. .
~IA~"'"'
~~17_
_ ..
',;,_".,.,/_,~ ,~
...
"
"
-:+-~ ...
'.
--m,1.I>
.......
__
: J
OM.IU~
.
..
. l
.)
J
,f,
,,
.,. ,
~~~;...,..:-;~.,.....~~
~'"
TltlO
J "lUI
'~j'1'~1\1
~
, , 1 ~
1I
,,/",
.
JOO"'"
JI
O.. Str
:;J\~(t .. ~.~
.., ~ III O.. {O"
31)0"
itridcmJl'OlmClOl\l)'E.~"*'
20
FIG. 9.13. Maximum sample thickness, d max at which the proton

scattered from the front side of the sample has a minimum
outgoing energy ofEmin = 200 keY. The calculation was done for
four different scattering angles and scaled to four different sets
of energy and thickness axis, as labeled. The projected range
data were taken from PSTAR (Berger el ai., 1993, 2005) for
various sample materials as labeled. The dotted line shows, as a
comparison, the vaJues calculated using Eq. (9.24) for silicon
(E",,"' 200 keY).
9.4.2 Detector Thickness
The incident proton energy has to be chosen from Fig. 9.13 so
that the required sample thickness is transmitted by protons
scattering at the steep angle. On the other hand, the maximum
possible energy is given by the- range of the fiat-angle-scattered
protons inside the detector for full energy analysis and depth
profiling. Stack detectors are JX)ssible. but they are expensive
and are not recommended for total energy resolution. Thick Si
detector material with a depletion depth of ~ = 1000 r..un or
above is available today with an acceptable energy resolution of
50-60 ke V even for large-area detector chips. Tbis is sufficient
for most accelerators, but at proton energies above 17 MeV, it is
a serious constraint because the scattered protons will not stop at
all. It is therefore much more favorable to use a parallel detector
mounting as shown in Fig. 9.1(b). The scattered protons with the
maximum energy Emu. are then stopped at a much greater
stopping length of do.,!cos 8"",.
Figure 9.14 shows the possible maxUnum incident energy,
for four different depletion thicknesses of Si detectors
depending on the scattering angle, so that both scattered protons
are stopped in the detector.
Eo,
30
40
Scattering Angle
b)
50
>
~ 40
50
60
Pam lle! Detector Mounting
,.,
)Depletion Depth 51:

300 500 1000 1500 JllTl
o+---',''='~;-'--~~Jor-. .,..","",~.,..,4TO~~-OO"T"--';--."'~
.S"'eri!i9 Angie a .
.
. " ,
FIG. 9.14. Maximum possible incident energy Eo.max so that all

scattered protons at a scattering angle of 8 are completely
stopped in the depletion region of a Si detector chip with typical
available depletion-region thicknesses of 300, 500, 1000, and
1500 !lID for (a) perpendicular mounting [see Fig. 9. 1(a)] and (b)
parallel mounting [see Fig. 9.1(b)J. Stopping force and range
data were taken from the PSTAR database (Berger el al. 1993,
2005), neglecting energy loss in the sample.
:'~ ~.
..
202
9.S ADDITIONAL REMARKS

9.5.1 Relativistic corrections to sum of scattering angles
The appropriate protan beam energy for pp-scattering
hydrogen analysis ranges from about 3 to 30 MeV, which gives
velocities of p= vIc = O.08-{).25. The "relativistic" or "apparent"
mass of the incident proton in the laboratory frame, given by lIlp'
== (Eir/mpc2 + l)mp, increases by about 0.3-3% compared to the
rest mass, ffip. The relativistic calculation with the target particle
at rest gives the dependence of the laboratory scattering angle ~
on the laboratory scattering angle of the other proton as
(1) The two scattered protons are indistinguishable. Thus, it

makes no sense to trace which proton was initially at rest, and
scattering by an angle of e cannot be distinguished from
scattering by an angle of x - 6 in the CM system [Fig. 9.16(a)].
In quantum mechanics, the scattering amplitudes f(6) and
f(S - 1t) have to be added for indistinguishable particles rather
than the cross sections themselves, so that,' including
consideration of exchange symmetry, one obtains
~ = arctan
-sine
]
[ y=(cos6=em -1)
(9.25)
with y= from Eq. (9.29) and the center-ofmass (CM) scattering

angle 9= from Eq. (9.30).
For proton energies of3 MeV and 30 MeV, Fig. 9.15 shows
the exact sum of the scattering angles + $, which is obviously
Dot exactly 90, but the deviation is negligible compared to the
effect of small-angle scattering in realistically thick samples.
(9.26)
for each state of total spin J, that is, J = 0 (singlet) or J = 1
(triplet) for proton-proton scattering [Fig. 9.16(b)]. In general,
the scattering cross section has contributions from all spin states.
If scattering is independent of the spin, s, of the parti<:les (no
spin flip) and no spin polarization occurs, the scattering
probability can be obtained by incoherent addition of the
different spin states according to their statistical weights a<;
(9.27)
The result is the typical interference term and, hence, an angular

dependence of the scattering probability, as discussed later in the
subsection on Matt scattering (Section 9.5.2.1).
,,,o'-~~~~~~~c.....
..
(2) The second effect results from the Pauli exclusion

principle. The wave function of the two prolons has to be
antisymmetric under their exchange. Hence, a symmetric space
wave function (S, D, G, etc., states) can only be associated with
an antisymmetric spin wave function (even J, i.e., pp singlet),
whereas an antisymmetric space wave function (p, F, ctc., states)
requires a symmetric one (odd J, i.e., pp triplet). Therefore, only
tbe singlet (J = 0) can be in the S state (angular momentum e =
0), and the triplet states have at minimum t = 1, meaning a P
state. At 10 MeV, the cross section is dominated more than 99%
by S-wave scattering (Bergervoet et 01., 1988, 1990). The
contribution of the P states increases with energy but is still a
minor factor below 30 MeV. In all cases, the P-wave scattering
cross section is zero at lab = 45 in the labomtory frame because
of the interference in the identical-particle system. Higher-order
even states (such as D states) have an even lower scattering
amplitude than the P states in the energy range of interest. In
conclusion, there is only significant scattering cross s~tion from
S-wave scattering, and hence, only the singlet state contributes.
The S-wave scattering cross section is nearly constant over a
large range of angles around Slnb = 45 (see data for Fig. 9.5).
_ _ _ _ _':""_ _-,J
O'(lab~)
FIG. 9.15. Swn of scattering angles of both protons, 8 + $, with

relativistic corrections for incident photon energies of 3 MeV
(dashed line) and 30 MeV (solid line).
9.5.2 Quantum effects due to identical particle scattering

Proton-proton scattering is scattering of identical particles.
The identity of the two protons has two effects (e.g., Joachain,
1984; Blatt and Weisskopf, 1991):
203
/
Chapter 9
b)
"Pm'
;,,'\' '
, . -.s
1\1,)"
'" . ,......"il4)
f7~l
I
T.2./
T
'/","'"
.
'P m =0 or'S m
'.
's
'",
'.
=0
FIG. 9.16. (a) Two different identical-particle collisions that are indistinguishable. (b) Plot of the four principle spin states of the pp_
scattering event, including the resulting lowest-order scattering states. The antiparallel adjustment of the spins is the only way to fonn,
at a 50% probability, the singlet IS state. Only this singlet contributes significantly to the s[2ttering cross section below E = 30 MeV.
The very different scattering cross sections from S and P
transfers (large scattering angle) and overlayed with some
states result in dramatic spin-dependent scattering count rates if
wiggles due to the interference effects from identical-particle
the spins of the protons in the beam and in the target are adjusted
before scattering. Parallel spin orientations result in a nearly zero
count rate, whereas antiparallel adjustment results in count rates
that are twice those. obatined in the unpolarized case: the
probability to obtain S-wave scattering is 50% for the
scattering (interference tenn~) and probably also interference of

Coulomb scattering and a strong interaction potential.
antiparallel situation, 0% for parallel orientations. and 25% in

the unpolarized case. Tb.is spin-selectivc scattering could be used
for experiments with polarized beams and targets in H analysis
as described in Dollinger el a/. (2006).
9.5.2.1 Mott scattering _.

For an ideal Coulomb potential, the scattering probability was
derived by Mott (19,0). It was first discussed for Coulomb
scattering of electrons and extensively studied for various nuclei
(e.g., Bromley el al.. 1961; Bethge el 01., 1967; Giinther el aI.,
1967; Richter et al. 1977). The "wiggles" in thc scattering cross
section due to the above-mentioned interference tenn are most
easily observed for heavy nuclei, and accelerator laboratories use
this effect for demonstrative student laboratory teaching of
quantum effects (e.g., UnivelSitllt zu Koin. Techniscbe
Universitat Mfinchen, and Ludwig-Maximilians-Universitat
MUnchen). The term Mott scattering is also commonly"used in a
~ore general way for spin coupling in elastic Coulomb
scattering. The Matt-scattering cross-section formula is given in
Appendix 14 in Eq. (AI4.1).
In fact. the actual scanering cross section, for example, at 20

Me V incident proton energy is enhanced by up to a factor of 500
compared to this Matt-scattering cross section because of
nuclear scattering effects. It is nearly constant for large energy
204
9.5.3 Relativistic transformations

For exact calculations of scattering cross sections, the
transformations from laboratory system to CM system have to be
done relativistically. We give here the transformation equations
for the specihl case that the target proton with mass Dlp = 938.27 .
2
MeV/c is at rest in the laboratory system and the incident proton
has kinetic energy, Em. in the laboratory system. Then, the
kinetic energy of the center of mass reads
Ecm
=__,.;;E~i',===;
1+
(9.28)
1+~2
2mp c
which equals Ein1'2 in the nonrelativistic case. The Lorentz factor

for the center of mass of the two protons is given by
rem=
E.
--m-2+1.
2mp c
(9.29)
"
The transformation of the scattering angle from the laboratory
system to the eM system is then calculated as
1- r =2 tan 2 e
San = arccos [
2
2
l+y= tan
(9.30)
.J
and the transformation from the eM to the laboratory system is

accomplished with
8::: arctan
I-cosO C I l lI_
[ 1+cos9 Yem
cm
em
l+co,O
(9.31)
sin 9cm + Ycm 2 (cos 9CI:l +1)2
J.H. (1948), Phys. Rev. 74 (5), 553.
Bromley, D.A., Kuehner, J.A., and A1mqvis~ E. (1961), Phys.

Rev. 123 (3), 878_
Burkig, J.W., Richardson, J.R, and Schrank, G.
Rev. 113 (I), 290_
J3/2 '
(9.32)
(9.33)
Hence, the differential scattering cross section transforms as
Cohen, B.L., Degnan, J.H., and Fink, C.L. (1972), J. Appl. Phys.
43 (I), 19.
dcr
dO
dcr
(6,m) = ill (6=) d[l[6(6,mll or
=
'm
~ (0) = { ~ [9= (e)f ~:m [0= (O)J.
E. (1959), Phys.
Bystricky, J., Carlson, P., Lechanoine, C, Lehar, F., Monnig, F.,

and Schubert, K.R (1980), Elastic and Charge Ecxchange
Scattering of Elementary Particles (Hellwege, K.-H., and
Schopper, R., eds.), Landolt-Bomstein Tables (New
Series, Group I), Numerical Data and Fu.nctional
Relationships in Science and Technology, Springer,
Berlin, voL 9a, pp. 110ff.
from which on-e obtains the simple nonrelativistic approximation
dO
i.M., Freier, G., Lampi, E.E., Sleator, W., and Williams,
Blatt, !.M.,and Weisskopf, V.F. (1991), Theoretical Nuclear

Physics, Dover, New York.
Finally, the transformation of the solid angle is given by
dO
ill =Y= [
Blair,
Dayton,I.E., and Schrank, G. (1956), Phys. Rev. 101 (4), 1358.

Dollinger, G., Reichart, p.. Datzmann, G., Hauplner, A., and
Kamer, H.-J. (2003), Appl. Phys. Lett. 82, 148.
Dollinger, G., Bergmaier, A.. Hauptner, A., Dietzel. S., Drexler,

G.A., Greubel, C, Hable, V., Reichart, P., K.n1ckcn, R.,
Cremer, T., and Friedl, A.A. (2006), Nucl. lnstrum.
Methods B249, 270.
(9.34)
REFERENCES
Allred, J.C, Armstrong, A.H., Bondelid, R.O., and Rosen, L.
(1952), Phys. Rev. 88 (3), 433.
Dujrnic, D., Jaksic, M., Soic, N., Tadic, T., and Bogdanovic, I.
(1996), Nuel. instrum. Methods Bill (1-2), 126.
Green, P.F., and Doyle, B.L. (1995), "Application of ion beam
analysis techniques to polymer science", in Scattering
Methods in Polymer Science (Richards, R.W., ed.), Ellis
Horwood, London, pp. 19'1-218.
Berdoz, A., Foroughi, f., and Nussbaum, C (1986), 1. Phys.

G12 (6), L133.
Berger, M.l, Coursey, IS., Zucker, M.A., and Chang, J. (1993),
Stopping Power and Ranges for Protons and Alpha

Particles, ICRU (International Com.mission on Radiation
GilDther, G., Bethge, K., and Schubert, K.R. (1967). Nucl. Phys.
AlOI (2), 288.
Units and Measurements) Report 49, Oxford University

Press, Oxford, UK.
Hofsiiss, H.C" Parikh, N.R., Swanson, M.L, and Chu, WoK.

(1991), Nucl.lnstrum. Methods B58 (I), 49.
Berger, MJ., Coursey, J., Zucker, M.A., and Chang, J. (2005),

EST AR, PSTAR, and AST AR: Computer Programs for
Calculating Stopping-Power and Range Tablcs for
Electrons, Protons, and Helium Ions (version 1.2.3),
National Institute of Standards and Technology (NIST).
Gaithersburg, MD; http://physics.nist.gov/Star (accessed
JuI13,2007).
IAEA Nuclear Data Services (2009), International Atomic

Energy
Agency,
Vienna,
Austria;
bttp:llwwwnds.iaea.org.
!mai, K., Nisimura, K., Tamura, N., and Sato, H. (1975), Nucl.
Phys. A246 (I), 76.
.
Jannie, N_, Jett. J.H., Detch, J.L., and Hutson, RI.. (1970), Phys.
Rev. Lett. 25 (I), 34.
Bergervoet, 1.R, van Campen, P.C., van der Sanden, W.A., and
de Swart, J.J. (1988), Phys. Rev. C38 (I), 15.
Jannie, N., Jett, 1.H., Detch, 1.1.., and Hutson, RI.. (1971), Phys.
Rev. C3 (I), 10.
Bergervoet, 1.R., van Campen, P.C., Klomp, R.A.M., de Kok, J.L., Rijken, T.A., Stoks, V.G.!', and de Swart, J.J. (1990),
Phys. Rev. C41 (4), 1435.
Bethe, H..A. (1999), Rev. Mod. Phys. 71 (2), S6.
Bethge, K., GilDther, G., Schubert, K.R., and Speth, E. (1967),
Phys. A202 (I), 70.
Jannie, N., and Jett,!.H. (1974), Phys_ Rev. CIO (I), 54.
Jannie, N., Jett, J.H., and Semper, R1. (1974), Phys. Rev. C10
(5), 1748.
z.
Jeong, T.H., Johnston, I..H., Young, D.E., and Waddell, C.N.

(1960), Phys. ReY. 118 (4), 1080.
205
/
Chapter 9
Jo.chain, C.J. (1984), Quantum Collision Theory, Elsevier
Science, Amsterdam.
Johnston, L.H., and Young, D.E. (1959), Phys. Rev. 116 (4),
989.
Ki1ruchi, S., Sanada, 1., Suwa, S., Hayashi, I., Nisimura, K., and
Fukunaga, K. (1960), J. Phys. Soc. Jpn. 15 (1), 9.
Knecht, DJ., MesseIt, S., Bernen;, E.D., and Northcliffe, L.C.
(1959), Phys. Rev. 114 (2), 550.
Komatsubara, T., Sac;a, K., Okumoto, K.., Ishii, S., Yamato, Y.

Satou, K., Furuno, K., and KurDsaw.. M. (2006), Nuel.
Instrum. Methods B251 (I), 237.
Martinsson, B.G., and Kristiansson, P. (1993), Nud. Instrum.
Methods B82 (4), 589.
Mot~ N.F. (1930), Proc. R. Soc. London A126 (801), 259.
NNDC Databases (2009), National Nuclear Data Center,
Brookhaven
National Laboratory,
Upton,
NY:
http://www.nndc.bnl.gov.
Paduschek, P., and Eichinger, P. (1980), Appl. Phys. Lett. 36 (1),
62.
Paduschek, P., and Eichinger, P. (1981), NucJ. Instrum. Methods
191 (1-3), 75.
Reichart, P., Dollinger, G., Bergmaier, A., Datzmann, G.,

Hauptner, A., Korner, H.-J., and Krticken, R. (2004b),
Nuel. Instrum. Methods B219, 980.
Richter, M., Henning, W., Komer, H.1., MillIer, R.. Rehm, K.E.,
Rother, H.P., Schaller, H., and Spieler, H. (1977), Nuel.
Phys. A278 (1),163.
Schiettekatte, F. (2008), Nud. Instrum. Methods B266, 1880.
Sigmund, P., and Winterhon, K.B. (1974), Nucl. Instrum.
Methods 119, 541.
Sjoland, K.A., Kristiansson, P, Elfinan, M., Mahnqvis~ K.G.,
Pallon, J., Utui, R.J., and Yang, C. (1997), Nuel. Instrum.
Methods B124 (4), 639.
Slobodrian, R.I., Conzet~ H.E., Shield, E., and Tivol, W.F.
(1968), Phys. Rev. 174 (4), 1122.
Slobodrian, R.I., Frois, B., Birchall, I., and Roy, R. (1976),
NucJ.Instrum. Methods 136 (3), 525.
Szihigyi, E., Paszti, F., Arosel, G. (1995), N"eI. Instrum.
Methods BIOO, 103.
Wassmer, H., and Muhry, H. (1973), Helv. Phys. Acta 46 (5),
626.
PanDfsky, W.K.H., and Fillmore, F.L. (1950), Phys. Rev. 79 (l),

57.
Wegden, M., Kristiansson, P., Skogby, H., Auzelyte, V., Elfinan,

M., Malmqvist, K.O., Nilsson, c., Pallon, J., and Shariff,
A. (2005), Nuel.Instrum. Me/hods B231 (1-4),524.
Reichar1, P., Dollinger, G., Bergmaicr, A., Datzmann, G.,

Bauptner, A., and KOmer, H.-J. (2002), Nud. Instrum.
Methods BI97, 134.
Worthington, H.R., McGruer, I.N., and Findley, D.E. (1953),.

Phys. Rev. 90 (5), 899.
Reich~ P., Datzmann, G., Hauptner, A., Hertenberger, R.,

Wild, c., and Dollinger, G. (2004.), Science, 306, 1537.
206
Yntcma, J.L., and White, M.G. (1954), Phys. Rev. 95 (5),1226.

Ziegler, 1.F., Biersack, J.P., and Ziegler, M.D. (2008), The
Stopping and Range of Ions in Matter, LuLu Press,
Morrisville,
NC.
See:
http://www.srim.org.
CHAPTER
1m
CHARGED PARTICLE
ACTIVATION ANALYSIS
G. Blondiaux and J.-L. Debrun
CNRS-CERI, Orleans, France
c.J. Maggiore
CONTENTS
10.1
INTRODUCTION ............................................................................................................... 209
10.2
THEORY ....................................................... ' .................................................................... 209

10.2.1 General principle of radioactivation ..................................................................... 209
10.2.2 Creation of radioisotopes byirr.diation ............................................................... 209
10.2.3 Stand.rdization for quantitation ........................................................................... 209
10.2.3.1 Approximations based on range .................................................................. 211
10.2.3.2 Numerical integration ................................................................................... 211
10.2.3.3 Approximations based on stopping power ............................................... 211
10.3
EXPERIMENTAL TECHNIQUES ................................................................................... 212

10.3.1 Reaction choice ........................................................................................................ 212
10.3.2 Irradiation procedures .......................................................................................... 214
10.3.3 Post-irradiation chemistry ..................................................................................... 214
10.3.3.1 Etching ..........................:........~...................................................................... 214
10.3.3.2 Radiochemical separations ........................................................................... 215
10.3.4 Measurement of the radioactivity .......................................................................... 216
10.4
SPECIFIC APPLICATIONS ......................................................................:....................... 216

10.4.1 Carbon ..................................................................................................................... 217
1004.1.1 12C(d,n)lJN ...................................................................................................... 217
207
/
Chapter 10
10.4.1.2 "q'He,a) II C .................................................................................................. 217

10.4.1.3 Otber reactions for carbon ........................................................................... 217
10.4.2 Nitro~en ................................................................................................................... 217
10.4.2.1 'N(p,n)"O ..................................................................................................... 217
10.4.2.2 'N(p,a)"C ...................................................................................................... 217
10.4.2.3 "N(d,n)"O ..................................................................................................... 217
10.4.2.4 'N(t,p)'''N....................................................................................................... 218
10.4.3 Oxygen ..................................................................................................................... 218
10.4.3.1 "O(p,a)"N ..................................................................................................... 218
10.4.3.2 "O(,He,p) 18F .................................................................................................. 220
10.4.3.3 'O(a,d) 18 F ...................................................................................................... 220
10.4.3.4 l60(t,n)18F ....................................................................................................... 220
10.4.4 Channeled activation for C and 0 ................:........................................................ 220
10.45 Analyses of other elements ..................................................................................... 224
10.4.6 Thin layer activation (TLA) for studies
on wear and corrosion ............................................................................................ 225

10.5 CONCLUSION ......................................................................................................................... 226
REFERENCES
............................................................................................................................ 226
208
Charged Particle Activation Analysis
10.1 INTRODUCTION
Charged particle activation analysis (CPAA) can, in principle,
be used to determine almost any element, in solids or liquids,
over a wide range of concentrations (from less than the part~per~
billion level to the percent level). However, taking into account
practical considerations and the availability of nwnerous other
analytical techniques, the important application of CPAA is to
anaI}'2e light elements such as C, N, and 0 at trace levels in
solids. For this application, CPAA is the best reference method
for the following reasons:
10.2.2 Creation of radioisotopes by irradiation

During irradiation, radioisotopes are created by :nuclear
reactions and immediately begin to decay, so that the
instantaneous rate of production for a thin target is given by
dW'I dt =
nm dx - A.7V ,
(10.1)
where J{ is the number of radioactive atoms al lime i, Q is the

number of atoms to be analyzed per unit mass, cr is the nuclear
cross section for the reaction of interest, i is the nlll1::lber of
incident ions on the target per unit time, dx is the target
thickness (glcm2), and)" is the decay constant of the radioisotope
of interest.
Chemical contaminations do not introduce significant errors

in the measurement, which is especially important in the case of
ubiquitous elements such as C, N, and o.
The decay constant is A= In(2)1', where, is the half-life of

the radioisotope.
Detection limits are good because, unlike "in~beam" analysis,

CPAA uses radioisotopes that can be separated from disturbing
backgrounds and then measured.
910, the number of radioactive atoms at the end of an

irradiation time lj, is obtained by integration as
As in other nuclear methods, absolute calibration is direct,

easy, and free from matrix effects.
(10.2)
This chapter presents only essential infonnation. More details

on activation analysis, radiochemical procedures, counting, and
so on can be found in De Soete et al. (1972). For more details on
charged particle activation itself, readers should consult
Vandecastccle (1988). Appendix 15 of this book contains a
compilation of cross sections and yield curves used in ,CPAA.
The activity at the end of the irradiation, the total number of

disintegrations per second, is
~
= 'A!NfJ (becque~els).
(IOJ)
The term (1- e -..~) is called the saturation factor. After a

rapid growth, it approaches 1 asymptoticaUy as shown in Fig.
10.1. The activity at time t". after the end of irradiation (the
origin of time is now the end of irradiation), is given by
10.2 THEORY
10.2.1 General prinCiple of radioactivation
The sample to be analyzed is "activated" by irradiation with
an ion beam of suitable type and energy. Artificial radioisotopes
such as in EXAMPLE 10.1 are created by nuclear reactions on
the matrix itself and o~ the impurities of Interest Qualitative
analysis is bascd on the identification of the created
radioisotopes, mainly using high~resolution gamma-ray
spectroscopy. Quantitation is based on the correspondence
between the stable isotope present in the sample and the
radioisotope produced by the beam. It depends on the direct
relationship that can be established between the concentration of
the element (isotope) of interest and the counting rate (activity)
of the corresponding radioisotope.
(lOA)
where d stands for decay.
Cross sections are functions of energy, and because the
incident ions lose energy as they traverse the sample, cr can also
be thought of as a function of depth, x. The activity for a thick
target at the end of irradiation is then given by
J,SeE)
A =ni(l-e-,-,,)E a(E) dE
o
(10.5)
EXAMPLE 10.1. Activation for analysis.

where E; is the energy of the incident ions in the sample and
SeE) is the stopping power, dE/dx. See Chapter 2 for a
discussion of stopping power.
Chromium can be analyzed by the (p,n) reaction on the stable

isotope, HCr, which produces 52Mn. This nuclide is radioactive
and emits characteristic gamma rays (1434 keY, 935 keY, and
744 keV) with a half-life of 5.6 days. Detection of 52Mn after
irradiation of the sample with protons 15 MeV) indicates the
presence of chromium because, below 15 Me V no other element
can produce S2Mn by a proton-induced reaction.
10.2.3 Standardization for quantitation

In principle, thc concentration, n (atoms per gram), of the
element to be analyzed can be obtained using Eg. (10.5), by
measuring both the absolute activity of the radionuclide of
interest and the integrated beam intensity. The absolute cross
section and the stopping power of the incident ion for numerical
integration of the integral must also be known. It is not easy to
209
Chapter 10
measure absolute activities and intensities, and the absolute cross

sections are not very welJ known either. Thus, for more accurate
unknO'ND. are used. in place of the activities in Eq. (10.6). By

knowing the concentration, Os, in the standard and measuring the
analysis, it is common to measure relative to the activity of a
ratio of the counting rates in the standard and the unknown, the
concentration in the unknown, Dx. can be calculated if the ratio
standard. The sample and the standard are irradiated at the same
energy with appropriate total charges, Qil and the ratio of the
activities after irradiation is compared. The ratio of the
concentrations is then given by
(I_e-
,),
fa(E) dE
Q"
S,(E)
Q, "fa(E) dE
A,
of the integrals, '1(, in Eq. (10.6) is calculated Several methods

~ The choice of an appropriate
method depends on the precision required and the information
available on the activation cross section as a function of energy.
have been proposed to calculate
.. .. , S, (E)
EXAMPLE 10.2. Corrections for counting times and delays.

Consider the case with no interferences and an activation
product with a half-life of 10 min. [1. = 1n(2)" = 1n(2)/10 =
0.693/10 = 0.0693]. The sample activation begins at 1:17 and ends
at I :39 for a sample irradiation time of lj = 22 min. If the beam
current is 0.1 J.lA, then the integrated charge is 132 ).LC. If the
counting of the sample begins at 1:44 and ends at I :55, then the
counting time is fc = 11 min., and the delay time is ~ = 5 min.
Then, according to Eq. (10.7), the count rate for the unknown at
the end of the irradiation is
(10.6)
A_-',"__ -'(I-e-"')
__
_ _---'"'-1 Q
(l-e-~'),
A,
Q,
I~.
End of Irradiation
C
Saturation
100
90
70
'-.
';; 60
... ~ 50
<t 40
30
Increase of
activity during
Irradiation
20
Decay of activity
after the end of
Irradiation
A, = C(1-e"'Ij
10
0L..l.-'---'----'--''--1.-'---'----'-L..L-'''=''--..J
o 1 2 3 4 56 1234567
e"'C
= 0 184C
~.
0.0693
(1- e-o.0693xJ)
eO. 06
..
x'
=1 199C
9)' tJ C
.._
n,
!'
O.184C,(l-e"'0691"'),xO.6
-;;- = (1- e
....0693.")
xl 199C
x'
I
The subscripts x and s refer to the unknown sample and the

standard, respectively. The end of irradiation is usually taken as
the reference time. Instead of the absolute activity. A, the
measured counting rate corrected for the decay during counting
time, te, and the delay time before measurement (EXAMPLE
10.2), lei. are used These corrections are given by
A.
O069),'C
time. Then, substituting the corrected counting mtes for the'

absolute activities in Eq. (10.6), we have
FIG. 10.1. Creation and increase of activity during irradiation,

followed by decay after the end of the irradiation.
(I-e "')
0.0693
(l_e-lJ.0693xll) e
where C, is the total count in the peak during the 3-min. counting
Number
of half-lives (T)
Co =
where ex is the total count in the peak of interest during the 11min. counting time. If the activation of the standard begins at 1:57
and ends at 1:58 with a beam cwrent of 0.01 ,UA, then the
integrated charge is 0.6 ftc. If the counting of the standard begins
at 2: 15 and ends at 2: 18, then the counting time is t., = 3 min., and
the delay time is" = 17 min. Equation (10.7) gives the corrected
count rate for the standard as
~ 80
~
x 132 IJ(.
C
=5.97xl0"'-' <J?.
C,
The substitution of the corrected measured count rates for the

absolute activities in Eq. (10.6) assumes that the sample and the
standard are counted under exactly the same conditions.
(10.7)
. where C is the integrated count acquired during the counting

time. The corrected Co values from the standard and the
210
10.2.3.1 Approximations based on range

The following approximations require no detailed knowledge
of the cross sections, but only relative stoppiiig powers or
ranges. Detailed justificatioll!:i can be found in Ricci and Hahn
(1965, 1967), Ishii ef ai. (1978a), and Chaudri ef ai. (1976). The
error introduced is less than 10% if the irradiation energy is a
reasonable amount higher than the threshold energy. The higher
the incident energy and the higher the reaction threshold, the
smaller the error introduced by these approximations. The error
is also smaller as the average mean atomic numbers of the
matrix and the sample are closer.
.
Activated range approximation (Ricci and Hahn, 1965):
R,(EJ - R,(E,,)
R,(E,)-R,(E m)
(10.10)
The systematic error from the approximation is small and can

be calculated (Ishii et al., 1978a), and just as for the numerical
integration, large variations in the shape of the cross section
have little effect.
A simple method of approximating !1\.suggested by Chaudri et

ai. (1976) is to use Eq. (10.10) at the mean energy, Em =(~ +
~h)/2, where Et is the incident energy and Eth is the threshold
energy for the reaction. This approximation is usually adequate
for most applications. EXANfPLE 10.3 shows the resull!:; for
different approximation methods.
EXAMPLE 10.3. Calculating the ratio of integrals.
The reaction 14N(p,a)l1C can be used to measure nitrogen in
materials, and nylon can be used as a standard. To measure the
nitrogen in a sample of vanadium with 6 MeV protons, the ratio
of integrals in Eq. (10.6) can be calculated. Tbe total range
approximation of Ricci and Hahn gives
Total range approximation (Ricci and Hahn, 1967):
R,(EJ
R,(E)
'1(.=--
In these approxim1'ltions Ei is the incident beam energy, and

Eth is the threshold energy for the reaction.
1{ ~ R Nylo. (6) _ 420.3 .llm><1.13 g/cml
S(E)
then (S)
(s)
0.575.
The reaction has a threshold energy of 5 MeV, so the mean

energy of Cbaudri et al. is
Em ~ (5 + 6)12 =
5.5 MeV.
Then, the mean stopping power approximation gives
10.2.33 Approximations based on stopping power

It was shown by Ishii ef af. (1978a) that <lC can be
approximated with a negligible error by the ratio of the stopping
powers at a calculable average energy. If the average stopping
power, (S}, is defined such that
fG(E) dE = _1 E cr(E) dE,
135.5 ~mx6.10 glcm'
R,(6)
10.2.3.2 Numerical integration

If the integrals in Eq. (10.6) are evaluated numerically, only
relative cross sections and relative stopping powers are needed.
Vandecasteele (1988) showed through calculations that the result
is quite insensitive to large variations in the shape of the crosssection curvc.
SNYlon(5.5)
=
4.55 xIO-' keV/(~g/cm')

7.66xl0- 2 keV/(",g!cm 3 )
0.594.
According to Vandecasteele (1988). the value of 1(for this

case by numerical integration is 0.603. Figure 10.2 shows a plot
of the deviations for the range and. StoPPh'1g power
approximations versus energy and the atomic number Z of the
matrix, where
(10.8)
= SeE,,), where Em is the average energy
E,
fEcr(E) dE
o
E,
(10.9)
Notice that the higher the incident energy and the higher the
reaction threshold, the smaller the error introduced by these
approx.imations. The error also is smaller when the average
mean atomic numbers of the matrix and the standard are closer.
f cr(E) dE
To a good approximation 1%), Em is matrix-independent,
and the ratio of the integrals, <.R.. is then given by
211
/
Chapter 10
0
0
20
AI V
40
Zr
60
80
10.3.1 Reaction choice

Cyclotrons and electrostatic accelerators are used primarily
with protons, deuterons, and 3He for activation analysis. Alpha
particles are of limited interest and tritons are interesting but
rarely available. Heavy ions (Z > 2) are of little analytical use,
and they are difficult to USC in practice because of sample
deterioration due to high stopping power. Table 10.1 lists the
ions, energies, reactions, and sampling depths commonly used in
activation analysis.
Pb
~M'V
9 MeV
6 MeV
~
'#.
Because there is a large choice of ions and energies, a given

analytical problem can sometimes be solved in various ways.
The strategy is to obtain an adequate sensitivity for the elements
to be analyzed while keeping the matrix activity to a minimum.
Nuclear interferences, that is, production of the same reaction
product by a different reaction on a different chemical element,
must also be avoided. In prnctice, this determines the reaction
that is used. EXAMPLE 10.4 is typical of the possibilities. When
analyzing an element of atomic number Z, one can use
radioisotopes of elements with atomic numbers Z - I, Z, Z + ~
and perhaps Z + 2 and Z - 2. Figure 1003 shows the Z versus
nucleon number, N, plot of the chart of the nuclides and the
nuclear reactions that can be used to reach adjacent nuclei that
might be of analytical interest. Notice that there are several
different ways to produce the same fmal radionuclide from the
same starting isotope. Similarly, there are possible interferences
for a particular radionuclide due to other reactions from isotopes
of other elements. Most of the time, thick targets are irradiated,
where "thick" means thicker than the range of the incident
particles. In that case, the higher the eDergy, the higher the
sensitivity because a larger volume of material is activated
<J
12 MeV
9 MeV
10
6 MeV
FIG. 10.2. Difference (%) between the ratio, ~ of the activation

integrals, obtained by numerical integration and by different
approximations, as a function of the atomic number of the matri)(
for different irradiation energies. The standard is nylon and this
is for the l"~(p.a)llC reaction (Vandecasteele, 1988). stopping
power approximation at tbe mean energy of Chaudri et af.

(1976) ... total mnge approximation of Ricci and Hahn (1967).
Note that Em can also be calculated from a thick-target yield
cwve (Ishii et al., 1979b). Using the average stopping power
method, Blondiaux et al. (1981) showed that no measurable
The nuclear interferences increase in number and importance

with the energy because more reaction channels become
available. Figure 10.5 shows an example of the possibiJjties for a
single isotope of copper, 63Cu, and an incident 3He beam. Few of
the radioisotopes have suitable half-lives, not too long and not
too short, and suitable decay schemes. Furthermore, only a
fraction of the "suitable" radioisotopes are free of nuclear
interference.
error is introduced when chemical compounds are used for

standards or when a compound matrix is analyzed. The BraggKleeman rule on the addilivity of stopping powers is valid in the
energy domain encountered in charged partic1e activation.
10.3 EXPERIMENTAL TECHNIQUES

Activation analysis usually consists of these independent
steps: (1) irradiation with the appropriate beam to activate the
Vande Graaff accelerators are preferred to cyclotrons

because they have higher beam quality and because they are
easier to operate, particularly when an energy cbange is needed.
For the higher energies, however, one needs a large tandem,
which might not be easily accessible; a small cyclotron might
then be an easier solution.
sample, (2) post-irradiation etching to remove activated surface

contamination, (3) radiochemical separation to isolate the
radioisotope of interest whcn it cannot be detected directly in the
sample, and (4) measurement of the activity.
212
Table 10.1. Ions, reactions, and sampling depths.

Energy
Sampling
ion
Reaction
(MeV)
depth
(p, n)
(p,2n)
(p, pn)
10 -30
100 IJm mm's
Incident
(p,)
d
(d, n)
(d,2n)
(d, p)
(d, a)
3 -20
(I, n)
(1, p)
(I, d)
3 -15
&3He. n)
( He,2n)
(,He, p)
3-20
(a, n)
(a,2n)
(a, 3n)
(a, p)
(a, pn)
((,3n
3He,n
,n
Z+1
p,n
p,y
d,n
3He,np
e<,np
I,n
'He,p
Original
p,pn
n,1
d,p
I,pn
101Jm-2mm
nucleus
Z-1
few
~m
p,"
Z-2
1545
I,)>
t,3 He
p:He
d,,,
p,2p
N-1
n,p
-100 ~m
(3He, a)
a,2n
Z+2
p,ap
10 lim 1oo'.s lim
N-2
FIG. 10.3. Reactions to obtain radioisotopes with atomic

numbers of Z - 2, Z - I, Z, Z + 1, and Z + 2, starting from a
target of atomic number Z.
(et, an)
EXAMPLE 10.4. Activation of nickel.
"-~"-~~"-~"'~~T'~--r-;j_,
:2:_ :
Several different reactions can be used to analyze for nickel,
among them "'Ni(p,n)"'Cu, "Ni(p,pnl"Ni, "Ni(p,)"Co,

"'Ni(d,n)"Cu, and "'Ni('He,pn)"Cu. If you wanted to use the
(p,a.) reaction that produces 55CO, you _would need to be
concerned about the interference from the ~c(p,2nisCo
reaction. Figure 10.4 shows the yields ofssCo from Ni and Fe as
functions of the incident proton energy. Clearly, the activation
should be carried out below 15 MeV, the tlm:shold for the (p,2n)
reaction on '(ip e.
Fe + p -> "Co ---<>
:::.
~
,~ 20
.l
~ 10
O~~~~~+~LL~~~~u-~~
10
15
20
25
Energy (MeV)
FlG._IOA. Interference of Fe by the s6pe(p,2n)S5Co reaction
s
above 15 MeV, when analyzing Ni by the '3Ni(p,a/ Co reaction.
213
"'I
Chapter 10
microamperes. Note that the samples might easily be fused or

vaporized, even below 1 J.,JA at several megaelectronvolts, if they
are insulating or not well cooled.
1
'H.,p2n
10.33 Post-irradiation chemistry
10.3.3.1 Etching
To demonstrate the importance of surface contamination,
Table IO.2lists tberange of surface concentrations ofC, N, and
o for various samples after chemical etching under the best
conditions, that is, the cleanest surface that can be obtained
(Quaglia et a/., 1976). This minimum surface contamination is
activated at the same time as the trace element in the bulk and
corresponds to a minimum detectable bulk concentration. The
table also includes the equivalent bulk concentration in parts per
million (ppm) that it would represent if not removed. This
equivalent is a function of the nature of the sample, the incident
energy. and the reaction used. It is clear that etching is an
absolute necessity at low levels, at least for C and O.
104
L -__~~______~____~__~
10
20
(MeV)
30
He
FIG. 10.5. Excitation functions for ~e-induced nuclear reactions

on "eu (Lamb, 1969).
103.2 Irradiation procedures

Two basic problems must be solved for accurate activation
_.analysis: current integration and specimen cooling. Figures 10.6
and 10.7 show typical irradiation setups. Direct current

integration and air cooling are usually preferred, but might not
always be feasible. Irradiation in air or in a controlled
atmosphere such as bellum simplifies the cooling of the samples.
At high energies, particularly for irradiations in air, the
activation in a thin foil can be used for monitoring the beam
current Note particularly the safety considerations described in
Chapter 17. The induced activity in the monitor foil serves as a
measure of the integrated charge, Q = it, where i is the beam
current and t is the irradiation time. Whenever possible, direct
current integration should be used because it is ,faster, simpler,
and more accurate than a thinfoil monjtor.
Cleaning by etching before irradiation is not consistently as

good as indicated in Table 10.2, and etching after irradiation is
delicate when low energies are used. Even though surface
contamination might be very thin (tens of angstroms). microns
of material must bc removed by etching because of the
phenomenon of recoil implantation. For instance. 18F nuclei from
the 160eHe,p)lBF reaction travel -0.7 J.lItl in GaAs for an
incident energy of 3 MeV and 4 J.l.Ill in Al for an incident energy
of 18 MeV. A limited and knO\VD. number of microns must b~
removed after irradiation to minimize the effects of surface
contamination. The etching must be studied for each particular
type of sample. To measure the etching, one can use mechanical
means or sometimes the activation of the matrix (Valladon et aT.,
1980a). The etching itself can be purely chemical if the vacuum
is very good in the accelerator. Usually, it must be a combination
of mechanical etching with diamond paste followed by chemical
etching because of the polymerization of hydrocarbons at the
beam spot (Blondiaux e/ al., 1980). This polymerized layer
inhibits the chemical etching and contains huge quantities of
carbon. For elements other than C, N, and 0, surface
contamination is not a severe problem. However, for trace
analysis, care should be taken to clean the sample surface before
irradiation because of possible diffusion under irradiation
(Blondiaux e/ al., 1984). It is always v,rise to control the sample
temperature during irradiation CWauters et af., 1986).
Table 10.2. Surface contamination
There is no rule for the irradiation times and intensities

because all cases are different. It is clear, however, that it is not
profitable to irradiate for more than one or two half-lives of the
radioisotope of interest because of the existence of the saturation
factor. Beam intensities for the standards are generally on the
order of nanoamperes. For the samples, beam intensities can
reach . several microamperes and in some cases tens of
Minimum
Equivalent
surface
concentration
bulk
Contaminant
(mlifcm2)
(ppm)
0.05-{J.5
<0.01
0.05-{J.5
0.1-10
<0.01
0.1-10
214
concentration
,.
:\
t
i,
Permanent magnets
Target
Air cooled sample holder
Beam-H~~~
Insulator
Faraday cup
FIG. 10.6. Setup for irradiation of samples in the vacuum at low power using direct earn current integration.
Beam
Water outlet
Target
vacuum
Water Inlet
Thin AI foil (catcher for recoil)

TItanium foil (25I'm)
Thin foil monitor (Havar)
FIG. 10.7. Setup for irradiation of samples at atmospheric pressure, either air or controlled atmosphere, at high power. Irradiation dose
measured by activation of a thin foil such as Havar.
Table 10.3. Semiconductor etches.
Matrix
.Bath composition
(acids are commercial
concentrated acids)
always a straightforward process. The polymerized carbon layer

must always be removed before chemical etching, and the
material should be removed in seveml different baths to avoid
recontamination problems. Furthermore, etching is a nonlinear
process (the amount of material removed is not proportional to
the time).
Approximate
etching speed
(I'm/min.)
Si
CH,COOH (2 voL) + HNO, (3

voL) + !IF (1 voL)
GaAs
!IF (1 yoL) + HNO, (1 voL) +

H2 S04 (1 vol.)
13
loP
Brl (0.5%) in methanol
10.33.2 Radiochemical separations

n favorable cases, when there is low matrix activation or the
production from the matrix of only short-lived radioisotopes
compared to the half-life of the nuclide of interest, the
radioactivity can be measured directly in the sample. This is
often the case for the analysis of a light element in a high-Z
matrix, where the Coulomb barrier effect can be used. Generally
speaking, however, the radioisotope of interest cannot be
measured directly. and it has to be isolated from unwanted
Table 10.3 lists recipes used for etching semiconductor

matrixes. The composition of the etching baths can be varied
depending on the needs of any particular analysis. Etching is not
215
/
Chapter 10
activities using radiochemical procedures. Numerous procedures
based on solvent extraction, precipitation, ion exchange,

distillation, or volatilization/fusion have been described (De
Soete et al., 1972). The methods (EXAMPLE 10.5) are not
particularly difficult, but they do require eltperience with
chemical techniques. This chapter does not discuss the details of
the methods. If you need to use radiochemical separation, find a
chemist to collaborate with you. With good chemical
procedures, the separation of the radionucJide of interest from
interfering radionuclides makes activation analysis a very

sensitive method compared to the otber "prompt methods" based
on the use of ion bearns.

One sb!)uld always be aware of the possible dangers arising
from the handling of radioactive samples. This is of course the
case for most matrixes at high energy (> 10 MeV). but it is also
the case even at a few megaelectronvolts for low-Z matrixes (B,
C, N, 0, F. etc.). Be swe to check the samples with a monitor
after irradiation. See Chapter 17, Radiological Safety.
103.4 Measurement of the radioactivity
Standard procedures are used for the analysis of the induced
activity of the sample (De Soete et 01., 1972). Gamma-ray
spectroscopy with a gennanium detector and multichannel
analyzer is all that is required for the detection of specific
gamma rays. For the case of pure positron emitters, where the
511 keY photons must be detected at very low levels, y_y
coincidence counting with sodium iodide is used. Figure 10.8
shows a typical coincidence detector geometry used to detect the
simultaneous back-to-back 5]1 keY gamma rays emitted when a
positron annihilates. Computer programs to analyze the
sometimes quite complex spectra can be found in most
laboratories. The exponential decay in the counting rate as
shown in EXAMPLE 10.6 should always be followed to confirm
that the proper activity is being measured. This is particularly
important in the case of positron emitters where the 511 keY
transition is not specific.
EXAMPLE 10.5. Radiochemical separation for oxygen in InP.
The reaction used is 160 (t,n)18F at 6 Mt:V (Bordes e/ 01.,
1987). Although indium itself is little activated at 6 MeV by
tritons, it is strongly activated by (n,n') and (n;y) reactions from
secondary neutrons produced by triton-induced reactions on
phosphorous. It is impossible to detect up directly because of
the overwhelming 11~ 1I4In, 115ln, and 116In activities. After
irradiation and etching of -5 f.1D1. the InP sample is dissolved in
HCI to which a few milligrams of fluoride (NaF) has been added
as an inactive fluorine carrier. ISF is then extracted using
diphenyldichlorosilane in diisopropyl ether. The organic phase is
rinsed and directly counted with a Ge detector. The oxygen
concentration typically varies from 6 x lOIS atJcm1 to 16 x 1015
1
atJcm , with a detection limit of 1014 at/cm3 The complete postirradiation treatment takes less than 30 min.
216
EXAMPLE 10.6. Following the decay curve.

The usual method of measuring the decay constant is to plot
the count per time interval versus time on a semilogarithmic
plot If thc data can be fit with a single straight line whose slope
is the decay constant of desired activity, then the correct activity
is being followed. Suppose that you intend to activate a sample
by producing a radioisotope with a 3-mm. half-life and you
measure the following activities in consecutive I-min. intervals:
10,100,7847, 6378,4929,4030,3092,2550, 1939, 1614, 1214,
1023,758,650,437,412,294. Figure 10.9 shows a plot of the
activity versus time and its exponential fit. The value of the
decay constant, l. ~ 0.2321 ~ In(2)lt ~ 0.69311t, imp~es, ~ 2.98
min. More complex situations with multiple activation products
will produce data with breaks in the curve and can be fit with
nonlinear least-squares procedures.
10.4 SPECIFIC APPLlCA nONS
At, mentioned previously, CP AA is the method of choice for
the trace analysis ofC, N, and 0 because it does nol depend on
the availability of blanks or standards with known
concentrations oftbese ubiquitous elements. Other methods such
as neutron activation or photon activation are not effective for
the analysis of C, N, or 0 at trace levels because no suitable
reaction is available or the cross section is too small for adequate
sensitivity. CPM is one oftbe reference methods accepted by
the European Community Bureau of Reference for the trace
analysis ofC, N. and O.
The following sections discuss the reactions that have been
used for the analysis of these light elements and their
interferences. Cross-section curves and thick-target yields for
reactions of interest and for interfering reactions are needed in
order to achieve the best accuracy when calibrating by the
numerical integration and average stopping power methods.
They also are valuable for locating resonances and maxiina to
control the analysis and evaluate sensitivities. Relative thicktarget yield curves are easy to measure; they are useful for
calculating average energies and for directly evaluating
sensitivities as a function of energy.
Table 10.4. Carbon reactions.
Threshold
Reaction
(MeV)
Interfering
reactions
uN(d, t) '3 N
14N(d, dn) 1~
1eO(d, an) 13N
9Be(-3He. nl
"
1OS(3He. d)l'C
I'B(3He.l)l1C
14NeHc, a.d) C
"
lliQ(lHe.2o:)l1C
Threshold
!MeV)
4.9
12
8.4
a
2.5
10
6.3
Charged Partide Activation Analysis
10.4.1 Carbon
The two main reactions are listed in Table 10.4. They allow
the determination of carbon down to the part-per-billion level
(10-') if a post-irradiation radiochemical separation is used.
Nondestructive analysis is possible well below the part-permillion level in matrixes with Z > 30, where it is possible to take
advantage of the Coulomb barrier effect. In the case of
radiochemical separations, the methods mostly used are based on
reducing fusion for DN and trapping N2 on hot titanium or on
oxidizing fusion for lie with absorption of 11eo2 on soda
asbestos. The usual standard is graphite.
10.4.1.3 Other reactions for cariJon

There are other reactions for carbon, but they are of li~e
analytical interest. The cross-section curves for 12C(p,pn)1i C
11C('He,pn)1",
)1'0 and 12C (a,an) 11 C are compIled
'
-1.'1, 11C('He,n,
in an International Atomic Energy Agency (IAEA) ll.andbook
(Albert ef al., 1987., 1987b).
References for actual analysis of carbon in various matrixes
are given in Table 10.5.
10.4.2 Nitrogen
For nitrogen, there is no obvious choice for the nuclear
reaction that should be used. The four reactions that we can
recommend are listed in Table 10.6. Other possible reactions are
I'N(d,an)IIC, I'N(,He,o;)I'N, 1'N(,He,d)I'O, and I'N(c<,o;n)IlN.
Data for most of these reactions can be found in the IAEA
handbook (Albert ef al., 1987a, 1987b). The preferred reaction is
14N(p,a)lIC, because thc-part-per-billion level can be reached
and lie has a half-life of 20.3 min., easily compatible with
radiochemical separations. The other radioisotopes have much
shorterhalflives
70.6 s; 1'0, 122 s; 16N, 7.1 s).
10.4.1.1 "qd,n)"N
The 12c(d,n)l~ reaction produces a pure positron emitter
with a half-life of 9.97 min. It is free of interferences below 4.9
MeV, and in practice, it can usually be used up to the 8.4 MeV
threshold for the oxygen reaction. The reasons are that the cross
sections for the deuteron-induced reactions on nitrogen are small
and the concentration of nitrogen in the sample is usually less
than the carbon concentration. The corresponding cross sections
are shown in Appendix 15. Figure A15.l shows the cross section
versus energy for the (d,n) reaction, and Fig. All.2 shows the
ratio of iJN produced from carbon to that produced from
nitrogen as a function of deuteron energy. The thick-target yield
curves for the reaction 00 carbon and the interfering reactions on
Nand 0 are shown in Figs. AI5.3, A1S.4, and AI5.5. These
figures were taken from Krasnov ef aI. (1970). EXAMPLE 10.7
shows the use of this reaction in metals. and EXAMPLE 10.8
shows its use with a radiochemical separation.
C'O,
10.4.2.1 "N(p,n)l'O
As shown in Fig. A15.12 (Appendix 15), the cross section is
not very high (Nozaki and Iwamoto, 1981; Kuan and Risser
1964). The sensitivity is therefore limited to about
ppm/weight. The interesting points are that the reaction is
interference-free up to -22 MeV and that 140 exhibits a
characteristic high-energy gamma ray (2313 keY). It is thus
possible to analyze nitrogen without radiochemical separation
and without complex decay curve analysis of the 511 keY peak
if the matrix does not activate too much. The drawback of this
reaction is its high threshold energy of 6.3 MeV.
10.4.1.2 "q'He,a)I1C
This reaction also produces a pure positron emitter, but with a
20.3-mm. half-life. Several measured experimental cross-section
curves are shown in Fig. 10.10 (Lamb, 1969). The differences
between the curves are typical of total cross-section
measurements, but they are not important for activation analysis
using standards. Beryllium and boron interfere at all energies.
The cross-section curves for these interferences are given in
Figs. All.7 and AIl.8 (Appendix Il). The thicktarget yield
cwves for the reaction on carbon and the interfering reactioDs on
Be and B are given in Figs. AI5.9, AI45.1O, and AIl.I1
(Appendix Il).
10.4.2.2 1'N(p,a)llC
If the s'ample does not contain too much boron, this is the
most suitable reaction for nitrogen analysis at levels of I ppm or
greater. Figure A15.!3 (Appendix Il) shows the crosssection
curve for the 1"N(p.a)lIC reaction (Jacobs et al., 1974). A
detailed curve up to 7 Me V was redetermined by Kohl et al.
(1990). Figure AIl.14 (Appendix 15) shows the cross section
for the interfering "B(P,n) l1 C reaction (Anders, 1981). The
thicktarget yield curves for the 1'N(p,a)I1 C and the IIB(p,n)IIC
reaction are shown in Figs. A15.15 and A15.l6 (Appendix 15)
(Krasnov e/ al., 1970). The usc of this reaction is demonstrated
in EXAMPLE 10.9.
It is clear that the 12CCHe,a)l1C reaction can be used only if

the concentration levels of Be and B are much lower than those
.for carbon. Note that this reaction should not be used for the
analysis of carbon in semiconductors such as GaAs or InP grown
by the liquid-encapsulated Czochralski (LEC) method because
these crystals are grown in the presence of molten B 10) and
present relatively high boron concentrations. Nitrogen and
oxygen can also interfere above -10 and -12 MeV, respectively.
The thick-target yield curves for these higher-energy
interferences are also given in Krasnov et aT. (1970). For
oxygen, see Fig. AlS.22 in Appendix 15.
I
...L
10.4.2.3 l'N(d,n)l'O
This reaction is interference free up to -11 MeV. Although it
is intrinsically sensitive enough to reach the part-per-billion
level, it cannot reach this level in practice because ISO must be
radiocbemically separated. which takes time. and a large amount
of activity is lost because of the short half-life of 1'0 (122 5).
The best use of this reaction is probably for concentrations of
217
/
Chapter 10
>0.1 ppm/weight The cross-section curve is shown in Fig.

AI5.l7 (Appendix IS) (KOhl ef al., 1990), and the thick-target
yield cUlVe is shown in Fig. A15.18 (Appendix IS) (Enge1mann,
1971.).
10.4.2.4 "N(~p)''N
The interferences from 13C and 170 should be negligible in

most cases because of the low isotopic abundances and low
cross sections. The drawback of this reaction is the requirement
for access to a beam of tritons, but it bas very good potential for
the analysis of nitrogen. It is the eqwvalent of the analysis of
oxygen with 14 MeV neutrons [160(n.p)1~. except that one can
obtain a much higher flux on the sample and use of 3 MeV
tritons results in low matrix activity compared to nitrogen.
Preliminary tests (unpublished results from the group in Orleans)
have sbo'\.VD. that sensitivities of less than 0.1 ppm can be

obtained in practice in less than I min. (the half-life of ''N is
7.1 s).
References for actual analysis of nitrogen in various matrixes
are listed in Table 10.7.
10.43 Oxygen
The main reactions for the analysis of oxygen are listed in
Table 10.8.
10.43.1 "O(p,a)"N
This reaction is a pure positron emitter with a lO-min. halflife. The cross section is shown in Fig. A15.19 (Appendix IS)
(Furukawa and Tanaka, 1960). The thick-target yield curve is
shown in Fig. A15.20 (Appendix IS) (Krasnov ef aI., 1970).
This reaction is not selective at any energy and should be used
only in selected cases (high fluorine concentration in the sample
and low carbon concentration). The importance of interfering
reactions is shown in Appendix IS in Fig. A15.21 for carbon and
in Fig. A15.1S for nitrogen.
EXAMPLE 10.8. Carbon in GaAs with radiochemical

separation.
Wei ef al. (1987) an.lyzed carbon in LEC-grown GaAs by
the "c(d,n)"N reaction at 8 MeV. After irradiation for 20 min.
with a 3-J.LA beam of deuterons, the samples were mechanically
pnlished and etched to remove 4 !lIIl from the surface. Then, the
sample was vaporized in a graphite crucible (reducing fusion)
with the "N trapped on titanium grains (sponge) at 900C. The
detection limit was 8 x 10" aVern' (OJ ppb/weight). By
measuring the carbon content from various parts of the crystal in
the 10-30 ppb range, it was pnssible to determine the
segregation coefficient of carbon. in a gallium-rich crystal of
GaAs grown by the LEC method.
EXAMPLE 10.9. Nitrogen in silicon, with radiochemical
separation.
Nozaki et al. (1970) analyzed nitrogen in silicon by the

14N(p,a)IIC reactioD. This was, of course, undoped silicon
(boron-free). The samples were irradiated for 20 min. at 13 MeV
witb a current of few microamperes (no precise value was
given). Nylon was used as the standMd in this case, but Kap10n
is also often used. After etching, the samples were dlssolved in
HF + Kl0 4; lIe was converted to 11e02 and absorbed in NaOH.
BaeD) was precipitated and counted (total time -30 min.),
The lowest concentration actually seen was -1 ppb/weight,
the most frequent concentration observed was a few parts per
billion, and the maximum was 20 ppb, This technique was used
to study the segregation of nitrogen during zone melting (clearly
toward the tail end of the ingot), as well as the purification
during this process (good purification). Other publications in
which carbon and o<ygen were also analyzed \Nozaki ef al.,
1974a, 1976) completed this study.
EXAMPLE 10.7. Carbon in Zr, Nb, Ta, and W.

Taking advantage of the Coulomb barrier effect,
Vandecasteele et aI. (1979a) were able to analyze carbon in Zr,
Nb, Ta, and W without chemical separation. The IlC(d,n)13N
reaction was used. The samples were irradiated for 2-5 min. at 5
MeV for Zr or at 7 MeV for Nb, Ta, and W, with an intensity of
1-2 !lA. After irradiation and etching, the 511 keY annihilation
peak was measwed in coincidence between two (3-in. x 3-in.)
NaI CI)'Stals, Concentrations ranged from a few tens of
ppm/weight in Zr and Nh, do'ND. to 1 ppm/weight in Ta, and less
than 15 ppb/weight in W.
ScIntillation detector
Scintillation detector
"t Sample
FIG. 10.8, Typical detector geometry for coincidence detection

of the 511 keY annihilation radiation from a positron emitter.
218
Table 10.5. References for carbon analyses.

Reaction
used
Matrix
Reference
AI
12c(d. n)13N
Goethals et al., 1979
s;
12C(3He,o.)11C
Nozakl at af. 1970, 1974a, 1975;

Engelmann, 1971b; Marschal et a/.,
1971; Endo eta/., 1972;
Vandecasteele et BI., 1974a; BOttger
et 81., 1980
Engelmann etaf., 1971b; Bottger
et aI., 1980; Martin and Haas, 1972;
Sann! at af., 1984; Bonger at aI.,
102~~~~~-L~~~~~
0.0
5.0
10.0
15.0
1975
20.0
Time (minutes)
FIG. 10.9. Semilogarithmic plot of activity versus time for a
radionllclide with a 3-min. half-life.
F.
12C(ct. n) \50
Mayolet et al., 1972
N;
12C(d, n)'3N
Strijckmans ot a/., 1980
Zr
12(;(0, n)'::lN
Vandecasteele et aI., 1979a
Zircalloy
12C(d, n)13N
Mortier et a/.. 1981
Nb
12C(d, n) 13N
Shikano st al., 1987a;

Vandecasteele et 8/., 1979a
Me
Vandecasteele et al., 1981
Vandecasteele et al., 1979a. 1981
Ta
12C(a, an)I1C
Engelmann. 1971b; Vandecast9f3le

et al. 1979a
Au
12C(d, n)'3N
Vandecasteela and Hoste, 1985
GaAs
12C(d, n)13N
-~- ~
Ga AI As
10
15.20
25
30
35
Misdaq st 81., 1986
Table 10.6. Nitrogen reactions.

Threshold
Beam energy (MeV)

FIG. 10.10. Cross sections for the I1CeHe,a)IIC reaction
compiled by Lamb (1969),
12C(d, n)1::lN
Wei et a/., 1987; Blondiaux and

Debrun. 1988: Valladon et al.,
1980b; MeyQr and Bethge, 1990;
Shikano et 81., 1987b
219
(MeV)
Interfering
reactlon(s)
Threshold
Reaction
14N(p. n) 140
6.3
160(p, t) 140
21.7
14N(p, a)l'C
3.1
(MeV)
tlB(p, n)lIC
''c(p. d)"C
17.9
14N(d, n) 150
r'O(d,l)''O.
10.6
14N(I, p) 16 N
13C{t. y)'6N
170(1. a)16N
0
0
/
Chapter 10
10.4.3.2 16 0('He,p)"F
Table 10.7. References for nitrogen analyses.

Matrix
Reaellon used
'''N(p. a)"C
Nozaki ef sl.. 1966
51
14N(p, a)l1e
Nozaki et al., 1970, 19748. 1976;

KOhl at aJ.. 1990: Krogner, 1975
li
14N(p, n) 140
Strijckmans 81 a1., 19n
AJ..TI
14N(P. n) '40
Vandecasteele ot 81., 1979b
Fe
14N(d.
n)'SO
NI . _ _~~(P. a):'C.. __
Zr
Zircatkly
Nb
Ta
W
Mayolet
st al., 1972
Strijckmans ef al., 1980
14N<p.o.)lle
Petit, 1978; StrIJckmans ef af., 1982
'''N(d, n}'SO
14N(ct, an)13N
Giovagnofi alai., 1979

Strljckmans et a/., 1992
14N(a, an)13N
Mortior et al., 1981
1~(p,
strijckmans et al., 19n

Sastri and Krivan, 1981
n)140
''N(p.2aJ'Be
Mo
Reference
14N(d, n) 15 0
Mayolet el al. 1972
14N(p,a)"C
Vandecasleele ctal., 1981
10.4.3.3 "O(ll,d)"F
Activation with alpha particles has been used in the past
when 3He was rare and expensive. The use' of alpha particles
should be avoided, however, because of the high value of the
fluorine interference and because thc high threshold values cause
high matrix activation. The cross-section curve (Nozaki et al.,
19740) and the thick-target yield curve (Krasnov el al., 1970) are
shown in Figs. AIS.25 and A15.26 (Appendix 15), respectively.
The thick-target yleld curve for the interference of fluorine is
shown in Fig. AIS.27 (Appendix IS).
10.4.3.4 "O(~n)"F
uN(p, n) 140
Strijckmans et a/., 19n
HN(p,o)'1e
Strijckmans et al., 1978
14 N(p,
Vandecasteele et a/.. 19B1
a)"'C
Helium-3 activation is perhaps the most widely used

activation method for the analysis of oxygen. ISF is a pure
positron emitter with a 110-min. half-life. The interferences of
fluorine and sodium exist at all energies and are only <5% for
concentrations identical to that of oxygen (Engelmann and
Marschal, 1971). Also, the thresholds of the reactions are low,
which limits the activation of the matrix, and the cross section is
quite good (see Fig. AIS.22 in Appendix IS; Lamb, 1969). In
Appendix 14. the thick-target yield is shown in Fig. AIS.23, and
the thick-target yield cwve for the interfering reaction on
fluorine (Krasnov 01 01., 1970) is shown in Fig. A IS .24.
This is certainly the best reaction for the analysis of oxygen at

trace levels. It has a good cross section at low energy, as shown
in Fig. AIS.28 (Appendix IS) (Revel e/ 01., 1977), and is
interference-free up to 4 MeV. In practice, it can be considered
as interference-free up to 12 MeV because Deon is a rare and
inert gaseous element whose concentration in most samples is
negligible. The thick-target yield Curve (Bordes 01 01., 1987). is
shown in Fig. AIS.29 (Appendix 15), along with yields for other
triton-induced reactions. EXAMPLE 10.10 is typical of the use
of this reaction.
Table 10.8. Oxygen reactions.
Reaction
Threshold
(MeV)
Threshold
(MeV)
''C(p.ll'3N
'"c(p, n) '3N
14N(P. d) 13N
0
3.2
8.9
19F('He, a) "F
3.8
23Na(3He. 20) I8F
0.4
.. 5.5
'50(3He, pi I8F
Intelfering
reactions
References for the analysis of oxygen mostly in metals and

semiconductors are listed in Table 10.9.
10.4.4 Channeled activation for C and 0
Channeling of the incident ions, in combination with
I6O(3He, nl '8N.
+e.C-7 'B F
'50(a, d)'8F
20.4
"Nla, nl 18f
8.1
'50(a, pn) 18F
23.2
19F(a, 1lIl)"F
12.6
'50(a,2n)I8Ne
p. "F
29.7
'"Na(a, 2an)'SF
25
2ONe(t, an)'SF
19f(t. tn)'SF
.~:-
Rutherford backscattering spectrometry (RBS), particle-induced

X-ray emission (PIXE) spectroscopy, or prompt nuclear
reactions, has been widely used to study thc location and role of
dopants or impwities in single crystals (Feldman et of., 1982).
These methods are usually applied at high concentrations,
typically above 0.1%. RBS and PIXE cannot bc used for light
elements such as C and 0, and prompt nuclear reactions are
intrinsically limited in sensitivity and subject to surface
contamination problems. Chapter 12 provides a general
discussion of channeling.
'-'
.~.
;'.,t.
:-'r",
4
12.1
The combination of channeling with charge~ particle

activation, although not free of problems, allows studies on the
."~;.,
~~
~,t
.,'~$'
..~~
220
'.~~
lattice location of C and 0 at levels of <1 ppm/weight. The
reasons for this arc thc high intrinsic sensitivity at low energy
and the possibility of removing surface contamination. Also note
that accidental channeling, as in other methods, can be a source
of Jarge errors in charged particle activation analysis. This has
been demonstrated up to factors of 2 in the work of Hanna
Bakraji e/ al. (199Ia).
EXAMPLES 10.11 and 10.12 demonstrate that charged
particle activation combined with channeling can be used as a
tool to study the lattice position of some light elements at the
sub-part-per-million level, probably down to 0.1 ppm/weight.
The interest of the method is that it is applicable to a number of
as-gro'WD. pure materials, not especially doped for the pmpose of
characterization. Also, the elements that can be investigated-B,
C, and O-are most often the major residual impurities. of the
semiconductor materials (8 for Czochralski crystals prepared
under liquid B20 3). A considerable amount of work remains to
be done to fully understand the reiC'ltionship between the light
elements cited above and the properties of semiconductors.
Charged particle activation is an interesting tool for these typcs
of studies,
EXAMPLE 10.11. Carbon in Gs1_xAlxAs.
The detailed work is pllblished in Misdaq et al. (1986). Ga J_",AI"As

samples were prepared by vapor-phase epitaxy with organometallic
compounds. The alWIlinum concentration, )t, was varied between 25%
and. 50% by varying the flow of trimethylaluminum relative tc:) that of
trimethylgallium. Thick (20-30 ~) layers were prepared esrecially for
the needs of the analysis. Carbon was analyzed using the 2C(d,D)I~
reaction at 2.7 MeV (I J.lA, IS min.). I~ could be detected directly
without chemical separation. Irradiations were performed in the (IOO),
the (II0), and random directions. The carbon concentration i~creased
with the AI content at fixed arsine flow and decreased with the arsine
flow at fixed Al concentration. Figure 10.11 represents the ClR.. ratio of
the channeled yield (el to the random yield (R) as a functi()n of the
carbon concentration for the <100) direction and for three different AI
concentrations (52%, 32%, and 25%). CIR. ratios were also measured in
the (110) direction for samples before and after annealing. Detailed
analyses interpreted these data as evidence of interstitial carbon and of
carbon in substitution on arsenic sites. Arsenic vacancies seemed to
play s major rolc concerning the proportion of carbon prescnt on the
different sites. The thennal rreaCInent revealed a change of site, from
interstitial to substitutional.
EXAMPLE 10.12. Ox.ygen in Gal_~AI1As.
Activation has several drawbacks, however. The first one is

that activation, in the case of carbon, takes place over 20--30 ~m
(although the activation of the last few microns is negligible). As
the ions penetrate the sample. dechanneling becomes
increasingly important, so that the [mal observed 13N yield is a
mixed response that is the sum of a channeled activation yield, a
partly channeled activation yield, and a random yield.
Furthermore, the best channeled part of the sample (the first 2
Work similar to tbat on carbon was performed on oxygen using the

u'OeHe,pisF reaction at 3 MeV (Hanna Bakraji et 01., 199 ta). The
samples were irradiated for 45 min. at 0.6 JJA. Oxygen ClIF) could be
measured directly without chemical s.eparation. The 18F yield for the
160eHc,p)18F reaction was measured after orientation along the (100)
axis. Figure 10.12 shows the ratio (CIR) Df this "channeled yield", C, to
the random yield, R. The results, as for carbon, could be explained in
terms of arsenic vacancies.
filll) is eliminated by etching. The second drawback is that
The detection limit of interstitial oxygen under these

conditions is estimated to be -100 ppb/weight. It would be -10
etching the same number of microns for the channeled sample

and the random sample leads to distorted CfR ratios because the
energy loss is not the same for channeled and random ions. The
method can be improved by trying to reduce the etching after
irradiation and by using thinner layers. Surface cleaning under
ultra-high vacuWD, using either ion sputtering or laser
ablation/desorption so that etching after irradiation would
becomc unnecessary , would be a major improvement
measured in a few cases. These high values, as indicated earlier

in this chapter, are due to different stopping powers for the
channeled and random ions. The effect is particularly large here
because 1.5 J..1IIl is removed by etching, whereas 900/0 of the
induced activity is confined within 4 mm of the surface.
EXAMPLE 10.10. Oxygen analysis in GaAs using the
EXAMPLE 10.13. Corrosioo-erosion phenomenon in nuclear power

plants.
ppb/weight using 2.5. MeV tritons and the l'O(t,n)"F reaction.

In the present case. abnormally high C/R ratios have been
160(t,n)18F reaction.
The material studied was unalloyed steel. and the reaction used was
~e(p,n)56Co(Bouchacourt el 01., 1989). The half-life of ~Co is 77.3
days, so the loss of activity due to the natural decay of the radioisotope
is small compared to the loss of activity due to corrosion. The
irradiation energy was 9 MeV; the corresponding activated depth was
40 JlOl, as shown in Fig. 10.13. The sensitivity was 2 x to-l l11m./year,
for a lOO-h test. Examples of results are given in Figs. 10.14 and 10.15.
The variation of the corrosion-erosion with the velocity and the
temperature of the cooling water and the effects of alloying with a small
readily
were
demonstrated.
percentage
of
chromium
The samples were irradiated for 2 h at 3 MeV and 0.5 flA

After irradiation, 5 f.LID was removed by etching, and the samples
could then be directly measured on a Ge(Li) detector. Oxygen
was Dot detected in these samples, and the experimental
detection limit was 6 ppb/weight This limit could have been
increased by a factor of 5-10 by separating 18F because there is
some background from the activation of GaAs.
221
Chapter 10
Table 10.9. References far oxygen analyses.

Matrix
Reaetlon used
Reference
Na
''a (p, n)'8F
AI
Matrix
Reaction used
Reference
Engelmann 1970a
CuZn
''o('He, p)'6F
Debrun and Barrandon. 1975
160(3He, p)18F
Vialatte. 1973; Vandecasteele and

Hoste, 1975a; Petri and Sastri,
ZnTe
160(3He, p)18F
Caneau 8t al., 1981
1975; Fedoroff at al., 1981
GaP
160(3He, p)18f
Kim, 1971
lSO(t, n)ISF
Vialatte, 1973; Valladon et al.
GaAs
160(3He, p)18f
Misaelides ef aI., 1987
160(1, n)16F
VaJladon at al., 1geOb; Va/ladon

and Debrun, 19n; Huber st aI.,
1978; Huberetaf., 1979
GaAIAs
''orHa, p)'8F
Hanna Bakraji
Ge
160fHe, p)1BF
Revel etal., 1977; Vandacasteele

and Hoste, 1975c
1978
lB()(p, n)leF,
Vialane, 1973
''o(p, a)13N,
''0(0., pn) 18F
et al.,
AIMg
160(3He, p) ls F
Fedoroff et al., 1974
AI SI
''orHe, p)18F
Vandecasteele and Hoste, 1975a
Si
160(3He, p)1Sf
Nozakl eta/., 1970; Nozaki etG/.,

1976; Sannj at al. 1984; Nozaki
160(t, n)18F
Valladon and Debrun, 19n
at af. 1966: Schweikert and Rook,
Zr
160(a. pn)18F
Revel and Albert, 1968
1970; Engelmann et al. 1970; Kim,

1971; Vandecasteele et al., 1974b;
Mo
16o(t, n)18f
Valladon and Debrun, 19n
RooK and Schweikert, 1969; Kim.
1I30eHe, p)18f
1969
160(a., pn)18F
1991a
Fedoroff et af., 1974; Revel 8t af.,
1968
Saito at al., 1963; Vandecasteele
at af., 1974b
160(a, pn)18F
Faure et a/., 1972a
lSO(p, n)laF
Schweikert 8t al., 1974
Ag
160(0:, pn)18F
Kahn et al., 1974
lSO(t, n) 1sF
Valladon and Debrun, 1977
InP
160(1, n) 18 F
Bordes st al., 1987
''o(,He, p)18F
Nazakl st a/., 1979
HI
'60(0:, pn)'8F
Revel and Albert, 1968
'SO(t, n)18F
Valladon and Debrun, 1977
Ta
160rHe, p)18F
Debrun and Barrandon, 1975
Cr
''orHe, p)18F
Debrun st al., 1969a
160(3tie, p) 18 F
Fa
160{cr., pn)tBF
Fedoroff et al., 1974; Debrun el al.,

1969a
Vandecasteele 6t Bl.. 1981; Debrun

and Barrandon, 1975
160(a, pn}16F
Revel and
Co
160(0.. pn)18F
Kahn 8t al., 1974
Pb
160(a, pn)16F
Ni
U'O(lHe. p)18F
Strijckffians st al., 1980; Debrun

8cal., 1969a
Faure ~ al., 1972b; Vandecasteele

and Hosie, 1975d
PI
160(3He, p)18F
Petri and Erdtmal)n, 1981
Au
16 0((1, pn}lBF
Defebve Bt a/., 1976
160{p, n)H'F
Mitachl at af., 1985
l60(d, n)17F
Bartha at Sl., 1990
SiN
Cu
''o('He, p)18F
Kahn 8t a/., 1974; lee st al., 1970;

Debrun Btal., 1975; Vandecasteele
and Hoste, 1975b
160(a, pn)18F
Vandocasteele st Bi., 1975;

DefebvB e/ al., 1976
Fluoride
glasses
Diamond
160rHe,
p),aF
Albert.
1968
Maggiore sf a/.. 1989
.,:1
~"ij
'.,:1
.'7-
?i{
,;.~
":1
;i
222
Charged Particle Acti\lation Analysis
Table lO.IO. References for proton activation analyses.
Matrix
Reference
Chemical
separat.ion
Na
Nordmann et sl., 1975
Y"
AI
Debrun et aI., 1969b; Dabney et al.,

1973; Vandecasteele et a/., 1980
Y"
Debrun Bt aL, 1973; Debrun et al.,

1976; Barrandon ~t a/., 1976; Shiba1a
Bt a/., 1979
no
Debrun at af., 1976; Barrandon st al.,

1976
no
Dabney et a/., 1973; Dams et aI., 1986
yes
Goethals et ai. 1986
no
Co
Debrun and Barrandon, 1973; Debrun

et <ill., 1976; Barrandon at aI., 1976;
Benaben at 80/. 1975; Krivan, 1975a,
1975b,1976
no
Ni
Vandecasteele etal., 1980
yes
Cu
Vandecasteele at 13/., 1980; Benaben

et al., 1979
Zn
De Brucker et al., 1987
Y"
Nb
Oebrun et al., 1975; 8arrandon sr al.,

1976; Barrandon et al., 1974; Krtvan,
1975c
no
Si
Fe
Chemical
Matrix
Biological
Tb
Dy
Ho
Oebrun et af., 1976; Barrandon et al.,
no
Debrun et aI., 1976; Barrandon et al.,
no
Debrun et al., 1976: Barrandon et al.,
no
1976
Ta
Krivan. 1975a: Barrandon st 81.,1976;

Barrandon /It a( 1974; Krivan st al..
1974
no
Sastri and Krivan, 1982; Schmid

et al., 1987
yes
Sastri and Krivan, 1982; 1983
yes
I,
Debrun et af., 1976; Barrandon et 80/.,

1976
no
Riddle and Schweikert, 1973
yes
Debrun ef 8/., 1976: Barrandon et aI.,
no
1976
Glass
Riddle and Schweikert, 1973
yes
Barthe et al., 1989
ye,
Landsberger, 1984
no
y.s
Zr F4
Rh
Debrun etal., 1976; Barrandon etal.,

1976; Debrun at af. 1975
no
Coal
Ag
Debrun and Barrandon, 1973; Debrun

ef al., 1976; Barrandon ef 81., 1976
no
Minerals
Delmas et sl., 1976
Environmental
Wauters et a/., 1987a
yes
InP
Lacroix et a/., 1984
no
p,
Debrun ef al., 1976; Barrandon ef al.

1976
no
Airborne
particulate
matter
Fly ash
Bonardi st a/., 1982; Zikovsky and

SChweikert, 1977
1976
Faa et .a/., 1979; Faix and Krivan, 19132
Delmas and Debrun, 1976
separatlon
1976
Au
Cd Te
Referenee
223
no
~ ..
no
Wauters et 81., 19S7b; Wauters at al.,

1987c
ye,
Krivan etal., 1974; Priest et 801.,1980;

Desaedeleer and Ronneau, 1976a;
Parsa et al., 1974
no
Strijckmans et al., 1985
Chapter 10
'.,
e: 0.5
(J
Bammdon e/ al., 1976) using 10 MeV protons. The evolution of

these sensitivities, which follows that of the thick-target yields
with the proton energy (assuming that there are no nuclear
interferences), is shown in Figs. AI5.32, AIS.33, and A15.34
(Appendix IS) for a number of elements (Albert e/ al., 19870,
IB 52% AI
032% AJ
o 25% AJ
,
.. m
I 987b) .
'
--~~~~
Table 10.11 lists actua! results of analysis for impurities in a

silicon matrix, which is a favorable matrix.
-----------------
Table 10.11. Proton activation analyses in Si.

"'Clean" sample
(at/e""')
Concentration (ppmlweight)
1.2x10 16
SSx10.12
S2x10 12
S6;.; 10 12
S 10 13
2 x 10 14
S4x10 13
2>: 10 13
:S 10'2
B
C
FIG. 10.1 I. Variation of the CIR ratio of the channeled yield (C)
of 13N in the (100) direction to the yield of 13N in a random
Ti
Cr
direction (R) as a function of the carbon concentration in Gal_
F.
lA1xAs, for three different AI concentrations (Misdaq et af.,
Ni
Cu
1986).
G.
Zr
... Arsine flow
10 16
3x 10U
7 x 10 1:3
S 6 x 10 12
2 x 10 13
2 x 10 14
:::; 4 x 10 1J
10 13
S 10'2
metallurgy.
Arsine now = 28 IIl1
C3
~4
(trtlcm")
,., Contamination by the wetting agent used during the
6 IIh
OArslneflow=1411h
OArsine flow 16 Vh
C:C
ContaminatecJ
sample*
1.0 ,..--,--,----,-----,----,
0.8
2
~0.6
.;;
O~LW~~~~~~~~~~
10
20
30
40
50
60
c(
Oxygen concentration (ppm/weight)

FIG. 10.12. C/R ratio oftbe 18F yield when channeling along the
(100) axis (C) to the random 18F yield (R) as a function of the
oxygen concentration in Gal_xAlxAs, for different arsine flows
(Hanna Bakraji e/ al., 1991a).
10.4.5 Analyses of other elements
Table 10.10 lists some publisbed reports of the use of proton
activation for the analysis of various matrixes (Vandccasteeie,
1988).
Figures AIS.30 and AIS.31 (Appendix 15) show the
calculated best detection limits for a number of elements in an
aluminum matrix under standard conditions (Debrun el al., 1976;
224
0.4
0.2
10
20
30
Depth (!lm)
40
50
FIG. iO.13. Relative total activity versus depth (total activity at

depth zero equal to 1) for a steel sample irradiated with 9 MeV
e.rotons at an angle of 22 and for the S6CO activity from the
>Ope(p,n)S6Co reaction.
10.4.6 Thin layer activation (TLA) for studies on wear and

corrosion
For thin layer activation (TLA), the sample to be studied is
irradiated with an ion beam in order to induce activation over a
well-defined area and to a limited depth. The loss of material
due to wrrosion or wear is related to the loss of radioactivity.
The advantages of this method are its speed and sensitivity and
the possibility of achieving continuous measwemcnts. For
instance, TLA is 100 times faster than methods based on
mechanical measurements and 1000 times faster than those
1.0
.!!
ca
based on ultrasonic measurements. Important fmancial .:savings

can be made in industry using TLA.
The potential of the thin layer activation method and its
different applications have been described in detail in review
articles, for instance, that of Conlon (1982). We limit ourselves
here to only one example (EXAMPLE 10.13): corrosion-erosion :
in heat exchangers. Note that most applications deal with studies
on wear in engines, turbines, and so on.
One phase
Twophase
.;;;0
0~
One phase: temperature 180"C

velocity SSm/s
Two phase: temperature 17S"C
velocity SOmIs
0.5
..
0
CI)
>
ca
Gi
It
0.0
2
1
Chromium content (%)
FIG. 10.14. Corrosion rate for unalloyed steel for different speeds of the circulating waler as a function of water temperature
(Bouchacourt el aI., 1989).
.
~
.c
8
<) V:1 mls
0
0
0
<:)
0
6 C-
V:3 mts
-E
CI)
V:4 rnIs
'"
.;;;
2 C-
t-~~
/~O...______0--o-------o~
t.
e
~
()
o~
),.~t.
______"_________
~,.~
~
/.0:
Sbaight tubes 8mm 1.0.

Ammonia pH:g.a
A V:2 mls
80
100
120
140
160
180
200
:----
0,
220
_L
240
260
Temperature (Oe)
FIG. 10.15. Relative corrosion rate versus the chromium content of steel for two experimental conditions (Bouchacourt et aI., 1989).
225
---------_.
--_._.
..~
-,
Chapter 10
Blondiaux, G., Valladon, M., Koernmerer, C., Giovagnoli, A..

and Debrun, J.L. (1981), in Analysis of Nan-Metals in Metals
(Kraft, G., ed.), de Gruyter, Berlin, p. 233.
10_5 CONCLUSION
Charged particle activation can, in principle, be used to
analyze ahnost any element at the part-per-billion (10-') leveL
H, He, and Si are notable exceptions. Except for C, N. and 0 at
very low levels, there are many excellent methods to achieve
trace analysis: inductively coupled plasma (ICP), rcp with mass
spectrometry (lCP-MS), secondary ion mass spectrometry
(SIMS), and glow discharge mass spectrometry (GDMS), among
Blondiaux, G., Albert, P., Giovagnoli, A., and Debrun, 1.L.

(1984), Anal. Chim. Acta 160, 289.
Blondiaux, G., and Debrun, J.L. (1988), in Proceedings of the
12th International Symposium on the Application. of Jon Beams

in Materials Science (1987), (Sebc, T., and Yamamoto, Y., eels.),
Hosei University Press, Tokyo, Japan, p. 85.
others. Therefore, charged particle activation should be used

only in difficult cases or as an independent method for the
calibration of standards.
Bonardi, M., Birattari, C., Gilardi, M.e., Pietro, R., and

Sabbioni, E. (1982),.1. Radioanal. Chem. 70,337.
and nondestructive analysis in the case of matrixes that are the

following elements or compounds and alloys of these elements:
Proton activation at E:::: 11 MeV provides a multi-elemental
Bordes, N., Blondiaux, G., Maggiore, C']" VaUadon, M"

Debrun, J.L., Coquille, R., and Gauneau, M. (1987), Nucl.
Instrurn. Methods B24-25, 722.
H, Be, C, N, 0, F, AI, Si, P, S, el, MD, Co, Nb, Rh, Ag, In, Pr,
Bottger, M.L, and Bimstein, D. (1975), Isotopenpraxis
Th, Ho, Ta, and Au (see Table 10.10). Proton activation is a

good complement to thermal neutron activation because
n, 127.
BOttger, M.L., Bimslein, D., Helbig, W., and Niese, S. (1980), J.

Radioonal. Chern. 58, 173.
elements that are very sensitive in one method are less sensitive
in the other. Activation with other particles such as 3He and 4He
bas been very seldom used and is of little interest in practice.
Bouchacourt, M., Marsigne, C, Dubai!. A., Blondiaux, G., and

Debrun, J.L. (l989),Nucl.lnstrum. Methods B40-4I, 1199.
REFERENCES
Caneau, C, Schneider, M., and Moreau, R. (1981), J. CrysJ.

Growth 53, 605.
Albert, P.. Blondiaux, G., Debrun. lL, Giovagnoli, A:,

ValladoD, M. (1987a), "Activation cross~sections for elements
Chaudri, M.A., Bums, G., Reen, E., Rouse, J.L., and Spicer,
8.M. (1976), in Proceedings of the International Conference on
from lithium to sulphur", in Handbook on Nuclear Activation
Modem Trends in Activation Analysis, Munich, Germany (Born,

eel), Kluwer Academic Publishers, Dordrecht, The
Data, IAEA Technical Report Series No. 273, IAEA, Vienna.

Austria, sec. 3-2, pp. 479----536.
H.~J.,
Netherlands, pp. 243-253.
Albert, P., Blondiaux, G., Debrun, J.L., Giovagnoli, A.,

Valladon, M. (1987b), "Thick target yields for the production of
Conlon, T.W. (1982), Contemp. Phys. 23, 353.

Dabney, S.A., Swindle, D.L., Beck, J.N., Francis, G., and
Schweikert, E.A. (1973), J. Radioanal. Chem 16,375.
radIoisotopes", in Handbook on Nuclear Activation Data, lAEA

Technical Report Series No. 273, IAEA, Vienna, Austria, sec. 33, pp. 537-628.
Anden;, B. (1981),
Dams, R., Alluyn, F., Vanloo, B., Wauters,

Vandecasteeie, C. (1986), Z. Anal. Chern. 325, 163.
z. Phys. A30l, 353.
G.,
,l'\
and
De Brucker, N., Strijckmans, K., and Vandecasteele, C. (1987),

Anal. Chirn. Acta 195, 323.
Barrandon, IN., Benaben, P., Debrun, I.L, and VaUadon, M.

(1974), Anal. Chim. Acta 73, 39.
Barrandon, J.N., Benaben, P., and Debrun, I.L (1976), Anal.
Chim. Acta 83, 157.
Debrun, J.L., Barrandon, J.N., and Albert, P. (19690), Bull. Soc.

Chirn. Fr. 3,1011.
Debrun, J.L., Barrandon, I.N., and Albert, P. (1969b), Bull. Soc.
Chim. Fr. 3,1017.
Debrun, J.L., and Barrandon, J.N. (1973), J. Radioanal. Chern.
17,291.
Barthe, M.F., Bajard, M. T., Giovagnoli, A., Blondiaux, G.,

Debrun, J.L., Lesergen~ M., Tregoa~ Y., and Barraud, J.Y.
(1989), Nucl.Instrum. Methods B40-4I, 1202.
Barthe, M.F., Giovagnoli, A., Blondiaux, G., Debrun, 1.L.,
Tregoa~ Y., and Barraud, J.Y. (1990), Nud. Instrum. Methods
.,,
Debrun, J.L, and Barrandon, J.N. (1975), in Proceedings of the.,.
7th International Conference on
B45,105.
Benaben, P., Barrandon. J.N., and Debrun, J.L (1975), Anal.
Chirn. Acta 78, 129.
Benaben, P., Barrandon, IN., and Debrun, 1.L. (1979),
Radiochem. Radiaanal. Lett. 37,241.
Cyclotrons and Their
~;
Debrun, lL, Barrandon, IN., Benaben, P., and Kouxel, C.

(1975), Anal. Chem. 47, 637.
,.
~~brun4~1'~6'7 Barrandon, J.N., and Benaben, P. (1976), Anal.
Blondiaux, G., Sastri, C.S., Valladon, M., and Debrun, J.L

(1980), J. Radioanal. Nud. Chern. 56, 163.
'.~
Applications, Birkhliuser, Basel, Switzerland, p. 507.
Defebve,
,
P.,
Lerch,
P.,
and
Vandecasteele,
Radiochem. Radioanal. Lett. 24, 51.
C.
(1976),
';I
f
,:l
1l'
'1./,0<W
:f?
~#
226
Delmas, R, and Debrun, J.L. (1976), Jiadiochern. Radioanal.

Lell. 27, 91.
Ishii, K., Valladon, M., and Debrun, J.L. (1978a), Nucl.I,..",trum.

Methods ISO, 213.
Delmas, R, Bmandon, J.N., and Debrun, J.L. (1976), Analysis

4,339.
Ishii, K., VaUadon, M., Sastri, C.S., and Debrun, J.L. (1978b),
Nucl. Insrrum. Methods 153, 503.
De Soete, D., Gijbels, R., and Hoste, 1. (1972), Neutron

Acrivation Analysis, Wiley Interscience, London.
Desaedeleer, G., and Ronneau, C. (1976a), J. Radioanal. Chern.
(1974), Pltys. RI!V. C9, 2134.
32,117.
Rev. lSI, 1428.
Jacobs, W.W., Bociansky, D., Chamberlin, D., and Ober,g, D.L.

Jaszczak, R.l., Macklin, R.L., and Gibbons, 1.H. (1969), Rhys.
Desoedeleer, G., and Ronncau, C. (1976b), Anal. Chern. 48, 572.
Kim, C.K. (1969), Radioanal. Lett. 2, 53.
Endo, Y., Yatsurugi, Y., Akiyama, N., and Nozaki, T. (1972),

Anal. Chern. 44, 2258.
Kim, CK (1971), Ana/. Chim. Acta 54, 407.
Kohl, F., Krauskopf, J., Misaelides, P., Michelmann, R., Wolf,
Engelmann, C. (1970),1. Radiaanal. Chern. 6,227.
G., and Bethge, K. (1990), Nucl.lnstrum. Methods B50, 1 9.
Engelmann, C, Gosset, J., and Rigaud, I.M. (1970), Radiochem.

Radioanal. Lett. 5,319.
Kohn, A., Barrandon, J.N., Debrun, J.L., Yalladon,

Vialette, B. (1974), Anal. Chem. 46, 1737.
Engelmann, C. (197Ia),J. Radioanal. Chern. 7, 89.
Krasnov, N., Dmitriyev, P.P., Dmitriyeva, S.P., Konstantinov,

l.0., and Molin, G.A. (1970), Uses oj Cyclotrons in Chemistry,
Engelmann, C. (197Ib),J. Radioanal. Chern. 7,281.
M., and
Metallurgy and Biology (Amphlett, C.B., cd.), Butterworths,
Engelmann, C, and Marschal, A. (1971), Radiochem. Radioanal.
London.
Lett. 6, 189
Krivan, V., Swindle, D.L., and Schweikert, E.A. (1974), Anal.

Chern. 46, 1626.
Faix, W.G., Mitchell, JW., and Krivan, V. (1979), 1. Radioanal.

Chem. 53, 97.
Krivan, V. (1975a),1. Radioanal. Chem. 26, 151.
Faix, W.G., and Krivan, V. (1982) Talanta 29, 285.
Krivan, V. (1975b), Anal. Chim. Acta 79, 161.
Faure, L, Boissier, M., and Tousset, 1. (1972a), J Rodioanal.

Chem. 10,213.
Krivan, V. (1975c). Anal. Chem. 47,469.
Faure, L, Giroux, 1., and Tousset, 1. (1972b), J Radioanal.

Chem. 10,223.
Krivan, V. (1976), TalantaZ3, 621.
Fedoroff, M, Debove, L., and Loos-Neskovic, C (1974), 1.

Radioanal. Chern. 21,331.
Ku"", H.M., and Risser, 1.R. (1964), Nucl. Pltys. 51, 518.
Kragner, K. (1975), Isotopenpraxis ll, 278.
Lacroix, R., Blondiaux, G., Giovagnoli, A., Valladon, M.,
Fedoroff, M., Loos-Neskovic, C., Roucha~d, J.C., and Revel. G.
Debrun, 1.L., Coquille, R., and Gauneau, M. (1984), 1.
Radioanal. Nucl. Chern. 83,91.
(1981), in Analysis oj Non-Metals in Merals (Kraft, G., cd.), de

'
Gruyter, Berlin, p. 243.
Lamb; J.F. (1969), Radioactivation by 3He Bombardment: A
Feldman, L.e., Mayer, J.W., and Picraux, S.T. (1982), Materials

Analysis by Jon Channeling, Academic Press, New York.
Practical Analytical System, Report UCRL-18981, Lawrence

Berkeley Laboratory, University of California, Berkeley, CA.
Furukawa, M., and Tanaka, S. (1960), J. Phys. Soc. Jpn. IS,

2167.
Giovagnoli, A., Valladon, M., Koemmerer,
c., Blondiaux,
Landsberger, S. (1984), Int. 1. Environ. Anal. Chem. 16,337.

Lee, D:M., Stauffacher, C.Y., and Markowitz, S.S. (1970), Anal.
Chem. 42,994,
G.,
and Debrun, J.L. (1979), Anal. Chim. Acta 109, 411.

Goethals, P., Yandecasteele,
Chim. Acta 108, 367.
c.,
Liebler, V., Bethge, K.., Krauskopf, J., Meyer, lD., Misaelides,

P., and Wolf, G. (1989), Nucl. Insrrum. Methods B36, 7.
and Hoste, J. (1979), Anal.
Maggiore, C.l, Blacic, J.D., Blondiaux, G., Debrun, J.L., Hage
Goethals, P., Bosman, B., and Vandecasteele, C. (1986), Bull.

Soc. Chim. Belg. 95,331.
Ali, M., Mathez, E., and Misdaq, M.A. (1989), Nud. Instrum.
Methods B4~I, 1193.
Hanna Bakraji, E., Ducouret, E., Blonrliaux, G., and Debrun, J.L.
(1991a),Nucl.lnstrum. MethodsB56-57, 819.
Marschal, A., Gosset, J., and Englemann, C. (1971), J.
Hanna Bakraji, E., Blonrliaux, G.,'Ducouset, E., and Debrun, J.L.
Martin, J., and Haas, E. (1972), Z. Anal. Chim. Z59, 97
(199Ib), Nucl.lnsrrum. Merhods B56-57, 896.
Mayotet, R, Reimers, P., and Englemann, C. (1972), J.

RadioanaJ. Chern. 12, 115.
Meyer, J.D., and Bethge, K. (1990), in Proceedings of the 6th
Conference On Semi-Insulating III-V Materials, Toronto,
Canada, cbap. 4.
.
Radioanal. Chern. 8,243
Huber, A., Morillo~ G., Linb, N.T., Debrun, J.L., and Valladon,
M. (1978), Nud. Instrum. Methods 149, 543.
Huber, A., Linh, N.T., Valladon, M., Debrun, J.L., Martin, G.M.,
Mitouneau, A., and Mircea, A. (1979),1. App/. Rhys. 50,4022.
227
/
Chapter 10
Sanni, A.O., Roche, N.G., Dowell, N., Schweikert, E.A., and

Ramsey, T.H. (1984), J. Radioanal. Nucl. Chern. 81, 125.
Michelmann, R. W., Krauskopf, J., Meyer, J.D., and Bethge, K.

(1990), Nucl. Instrum. Methods B51, I
MiSllelides, P., Krauskopf, J., Wolf, G., and Bethge, K. (1987),
Misdaq. M.A., Blondiaux, G., Andre, J.P., Hage Ali, M.,
VaUadon, M., Maggiore, CJ., and Debrun, J.L. (1986), Nuc1.
instrum. Methods B15, 329.
Mitacbi, S., Sakagucbi, S., Yonezawa, H., Shikano, K.,
Sbigematsu, T., and Takahasbi, S. (1985), Jpn. J. Appl. Phys. 24,
Sastri, C.S., and Krivan, V. (1981), Anal. Chern. 53,2242.

Sastri, e.S., and Krivan, V. (1982), Anal. Chim. Acta 141, 399.
Sastri, C.S., and Krivan, V. (1983), J. Trace Microprobe Tech.
1,293.
Schmid, W., Egger, K.P., and Krivan, V. (1987), J. Radioanal.
Nud. Chern. 119

Schweikert, E.A., and Rook, H.L. (1970), Anal. Chern. 42, 1525.
L827.
Schweikert, E.A., McGinley, J.R., Francis, G., and Swindle,

D.L. (1974), J. Radioanal. Chern. 19,89.
Mortier, R., Strijckmans, K., and Vandecasteele, C. (198l), Bull.

Soc. Chim. Belg. 90, 297. ..
Shibata, S., Tanaka, S., Suzuki, T., Umezawa, H., Lo, J.G., and
Yeh, 8.J. (1979),1nt. J. Appl. Radiat. Isotopes 30,563.
Nordmann, F., Flulu, A., Tinelli, G., and Engelmann, C. (1975),
Analysis 3, 171.
Shikano, K., Katob, M., Shigematsu, T., and Yonezawa, H.

(19878), J. Radioanal. Nuel. Chern. 119,237.
Shikano, K., Katoh, M., Shigematsu, T., and Yonezawa, H.
(I 98Th), J. Radioanal. Nuel. Chern. 119,433.
Strijckmans, K.., Vandecasteele, C., and Hoste, J. (1977), Anal.
Chim. Acta 89, 255.
Strijckmans, K., Vandecastcclc, C., and Hoste, J. (1978), Anal.
Chim. Acta 96, 195.
Nozaki, T., Okuo, T., Akutsu, H., and Furukawa, M. (1966),

Bull. Chern. Soc. Jpn. 39, 2685.
Nozaki,
T.. Yatsurugi, Y., and Akiyama, N. (1970), J.
Radioanal. Chem. 4, 87.

Nozaki, T., Yatsurugi, Y., Akiyama, N., Endo, Y., and Makide,
Y. (l974a), J. Radioanal. Chern. 19, 109.
Nozaki, T., Iwamoto, M., and Ido, T. (1974b), Int. J. Appl.
Radial, isotopes 25, 393.
Strijckmans, K., Vandecasteele, C., and Hoste, J. (1980),
Nozaki, T., Yatsurugi, Y., and Endo, Y. (1976), J. Radioanal.

Chern. 32,43.
z.
Anal. Chim. 303, 106.

Strijckmans, K., Motier, R, Vandecasteele, C, and Hoste, 1.
(1982), Mikrochirn. Acto 11, 321.
Nozaki, T., Iwamoto, M., Usami, K., Mukai, K., and Hiraiwa, A.
(1979), J. Radioanal. Chern. 52, 449.
Nozaki, T.M., and Iwamoto, M. (1981), Radiochim. Acta 29,57.
Pars" B., and Markowitz, 8.S. (1974), Anal. Chern. 46, 186.
Peti~ 1. (1978), Mern. Sci. Rev. Metall. 75,395.
Strijckmans, K., DeBrucker, N., and Vandecasteele, C. (1985),

J. Radioanal. Nuel. Chern. Lett. 96, 389.
Valladon, M., and Debrun, J.L. (l977), J. Radioanai. Chern. 39,
385
Petri, H., and Sastri, C.S. (1975), z. Anal. Chirn. 177,25.
VaUadon, M., Giovagnoli, A., and Debrun, J.L. (1978), Analysis

6,452.
Petri, H.; and ErdlIiIann, G. (1981), Analysis of Non-Metals in

Metals (Kraft, G., ed.), de Gruyler, Berlin, p. 253.
Valladon, M., Blondiaux, G., Giovagnoli, A., Koemmerer, C.,

and Debrun, J.L. (1980a), Anal. Chim. Acta 116, 25.
Priest, P., Devillers, M., and Desaedeleer, G. (1980),

Radiochemical Analysis (Lyon, W.S., ed.), Ann Arbor Science,
Ann Arbor, MI, p. 191.
Valladon, M., Blondiaux, G., Koemmerer, C., Hallais, 1.,

Poiblaud, G, Huber, A., and Debrun, J.L. (1980b), J. Radioanal.
Chern. 58, 169.
Quaglia, L., Weber, G., David, D., Van Audenhove, J., and
Pauwels, I. (1976), Ewisotop Report ITE No. 90.
Vandecasteele, C., Adams, F., and Hoste, J. (1974a), Anal.

Chim. Acta 72, 269.
Revel, G., and Albert, P. (1968),J. Nud. Mater. 25, 87.
Vandecasteele, C., Adams, F .. and Hoste, J. (l974b), Anal.

Chim. Acta 71, 67.
Vandecasteele, C., and Hoste, 1. (1975a), Bull. Soc. Chim. Belg.
84,673.
Vandeeasteele, e., and Hoste, 1. (1975b), Anal. Chim. Acta 79,
302.
Vandecasteele, C., and Hoste, 1. (1975c), Anal. Chim. Acta 78,
121.
Vandecasteele, C., and Hoste, 1. (1975d), J. Radioanal. Chem.
27,465.
Revel, G., Da Cunha Belo, M., Linck, I., and Kraw, L. (1977),
Rev. Phys. Appl. 12,81.

Ricci, E., and Hahn, R.L. (1965), Anal. Chern. 37,742.
Ricci, E., and Hahn, R.L. (1967), Anal. Chem. 39, 794.
Riddle, D.C., and Schweikert, E.A. (1973), J. Radioanal. Chern.
16,413.
Rook, H.L., and Schweikert, E.A. (1969), Anal. Chem. 41, 958.
Saito, K, Nozaki, T., Tanaka, S., Furukawa, M., and Cheng, H.
(1963), Int. J. Appl. RadlOt.lsotopes 14,357.
228
,
'-'1
r
I
I
c., Adams, F., and Hoste, 1. (1975), Ana!.
Vandecasteele, c., Strijckmans, K.., and Hoste, 1. (1979a), Anal.

Chim. Acta 108, 127.
Wauters, G., Vandecasteele, C., and Hoste, 1. (1986), 1.

Radioanal. Nuc!. Chern. 98,345.
Wauters, G., Vandecasteele, c., and Hoste, 1. (1987a),L Trace"
Microprobe Tech. 5, 169.
Vandecasteele, C., Strijckmans, K., Kuffer, R., and Hoste, 1.

(1979b), Radiochem. Radioanal. Lett. 38, 261.
Wauters, G., Vandecasteele, c., Strijckmans, K., and HCJste, 1.

(1987b), J. Radioana!. Nucl. Chem. 112,23.
Vandecasteele, C., Dewaele, J., Esprit, M., and Goethals, P.

(1980), Ana!. Chim. Acta 119, 121.
Wauters, G., Vandecasteele, C., and Haste, J. (1987c), J.
Vandecasteele, C, Strijckmans, K.., Engelmann, C., and Ortner,

H.M. (1981), Ta!anta 28 19.Vandecasteele, C., and Haste, 1.
(1985), J. Rodiaana!. Nucl. Chern. Lett. 89, 167.
Wei, L.C., Blondiaux, G., Giovagnoli, A., Valladon, M., and

Debrun, J.L. (1987), Nucl. lnstrurn. Methods 824-25, 999 .
Vandecasteele, C. (1988), Activation Analysis with Charged
WohUebran, K., and Schuster, E. (1967), Radiochim. Acta 8, 78.
Vandecasteele,
Chim.
Acta 76, 27.
Radioanal. Nucl. Chern. 110,477.
Wilkinson, D.H. (1955), Phys. Rev. 100,32.
Particles, Ellis Horwood Ltd., Chichester, UK. Vialane, B.
Zikovsky, L., and Schweikert, E.A. (1977), J. Radioana!. Chern.

37,571.
(1973),J. Radioano!. Chern. 17,301.
229
CHAPTER
PARTICLE-INDUCED
X-RAY EMISSION: PIXE
J. L. Campbell
University of Guelph, Guelph, Canada
CONTENTS
11.1 INTRODUCTION ........................................................................................................ 233
11.2 FUNDAMENTALS ...................................................................................................... 234
11.2.1 Intensity--concentration relationship ............................................................... 234
11.2.2 Bremsstrahlung background ............................................................................ 234
11.2.3 PIXE database ....,............................................................................................... 235
11.3 SAMPLE TYPES AND ANALYTICAL APPROACHES:.:: ....::.:.......................;.~.236 ....
11.3.1 Thin-sample approximation .............................................................................. 236
11.3.2 Thick-sample: Known matrix ........................................................................... 236
11.3.3 Intermediate sample; Known matrix ............................................................... 236
11.3.4 Intermediate sample: Unknown matrix ........................................................... 236
11.3.5 Thick samples: Matrix concentrations unknown ............................................236
11.3.6 Layered samples ................................................................................................. 237
11.3.7 Spectrum fitting ................................................................................................. 237
11.4 EXPERIMENTAL ASPECTS .................................................................................... 237
11.4.1 Accelerator and Ion beam .................................................................................237
11.4.2 X-ray detector ..................................................................................................... 237
11.4.3 On-demand beam deflection .............................................................................238
11.4.4 X-ray absorbing fIlters ...................................................................................... 238
11.5 SPECTRUM FITTING AND CONVERSION OF PEAK
AREA TO CONCENTRATIOtl'S ................................................................................ 238
11.5.1 Spectrum fitting ...........................................................................................:..... 238
231
I
Chapter 11
11.5.2 Standardization ..............................................................................................239

11.5.3 "Invisible elements" .................................................................................. ~ ....... 240
11.5.4 Accuracy and limits of detection ......................................................................241
11.6 TIllN FILMS, LAYERED STRUCTURES, AND DEPTH DISTRIBUTIONS ... 241
11.7 MICRO-PIXE ............................;.............................................................................242
11.8 WGIlER ENERGIES AND IlEAVIER IONS ......................................................... 243
11.9 FINAL REMARKS ...................................................................................................... 243
REFERENCES .....................................................................................................................244
..
232
-;';,
~.
Particle-Induced X-Ray Emission: PIXE
11.1 INTRODUCTION
Particle-induced X-ray emission (most often effected with
proton beams) is based on the ejection of inner-shell electrons
from target atoms and the spectroscopy of the subsequently
emitted X-rays. Figure 11.1 shows the many emission lines that
i
~epo~.~ib~l~e~~~~~~~~~~~~~~~~~~
i
N
lt~5
(f! f!J,'(2) ,,(,.

etc)
K ,._-
l2Serlt15
(11.~."S'''',.
'''')
USOrl=
(I.u"l'O:<I'~~1.5'1J.t.
.~)
- - - - -.. ,--------
KSer\elll
(~nt.flt,~,
etc)
helium ions, and a few explorations have been conducted willi

heavier ions. Compared to RBS and nuclear reaction analysis
(NRA), PIXE offers oDly very modest possibilities fOL" depth
profiling of elements within a target, but it can be a powerful
tool for the analysis of subsurface inclusions within a sample.
More than half of CurreDt PIXE work is done with micro."beams,
where the ability to produce quantitative trace-element
distributions in one or hvo dimensions has been valu<!ble in
diverse fields of application. For more information on
microbeams, please refer to Chapter 13, Nuclear Microprobe
Analysis.
Figure 11.2 displays a PlXE spectrum (Ryan ef al., 2001)

recorded with a high-purity Ge (HPGe) detector, chosen to
illustrate the high dynamic range that is usually encountered. It
has intense X-ray peaks from lighter elements of high
concentration (Fe :::: 10%) and much weaker peaks at higher Xray energies arising from the trace elements (typically tens of
parts per million). Invariably, this dynamic range, combined
with electronic limitations on achievable counting rates,
demands use of an absorber (200 f1IIl of aluminum in this case)
between sample and detector to strongly suppress the dominant
lower-energy X-rays while transmitting the majority of the
higher-energy X-rays. Because each element emits several X-ray
lines: even a small number of elements can result in a
complicated spectrum, with overlapping peaks superimposed on
a
bremsstrahlung
background
of
complex
shape.
(Bremsstrahlung radiation is electromagnetic radiation produced
by the acceleration of a charged particle when deflected by
another charged particle, such as an electron deflected by an
atomic nucleus.)
irl
100
tt
. I"
Atomic Numbe ... (Z)
FIG. Il.l. (Top) Atomic-level diagram showing the main K. L.

and M transition lines. (Bottom) Energy ranges of the main
series
...
."' .. .,...
"
iIIl
jJ
OJ
.~
"
Go
t};
"'
io
OJ
0.
... ..
Oll.
,o' """
I!1l\
" ''.1J
V .OJ
lib
Many of the lines are 'unresolved, so the main features of

particle-induced X-ray emission (PIXE) spectra are the
compound tines and line groups Ka., K~, La, L~, Ly. Ma, and so
As an ion beam analysis technique, PIXE shares with

Rutherford backscattering spectrometry (RBS) the advantageous
simplicity of having the elements of the sample appear in the
spectra in a monotonic and predictable way, while differing from
particle-induced gamma emission (pIGE), where the nuclear
reactions that are employed differ radically from one element to
the nex.t. An increasing amount of work is being done with
.. .,
jo
. i...,
"""'(a'i)
"
FIG. 11.2. Typical PlXE spectrum of a geological sample

showing major elements at low X-ray energy and trace elements
at high energy. [Reprinted from Ryan ef al. (2001) with
permission from Elsevier.]
233
Chapter 11
FIXE also belongs to the family of X-ray emission techniques

and shares with these techniques the property of being an
elemental analysis method, with no isotopic sensitivity and,
except in special cases. very little sensitivity to chemical
bonding, It is very' similar to electron probe microanalysis
(EPMA). a highly developed technique that bas been a
workhorse for several decades in many application areas. EPMA
interrogates a region about I fJ.m below the surface, and its
detection limits are a typically a little better than one part per
thousand; it is thus a method for major and minor elements.
PIXE interrogates a depth that is element-dependent through the
attenuation coefficient of the X -rays of the element of concern,
and this depth can reach a few tens of microns. Its detection
- _. ~- limits-can-reicli-down to the part-per-million leve~ and thus, it
often complements EPMA, especially in geochemical
applications. However, PIXE can also assume the major element
responsibilities: Campbell et a/. (1997) used both EPMA aod
micro-PIXE (Ji-PIXE) to determine major and minor element
concentrations in a suite of 30 geochemical standards and
reference materials and found excellent agreement, with the
accuracy of the PIXE results being 1-2%, X-ray fluorescence
analysis as an approach to the technology of target preparation
(drying, ashing, dige:;:tion) could improve levels of detection
(LODs). These plots must be interpreted with care, as they deal
with interference-free LODs. Their usefulness diminishes
somewhat as thick samples of geological and archaeological
materials containing both matrix and trace elements bring
increased complexity. Usually. such samples cannot be SUbjected
to any concentration procedures because J.1-PIXE is being used to
probe specific portions. A particular trace element can have a
favorable LOD in a given matrix where it is interference-free.
but as the matrix is vari~ different concentrations of another
element whose X-rays interfere with those of the element of
interest can then cause dramatic variations in the LOD. The best
way to detennine LODs is therefore to record spectra from real
--'--~amplcs or tol~te su~h in the literature.' .
Another competing method is X-ray fluorescence (XRF)

analysis. which has the obvious advantage of portability, with an
increasing number of handheld XRF devices becoming
available. It also has the advantage of a modest degree of
"tunability" of the exciting source to the specimen through
choice of the X-ray tube's anode elemenl In its synchrotron
variant (s)1lchrotron X-ray fluorescence, SXRF), the beam can
be focused to spot sizes that compete with those obtained using
Il-PIXE. In addition, the exciting X-ray energy can be adjusted
to lie just above the absorption edge of an element of specific
interest, thereby optimizing its detection limit; this tunability
confers a significant advantage over ~-PIXE. Another advantage
of SXRF is the ability to study chemical bonding effects by
probing structure near X-ray absorption edges. However, PIXE
too bas a "stable" of companion techniques including RBS and
nuclear reaction analysis, which give it advantages in particular
situations. In addition, beam time is often easier to obtain at
234
small ion beam analysis (IBA) installations as opposed to large

synchrotron facilities.
11.2 FUNDAMENTALS
11.2.1Intensity--concentration relationship
For the geometry shown in Fig. 113, with a homogeneous
sarople thick enough to stop the ion beam and with the justifiable
neglect of straggling (given that the bulk of the X-ray production
arises from the earUest part of the proton track), the intensity or
yield Y(Z) of the principal characteristic X.ray (Koi or La,) of
element Z (atomic mass Ai) having concentrations Cz is
_ N.. coL b zt;;(OJ47t) N C 'fO"iE)TI(E) dE
Y(Z)_
'
S (E)
At
Eo
M
(11.1)
Here, Eo is the initial ion energy, and Er is the final eoergy on

exit from the sample. S.,(E) is the overall stopping power of the
sample matrix, crz(E) is the ionization cross section, (I)z is the
fluorescence yield, bz is the branching fraction of the principal
line within the particular X-ray series (e.g., K.a within the K
series), tz is the transmission through any absorbers preseot, n is
the detector solid angle, and EZ is its relative efficiency. Np is the
number of incident protons, and Nay is Avogadro's number. The
dominant linear dependence of Yz on Cz is modified' by the
integral term, kno~n as the matrix correction, which can be
evaluated only if the concentrations of all elements present are
knovm. Within this matrix tenn,
Tz(E)=expH*ls:~:ls~~)l
(11.2)
where the matrix mass attenuation coefficient (WP)M is the

concentration-weighted sum of the mass attenuation coefficients
of the matrix e1ements,
Equation (11.1) neglects secondary contributions to the
intensity arising from the X-rays of a dominant element being
absorbed within the sample and fluorescing those of another
elemenl Secondary fluorescence must be included in any code
that generates X-ray yields from element concentrations.
11.2.2 Bremsstrahlung background

For non-interfering peaks, the detection ijrnits in PIXE are
determined by the continuous bremsstrahlung. The projectile
contribution to this continuum is very small. The main
component is secondary-electron bremsstrahlung (SEE), which
arises when an electron ejected from a bound state is decelerated
by Coulomb scattering in the sample; its maximum energy,
corresponding to all of the energy imparted to an electron in a
bead-on collision with a proton, is (4m.1m,.)JE" which is 6 keY
for the typically used 3 MeV proton. Quasi-free-electron
bremsstntblung (QFEB) is emitted in cases where the target
electron velocities are so low compared with those of the
projectile that the electrons are essentially free; for 3 Mey
protons, the upper energy limit for QFEB is 1.6 keY. Atorruc
instrumental properties that appear in Eq. (I 1.1). A brief

examination of the accuracy oftbe database is pertinent here.
bremsstrahlung (AB) is emitted by an ejected electron falling

back into its initial bound state; it is much less intense than SEE
but extends to higher photon energies. These contributions are
shown schematically in Fig. liA.
S"p"e<imen
updated on Ziegler's Web site (http://www.srim.org). appear to .be accurate at the level of a few percent and are well-suited to
. inclusion in PIXE analysis codes. For protons incident on the K
shell, the tabulated ECPSSR-DHS (energyloss Coulomb-
Interacthm Point
a'
'on
For proton and helium stopping powers, the semiempirical

fits to carefully selected data done by Ziegler et 01. (1985) and
deflection perturbed-stationary-state relativistic Dirac-HartreeSlater) theoretical ionization cross sections of Chen and
Crasemaun (1985, 1989) and the "reference" cross sect ions of
Paul and Sacher (1989) are in agreement and appear to be
accurate to within a few percent. Except at very low atomic
DWDbers, theoretical K-shell fluorescence yields (HubbeU et aJ.,
1994) and X-ray branching frnctions (Scofield, I 974a, 1974b)
agree closely with experiment. The most modcm tabulation of
X-ray energies is that of the National Institute of Standards and
Technology (NlST) (Deslattes et 01., 2005). There are many
tabulations and calculations of X-ray attenuation coefficients;
NISI', XCOM code (Berger et al., 2009), based on theoretical
calculations. is convenient and, with some exceptions, again
provides accuracy at the level of a few percent. One ex.ception
arises when ali X-ray energy value lies just above the K
absorption edge of a major constituent. An example is analysis
of FeNi standard alloys where the Ni K X-ray is strongly
absorbed by iron; analysis based on Eq. (11.1) gives an
erroneous re~lUlt even if a secondary fluorescence correction is
included (CampbeU et ai., 1993). When the XCOM attenuation
coefficients for Ni K X-rays in Fe are replaced by near-edge
measured values from synchrotron experiments, the analysis
agrees closely with the certified concentrations.
XRay
FIG. 11.3. Generalized geometry for PIXE
x- ray energy
For L-subshell ~ionization, ;ario~s UDc~rtainties remain

despite a plethora of cross-section studies. Again, the most
sophisticated theoretical cross sections are the ECPSSR-DHS
values of Chen and Crasemann (1985, 1989), and Odic et al.
(1994) provide a set of reference cross sections based on critical
examination of experimental data in the literature. Lapicki
(2002) provides a detailed discussion of various corrections to
the ECPSSR approach, but concludes that the scatter of
ex.perimental data is the limiting factor .in testing the theory. The
situation is greatly complicated by the need to assume values of
the subshell fluorescence yields (roj, i = 1, 2, 3) and CosterKronig probabilities (fij) in deducing subshell ionization cross
sections from measured X-ray production cross sections. The
values of these six quantities that were generally assumed
(Krause, 1979) for almost all published work have recently been
superseded by those in two new critical reviews (Campbell,
2003, 2009). Although some recommended values have changed
significantly. considerable uncertainties are still attached,
especially to the three L1 parameters, and DHS theory (Cben et
aI., 198]) is not satisfactory in the L 1 case. In contrast,
theoretical predictions of the relative intensities of LX-rays
FIG. 11.4. Componentsof the PIXE background as defined in

the text. [Reprinted from Ishii and Morita (1987) with
permission from Elsevier.]
A lengthy series of theoretical studies of these processes by
Ishii and co-worker> [most recently, !shU et 01. (2005)] has had
much success. A limited attempt was made by Murozono et a/.
(1999) to develop from this work an analytical description of
background shape for use in spectrum interpretation. However,
in practice, real specimens contain many elements spanning a
wide range of atomic number, and the projectile energy varies
from its initial value down to zero; therefore, the mathematical
description of background is very complex. In addition,
background is sometimes augmented by (p,y) reactions. Most
workers therefore adopt empirical appwaches to background
detennination, as mentioned in Section 11 J .1.
11.2.3 PIXE database

Depending on how standards are used, PIXE analysis has
greater or lesser reliance on the atomic physics database and the
235
Chapter 11
within each of the L I, L2, and L3 series are in excellent

agreement with experiment (papp et al., 1993). It follows that
where Y I is the theoretical X-ray yield per unit solid angle, per
unit concentration, and per unit beam charge. We have combined
the detector solid angle and any calibration factor for the
the L cross-section situation needs to be re-assessed.
integrated charge Q into a single instrumental constant, H.

On occasion, M X-mys of heavy elements appear in PIXE
spectra, and in such cases, caution is advised, especially when
lbis constant H can be determined with standards; the

advantage of this approach is that, if one acceptS both the
these lines overlap the K X-rays of light elements (for example,
database and the detector parameters as being sufficiently

accurate, then these standards need not have a matrix identical to
that of the specimens. Of course, the accuracy should be tested
with reference materials in any given analytical con.tcxt. When
this is done, small dependencies of H on X-ray energy are
sometimes discerned; these tend to occur at X-ray energies
below 5 keV, where the frontal absorbing layers of the detector
are not fuUy understood, and above 30 keY, where the nominal
detector thickness might not be precisely correct.
M lines of lead and K lines of sulfur). Ionization cross sections

and fluorescence yields are not well known for the M subshells,
and the situation is complicated by the steeply varying detector
efficiency at these low energies.
ECPSSR-DHS cross sections have not been computed for

deuterons or helium ions, and so recourse must be bad for these
ions to the older ECPSSR-SH treatment (Liu and Cipolla, 1996)

based on screened hydrogenic (SH) wave functions.
11.3.3 Intermediate sample: Known matrix
11.3 SAMPLE TYPES AND ANAL YTICAL

APPROACHES
Equation (11.1) can be exploited in severallirniting or typical
situations, as we describe here.
All of Section 11.3 ,2 applies to a target that is not thick
epough to reduce the beam energy to zero but cannot be

considered in the extreme case as "thin". Early in the
development of PIXE, Maenhaut et al. (1980) stressed the need
for appropriate thickness corrections in "nearly thin" biological
and environmental samples. The sample matrix and thickness
must be known a priori (or the latter measured by transmittedion energy loss) in order that the integral in Eq. (11.1) can be
evaluated between the proper limits of ion energy. This situation
prevails for many biological and environmentp samples
deposited as thin layers on thin polymer bacldngs.
11.3.1 Thin-sample approximation

For samples thin enough that the beam energy 10ss and the
ray attenuation are negligible, the matrix effects drop out, and
there is a direct linear relationship between Iz and Cz . A PlXE
system can then be calibrated simply by running thin-film
standards, such as those provided by MicroMatter Corp. (Deer
x-
Harbor, WA) and plotting Kn or La intensity versus areal

concentration as a function of Z, often referred to as a sensitivity
cwve. This measured curve subsumes all of the physics and all
of the detector terms in Eq. (11.1) and reduces the analysis to a

direct comparison with standards; the accuracy is then
determined by the accuracy of the standard concentrations and of
the spectrum fitting. The sample substrate must also be thin and
free of elemental contaminants; polycarbonate foils of a few
microns thickness are good candidates. This thin-sample
approach has been widely used in PlXE analysis of aerosols.
11.3.2 Thick-sample: Known matrix

This frequently encountered situation presents itself when the
major elements (comprising >99.9%) of a thick sample are
mown. together with their concentrations. A common example
is analysis of sulfide mineral grains, such as pyrite (FeS), for
silver and platinum-group elements. Here, the matrix (denoted
M) is known from stoichiometry, but in other cases, it may have
been determined by prior RBS or EPMA or even by a
simultaneous PIXE analysis using a second detector tuned for
major elements (see Section 11.4,2). The integrals in Eq. (11.1),
can be computed directly using the database and the matrix
concentrations. The problem is therefore reduced to the simple
linear equation
113.4 Intermediate sample: Unknown matrix

When there are several light elements of unknown identity
and concentration, as in biological and environmenthl samples,
the Q-factor method of Grime (1996) offers an elegant solution.

This involves simultaneous RBS and PIXE, with the RBS
spectrum being modeled to provide the majqr element
concentrations and the thickness; these quantities then provide
the necessary input for the PIXE code, which supplies the trace
concentrations. Tbis approach is advantageous for )l-PIXE
analysis of heterogeneous samples, as it determines the matrix at
each point of analysis.
11.3.S Thick samples: Matrix concentrations unknown
In this situation, the matrix eJement identities are known, but
their concentrations are not. It follows that both the matrix and
the trace-element concentrations must be determined. Provided
that the spectrum contains the X-rays of all of the matrix
element<;, this can be done by an iterative solution of Eq. (11.3).
A PIXE code would determine peak areas and thence a :first
estimate of concentrations ignoring matrix effects. These
preliminary concentrations would then be used to compute the
matrix effects for a second iteration including a refit of the
spectrum. This nested iteration scheme would continue until a
self-consistent set of concentrations was established. One might
(11.3)
236
or might not choose to enforce a 100% total concentration;

relaxing this constraint, as is the common practice in ElVIP A,
provides a useful check of the results. If the 100% constraint is
applied, the user can allow for the presence of one "invisible
element" in the matrix, for example, a light element such as
oxygen whose X rays cannot be observed. This is of value in
substrate), uniform beam intensity over the cross-sectional area

is required. This is usually accomplished by use of a very thin
diffuser foil (with necessary replacement at intervals) or by
defocusing the beam with a pair of quadrupole magnets; in each
case, a subsequent aperture provides a well-defmed beam area.
Areal mass density can be compared between sample and .
standard, or the actual mass of a sample can be determined if the
beam spot dimensions are measured (e.g., by burning an image
on a thin polycarbonate foil). The integrated charge Q passing
through the sample is measured on a Faraday cup, whicb. should
be protected from secondary electron loss by a negatively biased
suppressor grid and which should be within a grounded
enclosure.
PIXE analysis of oxide minerals, a topical example being found

in the u:se of the alpha-particle X-ray spectrometers carried by
the Mars Exploration Rovers for the pwpose of combined PIXE

and XRF analysis of Martian rocks and soils" in situ (Gellert et
aI., 2006).
11.3.6 Layered samples
The methods described in the preceding sections can be
extended to films on substrates and, indeed, to specimens
For thick, homogeneous samples, the issue of beam

uniformity is less important. Here, the sample must be
electrically isolated so that the incident beam current can be
integrated; nonconductive samples require carbon-coating.
Suppression of secondary electrons escaping from the sample is
necessary.
comprising several layers. The data that can. be extracted depend

on the data that are known, Several restrictions must be borne in
mind if a correct solution is to be obtained: For example. a given
element of rulicnown concentration cannot be present in more
than one layer. If the elements and their concentrations in a
given layer are known, the layer thickness can be detennined,
and vice versa. An analysis code needs to be flex.ible enough to
cope with a variety of situations. An application area of
particular interest is that of subsurface inclusions within a simple
matrix. Ryan e/ ai. (1991, 1995a) have had spectacular success
in analyzing the contents of fluid inclusions of different sbapes
within mineral matrixes.
Focused microbeams can be made using eotrunercially

available quadrupole magnet arrangements. An optical
microscope equipped with a charge-coupled device (CCD)
camera is a necessary adjunct for observing the beam spot on a
fluorescent target. Nuclear rnicroprobes usually contain
magnetic or electrostatic rastering arrangements, 50 that the
PIXE analysis can be spatially resolved in one or two
dimensions and concentration maps produced.
11.3.7 Spectrum fitting

In all of the above cases, the spectra have to be fitted to a
model in order to extract peak intensities. In samples other than
thin ones, the relative line intensities for any given element must
be adjusted within this model to. reflect matrix absorption .
effects. This demands that the fitting process be closely coupled
to the process of generation of concentrations: neither of these
two processes can be considered as independent of the oilier, as
discussed further below.
11.4 EXPERIMENTAL ASPECTS

11.4.1 Accelerator and ion beam
Most PIXE work is done with small single-end or tandem
machines providing 2-4 MeV protons. If spectra are required at
different proton energies. as in studies of layered samples,
frequent changes of terminal voltage and retuning of beam optics
might be inconvenient. Fern et al. (2002) solved this problem in
their external microbeam setup through the use of energydegrading foils close to the sample. They chose the polymer
Upilex, which is favorable on grounds of mdiation damage,
induced energy straggling, and trace-element content. With a 3
MeV proton beam, this method provided energies of2.78, 2.33,
1.82, and 1.18 Me V using available foil thicknesses.
11.4.2 X-ray detector

The detector is usually
Si(Li)~
but Ge detectors and silicon
drift detectors (SDDs) are increasingly used as well. The

thinnest beryllium window (8 J.Illl) allows detection down to
sodium K X-rays, but a polymer window gives a useful increase
in the efficiency at such low X-ray energies. The intrinsic
efficiency [ez in Eq. (11.1)J can be estimated from the
manufacturer's information, or it can be measured with standard
For analysis of thin samples that might not have a spatially

homogeneous element distribution (e.g., fluid drops dried on a
237
In external-beam PIXE, the beam or microbeam is extracted

into the laboratory milieu through a very thin window. This
enables analysis of objects that are too large or too valuable to
be placed in a vacuum chamber. Whereas the polymer Kapton
was a popular early choice for the exit window, Grassi (2006)
found 'greatly improved radiation resistance with 100-nm-thick
silicon nitride having an area of 0.5 x 0.5 mm?; the mean energy
loss for 3 MeV protons is about 3 keV. With external beam:s.
direct charge integration is no longer feasible. Sometimes, a
second detector records argon K X-rays from the intervening air
as a proxy for beam charge; otherwise} a rotating vane coated
with a metal film can be used in conjunction with a second
detector. Doyle et a/. (1991) provided a detailed overview of the
external-beam approach, including attention to associated
radiation hazards.
IChapter 11
The idea of a "funny filter" was suggested by Harrison and

Eldred (1973) to enable simultaneous recording in one detector
radionuclide sources as described by O'Meara and Campbell

(2004). Alternatively, the intrinsic efficiency factor can be
merged with the H value, which now would assume a
dependence on X-ray energy and would have to be measured
of both the matrix and the trace elements. In this approach, a

filter thick enough to suppress the former is chosen, and a very
small aperture is dril1ed in it to permit the passage of a small
fraction of the intense major-clcment X-rays. Of course, the
possibility of scattered protons entering the detector again
becomes an issue.
with analytical standards. There has been a trend in PIXE

laboratories toward simultaneous use of two detector-absorber
combinations, one optimized for low-Z matrix elements and the

other for high-Z trace elements, as descnoed in Section 11.4.4,
Critical absorption filters and cleverly designed compound

filters have been used to good effect in the analysis of
archaeological objects. Fleming and Swann (1988) describe a
range of such filters, each optimized for particular sample types.
The. GUPIX code allows the user to include up to five absorbers,
one of which can be a funny filter. Even a gas has been used for
filtering, as in the work of Guereca and Ruvalcaba-Sil (2005),
who used a compound filter of silver and argon in thc 8naJysis of
Ca-ricb materials; this yielded a factor-of-two reduction in the
With very thin windows, protons backscattered from the

target can enter the detector. In some detectors, this causes an
immediate degradation of line shape, and in others, it causes a

progressive degradation. Arrangements of small high-field
permanent magnets are therefore often used to deflect

backscattered protons onto a collimator in front of the detector
(Calligaro el al., 1996).
Most published PIXE work has been done with analog pulse
processors. However, digital processing can confer advantages.
For example, Papp et af. (2004) demonstrated a factor-of-eight
reduction in peak pileup intensity for iron K X-rays relative to an
excellent analog processor operating under the same conditions.
detection limits for A1 and Si.
u.s
SPECTRUM FITTING AND CONVERSION OF

PEAK AREA TO CONCENTRATIONS
U,S.I Spectrum fitting
In the 1-40 keV range of Xray energies examined by PIXE,
a typical element will contribute at least four K X-rays plus
radiative Auger satellites or approximately 20 L X-rays; each of
these peaks will have a silicon escape peak. The range of peak
heights tends, to be large, reflecting concentrations that vary
from percent to parl-per-million levels. This plethora of peaks is
superposed on a continuous background mainly due to electron
bremsstrahlung but sometimes including Compton events from
gamma rays. The usual approach to extracting ptincipal peak
areas is through nonlinear least-squares fitting. Before the fit
begins in any PIXE analysis code, the database of relative X-ray
intensities for each element must be corrected. for matrix
absorption, filter transmission, and detector efficiency:
11.43 On-demand beam deflection

Many PIXE systems contain an electrostatic beam deflector,
which "kicks" the beam off an aperture when triggered by a
logic pulse from the detector's pulse processor. This confers
three very useful advantages. First, the pileup continuum is
reduced in intensity. In addition, heating damage to tbe sample is
minimized because the beam is absent during pulse processing
intervals. It is for the latter reaSOD that the third advantage arises:
no dead-time corrections to the data are needed. Such a system is
described by Teesdale and Campbell (1990).
11.4.4 X-ray absllrbing filters

Relative to othcr cnergy-dispcrsivc X-ray (EDX) analysis
methods, PIXE is unique in its frequent use of an X-ray absorber
(filter) between sample and detector. This absorber suppresses
of
the
secondary-electron
the
low-energy
portion
bremsstrahlung, which would otherwise dominate the processor
throughput, and thereby raises the contribution of higher-energy
characteristic X-rays in the detected spectrum.. It also suppresses
the intense X-rays of light matrix elements, allowing the
spectrum to become dominated by the much less intense X-rays
of higher-Z trace elements. In the analysis of a silicate mineral,
for example, a detector with a thin window and no absorber
could be used for the X-rays of the matrix elements Na through
Ca; a second detector could be run simultaneously with an
absorber sufficient to suppress these X-rays so that it could
collect the K X-rays of elements Ca through Ba and the L X~rays
of heavy elements such as Pb. The first detector would need a
backscatter deflection magnet
Several PIXE codes are available, taking different approaches

to peak and background description and employing different
databases. Some take the peaks to be Gaussian, whereas others
adopt the more rigorous Voigtian approach. Some ignore the
peak asymmetry that arises from electron transport and charge
collection effects at low X-ray energy and from Compton scaner
at high energy, whereas others provide means to iilclude this
effect in the peak model. All include silicon or ge~ium X-ray
escape peaks. Continuous background can be modeled with
multiparameter semiempirical functions that are included in the
overall fitting function, but use of mathematical approaches to
strip off the background is Simpler and has been demonstrably
successful in the cases of the software packages GeoPIXE (Ryan
el al., 1990) and GUPIX (Maxwell el al., 1988, 1995). It is vital
to include peak pileup, and the tv:o .codes mentioned do so to
second order; the GUPIX code also offers the option of peakplus-continuum pileup (Maxwell and Campbell, 2002).
238
Particle-Induced XRay Emission: PIXE
An IAEA intercomparison of nine codes (IAEA, 2003) using

raLher simple spectra concluded that disagreements in fitted peak
areas were small except in special cases of weak peaks situated

on the low-energy tails of intense neighbors. This study did not
venture into derived concentrations. An intercomparison of
concentration detenninations is a more complex matter, covering

experimental and standardization procedures, as well as
spectrum fitting. The best way for a PIXE laboratory to test its
integrated methodology is through analysis of recognized

standard reference materials similar to the samples that arc being
analyzed; the literature contains many such examples (Maenhaut
et ai., 1984; Ryan et ai., 1990; Campbell e/ ai., 1997).
Mo K X-ray energy. Next, using this H value with. the Fe

concentration and peak. area, the absorber thickness can b~
deduced; for thicker absorbers that reduce the Fe intensity
dramatically, recourse is had to Ni. With this thickness vc:llue, the
Mo calcu1ation can be repeated, and a slight final adjusto:J.ent can
be made to the H value. The results, in Table 1L 1, show.
impressive agreement among the various standar-ds and'
absorbers, If a heavier element such as Te were present in the
steel standard, it could have been used to determine the effective
detector thickness, but recourse to a different SRM is usually
necessary for this.
Table 11.1. R values and AI absorber thicknesses forNIST
The choice among PIXE codes therefore rests more on issues
molybdenum steel standards. [Reprinted from Nejedly aad

Campbell (2000) with permission from ElsevieLJ
such as the range of sample type that can be handled, the

provision of rigorous error estimates and detection limits,
demonstrated accuracy in analysis of standard reference
materials (SRMs), and provision of clever tricks for handling
real analytical situations. Ar.J. example that combines the fITSt and
last of these is GeoPIXE's quantitative analysis of deep fluid
inclusions within mineral specimens (Ryan et al., 1991). Again
using the first criterion, GUPIX handles thin, intennediate, thick,
and multilayer targets; when the matrix elements of thick
samples are unknown, it uses an iterative approach to determine
them and thus provides a full analysis of major and trace
elements. GUPIX also offers a cboice among protons, deuterons,
JHe, and ~He (Hopman et al., 2002), together with cyclotronderived protons up to 100 MeV (Denkcr et al., 2004).
SRM
1155
CI296
1\57
1261A
1155
C1296
The various examples described in the remainder of this
chapter were chosen to go beyond the conventional trace1160
element analyses in thin or thick samples that dominate PIXE
and to illustrate some of the more specialized aspects. Several _
examples, both conventional and less so, are provided to
illustrate the capabilities of the GUPIX package.
1155
11.5.2 Standardization
CI296
In principle, PIXE could be a standardless method, given the
1160
accurate database (Section 11.2.3), but the analytical community

would not be convinced by such an approach. At the other
extreme, one could rely entirely on standards, but tbe effort to
create standards pertinent to every siruation encountered would
be tremendous. The GUPIX code provides, a pragmatic approach
that combines these extremes. using both' fundamental
parameters and a small set of standards; this approach has found
widespread use.
A calibration (Nejedly and Campbell, 2000) of a
R(Fe)
H(Ni)
H(Mo)
Absorber thickness = 250 f.Lm
0.0853
. 0.0003
0.0843
0.0003
0.0838
0.0003
0.0842
0.0002
Absorber thickness ~ 350 fUll
0.0853
0.0007
0.0847
0.0007
0.0850
0.0857
0.0012 ____ 0.0001
Absorber thickness
0.0843
0.0042
0.0857
0.0023
0.0835
0.0078
0.0847
0.00t8
0.0838
0.0005
0.0849
0.00l3
0.0876
0.001I
0.084-3
0.0007
0.0847
0.0055
0.0854
0.0002 -
=500 flm
0.0830
0.0022
0.0856
0.0003
0.0850
0.0007
0.0857
0.0007
0.0858
0.0003
Care must be taken with this approach at the extremes of Xray euergy. At higher X-ray energies, the detector's intrinsic
efficiency faUs below 1.00, and the approach is vulnerable to
error in detector thickness and to inadequate treatment of
Compton interactions in the efficiency model. At energies below
about 4 keV, entrance windows, contacts, ice buildup. the escape
of photo- and Auger electrons, and the poorly understood effects
referred to as incomplete cbarge collection (ICC) all conspire to
render calculated intrinsic efficiencies inaccurate. If H is
determined over a wide energy range, deviations from constancy
can occur at these extremes; database imperfections might also
contribute and have been shown (Campbell et aI., 2000) to cause
differences between H values measured u.-,ing K and LX-rays.
~PIXE
system for mineralogical applications using NIST molybdenum

steel standards illustrates this approach. Using manufacturer's
values for the detector dimensions and absorber thickness, the H
value for the system can be obtained by substituting the Mo
concentration and Kct X-ray peak area into Eq. (11.3);
uncertainties in the assumed quantities have little impact at the
239
/
Chapter 11
fractional aperture, H varies dramatically with X-ray energy. In

one of the most rigorous and Successful efforts at system
calibration, Gama et al. demonstrated how a least-squares
optimization of the system parameters leads to a constant H with
GUPJX permits the m;.er to defme H as energy-dependent and to

have different H(Ex) functions for K, L, and M X-rays.
Matters are more complex when a funny filter is used. Gama

et al. (2001) present H values measured with many geochemical
standards a1 many X-ray energies for the Orleans Jl-PIXE
system. With the nominal values of the absorber thickness and
sensible values for absorber thickness, hole area.. and so on, as
shown in Fig. \l.s.
0.03 +-+-+-- Nominal detection

0.025 -I-'--I~
0.02 t-'li----1r--t--r-'--'l..-r--
::J: 0.015 H~
0.01
t-+"--t--if--t-"'+--t---+--+---I
0.005
E[keV]
FIG. \l.5. H values measured by Gama e/ al. (2001) using many standards. The dashed curve w", obtained with the nominal
geometric parameters, and the solid onc was obtained from a nonlinear least-squares optimization procedure. Reprinted with
pennission from Elsevier.
_.11.53 "Invisible elements"
EXAMPLE 11.1. Trace elements in zircon, using invisible

elements.
As mentioned in Section 11.3.4, it often occurs that a

specimen whose matrix concentrations are unknown contains
light elements that contribute to the matrix effects but whose Xrays are not present in the spectrum. GUPIX can handle this
situation in two ways if there is one such invisible element: it
can he coupled to every visible element, as in a mineral where
The detailed work is publisbed in Halden e/ al. (1993). The

matrix is ZrSiO" and the spectrum contains peaks from traces
of Hf, Th, and U. The 125-Jlffi Mylar absorber necessary to
decrease the bremsstrahlung intensity eliminales the Si K peak,
so two elements are "invisible". SiO, is declared to GUPlX as
on invisible complex coupled to Zr through normal valences.
An iterative solution is performed for the Zr, lif, lb, and U
concentrations, with SiOo( making up the total concentration to
100%.
all elements can be assumed to be present in oxide form;

otherwise it can be accorded a concentration equal to 100% less
the sum of visible element concentrations. These approaches can
be eKtended to two or more invisible elements provided that all
the invisible elements are coupled in a known relationship.
240
11.5.4 Accuracy and limits of detection

The LOD is generally defIned as the concentration of an
element that would give rise to an X-ray peak having an
intensity equal to a three-sigma (30) fluctuation of the
100
Ep 2.4 MeV
660 ~m Mylar abS<lrbe r
background underlying the peak. That background comprises

both continuous bremsstrahlung and any overlapping X-ray
peaks from other elements. If it is to be scalable, an LOD must
be specified for a given sample type, a given sample-detector
geometry, and a given elapsed time or accumulated beam
charge. Statistical accuracy of the derived concentrations is
o DL : dry 4mgJcm2 ]
DL : acid digest
10D ~C
ash 1mg/cm2
Cl
~ DL:
OJ
::i
~
usually detennined by the spectrum fitting code from the

uncertainties that it estimates for the peak height and width
parameters. It can be estimated simply, however. by combining
in quadrature the 3" uncertainties in (peak + background) and in
:!::
10
E
:::i
l:
the estimated underlying background.
~CD
...
Early PIXE works contained many useful plots of delection

limits as a function of atomic number for thin and intermediate
samples of biological and aerosol materials. Maenhaut el at.'s
Q)
0
0
LOD plots for the latter (Fig. 11.6) were valuable in showing
how different approaches to the methodology of target
preparation (drying, ashing, digestion) could improve LODs.
Such U-shaped plots must be interpreted with care, as they deal

with interference-free LODs. Their usefulness diminished
somewhat as thick samples of geological and archaeological
0.1
K Ti Mn Ni Ga Se Rb
materials brought spectra of increased complexity and

variability, containing many peaks from major, minor, and trace
elements. Depending on the resulting peak overlaps, the LODs
for particular elements can depart dramatically from the basic
interference-free U-shaped curve. The best way to determine
LODs is therefore to record spectra from real samples or to
locate such spectra in the literature.
Mo
Cd
Pb
Element
FIG. 11.6. Detection limits for biological samples prepared by
various methods. Reprinted from Maenhaut et al. (1984) with
pennission from Elsevier.
Example 11.2. Detennination of the thickness of a thiD film
The concentration table that is output by the GUPIX code

lists, for every element specified by the user, the concentration.
containing a visible and an invisible element.
both estimates of statistical error, and the actual LOD. The last
of these is especially useful when the element's X-rays are
absent from the spectrum, as an upper limit on its cODcentration
can be a valuable piece of information.
The sample is a very thin film of platinum carbide,
supported on a very thin polymer film which shows no Xrays. The stoichiometry is known tobe PtlCn. Only the Pt L
X-rays are visible in the spectrum. The PtC is treated by
GUPIX as a layered specimen, in the special limit of just a
single layer. And in that layer the C is obviously an invisible
11.6 THIN FILMS, LAYERED STRUcrURES, and
DEPTH DISTRIBUTIONS
element. Assuming the given stoichiometry and a total

concentration of 100%, GUPIX determines the thickness and
the Pt concentration; the carbon concentration is 100%
minus the Pt concentration.
There are two reasons that PIXE does not lend itself to
accurate depth profiling in the way that NRA and RES do. First,
the observed proton-induced X-ray yield is integrated over a
wide range of depths within the sample and, therefore, does DOt
reflect any specific depth. Second, the characteristic depth
(defined here as the depth within which 90% of observed X-rays
originate) is element-dependent. However, PIXE can provide
accurate measurements of thickness and composition of single aT
multiple homogeneous thin layers, provided that the speclrum
processing code takes full account of secondary fluorescence
bernreen the layers.
In dealing with a layer on an infinite substrate, the GUPIX

code provides a choice between iterating the layer or the
substrate concentration to 100%. Accuracy for the K X-ray case
using 3 MeV protons can be assessed from a determination
(Campbell el al., 2000) of the thickness of nickel on a steel
substrate in NIST standard reference material 1132A. The
recorrunendcd thickness value is 8.36 0.4 mgicm2 . PIXE
241
I
Chapter 11
values obtained by the two iteration methods were 8.47 and 8.30
EXAMPLE U.S. Beam energy variation in the study of a

Leonardo painting.
mglcm 2, with an uncertainty in each case of G.I mglcm 2 from

the GUPIX fit. For the BCR (Community Bureau of Reference)
certified standard CRM328, which is a gold layer on an iron
substrate, measured thicknesses in the mglcm2 range differed
from the official values by 2-4%. Thus, the accuracy with L Xmys is slightly worse than that with K X-rays, pointing to the
The study was performed by Grassi e/ 01. (2005). Use of 1.4

Me V protons permitted accurate analysis of the light
elements comprising the protective varnish. Increasing the
proton energy until the L Xrays of the known pigment
elements Hg and Pb just appeared allowed estimation of the
varnish thickness. This information allowed the authors to
calculate the range of 2.8 MeV protons in the paint below
the varnish. Then, the relative Hg and Pb peak intensities in
the 2.8 MeV spectrum provided the relative amounts of the
lead white and cinnabar pigments and, thence, the paint
layer thickness. AnalysiS of the substrate in this example
would require higher penetrating power thrC?ugh the use of
higher-energy protons.
need to improve the knowledge of the L database.

EXAMPLE 11.3. Stoichiometry of thin films.
RBS is, of course, an excellent technique for detennining the
thickness and elemental composition of films. When the
elements involved have similar masses, however, the eDergy
resolution becomes an issue. PIXE, with its ability to resolve
neighboring elements, can offer an alternative. For example,

Campbell et 01. (1993) analyzed 10-~ Invar foils containing
64% iron and 36% nickel. In such cases, secondary
11.7 M1CRO-PIXE
All of the foregoing discussion applies equally to
conventional broadbeam PIXE or to microbeam PIXE. The
majority of current PIXE work is of the microbeam variety (}lPIXE), and in this context, the attachment of a high-power
optical microscope to analysis chambers becomes important for
selection of the region to be analyzed and for observation of its
structure. High optical magnification is desirable, which
necessitates a small working distance for the objective and thus
.encowages a design that has the objective within the chamber
vacuum. A standard approach using a beam incident
perpendicular to the sample has been to pass the bearp. through a
45 mirror and to have the objective close to the mirror (i.e., the
microscope is at 90 with respect to the beam); this
accommodates a detector at 45, which is favorable in terms of
minimizing bremsstrahlung background. To achieve greater
optical magnification (300x), Perujo et al. (1988) installed a
high-power microscope within the vacuum, with its axis along
the sample no1'll'Ul1; this forced a design in which the beam
arrived at 45 to the normal and the X-rai takeoff angle was also
45. A more recent design (Ryan et 01., 1999) made the first use
of catadioptric optics (i.e., an image-forming optical system
including both lenses and curved mirrors) by introducing a
Questar reflecting microscope with a 15-cm workitig distance
viewing a mirror that transmits the microbeam through a small
aperture.
fluorescence is significant; however, tabulated attenuation

coefficients close to absorption edges can be inaccurate, For
this example, Campbell el of. replaced the XCOM attenuation
coefficients in GUPIX by more accurate values from near
edge synchrotron measurements. This brought the PIXE iron
result to 63.4% D.3% and the nickel result to 35.3% 0.3%.
Various authors have used variation of the incident angle or
beam energy to obtain' semiquantitative information on depth
profiles. as in EXAMPLES 11.4 aod 1105.
EXAMPLE 11.4. Depth profiling through variation of
incident angle
Weber el al. (2005) studied the behavior of the X-ray yields
of iron, gold, and zinc in an archaeological artifact as the
incident angle of their proton beam was varied from 0 to
70. The iron was presumed to be in pigments at the surface;
the gold was presumed to be part of a subsurface gilding
layer; zinc was not expected to be present. The Fe Ka
intenSity increased continuously, suggesting that the iron
was located on the surface~ the Au LP intensity increased
and then decreased, confmning iliat it lay deeper, and the Zn
Ka. decreased continuously, suggesting that zinc was present
below the successive pigment and gilding layers. The
unexpected zinc was traced to a restorative procedure.
242
,.
A major portion of ).l-PIXE work involves the imaging of

element distributions using a microbeam that is rastered
unifonnly over the selected portion of the specimen surface, to
provide either one-dimensional line scans or two-dimensional
area maps. The data are recorded event-by-event, with each
event comprising the x and y coordinates and the detector X-ray
energy signal. From the sum spectrum of all events, "windows"
can be defined to span each peak of interest, and then the event
list call be translated into a positional display of the counts in
each window.
This approach is subject to the obvious error of contamination
of element windows by interfering X-rays of other elements,
escape peaks, pileup events, and background. Such element
maps are not fully quantitative, and caution is needed in their
interpretation; the windows of trace elements in a complex
spectnun can occasionally contain sufficient contamination that
the distribution results for that element are largely incorrect. One
solution is to revert to point ).l-PlXE analyses at particular
positions of interest on the sample. However, the invention of
dynamic analysis (DA), perfected by Ryan and co-workers over
several publications extending across a decade, provides a
powerful matrix procedure for true quantitative imaging. In its
first version (Ryan and Jamieson, 1993), DA llsed individual
element spectra ("signatures") and assumed a common matrix
and
therefore known relationships betwcen clement
concentration and X-ray yield at every point; a linear leastsquares approach determined the contribution of each element
(and its artifacts) in each window in real time. Ryan et a/.
(1995b) then showed that the latter assumption was problematic
when the major element composition changed across the
scanned area. Ryan et a/. (2000) dealt with this issue through
preliminary point analysis to identify the major phase
components (i.e., regions of different matrixes), enabling yieldconcentration relations to be derived separately for each phase.
More recently, Ryan et aJ. (2005) developed a means of dealing
with pileup effects, which can be significant in images of
geological samples. The GeoPIXE package provides spectrum
fitting and construction of the DA matrix for real-time dynamic
analysis and also provides an environment for image correction
and manipulation offline to provide fully quantitative displays of
concentration.
detectors, overall counting rates of up to I million CO"1..lD.ts per

second (Mcps) were achieved without serious degradation of
energy resolution. Removal of the Mylar proton absorber
necessitated limiting the overall rate to less than 40 kcps_ Further
developments in this direction are likely to heigll.ten the
attractiveness of J.1-PIXE for elemental analysis and imaging.
11.8 HIGHER ENERGIES AND HEAVIER IONS
Protons baving energies above 3 MeV have not featured
significantly in PIXE, partly because of accelerator voltage
limitations and partly because of concern about increased
nuclear reaction gamma-ray background. However, Denker et a/.
(2004) made systematic use of 68 MeV protons from a
cyclotron, thereby including K X-rays of heavy elements such as
gold in her spectra. This method requires picoampere currents
and use of ionization chambers for charge normalizatiol:l. It bas
been successfully app.1ied to a variety of archaeometric samples
and is uniquely suited for valuable samples that ha"e been
encased in a protective polymer layer that cannot be removed.
A modest but growing amount of use is made of He ions in
PlXE, sometimes in conjunction with RES. In a painstaking
comparison of 3 MeV protons and 5 MeV He ions on thin
organic samples, Johansson (1992) fOlUld that detection limits
improved by a factor of 2-3 under similar analysis conditions;
when the He energy was increased to 5.5 MeV, nuclear reaction
background increased mpidly. Beck (2005) showed elegantly
that helium ions confer benefits in the special context of light
trace clements in heavier matrixes. Because the detected X-rays
arise from very shallow depths, the X-ray yields from the two
ions at equal energy are quite similar. However, the secondaryelectron bremsstrahlung is shifted signicantly to lower
energies, and X-ray production from medium- and high-Z
elements is reduced. Both of these effects reduce the background _
and thus improve the detection limits. For Mg in calcite, using
1.8 MeV iOllli, the LOD was improved from 64 ppm to 30 ppm.
When yet-heavier ions are introduced, tabulations of theoretical
cross sections are lacking and multiple-vacancy satellites
complicate fitting of the spectra.
11.9 FINAL REMARKS
The author dedicates this chapter to tbe memory of Professor
Sven Johansson, widely considered to be the father of tlie PIXE
technique. The chapter has only touched on the myriad of clever
applications of PIXE by many colleagues in atmospheric
science, biology and medicine, geochemistry, art and
arcbaeometry, materials science, and so on. Much more
infonnation can be found in various international conference
series (e.g., Particle-Induced X-ray Emission and Its Analytical
Applications, Ion Beam Analysis, Applications of Accelerators
in Research and Industry, European Conference on Accelerator
Applications in Research and Technology. and Nuclear
Microbeam Technology and Applications), as well as in the
books by Johanssoo and Campbell (1988) and Johansson.
Campbell, and Malmqvist (1995)
Detection limits in f.1-PlXE imaging are obviously inferior to

those obtainable in analysis of a single point on the sample,
which provides an incentive to improve the solid angle of the xmy detector. In addition, the demand for quantitative ~-PIXE
analysis at spatial resolutions under 1 ~ points out the need to
offset the inevitably small beam currents by increasing the solid
angle. Doyle e/ al. (2005) took a major step in this direction
through use of an annular 12-element silicon drift detector. This
device provides a solid angle of 1.09 sr at a working distance of
4 nun from the sample to the detecting chip, with the beam
reaching the sample through the center of the annulus. Provided
that baekscattered P!otons wcrc excluded from the SOD
243
Chapter 11
REFERENCES
Beck, L (2005), X-ray Speelrom. 34, 393.
Zipfel, J., and Squyres, S. (2005), J. Geophys. Res. lll,

2005JE002555.
Berger, M.l, Hubbell, lH., Seltzer, S.M, Chang, 1., Coursey,

J.S., Sukumar, R., and Zucker, D.S. (2009), XCOM' Photon
Cross Sections Database, NIST Standard Reference Database 8
(XGAM), NIST, Gaithersburg, MD, available at
http://physics.nist.govlPhysReIDatalXcom!TextlXCOM.html.
Grassi, N. (2006), private communication (University of

Palermo, Palermo, Italy).
Calligaro, T., MacArthur, J.D., and Salomon, J. (1996), Nuel.

Instrum. Methods 8109-110, 125.
Halden, N.M., Hawthorne, F.C., Campbell, 1.L., Teesdale, W.l.,

Maxwell, JA. and Higuchi, D. (1993), Can. Mineral. 31, 637.
Campbell, JL, Higuchi, D., Maxwell, I.A., and Teesdale, W.J.

(19?3), Nuel.lnstrum. yethods B77, 95.
Hanison, J.F., and Eldred, R.A. (1973), Adv. X-ray Anal. 17,
560.
Campbell, IL., Czamanske, O.K, McDonald, L., and Teesdale,

W.1. (1997), Nucl.lnstrum. Methods B130, 608.
Hopman, TL, Nejedly, Z., Maxwell, J.A., and Campbell, J.L.

Campbell, lL., Hopman, T.L., Maxwell, lA., and Nejedly, Z.
Hubbell, J.H., Treban, P.N., Singh, N., Chand, B., Mehta, D.,
Garg, M.L., Singh, S., and Puri, S. (1994), J. Phys. Chern. Ref
Data 23, 339.
GOereea, G., and Ruvalcaba-Sil, lL (2005), X-ray Spectrom.

34,366.
(2000), Nucl.lnstrum. Methods B170, 193.
Campbell, I.L. (2003), At. Data Nud. Data Tables 85, 291.
lAEA (International Atomic Energy Agency) (2003),
intercomparison o/PIXE spectrometry SO/Mare packages,
Campbell, J.L (2009), At. Data Nud. Data Tables 95, liS.
Report lAEA-TECDOC-1342, lAEA, Vienna, Austria.
Chen, MR. Crasemann, B., and Mark, H. (1981), Phys. Rev.

A24,177.
Ishii, K., and Morita, S. (1987) Nucl. Instrum. Methods 822,68.
CbeD, M.H., and Crasemann, B. (1985), At. Data Nuc!. Data

Tables 33, 217.
Ishii, K, Yamazaki, H., Matsuyama, S. Galster, W., Satoh, T.,

and Budnar, M. (2005) X-ray Spectrom. 34, 363.
CheD, M.H., an(c:;!""el\lann, B. (1989), At. Data Nuel. Data
Johansson, S.A.E., and Campbell, JL (1988), PIXE: A Novel

Technique for Elemental Analysis, John Wiley and Sons,
. _.- Tables 4i, :i57.
Chichester, UK
Denker, A., Opitz-Coutereau, J., Campbell, lL., Maxwel~ lA.,

and Hopman, TL (2004), Nuel. Ins/rum. Methods 219-220,
130.
Johansson, S.A.E. (1992), Int. J. PIXE 2,33.

Johansson, S.A.E., Campbell, J.L., and Malmqvist, K.G. (1995),
Particle-Induced X-ray Emission Spectrometry (PIXE), John
Wiley and"Sons, New York.
Deslattes, RD., Kessler Jr., E.G., Indelicato, P., de Billy, L"

.- Lindroth, E., Anton, I., Coursey, 1.8., Schwab, D.1., Chang, 1.,
Sukurnar, R., Olsen, K., and Dragose!, RA. (2005), X-ray
Transition Energies Database, NIST, Gaithersburg, MD,
available at http://pbysics.nist.govlPhysRefDatalXmyTrans.
Krause, M.D. (1979), J. Phys. Chem. Ref Data 8, 307.

Lapicki, G. (2002), Nucl.lnstrum. Methods B189, 8.
Doyle, B.L., Walsh, D.S., and Lee, S.R. (1991), Nucl.lnstrum.

Methods B54, 244.
Liu, Z., and Cipolla, S.1. (1996) Comput. Phys. Commun. 97,
31~
.
Doyle, B.L., Walsh, D.S., Kotula, P.G., Rossi, P., Schulein, T.,
and Rohde, M. (2005), X-Ray Speclrom. 34, 279.
Fedi, M.E., Chiara, M., Giuntini, L., Lucarelli, F., and Mando,
P.A. (2002) Nucl. Instrum. Methods B189, 56.
Maxwell, I.A., Campbell, J.L., and Teesdale, W.l. (1988), Nucl.

Flemiog, S.1., and Swann, C.P. (1988), Nuel. InslrUm. Methods

B30,444.
Maxwell, J.A., Campbell, J.L., and Teesdale, W.l. (1995), Noel.

[ns/rom. Methods B95, 407.
Gama, S., Vollinger, M., Ramboz, Coo and Rouer, O. (2001),

Nuel. Instrum. Methods B181, ISO.
Maxwell, JA., and Campbell, J.L. (2002), X-ray Speelrom. 34,

320.
Gellert, R., Rieder, R, Bruckner, 1., Clark, B.C., Dreibus, G.,

Klingelbofer, G., Lugmair, G;, Ming, D.W., Wa.n.k.e, H. t Yen, A.,
Murozono, K.., .Ishii, K., Yamazaki, H., Matsuyama, S., and

Iwasaki, S. (1999), Nucl. Instrum. Methods 8150, 76:
Maeobaut, W., De Reo, L., and Vandeobaute, J. (1984),NucL
244
Nejedly, Z., and Campbell, J.L. (2000), Nuc!. Instrum. Methods

B160, 415.
Ryan, CO., Jamieson, D.N., Chunns, CL., and Pilcher, T.V.

(1995b),Nucl.Il111trum. Methods BI04, 157
O'Meara, J.M., and Campbell, J.L. (2004) X-ray Spectrom. 33,

14L
.
Ryan, e.G., Jamieson, D.N., Griffm, W.L., and Cripps, G.

Orlic, I., Sow, CH., and Tang, S.M. (1994), At. Data Nucl. Data
Tables 56, 159.
Ryan, CO, (2000), Int. J. ImagingSyst. Technol.ll, 219_
Papp, T., Campbell, J.L., and Raman, S. (1993),J. Phys. B: At.

Mol. Opt. Phys. B26, 4007.
Ryan, CO., van Achterberg, E., Oriffm, W.L., Pearson, N.J.,

O'Reilly, S.Y., and Kivi, K. (2001), Nucl. Il111frum. Methods
BI81,578-585.
Papp, T., Maxwell, lA., Papp, A., Nejedly, Z., and Campbell,
J.L. (2004), Nud. Instrum. Methods B219-220, 503.
Ryan, CO., van Achterberg, E., and Jamieson, D.N. (20()5),

Paul, H., and Sacher, J. (1989), At Data Nuel. Data Tables 42,
Scofield, J.R (1974a), Phys. Rev. A9, 1041.
105.
Scofield, J.R (1974b), Phys. Rev. AIO, 1507; (erratum) (1975),

Phys. Rev. A12, 345.
Pernjo, A., Riddolls, T.J., and Campbell, J.L. (1988), Nucl..

Teesdale, W.J., and Campbell, J.L. (1990), Nuel. Insfrum.

Methods B52, 93.
Ryan, C.G., Cousens, D.R., Sie, S.H., Griffin, W.L, and Suter,
O.F. (1990), Nucl. Instrum. Methods B47, 55.
Weber, G., Martinot, L., Strivay, D., Gamir, H.P., and George,
P. (2005), X-ray Spectrom. 34,297.
Ryan, C.G., Cousens, D.R., Heinrich, C.A., Griffm, W.L., Sie,

S.H., and Memagh, T.P. (1991), Nucl. Instrum. Methods B54,
Ziegler, J.F., Littmark, U., and Biersack, H. (1985), The
292.
Stopping and Range ofIons in Matter, Pergamon Press, New

York; also see http://www.srim.org.
Ryan, CO., and Jamieson, D.N. (1993), Nuel. Instnlm. Methods

B77,203.
Ryan, C.G., Heinrich, C.A., van Achterberg, E., Ballhaus,
c.,
'and Memagh, T.P. (1995a), Nud. Instrum. Methods BI04, 182.
245
CHAPTER
lEI
ION CHANNELING
M. L. Swanson
University of North Carolina, Chapel Hill, North Carolina, USA
L. Shao
Texas A&M University, College Station, Texas, USA
CONTENTS
12.1
INTRODUCTION ..................................:.................................................................249
12.2 THEORY OF CHANNELING ............................................................................... 249

12.2.1 Basic concepts of channeling ............................................................................. 249
12.2.2 Flux distribution and trajectories of channeled ions ...................................... 252
12.2.3 Defect scattering ......:..............:::...: .::...:.:..:.....;:;::~ .............................................253
12.2.4 Dechanneling by point defects .:..........................:............................................. 254
12.2.5 Dechanneling by dislocation loops ....................................................................255
12.2.6 Energy dependence of dechanneling ..................................................................... 255
12.2.7 Dechanneling by heavily damaged regions ........................................................... 256
12.2.8 Energy loss of channeled and random ions .......................................................... 257
12.3 DATA ANALYSIS ......................................................................................................... 257
12.3.1 Channel~nergy conversion ......:............................................................................. 257
12.3.2 Energy-depth conversion ........................................................................................ 258
12.3.3 Separation of dechanneling fraction ...................................................................... 258
12.3.3.1 Line approximation ..................................................................................... 258
12.3.3.2 Iterative procedure ...................................................................................... 259
12.3.3.3 Double-iteration proeedure ........................................................................ 259
12.4 LATTICE SITES OF SOLUTE ATOMS ....................................................................... 259
12.4.1 Principles ................................................................................................................... 259
12.4.2 Comparison of RBS, PIXE, and NRA methods.................................................:. 262
247
Chapter 12
12.4.3 Analytical versus Monte Carlo analysis ................................................................ 262

12.4.4 Numerical method to calculate angular scans .........................;..................... 262
12.4.4.1 Continuum approximatiou ........................................................................ 262
12.4.4.2 Half-way-plane approach ........................................................................... 263
12.5 EXAMPLES IN BULK MATERIALS ........................................................ 264
12.5.1 Ordered structures and compounds...........................................264
12.5.2 Low-symmetry interstitial sites ................................................................... 265
12.5.3 SmaIl displacements ............................................................................................ 265
12.5.4 Gas atom sites .................................................................. _................................. 265
12.5.5 Solute atoms associated with point defects ........................................ 266
~~=="-~~~.,,
...-=-'H.-.,.
..,...,.___ ~
.-..
12.6 SURFACE STUDIES ................................................................................................ 266

12.6.1 Basic considerations ...........................................................................................266
12.6.2 Surface relaxation ...............................................................................................268
12.6.3 Surface impurities ..............................................................................................268
12.6.4 Interface structures ....................................................................................268
12.7
EPITAXIAL LAYERS AND STRAIN ................................................................ .270
12.8 EXPERIMENTAL METHODS .... :.......................................................................271

12.8.1 Cbanneling equipment .......................................................................................271
12.8.2 Choice of ion species and energy .........................................................273
12.8.3 Cboice of RBS, X-rays, or nuclear reactions .........................................273
12.8.4 Ion damage .........................................................................................................274 _. ------..._. 12.8.5 Sensitivity ..........................................................;................................................. 274
12.9 OBTAINING CHANNELING SPECTRA ............................................................ 274
12.9.1 Stereographic projection ................................................................................... 274
12.9.2 Mount the sample and find a suitable channeJ... .......................... 275 12.9.3 Find the normalized yield ...................................................... 276
- ------12.9.4 Perform the angular scans ..............................;......................;.. ;... 276 _
ACKNOWLEDGMENTS ........................................................................... 278
NOTA TION
......................................................................................................... 278
REFEREN CES .................................................................................................................. 279
248
Ion Channeling
12.1 INTRODUCTION
Ion channeling is a powerful tool for the characterization of
crystalline solids. The method is based on the so-<:alled
channeling effect where, if a well-collimated beam of fast
charged particles impinges on a crystal in a direction aligned
with the crystal axis, the trajectories of incident particles are
excluded from a small region around the rows of atoms. The
radii of these forbidden zones are on the order of the ThomasFermi screening dIStance (-o.l A). Consequently, the yield of
close-encounter collisions is significantly lower than the normal
yield obtained under non-channeling conditions. If the crystal
contains defects involving atoms that are displaced from the
atomic TOWS, close-encounter collisions with these displacements
will increase the yield. The yield can be used to precisely
calculate the defect densities in imperfect crystals.
In channeling Rutherford backscattering spectrometry

(RBS), the yield is ligbt particles backscattered from heavy
atoms. In channeling elastic recoil detection (ERD) analysis, the
yield is forward-scattered light target aloms. In channeling
particle-induced X-ray emission (PIXE) spectroscopy, the yield
is X-rays emitted by the deexcitation of target atoms. In
channeling nuclear reaction analysis (NRA), the yield is emitted
gamma radiation or Charged particles from exciled nuclei. In
addition to obtaining defect concentrations in damaged crystals,
ion channeling techniques plus angular scans can provide
infonnation about lattice locations of displaced atoms. This is
extremely useful in the determination of stable defect
configurations. In a typical RBS angular scan, an ion beam is
introduced along a major crystallographic direction. The yield of
c1ose-encounter collision processes is measured as a function of
beam tilting angle. With the aid of modeling, defect locations
can be determined.
In this chapter, we describe the phenomellon of channeling,
its measurement, and its applications. General references in this
area include Morgan (1973), Gemmell (1974), Feldman et al.
(1982), Swanson (1982), Howe el al. (1983), and Mayer and
Rimini (1977) (the "Catania Handbook"). We do not discuss
high-energy channeling (i.e., above a few megaelectronvolts per
nucleon), low-energy channeling (below 50 keY), positron or
electron channeling, or emission channeling. Good references
for these topics are as follows: high-energy channeling,
UggerhDj (1980) and Andersen (1980); low-energy channeling,
Buck el al. (1983) and Aono el al. (1989); positron channeling,
Andersen el al. (l97Ia) and Schultz el al. (1988); and electron
and emission channeling, Gemmell (1974) and Hofsass and
Lindner (1991).
.
12.2 THEORY OF CHANNELING
from computer simulations of ion trajectories by RobinsOI:l. and

Oen (19633, 1963b). Cbanneling can occur for neutral, positive,
or negative particles, but herein, we discuss only the more
common case of positive ions. Channeling has been observe d for
a wide range of ion masseS and for energies from the
kiloelectronvolt to the ~igaelectronvolt range. We are concerned
here mostly with 'It, H+, and "He+ ions with energies of 0.32.5 MeV, which are used for most channeling analyses.
An energetic ion that is directed at a small angle, '1', to close~

packed rows or planes of atoms in a crystal is steered by a series
of gentle collisions with the atoms so that it is channeled int() the
regions between these rows or planes (Fig. 12.1). The
channeling effect is easily understood in terms of a corre lated
series of elastic two-body collisions (Lehmann and Leibfried,
1963; Nelson and Thompson, 1963; Erginsoy, 1965; Lindbard,
1965). The validity of the classical collision model was
demonstrated by Lindhard (1965).
oI
0
I
~d-{
FIG. 12.1. String model of channeling.

The following assumptions are necessary for this treatment of
channeling:
for a channeled ion, the scattering angles are small;
successive collisions are strongly correlated (because ions
passing close to an atom in a string must also come close to
the next atom);
the collisions are elastic t..\'o~body encounters; and
for the description of channeling, a real crystal can be
approximated by perfect strings of atoms in which the atomic
spacing, d, is uniform (Fig. 12.1).
Channeling requires that the ion beam must impinge on a crystal
aligned with its planar or axial directions. In Figs. 122(a}-{c),
channeling will occur if, from the viewpoint of incident ions, the
atomic configurations are similar to those .in Figs. 12.2(b) and
12.2(c).
For a series of correlated collisions, it was shown by Lindhard
(1965) that the ion can be considered to move in a lIansverse
. potential, Yr, that results from an averaging of the potentials of
each atom in a string_ This average (continuum) potential is thus
expressed as
12.2.1 Basic concepts of channeling
VT(P)=~ jV[(p' +X')lil]dx,
Ion channeling was predicted many years ago (Stark, 1912),

but was first detected only in 1963, both from the experimental
range studies of Davics and co-workers (piercy et al., 1963) and
d_
249
(12.1)
/
Chapter 12
where p is the distance of the ion from the string of atoms, x is the
distance traveled along the string, and d is the spacing between
atom, in the string (Fig. 12.1). If the string consists of different
atomic species, the potentials for each atom are different, and an
average potential is used.
(.,
..
f.(p/a)
= L a,K,(-Il,p/a),
(12.6)
i_I
where ai and Pi are the same eonstants as used in Eq. (12.3) and
~ is a modified Bessel function. Note that the value of fMIn: is
given by FRS of Fig. A17.2 in Appendix 17 (EXAMPLE 12.1).
EXAMPLE 12.1. Calculation of ft and fM for "He+ incident
along a (I I 0) channel in Si.
Z, = 2, Z, = 14, a = 0.0157 nm from Eq. (12.4).

FIG. 12.2. Model of lattice atoms, showing the atomic
configuration in the diamond cubic structure, viewed along (a)
mndom, (b) {llO} planar, and (c) (110) axial directions (from
Chu ef al., 1978).
[Note that the approximation of Table AI7.3 (Appendix 17)

gives a = 0.0194 nm.] Then, from Eq. (12.7), fL(p/a) = 0.69,
0.28, and 0.086 for pIa ... 1,2, and 3, respectively. From Fig.
A17.2 (Appendix 17), using fM = (FRS)', fM = 0.66, 0.31, and
0.12 for the same values of pia.

The two-body potential V(r) is generally taken to have the
Thomao:;-Fermi fonn
Z Z e'
VCr) ~ - '-'-<p(r/a) ,
(12.2)
r
where Zl and 22 are the atomic numbers of the ion and an atom
in the string, respectively; e is the electronic charge; r is the
nuclear separation distance; ([)(r/a) is the Thomas-Fermi screening
function; and a is the screening distance. The use of such a
potential is appropriate for the intennediate impact parameters
involved in ion channeling, as discussed by Lindbard el al. (1968).
It should he noted that such a two-body potential is independent
of the ion velocity; that is, inelastic scattering is neglected.
An analytical approximation for q>(r/a), developed by Moliere
(1947), is often used in calculating continuum potentials
J
cp(r/a) =
La,
,., exp(-Il;,/a),
(12.3)
where aj and ~i are constants. The screening length. a, is given

by an equation of the form (Firsov, 1958; Gemmell, 1974)
a ~ 0.8853., ( Z, 112 + Z, III )-UJ ,
(12.4)
where .. is the Bohr radius, 5.292 x 10-" m. Integration of Eq.

(12.1), using a potential of the type in Eq. (12.2), gives the string
potential (EXAMPLE 12.1)
These values of Ii. or fM can be inserted into Eq. (12.5) to find

the string potential. USiDg d = (0.5431/1.414) mn = 0.384 nm
from Tables A17.1 and AI 7.3 (Appendix 17) and e' = 1.44 x
lO-IJ em MeV, we obtain 2Z I Z 2e 2/d;:; 210 eV.
Another important expression for f(p/a), called the standard

potential, was developed by Lindbard (1965) (EXAMPLE 12.1)
fL(p/a) ~ O.5Jn[3(a/p)'
+ I].
(12.7)
Under the high-energy conditions that we consider here, the

approximation of the continuum string potential [Eq. (12.l)J is
valid if the incident angle, 'V. is Jess than a characteristic angle
'II" given by .
~
\jI,
( ')'"
2Z,Z,e
Ed
(12.8)
where E is the incident energy of the ion (EXAMPLE 12.2).

This condition results from the requirement that many collisions
are involved in the scattering of an ion as it approaches a string
of atoms.
EXAMPLE 12.2. Calculation of 'Vl for 2 MeV 4He+
VT(p)
2Z,Z,e' [(pIa).
d
along a (liD) cbannel in Si.

(12.5)
Z, - 2, Z, - 14, e' = 1.44 x W" em MeV, E = 2 MeV, d =
0.384 om (from EXAMPLE 12.1). Thus, 'II, = 0.0102 rad
= 0.584'.
When using Moliere's screening function, f(p/a) becomes
.1
250
Ion Channeling
For mixed strings of different atoms, Z1. is an arithmetic

average of the atomic numbers of the atoms in the string. If the
atoms are not uniformly spaced along the string, d is also taken
as an average spacing (EXAMPLE 12.3). Table A17.2
(Appendix (7) gives the values by which the lattice constant do
must be multiplied to give the different interatomic spacing d
and interplanar spacing dp in common diatomic compounds.
Surf-CIce
Surface
o
000
000
I
0000
0000000
000
Ion
A. Nearly perfect crystal
00
0 0
0 0
O~."~i'-~'''''
0
o 0 0
000
c. Amorphous surface Ieyer
EXAMPLE 12.3. Calculation of interatomic spacing d and

interplanar spacing cl.p for GaP (zinc blende structure).
Depth
00.0000
Z ~ (Z", + Z,Y2 = (15 + 31)12 ~ 23

do = 0.545 nm (Table Al7A, Appendix 17)
(Ill) axis: d ~ do(3,n14 + 3 x 31n14)12
= 0.472 !lID (Table AI7.2. Appendix (7)
{liD} plane: d p =d,/(2 x 2'n)~0.193 om
0.0
~
000000
B. Dechanneling
from point
c
X
o ~A==::::::::=;~_
defects
Figure 12.3 illustrates the effects of crystalline
BaekseattQfOd
imperfections on channeling {B0gh, 1968; Gemmell, 1974;
energy
Swanson, 1982; Feldman et al., 1982). For channeled ions in a
FIG. 12.3. Schematic diagram illustrating ion scattering from
nearly perfect crystal [Fig. 12.3(a)}, c1ose-encounter
defects: (a) little scattering for shallow depths in a nearly perfect
probabilities are small, and backscattering yields are
crySlal, (b) dechanneling by multiple scattering from pOinl
significantly reduced. If the crystal contains defects, even
defects, and (c) direct backscattering from an amorphous surface
slightly displaced. from atomic rows, the string potential is
layer. (d) Nonnalizcd yields contributed from the above three
disturbed, and ions will be dechanneled [Fig. 12.3(b)J. The
scattering mechanisms.
dechanneled ions 'begin to see lattice atoms similarly to
nonchanneled ions. If a near-surface layer of the crystal is
The nonnalized yield, x.~-), for alignment along an axial
completely disordered (amorphized), as shown in Fig. 12.3(c),
channel (uvw), according to the continuum model (Lindhard,
backscattering yields are significantly enhanced, and the close(965), is given simply by the fraction of the channel area that is
encounter probability becomes equal to that obtained in the
blocked by the vibrating strings of atoms (EXAMPLE 12.4), or
non-channeling mode. Figure 12.3(d) shows the normalized
yields contributed from the above three scattering mechanisms.
NormalizatioD was achieved by dividing the aligned yields by -'"-...... X~uvw} = Nd1t(U 2 +a 2 ) ,
(12.9)
2
the random yield.
where N is the atomic density; a is the screening length dermed
in Eq. (12.4); u, is the lwo-dimensioDal vibration amplitude, u, =
1.414u\; and Ut is the one-dimensional vibration amplitude. This
result agrees remarkably well with experimental data for
channeling of protons in a material such as tungsten where U2 is
small, but it is generally less than the experimental values in
The following quantities are used in this chapter to describe

different yield contributions:
X is the normalized yield, given by the ratio (aligned
yield)l(random yield).
Xn is the normalized yield from host atoms.
XI is the normalized yield for solute atoms.
other cases (Barrett, 1971). The experimental X~\l_) value is
x~....... ),X~uvw) are the normalized yields for host and solute
increased appreciably if the divergence of the ion beam exceeds

or if oxides or other impurities are present on the surface
of the crystal.
atoms, respectively, for alignment along (uvw) axial

channels.
~O.l 0
X~hkl) ,X!hkI f are the normalized yields for host and solute
atoms, respectively, for alignment along
channels.
XbD is the dechanneled fraction of ions.
The half-width of the channeling dip, \~112, defined as the

angle '" at which the normalized yield from host atoms, lh,
reaches a value halfway between its minimum value and the
random value (unity), is related to the Lindhard characteristic
angle'll' (defined in Eq. 12.8) by some constant c" (-1) that
depends somewhat on the vibrational amplitude of the atoms
{hkl} planar
251
/
Chapter 12
where,+, is the initial incident angle when the ions strike the crystaL
Pin is distance between the atomic row and the initial striking
(12.10)
Using the results of many numerical simulations, Barrett (1971)
showed that the critical angle half-width can be expressed as
position.
In analytical calculations, it is further assumed that statistical

equilibrium of ion positions has been reached. This means that
ions are uniformly distributed within an accessible area specified
by the equation
(12.11)
where ~ = 0.85u,/a and Fd~) is given in Fig. A17.2 in Appendix
17. Barrett gave an equation similar to Eq. (12.11) for planar

channels. Calculated values of 'VII2 are compared with
experimental values in Appendix 17 (Table A 17.5).
The continuum, single-string model of channeling, with the
inclusion of atomic vibrations, has been used to calculate the
shape of channeling dips. In Fig. 12.4, the calculations are
compared with experimental results for ions in W (Andersen,
1967). In this case, the aligned value of Xh. the angular width of
the dip, and the increase in Xtt above unity in the shoulder regions
agree quite well with the experimental data. The reason for the
high yields in the shoulders is that ions that are incident at angles
slightly larger than lJIll2 have a higher-than-average probability of
colliding with atoms in the strings. These calculations are valid
only near the surface of the crystal, because multiple scattering of
ions by electrons and other defects, as well as the energy loss of
ions as they pent:tIate the crystal, are neglected.
(12.13)
The ions are excluded from an area npl close to the string. It
follows th~t the ion flux is enhanced near the center of a channel.
Under these conditions, a simple estimate of the flux distribution
within a channel can be made for a given value of '1', using
potential contours of the type shown in Fig. 12.5. For ' = 0, the
normalized flux, F i, at the jth potential contour is given
approximately by (EXAMPLE 12.4)
A,
Fj = JdAlA=ln(A,IA,j,
(12.14)
where ~ is the cross-sectional area of the channel and Ai is the

accessible area (at a potential lower than that of the jth potential
contour)(Andersen et ai., 197Ib).
(a) <110> Channel
(b) Cells of strings
equl-potentlal como.urs
1.4
. '00
tOO
50
2D
10
5'----8-6-4-20246
Angle from <100> (degrees)
FIG. 12.4. Experimental and calculated (dashed line) backscattering

yields for (100) axial channeling of 480 MeV protons in W at 309
K (from Andersen, 1967).
Potential
minimum
12.2.2 FJux distribution and trajectories of channeled ions
Because of the steering action of cbanneling, ions are directed
toward the center of a channel, resulting in an enhanced ion flux
there. In order to calculate ion trajectories and, thus, the spatial
distributions of channeled ions within a given channel, it is
usually assumed that the transverse component, Eb of the ion
energy is conserved (energy losses are neglected). Thus. the ion
transverse energy remains constant and is given by
(12.12)
Center
of
channel
0.036 LLJ..l...L.......l_'-....L_......l_L..L-"o
0.10.512
5 10
20
50100200
FIG. 12.5. (a) Equipotential contows (in units of clcctronvolts)

for He + ions in AI for one-quarter of a (110) channel, as
indicated by the shaded area in (b). The results were obtained for
the 14 strings of atoms indicated by the solid circles in (b). Tbe
first and second ceUs of strings are designated by solid and
dashed circles, respectively (from Matsunami e/ aI., 1978).
.,
i
252
.4
.(
,,'
.li
Ion Channeling
Detailed calculations of ion flux distributions, using

continuum potentials calculated by summing several string
potentials and including thennal vibrations and multiple
scattering, have given flux profiles of the type shown in Fig. 12.6
(Van Vliet, 1971; Matsunami etal., 1978; Nielsen, 1988).
(12.15)
where n is the number of atomic strings bordering the channel,
pch is the distance from a string to the center of the channel, C is
a constant taken as 3 1n , a is the screening distance, and 'VI is the
Linhard characteristic angle (Van Vliet, 1971).
EXAMPLE 12.4. Calculation off;.
In Fig. 12.5, the areas outside the strings enclosed by the 20

eV potential contours are -0.6 of the total, giving F j
,I
- - Present analytical calculation

--__ Simulation calculation by BarreH
0.05 _?
---
12
<110> Channel
Equi-flux contours
0.1
1\
16
1n(l/O.6) ~ 0.7, in good agreement with Fig. 12.6.
0.01
x
x'
"
20
... xx
x
\
;0'
"('
I[
(a)
\ ,-, /
x~x'"
"E
,..,'"
II.
iII ....
~
x
(b)
4
2
0.5
1.0
1.5
2.0
400
)
Potential
minimum
Center
01
channel
I
2.37
2.34
\.\
~~'--
0.5
0.2 0,1.05
These depth oscillations are less regular for axial channels

than for planar channels, because axially channeled ions can
move to neighboring channels more easily than planar channeled
ions can. With increasing depth, the oscillations are damped out
rapidly because of multiple scattering and the anharmonicity of
potential contours. In Fig. 12.7, the, calculated axial depth
oscillations of the mid-channel flux are shown for I MeV He'" in
a (100) channel of Cu (Van Vliet, 1971).
Before statistical equilibrium is reached. the ion flux not only

varies across a channel but also exhibits strong depth
oscillations. This occurs because of the rather regular movement
of channeled ions from one side of a channel to the other. Thus,
peaks in the mid-channel flux (and minima in the flux at the
channel walls) occur at }J4, where i\. is the wavelength of the
channeled ion's path. An analytical expression for A., based on
the harmonic bowl approximation for the continuum potential is
12.2.3 Defect scattering

Damage to a crystal involves displaced lattice atoms and
broadly includes strain, point defects, dislocations, stacking
faults, twins, defect clusters, small precipitates, and amorphous
253
- .. -=-_. "--
1 MeV He+ ions in a {loo) axial channel of Cu. (a) No multiple

scattering included; (b) multiple scattering, eo.ergy losses,
Ihennal vibrations (T = 273 K), and beam divergence (0.06')
included; and (c) same as for (b) but with a beam divergence of
0.23. The full curves are best fits to the computed results (from
Van Vliet, 1971),
FIG. 12.6. Equiflux contours for 1 MeV He+ ions in AI at 30 K,

showo for one-quarter of a (Il 0) channel. Results of an
analytical calculation are shown as solid lines, and those of a
Monte Carlo simulation are shown as broken lines. The flux
values have been normalized to the value for random incidence
(from Matsunami et al., 1978).
1200
FIG. 12.7. Depth vanation of the calculated mid-<:hannel flux of
L---.l_.L..L:'--'...L_-1_lLLl..L'::'
2.0 1.5 1.0
800
Depth (A)
, 'i
Chapter 12
regions. Damage produced by an ion beam consists of simple

Frenkel pairs (a vacant lattice site plus the self-interstitial atom
ejected from that site) and clusters of vacancies and selfinterstitials, some of which might condense to form voids or
dislocation loops. High concentrations of defects caused by ion
irradiation can produce amorphous regions.
(12.16), g~~) is a geometric factor giving the fraction of each
site that projects into the given channel.

Because of the lack of knowledge about a stable
ccnfiguration of displacements. it is usually assumed that, for
single, randomly displaced atoms, the defect scattering factor is
given by g:UVW)F;{IlV") =1. Therefore, Eq. (12.16) is usually
As oullined in Section 12.1 and Fig. 12.3, displaced atoms

can affect channeling in two qualitative ways: by dechanneliog
(gradual deflection of the channeled ions out of channels) or by
direct backscattering of (;haoneled ions. Of course, in practice, a
combination of these two processes usually occurs. Limiting
cases are (1) "distortion"-type defects (e.g., a high density of
dislocations), which cause mostly dechanneling because they
involve small displacements of host atoms into the channels,
and (2) "obstruction"-type defects (e.g., amorphous regions),
which cause mostly direct backscattering because they occupy
central positions in the channels.
expressed as
n (x)
X,(x) ; X",,(x)+[I-X",(x)]_O_,
(12.17)
where nD(x) is the defect density at a depth ofx and n is the total
atomic density. In order to extract nD(x), the decbanneled
fraction XhD must be known.
Lattice vibrations can be considered to be another exari:lple of

distortion~type defects, and it should be noted that the amount of
dechanneling caused by dislocations or point defects is
dependent on the temperature of the sample; that is, the amounts
of dechanneling caused by defects and by thermal vibrations are
not, in general, additive (Matsunami and Itoh, 1973; Howe el al.,
1976).
The effect oflatt;ce defects on channeliog can be descnDed
most simply by considering that the ion beam consists of a
channeled fraction that is backscattered only from displaced
atoms and a dechanneled fraction that is backscattered from all
atoms in the same way as a randomly directed beam (Bogh. 1968;
Merkle el al., 1973; Feldman et 01., 1982; Swanson, 1982).
Channeled ions gradually become decbanneled (deflected out of
the channel) by multiple collisions with electrons and with
displaced atoms. including thermally vibrating atoms and lattice
defects (Gemmell, 1974). In backscattering geometry, the
Donnalized yield, lb.. is equal to the decbanneled fraction of the
beam plus the yield arising from the backscattering of the
channeled ions from displaced atoms. For alignment along a
channel (uvw), the nonnalized yield from host atoms, 'X~.......-),
12.2.4 Dechanneling by paint defects

As ions penetrate a crystal, their transverse energy
component, E.L, gradually increases because of mUltiple
collisions with e1ectrons and displaced atoms. This effect
eventually causes ions to be deflected out of channels
(dechanneled), producing an increase in the yield, X~UYW), and a
decrease in the channel half-width, 'l'l12, as the ions penetrate
deeper into the material (Fig. 12.3). For shallow depths, where X
< 0,2. dechanneling is caus~d mainly by nuclear multiple
collisions with thermally vibrating atoms and with lattice
defects, whereas multiple electronic collisions become
increasingly importaot at greater depths (Gemmel~ 1974).
In a commonly used phenomenological dechanneling modeL
the rate at which the channeled fraction (I - Xon) of ions
becomes dechanneled by a concentration Nd of irradiationproduced defects is given by (Bogh, 1968; Merkle ef al., 1973)
d(1-X",)
<Ix
(12.18)
where O'd is an effective defect cross section for dechanneling

and Pdi is the effect of thermal vibrations. Integration of Eq.
(12.18) gives
can thus be written as

(12.16)
(12.19)
wbere X:;W)
is the dechanneled fraction of ions and
(I-X:';'-') is the channeled fraction of ions. The total fraction,
where Xv is the value for the undamaged crystal,
4, of displaced atoms can be divided into components, Cui. that

are distributed among various sites i in the channel where the
In a more fundamental treatment of dechanneling, the

diffusion equation for the mUltiple electronic and nuclear
scattering of ions is solved to find the distribution. g(ELX), of
the transverse ion energy, E.Lo as a function of depth (Lindhard,
normalized ion flux has values F;{.vw). Trus normalization is

with respect only to the channeled part of the beam. In Eq.
254
Ion Channeling
1965; Bonderup cl at., 1972; Gemmell, 1974). A.n ion is

assumed to be dechanneled when its transverse energy exceeds
a critical value, E~. Rechanneling is neglected. Then, the'
It is clear from Fig. l2.8(b) that the spectra lack d~ecl

scattering peaks. Under the condition that direct scatt ering is
negligtble, the following approximation can be made
dechanneled fraction of ions is
(l2.22)
%bD(X)-X",,(O) =
Jg(EL'x) dEL
From Eq. (12.19), defect density can be calculated as
(12.20)
E,
(12.23)
One problem with the random-walk (diffusion) approach is

the dIfficulty of specifying an initial distribution, g(El.,O). at the
surface. In fact, for small values ofr., 0.2), it can be assumed
that most of the ions that are dechanneled had large values of Elo
at the surface. Thus, dEl./dx is much larger for those ions than
for well-cbanneled ions and is determined largely by nuclear
scattering. This leads to the "steady-increase" appr~ximation for
dechanneling, in which diffusion is neglected and the transverse
energy of a given ion at depth x is determined solely by its
initial transverse energy at the surface (Foti et aI., 1971). This
approach leads to dechanneling results that are similar to the
diffusion equation values when X < Q.2 (Booderup et al., 1972).
For dislocation lines, the planar dechanneling cross section

per unit dislocation length is lp (given in units of distance) and
is given approximately by (Mory and Quere, 1972)
Ap
v,
dO
J,
(12.24)
"D
J~ dO.
Eb,
cZ l Z 2 e Np
where b is the magnitude of the Burgers vector of the

dislocation loops, c is a constant dependent on the dislocation
orientation and type, and Np is the atomic density in the plane.
With
= Ap in Eq. (12.23), the obtained value ofnD is the total
projected length of dislocation lines per unit volume at a depth
ofx.
The cross section for dechanneling by point defects is the

close-impact collision probability of scattering through an angle
greater than the critical angle \jIh which is given by
aD C'JI.l =
=(
12.2.6 Energy dependence of dechanneling

The energy dependence behaviors of dechanneling from
different classes of defects should differ markedly. Thus, in
principle, dechanneling can be used to identify the type of
defects, although in most practical cases where a mixture of
defect types occur, the utilization of other techniques such as
electron microscopy is required. ~.
(12.21)
12.2.5 Dechanneling by dislocation loops

A good example of the dechanne1ing caused by dislocations
was gi ...'en by Picraux et al. (1980). In that experiment, a network
of misfit dislocations was created in St. The dislocation array is :
seen clearly in the transmission electron micrograph of Fig. - . - ~;:n:rgy dependence is expected to be EO for stacking
12.8(a) (picraux et al., 1980). The depth of these dislocations
faults, EII2 for dislocations, and E-I to FII2 for point defects and
(-450 nm) is easily determined from the position of the step in
defect clwters (Quere, 1974). In the simple case ofFig. 12.8, the
the backscattering spectra for both the (111) axial and {II0}
dechanneling step exhibited the ex.pected 112 dependence.
plaoar charmels [Fig. l2.8(b)]. Note that the step is much more.
pronounced for the planar channel because of the smaller
channel half-width for planar channels.
255
Chapter 12
(b) Channeling
1.5
Depth ()un)
1.0
0.5
8.
--"'r--..,..--~-~ ~v
Nonehenneled
Ii
....
Dfsloeatlon
(al TEM
0.8
1.0
1.2
1.4
1.6
Enl!rgy (MeV)
(e)
p. Implanted SI
3000 A
0.5
1.0
1.5
'-----'
E"2 (MeV)'12
FIG. 12.8. Analysis of pbosphorous-diffused Si by <a) transmission electron microscopy and (b) backscattering spectra for 2.5 MeV
He ions incident along {llO} planar and (Ill) axial channels and (e) energy dependence of the planar dechanneling. A misfit
dislocation network is observed at a depth of 450 nm (from Picraux et aT., 1980).
(
backscattering yield of2 MeV He ions reached the random yield

(the plateau region). A series of furnace anneals caused the
amorphous layer to recrystallize epitaxially onto the substrate.
The epitaxial growth was linear with time for a (100) surface
U.2.7 Dechanneling by beavily damaged regions

Channeling is commonly used to measure the epitaxial
regrowth of semiconductors that are amorphized by ion
implantation. A classic example of this is shown in Fig. 12.9,
where Si was implanted with Si in order to amorphize the
surface layer without introducing foreign atoms. The
implantation energy was varied from 50 keY to 250 keY, which
amorpbized the clystal to a depth of 460 run, This depth
was determined directly from the energy interval over which the
orientation of the crystal and is easily measured from the data of
Fig. 12.9. The regrowth rate was lower and nonlinear for a
crystal having a (Ill) surface, and considerable residual damage
(largely in the form of twinned regions) remained after complete
recrystallization.
256
',i
I
I
I
I
I
I
I
I
I
I
Ion Channeling
Ion-induced buried damaged layers have also been measured

in metals and semiconductors by channeling in cases where the
crystal has net been amorphized, using an analysis based on Eq.
(12.17). The variatioa of damage with depth has beea measured
and compared with theoretical range distributions of the
implanted ions.
Although single scattering accounts for dechanneling in a
reasonable manner for low amounts of disorder, plural or
multiple scattering of the channeled ions becomes important as
the amount of disorder increases. Further discussion on:
decbanneling caused by a thick amorphous layer can be found iIi
Appeadix 17, Eq. (AI7.11).
l!!
<
12.2.8 Energy loss of channeled and random ions

The energy losses of channeled ions and random ions can be
significantly different. This is due to the fact that channeled ions,
with a typical ion flux distribution peaked in the middle of
channel, have small probability of a close encounter with atom
rows and, consequently, have redLlced interactions with innershell electrons of target atoms. Therefore, channeled ions usually
have a stopping power less than that of random ions.
Both theoretical models and experimental studies that address
energy loss for channeled particles are lcss well developed. The.
energy difference between channeled and random ions can cause
considerable error in detennining depth scales afRBS spectrum.
Fig. 12".1 0 shows energy loss of He ions as a function of the
incident ion energy for the (IOO), (Ill), and (110) channeliag
directions ofsiIicon (Shao et al' J 2006).
.. 0
=.
~
1:)
150
m'ng~4
t~~~~~~'~6~'-~~~'~~'~>U~~~"~!~~~'
~~
~~~~.
S ...
amln.,;
1~ ~ \~"-.:--~
...min......
...
~-~<v...,.l
oo
.....
.i'
..
.;
~~.
mm ~f;O
~9
...In~"
.... /." .1;"";<:,. x

.4511'
...
..min.
~~v~..~..:
:~
'-.R-~."...,~ \;l;'?M
~.
'!.""oJ'''''''. ~
.0. . .o,,~'tJa
\.,_.
...
n.-.
'""'v \I.-
min
.:
.~,~:
'...
o~~
<111> _. _
-i
.",
'I.
840
"'<"t
__~____~__~~
Channeling stopping powers of limited ions (He and Li) can-

0.9
1.0
1.1
1.2
be fitted through a semiempirical expression frrst proposed by~:..:.. - - "..... - .,' ","~ .Energy (MeV) _~~__ . "._ ._ .~_. _ . .~ ..... -.
Nieman.n et al. (1996). Fining parameters from Azevedo et a/.
~IG: .12.9. Backscattenng sp:ctraJor 2 MeV He Ions channeled
I
are provided
2002).
in Appendix 17 Table A17.9 (Azevedo et a1.
dE
EII2 In(2. 71828 + IlE)
dx
(12.25)
where a. and f3 are fitting parameters and ~ is the ion energy.

It should be noted that reduction in stopping power for
cbanneledjons is a phenomenon that depends on ion penetration
depth (Araujo et al., 2004). For shallow penetratioa depths (a
few hundred nanometers), the stopping power is close to the '
random stopping power because of a nearly uniform ion flux
distribution. At deep depth, the stopping power approaches the
random value as a result of significant dechanneling.
257
m slhcon samples that were lDlplanted at 80 K, pre-annealed at

673 K for 60 min, and annealed at 823 K for the indicated times.
The upper portion shows (100) spectra for a sample with the
surface Donnal 0.3 0 off a (100) a.x.is. The lower portion shows
(11) spectra for a sample with the surface normal 0.3 off a
(111) axis. The depth scale was calculated assuming the bulk
density of Si (from Q;epregi et al., 1976).
12.3 DATA ANALYSIS
12.3.1 Channel-energy conversion
The type of detector widely used in RES analysis is the
surface-barrier detector.1t consists ofa silicon (n-type) SUbstrate
and an evaporated thin layer of gold on the silicon. The Structure
forms a diode whose depletion region increases with bias
voltage. the number of c:1ectron-hole pairs created in the
depletion region is proportional to the particle energy loss in this
region. The created electrons and holes cause a pulse of charge
that is then amplified and analyzed. The height of the pulse is
then converted to channel numbers. A good approximation gives
Chapter 12
till
E=c+-n,
12.3.2 Energy-depth conversion

For a particle incident with angle ~J with respect to sample
normal that backscatters from depth t for a scattering angle <Pl,
!he final energy is given by
(12.26)
where E is energy of the particle and n is the corresponding

channel number in the spectrum. For channel-energy
conversion, the values' of c and tlEIiID must be known. These
values. can be determined by linear fitting of !he above equation
using a calibration sample that contains at least two different
elements on the surface. The energy of baclcscattered particle

from these elements can be determined by E = kEo, where k is
the backscattering kinetic factor and Eo is the energy of the
projectile. [f element bas a deltalike spectrum, n is selected to
be the channel corresponding to the spectral peak. If the
spectrum is steplike or boxlike, n is the channel at which the
where (dEJdx)E, is the energy loss for the particle on its

outward. path. EJ is the energy of the particle immediately after
scattering. For near-surface analysis, the approximation EJ ;:: ~
can be used.
yield begins to reach !he half-height of !he spectrum.
Equation (12.27) is based on !he assumpfion of a constanl

dEJdx. A more accurate approach, a point-to-point conversion
routine, can be used to create the corresponding energy for a
":' backscattering event occurring at a specific depth. In this
approach, the sample is divided into many discrete thin layers
(slabs), Starting from the surface, the energy of the incident
particles at each slab boundary is calculated. Assuming that a
backscattering event occurs at the boundary of each slab, the
corresponding energy of the backscattered particle is calculated.
For both the inward and outward paths, stopping powers as 8
function of energy are needed to determine the energy loss in
,---r--,
'\.2 ; - .'-'1'--'---'!t'~:
1.. 0
.... ~.- .. ------...........................
0.8
0.' I0.'
1.2
".0
<100>
L:==",*=====l======l
~
.. _-...................
,0,8
each slab. A semiempirical expression of stopping power is

needed for Ibis process (see Table A17.9, Appendix 17).
,6f8
(oj)
12.3.3 Separation of dechann.ling fraction

To analyze displacement profiles quantitatively, the
backscattering of channeled particles from displaced atoms
needs to be separated from the backscattering of dechanneled
particles from lattice atoms. There are three approaches for this
separation, as descnbed in the Dext three subsections.
'
.J t,."
.~~ :~...-'",.
'~';", .
,.
J'~
.~.~
'" ~q-';'l,:-f~..;
',~~-:to>~:u'
~~~~~--~--~~
1'
E~Gf'~'(MQ!O
-'..9J ..... '..
.'
FIG. 12.10. The ratio of channeling to random stopping loss of

'He+ ions along !he (100), (Ill), and (110) Si axes as a function
of incident energy (Shao et al., 2006).
12.3.3.1 line approximation

The easiest way to separate the dechanneling component is by
using a stIaigbt line as shown in Fig. 12. l1(a). Usually, the line
connects two points. one corresponding to a minimum yield after
the surface damage peak and the other corresponding to a
minimum yield after the damage peak.
258
&
Ion Channeling
12.3.3.3 "Double-iteration procedure

A double ...iteration procedure has also been propo.s:ed that
can improve the accuracy in dechanneling separation (Sllao and
Nastasi, 2005). The method requires no knowledge al>out 0"0.
Furthennore, both 0"0 and XhO(X) can be determined thrc>ugh the
procedure. The detailed steps for this method are described next.
~ !"'q~pr~~t~d ~i
'... 'Y'irgi~. ~i
ta)
The double-iteration approach requires selecting a :point at

Xl giving DD{XI) = O. XI can be any poinl beyond which
direct backscattering centers diminish to zero. By cOOlbining
Eqs. (12.17) and (12.19), the normalized yield is obtained. as
depth
J! . ; .......... .
o~~~----.---.----.----r---~
o
-,300
600
x, (x,) =Xv (x,) + [I-Xv (x,}]x
Si Depth (tim)
.3
E'
"
i'"
jlii.
(b)
.!'!-
'"'"
'"'"
>='"
(f)
(12.28)
hD
'*" ~
"1.,
J,!
m=1
m:;;;2
m=3
--- . . n;=4
--m=10
-cxp - 'J' a o Xh(X)-XhD(x) n ]}.

[ o
X (x)
"I
Caculated Ino
2
{I
'"
I
For any point x < x], the dechanneling component,. lItO, is

given by
"r
(12.29)
. ..
"
..
:;00
where ro- represents the number of cycles of the itera.tive

calculation.
soo
Si Depth (nm)
FIG. 12.11. The channeling RES speclrum and tbe calculated

decbanneJing component for H-implanted Si. (a) Line
approximation. (b) Double...iteration procedure after different
cycles (m) of thc iterative process (Shao and Nastasi, 2005).
12.3.3.2 Herative procedure

In a standard iterative procedure, the substrate is divided
into multiple layers. The analysis starts from the surface layer by
assuming XhD(Z) = 0 in Eq. (l2.! 7), and no in the fust layer is
calculated fIrst. Then, the amount of dechanneling in the next
layer is calculated by using Eq. (12.19). This amount of
decbanneling is then used in Eq. (l2.! 7) to obtain the number of
defects in the second layer. This result is used, in turn, to
calculate the additional amount of dechanneling in the next
layer. The procedure is continued unW the defect density drops
to zero at a depth beyond the disordered region.
This iterative procedure requires that the dechanneling cross

section, aD, must fIrst be known. Theoretically, Go is given by
Eq. (12.21). However, the calculation did not consider the lattice
distortions around defects. The strain relaxation around defects
can enhance dechanneling signifIcantly.
259
First, an estimated Xno{x) curve (it can be any arbitrary

curve) is input into Eq. (12.28), so that aD can be cak:ulated. ~~
Both X.o(x) and aD are then used on the right-hand side of Eq.
(12.29) to obtain a more accurate estimate of XhD{X) on tbe lefthand side. The improved Xhl)(x) value is then substituted back
into Eq. (12.28) for a repeated calculation. The calculation is
repeated until both XhD(X) and ero fInally converge. A
demonstration of this technique is shown in Fig. 12.11(b).
12.4 LATTICE SITES OF SOLUTE ATOMS

12.4.1 Principles
One -of the most frequently used applications of channeling
has been to detennine lattice sites of impurity atoms in crystals
(Davies, 1973; Picraux, 1975; Swanson, 1982; Feldman et ai.,
1982). All shown in Fig. 12.3, impurity atoms that project into
axial or planar channels are "seen" by the ion beam, causing an
increased backscattering yield from the solute atoms, ;(5' as
compared to that from the host atoms, lh. By comparing yields
for different channels, the positions of solute atoms can be
determined quite accurately using a triangulation procedure. The
channeling method is one of the most direct ways of detennining
actual lattice positions of solute atoms, in contrast to other
methods such as the Mossbauer effect and electron paramagnetic
Chapter 12
resonance, which determine the electronic environment of solute

atoms rather than providing direct spatial information.
hcp structure would show profiles of type (a) for {1120) planar
The easiest lattice site to study by channeling is a highsymmetry interstitial site such as the tetrahedral interstitial site
in a face-centered-cubic (fcc) lattice (Fig. 12.12). This position
is the e/J,}/4,lh~ site, where do is the lattice parameter. It lies
exactly at the center of (100) channels, is one-half the distance to
the center of (110) channels, and is completely shadowed in
(III) channels (that is, it lies along (Ill) strings of atoms). The
projection of tetrahedral sites into each of these channels is
shown in Fig. 12.12. The normalized yields of backscattered
ions (or other close-encounter reaction products) from solute
atoms that are exclusively in tetrahedral sites will be
for (000 I),
r~'II)
'" rt') , X~'ID)
channels; of type (c) for
(10 TO)
axial channels; and of type (d)
(1120), {0001). and {I 0 To}
channels.
"1, and X~'OO) > 1. In fact, from (IOO)
angular scans, a clear peak in the yield from solute atoms should
be seen at perfect {l00} alignment because the ion flux is high at
the center of the channel.
Tetrahedral Interstitial site
FIG. 12.12. Tetrahedral interstitial (0) in a face-centered-cubic

structure. Projections of this site into thc (100), (liD), and (111)
channels are shown (from Swanson, 1982).
The angular scans of backscattering yields show

characteristic profiles when solute atoms project various
distances into the channels (Davies, 1973; Gemmell, 1974;
Feldman et aI., 1982; Swanson, 1982), as shown schematically
for a low-index axial channel in Fig. 12.13.
The general procedure for locating solute atoms is to

determine the type of displacement that predominates. using
selected angular-scan data of the type shown in Fig. 12.13, and
then to compare the channeling data with analytical or computer
calculations to determine precise lattice positions. Symmetry
considerations dictate how much infonnation of this type is
required. A considerable number of scans can be necessruy when
a low-symmetry position or several different solute atom
positions occur. However, if the solute atoms are in a highsymmetIy position. channeling data along only two or three axes
usually are sufficient to dctenninc the specific position (Fig.
12.12).
(a) For solute atoms in substitutional lattice sites, the yields

from solute and host atoms are identical for all channels
(assuming "the samc--vibrational amplitude).
(b) For solute atoms projecting a few hundredths of a

nanometer into a channe~ the dip in yield from solute atoms
is narrower than that from host atoms.
(c) For large projections ("'().l nm), small double peaks or a
small central peak is observed..
-- (d) For displacements into the center of a channel, a large,
narrow peak in yield from solute atoms is seen.
(e) For randomly located solute atoms, a flat solute profile is
observed
To detennine exact lattice siteS 'of solute atoms, it is

necessary to compare angular scans of the type shown in Fig.
12.13 with those calculated from flux contours (Fig.12.6).
Analytical calculations of (110) angular scans obtained from
the results of Fig. ]2.6 are sho\VD. in Fig. 12.14 for various
displacements of solute atoms into the (110) channel of Al
(Matsunami el al., 1978). Displacement of 0.2 nm along (100)
corresponding to the tetrahedral site of Fig. 12.12 gives a peakto-valley ratio of 2.2 in the (100) angular scan. When unique,
high-symmetry positions are involved, it is possible to specify
in this way the position of the solute atom to less than 0.01 nnl.
If a variety of sites occurs, this procedure inevitably leads to a
variety of possible combinations of lattice sites, and
considerable model fitting can be required. In such cases, the
physical insight gained by performing a systematic series of
experiments to alter the lattice sites is helpful.
Combinations of substitutional plus interstitial sites and
substitutional sites plus small displacements are also shown in

Fig. 12.13. For high-symmetry interstitial positions, the effect of
the crystal synunetry is to give profiles of types (a), (c), and (d)
for different channels; such an interstitial position often lies in
the center of one set of channels, off-center in another set, and
completely shadowed in another set (see Fig. 12.12). Lowersymmetry positions project into all channels. The projections of
tetrahedral and octahedral interstitial sites into major channels
for body-centered-cubic (bee), fcc, diamond-cubic, and
hexagonally close-packed (hcp) structures are shown in
Appendix 17. For example, the octahedral interstitial site in an
260
Jon Channeling
ljJiJ
o
2.4~-~-~-~-~-~--~-..,
20
2.2
<11fb. Ctlannel
'If
(a) Substitutional
2.0
Channel cross section
1.4
Host
'If
,.
(b) Small displacement
'If
:!! 1.2
(e) Random
1.0
l
1
[1[1
"""
"-""
'If
(c) Large displacement
'If
(d) Center of channel

(interstitial)
X
1
0.8
0
0
'If
(f) Substitutional
plus Interstitial
FIG. 12.14. Calculated yields in a (110) axial cbannel for atoms

displaced along the (100) direction as a function of the incident
angle 11', nonnalized to the Lindhard characteristic angle 'V1- The
o
'If
g) Substitutional
'plus small
numbers on the curves are the displacemeni values in angstroms.

The calculation was for 1 Me V He ions in a \ 11 O} channel of AI
at 30 K at a depth of 100 run and for a beam divergence of

0.048 (from Matsunarni e/ al., 1978).
displacement
Both qualitative and quantitative infonnation on the lattice

location of solute atoms is obtained from measurements of the
FIG. 12.13. Angular scans for different projections of a solute

atom into an axial channel (from Howe et ai., 1983).
nonnalized yields from host atoms, 'X.~UVW) , and from solute
261
1_Chapter 12
x;'''''').
atoms,
for perfect alignment along various (uvw)
channels. A useful empirical quantity from such measurements
is the apparent fraction of solute atoms displaced from lattice

sites into the given cbannel, f~-) (Swanson, 1982)
(uvw)
f(uvw) ::;:
X.
is important to realize, however, that
Ft
vw
)
is very sensitive to
factors such as surface oxides, beam divergence, and depth of

analysis.
{uvw)
-Xh
(12.30)
l-X~"""')
<h
side of Eq. (12.32) with calculated ratios on the right-hand side

of the equation for various possible sites i enables the position(s)
of the solute atoms to be determined (Matsunami et al.. 1978). It
12.4.2 Comparison of RBS, PIXE, and NRA methods
This would be the actual fraction of solute atoms that project

into the channel iftbe (normalized) ion flux at the solute atom's
position were unity. However, because the ion flux varies

greatly across a <::hannel (Fig. 12.6), f~uvw) is larger than the
true displaced fraction when the solute atoms are near the
The cboice ofRBS, particle-induced X-ray emission (PIXE),

or nuclear reaction analysis (NRA) (or a combination of these
methods) for lattice-site measurements is dictated by the
particular element to be analyzed. Careful attention must be paid
to the cross section for a particular reaction and to the depth
information available. Some examples of the use of these
methods will be given in Section 12.8.3.
center of the channel, and f~uvw} is smaller than the true

displaced fraction when the solute atoms are near the edge of
the channel. If a fraction C si of solute atoms is present in each
12.4.3 Analytical ver.;us Monle Carlo analysis

There are four degrees of sophistication for analysis of lattice
site data, in order of increasing acewaey:
of several different lattice sites labeled i, then
(1) Examine the angular scans (or just the apparent displaced
fraction 'rd~U""') to estimate lattice sites from Fig. 12.13.
(1231)
(2) Use the statistical equilibrium approximation in Eq. (12.14)

for the ion flux at different positions in each channel and
g:UVW)F;( uvw) is the effective ion flux at each of those sites
compare with angular scan or f~"""'} data (Andersen et al.,
[compare Eq. (12.16)]. Here, g~uyw} is a geometric factor equal
1971b; Swanson et al., 1973).

(3) Use analytical calculations with multistring Moliere-type
potentials to calculate equipotential contours and- ion fluxes,
and then calculale Iillgular scali yields [Eq. (12.14) and Figs.
12.5, 12.6, and 12.13] (Matsunami et al., 1978; Nielsen,
1988).
(4) Use Monte Carlo calculations of flux profiles and angular
scans (Barrett, 1990; Smulders and Boerma, 1987;
Smulders et al., 1990).
where
to
the fraction of solute atom sites that project into the given
(uvw) channel, and
~IIVW)
is the normalized ion flux for each"
site calculated with respect to only the average channeled flux.
In other words,
F;(m,w}
is the ratio of the channeled beam
intensity at the solute atom site to the total channeled beam

intensity. This is the quantity normally calculated, as in Fig.
12.6, where decbanneling is neglected.
The choice of analysis depends 00 the amount of infonnatiol1

required and the complexity of the lattice sites and crystal
structures. Many programs are available for both analytical and
Monte Carlo calculations.
As an example, consider that one-half of the solute atoms are

in substitutional lattice siteS (Csl = 0.5 and Fsi = 0) and one-half
are in interstitial sites (52) where the flux Fa
2; then f~D""')
1. From flux calculations of the type sho\lm in Fig. 12.6,

empirical values of f~\IVW)
12.4.4 Numerical meth.od to calculate angUlar scans

12.4.4.1 Continuum approximation
the angular dependence of backscattering yields can be
calculated numerically by using continuum approximation. The
procedure consists of four steps to calculate (I) the transverse
potential as a function of location across the channel, (2) the
probability distribution of the transverse energy of the ions, (3)
the ion flux across the channel, and (4) the angular distribution
of the backscattering yields (Tang and Zhang, 1988).
for various channels ~an be
compared with calculated values of F~IIYW) for different sites. If

the solute atoms are in only substitutional lattice sites and one
unique interstitial site, then
(12.32)
Transverse potential. The axial channel can be divided into
tiny square regions of area M. The potential for each square is
where (uvw) and (uvw)* are two different channeling directions

or planes. A comparison of the observ~d ratio on the left-hand
calculated by adding potential contributions from multi-atom
262
Ion Channeling
rows. Considering periodic boundary conditions, an appropriate

unit cell sbould be selected. Taking Si (100) as an example (see
Fig. 12.15), a group of 12-atom rows with a square unit cell is
appropriate for the calculations associated with displacements
close to the channel center. A group of 9-atom rows with a
circular unit cell is appropriate for problems associated with
Ion flux distribution. The allowed area for channeled ions

with transverse energy EJ.,i is the area baving a potential equal to
or less than El..,i. The accessible area is given by A(EJ..,i) = (ml +
m2 + ...
small displacement from the atom row (Tang and Zhang, 1988).
(]2.34)
+mJ andEJ..i =llt1.
Because ions are allowed to be uniformly distributed over

the accessible area, the ion density at rh (the position.
corresponding to potential Ub,) is given by
The ion flux F(rtJ at rtl is contributed by only those ions with EJ.
~ ub' Under the condition that El..,i = E~ + Ui = Uh, ion flux is
given by
F(r.)=x
f f,(r) g(E,) dE,

EJ. ,.....
(12.35)
(b)
(a)
potential. (a) A group of 12-atom rows with a square unit cell
Angular distribution. When ions are incident with an angle of

W, the yield of close encounters is given by
and (b) a group of 9-atom rows with a circular unit cell (Tang
and Zhang, 1988).
x('I') = fF(r) dP(r) ,
FIG. 12.15. Continuum approximation for calculation of string
The potenlial for each small area M. is calculated and listed

in a table in order from the smallest to the largest. As shown in
where Per) is the displacement probability. Selecting 'V = 0,

O.JO, .. " the angular distribution is obtained by repeating the
the Table 12.1, Ui and E.l.,i represent the potential and transverse
energy, respectively, for M. mi represents the number of areas
6A that have the same potential energy (Tang and Zhang, 1988).
above calculation procedure. The last step is the normalization.

The angular distribution can be nonnalized by the yield obtained
with'll" 50/1 (Taog and Zhang, 1988).
Table 12.1. Listofpotentials from the first step of the continuum

approximation to calculate
backscattering yields.
u,
... U;
... ELi
... mpA
ELI
mlM
the
angular
dependence
b
U,
.E.Lb
m,,6A
. 12.4.2 Half-way-plaue approach-. --::-::"._:...::'-,.~-...:::,, _ _ _
of
.1..1
m.M
A
rnl
N
.c.
In the balf-way-plane approach, a displaced vibrating atom
L
UL
E.L.L
mLM
Probability distribution of transverse energy. For an ion

hitting an area with a potential of 0;, the transverse energy is
given by E.u = Eo/' + U;. Therefore, the possibility of obtaining'
the value EJ.,i is given by (Tang and Zhang, 1988)
g(E)=-'- = - ,
(12.36)
is considered to be emitting particles isotropically, and the

distributioll in the transverse energy of the particles with respect
to the a10m row is calculated at a plane halfway between the
atoms in the row. Changes in the transverse energy distribution
of the particles after transmission through the suJface are
neglected. The calculated angular distribution is directly related
to the probability of one ion hitting another atom with the same
displacement. A systematic approach to calculate the angular
distributions of backscattering was given by Picraux. et al.
(1972).
For an emitting atom that is displaced from the atomic row,
the distribution probability in cylindrical coordinates is given by
(12.33)
dP(r',~)=exp[-(;J -(:J + 2x,rt~]x
where Ao is the size of the channel and ~ = NM.

de' d{"')
p'
263
(12.37)
/
Chapter 12
where Xo is the equilibrium displacement of emitting atoms from

the row. p2 is the mean-square vibrational amplitude. See Fig.
12.16 for the geometry. If an emitting .tom emits particle in
the direction of (~,8), the emitted particle will meet the balf-way
plane at a position (rh'Y). The value of II is the perpendicular
distance of the emitted particle from the row after traveling a
distance of dl2 in the z direction. where d is the average spacing
of atoms along the row. II is detennined from the values of "0. r',
e',~, and 8 by using the Eqs. 12.38-12.41
y
where E is the energy of the emitted particle and U is the

average potential of the row of atoms. If the particle emission is
isotropic, the transverse energy distribution is given by
(12.43)
where [0 is the upper limit of atom displacement, which is given
by mo' = (NdrJ , with N as the atomic density.
x'
The emitted particle will ultimately reach the surface and be

detected. The distribution in the angle, \Ve1 of the emitted particle
is
, 'r d(r')
EMITTED
PARTICLE
P.(EIV. ) = r,'--:,rI[EIV. + U(r)].
where
(12.44)
,r,-r
is determined by Uri) = Elf.'
+ U(r).
The obtained
distribution can be nonnalized with the value obtained for 'Ve :::;
5'Vl. Fig. 12.17 shows the results obtalned by Picraux l al. for I
MeV He .Iong the Si (110) axis. The angular width of the
channeling dip increases with increasing displacement from the
lattice row (picraux et aI., 1972).
FIG. 12.16. Geometry for calculation of the cbanneling angular

distribution (picraux et al., 1972).
10 an ordered lattice or compound, a given channel can be

bordered by strings of only one element or by mixed strings of
different elements. In the former case, a difference in angular
half-width, 'VIa. occurs for the different elements because of the
relation in Eq. (12.8), namely, If' = (2Z,Z,e'IEd)'o, where Z,
and Z2 are the atomic numbers of the projectile and target,
respectively; E is the projectile energy; and d is the atomic
spacing along the strings of atoms bordering the channel This
difference has also been observed (Eriksson and Davies, 1969;
Picraux et 01., 1969) in many compounds containing elements of
very different Z, values, sucb as Ua, and GaP.
(12.38)
(12.39)
(12.40)
Both axial and planar channeling have been used to measure

the sites of In and Sn in GaP and GaAs. The principle of the
planar method (Swanson eJ al., 1989) is th.t a (Ill) plane in
GaP is bounded on one side by Ga atoms and on the other side
by P atoms. Thus, a beam of ions directed toward the Ga wall of
the channel is steered so that it encounters only Oa atoms at a
specific depth (equal to approximately A/4, where t.. is the {'
wavelength for planar depth oscillations) corresponding to the "l
critical angle 'Vt with respect to the petfectly aligned direction. 'l.
cosy
-G~ )(~ )sine sine'.
12.5 EXAMPLES IN BULK MATERIALS

12.S.1 Ordered structures and compounds
(12.41)
Once T, is determined. the transverse energy of the emitted

particle is given by
(12.42)
264
Ion Channeling
Similarly, a beam directed toward the P wall of the channel
dip occurs for several channels, it can be concluded that tbe

solute atoms are displaced only small distances (<D.OS IlIO.) from
normal lattice sites. On the other hand, if the projections are'
small into some low-index channels and large into otheys, one
can conclude that the solute atoms are displaced larger distances
into well-defmed interstitial sites.
encounters only P atoms at a similar depth. Thus, oscillations in

yield from Ga atoms and P atoms vary with depth in a
complementary way. The ,ion beam will interact with implanted
dopant atoms in the same way as with the host atoms that the
dopant atoms replace, so that the sites occupied by the dopant
atoms can be identified at any depth where the oscillations are
measurable. The tilt angle and beam energy can be tuned to a
specific depth. The axial channeling method (Andersen et at.,
12.5.3 Small displacements

A classic example of studying small displacements was given
for the case of Bi implanted into Si by Picraux et ai, (1972).
They used the following strategies to alleviate the issues in data
analysis: (1) It was assumed that the thermal vibration amplitude
of Bi is similar to that of Si. Because Bi atoms are much heavier
than Si atoms, the probability for Bi atoms to exhibit large-
1982) uses the same principle, but because axial depth

oscillations damp out much faster than planar oscillations, the
axial method is applicable only for solute atoms distributed very
close to the surface.
I~'
( \ r-,
/ 1. \
12
~ / l\ \
L
f
Le
"
:Z
amplitude vibrations in Si is small. (2) Because X caLculated

from the half-way-plane approach was less than the value from
experiments, the theoretical X value was adjusted by considering
simulation results from Barrett (1971) and also the contribution
from the surface oxide layer. (3) Because the calculated value of
\Vlf.! was larger than the experimental value, the theoretical value
of \f'1J1 from Si was adjusted by a factor to match the
experimental value. The same factor was used to adjust the 0/11:2
value for Bi.
Si {lW}, l'/IIK
"! /
' ...
G><::?-S~~=::::;:::;::::~l
- I I
II I '
iUfI
MID.,
I
I
,.
12.5.4 Gas atom sites

Small gas atoms often occupy octahedral or tetrahedral
interstitial sites in metals, as shown by cbanneling experiments
(Bugeat and Ligeon, 1979; Carstanjen, 1980; Yagi et al., 1985),
One example that illustrates the use of the channeling method
for detennining these high-symmetry interstitial positions is the
lattice location of deuterium in Pd (Fig. 12.18). The nuclear
~ 'reaction 2HeHe,p)4He ' _was. used to detect the deuterium
,. (Car.;tanjeo et ai., 1978), The angular scans of Fig, 12.18 clearly
demonstrate that the deuterium atoms occupied octahedral
interstitial sites because those sites are completely shadowed in
the (100) axial and {I OO} planar channels, but are exactly in the
center of the (110) axial and {Ill} planar channels. (See
Appendix 17 for projections of different interstitial sites into
various cbannels.)
'0 0.0
Small gas atoms such as H or D are often retained in metals

only in a trapped state, Picraux (1981) showed that the position
ofD implanted in Fe was -{),04 run from octahedral sites. This is
an example of sophisticated analysis involving computer fitting
of both axial and planar channeling data, The results indicated
that the D was trapped by single vacancies.
FIG. 12.17. Calculated single-alignmeJ;lt angular distribution as a

function of equillbrium displacement distance (in angstroms)
from the row for I MeV He+ along the (1l0) axis in Si at 296 K
(picraux ef aI., 1972),
Many experiments with ion-implanted Si have shown that the

implanted solutes can occupy a variety of lattice sites, but it is
often not clear whether these are equilibrium or defectassociated sites. One example is Yb implanted into Si: the solute
atoms lie near the tetrahedral interstitial sites, produCing strong
peaking in the Yb yield for the (11 0) channel and small dips in
12.S.2 Low-symmetry interstitial sites
Low-symmetry interstitial sites can involve small projections

[Fig. 12.13(b)] and/or large projections [Fig. 12.13(c)1 of solute
atoms into different channels. When narrowing of the channeling
265
."
Chapter 12
the Vb yields for the (100) and (lll) channels (Andersen el 01.,
1971b).
..
:!!
~1.8
c
~ 1.&
Q. 1.4
1.2
1.0
L:.2!;.-.-;:,.:-:o:-.-,0:':-:2:-'.J
Angle-
I ++
E
~
>'1.0
~~O8
j'
a: 0.6
{'OO)
1m!
T
.I
f.
II
[J
{111}
iWOli
T
V .Jl
'D
1.8
>.1.6
c
21.4
eCL1.2
~
-1.0..0.5" O 0.5 1.0

AnEJle0
'"
1.0
L----:;-_~~-----!----.J
0
1
~1
Angle from <100> direction (degrees)
., .0.-0.5 O 0.5 1.0

0
Angle-
FIG. 12.18. Analysis of D in Pd. Angular scans for D, using

proton yields from the nuclear reaction D(,He,p)'He for the
(100) and (llO) axes and the {loo} and {lll} planes in
pdo.",oO.OO7 at 295 K, using 0.75 MeV 'He ions (from Carstanjen
el aI., 1978).
12.5.5 Solute atoms associated with point defects

In genera!, the presence of nearby point defects will pUll
solute atoms out of their nonnallattice sites, resulting"in either
small or large displacements. For example, for an alloy of AI
containing 0.2% In that was irradiated to produce vacancies,
vacancy clustering about the In atoms caused a large fraction of
the In atoms to occupy tetrahedral sites surrounded by four
vacant lattice sites (Fig. 12.19 inset).
The growth of the peak in yield from In atoms for a (100)
angular scan as a function of irradiation fluence is shown in Fig.

12.19 (Swanson and Howe, 1983). This peak is strong evidence
that the In atoms occupy the tetrahedral interstitial sites, which
lie in the center of the (loo) channels (Fig. 12.12). For the
highest fluence, the peak-to-valley ratio was 2.3, which is close
to that calculated for tetrahedral sites. Thus, a large fraction
(-0.60) of the In atoms were in tetrahedral interstitial sites'. Most
of the remaining In atoms occupied substitutional or random
sites. Only the central part of the angu1ar scan was measured
(Fig. 12.19) in order to reduce the effect of damage from the
analyzing beam.
266
FIG. 12.19. (100) angular scans for irradiated Al..().02 at. % In.
Backscattering yields (X) of 1 MeV He ions from Al and In
atoms are shown as a function of the angle from a (100) axis.
The sample was irradiated with 1 Me V He ions to the indicated
fJuences at 35 K and then annealed for 600 s at 220 K before
each of the angular scans. The yields were measured at 35 K for
a depth interval of 50-280 om from the surface. Because the Ai
yields were almost the same after each irradiation, only one set
of AI data is shown (from Swanson and Howe, 1983). The inset
show5 the deduced ~trahedral interstitial position of an In atom
(0), surrounded by foor vacancies (D).
12.6 SURFACE STUDIES

12.6.1 Basic considerations
In a good channeling direction, a clearly resolvable surface
peak in the backscattenng yield can be observed because the ion
beam is scattered from the exposed surface atoms with the same
intensity as from a random array of atoms, that is, the surface
atoms are not shadowed The second atom in each string of
atoms could, however, be completely shadowed in a perfect rigid
lattice, as shown in Fig. 12.20, in which case the surface peak
would correspond to the density of the top atoms of each row or
.
1y 1015 atoms cm-i (usually denoted as one atom per
approXlmate
string). The second atom in a string is no longer shadowed if it is
displaced out of the shadow cone of the incident ions for
example, by therma1 vibrations or by a shift in lattice poSitions
caused by surface reconstruction or relaxation. Thus, channeling
can be used to measure lateral reconstruction, surface relaxation,
Ion Channeling
vibrational amplitudes, and premelting of atoms in the surface

and near-surface planes (van dec Veen, 1985; Swanson, 1982;
Howe et al. 1983).
~
C
:::J
[111'.:
200
[110]~
.....
[Y ''''
">.
;;
'"c
100
.!lo
III
300
-s"!,
FIG. 12.20. An example of a shadow cone behind a surface atom
bombarded by energetic ions. R is the radius of the cone at the .
next atom (Feldman e( al., 1982; Swanson, 1982).
A surface peak is seen in channeling spectra even for a
perfect, uncontaminated surface because the outennost layer of
atoms is not shadowed. An example of a surface peak. for clean'
Pt is shown in Fig. 12.21 (Davies et al., 1978). The presence of a
surface oxide will, of course, erihance the surface peak:, so that
measurements are usually done in ultrahigh vacuum. Depending
on the radius, R of the shadow cone (Fig. 12.20), more than one
atom per string might be seen. For Coulomb scattering, the
radius, R:. of the cone at the second atom is
R, ~ 2(Z,Z,e'dIE).
Surface
o,,'\o~
On.
'... "
o
atom
, '
Q.,~P.~
o 0 0""",,, strmg
00
One
o,~, 0
a-atom
0, 0 ... 0 -
o 0'-':-0... 0
0 ~''::~C string
0--
0, 0
0000'
00000""-
(a) Perfect lattice _
(e) Re.laxed
per
Two
atoms
SUrfa~::~ng
o
00000
where d is the spacing of atoms in the string and E is the energy

of the incident ion (EXAMPLE 12.5) (Bagh. 1973; Feldman et
at., 1982). If the thermal vibrational amplitude ex.ceeds R..;, a
second atom is seen by the ion beam.
0---
Two
atoms
OODOo-per
o
000
o 0
string
(b) Reconslructed surface
A
Energy _
FIG. 12.22. Schematit: illustration of measurements of

channeling surface peaks to study displacements of surface
atoms: (a) perfect lattice, (b) reconstructed surface, and (c)
relaxed surface (from Swanson, 1982).
EXAMPLE 12.5. The shadow cone radius for a Au layer

on Ag (Fig. 12.19).
For a 1 MeV He beam on Au, 2, = 2, Z2 = 79, d{1I0) =
0.288 run (Tables A17.1 and AI7.3. Appendix 17), e' =
1.44 x 1O-1l em MeV. E = I MeV. Thus. Rc = 0.0162 run,
whit:h t:an be compared with U2 = 0.0 12 run.
450
FIG. 12.21. Backscattered energy spectra for 2 MeV He ions

incident on a Pt(111) surface along a (IIO} direction at 60 K.
The energy scale has been expanded with a large zero Off.5et. The
full width at half-maximum (fwhm) of the surface peak is
equivalent to a depth resolution of 3.5 run. Its area is equivalent
to 1.39 Pt atoms per (110) row (from Davies et al., 1978).
400
350
Channel number
Fig. 12.22 shows schematically how the yield of the surface

pcak can be increased beyond one atom per string:
by displacement of the surface laterally, that is, in a
reconstructed or melted surface [Fig. 12.22(b)];
267
--~--------------------------------Chapter 12
12.6.2 Surface relaxation
by inward or outward relaxation of the surface layer

(observable only for nonperpendicular incidence [Fig.
12.22(c)]; and
by an increase in the thermal vibrational amplitude U2 beyond
1\.,.
The measurement of the relaxation of surface Ni atoms by the

blocking method is shown in Fig. 12.23(d) (van der Veen el al.,
1979,). In this casc, the (110) surface atoms for a clean surface
relaxed inward by (4 1)% of the interplanar spacing. When the
surface was contaminated by oxygen, an outward relaxation of
-1% was observed. Similarly. a relaxation of <0.003 om (1.5%)

was observed for (Ill) PI surface atoms both by this method
(van der Veen el al., 1979b) and by single atigrunents of the Pt
surface peak as a function of the incident angle of the probing
ion beam (Davies el al., 1980). These results are in general
agreement with low-energy electron diffraction (LEED) data,
but the ~ccuracy of the channeling method is considerably.
greater. The technique of Davies et al. (1980) has been extended
Each of these effects is more pronounced with increasing ion

energy, which reduces I\., according to Eq. (12.45).
To calculate the actual number of atoms per string that
contribute to the surface peak, Monte Carlo computer
simulations (Barrett, 1971; Tromp and van der Veen, 1983) are
-" =':-req"uired..-.. In :...-such-'. calculations, a Thomas-Fermi screened

Coulomb potential and the Debye approximation to vibrational
amplitudes are often used. DDe refinement is the inclusion of
to provide information about surface Debye tempcratw"es as
correlations between thennal vibrations of adjacent atoms,

which increases the shadowing effect (Jackson and Barrett,
well.
1979).
12.6.3 Surface impurities
One special feature of the combined shadowing and blocking

12.22, a double-aligrunent technique is extremely useful for the
technique is the ability to locate the positions of impurity atoms

adsorbed on a crystal surface. An excellent example of this is the
study of surface structure (Turkenburg et aT., 1976). In this
study of sulfur on Ni(ll 0) by van der Veen el al. (1979c).
In addition to the single-alignment method, illustrated in Fig.
approach [Fig. 12.23(a)], the incident beam is directed along one

crystallographic axis, and the detector is aligned with another
axis. The backscattering yields from both surface atoms and
atoms well below the surface [the bulk yield, Fig. 12.23(b)] are
measured. Along the incident beam direction in Fig. 12.23(a),
. only 50% of the atomic rows originate in the swface plane; the
other 50% originate in the second plane. Consequently. at least
12.6.4 Interface structures

The nature afthe interface between a crystal and an overlying
oxide can also be studied by channeling. In the case of Si
covered by an oxide layer, it was shown that the oxide was
. stoichiometric SiOl and that one or two monolayers of Si at the
Si-Si0 2 interface were displaced from DorriJal lattice sites
(Feldman
50% of the backscattered ions contributing to the surface peak

can be blocked along certain emission directions by the surface
el al.,
1978; Jaekman el al., 1980). For these
experiments, the total areas of the surface silicon peak. and the
oxygen peak were measured as the oxide was gradually removed
by etching in a HF solution. The variation of the total area
density of Si atoms in the swface peak as a function of the
oxygen coverage (Fig. 12.24) demonstrated that the oxide was
stoichiometric SiOz. In addition, the density of Si atoms
extrapolated to zero oxygen atom density was 8.6 x lOU atoms
cm-2, which was greater than the density of unshadowed Si
atoms, 6.4 x lOIS cm-2 , in the surface region of a clean Si{IlO}
crystal. The excess 2.2 x lOIS atoms cm-2 were attributed to
displaced Si atoms at the Si-Si02 inlenace.
plane of atoms, thus producing a blocking dip in the surface

peak yield. Ally relaxation, ~d, of the surface plane will rotate
the position of this blocking dip through an angle of 69 relative
'--to the uoderlying bulk blocking direction [Fig. 12.23(c)]. Thus,
the magnitude and sign of lid can be obtained from purely
geometrical considerations. Another great advantage of this
blocking technique is that it can be readily adapted to provide
information on the location of adsorbed atoms because these too
can produce observable blocking dips in the backscattered yield
from the underlying crystal.
.1
.l
1
;
268
Ion Channeling
(a)
(d)
[ [ [[
Surface
Bulk
"
Beam direction [314]

scattering plane (i11)
W-NI(110)
I<eV
Eo = 101
})))) 6.d
Surface
oJ
06
~~~~\;
(b)
If
t
."
"\
Oxygen
covered
Gi
:;:
Bulk
Surface
Backscattered energy
(c)
I
I
:-'~:-'~\SJ'-:""
_.:-,-- . -~-
I
fl'r-f.::::::::::-_ Surface
Bulk
,I lI
.......
'A6
i!
,
,
Bulk
'
40
43
46
49
52
Scattering angle e (degree)-
Scattering angle
,
FIG. 12.23. An illustration of the method to study surface relaxation experimentally, based on channeling and blocking.
(a) The scattering plane coincides with the crystallographic plane. A displacement of the first layer results in different angular
, ,
I
.!
positions of the blocking cone for surface and bulk atoms.

(b) An energy scan of backscattered ions detected at an angle e by means of an electrostatic energy analyzer.
(c) Angular scans of backscattered. intensities from surface and bulk atoms. The displacement caused by the relaxation effect causes a
shift,
in the surface blocking minimum (from Turkenburg <10/" 1976),
(d) Blocking minima for 0.1 MeV protons scattered from bulk and surface Ni atoms for a clean (110) surface arid for an oxygen
coverage of one-third of a monolayer. The Ni yield was normaliZed by the Rutherford cross section, GR' The solid symbols represent
the clean surface, and:the open symbols represent the oxygen-covered swface (from van der Veen ef al., 1979a).
"a,
269
Chapter 12
Fischer el aJ. (1988) studied the bonding at a CoSirSi
An example of epitaxial layers with a considerable lattice
interface by using the thin CoSh overlayer as an "jon lens" to

focus the ion beam onto the interrdce atoms of the Si(lll)
substrate. They found that the Si atoms at the interface bonded to
mismatch is the GeJSi system. where the lattice parameter of Ge

is 4.4% greater than that of Si. The epitaxy of Ge on Si is easily
measured by RBS. It is predicted theoretically that only about
six monolayers of (100) Ge could be grown OD Si without the
creation of mismatch dislocations or twins. However, because
growth by molecular beam epitaxy (MBE), for example, is not
an equilibrium process, considerable latitude exists in the film
quality and the: critical thickness for creation of mismatch
the Co atoms of the silicide.

12.7 EPITAXIAL LAYERS AND STRAIN
The quality of epitaxial layers and the strain in such layers or
in multilayers can be determined by channeling measurements at

normal incidence or at oblique angles. In general, good epitaxy
can be achieved only when the lattice parameters of the substrate
and overlayer are almost the same, For a small lattice-parameter
mismatch, perfect epitaxy can be accommodated by vertical
elastic strain in the overlayer. When the slress in the strained
epilayer exceeds the elastic modulus. the strain is
accommodated by the creation of "misfit" dislocations.
dislocations.
A thin epitaxial Dverlayer of Ge on (100) Si with a capping
layer ofSi is shown schematically in Fig. 12.25 (Feldman et al.,
1987). For a lattice match in planes parallel to the surface, a
considerable strain is introduced in the normal direction. In the
case of a (100) surface, this strain should not create an increase
in X~OO) but it does cause Ge atoms to project into the (110)
24
channels at 45 0 to the surface. This produces some bending of

the ion beam in these channels as the ions are steered into the Ge
(110) channels. However, for very thin layers (Fig. 12.25), the
projection ofGe atoms into the paths of ions that have been well
channeled by the Si overlayer causes unusual asymmetric
channeling effects (Fig. 12.26). An ion beam directed at an angle
slightly greater than that required for (110) alignment (>45') is
steered more toward the displaced Ge atoms, giving a large Ge
yield, whereas a beam at <45 0 is steered away from the
displaced Oe atoms, giving a lower-than-normal Ge yield in the
shoulder region. As control experiments, Fig. 12.26(c) shows' a
(I (0) angolar scan, which gives a symmetric scan, as expected,
and Fig. 12.26(d) shows the (110) angular scan for 100
monolayers of a SiO.99Geo.Ol epitaxial film, which shows very
little strain, again as expected
E"OJ!leV
~tll>~
.. 20
E
~
.Ea 16
~
~
,..0 12
~
iii
Computer simulations of the asymmetric angular scans of

Figs. 12.26(a) and 12.26(b) give good agreement when the
perpendicular displacement of Ge atoms in the nth layer of Ge at
an off-normal angle e is given by
(n- Y,)(d," -d..,,) sin 9
6
8 10 12
Oxygen (1015 atoms/em2)
14
(12.46)
16
where dosi is the Si lattice constant and
FIG. 12.24. Measured densities of Si and 0 atoms for a Si(llO)

surface that was covered by oxides of varying thicknesses. The
straight line corresponds to stoichiometric SiD2 plus 8.6 x lOiS
atoms cm-1 of excess Si. A contribution of 6.4 x 1015
is the perpendicular lattice constant of Ge. This equation follows

from application of Poisson's equation. The POiSSOD ratio is u ==
0.273 for Ge.
01
unshadowcd Si atoms per em2 from the Si substrate is indicated

(from Jackman el al., 1980).
=d.+ (i+u)(d""-d,,,)
oSl
l-u
(12.47)
.~
{
1
270
Ion Channeling
Transmission channeling is also a very useful way to study

thin epitaxial layers, for example, a few mono layers of Ge on Si
The goniometer is obviously a crucial part of the cb..anneling

equipment. Early channeling experimeDts were perforD:1ed with
manual two-axis goniometers (Davies ef al., 1967). (For
noncomputerized goniometers, tilt and azimuthal rolations are
the most convenient) Although angular scans are tedic>us using
manual angular manipulations, the superior flexibility fr.r control,
of the angular spacing and ion dose during the scans is a. decided
advantage. In addition, alignment can be done more rapidly
manually, thus introducing less ion damage to the sampL e.
(Lyman etal., 1991).

12.8 EXPERlMENTAL METHODS
12.8.1 Channeling equipment
The experimental equipment and ion beam analysis methods
for channeling studies are basically the same as described in
Chapters 4, 6, 7, 8, 11, and 16, but some details that are specific
to channeling are discussed here.
To obtain collimated beams of light ions in the 0.3-2.5 MeV
energy range, an electrostatic accelerator is normally used. A
[1001
beam divergence of 0.10 or less (achieved by placing apertures

at least 1-2 m apart) is suitable, and the beam at target is
approximately 1 mm in diameter. Micrometer-sized beams can
be used, but then ion damage can become a severe problem. A
single-crystal sample is mounted on a two- or three-axis
gODiometer in a target chamber containing suitable solid-state
detectors (Fig. 12.27). Usually, a backscattering geometry is
used, with a scattering angle of approximately 150 0 (Chu et aI.,
1978). Increased depth resolution can be obtained by using
beams that are incident at glancing angles to the sample surface
or by placing the detector at a glancing angle. Double-aligrunent
geometry is most easily achieved by placing an annular detector
in the beam path. For single-alignment measurements. the solid
angle of the detector is large enough to avoid accidental double
alignment.
150!
SI eap
Ideally, the sample is surrounded by a cold shield at liquid-
nitrogen temperatures or lower to avoid condensation of

hydrocarbons and other contaminants on the sample. This shield
is biased negatively with respect to the sample in or-der to repel ~secondary electrons, and the ion fluence is best monitored by .
measuring the total charge accumulated on the sample holder
and shield. For defect studies, it is useful to cool the sample as
well. This is conveniently done by connecting the sample to a
nitrogen bath or a cryogenic cooler by means of a copper braid
(Bottiger et ai., 1973). A small electrical heater attached to the
sample holder is then used to vary the sample temperature
between about 20 K and 400 K. For higher temperatures, the
braid is disconnected. An ultrahigb-vacuWD target chamber is
necessary for surface studies. Associated equipment is desirable
for complementary studies using low-energy electroD diffraction
(LEED), Auger spectroscopy, and other techniques (Davies and
Norton, 1980).
271
-"
51 (IDOl
substrate
FIG. 12.25. Schematic diagram of the structure of a strained

epitaxial film of Ge capped with the substrate material. Si. The
channeling directions [100] and [110] are shown. The
displacements, .xn. of the Ge atoms from the substitutional sites are
indicated (Feldman ef aI., 1987).
Chapter 12
1110)
~' MeVHe.
1.6
...
Qj
1.4
6~200A
1.2
'$.
... 1.0
Q)
.!:! O.B
1.6
... 1.4
"i 1.2
S[ CAP
'$.
... 1.0
51
Ge
IS
= 0.8
iii
CD
E 0.6
E 0.6
-. .-.-.:..~-~ ~ -0.4
0.2
0.0 ~-'~.L..L.......Jlo~-LL~-'--_
47.0 46.0 45.0 44.0 43.0
Angle from normal (degrees)
1l1li1 (100)
'ciW!
~10
SMLGe
051
"tl
Gi
's:.
1.2
0.4
0.2
0.0
~--'~.........J!O~""""",~,---_
43.0 44.0 45.0 46.0 47.0

(b)
~'.8M.VH
ML Ge
1SO
1.4
A 51 CAP
1.2
:g ,
~ 1.0
1.0
"tl
.~ 0.8
c;; 0.6
=a
E
(; 0.4
0.8
0.6
'.BMeVHe.
;S1100
100ML
SI+10t
%Go
051
- Ge
136..1.51 CAP
E
(; 0.4
z 0~2
.----z .
0.2
0.0
He
200ASICAP
Ge
(a)
1.4
1.8 MeV
L---1-LL.L..!...<........:'~L.L.L-'-'--'~_
0.0
-2.0 -1.0 0.0 1.0 2.0

43.0 44.0 45.0 46.0 47.0

(e)
(d)
FIG. 12.26. Channeling angular scan of GelSi structures, using 1.8 MeV He ions (Feldman ef at., 1987). (a) Scan through the [110]
axial direction in the (l00) plane for an epitaxial structure consisting of Si(IOO)/Ge [6 monolayers (ML)]lSi (20 run), (b) s<:an through
the [110] axial direction in the (100) plane for the same sample as in (a), (c) scan through the [100] (crystal normal) channeling
direction for the same sample as in (a), and (d) scan through the [110] channeling direction in a sample of Si(lOO)lSio."G""" (100
ML)/Si (13.6 nm). The stereogram in the inset indicates the direction of the angular scan for (a) and (b).
272
ton Channeling
r"~~~""-_
.-Goniometer
In addition, the damage introduced by heavier ions is much

greater (Swanson and Maury, 1975; Schir,:,tt and Thomsen"
1972).
The choice of ion energy is determined by considcra.tions of
cross section and mass resolution. The cross sect jon for
backsc~ttering generally decreases with increasing energy,
Scattered
beam
C
i
',"
,
.
"
Cryst I
Solid state
~ detector
He beam
Collimators
accordmg to the Rutherford relationship (Chapters 3 and 4);

however, at high energies (e.g., above 2 MeV (He+ ions on light
elements such as 0, N. and C), the cross section for
backscattering can increase by a factor of up to 10 because of
non-Rutherford behavior. In addition, several re:lonaDces in
(a,a.) reactions occur at higher He energies. Thus, higher
energies can be used to good advantage in order to increase
sensitivity, but care must be taken to avoid unwanted
resonances, for example, when studying Si. Because channeling
studies are dependent on the channel width, "'1f2, it must also be
remembered that the width varies as E 111 , so that detailed angular
scans can be more difficult to perfonn at higher energies,
FIG. ,12.27. Schematic view of the setup for channeling
expenments. The ion beam impinges on a crystal mounted

on a two-axis goniometer, with tilt angle 9 (rotation about
the vertical axis), and azimuthal angle cp (rotation about the
~or the use o~ nuclea~ reactions (see Chapters 6-8), the

chOices of both Ion species and energy are clearly crucial.
Considerations of the nuclear cross sections, depth of analysis,
resonance versus nonresonance reactions, and uniqueness of the
nuclear reaction Q values are dictated by the speCific goal. For
example, the lIB(p,a)8Be reaction has a high cross sec1ion but
provides no depth resolution because the emitted alpha particles
have a variety of energies. Thus, when using that reaction for the
lattice location ofB, the depth ofB must already be known.
crystal normal). The nuclear particle detector measures the

energy of the backscattcrcd ions (from Feldman et 01.,
1982).
<?ommercial computerized goniometers are now readily

available, but because the sample is usually mounted on a
massive plate, in general, they offer less fleXlbility in the sample
geometry (e.g" compensation for sample thickness to retain the
axes of rotation at the same spot on the sample surface); in
sample beating and cooling; and in use of cold shields for
12.8.3 Choice of RBS, X-rays, or nuclear reactions
reduction of contamination and as partial Faraday cups for better; _ For channeling studies of crystal perfection, RBS in the
current integration .. Thus, a self-designed ,manual .... or :.::... Rutherford scattezil:lg t:.eg ime . is the - best choice. For sulface ~-,
computerized instrument might be better (and cheaper) for many
studies involving some elements that are difficult for RBS, a
applications (Gemmell, 1971). When designing a goniometer,
combination of RBS and nuclear reactions is often chosen.
provision should be made for tbennal and spatial separation of a
These ~o methods can be used simultaneously with the same
low~mass target holder, so that it can conveniently be heated and
Ion speCIes and energy, or a combination of RBS with one
cooled and surrounded by a cooling shield. Such a design was
particle and NRA with another can be employed. For example,
used in the manual goniometer of Bottiger et aT. (1973), An
the surface coverage of 0 has been studied using the 160(d,p)170
angular precision ofO.O} Q is desirable for the tilt angle.
reaction,
12.8.2 Cboice of iDn species and energy

For most channeling applications. the use of 4He+ beams (at
1-2 MeV) is suitable because of the moderate mass resolution
beam damage, channel width, and depth resolution. If on~
desires to probe more deeply into the sample. IIr beams are
suitable, but the mass resolution is poor. Thus, W beams are
suitable for probing the distribution of deep damage, such as that
caused by high-energy implantations to -1 J.ffil. As discussed in .
Chapter 4, the use of heavier ions such as C increases the mass
resolution and the sensitivity for detection of small quantities of
heavy elements in a light-element host material. However, the
depth resolution is not improved when surface-barrier Si
detectors are used because of the decrease in detector reso1ution.
273
For lattice location of impurities, RBS is the best choice

when the impurity mass is sufficiently larger than the host atom
mass; that is. the energy separation of the ions backscattered
from impurity atoms and host atoms will give enough depth of
analysis. For example, to measure the lattice position of As in Si
using I Me V "He+ ions at a scattering angle of 1500 , the energies
ofbackscattered He from surface Si and As atoms are ESi = 0.58
MeV and EAs = 0.82 MeV, so that the separation in energy is
240 keY. This corresponds to a depth of 400 run, given that the
energy loss is about 600 eY/nm (Chapter 2). Thus, He' RBS is
suitable for As lattice location up to a depth of 400 nm.
However, in less fa"orable cases, such as Ag impurities in Zr,
the difference in backscattering energies for I MeV He+ is only
Chapter t2
The amount of ion damage is obtained directly from the

fraction of the incident ion energy that goes into nuclear
collisions. This is obtained best by a Monte Carlo simulation
method, such as the widely used TRIM program (Ziegler, 1991).
Approximations such as the Kinchin-Pease method or the
Sigmund approach (Swanson and Maury, 1975) can also be
used.
20 ke V, giving a depth region of only 20 nm. This can be
increased two-fold by wing 2 MeV He+.

For light impurity elements in a beavy-element bost lattice,
accurate measurement ofbackscattering yields from the impurity
element is difficult because the signal is buried in the signal
from the heavier host element In such cases, analysis by nuclear
reactions or characteristic X-rays is used. There are often several

ray yields for a given impurity species. Important considerations
Because the damage is caused by nuclear collisions, it is

caused only by the dec1:Janneled part of ilie beam and is
for lattice location studies are the amounts of damage produced
proportional to X~UVW). This bas been shown to be true
ion energies and species that give suitable nuclear reaction or X-
and the depth of analysis. The damage depends on the reaction
experimentally (Swanson et al., 1979). Thus, if channeling

analyses are done only near good channeling directions, damage
is reduced. In practice, this can reduce the extent of angular
scans for lattice location of impurities. Alternatively, the beam
spot can be moved often, but this can cause problems if the
sample is not unifonn in crystal quality, impurity concentration,
or defect distribution.
cross section, the ion species, and the ion energy. Nuclear
reactions emitting alphas (of a unique energy) give much better

depth resolution than those producing protons. Becawe of the
large penetration of characteristic X-rays, the depth from which

they emanate is difficult to determine. Thus, because lattice
location studies require a comparison of yields from impurity
atoms and host atoms at the same depth, few such studies have
been done using X-rays (Ecker, 1983; Lennard et aI., 1988).
However, there are ways of overcoming this problem, such as
introducing the impurity into a shallow layer near the surface.
Reducing damage by using high-energy or low~mass ions in

ordcr to lower the collision cross section also reduces the yield
ofbackscattered ions, so that tbe benefit is questionable.
In general, for lattice location studies, heavy impurities in a

light matrix are studied by RES, very ligbt impurities (H to S) in
a heavier lattice are studied by NRA, and intennediate-mass
impurities in a heavy host lattice are studied by characteristic xray yields.
12.8.S Sensitivity
For lattice location of impurities, the impurity concentration
must be quite high, typically between 10-3 and 10"'" atomic
fraction.
12.8.4 Ion damage

The damage caused by energetic ions is mainly due to elastic
collisions with target atoms. Because damage from the analyzing
beam can interfere with the results of channeling analysis,
especially for studies of low levels of implantation damage or
lattice location studies of low concentrations of impurities, it is
advisable to minimize that damage by
12.9 OBTAINING CHANNELING SPECTRA

12.9.1 Stereographic projection
To align a crystal for channeling measmements, it is
necessary to understand the essentials of stereographic
projection. Good introductory treatments of the method are
given by Barrett (1952), Cultity (1967), . and Amoros et al.
(1975). Standard stereographic projections of cubic crystals
((100), (110), and (III) normals) and hexagonal crystals are
shown in Appendix 17, Figs. AI7.7-AI7.11. Stereographic
projections of other crystal structures such as tetragonal.
orthorhombic, and monoclinic are given in the book by Smaill
(1972).
increasing the detector solid angle,

analyzing only near channeling directions,
moving the analyzing spot frequently, and

using high-energy and low-mass iom.
The simplest ways to increase detector solid angle are to use
larger detectors and to move the detector closer to the target.
These methods cause some loss of depth and mass resolution,
but there are two methods of reducing that loss. One method is
to use multiple detectors or specially constructed detectors. A
second method is to use coincidence techniques, where the
energies of the scattered ion and the recoiled target atom are
measured in coincidence (Hofsass et ,01., 1991; Chu and Wu,
1988). The coincidence method is suitable only for lowmass
elements and requires a unifonn thin foil target (approximately I
JllIl thick) because the detectors are mounted on opposite sides
of the sample.
.
The stereographic projection provides an accurate

representation of the angles between the plane...:;. and axes of a
crystal. For simplicity, consider first a cubic crystal, for which
each axis [uvw] in the Miner index notation is petpendicular to
the plane (hkl) of the same index.. Consequently, this axis is also
perpendicular to every direction or axis contained in that plane.
(If we denote One of the axes contained in the plane as [hikiIJ.
then the sum of products hiu + ~v + ljw = o. This is just the
criterion that the stun of the products of direction cosines of
perpendicular lines equals zero.) As the simplest example of this
notation, the [100] axis is defined as the direction from the
274
.:,:
Ion Channeling
origin in Cartesian coordinates to a point one positive unit along

the x axis and zero units along the y and z axes. The {I001 axis is
perpendicular to the (100) plane, which is the plane intersecting
the x axis at unit distance and intersecting the y and z axes at
infinity; that is, the (hkl) planar notation corresponds to
intersections of the plane at Ilh, 11k, and 111 points on the x, y,
and z axes, respectively. In Fig. 12.28, the normals to the {100}
planes of a cubic crystal are shown as they intersect a
surrounding sphere at points called poles. The positions of the
poles detennine the orientations of the {IOO} planes. Note that
the normal to a (100) plane intersects the sphere in two
stereographic projection, preserving angular relatioD.ships but

distorting areas. The pole A represents the plane NESW, which
is nOImal to the diameter AB. The great circle NESW P:t"Ojects to
fonn the basic circle N'E'S'W on the proj~tion. The plane
sketched at C bas a pole at P that is projected to the point P'.
Thus, the angle between the planes C and NESW is giv<on by the,
distance between P' and A.
12.9.2 Mount the sample and fInd a suitable channel
The crystal is mounted on a goniometer in an accelerator
target chamber. The experimental equipment required to obtain a
channeled RBS spectrum is the same as that described in
Chapter 4. with the necessary addition of a gOnlometer to align
crystallographic channels with the ion beam. A two-axis
goniometer with two tilts, or preferably one tilt 9 and one
azimuthal rotation~. is suitable, as indicated in Fig. 12.2 7.
diametrically opposed poles, namely, the [TOO] and [1001 poles.

The planes are also represented by great circles on the sphere, as
shown. Note that the (IOO) planar great circle cuts tbe [0 TO],
[OWl, [001], and [00 T] poles. The angle between two planes
equals the angle between their nonnals or between their great
circles. These angles can be measured precisely by the distance
between the poles along a great circle connecting the poles.
100
--
'001
01 0 t----+---i.,~'-+_--_" 010
I
.1
I
-...
".....
"",,"-
As an example of aligning a crystal for channeling, consider a

(110) Si slice. [Note that this is the common notation for a Si
crystal with a {11O) planar surface, or a (110) axial normal.]
The sample is mounted on the goniometer, which is tilted to a
small angle, typically 6, so that the beam is incideut on the
target at approximately 6" from a (110) axial channel (which is
typically 1_2 from the nonnal to the surface); see Appendix 17
for the stereographic projection of a (110)-aligned crystal. A
random RES spectrum is then obtained, as shown in Fig.
12.30(a), top curve, and using a single-channel analyzer, an
energy window is set at just below the maximum energy of the
He ions backscattered from the Si crystal. The RBS Counts in
this window are then collected for a given ion dose as the
azimuthal angle is incremented in steps of 1_2. The RES counts
are recorded as a function of the azimuthal angle 4>, as shown in
Fig. 12.30(b). The angles ~ at wbicb major dip, in yield occur
correspond to alignment along planar channels.
100
The angles at which these planar dips occur are then plotted
on a polar net for the given tilt angle 8, as sholMl in Fig.
l2.30(c). The procedure is repeated for one or more different
tilts (e.g., ~ 5), the angles for the planar dips are plotted on
the polar diagram, and the points are connected by straight lines.
These lines represent {I!o}, {IDO}, and two {Ill} crystal
planes, and they interneet at a (110) crystallographic axis located
near the origin, 8 = 0. The angles 1jJ and 8 at this intersection
point arc the goniometer settings for (110) alignment. It is
sometimes difficult to distinguish between OlO) and (100)
crystal alignments, because in both cases, major planes intersect
at 90. However, a close inspection of Fig. A17.6 (Appendix
17), which shows the major planes: for cubic crystals having
different alignments, is helpful. In the case of a (lIO) surface
normal, the {HO} and {IOO} planes are perpendicular, but the
{Ill} planes are about 35 from the {lID} planes, as is also the
case in Fig. 12.30(c).
FIG. 12.28. {IDO} poles of a cubic crystal (from (ulIity, 1967).

In order to transfer this spherical crystallographic information
to a planar representation, the stereographic projection is used

(Fig. 12.29). A projection plane is placed at the end of any .
diameter of the sphere, and the point of projection is chosen as .
the other end of that diameter (the line AB in Fig. 12.29). If a .
light source is placed at B, the shadow of the pole P is projected
onto the projcction plane at P'. This gives the equiangular
275
/
Chapter 12
planes, whereas the opposite is true for face-centered-cubic

crystals (see Appendix 17, Fig. AI7.6).
ProJection plane
12.9.3 Find the normalized yield

The nonnalized yield X(UYW) is the ratio of the RBS yield for
good (uvw) alignment to the yield. for a randomly-oriented
crystal as a function of the depth x along the beam path. After an
aligned RBS spectrum is recorded for a given integrated
current, the sample is tilted about 5 from the aligned tilt, and
then rotated continuously to obtain a good random spectrum
for the same ion dose. A random spectrum is one for which
the ion beam is directed in a random crystallographic
direction, which does not exist except for an amorphous
material. Because the use of a fixed off-axis alignment
normally gives a yield that deviates by 5-10% from the
random or average yield [Fig. 12.30(b)], it is usual to rotate
the crystal to get an accurate random spectrum. Finally, the
normalized yield Can be obtained by dividing the aligned
yield by the random yield. Normally, the beam intensity (ion
flux) should be reduced for the random spectrum in order to
avoid pulse pile-up or dead time, because the backscattering
yield is much greater for random alignment.
PoInt of
o<OiOd!on
-4--------
seellgn through
JIB and PC
A~C
P'
FIG. 12.29. Illustration of the concept

stereographic projections (from Cullity, 1967).
To find the precise angles
The alignment procedure outlined here can be done

manually or using alignment programs with computer
controlled goniometers. The alignment can also be done by a
different quick manual method by rotating the sample
continuously and observing the planar dips on a ratiometer,
which displays the ratio of the number of counts in the energy
window to the beam current. If the beam current is very
stable, it is sufficient to observe the RES counts in the energy
window. This method is often the best in complicated cases
where the crystal orientation is uncertain because both the tilt
and azimuth angles can be adjusted to follow a given planar
dip. This avoids the problem of the false assignment of
different individual dips in yield with a given plane.
of
e and $ for (110) alignment, an
angular scan similar to Fig. 1230(b) is performed, but over a

smaller angular range of about 2-40 by varying the tilt angle e
in small increments (about 0.05) with cp set at the approximate
value 60 from Fig 12.30(c). The result is a curve similar to that
shown in Fig. 12.30(d). Fine tuning of the aligned value of $
can then be done by setting e at the minimum value of Fig.
12.30(d), 0.90, and performing a similar scan with varying $.
In practice, for crystals already cut along a major

crystallographic plane, such as our (I 10) Si slice, it is
convenient to insert a small wedge below the sample in order to
tilt the crystal by about 2 from perfect alignment; this
12.9.4 Perform the angular scans

An angular SCall through a planar cbannel is best done by
traversing the plane at right angles, and it should be done at an
angle of about S from a major axis in order to avoid high-index
crystallograpbic directions. For example, to scan through the
(011) plane in a (IOO)-oriented Si crystal, the erystal would
simplifies the alignment fme tuning because the alignment is

almost independent of, when the tilt angle is very small.
To fmd a non-normal channel. it is convenient to tilt the
first be aligned with the [Oil] axis, then rotated about S from
crystal along a good planar channel the required amount to
that axis along the (011) plane, and then tilted to scan across
reach the desired channel. The amount of tilt is given by the

known angles between axes in a crystal For cubic crystals,
these angles are listed in Table A17.8 in Appendix 17. For
example, 10 find the (100) axis in our (110) Si crystal, the tilt
from (110) is 45. It is, of course, necessary to tilt along the
correct plane joining these axes, which is a {I oo} plane. This
plane can be chosen by the strength of its dip compared with the
other planar dips. Note that, in diamond-cubic crystals, the
{IOO} plaoes are weaker (give smaller dips) thao the {liD}
the plane, as shown in Fig. 12.31(a). Before doing the scan, it is

best to measure the (011) planar yield for several positions
near 5 from the [011] axis in order to find a flat region in the
planar yield.
Axial scans can be done in several ways: (1) along planar
channels; (2) along arbitrary (nonplanar) paths; and (3) along a
276
Ion Channeling
random path, by rotating about the axis for each angular
increment. Computer programs must be used for the last

approach.
8 10 12 1416 20 2428
42 56
112224
,,
, ,
\
\
\
\
\
\
\
1~L-~~~~~~__L-~__~~__~__
0.2Eo
0.4Ec
O.GEo
Energy (Eo
(oj
=1.0 MeV)
0.8Ec,
1.0Eo
(e)
x 5 I''''
~
..... J
....
' . ,. ...
4
.l!la::
I:
..e~2m.
1
.-
Fixed lilt angle, tJ= 6
>=
~)
g
.. Aligned at 0.9: ~.
{110}
0 40 50 60 70 80 90 100 110120
Rotation angle", (degrees)
(b)
FIG. 12.30. Alignment of a Si (110) crystal for channeling: (a)

RBS spectra for a randomly oriented crystal (top) and an' aligned
crystal (bottom). A narrow energy window is shown below the
surface peak. (b) RBS yield as a function of the azimuthal angle $
for a tilt angle of 6". (c) Polar plot of the planar minima from
panel (b). (d) Precise determination of the value of 9 for (110)
-2
o
Tilt angle 8
(d)
alignment, with set at 60.
-1
277
Chapter 12
For comparison with calculations, methods (1) and (3) are

prefcmblc. To scan through an axis along a plane, it is sufficicnt
'VI = characteristic angle for channeling

'l'ln = half-width of the channeling dip
to note the e and $ settings for the axial alignment and for two
points in the plane on either side of the axis and then to
interpolate the intermediate e and ~ settings for the scan. If the
alignment is slightly off the axial setting, the plane will be
em::ountered on one side of the axis, giving an asymmetric scan
[Fig. 12.3 I (b)]
(ioo]
(100)
(011)
plane
plane
ACKNOWLEDGMENTS
The authors are grateful to l.A. Davies, L.e. Feldman, L.M.
Howe, S.T. Pieraw<, R.P. Sharma, J.C. Soares, and D.O. Boerma
....;. for supplying figores .. D.O. Boerma and P.l.M. Smulders

provided useful comments and offered to supply computer
programs. Special thanks are given to J.H. Barrett for helpful
comments and suggestions for improving the manuscript.
NOTATION
a = Thomas-Fermi screening length
11
llo = Bohr radius = 5.292 x 10- m
d = atomic spacing along axial direction
do = lattice constant
~ = interplanar spacing
e 2 = electronic charge squared = 1.44 x 10-13 cm MeV
E = ion energy
dEJdx = energy loss along the ion beam path
FRS =square root of the Moliere string potential
Fps = square root of the planar potential
H "" number of counts per unit energy loss in an RBS spectrum
Mh M2 = atomic masses of ions and target atoms, respectively
N = atomic density per unit volume
Nd = defect concentration per unit volume
(a)
[100]
(b)
o'
T = crystal temperature
._._~ UI= one
dimensional vibrational amplitude
= two:.dimensional vibrational amplitude (= 1.414u,)

x = distance along ion beam path
U2
Zl, 22 = atomic numbers of ions and target atoms, respectively
X. = normalized yield= (aligned yield)l(random yield)

XII = nonnalized yield from host atoms
Xa = normalized yield for solute atoms
<001>
X~UVW) ,x.~uyw} = nonnalized yiclw for host and solute atoms,
respectively, for alignment along (uvw) axial channels
,,'.
'X.~lOOf ,x.!hkl.1 "'" normalized yields for host and solute atoms,
(e)
respectively, for alignment along {bkl} planar channels

XhD = dechanneled fraction of ions
$ = crystal azimuthal angle

~D = Debye function
A." = dechanneling length caused by dislocations [Eq. (12.33)]
e =crystal tilt angle
FIG. 12.31. (a) 1llustrntion of an angular scan path for a (011)

planar scan, (b) symmetric ax:ial angular scan and asymmetric
scan caused by a slight misalignment, and (c) scan paths for
eo = Debye temperature (K)
symmetric (-) and asymmetric (- - -) axial scans.
= cross section for dechanneling by defects

Otb = cross section for dechanneling by thermal vibrations
ad
278
Ion Channeling
REFERENCES
Uggerh.j, E. (1980), Nucl. lnslrum. Methods 170, lOS.

Channeling at gigaelectroDvolt energies.
GENERAL
Aono, M., Katayama, M., Nomura, E., Chasse, T., Choi, D., and
Kato, M. (1989), Nucl. Inslrum. Methods 37-38, 264. Coaxial
ion scattering spectroscopy (CAlCISS).
THEORY AND COMPUTER
SI~fULA TIONS
Andersen, J.u. (1967), Mal. Fys. Medd. K Dan. Vide",k. Se/sk.

36 (7). Comparison of experimental channeling yields with
simple theory (for H in W; Fig. 12.6).
Andersen, J.U., Augustyniak, W.M., and Uggerlillj, E. (197la),

Phys. Rev. B3, 705. Megaelectronvolt positron channeling.
Andersen, J.U. (1980), Nucl. Instrum. Methods 170, 1, and
following papers. Channeling radiation.
Andersen, J.U., Andreasen, 0., Davies, J.A., and Uggerlillj, E.

(197Ib), Radiat. Eff. 7, 25. First observation offlux peakingfor lattice location ofYb in Si.
BOgb, E. (1968), Can. J. Phys. 46, 653. Study oflattice damage.
Barrett, J.H. (1971), Phys. Rev. BJ, 1527. Monte Carlo computer
Buck, T.M., Wheatley, O.H., and Jackson, D.P. (1983), Nucl.
simulations of channeling.
Instrum. Methods 218, 257. Low-energy ion scattering from

Cu,Au (see also adjacent articles in that journal).
Barrett, J.H. (1990), Nuc!. Instrum. Methods 844, 367. Short

review of computer simulation of channeling.
Chu, W.K., Mayer, J.W., and Nicolet, M.-A. (1978),

Backscattering Spectrometry, Academic Press, New York.
Experimental setup and alignment description.
Bogh, E. (1968), Can. J. Phys. 46, 653. Study of lattice damage.
Bonderup, E., Esbensen, H., Andersen, J.u., and Schiett, HE.

(1972), Radiat. Elf. 12,261. Theory of dechanneling.
Culbertson, R.l .. Feldman, L.C., Silvennan, P.l., and Boehm, H.

(1981), Phys. Rev. Lett. 47, 657. Epitaxial growth [see Fig.
12.4(b)).
Erginsoy, e. (1965), Phys. Rev. Lett. 15,360. Early theory.

Feldman, L.e., and Appleton, B.R. (1969), Appl. Phys. Lett. 15,
305. Double alignment.
Feldman, L.C., Mayer, J.W., and Picraux, S.T. (1982), Motenols

Analysis by ion Channeling, Academic Press, New York. Good
general review.
Firsov, O.B. (1958), Sov. Phys. JETP 6, 534. Screening

functions.
Gemmell, D.S. (1974), Rev. Mod. Phys. 46, 129. Good general,
Fori, G., Grasso, F., Quattracchi, R., and Rirnini, E. (1971),

Phys. Rev. B3. 2169. Dechanne1ing.
review of channeling.
Hofsass, H., and Lindner, G. (1991), Phys. Rep. 201, 121.
Emission channeling ...
Gemmell, D.S. (1974), Rev. Mod. Phys. 46, 129. Good general
_ revie~ C?f ~banneling.
.~
"_ . . __ -
Howe, L.M., Swanson, M.L., and Davies, LA. (1983), Methods .

of Experimental PhySiCS, Academic Press, New York, vol. 21, p.
275. Review of defect and surface studies.
Howe, L.M., Swanson, M.L., and Quenneville, A.F. (1976),

Nuc!. lnstrum. Methods 132, 241. Channeling yields and
dechanneling experiments in AI, Si, Cu, and Ge.
Mayer, J.W., and Rimini, E. (eds.) (1977), Ion Beam Handbook

for Material Analysis, Academic Press, New York. This is the .
old Ion Beam Handbook, which also includes a section on X-
Lehmann, C., and Leibfried, G. (1963), J. Appl. Phys. 34, 2821.

Early theory.
rays ("Catania Handbook").

Morgan, D.V. (ed.) (1973), Channeling; Theory, Observations
and Applications,
Wiley (Interscience), New York.
Lindhard, J. (1965), Mat. Fys. Medd. K. Dan. Vidensk. Se/sk. 34

(14). Comprehensive theory of channeting-the classic in the
field.
Lindhard. J., Nielsen, V., and Scharff, M. (1968), Mar. Fys.
Medd. K. Dan. Yidensk. Selslc. 36 (10).
Schultz, P.J., Logan, L.R., Jackman, T.E., and Davies, J.A.

(1988), Phys. Rev. B38, 6369. Positron channeling at <50 keY.
Matsunami, N., and Itoh, N. (1973), Phys. Lett. A43, 435.

Theory of dechanneling.
Swanson, M.L. (1982), Rep. Prog. Phys. 45, 47. Review,

including channeling studies of lattice defects.
Matsunami, N., Swanson, M.L., and Howe, L.M.,(1978), Can. J.

Phys. 56, 1057. Analytical calculations of equipotential contours
and ion flux distnbutions (Figs. 12.7 and 12.8).
Tang, J.Y., and Zhang, Z.H. (1988), Stopping Power, Range ond
Channeling Effect of Ions in Solids, Atomic Energy, Beijing,
P.R China.
279
/
Chapter 12
Merkle, K.L., Pronko, P.P., Gemmell, D.S., Mikkelson, R.e.,

and Wrobel, I.R. (1973), Phys. Rev. B8, 1002. Simple
dechanneling theory and dechanneling periments.
Davies, lA., Denhartog, 1., Eriksson, L, and Mayer, lW.

(1967), Can. J. Phys. 45, 4053. Early channeling studie, in
Moliere, G. (1947), Z. Naturforsch. Al, 133. Formulation of

interatomic potential.
Davies, lA, and Norton, P.R. (1980), Nucl. Instrum. Methods
implanted Si using a manual two-axis goniometer.

168, 611. Combinatiqn of channeling and nuclear reaction
analysis with other surface science methods.
Morgan, D.V., and Van Vliet, D. (1972), Radiat. Eff 1Z, 203.
Ecker, K.H. (1983), J. Nucl Mater. 119, 301. Lattice location in

metals using characteristic X-rays produced by channeled
protons.
Computer simulations.
Nelson, RS., and Thompson, M.W. (1963), Philos. Mog. 8,
1677. Early theory.
Feldman, L.C., Mayer, J.W., and Picraux, S.T. (1982), Materials

Analysis by Ion Chonneiing,.Academic Press, New York. Good
Nielsen, B.B. (1988), Phys. Rev. B37, 6353. Analytical

calculations of ion flux distributions:
general review and Fig. 12.10.
Piercy, G.R, Brown, F., Davies, I.A., and McCargo, M. (1963),

Phys. Rev. Lell. 10, 399. First experimental evidence for
channeling-from range studies in AI.
Gemmell. D.S. (1971), Nucl. Instrum. Methods 91, 15. Early

computerized goniometer.
GelDIDell, D.S. (1974), Rev. Mod. Phys. 46, 129. Channeled
Robinson, M.T., and Oen, U.S. (1963a),Appl. Phys. Lett. 2, 30.
versus random stopping powers.
Robinson, M.T., and Oen, U.S. (1963b), Phys. Rev. I3Z,2385.
Hofsliss, H.e., Parikh, N.R., Swanson, M.L., and Chu, W,K.

(1991), Nucl. lnstrum. Meihods B58, 49. Coincidence method
First predictions of channeling by computer simulations.
for enhancing the sensitivity for elastic recoil analysis.
Smulders, P.1.M., and Boerma, D.O. (1987), Nucl. Instrum.

Methods 829, 471. Computer simulations of channeling for
Lennard. W.N., Swanson, M.L, Eger, D., Springtborpe, AJ.,

and Shepherd, F.R. (1988), J. Electron. Mater. 17, I. Lattice
precise lattice location of impurities.

Stark, I. (1912), Phys.
channeling.
z.
positions of Zn in InP using characteristic X-ray yields.
13, 973. First prediction of ion
Picraux, S.T., Follstaedt, n.M., Baeri, P., Campisano, S.u., Foti,

G., and Rimini, E. (1980), Radiat. Eff 49, 75. Dechanneling
effect of dislocations, using pIanar channeling.
Tromp, RM., and van der Veen, I.F. (1983), Suif. Sci. 133, 137.
Computer simu1ations for surface studies.
Picraux, ST. (1981), Nud. Instrum. Methods 182-183, 413.
van der Veen, I.F. (1985), Surf Sci. Rep. 5, 199. Computer
simulations and review of surface studies.
Lattice location of gas atoms in metals using axial and planar

angular scans.
Van Vliet, D. (1971), Radial. Eff 10, 137. Analytical

calculations of ion flux distributions and depth oscillations of
Schiett, H.E., and Thomsen, P.V. (1972), Radiat. Eff 14, 39.
ion flux.
Calculation of ion damage.

Smulders, P.I.M., and Boerma, D.O. (1987), Nud. Instrum.
Methods B29, 47 L Computer simulations of channeling for
EXPERIMENTAL METHODS
Barrett, I.H. (1978), Nucl. Instrum. Methods 149,341. Monte
Carlo computer simulations for precise lattice location of
Swanson, M.L., and Maury, F. (1975), Can. J. Phys. 53, 1117.

Calculation of ion damage in Al and channeling studies of
defect-induced solute atom displacements.
impurity atoms.
Bottiger, J., Davies, J.A., Lori, J., and Whitton, J.L. (1973),
Nucl. Instrum. Methods 109, 579. Low-temperature setup for
Swanson, M.L., Offermann, P., and Ecker, K.H. (1979), Can. J.

Phys. 57,457. Reduced damage from channeled ions.
channeling measurements.
Chu, W.K., Mayer, I.W., and Nicolet, M.-A (1978),
Backscatterlng Spectrometry, Academic Press, New York.
Ziegler, I. (1991), Transport of Ions in Maffer. TRIM-9I, IBM

CoIp.) Yorktown, NY. Software for Monte Carlo calculations of
ion ranges and damage.
Cbu, W.K., and Wu, D.T. (1988), Nucl. Instrum. Methods B35,
518. Coincidence methods Qf Clflalysis.
280
Ion Channeling
CRYSTAL ALIGNMENT
Lyrrum, P.F., Thevuthasan, S., and Seiberling, L.E. (1991), J.

Cryst. Growth 113, 45. Transmission channeling studies of Ge
layers grown on Si.
Amoros, J.L., Buerger, M.J., and de Amoros, M.C. (1975), The

Laue Method, Academic Press, New York. Description of Laue
X-ray diffraction patterns.
Picraux, S.T., Dawson, L.R., Osbourn, G.G., and Bieliel<:l, R.M.

(1983), Appl. Phys. Lell. 43,1020. Strain in superlattices.
Barrett, C.S. (1952), Structure of Metals, McGraw-Hill, New

York. Elementary crystallography.
Picraux, S.T., Chu, WK., Allen, W.R., and Ellison, J.A. (1986),
Nuc1. instrum. Methods B15, 306. Strain in superlattices.
Cullity, B.D. (1967), Elements of X-ray Diffraction, AddisonWesley, London. Description of stereographic projection and
crystal orientation using Laue back reflection X-ray patterns.
DAMAGE MEASUREMENTS
Baeri, P., Campisano, S.U, Ciavola, G., and Rimini, E. (1976),
Nucl. lnstrum. Methods 132, 237. Defect displacemeu"tS from
channeling data.
Quenneville, A.F. (1979), Report CRNL 1926, Atomic Energy

of Canada, Mississauga, Canada. X-ray patterns illustrating
mosaic spread and surface defonnation.
Bogb, E. (1968), Can. 1. Phys. 46, 653. Study of lattice damage.
Sm.ill, 1.S. (1972), Metallurgical Stereographic ProjectIOns,

Wiley (Halsted Press), New York. Many stereo graphic
projections of noncubic structures.
Bonderup, E., Esbensen, H .. Andersen, J.u., and Schial!, H.E.

(1972), Radiat. Eff. 12,261. Theory of dechanneling.
Whitton, J.L. (1965), J. Appl. Phys. 36, 3917. Vibratory

polishing methods.
Carstanjen, H.-D. (1980), Phys. Status Solidi A59, II. Review

of lattice location of gas atoms.
Whitton, J.L. (1969), Proc. R. Soc. A311, 63. Polishing

methods.
Carstanjen, H.-D. (1989), Z. Phys. Chern. 165, 141. Effect of

lattice vibrations on channeling yields from deuterium.
EPITAXIAL LAYERS AND STRAIN
Chu, WK, Mayer, J,W., and Nicolet, M.-A. (1978),

Backscattering Spectrometry, Academic Press, New York.
Chiu, K.c.R., Poate, 1.M., Feldman, L.c., and Doherty, C,J.
Csepregi, L., Mayer, J.W., and Sigmon, T.W. (1976), Appl.

Phys. Lett. 29, 92. RBS analysis of epitaxial growth of
amorphized Si.
(1980), Appl. Phys. Lett. 36, 544. Epitaxy of NiSi, on Si.

Culbertson, R.J .. Feldman. L.C., Silvenman, PJ., and Boehm, H.
(1981), Phys. Rev. Lett. 47, 657. Measurement of the initial
registry of epitaxial layers.
Eisen, F.B. (1973), Channeling: Theory, Observations and

Applications. (Morgan, D.V., ed.), Wiley (lnterscience), New
York, p. 415. Dechanneling by buried damaged layers. '.'.' .-
Ellison, J.A., Picraux, S.T., Allen, W.R., and Chu, W.K. (1988),
Phys. Rev. B37, 7290. Also see: Picraux, S.T., Biefeld, RM..
Allen, W.R, Chu, W.K., and Ellison, J.A. (1988), Phys. Rev.
Feldman, L.e., Mayer, 1.W., and Picraux, S.T. (1982), Materials

Analysis by Jon Channeling, Academic Press, New York Good
geneql.i review.
B38, 11086. Allen, W.R., Chu, W.K., Picraux, S.T., Biefeld,

R.M., and Ellison, l.A. (1989), Phys. Rev. 839, 3954. Chu,
W.K., Allen, W.R., Picraux, S.T., and Ellison, J.A. (1990), Phys.
Rev. B42, 5923. Planar cbanneling in superlattices; the second
article gives experimental details of catastrophic dechanneling.
Gemmell, D.S (1974), Rev. Mod. Phys. 46, 129. Good general
review of channeling.
Howe, L.M., Swanson, M.L., and Quenneville, A.F. (1976),
Nucl. Instrum. Methods 132, 241. Channeling yields and
dechanneling experiments in AI, Si, Cu, and Ge.
Feldman, LC., Bevk, J' I Davidson, B.A., Gassmann, H.-J., and

Mannaerts, l.P. (1987), Phys. Rev. Lell. 59, 664. Epitaxy ofGe
on Si.
Howe, L.M., Swanson, M.L., and Davies, lA. (1983), Methods

of Experimental Physics, Academic Press, New York, yol. 21) p.
275. GeDeralreview.
lshiwara, H., and Furukawa, S. (1976), J. Appl. Phys. 47, 1686.

Mosaic spread.
Keil, E., Zeitler, E., and Zinn, W. (1960),

1031. Plural scattering theory.
lshiwara, H, Hikosaka, K." Nagatomo, M., and Furukawa, S.

(1979), Suif Sci. 86, 711. Channeling study of the mosaic
spread of silicides (see Fig. 12.25).
z. Naturforsch. A15,
Matsunarni, N., and Itob, N. (1973), Phys. Lett. A43, 435.

Theory of dechanneling.
281
Chapter 12
Merkle, K.L., Pronko, P.P., Gemmell, D.S., Mikkelson, R.C.,

and Wrobel, JR. (1973), Phys. Rev. BS, 1002. Simple
LATTICE SITES OF SOLUTE ATOMS

Alexander, RB., and Poate, 1.M. (1972), Radiat. Eff. 12, 211.
Angular scans for substitutional impurities: 2% Au in Cu_
Andersen, J.u., Andreasen, 0., Davies, J_A., and Uggerhej, E.
(1971b), Radiat. Eff. 7, 25. First observation of flux peakingfor lattice location of Yb in Si_
decbanneling theol)' and decbanneling experiments.

Mory, J., and Quere, Y. (1972), Radiat. Eff. 13, 57. Theory of
dechanneling by dislocations.
Picraux, S,T., FoUstaedt, n,M., Baeri~ P., Carnpisano, S.U, Foti,

G., and Rimini, E. (1980), Radiat. Eff. 49, 75. Dechanneling
effect of dislocations, using planar channeling.
Andersen, J.U., Cbecbenin, N.G., and Zhang, Z.H. (1982), N.eI.

Instrum. Methods 194, 129. Lattice sites of solute atoms in GaP.
Barrett, JE. (1978), Nucl. Instrum. Methods 149, 341. Monte
Quere, Y. (1974), J. N.el. Mater. 53, 262. Energy dependence

of dechanneling.
''''''QUcie;'' v;' (1978), ~ Radiat.
Carlo computer simulations for precise lattice location of

irnpwity atoms,
Eff. 38, 131. Dechanneling by
dislocations.
Barrett, lR. (1988). Phys. Rev. B38, 5069. Analysis of lattice
Rimini, E. (1975). in Material Characteri2ation Using Ion Beams

(Thomas, J.P., and Cachard, A., eds.), Plenum, New York, p.
Barrett, J.H. (1990), N.eI. Instrum. Methods B44, 367. Short
455. Damage distributions from dechanneLing data.
review of computer simulation of channeling.
vibrations in eu: 2% Au.
Sharma, R.P., Rebo, L.E., Baldo, P.M., and Liu, J.Z. (1988),
Phys. Rev. B38, 9287; (1989), Phys. Rev. Lett. 62, 2869.
Discontinuity of channeling widths in high-Tc superconductors
at the transition temperature.
. Bugeat, J.P., and Ligeon, E. (1979), Phys. Lett. A71, 93. Lattice
sites of H in fcc metals.

including cbanneling studies oflattice defects.
Carstanjen, H.D. (1980), Phys. Status Solidi A59, II. Review of
Swanson, M.L., Howe, L.M., Jackman, T.E., and Moore, I.A.

(1982), N.eI. Instrum. Methods 194, 165. Dechanneling by
irradiation-induced defects.
Applications (Morgan, D.V., ed.), Wiley (Interscience), New
C,rstanjen, H.D., Dunst!, I., lobi, G., and Sizmann, R (1978),

Phys. Status Solidi A4S, 529. Lattice location ofD in Pd
lattice location of gas atoms,

Davies,
(1973), Channeling: Theory, Observations and
York, p_ 391. General lattice location methods,
Shao, L., and Nastasi, M. (2005), Appl. Phys. Lett. 87,064103.
Eriksson, L., and Davies, J.A. (1969), Ark. Fys. 39, 439.
Variation of channel width with the atomic mass of constituent
elements in compounds.
Methods for the accurate analysis of channeling Rutherford

backscattering spectrometry.
_
J.c.
Feldman, L.C., Mayer, J.W., and Picraux, S.T. (1982), Materials
CHANNELING
AND RANDOM
STOPPING POWER
..---_.
--->.
--'~-"-
Analysis by Jon Channeling, Academic Press, New York. Good

general review.
Azevedo, G. de M., Behar, M.., Dias. I.F., Grande, Pol., and da

Silv", D.L. (2002), Phys. Rev. B 65, 075203. Stopping power.
Knapp, J.A., and Follstaed~ DM. (1981), Nucl. Instrum.

Methods 182-183, 1017. Channeling study of solute atom
location in precipitates in steel_
Araujo, L.L., Grande, P.L., Behar. M_. Dias, J.F., Lifschitz, A_F_,
Arista, N.R., and Schiwietz, G. (2004), Phys. Rev. A 70, 032903.
Channeling stopping power_
Gemmell, D.S. (1974), Rev. Mod. Phys. 46, 129. General review
of procedures.
Niemann, D., Konac, G., and Kalbilzer, S. (1996), Nud.

Instrum. Methods Phys. Res. BII8, 11. Stopping power.
Howe, L.M., Swanson, M.L., and Davies, J.A. (1983), Methods
Shao, L.. Wang, Y.Q_, Wetteland, C.J., Nastasi, M., Thompson,

P.E., and Mayer, J.W. (2005,), Appl. Phys. Lett, 86, 221913.
of Experimental Physics, Academic Press, New York, vol. 21, p_
275_ General review.
Optimized energy window of He beams for accurate

determination of depth in channeling Rutherford backscattering
spectrometry,
Matsunami, N., Swanson, M.L., and Howe, L.M. (1978), Can. J.

Phys. 56, 1057. Analytical calculations of equipotential contours
and ion flux distributions.
Shao, L., Wang, Y.Q., Nastasi, M., and Mayer, J.W. (2006),
Nud. Instrum. Methods Phys. Res. B249, 51. Stopping power of
Nielsen, B.B. (1988), Phys. Rev. B37, 6353. Analytical

calculations of ion flux distributions, .
channeled ions.
282
Ion Channeling
Picraux, S.T. (1975), New Uses of [on Accelerators (Ziegler,

J.P., ed.), Plenum, New York, p. 229. General procedures.
Watkins, G.D. (1975), Phys. Rev. B12, 5824. EPR studies of

interstitial B atoms in Si.
Picraux, S.T., Davies, J.A., Eriksson, L, Jobansson, N.G.E., and

Mayer, l.W. (1969), Phys. Rev. 180, 873. Variation of channel
width with constituent elements in compounds.
SURFACE AND INTERFACE STUDIES

Barrett, l.H. (1971), Phys. Rev. B3, 1527. Monte Carlo
computer simulations of channeling.
Picraux, S.T., Brown, W.L., and Gibson, W.M. (1972), Phys.

Rev. B6, 1382. Lattice position ofBi implanted into Si.
80gb; E. (1973), in Channeling: Theory, Observations and

Applications (Morgan, D.V., ed.), Wiley (Interscience), New
York, p. 435. Revi~w.
Picmux, S.T. (1981), Nud. Instrom. Methods 182-183, 413.

Lattice position of trapped D in Fe.
Culbertson, RJ., Feldman, L.C, Silverman, P.J., and Boehm, H.

(1981), Phys. Rev. Lell. 47, 657. Epitaxial growth of Au on Ag.
Rebouta, L, Soares, J.C, daSilva, M.F., Sanz-Garcia, J.A,

Dieguez, E., and Agullo-Lopez, F. (1990), Nud. Enstrom.
Methods 850, 428.
.
Rebouta, L., Soares, I.e., daSilva, M.F., Sanz-Garcia, lA., ..
Dieguez, E., and Aguila-Lopez, F. (1992), J. Mater. Res. 7, 130..
Lattice positions ofTa and Hf in LiNb03 (Fig. 12.36).
Shao, L., Wang, Y.Q., Lee, 1.K., Nastasi, M., Thompson, P.E.,
Theodore, N.D., and Mayer, J.W. (2005b), Appl. Phys. Lell. 87,
131901. Lattice location ofR in Si.
Smulders, P.J.M., and Boenna, D.O. (1987), Nucl. Instrom ..
Methods B29, 471. Computer simulations of cbanneling for
Smulders, PJ.M., Boenna, D.O., Nielsen, B.B., and Swanson,
M.L. (1990), Nue!. Instrum. Methods 845, 438. Lattice location'
of B in Si using analytical and computer simulations of
channehng.
Davies, lA., Jackson, D.P., Matsunami, N., Norton, P.R., and

Andersen, J.U. (1978), Surf Sci. 78, 274. Surface relaxation of
PI.
Davies, J.A.~ Jackson, D.P., Norton, P.R., Posner, D.E., and
Unertl, W.N. (1980), Solid State Commun. 34, 41. Surface
relaxation ofPt
Davies, J.A., and Norton, P.R. (1980), Nuc!. Instrum. Methods
168, 611. Re(.;onstruction of Pt surfaces.
Feldman, Le., Kauffman, R.L., Silverman, P.J., Zuhr, R.A., and
Barrett, l.H. (1977), Phys. Rev. Lell. 39, 38. Effect of thennal
vibrations on channeling surface peak.
Feldman, L.C., Silverman, PJ., Williams, LS., Jackman, T.E.,
and Stensgaard, I. (1978), Phys. Rev. Lett. 41, 1396. Si-Sio,
interfaces in thin Si crystals.
,Feldman, L.C. (1979), ISISS 1979 Surface Science, CRC Press,

Swanson, M.L., Maury, F., and Quenneville, A.F. (1973), Phys.
Rev Leu. 31, 1057. Lattice sites of "mixed dumbbell'~: Cleveland. Effect of surlace impurities on reconstruction.
"
interstitials in AI.
. -.,~" . "-,-- . . , _zoo .. -, . --. ~-. -~,-.,-.oFeldman,
LC., Mayer, lW.;and Picraux, ST. (1982), M~t~rjals
0
"
Swanson, M.L., Howe, L.M., Saris, F.W., and Quenneville, A.F..

(1981), in Defects in Semiconductors (Narayan, 1., and Tan,
T.Y., eds.), North-Holland, Amsterdam, p. 71. Lattice sites ofB
in Si.
including channeling studies of lattice defect::>.
Analysis by Ion Channeling, Academic Press, New York. Good

general re.view.
Fischer,. AE.MJ., Gustafsson, T., and van der Veen, IF.

(1988), Phys. Rev. B37, 6305. SilicideiSi interface bonding.
Frenken, l.W.M., and van der Veen, l.F. (1985), Phys. Rev. Lett.
54,134. Surface melting ofPb.
Swanson, M.L., and Howe, L.M. (1983), Nud. Enstrom.

Methods 218, 613. Tetrahedral interstitial sites of In atoms in Al
caused by trapped vacancies.
Frenken, l.W.M., Maree, P.M.!., and van der Veen, l.F. (1986),
Phys. Rev. B34,7506. Surface melting ofPb.
Swanson, M.L, Parikh, N.R., Sandhu, O.S., Frey, E.C., Zhang,;

Z.H., and Cbu, W.K. (1989), Mater. Res. Soc. Symp. Proc. 144,
409. Using depth oscillations of planar channeling yields to
locate solute atoms in compound semiconductors.
Yagi, E., Kobayashl, T., Nakamura, S., Fukai, Y., and.
Watanabe, K. (1985), Phys. Rev. B31, 1640. Lattice position of
H in V-effect of stress.
283
Howe, L.M., Swanson, M.L., and Davies, lAo (1983), Methods
of Experimental PhySiCS, Academic Press,l\ew York, vol. 21, p.

275. General review.
Jackman, T.E., MacDonald, LR., Feldman, L.C., Silverman,
P.J., and Stensgaard, I. (1980), Surf Sci. 100,35. Study of thin
layers of SiOl. and the Si-Si02 interface.
'. '!.
Chapter 12
Jackson, D.P., and Barren, J.H. (1979), Phys. Lett. A71, 359.
van der Veen, J.F. (1985), Surf. Sci. Rep. 5, 199. Computer
Effect of correlated thennal vibrations on the surface channeling
simulations and review of surface studies.
peak.
van der Veen, J.F., Smeenk, R.G., Tromp, R.M., and Saris, F.W.
(19790), Surf. Sci. 79,212. Relaxation ofNi sunaces.
Swanson, M.L. (1982), Rep. Frog. Phys. 45, 47. Review,

including channeling studies of lattice defects.
van def Veen, J.F., Smeenk, RG., Tromp, R.M, and Saris, F.W.
(I979b), Surf. Sci. 79, 219. Relaxation ofPtsunaces.
and
van der Veen, J.F. (1983), Surf. Sci. 133, 137.

Tromp, R.M.,
Computer simulations for surface studies.
Turkenburg,
van der Veen, J.F., Tromp, R.M., Smeenk, R.G., and Saris, F.W.
(1979c), Surf. Sci. 82, 468. Position of S on Ni sunaces.
w.e., Soszka. W.,
Saris, F.W., Kersten, H.H., and

Colenbrander, B.G. (1976), Nud. Jnstrum. Methods 132, 587.
Double alignment for surface studies.
284
CHAPTER
NUCLEAR MICROPROBE ANALYSIS

M. B. H. Breese
National University of Singapore, Singapore
CONTENTS
13.1 MICROPROBE OPTICS ............................................................................................ 286
13.1.1 Introduction ........................................................................................................ 286
13.1.2 High-demagnification nuclear microprobes ...................................................288
13.1.3 Comparison between megaelectronvoit ions and
kiloelectronvolt electrons ................................................................................... 289
13.1.4 Ion beam damage and ion beam modification ................................................ 289
I
I
I
13.2APPLICATlONS OF NUCLEAR MlCROPROBES ................................................290

13.2.1 High- aud low-beam-current applications .......................................................290
13.2.2 Microprobe applications of high-beam-current techniques .......................... 291
13.2.3 Spatially resolved channeling microscopy .......................................................292
13.2.3.1. Altering the beam4:rystal alignment.. ................................................. 292
13.2.3.2.-- Channeling contrast microscopy ........................:...::..:.:.............. :......293
13.233. Transmission channeling ....................................................................... 293
13.2.4 Scanning transmission ion microscopy (STIM) .............................................. 294
13.2.5 Ion beam induced cbarge (ffiIC) microscopy ................................................. 296
13.2.5.1. Time-resolved ffiIC (TRIBIC) ..............................................................297
13.2.6 Single event upset imaging ................................................................................ 298
13.2.7 Ion beam induced luminescence (mIL) or ionoluminescence ....................... 298
13.2.8 Ion induced electron imaging ............................................................................ 300
13.2.9 Cell irradiation ................................................................................................... 300
133 SUMMARY ..................................................................................................................300
REFEREN CES ..................................................................................................................30 1
I
I
I
285
Chapter 13
13.1 MICROPROBE OPTICS
mlcroprobes use combinations of two, three, or four quadrupole

lenses with alternating polarities and different strengths to focus
megaelectronvolt ions to a small spot. On the other hand,
superconducting solenoid lenses such as the one at Los Alamos
National Laboratory, Los Alamos, NM (Maggiore, 1982), is
simple in design and has no intrinsic aberrations suffered by the
quadrupole systems, but the beam spot size is limited to no less
13.1.1 Introduction
A very brief introduction to microprobe optics is given here
because the subject is well-covered in a variety of other texts.
General accounts of quadrupole optics can be found in many
sources, sucb as Dymnikov and Yavor (1964), Septicr (1967),
and Hawkes (1966), as well as those specific to microprobe
probe-fonning systems such as Maggiore (1982), Grime and
Watt (1984), Wall and Grime (1987), Breese of al. (1996), and
Dymnikov and Osetinskij (1989).
than a few microns.

Figure 13.1 shows a schematic of the important components
of a microprobe. The beam from the accelerator passes through
an object aperture, and the quadrupole lenses, excited with
alternate polarities, fonn a demagnified image of the object
aperture on the sample surface. Larger demagnitieation results in
smaller beam spot sizes if the effects of intrinsic and parasitic
aberrations are negligible. Demagnification is roughly given by
the object distance divided by the image distance. For lowdemagnification microprobes, the image distance is normally
limited to is em by the need to have various detectors and
microscopes viewing the sample surface and also by the lens
currents required to produce suitably strong quadrupole focusing
fields. To achieve reasonably large demagnifications, most
microprobes have lengths of 5-10 m. The sizes of the object and
collimator apertures control the beam current within the focused
spot The collimator aperture defmes the beam angular
divergence passing into the lenses and largely defmes the extent
to whic;:h the beam spot is affected by intrinsic aberrations of the
quadrupoles, such as chromatic and spherical aberrations.
-..=-, .... ,. The' magnetic -rigidity- of charged particles is equal to their

momentwn per unit charge and is a measure of how difficult
they are to bend through a given radius in a magnetic field.
Megaelectronvolt ion beams have a high magnetic rigidity, so
solenoid lenses are usually Dot strong enough to focus them.
Their electric rigidity, equal to their energy per unit charge, is
also large, so neither are cylindrical electrostatic lenses strong
enough to focus them. Quadrupole lenses have a strong focusing
effect because their field lines are perpendicular to the ion
velocity. Magnetic quadrupole lenses are used in preference ~o
electrostatic quadrupoles, which require high field strengths and
small bore radii, resulting in electrical breakdown problems and
parasitic harmonic fields. There are, however, drawbacks to the
use of any type of quadrupole lenses, such as greater complexity
of the focusing system. A single quadrupole lens focuses in one
plane and defocuses in the orthogonal plane, so more than one
quadrupole lens is needed to fonn a point focus. Most
imllge
__________""'\ibiEici _________-'00
..'
...~dlstii!tce:
!><iam from . ~..::::;==-..:.._~

ac\'6leralor
MeV ion
disiiinct!
.-
. t
,sah~iai~s
~"~~~;;~~~~~~~~~~~:::::::llflll
.Qbjed
:slitS
COflitnlllor
-'.IM
1tj1~ge,
~~1~'
FIG. 13.1. Schematic of the major components of a microprobe, in this example using a quadruplet focusing system.
The focused beam spot is scanned over the sample surface

using two sets of magnetic or electrostatic plates located before
or after the quadrupoJes. Different microprobes use different
types and configurations of scan plates, depending on their
primary requirements. Electrostatic scanning can be faster than
magnetic scanning as there is no hysteresis, but the plates need

to be mounted within the vacuum envelope. A scanning system
located before the lenses enables the microprobe to have a
shorter image distance, but such a serup has a limited scanning
range and requires higher currents or voltages owing to the
286
Nuclear Microprobe Analysis
demagnifying effect of the lenses. In all configurations, the beam
scanning coordinates are fed to the data acquisition computer

along with the pulse height of each measured event, enabling
spatially resolved images to be recorded.
produce a point focus. The plots show the different effects on the
position and angle, (x,9) in the horizontal plane and (y,p) in the
vertical plane, of each quadrupole system. The features of each'
focusing system are summarized in Table 13.1, based on a
quadrupole lens length of 10 em and a bore radius of7.5:mID.
Figures \J.2(aHc) show the different focusing actions of a

microprobe that uses two, three, and four quadrupole lenses to
Table 13.1. Features ofthefocosing systems shown in Figs. 13.2(aHd).
(a) Doublet
(b) Triplet
(c) Quadruplet
(d) Highdemagnification
triplet
-10,-62
25
213,378
77,-20
39
504, 525
-14, -14
14
112,274
230,-55
112
1024,1286
130,150
340,960
101, 160
314,830
-85,-99
133, 1663
-64, -61
2000,-7700
Demagnifications
(919), (~I$)
Mean demagnification D
Lem strengths
(ampere turns)
Chromatic aberration
(,.unlauad% )
(x/SO), (ylejli)
Sphericaiaberrations
(I'm/auad')
(x/Sf), (yI9'?)
*Only the most important chromatic and spherical aberration coefficients are included for brevity .
.(y,$)
-'4
-'""
6>20
."
,.W
:Z-'iJ~;"
{m)
.'"
;'.00
,
,
>-
i;g
-~ .!
~.-I,or
:~t
.
~cc
Il~~
0 D0 0
:ti.+CI
11,'50: _ (;.6p
z-"'''1~,'f''1 ..
7.~
FlG. 13.2. Ion optical trajectories of an ion with a divergence of +0.1 mrad and -0.1 mrad. respectively. in the horizontal and vertical
planes of (a) doublet:(b) bigh-c::x(;itation triplet, and (c) "Russian quadruplet" microprobes. The microprobe length of 7.1 m and the
image distance of 15 em are kept constant. In (d), the triplet in the image distance is reduced to 7 em, and the lens lengths are halved
to 5 em. Produced using the matrix code PRAM (Jamieson and Legge, 1987).
287
13.1.2 High-demagnifitation nuclear microprobes

A shorter quadrupole lens has been designed to overcome
A simple lens has a negative demagnification, whereby the

beam out of the lens is deflected in the opposite direction to that
entering it. Because the triplet produces a beam crossover in one
plane, it has ODe positive and one negative demagnification. To
produce a I - m beam
spo~
the object aperture of the doublet
would therefore need to be 77 m in the horizontal plane and 20

m in the vertical plane, if aberrations did not degrade the beam
spot size. Because the effects of chromatic and spherical
aberrations on the focused beam spot size mainly depend on the
beam angle entering the lenses, the coefficients in Table 13.1 are
given as the increase in x or y in the image plane as a function of
the angles
e and ~.
A doublet is the simplest system, but it produces highly

asymmetric demagnifications. A quadruplet produces an
orthomorphic focus (equal demagnificanons in the x aod y
planes) but at the expense of the lowest demagnification. A
triplet has a smaller asymmetry in the dernagnifications than a
doublet. It gives the highest mean demagnification,
D=
,JIce I 8) x (<I> 1<11 '
these limitations, achieving a similar overall focusing strength

but with a lens main body length of only 5 em (Breese et ai.,
1999). This new lens is asymmetric, with a "snout" on one side
so it is capable of achieving the
so that the pole pieces project 15 1DD1 beyond the main lens body
(Fig. 13.3(0)]. This moves the effective exit face of the fin.llens
as close as possible to the wall of the target chamber in order to
reduce the incage distance (Fig. IJ.3(b)]. Figure 1l.2(d) shows
the focusing action of a triplet of these new lenses, with the total
microprobe length kept the same as for the other configurations.
The image distance was reduced from 15 em to 7 cm as a result
of the new lens design and a redesigned target chamber,
resulting in a maximum demagnification of -230 and a mean
demagnification of -11 0 in Table 13.1. The beam converges on
the sample surface more steeply owing to the higher
demagnification, giving a smaller depth of focus over which
regions of the samples with different heights are in focus. A
drawback of suc~. highdemagnification microprobes is that there
. is little space for detectors to be located to view the front surface
of samples, requiring better chamber design.
smallest beam spot size for a given beam brigbtness and current.
This system is called the high-excitation triplet (Grime et al.,
1982; Grime et al., 1991), because there are other possibilities of
fanning a point focus with lower lens currents, although with
lower demagnifications. The lens currents for the triplet are
higher than those for the other configurations, so the lenses tend
to become hotter and can cause thermal drift effects ifheat is not
adequately removed.
In keeping with its higber demagnifications, the chromatic

and spherical aberration coefficients for the triplet are higher
than those for the other two systems. The higher
dem.gnifications of the triplet produce smaller beam spots only
when the effects of its large aberrations do not limit the beain
spot size. Thus, the accelerator needs to be well-stabilized to
produce a beam with a low energy spread so as to minimize the
effects of chromatic aberrations. For the triplet, the beam spot
size increases rapidly with larger beam divergence angles into
the lenses owing to the large spherical aberrations. The
quadruplet is the most "forgiving" of systems, having the lowest
aberrations. Similar conclusions are reacbed for excitation and
rotation and translation aberrations oftbe quadrupoles; the triplet
is generally the most sensitive, so it requires the most careful
setting up in order to take advantage of its higher
demagnification.
FIG. 13.3. Photographs of(.) a new, thin quadrupole lens with a

projecting snout aod (b) the inside ofhigb-demagnification
microprobe chamber, sbowing the short image distance.
In the three low-demagnification systems in Table }3.1, a

maximum demagnificatioD of -80 and a mean demagnification
of -40 are achievable in reality. Higher demagnifications are
Highdernagnification microprobes have larger chromatic and

spherical aberration coefficients. Their ability to focus to a
smaller spot size is diminished if a large beam current is focused,
as this necessitates larger divergence angles into the lenses and
thus a more aberrated focus. This same problem can be
identified with other types of aberrations, such as lens
excitations, rotational and translational misalignments, and so on,
so bigh-demagnification microprobes need to be well-aligned if
higher performance is to result. With careful optimization,
bowever, beam spot sizes of about 35 om have been attained at
low currents, as shown in Fig. 13.4.
limited by the long quadrupole lenses and by the excitation coils

used to induce the magnetic flux. These nonnally protrude
several centimeters beyond the physical lens, which also limits
the use of a shorter image distance.
Several groups have tried 8 variety of other schemes to

increase demagnification. More comptex quadrupole focusing
systems have been developed, such a.t:;; a spaced quadruplet
288
(Shariff et al. 2005), a quintuplet (Ryan and JamiesoD, 1999),

and a 5ix~lens system utilizing two triplets (Dymnikov and Glass,
2004; !ncerti et aI., 2005). Thes< systems do bave large
demagnifications, but they also have large intrinsic: and parasitic
aberrations, and none to date have produced a beam resolution
that matches that produced by the simpler triplets and
quadruplets.
ions require focusing systems that are more complex a.:nd more
expensive and have higher aberrations. ~e payoff for tlJ.is extr~
complexity is that the megaelectronvolt Ion beam tends to stay
focused through many microns of the material. because the
greater ion momentum makes large-angle scattering unlikely. In
comparison, lighter electrons undergo a strong increase in the .
lateral b.eam spread with distance into the materiaL To jllustrate .
this effect, Fig. 13.5 shows Monte Carlo plots of the trajectories
of megae1ectronvoll protons and 50 keV electrons through 10
~ of the polymer poly(methyl methacrylate) (PMMA). The
vertical scales are the same on both plots, but the h.oQnzontal
scales differ by a factor of 100. This means that focused
megaelectronvolt ion beams can be used to attain higher spatial
resolution analysis than kiloelectronvolt electrons in layer
thicknesses of greater than 1 I-UD.
<00
350
joo
25'
"
e
::. .200
o
13.1.4 Ion beam damage and ion beam modification

An important consideration for all microprobe usc is beam
damage and heating effects at the high current densities required
100
for microanalytical work. Even a low-dernagnification
so
microprobe is capable of focusing 1 nA of proton beam current
into a spot size of 1 Junl (Watt et al., 1995), equivalent to more
than 10 17 ions/(cm2/s). Such high current densities can easily
1100
1400
1500
eoo
900
1000
1200
1300
cause sputtering, damage, and heating effects in sensitive
X coordinate (om)
biological tissues, polymers, and semiconductors that are
FIG. 13.4. Line scan of highly focused rnegaelectronvolt proton
exacerbated with the higher current densities attainable with
beam over a sharp edge, showing resolution of35 run. Reprinted
high-demagnification microprobes. However, in ooe study, by
from Watt et al. (2003).
improvements of the X-ray detection efficiency and a sample
cooling system, no elemental loss was observed during the
A major factor that determines whether a higher
irradiation at a proton beam current of 100 pA (Sakai et at.,
demagnification will result in a significant reduction in the beam
2005). This limitation is another reason why the usc of
spot size is the beam energy spread, which is largely determined
microprobes for high-resolution microanalytical work using
by the stabilization system utilized by the accelerator. The ~:: particle-induced. X-ray emission (PIXE) and Rutherford
energy Spread limits the attainable beam spot size because of the
backscattering spectrometry (RES) has been little developed so
chromatic aberrations of the quadrupole lenses, with ions of
far. This has not posed a major limitation to their use to date, as
slightly different energies being focused to different distances
most
microanalytical work with high-demagnification
along the beam axis. The energy spread also results in a
microprobes has concentrated on low-beam-current techniques
fluctuating beam current, which makes it difficult to scan the
such as scanning transmission ion microscopy (STIM),
focused beam uniformly over an area. A new generation of
transmission channeling, ion beam induced charge (ISlC), ion
megavolt accelerators (Visser et al., 2005), achieve very small
beam induced luminescence (mIL). and single-event upset (SEll)
tenninal voltage fluctuations of lens of volts. which is at least 10
imaging where beam damage is not such a major issue.
times better than that of Van de Gra.aff accelerators. The
resultant beam current hitting the sample in a microprobe is. very
This same feature makes the use of high-demagnification
stable, with fluctuations as low as 2% reported. when the
microprobes for focused-beam irradiation of semiconductors and
accelerator is properly optimized (Teo et al., 2004). Such singlepolymers very attractive, as the higher current density produces
ended accelerators also produce high-brightness beams from
the desired consequence of more sample modification in a
radio-frequency ion sources, making them ideal for highshorter period. This type of application is becoming increasingly
resolution microprobe use.
.
widespread for microprobes. Lithography applications in
polymers such as PMMA and SU-8 make use of the ability of
13.1.3 Comparison between megaelectronvolt ions and
high-energy ions to create polymer chain scission to fabricate
kiloelectronvolt electrons
high-aspect-ratio
microstructures.
Many
semiconductor
A scanning electron microscope can easily focus
applications are based OD using focused megaelectronvolt ions to
kiloelectronvolt electrons with a simple solenoid lenses owing to
selectively damage areas that are electrochemically anOdized at a
their low magnetic rigidity. In comparison, megaelectronvolt
different rate, producing patterned silicon micromachined
"
15'
Sigma:i6.5i.4
FWHM=34.S%3,3
289
'., ~
Chapter 13
surfaces, controlled photoluminescence porous silicon, and

Bragg reflectors. Focused proton irradiation was recently used to
create ferromagnetic microstructures in graphite (Esquinazi el 01.,
2003).
13.2 APPLICATIONS OF NUCLEAR MlCROPROBES
13.2.1 High- and low-beam-current applications
Microprobe analytical applications can be divided into two
regimes. In the high-current regime of typically 10 pA to 10 nA,
micrometers, although a beam spot size of 1 J.llD containing I nA
of current is possible. The data acquisition rate is DonnaUy low,

typically a maximum of a few kilohertz,' and measurement
periods are long in order to collect sufficient statistics. To
increase the counting statistics, detectors in microprobe
chambers are usually larger than those used for conventional
analysis, such as 80 mm' Si(Li) detectors for PlXE analysis and
300 mm' area surface barrier detectors for RBS analysis.
Associated drawbacks of lower energy resolutions owing to
larger kinematic spreads and increased detector noise levels are
an inevitable compromise for microprobe analysis.
the focused beam is used for spatially resolved analysis using

conventional iOD beam analysis (IBA) techniques. The beam
spot size is detennined by the b~ curr~.t .~d,.i~-,-~~g~~~,!:~.. ",.,._.,.. ,.
l (3)
6 .....
"'j"
8i .. _ .. _' .. .H
"Literal
.'. , b);pli1ceiije[ll(nrn)
'., ., ..... ",... .
'
FIG. 13.5. Monte Carlo plots of the trajectories of (a) 2 MeV protons (and also the subsequent generation of secondary electrons) and
(b) 50 keY electrons through 10 ~m of PMMA. Reprinted from UdaI.gam. (2006).
The low-current regime uscs a beam current of typically 1 fA

to 1 pA for analysis using techniques such IBIC, STIM, IBIL,
SEU imaging, transmission channeling, and ion-induced electron
imaging, as sbovvn in Fig. 13.6. At such low beam currents, there
is less damage and sputtering and fewer heating problems, and
spatial resolutions of 100 run are routinely achieved. Analysis at
such low beam currents is possible because cach incident ion can
produce a measurable event, in comparison to the small X-ray
production and ion scattering cross sections for FIXE and RBS
analysis. For example, in IBIC analysis, each incident ion
generates sufficient electroo-bole pairs to produce a measurable
charge pulse, and in STIM, the energy of each ion transmitted
through a thin sample is measured.
The low-current techniques tend to suffer from the opposite

problem of the bigh-current techniques in producing too many
measurab1e signals per second, so the beam current has to be
very low to reduce the measured count rate to several kilohertz.
Measurements in this regime are therefore limited by the data
acquisition system, and there are severa1 efforts to build faster
systems capable of counting at much higher rates.
These low-current techniques tend to have most relevance for
microscopy applications with a scanned focused beam, rather
than spectrometry applications with an unfocused beam, that is.
they are mainly used for producing images rather than spectra.
For example, a transmitted energy spectrum in STIM or
290
r
I
I
I
I
low sensitivity being due to the large electron brehmstr ahhlung

background PIXE is the most widespread of microprobe
microanalytical techniques. with many imaging ancl point~
analysis applications in trace-element analysis in 'hiology,
biomedicine, geology, environmental science, and archaeology.
The theory and applications using focused and unfocused beams
are given Johansson and Campbell (1988) and Johansson et al.
(1993). One noteworthy development of PIXE for microprobe
use has been the use of dynamic PIXE analysis (Ryan and
Jamieson, 1993; Ryan et al., 1995). which eliminates problems
due to overlapping X-ray lines. This couLd otherwise produce
misleading artifacts in PIXE images collected in the normal
manner by counting the number of measured X-rays -within a
defmed energy window around a given peak. Another
development was to set up external rnicrobeams, wbich has
many advantages, including reducing damage by heating and
charging and allowing in vivo sample analysis, although at a
lower spatial resolution owing to a large image distance and
scattering of the beam outside the chamber vacuum (Sakai et ai.,
1999).
transmission channeling is difficult to quantitatively interpret in

terms of sample thickness or density and is mainly useful in
producing energy loss images from different "slices" of the
transmitted energy spectrum. The same applies to electron
imaging, where the measUred ion-induced electron flux from a
fixed location bas limited interest but provides a very useful
signal for spatially resolved imaging. This distinction is less
clear for !BIC, !BIL, and SEU analyses, where the spectra can
be quantitatively interpreted and can be as important as the
corresponding microscopy images. For example, the intensity
and wavelengths recorded in an IBIL spectrum can be
interpreted in terms of the elements and defect states present.
Focused Mev icinbeam

1
ion induced
luminescence ~
IBIL
Ion induced
l,S
$ ; Electron-hole pair
-=.:!
creation - ISIC
PIXE microanalysis and imaging of biological tissues is one

of the most widespread uses of microprobes (Watt and Grime,
1987; Llabador and Moretto, 1998). One important aspect of
such studies is that they rely on having some means of
recognizing the tissue sample area under analysis. There are
various ways of doing this using optical microscopy or using

STIM as described in Section 13.2.4. The example in Fig. 13.7
shows microprobe images of cells, combining STIM images for
ions transmitted through thin samplefcrystal
location and PIXE images of the potassium and cesium
distributions (Ortega, 2005). By further processing of these two
- BTIMI transmission channeling
PIXE images it can be deduced that the potassium and cesium
FIG. 13.6. Schematic of the various low-current techniques used
are colocalized, with both elements showing the same
on a nuclear,microprobe.
_~.' ~.--oc-=-"':"'_+ ___ intracellular distributions. - .-. '":.~:-:: .-:-..=-._ :':...-~.~.~.::---=< ::.~ ~..:....~
",-",-=.,.-",
..
13.2.2 Microprobe applications of high-beam-current
techniques
Most applications of these techniques use unfocused
megaelectronvolt ion beams, as described in other chapters of
this book. PlXE bas relatively large production cross sections
compared with R;BS and nuclear reaction analysis (NRA), so it is
quite easy to generate spatially resolved images of elemental
ctistributions at beam currents as low as 10 pA with a microprobe.
Most microprobe applications of RBS, NRA, and elastic recoil
detection analysis (ERDA) rely on recording spectra from
spatially resolved points across a surface rather than producing
images that require larger beam currents. Many microprobe
applicatioIlli of these techniques are given in Grime and Watt
(1984), Watt and Grime ([987), and Breese et al. (1996).
PIXE has several advantages compared with other techniques

for measuring trace-element distributions. It is capable of
quantitatively imaging elemental distributions to a sensitivity of
parts per million, whereas EFMA (electron probe microanalysis)
typically detects concentrations of parts per thousand, with this
291
RBS spectra have been collected using a focused

megaelectronvolt ion beam from micron-size areas and depthresolved images formed of integrated circuits, TiC-coated
graphite limiter tiles (Doyle and Wing, 1983), and geological
samples (JaIDieson and Ryan, 1993). For biomedical applications,
RBS images of light elements present in thin tissue sections can
be recorded along with distributions of heavier elements using
PIXE (Landsberg et al., 1992), as there is no overlap in
backscattering signals for carbon and nitrogen, for example. An
example of RES microanalysis is shown in Fig. 13.8. Here, RBS
and PIXE were used to study the oxidation rate of silicon carbide
composite samples that were exposed to lithium titanate as part
of a study on materials for fusion reactors (Alves et aI., 2001).
The material was analyzed with a focused proton beam at
energies between 1.7 and 2.0 MeV to measure backscattering
spectra below, at, and above a carbon resonant scattering peak at
1.75 Me V. RES spectra were collected from two differen1 points
a few microns in diameter that were identified using the PLXE
images where titanium precipitation was observed. The RBS
spectra collected from different areas exhibit widely differing
. --
Chapter 13
thrcugh thin crystals to image di,locatio", and lattice defects
behaviors that cannot be readily observed using analysis with an

unfocused ion beam, thus enabling variations in the local rate of
diffusion and oxidation to be studied.
using transmission channeling.
SiC 3D Li2 Ti03 lOOOh-J.75 MeV
I~
FIG. 13.7. (Top) scanning transmission ion microscopy (STIM)

and (bottom) particle-induced X-ray emtSSlon (PIXE)
microprobe images of Arabidosis thaliana cells. Scale bar is 20
!J.IIl. The lower right image shows the measured CsIK ratio.
Reprinted frDm Ortega (2005).
2O!l
NRA to provide spectra of light elements in metallurgical

samples from resolved locations with beam currents of tens of
:C ii:o
nanoamperes or more (e.g., McMillan, 1988; Pierce el 0/., 1974).
100
. '50
2DO
250
300
350
40[,
Choiulnel
FIG. B.S. (Top) 500 x 500 m'p3rtic1e-induced X-ray emission
(PIX) images showing the Si and Ti distributions in a silicon
carbide composite sample exposed to LhTi03 for 1000 h.
(Bottom) Rutherford backscattering spectrometry (RBS) spectra
measured from the two points identified in the FIXE images.
Reprinted from Alves el 0/. (2001).
The increase in resonant scattering cross sections for light

elements makes imaging applications feasible with a beam
current of a few nanoamperes (e.g., Llabador el 0/., 1993; Wang

et al., 1993). Another imaging example showed the deuterium
concentration on the surface of a limiter tile from a fusion
reactor (Doyle ef 0/., 1991).

Because of the low cross sections for ERDA, large beam
currents of tens of nanoampercs arc required for analysis. This,
glanciog~angJe
I" ~Or
400
There have been many applications of microprobe use in
together with the
-'-=; .
13.2.3.1 Altering the beam-uystnl alignment

A microprobe provides several alternatives for varying the
beam-crystal alignment for channeling analysis other than use of
a goniometer. Indee~ there is little space in the chamber of a
high-demagnification microprobe to install a goniometer, and
other approaches have to be used. A beam-rocking system
provides the ability to tilt a focused beam spot through angles of
several degrees, with the beam remairllng stable on the sample:
surface to within 2 JllD (de Kerckhove el aI., 1997, 1998).
Another solution is to shift the coUimator aperture of the
microprobe away from the beam axis, resulting in the beam
being tilted with respect to the crystal surface (Breese et 01.,
2004). A collimator displacement 0(50 IJlIl typically produces a
change of 0.10 in the beam tilt with respect to the sample
surface, depending on the dcmagnification. This is comparable
geometry, makes a spatial
resolution of a few microns the best that can .be achieved with
microprobe analysis. Typical microprobe applications include its
use for point analysis in measuring the distribution of hydrogen
across catalyst pellets (Sofield e/ or., 1981) and in melt
inclusions (Mosbah eJ 01., 1991).
13.2.3 Spatially resolved channeling microscopy

The ability to focus megaelectronvolt ion beams to small
spots enables several modes of spatially resolved ion channeling
analysis. These range from the use of channeling contrast
microscopy (CeM) to record backscattering spectra in
channeling alignment from micron-size areas, in order to obtain
images of channeled yield variations across the sample surface,
to measuring the energies and intensities of ions transmitted
292
.1
Figure 13.9 shows transmission channeling images of a single

stacking fault in silicon at different tilts with respect to the
vertically running (Oll) planes. Dark and bright regions
represent regions of high and low energy loss, respectively. The
fault appears dark in Fig. 13.9(a) because many protons are
dechanneleti OIL passing tluough the fault plane. The additional
dechanneling effect of the partial dislocation that bounds the
right-hand fault edge is visible as a darker band in the lower
right of the image. In Fig. J3.9(b), the left-most portion of the
fault that is closest to the swiace appears as a bright band
because of the blocking-to-channeling transition (King et aI.,
1995). Figure 13.9(c) shows a smaller region of the central fault
portion at a tilt of 0.10, in which periodic intensity oscillations
are observed as bright and dark bands.
the channeling critical angles of 0.10--0.400 for 2 MeV

protons or helium ions. With this procedure, no loss of spatial
resolution has been observed in a beam spot size of 60 nm for a
beam tilted by up to 0.5'.
10
13.2.3.2 Channeling contrast microscopy

1n CCM, the backscattered ion, or PIXE yield, is measured as
a function of beam position within an area, and an image is built
up showing variations in crystalline qUality. The flrst use of
CeM was to image a region of an ion-implanted silicon crystal
that had been laser-annealed (McCallum ct al., 1983). More
recently, it has been used to produce depth-resolved images of
lattice plane rotation in partially relaxed silicon-germanium
layers grown on a buffered silicon substrate (Seng et 01., 2005).
Backscattering analysis from smail areas requires a beam current
of more than I nA, and care is needed to avoid beam damage or
ion-induced swelling of the crystal lattice, which would affect
the results. Because a high beam current is required, the
collimator and/or object apertures need to be open fairly wide.
However, opening the collimator aperture too far means that the
beam divergence angle on the crystal. -surface could be
comparable to the channeling critical angles, so it is preferable to
open the object aperture, even though this limits the beam spot
size for CeM analysis to a few micrometers. For the same
reason, it is easier to perform CCM with helium ions rather than
protons because of the greater scattering cross sections and
larger critical angles of the fonner.
13.23.3 Transmission channeling
Focused megaelectronvolt proton beams can be used to
produce spatially ,resolved transmission channeling images of
defects in thinned crystals. The energy spectrum of the
transmitted protons is recorded with a semiconductor detector on ~.
the beam axis while the focused beam scans over the crystal .
surface (Cholewa et al., 1990). Any lattice disruption causes
additional dechanneling, and transmission channeling images
showing variations in transmitted energy reveal the distribution
of defects such as dislocations, stacking faults, precipitates, areas
of elastic strain, and exfoliated regions.
A transmission channeling setup can also be implemented on
a high-demagnification microprobe (Breese et al., 2004). This
gives a focused beam convergence angle of only -0.01 on the
crystal surface at very low beam cunents of between 1 fA and 1
pA, which is much less than the planar channeling critical owing
to the very low beam current. Even at spot sizes of 60 nm, beamdamage effects can be ignored owing to the very low proton
beam current used for analysis.
293
FIG. 13.9. Transmission channeling images of a stacking fault at

tilts of (a) 0.00', (b) 0.22, (e) 0.10'. In (a) and (b), the
horizontal x vertical scan area is 4.4 x 22 J.UD2, and (c) was
recorded from a 3.3 x 3.3 J.UIl"l area close to the central fault
portion. Reprinted from Breese et al. (2004).
High.:.contrast transmission channeling images and line scans
of isolated bunches and individual 60 misfit dislocations in
thic~ partially relaxed silicon-gennanium layers are shown in
Fig. 13.10. The Nomarski image in Fig. 13.IO(a) sbows a crossann running along the (110) and
(1 TO) planes, and the
transmission channeling images were recorded from within the

boxed area. The wide, horizontally running band produced by
the bund of five dislocations lies along the (110) planes, with
two etch pits visible. A change in contrast is observed at the
bunch of five dislocations in Figs. 13.10(b)-{d). All components
of the Burgers vector showing the magnitude and direction of
the 60 dislocations were considered in interpreting these results,
with the lattice plane rotation calculated as a function of lateral
position away from the dislocation core and depth below the
epilayer surface due to the combined effects of the two edge
components, based on Mazzer et al. (1997).
Chapter 13
FIG. 13.10. (a) Low-magnification No~ki image of partially relaxed silicon--germani~ surface. (b}-{d) Transmission channeling
images of the 11 x 11 "",' box area shown in (a). The tilt angle with respect to the (110) planes is (b) +0.16' , (c) 0.00' , (d) ...;}.16'.
Reprinted from Breese et al. (2005).
As an exnmple, Fig. 13.11(a) shows a typical schematic of the
STIM setup (Minqin el a/., 2007). with cell samples mounted on
a thin silicon nitride window. Behind this is a charged particle
detector, which can be either a silicon surface bamer detector or
a PW diode (i.e., a diode with a nearMintrinsic semiconductor
region between regions of n- and p-type semiconductors). PIN
13.2.4 Scanning transmission iDn microscopy (STIM)

Scanning transmission ion microscopy (STIM) , is a simple
microprobe technique that relies on measuring the energy of ions

that are transmitted through thin snmples (Overley et 01., 1988).
It is used to image variations in sample areal density, whether
these variations arise from changes in stoichiometty or changes
in thickness. It is very useful for sample imaging and location of
particular features for microanalysis with other ion beam
teChniques such as PIXE. One of the most widespread uses of
STIM imaging is for biomedical analysis in conjunction with
PIXE Or RBS. Because almost every incident ion produces a
measurable event for STIM, the beam current in the focused spot
is usually only a few thousand ions per second, and the spot size
is consequently typically less than 100 nm.
diodes have the advantage of being considerably cheaper than

surface barrier detectors, so it does not unduly matter if they are
damaged by the ion hewn during measurements. Figure 13.l1(b)
shows an optical micrograph of cells under analysis, located on

top of a thin 500 x 500 1-UD2 silicon nitride window. The resultant
STIM energy spectrum produced by !he scanned beam of 1 MeV
helium ions is shown io Fig. l3.11(c).
294
.,
Nuclear Microprobe A:nalysis
ft?cl!seqion,
be"m'
~arn;!.rnatsu Pin
diode 51223
100000
(c)
19000
II
<
1000
0
0
.r:
0.
100
<e
lQ
1
500
650
950
1100
En~rgy (keV)
FIG. 13.11. (a) Scanning transmission ion microscopy (STIM) setup for imaging cells. (b) Optical micrograph of cells to be analyzed.
(c) STllvl ene"rgy spectrum obtained using a focused I MeV helium ion beam s~anning over the area shown in (b). (d) STIM. image
from the 50 x 50 ~2 box area sbown in (b). Reprinted from Minqin el al. (2007).
Different approaches can be used to produce STIM images

from the measured ion energies at each pixel. The simplest is to
divide the energy spectrum into different windows and display
the number of cOunts in each window as a separate map. This
approach is able to produce online images as the data are being
collected, which is very usefuL Another approach is to combine
all of the data together to produce a single image showing the
median ion energy loss at each location within the scanned area.
This approach typically produces better images with lower noise
but can be done only after data collection bas finished. Figure
13.1I(d) sbows a median-energy-Ioss STIM image from the
boxed area in Fig. l3.11(b), in which dark and bright regioos
correspond to low and high energy losses, respectively,
rate of energy loss, whereas helium ions tend to provide greater

energy loss contrast between two areas of slightly differing areal
density owing to their higher rate of energy loss. The use of
STIM in channeling aligoment to produce transmission
channeling images of crystal defects is described in Section
13.2.3.3.
STIM tomography is an ex.teDsion of conventional STIM
measwements in which images are recorded over a wide range
of rotational aligoments of the sample (pontau et al., 1989;
Fischer and Muhlbauer, 1990). The entire data set is then
processed to produce three-dimensional renditions of the internal
sample structure with bigh contrast This technique has been
used for reconstructing three-dimensional images of silicon
pillars, fusion pellets, and biological ceUs, among other samples
(Michelet-Habcbi el al., 2005). A rew problems must be
The most commonly used ions for STLL\1 imaging are

megaelectronvolt protons and helium ions. Protons provide an
imaging capability through a greater depth owing to the lower
overcome for successful reconstruction, such as very careful
295
Chapter 13
energy ion produces charge carriers along its trajectory. If

carriers are produced within a depletion region of a device
junction, they arc rapidly separated and can be collected with
high efficiency. The carriers that are generated in field-free
regions, either beneath or to the side of device junctions, slowly
diffuse away from their point of generation. Many are trapped
and recombine at point defects created by the wafer doping. The
rate at which this occurs is given by the minority-carrier
diffusion length. A reduced number of carriers arrive at a device
junction, where they are collected along with the drift charge. In
Fig. 13.l2(b), two different irradiation directions are shown,
either from the top or from the side, where the latter is called
laterallBlC (Vitlone e/ al., 1999; Lu e/ a/., 2002). A depletion
layer is produced at thc surface by deposition of a metal layer
that forms a Schottky barrier, or an external bias voltage can be
applied When the beam scans over a grain boundary, this acts
an efficient center for carrier recombination, so few carriers
diffuse back to the collecting junction at the surface. In lateral
mIC analysis of a cleaved sample, the beam can induce carriers
at different depths beneath the surface, enabling more in-depth ----'
studies of the motion of induced charges to either collecting
surface.
mounting and compensation for any sample movement during

rotation. The large number of recorded images also increases the
probability of beam-induced damage occwring, particularly in
biological samples.
13.2.5 Ion beam induced charge (IBIC) mic["oscopy
IBIC microscopy uses a focused megaelectronvolt light-ion

beam to create electron-hole pairs that are separated in the
presence of an external electric field, or the electric field of a po junction, and are detected. The measured charge depends on
the amount of trapping and recombination at point and extended
defects, the doping density, the minority-carrier diffusion length,
_. __ and the electric field strengths.
Figure 13.12 shows the basic analytical geometries used for
lBIC analysis of devices and planar semiconductor or insulator
wafers that might cont.ai.D. grain boundaries, dislocations, or other
types oflattice defects. IBIC microscopy uses a current of 1-10
iA of megaelectronvolt ions and measures individual cbarge
pulses. Each ion normally generates enough electron-bole pairs
in semiconductor and insulating materials to produce a charge
pulse that is measurable above the device noise level. The highMeV ions
P~amptifierJ
amptifieriDAQ
(a)
MeVions;
Side
. - - - - - - . - - - - - - - irradiation
with
Pr&'amprifieri
ampJifierlDAQ
MeVioris
(b)
FIG. 13.12. Geometries used for ion beam induced cbarge (IBIC) analysis of (a) devices and (b) planar semiconductor or insulator
wafers, containing grain boundaries, dislocations or other types of lattice defects.
A one-dimensional model of the measured charge pulse

height was initially developed (Breese, 1993) to enable the
quantitative analysis of the charge pulse height. The analytical
theory of IBIC generation and (;Ollection has since been
comprehensively developed [Vittone e/ al., 2000; Vitton" 2004).
One model was based on an extended Shockley-Ramo theorem,
and a more recent one was based on Gunn's theorem, which
removes problems due to space charge uncertainties within the
induced column of charge carriers. Another approach to studying
the motion and time evolution of ion-induced charges in
296
semiconductor devices is to use computer codes such as

MEDICI or DAVINCI. These codes were mainly written for
devi~e modeling applications and solve the Laplace equation in
two or tluee dimensions, thus enabling the study of charge
transport in arbitrary device geometries.
IBIC microscopy has several features that differentiate it
from other forms of charge collection microscopy. First, it can
provide high-resolution analysis through thick layers, whereas
kiloelectronvolt electrons used for electron beam induced current
(EBIC) microscopy (Leamy, 1982) are strongly scattered, as
shown in Fig. 13.5. Second, the trajectories of megaelectronvolt

ions are not differently influenced by the passage through metal
layers compared to semiconductor or insulator layers, so mIC
can perform quantitative analysis underneath metal layers. Third,
there is little effect due to charging by the incident beam current.
Another feature of IBIC is sample damage. Megaelectronvolt
ions have enough momentum to displace atoms within
crystalline materials away from their lattice sites to interstitial
locations during a collision, creating vacancy-hole pairs (i.e.,
Frenkel defects). Beam damage acts as trapping and
recombination sites for subsequent ion-induced carriers,
complicating quantitative analysis by reducing the charge pulse
heights.
IBIC microscopy was originally developed as a means of
imaging device depletion regions (Breese et 01., 1992) and
dislocations (Breese at al., 1993) and as a means of identifying
regions of devices that are sensitive to single-event upsets
(Sexton ef al., 1993). It has since found widespread applications
in many microprobe laboratories, and its use has greatly
ex:panded to cover the analysis of many types of semiconductor
radiation detectors, high~power transistors (Osipowicz et 01.,
2001), charge-coupled arrays, solar cells, and light-emitting
diodes (LEOs).
increasing magnification of a region at the tip of a singIe 100).lIIl-wide electrode. The images were acquired for 20 min. each,
using a focused 2 MeV proton beam and a diamond bias ,",oltage
of 200 Von alternate electrodes. In Fig. 13.13(0), the d.etector
active area is in the lower portion. The vertically running dark
areas correspond to the locations of the metal electrodes;=- where
the collection efficiency is lowest. The vertically running white
regions correspond to the regions where the largest charge pulses
are measured from the exposed diamond surface between the
interdigitated electrodes, where an electric field gradient exists.
The highest regions of charge collection are from the tips of the
electrodes, consistent with this being the location of the highest
electric field gradient. In the higher-magnification images [Figs.
13.13(b)-{d)l, individual grains are resolved, where their widely
differing collection efficiency can be observed.
The most widespread application of IBIC has been for the

analysis of semiconductor and insulator materials for radiation
detectors. Manfredotti and co~workers used IBIC extensively,
combined with IBIL, to study polycrystalline chemical-vapor~
deposited (CVD) and natural diamond (Manfredotti el al., 1995),
including the electrical properties, radiation~induced effects, and
the homogeneity, for detector tracking applications (Manfredotti
ef al., 2002). Tbey also used IBIC to study physical and
chemical inhomogeneities in semiconductor and insulating
materials in combination with PIXE, radiation detectors (Jaksic ~-
er al., 1999), and silicoo carbide (Jaksic el al., 2002). Sellin and
co-workers used IBIC to analyze polycrystalline CVD diamond,
including imaging charge-transport properties (Sellin et ai.,
2000), in combination with IBIL microscopy, imaging
FIG. 13.13. Ion beam induced charge (IBIC) analysis of a
intracrystallite charge transport at reduced temperatures, and
chemical-vapor-deposited (CVD) diamond detector with
studying coplanar CVD diamond radiation detectors (Galbiati et
increasing magnification. The size of the scanned areas is
al., 200 I) and their spectroscopic response. Other materials such
indicated in each case. Light and dark areas correspond to low
as CdZnTe radiation detectors (Rath et al., 2003; Vizkelethy et
and high levels of charge collection, respectively. Reprinted
aI., 1999), epitaxial GaN detectors, and CdTe bave also been
from Sellin el al. (2000).
studied The microprobe group in Melbourne has used mIC in
many applications, such as imaging the charge collection and
temperature-dependent emptying of grain-boundary charge traps
and the role of charge trapping at grain boundaries in CVD
diamond (Hearne 01 al., 2004, 2006), characterizing particle
detectors, studying detectors for single-atom doping experiments
(Yang el al., 2003), and analyzing solar cells and LEDs. Many
of these applications arc reviewed in Jamieson (1997) and
Jamieson ef al. (1999).
13.2.5.1 Time-resolved IBIC (TRIBIC)

Another mode of IBlC analysis is a time-resolved version in
which the time evolution of the charge pulse on the nanosecond
timescale is recorded, in addition to the total amount of charge
produced. This enables the differentiation of fast charge drift
within electric fields and a slower component from charge
diffusion through regions free of electric fields (SchOne et al.,
\999; Guo ef al., 2001). TRIBIC has been used to measure the
rate at which charge is collected by individual device nodes due
As an example of IDle analysis of radiation detectors, Fig.

13.13 shows IBIC images of a coplanar diamond detector with
297
Chapter 13
to the passage of high-energy heavy ions. TRIBle has also been

used in conjunction with a temperature-controlled sample stage
to study the time evolution of the charge pulses as a function of
temperature to detennine the trap levels within the
semiconductor or insulator bandgap in a manner similar to deeplevel transient spectroscopy (Laird et al., (999).
broad beams of high-atomic-number (bigh-Z), high-energy

(HZE) ions and observing any increases in the overall upset rate,
but this does not enable the easy detennination of which portion
of the device is causing the upset. Single-event upset (SEU)
imaging (Hom et aI., 1992, 1993) provides a means of producing
spatially resolved images of the upset sensitivity ofa device area
by irradiating it with a focused beam of HZE ions in a specially
An example of a TRlBIC analysis is shown in Fig. 13.14,
designed microprobe for focusing such ions. By correlating. the
where time-resolved IBIC signals (current transients) were
spatia1ly resolved upset sensitivity in conjunction with mlC
recorded using a focused 12 MeV carbon ion beam at s'patially
images and design masks, the components which are most likely
resolved locations across a single complementary metal oxide
to cause upsets can be identified. SEU imaging uses a similar
semiconductor (CMOS) transistor. Complete current transients
collecting scheme to that shown in Fig. 13.12(a), with the
induced by inruvidual ,ions were recorded using a 5 GHz
difference that it uses a focused beam of HZE ions and measures
bandwidth with resolved components due to charge diffusion"_
.- any changes in device logic states whereas IBIC records the ion
and drift. Figure 13.14(a) sbows the area of interes~ which
induced charge arriving at device junctions.
comprises the drain area in the center and the gate area below. In
the top image, it can be seen that most charge is collected in the
The example in Fig. 13.15 demonstrates the power of SEU
center of the drain, with ions striking within 1-2 J.U11 of the drain
imaging. Figure 13.15(a) shows measurements of SEU upset
diffusion area or close to the gate also producing large charge
probability recorded by irradiating the whole of a silicon-ontransients. However, the total charge collected is Dot a sufficient
insulator (SOl) static random access memory (SRAM) device
indicator as to whether the region is sensitive to being upset
with an unfocused beam of HZE ions, giving the average SEU
(SEU imaging is described in the next section), so the bottom
probability but no information on which part of the device is
image is more useful in predicting this. The measured time
most likely to be causing the upsets. Historically, SOl circuits
evolution of the current transients is shown in Fig. 13.14(b).
have proven to be less prone to SEU from heavy-ion strikes than
bulk silicon circuits. In a bulk Si transistor, the cbarge generated
by the incident ions is collected from the substrate below the
device, regardless of whether the gate or the drain is hit. It was
once thought that only gate strikes cause SEUs in SOl transistors
<.m
because the electric fields required to induce charge collection
... are present only near the gates. The charge deposited in the
l~
buried oxide, substrate, and drain region was believed not to
I::
contribute to the charge collection process. The SEU image in
Fig. 13.15(b), however, clearly sbows that the SEUs are
occwring not only in the gate areas but rather in the combined
draln-gate areas, indicating that charge collection probably
occurs from the substrate below the buried oxide in this SOl
(a)
(b)
device. Other examples can be found in Fischer et al. (1997),
FIG. 13.14. (a) ~Tw(}-dimensional scan over a 8.3 x 30 1llD' area
Aton et al. (1997), and Vlzkelethy et al. (2005).
of a metal oxide semiconductor fie1d-effect transistor
13.2.7100 beam induced hlminescence (IBIL) or
(MOSFET): (top) total cbarge collected and (bottom) peak drain
ionoluminescence
current. (b) Current transients induced by ions striking near the
Ion beam induced luminescence (IBlL) or ionoluminescence
drain edge, within the region and through the gate. Reprinted
relies on the ion-induced generation of excited electronic states
from ScMn. er al. (1999).
in semiconductors, insulators, and minerals. De-excitation can
result in emission of ultraviolet, visible, or infrared light that is
13.2.6 Sing1e-event upset imaging
recorded by a spectrometer (Homann et al., 1994). mIL can
Both solar and cosmic radiation result in the production of
provide information on the chemical bonding of elements
electron-bole pairs in the semiconductor and insulating materials
prc:seQt or their defect states, which is manifested by light
in satellite and terrestrial microelectronic devices iliat are used in
emission of different wavelengths and intensities. The light
satellites and other high-radiation environments. This excess
coUection efficiency is relatively low, as luminescence is emitted
charge can result in various forms of upsets to device memories.
isotropically but collected over only a eertain solid angle, so
and functioning, either as permanent damage (bard upsets) or
analysis might require a beam current of several picoamperes or
transient damage such as corrupting the information. stored in a
more. Where both lBIC and IBIL analyses are performed on the
memory cell (soft upsets). There has been much work on
same sample, IBIC should be carried out first so as to minimize
designing devices that have the lowest possible sensitivity to
damage. mIL is particularly useful when used in conjunction
upsets. Much work has been carried out irradiating dcvices with
wilh PIXE for trace-element analysis or RBS for depth profiling.
..
298
.'.
B If
~.
Ii
p.olt~.t..
n-origat.s
and drains
~iJ.d
(a)
dnlilis
(b)
FIG. 13.15. (a) Single-event upset (SEU) cross section of a silicon-an-insulator (Sal) static random access memory (SRAM) device
from broad-beam tests carried out at Sandia National Laboratories (SNL) and Brookhaven National Laboratory (BNL). (b) SEU image
of the same device stnu;ture recorded with the nuclcar microprobe and overlaid on the circuit layout.
The most cornmOD application of mIL is in the study of

semiconductor and insulator materials such as diamond
(Manfredotti ., 01., 1999a, 1999b), synthetic quartz, corundum
and spinel (Bettiol el al., 1997), and boron nitride (Manfredotli
e/ 01., 2001). IBIL bas been used in combination with PlXE for
rare-earth-element analysis and in conjunction with lBIC to
image Ni segregation in synthetic diamond (Bettio! et a!', 2001),
diamond meteorites (Lo Giudice el al., 2003), and CVD
diamond(Manfredotti ., 01., 1998; Sellin " 01., 2000): It lias
also been used for the identification of cytological structure
(Rossi et 01., 2001), and other work in materials and earth
sciences is described in Vittone et al. (2oo1a). Other work is
reviewed in PalloD et al. (1997) and Jamieson el 01. (1999), and
more comprehensive accounts are given in Breese et af. (1996)
and Townsend el 01. (1994).
XAXIs
4,OXl o\!-Tr'"",,,,,,,====,,=~,.--.:--:<::;;;;:;;':=-----.
3.Sx10'5
'3.Ox.,o'~1L:=:'::===
2,:5;11;10"
-~
Figure 13,16 shows an example ofreIL analysis. Diamond is

an ideal material from which to make radiation dosimeters
because it is tissue-equivalent, chemically inert, nontoxic, and
radiation-hard, Vittone et af. (2001b) fabricated diamond
microdosirneters on tungsten wires coated with gold and used
IBIL, IBIC, and PIXE to analyze these structures. IBIL was used
to image radiative recombination centers, and when combined
with PIXE, it enabled the evaluation of the shape of the diamond
film covering the tungsten needle. In Fig. 13J6(c), the main
luminescence peak centered at 2.8-3.0 eV is the A-band of
diamond. After a very high irradiation dose of 1013/cm2, the Aband peak bas been quenched, and a peak at 2.41 eV has
appeared, attributed to the nitrogen~related H3 band.
2,OX10:
1.5x'10
5,0;1(10'"
(c)
0.0-:_
.,
<
~tl
......
_~---<i!I
'.
'
1.61.82'.'02,:2 2.4 2.6 2.8 3;0 3.23.43,6 -3:8
En"tgy [EN]
FIG. 13.16. (a) Scaoning electron microscopy (SEM) image of.
diamond microdosimeter tip. Ion beam induced luminescence
(IBIL) and particle-induced X-ray emission (FIXE) images were
recorded from the 100 x 100 }illl' box area shown in (b).
Superposition of the tungsten PIXE image (dark region) and a
panchromatic IBIL im.ge (light gray region). (c) IBIL spectra of
a small region of the microdosimeter tip. Reprinted from Vittone
e/ al. (200Ib).
299
Chapter 13
al., 2007). A microprobe is used to target individual cells or sub~

cellular componects such as cell nuclei, cytoplasm, and
membrane and deliver a specific ion dose. This enables
radiobiological studies of cells that cannot be easily performed
by other approacbes using unfocused ion beams or X-rays.
Applications include studying the response of cells to internal
and external stimuli and radiation, cancer therapy, and
occupational risks of exposure to radiation.
13.2.8 lon-induced electron imaging

Imaging with secondary electrons induced by a focused
megaelectronvolt high-energy ion beam provides topographical
information about samples, which is important for the
identification of regions of interest on samples and the
interpretation of spatially resolved images collccted using PIXE
or RBS, for example.
In pulse-counting mode, the image contrast is produced by

counting secondary electron pulses recorded at each scan
position with a channel electron multiplier. The signal output
can be easily integrated into existing data-collection electronics.
The image is acquired randomly over periods of many seconds.
- <-- ='.c.. -. Variations in signal mtensity due to current fluctuations tend to
average out with time, so the images become less sensitive to
fluctuations and reasonable topographical contrast can be
obtained (Morris el al., 1985; Kneis et al., 1982). However, the
number of electrons contributing to the image is fundamentally
limited by the counting rate, resulting in statistically noisy
images.
In voltage-modulated mode, the image contrast is controlled
by variations in a large, continuous secondary-electron flux ali
the beam scans across the sample, similar to a scanning electron
microscope. Although this mode was implemented early on
nuclear microprobes (Younger and Cookson, 1979), it has been
little used since because of the fluctuations of the accelerator
terminal voltage, leading to beam current fluctuations and,
therefore, great- difficulties in image fonnation. This v.ras
recognized early on, and schemes based on normalizing the
secondary-electron signal strength to fluctuations in the beam
current were developed, but this resulted in complex signal
processing and limited success.
FIG. 13.17. Ion-induced secondary electron images of (a), (b) a

(c) a patterned silicon
copper grid with a period of 12.7 _
wafer, and (d) a~50 x 50 ~2 area showing a single diamond
crystallite. Reprinted from Teo el af. (2004).
This limitation can be overcome using the new generation of

highly sta.hili=! acc.elerators, enabling highly detailed images to
------be produced. The hardware and performance-ofsuch systems are
described in Teo el al. (2004), along with examples of the types
of images produced and optimum. collection conditions.
Secondary electrons are dctccted by a photomultiplier tube
through a scintillator screen placed in front of it The current
output of the photomultiplier is amplified by a current-to-voltage
amplifier, before being fed into a specialized image collection
and processing system. With careful optimization of the
accelerator, the beam current fluctuations in the microprobe
chamber can be reduced to 2%, so that high-contrast, low-noise
secondary-electron images can be produced. Figure 13.17 shows
ion-induced secondary-electron images recorded under these
conditions whcre thc image quality is comparable to that
obtained using a scanning electron microscope.
There are different approaches to constructing the beam

delivery and focusing system for single-ion cell-irradiation
facilities. The simplest relies on collimating a beam using a
narrow glass capillary located close to the cells to be irradiated
(Fo\kard el al., 1997). More complex systems involve the use of
quadrupole focusing systems to provide a more controllable
scanning-beam irradiation facility. The beam is brought into air
though a thin window, as for an external-beam facility.
Fluorescence microscopy is typically used to identify and
poSition cells at the focused beam spot While excellenl
performance on horizontally mounted microprobes can be
achieved (Dollinger ., al., 2006), the dish holding the cells
should ideally be horizontal, and severa] dedicated, vertically
mounted focusing and deJivery systems have been constructed.
13.2.9 Cell irradiation

An increasingly widespread application of microprobes is to
controllably deliver single high-energy ions within a focused
beam spot for in vitro cell irradiation (Bigelow et ai., 2005;
Dollinger el al., 2006; Arteaga-Marrero et ai., 2007; Nilsson et
13.3 SUMMARY
There are about 50 microprobe research groups around the
world, involved in a wide range of 'studies of microanalytical
work in biology, biomedicine, materials analysis, geology, and
300
archaeology, for example. Some microprobes are based on

focusing high beam current for more traditional ion beam
analysis applications using PXE, RBS, NRA, and ERDA,
7"here~ many rely more on low-beam-current applications
lDvolvmg STIM, IBIC, IBIL, and SEU imaging, where higher
spatial resolutions of less than 100 nm achievable. The best
reported resolution to date is about 35 run, with several new
n:Ucrop~ob~ ~esigns presently under construction that will try to
push this
toward IODIn. This will provide even more scope
for developmg new and existing applications of highly focused
megaelectronvolt ion beams.
w:ut
REFERENCES
Alves, L.c., Alves, E., da Silva, M.F., Soares, J.e., Paul, A., La
Barber., A., and Riccardi, B. (2001), Nuel. Instrum. Methods
BI81,377
Arteaga-Marrero, N., Pallon, I., Olsson, M.G., Auzlelyte, V.,
Elfman, M., Kristiansson, p" Malmqvist, K., Nilsson, C., and
Wegden, M. (2007), Noel. Instrum. Methods B260, 91.
Aton, T., Seitchik, I., Yang, S.-H., and Shichijo, H. (1997), Nuel.
Bettiol, A.A., Nugent, KW., and Jamieson, D.N. (1997), Nuel.
instrum. Methods 8130, 734.
Bettiol, A.A., Ryan, e.G., Jamieson, D.N., and Prawer, S. (2001),

Nuc/.lnslrum. Methods B181, 225.
Bigelow, A.W., Ross, GJ., Randers-Pehrson, G., and Brenner,
OJ. (2005), Nuc!.Instrum. Methods B231, 202.
Breese, M.B.H, Grime, G.W., and Watt, F. (1992), J. Appl.
Phys. 72,2097.
Doyle, B.L., and Wing, N.D. (1983), IEEE Trans Nuel. Sci. 30
1214.
'
Doyle, B.L., Walsh, D.S., and Lee, S.R. (1991), Nuel. Instrum .
Methods B54, 244.
Dymnikov, A.D., and Yavor, S.Y. (1964), SOy. Phys.-Teeh Phys.
(USA) 8 (7),639.
DymnikOY, A.D., and Osetinskij, G.M. (1989), Sov. J. Pa..-/. Nucl.
(USA) 20 (3), 293.
Dymnikov, A.D., and Glass, GA. (2004), Nucl. IMStrum.
Methods B219-220, 994.
Esquinazi, P., Spemann, D., Hohne, R, Setzer, A., Han, K.-H.,
and Butz, T. (2003), Phys. Rev. Lett. 91,227201.
Fischer, B.E., and Muhlbauer, e. (1990), Nucl. Instrum. Adethods
847,271
Fiscber, B.E., Schlagl, M., Barak, I., Adler, E., and Metzger, S.
(1997),Nucl.lnstrum. Methods B130, 478.
Folkru:d, M., Vojn~vic, B., Schettino, G., Forsberg, M., 13owey.
G., Pose, K.M., MIchael, 8., Michette, A.G., and Pfauntsch, SJ.
(1997), Nucl. Instrum. Methods Bl30, 270.
Galbiati, A., Breese, M.B.H., Knights, AP., and Sellin, P.J.
(2001), Nllc/. Instrum. Methods A466, 52.
Grime, G.W., Watt, F., Blower, o.D., Takacs, l, and Jamieson

D.N. (1982), Nuel.lnstrum. Methods 197, 97.
'
Grime, G.W., and Watt, F. (1984), Beam Optics oj Quodrupole
Probe Formjng Systems, Adam Hilgcr, Bristol, UK.
Grime, G.W., Dawson, M., Marsh, M., McArthur, I.C., and Watt
F. (1991), Nuc!.Instrum. Methods B54, 52.
'
Breese, M.B.H, Killg, PJ.C., Grime, G.W., and Wllshaw, P.R.

Guo, B.N., EI Bouanani, M., Renfrow, S.N., Nigam, M., Walsh,
_
_~
D.S., Doyle, B.L., Duggan, I.L., and McDaniel, F.D. (2001),
Breese, M.B.H. (1993), J. Appl. Ph),s~74: 3789.";;:"
. Nuc!.Instrum. MethodsBl81, 315. - - - . '.
-Brcese, M.B.H., Jamieson, D.N., and Killg, PJ.C. PW, 09(6):-Q;;adruporeo;,:;,
Materials Analysis Using a Nuclear Microprobe, Wiley, New
odem OptIcs 42, ~pnnger, Berlm.
York.
Hearne, S.M., Janueson, D.N., Trajkov, E., Prawer, S., and
Butier, I.E. (2004) Appl. Phys. Lett. 84,4493.
Breese, M.B.H., Grime, G.W., Linford, W., and Harold, M.
Hearne, S.M., Trajkov, E., Iannieson, D.N., Butler, I.E., and
(1999), Nuel. lnstrum. Methods BI5S, 48.
Prawer, S. (2006),J. Appl. Phys. 99, 113703.
Breese, M.B.H., T.eo, E.J., Raua, M.A., Huang, L., van Kan, lA.,
Homann, N.P.O., Yang, e., and Malmqvist, KG. (1994), Nucl.
Watt, F., and Killg, PJ.e. (2004), Phys. Rev. Lett. 92,045503.
Instrum. Methods A353, 610.
Breese, M.B.H, Huang, L., Teo, E.J.,-King, P.J.C., and Wilshaw,
Hom, K.M., Doyle, B.L., and Sexton, F.W. (1992), IEEE Trans.
P.R. (2005), Appl. Phys. Lett. 87, 211907.
N.c!. Sci. 39, 7.
Cholewa, M., Bencb, G., Legge, G.J.F., and Saint, A. (1990),
Hom, K.M., Doyle, B.L., Sexton, F.W. Laird, J.S., Saint, A.,
Appl. Phys. Lett. 56, 1236.
Cholewa, M., and Legge, GJ.F. (1993), Noel. Instrum Methods
de Kerckhove, D.G., Breese, M.B.H., and Grime, G.W. (1997),
B77,355.
Nuc/. lostrum. Methods B129, 534.
Incerti, S., Smith, R. W., Merchant, M., Grime, G.W., Moot, F.,
de Kerckhove, D.O., Breese, M.B.H., Wilkinson, A.G., and
Seram, L., Moretto, P., Touzeau, C., Barberet, P., Habchi, C.,
Grime, G.W. (1998), Nuel. lnstrum. Methods BI36-138, 1240.
and Nguyen, DT. (2005), Nuel.lnstrum. Methods B231, 76.
Dollinger, G., Bergmaier, A., and Hauptner, A. (2006), Nuel.
Jaksic, M., Pastuovi, Z., and Tadi, T. (1999), Nucl. Instrum.
Methods B158, 458.
(1993), Appl. Phys. Lett. 62, 3309.
~9~::---~;;kes:
301
L.
s~rin-;'~ Tra-';;~
Chapter 13
Jaksic, M., Gracin, D., Meduni, Z., Pastuovi, Z., Vittone, E., and
Nava, F. (2002), Nud. Instrum. Methods B188, 130.
Manfredotti. C., Fizzotti, F., Lo Giudice, A., Paolini, c., Vittone,

E., aod Lu, R. (2002), N.cl. Instrum. Methods B187, 566.
Jamieson, D.N., and Legge, GJ.F. (1987) Nud. Instrum.

Methods B29, 544.
Mazzer. M., Drigo, A. V., Romanto, F., Salviati, G., and

Lazzarini, L. (1997), Phys. Rev. B56, 6895.
Jamiesnn, D.N., and Ryan, e.G. (1993) Nud. Instrum. Methods

B77,415.
McCallum, J.C., Mckenzie, C.D., Rossiter, KG., Sho~ KT.,

and Williams, J.S. (1983), Appl. Phys. Lett. 42, 827.
Jamieson, D.N. (1997), Nucl. Instrum. Methods B130, 706.
McMillan, I.W. (1988), Nucl Instrum. Methods B30, 474.
Jamieson, D.N., Beckman, D.R., Bettiol, A.A., Laird, J.S., Lee,

KK, Prawer, S., Saint, A., Witham, L.C.G., and Yang, e.y.
(1999), Nud. Instrum. Methods B158, 628.
Miebelet-Habcbi, c., Incerti, S., Aguer, P., Barberet, Ph.,

Gontier, E . Grente, K... Moretto, Ph., Nguyen., D.T., Pouthier,
A., Pouthier, T., Rebilla~ F., and Smith, R. W. (2005), Nud.
Johansson, S.A.E., and Campbell, J.L. (1988), PIXE. A Novel

Technique/or Elemental Analysis, Wiley, New York.
Johansson, S.A.E., Campbell, J.L., and Malmqvis~ KG. (1995),
Particle Induced X-ray Emission Spectrometry, Wiley, New
York.
King, PJ.C., Breese, M.B.H., Smulders, P.J.M., Wilshaw, P.R.,
and Grime, G.W. (1995), Phys. Rev. Letl. 74,411.
Kneis, H., Martin. B., Nobiling, R., Povh, B., aDd Traxel, K
(1982), N.cl. Instrum. Methods 197, 79.
Laird, l.S., Bardos, R.A, Jagadish, C., Jamieson, D.N., and

Legge, G.J.F. (1999), Nud. Instrum. Methods B158, 464.
Landsberg, J.P., McDonald, B., and Watt, F. {I 992), Nature 360,
65.
Minqin, R., van Kan, J.A., Bettiol, A.A., Daina, L., Yee Gek, c.,
Boon Hua~ B., Whitlow, H., Osipowiez, T., and Watt, F. (2007),
Mosbah, M, Clochiatti, R., Tirira, J., Gosse~ 1., Massin~ P., and
Troeellier, P. (1991),Nucl. Instrum. Methods B54, 298.
Morris, W.G., Katz, W., Bakhru, H., aod Haberl, A.W. (1985), J
Vac. Sci. Techno!. B3 391.
Nilsson, C., Petriconi, S. Reinert, T., and Butz, T. (2007), Nucl.
lnstrum. Methods B260, 71.
Ortega, R. (2005), Nucl.lnstrum. Methods B23I, 218.
Leamy, H.J. (1982),.1: Appl. Phys. 53, RS1.
Osipowicz, T., Zmeck, M., Watt, F., Fiege, G., Balk, L.,
Niedemosfueide, F., aod Schulze, H.-I. (2001), Nuel. I.strum.
Methods B181, 3 II.
Llabadnr, Y., Moretto, P., and Guegan, H. (1993), Nucl I.strum.

Methods B77, 123.
Overley, J.C., Schofield, R.M.S., MacDonald, J.D., aod Lefevre,

H.W. (1988), Nuel. Instrum. Methods B30, 337.
Llabador, Y., and Moretto, Ph. (1998), Applications of Nuclear

Microprobes in the Life Sciences, World Scientific, Singapore.
Pallon, J., Yang. C., Utui, R.I., Elfman, M., Malmqvis~ K.G.,
Kristiaosson, P., and Sj6land, KA. (1997), Nucl. Instrum.
Methods B130, 199.
Lo Giudlce, A" Pratesi, G., Olivero, P., Paolini, c., Vittone, E.,
Manfredotti, C., Sammjceli, F., and Rigato, V. (2003), Nucl.
Pierce, T.B., McMillan, J.W., Peck, P.F., and Jon"" I.G. (1974),
Nuc!. Instrum. Methods lI8, 115.
Lu, R.R., Manfredotti, C., Fizzotti, F., Vittone, E., and Lo

Giudice, A. (2002), Mater. Sci. Eng. B90, 191.
Maggiore, CJ. (l982), Los Alamos Sci.. 3, 26.
Pontau, A.E., A.ntolak, AJ., Morse, D.H., Ver Berkmoes, A.A.,

Brase, J.M., Heikkinen, D. W., Martz, HE., and Proctor, ill.
(1989), Nud. Instrum Methods 840-41, 646.
Manfredotti, C., Fizzotti, F., Vittone, E., Boero, M., Poleselln, P.,
Galassini, S., Jaksic, M., Fazinic, S., and Bogdanovic, 1. (1995),
Nucl. Instrum. Methods BIOO, /33.
Manfredotti, C., Apostato, G., Cinque, G., Fizzotti, F., Lo

Giudice, A., Polesello, P., Truccato, M., Vittone, E., Egeni, G.,
RudeUo, V., and Rossi, P. (1998), Diamond Relat. Mater. 7,742.
Rath, S., Sellin, P., Breese, M.B.H., Herman, H., Alves, L.e.,
aod Hollaod, A.H. (2003), Nuc/. Instrum. Methods A512, 427.
Rossi, P., Di Maggio, C., Egeni, G.P., Galligioni, A., Gennaro,
G., Giacomelli, L, Lo Giudice, A, Pegoraro, M., Pescarini, L.,
Rudello, V., and Vittonc, E. (2001), Nuc!. lnstrum. Methods
BI81,437.
Manfredotti, C., Fizzotti, F., Polesello, P., and Vittone, E.

(1999a), Nuc!. Instrum. Methods A426, 156.
Ryan, C.G., and Jamieson, D.N. (1993), Nuel. Instrum. Methods

B77,293.
.
Manfredotti, C., Fizzotti. F., La Giudice, A, Polesello, P.,

ViUone, E., Truccato, E.M., and Rossi, P. (l999b), Diamond
Relat. Mater. 8, 1592.
.
Ryan, CG., Jamieson, D.N., Churms, C.L., and Pilcher, l.V.

(1995), Nucl. Instrum. Methods BI04, 157.
c.,
Manfredotti,
Vittone, E., La Giudice. A., Paolini, C, Fizzotti,
F., Dinca, G., Ralchenko, V., and Nistor, S.V. (2001), Diamond
Relat. Mater. ]0,568.
302
Ryan, C.G., and Jamieson, D.N. (1999), Nuel. Instrum. Methods

BI58,97.
Sakai, T. Kamiya, T., Oikawa, M., Sato, T., Tanaka, A., and
Ishii, K (1999), Nud. Instrum. Methods BIS8, 275.
Nudear Microprobe Analysis
Sakai, T., Oikawa, M., Sato, T., Nagamine, T., Moon, H.D.,
Nakazato, K., and Suzuki, K. (2005), Nuc/. lnstrum. Methods
B231, 112.
Schone, H., Walsh, D.S., Sexton, F.W., DOYle, B.L., Dodd, P.E.,
Aurand, J.F., and Wing, N. (1999), Nucl. lnstrum. Methods
BI58,424.
Sellin, PJ., Breese, M.B.H., Alves, L.e., Sussman, R.S., and
Whitehead, AJ. (2000), Appl. Phys. Lell. 77, 913.
Seng, H.L., Osipowicz, T., Zhang, J., Tok, E.S., and Watt, F.
(2005), Nucl. Ins/rum. Methods B231, 446.
Soptief, A. (od.) (1967), Focusing of Charged Particles,
Academic Press, London, vols. I and II.
Sexton, F.W., Hom, K.M., Doyle, B.L., Laird, J.S., Cholewa, M.,
Saint, A., and Legge, G.J.F. (1993), Nud. [nstrum. Methods B79,
436.
Shariff, A., Auzelyte, V., Elfman, M., Kristiansson, P.,
Malmqvist, K., Nilsson, C, Pallon, 1., and Wegden, M. (2005),
Nucl. Instrom. Methods B231, 1.
Sofield, C.1., Bridwell, L.B., and Wright, C.1. (1981). Nucl.
instrum. Methods 191, 79.
Teo, E.1 .. Breese, M.B.H.. Bettiol. A.A., Watt, F., and Alves,
L.C. (2004), J. Vac. Sci. Technol. B22, 560.
Townsend, PD., Chandler. P.1., and Zhang, L. (1994), Optical
Effects of Ion Implantation, Cambridge University Press,
Vittone, E., Fizzotti, F., Lo Giudice, A, Paolini. C, and

Maofredotti, C. (2000), Nucl.Instrum. Methods BI61-163, 446 .
Vittane, E., Lo Giudice. A., Manfredotti, C., Egeni, G., !Ludella,
V., Rossi, P., Gennaro, G., Pratesi, G., and Corazza, M. (200la),
Vittone, E., Ricciardi, c., Lo Giudice, A, Fizzo"tti, F.,
Manfredotti, C., Egeni, G., and Rudello, V. (2001b), Nucl.
[nstrum. Methods B181, 349.
Viltone, E. (2004), Nud.lnstrum. Methods B219-220, 1043.
Vizkeletby, G . Brunet~ B.A., Walsh, D.S., James, R.B., Olsen,
R W., and Doyle, B.L. (1999), Nud. Instrum. Methods B158,
437.
Vizkelethy, G., Doyle, B.L., Brice, D.K., Dodd, P.E., Shaneyfelt,
M.R., and Sehwank, J.R. (2005), Nucl. [nstrum. Methods B231,
467.
Wang, W., Zhu, J., Le, A., Lu, R, Lin, S., Bao, 1., Jiang, D., She,
J., and Ji, A. (1993), Nuc!. [nstrum. Methods B77,349.
Watt, F., and Grime, G.W. (eds.) (1987), Principles and
Applications of High Energy 10n Microbeams, Adam Hilger,
Bristol, UK.
Watt, F., Choo. T.F . Lee, K.K.. Osipowicz, T.. Orlie, I., and
Tang, S.M. (l995),N"cl. [nstrum. Methods Bl04, 101.
Watt, F., van Kan, 1.A., Rajta, I., Bettiol, A.A., Choo, T.F.,
Breese, M.B.H., and Osipowicz, T. (2003), N"c/ Instrum.
Methods B210, 14.
Yang, C., Jamieson, D.N., Pakes, C.L, George, D.P., Reame,
S.M., Dzurak, A.S., Gauj., E., Stanley, F., and Clark, R.G.
(2003) Nucl. [nstrum. Methods B210, 186.
Younger, P.A., and Cookson, J.A. (1979), Nuc!. [nstrum.
.... Methods B158, 193.
London.
Udalagama, C.N.B. (2006), PhD thesis, National University of
Singapore, SingafXlre.
Visser, 1., Mous, DJ.W., Gottdang, A., and Haitsma, RG.
(2005), Nucl.Instrum. Methods B231, 32.
Vittone, E., Fizzotti, F., Gargioni, E., Lu, R., Polese\10, P., Lo
Giudice, A., Manfredotti, C., Galassini, S., and JaksiC; M. (1999), ~j.71""'~'-~"'"""""'"
Nucl. Instrum. Me/hods B158, 476.
303
I I.
IL
..... --"""~., ... -.....
':."'-~ '"-----.-"."'-"'><~~..,....,."""--.~-
CHAPTER
DATA ANALYSIS SOFTWARE

FOR ION BEAM ANALYSIS
N. P. Barradas
Instituto Tecnologico e Nuclear, Sacavem, Portugal
Centro de Fisica Nuclear da Universidade de Lisboa, Lisboa, Portugal
E. Rauhala
CONTENTS
14.1 INTRODUCTION..~ ............................................ ! 307
14.1.1
Data analysis software .............................................:.. :.. :..:.:.: ... ,.............::.:.307
14.1.2
Scope of the chapter .....................................................................................307
14.1.3
Historical development and reviews of software programs ..................... 307
14.2 TYPES OF CODES .................................................................................................307
14.2.1
The direct calculation method .................................................................... 307
14.2.2
Simulation by successive iterations ............................................................ 308
14.2.3
Interactive versus automated ......................................................................311
14.2.4
Deterministic versus stochastic (Monte Carlo) ......................................... 311
14.3 CAPABILITIES OF CODES .................................................................................312
14.3.1
Design basis ................................................................................................... 312
14.3.1.1. Techniques implemented ............................................................. 312
14.3.1.2. Experimental conditions supported ...............................................312
14.3.1.3. Description of samples ..................................................................... 312
14.3.2
Databases implemented ...............................................................................313
14.3.2.1. Stopping power ................................................................................. 313
14.3.2.2. Scattering cross section .................................................................... 314
14.3.3
Basic physics ................................................................................................. 315
305
.-
"~~r
Chapter 14
t
!
14.3.4
Advanced physics .........................................................................................316

14.3.4.1. Straggling models .............................................................................316
14.3.4.2. Electron screening ............................................................................316
14.3.4.3. Plural scattering ...............................................................................316
14.3.4.4. Multiple scattering ........................................................................... 316
14.3.4.5. Simulation of resonances .................................................................318
14.3.4.6. Surfaceand interface ronghness ..................................................... 318
14.3.4.7. Channeling ........................................................................................ 318
14.3.4.8. Pulse pilenp ....................................................................................... 318
14.3.5
Automated optimization .............................................................................. 318
. 14.3.5.1. Fitting ................................................................................................319
14.3.5.2. Bayesian inference ...........................................................................319
14.3.6
Usability and usage ......................................................................................319
+I
t
t
14.4
ACCURACY ............................................................................................................320
14.4.1
Numerical accuracy of codes .......................................................................320
14.4.2
Intrinsic accuracy of IDA experiments ......................................................320
14.4.2.1. Models and basic physical quantities .............................................320
14.4.2.2. Experimental conditions ..................................................................321
14.4.2.3. Counting statistics ............................................................................321
14.4.3
Physical effects on data analysis ................................................................. 321
14.5
RUMP AND FIRST-GENERATION CODES ...................................................... 321
14.6
NEW-GENERATION CODES: SIMNRA AND NDF ......................................... 326

Further capabilities ......................................................................................326
14.6.1
14.6.2
SIMNRA .......................................................................................................326
14.6.3
NDF ............................................................................................................329
14.6.4
Issues ............................................................................................................332
14.7
MONTE CARLO SIMULATION .......................................................................... 333
14.8
OTHER TEClINIQUES .........................................................................................334

14.8.1
PIXE ............................................................................................................334
14.8.2
Resonant NRA, PIGE, NRP, MEIS, channeling, microscopies,
and so on .......................................................................................................336
14.9 THE IDA DATA FORMATIDF ............................................................................336
ACKNOWLEDGMENTS ...................................................................................................337
.i
SOFTWARE WEB SITES ................................................................................................. 343
1
!,
REFERENCES ............................................................................................................................... 344
,
;
";
306
Data Analysis Software for Ion Beam Analysis
14.1 INTRODUCTION
14.1.1 Data analysis software
The data analysis software in ion beam methods are computer
programs designed to extract information about samples from
the measured ion beam spectra. The desired information includes
identification of sample elements, their co~centrations, areal
densities, and layer thicknesses. At best, one spectrum can be
converted to concentration--depth distributions of all elements in
the sample, Often, however, such a full description of the sample
based on a single experiment is not possible. The analyst can
then perfonn additional experiments using different
experimental parameters such as ion energies or a different
measurement geometIy or including information from one or
more other complementary techniques.
profiling (NRP) [see Arnsel (1996) fOf a discussion on

acronyms]. Short sections on particle-induced X-ray emission
(PIXE) spectroscopy and other techniques such as N"RP and
channeling are also included.
A list of Web sites relevant for rnA software a.nd data

analysis is provided at the end of the chapter, just before the li.st .of references. Note that this list is by no means comprebensive,
and in particular, that several of the codes mentioned in this
chapter have no dedicated Web site.
The software can also be used for designing relevant

ex.periments. Depending on the assumed sample structure, the
appropriate ion beam technique can be chosen by doing test
simulations prior to experiments. For example, the most suitable
parameters for the measurements can be tested.. The choice of
ions, energies, geometries, and so on can be examined without
performing often ex.tensive and complicated experiments.
The third main function of the software is found in education
and research. By using simulation computer programs, the
researcher can readily scrutinize the characteristics of various
ion beam techniqlJes. The dependencies between stopping
powers, cross .sections, energies, and so on ean be studied. Novel
reaction cross sections and stopping powers ean be implemented
in the programs to investigate the agreement between simulatioD
and ex:perimental data. Fine details, such as double and multiple
scattering, screening, pileup, and the properties of various
detector systems can be analyzed with the help of data analysis
...
...
programs.
'b '
h"
-d'-th"' ".' ftw'
'd
d'"
It cannot e overemp aslZe
at so are IS an al to ata "".-.
analysis and does not replace the judgment of the analyst.

Sofuvare that is not correctly used, or that is used outside its
scope of application, leads to wrong data analysis. Furthermore,
it should be noted that the scope of application of a given
program might not be clear and will depend on the details of the
experiment being analyzed: for instance, Bohr straggling is
adequate for analysis of nonnal-incidence proton Rutherford
backscattering spectrometry (RBS) with a surface barrier
detector, but it is completely inadequate for heavy-ion elastic
recoil detection analysis (ERDA) or for high-resolution nearsurface analysis, where the straggling is asymmetric.
14.1.3 Historical development and reviews of software

programs
Computer programs to analyze data from RES, N:.EA, and
ERDA date back to the 1960s and 19705. These techniques
developed in parallel with the beginning of tlle new
semiconductor and other high-technolOgy industries, where new
needs for ion beams in the production, modificati on, and
characterization of novel materials were arising. rnA techniques
for quantitative depth profiling in the micron rauge and
determination of low-concentration elemental impurities were
quickly recognized and widely applied. As these analytical tools
became more versatile, the samples, spectra, and data analysis
problems also became more complicated. By the eud of the
1990s, the collection of software bad developed in various
directions: codes exist that can handle very general data analysis
problems and various IBA techniques; some are highly
alJtomatic, whereas others treat specified problems with great
ex:actness.
The historical development of the software was reviewed
recently by Rauhala et al. (2006), which presents the
characteristics and status of 12 software packages as of 2006. An
intercomparisoD of several software packages for simulations
and real data-analysis applications is presented in Banadas el al.
(200Th). Relevant developments up to 2009 are also included in. ~

this cha .
~--
ter
14.2 TYPES OF ,CODES
14.2.1 The direct calculation method

In direct spectrum analysis, the yields from separated signals
of the spectrum are transformed into concentration values by
closed-form analytical calculation. Other forms of direct analysis
include conversion from signal areas and heights to total
amounts or concentrations. This is the opposite of spectrum
simulation, where a theoretical spectrum is generated from an
assumed depth prome and compared to the actual data
This approach was introduced by Borge>en el al. (1982) with

the code SQUEAKIE. This approach is straightforward and
effective in many cases. However, it has problems becall..<:;e of
the implementation of straggling and the stability of results as a
result of uncertainties in the stopping database and measurement
noise. The direct method is still used, nonnally as a complement
to full simulations, and often in heavy-ion ERDA (Spaeth el aI.,
1998).
14.1.2 Scope of the chapter

This chapter deals mainly with the data analysis software of
particle-particle ion beam analysis (IBA) techniques: Rutherford
backscattering spectrometry (RBS), elastic recoil detection
analysis (ERDA), and nuclear reaction analysis (NRA). In the
last case, we limit ourselves to techniques utilizing one single
beam energy, as opposed to resonant NRA or narrow resonance
307
Chapter 14
14.2.2 Simulation by successive iterations

Almost all of the presently available rnA data anaJysis
software is based on simulation by successive iterations. The

simulation modeling assumes that the underlying physics,
mathematics, and nuclear and atomic data are valid and
adequately describe the physical processes involved. Starting
from a known sample structure, the corresponding experimental
energy spectrum can be theoretically simulated from a few basic
data and the knovm formalism of the reaction spectrometry. by

comparing the experimental and theoretical spectra and
iteratively modifying the assumed composition of the sample, a
close similarity of the spectra is typically accomplished after a
few iterations. The sample structure leading to the theoretical
spectrum is then taken to correspond to the material's sample
structure. Erroneous results or misinterpretations of the
material's structure can result from incorrect science, ambiguous
data, or inadequate software documentation and guidance for
analysts to extract the correct information.
The typical iterative simulation proceeds as follows:

1. Assume the experimental parameters and the known
experimental spectrum.
2. Assume an initial sample composition.
3. Calculate a theoretical spectrum corresponding to the
experimental parameters and the assumed sample composition.
4. Compare the simulated spectrwn to the experimental

spectrum.
5. Identity Ihe differences and modity Ihe assumed sample
composition accordingly.
6. Repeat iteration steps 3-5 until an adequate fit between the

simulated and experimental spectra is accomplished.
7. Take the sample parameters corresponding to the best
simulation to be the sample Structure resulting from the data
analysis.
This procedure is illustrated in EXAMPLE 14.1.
EXAMPLE 14.1. Convergence ofitemtion.

Assume that the experimental parameters and the elements in the sample are known. Start by making a guess about the sample
composition. How do the iterations converge?
,0000
2.0 MeV 4He

LaGaO
16000
o~
1"
r-
.Si
:;;;.
12000
.i
4"
La
i~O"
..;;.;i;;
:rj . ~
2000
P,~~~~r-~~~~
~5ilo 600
to() Iioo
900 1000' hOt! . j'2(j()
Scattered ion energy (keV)
4i'
.fa'\
j,
~~tt~red i.ori
.~
"
enefgy{keY)
Fig. 14.1. Backscattering spectrum (0) for 2.00 MeV ~e ions incident on a LaGaO layer on a silicon substrate.
308
....
..
.;' ..
. " .. " .'
1<00. 1450 1500 " 1550,1600 1650 :1700'1750
laoo 1850
Data Analysis Sofu,vare for Ion Beam Analysis
Figure 14.1 shows the spectrwn for 2.00 MeV 4He ions incident
on a LaGaO layer on a silicon substrate. In the experiment, the
following parameters were used:
Incident 4He-ion energy, EHe = 2.00 MeV

Scattering angle, 9 = 170 0
Angle of incidence, 8, = 50
Solid angle x accumulated charge, Qrl = 69.0 IDSr J.1C

Energy per chanoel, BE = 3.475 keY/chanoel
Energy in zero channel, = 27.3 ke V
Zeroth iteration: Initially, as a guess of the LaGaO film
composition, equal concentrations of the elements' in a layer of
200)( lOIS at.lcm1 areal density were assumed.
Initial assumptions:
Equal amounts of the three elements, 33% La, 33% Ga, 33% O.
Layer areal density, 200 x IOu at./cm2.
:::::) Oth iteration.

The resulting simulated spectrum is denoted as the Oth
iteration in Fig. 14.1. The simulated signal heights of La and Ga
and tbe signal widtb of La were then compared to those of the
experimental spectrum. Concentrations and layer areal density
were changed assuming the same linear dependence between the
signal heights and concentrations and between the signal widths
and areal density of the layer.
The La concentration was thus multiplied by Lhe ratio of the
experimental La signal height to the simulated La signal height.
Similarly, the areal density of the fihn was multiplied by the
ratio of the experimental La signal width to the simulated La
signal width.
Reduce the La concentration to 85% x 0.25 = 0.2l.

Reduce the Ga concentration to 88% x 0.26 = 0.23.
1
Increase the areal density of the layer to 1.2)( 480 )( 10 .3 at.lcrni
2
15
= 580 x 10 at/cm
::::::) 2nd iteration.
Third iteration. The height of the experimental La signal was

observed to be 95% of the height of the simulated signal. The
same ratio for Ga was observed to be 93%.
The experimental signal was observed to be 1.06 tim.es wider
than the simulated signal.
Reduce the La concentration to 95% x 0.21 = 0.20.
Reduce the Ga concentration to 93% x 0.23= 0.21.
Increase the areal density of the layer to 1.06 x 580 x 10 15
aUcm 2 = 615 x lOIS.at/cm2.
=> 3rd iteration.

Fourth iteration. The height Of the experimental La signal was
observed to be 97% of the height of the simulated signaL The fit
of the Ga signal was acceptable.
The experimental signal was observed to be 1.04 times wider
than the simulated signal.
Maintain the Ga concentration at 0.21.
Increase the areal density of the layer. to 1.04 x 615 x lOIS
2
aUcm2 = 640 x 10 1s atJcm
::::::) 4th iteration.
As a result of the fourth iteration, both the signal heights and

the signal widths of La and Ga seem to be fitted adequately
This process was then repeated. The solid lines denoted as ...._- within the 'plotting accuracy of the figure. In practical terms,
1st, ... , 4th, in Fig. 14.1, show the results of subsequent
users tend to adjust the concentration and thickness values
iterations. In detail, the iteration converged as follows:
manually, without doing the exact procedure shown nere. This
leads to a larger number of iterations. which most users find
First iteration. The height of the experimental La signal was
acceptable given that the calculations are very fast with modern
observed to be roughly 75% of the height of the simulated
computers.
signal. The same ratio for Ga was observed to be 80%.
The experimental La signal was observed to be roughly 2.4
The crucial point in the iterative procedure is to describe the
times wider than the simulated signal.
sample composition. Several codes let the user introduce
functions that describe the depth profile of a given element.
However, the most popular method to describe a sample is to
Reduce the Ga concentration to 80% x 0.33 "" 0.26.
divide it into sublayers. Each sublayer has a different
2
Increase the areal density of the layer to 2.4 x 200 x lOiS atJcm
composition, which is iteratively changed. Also, the number of
sublayers is increased or decreased during the iterative
= 480 x lOiS at/cm1.
procedure: If the current number of layers is not sufficient to
=> 1st iteration.
reproduce the changing signals observed, the user introduces a
new layer. If. on the other hand, the user can obtain an
Second iteration. The height of the experimental La signal was
equivalent fit by using fewer layers, this is normally preferred.
observed to be 85% of the height of the simulated signal. The
In practice, the depth resolution of the experiment limits thc
same ratio for Ga was observed to be 88%.
number of layers needed. This is illustrated in EXAMPLE 14.2.
The experimental signal was observed to be 1.2 times wider than
the simulated signal.
309
i:
l"'-. _
Chapter 14
of constant composition. The simulated spectrum as a result of

data analysis by the program GISA is illustrated as the thick
solid line connecting the experimental data points. The
theoretical signals of the elemental components in each of the
EXAMPLE 14.2. Simulating composition changes by division

into sublayers.
Changes of concentrations of components within a layer can be
treated by dividing the layer into sublayers of constant
subl.yers are shown as thin solid lines. The thick solid line thus
represents the sum of the thin lines of the elemental signals in
composition. Figure 14.2 shows. spectrum for 2.75 MeV 'Li

ions incident on a PbZrSnO thin-film sample on a Si substrate
the subl.yers.
(HaIjuoj. el a1., 2007). When ttying to simulate the spectrum,

one finds that the interface between the film and the substrate is
The experimental parameters, the concentrations of the elements,

and the areal densities of the sublayers are given in Table 14.1.
The concentratioIlS of the elements versus areal density are
not well defmed and that the concentrations change with depth
within the thin film.
shown in Fig. 14.3 .
. __ ....Jhis spectrum bas .Reen simulated by assuming nine sublayers

7000"i if:...
'-r>'-T--.-.-'-i'4--i-o-;.c,....oH--r-,
''''0
HiM
2~i50 MeV iU
PbZr(Sn)O/S(
2S[0"~t1M"""'"
2400
2000
Ph
1600
1400
1200
1lO0
'G;
1000
600
(5
z: "'0
,!,!
"
600
Si
"
:~ 600
" ..... <J ..
200
. '''''-cd+~'..
~If-""'T'"......"..,..=--='H
OiO ~o GI aD 'roo i2ii
FIG. 14.2. B.ckscattering
'.0 JO;<~",*~
"300
'.3.ii
';00 '410"
li<rl:' ~B1l'tIll 500
spec~~:;~~r:e~r7Li ions incident on. PbZrSnO thin'~=;I~~:!Si substrate. The thick solid f,'
line connecting the experimental data points is the simulated spectrum
None of the layers was assumed to contain any undetected
light elements. The oxygen signal was not fitted; the oxygen
concentration follows from the concentrations of the heavier
elements, as it is presumed to make up the total fractional
concentration of each sublayer to 1. Bohr straggling and a scaled
stopping of 1.08 times the TRIM-91 stopping value were
assumed for the metal oxide layers.
The example is intended to demonstrate how the composition

variations can be treated as constant composition sublayers. ~
. discussed in Section 14.3.4.6, the interdiffusion and interfacial
roughness can usually not be distinguished by using a
backscanering spectrometry. In the present case, the probable
cause for the mixing of the elements at the interface is diffusion.
as the atomic-layer deposition of the metal oxide film should not
change the smooth Si surface.
310
:~ ~
-~
.-
k i
:f
Data Analysis Software for Ion Beam AnalYSis
14.2.3 Interactive versus automated

The iterative procedure can be executed in an interactive way,
where the user repeatedly refines the depth profile until it is
com:idered that better agreement cannot be reached within the
time and patience limits of the user; or with an automated
process based on optimization of some target function such as a
X2 or likelihood function.
In any case, some fonn of automation is desirable,

particularly if large quantities of data are to be analyzed~
O,6
5'
..
.~0.20
Pb
115
---ZI
f!
[5
... _
C 0.4
Caution is needed when using automated optimization.

Different depth profiles can lead to the same simulated spectrum
(Butler, 1990), and the code has no way of knowing which of the
possible solutions is the correct one. The user must often restrict
the solution space, for instance, by pcrfonning only a local
optimization on a limited number of parameters, one at a time.
A second problem is that the
o.s .. '"_ .. ~ . . . . .
1.t
.,
:
gO.10
0.05
-I function is very sensitive to
100
200
300
400
Areal
the quality of the simulation: for instance, if straggling is not

included, an automated procedure will return interfacial mixing
layers that might not exist. This can be solved naturally by
including straggling, but there are some physical effects for
which no accurate analytical simulations have been developed so
far. in such cases, fully automated processes will lead to artifacts
in the depth profiles derived, unless the user restricts the solution
space to physical solutions.
500
600
701)" 8()0
density(1015at1.cm~
900
1060
FIG. 14.3. The constant concentrations of the elements (see

Table 14.1) versus areal density. Incident particle 7Li, Eo = 2750
keY, liE - 4.424 keY/channeL Q!l - 9.687 ~C mSI, scattering
angle 165, angle of incidence 60, exit angle 45, system
resolution 25 keY full width at half-maximum (fwbm).
Table 14.1. Experimental parameters and concentrations of the elements and areal densities of the sublayers.
Sublayer
lOIS at/em:!:
Pb
Sn
Zr
Si
100
0.16
0.0075
0.145
100
0.173
0.0128
0.163
0.6875
0.6512
100
0.168
0.0155
0.165
0.6515
0.1
0.1
0.611
4
-:5
0.17
50
--
-50
0.11
0.019
-0
0.022
0.31-='=- 0.558
50
0.036
0.0153
0.66
50
0.023
0.977
200
0.0015
0.9985
5000
14.2.4 Deterministic versus stochastic: (Monte Carlo)
;...~
~~--
...".
.. ~
----
0.2887
procedures descnbed above, and they are, in fact, the standard

approach.
Deterministic codes follow a calculational procedure that is

defmed a priori. In simulations, the samp1e is divided into thin
subl,yets. Both the incoming beam energy and the detected
energy for scattering on each element present are calculated for
each sublayer. The simulation is then constructed based on these
quantities and on the concentration of each element in each
sublayer. Straggling and other physical phenomena can be
included. However, individual collisions between the analyzing
ion (and its electrons) and the target nuclei and electrons are not
modeled.
Monte Carlo codes work in a completely different, more

fundamenta1, way. They model individual interactions between
particles. In this way, phenomena that have traditionally been
difficult to include in the standard method are included in a
natural way. The quality of the simulations that can be reached
is, in princip1e, much superior.
However, there are issues with the calculation times required.
Even with modern computers. acceleration techniques to speed
up the ca1culations must be used. These include, for instance,
disregarding collisions with scattering angles smaller or larger
These deterministic simulations are nonnally very fast,

lending themselves very well to the interactive and automated
311
--_._-----------,...--
'" ,,1
"":
Chapter 14
than values defined a priori and artificially increasing the

scattering cross section or the solid angle of the detector in order
to have more events. In some cases, such techniques can affect

the quality of the simulations.
A second issue is related to the fact that Monte Carlo codes

often require very detailed knowledge about the physics
involved, and so far, their use bas remained mostly confined to
the code authors. Until recently, almost all data analysis was
performed using deterministic codes (the largest exception was
the use of Monte Carlo codes for the analysis of channeling
data).
, Nevertheless, improvements of the calculation times, coupled
with the development of user-friendly interfaces, mean that
Monte Carlo codes are becoming more widespread and might
become dominant in areas such as heavy-ion ERDA, where
multiple scattering and special detection systems play a large
role.
14.3 CAPABILITIES OF CODES
In this section, we discuss some of the considerations that are
important when selecting an IDA code. We note that there is no
such thing as a "best code" for all situations. All codes discussed
here do well at what they were designed to do, as shown by an
intercomparison exercise organized by the lnlernational Atomic
Energy Agency (lAEA) (Barradas el al., 2007b). No code is
complete in the sense that it can analyze all possible IBA data.
The codes that took part in the IAEA intercomparison
exercise were compared numerically and validated. Some of
their most important characteristics are shown in Tables 14.214.8. Characteristics of several other codes are described in
Rauhala el al. (2006). We stress that all Information given was
valid at the time of writing, as several codes are being actively
developed.
'When choosing which code to use, users must consider
whether a given code is capable of extracting the information
required and how easy it is to do so with each code. A complex
sample where interface roughnes.s and multiple scattering playa
crucial role might need a last-generation code, but to determine
the thickness of a thin film with one single heavy element on a
light substrate, a full Monte Carlo calculation is probably too
much work without extra benefit. Also, in some cases. manual
calculations with careful consideration of error sources might be
the most accurate method (see, for instance, Boudreault et aI.,
2002).
14.3.1 Design basis
The fIrst consideration is obviously whether the data to be
data, but it might be possible to adapt it to analyze RBS with

heavier ions, particularly if the source code is available.
In a similar way, even if a code includes RBS, for example, it
might be limited to given experimental conditions. One example
is the detection geometry: some codes include not only the well-
mM and Cornell geometries, but also general geometry

where any location of the detector is allowed (including
transmission). Another consideration is the detection system,
with most codes including only energy-dispersive systems,
whereas some can include velocity and time-of-flight spectra,
magnetic spectrometers, and others.
known
14.3.1.1 Techniques implemented

As stated in Section 14.1.2, this chapter is aimed
predominantly at RBS and ERDA software packages. Elastic
backscattering spectrometry (EBS) is, in practice, RBS with
non-Rutherford cross sections that must be provided by the user
(or included in the code) for each given ion/target nucleus pair.
Nonresonant NRA is also included here because it also uses

one single beam energy (as opposed to resonant NRA, where the
energy is scanned), and it is common that both the scattered
beam and some reaction product are detected simultaneously.
Techniques such as channeling, PIXE, and resonant NRA are
included in some of the codes.
Table 14.6 lists the techniques implemented in each code. All
implement RBS, but some do not include ERDA or NRA. The
code NDF is the only one to include PIXE. Note, however, that
some laboratories perform only RBS, and it might be convenient
to use a code with fewer options (and thus possibly easier to use)
that can do what the user needs and nothing else. However, most
laboratories employ several techniques, and using the same code
to analyze all of the data has obvious advantages.
14.3.1.2 Experimental conditions supported
Tahle 14.6 also includes the experimental conditions
supported by each code. They are similar for all codes, which
handle most situations. Some differences exist, however, and
users who need special conditions should consider carefully the
vmying capabilities of the codes. Also, it might be possible to
convince authors of codes still under development to implement
some specific experimental conditions not previously covered. If
the source code is available, expert users can modify the
supported experimcntal conditions according to their
convenience.
., ,
l
14.3.1.3 Description of samples

'';' ,.
One of the most important points in codes is boW
conveniently the sample can be described. This is illustrated in
Tahle 14.8.
.;~ :~
'r
analyzed are within the design basis of the code considered. The
two main points are the techniques implemented by each code
and, for each technique, the experimental conditions supported.
;<~I'
.~
Homogeneous layers are considered in all codes that .' - ~-i

participated in the IAEA intercomparison exercise and in almost ~~~ ;}
~\.~
~.::$
l~
For instance, jf a code can only analyze RBS data with H and
He beams, there is no point in trying to use it to analyze NRA
312
~N
,-.l
''>'"
Data Analysis Software for [on Beam Analysis
all codes known to us, In some codes. functions describing

continuous profiles can also be introduced, which is important if,
for instance, diffusion of implant profiles are being analyzed,
6000
Sometimes, a maximum number of layers or of dements

within a given layer is imposed. When restrictions on the
number of layers exist, problems can arise when a very thin
layer description is required. This is the case for samples with
many thin layers and for high-resolution experiments (fOf
instance, using a magnetic spectrometer Of grazing-angle
geometry).
b) 45<
'10'
RBS spectra are ambiguous whenever more elements are

present than spectra were collected (Butler, 1990; Alkemade et
al., 1990). In that case, more than one different depth profile can
lead to exactly the same simulated spectrum. To remove
ambiguities, mOfe spectra should be collected, with different
experimental conditions. However, often, the elements are
bound in molecules, allowing the analyst to relate the respective
signals to each othcr; for instance, if SiOl is present, the small 0
signal can be related to the much larger Si signal by imposing
the known stoichiometry. One very useful function in a code is
therefore to allow the user to use as logical fitting elements. not
only atomic species, but molecules as well. The molecules can
be of known or unknown composition. In the latter case, such a
code would fit the molecular stoichiometry together with its
concentration in each layer. One example of such an approach is
discussed in Chapter 15, Pitfalls in laD Beam Analysis.
FIG 14.4. Data and fit (solid line) obtained for a 21-layer
ZrOz/SiO z antireflection coating with a top Au laycr, shown for
the data collected at (a) nonnal and (b) 45 0 incidence, The fitted
partial spectra for Si and 0 are also shown.
In Fig. 14.4, we show the RES spectra ofa 2l-layer

Zr02/SiOl antireflection coating with a top Au layer (Jeynes e/
al., 2000a). The Zr had a small (known) amount of Hf
contamination. Both nonnal incidence and a 45 angle of
incidence were used. The 0, Si, and Hf signals are small and
superimposed on the much '-larger Zr signal.- Il- is ,virtually l . '
impossible to analyze these data meaningfully without
introducing the chemical infonnation in some way.
Consequently, the two spectra were fitted simultaneously, with
the knO\VIl molecules imposed as logical fitting elements. The
small-5% misfits might be due to inaccurate stopping or small
deviations from stoichiometry. In any case, complete
disambiguation of the 'data is achieved.
14.3.2.1 Stopping power

The greatest advances in the knowledge of stopping power
achieved in the past decade were for heavy ions. This was
possible because of a wealth of new experimental data becoming
available, which was integrated, for example, into both the most
widely used stopping-power semiempirical interpolating scheme
used in lEA work, SRIM (Ziegler, 2004; Ziegler ef al., 2008),
and MST AR (paul and Sehinner, 200 I, 2002), a prognnn that
calculates stopping powers for heavy iODS. See Chapter 2 for
further information on comput~r calculations of stopping powers. :- _.
14.3.2 Databases implemented
. '
.,
. For in;i;mce. ill Fig: '14.5, ;;"e show the calculated spectrum
for a 3 MeV 'Li beam impinging normally on a SilSiO:z (2 x lO l8
atJcrn')fTi (1.5 x 10" at./em') sample and detected at .160', as
obtained using 1985 stopping-power values and SRfM version
2006.02. It is clear that the very large changes in the simulation
will be reflected in the accuracy of analyses of real data, with
errors of up to 10% being made in thickness values, depending
on the stopping values used.
Large changes also occurred for light ions in some common
systems. For instance, the stopping of He in Si is now known
with an accuracy of better than 2%, and this update has been
integrated into SRIM and other databases.
One of the main conclusions of the IAEA intercomparison

exercise of rnA softwares was that the greatest differences in the
results arise from differences in the fundamental databases used,
namely, the stopping-power database, as well as the nonRutherford database in the case of EBS, ERDA, and NRA.
Also, many experiments in insulating compounds have shawn

that the Bragg rule does not always reproduce the experimental
stopping, and the capability, of including molecular stopping
powers in codes is very important.
In the past decade, important advances have been made in

terms of both new and more accurate experimental data and new
theoretical and semiempirical schemes. Whlch databases are
included in each code, as well as the possibility of loading new
values, is thus an important factor.
The stopping power along crystalline directions is ~ually

smaller than that in a random direction or in an amorphous
material with the same composition.
313
Chapter 14
Ideally, codes should include the most recent stopping-power

databases available, as wen as the possibility of loading new
stopping values for particular systems. In any case, lhe users of
data analysis software must be aware of the stopping used in a
given analysis, along with its implications for the accuracy of the
results obtained.
143.2.2 Scattering cross section

When analyzing EBS or NRA data, or 'He ERDA
measurements of hydrogen isotopes, for example, the nonRutherford cross section used is of P8ramount importance. Some
codes, namely SIMNRA, include a wide database of crosssection values for given ion/target pairs, whereas others read in
external data files with the required cross sections.
200
The code SigmaCalc (Gurbich, 1997, 1998, 2005) for

calculating cross sections from tabulated theoretical model
parameters was developed by A. Gurbich. Sigma Calc is usually
very reliable for those ion/target pairs that it includes, and it
allows one to interpolate for scattering angles for which no
measured cross sections are available. This action has also led to
the integration of NRABASE and SIGMABASE into a single
database called ffiANDL (Ion Beam Analysis Nuclear Data
Library), under the auspices of the IAEA.
3 MeV 7U on Si/SiO 2x1018at/cm'lTi 1, 5x10 18 at/em'

--ZBL85
- - - SRIM 2006-02
I
I_lOin SiO,
150
III
'2
:::l
.ri
~
ttl
100
"iii
Ti
Si bulk
---
>=
50
a
500
1000
1500
E (keV)
FIG. 1'4.5. Calculated spectra for 3 MeV 1Li beam incident on a SilSi0 2 sample. The onJy difference between the simulations is that
different stopping powers were used: values from 1985 (solid line) and the most recently available values (dashed line). for accurate
analyses, users must be aware of the stopping powers used by data analysis codes.
Data with slowly changing non-Rutherford cross sections can

be analyzed manually, as shown, for instance, in Chapter 5,
section 5.3.4 of the first edition of this handbook (Tesmer and
Nastasi, 1995) for 'He ERDA aoalysis of hydrogen in Si,N,(H).

However, this is a painstaking procedure that can be done much
faster, and more accurately, by computer simulation.
314
~I
I
I
I
I
I
I
I
I
Data Analysis
Softw~re
When superimposed signals with fast-changing cross sections

are present, it becomes almost impossible to analyze the data
manually, and computer codes must be used, along with the
appropriate cross sections. For example, in Fig. 14.6, we show
1.75 MeV 'H backscattcring off a TaNiC film on Si (Jeynes et
al., 2000b). The C signal is superunposed on the Si background,
and both signals change rapidly because of the presence of

resonances in both cross sections. The simulation includes one
single homogeneous TaNiC layer and is the result of an
automated fit that took less than 1 min. The scattering cross
sections for C and Si were taken from SigmaCalc, and those for
Ta and Ni are screened Rutherford cross sections.

14.3.3 Basic physics
All IBA codes (except the Monte Carlo ones) employ similar
principles to perfonn a basic simulation. The ingoing beam
follows a straight trajectory, while losing energy, toward the
sample; it interacts with a target nucleus; and the outgoing beam
for Ion Beam Analysis
follows a straight trajectory, while losing energy. on its way to

the detector. The energy spread of the beam (e>ften called
straggling) is also calculated.
From the point of view of the user, the fact that (he different
codes implement these steps in very different ways is not
important.
The user wants to know, first, that all codes perform good
simulations (at least for simple everyday spectra; complex
samples might require particular codes). Confmning that a code
is correct would imply a deep analysis of the physics included
and of its implementation in the algorithms employed. Users do
not want to do this, and often, they cannot (either because the
source code is not available or because they do not ba ve the time
or the necessary background). Until recently, users simply
assumed that codes were correct.
I
I
I
r
I
I
.75 MeV lH on Si/TaNIC

20000
!1
15000
Single layer fitled:

Ta 16.8 a1.%
Ni 72.1 at.%.
C 11.1 at.%
18
6.59x10 at.lcm 2
.e-!!.
::Q 10000
~
;;:
Ni
.~."\
I
I
5000
Ta
o~____~__~/__~__~~__~__~-.__~__~'~__~__-L~
1000
1500
1250
1750
E (keV)
FIG. 14.6. Spectrum for 1.75 MeV 'H backscattering off a TaNiC film on Si.
1%) and much smaller than the error in the stopping-power
values.
The lAEA intercomparison exercise showed that all of the

analyzed codes produced similar results for a 1.5 MeV 'He RBS
simulation of a simple sample. Signal total yields and signal
heights were calculated with a 0.1-0.2% standard deviation
between codes (as long as the same stopping powers and
scattering cross sections were used in all cases). This is thus the
error due to the differences in implementation of the same
physics. It is smaller than the typical experimental errors
coming, for example, from counting statistics (rarely better than
With the assurance that the codes (at least those that
participated in the IAEA intercomparison) are correct for simple
cases t the user might be concerned with how fast the simulations
are. With modem computers, the simulation of simple spectra,
even with straggling included, is always fast. Efficiency
becomes important only when some further

plural scattering, is included.
315
physics~
such as
Chapter 14
143.4 Advanced physics

Although all simulation codes treat basic ion stopping and
scattering phenomena, many of the subtle features in spectra
arise from more complex interactions. The main issues involve
energy spread, multiple and plural scattering, the handling of
State-of-the-art calculations of straggling, including

geometrical straggling and multiple scattering~ are performed by
the code DEPTII (Szilagyi ., al., 1995; Szilagyi, 2000). The
recently presented RESOLNRA code (Mayer 2008) leads to
equivalent results.
non-Rutherford cross sectlons, surface and interface roughness,
No code dedicated to general-pwpose rnA data analysis

implements accurate straggling functions such as the LandauVavilov function. See, for instance, Bichsel (2006) and
references therein for a detailed discussion of this issue.
and channeling.
The first comment is that, in principle, the best way of
dealing with all of these phenomena is full Monte Carlo
simulations. MCERD, for instance, is a Monte Carlo code for the
analysis of ERDA data, but it can also handle RBS. It can
include all of these effects, except for channeling. Different
"---~codes have been 'dedicated to the analysis of channeling in
specific systems.
However, Me codes are mostly used by their authors, and
calculation times can be an issue. Development of intuitive user
interfaces and continuing gains in efficiency might change this
situation, but for the moment, traditional codes are still the most
often used. The following discussion is therefore related to
traditional codes
The codes NDF and SlMNRA implement or are developing

asymmetric straggling functions that might not be physically
correct but that do represent reality better than the usual
Gaussian distributions.
14.3.4.2 Elect.-on screening
Electron screening at low energies decreases the effective
charge of the target nucleus seen by the analyzing ion, leading to
a smaller scattering cross section (L'Ecuyer el al., 1979;
Andersen et al.. 1980).1t is greater for heavier target species and
heavier ions, as well as for lower energies. For 1.5 MeV 'He
RBS off gold, it leads to a 2% correction to the cross section.
Most codes implement this effect.
We are concerned with determining the situations in which

the basic simulations discussed in Section 14.3.3 are adequate.
In situations where they are not adequate. we want to know
whether codes include enough further pbysics to lead to
meaningful analysis. The capabilities of the different codes are
summarized in Table 14.7.
14.3.4.1 StraggUng models

Practically all of the rnA codes implement the Bohr model
(Bohr, 1948); with or without the Cbu/Yang correction (Cbu,
1976; Yang el a/., 1991), important for He and heavier beams;
some also implement the Tschalar effect (TschaHir, 1968a,
1968b; TschaUir and Maccabee, 1970), which is important in
thick layers (for instance, to calculate the energy spread in a
buried layer or after a stopper foil). Some codes let the user scale
the ealculated straggling. We note that pure Bohr straggling
(without the Tschalar effect) is valid in a rather narrow range of
energy loss, typically when the energy loss does not exceed
around 20010 of the initial beam energy.
However, straggling is important only if the detailed signal
carnes relevant infonnation. To determine the
stoichiometry and thickness of films, straggling is very often
irrelevant, and the user can either ignore it or adjust the
straggling to match observed signal widths. In this case, all
codes are adequate,
shape
14.3.4.3 Plural scattering

Plural scattering involves a few large-angle scattering events.
1t leads to a low-energy background, to an increased yield of
bulk signals at low energies, and potentially to yield above the
surface energy. Plural scattering can normally be disregarded,
even when it is sizable, because it changes slowly with energy,
and ad hoc backgrounds or signal subtraction techniques are
often used.
Plural scattering is important on the rare occasions where the

yield at low energies or in a given background region must be
understood quantitatively. Also, at grazing angle (of incidence or
detection), it can make a large contribution even to the signal of
fairly thin surface films. Both SIMNRA (Eckstein aDd Mayer,
1999) and NDF (Barradas, 2004) can calculate double scattering
(two large-angle cvents. which accounts for most plural
scattering). The two algorithms are not equivalent, but they lead
to similar results in most situations. 1n some cases, such as in
grazing-angle geometries, large differences might be observed,
so the user should study the respective literature. Calculations
are slow and not as accurate as for single scattering, but
excelle_nt reproduction of observed signals can be obtained, as
seen in Fig. 14.7 (Mayer, 2002) for a normal incidence
spectrum.
If, on the other hand, interdiffusion or roughness between

layers is being studied, or if a changing depth profile must be
derived with maximum achievable accuracy, then any error in
the straggling will directly lead to an error in the results. In such
situations, the best available theory (i.e., a code that implements
it) must be used.
14.3.4.4 Multiple scattering

Multiple scattering involves a large number of low-angle
scattering events. It changes the trajectories of the beam ions,
making them quite different from straight lines. This has many
consequences. some of which standard code.... are unable to
316
At grazing angles (for instance, in ERDA), multiple

scattering is often the most important contribution toQ energ
spread. If energy spread is important (see Section 14.3.4.1), then
multiple scattering must be ca1culated.
calculate. It is therefore critical to know when it is important and

when its effects can be calculated adequately.
Multiple scattering leads to aD extra contribution to the
energy spread of the beam, a change in the shape of signals
because this contribution is not Gaussian-shaped, and a change
in the observed yield.
The best available calculations of multiple scattering (but not

necessarily the best simulation of energy spectra) are made by :
the code DEPTH (Szilagyi ef al., 1995). Some codes implement
the extra energy spread, but assume it to be GallS:si~shaped.
Most codes do not calculate multiple scattering. For heavy ions
at fairly low velocities in grazing-angle geometries, none of the
standard codes compares very well with Mon"te Carlo
calculations, as of the time this was written.
It is more important for heavier ions at lower energies in

heavier targets, as well as for grazing angles. In fact, at nearnormal incidence, for the detection of ions with energies in the
typical IBA range, multiple scattering can almost always be
ignored.
400
Experiment al
...... Smooth + plural scattering
_._._. Rou:Jh
- - ROllilh + plural scattering
300
~:l
o 200
U .....
300
400
500
600
700
800
900
Channel
FIG. 14.7. 2.5 MeV protons back.scattered from a 3.5-~ W layer on a rough carbon substrate, at normal incidence and with a
scattering angle 165'. (CiIcles) Experimeotal data, (dotted line) calculated spectrum for a smooth W layer (3.6 l'Il') on a ,mooth C
substrate including plural scattering, (dashed line) calculated spectrum for a rough W layer (3.5 Jlm, cr = 0.30~) on a rough substrate
(fwhm 20'); (,olid line) ,ame as d""hed line hut including plural ,cattering. (From Mayer, 2002.).
317
<..\~~. ..
~"
Chapter 14
14.3.4.5 Simulation of resonances

It is quite diffieult to simulate buried sharp resonances. The
combination of "booed" and -"sbarp" means whenever the
energy spread of the beam before scattering is comparable to the
resonance width. For bulk signals, the shape of the resonance
signal is considerably affected; for thin films, the actual yield is
also affected. This is important whenever a sharp resonance is
used to enhance the yield from a buried element.
For instance, for protons crossing a 4.2-JlID- Ni film (leacting
to a high energy spread) on top of a l-f1IIl Mylar film, the
calculated C yield frOID the Mylar can be wrong by a factor of 10
if the energy spread of the beam is not taken into account (see'
Fig. lOaf Barradas el 01., 2006).
To the best of our knowledge, the only code for analysis of

general RBS and ERDA data that is currently able to calculate
correctly the shape and yield in the presence of buried sharp
resonances is NDF (Barradas er al., 2006). Dedicated codes for
specific systems (Tosaki el aI., 2000) and for resonant NRA
(Vickridge and Arnsel, 1990; Pezzi el 01., 2005) also exist.
14.3.4.6 Surface and interface roughness
There is no general definition of roughness. Many different
types of roughness exist. In the context of IBA, the most
important point is that, in general, RBS experiments are not
sufficient to determine which type of fOuglmess is present. In
fact, they are not even capable of distinguishing between layer
interdiffusion and interfacial rou~ess. In almost all cases,
extra information from other techniques [usually transmission
electron microscopy (TEM) or atomic force microscopy (AFM)]
is requiTed. That said, RBS can be very useful for quantifying
roughness in some cases.
Several methods exist to simuJate spectra including
roughness. The most accurate is to perform a Monte Carlo
simulation, modeling the d~ired surface or interface. In general,
roughness is not amenable to routine data analysis.
A typical approach in analytic codes is to sum partial spectra

over a distribution function of some sample characteristics, such
as surface height or film thickness. RUMP and SIMNRA
implement different variations of this method.
A second approach is to calculate analytically the effect on
signal width due to a given type of roughness and to take the
result as an extra contribution to energy spread. in this approach,
spectra can be calculated in the usual way. NDF implements this
method, which has stringent but well-defmed conditions of
applicability, for different types of roughness and also for
inclusions, voids, and quantum dots. Both approaches ignore
correlation effects and fail for features with large aspect ratios
leading to re-entrant beams.
An analysis of a rough sample performed using SlMNRA is

shown in Fig. 14.7 (Mayer. 2002), where plural scattering also
318
plays an impor1ant role. The sample is a rough W layer on a

rough C substrate, which is a very complex situation.
Quantitative information on the roughness parameters is
obtained in a relatively simple way, as the user needs only to
specify the type and amount of rouglmess and the code takes
care of all of the calculations.
A new algorithm to take correlation effects and re-entrant

beams into account was recently published (Molodtsov e/ aI.,
2009). It is, strictly speaking, for the published fonnulation,
valid only for single-layer targets, but once it js included in data
analysis codes, it will expand the range of samples that can be
meaningfully analyzed.
14..1.4.7 Channeling
Specific methods have been developed to analyze channeling
data. Most standard analytic codes are simply not capable of
analyzing channeling. Some implement ad hoc corrections to the
channeled yield in order to be able to derive information from
unchanneled parts of the spectrum. Only one of the standard
codes that participated in the lAEA intercomparison exercise,
RBX, is actually capable of simulating the channeled RES
spectra of virgin crystals, even with point defect diStributions.
14.3.4.8 Pulse pileup
Pileup leads to some counts being lost and some counts being
gained in the collected spectrum. Without exception, more
counts are lost than gained, and so the total yield, integrated over
the entire energy range, is smaller than it would be for a perfect
detection system.
Some codes calculate pileup based on system parameters
such as amplifier type, amplifier shaping time, beam current,
collection time. and characteristic time of a pileup rejection
circuit ifpresent (Wielopolski and Gardner, 1976, 1977; Gardner
and Wielopolski, 1977; Moladtsov and Gurbich, 2009). This can
be important in some cases, particularly given that pileup is not
linear and can be a correction of severa1 percentage points to the
yield. It should always be calculated. except for very low count
rates. Chapter ]5, Pitfalls in Ion Beam Analysis, describes an
example.
14.3.5 Automated optimization
Practically all codes can work interactively. The user
examines the data, makes an initial guess of the sample
structure, calculates the corresponding theoretical spectrum, and
compares it to the data. Differences drive modifications in the
defined structure until the user considers the agreement to be
adequate (or, rather often, until patience runs out).
In automated optimization, it is the computer code that

controls this procedure. Users should be aware that an objective
criterion of goodness of fit is needed, usually a X2 or likelihood
function. It is this functioIl that the codes try to optimize. This
can lead. to a serious pitfall. If the model does not describe all of
the relevant physics (for instance, if multiple scattering is
important but n01 calculated or calculated poorly) or if the user

introduces too few or too many elements in the sample
descriptioD, a good fit that also corresponds to the true sample:
structure cannot be obtained. The code will output something,
but what it outputs will not be a good solution. We cannot
overemphasize the fact that automated optimization is a
wonderful feature, but it can lead. to substantial errors unless
great care is taken in checking the results.
If several spectra were collected, the possibility of fitting

them all simultaneously, with the same sample structure, is
essential to ensure that the results are consistent and that all of
the information present in the spectra is taken i:nto account to
generate the final result
14.3.5.1 Fitting
Automated fitting is designed. to relieve the user from this
task. Not all codes do it. Some, such as RUMP and SIMNRA,
perfonn a local optimization on a limited number of parameters,
starting from an initial guess that usually must include the
correct number of layers.
NDF uses an advanced algorithm, simulated annealing
(Kirkpatrik et aI., 1983), complemented by a local search, to find
an optimum solution without any need for an initial guess,
except for the elements present. All parameters can be fitted.
perform a BI run than to do a least-squares fit. For the :moment,

BI is still not suitable for the routine analysis of large amounts of
data. A few samples per day can be analyzed, however, and BI
the only method that can provide reliable error bars in a general
way.
is
14.3.6 Usability and usage

Usability of a code refers to how easy it is to use the code
correctly. This depends on how intuitive and simple the user
interface is, but also on how much knowledge the user bas about
the technique used. Nonexpert users might not know that scaling
a given stopping power by a factor of 2 might not be jUstifiable
or that intermixing and surface roughness can have the same
effect on measured data. Many other examples can easily be
found by carefully reading published work, where mac curate or
wrong results due to bad data analysis are unfortunately not
uncommon.
The easiest code to use is often considered to be RUl\1P,
whether in its command-line or windowed incarnations, and no
chapter on IBA software could go without citing its "beam me
up Scottie" command line (see EXAMPLE 14.3). Firstgeneration codes such as GISA already included strong
graphical capabilities and easy-to-navigate menus. RBX allows
quick editing of parameters for analysis of multiple spectra.
The new-generation codes have more user options and, as a

consequence, are perhaps not as intuitive. To use advanced
physics features, one often must have knowledge of the issues
involved and of an increased number of system parameters. As a
consequence, usage (particularly by novices) is more prone to
errolS. SIMNRA is considered easier to use than NDF. Their
manuals state they are "easy to use". (Mayer, 2007) and
. "designed to be a powerful tool for expenenced analysts of Ion
In the past decade, Bayesian inference (BI) methods have - - Beam Analysis data" (Jeynes el al., 2000<:). respectively. MC
been applied to IBA data analysis, providing a tool to
codes such as MCERD are not yet widely used except by the
systematicaUy determine confidence limits on the results
authors and their collaborators. TIlls might change soon as one
obtained (Barradas el al., 1999; Padayachee el al. 2001;
of the MC codes, Corteo, now has a user-friendly windowed
Neumaier e/ ai., 2001; Mayer at al., 2005; and Edelmann e/ aI.,
interface (Schiettekatte, 2008).
2005). The mathematics is somewhat involved, but the users of
The IAEA intercomparison exercise originally planned to test
codes do not need to be concerned with the details.
usability by novices and expert users, but this has DOt yet been
In practical terms, instead of producing one best fit, for
done.
instance, millimlzing the X2 BI performs a series of simulations,
RUMP is historically the mosH:ited IBA data analysis code.
for very many different sample structures and depth proflles, all
with around 120--150 citations per year in the past decade.
of which are coosistent with the data. Statistical moments. such
SI.l\1NRA came a close second in 2007, and if the trend
as the average and standard deviation, can then be calculated
continues, it will overtake RUMP as this book is being
published. NDF, the third most cited code, receives about oneDuring the BI calculations, the known experimental errors,
half the citations that SIMNRA does. SIMNRA and NDF
including those in the energy calibration, solid angle, beam
combined now account for more citations than RUMP does.
fluence, and even beam energy or scattering angle, can be
introduced. These errors, together with the statistical counting . Citations might be misleading when trying to assess real use of a
code, however, as users might not always cite the code they we.
error, are then reflected in the :final error calculated for the depth
profile.
14.3.5.2 Bayesian inference

IBA is often fully quantitative without needing standards.
This is arguably its greatest strength. However, actual error bars
or confidence limits on the concentrations, thickness values, and
depth profiles determined are almost never presented or
published. This is arguably the greatest weakness ofIBA work.
Tbe codes MCERD and NDF both implement BI. It is

computationally expensive; that is, it takes much longer to
319
.. -
--------'-------------....
., ,
.,
,~
Chapter 14
14.4 ACCURACY
The accuracy, that can be achieved in an IBA experiment
depends on many parameters. For each given ex:periment, an

uncertainty budget can be made, including all of the different
sources of error. A detailed example ofuncertai.nty estimation is
provided in Chapter 15, Pitfalls in Ion Beam Analysis.
Data analysis codes cannot lead to results that are more

accurate than the models applied and experimental data
analyzed. Furthermore, most of the codes usually do not produce
an error analysis, just numbers without associated accuracies.

The consequence is that it is almost universal practice to quote
the results provided by the codes, such as concentrations, layer
.. ~--,," thicknesses; or depth profiles, without quoting the associated
uncertainties. Given that IBA techniques such as RBS and
ERDA are inherently quantitative, this is not really justifiable.
Bayesian inference can be used as a means of error analysis
(Barradas ef 01.. 1999), but it is Dot implemented in most codes,
and it is computationaUy intensive. The uncertainty budget is a
valid alternative, but it requires detailed knowledge 'that many
users do not have.
14.4.1 Numerical accuracy of codes

The 'codes themselves have an associated accuracy. For
exactly the same sample structure and experimental conditions
and including exactly the same physics, no two codes will
calculate numerically the same theoretical spectrum. Differences
in implementations of the physics and algorithms and even in
floating-point representation do lead to differences in the
calculated spectra and, thus. to different fmal results of the data
analysis. The issue is how large these differences are.
The codes that took part in the IAEA intercomparison
exercise (Barradas et al., 2007b) were compared numerically and
validated. For ~He RBS, differences of uf to 0.2% were found
For 7Li RBS, this increased to 0.7%. For He ERDA, differences
between 0.5% and 1.3% were found in calculated yield values.
This is a further source of inaccuracy that should be induded in
the uncertainty budget We note that experimental uncertainties
down to about 0.5% are difficult to reach, but have been reported
(Jeynes et al., 2006; Barradas et 01., 2007b). This means tha~ at
least for ERDA and heavy-ion RBS, the accuracy of the
calculations can be an issue.
basic physical quantities, such as cross sections or stopping

powers the accuracy with which the experimental parameters are
known; and the counting statistics.
14.4.2.1 Models and basic physical quantities
Many different models are used in IBA data analysis to
describe the various physical phenomena involved. The accuracy
of these models is often difficult to assess. Examples are
stopping and scattering cross sections, plural and multiple
scattering, screening, the energy spread, the detector response,
pileup, and so on. A few examples are briefly considered in this
section.
The accuracy of scattering cross sections is limited at high

energies by the occurrence of elastic nuclear reactions. In this
case, experimental cross sections or cross sections evaluated
with Duclear models must be used (see, e.g., IBANDL). The
errors involved depend on the error of the cross-sectiOD
measurements. Uscrs interested in accwacy must check the
original literature for each reaction, but accuracies better than 12% are almost never achieved, and 10% is common. Given that
many cross sections have a strong angular dependence, the
uncertainty in the scattering angle leads to an even larger error
that is difficult to evaluate. Nuclear models can predict
scattering at angles where experimental data are Dot available,
and SigmaCalc should be used for all reactions where it is
available at the required energy iange.
I'
In ~e low-energy limit, electron screening becomes the issue.

For 1 MeV 4He+ backscattered off Ta, the cross section is "
already 3% smallcr than the Rutherford fonnula (Raubala,
1987). For 0.3 MeV 'He' forward-scattered off Au at a 15' " .
angle, the correction is about 33% (Andersen et 01., 1980), ~~
However, in both cases, this factor can be caLculate~ for
instance, by the angular-dependent screening correction given by
Andersen et al. (1980), which leads to cross sections accurate t
within I % for scattering angles above 15" and He+ energies ~~;
above OJ keY. For the typical velocities in RBS, the accuracy is
not better than 0.2% for heavy target elements such as Au.
,t "
The limited accuracy of stopping powers is a fundamental
]~:
limit to the accuracy with which layer thicknesses and depth A"
profiles can be determined. A statistical analysis by Ziegler
-1 fi,
:., ., ';,._-: .:~.
""
However, if we consider only the codes RUMP, NDF, and

SIMNRA, which together account for over 80% of the citations
in work publJshed in 2006 and 2007. then the differences are
0.1% for 4He RBS. 0.3% for 7Li RBS, and 0.1% for 4He ERDA.
These results are Significantly better than the achievable
experimental uncertainty and much better than the uncertainty
with which stopping powers are cw-rently known.
showed that the average standard deviation of SRIM-2003

stopping-power calculations relative to experimental values is
4.2% and 4.1% for H and He ions, respectively, and 5.1% and
6.1 % for Li and heavier ions, respectively (Ziegler, 2004). For H
and He ions, 25% of the calculations are off by more than 5%
relative to the experimental values. For heavy ions, this increases
to 42% of all SRIM. stopping powers, and 18% of the
calculations having an error larger than 10%. Very few stopping
powers are known with an accuracy better than 2%.
14.4.2 Intrinsic Rccuracy ofIBA experiments

The intrinsic accuracy of rnA experiments is limited by three
main factors: the accuracy of the models applied and the related
For light compounds, particularly insulators, the Bragg rule }~)

can lead to large errors (above 10% at the stopping peak).
:'j. ~J
;:~ '>
-~ '"
:i~,
,,'~_:
~'~ ~~
~ :J
1~g"
:E~
~~~
320
~l
One of the sources of the low-energy background is double

scattering. If an accw-ate low-energy calculation is important,
then use a code that can calculate double scattering.
However, other sources, such as slit scattering, also lead to
. similar backgrounds and might be impossible to calculale.
Calculate pileup whenever the count rate is not very low or ..
14.4.2.2 Experimental conditions

The experimental conditions, such as beam energy, incidence
and scattering angles, beam fluence, solid angle of the detector,
multichannel analyzer (MeA) energy calibration, and others, all
have an associated accuracy, which is often not known. The
reader is referred to Chapter 15, Pitfalls in Ion Beam Analysis,
whenever a high-energy background is observed.

All contributions to energy spread, including geometrical
straggling and multiple scattering, must be calculated
whenever the broadness of signals is relevant to the analysis.
Calculate the effect of surface and interface roughness
whenever relevant; very often, surface roughness i.s ignored
and falsely assigned to layer intermixing.
Read carefully Section 14.3.4 (Advanced physics) and
Section 14.6.1 (Further capabilities), some of the issues
discussed might be relevant to the analysis of your data.
Read carefully Chapter 15 of this handbook, Pitfalls in Ion
Beam Analysis, particularly the sections on accuracy and the
error budget.
for a further discussion of this issue.

14.4.2.3 Counting statistics
Experiments cannot last indefinitely. Often, a limited amount
of beam time is allocated for a given number of samples that
must all be measured. More generally, counting statistics is
limited by the damage caused to lhe sample by the beam. In any
case, Poisson statistics (often approximated as Gaussian
statistics) is well understood and is used in most error analyses
that are actually publishcd. One common example-is to calculate
the error of an isolated signal as the square root of the integrated
yield. This leads to an underestimation of the true accuracy,
which is often much worse than what would be granted only by
statistics.
14.4.3 Physical effects on data analysis
The effects that some physical phenomena have on data
analysis are described in Table 14.9. Where possible, the
resulting errors were quantified. Note that the error values given
are indicative only, and a detailed error analysis must be made
on a casc-by-case basis.
14.5 RUMP AND FIRST -GENERATION CODES

In 1976, Ziegler and co-workers published the first code,
called IDA, that performed full simulations of RES spectra
(Ziegler et al., 1976). In the next decade, a series of codes
surfaced, the most popular of which was RUMP (Doolittle,
.1985).
Many versions of RUMP are available, induding some very

early ones. Users are cautioned that they should regularly
download the new version (along with the companion code
Genplot): RUMP is being actively developed, and new versions
Always include electron screening in the calculations. The
angular dependence of screening is important at small
incorporate such updates as recent stopping powers, heavy ions,
scattering angles and low energies, so it should be included... and improved algori~.
[in practice, this means Andersen e/ af. (1980) screening].
RUMP is very reliable for whatit is designed to do.
f~~t.
For proton beams at any energy and 4He beams at energies as
together with its being easy to learn and use, is the main reason
low as 2.0 MeV, always check the literature arid the IBANDL
for its popularity. Novicc uscrs can obtain accurate results from
database for possible nuclear reactions and the corresponding
simple spectra without too much work. In the RUMP example
cross sections.
(EXAMPLE 14.3), we show how information can be extracted
Always use the most recent stopping-power data available. In
from a fairly complex spectrum by an expert user. This is one of
some cases, particularly for heavy ions, this can imply using
the samples that was analyzed in the IAEA intercomparison
literature values and not one of the popular interpolative
exercise of IBA s.oftware.
schemes.
Be aware of the accuracy with which you know the
experimental parameters. The nominal values cannot be taken
EXAMPLE 14.3, RUMP.
for granted unless a strong effort has been made to determine
In tbis example, sample E3 from the IAEA intercomparison
them.
exercise of IBA software (Barradas et al., 2007b) is used. It is a
hafniwn oxide layer with impurities on a Si substrate, analysed
by RBS with a 2.5 MeV 'He+ beam. The fIrst step is to load the
Furthennore, depending on the experiment at hand, the
data and define the experimental conditions. This is done in
following points might need to be included in the analysis:
RUMP by using an edit window or by commands given in the
best knO\VD. screen in IBA (Fig. 14.8).
With this table in mind, the following <!dvice can be given:
TIili
321
Chapter 14
FIG. 14.8. First step of a RUMP analysis: load the data and define the experimental conditions.
322
O~ta
Analysis Software for Ion Beam Analysis
The second step is often to configure some simulation parameters using line commands. In this case. SRIM-2006 stopping powers are
used. the Don-Rutherford scattering cross section for the 160(a,a)160 is loaded, and straggling and multiple scattering are enabled. By,
default, all isotopes are calculated separately (as opposed to making one :lingle calculation for the average mass of the target element),
and Andersen electron sCT7ening (Andersen et oZ., 1980) is included.
.
FIG. 14.9. Second step ofa RUMP analysis: Configure ;imulation parameters.
The third step is to defme the sample structure. This is done using inruitive, easy-to-Iearn commands. The corresponding first guess is
then calculated.
323
Chapter 14
Energy (MeV)
'\
\~..-.
-u
..........
__ _____ -:o
~
O~~~-r~~~~~~~~~~
500
1000
1500
FIG. 14.10. Third step ofa RUMP aoalysis: define the sample structure, perform the simulation.
324
2000
Data Analysis
Softwar~
for Ion Beam Analysis
A manual iterative procedure as described in EXAMPLE 14.1: would follow the first guess. Alternatively, RU1vlP inc ludes an
optimization algorithm. The window shows the definition of the fitting space and the final results. In most situations, [lot all -variables
would be fitted at the same time. A combination of manual iteratio.Ds and fitting of some parameters is typical.
.:...~
"
'";.':
.f,
.'
Edqyo.1cV}
OJ
1.0
Jl
"L'
....
_0,
__ .
~_..-~
_ _ --...... _e-.
--.,
,
,,
,.,
,,..
'"''
'''''
FIG. 14.l1. Final result of a RUMP analysis after manual or: automated iteration until agreement between simulation and fit is
achieved.
Chapter 14
However, the rnA samples that need to be analyzed are

becoming mcreasingly complex, making stronger demands on
data analysis. Advanced physical effects such as roughness,
quantum dots and inclusions, and multiple and plural scatlering,
as well as slmultaneous use of multiple techniques or different
experimental conditions, need to be included in routine data
Both codes have some roughness capabilities. SIMNRA

calculates very many different trajectories for actual surfaces, as
dermed by the user. This approach is accurate, and slow, and it
requires detailed knowledge of the sample surface. NDF uses
models for a few types of roughness, as well as for inc1usions
and quantum dots. This approach is fast and requires less
knowledge by the user. but it is not as accurate_ In practice, the
SIMNRA approach is best for detailed studies of a few very
important samples, whereas the NDF approach is adequate for
routine analysis ofroughness..parameters_
analysis.
RU1v[P lacks many of these capabilities, which is probably
the reason why new-generation codes such as SIMNRA arc

overtaking it.
SfMNRA includes RBS, ERDA, and nonresonantNRA for

any ion detected at any angle (including forward scattering and
transmission) and for any target. NDF includes those techniques,
as well as PIXE (pascual-Izarra, 2006a) and resonant NRA
Other popular first-generation codes, used by differenfgroups
(as opposed to the many codes that are used mostly by their
author only), are elSA (Raubal., 1984; Saarilabti and RaubaIa,
1992) and REX (K6tai, 1994), which places strong emphasis on
working with sets of spectra from the same sample.
SIMNRA offers some limited support for multiple spectra

collected from the same sample. The user must write a script or
small program and use OLE (object linking and embedding)
autOmation, supported by SlMNRA. NDF is designed for
analyzing any number of ~'Pectra from the same sample.- The same depth profile is used to fit all of the spectra simultaneously_
There are no restrictions on using data from different techniques,
and all complementary information is integrated into the final
14,6 NEW-GENERATION CODES: SIMNRA AND NDF

Several IBA data analysis codes were developed in the 19905.
Of these, SlMNRA (Mayer, 1997) and NDF (also known as
DataFumace) (Jeynes el al., 2003) placed strong emphasis on
implementing advanced physics and experimental conditions.
Another code, DEPTH (Szilagyi et 01.,1995), was developed
specifically to make state-of-the-art calculations of depth
resolution. It is not geared toward the analysis of data, but it (or
a different implementation of the same algorithms) should be
used whenever depth resolution is importanl
result
Barradas and co-workers publisbed algorithms to improve the
simulation of buried .resonances (Barradas et al., 2006), double
scattering at grazing angles (Barradas, 2004), the yield at very
low energies (Barradas. 2007). and the shape of surface signals
in high-resolution experiments (Barradas e/ aI., 2007a). As of
the time of publication of this handbook, these effects are
included only in NDF, but other codes can easily include them
14.6.1 Further capabilities

The reader is referred to Section 14.3 for information about
when the effects mentioned in this section are important.
~~
SlMNRA and NDF both implement plural and multiple

scattering effects to some extent Multiple scattering is included
only insofar as it affects the depth resolution.
~,
~,;
\.
.
~f'J.
14.6.2 SlMNRA
.Ii'
Development of SIMNRA started in 1996. It is a Windows l'"
code with intuitive menus and windows.
SIMNRA includes its O\VD. algorithms to calculate the depth

resolution. NDF uses a different approach. which is to call
DEPTH in run time (NDF writes the required input files
automatically) and use the DEPTH results. This has the
edvantage that DEPTII is the state of the art in terms of depth
resolution and is still being developed and update~ and it bas
the disadvantage that the user must also have DEPTIi (which is
free of charge) installed.
.~,
,.1
We show in the SlMNRA example (EXAMPLE 14.4) exactly ,

the same spectrum as was used for the RUMP example 1.
(EXAMPLE 14.3). ]t is easy to foIlow how a rather different l'
procedure leads to equivalent results.
~.
326
EXAMPLE 14.4. SlMNRA

In this example, sample E3 from the IAEA intercomparison exercise of rnA software (Barradas et al., 2007b) is used. It is a bafnium
ox:ide layer with impurities on a Si substrate, analysed by RBS with a 2.5 MeV 'He+ beam. The fIrst step of the analysis, after loading
in the data, is to defme the experimental conditions. This is done in an intuitive windows called from the Setup menu. In the example
shown, the real and live times are used to apply a live-time correction and to calculate the pileup contribution. The fmal result of the
analysis is already shown here; in a situation where analysis is just starting, only the data would appear.
FIG. 14.12. First step of a SlMNRA analysis: load the data and define the experimental conditions.
327
Chapter 14
The second step is to define the physics included in the calculation. In this case, all isotopes are calculated separately (as opposed to
making one single calculation for the average mass of the target element), Bohr straggling with the ChuIYang correction and including
multiple scattering is calculated, SiuM-2003 stopping powers are used, the scattering cross section for the l'O(a,a)"O is calculated
with SigmaCaic, and Andersen electron screening (Andersen et aI., 1980) is calculated for the other elements .
.~
FIG. 14.13. Second step ofa SMINRA analysis: Configure simulation parameters.
328
Finally, the sample is defmed: Layers are created, with any number of elements. The user inputs the thickness and concentration
values, as well as roughness parameters if required. Then, a simulation is generated and compared to the data. The user can then
iteratively refme the sample description until a good fit is reached or use a fitting routine to adjust the thickness and concentration of
one layer at a time.
FIG. 14.14. Third step ofa SIMNRA analysis: define the sample structure, perform the simulation. Here the final result is shown, after
manual iteration until agreement between simulation and fit is achieved.
14.6.3 NDF
Development of NDF started in 1997. It can be run in

Windows or DOS mode. in both Windows and UNIX.
We show in the NDF example (EXAMPLE 14.5) exactly the
same spectrum as was used for the RUMP and SIMNRA
examples. The main difference is that the final depth profile is
reached without an initial guess being defined by the user.
EXAMPLE 14.5. NDFlDataFurnace.
In this example, sample E3 from the IAEA intercomparison

exercise of IBA software (Barradas et af., 2007b) is used. It is a
hafnium oxide layer with impurities on a Si substrate, analysed

by RBS with a 2.5 MeV "He" beam. The first step after loading
in the data is to defme the experimental conditions. This is done
by creating' geometry fIles, one per each different set of
experimental conditions used. In this case. only one
measurement was made, so only one geometry file is required.
. The physics included in the calculation (the same as in RU1\1P
and SIMRNA) is also input in the geometry file. Note that NDF
does not make its own multiple scattering calculations; instead,
it uses DEPTH in run time. The flnal step is to associate each
spectrum with its corresponding geometry. in the main rnA
DataFumace window.
!,ih _
329
Chapter 14
n
C!
be
If
"
.. FIG. 14.15. First step of a NDF analysis: load the data and define the experimental conditions, configure simulation parameters.
~
,.
<
\'
..
"
330
'.
~;
',~
.\.
,,
The second step is to define which elements exist in the sample. This is done by creating a structme file with logical elements, which
can be simple elements such as Si, molecules such as Si02 with a given density, or complex molecules with unknown stoichiometry t(J
be determined in the fit Restrictions on the depth and concentration ranges where each logical element can exist might also be given.
If many completely different samples are tobe analyzed in batch mode, one structurc flie is created for each sample. If man)' similar
samples are to be analyzed, the same structure file is used for all of them.
'ndffOl01.dat.Expt
,.- ~dff0101.d.at.~f
FIG. 14.16. Second step of a NDF analysis: defme which elements or molecules are present in the sample.
331
b......
Chapter 14
The final step is to run an automated fit. In this case, the user does not have to define a layer structure; rather, this structure will be an
output of the fit. The algorithm used, simulated annealing, tries. to find the simplest structure consistent with the data. AI; the
calculation is completely automated, the user must check the results, because it can find unwanted solutions. Alternatively, the user
can also make an iterative analysis by defining a layer structure in the usual way. This can also be the initial guess for a local search
on all parameters.
FIG. 14.17. Third step of a NDF analysis: run a fully automated fit, or, alternatively. perform a iterative analysis.
14.6.4 Issues
Both codes are only available cornmerciaUy. Free test

versions can be obtained, but to use the codes in research or
Both SIMNRA and NDF are being actively developed and
industry, they must be purchased. This limits their availability. It
supported. This means tbat users <::an rely on advice and help
from the developers.
also means that reliable first-generation codes will continue to be

used in many cases. This is not a bad thing, as those codes are
Both codes are strictly proprietary. The source code is

restricted to the authors. This means that users cannot check
whether the algorithms used are correct or not. Most users do Dot
want to do this anyv.ray, and the IAEA intercomparisoD of rnA
software showed that the codes are essentially correct
perfectly capable of handling many experimental data.

Some modules recently introduced in NDF (e.g., for PIXE
and resonant NRA) are open source.
332
New-generation codes are more complex and difficult to use,

which is normal given that they include many further options.
This is another reason for the continuing use of fIrst-generation
codes in cases where the extra options are not necessruy.
14.7 MONTE CARLO SIMULAnON
The standard data analysis procedure described in the
sections above uses detenninistic algorithms. These can include
physics through accurate or approximate implementations. All of
the algorithms have limitations coming from two main sources:
First, many phenomena are treated in a statistical way with
disregard for the real particle-particle interactions that occur.
Energy loss, energy-loss straggling, and multiple scattering are
the three main phenomena in this category. In some cases, such
as multiple scattering in grazing-angle conditions, the best
models available (Arnsel et al., 2003) reach their limit of
validity. Second, the details of the experimental setup are
nonnally disregarded. For instance, calculations of geometrical
straggling often disregard its influence on the scattcring cross
section (Szihigyi e/ al., 1995; Raubala e/ al., 2006). Details of
detection systems other than solid-state dispersive energy
detectors, such as time-of-flight (TOF) systems, are not
implemented in most standard codes.
EXAMPLE 14.6 of data analysis by MC illustrates very well

the capabilities and limitations of Me simulations. All spectral.
features are reproduced, whicb allows the user to ascertain
whether a given signal is relevant or not. In the example
presented, it is possible to exclude the presence of 0 in the TaN
layer with accuracy much better than what could be acbicvcd by
standard codes, which do not calculate plural and multiple
scattering as accurately calculated as does Me. On 1:he other
hand, statistical oscillations are observed in the calculation. They
represent the accuracy of the Me calculation, which can be
improved by using longer calculation times. Nevertheless, the
calculation of backscattered ions, as well as of lo~ energy
signals, is inherently slower in MC simulations.
The main barrier to the general use of Me codes m routine
data analysis, in addition to calculation times that are still longer
than those required by standard codes, is that they are considered
difficult to usc. In some cases, no graphical user interface is
available, and the user must edit input files and look at output
files, which is very efficient for advanced users and rather
difficult for everyone else. MC codes are not yet widely used
except by the authors and their collaborators. This might change
soon, as one of the MC codes, Corteo, now has a user-friendly
windowed interface (Schiettekatte, 2008).
In particular, standard codes have problems in the analysis of

heavy-ion TOF-ERDA data, particularly at low energies where
multiple scattering plays a very important role. Not only are
signals broadened. but the actual yield is also Dot what would be
expected from single scattering.
Nevertheless, the extraordinary quality of the simulations that

can be achieved, particularly wben plural and multiple scattcring
are important, means that MC simulations might be the basis for
the next generation of codes, once issues with the user interface
and computer speed are solved.
The alternative is to develop a Monte Carlo (MC) simulation

of the in~ividua~ particle-particle interactions. In practice, i~n-
EXAMPLE 14.6. Monte Carlo.

electron mteracttoDs are not. .
calculated,
and
tabulated
stoppmg
W es howm
. F'Ig.14
.
. 18 a TOF measuremen t 0 f a S"S'O
1/
I 2 (100
powers are used. Nev.erth~less, c~mplex phYSIcal processes ~uch ~ ~ nm)!TaN (30 nm) sample measured with a 16 MeV 63 CU beam ..
as double
and multIple
as well as the. full tural
IOn'~
. . scattenng,
tak'
ana1yz ed WIth MCERD . RecOl'1s an d b at ks ca rt eredu e
atoms
.
detectlOn system Interaction, are
en mto account .mana
.
.
.
were separated b y th e d e tec "....on system.
way, WIthOut the approxunations that standard codes IDvolve.
The analysis, shown in Fig. 14.19, leads to the
Three Me codes for general-pw-pose analysis of ERDA data
stoichiometries Si, 1.00 0.02; 0, 2.00 0.02; and Ta, 0.47
have been presented and applied to the analysis of different
0.02; N, 0.53 0.02. The first comment is that the simulations
systems. MCERD and Corteo can also handle RBS data
include the statistical fluctuatiom expected from Me. The most
(Sajavaara e/ al., 1998; Arstila e/ al., 2001; Schiettekarte, 2008).
important point is that extraordinarily good agreement with the
The third code, FTHIE, is a fast Fortran version of TRIM fqr
data is obtained, even in the shape of the signal back edges and
heavy-ion ERDA (Johnston et al., 2000; Franich e/ al., 2004) .. the low-energy backgrounds. This is essential, for instance, to
Other codes, such as RBSIM (Smulders and Boenna, 1987) and
rule out the presence of 0 in the TaN film. No standard code
GEANT4 (Geil ef al., 2007), also exist.
could achieve SUCh results.
333
Chapter 14
Time-of-flight
1GMaV"Cu*
Tart(3OnmVSi~100nm)1Si
81191.... 3:8.&.,19.:2" .
,!
~
'J
FIG. 14.18. Time-of-flight ERDA ofa SifSiO, (100 nm)!TaN (30 nm) sample measured with a 16 MeV "eu beam.
.~
!\
The statistical fluctuation of the simulation is high for the Ta

recoil signal. This is due to the low recoil cross section for this
intercomparison of PIXE codes (Blaauw et al. 2002), but only!'

vel)' simple spectra were tested.
~
;j
heavier target element. Also, the backscattcring cross section is

low, which means that the calculation of the Cu recoils can be
The best-known PIXE code is GUPIX (Maxwell e/ al., 1995),

which is used in many laboratories around the world. GUPIX is
geared toward analyzing PIXE energy spectra accurately, :J
including those from multilayered targets. GUPIX is not "1
~
designed
to
determine
quantitative
two-dimensional J
concentration ma,ps from microbeam experiments.
orders of magnitude slower to reach the same statistical
accuracy. These aspects are inherent to Me data analysis.

Further improvements in the efficiency of the algorithms and in
the computer speed can lead to faster simulation times, which
are currently in the minute to hour range.
The Ta fe(:oil signal is not as well simulated as the other
elements, and a good simulation requires a slightly smaller layer
thickness., which is not consistent with the data from the other
elements. This is due to inaccuracies in the heavy-ion stoppingpower databases, whieh are sparse and not as accurate 'as those
for H and He, particularly at low velocities. This is a problem of
the technique and not of the data analysis method. as' all codes'
rely on the same stopping databases.
14.8 OTHER TECHNIQUES

In this section, only a brief mention of methods and codes
AXIL 01an Espen et ai., 1977) and its free-of-charge version

QXAS are also widely known.
:J
The code GeoPIXE was developed more recently (Ryan et fi
al.. 1990; Ryan. 2000). It has proved to be very powerful not

only in the analysis of energy spectra, but also for quantitative
iWo-ciirneilsioilai analysis. It is also capable of analyzing
-multilayered targets, as well as three-dimensional inclusions by
modeling FIXE yields from buried three-dimensional objects.
%.1','
;~
(,
.~
An alternative approach, AXSIA (Doyle et ai., 2006), uses d

multivariate spectral analysis. It has the advantage of being
automated and nonbiased and can also, in principle, be used to
combine different data sets, such as simultaneously collected
micro-PIXE and micro-RBS data.
.~
used for data analysis' of other techniques is made.
14.8.1 PlXE
Over the years, the PIXE community developed its own
)t~
methods and codes for data analysis. The IAEA organized an
lj
:;;l
.,
:t!
334
:~
5COO 16 MlaV '"cu =>'IlIN(30nmyslO,.(100nm)lSl
Neoq<.
CU .....
CU slmu ~ 1m
Nnw.aZinn
-.NdJaLS3nn
-Namu.Unm
'000
...
4.G
Oe=~~~--~~~~~
1Q.Q 10.5 ".D 11.5 12.0 12.5 13.0 13.5 , .... 0
Energy (MeV)
8CIO
16 NaV "cu q'1llll{90....)ISIO,.(1GOnm)lS1
7IlO
~soa
11-
"Ii
c
~15OD
"
~1000
~S-
SI .....
51""" B2 rill
2IlO
...
'00
0
ao
'200
as
4.
4.5
"0
5.5
S.G
....
I.!
Energy (MeV)
16 MoV '"cu =>1IIN(un}lSlO,(11lOm1)1S1
'000
e
EmrgJ(MeV)
Trtstnu. v rm
. 0 .....
2IICI
TaupL
Ii
lam
r~
-ORInTaN
0 t05In 'Ta\!
"
8-
8--
Ii
5800
loa
50
200
Oanv.D2nm
- - OftiTTaN
Enorgy (MeV)
\.
a.75
flO
1.2!
8.!1
Energy (MeV)
FIG. 14.19. MCERD analysis ofTOF-ERDA ofa SilSi02 100nrnfTaN 30nm sample measured with a 16 MeV "Cu beam.
The open~source LibCPIXE library (pascual-lzam. et al.

2006b), based on the DA1TPIXE code (Reis and Alves, 1992),
also supports multilayered targets. It has been integrated in the
general-purpose IBA data analysis code NDF (Jeynes et a/.,
2003). Any number of PIXE spectra can be analyzed together
with any number of RBS, ERDA, and NRA spectra, leading to
one single depth profile consistent with all of the data. Similarly
Already in the 1990s, PIXE was combined with RES in the

same code, through the so-called Q-factor method implemented
in the user-friendly DAN32 package (Grime, 1996). which uses
RUMP and GUPIX. A fit to the prolon RES spe<;trum collected
simultaneously with the PIXE data is made. This fit is used to
determine the actual collected charge, which is then used in the
quantitative analysis of the PIXE data. This very useful
improvement, however, does not ensure complete consistency
to GUPIX. this approach is geared toward spectra and not twodimensional maps.
between the analysis of the RBS and PIXE data, and it also does
not fully tak.e advruitage of the combination of the depth
sensitivity of RBS with the elemental and trace sensitivity of
PIXE.
335
Chapter 14
14.8.2 Resonant NRA, PIGE, NRP. MEIS, cbanneling,

microscopies, and so on
This short section is dedicated to techniques and experimental
conditions not yet covered in this chapter.
They all have dedicated methods of data analysis. Particular

applications often require the development of specific methods.
In several different techniques, the yield of a given nuclear

reaction is determined as a function of the initial beam energy.
Depth profile infonnation can be extracted from the shape of the
excitation curve obtained by scanning the beam energy.
Vieira, and co-workers for the unattended automatic analysis of
The nuclear reaction is often a (elastic or inelastic) particleparticle reaction or a particle-gamma reaction. These techniques
are often called resonant NRA and particle-induced gamma
emission (FlOE), respectively. Different codes have been
'.--preserited, Such 'aii-iu'lALNRA by Johnston (1993), WinRNRA

by Landry and Schaaf (2001), and ERYA by Mateus 0/ al.
(2005). NDF also implements this technique.
If a narrow resonance is used, normally isolated and
background-free, together with a correspondingly small initial
beam energy spread (each with a fwhm typically smaller than
0.5 keY), the term narrow resonance profiling (NRP) is used
independently of the reaction type. In that case, it is necessary to.
take into detailed account the individual energy-loss events to
correctly calculate the excitation curves. The best-known code
for that pwpose is SPACES by Vickridge and Amsel (1990). It
implements the stochastic theory of energy loss of Maurel et al.
(1982), which allows the user to retrieve accurate highresolution depth profiles. Codes that assume a Gaussian energy
distribution of the beam will lead to inaccurate depth profiles,
-- "particularly 'near the surface. More recently, another code,
These are beyond the scope of this chapter.

Artificial neural networks have been developed by Barradas,
IBA data and applied to a variety of systems, from the very
simple to the very complex (Barradas and Vieira. 2000; Nene et
al., 2006). This approach can be of practical importance only
when large numbers of similar spectra are collected, as in
quality-control applications, for example, or in the analysis of
data collected in real-time RBS experiments, where hundreds or
thousands of spectra are collected within a few days
(Demeulemeester of al. 2008).

Finally, many codes and data analysis codes have been
developed over the years for very specific applications and
experimental conditions. We refer the reader to the review by
Rauhala et al. (2006), and the references therein, where several
of these codes are discussed.
14.9 THE IBADATA FORMATIDF

IBA data analysis codes such as RUMP, SIMNRA, NDF, and
others all use and implement various fonnats to store the data
and to describe the experimental conditions and simulation or fit
parameters. RUMP reads and writes both specific ASCII and

binary fonnats. SIMNRA reads and writes a specific ASCII
based format. NDF is user-driven and reads over 30 different
ASCII and binary formats, but it writes specific files.

This is a strong barrier against interchange of data between
people from different laboratories, or that use different set-ups to
measure their data, different computers to store their data, and
different data analysis programs to analyse their data.
FLATUS, was presented by Pezzi 01 al. (2005).

Medium-energy ion scattering (MEIS) could be seen as a
particular case of RBS at low energies, but in facl, several
A new rnA data format, called IDF. that aspiTes to support all
forms of rnA data in a well defined simple format permitting
easy exchange between sites and analysis programs, was
presented in the 19th International Conference on Ion Beam
important points must be considered, including special detector

arrangementsi the screened scattering cross section used; and
also the fact that, for the high-resolution profiling that can be
achieved with MEIS, stochastic energy loss must be considered
(pezzi 01 aI., 2007).
Specific methods have been developed for the analysis of
channeling data. These include dedicated Monte Carlo codes

(sometimes limited to specific systems), as well as different
implementations of analytical theories. Monte Carlo codes such
as FLUX by Smulders and Boerma (1987) and BISIC

(Albertazzi 0/ al, 1996) are adequate for detailed studies of
crystalline lattices, including lattice locations of imp'urities,
defect configurations, and damage. Analytic codes stch as
DICADA (Giirtner, 2005) and REX (K6tai, 1994) are less

accurate but possibly easier to use in the study of defect profiles.
Analysis (September 7-11, 2009, Cambridge, UK), and will be

implemented in NDF, RUMP and SIMNRA codes by the end of
2009.
The main charncteristics of the IDF are: 1) It can be read and
understood by human inspection of the file, i.e. it is Ascn. This
makes it portable amongst all hardware platforms and operating
systems, and it is self-explanatory; 2) It is flexible enough to
cater to all IBA data needs, so a file entry can correspond to one
given sample, or to a group of similar saDlples. or to one
experimental run, or it may contain simulations only, or anything
else that the user may need; and 3) It is easily extendable; most
REX is the only standard all-pwpose data analysis code that

includes channeling.
commonly used features are already included, but special needs
Different ion-based microscopies have been developed. These

range from the two-dimensional maps collected with PIXE to
tecbniques such as scanning transmission ion microscopy (SlIM).
The IDF was implemented as an extensible mark-up language

(XML) format. XML was developed by the World Wide Web
are easy to incorporate by any use.
Consortium (W3C) to structure, store, and transport data. It was
336
chosen as the IDF development platform because it is a

recognized standard that fulfills the specifications dermed.
ACKNOWLEDGEMENTS
We thank Mike Thompson, Matej Mayer, and Kai Arstila for
providing the figures for the RUMP, SIMNRA, and MCERD
examples, as well as all those who commented on early drafts.
The definition of the IDF, including full documentation and

examples, is given in bttp:/lidf.schemas.itn.pt .
Table 14.2. General infonnation about analysis programs.

Analysis
Technical contact
Current status
Distribution mode
program
Status of
source code
Operating
syslem(s)
DEPTH
Edit Szilagyi
KFKl Research Institute for Particle and
Nuclear Physics
Budapest, Hungary
szilagyi@nnki.klki.hu; www.kfki.hU/-ionhp/
Active
development
No charge;
downloadable: from
Web site
Restricted to
author/not
available
DOS and
Windows
GISA
Eero Raubala
and Jaakko Saarilahti, Technical Research
Center of Finland
Eero.rauhala@helsinki.fi;
laakko.Saarilahti@vtt.fi
Kai Arstila
IMEe, Leuven, Belgiwn
kai.arsti.Ja@iki.fi
In use
No charge; "",'lite to
author for copy
Restricted to
author/not
available
DOS (or
emulators)
Active
development
No charge; write to
author for copy
Source code
available
Linux
(Windows)
Nuno Barradas
Technological and Nuclear Institule
Sacavem, Portugal
nunoni@itn.pt;
www.ee.surrey.ac.uklResearchiSCRIBNndf7
Active
development
Commercially available
through Univ. of
Surrey; evaluation
copies available upon
request
Windows,
DOS,
Endre K6tai
KFK.I Research Institute for Particle and
Nuclear Physics
Budapest, Hungary
kotai@rrnki.kfki.hu
Mike Thompson
Dcpt. of Materials Science, Comen University
Ithaca, NY USA
motl@comell.edu;www.genplot.com
Active
development
No cbarge; 'Write to
author for copy
Restricted to
author/not
available;
some
modules
open source
Restricted to
author/not
'available
Active
development
Commercially available
through Computer
Graphics Service;
evaluation copies
available from Web site
Commercially _available
through MPI for
Plasma Physics;
evaluation copies
available from Web site
Source code
available
Windows,
Linux,
MCERD
NDF
(DataFumace)
RBX
RUMP
SIMNRA
Matej Mayer
MPI for Plasma Physics
Garching, Germany
MateiMayer@ipp.mpg.de;
www.rzg.mEgdel-marnl
Active
development
337
UNIX
Windows
UNIX,OS2
Restricted to
author/not
available
Windows
Chapter t4
Table 14.3. User interface properties ofllie analysis programs.

Analysis
program
Nature of
the primary
Primary simulation
Graphical
output
modes
DEPTH
interface
Interactive
Manual iteration
N/A
GlSA
or batch
Interactive
Manual iteration;
Screen only
automated parameter
search
MCERD
NDF
(DataFumace)
Batch
Batch
directed or
Manual iteration
N/A
Fully automated search;

manual iteration
Publication
Manual iteration
Draft
quality
quality
interactive
RBX
Interactive
RUMP
Interactive
SIMNRA
Interactive
Manual iteration;
Publication
automated parameter
search
quality
Manual iteration;
Draft
quality
automated parameter
search
Table 14.4. Fundamental databases used by analysis programs.

Analysis
Stopping powers
program
Cross sections
DEPTH
ZBL'95
User-defmed from file
GISA
TRIMI991 or Ziegler and Chu;

user-defmed correction per layer
ZBL'96 or user-defined
ZBL'95, SlUM, KKKNS (Si), MSTAR
Ability to Joad experimental values; user-defined (;orrection
User-defmed from file
MCERD
NDF
(DataFurnace)
Rutherford, user-<lefined from file to be included

User-<lcfined from file
per ion/element
RBX
ZBL'95 with channeling correction possible; user-defined

. correction per layer "
"
Internal (modifiable) library of cross sections; external cross
section Library Editor (import Sigmabase R33; user-deflDed)
RUMP
ZBL'92, SRJM, KKKNS

User-rlefined compounds and databases; user-defined
correction per layer or per element
mANDL, SigrnaCalc; user-defined
SIMNRA
SRIM-1997, SRlM,orKKKNS
User-defiDed" correction per layer and ion
mANDL, SigrnaCalc; user-<lefmed
338
Data AnalysIs Software for Ion Beam Analysis
Table 14.5. Fundamental physics handled in the simulations.

Analysis
program
DEPTH
Isotope calculation
Single
i~otope
Screening calculation
Yes
Plural scattering
Multiple scattering
Geometric
stmggling
Channeling
None
Yes, Pearson VII distribution
Yes
No
None
None
No
No
Full Me calculation
Full Me calculation
Yes
No
Bohr. Chu., Yang.

TschaIMr
Dual scattering approx.

(run-time option)
Yes, Gaussian approximation

from DEPTH calculation
Yes. from DEPTH
No
calculation
Bohr. Chu.' Yang.
None
Yes (same model as DEPTH)
Yes
Defect
calculation;
simulation of
channeled spectra
None
None
No
No
Dual scattering approx.

(run-time option)
Yes (DEPTH model
Yes
No
approximated as Gaussian)
Straggling
models
Bohr, Chu, Yang,
Tschalilr
Natural abundance
andlor specific
isotopes
Energy/angle; external
tables by users
Bohr +
Natural abundance
and/or specific
isotopes
No
Bohr. Chu: Yang
Natural abundance
and/or specific
i:mlopc:)
Natural abundance
and/or specific
isotopes
Energy/angle;
Andersen and L'Ecuyer
RUMP
Natural abundance
andlor specific
isotopes
Energy only; L'Ecuyer
Bohr
SIMNRA
Natural abundance
andlor specific
isotopes
Energy/angle;
Andersen and L'Ecl1yer
TschaHlr
OlSA
MCERD
NDF
(DataFumace)
RBX
LindhardlScharff
!
!
Yes
TschaUir
Bohr. Chu., Yang.
~ter14
Table 14.6. EXI!erimental conditions and simulation caEabilities.

Scattering
Incident
Analytical technique(s)
ions
geometry(ies)
Analysis
program
DEPTH
All
RBS, ERDA, NRA
IBM, Cornell
Pileup
correction
No
,
Stopper foils
Energy calibration
Energy-dispersive
(magnetic spectrograph
coming)
Simulated; including
Linear
Detection system(s)
inhomo~eneities
Quadratic
GISA
All
RBS
IBM
No
Energy-dispersive
N/A
MCERD
All
ERDA,RBS
IBM
No
Energy-dispersive;
TOF; different layers in
detector for gas
detectors, for example
Simulated; equivalent
treatment to sample
Linear
NDF
(DataFumace)
All
mM, Cornell,
Yes
Energy-dispersive
treatment to sample
Quadratic; varying by
ion species
RBX
All
RBS,ERDA,
nomesonant NRA, PIXE,
resonaot NRA (PIGE,
NRP)
RBS,ERDA,
nomesonant NRA
IBM, Cornell
Yes
Energy-dispersive
Simulated;
homogeneous foils only
Linear
RUMP
All
RBS, ERDA
Cornell, IBM,
General
Yes
Energy-dispersive,
partial TOF
Simulated or from user

calibration
Linear
SIMNRA
All
RBS,ERDA,
IBM, Cornell,
General
Yes
Energy-dispersive,
TOF, electrostatic, thin
solid-state detectors
with transmission of
particles
treatment to sample
Quadratic; varying by
ion species
General
<!';
nonresonant~
~"""':""""""'~""<",,"o.lro>'''';''''"''.''''~,"'''''''''''P .
.~.- ... -""-~.~.,~-,- "
.. -~>-< .... ~-~ ..
.J
Data AnalYSis Software for Ion Beam Analysis
Table 14.7. Fitting capabilities DEPTH and RBX have no fitting capabilit~~).
Analysis
program
GISA
Starting conditions
Reasonable guess
Optimization
method
Error estimation
Xi
None returned
Searchable
experimental
parameters
Autorefinement
oflayers
None
No
One layer
at a time
Poi~~on
Eo. ECD1 charge. 9,

<l>,<p
Yes
A11 parameters
variable, user
controls which
ones change
Poisson
Eo, Eeal charge,

current, 9, Ill, t:p
No
No internal limit,
practical of 30
parameters at a
time
Poisson
E e l, charge
No
One layer at a
time, all
characteristics
Statistics
used
Limitations
minimization
NDF
(DataFurnace)
Elements only; guess

can be used but not
required
Simulated
annealing
plus grid
search
Bayesian inference with Markov

chain Monte Carlo integration
(time-intensive)
RUMP
Reasonable guess
Marquart
search
Cu~ature of matrix; full

correlation of error sensitivities
(intrinsic to search method)
x:
SIMNRA
Reasonable guess
Simplex
search
Additional search to determine

curvature near best fit
(comparable to fit time)
!,
I,
Table 14.8. Sample definition and complexity handled,
Analysis
Sample description
Continuous profiles
Substrate
roughness
Slab limitations/elements
I'
, "
program
Layer
roughness
None
10 layers; 10 clements; user can
define so layer"'element is constant
No
No
No
No
Effective error function

interdiffusion profiles between
layers
t ~'
User-defined functions (Gauss,
error functions, etc.); channeling
defect distributions
250 layers in description; up to 92

constituents
Yes; approximated
as energy
broadening
Yes; approximated
as energy
broadening
None
No
No
Layer definition with equation

overlays
Gaussian implants. error function'

diffusion, Pearson IV profiles,
one-sided arid two-sided diffusion
None
No
Yes; single- or
dual-sided, all
interfaces possible
Layer definition; use of profile

function
Slabs only
100 layers in description; 40

elements per layer
Yes; Lorentzian or
Gaussian angular
distribution
Yes; all interfaces

possible
DEPTH
GISA
Layer definition
Layer definition or profile
function
Slabs only
Maps continuous profiles onto
slab structure'
NDF
(DataFumace)
Layer definition; use of profile

function
REX
Layer definition
RUMP
SIMNRA
ii
II .
~
I,:
Chapter 14
Table 14.9. Effects of some phenomena on data analysis.

phenomenon
Ion
Energy
Scattering
Target clement
Effect on spectrum
Error
Effect on analysis
al1sl~
cross section
--electron'screening--lXil---------Aii----------<is o ------------Aii-----------------D~-rease(fyiefdAll
~ 150
> I SC
Heavy
keVIZ J
All
>.500
>90'"
Heavy
--_. -Unk-nown
(iarge) ---------------------------f~ts_e ~u_a_'!t!~l!t~ye_1!t}1!!:t~i..s_ . 1%
~.2%
.. _... ____ .. ______ . ___ ... _____________If~Y-'?_l ________ . _. _________________ .. ____ . ___________________ . _... ________ . _______________________ . _. ___________________________ . ______ . _. _. _. ____________ .
nuclear effects
2:100 keY
All
>2.0 MeV
All
Light (medium
heavy)
Light
Increased or decreased
yield
Depends on reaction (can be large)
stopping power
false quantitative and depth
. ______ _. ______________________ ___ __________________________ . __________ _. _._. _____________________________ _. __ ___________________ ~!.l.!'ll'~!!!. _ _. _. ___ ._._. ______ .
general
uncertainties
All
All
Si
All
Li
Heavier
ions
Bragg rule violations
--ph;;i~;i;~te~ffe~tS-
energy spread,
All
4,2% average
(>10% for 13% of target atoms)
4.1% average
(>10% for 11 % of target atoms)
9%
5.1% average
(>10% for 17% oflarget atoms)
6.1% average
(> 10% for 18% of target atoms)
-Ail-- _... ----- --- -- ------ -- -... --- -- -------_. _. --- i 5%" --- ---- _... -. ----- -- -" ------- -----. _. _. _.. ---. ---------------_. _.
.______ . _. _. ______ ._._._,._._. ________ .______0!1~l'J~~C!'JO_______ . _..l:teJ!i..s_,!l)~_ ~i~.t~~ ______ ......... __ . __ .. __________ . ______ _. _._. _______________________ . _. ___ .' __ .. _. _. _.
--Lightc~-~p;~~d~-
Light solids versus

liQuid or gaseous
All
Heavy
Grazing-angle
geometry
Nonna1
incidence
All
mUl1iplc scattering
surfnce roughness
yields and widths
All
All
-i~~~;s-e-d ~~'d~~~~~~d
yields and widths
edge and peak broadness
-~
o~i
:55~ 1 0%
Large
Small
All
Increased edge and peak
DeQends on details (can be small or large}

Depends on details (~n be small or large)
broadne!l~
channeling
All
false interdiffusion, mixing,

depth profile, and roughness
analysis
Incidentally
Decreased yields
Depends on details (can be small or large)
;rie\ds l backsround
Increased background
aligned
false analysis of intermixing

and interdiffusion
false quantitative and depth
analysis
gcornet!),:
pulse pileup
All
low-energy
background
All
Heavy (all)
Lew (all)
"'The errors quoted refer to SRIM-2003.
Grazing
geometry (all)
Heavy (all)
5% for count rates below -5 kHz
Depends on the details; can be dominant at

low energies; small in most analyses
false quantitative analysis,

reduced sensitjvi~
false quantitative analysis,
reduced sensitivity
SOFTWARE WEB SITES
This list of web sites rele",ant for IBA software and data analysis is by no means complete. Several of the codes mentioned in this
chapter have no dedicated Web site.
Corteo
hnp:llwww.lps_umontreal.ca/~schiette/index_php?n-Recherche.Corteo
Monte Carlo'(MC) simulation DIOgram for ion bearo analvsis soeclra

DEP1H
http://www.kfki.hU/-ionbp/dodprog/wdepth.htm
Windows-based program for calculating the energy ~d depth resolution of ion beam analysis methods such as
Rutherford backscattering spectrometrY (RBS) and elastic recoil detection analysis (ERDA)
FLUX
http://members.home.nllp.j.m.smuldersIFLUXlHTMll
Program for simulating the trajectories of high-energy ions in single crystals in channeling, or near-channeling,
directions
GeoPIXE
bttp://www.vrop.csiro.aulGeoPIXE.html
Software package that perfonns real-time particle-induced X-ray emission (PIXE)/synchrotron X-ray fluorescence
(SXRF) spectral deconvolution using dynamic analysis
.
GUPIXI
GUPIXWIN
bttp:llpixe.physics.uoguelpb.ca/gupixlmain!
Software packsge for fitting particle-induced X-ray emission (PIXE) spectra from thin, thick, intermediate, and layered
specimens
IBANDL
http://www-ndsjaea.orglibandli
Library of experimental data on nuclear cross sections and excitation functions relevant to ion beam analysis (IBA);
resulted from the merger Sijffi]aBase and NRABASE
IBIS
http://www.klki.hul-ionbp/
Collection of resources for the ion beam community (called the Ion Beam Infonnation System) that includes links to
several programs and other useful sites, as well as an archive of documents
IDF
bttp:l/idf.schemas.itn.pt
Defmition of the IBA data fonnat IDF including documentation and examples
LibCPIXE
htlp:l/sourceforge.netlprojects/cpixe _~. __: ._: ~.~ _0:- ~___

. _._ ... __.. :-.:... ."",-,,-,~_._:: .
Library for the simulation of particle-induced X-ray emissioncpIXE)spectra
-,<-.
~~----~-~-~-------
MSTAR
http://www.exphys.uni-linz.ac.atistopping/
Collection of data on stopping powers of light and heavy ions
NDF
http://www.itn.ptifacilities/lfJind'ukJfi_ndf.htm
General-purpose cD.de for the analysis of Rutherford backscattering spectrometry (RBS), elastic recoil detection
analysis (ERDA), particle-induced X-ray emission (FIXE), non-resonant nuclear reaction analysis (NRA), and neutron
depth profiling (NDP) data for any ion, any target, any geometry, and any number of,pectra
http://www.ee.surrey.ac.uklSCRIBAlndfi
Fitting code (not a simulator) that extracts elemental depth profiles from Rutherford backscattering and related ion
beam analysis spectra
DataFurnace
QXAS
http://www.iaea.org/OurWorkiSTINAlNAAlJpcilinshrrf/poiXRFdown.php
.
DOS-based software nacka.e (called Quantitative X-rav Analvsis Svsterol for auantitative X-ray fluorescence analysis
RUMP
http://www.genplot.com
Graphical data analysis and plotting package that pro-vides comprehensive analysis and simulation of Rutherford
backscatterin~ spectrometry-(RBS)-and elastic recoil detection reRD) spectra
Sigmacalc
http://www-nds.iaea_orgfsigmacalC/
.
Service providing evaluated (recommended) differential cross sections for ion beam analysis
343
Chapter 14
SIMNRA
http://www.rzg.mpg.del-mami
Windows-based program for the simulation ofbackscattering spectra for ion beam analysis with megaelectronvolt
ions, mainly intended for non-Rutherford backscattering, nuclear reaction analysis (NRA), and elastic recoil detection
anal sis ERDA
SRlM
http://www.srim.org
Group of programs that calculate the stopping and range of ions in matter using a quantwn mechanical treatment of
ion-atom collisions
Doyle, B.L., Provencio, P.P., Kotula, P.G., Antolak, AJ., Ryan,
C.G., Campbell, J.L., Barrett, K. (2006), Nucl. lnstrum. Methods
B249,828.
REFERENCES
Albertazzi, E., Bianconi, M., Lulli, G., Nipoti, R, and Cantiano,
M. (1996), Nuc!. Instrum. Methods B1I8, 128.
Alkemade, P.F.A., Habrakeu t F.R.P.M., and van der Weg, W.F.
- (1990), Nucl. Instrum. Methods B45, l39.
Eckstein, W., and Mayer, M. (1999), Nucl. Instrum. Methods

BI53,337.
Edehilano, E., Arstila, K., and Keinonen, J. (2005), Nud.
Franich, R-D., Joboston, P.N., and Bubb, I.F. (2004), Nue!.
Instrum. Methods B219--220, 87.
Gardner, R.P., and Wielopolski, L. (1977), Nuel. Instrum.
Methods 140,289.
Gartner, K. (2005), Nuel. Instrum. Methods B227, 522.
Amsel, G. (1996), Nuc!.lnstrum. Methods B118, 52.
Amsel, G., Battistig, G., and L'Roir, A. (2003), Nucl. Instrum.

Methods B201, 325.
Andersen, H.H., Besenbacher, F., Loftager, P., and Moller, W.
(1980), Phys. Rev. All, 1891.
Arstila, K., Sajavaara, T., and Keinonen, 1. (2001), Nucl.
Barradas, N.P., Jeynes, c., Jenkin, M., and Marriott, P.K.
(1999), Thin Solid Films 343--344, 31.
Barradas, N.P., and Vieira, A. (2000), Phys. Rev. E62, 5818
Barradas, N.P. (2004), Nuc!. Instrum. Methods B225, 318 .
Geil, R.D., Mendenhall, M., Weller, R-A., and Rogers, B.R.

Grime, GW. (1996), Nuel. Instrum. Methods B109--110, 170.
Gurbich, A.F. (1997), Nucl.Ins/rum. Meth~ds Bl!?, 311.
Gurbich, A.F. (1998), Nuc!. Instrum. Methods B145, 578.
Gurbich, A.F. (2005), in Nuclear Data for Science and
Technology (Haight, R.C., Chadwick, M.B., Kawano, T., and .
Talou, P., eds.), AlP Conference Proceedings, AIDerican
Institute of Physics, New York, vol. 769, p. 1670.
.. Barradas, N.P., Alves, E., Jeynes, C., and Tosaki, M. (2006),

Barradas, N.P. (2007), Nucl.Instrum. Methods B261, 418.
Barradas, N.P., Pezzi, R.P., and Baumvol, lJ.R- (2007a), Nucl.
Barradas, N.P., - Arstila, K., Battistig, G., Bianconi, M.,

Dytlewski, N., Jeynes, C., K6tai, E., Lulli, G., Mayer, M.,
Raubala, E., Szilagyi, E., and Thompson, M. (2007b), Nud.
Hrujuoja, J., Vayrynen, S., Putkonen, M., Niinist6, L., and ..

Raubala, E. (2007), Appl. Surf Sci. 253,5228.
Bicbsel, H (2006), Nucl. Instrum. Methods A562, 154.
Jeynes, C, Barradas, N.P., Wilde, l.R., and Greer, A.L. (2000b),.

Nuel. Ins/rum Methods BI61-163, 287.
Jeynes, C, Barradas, N., and Webb, R. (2000e), The University
of Surrey Ion Beam Centre IBA DataFumace Users Guide,
University of Surrey Ion Beam Centre, Guildford,
Jeynes, C., Barradas, N.P., Raila-Yuan, H., Hichwa, B.P.,

Close, R. (2000a), Surf Interface Analysis 30,237.
Blaauw, M., Campbell, 1.L., Fazinic, S., JakSic, M., Orlic, I., and
Van Espen, P. (2002), Nucl. Instrum. Methods B189, 113.
Bohr, N. (1948), Milt. Fys. Medd. K DW!. Vidensk. Selsk. 18 (8).
B0rgesen, P., Behrisch, R-, and Scherzer, B.M.U. (1982), Appl.
Phys. Al7, 183.
http://www.ee.surrey.ac.uklIBC/ndflwindfinanuaI70.pdf.
Jeynes, C., Barradas, N.P., Marriott, P K., Boudreault, G." .
Jenkin, M., Wendler, E., and Webb, R.P. (2003), 1. Phys. D: .
Appl. Phys. 36, R97.
Jeynes, C., Peng, N., Barradas, N.P., and Gwilliam, R.M. (2006),
Nud. Instrum. Methods. B249, 482.
Boudreaul!, G., Jeynes, C., Wendler, E., Nejim, A., Webb, R.P.,
and W~~en, U. (2002), Surf Interface Anal. 33, 478.
Butler, J.W. (1990), Nucl. Instrum. Methods 845,160.
Chu, W.K. (1976), Phys. Rev. A13, 2057.
Demeulemeester, 1., Smeets, D .. , Van Boekstael, C.,

Detaveruier, C, Comrie, C.M., Barradas, N.P., Vieira., A., and
Vantomme, A. (2008), Appl. Phys. Lett. 93,261912.
Jobo,ton, P.N. (1993), Nuel.Instrum. Methods B79, 506.

Johnston, P.N., Franich, R.D., Bubb, I.F., EI Bouanani,
Cohen, D.D., Dytlewski, N., aod Siegele, R- (2000),
Instrum. Methods BI61-163, 314.
Doolittle, L.R. (1985), Nuel. Instrum. Me/hods B9, 344.
344
Kirkpatrik, S., Gelat!, e.D., and Vecchi, M.P. (1983), Science

220,671.
Raubala, E., Barradas, N.P., Fazinic, S., Mayer, M., Szilagyi, E.,
and Thompson, M. (2006), Nucl. Instrum. Methods B144, 436.
Kotai, E. (1994), Nuc!. Instrum. Methods B8S, 588.
Reis, M.A., and Alves, L.e. (1992), Nuc!. Instrum. Methods
Landry, F., and Schaaf, P. (2001), Nuc!. Instrum. Methods B179,'

262.
'
J .A
. N (1979) , N uc.I
L 'E cuyer, J ., D
aYles,
., d
an MatsunmDl,.
instrum. Methods 160,337.
B68,300.
Ryan, e.G., Cousens, D.R., Sie, S.H., Griffm, \V.L., ud Suter,
G.G. (1990), Nuc!. Instrum. Methods B47, 271.
Ryan, e.G. (2000), Int. 1. Imaging Syst. Technol.ll, 219.
Mateus, R., Jesus, A.P., and Riheiro, J.P. (2005), Nud. Instrum.
Methods B229, 302.
Saarilahti J., and Raubala E. (1992), Nuc!. Instrum. Methods

B64, 734.
Maurel, B., Amsel, G., and Nadai, J.P. (1982), Nud. Instrum.
Methods 197,1.
Sajavaara, T., Jokinen, J., Arstila, K., and Keinonen, J. (1998),

Nuc!. Instrum. Methods BI39, 225.
Maxwell, lA, Teesdale, WJ., and Camphell, JL. (1995), Nuc!.

Instrum. Methods B9S, 407.
Mayer, M. (1997), SJMNRA User's Cmide, Technical Report
JPP9/113, Max-Planck-Institut fUr Plasmaphysik, Garching,
Germany.
Mayer, M. (2002), Nuc!. Instrum. Methods B194, 177.
Mayer, M., Fischer, R., Lindig, S.. von Toussaint, U., Stark,
R.W., and Dose, V. (2005), Nucl. Instrum. Methods B228, 349.
Schiettekatte, F. (2008), Nuc!. Instrum. Methods B266, 1880.

Smulders, P.J.M., and Boerma, D.O. (1987), Nuel. Instrum.
Methods B29, 471.
Spaeth, C., Richter, F., Grigull, 5., and Kreissig, U. (1998),
Nuc!. IrlStrum. Methods B140, 243.
Szilagyi, E., Paszti, F., and Arose!, G. (1995), Nucl. lnstrum.
Me/hods. BIOO, 103.
Szilagyi, E. (2000), Nuc!. Instrum. Methods BI61-163, 37.
Mayer, M. (2007), SIMNRA User's Guide, MaxPlanck-lnstitut

Tesmer. J.R., and Nastasi, M. (eds.) (1995), Handbook of
fur
Plasrnaphysik,
Garching,
Germany,
Modem fon Beam Materials Analysis, Materials Research
http://www.rzg.mpg.de/%7Emam/Soc iety.Pittsburgh.PA.
Mayer, M. (2008), Nuc!. Imtrum Method, B266, 1852.
Molodtsov, S., Gurbich, A.F., and Jeynes, C. (2009), J. Phys. D:
Appl. Phys. 41, Article Numher 205303.
Tosaki, M., Ito, S., and Maeda, N. (2000), NuC!. instrum.

Methods BI68, 543.
Tschaliir, C. (1968a), Nud. Instrum. Methods 61, 141.
Tschaliir, e. (I 968b), Nuel. Instrum. Methods 64, 237.

Tschalar, C., and Maccabee, H.D. (1970), Phys. Rev. Bl, 2863.
V
E
P N II
H
d Ad
F (1977) N I
_ an spen, ., u ens, ., an
ams,.
,UC.
--jnstrum. Methods 145, 579.' -.~ ~Neumaier, P., Dollinger," G., Bergrna-ier, -A.,- Genchev, I.,". V'!Ckridge,.,
1 an d Amse I" G (1'99-0-),ue.
-N "/ 1ns-trum.'~-)I~th
J.me 0=
Gorgens, L., Fischer, R., Ronning, C., and Hofsiiss, H. (2001),
B4S 6
Nuc!. Instrum. Methods B183, 48.
' .
Wielopolski, L., and Gardner, R.P. (1976), Nuel. lnstrum.
Padayachee, J., Meyer, K.A., and Prozesky, V.M. (200 I), Nuc!.
Methods 133, 303.
Wielopolski, L., and Gardner, R.P. (1977), Nuc!. Instrum.
Pascual-Izarra, C., Reis, M.A., and Barradas, N.P. (2006a), Nuel.
Methods 140, 297.
Yang;Q., O'Connor, D.1., and Wang, Z. (1991), Nuel.instrum.
Pascual-Izarra, C., Barradas N.P., and Reis, M.A. (2006b), Nuel.
Methods B61, 149.
Ziegler, J.F., Lever, R.F., and Hirvonen, J.K. (1976), "Computer
Paul, H., and Schinner, A. (2001), Nuel. Instrum. Methods BI79,
Analysis of Nuclear Backscattering", in ion Beam Surface Layer
299.
Analysis (Meyer, 0., Linker, G., and Kappeler, F., eds.), Plenum,
Paul, H., and Schinner, A. (2002), Nuel. Instrum. Methods BI9S,
New York, voL I, p. 163.
166
Ziegler, J.F. (2004), Nuel. Instrum. Methods B219-220, 1027.
Pezzi, R.P., Wallace, R.M., Copel, M., and Baumvol, Ll.R.
Ziegler, J.F., Biersack, J.P., and Ziegler, M.D. (2008), SRlM-'
(2005), AIP Conf Proc. 788, 571.
Tne Stopping and Range of Ions in Mat1er, Lulu Press,
Pezzi, R.P., Grande, P.L., Copel, M., Schiwietz, G., Krug, C.,
Morrisville, NC.
and Baumvol, U.R. (2007), Surf Sci. 601,5559.
Molodtsov, 5., and Gurbich, A.F. (2009), Nuc!. Instrum.

Methods B, doi:10.1016/j.nimb.2009.08.008 .
Nene, N.R., Viein!, A., and Barradas, N.P. (2006), Nuel..
Instrum. Methods B246, 471 _
Raubala, E. (1984),1. Appl. Phys. 56,3324.

Raubala, E. (1987), J. Appl. Phys. 62,2140.
345
CHAPTER
PITFALLS IN ION BEAM

ANALYSIS
C. Jeynes
University of Surrey Ion Beam Centre, Guildford, England
N. P. Barradas
Instituto Tecnologico e Nuclear, Sacavem, Portugal
Centro de Flsica Nuclear da Universidade de Lisboa, Lisbon, Portugal
Contributors: M. Dobeli, J. A. Davies, W. N. Lennard, and I. V. Mitchell
.'
15.1
_.
w __
.-
._-~
..... -
__
. _ -
-,--"
INTRODUCTION ....................................................................................................349
15.2 LOST BEAM AND EVENTS .................................................................................351

15.2.1 Charge measurement .........................................................................................351
15.2.2 Beamline charge exchange ................................................................................352
15.2.3 Particle identification .........................................................................................352
15.2.4 Dead time and pulse pileup ...............................................................................353
15.2.5 Electronic noise and ground loops ....................................................................355
15.3 CALmRATION OF FIXED PARAMETERS ...................................................... 356
15.3.1 Energy ...............................................................................................................356
15.3.2 Solid angle ...........................................................................................................358
15.3.2.1. NRA Standards ...................................................................................... 358
15.3.2.2. Hydrogen isotope standards ..................................................................359
15.3.2.3. BAM hydrogen standard .......................................................................359
15.3.2.4. Bi-implaoted (Harwell) standard ......................................................... 359
15.3.2.5. BAMIIRMM Sh standard ..................................................................... 359
347
.-~
...
#_~.
Chapter 15
15.3.2.6. a-Si standard ....................:...................................................................... 360

15.3.3 Electronic gain ....................................................................................................360
15.3.3.1 Fractional channel numbers .................................................................361
15.3.3.2 Correlated gain and offset .....................................................................361
15.3.3.3 Detector resolution .................................................................................361
15.3.3.4 Pulse height defect ..................................................................................361
153.4 Scattering angle ..................................................................................................362
153.5 Detector resolution .............................................................................................362
15.4 ALGORITHMIC ISSUES ...................................................................................... 363
15.4.1 The Rutherford cross section ............................................................................363
15.4.2 Electronic stopping cross sections ....................................................................364
.....- . -.. -.-.~- 15.4.3 Plural and multiple scattering .......................................................................... 365
15.4.4 Depth resolution ................................................................................................. 366
15.5 ACCURA TE IBA ............................................................................,........................366
15.5.1 Uncertainty estimation ...................................................................................... 367
15.5.1.1 Type AaiId Type B ................................................................................367
15.5.1.2 Uncertainty budget ................................................................................367
15.5.2 Spectral ambiguity .............................................................................................368
15.5.2.1 Multiple spectra .....................................................................................369
15.5.2.2 Restricting the state space .....................................................................369
15.5.2.3 Butler's example re-analyzed ................................................................369
15.5.3 Model-free analysis and Occam's razor ..........................................................370
15.5.4 Common pitfalls in data analysis ..................................................................... 371
15.6 '.' UNWANTED TARGET-BEAM INTERACTIONS ............................................374
15.6.1 Beam-induced heating .......................................................................................374
15.6.2 Beam-induced radiation damage ......................................................................375
15.6.3 Beam-induced sputtering ..................................................................................376
.15.6.4 Charging in insulators ....................................................................................... 376
~~-~:~--=--_~".~~_?:6.~_!~oton emission in insulators .........................:............
.378
u .....................................
15.7 OTHER EFFECTS ..................................................................................................378

15.7.1 Surface and interface roughness ............................................................................ 378
15.7.2 Target nonuniformity .............................................................................................. 378
15.7.3 Thin film units .......................................................................................................... 378
REFERENCES
..................................................:............................................................ 380
348
Pitfalls in Ion Beam Analysis
15.1 INTRODUCTION
Ion beam analysis (iliA) IS a quantitative analytical
technique, and in this chapter, we intend to show how to avoid
many pitfalls when determining elemental depth profiles
accurately with light-ion Rutherford backscattering spectrometry
(RBS) using megaelectronvolt ion beams. IBA can, of course,
use various other beams and use these for a variety of other
purposes, including profiling of crystalline defects. The present
discussion is usually, but not always, easily generalized to these
other cases. We will therefore also cover pitfalls in a number of
other important examples.
We will, in general, be guided by Chapter 12 of the previous
edition of this handbook (Tesmer and Na~tasi, 1995). However,
that chapter contains many subtle discussions that we will avoid
or simplify, referring the advanced IBA user back to it. Today,
most laboratories tPat are starting to usc IBA have the new
generation of accelerators, eliminating many of the problems
associated with the older accelerators that the previous chapter
addressed.
A new generation of software is available that renders
obsolete many of the nwnerical examples of the previous
handbook. We will assume that users interested in accurate work
will have access to one of the codes described and evaluated in
the 2006 International Atomic Energy Agency (IAEA)
sponsored IBA software intercomparison ex:ercise (Barradas et
al., 2007, 2008). where it is shown that IBA spectra can be

treated numerically with a confidence of about 0.2% and that,
currently, the best ab~olute experimental accuracy available is
0.6%. Data analysis software for IBA is discussed in Chapter 14.
We therefore intend to describe sources of error (pitfalls) greater
than about 0.25%; we believe that users should be aiming at
accuracies of 1% or better. -- _:.
..; -... --.
-
- .
-~~~~.'~
- .
is a powerful and sensitive technique, but it is usually only

semiquantitative, and it is generally unreliable at interfacc~
(unless the system is well known)_ IBA, on the other hand, is
particularly good at interfaces and has an accwacy that is easily
traceable to international standards, making it suitable for
standards work. We show in this chapter how to best capitalize
on these strengths.
For parts of this chapter, we take a measurement of the

implanted fluence of As into Si (Jeynes et aI., 2006) as a worked
example of various pitfalls in doing accurate IBA. The collected
spectra are shown in Fig. IS.1. Two important equations are
relevant to this discussion: Equation 15.1 simply describes the
number NA (in atomslcm2) of As atoms, which is given by the
measured area AA of the signal of element A (in this particular
example. A is As) (in counts)
(15. J)
where Q is the number of incident particles (the '~collected
charge") and.n is the detector solid angle (in steradians, sr). cr' A
is the differential cross 'section (in cm2/sr) of element A and is a
function both of the scattering angle and of the beam energy E
as it decreases from its initial energy, Eo, as it penetrates into the
sample.
In Fig. 15_1, there are two signals of analytical interest. The

ftrst is obviously the As signaL However, in Eq. (15.1), we also
need to know the product of the charge and the solid angle. Qfl.
This is detennined from the Si signal and the known energy loss
of the He beam in a silicon matrix. For backscattering with the
detector at an angle of 8 to the beam and normal beam incidence,
the surface yield. Yo (in counts/channel), for element A in matrix
AB is given by the equation
..
.
.--~
~g
.. ------ _ . .
-----.-~-.-
.. --~----
..
~--
- -...
~----
-~"(J5.2)-The nuclear physics background of the old generation of IBA~'" .-.

practitioners has become inaccessible to many of today's users;
rnA is of great value in materials characterization and is used
where fA is the fraction of the matrix that is element A, f1 is
the electronic gain (in eV/channei of the analog-digital
increasingly by materiaLs scientists who have not had a nuclear
physics training. This chapter is intended to be of particular
converter), and [.o]~ (in eV em2) is the energy-loss factor for
value to this type of user.
the signal of A from the surface of matrix AB given the beam
Accurate elemental depth profiling by IBA is of great value
energy and the scattering angle (see Chapter 4).
in many modem thin-film technologies: the most obvious
comparable technique is secondary ion mass spectrometry
(SlMS), which is often used for these types of problems. SlMS
349
Chapter 15
5O
a
'. <:hanhei
.<50
40
..
450
t:
"o
.".
320
Channel
420
. Depthn,"
240
..
d
,-ao
FIG. 15.1. Accurate analysis of As implanted fluenee in Si. Ca) Nominal implant 5 x 10" As/cm' at 100 keY. L5 MeV He RES
spectra from two detectors collected simultaneously, and 0.5 mC collected charge, with fits. The beam is incident normally on
sample such that it channels in the substrate. and the near-surface region of silicon amorphized by the ~jon beam is clearly seen.
small 0 peak from the surface oxide is also clearly distinguishable. The As, Si, and 0 edges are marked for the detector with the
solid angle. (b) Detail of the As signal. (c) Expanded detail of the As signal, with the nonlinear pileup correction function
Pileup-corrected As signal from each detector plotted on a concentration versus depth scale (using a Si density of 5 x 'n"I,,~'\
(Reproduced from Jeynes et al., 2006.)
Equations (15.1) and (15.2) have been simplified without any

loss of generality. They should both really be integrals over the
depth of the signal, noticing that cr' varies as llE2, and that [E] is
also a function of depth. The standard IBA codes have been
validated to do these integrations correctly (Barradas

2007. 2008).
For the example of Fig. 15.1. determining the number of
atoms present [using Eq. (IS.I)] depends on an
350
measurement of the product of the charge and the solid angle,

QQ, from the height, Y, of the silicon substrate signal [m.ing Eq_
(15.2)], where the number of Si scattering centers is implicit in
the tJ[c:] ratio. In the rest of this chapter, we elaborate on
various of the issues raised by these equations (as well as some
others).
significant number will have high energies that could be on the

order of kiloelectronvolts. Some electrons will therefore escap~
the Faraday cup (S) through the secondary aperture (A; this
aperture is not struck by the primary beam) and must be
suppressed. There is also the possibility that secondary positive
ions (or low-energy multiply scattered primary ions) could also_
escape the' Faraday CUpj they will be attracted by the suppression electrode and will give rise to tertiary currents (which will be
amplified by secondary electrons liberated and repelled from the
suppression electrode by the ion impacts). Tertiary currents are
not seen by the current integrator (1) in this design. These
secondary particle effects can be very large for some insulating
samples and at low (kiloelectronvolt) beam energies.
15,2 LOST BEAM AND EVENTS
15.2.1 Charge measurement

The collected charge Q in Eqs. (15.1) and (15.2) is a very
important quantity that it is not very easy to measure. In the
implantation of Fig. 15.1. the As implanted fluence was
detennined during the implantation by collecting charge in
Faraday cups of the design shown in Fig. 15,2. This is a
demanding application because large heavy-ion beam currents
are used in implanters and the beam being measured usually
faUs only partially on the primary aperture. Jeynes et al. (2006)
shows that this design js capable of charge measurement in this
application with an accuracy of around 1%.
For IBA applicatio~s, such a well-designed Faraday cup is

usually impractical, as the target might be large or part of a more
or less complicated manipulator and., in any case, access is
needed to the detector, which could be of various types (particle,
photons including X-rays, time-of-flight, etc.) or in various
geometries. However, when direct charge measurements are
made, the geometry of the actual apparatus should be critically
compared with the "ideal" Faraday cup described. It should be
remembcrcd. thaI photon impact can also liberate secondary
electrons, and for samples that generate many photons, this
effect can be large (Venkatesan el al., 1984).
Sl
Electron suppression can work just as well with magnetic

fields as with electrostatic fields, and small, strong permanent
magnets are now readily available. However. the use of
electrostatic and magnetic fields together is to be avoided
because, when both fields are present, electron paths can be very
~
complicated and any possible escape paths will probably be
I ~
found. With electrostatic fields, the suppressor action should be
' -_ _ _~~_~_ _ _ _ _ _ __"_'_ __,._::_:--__,_--'-~ cheeked pejiooically:.as insulatiIig depo'sits can build up on the
-- surfaces over time, - attracting fIxed charges that can entirely ~
FIG. 15.2. Faraday cup design. The sensor S is an equipotential
compensate the applied field.
volume in a grounded case. Secondary electrons from the
primary aperture P are suppressed by electrode S 1, and any
A variety of methods have been used to determine collected
secondary electrons escaping through lhe secondary aperture A
charge apart from direct measurement from the target.
are suppressed by electrode S2. Suppression voltages are
Transmission Faraday cups were described by Sitter et aL
typically 300 V.
(1982) and by Ptiszti el al. (1990). In both of these instruments, a
rotating or reciprocating vane cuts the beam periodicallyJ and
At issue is not the simple measurement of charge in a current
the charge deposited on the vane is measured with a precision of
integrator, which can (and should!) be calibrated against
about 1%_ The pitfall to avoid in this sort of measurement is
standard current sources with picoCoulomb accuracy. There are
forgetting that the secondary electron yield is enormous when
two points to note from Fig. 15.2. First, the incident beam will
the beam strikes the edge of the vane and that suppression of
generate a large number of secondary electrons at the primary
these electrons is as important on the downside as on the upside
(beam-defining) aperture (P), many of which will have a
of the vane. A passive instrument where a fIXed mesh is uscd
significant momentum toward the target. The secondary electron
instead of a chopper was described by EI Bouanani ef al. (2006).
current is usuaUy comparable to the incident beam current and
Rotating vane instruments were reported by Piel et af. (1994)
often exceeds it, in the case of heavy-ion beams, by orders of
and by Giuntini and Manda (1994) that collected backscattered
magnitude. It is important to carefully suppress such electrons.
particles from the vane, instead of charge. For these instruments,
the vane is coated with gold, and the particles scattered from
Second, the beam striking "the target will liberate many
secondary electrons and photons (X-rays). Most of the
gold are uis~Timinated in a single-channel analyzer (SeA). Thi!i
secondary electrons will be low in energy 10 eV), but a
is quite a troublesome arrangement because the signal changes
351
.""
i~
Chapter 15
for different beams and one really needs to see the spectrum to
to unwanted contaminant beams that arrive at the target together
set up the SeA correctly. Users have also found that a gate is
needed to exclude edge signals from the detector. It is possible,
but not easy, to obtain excellent precision (1 %) with such a
design.
with the desired species. Such problems are caused by

atomic or molC(:ular particles whose kinematic properties
mimic the acceleration and deflection kinematics of the
desired particle,
Transmission detectors take a fraction of the beam, of course,
and this fraction must be measured directly with a downstream
unwanted particles that reach the target because of a low~

probability charge exchange sequence occurring upstream of
the target, or
Faraday cup (which can be of ideal design). The great virtue of

transmission detectors is that they are completely independent of
the target, which can then be manipulated, be heated or cocled,
particles that have completely incorrect kinematics but arrive

at the target because of (wall) scattering.
or be made of any material without affecting the charge

measurement in any way.
One of the simplest ways to test for a mixture of "equalmass" beams in single-ended machines is to detect elastically
scattered projectiles from a thin self-supporting Au target. An
energy spectrum analysis with a standard surfacebanier
detector system will readily distinguish the fragments of
molecular im,Euritics, such as mixtures ofR3+, HD+, and JHe+ or
mixtures of He" and D 2", which can arise through ionsource :.;
. memory effects.
1-
. '-- -'15.2.2 Beamline-charge 'exchange

In modern systems with good vacuums, beamline charge
exchange is an unusual problem. Users should be aware,
however, that there is a possibility of significant beam charge
exchange where the vacuum is poor. This problem was reviewed
in considerable detail by Allison (1958). when it does occur,
there is a probability both of lower and higher charge states.
.~
The consequences of scattering the three ion species H3+,

lID+, and JHe+ from a thin Au scattering foil, when the incident
species have equal energy (l MeV), are listed in Table 15.1. We <1
assume that the detector is located at 90 0 (laboratory angle). The
three beams would give rise to three distinct peaks:
1
Accelerators used for implantations often scan the ion beam

to implant the sample uniformly, and in such beam lines,. a
"neutral trap" is always incorporated, where a neutral trap is a
voltage applied to transmit the beam past a kink in the beam
tine. The neutral beam is not deflected and is intercepted by an
aperture. For rnA, the beam lines do not always have neutral
traps, and the beams are not usually scanned, so that neutral
beams will give an unchanged IBA signal but will not be
detected by the current integrator. Because the beam lines are
for lID', two peaks corresponding to 0.33 MeV protons arid :~

0.67 MeV deuterons; and
,,~
usually long, the neutral fraction of the beam can be significant

if the vacuum is poor.
:'~
The relative intensities are distinctive. The entries in Table "'-,J
15.1 are based on a simple (Z,Z,IE)' dependence of the elastic '~

~
(Rutherford) cross section for the dissociated fragments, with the
I?nematic factors approxfuuted as unity.
. if th~ beamline pressure is 1 mPa (10-' mbar) and the beamline
15.2.3 Particle identification

A precise knowledge of the parameters of the incident
particle is essential for all ion beam experiments. We assume
here that the accelerator has only an analyzing magnet [with
relatively poor mass resolution compared to that of accelerator
mass spectrometry (AMS)] to perform momentum selection of
the beam. Many accelerators suffer from periodic problems due
for 3He+. a single peak at 1 Me V energy.
Conversely, a He+ beam can suffer further charge-stripping

collisions, giving a significant He2 + component. As an example,
---'is- 4 m long, then the'~euir8lization probability for a 450 keY
He+ particle is about 1%, but the probability of electron
stripping to form He2+ is higher at about 2%. For a 450 keY He2+
beam in this beamlioe, the probability of He+ is 5%, but the
probability of the neutral is only 0.5%. For three times the
energy, one might expec~ the neutralization cross sections to fall
by a factor 3 or so. The cross section for He+ to change to He2+
falls from 66 megabarns (Mb) to 37 Mb as the He+ energy
increases from 450 keY to 1700 keV, but the cross section of the
reverse charge exchange, of He2+ changing to He+, falls from
170 Mb (for 450 keV He') to 5 Mb (for 1700 keV He').
for H,, a single peak at 0.33 MeV energy;
- - - - - - - - - - - 'j
Table 15.1: Relative particle yields and energies for several .~
~
mass 3 ions incident on Au at 1 MeV.

Relative yield(s) of scattered particle(s)
Beam
H
0.33 MeV
0.67 MeV
'He
1.0 MeV
.~
o
i
,
";':
H,'
27
"
IID+
_ _ _ _ _ _ _ _9_ _ _ _ _ _2_._2_5________
3
He
------------------------ i
.>::->
'~~
Another example of the use of this Au foil scattering

technique is the identification of 160 ions in the presence of ":He.
ions-a curiosity observed by Henunent et al. (1975) when poor
352
beamline vacuum developed, thereby allowing a 160+ impurity

beam co-accelerated to 2 MeV with the 4He+ beam to undergo
charge exchange to form \60 2+ in the drift section between the
accelerator and the analyzing magnet. This process allowed two
beams of "identical" rigidity, MEJq' = 8 MeV arnu, to be passed
by the magnet.
and there is a nonzero probability for pulses to occur at any time

separation, even simultaneously. Surface barrier detector
preamplifiers, for example, can have a rise time of about 30 DS:
and a typical resolution time for an amplifier PUR circuit is 500
The molecular interferences are not problematic for tandemtype accelerators, where there is an injection magnet at the
source, an analyzing magnet following acceleration, and a'
(collisional) stripping event at the high-voltage tenninal.
.
Wall scattering can be minimized using a series of apertures;
however, care must be exercised so as not to restrict beamline
pumping speed excessively_
15.2.4 Dead time and pulse pileup

We restrict our comments to "singles" experiments (as
opposed to coincidence experiments) and solid-state detectors,
which are IDost common for ion beam analysis applications,
although much of this section is common to all types of analysis_
ns. However, where PUR is used, any pileup events that are not
rejected are considered nearly simultaneoWi, that is, the
amplitude of the resulting sum pulse is nearly the sum of the
amplitudes of the pileup pulses. Where PUR is not used, the sum
pulse can be of (nearly) any amplitude.
It is possible to eliminate pileup almost completely by
detennining both the momentum and the energy of the detected
particles using a AE-E detector telescope (Gurbich, 1996) or
using a time-of-flight detector with a pulsed incident ion beam
(Gurbich and Komilov, 1991). However, these approaches are
significantly more complex, and the llE-E method requires very
thin 6.E detectors for lower-energy He beams. Instead, we

concentrate here on correcting spectra collected with the
simplest detectors.
Figure 15.3 shows a spectrum with significant pileup (and
Dead time is generated in the counting system by the time

taken to process each pulse. This is typically dominated by the
analog-to-digital conversion (ADC) time (on the order of 10 ~s)
for particle detectors and by the preamplifier pulse processing
time for Si(Li) X-ray detectors (typically greater than 10 J.LS). ill
all systems, the ADC should be capable of reporting to the Wier
the actual dead time for each counting channel. This dead time
must include any time spent in handshaking between the various
electronic components. Where the ADC is a multichannel
analyzer (MeA) or where the host computer is multitasking,
there could be significant extra system dead time that affects all
with PUR). The Au signal appears at channel 150, and the

electronic pulser signal appears at channel 480. AU of the counts
above the Au signal are pileup cotmts (except for the pulser),
and the shape of the pileup signal is recognizably an
autoconvolution of the spectrum. The two-pulse pileup is
accurately calculated (with or without PUR) with the algorithm
of Wielopolski and Gardner (JV &0) (Wielopolski and Gardner,
1976, 1977; Gardner and Wielopolski, 1977). Tbis is a slow
calculation, involving Deb......,I' operations, where ~hmn~1 is the
number of channels. The autoconvolution method (van Lieshout
et al., 1966), improved by Amsel et 01. (1992), is faster given
that it involves only Iletwm~? operations, but it is not as accurate.
channels. In .any . case, where multiple detectors are used
simultaneously, it is more important to accurately determine the -:.:.:""- Where the pileup correction is sigillficant,best calculation
relative live charge between the detectors than it is to determine
should be done: refer to the manual for the IDA code being used
the charge that fell on the sample (the true charge) during the
to determine what this is. Some codes will also make an
collection time. For accurate work, it is essential to understand
approximate calculation of three-pulse pileup (e.g., Barradas and
the operation of your spectroscopy system in detail.
Jeynes, 2008). We note that the best pileup calculations for most
current particle-induced X-ray emission (PIXE) codes do not use
Pulse pileup occurs when separate pulses come too close
the best algorithms (Barradas and Reis, 2006). We note also that
together in time and are not recognized as separate by the
the W &0 algorithm assumes parabola-shaped pulses but can be
spectroscopy system. There are two cases for pulse pileup
extended to other shapes as well, and it requires knowledge of
depending on whether a pileup rejection (PUR) circuit is present
the time for a pulse to reach the maximum. which depends on
or not Where PUR is used, pileup events will still occur because
the amplifier used (normally given in the amplifier
every spectroscopy system has a pulse detection time resolution
documentation).
the
353
--~
Chapter 15
Au
.C
Mg
i1
o .
.12
~~l
.~.
:.
. .. .
,.
..
..~
...
o.
512
Channel
FIG. 15.3 .. Pileup with pileup rejection. Data obtained from 707 keY 'It RBS of a AulMg multilayer on C, with a count rate of 4.1
kHz, an amplifier shaping time of 500 ns, and a PUR time resolution of 550 ns. A distinct 0 contamination peak can also be seeD. Fit
including pileup calculated with Wle1opolski and Gardner's algorithm is as a solid line.
When doing accurate work that depends on a good pileup . ~

calculation, it is important to use a constant beam current. The :~
algorithms used by IBA codes take the live time as the input ~
from which the average count rate is calculated. If the count rate
fluctuates, the pileup cannot be calculated accurately. The W&G . ~
calculation has no free parnmeters, although the collection time,
the shaping time, and the PUR time resolution can be treated as
fitting parameters. Analysts should note that large, rapid ;
fluctuations in the incident beam intensity described by Sjoland ~~
el al. (1999) invalidate the simple W &G calculation, aod every \
effort should be made to avoid fluctuating beams. Thankfully. A
modern accelerators generate very constant beam currents. .J.;
Sj51and et al. (1999) also describe the more difficult case of .1
scanning microbeam PIXE, where, even if the beam current is ,1
constant, the count rate generally is not
"1-
The W&G algorithm was applied in Fig. I 5.1 (c) to determine

the pileup correction function of the As signal. This figure
illustrates a dangerous pitfall: One might think that this pileup
background is nicely linear from the pileup behaviors above and
below the signal. However, it is very clear from the calculated
pileup signal shown that the pileup is actually strongly
nonlinear. Of course, the pileup at any channel is the sum of the
pileup events that are detected in that channel and events that
should be in that chancel but that are actually piled up and end
up in higher channels. For strong signals, the pileup background
can be negative because the latter events outweigh the former.
For example, the Au signal of Fig. 15.3 is underestimated by
1.5% even for the relatively low count rate of 4 kHz.
Notice in Fig. 15.3 tha~ below about channel 30 there is a
large "noise" signal. Usually. this is discriminated by the lowerlevel discriminator (LLD) of the amplifier to reduce the count
rate into the AnC, but pileup will occur anyway, regardless of
whether the (real) counts appear in the spectrum. In this case, the
LLD was on, but ao assumed (ad hoc) signal was arbitrnrily
included in the spectrum shown, such that the pileup above the
Au edge and above the pulser signal were correctly reproduced.
The extra (unfitted) counts above about channel 505 are an
artifact of the ADC, which integrates all of the high-energy
counts instead of truncating them. The pitfall to avoid is relying
on even a good pileup calculation such as W &G too near a
leading edge where the LLD will distort tbe calculated pileup
signal.
Dead time and pileup arise in all of the electronic

components (preamplifier, amplifier, ADC. etc.) because of the
Pulse rate. Solid-state detectors should he operated at pulse mtes
such that the pulses are separated by an average time interval
that is much greater than the dead time inherent in the rest of the
electronic circuitry. A 10-J.1.S dead time per pulse implies a
maximum counting rate of 100 kHz, but of course, the pulses
come randomly in time, and dead time will increase rapidly (and
nonlinearly) for count rates higher than about 10 kHz.
j
,.
_. .:.~.,
>.J
.i
7.
.,~
~
,}
use a test pulse (see the pulser .~
It is a good practice to always

signal at channel 480 of Fig.
15.3) at the appropriate
),'J
'.~
354
preamplifier input, that is, to introduce a signal that is periodic

with known frequency (typically about 60 Hz). Thus, the total
number of "missing" pulses due to count-rate losses during the
acquisition time can be calculated from the measured area under
the pulser peak.. If the pulser is gated for dead time, then it will
reflect pileup los.ses. When using a Faraday cup that rotates into
the beam periodically! there are no pulser losses when the beam
is hitting the cup. Here, the precise duty cycle of the cup must be
measured, and its effect must be factored into the correction.
Another way to measure counting losses using a pulser is to
trigger the pulser with an output (usually scaled) from the
current integrator. This way, if the beam goes off, so do the
pulses, and because one usually counts for a preset charge, the
number of Jlulses is also known. The pulser is in any case
valuable for monitoring thc stability of the electronics during the
data collection run and also for monitoring the noise
performance of the system.
In general, counting-rate losses will be nonnegligible for total

pulse counting rates exceeding 1 kHz, which applies to almost
all RBS measurements. Raising the LLD of the amplifier and/or
ADC can decrease the dead-time problem, provided that the
low-energy region of the spectrum is extraneous to the
experiment (but doing so will also distort the pileup calculation
as described in the preceding paragrapbs). See Knoll's exceUent
text (1989) for more detail.
resolution, the performance can be improved by reducing the

capacitance of the detector and connecting cable and by using a
preamplifier with a cooled field-effect transistor (FET). Ii the
specified noise level (measured by a precision pulser) is Dot
obtained, periodic noise is most certainly involved. Additional
infonnation can be found in Goulding and Landis (1982),
Radeka (1988), and Morrison (2007).
.
Tracking and removing periodic noise can be very tedious,

and there is no single recipe that reliably leads to success.
Basically, there are two strategies for doing so: on one hand,
removing the sources of noise and, on the other hand, improving
the shielding and reducing the sensitivity of the detector chain
againSt the disturbance. In the following list, we describe tests
that can be done and measures that can be taken in pursuing both
of these strategies. It has to be mentioned that some actions that
improve conditions in one situation might be detrimental in
another, depending on the electrical grounding scneme used and
the nature of the noise source.
Check cable shieldings and connectors; the ground lead of
the preamplifier input connector especially has to be finn
(check by careful twisting). Do not use unnecessarily long
cables.
Beamlines and vacuum vessels are not necessarily good
conductors and on a well-defined ground potentiai. Try to
connect the flange that carries the electrical feedthrough to
the detector with a solid copper cable to the electrical ground
of the detector electronics.
To conclude this section, we urge the reader to avoid the

pitfall of counting too slowly in an attempt to avoid pileup and
dead-time effects. It is a misuse of spectroscopic electronics of
excellent linearity-and of highly expensive accelerator
facilities-not to count as fast as is feasible. We emphasize that
the W &0 pileup calculation is exact for two-pulse pileup (for
parabolic pulses) and good-quality electronics will also give
very precise dead-time values, so there is no need to reduce the
count rate to avoid these effects, except, of course, where the
signal-to-noise ratio is critical, such as when looking for trace
heavy elements in a light matrix where the background is all
from pileup. In any case, there are diminishing returns for
decreasing the counting rate, as pileup is linear with counting
rate and so is dead time for low count ratcs.
Extra ground cables have to be of large cross section (10

mm2 rather than 1 mm 2) and connected fIrmly. If other cables
carry ground along the same line, avoid producing ground
loops by ~lightly twj~~g :abl~s aroun~_each .other..::::.
If you have three-phase electrical power in the laboratory that

is split into three two-phase current networks, never mix
phases for one experimental setup. Acquisition electronics
(including data-collecting computers) shquld be powered by
the same phase. If necessary, install an extra power cable to
the computer or control room.
If possible, try "star grounding" of your electronics to avoid
ground loops; that is, llse only one single connection to
ground potential in the whole system. In this case, the
shielding of the detector cable has to be insulated from the
vacuum vessel.
Count fast and do the corrections correctly!

15.2.5 Electronic noise and ground loops
State-of-the-art ion-implanted silicon detectors reach an
energy resolution of 10 keY [full width at half-maximum
(fwhm)] for helium ions and 5 keY for protons. The contribution
of electronic noise from a standard preamplifier is about 3 ke V.
For high-resolution measurements, it is therefore important to
avoid picking up any other disturbance from the electromagnetic
environment. If a gas ionization chamber is used, the charge
produced per kiloelectronvolt of panicle energy can be more
than a factor of 6 lower than in a silicon detector, and electronic
noise becomes more important by this same factor. If capacitive
noise from the preamplifier is the litpiting contribution to the
,)
. Jtk:!...........
Avoid strong high-frequency emitters close to the

experiment. These are often chopped power supplies,
turbopump controllers, PC monitors, and so on. If emitters
are inside the vacuum vessel (for example. turbopumps), the
detector can act as an antenna: rethink. the shielding scheme.
To track the sources of periodic noise, run a fast Fourier
transform on an oscilloscope. In some cases, a simple ferrite
filter can help.
355
Chapter 15
Microphonic noise (i.e., mechanical vibrations of the

detector) can produce similar disturbances. If produced by a
turbopump, it can even have the same frequency as the noise
induced by the electromagnetic field.
15.3 CALIBRATION OF FIXED PARAMETERS
In Eq. (15.1), the unknown experimental parameter is the
product of the charge Q and the solid angle Q. The measurement
of charge was treated in Section 15.2.1. In Eq. (15.2), the
electronics gain 6. appears, as welJ as the product of the charge
and the solid angle, Qn. In both equations, the scattering angle e
is implicit in both the scattering cross section cr' and the
integration path lengths. This section treats the separate
.. __ detennin.tions of n,. 8, and e, whicb are (usually) fixed in any
one run.
Solid angle is troublesome to measure accurately at the 1%
level, as is routine charge collection in RBS systems, and
analysts bave used standard (certified) saroples for decades to
avoid routine use of absolute values of charge and solid angle. It
is difficult to determine the detector-to-target distance with a
precision of 1% (generating a 2% uncertainty in the solid angle).
It should be noted tha~ if the product of the charge and the solid
angle is calibrated on one beam, then care should be taken that
no relative error is introduced through current integration when
perfonning an experiment in the same geometry with a different
beam. The comments on determination of the solid angle apply
to particle detectors, where the intrinsic detector efficiency is
. ".known to be 100% for energies exceeding -30 keY. For I-ray
and neutron measurements, it is more difficult to achieve an
accuracy of .... 1% for the overall detection efficiency.
15.3,1 Energy
Accelerator energy calibration is critical for accurate work.
Clear pitfalls here were highlighted in the first edition of the
___ handbook (also s~e. Section 15.3.2), which pointed out that the
fust RBS round-robin experiment had vel)' disappointing 'results
partly because of very poor energy calibration (see Section
15.3.2.4). For RBS, the spectra have similar shapes for all
energies. Of course, the energy loss is sensitive to the beam
energy, but it is not nearly well enough known for use in energy
calibration. In any case, the energy loss for any given layer can
rarely be measured with a precision exceeding I %. However,
because the cross section is proportional to llE2 but linear in
gain, an energy error wilJ propagate through to errors in the
determined layer thicknesses. For elastic (non-Rutherford)
backscattering spectrometry (EBS) [and nuclear reaction
analysis (NRA), there are very often strong resonances that, in
some circumstances, can lead to very large errors consequent on
an energy calibration error. PIXE is the one case where energy is
a second-order effect, and one often flnds that PIXE laboratories
arc rather cavalier about energy calibration. This is a potentially
serious pitfall for those laboratories if they try to do a more

complete analySis using simultaneous particle scattering, as is
DOW desirable with the new codes that handle these mUltiple
techniques self-consistenUy (such as the DataFwnace code,
discussed in Chapter 14).
One might think that, because a1l accelerators have resistor
chains to distribute the terminal potential down the accelerator
column through the column curren~ the terminal potential could
be easily determined through Ohm's law, because the resistol1l
have ftxed values. However, these resistors have extraordinarily
high values (typically -10
and operate at very high voltages
(typically -40 key), and high.value high-voltage resistors are
known t%llow Ohm's law only approximately and, in any .
case, Dot reproducibly at the sub-l % level.
an)
Therefore, it is necessary to calibrate indirectly, using

establisbed standards (see Tables 15.2 and 15.3). The two
primary accelerator energy calibration points are the
27 AI\!"y)"Si reaction at 991.90 O.04 keY and the threshold of
. the Li(p,o)'Be reactioo at 1880.60 0.07 keY. These values
were taken from the authoritative work of Marion (J966) and are
still the currently accepted values [but note the more precise
value from White el at. (1985) for the latter]. The previous
edition of this handbook also included a table of accelerator
calibration energies (Tesmer and Nastasi, 1995~ Table A19.1,
Appeodix 19), but it should be made clear that many of those
energies were provided for convenience only. are not the
accepted values, and should not be used in critical work. For .
example, ~the previous edition of this handbook listed a
16
resonance energy of 3035.9 2.3 keV for the 16 0 (0.,0.) 0
reaction, but the IAEA-recommended evaluated cross-section
code SigmaCalc (2009; also see IBANDL, 2009) gives a
resonance energy of 3044 ke V.
Secoodary calibration points can frequeotly be used 10 verify
the reproducibility of the accelerator energy, usually at a lower
precision and almost always at a lower absolute accuracy. EBS
resonances are very convenient faT this purpose, especially given
that there are a number of resonances at high energies and for
'He beams. See Table 15.4 for suggestions, but note that the
energy positions of the resonances do Dot (usually) have an
evaluated uncertainty, even though the SigmaCalc CJOSS sections
are evaluated. The exception to this is the C(p,p)C reaction, for
which the energy is very well knO\VIl. The estimated
uncertainties_ were taken from the original cross~sectiOD
measurements. To make best use of these resonance energies,
considering that the line shapes can be strongly skewed, one of
the rnA codes that can handle EBS correctly should be used t.o
fit the spectra obtained. Very high precision can be obtaine~ 1f
the rnA codes are used carefully, even with the very Wide
resonances.
356
tr
Table 15.2. Primary (recommended) accelerator calibration points for protons [from Marion (1966), except where
otherwise stated]. Note that, for threshhold reactions, the 2003 masses are used (Audi et aI., 2003); also see the
cODvenient calculator at www.nndc.hnl.gov/qcalc.
Energy
(keY)
340.46
872.11
991.90
1747.6
1880.60
1880.443
1880.356
3235.48
4234.43
5796.9
5803.73
5803.621
6460.47
7026.56
9193.85
9510.55
12728
14233
14230.75
19641.9
0.04
0.02
0.04
0.9
0.07'
0.020'
0.081
0.29
0.31
3.8'
0.12'
0.100
0.15
1.6
0.20
1.46
10
8
0.02
1.1
Reaction
fwhm
(key)
19F(p,ay)160
I'F(p,ay)160
27 Al(p,y)"Si
IJC(p,y)I'N
'Li(p,D)'Be
'Li(p,n)'Be
'Li(p,n)'Be
IJC(p,n)I'N
I9F (p,nt'Ne
27Al(p,n)" Si
" A1(p,n)"Si
27Al(p,ni7Si
"S(p,n)"Cl
"'Ni(p,n)6IlCu
"Fe(p,n)"C
s8Ni(p,ni 8Cu
I6O(p,p)160
l2C(p,p)12C
12C(p,p)12C
12C(p,n)12N
2.4
4.7
0.1
0.8
Comment
Marion, 1966
White ef al., 1985
Marion, 1966
White e/ 01., 1985
<2
<1
1.2
. Huenges et aJ., 1973
*Both 1966 and 1984 values given for comparison.

Table 15.3. Accelerator calibration points from He threshold reactions. Note that 2003 masses are used (Audi e/ aI.,
2003), unless otherwise stated; see the convenient calculator at www.nndc.bnl.gov/qcalc
Energy(keY)
1436.7
1437.9
1436.6
1435.82
2966.1
2965.0
2964.9
3796.54
4380.035
6088.04
6620.54
8131.37
11337.86
15171.49
0.5
0.6
0.5
0.14
1.7
1.5
1.5
033
0.612
0.32
1.6
0.67
0.67
0.36
. Reactfon--.:"- ._ ..,,....,
"C<'l"1e,n)"0
l2C('He,n)I'0
12C ~e,n)140
"c( e,n) 14 0
'Li{'He,n)'B
'Li{'He,n)'B
'Li('He,n)'B
l'O(,He,n) 18Ne
'Li(a,n)l'll
I'N(a,n)I'F
'Li(a,n)'B
1'N(a,n)I'F
1'C(a,n)uO
l60(a,n)I'Ne
.. - -.--....."- Comment
Experimental (two measurements; Marion, 1966)

Roush et aI., 1970
Theoretical (Mattauch et aI., 1965)
Theoretical
Experimental (one measurement; Marion, 1966)
Theoretical (Mattauch et aI., 1965)
Theoretical
Theoretical
Theoretical
Theoretical
Theoretical
Theoretical
Theoretical
Theoretical
357
.I~ ..
~ ~~
Chapter 15
Table 15.4. Secondary energy calibration points based on SigmaCalc, version 1.6. Note that uncertainties were derived from
the original literature, not from SigmaCalc. and are standard deviations, not fwhm.
Energy
(keY)
3044
3885
4265
3470
1483
1748.5
3198
1734
Reaction
Reference
(keY)
5
5
5
16O(a,a)160
"O(a,a)I'O
"C(a,a/'C
"O(p,p)I'O
"Mg(p,p)''Mg
1'N(p,p)I'N
1'N(p,p)I'N
0.2
12CSpp'tC
fwhm
10
2
27
2
0.3
9
6
45
SigmaCalc (unpublished to date)

SigmaCaic (unpublished to date)
Gurbich (2000)
Gurbich (1997)
Gurbich and Jeynes (2007)
Gurbich (2008)
Gurbich (2008)
Gurbich ( 1998)
A competent operator should be able to obtain a terminal

voltage calibration at an absolute accuracy better than 0.1 %, but
this is not easy, partly because this level of accuracy is never
easy to achieve, partly because precision at this level is usually
limited by counting statistics (so that significant beam time_ is
required). and partly because such accuracy requires careful
attention to the quality of tbe sample surface used. An
aluminium sample should be well polished (shiny) and clean,
and a lithium sample (which is very reactive) should have a fresh
surface prepared each time in situ by being cut with a knife
under an argon atmosphere. Bear in mind that one needs only a
13-nm (5 J.!g/cm') alumina layer or a 4 J.!g/cm' carbon layer to
give a I ke V energy loss for protons at 1 MeV.
15.3.2 Solid angle

In this section, we consider the detennination of solid angle,
one of the critical parameters in Eqs. (I5.1) and (15.2). It is
mostly detennined indirectly using a standard, and we consider
first NRA standards and then various RBS standards.
Whenever the cross section varies significantly across the
finite solid angle of the detector, as is the case for many nuclear
reactions, as well as for non-Rutherford elastic scattering, care
must be used to properly integrate over the sol.id angle with
respect to the angular dependence of the cross section. Care
should also be taken (especially for higher-energy incident .~
beams, which can give rise to unexpected nuclear reactions) in 'J
choosing an aperture that is opaque not only to megaelectronvolt '.1
a-particles but also to more penetrating particles, such as higb- .~
energy protons.
~i
Moreover. many modern accelerators are tandem machines,

where a 0.1 % uncertainty in the tenninal voltage translates into a
0.2% uncertainty in the beam energy. Analysts should not forget
that there is also the injection voltage of the ion source into the
accelerator, which is a voltage additional to the terminal voltage,
so that, when the linearity of the accelerator is detennined, there
should always be a zero offset (of the right sign!) equal to the
injection voltage. Of course, this -20 kV injection voltage can
(and should) be measured directly with readily available
If there is an accurately calibrated radioactive source of small .:~

area (for example, 141 Am), then the solid angle can be :~
determined directly by placing the source at the position of the
target. However, it is usually difficuJt to reduce the uncertainty '.1
in positioning the a -source to the desired 1-2% level.
{1
i~
instruments.
15.3.2.1 NRA standards

"1
In discussing reference standards. we are also talking about ..;,
those nuclear reactions for which the corresponding crosS ~.
sections are wen known at the level of a few percent. To date, /:
probably the most widely used standard for NRA applications "1
has been that for 160. specifically anodically grown thin Tapj . ':'~~
targcts on a Ta substrate. The Paris group (Amsel and Samuel,
1967) pioneered their use through the 160(d,Pl)170 reaction at a ~i
bombarding energy of972 keV and a detector angle assumed to r:~
be 1500 but more likely to be 164. Here. there IS a broad (-85 1~.,1
keY) resonance in the reaction cross section. The protons are _~'
emitted with an energy of 1.6 MeV, and the crOSS section has ~\b
been measured with a preciSion of :1:3% by Davies and Norton .. ,;;;~
(1980) to have a value of 13.3 mblsr and by Lennard el aL:'<!-
The good news is that generating voltmeten; (GYMs), the

instruments that monitor the terminal voltage) are usually both
very stable and astonishingly precise. Modem tandem machines
have GYMs that are reproducible and stable over months at the
0.05% level. However, the GVM is a variable-capacitance
device that uses the induced ac voltage as a monitor for the
tenninaJ voltage. The voltage induced on the GYM is directly
proportional to its distance from the tenninal and is therefore
subject to a temperature coefficient. which, for example. is
0.031'C on the Swrey HVEE 2 MV Tandetron. For accurate
work using energy feedback based on a GYM, the accelerator
hall temperature must be closely controlled.
358
achieved. Several Si wafers were implanted at the UK Atomic

2
Energy Agency CUKAEA, Harwell site) with -5 x lOIS Bilcm a~
an energy of 40 keV, which locates the Bi distribution at a depth
of -20 nm. These samples, which became known as the Harwell
Series I standards. were then partitioned into l-cm~ pieces and
distributed to -50 RES laboratories around the world. The
results, reported by Baglin at the 1975 Karlsruhe IDA
conference, were disappointing. Even among experienced
laboratories, discrepancies of 20% were commOD. and some
results varied by 50% or more. The individual causes were never
fully resolved, but from the shape of the 1.9 MeV H' RES
spectra (includmg the strong 1.6 MeV resonance in Si), Baglin
concluded that several laboratories did not even have a correctly
calibrated accelerator energy scale.
(1989) to bave a value of 13.6 mb/sr. Seab et al. (1988) reported

a successful round-robin intercomparison of absolute
measurements in different laboratories using this reaction.
Gurbich & Molodtsov (2004) reported the excitation function for
this reaction at 150".
For those not baving access to deuterium beams, substitution
of ISO in the electrolyte allows Tal1HOS targets to be fabricated.
The IgO(p,a)l~ reaction at E = 750 kcV could then be used,
although the cross-section value is not known as well as for the
"O(d,p,)"O reaction (Christensen et al., 1990). The
160eHe,a)ISO and 160eHe,p)18F reactions are also useful, and
some cross-section values have been reported (Lennard et al.,
1989; Abel et al., 1990). It should be noted that oxygen from a
TazO$ oxide layer will begin dissolving back into the Ta
substrate at a temperature of-700 K (Smyth, 1966). Indifference
to beam heating of standards is therefore not encouraged.
Subsequently, a collaboration between ion beam groups at

Chalk River (Ontario, Caoada), I'Univers;t" de Paris, Harwell,
aod Gee1 (Belgium) developed several independent methods of
calibrating the Bi content of the Harwell standards, eventually
obtaining very satisfactory (2% or better) agreement among the
different laboratories (Cohen et aJ., 1983).10 the course oflhese
studies, several of the pitfalls noted in this chapter became
recognized for the flrst time. For example, the energy
dependence of the Bi yield was observed [0 deviate slightly from
the expected lIE2 Rutherford law, and this observation led to the
recognition of the electron screening correction. Many pitfalls
associated with poor Faraday cup techniques were identified and
corrected. Also, the problem of charge exchange along the
beamline was found to be responsible for significant fluctuations
(4%) in some of the Harwell Series II Bi implantations (Davies
eI al., 1986).
The cross section for the 3He(d,p)4He reaction (Q value of

18353 keY) is very well known (2%) for a center-of-mass
energy of 250 keV (Glib = 58 mb/Sf) (Davies and Norton, 1980;
Moller and Besenbacher, 1980). Unfortunately, lHe targets
fabricated by ion implantation are sometimes not stable under
prolonged ion beam bombardment (Alexander et ai., 1984;
Geissel el al., 1984).
15.3.2.2 Hydrogen isotope standards

Stable hydride or polymer targets suitable for hydrogen (or
deuterium) thin-film standards are still few. There are problems
in the preparation of reference targets by ion implantation
(current integration, dose un.iformity, depth profiles, and
stability: see Amsel and Davies, 1983).
Recently, Banks et a/.. (2004) ~howed that standard samples _
could be made from hydrides of Er and that these samples were ~.stable to .~e be~, but the sa;nples were both rather rough ~d
rathe.r thlck~ needm~ :>5 MeV He to look all ~ou~ them ",:th
elastlc .recOII detectIOn (ERD; also ca~led elastic reCOil detectIon
analysts (ERDA) and fonvard reCOil spectrometry (FRES or
FRS)]. However, an absolute H (and D) content was determined
independently of IBA to better than 2%, and ERD and NRA
have an uncertainty of 3.3%. Boudreault et al. (2004b) reported
an interlaboratory reproducibility for H implanted in Si at 2.2%,
althougb the expanded uncertainty (k ~ I) was estimated at 6%.
15.3.2.3 BAM hydrogen standard

Reinholz et al. (2008) established a hydrogenated amorpbous
silicon (a-Si:I{) certified reference material (CRM) grown by
chemical vapor deposition (CVD) with about 10 at. % H in a 1-2~m-thick a-Si layer (on a Si substrate) with an expanded
uncertainty (k ~ 2) in the H content of about I % (BAM, 2009).
The resulting comparison, with the same Bi-implanted wafer

being used in all threelabora_tori,es, showed a 2% agreement ..
and th-erefore demonstrated that RBS can achieve
expected absolute accuracy without the use of any calibrated standard. In
practice, because the detector solid angle and the scattering angle
are difficult quantities to measure accurately, it is simpler to use
a previowly calibrated standard instead.
the
Harwell Series I wafers have a Bi content of 4.87 (0.08) x

10" Bilcm' (Watjen and Bax, 1994). Series II wafers have the
same nominal Bi content as Series I, but fiuchlations as large as
::4% have been found (as noted above). Hence. whenever greater
accuracy is required, one should obtain a wafer that has been
cross-calibrated against a Series I standard. These samples can
still be found; however, they have now been superceded by the
a-Si standard. which is readily made. and also by the much more
accwate Sb-implaoted standard. These new standards are
described in the next two subsections.
15.3.2.5 BAMlIRMM Sb standard
Sb-implanted samples are available from the Institute for
Reference Materials and Measurements (IRMM), Gee!,
Belgium, and the Bundesanstalt fUr MaterialfoISchung und-
15.3.2.4 Bi-implanted (Harwell) standard

In 1975. an RBS round-robin experiment was conducted to
test whether the often-claimed 2% accuracy was actually being
359
-.~.-
Chapter 15
priifung (BAM), Bedin, Gennany, and are certified at 0,6%

(registered as JRMM-302fBAM-LOOI: Pritzkow ef al., 2001;
Ecker e/ al., 2002). These samples supercede the Harwell Bi
implant samples, but are used in exactly the same way.
The samples are an implant of Sb into Si with a range of 160

and a certified Sb fluence of 48.1 (O.3) x 10"/cm' (lcr
uncertainty). The Si wafer has a 90-nm oxide (of uncertified
thickness). This standard has been used once to date in publisbed
work for a traceable RBS analysis of some As implants
(Boudreault et al., 2004a), and it has also been u.sed to validate
Si stopping powers (see next subsection). In both, of thes,?
applications, the Sb standard was transferred through the a-SI
substrate yield, and to do this more accurately, the Sb standard
. was amorpbized to a depth of 630 om with a cold 500 keY Si
self-implant of 5 x lOu/em2 fluence.
DID
\5.3.2.6 a-Si standard

Useful work has been done on new standard samples
involving measurement of the surface yield of an implanted
(amorphized) silicon sample (Lennard et aI., 1999; Bianconi el
aI" 2000). This is nearly equivalent to an absolute measurement
of the energy loss of 4He in Si if a series of measurements are
made at different energies, Konac et aI, (1998) made direct
measurements of the energy loss of4He in Si (also known as the
inelastic electronic energy-loss cross sections or the stopping
power) with an uncertainty of about 2% (these important
measurements are often referred to in the literature as KKKNS),
and Niemann et aJ. (1996) also made measurements of energy
loss in Si with an uncertainty approaching I %. The stoppingpower function was systematically extracted using Bayesian
methods, along with uncertainty estimates, from the
measurements of Bianconi e/ al. by Barradas et al. (2002).
In their study, Bianconi et al. (2000) made absolute

measurements of the product of the charge and solid angle, at
.. nominally 1% accuracy, and obtained agreement within the
stated error. The traceability of Lennard ef al.'s (1999) work is
not as easy 10 establish, but they obtained the same values and
consistency with KKKNS, The conclusion of all of this work is
that the measured Si stopping cross sections have an uncertainty
(Icr) of 2%.
We sbould bigblgbt the fact that the Si stopping powers
stored in the SRlM-2003 database (Ziegler, 2004) were used to
determine the Sb content of an IRM:M-certified Sb sample (see
previous subsection) for 1.5 MeV He RBS, and it was found that
the certified value was obtained within 0.3% (with a 10uncertainty of 0.2%) (Barradas ef al., 2007, 2008). This
demonstrates that, for this He beam energy, the SRlM-2003
stopping powers for Si are correct to within the 10" uncertainty of
the certified sample: 0.6%.
The point to empbasize here is that Bi or Sb implant samples
are specific artifacts, but every laboratory can make its own
amorphized Si samples on demand. Secondary standards must be

used systematically with the Bi or Sb certified standards, along
with the associated error and complexity; not so for the
arnorphized Si. The difficulty with certifying the Bi implants has
been in establishing the real variation across the implant batch,
but the unifonnity and purity of modem production silicon
ingots has been established at extraordinary sensitivity and
accuracy: modem standard RBS samples can now take
advantage of this fact
15.3.3 EJectronic gain
It is harder than generally supposed to establish the
electronics energy calibration (the gain, in keY/channel) with an
accuracy better than 1%, and it is easy to make errors of 2% and
more. Whenever a spectral area is being evaluated [Eq. (l5.1)J,
only the product of the charge and solid angle, Q!l, is required
(this is the case when using the Bi or Sb standards), but if the aSi yield (or other stopping power) IS bemg used to calibrate the
solid angle, then Eq. (15.2) is invoked, which uses the product
Qnt., and the gain is also required to evaluate the product QO
needed for interpreting spectral areas. The Si stopping powers
have been compared directly with the Sb standard and are
therefore very accurately known (for 1.5 MeV He); other
stopping powers are not usually kno~ to t?uch better tha.n 4%.
Nevertheless it is a bad analytical mistake to omit the
determinatio~ of one of the local parameters of the analysis
because there might be a systematic error in ODe of the global
parameters. After all, the stopping powers of a material ":"
constant, even if they are not known very well, but the analYSIS
is not repeatable if the local parameters are not accurately
determined. 10 this section, we therefore dwell on the
detcrmination of the important gain parameter 0. at some length.
.:
I
~
:1
.,
,j
~'1
:;1
;
:1
.'4
l
..
i~
~
'-."'1I
To determine the gain, one must flnd a sample that bas

several (at least two!) elementa of both high and low Z at the J~
swface. We consider a AulNi/SiOzlSi sample (Fig. 15.4) (Jeynes ;:I;1?
et al., 1998), where the Au and Ni films are IS TFU thick
:~i
the oxide is 1500 TFU thick [I TFU (thin film unit) = 1D~(:t
2
atoms/em ; sec Section 15.7.3]. The Ni film is used (0 wet the ~l~
oxide for.~e Au d~osition, and thi~ sampl;, has eX~,e~ent long~ :;;~
tenn stabIlity. For thiS sample, the Sl and 0 surface Signals are ;:t~
actually buried under the metal layers, and an acemate ::j,
calibration must take this fact into account.
""11
t:.
fr:.
This discussion will be in tenos of the above-described ;~

sample, and the reader can substihlte his own favorite caLibmtion t~~~~.
sample. Consider the following examples of what is possible: ,In ;A~
the very careful work of Bianconi eJ 01. (20~), a seveD-po~t :t....
calibration was used, with a Type A estunate of 0.2 f\
uncertainty (see Section 15.5.1) (this estimate was published ID 1'1"
Barradas et al., 2007). Munni1c ef al. (1996) des<::tibed stoppmg- ~1
power measurements usmg RBS WIth an UDcertalllty aD. the ~aUl :;:t.
of 0.16%. Gurbicb and Jeynes (2007) detennined the gam WIth. ,"
Type A uncertainty of<O.l%.
'.:'~
.1)
ctiw-,
.y~\'
360
-----
.---------------------------------------------------------------------------------------------------------Pitfalls in Ion Beam Analysis
To detennine the gain, we have to identify channel numbers
15.3.3.3 Detector resolution

Fitting codes usually use some sort of
funCtiOil to obtain,
miDimum values. Because the positions of edges and peaks can
be determined with extraordinary accuracy from RBS spectra,
RBS is very sensitive to the effect of the detcctor resolution on
the widths .of these edges or peaks. Therefore, the analyst often
finds that noticeably different results are obtained when the
detector resolution is Dot correctly detennined (see Section
15.3.6 below).
corresponding to wellknown energies and then do a linear

correlation to obtain the gain (II) and offset (0) values: E" = lie
+ 0, that is, the energy Ec is represented by a particular channel
C. The surface signal of element E is given by ~ = keEo. where
Eo is the incident beam energy and kc is the kinematic factor for

element E. The kinematic factor is known analytically of course
and depends (fairly weakly at large angles) on the scattering
angle a. Thus, for an accurate gain calibration, both the beam

energy and the scattering angle must be known rather precisely.
This discussion ignores the pulse height defect (see Section

15.3.3.4).
15.3.3.4 Pulse height defect

Allhaugh the accelcmtar energy (;an be established readily at
about 0.1%, the absolute electronics calibration depends on an
accurate knowledge of the pulse height defect of the detector.
Semiconductor radiation detectors are known to respond in a
nonlinear fashion to particle energy. Nonnally, the deviations
are lumped into the so-called pulse height defect (PHD), which
includes the effects of the detector entrance window and dead
layer, as well as the nuclear (nonionizing) energy loss of the
projectile arising from elastic collisions with the atomic lattice of
the detector material.
15.3.3.1 Fractional channel numbers

To identify the channel numbers, the user chooses the signal
peak for very thin layers and the signal half-height for thick
layers. For an accurate gain determination, it is essential to find
the channels as real numbers, Dot as integers. The reader might
demur here, saying that the detector resolution is (typically) 16
keY and the channel width is only (say) 3 keY/channel, so we
are demanding an unphysical precision. However, the reader
should remember that, first, in accurate work, one should always
use one or more significant figures than is justified by the
intrinsic experimental precision, to avoid rounding errors, and,
second, some parameters of the data can be established with a
preciSion much greater than the nominal system resolution. This
latter point was made very strongly by Jeynes and Kimber
(1985), who showed that the peak of a distribution in an RBS
spectrum can be determined with a precision of about 300 eV
where the detector resolution is 15 keV.
The preceding discussion ignored this effect, treating the

energies of the sunace signals as being given only by the
kinematics. However, in reality, the detector sees only that
fraction of the energy of the particle that is deposited into
electron-hole pairs (that is, after passing through the entrance
window and dead layer) and for whic;b the electrons do actually
reach the detector anode and are not lost in recombination sites
(defects). Old detectors could have significant radiation damage,
providing a significant number of recombination sites. These
detectors will no longer behave linearly and cannot be used for
It is therefore clear that the best gain determination will not
be made unless numerical_methods -are used to determine real.::.:. accurate work, although all rnA codes allow nonlinear gains to .
be used'-- ,_.--~~~ ..
_ _" _
(fractional) channel nwnbers precisely. This is, incidentally, the
only reason for using ADCs at high resolution for collecting
However, all detectors, even high-quality (new) ones, will
relatively low-resolution RBS spectra: In Fig. 15.4, we show
detect a particle energy significantly less than the particle has
512--channel spectra, but we could easily use 4000 spectra, in
when leaving the sample, as a result of the PHD. Usually, this
which case integer channel numbers would be at sufficient
effect can be ignored because the absolute value of the
resolution for determining the gain as precisely as is possible.
correction is small. For instance, Bianconi et a1. 's (2000) highHowever, standard theory (the Nyquist theorem) says that
precision gain determination referred to in Section 15.3.3 did not
sampling at greater than 3 times the resolution gains no extra
correct for the PHD, but this introduced an error of only 0.2%
infonnation, so strictly speaking, 4000 spectra are a waste of
(Barradas et a/., 2007, 2008). Moreover, the nonlinearity in the
space!
detector response that must exist because the PHD varies with
particle energy is so small that it cannot be detected in standard
15.3.3.2 Correlated gain and offset
cases (Jeynes e/ aI., 1998), being mimicked by a (slightly
The rnA codes all have facilities for fitting spectra to extract
different) linear gain. However, in cases where different spe1:tra
the electronic gain. Of course, it is necessary to determine both
collected at significantly different beam energies must be
the gain and the offse~ even though the value of the offset has
compared, it is essential to apply the PHD correction for the best
no importance and the gain and offset are strongly correlated.
accuracy.
Therefore, an error .in the offset gives a corresponding error in
The PHD was comprehensively described by Lennard and
the gain. This is why it is so easy to have large errors in the gain.
Massoumi (1990), and Jeynes el al. (1998) presented a full
The fitting used by the codes might not handle this correlation
(manual) analysis in a particular case where the best demonstracorrectly, giving the user a false estimate of the
accuracy.
true
361
~ ..
Chapter 15
ble accuracy for the electronic gain, .6., was only 0.5%. Lennard
e/ al. (1999) did not make a comparable accuracy estimate
explicitly, but their work is also consistent with an accuracy in A
of about 0.5%. The extremely accurate analysis of Munnik et al.
(1996) includes a full PHD correction, and Gurbicb and Jeynes'
(2007) benchmark measurement of the ""Mg(p,p)m'Mg elastic
scattering cross sections depends on the analysis of a PHDcorrected set of spectra with a single gain across the whole
energy range (see Fig. 15.3 for one of these spectra).
might not accurately represent the data. This is actually another

good way to determine true effective scattering angles. For
example, for the 1735 keY resonance in the 12C(p,p)"C EBS
reaction, the maximum cross section relative to the Rutherford
value for 150 is 39.06, but for 150.2, it is 39.34, a change of
nearly 1%. Conversely, if EBS resonances are being used to
analyze light elements, the actual angles must be verified for
accurate work..
A further effect that has been studied in some detail (Lennard

et al., 1986; Bauer and Bartels, 1990; Comedi and Davies, 1992)
in Si for light ions e.2H, 3,4He. 7Li) ari:o:es 'as a consequence of
the differing ionization densities produced by the incident
particles. 10 principle, an rnA spectrum can easily be obtained
using higher-energy beams for which various nuclear reactions
occur and various types of particle all enter the detector; for such
a spectrum, one could perbaps use a single absolute electronics
calibration if the PHD correction could be done correctly.
However, nO PHD-corrected data have yet been published for
these types ofNRA spectra that also correct for the particle type,
although Pascual-Izarra and Barradas (2008) describe a full PHD
correction in detail.
15.3.5 Detector resolution

We already noted (Section 15.3.3.3) that fitting prognuns are
sensitive to the value used for detector resolution because the
shapes of edges and peaks are detennined very precisely in RBS
spectra. Figure 15.4 shows a AulNi/SiO,/Si calibration sample
spectruni with fits assuming various detector resolutions. The
signals from the thin metal surface layers have a shape entirely
determined by the detector resolution. To determine the detector
resolution, the height of the signals must be matcbed, given that
the signal arca is also matched. The fitted gain is given in all
cases in Fig. 15.4, gjven that the metal thickness is fixed. A 4% i
change in resolution gives a maximum 0.25% change in gain. ~
The fact that the gain can be affected at all by the resolution,
albeit at second order, is surprising at first sight
}
15.3.4 Scattering angle

The scattering angle, e, is: the angle the detector makes with
the incident beam. This should be measured as accurately as
possible because, even at = 1500 and normal beam incidence,
an uncertainty of only 02 gives an uncertainty of 0.3% in
fluence measurements (see Section 15.5). For glancing-beam .
exit geometries, any uncertainty in the scattering angle greatly
amplifies the uncertainty of the results.
Also shown is the fitted metal thickness: In this case, a 4% \

change in the resolution gives a 1.5% change in the fitted
thickness. This is almost a first-order effect even though one ':~
might have thought that there should be no change. The size of .
this sort of effect depends strongly on how the fitting algorithm
(in this case, a grid-search X' approach) works, and the reader ;~
should be warned that fitting algorithms do not always behave as ':1
one might intuitively expect.
In principle, the detector resolution function is neither ~

?~ussian nor even symmetrica~, and in ~e mo~t ac~urate work, ,~~
Where a goniometer is used (presumably including most

applications of glancing exit geometries), it is easy to measure
scattering angle: One only has to shine a laser down the beam
path and reflect it onto the detector from a mirror at the sample
position. The scattering angle can then be read from the
goniometer. Care must be taken in precision work to specify the
plane of the sample because this affects the scattering angle. For
small detectors, the blocking effect in single crystals can also be
used to detect the scattering angle. In this method, one aligns a
major axis of the crystal with the detector (not the beam),
lOOking for a dip in the yield. The beam energy bas to be such
that the detector solid angle is not too much larger than the
channeling critical angle.
necessary to account for this aspect ill detail. ThIS should be :~'i
done as a matter of course for PIXE, where the detectors have ,,~
relatively high resolutions (typically <200 eV) and the shape of
the instrumental function is well-known to be non-Gaussian. The'.'~
GUPIX program (see Blaauw e/ al., 2002), for instance, assumes X
that the instrumeotal function is specified in detail by the user.. :~
For PIXE, the instrumental function must be measured as a :~
function of X-ray energy over the whole energy range using a S1~:.
series of pure targets. If this is not done, the ratios of the vari~us '~.;.
K and L lines cannot be detennined with precision. Particle :."
scattering detectors have much lower energy resolution
(typically -15 keY), and the instrumental functions of these ,t;,
detectors are not usually measured in detail, although remarkable ..~
results can be obtained when it is. Fischer e/ of. (1997), for. ~,
example, demonstrated that they could reliably resolve the. i~
isotopes of Co using a full Bayesian deconvolution code! ,"!,~
However, none of the routine IBA codes currently have facilities~ ~~~
for a particle scattering detector resolution that is a function of t
energy, and the highest-resolution work yet reported with these :)'_
detectors (determining interface roughness of <1 nm) used 8.:::'-;
simple Gaussian (symmetrical) function (Barradas, 2002).
...""
It IS
.:ii
In some cases, the angle subtended by the detector has a

significant effect on the result. For glancing-beam-exit work, the
detector must be collimated in one direction to restrict the
possible variation in scattered-particle exit path length to the
detector. For these geometries, it is worth noting that curved
collimation slits are optimal (Bric;;e and Doyle, 1990). For some
elastic (non-Rutherford) backscattering (EBS) reactions, the
cross section varies strongly (and nonlinearly) with scattering
angle. and for larger detectors, the average scattering angle
,::..1:.
..~l~.
~
362
J~t.'
160
Vl
C
::::J.
O
()
a
Channel
160
160
350
160
'"
E
:::l
d
290
Channel
330 290
. Channel
330 290
Channel
330
FIG. 15.4. Detector resolution and its effect on the calibration of gain. (a) Spectrum for 1.557 MeV 4 He+ RBS with scanering at
149.2". Sample is Au/Ni/SiO,/Si (l2.8Il4.8/l607 TFU). (b) Delector resolution of 25 keY, fitted gain for fixed metal thickoess =
4.457 keV/channel (fitted metal !bickoess = 28.55 TFU); (c) 24 keY, 4.472 keY/channel (28.07 TFU); (d) 26 keY, 4.460 keVlchannel
(28.96 TFU).
these equations. Here, we consider how well-known the

Rutherford (and Don-Rutherford) elastic scattering cross sections
However, in general, the lesson is that, for accurate work, all
of the parameters should be correct.
and the inelastic electronic stopping cross sections are. We also

discuss the limits within which the basic single-scattering
15.4 ALGORITHMIC ISSUES

In this section, we consider a number of issues that affect the
interpretation of spectra and provide occasion for various subtle
pitfalls for the unwary. To perform a traceable analysis, one
must be able to accurately estimate the parameters in Eqs, (15.1)
and (15.2) (considered in turn in the preceding section). but one
asslllDption is valid and how well the user can calculate the
variation of the energy resolution with depth.
15.4.1 The Rutherford cross section
How accurate is the Rutherford scattering law? This is often
the key question in achieving quantitative RES analysis, because
must also be able to accurately do the integrations implicit in
363
N-k
Chapter 15
above should not be relied on for accurate work; instead

measured cross sections should always be used.. The IAEA has
sponsored a database of EBS cross sections (!BANDL, 2009).
The analyst should be aware that these cross sections are a
strong function of scattering angle, so that, if measurements are
not available for the angle used, then accurate EBS work cannot
be done. However, for some nuclei, the EBS croSs sections were
evaluated by Gurbich, who determined the quantum mechanical
parameters of the nuclear scattering by a fit to all available data
for these nuclei using the program SigmaCalc (SigmaCaIc,
2009). These cross sections are the most reliable and are valid
for (almost) all scattering angles. Both the IBANDL database
and the SigmaCalc calculator are under active development, and
analysts, should check whether cross sections they need are
included yet. Where SigmaCalc has cross sections, they should
always be used. Note that the IBA codes always give the user
access to these cross sections even if they have default values,
and the user should know which cross sections are being used by
the codes.
its main advantage is the assumed existence of a universal and

predictable scattering cross section, O"R (cm 2/sr), whose
dependence on beam energy E (MeV), scattering angle O. and
atomic numbers Zl and Z:z of the beam and target atoms,
respectively, can be accurately described by Rutherford's
formula obtained from the Coulomb potential (see Chapter 4).
.. 0
o.
The validity of this point-charge Rutherford scattering law

requires the distance of closest approach (collision diameter or
impact parameter), b (pm) - I.44Z,Z,i[E (keV)], between the
projectile and target nuclei to fall within the Rutherford window
(although Dote that this window does not reaIly exist: see
Gurbich, 2004). This means that b must be considerably larger
than the nuclear radius of the target alom [r. (fin) = 1.4A,lIJ). At
the same time, b must also be much smaller than the atomic Kshell radius, TK (pm) = 50~, in order to minimize electron
screening effects. A qUick and simple estimate of the upper and
lower energy limits within which the scattering cross section
should be within 4% of the O"l value can be estimated by
requiring b to fall in the range rKI2 > b > 3rn
If the energy is below the lower limit for RBS, then the major
part of the collision takes place before the projectile has
penetrated all of the electron shells of the target atom; that is, the
unscrecned Coulomb field assumed in Rutherford scattering
plays a minor part in the collision. With megaelectronvolt
accelerators and surface-barrier detectors, this almost never
happens, even for the heaviest target nuclei. However,
backscattering studies in the medium-energy ion scattering .'-;
(MEIS) regime involve energies well below the limit for
Rutherford scattering. For quantitative work, suitable algorithms .~
are given by Mendenhall and Weller (1991); however, they
point out that a fairly accurate (2-3%) approximation to the .. 1.
screened Coulomb cross section can still be obtained by
applying the simple screening correction of Eq. (15.3).
"
.,,~
Note, however, that a small screening correction to 0"11: is

always necessary, even though most of the (;ollision is
completely unscreened. This arises because the initial part of
each scattering trajectory is fully screened (i.e., from ro to 5022
pm from the target nucleus). Consequently, the incident
projectile penetrates into the UDscreened region with somewhat
higher energy than would OCCllI if the target atom were a bare
nucleus. The resulting decrease in cross section below O"R is
usually mucb less than 4%. Furtbennore. experimental studies
by L'Ecnyer ef al. (1979), Ande",en ef al. (1980), and Hautala
and Luomajarvi (1980) showed that its magnitude is predicted
(with reasonable accuracy in most cases) by the relationship
:1
I
,
(15.3)
15.4.2 Electronic stopping cross sections

'J
Almost all IBA work relies on electronic stopping crosS ::.
sections. In particular, quantification of depth profiles depends >.I
on the stopping-power values used. The same is true of
elemental concentrations in most cases.
:~
~.~:~
A
where E is the projectile energy in keV. Strictly speaking, the

-"~center-of~mass energy should be inserted into Eq. (15.3), but the
screening correction is usually small enough that the use of
laboratory coordinates introduces negligible error. This
L'Ecuyer et al. correction is incorporated in the main IBA
codes. The uncertainty in the screening correction is cited as
0.5% by Wlltjen and Bax (1994) for the case of 1.5 MeV 'He on
Bi, and this uncertainty can be scaled for other cases from Eq.
(15.3). However, for small scattering angles or for beavy-ion
beams, L'Ecuyer el al.'s correction is Dot as accurate as that
proposed by Andersen et al., which takes into account the
dependence of screening on the scattering angle. Andersen et
al. 's correction is also incorporated in the main IBA codes and
should be used.
~~
Different sources of stopping powers are available, from Db ....~

initio theoretical calculations to experimental data covering a .1l
.~
limited energy range of a given ion in a given material. The most ~U.'
widely used source is SRIM (Ziegler, 2004), which is. "
semiempirical interpolation scheme that covers all ions at all ~~
energies in all materials. Users should be aware that SRlM has r:J
evolved over the years, and different versions provide different 'j
stopping values. The same is true of data analysis codes. where ~
different versions might incOIporate different SRIM versionI s. .~~
Other sources of tabulated stopping powers exist and are a so :~~
used, namely, ICRU (1993) and MSTAR (paul and Schinner,
2001,2002) (the lalter for heavy ions only).
If the energy exceeds the upper limit for RES, then we are in
the regime
of non-Rutherford elastic backscartering
spectrometry (EBS). As examples, proton scattering on C is
significantly non-Rutherford at 380 keY, and He scattering on C
is significantly non-Rutherford at 2.2 MeV. The rough estimates
A statistical analysis performed by Ziegler (2004), including

all of the over 25,000 data points in the SRIM-2003 database,
364
Rutherford backscattering event, is occurring. However, the

mean free path for unscreened scattering events [(N1taTFr 1] is,
only .... 100 run, where N (atoms/cmJ) is the atom density of the
target and aTF ::::: 10 pm is the appropriate Thomas-Fermi
screening length (and note that, at megaelectronvolt energies,
only unscreened scattering can contribute significantly to the
mean deflection angle). Hence, even at 'quite shallow depths, a
fraction of the beam undergoes significant secondary deflections
along the incoming/outgoing trajectories. If not taken into
account, these secondary deflections can lead to an incorrect
interpretation of data.
showed that the accuracy of SRIM-2003 stopping calculations

was 4.2% and 4.1% for H and He ions, respectively. The
accuracy was fOWld to be only 5.1% and 6.1% for Li and heavier
ions, respectively. However, these values are average estimates
and do not represent the wieertainty for any given systcm. In the
same analysis, it was shown that SRIM-2003 predicts around
75% of stopping values within 5% of ex.perimental values for H
and He ions, but only 58% for heavy ions. That is, for heavy
ions, 42% of the stopping-power values calculated with SRIM2003 differ from the ex.perimental values by more than 5%. This
decreases to 18% for errors larger than 10%. It is worth noting
iliat the experimental measurements for stopping power are
notoriously hard to make, so that, even if SRIM reproduces the
tal d a ta, the re 15
. no guaran tee th at th e data are correc t.
.
expenmen
Full quantitative treatment of sucb effects .requires a Monte

Carlo simulation of each particle trajectory. Such calculations
have undergone major improvements in terms of accuracy and
efficiency (Arstila et a1., 2001), but they are still difficult to use
in routine work. Some traditional codes now include both
multiple and double scattering.
The outcome is that, for H and He ions (and increasingly for

Li as well), the tabulated stopping values are nonnally accwate
(at the 5% level), but in some particular systems, large errors can
be made. Users interested in high accuracy should check the
literature for experimental values and compare them with the
calculations. In some systems, highly accurate data are ayailable.
On the other hand, for heavier ions, the ex.perimental data are
The term plural scattering describes trajectories in which the

ion suffers several large-angle scattering events before being
detected. A particular case is double scattering, corresponding to
two large-angle events. The combined kinematic factor can be
larger than the kinematic factor for single scattering, leading to
yields above the nominal surface signal energy. The most
important effect, however, is an increase in the yield at low
energies, along with a low-energy background.
sparse, and the calculations can be inaccurate for many systems.

Advances in the accuracy of heavy-ion stopping are
continuously being made, so the situation might impr6ve in the
coming years. In the meantime, reliance on tabulated, often
interpolated, stopping valUe:) might be the only practical
solution, but it can lead to large errors.
At least two well-known data analysis codes, SIM:N'RA
Paul and Schinner (2003, 2006) conducted statistical analyses

of the a.ccuracy of stopping powers from different sources,
. 1udmg SRIM-2003 (an d 0 Ider vemons
.
f SRIM) , ICRU , and
mc
0
(Eckstein and Mayer, 1999) and DataFurnace (Banadas, 2004),

include the calculation of double scattering in RES as a user
option. Calculation times are much longer than for single
scattering. One example is sbown in Fig. 15.5, where the
MSTAR, for light and heavy ions in solids, gases, and
SlMNRA simulation matches both the data and a Monte Carlo
compounds. The main conclusion is that the three sources are .-..
.
4
"~.simulation for 0.5 MeV He ions backscattered from about 115 equally accurate in most cases, as long as the most recent
versions are used.
Din of Au on Si at a scattering angle of 165. A comparable RBS
example for DataFwnace shows a TiAlNlMo multilayer
Great care should be exercised when analyzing compounds,
measured at grazing incidence, with a good fit including the
particularly insulators, because most codes use the Bragg rule to
effects of a large double scattering signal ext~nding to low
calculate the compound stopping from the elemental stopping
energies (Barradas and Joynes, 2008).
powers. This approach leads often to larger errors. SRIM-2003
Multiple scattering refers to the succession of many smallimplements the so-called cores-and-bonds (CAB) correction,
angle
scattering events, leading to an angular broadening in the
which is, however, limited to some tabulated compounds. The
beam path. Such secondaIy deflections obviously change the
alternative is to use the molecular stopping powers if available
scattering angle involved in the main RBS collision, thus
from experiments, which some lBA codes can use.
affecling the magnitudes of both the cross section and the
In all cases, we should emphasize that, when Eq. (15.2) is
kinematic energy-loss factor. Furthennore, it also affects the
used, the dominant uncertainty almost always comes from the
depth-to-energy conversion scale.
stopping-power database.
However, the net effect of multiple scattering is, to first
order, an extra contribution to the energy spread of the beam
15.4.3 Plural and multiple scattering
(Szilagyi eJ 01., 1995). This can be the largest contribution to
One major assumption in most RBS and IBA studies is that
straggling in some situations, particularly in grazing-angle
the incoming and outgoing trajectories are completely linear,
experiments, and it must be taken into account for a correct
that is. only one significant angular deflection, namely. the
interpretation of the data.
365
I
h .. _ .
Chapter 15
SOOkoV
-"'"
"0Il0r;::=====~-~~""l
- - SlMNRA. "'"
------ SIMNRA. $lodll
--11UM.SP
10000
contribution of multiple scattering to straggling must be

calculated and included in the analysis
Note that current vernions of a popular computer code for the
simulation of RBS profiles, RUMP (Doolittle, 1985, 1986), dQ
not include any treatment of plural or multiple scattering.
Depth resolution is defined as the minimum separation in
depth between two layers such that a maximum in the obsetved
signal is observed for each layer. It determines the capability to
separate signals arising from different layers.
sooo
In RBS, depth resolution is best near the surface and degrades
in deeper layers as a result of energy straggling. In conventional'

range-foil ERD, depending on the beam-sample--<ietector
Energy (koV)
,
,
,,
'
-'
geometry, the best depth resolution can be below the surface

because of the large geometrical energy broadening encountered
in ERD (Wang e/ al., 1990), There are several different .;
contributions to depth resolution: resolution of the. ~etection . ,,;
system, energy-loss straggling, multiple scattering, beam energy .'
and angular spread, finite size of the beam spot on the sample, )
and finite size of the detector. The last three terms together are
usually called geometrical straggling and influence the energy'
resolution because they lead to a spread of possible beam ~:..
trajectories. The code DEPTH (Szilagyi et al., 1995) uses state
of-tbe-art theory to calculate the different contributions. Given
that very few energy-spread measurements are available, this ~
program is currently the most reliable source for depth ~
resolution. SIMNRA implements the same straggling f~
calculations independently.
~;
Jf
'"
H.:
Different pitfaUs often arise when depth resolution is not well

understood If the energy spread at a given depth is not well ft
calculated (for instance, because of the neglect of multiple -'?~.
scattering or geometrical broadening), then the simulation for -~f~
sharp interfaces will be sharper than the observed data; the user .$
~uld. the~ wrongly conclude that. the sample has strong ::l~.
mterdiffuslon between layers or rough mterfaces.
. ,,~~1
Energy (keV)
FIG. 15.5: Double scattering effecls in 'He RBS of Au on Si.
Comparison of the experimental energy spectrum (dots) at a
polar emission angle of 15" for 0.5 MeV "He. The histogram
gives the TRlM.SP result; the dashed and solid lices represent
the single-collision model and the dual-<:ollision approximation,
respectively, both calculated with SIMNRA. The lower part of
the figure shows the background between the Si edge and the
lower Au edge in more detail. (Reproduced from Eckstein and
Mayer, 1999.)
Analysts might also be tempted to retrieve more infonnatI~ ~~,

than the data justify. This is nonnally done by imposing therr".
favorite model on the data, mthout testing other possibilitie5
that could also lead to a good fit within the depth resolution ata_ ~f.i
given depth. In this respect, overreliance on X2 goodness of fi~
for ex.ample, can lead to ultimately incorrect data interpretation, '.::-' ,
because, as seen in section 15.3.7, the depth resolution used lioE::
directly affects the results.
"'1,~
J
J\
't
A second effect of multiple scattering is that the shape of the

energy spread is no longer Gaussian. This can become important
for heavy ions in beavy targets, at low and intermediate

energies. Some limited treatment of this effect is included in
some lBA codes.
,-
;-:':~
15.5 ACCURATE IBA

:~j
The ultimate consideration of all analysis is the accura~ ~it
available. We use accuracy here in the critical sense, that 1S, :,'~~
where a measurement can be traced back to international :~~
standards of mass, length, and time with a specifi~ble j'f~
uncertainty. Because the Rutherford cross section is analytical, .,~._
the accuracy of RBS is potentially unlimited-except for tb e
Multiple scattering is very small for light ions at ncar-normal

incidence (and detection). It can be very large for heavy ions or
for grazing-angle incidence or detection even for He beams. In
that case, whenever depth profiles, roughness, or other sample
features that affect the sharpness of signals are studied, the
.&
~t
"\j;
.:;~
366
major problem in alllBA, the limited knowledge of the energy

loss of ions in matter, as discussed in Section 15.4.2. However,
there are certain sorts of analysis where the energy loss enters
only in second order: one of these cases was treated in detail by
Jeynes et 01. (1997) with the conclusion that, even in this ideal
case, several small effects have to be considered (at the 0.25%
level) that will cumulatively make an accuracy better than 1%
hard to achieve. (We quote all uncertainties here at the 10confidcnce level.)
According to GUM, there are two types of uncertainty

estimation: Type A and Type B. Uncertainties are of Type A
when they can be calculated as a standard error from a set of
measurements. They are of Type B when the statistical data
needed for Type A calculations are not available and the user
has to make a more infonnal estimate of the probable
measurem~t error. Thus, for example, a measurement of length
might be made with a ruler. The experimenter could estimate the
uncertainty of a single measurement by assuming a maximum
reading error of, say, one-half of a graduation. This would be a
To our knowledge, the only doubt about the potential
Type B estimate. On the other hand, to obtain a Type A estimate
the experimenter could make a series of measurements and then
accuracy of RBS is the interpretation of the low-energy tails in
backscattering spectra. Tails are certainly caused by multipledetermine the average and standard deviation. It is worth noting
and plural-scattering effects and have been calculated
that Type A estimates are not always better than Type B
estimates. In the case of a ruler measurement, for example, there
successfully with Monte Carlo techniques by Bauer et al. (1992,
1993) and by Eckstein and Mayer (I 999) for low-energy beams,
might well be a systematic reading bias from an observer that
for which the effects are large. Such tails could also be caused
could go unnoticed. This is not such a trivial point as it seems.
both by slit scattering and any low-energy component that might
Polanyi (1958) rehearses the famous case
be present in the beam. However, it has been claimed by
. of the Astronomer Royal, [Nevil] Maske1eyne, who
dismissed his assistant [David] Kinnebrook [in 1796]
Gurbich (1995), on the basis of time-of-flight experiments with
for persistenlly recording the passage of stars more than
a 2 Me V pulsed proton beam and very thin (10 ke V) selfhalf a second later than he, his superior. Maskeleyne
supporting gold foils, that the low-energy background cannot be
explained by plural, multiple, and slit scattering alone, which
did not realize that an equally watchful observer may
would mean that some significant physical phen(Jmenon is still
register systematically different times by the method
employed by him; it was only Bessel's realization of
left unaccounted. In any case, the single-scattering
approximation certainly fails sometimes: Barradas et al. (2007,
this possibility which 20 years later [after considerable
2008) pointed out that the calculation of some heavy-ion ERD
work by Bessel, who was the ftrst, in 1838, to observe
stellar parallax] resolved the discrepancy and belatedly
spectra by the IBA available codes is not consistent with a
Monte Carlo calculation; however, it is not yet known absolutely
justified Kinnebrook.
We note further that GUM speaks of "uncertainty"
how reliably these calculation results agree with experiment.
Gurbich also does not provide sufficient detail of his Monte
(indicating that our knowledge is limited) rather than "error"
Carlo calculation for us to evaluate his result.
(which implies that a mistake has heen made). It also uses the
In this section; we consider' a systematic approach -to the~~ .idea of "cov.erage", so that, instead of saying "the 10" error was
2%",',,: GUM ..:uses~~the more generalized .:-terminolcigy . "the
estimation of uncertainty, including a discussion of the explicit expanded uncertainty was 2% with a coverage factor k = 1". Of
uncertainty budget that must be included in any standards work
course, in all cases, both Type A and Type B uncertainties must
that claims to be traceable to international standards. An
be estimated and combined as necessary.
important benefit of IBA, and the focus of this chapter, is the
availability of absolute accuracy, but no analysis has any value
15,5.1.2 Uncertainty budget
unless the analyst is able to reliably specify its uncertainty.
The uncertainty budget is specified by GUM as a formal
approach to the systematic evaluation of the uncertainty of a
We also include an extended discussion of analytical
measurement, which is essential when a critical result is
approaches to the recognition and avoidance of ambiguity in
presented whose traceability must be explicit. This concept was
IBA data. It is of great importance for the analyst to recognize
described with special reference to IBA by Wa.tjen and
the difference between necessary conclusions (required by the
coworkers (SjOland et 01., 2000). We emphasize again that rnA
data) and valid ODes (permitted by the data). In many materials
is a quantitative analytical technique that is capable of great
problems to which IBA is applied, the user will want definitive
precision and that bas well-known uncertainties; we do our users
answers from the analyst. The analyst therefore needs to gain
a disservice by not properly estimating these uncertainties.
skill in seeing objectively what information is really contained in
The uncertainty budget for the analysis of Fig. 15.1 is
the data and how to apply prior information from the user to
presented in Table 15.5 as an example. All of the sources of
eliminate various valid solutions of the data.
uncertainty are listed, making the origin of the claimed
combined standard uncertainty very dear and, hence, easy to
15.5.1 Uncertainty estimation
evaluate critically. It is clear that the stopping power uncertainty
IS.5,LI Type A and Type B
We start by referring the reader to the Guide to the
dominates the total combined uncertainty.
Expression of Uncertainty in Measurement (GUM, 1995).
367
-,
Chapter 15
Table 15.5. Uncertainty Budget. Coverage factor k = I. (Reproduced from Jeynes etal, 2006)
TypeAorB
Cornell
IBM detector
Comment
2.60%
0.80%
2"10
Uncertainty of pileup correction
2%
Counting statistics, As signal
0.28%
0.47%
Counting statistics, a-Si signal
0.08%
0.13%
ScaHering angle ~. -
0.28%
0.07%
i,
,I
From shape-fitting accuracy
0.2 and -1I[sin' (e/2))
Electronic gain
0.5%
0.5%
Pileup correction
0.05%
0.02%
Relative uncertainty
0.64%
0.70%
Relative uncertainty of average of two

detectors
0.48%
averagel.fi
See Section 15.2.4
:.--
Beam energy
0.20%
Same for both detectors
Rutherford cross section
0.16%
Screening correction
Combined standard uncertainty
rnA code uncertainty
0.54%
Relative accuracy
0.2%
From software intercomparison

(Barradas., al., 2007,2008)
0.7% for HI RBS; 0.4% for He
ERD
Si stopping power
0.6%
From software intercomparison

(Barradas ., al., 2007, 2008)
because SRIM-2003 stopping

powers were used
Total combined standard uncertainty
0.83%
It is this sort of treatment that is able to establish the

traceability cf a measurement and, thus, justify the weight put
upon it. We note th.at IBA is very simple, and specifying its
traceability to international standards of weights and measures is
relatively easy. The present treatment is a significant
simplification of the GuM recommendations, which can he
applied successfully to much more complicated cases.
Absolute accuracy
different conditions to constrain the solutions found (where a

"solution" is an elemental depth profile that has a spectrum that
fits the data), and the second is to rule out, a priori, certain types
of solution, which we call "restricting the state space", where the
state space is the multidimensional space containing all possible
depth profiles.
We say that an optimal solution is a depth profile from which
a spectrum can be calculated that fits the data well and that a
spectrum is ambiguous where more than one optimal solutions
15.5.2 Spectral ambiguity

In this section, we follow the treatment of Jeynes et al.
exist.
(2003), which also includes further examples. We show that IBA
For the AF, implant of Fig. 15.1, for example, one element in
the state space is a pure As sample, another is a pure Si sample,
and there are a very large number of intermediate elements. The
state space is large, but it is not infinite because the energy
spectra are grossly ambiguous in general, and we re-analyze the

ambiguous spectrum previously discussed masterfully by Butler
in 1990 (see Fig. 15.6). There are two approaches to overcommg
this ambiguity: the first is to collect multiple spectra under
368
detector
Pileup correction
I,i
resolution limits the number of layers that must be considered
substrate is pure silicon. However, Barradas e/ al. (1999aj

denoted BlJM) demonstrated that the most probable solution of
this type of spectrum, assuming that some oxygen signal is'
present and that DO prior infonnation is available, is that there is
a significant quantity of the heavy element in the substrate. This
is because any particular spectrum can be reproduced by many
mixtures of the three elements, given that the sensitivity to the
light element is limited and there is always some uncertainty
about the coUected charge.
and the detection sensitivity limits the number of layer

compositions that must be considered.
15.5.2.1 Multiple spectra
Every analyst is familiar with the practice of tilting the
sample and taking another spectrum to determine which features
of the spectrum come from the surface: the surface signal
position does not vary with beam incident angle, whereas signals
from below the surface will appear 10 move as the geometry
changes. An equivalent way of doing this is to use two detectors
at different scattering angles. This is not a new idea: Williams
and M611er were using two (or more) detectors in 1978
(although for a rather different purpose; Williams and Moller,
1978), and Edge (1983) repnrted calculations emphasizing the
value of spectra from two detectors but using an iterative
method of calculation that is not easy to extend to three or more
spectra. Butler (1990) emphasized the value of multiple
detectDrs, and Alkemade el al. (1990) demonstrated tha~ for a
sample with n elements, one needs, in principle, to collect n - 1
different spectra to eliminate ambiguity.
It is important to be objective about what we know about the

sample a priori. If we assume nothing about the sample, then we
have to report a range of possible solutions, consistent with the
data. Interestingly, BlJM also demonstrated that, provided that
the state space is suitably restricted, RBS data are remarkably
unambiguous with respect to collected charge (tolal number of
counts). It is a common pitfall for analysts to collect large
amounts of charge to obtain "smooth" data, even though very
small charges (BJJM considered 0.1 ~C wilh a 2.5 mSr detector
solid angle) can give objectively quite wcll-determined solutions
even for difficult cases, with the right number of layers and
qualitatively the right stoichiometry in the layers. Of course,
with fewer counts in Ihe spectrum, the statistical uncertainties of
the fitted stoichiometry and thicknesses of the layers increase, as
expected. It is worth pain ling out that this discussion emphasizes
the value of backscattering spectra collected simultaneously with
microbearu PIXE spectra: Microbeams typicaUy use only 100
pA, and the total collected charge is often a small fraction of a
microCoulomb. Do not fall into the related pitfall of neglecting
the potentially large amount of information in a noisy spectrum!
Figure 15.1(d) shows an example of the independently

collected data from two detectors directly compared. In this
case, exactly the same quantity (the depth profile of As) is
measured by each detector, a.nd the profiles should overlap. That
they do is an indication that the independently determined
electronic gain calibrations are reliable. The fwhm values of the
As signals are different because the detector energy resolutions
are different. The signals at negative depths appear because the
channel data are being replotted directly on a depth scale,
without correction for broadening due to the detector resolution
IS.5.2.3 Butler's nample re-analyzed
and straggle.
We show Butler's. e~ample of an oxidized NiCrAl alloy in
--.:' -_. -- - . . --:-::::----Fig. 15.6. He pointed out-that,'for~this example, the false_~
15.5.2.2 Restricting the state space
-.-.~--.----- - ~ . solutions can be eliminated if prior chemical information is
It is very easy to demonstrate that RES spectra are
taken into account. Thus, he knows that the oxygen comes from
ambiguous. Butler showed an example that is ambiguous in the
the oxidizing process and, therefore, enters through the surface.
sense that different depth profiles exist where different partial
(Actually, most of his false solutions can he eliminated simply
spectra add up to the same total spectrum; we discuss this
by excluding 0 from the substrate.) Moreover, the 0 binds with
interesting case in the next subsection. However, we have not
the metals in well-known ways. Therefore, the ambiguity is
(yet) found any examples where rnA data are systematically
greatly reduced if the analyst can manipulate molecular (rather
ambiguous in the sense that the system is ":frustrated" in
than atomic) depth profiles.
Kirkpatrick e/ al. 's (1983) terminology, that is, where a number
Figure 15.6(a) shows the elemental depth profile of Butler's
of optimal solutions exist with large potential baniers between
example, with a spectrum calculated from it shown in Fig.
them. Other systems that are frustrated are easy to find: see the
15.6(b). Figure 15.6(c) shows the solution obtained dosest to the
related discussion of ellipsometry data by Barradas e/ al.
original profile, and we discuss this result further below. We
(1999b) as an example.
point out here that this result is essentially identical to the
original, except for some interface broadening (we have not
When interpreting such data as the As implants descn'bed in
deconvoluted the straggle).
Fig. 15.1, analysts are used to tacitly ruling out the possibility of
As deep in the sample because they know that, in this case, the
369
..
Chapter 15
(.)100
,Cb)
1
Orlglilal
80
..
'#
40
+Ir~\~v-~,~'~~-'~===:I~~
20
o
EI,,--;;1f\~.!l~~.~.="=:'j"
o
2500
ThIn FIlm
100
'5000
UnJ"
(e)
(d)
60r-~~--------------~
S0t--1~--------------~
.-
30
t"":.:.....t=-"""'''';;::-------1
: t=t;~:;~rP=--_l
__
c-
. .. l ..-...
"-,
40t-~~--------------~
o~~~~~~
-- -
C,"lJl1al
o.\oInoI
- - 8"""",, 2
~-,'
-~.1
5000
FIG. 15.6. Butler's (1990) ex~ple of an o,udiz.ed NiCrAI alloy re-analyzed (reproduced from Jeynes el 0/., 2003). Ca) The original
profile from which the spectrum was calculated; (b) spectrum (symbols) and fit (line); (c) atomic profile fitted to data assuming
molecules and complete oxidation from the surface, using two spectra at different detector angles and excluding alwnina from the
surface; (d) comparison with the original profile of the Cr profile calculated under various assumptions. Specifying only elements
barely constrains the profi1e, and even with two detectors, the profile is not recovered at intermediate depths. Using only one detector
with the assumption of molecules is also not sufficient. The molecules used .are NiO. Cr20}. AhO), and NiI9~CrHI~119.
It turns out that the Cr profile is the most sensitive to the prior
assumptions of the analysis, and Fig. 15.6(d) shows the Cr
profiles obtained under four different asswnptions. To retrieve
Butler's initial proftle unambiguously, one must specify not only
the molecules present, but also the following conditions: only
oxides are present near the surface, oxygen is excluded from the
substrate, A1 is excluded from the near-surface region. and two
independent spectra were recorded (at different scattering angles
in our example). Butler did not point out this last condition for
this example, although he noted that, in general multiple speclTa
are always a help. These particular data are very ambiguous:
many different solutions, all of them as good as that shown in
Fig. 15.6(b), can be obtained without aoy of the conditions
mentioned.
about the chemistry to investigate whether they are consistent

with the data. If they are Dot consistent, they can be ruled out.
Thus. RBS is an effective tool not just for obtaining a solution to
a spectrum, but also for testing a variety of assumptions about
the sample against the data.
:'i
J
.~
15.53 Model-free analysis and Occam's razor
OJ
In the presence of ambiguity, the analyst must be careful to -~

control his assumptions. The philosophical maxim of William of
Ockham from the 14th century known as Occam's razor should '~
be kept in mind by aU analysts. This maxim is given variously as ~
Pluralitas non est ponenda sine nec'cesitate (plurality should not :~
be posited without necessity) or Non sunl muitiplicanda entia
praeler necessitaiem (entities are not to be multiplied except of I
necessity). Consider how this principle applies to ion beam "oJ
analysis.
:~
'1
'l,
In this example, we have allowed the 0 to exist only when

bound to metals, and we have allowed only free Ni to exist. The
substrate is specified by a molecule representing the starting
alloy composition. It is easy to specify various assumptions
There are many alternative ways to fit any given set of~ .:;.i
If a good fit to the data bas been found, it might represent a valid .J
'~11
,1
370
.
j{
,it.
,
solution. Whether or not it is valid depends on whether correct
parameters have been used. Sometimes, this might be an
intricate question if one is interpreting the detail of a spectrum.
Even if the solution is valid, in the presence of ambiguity, it is
not necessarily true. Even if the solution is not false, it might not
be useful in the case where Occam's razor has not been wielded.
In Butler's example, the spectrum is consistent with a variety of
Cr profiles; one can discriminate among these profiles to
detennine the real one only by restricting the number of
parameters describing the system (using Occam's razor), for
example, imposing prior chemical assumptions on the system. In
this example, four molecules are used instead of four elements,
but the light element is correlated with the heavy elements so
that, in addition to the number of stoichiometric possibilities
being reduced, the insensitivity of the system to the light
elements is not allowed to seriously distort the results.
Moreover, analysts sometimes overlook assumptions that
they tacitly impose on data when interpreting the data according
to a model. For example, the As implant profile of Fig. IS.!
could be fitted with a Gaussian distribution, and many IBA
codes facilitate such a fitting. Now, there is nothing wrong with
this approach, as long as it is borne in mind that an explicit
assumption is being imposed on the data, which can lead to error
(that might be large) if the profile is not, in fact, Gaussian. A
model-free fit, imposing no prior assumption on the data, would
instead involve a number of layers of varying stoichiometry.
Occam's razor requires that the number of layers be restricted in
order to restrict the number of fining parameters. Then, there is
an explicit tradeoff between the goodness of fit and the number
of layers. Clearly, for any given spectrum, the number of fitting
other implicitly. All of these assumptions have the potential, if

incorrect, to lead to severely wrong results. The assumptions can
be divided very roughly into four categories: those related to the'
sample, the experimental parameters, the physics, and the
databases.
Assumptions about the sample include the identities of the
elements present; a given depth profile or elemental distribution;
and the existence of roughness, inclusions, and other
inhomogeneities. The experimental parameters are known with
an accw-acy that varies widely for different setups. The user
might be aware of this accmacy (for instance, a 1 keY beam
energy spread) or not (for instance, a 5 keY energy drift with
accelerator temperature during an experiment). Assumptions
about the physics involved include the interactions both between
the analyzing beam and the sample and between the detected
particle and the detection system. The user might include only
the basic physics given in Section 15.3.3 in situations where
some of the phenomena described in Section 15.3.4 playa vital
role. Using a given stopping or scattering cross section is an
implicit assumption that the data base used are correct for that
case. Moreover, the prior assumptions are often related to each
other. For instance, an assumption about the scattering angle
implies a consistent assumption about the scattering cross
section.
We list in Table 15.6 the consequences for data analysis of
some of the problems that commonly arise as a result of
incorrect assumptions, along with steps that can be taken to
prevent or remedy such problems. These preventive and
corrective actions can be divided into four main groups: First
and foremost, know ledge about the setup used and its
layers can be increased uselessly to follow the statistical

variation of the spectrum. This underlines the value both of
characteristics should be as complete and accurate as possible .

. This often involves,~ elaborate .. experiments . or even - the
Occam's razor, which highlights and deprecates this uselessness,_-o .~ installation of new hardware (for lnstance, to measure the beam
and of multiple spe<:tra, which have independent statistical
variation and thus naturally discourage the use of too many
charge) and is generally not feasible in the short run. Second,
alternative experiments, using different beams, energies, and
layers.
other parameters or even completely different techniques, might
be required. Additional beam time or other techniques, however,
15.5.4 Common pitfalls in data analysis
are not always available. Third, programs that include the be~t
Here, we are concerned with some of the problems and issues
physical models and databases available for the problem at hand
that are particularly relevant to data analysis. Many prior
should be used. This is often the only practical alternative for the
assumptions are made in all data analysis, some explicitly and
data analyst, who is thus often confronted with problematic data.
371
Chapter 15
Table 15.6. Pitfalls and remedies in data analysis.

Assumption about
Elements present
Possible
What went wrong
Impurities ignored
cons~guences
What can be done

Use a complementary
technique that can detect the
impurities; link the impurity to
a major element.
Impurity signal wrongly

assigned to some other element
(possibly one with a small
cross section, leading to large
effects on the overall solution)

Wrong fitted charge to
accommodate the unobserved
element, with consequences for
"Missing element" falsely

postulated or ignored
.... --,-,,--_._----
Use a complementary
technique that can detect the
missing element; collect
spectra under different
experimental conditions.
Collect spectra under different
experimental.conditions; test
all other elements

-.~~~-.
Depth profile
Completely wrong results
False depth profile chosen,

leading nevertheless to a good
fit
is often ambiguous)
Profile function (such as
Gaussian) imposed on some
element when real profile is
different
QBA
-.
._--- ----_.,- -
different models.
Sample roughness ignored

(often the case in interdiffusion
studies)
Wrong type of sample
roughness considered
-~---
Beam energy
Calibration of accelerator
inaccurate
Energy drifted during the

experiment
Scattering angle
Poor original measW"ement;

problem with moving detector
Angle of incidence
Poor alignment of sample

holder; problem with
goniometer
Energy resolution
Not determined recently
Solid angle
Not well determined; might
change with sample distance
Inaccurate results for that

element; if signal is
superimposed on other fastchanging signals, other
elements possibly affected as
well
Obtained depth profile too
broad or completely wrong in
severe cases
Derived roughness parameters
meaningless
_.
Systematic error in analysis
(can be partially compensated
for RBS by MCA calibration
and stopping); EBS cross
sections wrong; depth scales
wrong
Apparent energy resolution
worse than expected; apparent
gain changes; slight changes in
cross sections, leading to
inaccuraCY in the quantification
Completely wrong results;
near-normnl incidence and
grazing-angle experiments
inconsistent
Completely wrong results;
near-nonnal incidence and
grazing-angle experiments
inconsistent
Small error in results; can be
crucial if interdiffusion or
roughness are important
Nonnally assigned to collected
charge, leading to an extra
errOf in the results
372
Perform a model-free analysis;

test different models.
Measure roughness; perform

grazing-angle experiments.
Determine roughness type;
include it in analysis if
possible; test different models_
Calibrate the accelerator
energy!
Measure energy drift by

collecting calibration spectra at
the beginning and end of each
run; discard badly affected
data.
Detennine accurately (might be
difficult); treat as a fit
parameter.
Align sample holder (might be
difficult); treat as a fit
parameter.
Detennine energy resolution in

each run; treat as a fit
parameter.
Determine accurately (might be "
difficult); adjust for sample
distance.
Beamcbarge
Not well determined (or not

measured)
Stopping power
Database inaccurate
Bragg rule for molecular

stopping inaccurate
Scattering cross section
Required non-Rutherford (EBS

or NRA) cross section might not
be measured or determined yet
or might be inaccurate
Calculated cross section for an
EBS (or NRA) resonance at a
given depth for the average
beam energy at that depth
Straggling
Used Bohr straggling where

Bohr model is not valid
Did not include TschaHir effect
I~o:~d_~ben relev~t
Plural scattering
--
--- - -
Multiple scattering
--
--- -
--
Ignored when relevant
Not well calculated,

particularly at grazing angles
Channeling
Unnoticed accidental
channeling overlooked in
analysiS
Pulse pileup
Ignored when relevant
Poorly fitted data
Could not find correct model
Not very important if the

spectrum includes some signal
(normally substrate) that can
serve as internal calibration;
completely wrong results
otherwise
Inaccurate elemental
concentrations and layer
thicknesses; completely wrong
analysis in severe cases
Detenmne accurately (might be

difficult or impossible); rely on
internal normalization (might
be difficult or impossible);
treat as a fit parameter.
Use a better database ifavailable;

determine the relevant stopping
powers if possible; perform
experiment with a different beam;
treat as a fit parameter.
Use molecular stopping powers
if available; determine them if
concentrations and layer
thicknesses; completely wrong possible: performing an
experiment with a different beam
analysis in severe cases
is unlikely to help.
Determine the relevant cross
concentrations; completely
section; exclude the element
wrong analysis in severe cases
involved from the analysis;
perform experiment with a
different beam/energy.
Use codes that correctly
concentration and depth profile integrate cross section over the
of affected element; completely entire energy distribution of the
wrong analysis in severe cases
analyzing beam before
scattering.
Incorrect diffusion, mixing,
Always use the best models
and rougbness results
available.
Incorrect diffusion, mixing, and
Always use the best models
roughness results for depths
available.
where energy loss is 10-20%
larger than initial beam energy
(Heavy) Elements introduced
Use the best models available
at depths where they do not - -=. whenever plural.scattering is .relevant.
exist
Incorrect diffusion, mixing,
Use the best models available
whenever multiple scattering is
and roughness results
relevant.
Monte Carlo techniques might
Incorrect results
be the only available alternative;
perfonn experiment with a
different beam.
Incorrect results; but if only the
Adjust incidence angle; rotate
substrate signal is affected, there sample while measuring;
might be no consequences unless calculate channeling.
the substrate signal is used for
normalization
(Heavy) Elements introduced at
Accurate models exist-use
depths where they do not exist;
them!
signal heights and areas affected;
dramatically degraded accuracy
Reliable analysis requires weUCompletely wrong results
fitted data.
Acronyms: EBS. elastic backscanering spectrometry; MeA, multicharwc1 analy:zer; NRA. nuclear reaetioD analysis; RBS, Rutherford backscattering
spectrometry. Caution should be used where the suggestion is to treat a variable as a fit parameter, as this could lead to evenlarger erron!.
373
Chapler 15
correct. A particle scattering spectrum is an object with some

"holographic" properties: its parts are interdependenL Of course,
no spectrum is perfectly fitted, but the analyst should ensure
both that the fit is qualitatively correct for the entire spectrum
and that misfitting regions are understood properly. For
example, the low-energy signal is not usually well fitted because
IDultiple-, plural-, and slit-scattering effects are generally
neglected.
This leads to the fourth and most important group of

preventive actions, which is to avoid both overinterpretation and
underinterpretation of the data. Overinterpretation occurs
whenever the data analyst unjustifiably imposes a given model
on the data. This often leads to tuning the experimental
parameters beyond their known accuracy (such as increasing the
charge while scaling the stopping power for a given element in a
given layer). The user must consider different models that might
lead to an equivalent or better solution. Of course, the prior
assumptions that the analyst can impose on data to exclude
certain valid solutions. discussed in Sections 15.5.2 and 15.5.3,
are also strictly overinterpretations, but in this case, they are
explicitly justified.
Two main conclusions can be made: Reliable analysis

requires well-fitted data. However, a well-fitted spectrum
implies only the validity of the model, not its truth.

15.6 UNWANTED TARGET-BEAM INTERACTIONS
15.6.1 Beam-induced heating
The power deposited in a target under beam impact can be
expressed in watts through the product of the particle energy (in
megaelectronvolts) and the number of particles per second (in
microamperes), where the latter represents particle current, not
electrical current. If the cross-sectiona'l area of the beam is
known, then the power density (watts per unit area) is also
readily calculated. Basic heat-flow calculations can then be
made.
Underinterpretation occurs when the analyst does not extract

all of the information that is actually present in the data. For
instance, signal Ytidths, which are often disregarded, can provide
information on both intermixing and roughness. The analyst
might not be interested in that further information, but often, it is
simply due to lack of knowledge (for instance, that newgeneration codes can extract roughness information with
considerable ease). However, most often, undcrintcrpretation
occurs when the data are not well fitted Modem codes are now
available that facilitate excellent fitting of spectra, enabling the
analyst to put much greater weight on the results.
For a target of semi-infinite thickness, the validity of this

calculation rests on the fact that the fraction of incident particles
reflected or scattered out of the target is negligible 1 %). which
is true for all beams at megaelectronvolt energies. It further
requires that secondary particle emission (sputtered particle flux,
nuclear reaction product flux, electron emission, etc.) produces
no secondary heating effects. Again, for megaelectronvolt
particles, this assumption is usually satisfied. For thin targets,
As a general rule, it is always desirable to collect multiple

spectra from the same sample. Thia can be done using multiple
detectors installed in the experimental chamber or by performing
different experiments including different techniques. Although
this is not necessary in simple cases. to analyze complex
samples meaningfully, multiple spectra and multiple techniques
are aimo,t always required. The typical example is to use RBS,
EBS, ERDA, and PIXE to be sensitive to the heavy e1ements,
specific light elements, hydrogen, and minor contaminants. All
spectra collected should be analyzed -simultaneously with the
same depth profile, to ensure a self-consistent analysis.
the heat input is determined by the energy deposited by the beam

within the target, which might be much less than the incident
energy. Although it is usually of no interest in iOD beam analysis
applications. heat generation in the target proceeds through thesame fundamental interactions that have been identified in
discussions of "spike" phenomena. Beam energy expended
through electronic stopping is coupled to electrons. that is, the
electrons are "heated", and this excitation must then be coupled
to the lattice through electron-phonan interactions, thereby
producing macroscopic sample heating with a characteristic time
constant of picoseconds. The' route to lattice heating is more
direct in the nuclear stopping regime, which dominates near the
end of the particle range. As noted above, however, the quantity
of interest is usually the mean power deposited. The variation in
rate of thennaJization of the lattice with depth is rarely of
concern.
A major pitfall is the use of the wrong value for the product
of the charge and solid angle, QQ, where normalization of this
quantity is not easily available from the spectrum. QO is
implicitly determined by the composition of the sample (where
there is no cbanneling). so that, if the wrong value is imposed on
the data, the results can be severely distorted. An illuminating
example in which the uniqueness of Q!l was used to determine
the sample composition with great elegance was reported by Lee
et al. (2006)
Finally, an outstanding pitfall to bc avoided is the
overinteIpretatian of poorly fitted data, which must always be
treated with great caution. Given that particle scattering spectra
are not easily inverted into depth profiles, because the signal at a
particular channel can come from various elements at different
depths, if the spectrum is poorly fitted, the model could be
entirely wrong! In particular, the fact that part of the spectrum
The sample response to beam heating depends on

experimental conditions. If the beam spot is small in area, radial
symmetry can be assumed. At temperatures above a few
hundred degrees Celsius, cooling is dominated by the radiative
mechanism ~), rather than by conduction. (When conductivity
within the sample is good, temperature profiles are rather
unremarkable.)
can be fitted petfectly does not mean that the model is partly
374
,j
j
Ji"
permanent damage effects. In such materials, the major source

of radiation damage is the small nuclear stopping component
(first mechanism, above). An upper limit to the resulting defect'
density can be obtained by multiplying the nuclear stopping
power, (dE/dx)n (in eV lO-u cm2), by the total ion fluence and
dividing by twice the displacement energy, ~. At IBA energies,
15
2
the nuclear stopping power is usually between 0.03 eV 10- cm
and 0.3 eV 10-1 cm?, and Ed is typically-3D eV_ Hence, a beam
fluence of lOiS iOllS/cm2 produces, at most, a defect density in
the 0.05--1).5% range.
However, mechanical contact does not ensure good thermal

contact, and tbennal contacts are notoriously bard to make in a
vacuum. The existence of a hot spot can lead to
differences in electron emission;
differences in sticking probabilities for adsorbates (e.g., O2 .
CO, H20) and, thus, lower coverages relative to cooler
surrounding material;
varying decomposition rates for adsorbed hydrocarbons; and
annealing of beam-induced disorder.
Thennal gradients can result in the redistribution of mobile
impurities, including embedded gas atoms from the beam, and
can cause decomposition of materials. The question of how best
to determine the target temperature under the beam lies beyond
the scope of these comments. Schultz ef al. (1992) showed how
important temperature control can be: they reported large
changes (up to an order of magnitude) in the level of residual
disorder in megaelectronvolt energy self-irradiation of Si over
the temperature range of 27Ji-320 K for otherwise identical
irradiation conditions. Hence, significant changes in beaminduced disorder can result from beam heating of even a few
degrees.
It is worth emphasizing that beam-induced heating effects are
exacerbated when a microbearn is used because the beam current
density is significantly increased during the mirobeam analysis.
15.6.2 Beam-induced radiation damage

In many rnA applications, radiation damage produced by the
ion beam is a key limitation. ID general, such effects depend not
only on the primary energy-loss process involved, but also on
subsequent solid-state diffusion effects. Hence, the nature and temperature of the target, the dose rate, and the total fluence are
all relevant parameters. The review literature on radiation
damage is extensive (see, for example, Grubb, 1974; Weber et
al., 1997; Averback and de la Rubia, 1998), and we provide here
only some simple guidelines to assist newcomers in the lBA
field.
On the other hand, in insulators and other molecular

compounds, damage production rates are often considerably
greater because, in this case, the electronic stopping power
(second mechanism, above) can also cause bond breakage and,
hence, permanent chemical and structural changes in the target.
Radiation chemistry studies in a wide variety of inorganic and
organic molecular solids, using electron, y-ray, and
megaelectronvolt ion bombardment, have shown that the
nwnber of bonds broken per 100 eV of deposited energy (the 50called g factor) is roughly 10, indicating that, in such materials,
electronic stopping processes are usually more effective in
breaking bonds than the nuclear collision cascades. For example,
Benyagoub (2006) showed that the fraction, g, of electronic
energy loss converted to heat is 0.124 for zirconia and 0.129 for
hafuia. Because the electronic stopping power at IBA energies is
two to three orders of magnitude greater than the nuclear
stopping power, it is evident that the total rate of damage
creation in insulato_rs can be as much as 10) times greater than
the nuclear collision estimates in the previous paragraph.
Radiation damage processes arise from two widely different

mechanisms of energy transfer:
In practice, the observed levels of damage vary widely from

one type of insulator to another. We have presented here only a
rough cstimate of the maximum-rote-at which damage is created;
in many materials, self-annealing reduces -the damage. For
example, most ceramic materials
BN) and certain
inorganic oxides (MgO, Al z0 3, Si02, U0 2) are fairly resistant,
whereas alkali halides, polymers, and most organic and
biological materials are rapidly and permanently decomposed by
electronic stopping processes. In particular, the beam is actually
capable of rapidly drilling holes in poly(tetraf1uoroethylene)
ewe,
(PTFE) (Grime ef al., 2005).

Damage creation due to electronic stopping is not restricted
to insulators; rather, it can occur in any polyatomic molecular
solid, including high-Tc superconductors.
In nuclear (or atomic) stopping, (dFJdx)~ scattering of the

incident ion by the (partially) screened target nucleus results
in significant momentum being transferred to the whole
atom This contribution is the dominant energy-loss process
at low (kiloelectronvolt) energies, but at energies common in
rnA (-1 MeV/nucleon), it is only a very small fraction
(...().I%) of the total stopping process.
Finally, it should be emphasized that, even when a large

amount of damage is created by the analyzing beam, this does
~ot necessarily introduce significant error into the quantities
being- analyzed. However, if some of the resulting defect species
are mobile (radiation-enhanced diffusion) or volati1e (for
example, Hz), then significant changes in depth distribution or
stoichiometry might result. In some cases, low-temperature
analysis can reduce the problem.
In electronic stopping, (dE/dx)e, energy is lost to various

electronic excitation and ionization processes.
In metals and most semiconductors, electronic excitation and

ionization both decay almost instantaneously without producing
375
>
Chapter 15
SingJe-crystal targets, which involve RBS/ch~el~g

analyses, are particuJ.arly sensitive to all types of radiatIon
damage, even when the resulting defect species are immobile.
Interactions between solute atoms and point defects (vacancies
or interstitials), for example, can cause solute atoms to move
into a completely different lattice site configuration (see Chapter
12, Section 12.5.5).
Again, it is worth emphasizing that beam-induced damage
effects are exacerbated when a microbeam. is uc;ed.
15.6.3 Beam-Induced sputtering
As part of the radiation damage process, near-surface atoms
occasionally receive sufficient kinetic. energy or electronic
excitation to be ejected-sputtered--from the target surface.
Obviously, if the amount sputtered during rnA beco~es
2
significant (lOIS crn- ;:;:; one monolayer), then the such properties
as the surface structure, stoichiometry, and film thickness could
be irreversibly changed Again, as in the radiation damage
section, we divide the discussion into two widely different types
of behavior
In metals and most semiconductors, only the nuclear stopping
component contributes to sputtering, and the observed yield is
.directly proportional to (dFJdx)n evaluated at the target surface,
that is, at the incident beam energy, Eo. The relationship between
the sputtering yield, Y (atoms/incident ion), and (dFJdx)"
[eV/(1015 atoms cm-2)] can be expressed as
O.l(dEIdx)n ~ __ , .
U, cos<j>
(15.4)
The numerntor [O.I(dFJdx),J is a rough estimate of the total

energy (in eY) contnbuting to the sputtering process, U, (the
surface binding euergy of the target) is usually between 2
eV/atom. and 10 eV/atom, and $ is the angle between the
- ~-~ incident beam. -direction and the surface normal. Andersen
(1987) showed, fortuitously, an exceptionally goOO theoretical
description of the sputtering behavior ofNi (a typical metal) as a
ftmction of the energy and atomic number of the incident beam.
The maximum sputtering yield is about 10-2 Ni atoms/incident
ion for about 1 ke V H and about 8 Ni atoms/ion for 100 keV Xe.
376
In general however, the agreement between predicted and

measured sputtering yields is no better than a factor of 2.
In the low-energy regime (10-100 keY), where (dFJdx)"

reaches its maximum value, sputtering yields as large as 10
atoms/ion are often encountered. However, at IBA energies,
(dFJdx)" is in the range ofO.O~.3 eV/(IO" atoms cm-'); hence,
the sputtering yield is less than ) 0-2
In insulators, the collision cascade mechanism of sputtering
(described above) still occurs, and hence, sputtering yields of at
3
least 10- atoms/ion occur in all rnA studies. In insulators,
however the much larger (dEldx)e process can also produce
signific~t sputtering through a variety of mechanisms: for
example, bond breaking, "Coulomb explosion" effects, and
fonnation of volatile products (H" 0" etc.). The exact
mechanisms are, in general, not well characterized, and they
depend markedly on the type of insulating material involved.
Nevertheless, experimental evidence (Tombrello, 19&4) shows
that the sputtering yield in insulators can be ~ h~gh ~ 10
atoms/incident ion (in rare cases, even 100 atoms/mcldent Ion).
In such materials, the electronic st~ping process bas enhanced
the sputtering rate by as much as 10 .
Note that, in polyatomic targets, preferential sputtering of
certain atomic species can often occur. Hence, whenever the .
total amount sputtered is greater than one monolayer (_ 1015 cm-2),
Significant changes in near-surface stoichiometry can als.o result
An extreme example is the rapid dehydrogenatIon (or
graphitization) of ~lymers that occurs during IDA with
megaelectronvolt He beams [see, for example, the reVJew of
polymer damage by Brown (1986)].
.
.
15.6.4 Charging in insulators
Compared to metals and semiconductors, m~st msulatmg .....
materials present a series of special problems WIth respect to
IBA In the preceding section on beam-induced ~ge, the .. '
greatly enhanced sensitivity of insulators to beam-mduced '..
effects such as radiation damage and sputtering was alr~dy..
emphasized. Here, we briefly Dote two additional complicatJ~ns..
that arise in certain types of insulating materials: target chargmg
and photon emission.
Energy (MeV)
0.3
0.4
0.5
0.6
0.7
800~--~~-----,------~------.-----~-.
~600
a;
'"
!'"
~c:
:::I
...o
400
Grounded
'C
a;
:>
200
o
100
150
200
250
300
350
Channel
FIG. 15.7. Surface charging effet:::t. Comparison of RES spectra from a quartz target using 1 MeV 4He: ungrounded and grounded
through a thin conductive surface layer of graphite by rubbing a pencil lightly across the swface (Almeida and Macauley-Newcombe.
private comm~nication 1991) ..
Depending on the geometry, thickness, and resistivity of the

target material, surface charging under megaelectronvolt ion
bombardment can, in severe cases, reach several tens of
kiloelectronvolts (Kim e/ al. 2(03). (This value is. of course.
still small compared to the incident beam energy, unlike the
inswator situation encountered in Auger and secondary ion mass
spectroscopy.) However. because the charge states of the
incident and backscattered beam are not necessarily the same,
surface charging can seriously distort the energy and, hence, the
shape of the observed RBS spectrum, as sho"\llD for the case of
quartz in Fig. 15.7. Because the backscattered He particles are
mainly 4He2+, one can use 'He2+ as the incident beam to reduce
One effective way to neutralize surface charging effects is to

provide a supply of low~E electrons from a small, hot filament
located nearby. If such a filament is powered through an
isolation transformer and is electrically connected to the target
holder through a suitable (+50 V) bias to prevent electrons from
escaping to the chamber walls, then quantitative current
integration can still be achieved. Generally, the use of a
transmission Faraday cup is a better way to solve these cunent
integration problems.
.
the charging-induced energy shift in the RBS analysis of
coating the surface at least partially with a very thin layer of
insulating specimens.
conducting material, such as graphite (for example, rubbing a

pencil lightly across the surface sometimes works) or an
evaporated metal (or carbon), and
Other possible methods that have been used to reduce surfa.ce
charging include
Furthermore,
surface charging sometimes produces
sufficiently high electric fields to interfere with the performance
of the Faraday cup system. 'Also, in certain insulators (for
example, BaTiOJ ), excessive surface charging can even cause
the target to disintegrate.
sweeping the beam so that it also bombards the adjacent

(metal) target bolder or a suitably placed metal grid in front
377
L>.o...... __
Chapter 15
However, it is normally impossible to determine, from IBA

techniques alone, whether the broadening of a given peak Or
edge is due to roughness, interdiffusion. or an unexpected depth
profile of the affected element. Intcrpretation of the data almost
always requires extra information about the sample at band from
some other source. Different microscopies are often employed
for that pwpose.
of the target, which generates a supply of secondary electrons to

neutralize the positively charged swface of the insulator.
In many insulators, charging effects are small enough to be

ignored, or ,discharge by surface tracking can be encouraged by
performing the analysisclose to a contact.
15.6.5 Photon emission in insulators
In insulators, much of the electronic excitation resulting from
the (dEJdx)e stopping mechanism is eventually converted into
optical emission .in the visible and u1traviolet regions. Polymers,
alkali halides, and other materials are transparent to such
radiation. Consequently, photons emitted even from a depth of
several microns can readily escape from the target. In such
cases, hundreds or even thousands of photons can be emitted per
incident ion. We have already discussed how these photons can
generate a flux of secondaIy electrons at the chamber waIls.
Furthermore, the energy resolution of charged-particle detectors
can be seriously degraded by photogeneration of carriers in the
depleted region. (The standard surface barrier detector has a Au
electrode with a thickness of -20 run that is transparent to
visible light.)
15.7.2 Target nonuniformity

Care must be taken in basing the interpretation of RBS
spectra on limited data. Apparently simple features can be
misleading if it is assumed that tlie target has lateral uniformity
under the analyzing beam. As shown by Campisano et al.
(1975), the shape ofa Pb spectrum recorded from a nonuniform
surface film of Pb on Si can c1ose1y resemble that of a laterally
uniform diffusion profile ofPb in Si.
II
Ii
This confusion is removed by supplementary measurements '."'j

aIld aIlalysis (Campisano ef al., 1978). As shown in fig. 15.8(a), ,i
the consequences of varying the target angle, $, with respect to
the incident beam for a Pb film of 50-nm thickness overlaying a )1
Si substrate are to give a familiar (cos $r' broadening of the Pb .j
profile in the RBS spectrum and a correlated shift in the leading ,1
edge of the Si continuum. Completely different profiles are . .,
obtained for samples that have been annealed to 280'C for 20
min., after deposition. In this case (Fig. 15.8(b)], the decrease in
the scattered particle yield, Hsit from Si atoms at the surface as ,:i
the tilting aIlgle ($) increases rules out interdiffusion. The results./
were modeled in terms of raised Ph features (islands), and good .-;
agreement was found with island dimensions and spacings -~~
measured by scanning electron microscopy (SEM) analysis,
which displayed the altered surfuce morphology. RBS methods
have been used very effectively by Zinke-Allmang and co- . 'l.~,
workers (Carlow el al., 1997) to study the development of layer
by-layer and island growth (Stranski-Krastanov films).
l.~,:
15.7 OTHER EFFECTS

lS.7.1 Surfnccand interface roughness
In experiments where the incident or emergent charged
particle makes a small angle (for example, <15) with a target
surface that is not smooth, problems in the shape of the energy
distribution can arise as a result of the surface topography. Th..is
is very often the case for ERDA measurements (see Chapter 5).
Even at near-perpendicular incidence, materials such as porous
targets and sintered powders can introduce very serious surface
roughness complications, leading to data that are either
impossible or troublesome to analyze or that do not carry useful
infonnation. Recently. Molodtsov et al. (2008) reported a
method for haIldling spectra from very rough samples, but this
algorithm is not yet implemented in rnA codes.
.1
{I
"1
.J
15.7.3 Thin ftlm units

'"
Thin film units (lOiS atoms/em2) are the natural units for IBA .~~
2
because the energy loss is measured in eV/(atomslem ), and ODe :~
monolayer is on the order of lOIS atoms/eml
When roughness is moderate (that is, when it is not so great

that it destroys spectral features),IBA can be used effectively to
study surface and interfacial roughness. Different methods have
been developed. The simplest one is to average spectra
calculated for different sample structures, for example, if the
roughness leads to a layer with changing thickness. Full Monte
Carlo simulations have also been perfonned. A fast and often
efficient approach is to calculate the broadening of observed
features due to specific types of roughness. Calculations of the
effect of voids, inclusions, and quantum dots are also included in
some codes (Barradas, 2001; Mayer el al., 2005).
.~
.'.
Energy loss is measured in these units because an accurate ,:,.;

measure of thin-film thickness is in terms of ~g/cm2 (equivalent ~.:.'f,i<
to atorns/cm 2 if the stoiChiometry js known) and the mo~r ~;f
reliable way to measure the thiekness of a thin film is to weigh It ,.~.,;
and measure its area. Linear thicknesses (in nanometerss) are ;".
unreliable because the densities of thin films can be markedly ~;;
different from the bulk values and because surface
contamination. (including oxides) and other surface effects can
be large for thm films.
_
.
f'iil' .
,"Ii-
. "'fr'
.~:'
:,,;.1 .
.~,.;~
378
6.
'.~
.
. 6 .....
51
. 2.0 MeV H"G+
a. 4
..
.ooi.
. As dep()sited
'in'
<::
_____ .
D
~~8.
70"
b
4
'"0
:.'\]HSl
2t
Pb
..
~
....
D
o----~
~
(a)
..
."
CD 0
(b)
>.
."
<:
.;:
.,
co
.,u
...
/v
..
ID
Pb
Annealed at 280'C
51
(20 min.)
2
8Q'
1.0
IPb
1.5
Energy (MeV)
FIG. 15.8. Ambiguity in RBS between roughness and diffusion, Backscattered energy spectra for several tilting angles: (a) from a
uniform Pb layer (50 run) on a Si substrate and (b) from Ibe same layer after being annealed for 20 min. at 280'C (Campi,ano ef 01.,
1978).
The pitfall that must be avoided is forgetting Ib,t presenting
IBA is an ideal technique for measuring thin-film profiles

where the composition changes with depth. In these cases, the
depth profiles with a linear depth scale (in DaDometers) hides

assumptions about the density of the sample, which might not bc
well known and might be greatly different from what is
expected. Some rnA codes allow the analyst to fit the data using
molecules and mixtures of molecular densities. In Fig. 15.6, for
example, the depth scale is in thin film units. To plot the profiles
in nanometers, one must correctly assign the oxide densities,
which is not an entirely trivial exercise.
appropriate deosity to use depeods on Ibe chemistry of Ibe

sample, and assuming different chemistries will lead one to infer
a different density profile, Therefore, the most neutral
representation of IBA-measured depths is in terms of Jlg/cm1j
thin film units are very convenient and do not involve the
sample density, but do assume that the stoichiometry is knmvn.
379
~.
.. _.. -
..
_-------Chapter 15
Barradas, N.P., leynes, e., Webb, R.P., and Wendler, E. (2002),

ACKNOWLEDGEMENT
This work is a revision of, and incorporates part of the

previous 1995 Handbook of Modern Ion Beam Materials
Analysis (the "Black Handbook") Chapter 12 by l.A. Davies,
W.N. Lennard, and 1.V. Mitchell. The section on electronic
noise is by Max Dobeli (Zurich).
Barradas, N.P. (2004), Nuel. Instrum. Methods B225, 318.

Barradas, N.P., and Reis, MA (2006), X-ray Spectrom. 35 (4),
232.
Barradas, N.P., Arstila, K., Battistig, G., Bianconi, M.,

Dytlewllki, N., Jeynes, C., K6tai, E., Lulli, G., Mayer, M.,
REFERENCES
Abel. F., Amsel, G., d'Artemare, E., Ortega., C, Siejka, J. t and

Vizkelethy, G. (1990),Nuel. Instrum. Methods 845,100.
Raubala, E., SzilAgyi, E., and Thompson, M. (2007),

Alexander, T.K.., Ball, G.C., Lennard, W.N., Geissel, H., and

Mal<, H.-B. (198'2, Nuc!. Phys. A427, 526.
Barradas, N.P., Arstila, K.., Battistig, G., Bianconi, M.,

. Dytlewski, N., Jeynes, c., K6tai, E., Lulli, G., Mayer, M.,
Alkemade, PEA., Habraken, F.H.P.M., and van der Weg, W.F.

(1990), Nuel. Instrum. Methods 845, 139.
Raub.la, E., Szilagyi, E., and Thompson, M. (2008),

"
'1,'
Barradas, N.P., and leynes, C. (2008), Nuel. lnstrum. Methods

B266, 1875.
Allison, S.K.. (1958), Rev. Mod. Phys. 30 (4),1137.

Amse\, G., and Samuel, D. (1967), Anal. Chem. 39, 1689.
Bauer, P., and Bortels, G. (1990), Nucl.lnstrom. Methods A299,

205.
Amsel, G., and Davies, A. (1983),Nucl. Instrum. Methods 218,

177.
Bauer, P., Steinbauer. E., and Biers.ck, l.P. (1992), Nucl.

Instrom. Me'hods B64, 711.
Amsel, G., Girard, E., Vizkelethy, G., Battistig, G., Girard, Y.,
and Szilagyi, E. (1992), Nuel. Instrum. Methods B64,
811.
l
~
{,
Bauer, p,. Steinbauer, E., and Biersack, J.P. (1993). Nucl.

Ins/rom. Methods B79, 443.
Andersen, H.H., Besenbacher, F., Loftager, P., and Meiiler, W.

(1980), Phys. Rev. A21, 1891.
Benyagoub, A. (2006), Nuel. Instrum. Methods B245, 225.
~dersen: H.H.-(1987), M,c/.lnstrum. Methods B18, 321.

Arstila, K.., Sajavaara, T., and Keinonen, 1. (2001), Nuel.
Audi, G., Wapstrab, A.H., and Tbibaulta, C. (1003), Nud. Phys.
....._._A729, 337 ..--"' ....
Bianconi, M., Abel, A., Banks, le., Climent Font, A., Cuben,
C., Doyle, B.L., Lotti, R, Lulli, G., Nipoti, R., Vickridge,
I., Walsh, D., and Wendler, E. (2000),Nuel. Instrum
Methods BI61-163, 293.
..
Blaauw, M, Campbell, J.L., Fazini6, S., Jaldie, M., Orlie, I., and
Van Espen, P. (2002), Nuel. Instrum. Methods B189, 113.
'.
Averback, R.S., and de la Rubia, T.D. (1998), Solid State Phys.

51,281.
Brice, D.K.., and Doyle, B.L. (1990), Nuel. Ins/rum. Methods

845,265.
BAM (2009), BAM-S107, BAM-SI08, BAM-SI09, and BAMSIlO. in Cawlogue o/Certified Reference Materials,
Bundesanstalt rur Materialforschung und -priifung
(BAM), Berlin, Germany.
Brown, W.L. (1986), Radial. Eff 98, liS.

Butler, l.W. (1990), Nuel. Instrum. Methods 845, 160.
Boudreault, G., Claudio, G., Jeynes, C., Low, R., and Sealy, BJ.
(2004.), Nuel. Instrum. Methods B217 (I), 177.
Banks, J.c., Browning, J.F., Wampler, W.R., Doyle, B.L.,

LaDuca, CA, Tesmer, l.R, Wetteland, e.J., and Wang,
Y.Q. (2004), Nuel. Instrum. Methods B219-220, 444.
Barradas. N.P., Jeynes, c., Jenkin, M., and Marriott, P.K.

(1999a), Thin Solid Films 343-344, 3 I.
Boudreault, G.; Elliman, R.G., Grotzsebel, R, Gujrathi, S.c.,

Jeynes, C., Lennard, W.N., Rauhala, E., Sajavaara, T.,
Timmers, H., Wang, Y.Q., and Weijers, T.D.M. (2004b),
Barradas, N.P., Keddie, 1.L., and Saekin, R. (1999b) Phys. Rev.

E59 (5), 6138.
Campisano, S.U., Foti, G., Grasso, F., and Rimini, E. (1975),

Thin Solid Films 25, 43 L
Barr.das, N.P. (2001), J. Phys. D: Appl. Phys. 34,2109.
Campisano, S.U" Ciavola, G., Costanzo, E., Foti, G., and

Rimini, E. (1978), Nuel. Instrnm. Methods 149, 229.
Barradas, N.P. (2002), Nuel. Instrum. Methods B190, 247.
380
!,
:!"
:,~
:~~
BlPMI1ECIIFCCIlSOIIUPACIIUPAP/OIML, identical to
Carlow, G.R., Barel, R.J., Zinke-Allmang, M. (1997),

Phys.Rev,B 56(19), 12l 19
EN 13005:1999. See also the valuable Web site

http://www.npl.co.uklmathematics-scientificcomputing/measurement-uncertainties!.
Christensen, N.S., Jensen, F., Besenbacher, F., and Stensgaard,

I. (1990), Nucl. Instrum. Methods B51, 97.
Cohen, c., Davies, J.A., Drigo, A.V., and Jackman, T.E. (1983),
Nucl.lnstrum. Methods 218, 147.
'Cornedi, D., and Davie., JA (1992), Nucl.lnstrum. Methods
B67,93.
Davies, JA, and Norton, P.R. (1980), Nile/. Instrum. Methods
168,611.
Gurbieh, A.F., and Kornilov, N.V. (1991),Nucl. Instrum.

Methods 862, lSI.
Gurbich, A.F. (1995), Nllei. Instrum. Methods A364, 496.
Gurbieh, A.F. (1996), Nucl. Jnstrum. Methods 8111, 137.
Gurbieh, A.F. (1997), Nllcl.lnstrum. Methods B129, 311.
Gurhieh, A.F. (1998), Nllel. Instrum. Methods 8136-138,60.
Davies, J.A., Jackman, T.E., Escbbach, H.L., Domba, W.,

Watjen, U., and Chive"" D. (1986), Nuel.lnstrum.
Methods 815, 238.
Gurbich, A.F. (2000), Nllel. Jnstrum. Methods BI61-163, 125.

Gurbieb, A.F., and Molodtsov, S.L. (2004), N.d. I.strum.
Methods B226, 637.
Doolittle, L.R. (1985), Nucl. lnstrum. Methods B9, 344.
Gurbich, A.F. (2004), Nllcl.lnstrum. Methods 8217,183.
Doolittle, L.R. (1986), Nucl. Jnstrum. Methods B15, 227.
GurblCh, A., and leynes, C. (2007), Nucl. lnstrum. Methods

8265,447.
Edge, R.D. (1983), IEEE Trans. Nucl. Sci. NS-30 (2), 1685.
El Bouanani, M., Pelican, P., Razpet, A., CadeZ, I., Budnar, M.,
Simcic, J., and Markelj, S. (2006), Nuel. lnstnlm.
Methods B243, 392.
Gurbich, A. (2008), Nllc/. Instrum. Methods B266, 1193.

Hemment, P.L.F., Singleton, J.F., and Stephens, K.G. (1975),
Thin SoUd Films 28, L1.
Eckstein, W., and Mayer, M. (1999),Nuci. Ins/rum. Methods

8153,337.
Hautala, M., and Luomajarvi, M. (1980), Radiat. Elf 45, 159.
Ecker, K.H., Watjeo, U., Berger, A., Persson, L., Pritzkaw, W.,
Radtke, M., and Riesemeier, H. (2002), Nucl.lnstrum.
Methods B188, 120.
Huenges, E., Rosier, H., and Vonaeh, H. (1973), Phys. Lett. 846
(3),361.
lBANDL (2009), database at www-nds.iaea.orgiibandl
maintained by AF.Gurbich.'::::- ~- -....-.---.-~-.- .:.-:.
Fischer, R., Mayer, M., von der Linden,.W., and Dose, V.

(1997), Phys. Rev. E55, 6667.
.
~.
--
..
ICRU (1993), Stopping Powers and Rangesfor Protons and Alpho

Particles, lCRU Report 49, ICRU, Bethesda, MD.
Gardoer, R.P., and Wielopolski, L. (1977), Nile/. Instrum.

Methods 140, 289.
Jeynes,
Geissel, H., Lennard, W.N., Ball, G.c., Forster, J.S., Lone,

MA, Milani, L., Phillips, D., and Plattner, H.H. (1984),
Nllei. Instrum. Methods 230, 770.
c., and Kimber, A.C. (1985), J. Phys. D. 18, L93.
Joynes, C., Jafri, Z.H., Webb, R.P., Ashwin, M.J., and Kimber,
A.C. (1997), Suif. Interface Anal. 25, 254.
Giuntini, L., and Mando, PA (1994), Nllel. Instrum. Methods

885,744.
Jeynes, c., Barradas, N.P., Blewett, M.J., and Webb, R.P.

(1998), Nuc/. Instrum. Methods B136-138, 1229.
Goulding, F.S., and Landis, D.A. (1982), IEEE Trans. Nuc/. Sci.
NS-29, 1125.
Jeynes, c., Barradas, N.P., Marriott, P.K., BoudreauIt, G.,

Jenkin, M., Wendler, E., and Webb, R.P. (2003), J. Phys.
D: Appl. Phys. 36, R97.
Grime, O.W., Sofield, CJ., Gomez-Morilla, I., Gwilliam, R,

Ynsa, M.D., and Enguita, O. (2005), Nile/. Instrum.
Methods B231, 378. This effect has also been observed in
a poor vacuum.
Jeynes, C., Peng, N., Barradas, N.P., and Gwilliam, R.M.

(2006), Nllcl. I"strum. Methods 8249, 482.
Kim, J., Hong W, Woo Hi, Eum CH (2003), J. Koreom Phys.
Soc. 43, 582.
Grubb, D.T. (1974),J. Mater. Sci. 9,1715.

GUM (1995), Guide to the Expression of Uncertainty in
Kirkpatrick, S., Gelatt Jr., C.D., and Vecchi, M.P. (1983),

Science 220, 671.
Measurement,
381
- ---_._-----==
Chapter 15
Kon8C, G., Kalbitzer, S., Klatt, C., Niemann, D., and Stoll, R.
Paul, H., and Schinner, A. (2002), Nuel. Instrum. Methods Phys.

Res. B195, 166.
(1998), Nuel.lnstrum. MethodsB136--138, 159.

Knoll, G.F. (1989), Radiation Detection and Measurement, 2nd
Paul, H., and Schinner, A. (2003), Nuel. instrum. Methods Phys.

Res. B209 252; see also Paul, H., and Schinner A. (2005),
Ed., John Wiley and Sons, New York.

L'Ecuyer, J., Davies, J.A., and Matsunami, N. (1979), Nuel.
Paul, H., and Schinner, A. (2006), Nuel. instrum. Methods Phys.

Res. B249, 1.
Lee, W.P., Gundabala, V.R., Akpa, B.S., Johns, M.L., Jeynes,
Piel, N . Berheide, M., Polaczyk, C., Rolfs,
C., and Routh, AF. (2006), Langmuir 22, 5314.
c., and Schulte,
W.H. (1994), Nucl.lnstrum. Methods A349 (I), 18.
Lennard, W.N., Geissel, H., Winterbon, K.B., Phillips, D.,
Pritzkow, W., Vogl, 1., Berger, A., Ecker, K., Grotzschel, R.,
Klingbeil, P., Persson, L., Riebe, G., and W!l.tjen, U.
(2001), Fresenius J. Anal. Chem. 371,867.
Alexander, T.K, and Forster, J.S. (1986), Nud. lnstrum.

Methods A248, 454.
Lennard, W.N., Tong, S.Y., Mitchell, I.V., and Massoumi, G.R.
(1989), Nucl.lnstrum. Methods B43, 187.
Polanyi, M. (1958), Personal Knowledge, Chicago University

Press, Chicago, IL. I quote from the 1974 paperback
Lennard, W.N., and Massoumi, G.R. (1990), Nuel.Instrum.

Merhods B48, 47.
edition, chap. 2, p. 19. A beautiful book, invaluable for
those who want to know how we know what we know.
Lennard, W.N., Massoumi, G.R., Simpson, T.W., and Mitchell,

I.V. (1999), Nuel.lnstrum. Methods B152, 370.
Radeka, V. (1988), Annu. Rev. Nuel. Part. Sci. 38, 217.
Marion, J.B. (1966), Rev. Mod. Phys. 38, 660.
Reinholz, V., Bremser, W., Brzezinka, K W., Strub, E., Weise,

H.P., and Merchel, S. (2008), Nuc!.lnstrum. Methods
B266,2418.
Mattauch, J.H.E., Thiele, W., and Wapslra, AH. (1965), Nuel.

Phys. 67, 73.
Roush, M.L., West, M.A., and Marion, J.B. (1970), Nuel. Phys.
A147,235.
Mayer, M., Fischer, R., Lindig, S., von Toussaint, V., Stark,
R W., and Dose, V. (2005), Nuel. Instrum. Methods
Schultz, PJ., Jagadish, C., Ridgway, M.e., Ellirnan, RG., and

Williams, J.S. (1991), Phys. Rev. B44,9118.
B228,349.
Mendenhall, M.H., and Weller, RA. (1991), Nucl.lnstrum.
Methods B58, II.
Seah, M.P., David, D., Davies, lA., Jeynes, C., Ortega, C.,
Read, PM., Sofield, C.J., and Weber, G. (1988), Nucl.
Moller, W., and Besenbacher, F. (1980), Nuel.lnstrum. Methods

168, 111.
SigmaCalc (2009), database and program at wwwnds.iaea.orgl'igmacalc built by A.F. Gurbicb.
Molodtsov, S.L., Gurbich, AF., and Jeynes, C. (2008), J. Phys.

D: Appl. Phys. 41,205303.
Sitter, C., Davies, 1.A., Jackman, T.E., and Norton P.R. (1982),
Rev. Sci. Instrum. 53,797.
Morrison, R (2007), Grounding and Shielding: Circuits and
Interference, Wiley, New York.
Sjoland, KA., Munnik, F., Chaves, C., and Wdljen, U. (1999),

Nucl. instrum. Methods B150, 69.
Munnik, F., Plompen, A.J.M., Rliislinen, J., and Watjen, V.

(1996), Nuel.Instrum. Methods B1I9, 445.
Sjoland, K.A., Munnik, F., and Wdtjen, U. (2000), Nuel. instrum.

Methods BI61, 275.
Niemann, D., Konac, G., and Kalbitzer, S. (1996), Nuel.

inslrum. Methods B118, II.
Smyth,D.M. (l966),J. Electrochem. Soc. 113,1271.
Pascual-Izana, C., and Barradas, N.P. (2008), Nuel. Instrum.

Methods Phys. Res. B266, 1866.
Szihigyi, E., Paszti, F., and Arosel, G. (1995),Nucl.lnstrum.

Methods BI00, 103.
Paszti, F., Manuaba, A., Hajdu, C., MelD, A.A., and Da Silva,
M.F. (1990), Nuel. instrum. Methods B47, 187.
Tesmer, J.R., and Nastasi, M. (eds.) (1995), Handbook of
Modem ion Beam Materials Analysis, Materials Research

Society, Pittsburgh, PA.
Paul, H., and Schinner, A (2001), Nuel. instrum. Methods Phys.

Res. B179, 299.
Tombrello, T.A. (1984),Nucl. Instrum. Methods B2, 555.
382
Pitfalls In Ion Beam Analysis
L.W., Navrotsky, A., Price, D.L.. Stoneham, A.M., and

Weinberg, W.C. (1997). J. Mater. Res. 12 (8), 1946.
van Lieshout, R., Wapstra, A.H., Ricci, R.A.. and Girgis, R.K.
(1966), "Scintillation spectra analysis", in Alpha-, betaand gamma-ray spectroscopy (Siegbabn, K., cd.), NorthHolland, Amsterdam, p. 515.
Wielopolski, L., and Gardner, R.P. (1976), Nud.lnstrum.

Methods 133 303.
Vonkatesan, T., Brown, W.L., and Wilkens, B.1. (1984), Nud.

[nstrum. Methods BI, 605.
Wielopolski, L., and Gardner, R.P. (1977), Nucl. Instrum.

Methods 140, 297.
Wang, Y.Q., Liso, C.G., Yang, S.S., and Zheng, Z.H. (1990),
Williams, 1.S., and Moller, W. (1978), N"d. Instrum. Methods

157 (2),: 213 ..
Wtitjen, U., and Bax, H. (1994), Nuc/. Instrum. Methods B85,

627.
White, R.E., Barker,P.H., and Lovelock, D.M.1. (1985),

Metr%gia 21, 193.
Weber, W.1., Ewing, R.C., Angell, C.A., Arnold, G.W.,
Ziegler, 1.F. (2004), Nucl. Instrum. MethodsB219-220, 1027;
Connack, A.N., Delaye, J.M., Griscom, D.L., Hobbs,
also see www.srirn.org .
-- ..
383
. ..
~.~
-_..
....-.-_.--,", --
----.-----.---...-..-.....,--...........- ...
_,--"
~~-
-'-
CHAPTER
INSTRUMENTATION AND
LABORATORY PRACTICE
RA. Weller
Vanderbilt University, Nashville, Tennessee, USA
Contributor: Y. Q. Wang
CONTENTS
16.1
INTRODUCTION ....................................................................................................387
16.2 ION SOURCES, ACCELERATORS, AND BEAMS ........................................... 387

16.2.1 Ion sources .......................................................................................................... 387
16.2.2 Accelerators ........................................................................................................ 388
16.2.3 -Beams ............... :........;...................................;......... :.....:..:............ :;.;.................. 389
16.2.4 Beam diagnostics ............................................................................. :... ~ ..............390 ;:~
16.3 DETECTORS ........................................................................................................... 391
16.3.1 Overview .............................................................................................................391
16.3.2 Electron multipliers ...........................................................................................391
16.3.3 Scintillators .........................................................................................................392
16.3.4 Semiconductor detectors ...................................................................................393
16.3.5 Detector systems .................................................................................................399
16.4 ELECTRONICS ...................................................................................................... 400
16.4.1 Introduction ........................................................................................................400
16.4.2 Nuclear eleclronics ............................................................................................. 401
16.4.3 Basic energy spectroscopy ................................................................................. 404
16.4.4 Preamplifiers ......................................................................................................405
16.4.5 Spectroscopy amplifiers .....................................................................................406
16.4.6 Multichannel analyzers .................................................................................... .408
16.4.7 Cahles and connections ...................................................................................... 410
385
Chapter 16
16.5 VACUUM TECHNOLOGY ...............................................................................418

16.5.1 Vacuum practice ................................................................................................ 418
16.5.2 Thiu ftlms ............................................................................................................ 418
16.6 STATISTICS AND DATAHANDLING .............................................................. 419
16.6.1 Counting statistics .............................................................................................. 419
16.6.2 Error propagation ..............................................................................................419
16.6.3 Dead-time correction .........................................................................................420
16.6.4 Random coincidences......................................................................................... 420
16.6.5 Advanced data manipulation ............................................................................ 422
REFERENCES ......................................................................................................................422
386
j
1
Instrumentation and Laboratory Practice
16.1 INTRODUCTION
Although many aspects of mstrwnentation are important for
specific tasks that arise in connection with ion beam analyses,
five general areas are of special significance. These are ion
beam production, transport, and diagnostics; particle detectors;
pulse electronics; vacuum technology and thin-film fabrication;
and statistics and data handling. Anyone who is active in the
field of ion beam analysis can ex.pect to have to deal with some
or all of these areas routinely. This chapter consists of brief
overviews of each of these areas of technology, emphasizing
the relevant topics for ion beam analytical work and suggesting
more detailed references for further study.
Most of the technology that is in current use by materials
scientists for ion beam analysis was originally developed for
the study of nuclei and can be accessed bibliographically
through such key words as "nuclear electronics" and "radiation
detection" and in journals such as Nuclear Instruments and
Methods in Physics Research, Nuclear Physics, and IEEE

Transactions on Nuclear Science. Because much of the
instrumentation needed for ion beam analysis is also widely
used for fundamental nuclear studies, environmental
monitoring, and nuc1ear medicine, it is possible to build a
laboratory almost exclusively with commercial equipment. The
manufacturers of this equipment are often excellent sources of
technical infonnation on both the theory and applications of
their products. Students beginning careers in materials analysis
should especially be encouraged to seek out manufacturer's
catalogs and other literature for the educational value that can
be derived from them.
16.2 ION SOURCES, ACCELERATORS, ANDBEAMS

The steps in the generation of an ion beam suitable for
analytical work are
the production of ions from neutral atoms or molecules,
the acceleration of these ions by electric fields.
the selection of ions with specific mass and energy, and

the transport of the ions to the specimen to be analyzed.
16.2.1 Ion sources

Many ion sources are in common use in ion beam work;
they depend on a wide range of physical principles for their
operation (Brown, 1989). The polarity of ions that is needed
depends on the type of accelerator used. In general, a positive
ion can be produced from any ion in the periodic table, and
some ion sources such as the electron cyclotron resonance
(ECR) source are capable of producing multiply charged ions
of high order. This can be quite useful because, for a given
accelerating VOltage, the energy of a projectile increases with
its charge state. Negative ions are generally more difficult to

obtain, and in some cases, such as nitrogen, they are so unstable
that. for practical purposes, they cannot be produced. Singleended accelerators, those with only one stage of acceleration,
always use positive ioDS. If they did not, they would produce a
parasitic beam of electrons that would generate a very large
background of X-ray radiation. Tandem accelerators of the kind
often used for ion beam analysis require ions to be initially
negative.
Most ion sources used for ion beam analytical work depend
on collisions between particles to generate ions. In sources in

which the original atoms or molecules are gaseous, electron
bombardment is the primary process that leads to ionization.
These sources can be used, however, only for elements that are
volatile or have volatile compounds. Sources that produce ions
by sputtering can be used to produce a wider assortment of
beams but often at reduced current. Sputter sources can be
designed to produce either positive or negative ions. Plasma
SOillces are generally mucb marc efficient at producing positive
ions. One gas-fed source that can produce large quantities of
both polarities is the duoplasmatron. This general-purpose
source is in widespread use. Another commonly encountered
gas-fed source is the radio-frequency or rf source. In these
devices, a plasma is produced in a bottle of low-pressure gas by
the application of a radio-frequency signal of around 100 MHz.
These rf sources are often found in the tenninals of singleended electrostatic accelerators.
Tandem ion accelerators commonly used in ion beam
analysis are typically equipped with tw"o .types of ion sources:- ~:
. either an rf plasma source (alphatross) or a duoplasmatron
source for He ion production and a source of-negative ions by
cesium sputtering (SNICS.II) for all other ions. In the
alphatross, the rf generator couples the power into the helium
gas-fed quartz tube to produce several milliamperes of 4 He.fo
ioDS first. Then, the ~'He.fo ions pass through a rubidium vapor
chamber where a fraction of them acquire two electrons
through belitun-rubidium collisions to become ~e- ions. In the
duoplasmatron ion SOlUce, electrons emitted from the filament
ionize the He gas into 4He.fo ions, which then undergo ion
exchange with lithium vapor to produce 'He beams. In the
SNICS-II source, a tungsten filament ionizes cesium first, and
then, the Cs + ions sputter a cathode to produce negatively
cbarged ions of the species of interest in the cathode materials.
For example, fr ions are efficiently produced by the SNICS-II
approach with the use of a titanium hydride cathode. Other
abundant negative ions that can be produced by the SNlCS-II
method include C, F, Ni, and Au. It should be noted that
negatively charged nitrogen ions are not stable, so instead,
molecular fOIlDS of NC" or TiN" are produ~ed first and then
387
Chapter 16
broken into
terminal.
ions during the stripping stage in the tandem
The capacitance, C, of a Vande Graaff terminal is typically

on the order of 100 pF. The cbarge, q, required to establish a
given voltage, V, follows from the usual relation q = CV. The
voltage is distributed uniformly between the terminal and
ground by a chain of high-voltage resistors and specially
engineered equipotential swfaces. These are designed to
minimize sparks, which can be quite destructive in accelerators
with large stored energy (CV 212). A movable electrode covered
\\'ith needles .called corona points is provided to stabilize the
voltage. The voltage of the terminal can be measured by a
device called a generating voltmeter. This device is a variation
on the concept of a vibrating reed electrometer that is mounted
inside the accelerator lank.. It is useful for measurements at the
level of a few percent, but it is not accurate enough to serve as
the principal reference for thc tcrmina1 voltage, although it is
sometimes used in this way.
16.2.2 Accelerators
The next stage in beam production is the accelerator.
Exhaustive discussions of the types of accelerators now used
widely for materials analysis can be found in the book by
England (1974). Voltages below approximately 500 kV can be
produced and maintained with reasonable success using only
air inswation. Above this value, it becomes necessary to
surround the terminal of the main high-voltage power supply
with a pressurized gas. SLIlfur hexafluoride. SF6 is perhaps the
best insulating gas; althougb various mixtures of CO, and N,
are also lL~ecL The vessels that contain the insulating gas give
accelerators their characteristic tanklike look. Voltages up to
about 400 kV can he generated with relatively conventional
power supplies that use voltage multiplication circuitry similar
to tbe original design of Cockcroft and Walton (England,
1974). Above about 500 kV, other techniques are often used to
generate high voltage. One of the most widely used devices for
the production of megaelectronvolt beams is the Vande Graaff
generator, named for its inventor, Robert J. Van de Graaff. An
engaging history of this ingenious device has been compiled by
Bromley (1974). We limit our further discussion of accelerators
to electrostatic accelerators of this type. Discussions of the
principles by which other accelerators operate can be found in
England's book or in the more recent monograph by Hwnphries
(1986).
., i
~'
All accelerators that are routinely used for materials analysis . ~;"
have tenninals that' are positively charged. In single-ended .:~j

accelerators [Fig. 16.l(a)], the ion source is placed inside the
'.~
high-voltage terminaL A modest voltage (typically from a rew ,,'
kilovolts to perhaps 150 kV) is used to inject ions into the ;
accelerating column. The energy with which ions arrive a1
ground potential is just nV in electronvolts, where n is the ion's ;;
charge. The design of single-ended accelerators is complicated' ~
by the need to provide power to the ion source while it is ,:-.
il
thousands to millions of volts above ground potential. In lower- ~.&;
voltage accelerators, the ion source is powered directly from' ..1"'
the ac power grid through an isolation transformer or by a
generator in the terminal driven by a motor at ground potential "&.
througb an insulating sbaft. In higber-voltage macbines, the .u'
terminal generator is driven directly by the belt or charging ~X
chains. Access to the ion source is also difficult because it is .-.SI
inside th~ pressure vessel. As a result, single-ended accelerators .~"\'!
are usually limited in the range of ions that can be conveniently ~~~
&.!D
produced. Additional flexibility can be ~eali?-Cd by installing an :.~-t
insulated gas feed line to the ion source so that gas changes can
~:
.f.
Van de Graatf generators operate by mechanically moving

electrica1 charge from ground potential to the high-voltage
terminal. For their operation, they depend crucially on the fact
often en~~:)Up.~~4il! a ftrst course on electricity and magnetism
~-=~-that charge always resides on the exterior surface of a
conductor. In older Van de Graatf accelerators, charge is
sprayed from sharp needles raised to a potential of a few tens of
kiloelectronvolts onto a wide rubber-impregnated endless belt
that extends from ground to the high-voltage tenninal. 1bis belt
moves continuously on rollers in such a way that charge is
carried up to and physically inside the higb-voltage tenninal.
The energy to do this is supplicd by an electric mofor that
propels the belt. Once the charge on the belt has entered the
tenninal, it no longer feels a force. It is removed from the belt
by another set of needles that are electrically connected to the
terminal and flows to the terminal's surface. The pelietron
accelerator, a trade name of National Electrostatics Corporation
(Middleton, WI), is a variant of the Van de Graaff design that
uses chains of alternating metallic and dielectric links (pellets)
in place of continuous belts to transport the charge. 1b.is
scheme avoids the problem of patchy charging of the belt and
has substantially reduced the associated voltage ripple on the
terminal.
-.
be made at a manifold external to the pressure vessel.

'.....~~,;~,~
Tandem accelerators [Fig. 16.1(b)] use a clever trick to 0 multiply the energy of the beam. Negative iom are produced at .~~t
ground potential and accelerated to the positive terminal .:~",
located midway along the beam tube structure (Rose and '(Of ~
Galejs, 1967). Inside the terminal, they encounter either a thin '::,3",.,!. :,.
carbon foil (typically 2-5 ~glcm2) or a tenuous gas (usually N, ~
or H2 to minimize scattering). Because of their high velocity, .~
these ions are stripped of electrons through collisions in this
region and emerge from the terminal with positive charge. , \1,~.
Because the terminal polarity is also positive, they are again .:;~~'5:
accelerated. The total energy they acquire is (n + I)V, where n ":.~'.:~i
. al tiL
is the final charge state of the ions and V is the temun
.. n
voltage. Ordinarily, a range of final charge states is produced I
a given terminal voltage. To obtain a desired final beam energy, ','~
ODe must choose the terminal voltage and select a specifiC t~}~.
;:~
at tt
".
388
'
Instrumentation and laboratory Practice
analysis, where high energies and heavy-ion beams such as l~

or 19p might be required.
charge state for the fmal ion. Tandem accelerators can produce
ion beams that are far more energetic than those available from
single-ended accelerators with conventional ion sources,
because n can be much larger. As a result. tandem setups are
more appropriate for techniques sucb as nuclear reaction
Single Ended
2MV
Ground
potential
(a)
Source
Accelerator tube
Wafer
Terminal
Double Ended (Tandem)

1MV
Ground
potential
Ground
potential
Ie: ~r-i
++
r-T""O,.-r--,
Source
,.
--~-.....;.
--
--- .
Stripper
----.;.-_.
- - ....
- -/
(b)
Wafer
Terminal
FIG. 16.1. Schematic diagram of (a) single-ended and (b) tandem Van de Graaff electrostatic accelerators. Terminal voltages up to
approximately 20 MY have been achieved with tandem accelerators. Voltages of 1-3 MY are typical for machines designed for ion
beam analytical applications. Figure courtesy of National Electrostatics Corporation, :Middleton, WI.
not always the case. Magnets actually filter the ions by
momentum per unit charge or, equivalently, by the magnetic
rigidity
16.2.3 Beams
The ion beam that emerges from the accelerator typically
contains a range of energies and even a multiplicity of species.
These variations result from the imperfect vacuum of the beam
tube, poor quality of the beam injected into the accelerator. and
processes such as scattering and charge exchange that occur
during acce1eratioD. To extract a beam suitable for materials
analysis, it is necessary to reject all ions that do not have the
desired mass and energy. This is almost universally accomplished
by passing the beam into the field of a magnet, called an
analyzing magnet, whose field has been tuned to pennit only the
desired species to pass through its narrow entrance and exit slits.
Under most circumstances, this means that it is possible to select
a unique beam with the de:sired mass and energy. However, this is
"'
( q' J
Bp= 2mE
(16.1)
where B is the magnetic induction; m is the particle's mass; E is

its energy; q is its charge; and p is its radius of curvature, which is
established by the magnet's entrance and exit slits. As a result,
interferences can occur. Perhaps the most common of these is
encountered when trying to accelerate He 2" ions in a .single-ended
accelerator. In this case, the hydrogen molecular ion, Ht. with
389
Chapter 16
one-half the charge, one-half the energy, and one-half the mass of
2
the He + ion, cannot be distinguished by a magnet. It is sometimes
bard to detect this condition by looking at backscattering spectra
alone, as backscattered protons can be interpreted as ordinary
structwe in an otherwise unremarkable He spectrum.
The careful calibration of analyzing magnets is usually done

using resonant nuclear reactions if the accelerator's energy is
adequate (Marion and Young, 1968; Overley et 01., 1969). See
Appendix 19 for a listing of many of these reactions and
be drawn between the effect of ordinary glass lenses on light and

the effect of these carefully designed electromagnetic fieldS on
charged particles whose motions in phase space are constrained
by Liouville's theorem (Lawson, 1977). The tecbnology of
designing and placing electromagnetic lenses is called ion optics
and has been elevated to a high art by the designers of electron

microscopes. An introduction to the subject can be found in
Moore et al. (1989), and more detailed treatments are available in
Lawson (1977), England (1974), and Humphries (1986).
techniques for energy calibration. For lower-energy accelerators,

it is necessary to rely on a precision voltage divider or a
Most of the cbarged-particle lenses encountered by materials

scientists will be either magnetic or electrostatic quadrupole
combination of partial knowledge of the expected beam and Eq.

(16.1). One proceeds by sweeping the magnetic field, B, and
noting the values at which beams are found. The most intense
lenses or electrostatic gap lenses (Harting and Read, 1976). A

detailed discussion of the operation of these devices is beyond the
scope of this chapter. In general, however, both behave ill a
beams usually are the ones expected on the basis of, for example,
the composition of the source gas. By making some initial
assumptions and iteratively assigning masses to other beams, it is
possible to arrive at a consistent relationship between B and m at
a given energy. All of this is necessary because ion sources
ahnost universally produce many more beams than the uninitiated
operator would reasonably expect These arise from impurities in
the feed material and sometimes from the materials comprising
the source itself. In addition, molecules can be manufactured in
the source. For example, chlorine feed gas is especially effective
at generating volatile chlorides from solid constituents within thc
source. The process is so efficient, in fact, that it is used to
produce beams of some species that are otherwise very difficult to
vaporize.
The strength of the field of the analyzing magnet is measured

using either a Hall effect probe or a nuclear magnetic resonance
probe. Both are accurate enough to enable the operator to
generate an accurate calibration. In some smaller accelerators, the
magnetic field probe is omitted on account of its east. In this case,
one obtains an approximate measure of B by inference from the
current in the magnet to which B is nominally proportional.
However, using current in place of a direct measure of B can lead
to larger errors because of the hysteresis that is characteristic of
large electromagnets. We note finally one additional characteristic
of analyzing magnets that has swprised more than one
unsuspecting operator. Always slowly reduce the current in a
large electromagnet to zero before switching off the power
supply. To do otherwise is to court disaster, because of the
magnet's huge inductance. A high-capacity, large-area diode
placed (reverse-biased) across the power supply's terminals is
good insurance against the potentially catastrophic damage that
can be done to a supply that has not been specifically engineered
to drive large inductive loads.
The beam is transported to the target chamber through an

evacuated pipe. It is steered to its destination by a series of
electric and magnetic fields produced by devices called lenses.
This nomenclature is appropriate because a powerful analogy can
390
manner analogous to (thick) convex glass lenses and, so, are
capable of focusing a parallel beam or producing the image" of

an "object". Quadrupole and gap lenses differ for practical
purposcs mostly in the ranges of energies that they can focus.
Quadrupole lenses are generally stronger, in the usual sense, than

gap lenses, and magnetic lenses are required for high-energy
particles. For low-energy ion worle, a commercial vendor has
created a do-it-yourself kit for electrostatic optics that retains
much of the romance of the "Erector Set" that launched many
scientific careers.
16.2.4 Beam diagnostics
The manner in which one uses lenses to transport and shape
an ion beam follows quite closely thc analogous procedure for
visible-light optics. However, to perfonn this operation, one must
be able to see the ion beam's position and its cross-sectional
shape. This is done using a series of beam viewers that fluoresce
when the beam strikes them. Fused quartz, sapphire, and

magnesium oxide are among the most commonly used viewer
materials.
..
It is reconunended that these bighly insulating materials be
coated with a thin layer of graphite to provide a path for the

electrical charge to reach ground. Otherwise, electric potentials of
truJy astonishing proportions can be generated, along with
significant quantities of X-rays when sparks discharge the
surface. Colloidal graphite, available commercially in spray cans,
is also useful in eliminating patch potentials on electrostatic
lenses and as an ultrahigh-vacuum~ompatible solid lubricant
The quartz viewer emits a blue glow when struck by most ion .
beams of more ilian a few nanoampere.s and energies exceeding
about 100 keY. For lower currents, conventional phosphors can
be used. In many cases, mechanical sensors, such as wires
sweeping through ilie beam, are used to generate an "electrical"
view of the beam. By placing a series of viewers along the beam
line, it is possible to propagate an image of the aperture frorn
which the ion source emits the beam, all the way to the plane of
the target
For proper behavior. the trajectory of the beam should be

along the optical axes of the lenses. Accelerator beamlines have
electrostatic or electromagnetic deflectors (termed steerers) to
accomplish this. -When a beam is properly steered into a lens,
adjusting the lens alters the focus but not the average position of
the beam. For best results, an ion beam should be controlled
primarily with steerers and lenses. It is common practice to use
slits to define the cross section of an ion beam, but the practice
should be used with considerable discretion. The role of slits in
ion beam transport sbould be restricted to eliminating that portion
of the beam that is manifestly aberrant. (Slits are, of course, an
integral part of analyzing magnets and other dispersive elements.)
If used to defme the beam area, scattering from the slits can
reduce the quality of the delivered beam.' In any event, slits
should be very thin at their edges, but not thinner than the range
of the ion beam. The fmal size of the beam on the target can be
increased or decreased by the lens system, subject to the
restrictions imposed by Liouville's theorem. Conventional
accelerator beam spots, when well focused, are typically from 0.1
rom to 3.0 rom in diameter. To obtain beam spots smaller than
about 0.1 mm, it is necessary to use special optical elements and
ion sources. As an example, see Jamieson et a/. (1989). For a
much smaller beam spot size (microns or less), one must typically
use multiple electromagnetic quadrupole lenses in series mode.
Details on microbeam production are discussed in Chapter 13.
importance. Because the ion beams produced by accelerators vary

in intensity (often as a result of the action of the system used to
stabilize the energy) on a time scale that is very short compared
with the typical time for an experiment, it is necessruy to measure
the cbarge directly rather than to measure current and time
separately and multiply them after the fact. Instrwnents to make.
this measurement are called current integrators and are
manufactured by several suppliers of general nuclear
. instrumentation. They produce a train of output pulses, each of
which corresponds to a fixed amount of input charge. Current
integration is, thus, reduced to pulse counting. Alternatively, one
can-purchase a linear voltage-to-frequency converter (sometimes
called a voltage-controlled oscillator) from any of a number of
electronic component manufacturers, and with the addition of an
operational amplifier wired as a current-to-voltage transducer,
produce a simple current integrator quite inexpensively.
16.3 DETECTORS
16.3.1 Overview
When an ion beam strikes the surface of a specimen, it initiates

of processes that usually resuit in the emission of
radiation. The detection and measurement of this radiation is the
crucial step in ion beam analysis. An exhaustive treatment of the
subject can be found in Knoll (1989). Additional infolffiation on
devices such as magnetic spectrograpbs and other specialized
systems can be found in the book by England (1974) and in the
reviews contained in the special issue of Nuclear Instroments and
Methods edited by Bromley (1979).
a series
The quantity of beam that is obtained on the target is

measured by carefully measuring the electrical current' that flows
between the target and ground. From the point of view of circuit
theory, the particle accelerator is a constant-current source with
In this chapter, we ~strict our attention to three kinds of
one side grounded. The ions in the beam are the charge carriers,
and the circuit is completed when the ions are intercepted by a
devices: semiconductor detectors, scintillators, and electron
material object connected to ground. A simple ammeter can. be -~--multipliers: Gas-filled detectors: and neutron. detectors are
used to me~~e the curre~t. However, a number of physical . discussed in 'Knoll (1989) .. Historically. ~the development of
phenomena lDlhated by th~ IOn bombardm~nt can complIcate the
detectors with a potential for application in materials analysis has
me~~ment. The ~ost .unportant one. IS secondary electron
been driven mostly by the demands of nuclear physics. Reports of
emISSl~n. ~en the I~ns ill ~e beam strike a surface, electron~
recent advances in detector technology are found. for the most
are e1ll1tted With energIes rangmg from a few electronvolts to over
. th lit
f th t fi Id C
t 'w
ti
. tit
1 keY. This flow of electrons awa from the tar et constitutes a
part, lD. e erature ~
a Ie . urren 1 0:ma 00. on e
d
propertLes of commerCial detectors can be found m the literature
't
y th b
g
t
current 0 f th e same poIan y as
e eam curren an IS
indistinguishable from it by a current measurement alone. One
of the manufacturers.
might then suggest that, by biasing the target at a comfortable
163.2 Electron multipliers
level of, say, +2 kV, thesc electrons could be restrained and
'Electron multipliers are among the simplest yet most versatile
accuracy of the measurement restored. However. it would then be
of
particle detectors. A schematic is shown in Fig. 16.2. An
noticed that every stray electron in the vacuum chamber streamed
electron multiplier consists of a cathode that is sensitive to
to the target. Thus arises the subject of Faraday cups. These are
radiation and a set of electrodes, called dynodes, that are biased at
electrodes specially designed to intercept ion beainS with minimal
progressively more positive voltages. The last electrode is called
interference from secondary processes. The general theory of
the anode, as in all vacuum tubes. When radiation in the form of
Faraday cups is given in the book by England (1974). Some of the
photons, electrons, or other particles strikes the cathode. electrons
problems with their use in materials work are discussed in
are emitted. These electrons are accelerated toward the first
Chapter 15.
dynode, where they strike and liberate additional electrons. In
Because most ion beam analytical work depends on the total
commercial electron multipliers. the process is repeated as many
number of ions striking a target, rather than their rate of arrival,
as a dozen times, giving gains up to 109 for single events at the
charge rather than current is usually the quantity of most
cathode. Pulses from electron multipliers are typically very short,
391
Chapter 16
with 5-10 os being common. Rise times in the range of 100--200

ps are obtained with the best devices designed for timing
16.3.3 Scintillators
applications.
by radiation. An example that has already been mentioned

fused quartz. However, there are much better materials, and
Electron multipliers are used in two distinct applications in

materials analysis. In one, they are called photomultipliers and are
used in combination with a scintillator to measure "(- and X-ray
radiation. In the other, they are used directly to detect photons or
Some materials emit brief pulses of light when they are Sln,_' ",1)':;
form the basis for the detection of 1-rays and X-rays and,
frequently, electrons and particles. The qualities that distin:gui,;h a'
good scintillator are a large absorption cross section
incident radiation, efficient generation of scintillation
and a long mean free path for these photons to be reabsorbed.
most common scintillators used in materials work are IIlO'l:ani(
crystals such as sodium iodide activated with thallium,
particles in, for example, time-or-flight measurements of particle

velocity.
The primary characteristic that distinguishes the device that is
chosen for each application is the composition of the cathode. For
measurements of light, the cathode is covered with a material that
emits electrons when irradiated and is called a photocathode. For
NaI(l1), and bismuth germanate,
B~Ge,012'
which is
referred to as BGO. Large varieties of plastic and

scintillators are also available, but they are used less frequently
materials work. One exception that should be noted, however, .
in timing or coincidence s.tudies, where certain plastics
particle detection, high sensitivity to light would yield .an

unnecessarily high level of background counts, and so, cathode
materials with relatively high work functions are chosen. The best
source of information on photomultipliers is the literature of the
manufacturers. Knoll (1989) provides a summary table of data on
a number of commonly encountered photomultiplier tubes, but in
some cases, bigher-perfonnance replacements are now to be
preferred.
distinguished by their extremely fast response times,

properties of some common scintillators are included in
16.1 and 16.2.
The physical principles by which scinlillators operate:
relatively well known. Energetic photons interact with
matter through the photoelectric effect, Compton scattering,
and pair production (Fig. 16.3). Materials containing
elements such as iodine and bismuth are, thus,
scintillators because of their large photoelectric cross
The electrons produced by these processes lose energy
further collisions until their energy is fully dissipated
scintillators such as sodium iodide, some of the energy
For work inside a vacuum system, a specialized form of

electron multiplier is now in widespread use. It is termed a
continuous dynode or channel electron multiplier and is often
generically called a cbanneltron after the trad,e name used by the
original manufacturer, Galileo Electro-Optics Corp. These
devices do not have discrete dynodes, but rather have cUIVed
internal surfaces of a resistive material (a reduced Pb glass) that
both grades the potential and acts as a source of stored charge.
Pulses from channel electron multipliers are extremely fast,
giving these devices better characteristics than any other detector
absorbed by intentionally created defects (thallium impurities .

the case ofNal) called activators whose energy states lie
the bandgap of the host Cl)'stal. When these excited
decay. they do so by the emission of photons whose
are lower than the bandgap of the crystal. These photons,
are usually in the visible region of the spectrum, are
travel significant distances in the crystal with
likelihood of reabsorption. Photomultiplier tubes
crystal collect the scintillation photons and generate
for timing applications such as time-of-flight spectrometry. They

. are also manufactured in the fonn of flat plates with, a high
density of microchannels, each of which is an active multiplier
(Wiza, 1979).
Originally developed as light amplifiers for use in night-vision
devices, microchannel plates provide large-sensitive-area
pulses from them
detection and extremely fast response times. They are widely used
Ideally, the number of scintillation photons is proportional
as the active detecting elements for particle timing applications.

Recently, they have also been combined with photocathodes in
conventional vacuum tubes, giving photomultiplier tubes with
remarkable characteristics. The Hamamatsu microchannel-plate,
photomultiplier tube, R2566U, has a rise time of 100 ps, a fall
time of 120 ps. and a total pulse width of 200 ps when operated
under standard conditions. Its frequency response is 3 GHz.
the energy of the initial radiation. However, a number

processes intervene to complicate this simple
Consider the case of an energetic gamma ray produced. .
example, in the nuclear reaction lH('~,(1"y)12C, which'
sometimes used for the measurement of smace
Suppose that the y-ray interacts with the detector through
production of an electron-positron pair and that each of these
392
Table 16.1. Properties of inorganic scintillators. Sources: Radiation Detectors Catalog, Solon Technologies, Inc.; Sakai (\988), Holl ef
al. (1988), Knoll (1989).
.
Scintillator
Den~
Ig/e )
Principal
decay
time
constant
Ips)
Pulse
rise
Hm.
1()'"
9.%
Ips)
Ught
output
(photons!
Absolute
scintillatlon
efficiency
(nm)
Index of
refractlon at
'mn
MeV)
I")
Hygroscopic?
0.23
0.5
38000
11.3
Yes
Wavelength or
Ughl
output
relative 10
Nal(TI)
emission
I")
m~.
l",~
Na(TI)
3.67
100
415
1.85
CsI(Tl)
4.51
45
550
1.80
1.0
52000
11.9
Slightly
Csl(Na)
4.51
.5
420
1.84
0.63
39000
11.4
Yes
'.8
8200
2.1
No
BI.Ge 30'2
7.13
CdWO"
7.9
CaF~Eu)
BaF 2
Fast
3.18
4.88
Slow
12-14
...0
2.15
0.'
40
540
2.2-2.3
5.0
50
435
1.44
4
20
225
310
325 nm
1.49@
0.94
0.0006
0.62
CsF
4.54
390
1.48
589 nm
0.005
UI(Eu)
4.08
35
470
1.96
1.4 n
radiation
0.6y
radiation
15000
3.8
No
24000
6.7
No
10000
4.S
No
Yes
11000
2.'
Yes
particles comes to rest within the crystal, depositing all of its

16.3.4 Semiconductor detectors
Ordinarily, a reverse-biased semiconductor diode does not
kinetic energy. At this stage, all of the energy except that
required to creatc the particles' rest masses has been accounted
conduct electricity because it contains a region from which
for. Immediately upon being stopped, the positron will annihilate
charge carriers have been depleted. However, when electronwith an electron from the crystal, generating two y-rays, each
hole palrs are formed randomly in the depletion region of such a
with an energy of 511 keY. There are now.three possibilities: .. diode~ following. the usual laws' of thermodynainics, the boles'-'Either both of these "(-rays are absorbed by the crystal, or one or" and electrons are swept in opposite directions by the established
both escape. It turns out that., for practical detectors, each of
electric field and produce the small, temperature-dependent
these possibilities is reasonably likely. Therefore, one can obb3.in
leakage current that is a characteristic of the device. It is well
three different depOSited energies and three different
known that light can also cause the generation of electron-hole
photomultiplier pulse amplitude groups from the absorption of a
pairs by directly promoting electrons from the valence to the
flux of monoenergetic y-rays. Other physical processes lead to
conduction band of the material. This is the principle of
additional idiosyncrasies in gamma-ray spectra (Fig. 16.4). For a
operation of the photodiode. Semiconductor detectors for
particles and high-energy photons, most of which are fabricated
complete discussion of these issues, see Knoll (1989) and Heath
(1964).
from Si or Ge (Table 16.3), work in exactly the same way. See
McKenzie (1979).
393
Chapter 16
Table 16.2. Physical constants of p1astic and liquid scintillator'S. Also included are glass scintillators, and for comparison, anthracene
and Nal(TI). Source: Table of Physical Constants of Scintillators, NE Technology, Inc., 9 Deer Park Drive, Monmouth, NJ 08852.
Used by permission.
MoWng
soflenlng
Aefrac or boiling
live
Scintillator
Plastic
NE lOlA
Plastic
P&aslic
Plastic
D,nslty
index
1.032
1.581
1.58
1.58
1.58
1.58
1.582
1.58lI
1.594
1.581
1.58
1,581
PiIoIU
PHot425
P~sUc
1.032
1.032
1.032
1.032
1.032
1.032
1.020
1.14
1.032
1.032
Plastic
1.19
1.49
NE:213
Uqu~
0.874
NE224
NE226
o.an
1508
1505
1.38
150
1.43
NE236
Uquid
Uquid
Deutemted liqllid
Oeulerated liqllid
Uquid
Uquk!
NE237
Uquk!
0.813
NE104
NE 108
NE 110
NE111A
NE114
NEIlS
NE 118
NE 125
PiIotF
Liquid
Type
NE230
NE 232
NE135
Plastic
Plastic
Plastic
Plastic
Plastic
Plastic
Plastic
1.61
0.945
0.89
0.858
0.796
1.47
1.444
1.459
point
re)
WaveHIt
length 01 Contenl (no. H
output main com- maximum of loading atoms!
emission element no.ofC
(%an ponen!
fhratene)
(ns)
(nm)
(% by wt) atoms}
Decay
Ugllt
75'
75'
75'
75'
75'
75'
75'
65
66
65
60
55
50
35
60.
55
65
99'
SO'
_ 75'.
75'
100'
or
78
80
20
141"
169'
SO'
81'
81'
350'
192'
192'
80
80
40
48
61
constant,
2.4
1."
15
3.3
to
4.0
320
3.3
2.7
21
1.36
3.7
26
3.3
3.0
4
4
3.5
3.15
423
400
545
434
370
434
395
434
425
425
391
425
.25
.25
430
425
43<l
013.8%
1.104
1.100
1.103
1.104
1.103
1.109
1.108
T." ~ 1ast n
u1!ra-fast counting
with pholodiodes
1. a. p, last n, etc.
utJatast timing
asforNE110
pIiOswidIes
1.104
high temperatures
n
T, a, p, last n
ultra fast timing . ~.
Cheren~ov detector
0.957 O/C
1.103
1.100
1.6
014.:2%
024.5%
420
425
425
Principal
applications
1.213
fasl n (1'.S.o.)
1330
1. fast n
1. Insensitive 10 n
0
0.984 O/C Spocial applicalions
1.960/c Special applicalioM
large tanks
2.0
1.92
low temperatures
high flash porn.
1.92
general
Loade<t
"lui<!
Neutron
NE311 &311A
NE314A
NE316
NE320
NE343
6 badad IIqu~
Pb loaded fiqllid
Sn loaded 6qui<!
'u loaded fiquid
GIl loaded fiquld
NE '22& 426
'L>ZnS(Ag)
ZnS(Ag) plastic
. (ZnS-type) NE 451
nnd
glass
C')"IaI
NE 901. oo~ 903 G~ss

NE 904, 905, 906 Glass
NE 907, 908
Glass
NE 91~ 913
Gbss
Anthracene
N.I(!1) .
Crystal
Crystal
0.91
1.411
65'
0.96
1.53
1.496
1.497
1.502
141"
O~
0.906
0.884
2.36
1.443
2.64
2.5
2.42
65
2'
35
148.5'
160'
168"
34
65
3.8
2.0
4.0
2.19
3.0
300
200
110
200
300
c.1200 2B 20 8 60
c.l200" 25 20858
110'
~'2
1.58
1.55
1.566
1.55
1.25
3.67
162
1775
c.1200
20
c.l200
25
217'
100
230
6SO'
18862
18855
30
230
425
425
425
425
425
450
450
395
395
399
397
447
413
65%
Pb7.5%
Snlo%
'uO.I5%
GIl 0.5%
1.701
1.261
1.411
1.428
1.360
n,p
T.X-reys
l.x.;ays
n
sIJw n
US'll.
last n
n. p
U2.3%
U6.6%
U7.S'/.
U7.7%
" P(low background)

0.715
l,a,
.-
last n
:,;
i
1, x-rays
i,
-1
'i
.~,
,~
"
;~
\
',:i{
>g
;~
,j
394
11
.....
hotocathode
focus
120
..,"
0
~
..,.."
Dynodes
100
Photoelectric
Pair production
dominant
effect
dominant
80
60
40
Compton effect
dominant
20
0
0.01
0.5 1
0.05 0.1
5 10
50100
r Energy (MeV)
FIG. '16.3. Relative importance of the photoelectric effect,
Compton effect, and pair production as a function of the )'-ray
energy and atomic number of the absorber. (From Evans, 1955.
Copyright McGraw-Hill; used with permission.)
0,,-
SI~j.HV_1-i'VIM Anode
~
Anode Signal
FIG. 16.2. A typical bias arrangement for a photomultiplier tube

showing the high volla$e (H.V) input, the resistive divider chain
for the d)oTIodes and Signal outputs at both the last dynode and
the anode.
Sowce
PO
~hleldin9
U.V. PhotonS-J"Ii;,,:h....~""1I-produced from local
e)lcltea states
follOwIng IoniZatIon
PhotoefeC1ron
Dynode ____--~C1"
(secondary
emlttedfrom
cathode
elec:tron
emission)
Anode
FIG. 16.4. Physics processes.affecting a y-ray spectrWn. (From Heath, 1964.)
395
Chapter 16
Table 16.3. Properties of pure silicon and germanium.
Sources: England (1974), Knoll (1989), Elmsley (1989).

Go
"
Alonie mass (1.1)
28""5
Stab!e 1soIq>es (u) (%)
23 - 92.13
29 - 4.67
3O-S.n
loss and energy straggling in the entrance layer are further

70-20.5
n-27.4
73-7.8
74-36..5
76-7.8
"-f9'm'I
.",
5.323
Alomic density (l0Z2Jcmll)
.96
4.41
. _ Lilearthermal expansion _
coeffide1ll (/('"1)
"""".--
laltIce IX'fIStanI 0 300 K(nm)
"""""''''''''''
...... ""(K)
1683
mO.6
""""""""""
12
18
1.115
1.165
0.665
0.746
Intrinsic rezisIMly' 300 K (Qen)
2.a<105
47
EIedron mobiBty (rdN-!)

>lIlK
1350
"'"
<SO
1.1xlo'
1900
8.2xlo'i
1()1
5.9xlot'
9.SxloG
3..62
3.76
HoI appecable
2.96
0.5430
0.5658
Forbidden tnergy gap (IV)
""lK
OK
Intrinsi: carrier density 0

300K (arr!)
..,
nK
2.1xlO"
The most common particle detector used for ion beam

materials ana1ysis is the Si surface-barrier detector, where a thin
gold layer is used as an electrode. Newer detectors use ion_.
implanted silicon layer to replace the Au electrode, so the energy "
reduced. For light ions such as protons or helium particles, the
quantity of collected charge is quite linear (about 3.6-3.7 eV per
electron-hole pair) in the total energy deposited in electronic
processes. The overall resolution is typically in the range of IllS keY. For heavier ions, the energy lost to nuclear collisions
gives rise to a nonlinearity known as the pulse-height defect,
which can become qwte large. This phenomenon also worsen .
the energy resolution of the detector. Some energy is also lost as .
the ion passes through the Au entrance window and the dead .~
layer beneath it. For dense ionization, recombination can also'
occur before the electrons and holes are fully separated, but this .
effect is usually quite small. For very precise wOIk with
barrier detectors, it is necessary to consider carefully the pbysics
of their operation in order to understand the circumstances under
which their linearity breaks down (Lennard el aI., 1986).
Caution should also be exercised when detecting heavy ions (A
> 4) because of their larger pulse-height defect and because of
the irreversible radiation damage that such ions cause to the
detector.
When choosing a surface-barrier type of silicon detector, one
can refer to vendors' detector catalogs, which usually provide
advice on, the style and operating characteristics to choose
specific applications such as high-resolution spectroscopy, fast'
timing, or rugged service. Be sure to specify a detector with a
depletion region thick enough to stop the highest-energy particle
that you expect to encounter in your experiments. Otherwise,
3.6xW
. lice mctiBty (enlN.,)
nK
Carrier satl/raIion velocity (etMJ

300 K
nx
Energy pel eleclron-hole ~ (eI\1

-. 31)0 K .
nK
~.2xl04
energy of these particles will not be measurable. TTb;le:I~:~:~r.;:
shown in Fig. 16.5 can be used to detennine the d

depth in 8 Si surface-barrier detector that is required to stop
particles, protons, and_ electrons of various energies and to

When an energetic ion enters the depletion region of a
semiconductor detector, it begins to lose energy through
collisions with the electrons and nuclei of thc constituent atoms.
Energy deposited in collisions with nuclei generates phonons
(beat) and is effectively lost, but energy used to excite electrons,
the so-called electronic component of stopping power, is
available to produce cbarge carriers at a rate of a few
electronvolts per pair. The ion comes to rest in the detedor
(provided, of course, that the detector is sufficiently thick) in a
matter of a few picos.econds and leaves in its wake a large
number of electron-hole pairs. As with their thermally gcncmted
counterparts, these charge carriers are swept from the depletion
region and appear as a pulse of current at the detector's
terminals. The length of this pulse is typically a few nanosecond.
It is collected by a charge-sensitive preamplifier for further
processing.
estimate (as a function of the resistivity of the Si) the detector

bias required to obtain this depletion depth.
Surface-barrier detectors are used primarily
electrons and ions. For detecting x- and l'-radiation,
drifted detectors and high-purity or intrinsic gennanium are
solid-state detectors of choice. The principal advantage
these solid-state photon detectors have over
detectors is high resolution. Some are conspicuously less easy to
me, however, because they must be cooled to liquid nitrogen
temperatures, They are also more costly. The detector in this.
class that is most commonly used for materials work is the .
lithium-drifted silicon X-ray detector, Si(Li), which is.
pronounced "silly". They are frequently found on scanning,
electron microscopes where they arc used for energy dispersive. "
X-ray analysis of the elemental constituents of~e sample. The
396
Ge(Li), pronounced 'Jelly", require continuous refrigeration by

liquid nitrogen (77 K) to avoid destruction by lithium
precipitation and have been made obsolete as the technology for'
producing high-purity germanium (HPGe) bas been perfected.
[Si(Li) detectors are also subject to some deterioration when
warmed to room temperature and should be continuously
maintained at 77 K for the best results and longest life.] HPGe
has a good energy resolution (-1.8 keY for 661 keV'Y-rays) and
requires a liquid nitrogen cooling only when it is in use.
Si(Li) detector is also an essential choice in particle-induced

X-ray emission (PIXE) analysis. Si(Li) detectors operate on the
same physical principles as other solid-state detectors and
in~er~t with ele~tromagnetic radiation in the same way as
scmhllators descnbed above. Figure 16.6 shows transmission
efficiency curves of a typical Si(Li) detector with different
windows and the energy resolution as a function of X-ray
energy. For r-roy detection, an intrinsic gennanium detector is
the preferred solution. Lithium-drifted germanium detectors,
SpeclflC
Specific
...... J_na ..
(ohm-c:m)
n"""
so
.....,.,-,,
Depfc1J"n
cap_llan_
{pFlmm~
~"'~
s.
..
,0-- -.
,0'
~.
". " u
,. ,..
'"
.,.
so "
. ..
..
~.o'
,. ,"
_2.~
._""
...
,,.
:- 10
"
-. .
.,
I-
.1!;!
--- 10-''''
....
m
,. ..,.
,,~
1-""
I:
--
=-'0'
.,
f- ",:
,.,. ,
---
(M"Y)
0.1 ~
B~
,0'
......"
P .uclo-encrgy
lor 1'8t\9f' ot
depletion dopth
.-~
.. ---
"
b"
'~
"
..
f- :C " .
f- - ,
..
"-'
.00 ,
...
.~
"-'
.05
.M
"
-'"- - ~-----
... -
- -
--
...
.~
1.0
~.
1 ~ or 5J. CI..232.5 mglcom~

1 mglem l of Sl 4.3 ~
A .. trDfgtd edge 1~"'9
t_......- -rtf""III"a lit !he ~ulrod deplelJon depth ......11
gJ_ comblnatJons of n=lat!vfty and deto<:tor bills that may be used to achieve that
dqth. (Shown.
t,,' .. ~ "" thO votrage ttI.!lt ........ be appl_ l<> .. 13,COO "hm-c'"
4500 oto .......... ......typo _mean .... ~ to atop" 23-6k0Y ..Ipha, II &-MaV
proton, Of'1I :e$O-k"y electrOn wfthln I:hCI depletion depth..)
p-typo _
FIG. ~6.5. S~co~ surface-barrier nomogram, showing the relationship between bias, depletion depth, resistivity, and specific
capacItaDc.e. Sunllar to the nomogram reported by Blankenship (\960). From the Ortee Catalog, copyright EG&G Ortec. Used
Wlth penmsslon
397
Chapter 16
100r--------------=~~,---~~------,
~
0.4
(a) C
Polymer
~ 80
c:
3mmThick
'u
'"
IE 60
5mmThick
UJ
c:
--'"
'"
.Q
<>
40
8pm
Cl 20
>-
l'---.13
01
~
'c:"
UJ
'------ 25
~m
~m
I I I I III
:;
u.
"1
0.1
10
100
Energy (keV)
10,000
(b)
Jt
1,000
:a
J:
:;:.
lL
> ~ ~;~
Ie'
Q)
100
:""~
1=80
,,0 '.b.J
~ ~'"
J;'
1/
FWHM (eV) :1(FWHM)'NOISE +
[2.351ffi"(
F=O.1
L...-
E Photon """"" (eV)

e = 3-9 t;N (m:ernge energy 10 create
V :Oe
10
0.1
..
1'0
keY
<0
~""Iedron pair at
HiD
1,000
FIG. 16.6. (a) Transmission efficiency curves of a typical Si(Li) detector for various types and thicknesses of windows. The
polymer window curve does not show the effect of the support grid on the overall efficiency. (b) SiCLi) energy resolution,
including effects of intrinsic and electronic noise, as a function of X~my energy. From the Canberra Catalog, copyright Canberra.
Used with permission.
398
163.5 Detector systems

Two complex detector systems are becoming increasingly
important for ion beam analytical work.. These are particle
telescopes and time-of-flight spectrometers. A particle
telescope is a composite detector that simultaneously measures
the energy lost by a particle as it first passes through a thin Si
detector (or a gas detector) and then stops in a conventional
solid-state detector (England, 1974). This configuration
returns two parameters for each event. The energy lost in the
thin detector is approximately proportional to the stopping
power of the ion, and the sum of the energies recorded in the
two detectors is the ion's total energy. By extending the
conventional concept of an energy spectrum to two.
dimensions and plotting the number of counts against both the
energy and the stopping power, it is possible to separate events
generated by ions of different species. The time-of-flight
spectrometer, one variation of which is shown in Fig. 16.7, is
a.1so a composite device consisting of two discrete particle
detectors. Its function is to measure the time required for a

single particle to pass between two points. Such systems had
their genesis in detectors developed during the 1970s for
heavy-ion nuclear spectroscopy (Gabor et al., 1975j Betts,
1979). Their application to ion beam analysis was reviewed in
1990 (Whitlow, 1990) and continues to be refined. In all of the
designs, the energetic particle first passes through a earnon foil
about 1-5 pglcm2 thick that is either self-supporting or
mounted on a high-transmission mesh. Electrons emitted from
the carbon foil are collected by a channel electron multiplier or
a microchannel plate (MCP), and a "start" pulse is generated.
After drifting a fIXed distance, the particle is again examined
by another carbon foil, by a semiconductor detector, or by
direct impact on the cathode of a channel electron multiplier.
This generates a "stop" pulse, and the interval between the
start and stop pulses is measured electronically. To understand
the details of this process, consider an example.
Pulse-height t - - - - - - I Time-to-amplitude
converter
analyzer
+1600 V
CFD
CFD
Preamps
.." ,,
~ro.:~
.:(0 .-
c;,~ ~""
~,e.
..
,,'.., \ ' ...",'"
""""
\
. . Ions ~.:c:' - - - _ - - _ - - _ - - - - - - - - \>~.

2 IJgJcm2 carbon foil
->
Q:::J=I--;S~top
pulse
-800 V bias
FIG. 16.7. A time-of-flight detector system. Note: CFD stands for constant-fraction discriminator, and MFP stands for microchannel
plate.
comparable to the 20 keY being obtained using a silicon surfacebarrier detector. Assume that the tota1length between the carbon
Let us suppose that a time-of-flight system is being designed

to give an energy resolution for a 2 MeV a-particle that is
399
L
..
Chapter 16
shoWn to be of particular value when used to measure

backscattering of low-energy ions (Mendenhall and Weller,
1991) or of heavy ions at high energies (Dobeli et al., 1991, and
references therein). Time-of-flight spectrometers are superior to
surface-barrier detectors in low-energy work because they offer
superior energy resolution. For high-energy heavy ions, they are
substantially more robust than solid-state detectors, whose
lifetimes can be extremely short (see Chapter 4). Time-of-lligbt
spectrometry is also useful for (folVlard) elastic recoil detection,
as a time-dispersive detector is fundamentally more appropriate
than a single energy-dispersive detector in this application (see
Chapter 5). A mathematical description of the effect of random
coincidences in a time-of-flight spectrometer on observed
s.pectra is given in Section 16.6.4.
foil and stop detector is 1 m. The flight tirrie required for the a
particle to cover this distance at an energy of 2 Me V is about
100 ns. An energy resolution of 20 keY is equivalent to 1%, so
that a timing resolution of 0.5% is required (E - r). This means
that a timing resolution of 500 ps is needed Consider first the
implications afthis resolution for the start and stop pulses. These
pulses are produced by very fast electron multipliers, and
although there is no directly useful information in their
amplitudes, these amplitudes do vary. The pulses might also
need to be amplified by a voltage~sensitive preamplifier. As a
result, they can have durations of up to several nanoseconds and
rise times approaching 1 os. These relatively large values

complicate the electronic task of consistently deciding just when
...- the antecedent pulse occurred.
The problem of generating consistent time markers is solved
with a device known as a constant-fraction discriminator (CFD).
It analyzes the analog pulse from the detector or preamplifier
and emits a standard fast negative logic pulse when the
amplitude has reached a specific fraction of its maximum value.
It does this by splitting the input signal into two components,
attenuating one component, inverting and delaying the other
(typically with a few centimeters of cable), and then
recombining the two into one fast bipolar pulse. The zerocrossing of this bipolar pulse occurs at a time that is. independent
of the amplitude of the input pulse and is used as the trigger
event for the generation of the fast-negative-Iogic time marker.
Because logic pulses have a consistent amplitude and shape, the
.. time of the original analog pulse can be established with the
required precision even though the amplitude of the original
pulse might vary by an order of magnitude.
16.4 ELECTRONICS
16.4.1 Introduction
The vast majority of all modem scientific measurements
follow the same general form. The physical quantity of interest
is related to an electrical quantity through a transducer, the
electrical quantity is processed using the standard techniques of
analog electronics; the analog signal is digitized with an
appropriate analog-to-digital converter; and the resulting values
are stored digitally by a computer, possibly after additional
digital processing. The procedures used for ion beam materials
analysis follow this scenario quife closely.
The transducers used in ion beam work are the various
detectors that were discussed in the previous section. The
pwpose of this section is to introduce the remainder of the chain.
All of the significant signal-processing and analysis tools
necessary for ion beam studies of materials are available from
commercial vendors. However, this does not imply that it is
possible to treat these t~ls as "black boxes". Quite the contrary
is true. 'If one is to obtain the best results in routine
measuremeots and have the flexibility to push into new areas, at
least the functioning of standard electronic modules must be
understood
-
Consider now the method by which the interval is timed. If

500-ps timing precision is required, then to time the interval
. directly, a clock ticking with this period is needed. This is
equivalent to 2 GHz, and counting this frequency directly is a
lather formidable task. More commonly, another method is used,
which has the added virtue that a multichannel analyzer can then
be used for time-of-llight spectrometry in the same way that it is
used with the detectors discussed above for energy spectrometry.
The start and stop pulses are connected to a device called a timet<ramplitude converter. A time-ta-amplitude convertor is a
specialized, single-shot pulse generator whose output is
proportional to the time interval between its start and stop
inputs. The TAC converts the problem of accwately measuring a
very short time interval to the problem of accurately measuring a
pulse amplitude. Because pulse-height analysis is a welldeveloped art. the time-of-flight problem is solved. Timing
resolutions of about 150 ps can be obtained with present
technology. This is more than adequate for most ion beam
analytical applications.
In addition, it is important to have enough understanding of

elementary analog electronics to be able to recognize the need
for simple devices and to build them. Perhaps the best reference
to guide the reader in achieving such skills is the excellent ~d
aptly named text of The Art of Electronics (Horowitz and Hi~
1989). The approximate minimum level of proficiency 1D
electronics that is desirable for materials scientists in ion beam
work is the ability to build simple active circuits with
operational amplifiers and passive components. SOUle
knowledge of discrete transistor circuits is also useful, but has
become less important. as the power and variety of integra~
analog electronic components has increased. Horowitz and J:lill
provide a number of tables that are very useful for selectlDg
proper components for simple electronic projects around the
laboratory.
Although they are less convenient and much more expensive

than surface-barrier detectors, time-of-flight systems have been
400
manufacturers' catalogs, lists a number of standard modules,

along with their inputs, outputs, and functions. Signal and power
inputs to these modules and their outputs have been dearly
defined. NIM modules can occupy one or more slots in any
position of a NTh-I bin. Bins can stand alone or be mounted in
19-in. "relay" racks. NIM bins contain dc power supplies (12
Y, 24 Y, and now often 6 V) for modules that can operate
within the standard current limitations established for the
number of slots that they occupy, as well as ac power for such
items as detector bias supplies. The modules connect to the bin
through a standard connector on the rear of the module. The pin
assignments are shown in Fig. 16.8 for the NIM: bin connector
and in Table 16.5 for the preamplifier power connector.
16.4.2 Nuclear electronics

The best places to learn about the fundamentaL elements of
nuclear electronics are the tutorial sections of fue manufacturers'
catalogs and their application notes (Orlee, 1984; Canberra,
1978; Duggan, 1988). The text by Nicholson (1974) is
comprehensive but more advanced. In this chapter, we discuss
only the simplest energy spectroscopy system for use with
surface-barrier or photon detectors. Our principle aim is to
introduce the nomenclature of the field and to place it in context
orion beam analysis instrumentation.
Commercial nuclear instrumentation was standardized by the
U.s. Atomic Energy Commission (AEC), the forerunner of
today's Department of Energy (USDOE, 1990). The standard is
basically for analog electronics, although there are also
definitions for logic signals. The basic ccmcept is of the nuclear
instrument module (NIM), from which the standard gets its
narne. The original NlM standard was drafted by the AEC's
Committee on Nuclear Instrument Modules. By the time of the
CA1VlAC standardization, the name National Instrumentation
Methods Committee appeared on documents. Now, because the
NIM standard is international, it shoLlld simply be considered to
be a name. "CAMAC" was originally conceived by the Europeans
as a name, not an acronym. but by popular demand, in the United
States it was embellished to "Computer Automated
Measurement and Control". Table 16.4, adapted from
NIM analog pulses can be of either polarity. The preferred

range for positive signals is 0-10 V. Fast analog signals are
preferred to be negative and in the range from 0 V to -1 V or
from 0 V to -5 V. There are two logic families: medium-speed
positive logic and fast negative logic. These logic pulses have
rigorous specifications that are given in Table 16.6. NIM
positive logic has states that are defIned by voltage and is
typically useful for rates from dc up to about 20 MHz. Fast
negative logic has states that are defined by current. With rise
times of - 2 ns, fast negative logic is capable of handling rates
up to and beyond 100 MHz. Fast timing applications use fast
negative logic exclusively.
.. .,..'
-"'~-----"-=':"
401
Chapter 16
Table 16.4. Swnmary of some commonly available NIM' modules, their typical inputs, outputs, and functions.
KIM module
Preamplifier
Input
Output
Camments
Unear pulse from a

detector
amplified pulse
Spectroscopy
amplifier
Linear tall pulse
Shaped linear pulse
Final amplification and shaping for analog-ta-digital

conversion.
Biased amplifier
Shaped linear pulse
Shaped linear pulse
Output pulse amplitude proportional to input amplitude less

bias level. Used to expand a region of a spectrum.
Timing amplifier
Fast linear pulse
Amplified fast Unear pulse
Broadband voltage amplification for timing applications
Delay-line amplifier
Unear tall pulse
Delayline shaped linear

pulse
A speClroscopy amplifier with delay-line pulse shaping.

Used for high-count-rate spectroscopy.
Summing amplifier
Shaped linear pulses

and often positive
logic pulses
Sum of pulses In two or
Sums linear or logic signals. Useful for mtxing signals.
Unear lail pulse or voltage
Initial amplification and shaping for spectroscopy or timing.

'Preamps mount near the detector, not in a NIM bin.
more channels
Delay or
delay amplifier
Linear or logic pulses
Delayed linear or logic

pulses of approximately Ihe
same amplitude and shape
Introduces an adjustable delay up to a few 115. Short delays

are acl1ieved by measured lengths of cable. Longer delays
use delay lines with buffer ampfifiers.
Single channel
ana!y2er
Shaped linear pulse
NIM positive logic pulse
Generates an output pulse only if the input pulse amplituda lies between adjustable upper and lower thresholds.
Constant fraction
discriminator
Fast linear pulse
NtM fast negative logic

pulse
Generates the output pulse a fixed period after the input

pulse reaches a specific fraction of its maximum ar'plitude.
Used for highesl resolution timing applications.
Unear gate
Linear pulse at the -[npur. Unear pulse of the same

NIM positive logic pulse
amplitude and shape as the
at the -gat9
input
Linear pulse is passed only" the gate Is enabled.
Gate and delay

generator
NIM positive or fast

negative logic pulse
Logic pulses with adjustable amplitude, width and

delay, often with fast negative logic marker pulses
A handy tool for working with logic signals. Can be used

for conversion between HIM positive and tast negative
logiC or between NIM and other logic families.
Coincidence J
anticoincidence unit
Two or more NIM
NIt.4 positive logic pulse
Generates an output pulse if input pulses occur

coincidentally .in user-designated logic states.
Scaler. counter
(timer)
NIM positive or fast negative logic pulses 8J1d

shaped linear pulses with
amplitudes typically
posilive logic pulses
Aatemeter
Pulser
Current integrator
Visual
or computer output
of accumulated number
of pulses
Counts pulses. Commercial units often contain intema!

time bases for slow timing applications and deliver logic
output pulses In proportion (e.g., 0.1) to the number of
input pulses.
100 mV
NIM positive or fast negative logic pulses or linear

pulses with amplitudes
typically .c: 100 mV
Visual indication of pulse

rale. Analog vottage or
CUrTent proportional to
input pulse rate
None required but analog pulse amplitude input

and logic trigger are
optional on some units
Shaped linear pUlse, tail

pulse, NIM poSitiVe or fast
Ion beam or other current.
NIM positive logic pulses at a

rate proportional to charge
accumulation
negative logic pulses
402
Gives an analog readout of pulse rate. Units typically

generate a fixed Charge pulse intemaJly for each input
pulse edge then filter and deliver the resulting current
(or a proportional voltage) as the output.
Pulsers generate test pulses for tests and calibration.
SOme even generate randomly occurrlng pulses (by
triggering on an internal noise source).
Typically used to integrate accelerator beam. An analc;>9
output proportional to input current is usually also available.
,
,
Bin
tonn~or
0 1001
034
Top
28
,,0 HOd
dc5 O
o,,0"Od d
dOd
18
DgO
T""
16
-6volts
Reserved
Re .... rved
.12vo!t5
-12 .. oIls
OS
"""
"
"
"
,.
..
4bcS~o
32
OS
"",.
Ground
guide
'""'"
GGS
..
"
42
G
28 34
'oga
'oga
19
..
.Pb,O'b'b
()"
0" t)()0
bQOb O
"
+6 volts
22
Reserved
0 0'
0"0 "0
Rear vI_
0
b10
Spare
.".~
Uc?,'
O
0
t;
,
"
"
"
"
"
"
"
" 5_
" R_""'"
'"
" "'.......
"" R-"'"
s""~
D~O
0 '(5 .
0 0
D.C.
22
"
26
Reservecl
S".~
19
41
Module conl'lllC'ttlr
Sp...
.2110vo~
16
36
F"netlon
Ru._
R........d
Fleserved
S,..~
''''~
'O?O
.:!4volts
-24 volts
S... ~
'P"~
"090 6
O"Ob
,
0
d
0.,,0
'P~
117 volts A..C. (hot)
ll_fMUmGND
R....
....
Sporn
..
J.
-,.~
1-0
0'0
"""""....
.cOottom
,~
guide
117 vaH"
High quality GND
Ground guide pin
GGP
" MIZSt De bussed to all bin

CCl~ors PG18 lJtrough
PO".
Nates;
1. Rl!served pins 1m! for future assignment by the CCIfJ1mlttee mel shall notbe usl!d until sucf1
IIIslgnmenb are made.

2. GP-1 .. Guide Pin
~1 .. Guide Soc:ket
~,~P'~:~:-tf,
GGP .. Ground Guide PIn

GGS " Ground Guide Socket
~~1B~1~
GGS
GOP
3. The POWER RETURN GROUND (bIn pin 34) Is the return b~ fOr III cle supplies.
The HIGH QUAUTY GROUND (blf\ pin 42) 131n~ud iI3 uro potential rde~". High quality
Bin conn.
Mod. conn.
gr.... nd aum:nt load by any mod,,1e ahould b<lllmlted t61 mA. AIsft, pulses 01' varying loads
ahould tie Omlled to riO mORl than 100 mieroamperes. cant mould be takefI not to
Sh.-lng guido pins Emd eeckets
capadtlvely coupll!l high quality ground to the loc:aJ grouncL Any capacitive coupling to local
grourw:l should be IsoIatecI by a ,,"1slanc1t of at least 1000 ohms. The 01ASSIS GROUND I..
fIOnna\tJ conne.::tecllo the bulldhlg grolln4. CClnnll<illoroa batw'cen the HIGH QUAUTV
GROUND bus [bIn pin G). !he powet' RETURN GROUND BUS (bIr1 pin 34), and the ehassls are
rnacle at the ground guide pin of PG1B.
""m-.
4. Slgnal3an pins 35 ancl36 st\aD UtlIlm Iha -lOgIc levelS 10f transrnlS3kln Cli digital data down to
OC- ,,' tha nlm .. peclfl...atlQllS. A 1091<;; "'" gate .. ON Ilnd RESETS.
FIG. 16.8. The pin =ignments for the NIM _dard connector (from USDOE, 1990. p. 25).
403
Chapter 16
Table 16.5. Pin assignments for Orice and Canberra amplifiers.

Pin
Signal
Description
GND
Power-supply ground
GND
High-quality ground
+12
+12 V dc supply
-24
-24 V de supply
--,-_~c_.L. __ "c:~~_.,::?4_
16.9). This circuit is also useful as a simple 50 Q cable driver

for TTL logic. NW fast negative logic most closely resembles
emitter-coupled logic (ECL). and modules to interface fast
negative NIM to EeL are commercially available (e.g., the
Ortec ECI600 Hi-<:hannel ECL-NIM-ECL converter). If you
need to build such an interface. an (aging) example is given in
Althaus aod Nagel (1972). Before beginning. become familiar
with the latest bigh-speed interface circuitry from the leading
manufacturers of integrnted circuits.
+24 V de supply
12
+ 12V
12 Vde supply
TTL
Table 16.6. NIM standard logic levels.

NIM logic
family
Positive logic
Logic
level
Output
(must
deliverl
NIM
--I
Input
(must accept)
Logic 1 +4 V to +12 V +3 V to +12 V
(1/2) ICL7667
or
(1/2) 050026
Logic 0 -2Vto+1 V -2Vto+1.5V
Fast' . negailveLofjie 1 -14 rnA to -12 rnA to--36

18 rnA
rnA
logic (50 Q)
+5V
Logic 0 +1 rnA to -1+20 rnA to-4

rnA
rnA
Because of the age of the NIM standard, its logic and voltage
___. levels .,do no! correspond with any standard logic family used
today or with the :15 V power supplies that are now commonly
used 10 power operational amplifiers (op amps) for analog
circuit functions. For the occasional analog construction project,
it is usually easiest to simply operate op-amp circuitry at the
TTL
NIM
TTL Gate
reduced 12 V that is available from the NIM hin. Interfacing
-12V
logic signals presents a bit more of a challenge. By far, the most

common integrated logic in use today is transistor-transistor
logic (TTL) and its compatible complementary metal oxide
semiconductor (CMOS) successors. To drive a standard TIL
circuit from a positive NIM output, you can use a commercial
module such as the Ortec 416A gate and delay generator.
Alternatively, you can use a comparator or simply resistively
couple the signal to a TTL gate with Schmitt trigger input that
has been diode-damped to ground and the +5 V supply, as
shown in Fig. 16.9. To boost a TTL output 10 drive a positive
Nllv1 input, use a driver with TIL-compatible input such as the
ICL7667 dual monolithic high-speed driver (an updated version
of the DS0026 driver) connected to the +12 V supply (Fig.
FIG. 16.9. Simple circuits to interface medium-speed positive

nuclear instrument module (NIM) logic and transistor-transistor
logic (TTL).
lCi.4.3 Basic energy spectroscopy
The simplest form of an energy spectroscopy system that is

useful for ion beam analytical work is shown in Fig. 16.10.
which presents a block diagram of the modules Co?
commercially available) that are needed The detector
indicated by a stylized diode symbol.
404
15
Detector bias
supply
+ High voltage
Spectroscopy
amplifier
MCAorADC
and other
preamplifier!
.........:~_ _ _...J' 1..-_ _---1_
signal
rocessing
.---lL
Particle
or
Photon
or
fIG. 16.10. A simple energy spectrometry system.
The voltage required to establish the reverse bias of the

detector is supplied by a bias power supply that is
conventionally a separate module, often with several
independent outputs. The high voltage is usually physically
applied to the preamplifier module, which contains circuitry for
limiting the current that ean flow in the detector and for
capacitively shunting the pulses into the charge-sensitive
preamplifier circuitry, which operates with ground reference.
16.4.4 Preamplifiers
Preamplifiers. come in two types, charge-sensitive and
VOltage-sensitive. Energy spectroscopy systems, such as the one
in Fig. 16.10, use charge-sensitive preamplifiers. As the diagram
implies, these devices integrate the pulse from the detector. The
reason for this is thai"' the - mforniation about the radiation's
energy is encoded in the number of electron-hole pairs that are
created by the detector. By integrating the pulse from the
detector, the energy can be measured directly. The resistorcapacitor (Re) time constant in the preamplifier's feedback
circuit is chosen to be on the order of 100 ).IS, which is vastly
longer than the pulse generated by the detector. For this reason,
the preamplifier's output resembles the graph of e-l for t:> 0 and
is called a tail puLse. The information is contained in the height
of the step at t = O.
The pulse from the preamplifier is further amplified and shaped

by the main spectroscopy amplifier and digitized and displayed
by a multichannel analyzer. Schematics of the shape of the pulse
are shown in Fig. 16.10 at various stages ofprocessing.
It is very important to emphasize that the preamplifiers and

amplifiers that are used for nuclear spectroscopy are very
different from those that are encountered, for example, in high-
fidelity audio applications. Although both are, io the strictest
During the sehlp and calibration of the system, it is desirable

to have a way to produce pulses without having to provide a
source of radiation. This is done with a special pulse generator.
Commercial spectroscopy preamplifiers provide an input for this
signal. If the pulser output is very linear with its amplitude
control, then the overall linearitY of the spectroscopy system can
be measured. This is particularly important for high-resolution yray systems because of the large number of very narrow peaks
that appear in the spectra. The resolution of the electronic system
. sense, linear, the audio amplifier is designed 10 have essentially

flat frequency response from near-de to several times the
maximum frequency that humans can hear. Spectroscopy
amplifiers, on the other hand. have frequency responses that are
carefully tailored to preserve the information content that resides
io the amplitude of the detector's pulse
405
~-
Chapter 16
can also be observed by disconnecting the detector and recording

the width of the pulser line.
Voltage-sensitive
preamplifiers
are
used
for
input and produce output pulses that are approximately Gaussian

in shape. proportional in amplitude to the input pulse step. and
approximately 1-2 ~ in duration. They accomplish this by
differentiating and then integrating the input signal. A diagram
of a modem semi-Gaussian shaping amplifier is shown in Fig.
16.11.
timing
applications such as the time-of-flight spectrometer discussed in

Section 16.3.5. These preamplifiers simply increase the
amplitude and power of the pulse without integrating it or

materially altering its shape (except for a possible polarity
reversal). It is most important to match the frequency response
of the preamplifier and its rise time with the detector signal
being processed. For example. microchannel plates need faster
preamplifiers than surface-barrier detectors. Manufacturers'
The first stage of a spectroscopy amplifier is

schematically in Fig. 16.11 as a capacitor-resistor (CR)'
differentiator. This -configuration of capacitor and resistor
(ignoring R2 for the moment) is referred to as a differentiating
circuit because. for signals that vary more slowly than the time
constant, the voltage at point 2 is approximately the derivative of .
the input. For the tail pulse whose rise time is much shorter than .
the time constant, the resulting output is a much more quickly
decreasing exponential as shown. This signal is amplified by
and fed to an integrator circuit, which, in its simplest form, is
just another capacitor and resistor wired as shown. This circuit
slows the very rapid rise of the pulse and yields the shape shown
at point 3. A second amplifier, A 2, provides isolation for this
stage and adds additional gain. In a modem amplifier, the simple'
RC integrator is replaced by a more sophisticated active
integrator in order to improve the signal-to-Doise ratio for low- .
level signals from semiconductor detectors. However, the
principle is the same.
literature provides guidance on the proper selections for different

applications. The pin assignment for the preamplifier power
connector is listed in Table 16.5. Power to NIM preamps is
typically supplied from 9-pin D connectors that are installed on
spectroscopy amplifiers for this purpose. This CODDe(;tor is not
part of the NIM standard but is used by several manufacturers.
However, be sure to verify that your equipment uses compatible
pin assignmentS. When this connector was flrst introduced.
incompatible pin assignments used by different manufacturers
led to much inconvenience.
16.4.5 Spectroscopy amplifiers

Modem spectroscopy amplifiers are extremely versatile and
powerful tools. Their basic function is to accept tail pulses as
J--
Input
from
preamp
I~~Rf-2.~. .-R' -l
A,
Differentiator
and pole-zero
cancellation
FIG. 16.11. A simplified schematic of a spectroscopy amplifier. Adapted from the Ortec Catalog. copyright EG&G Ortec. Used with
permission.
Most modern spectroscopy amplifiers have additional circuitry
that aids in improving performance at high count rates. Three
such features are pole-zero cancellation, baseline restoration, and
pileup rejection. A pole~zero circuit is also shown in Fig. 16.11.
It derives its name from the cancellation of a pole and a zero in
the complex plane in the Laplace transfonn analysis of the
circuit. Its operation is quite simple: If R2 were not present in the"."

circuit, then point 2 would be ac-coupled. to the input, and ~e voltage there would always average to zero. At moderate to high
count rates, pulses would be shifted by the undrnhoots of
previous pulses, and most disastrously. this shift would be ratedependent. R2 allows a small fraction of the input pulse to
406
is high and these overlaps are frequent. In order to reduce pileup,

many spectroscopy amplifiers have pileup-rejection circuitry.,
The principle is very simple: Within the amplifier, a parallel
signal path is established from the input. By differentiating the
input signal, it is possible to create very narrow pulses that
correspond to the leading edges of the input tail pulses. When a
pulse anives at the amplifier's input, a flag is set in digital logic
indicating that the amplifier is busy. If a second input pulse
arrives during the time (equal to the width of the shaped output
pulse) that this flag remains set, a second digital logic pulse is
generated coincident with the output pulse. This digital pulse is
provided as an output of the amplifier and can be used to gate a
multichannel analyzer or other modules and prevent the piled-up
analog pulse from being processed. Pileup is thus reduced at the
expcnsc of additional dead time in the spectroscopy system.
Figure 16.12 shows a ganuna-ray spectrum recorded with and
without pileup rejection. The procedure for correcting for dead
time is discussed below. For further- discussion of pileup, see
Knoll (1989) and references Iherein.
circumvent the blocking capacitor C. "''hen R2 is properly

adjusted. the undershoot is exactly canceled, and the ratedependent error is eliminated This is what you are doing when
adjusting the pole-zero cancellation potentiometer of a,
spectroscopy amplifier. Another feature of good amplifiers is a
baseline restorer. Because modem amplifiers are multistaged
and dc-coupled, the later stages amplifY the de offsets of
previous stages. The baseline restorer deals With this source of
offset. At low count rates, a simple CR differentiator (which is
equivalent to ac coupling) would suffice. However, this circuit
has problems at high pulse rates, as just described. The problem
is solved at the expense of more complex circuitry. The
amplifier's baseline-restoration circuit should be used when
trying to do precise spectroscopy at high count rates.
Pileup rejection is also very useful when high count ratcs arc
encountered. if two pulses arrive within a time interval that is
less than the approximate half-width of the main amplifier's
output pulse, they are sununed and are subsequently
indistinguishable from a single pulse with larger amplitude. This
leads to significant distortion of a spectrum when the count rate
Low counting rate (100 CPS)

High counting rate (8000 CPS)
103
-.,
I::::. ...
I
.~
II
I:
I
I
I
I
I
I
I
a;
Ir-
II:
101
__
Pulser
.,"
102
I:
:l
0
-.,
..'"
iI
I
I
I
I
I
I
I
\."
100~~~\t---L----L____~__~____~__~
Channel nuinber
FIG. 16.12. Spectra of 55 Fe showing distortion caused by pulse pileup. From Wielopolski and Gardner (1976). Used with permission.
407
Chapter 16
16.10. If your fmal pulse is negative, use the amplifier's input

polarity switch to invert it. Adjust the shaping constants
according to the instructions given in your multichannel analyzer
manual, or if the manual is not available, adjust the shaping
constants to give an aesthetically pleasing, bell-shaped pulse
with a width of about 1-2 ~s, as illustrated in Fig. 16.13. If you
need to use bipolar pulses, be sure that they swing positively
first and then negatively. Read your amplifier's manual and
follow its directions to adjust the pole-zero potentiometer.
Finally, if any of the connectors or controls on your amplifier
become loose, fix them. There is little else to go \IIIOng with a
. good spectroscopy amplifier.
Many spectroscopy amplifiers permit the selection of either

unipolar (semi-Gaussian) or bipolar output pulses (Fig. 16.13).
The latter are typically obtained by an additional differentiation
of the former. Consult the manual for your multichannel
analyzer to detennine which it prefers to receive as input. If you
cannot frod a preference stated, use unipolar pulses. Adjust the
amplifier using an oscilloscope, and keep the oscilloscope
connected and turned on throughout the experiment. -Seeing
badly shaped pulses is often the fIrst clue that something bas
gone wrong. For most additional processing. the amplifier's

output pulses should be positive and have amplitudes in the
range of 1-10 V. Adjust the gain to make it so. Preamplifier
._. pulses are sometimes inverted relative to the case shown in Fig.
8
6
J'\/ I'\.
Unipolar
/A \
/ 1/ \
J/ \
"- "-
.\
-2
-4
.'-V
3
Bipolar
4
Time (liS)
FIG. 16.13. Characteristic shapes of unipolar and bipolarNThninear pulses from a spectroscopy amplifier. The
bipolar pulse is obtained as the derivative of the unipolar pulse.
data channels, computer-based systems such as CAMAC (Clout,
1982), which was originally developed for high-energy pbysics,
are used. Here. we describe a standalone MeA with a Wilkinson
analog-to-digital converter for the sake of simplicity. Data
acquisition and handling are the most individualized aspects of
nuclear spectroscopy. The best advice that can be given is to
understand your own system, whatever its details. and not to
leave those details to someone else.
16.4.6 Multichannel analyzers

A properly adjusted spectroscopy amplifier's output pulse is
ready for analog-to-digital conversion. This is most often carried
oul in a multicharmel analyzer (MCA) (Fig. 16.14). The MCA
has been evolving rapidly in the past few years. It was originally
invented by D.H. Wilkinson for nuclear studies, and his analogto-digital conversion technique is. still widely used in
commercial equipment (Wilkinson, 1950). Both standalone
MCAs and personal computer plugin cards are now widely
used. In addition, for high-count-rate applications or multiple
408
Instrumentalion and laboratory Practice
To
computer
Display
'-"'e"'m"o"","-' Inte~ce
Input
"~l
buffer
.A @
Analog to
I1lglhll
converter
Shaped
pulse
.rL
Busy
out
FIG. 16.14. Schematic diagram of a multichannel analyzer.
dom rather than constant-frequency test pulses should be used.

However, for practical ion beam analytical work, the difference
is irrelevant if the pulser rate is below about 10% of the real
count rate. In modern MeA designs, the tolal real time and live
ti.J;ne are usually made available to the user. However, these
values are often less accurate measures than can be obtained
using a pulser. As a result, for the most precise work, you should
do your own dead-time estimation. The use of the measured
dead time to correct data is described in Section 16_6.2.
The heart of the MeA is the analog-to-digital converter

(ADC). The ADC analyzes the maximum amplitude of each
pulse with a precision (called conversion gain) that typically
ranges from one part in 256 to one part in 8192. Figure 16.15
illustrates the principle of the Wilkinson ADe. The input pulse
is wed to charge a capacitor until the:: peak of the:: pulse is
detected (by a zero crossing of its derivative). At this time or
with a small delay, a constant current source is turned on and
begins to linearly discharge the capacitor. At the time that the
discharging begins, a high-frequency clock is also started, and

its ticks are counted until the voltage on the capacitor crosses
zero. This stops the discharge, stops the clock, and makes the
total number of counts available as the digitized pulse amplitude.
Good Wilkinson ADCs use clock frequencies of 100 .MHz or
higher and, therefore, can do an 8k channel conversion in a time
on the order of 100 J.lS and a lower-precision conversioD
proportionally more quickly. During the conversion, the
analyzer is unavailable for processing additional pulses. It is
necessary to take this dead time along with pileup dead time into
account when doing quantitative work, One way to measure the
ADe dead time is to gate an oscillator using the circuitry in the
ADe that inhibits additional pulses while one is being
processed.. By inhibiting counts while the ADC is "dead". a
direct measure of "live time" is produced. A better way is to
leave a stable pulser attached to the preamplifier and adjust the
amplitude to insert test puLses- into a portion of the spectrum that
does not contain real features of interest. By comparing the
number of these pulses observed in the MeA spectrum with the
total number generated, a direct measure of counting losses due
to both ADC dead time and pileup rejection can be obtained. For
most applications, a rate of test pulses as low as 60 Hz is
It is assumed that the quantity being measured, for example,

the energy E or time T, is related to the MeA channel, n,
through a linear relationship such as
E=an+z.
=
(16.2)
where and z are cons~ts. The constant z is called the zero

offset. Both z and a must be obtained by calibrating the analyzer.
This is done by using standards in which spectral features with
known energies can be identified. Tables 16.7, 16.8, and 16.9

contain useful X-ray, y-ray, and a-particle calibration sources.
[For an electron source with discrete lines you can use a
radioactive source that decays by internal conversion (the
nuclear analog of Auger electron emission). See Jardine and
Lederer (1974)] Clearly, two spectral features with known
energies are sufficient to calibrate a linear analyzer. However, it
is better to have more so that a and z can be obtained with
probable errors by linear regression. Subsequently, one obtains E
corresponding to a channel number n from Eq. (16.2). For very

precise .,.-ray work with gennanium detectors, it is necessary to
verify the linearity of the analyzer. For routine ion beam
analysis, all commercial equipment is sufficiently linear that Eq.
satisfactory. Some authors argue (see Nicholson, 1974) that ran-
(16.2) can be assumed to hold.
409
--:1'
~
Chapter 16
(KnoU, 1989) ar multiparameter particle detection systems

(England, 1974), such as particle telescopes (which
simultaneously measure total energy and stopping power of an
ion by using a thin surface-barrier detector in combination with a
thick one) are two such cases. When more than one parameter
must be stored in connection with each event, the best generalpurpose approach is to use a CAMAC data acquisition system,
because they were developed for just such a pwpose.
Commercial hardware and software are available to implement
multidimensional histogranuning or to store data on an event-byevent basis for later computer analysis.
iLl
~
TIme
(a) Amplifier :
output_
pulse
I
Lower level
discriminator
UI
(b) Lower
level discriminator I
output
I
(c) Signal on
rundown
capacitor
(d) Address
clock _--,_~!u
:1
(el Memory
cycle -~f--i--------,I
(I) Linear
gate
closed
~I_ _ _ _ _ _ _ _,
(g)Deadt~
gate
L
L
FlG. 16.15. Timing diagram of a Wilkinson analog-to-digital

converter (ADC). From the Ortec Catalog, copyright EG&G
Ortec. Used with permission. The output of the ADe is a
number that is proportional to the height of the input pulse. An
output event is produced, of course, only when a pulse appears at
the ADC input. The MCA takes the number that is computed by
the ADC and uses it as an address to access a memory bank (Fig.
16.14). When a pulse arrives, the MeA addresses the indicated
memory register and increments it by 1 to record that one pulse
of that amplitude has been measured. In this way. a histogram of
measured pulse amplitudes is accumulated. Commercial

standalone MCAs display this histogram in real time on a VIdeo
monitor and offer several display options and often an
assortment of analytical capabilities. They usually also provide
some mechanism for downloading the information to a computer

for storage and further processing. Plug-in cards for personal
computers use the computer's display under control of
specialized software.
Although multichannel analyzers will suffice for most ion
beam analytical work, there are some instances where the
histagramming of a single parameter to produce a spectrwn is
inadequate. Experiments usmg position-sensitive detectors
410
16.4.7 Cables and connections

The analog and logic signals, collected beam current, detector
biases and most other electrical signals are routed from module
to module and throughout the laboratory using coaxial cables
(Bryant, 1984). These cables are used so widely for electronic
equipment that their specifications have been standardized by
the U.S. Government. Table 16.10 contains a detailed listing of
cables and their specifications by their RG numbers (redlo
guide-no.luniversal). The three most widely used cables for
nuclear spectroscopy are RG-621U, RG-58A1U, and RG-174/U.
In addition RG-SAIU and RG-S9AIU are ofien used as highvoltage cables. Analog and positive logic NW signals are routed
using RG-62flJ cable, which has a characteristic impedance of
93 n. Module inputs for these signals typically have an
impedance of around I ill, so that several inputs can be daisychained together, driven in parallel from the same output signal.
Because of the mismatch between the impedance of the cable
and the instrument input, it is necessary to terminate the cable in
its characteristic impedance. This is very important. Improperly
tenninated cables, especially long ones, arc one of the most
common sources of poor or erroneous signals.
For routing fast negative logic and high-speed analog signals,
one typically uses either RG-58AIU or RG-174fU cable. The
forme)" is more appropriate for long cable runs. The latter cable
has a much smaller diameter. but also has greater attenuation per
unit length. Even so, RG-174/u is more convenient for short,
dense interconnections. It is not usually necessary to provide
external tennination for a standard NlM fast negative input,
because it is nonnally terminated internally in 50 Q. However, if
fast signals are being routed through equipment that might not
be internally terminated, it is vital that proper termination be
provided externally. Signals propagate in coaxial cable at about
60% of the speed of light [propagation velocity = (z,cr', where
Zo is the characteristic impedance and C is the capacitance per
meter]. Fast negative logic pulses can be as short as 5 os. As a
result, it is possible, even in relatively short cable runs, to get
distinct, multiple pulses from improper termination. Do not step
on coaxial cables or bend them in arcs shorter than about 10
times their own diameters. If you need to make a 90 0 bend at a
I
!
I
j
II
i
j
I
i
i
junction, do it with a 90 elbow connector, not by kinking the

cable.
Coaxial cables are attached to NIM modules (and most other
laboratory equipment) by standard interlocking connectors
(Table 16.11). These too have standard designations (UG, in this
case, for union guide-no.luniversal), but these are not used as
frequently as the RG numbers of cables. The three most
hnportant connectors are the BNC [baby (or bayonet) NeillConcelmanl. the SHY (safe high-voltage). and the LEMO (the
company's name). BNC connectors are rated for voltages up to
500 V dc and are by far the most common connector. SHY
connectors are used for handling voltages above the limit for
BNC connectors (up to 5 kV dc) and for ultralow-Ievel currents
where the leakage at a BNC connection could be significant.
Another high-voltage connector, the MIN (miniature highvoltage), is also still encountered on some equipment However,
it is possible to force an MHV plug onto a BNC socket 'With
potentially disastrous consequences for equipment and people.
For this reason, do not use MHV connectors. Tfyou have a piece
of equipment that has MHV cOnOcctors on it, trcat it with special
care and replace the MHV connectors with SHY connectors the
next time that maintenance on the unit is required. LEMOs are
elegant little connectors that make possible very high connection
densities, especially when used with RG-1741U cables. They are
found most frequently on designs such as timing discriminators
where large numbers of 50 n cable connections are needed on
narrow modules.
signal ground as part of the construction. Such a system consists

of a large conducting mesh that is planted in the earth near the ,
laboratory and is connected to various strategic points through
heavy conductors. Do not underestimate the time and effort that
can go into eliminating ground loops or the damage to your data
that they can cause.
Quoted X-ray energies should not be regarded as standards
because of uncertainty caused by line shape (many of the lines
have complex structure), chemical shifts, and other shifts.
Sources: Campbell aDd McNelies (1975), Helmer et af. (1979).
Tuli (1985), and Lederer and Shirley (1978).
Table 16.7. Calibration sources for solid-state X-ray
detectors, including the intensity ratio of specified X-rays to the
indicated X~ray. Uncertainties in the final digits are shown in
parentheses.
Intensity
Half-life
Ex
(keV)
Ey
Isotope
(keV)
(XJy}
54Mo
312.5 d
S.47 (Kapl
834.84
0.251 (11
57Co
270.9 d
122.06
1157 (1)
0.79 (21
0.tt2 (21
6'zn
243.9 d
6.4 (K.)
7.1 (Kpl
14.4 t (yl
B.04 (Kal
1,115.5
B.9(~1
85S r
Bay
64.84 d
106.6 d
The most conunon problems that can occur in interconnecting

systems are incorrect routing of signals, poor-quality or
109Cd
462.9 d
improperly maintained cables and connectors, improper
termination of cables, and ground loops. The last is the most:----, ._--- - ----, -_."
113S n
115.1 d
insidious. All grounds are not equal. In an accelerator
laboratory, some of the equipment is usually located in the target
137Cs
30.17~
room, and some is in the control room. With such distributed
systems, it is possible, even likely, to encounter situations where
139ee
137.7 d
there are large loops in the ground wiring. These can be the
sources of offsets up to several volts. Get to know ground loops and
198Au
design your systems to avoid them. For help, consult Morrison
2.696d
(I977). Amazingly. he has made the subject interesting. If you
are building a laboratory, consider specifying a high-quality
203Hg
46.60 d
-~
0.660 (51
0.091 (2)
t3.38 (K.l
15.0 (~I
514.00
0.502 (3)
0.088 (1)
14.12 (KcJ
15.B5 (Kp)
89B.04
0.549 (7)
0.9B9 (2)
22.10 (K.)
25.0 ("pl._
8l30037
24.14 {Kal
27.4 (Kp)
391.69 ~
1.22 (4)
0_27 (1)
32.1 (t<.)
36.6 (K,I
661.66
0.067 (21
0.Ot59 (5)
33.29 (K.)
165.85
0.81 (9)
0.20 (2)
70.15 (K.l
80.7 (Kpl
411.80
0.0229 (51
0.0064 (2)
72.11 (K.)
B3.0 (Kpl
279.20
0.125 (3)
0.03-48 (B)
38.0(~)
22 (1)
4.7 (2)
411
~-.-----:=-~-
---_..
!
I~
Chapter 16
Table 16.8. Suggested y-ray calibration sources including half-lives and principaly-ray eoergies. Uncertainties in the final digits are
shown in parentheses. Sources: Helmer 0/ 01. (1979), Tuli (1985) (continued).
(k~
Source
Half-lIfe
(k7#)
Source
Half-life
'Be
53.29 d
477.605 (3)
10SmAg
127y
11(mAg
249.76 d
124Sb
60.20 d
602.730 (3)
645.855 (2)
713.781 (5)
722.786 (4)
790.712 (7)
968.201 (4)
1,045.131 (4)
1,325.512 (6)
1,388.164 (7)
1,436.563 (7)
1.690.980 (6)
2.090.942 (8) -
137CS
30.17y
661.660 (3)
13C
+ ex
22Na
a source
2.602 y
6,129.27 (5)
t,274.54 (1)
24Na
15.02 h
1,368.63 (1)
2,754.03 (1)
"50
83.83 d
889.277 (3)
1,120.545 (4)
51Cr
27.704 d
54Mn
312.5 d
320.064 (')
834.643 (6)
"Fe
44,496 d
1,099.251 (4)
1,291.596 (7)
56co
78.76 d
646.764 (6)
1,037.644 (4)
1,175.10(1)
, .238.20 (')
1,360.21 (')
, ,771.35 (2)
1,810.72 (2)
. ',963.71(')
2,015.18 (1)
2,034.76 (1)
2,113.11 (1)
2.212.92 (1)
2.598.46 (1)
3.009.60 (2)
3.201.95 (1)
3.253.42 (1)
3.273.00 (1)
3.451.15(1)
57Co
6Oeo
270.9 d
5.271 Y
144Ce
122.0614 (3)
136.4743 (5)
152.Eu
1.173.238 (4)
1.332.502 (5)
2B4.4d
13.33y
433.936 (4)
614.281 (4)
722.929 (4)
446.8tl (3)
620.360 (3)
657.762 (2)
677.623 (2)
687.015 (3)
706.682 (3)
744.277 (3)
763.944 (3)
616.031 (4)
664.685 (3)
937.493 (4)
' .364.300 (4)
1,475.788 (6)
1,505.040 (5)
1,562.302 (5)
696.5'0 (3)
1.469.160 (5)
2.165.662 (7)
121.7824 (4)
244.699 (1)
344.281 (2)
153Gd
2".6d
69.6734 (2)
97.4316 (3)
103.1807 (3)
898.042 (4)
1.636.06(1)
'1>fm
128.6 d
84.2551 (3)
64.02 d
724.199 (5)
'82'fa
114.5 d
2.03x10 4 y
702.645 (6)
871.119 (4)
67.7500 (2)
84.6808 (3)
100.1065 (3)
113.6723 (4)
SSZn
243.9 d
1.115.546(4)
say
106-54 d
"Zr
"Nb
412
I
i
Table 16.8. Suggested y-ray calibration sources including half-lives and principal y-ray energies. Uncertainties in the final digits are
shown in parenlheses. Sources: Helmer el al. (1979), TuU (1985) (continued).
Source
'''Ta
73.821 d
b....-
Half.llfe
(ConI)
2.696 d
413
411.804 (1)
675.888 (2)
1,087.691 (3)
46.6d
279.197 (1)
32.2y
569.702 (2)
1,063.662 (4)
l,nO.24 (1)
1.9131 Y
136.3434 (5)
205.7955 (5)
295.958 (1)
308.457 (1)
316.508 (1)
(k~
416.472 (1)
468.072 (1)
484.578 (1)
588.585 (2)
604.415 (2)
612.466 (2)
884.542 (2)
1921r
116.4186 (7)
152.4308 (5)
156.3B74 (5)
179.3948 (5)
19B353 (1)
222.11 0 (1)
229.322 (1)
264.076 (1)
1,121.301 (5)
1,189.050(5)
1,221.408 (5)
1.231.016 (5)
1,257.418 (5)
1,273.730 (5)
1,289.156 (5)
1,373.636 (5)
1,387.402 (4)
(ConI)
,
!
SOurce
Half-life
238.632 (2)
563.191 (2)
B60.564 (5)
B93.408 (5)
1,620.74 (1)
2,61453 (1)
-~
Chapter 16
Table 16.9. Alpba particle energy standards, including haf-lives, energies, and relative intensities. Uncertainties in the last digit of tbe
energies are sbown in parentheses. Vel)' thin 212Si and 212po sources can be f,repared by collecting lWRn on a sample surface in the
vicinity of a thorium surface. The controlling halflive is 10.64 b, that of" Ph from which mBi forms by ~ decay. Sources: Rytz
(1979), Ledereer aod Shirley (1978), Tuli (1985).
Isotope
Half-life
'''Sm
1.03x10 By
Eo (MeV)
2.470 (6)
Itt (%)
Isoto~
Half-life
100
2-4Pu
6570y
5.1682 (2)
5.1237 (2)
73.5
26.4
100
2~Am
738() y
5.350 (I)
5.2753 (5)
5.2330 (5)
0.22
87.4
210pO
138.376 d
5.3044 (1)
99+
(calculated)
1.8:><106 '1
150Gd
2.719 (8)
(calculated)
148Gd
75.:l:30y
""Tn
1.41xl0 1O y
... _. ';
.~.-.
:. .. ......_.. ..........
238U
4.468x10 9 y
23SU
7.038x10 By
236U
2.342x107 'I
3.18271 (2)
100
22BTh
1.9131 Y
4.197 (5)
4.150 (5)
n
23
5.4232 (2)
5.3403 (2)
12.7
26.7
241Am
432.2 y
5.544 (1)
5.4856 (1)
5.4429 (1)
0.34
85.2
13.1
""'Pu
87.74 y
5.4991 (2)
5.45S3 (2)
71.5
28.5
24-4Cm
18.10y
5.80482 (5)
5.76270 (5)
23.6
4.599 (2)
4.400 (2)
4.374 (4)
4.368 (2)
. 4.218 (2)
1.2
57
6.1
12.3
6.2
4.494 (3)
4.445 (5)
74
26
4.688 (2)
4.621 (2)
75.5
24.5
234U
2.45xl0 5 y
4.n5(1)
4.723 (1)
12.5
27.5
209pO
102 Y
4.881 (2)
99.52
23 1Pa
3.28xl04 y
5.059 (1)
5.028 (I)
5.014 (1)
4.952 (1)
4.736 (1)
11.0
20.0
25.4
22.B
B.4
239pu
2.412x10 4 y
5.115 (2)
5.1566 (4)
5.144 (1)
5.105 (1)
'43Cm
28.5 y
242Cm .
162.8 d
73.3
15.1
11.5
6.066 (I)
6.058 (1)
5.992 (1)
5.7850 (5)
5.7411 (5)
76.4
1.0
5.0
5.8
73.54
10.65
6.1128 (I)
6.0694 (I)
26.2
73.8
254Es
275.5 d
6.429 (1)
93.2
2S3es
20.47 d
6.63257 (5)
6.5914 (5)
90.0
6.6
212Bi
SO.55 m
6.08994 (4)
6.05083 (4)
27.2
69.9
212pO
0.2981'S
8.7842 (1)
100
414
11.0
n
23
7.S4xl04 'I
.:==.:: 2.898 'I
I.{%)
4.013 (3)
3.954 (8)
""rh
_ :~:~_ ~Po :.:--...
Eo (MeV)
100
15':;;'
~il=
p -
!>'
!='
."
Cable
Designation
(Badlo
gulde--ll
1JJ.nlversal)
conductor
styl,e
(AWa .
stranding}
RG-BIU
RG-8AJU
13(7)(21)
RG-9/U
13(7)<21)
RG-11/U'
RG-11A1U
1B (7)(26)
.g
Signal
Center
.lnsulaUon
propagation
velocity
Nominal
Nominal
Cable Maximum
Impedance
capacitance
(% of sptI&d
voltage
O.D.
(MHz,
C
(pFlm)
of light)
(mm)
Z.
(a)
(V .,)
Polyethylene 10.29
Polyethylene 10.67
Polyethylene 10.29
5,000
52
51
5,000
75
5,000
96.8
98.4
67.3
66
66
66
100
400
1000
Nominal
~.
signal
attenuation
Governing
(dBI100m)
specification
Comments
0
...,
7.2
JAN-C-17A
Black pvc Jacket.
15.4
MIL-C-17D
AG-eAIU has non-
contaminatIng jacket.
29.2
7.2
JAN-C-17A
100
400
1000
29.2
100
400
1000
6.6
13.8
23.3
JAN-C-17A
100
400
1000
14.8
32.8
55.8
JAN-C-17A
100
400
16.1
37.7
70.5
JAN-C-17A
15.4
MIL-C-17D
Gray. non~
contaminating
PVC
1>
~
"n
Jacket. Double
shielded.
Black PVC Jacket.

RG-11AA.J has nan-
contaminating jacket
~.
e-
~
~
AG-S8AJ
...V;
RG-58AJU
20 (solid)
20 (19 x 32)
polyethylene
Polyethylene
4.95
4.95
1.900
53.5
50
1,900
93.5
101.0
66
66
1000
RG-58C/U
20 (19 x .0071) Polyethylene
4.95
1,900
50
101.0
66
Polyethylene
6.15
2,300
73
69.9
66
100
400
1000
RG-59/U
22 (solid)
100
400
1000
RG-59B/U
23 (solid)
Polyethylene
6.15
75
2,300
67.3
66
;i
RG-62IU
RG-62BJU
22 (solid)
24 (7 x 32)
Seml-sond
. polyethytene
Semisolid
polyethylene
6.04
6.15
700
700
Ii 93
93
44.3
44.3
. 64
64
100
400
1000
100
400
Black pvc jacket.
~.
Black pvc jacket.

lor fast NIM
signals.
Used
16.1,
37.7
70.5
MIL-C-17F
Black, nonconta,mlnating PVC

jacket.
11.2
23.3
. 39.4
JANC-17A
Black PCV jacket.
11.2
23.0
MIL-C'17D
Black, noncontaminating pvc

jacket.
39.4
6.9
1000
17.7
2B.5
100
400
1000
20.0
36.1
9.5
."
JAN-C17A
MIL-C-17D
Black PVC Jackel.

Used for NtM linear
and positive logic
signals.
Black, noncontaminating PVC
jacket. Also used for
NIM applications.
"'e-.
~
~
~
~...
,
~
"""
00
00
.'"
ro
..S:
...
0
:t!.
P(")
i'l-
po
"'"2"
3
"~
0
""
"....""
0-
~
"gOJ
.
n
5';:7
&ocr "
? ~
~,
0_
00
o '
~
S'
" 0
8.//
~a.
n
~
0
....
Cable
Designation
(Badlo
g,ulde-#
IUnlversal)
Center
conductor
style
(AWe]
strandIng}
Insulation
Signal
propagation
Cable Maximum
Nominal
Nominal
velocity
0,0.
voltage Impedance capacitance (% of speed
(mm)
(Y .,)
C (pFIm)
'Zo(n)
of light)
(MHz)
90
Nominal
signal
attenuation
(dBl100m)
Governing
specification
29.2
MIL-C-170
Comments
RG-174N
26(7:<34)
Polyatl1ylane
2.56
1,500
50
101.0
66
100
400
1000
57.4
Blaek PVC Jacket.

Used tor high
98.4
speed NIM signals,

ollen with LEMO
connectors.
RG-17BB/U 30 (7 x 38)
Teflon
1.83
1,000
50
95.10
69.5
100
400
1000
45.9
MIL-C-170
91.9
Brown. fluorinated
ethylene-propylene
150.9
jacket High
temperatura.
"'O;"
RG-213/U
RG-2231U
13(7x21)
19 (SOlid)
Polyethylene 10.29
Polyethylene
5.38
5,000
1,900
50
' 50
101.0
101.0
66
66
7.2
100
400
1000
29.2
10'0
400
1000
14.B
30.2
53.5
MIL-C17D
15.4
MIL-C-17F
Black, noncontaminating
PVC jacket.
Black; non_
contaminating
PVC Jacket.
DotJble shielded.
RG-31atU
2077t
26 (7 x, 0067)
4 (solid)
Teflon
High molecufarwelght
2.49
38.1
1,200
300,00D
VDC
polyethylene
50
67
95.2
83.0
69.5
60
100
400
;000
34,1
54.1
101,7
MIL-C-170
Brown, fluorinated
elhylene-propvlene
jacket. High
temperature.
Very high vollage

coaxial cable.
iManufactured by Dielectric ScIences, Inc" Chelmsford, MA.
0
0
0
~,
eo.
~
~
'"d
~,
0
0
'6'
e>~
"
~
&
D
"
is'
OQ
&;
-'"
'"0:
0
"
~
'i'
n
Po
n
!'l-
-"
:::J~~,~,~(;~:,;~-~~:)~(';.:,{i,~\ ';'" <:i.:};~::::.i .~_;.~~~..;~\:~:;~'j$2~'i~~~~:~~J~~:~;~~~~i:~~4:~~.'.iii~~l~i\;_~-i~::~i;t~yil.Mir~~:ig1~t\;:}d~'tJ:$it.fi~~;\~;~~fil~jt~?t~~;,~~;~;$~*jAAI).~.r,ru
r'
tJ>-l
3: 2:
_J
!!l.~
P~
~::::
(fl)
Nominal
cable size
range
(Inches)
eNC
50
SHV
MHV
g-.g
rfl
Maximum
VoUage
Diolectrlc
withstanding
Coupling
style
fraquency
rallng
(VRMsl
voltage
(VRMS)
Governing
specification
0.150 - 0.250
Bayonet
500
1500
MIL-C-39012
Universal, Inexpensive.
"Baby Meill.concelman:
Not constant
0.054 - 0.440
Bayonet
3500
10,000 VDC
MIL-C-39012
NBS-ND-545
Preferred HV conneclor.
Not constant
0.054 - 0.440
Bayonet
3500
6500 VDC
Mll-C-39012
Potentially dangerous.
Nominal
Impedance
Series
Q
o "
(GHz)
Comments
s<o 01
"'...,
o 0
p~~
e~
p."
Avoid rigorously.
"
na
>0
UHF
Microdot
....
...,
Not constanl
0.199 - 0.432
Threaded
0.2
None
500
The original pre-WWII rt

connector. lns'lI:penslve, but
of low quality for signals.
93,50,70
< 0.154
Threaded
750
1000
50eM
LEMO (sarles 00)
K-Loc
50
0.150 - 0.250
Locking
2 (typical)
500
1500
TNO
50
0.150 - 0.250
Threaded
11
500
1500
MIL-C-39012
Essentially a threaded high

frequency BNG.
50
0.350 - 0.450
Threaded
11
1000
2500
MIL-C-39012
Patriarch of the family of rf

connectors. (Born, 1942.
Named for Paul ~eiil.)
HN
C
50
50
0.195 - 0.945
0.350 - 0.450
Threaded
11
Bayonet
1500
5000
1000
3000 (HV type)
3000
NBS-ND-549
MIL-C-39012
MIL-C-39012
..
"'
~a
~.
930 version used frequently

on surface barrier detectors.
~~
Very elegant wl1en used

with RG174/U cable.
High voltage version of type

N rf connector.
Common. High voltage type
useable to 2 GHz.
(Designer: Carl Qoncelman.)
sc
50
0.350 - 0.450
Threaded
11
1000
3000
MIL-C-39012
A threaded C for hostile

environments.
SMA
50
0.047 - 0.250
Threaded
1.2-18
, (some versions
to 27 GHz)
350 - 500
500 -1500
MIL-C-39012
Most widely used

microwave connector.
Sub-Miniature (A), very
high frequency. Maximum
ratinQs depend upon cable.
~ ~
0
S" ~
~
b'
!?So C-
C-0Q
~'"
... :5
~i;g
c::~
en",
>M
en_
o '"
~
0
rtg
~ 2.
ID &l
()
~.o
's<
?
>-l
~
,.,.
0
p-
."
.~
;::
~ ~ ::
I! '!
i ,I
!'C
0
:>
M
"
~.
0>
~
"r
0>
cr
"
~.
~
Chapter 16
16.5 VACUUM TECHNOLOGY
16.5.1 Vacuum practice

All ion beam analyses are conducted in vacuum because
vacuum is required for the production and transport of ion beams
and because surface properties are highly dependent on surface
cleanliness. In addition, vapor-deposited thin films must often be
manufactured in support of the routine practice of ion beam
analysis. A full discussion of vacuum systems is beyond the scope
of this chapter. However, some graphs are included because of
their general applicability to problems arising in the course of ion

beam analysis. For a complete discussion of all aspects of vacuum
technology, see Roth (1982). For ideas on producing atomically
:.... clean surfaces on elemental targets, consult Musket et af. (1982).
Probably the most useful general advice that could be given
about vacuum practice would be to always design systems
conservatively and to aim for the lowest pressure and cleanest
system that time and resources pennit. Here, "clean" refers to the
quantity of bydrocarbon vapors that enter the system from oil
diffusion pumps and mechanical roughing pwnps. If present in
sufficient quantities in the sample cbamber, molecules of these
hydrocarbons can be cracked by the action of the beam while
they temporarily reside on the sample surface_ This leaves
behind a film of nonvolatile fragments that can interfere with
surface analyses or surface modifications_
To avoid this problem, the sample area should be pumped by
an oilfree pump. Ion pumps are a good choice but not
necessarily the best choice for -all sifuations. It is essential to
always remember that the monolayer fonnation time at 10-6 Torr
is about 2 s (see Appendix 18, Fig. AI8.1), so that significant
experiments on surface properties require ultrahigh vacuwn
10-9 Torr). When designing vacuum systems aod internal
components, it is also important to remember that some
materials work well in vacuum and some cause problems. An
example of the latter is cadmium, which has a high vapor
pressure (see AppeJldix 18, Fig. AI8.2). Although not used for
structural purposes, cadmium is commonly used to plate steel
fasteners to suppress corrosion. If these fasteners are heated,
cadmium can be spread throughout the vacuum chamber,
contaminating samples and hardware alike_ Develop the habit of
using stainless steel frulteners exclusively inside your chamber,
and you will a\ioid this problem. If you have a diffusion pumped
system, you can still create a localized hydrocarbon-free volume
with a liquid-nitrogen-cooled shield around the sample. For best
results, use metal-to-metal external seals such as the now
standard "conflat" design or compatible alternatives from other
suppliers_ For exhaus1ive information on the properties of
elastomer seals, consult the manufacturers' literature.
lazy to do the job correctly. Now having said this, we know that
you will do something like this anyway. Therefore, obtain a
small bottle and spray can of Vacseal, as well as a tube of a
solventless epoxy such as Torr-Seal from Varian while your
system is still leakfree so that, if you must cast good vacuum
practice to the wind, at least your crude patches will work. You
should also know that, if you must glue something that will be
used in an ultrahigh-vacuum system, the thennal-curing epoxies
H2ID (electrically conducting) and H72 (electrically insulating)
from Epo-tek are useful alternatives to Torr-Seal.
16.5.2 Thin fIlms
Thin filins for use in ion beam work can be produced by
electrochemistry, eIectron-beam evaporation, resistiveheater
evaporation, sputtering, ion implantation, molecular beam
epitaxy, or laserassisted deposition. They are used in
applications rul diverse as beam stripper foils for tandem
accelerators, entrance windows for gas detectors, and
occasionally even exit windows for bringing ion beams into the"
air to perform analyses on samples that are incompatible with
vacuum. The classic reference on thin films is Maissel and
GIang (1970), but more recent comprehensive reviews are also
available (Schuegraf, 1988). Much recent work h;'" been
undertaken in conjunction with the development of
semiconductor technology and is covered in conference
proceedings such as those edited by Poate ef al. (1978) and
Lucovsky (1988). The most common technique for making thin
films is by evaporation. An extensive listing of the physical
properties of compounds, including their vapor pressures, was
compiled by Graper (1990) to assist in this task. This table is
presented in Appendix 18 as Table A18.1.
Several factors must be considered in selecting a foil for a
specific purpose. These include the required thickness;
constraints on the foil's composition such as the likelihood of
nuclear reactions; the area, if the foil is self 5uppOrting; the
differential pressure that the foil must support; and the level of
radiation that the foil will experience. Beam stripper foils
mounted in the terminal of a tandem Van de Graaff accelerator
are universally composed of carbon and are usually about 3-5
~glcm2 thick. At the other extreme are vacuum-system exit
windows, which must support atmospheric pressure on ODe side
and yet permit an ion beam to penetrate. In the early days of
nuclear physics, these windows were almost always made of Ni.
Now, however, AI, Ti. and the radiation-resistant organic
material kapton are also used_ The required thickness depends,
of course, on the area of the window, but it is typically a few
mglcm 2. Ti.is also commonly used for exit windows in highcurrent electron accelerators that manufacture plastic films by
radiation-induced polymerization_ An analysis of the mechanical
behavior of thin films subjected to a differential pressure was
published by Spivack (1972).lfthe required thickness ofa fi1rn
A final observation: When your vacuum system leaks, as it

will eventually, find the leak and fix it properly. Do not use wax,
glue, or other temporary measures because you are too busy or
418
is incompatible with the area that must be covered, it is often

possible to mOlIDt the film on a high-transmission mesh for
support.
of X per second, then the probability Pen) of observing exactly n

events in a time t is given by
One. of the most serious complications that arises in the

production of thin films by deposition is the phenomenon of
islimding (Feldman and Mayer, 1986). If deposited atoms are
mobile on the surface, they can coalesce into noncontiguous
clumps called islands that can be much thicker locally than
would be assumed on the basis of the average coverage of the
surlace as a whole. The presence of islanding can be detected by
high-resolution backscattering by. for example, insufficient
shifting of the thick-target edge attributable to the atoms of the
substrate. See also Chapter 15, Fig. 15.8(b). Islanding can
diminish the value of submonolayer films intended for use as
backscattcring energy-resolution standards. For better results,
use the thiCk-target edges of monoisotopic elements in this
application.
Pen)
(16.3)
nl
where m = A.t is the mean number of counts expected in the

interval.
The mean and variance of this distribution are defined as
m= InP(n) and cr' = I(n -m)'P(n)

n=O
(16.4)
n_O
According to a mathematical result known as the central limit

theorem (Mathews and Walker, 1970), if m is large, then P(n)
assumes the approximate form
e{n-m)~
Pen)
16.6 STATISTICS AND DATA HANDLING
!2a 2
(16.5)
,jZncr'
Thus, for large m, the probability of obtaining exactly n

counts is distributed normally with variance cr 2 = m. This is an
extremely important result, because it provides the rule for
assigning the statistical error to a measurement. For example, if
the number of counts measured in a backscattering experiment i:i
100, then the standard deviation (J =
=10. or 10%. If a
1% result is desired, then the required number of counts is 10".
16.6.1 Counting statistics

The data obtained in the course of ion beam analysis of
materials have intrinsic variability, and this is reflected.in the
uncertainty in the [mal result of the analysis. Errors come in two
JlOO
classes, systematic and random. Systematic errors are errors that

affect each data point in the same way. Examples are improper
calibration of a multichannel analyzer or other instrument, a
Because one way to obtain more counts is to run an experiment
poorly designed Faraday cup that suffers from significant
longer and because achieving one more decade of precision
secondary electron emission, an uncertainty in the thickness of a
d'
.
f fl
requires 100 times more effort, a certain tyranny is implied by
e '1
b
...
.
l
.
car on lOI use m a tune-o - 19ht spectrometer system, a .
- this result. Usually, It IS satISfactory to .obtalll about 10 counts m
poorIy constructed aperture de fin mg th e solid ang Ie subtend ed
.
()
...
.. ~
b . d
I
drifts m
. ~eIectromc - - the spectral feature of mterest. The resultIng 3 % .preCISIon
IS as
b y a SWlace- arner etector ong-tenn
..
.
h
tak ' .
d h
good as or better than most other sources of error m a typiCal Ion
Instruments t. at are not
en mto accoUD, an . uman errors
beam ex eriment
such as readmg the wrong scale on a meter WIth two range
p
scales. Some systematic errors can be eliminated by proper
16.6.2 Error propagation
design of an experiment, and some cannot. Careful attention to
The [mal result of an experiment is usually obtained by
detail usually minimizes the former errors. If the systematic
mathematical manipulation of individually measured quantities.
errors in a measurement are small, the measurement is said to be
For example, the true number of counts in a peak is the
accurate.
difference between the total measured counts in the region of the
peak. and the background upon which these are superimposed.
Most of this section is focused on a discussion of random
errors. Although these can include such things as errors in
To obtain the correct result, you must subtract the background.
judgment in reading analog meters, line voltage fluctuations,
Alternatively, one might wish to compute a cross section
mechanical vibrations, and other small disturbances, we confme
obtained by dividing a measured number of counts by the
our attention to the most important random error in ion beam
product of the target thickness, integrated beam current, and
analysis: counting statistics. Experiments with small random
detector solid angle, each of which is itself uncertain. How good
errors are said to be precise.
is the flnal answer? To find out, you need to consider the subject
of
error propagation.
The counting of events generated by quanta of radiation is
central to all ion beam analyses, and the fundamental statistical
property of these events is that they occur with constant
probability per unit time. As a result, they obey Poisson
statistics, which is to say that, if the events occur at a mean rate.
419
Rules for error propagation can be found in standard texts

that discuss data handling, such as that of Bevington (1969).
They are obtained by observing that, if z = f(x,y), where x and y
are random variables and f is some function, then
-,
Chapter 16
i'if
i'if
(16.6)
dz =-dx+-dy.
Ox
iJy
The quantities dx, dy, and dz are then interpreted as

differences of individual measured values from the mean, and
the expression is squared and averaged over the complete data
set For example, dz ~ oZj = Zj - (z), and similarly for dx and dy.
Here, the angle brackets, 0, imply the average value of the
quantity enclosed and i denotes the ith data poinl Thus, because
(Z; - (z'" cr/ (and similarly for x and y), ODe obtains
(16.7)
It is clear from this example that, whenever possible, steps

should be taken to physically reduce background rather than
relying on manipulation of data after the experiment to subtract
it.
(16.8)
cr,,' =(Cx-(X)(y-(y)).
If (Jz2
0, then the errors in x and y are said to be
uncorrected. In charged-particle spectrometry, errors are
usually either known to be uncorrected or are assumed to be.
Some counterexamples are described in Nicholson (1974).
=;;
Table 16.12 lists the rules for propagating eITors for several
specific functional fonns of f. Notice, in particular, the form z
= x. - y, which is used for background subtraction of y counts
from a total ofx. In this case, (IE' = OJ,: 2. +
= x + y. Thus, if a
peak contains 400 total counts of which 100 are background,
0'/
If the complete probability clislributions P,(x) of< and P,(y)

of y are known, then the distribution of values of 1. = f(x,y) can
be obtained from
P(z) = JIp,(x) P, (y) li[z -f(x,y)] dx dy
(16.9)
where Sex) is the Dirac delta function and the integral covers ali
of (x,y) space.
The problem of dead time was discussed in the section on

multichannel analyzers (Section 16.4.6) and in conjunction with
pileup. Here, we quantify the discussion. Suppose that).. is the
actual rate of events and that )..' is the measured rate in the
presence of dead time. Let the dead ti,me per pulse be 'to Then, in
1 s, the total_ dead time is A.'or. The nwnber of COWlts missed
during this time is just the true rate times this interval. Because
the number of missed counts in 1 s can also be expressed as A."A', we have
A-A' =U',.
A'
Function
z= axy
Uncorrelated errors
cr ,
cr
cr '
-'-=-' +-'z1
Xl
yl
ax
cr
. .
. ,
1
cr
cr
cr
....!... =-'-+-'-+2--!!,
z
cr
cr
The problem of dead time in pulse coDversion is made worse

by the randomness of pulses (England, 1974). Accurate counting
can occur at a much higher rate if pulses are regularly spaced.
Some advanced systems have been designed to regularize the
rate by buffering the analog pulses until the ADC can accept
them, but this is not common practice. If yOll have measurable
dead time, you should simply plan to make the above correction.
lfthe dead time exceeds about 5%, it is excessive and should be
reduced by, for example, reducing the ion beam current.
xy
=ax,t,b
xy
~ .. b'~
!.'
z= aeb:c.
(16.11)
Notice that, for a measurement that takes time t to accumulate

n' counts, the rate is ')..' = n'/L Therefore, the total dead time is
given by,,', = n'xlt. Because T is the (mean) dead time per couat,
one can see that )..'t is just the fractional dead time of the
analyzer, which is, of course, equal to the fractional number of
missed pulses obtained using the pulser technique discussed in
Section 16.4.6.
....!...=_-+---2....!L
,
z=-
Correlated errors
A=-- or A'=--.
I-A',
1+ A,
7.5% error.
Table 16.12. Error propagation formulas for some simple
functions of the random variables < and y. The parameters a and
b are constants.
(16.10)
From this expression, it follows that
one obtains a net couat of300 with an error of..f500 or about
16.6.4 Random coincidences

Another place' in which the randomness of pulse arrival has
important consequences is in time-of-fligbt measurements. For
low count rates, the start and stop pulses are almost all
In(bx)
420
I!
!
16.63 Dead-time correction

where
iI
i
I
I1
I
I!
L
correlated. However, as the count rate increases, events overlap.

For example, a fast particle can overtake a slower ODe that has
initiated the start pulse and cause a premature stop pulse.
Similarly, a start pulse can be initiated just before an earlier
particle generates a stop. thus producing an event of impossibly
short duration. Before considering the specific problem of time
offlight spectrometers, let us consider more generally the rate of
random coincidences that accompany the measurement of true
events in a coincidence experiment.
Suppose that the rate of events in one detector is At and that
the rate in another is 1...2 and that each of these values i~
somewhat less than the true rate, J.., of correlated events because
of detector efficiency or other experimental considerations. Let 't
be the resolving time of the coincidence apparatus. From Eq.
(16.3), the probability of at least one event in detector 1 dnring
the time x is PI = 1 - exp(-hl't), and the probability of at least
one event in detector 2 is P2 = 1 ~ exp(-A2't). The probability of
2
a coincidence event during time 't is thus P = P1P2 :::: AIA.2't In
time T, there are TIT: such intervals, so that the total number of
random events is B:::: Ali\.lTt. Now, in many situations, one can
assume that, AI:::: 1...2 :::: A, so that B:::: A2T't. However, AT is just
the total number of truly coin~ident events. Ther~fore, one finds
that B/(AT)::::: A't, which says that the ralio ufbackgroulld to true
counts goes up with the rate, 1..., of events. In other words, the
signal-ta-noise ratio in coincidence ex.periments (such as time
of flight measurements) improves as the beam current, and
therefore the event rate, is reduced. Obviously a measurement
entails a compromise between data quality and running time.
Time-of-flight spectra can be significantly altered by random
coincidences between start and stop detectors at high counting
rates. It is possible to compute this effect if one makes some
assumptions about the apparatus (Weller, 1993).~Here, "we
present a fonnula applicable to a spectrometer such as the one
shown in Fig. 16.6, which bas distinct start and stop detectors.
We assume that the true rate of particles entering the detector is
A; that the nonnalized intrinsic distnbution of flight times t is
f{t); and that the probabilities of triggering the start and stop
dete,ctors (the intrinsic efficiencies) as a function of flight time
are Th(t) and 1')2(t), respectively. We further dcfinc the -mean
intrinsic efficiencies of the detectors to be
'1, =
J'1' (t) f(l) dt
time t in terms of the intrinsic distribution f(t) and the true

event rate A.. This fonnulation includes the effect of random
pulses in the stop detector. The expression is
(16.14)
Here, p is the probability that the system is not busy with a

previous pulse at the time that a start pulse arrives. Note that this
equation predicts a component of additive random background
proportional to A and a multiplicative exponential damping
factor e -I'hlt that tends to attenuate the spectrum at Long arrival
times. In time-of flight backscattering spectra, this attenuation is
ordinarily negligible in the portion of the spectrum that has
features of interest and so can be ignorcd. Moreover, the
principal constituent of the additive background, at least for the
. commDn case ofbackscattering analysis of heavy constituents on
a lighter substrate, is constant and can be measured easily by
examining regions of the spectrwn that do not contain
kinematically
allowable
backscattering
events.
Thus,
background subtraction can be done quite easily. Notice too that,
as the rate of events is increased, the background increases
linearly. This is in accordance with the expectation gleaned
above by considering a simple coincidence apparatus. Note,
[mally, that, even at low counting rates, it is necessary to
measure the quantity 1'11(t) 1'12(t) in order to recover f(t) from the
measured spectrum. This is a disadvantage of timeofflight
spectrometry when _compared with the near unit quantum
efficiency of energy spectrometry using surfaccbarrier
detectors.
Equaticin -l16.14) cin be~coiiyert~d iiit; a-di'fThrenti;U ~~-;tion
and formally solved to obtain the best achievable approximation,
'1,(t) ~2(t) f(t), to the true distribution, f(t), from the measured
distribution P(t). The result is interesting but not useful for
practical work, because, in a real experiment, one does not
usually sample the distribution for a large enough range of t
values. The fonnal solution ofEq. (16.14) is
~,(t) '1, (t) f(t) ~ p-'e"'" J[~,)"P(t') + P'(t')] dt' ,
where P' denotes the derivative: Using the simple a!-iswnption

that the time-<lffught spectrometer's mean dead time, 1', is the
me~ duration of observed events, one obtains for 't
and
'1, = J'1,(t) f(t) dt.
(16.15)
(16.12)
(16.13)
JtP(t) dt
With these definitions, it is possible to express the

mathematical form of the probability, P(t), that an ion entering
the timeof-flight spectrometer generates an event with flight
fp(t) dt
o
421
(16.16)
Chapter 16
and, as in Section 16.6.3, for the Jive-time factor p

(16.17)
The random coincidence background sets a lower limit on the
sensitivity of time-of-flight measurements, but it should be Doted
from Eq. (16.14) that this background depends linearly on the
rate of events Tt2A, in the stop detector. If a significant portion of
these counts is a result of uninteresting, low-energy particles,
then the signal-to-noise perforqlance of a spectrometer can be
innproved by placing a range foil (see Chapter 5) just before the
-~ stop detector. In this location, the foil causes negligible loss of
timing resolution for the particles that pass through it, and pin
holes in the foil produce only edditional background, not
distortions in the spectrum.
16.6.5 Advanced data manipulation

The Dearest thing to a cult classic in the literature of statistics
for scientists is the set of notes of Orear (1958). which he based
on unpublished work of Enrico Fenni. Most of the essential
contents of these notes. as well as many other topics, are now
contained in the chapters on statistical handling of data in the
book by Press et al. (1992). This
reference is particularly valuable becau.."e it provides executable
C-language functions. FORTRAN and Pascal editions of the
book are also available. For most analytical tasks, curve fitting is
the most appropriate approach to reduce ion beam data. The
routines in Press el al. (1992) have been designed to propagate
errors properly, even through complicated nonlinear fitting
procedures, and they are the preferred approach eveD for such
mundane tasks as background subtraction. Functions are also
_. presented for performing sophisticated statistical tests on data
'0_0 sets, as well as such useful procedures as Fourier transforms (by
the fast Fourier transform algorithm).
The reader is cautioned, however, to use these tools with
discretion. For example, it is sometimes desirable to "smooth"
data for presentation. There are many ways to do this, but our
recommendation is to usc the procedure of Marchand and
Marmet (1983) or, perhaps even better, to perform a Gaussian
coiIvolution. The Gaussian function is well known to have the
same functional form in both real and reciprocal spaceo As a
result, it is easy to perfonn a convolution by simply taking the
Fourier transform of the data, multiplying by the transfonn of a
Gaussian with a width of a few channels, and (Fourier) inverting
the product. We find this result superior to the smoothing
procedure suggested in Press el al. (1992).
function based on an analysis of your experiment and fit your

dnta using the procedures presented in Press el al. (1992). This
approach is stable and wiJIlead to satisfactory results if you
have developed a good model.
REFERENCES
Althaus, R.F., and Nagel, L.W. (1972), IEEE Trans. Nucl. Sci.
NS-19 (1),520.
Betts, R.R. (1979), Nucl. Instrum. Melhods 162, 531.
Bevington, P.R. (1969), Data Reduction and Error Analysisfor
the PhySical Sciences. McGraw-Hill, New York. A standard
introductory text on the subject of data analysis.
Blankenship, J.L. (1960), Trans. Nucl. Sci. NS7 (2-3), 190.
Bromley, D.A. (1974), Nucl. Instrum. Methods 122, I. See also:
Bromley, D.A. (1984), "The Development of Heavy-Ion Noelear
Physics'\ in Treatise on Heavy-Ion Science (Bromley, D.A.,
ed.), Plenum, New York, voL I, p. 3. A higbly readable
historical article by a person with broad knowledge of both the
field and the personalities who made it.
Bromley, D.A. (ed.). (1979), "Detectors in Nuclear Science',
NucJ. lnsfnlm. Methods 162. This entire issue is devoted to
invited reviews of detector technology. The articles contain
much valuable technical and historical infonnation.
Brown, I.G. (1989), The Physics and Technology ofJon Sources,
Wiley, New York.
Bryant, J.H. (1984), IEEE Trans. Micruwave Throry Tech.
MTI-32 (9), 970. This issue also contains a number of
fascinating review articles on the history of microwave
electronics. See also: Maury, M.A. (1990), 'Microwave Coaxial
Connector Technology: A Continuing Revolution", in 1990
State of the An Reference (a special issue of MicrowuYe
Journal), Horizon House, Norwood, MA, p. 39.
Campbell, J.L., and McNelles, L.A. (1975), Nucl. Instrum.
Methods 125, 205.
Canberra. (1978), Canberra Laboralory Manual, Canberra
Industrics, Meriden, CT.
Clout, P. (1982), A CAMAC Primer, Report LA-UR-82-2718,
Los Alamos National Laboralory, Los Alamos, NM ThiS
document can also be obtained from a CAMAC module
manufacturer,
KineticSystems
Corporation,
bttp:llwww.kscorp.comIProductsicamacJ.
While on the subject of convolutions, and in closing, we issue

a warning: Be very careful about trying to "deconvolve" data by
manipulating the Fourier transform. The process has practical
difficulties and should be avoided. Instead, create a fining
422
0,0;.
_'~;iil!
D5beli, M., Haubert., P.C., Livi, R.P., Spicklemire, SJ.,

Weathers, D.L., and Tombreilo, T.A. (1991), NucJ. Instrum.
Methods B56--57, 764.
Duggan, 1.L. (1988), Laboratory Investigations in Nuclear
Science, Application Note TN-2000, Tennelec-The Nucleus,
Oak Ridge, TN. Tennelec, now a division of Canberra

Industries, is also a significant supplier ofNLM instnunentation.
Elmsley, 1. (1989), The Elements, Oxford University Press,
Oxford, UK.
England, 1.BA (1974),. Techniques in Nuclear Structure
Physics, Halstead Press, New York. A two-volume treatise on all
aspects of experimental nuclear physics. Very useful reference at

an advanced level.
Humphries Jr., S. (1986), Principles of Charged Particle

Acceleratjon, Wiley, New York.
Jamieson, D.N., Grime, G.W., and Watt, F. (1989), Nuel.
Instrum. Methods B40-41, 669. Other articles on this subject are
found frequently in this and other proceedings of the Conference
on the Application of Accelerators in Research and Industry,
which has been held biennially at the University of North Texas
for many years. For a number of years, these proceedings have
been published as special issues of Nuclear Instruments and
Methods in Physics Research.
Jardine, LJ., and Lederer, eM. (1974), Nucl. instrum. Methods

120,515.
York, p. 712.
Knoll, G.F. (1989), Radiation Detection and Measurement, 2nd

Ed., Wiley, New York. An essential reference for persons who
work with radiation.
Feldman, L.e., and Mayer, J.W. (1986), Introduction to Surface

and Thin Film Analysis, North-Holland, New York. p. 136.
Lawson, J.D. (1977), The Physics of Charged-Particle Beams,

Oxford University Press, Oxford, UK, p. 172.
Gabor, G., Schimmerling, W., Greiner, D., Bieser, F., and

Lindstrom, P. (1975), NucJ. Instrum. Methods 130,65.
Lederer, CM., and Shirley, V.S. (ed,.). (1978), Table ofIsotopes,

7th Ed., Wiley, New York.
Graper, E. (1990), Thin Film Handbook, Lebow Corporation,

Goleta, CA.
Lennard, W.N., Geissel, H., Winterbon, K.B., Phillips, D.,

Alexander, T.K., and Forster, 1.S. (1986), Nucl. Instrum.
Methods A248, 454.
Evans, R.D. (1955), The Atomic Nuelem, McGraw-Hill, New
Harting, E., and Read, F.H. (1976), Electrostatic Lenses,

Elsevier, New York. Includes many figures and tables for
persons who need to design electrostatic _optics ..

Heath, R.L. (1964), Scintillation Spectr~';'e'irr--<;~';';"a ROy
Spectrum Catalog, Report IDO-16880-1, U.S. Atomic Energy
Lucovsky, G. (ed.). (1988), Deposition and Growth: Limits for

Microelectronics, American Institute of Physics, Conference
--Proceedings No. 167, New York. _
.
Maissel, L.l., and Giang, R. (1970), Hondbook of Thin Film
Technology, McGraw-Hill, New York.
Commission Idaho Operations Office Research and

Development, Idaho Falls, ID. This reierence is dated, but it
contains an instructive introduction that is interesting from a
historical perspective. It is the origin of the often-<:opied figwe
shown in this chapter as Fig. 16.4.
Marchand, P., and Marmet, L. (1983), Rev. Sci. lnstrum. 54,

1034.
Marion, J.B., and Young, F.C. (1968), Nuclear Reaction
AnalYSis, Graphs and Tables, North Holland, Amsterdam, p.
Helmer, RB., Van Assche, P.H.M., and VanDer Leun, C.

(1979), At. Data Nucl. Data Tables 24, 39.
145. A classic handbook for experimental nuclear physicists.

Has many tables (some out of date, of course), including a listing
of accelerator calibration reactions.
Holl, I., Lorenz, E., and Mageras, G. (1988), Trans. Nucl. Sci. 35
(1), 105.
Mathews, J., and Walker, R.L. (1970), Mathematical Methods of

Physics, 2nd Ed., WA Benjamin, New York, p. 383. Graduate
text in mathematical physics with an introduction to many useful
topics in statistics and other areas.
Horowitz, P., and Hill, W. (1989), The Art of Electronics, 2nd

Ed., Cambridge UDiversity Press, Cambridge, UK. An essential
text and reference for physical scientists. Readable, wideranging, and full of practical experience and insight
McKenzie, J.M. (1979), Nucl. Instrum. Methods 162, 49.
423
Chapter 16
Mendenhall, M.H., and Weller, R.A. (1991), Nucl. Instrum.

Methods B59-60, 120.
Moore, I.H., Davis, C.C, and O>plan, M.A. (1989), Building
SCientific Apparatus, 2nd Ed., Addison Wesley, New York.
Chapter 5, p. 305. An essential text and reference for students of

experimental physical sciences.
Progress in Nuclear Techniques and Instrumentation (Farley,

F.I.M., ed.), North Holland, Amsterdam, pp. 1-116. A classic
article to introduce serious students to acce1erators.
Roth, A. (1982), Vacuum Technology, 2nd Ed., North Holland,

Amsterdam. Excellent reference on all aspects of vacuum
technology,
Morrison. R (1977), Grounding and Shielding Techniques in

Instrumentation, 2nd Ed., Wiley, New Yolk. A highly readable
little book on a subject that might at first glance seem to be very
dull. Recommended for anyone who is serious about making
Rytz, A. (1979), At. Data Nucl. Data Tables 23, 507.
precise electrical measurements.
Schuegraf, K.K. (ed). (1988), Handbook of Thin-Film

Depqsition Processes and Techniques, Noyes Publications, Park
Ridge, NJ.
Musket, R.G., McLean, W., O>lmenares, C.A., Makowieclci,

D.M., and Siekhaus, W.I. (1982), Appl. Surf Sci. 10, 143.
Sakai, E. (1988), Trans. Nucl. Sci. 43 (1), 418.
Spivack, MA (1972), Rev. Sci. Instrum. 43, 985.

Nicholson, P.W. (1974), Nuclear Electronics, Wiley, New York.
Specialized volume on nuclear electronics. A good review.
Tub, I.K. (1985), Nuclear Wallet Cards, National Nuclear Data

Center. Brookhaven National Laboratory, Upton, NY.
Orear,1. (1958), Notes on Statistics for Physicists, Publication

UCRL-3417, University of California Radiation Laboratory,
Berkeley, CA, Aug. 13.
USDOE {I 990), Standard NIM instrumentation System, Report

DOElER-0457T, U.S. Department of Energy, Office of Energy
Ortec (1984), Experiments in Nuclear Science. AN34.

Laboratory Manual, 3rd Ed., EG&G Ortee, Oak Ridge, TN.
Nuclear lnstromentation Modules, Report TID20893, Rev. 4
Overley, I.C, Parker, P.D., and Bromley, D.A. (1969), Nuc!.

Instrum. Methods 68, 61. O>ntains reactions that are particularly
useful for calibrating large accelerators.
Welter, R.A. {l993),Nucl.Instrum. Methods B79, 817.
Research, Washington, DC. This document supersedes Standard
Poate, I.M., Tu, K.N., and Mayer, I.W. (eds.). (1978), Thin-Film
InterdifJusion and Reactions, Wiley. New York.
Press, W.H., Teukolsky, S.A., Vetteriing, W.T., and Flannery,
B.P. (1992), Numerical Recipes in C, 2nd Ed., Cambridge
University Press, Cambridge, UK. See especially Chapter 14, p.
609, and Chapter IS, p. 656. An instant classic in the field of
numerical analysis, written by and for practicing scientists. An

essential reference.
Rose, P.H., and Galejs, A. (1967), "The Production and

Acceleration of Ion Beams in the Tandem Accelerator'\ in
424
(1974),
u.s. Atomic Energy Commission.
Wielopolslci, L., and Gardner, R.P. (1976), Nuel Instrum.

Methods 133, 303.
Whitlow, H.J. (1990), "Time-of-ftigbt Spectro,copy Methods for
Analysis of Materials with Heavy Ions: A Tutorial", in High
Energy and Heavy Jon Beams in Materials Analysis (Tesmer,
l.H., Maggiore, C.J., Nastasi, M., Barbour, J.C, and Mayer,
I.W, eds.), Materials Research Society, Pittsburgh, , PA, p. 243.
Wilkinson, D.H. (1950), Proc. Cambridge Philos. Soc. 46 Pt 3,
508.
Wiza, I.L. (1979), Nucl. Ins/rum. Methods 162, 587.
CHAPTER
RADIOLOGICAL SAFETY
P. M. DeLuca, Jr.
University of Wisconsin, Madison, Wisconsin, USA
J. R. Tesmer
P. Rossi
University of Padua, Padua, Italy, and Sandia National Laboratories, Albuquerque, New Mexico, USA
CONTENTS
17.1
INTRODUCTION ............................................................................................................. 426
17.2
REGULATORY AGENClES ............................................................................................. 426
17.3
RADIOLOGICAL QUANTITIES ..................................................................................... 426
17.4
DIRECT EXPOSURE ......................................................................................................... 429
17.5
INDIRECT EXPOSURE ..................................................................................................... 430
17.6
ENVIRONMENTAL IMPACT .......................................................................................... 431
17.7
RADIATION DETECTORS ............................................................................................... 432
17.8
CONCLUSIONS .................................................................................................................. 433
REFERENCES .................................................................................................................................. 434
425
---
Chapter 17
Further, the U.S. Department of Energy (USDOE) provides

the regulatory environment for those institutions that it operates.
USDOE generally adopts the position of USNRC on matters of
radiation control. However, USDOE bas taken positions of
greater conservatism with respect to radiation than USNRC.
Two regulations are of particular interest to operators and users
of ion beam analysis facilities: USDOE Order 5480.25 (DOE
5480.25. 1992), Safety of Accelerator Facilities, and Part 835 of
Cbapter 10 (Energy) of the Code of Federal Regulations
(IOCFR835, 1993), "Occupational Radiation Protection"
(formerly USDOE Order 5480.11) and their associated guidance
documents.
17.1 INTRODUCfION
Although a complete discussion of the many aspects of the
radiation environment encountered near low-energy particle
accelerators is beyond the scope of this chapter, general
information is provided here to guide the user. More detailed
information is available in the references listed at the end of this
chapter.
The following sections discuss the various agencies involved

in regulation and development of standards, briefly review the
more common radiological pafarneters and the typical radiation
environment encountered and its potential hazards, and provide

general guidance in the development of suitable radia~ion
protection for an ion beam analysis facility.
The U.S. Department of Transportation provides the'

regulatory environment for the transportation of radioactive
materials, those materials controlled by both USNRC and State
Agencies. USEPA is also involved in regulating transportation
insofar as it affects the environment occupied by the general
public.
17.2 REGULATORY AGENCIES

Unfortunately, current research is conducted in a complex
and diverse regulatory environment. Frequently, the same aspect

of radiological safety is controlled by two or more agencies, and
some effort is required to determine which is actually the
Although this description is specific for the United States, it

suggests a rather bleak and imposing situation. Readers are
cautioned to carefully detennine who the controlling parties are
for their facilities and resolve any potential areas of conflicting
or overlapping interest.
controlling agency. The following discussion outlines some of

the agencies involved in the control of radiation near
accelerators.
Although special nuclear materials (SNMs), namely, plutonium,
uranium enriched in 233U or 23SU, and tritium,. are not usually
found in most ion beam analysis laboratories, for the purpose of
a more complete overview, they are included in this discussion.
In the United States, the Nuclear Regulatory Commission
(USNRC) bas regulatory control of the usc of all SNMs and all
byproduct materials (i.e., materia1s produced as a result of using
special nuclear materials). The rules and regulations by which
the USNRC controls byproducts and SNMs are contained in
_Chapter 10 (Energy) of the Code of Federal Regulations, Parts 0
through 50. Most of the essential aspects are discussed in Part 20
(lOCFR20, 1993).USNRC does not usually control the use of
radiation-producing apparatuses other than those involved with
the aforementioned materials, nor does it control naturally
occurring or accelerator-produced radionuclides.
Although the regulatory environment is fraught with

confusion and some uncertainty, the scientific community that
provides guidance to the several regulatory agencies does 50 in a
coherent and comprebensive manner. In the United States, the
National Council on Radiation Protection and Measurements
(NCRP), based in Baltimore, MD, was organized in the early
1930s to provide crucial guidance on the proper and safe use of
radiation. As such, the NCRP has promulgated these findings in
a series of reports organized by topic and use. Considerable
information is contained in these brief reports. On an
international scale, the International Commission on
Radiological Protection (ICRP) and the International
Commission of Radiation Units and Measurements (lCRU)
provide similar guidance. All of these agencies coordinate their
activities to provide a comprehensive and cohesive presentation
to the scientific community.
Control of radiation-producing apparatuses and naturally

occurring or accelerator-produced radionuclides is usually
provided by state agencies. The most common of these devices
are radiographic apparatuses used in medical procedures. Some
states, called Agreement States, have assumed the regulatory
activities of USNRC, and those states control radiation
consistent with the regulations of USNRC. Beyond these
regulatory bodies, the U.5. Environmental Protection Agency
(USEPA) regulates radiation in the environment. generally
occupied by the public. Obviously, USEPA control overlaps
USNRC and Agreement State activities.
17.3 RADIOLOGICAL QUANTITIES

Various radiological units and quantities are used to descnoe
the effects of radiation. The radiation found near low-energy
accelerators can be classified as either directly or indirectly
ionizing radiation. Protons and alpha particles are directly
ionizing, whereas X-rays, "(-rays, and neutrons are nol Hence,
except for beams brought into the air for use (external beams).
most of the radiation hazards near low-energy accelerators are
from indirectly ionizing radiation.
426
Radiological Safety
De~pite .several alternate panlineterizations, the most accepted

and well-understood parameter is absorbed dose. As stated by
the ICRU (1992), absorbed dose D in a material is given by
D=~
dm'
(17.1)
where dE is the mean energy imparted by ionizing r.adiation to

matter of mass dm.
The unit of absorbed dose is joules per kilogram and is

denoted as the gray (Gy). Note that absorbed dose is defmed for
a specific material, leading to the usage of referring to the gray
as the "dose in tissue". Absorbed dose is a nODstochastic
quantity. Absorbed dose is the physical result of ionizing
radiation. The unit Gy replaces the older unit rad, where 100 rad
= 1 Gy.
For a better description of the effects of indirectly ionizing
radiation, an additional quantity is defmed, the precursor to the
absorbed dose, called the kerma, an acronym for kinetic energy
released in matter. The ICRU (1980) defines kenna as

K= dEn-
dm'
where D is the absorbed dose and Q is the quality factor at that

point. The unit of dose equivalent is the sievert (Sv), where 1 Sv
= 1 Jlkg. In the United States, an older unit, the rern (roentgen
equivalent in man), is still used. The simple conversion is 1 Sv =
100 rem, or 1 nuem = 10 J.1Sv. Note that dose equivalent and
absorbed dose differ only by the scaling of the biological effect,
and hence, they have the same base units, namely, joules per
kilogram. For X- or r-radiation, absorbed dose and dose
equivalent are essentially equaL For neutrons, the biological
effect varies with the neutron energy. Table 17.1, taken from
Table 1004(b).2 of section 20.1004, ''Units of radiation", of

Chapter 10 (Energy) of the Code of Federal Regulations
(10CFR20, 1993), indicates the variation of the biological effect
for neutrons with neutron energy. The introduction of a
biological effectiveness factor considerably complicates the
detennination of dose equivalent for fast neutrons because
physical measurements are usually associated with energy
absorbed per unit mass. Figure 17.1 shows a plot of the dose
equivalent per .unit fluence values as a function of neutron
energy. This infonnation is useful for estimating the dose
equivalent that would result from a known reaction.
(17.2)
.?'
",'b
where dEli is the sum of the initial kinetic energies of all

charged particles liberated by uncharged ionizing particles in a
volume of material with a mass of dm.
[I[ICI:IJ
[I
100
10~ 103
The unit of the kerma is the same as that of the absorbed

dose, the gray. The kerma per unit particle fluence is called the
kenna factor. Note that the kerma is not directly measurable,
but can be calculated from basic particle interaction
probabilities. ~ Hence, for indirectly ionizing radiation, the
absorption of energy consists of a two-step process: (1)
indirectly ionizing radiation cdnverts some energy to directly
ionizing charged particles (kenna) and (2) the resulting ionizing
charged particlcs impart energy to matter by excitation and
ionization (absorbed dose).
10~ -'0-7 10-- l~ 10.( 1cr 10'~ 10.1
FIG. 17.1. Neutron dose equivalent per unit neutron fluence

plotted as a function of neutron bombarding energy.
As a point of information, the difference in biological effects

among various types of radiation is in large part due to the rate
of energy transfer by charged particles at the microscopic level.
Specifically, an energy transfer rate of about 100 keV/j.lID
Although the biological effects of 1 Gy are largely similar for

all types of ionizing radiation, differing by perhaps less than a
factor of 10, the differences are sufficient to warrant the scaling
of the absorbed dose by a biological effectiveness parameter, Q.

This leads to a quantity called the dose equivaleot. The lCRU
produces the largest observed biological effect OCRU, 1992).

Thus; fast neutrons that produce recoil protons and other heavier
charged particles are observed to have a larger biological effect
than, for example, r-rays, which produce electrons with an
energy transfer rate less than about 5 keV/l1m.
(1992) defmes the dose equivalent (H) at a point in tissue as the

product
H=QD.
10'
Neutrorl Energy (MeV)
(17.3)
427
Chapter 17
Table 17.1. Mean quality factor, Q, and fluence per unit dose
equivalent for monoenergetic neutroDS.
Neutron
energy
Quality
factor*
(Me\')
(Q)
2
2
2
2
2
2
2.5
7.5
11
11
9
2.5 x 10->
1 X 10-7
1 x 10-<
1 x 10-'
1 X 10-4
1 X 10-3
1 X 10-2
1 X 10-1
5 x 10-1
1
2.5
7
10
14
20
40
60
4 xla'
4 X 102
4 x 102
4
10'
7
6.5
7.5
8
7
5.5
4
3.5
3.5
3.5
Table 17.2. Radiation dose limits established by the International

Commission on Radiological Protection (ICRP, 1990)
Occupational Dose Limits
Fluence per unit dose

equivalent'"
(cm-1 Sv-1)
980 x 10'
980 X 10'
810 x 10'
810 X 10'
840 x 10'
980 x 10'
1010 x 10'
170x 10'
39 x 10'
27 x 10'
29 x 10'
23 x 10'
24 x 10'
24 x 10'
7 x lOS
16 x 10'
14 x 10'
16 x 10'
20 x 10'
19 x 10'
16 x 10'
14 x 10'
Stochastic
(whole body)
Deterministic
on leos of eye
Deterministic
on sIQn, bands, and
effects
Ann. Eff-D - 50 mSv and
5-year Cum. Eff-D = 100 mSv
effects
Ann. Eq-D = 150 mSv
effects
feet
Ann. Eq-D = SOO mSv
Public Dose Limits

(elcluding medical and natural background exposure)
Stochastic
effects
(whole body)
Ann. Eff-D = 1 rnSv
Deterministic
effects
on lens and eye
. Ann. Eq-D = 15 mSv
Deterministic
effects
on skin, hands, and feet
Ann. Eq-D = SO mSv
On
embryo-fetus Eq-D
2
rnSv
(woman's abdomen)
-- (during entire pregnancy)
Legend
Eq-D: Equivalent do!';e, defil1ed a" the average absorbed

dose in a specified tissue or organ, weighted by the
radiation weighting factor (WRJ for the specific radiation
impinging on it.
Eff-D: Effective dose (intended to provide a means for
handling Donuniform irradiation), defined as the sum of the
weighted equivalent doses for all irradiated tissues or
organs. The tissue weighting factor (WT) takes into
account the relative detriment to each organ and tissue,
including the different mortality and morbidity risks from
cancer, the risk of severe bereditaIy effects, and the length
of life lost
Cum. Eff-D: Cumulative effective dose.
*Value of Q at the point where the dose equivalent is a

maximum in a 30-cm-diameter tissue-equivalent cylinder
(phantom).
*-':Monoenergetic neutrons incident normalJy on a 30-cindiameter tissue-equivalent cylinder (pbantom).
Ann.: annual.
Whole body: defined as extending from the top of the head
down to just below the elbow and just below the knee
In Tables 17.2, 17.3, and 17.4, we report the dose limits

required by various agencies, regarding only radiation exposures
and not taking into account the intake of radioactive material.
Different facilities establish their own limits that, in the United
States, cannot be less restrIctive than the USDOE limits. These
tables employ the quantity "Equivalent Dose" that slightly
differs from the "Dose Equivalent" previously introduced.
428
Radiological Safety
Table 17.3. Radiation dose limits* established by the National

Council on Radiation Protection and Measurements (NCRP,
1993)
Stochastic
(whole body)
Ann. Eff-D - 50 roSy and

Cum. Eff-D = 10 roSy x
age (in years)
Ann. Eq-D = 150 roSy
effects
Detenninistic effects
on lens of eye
Deterministic effects
on skin and extremities
Ann. Eq-D = 500 roSy
Public Dose Limits

(excluding medical and natural background
exposure)
effects
Ann. EfI-D - 1 roSy for
Stochastic
(whole body)
Deterministic
effects
on lens and eye

Detenninistic effects
on skin and extremities
On
embryo-fetus
(woman's abdomen)
For a particle flux density incident upon skin, the absorbed

dose rate is given by
/) =.S",Qx1.602xIO-'" Gyfs
(17.4)
where mStiss is the mass collision stopping power for the

particles of energy E in tissue in units of MeV cm2 g-l and $ is
the Charged-particle flux density in units of cm-2 5- 1 Figure
17.2 shows a plot of mStiss for tissue bombarded by protons and
a-particles with an energy of less than I 0 MeV. In a similar
manner, Fig. 17.3 shows a plot of dose versus charged-particle
energy nonnaiized to a tluence rate of 6.24 x 1011 S-I,
2
corresponding to a beam current of 1 nA with an area of 1 mm
The mStiss values were taken from the tabulations of Andersen
and Ziegler (1977). Such energy absorption rates are sufficient
to cause massive skin damage, albeit local to the surface of the
skin.
continuous ex.posure and

Ann. Eff-D = 5 roSy for
infrequent exposures
Ann. Eq-D = 15 mSy
4000
1000
/
800
"m
N
E 600
.. ....
400
/
/
3200
....
>
Month Eq-D = 0.5 mSy
""\
Ann. Eq-D = 50 mSv
"
", ,
.-
24<10
1600
200
,,
aoo
"'See legend of Table 17.2 for deftnitions of abbreviations.
Table 17.4. Radiation dose limits* established by the U.S.

Department of Energy (10CFR20, 1993)
Whole body'
Ann. EfI-D - 50 roSy
Lens of eye
Extremities,
other organs
skin,
and
10
10
10'
Particle Energy (MeV)

FIG. 17.2. Mass collision stopping powers for protons and
alpha particles as a function of particle energy for interactions
.wIth muscle tissue. . _. .__ _ .__
__ _ .
.
-. -- .-----~""-.>~--..~--~~ ...- ~.-..:--~--.--
Ann. Eq-D = ISO roSy

Ann. Eq-D = 500 mSy
100
/
Dose Limits for Facilities

(visitors and public)
Whole body
Ann. EfI-D - 1 mSy
Embryo-fetus
(woman's abdomen)
10-1
10.2
"e
80
"'See legend of Table 17.2 for deflnitions of abbreviations.
/
/
/
20
"'"
\
\
..
........
F1u.ln>rnlllAlnl~
0
10-2
17.4 DIRECT EXPOSURE

Exposure to the charged particle beam poses an extreme
hazard because these particles are directly ionizing. This is one
of the major hazards for researchers using external beams. The
specific hazards aSsociated with partide-induced-X-ray
emission (PIXE) in air are discussed by Doyle et al. (1991).
/
/
.0
; 4<1
320
60
400
\
/
/
Eq-D
5
mSy
during entire pregnancy
,,-, \
10-1
10'
,,
240
160
",
so
0
10'
FIG. 17.3. Absorbed dose rates for protons and alpha particles
<l;S a fimction of particle energy for interactions with m~cl~
tJss~e. Values correspond to a partJ.cle flux rate of 0.6 x 10 scm .
429
.-
Chapter 17
Whereas the energy absorption rate is very high for these

directly ionizing particles, the range ,in tissue for particles
below 10 MeV energy is modest. Figure 17.4 shows a plot of
proton and alpha-particle range values in tissue for similar
particle energy ranges. the particle range values were
determined from integration of the same data as used for the
stopping powers (Andersen and Ziegler, 1977). Although not
affording complete protection, the outer dead layer of the skin
(the epidermis) is about 50 ~ thick (ICRP, 1974),
corresponding to the range of approximately 1.5 Me V protons
and 6.4 MeV u-particles.
Clearly, the hazard from direct exposure to a charged
17.5 INDIRECf EXPOSURE

For many analyses, the incident particle energies are
typically a rew megaelectronvolts and generally below the
nucleon binding energy. Therefore, the potential hazard from
neutron ~xposure is often assumed to be small. This is not
necessarily the case. Significant quantities of neutrons are
generated at low particle bombardment energies when low-Z
projectiles strike low-Z targets. Some of the important neutronproducing reactions below 5 Me V particle bombarding energy
are listed io Table 17.5. A program by Drosg was used to
estimate these values from thick isotopically pure targets
(Drosg, 1992).
Table 17.5. Various reactions that produce significant

quantities of neutrons for bombardment by protons and
deuterons below 5 MeV. Values ofthick-targct neutron yield Y,
mean neutron energy Ea , and neutron dose equivalent rate H at
0 and 90 with respect to the bombardment direction are listed.
particle beam is eliminated by the introduction of a modest

thickness of material to stop the beam. Moreover, direct-
exposwe damage is confined to superficial layers of the skin,

with the greatest hazard being exposure of the eyes.
AlplUl
102
,-
j
~
'"
c~
'"
la'
.- .-
10
10.1
,.- .-
.- ,-
,-
,-
10
,- .-
,-
,-
Y
{sr-l I!C- I }
Reaction (MeV)
'H(p,n)'He
'H(d,n)'He
2.2
0
5.93
2.26
90
3.35
0.27
'H(d,o)'He
2H~t,ntHe
0.59
0.43
0.35
0.40
17.8
E.
(mSv h-l
(MeY)
0
1.2
5.24
18.3
18.7
90
0.42
2.95
14.8
14.4
nK')
0
0.64
0.24
90
0.30
0.03
0.10
0.08
0.06
0.07
Typically. these reactions are not deliberately used, but they

might occur inadvertently. For example, a sample bombarded
with deuterons will embed deuterons at the depth of their range,
and because hydrogen readily diffuses in many materials,
continued bombardment will produce neutrons from the 2H (d,n)
reaction. Increasing the bombardment energy will produce an
even more prolific source of neutrons as the bombarding
deuterons can reach the embedded deuterons with sufficient
energy to have a significantly increased interaction probability.
Also. mixing beams of deuterons and tritons on the same
sample will produce substantial amounts of 15 Me V neutrons.
Note that the "sample" discussed here could ioclude the beam
collimation apparatus or any oilier components exposed to the
beam! In fact, the 2H(d,n) reaction can be a significant source
of neutrons in older ion beam analysis facilities and can
eventually curtail (if unshie1ded) the use of deuteron beams for
analysis. Baking out the componcnts with the embedded
deuterium can reduce this problem.
la'
FIG. 17.4. Ranges of protons and a-particles as a function of

particle energy for muscle tissue.
, For example, the NCRP recommendation for superficial
exposure is 0.5 Sv year-I (NCRP, 1987), which can be produced
by a fluence of 1.2 x 10.5 mm-2 of 1 MeV protons. This fluence
can easily be achieved from protons backscattered at 1200
during bombardmeo1 of a IO-~-thick gold layer with 200 nC
of protons at a 10-cm target-to-skin distance. The exposure
time would be 2 s at a 100 nA beam current!
Although the most important danger of direct exposure to
cbarged particles is macroscopic organ damage, specifically,
damage to the skin, stcx:hastic effects are stiU possible, with
skin cancer being the most dangerous. Such a possibility would
become pronounced at exposures in the range of 0.1-3.0 Gy,
corresponding to fluences of between 2.4 x 10 4 mm-2 and 71 x
4
10 rom-2 of I Me V protons. However, exposure to external
beam.o; of vastly higher particle fluences and energies ate more
likely and are hence the principle concern.
Another significant source of neutrons is produced by

deuteron breakup on beam transport equipment and samples
when the deuteron bombarding energy exceeds the deuteron
binding energy in the center-of-mass frame, that is, a 2.2 MeV
laboratory energy for heavy targets. Fortunately, most analysis
reactions use lower bombarding energies.
430
Radiological Safety
Elastic resonance reactions for the analysis of light elements

with alpha particles or protons might also produce neutrons. As
the beam energy is increased above the Coulomb barrier (see
Appendix 8), the production of neutrons and gamma rays
increases dramatically for light targets. Data for (a,n) reactions
up to 10 MeV are given in Appendix 21, and those for (P,n),
(P,Y), and (d,n) reactions are in Appendix 20. Both neutrons and
gammas from higher-energy proton beams might require the
use of shielding. Fortunately, mostuseful proton resonances are
at lower energies. Nevertheless, in these cases, a knowledge of
what you are about to analyze is important for preventing
radiation exposures. In addition, the sample materials could
become activated with short-half-life beta emitters (or internal
conversion gammas). These are produced on purpose in
activation analysis (see Chapter 10), but are not welcomed by
the unknowing experimenter. These radionuclides can be
produced by (p,n), (a,p), and (a,n) reactions. For example,
"Ti(a,n) yields "Cr with a 28-day half-life. Hence, the sample
with Ti in it thus becomes a radioactive source with a relatively
long -half-life and must be treated as such. Another common
example is 29.3Si(a.,p) 32J3 p with half-lives of approximately 14
and 25 days, respectively. It is good practice to monitor for
radioactivity all samples that have been irradiated with
deuterons or any particles with energies above the Coulomb
barrier of the elements in the sample.
kept small to reduce costs. so that more radiation (essentially

neutrons) is emitted in the vertical direction.
. As the threshold of neutron-induced nuclear reactions with '
the air elements (N, 0, Ar) lies near or above 20 MeV,
interactions below this energy are restricted to elastic
scattering. The high-energy nuclear interaction lengths for N2
2
and O2 are on the order of 90 g cm- , whic~ for the density of
air (1.2 x lO" g cm-'), corresponds to 750 m, and hence, the
high-energy neutrons effectively escape to great distanceS_
Because only low-energy neutrons can be backscattered, near
the source, only these neutrons are important, whereas highenergy neutrons will predominate at distances comparable to
the nuclear interaction length. Because they are lighter than the
nitrogen and oxygen nuclei in air, the neutrons lose their energy
only after many elastic scatters, and the primary reduction in
fluence out to a few hundred meters mainly derives from
geometrical factors. A~ particle number must be conserved, the
dose is inversely proportional to the area over which the
particles are dispersed (a hemisphere of area 27tr\ There is also
some attenuation, expressed by an exponential factor, that
becomes important at distances of several hundreds of meters.
Thus, a rather complete expression for the dose H is
External (in-air) beams can produce radionuclides such as

ISO, 13N, and lSF, that might then be inhaled. Doyle et of.
(1991) discussed the hazards for the case of FIXE. The
allowable concentrations are listed in Part 835 of Chapter 10
(Energy) of the Cnde of Federal Regulations (IOCFR835,
1993).
where Q = baA is the dose, ha, averaged oyer the roof area, A; I.
is the attenuation length of neutrons in air; and r is the distance
from the source to the field point of interest. This simple
expression ignores the fact that neutrons are not
monochromatic_ and. moreover, that their spectrum is affected
by scattering off the air and even the ground, a.<; the high-energy
_neutrons are likely to survive to greater distances ...
(175)
!le,
11-6 ENVIRONMENTAL IMPACT
Both the prompt or direct radiation and""the- p"O~~ibk

emission of radioactive effluents from proton accelerator
operation may have off-site environmental impacts. A detailed
review article (Moritz, 200 l) covers all facets of the
environmental impact of proton accelerators- up to 1 GeV
energy, including isotope production facilities. Here, we restrict
ourselves to "analytic" facilities up to 15 MeV. Airborne,
ground, and surface water contamination can become important
in facilities for the production of radioisotopes or exotic beams
and were examined in the NCRP (1996) and IAEA (1980)
guides, The quantitative evaluation of effluent emissions relies
on complex pathway models, and a number of countries have
devised their own specific models and regulations requiring
compliance. Here, we address only the prompt or direct
radiation spreading around the facility for hundreds of meters
that bas to be taken into account in its design. This off-site
radiation is usually referred to as "skyshine", because, in most
cases, sufficient shielding is provided along horizontal
directions (by thick concrete walls), but the roof shielding is
..
_.
__ . .
.'~'~-'
---~.--~--.--~-
~~
..
Although the only precise way to calculate sky shine doses

requires a tracking program, such as MCNPX (2008) or
SCALE (2009), and mndeling of the facility and its
environment, it is possible to work with some approximate
expression. In this respect, a very thorough analysis was
performed by Stapleton et al. (1994), who used the "innportance
functions" calculated by Alsmiller et al. (1981) aod combined
them with a composite spectrum approximating the sea-level
cosmic-ray neutron spectrum that has an angular distribution
varying as cos(8). They obtained a function of the form
431
H(r)=( __
a_)
(r+b)'
exp(~)
perskyshineneutron
A[E,]
(17.6)
where E c, the cutoff energy, is the highest energy of a given

neutron spectrum. The best fit obtained by the same Stapleton
et al. for experimental and simulated data offered the values for
a, b, and ')..listed in Table 17.6. The average value for a is (2.41
0.29)fSv m', and that for b is (49.1 3.7) IlL In this method,
if the source Q is known only in terms of the equivalent dose,
-~,
Chapter 17
then it must first be converted to neutrons using the g factors of

the last column of Table 17.6. These factors are the average
equivalent doses over the composite spectrum with cutoff
energy Ec. At large distances and high cutoff energies, the
difference between Eqs. (17.5) and (17.6) is negligible, whereas
at short distances, the offset b limits the magnitude of H(rJ and
gives a more believable shape to the attenuation curves. For
low-energy accelerators. which are compact and sited such that
their boundaries are often within less than 100 m of the source,
the geometrical facton; always dominate the reduction in
equivalent dose with distance, and it is usually not necessary to
include the exponential attenuation term.
Table 17.6. -The parameters a and b and the attenuation
length ~ to be used 10 Eq. (17.6) 10 the energy range of our
10terest (Stapleton ef al., 1994).
E
).
(MeV)
{fSv ml}
(rn)
(m)
~fSv m'}
1.1
1.96 0.28 47.1 5.4
1424
4.5
2.780.16 53.1 2.5
183 2
5.7
12.2
2.940.15 54.2 2.2
213 3
7.4
45
2.81 0.14 53.1 2.1
267*4
9.6
125
2.44 0.11 49.2::10 1.8 355 7
11.3
400
2.24 0.27 47.1 4.5 467*33
13.2
1 x :10 3 2.24 0.18 47.3 3.0 532 28
14.1
5 x 10 3 2.230.18 46.8 3.1 597*36
14.6
1 x 10
2.23 0.24 46.8 4.0 604 49
14.7
3 x 10
2.22 0.26 46.4 4.4 617 57
14.7
several types that serve different purposes and have different

applications.. The principle difference is whether they respond
actively or require a subsequent readout to determine the
absorbed dose. An example of a small active device is the
pocket ionization chamber. Modem instruments use a small
silicon diode operated in pbotovoltaic mode whose response is
interpreted by a microprocessor for active indication of
absorbed dose and dose rate.
Passive instruments include the so-called film badge and the
thermoluminescent dosimeter (TLD) of the ring or body type.
TLDs are usually supplied as a means of monitorlng long-term
(monthly) exposures to gamma, beta, and oeutron radiation.
Their advantages include ease of readout (by photomultiplier
tubes; see Chapter 16), reasonable sensjtivity, long-term
stability, and environmental ruggedness. They do, however,
have some disadvantages such as insensitivity to low levels of
radiation, perhaps 10 mrem per month; poor sensitivity to beta
radiation; and a neutron respoose that is Dot proportional to
either absorbed dose or dose equivalent. Neutron response is
. accomplished by using a lithium fluoride TLD element.
. Neutrons are moderated by the body and diffuse back into the
TLD element, where the "Li(n,a) reaction produces energy
deposits in the TLD element. Clearly, the TLD response is
sensitive to the neutron energy impingillg on the individual,
hence requiring a knowledge of the bombarding neutron
spectrum. Generally, the neutron albedo from the body
decreases with n,eutron bombarding energy and becomes very
small above a neutron energy of a few megaelectronvolts.
Hence, the LiF TLD neutron response falls in a similar manner.
Thermoplastic materials, usually CR-39, have served as
neutron-absorbed dosimeters. These materials record the
radiation-absorbed dose by using an etching technique to
display damage to the plastic produced by neutron-1odnced,
heavy-charged particle recoils. Their neutron sensitivity
exhibits an effective threshold of about 0.5 MeV with a good
response to tens of megaelectronvolts. Some suppliers of
personnel dosimeters now combine track-etch materials and
TLDs in the same package to more completely record the
neutron-, photon-, and beta-absorbed dose. The LiF TLD
neutron albedo response provides coverage below 1 MeV, and
the track-etch detector extends the neutron response to 10-20
MeV.
..
17.7 RADIATION DETECTORS

Most accelerators used for materials analyses are capable of
producing unwanted ionizing radiation. This situation, coupled
with the desirability of having the researcher near the sample
analysis chamber and the accelerator controls. requires that
great attention be given to reducing or limiting the possibility
of exposure to mdiation. In such situations, mdiation detectors
play a leading role in keeping the working environment safe.
All the beam energy available for analyses goes up, the hazard
increases from both prompt radiation and activated samples,
making radiatioo detection systems mandatory.
The other common types of detectors are survey instruments

that consist of a radiation-sensing element, sensitive to a
particular type of radiation, connected to electronics that
convert the signals from the detector into readings 00 the
instrument. These detectors are usually much more sensitive to
radiation than the dosimeters discussed earlier.
Several types of detectors are usually encountered in

accelerator laboratories used for ion beam materials analyses.
They can be divided into two classes of devices: (1) those to
assess the dose absorbed by individuals and (2) those to survey
the environment occupied by workers. The most common are
personnel dosimeters. AU accelerator laboratories should
require personnel to wear dosimeters. These detectors come in
Portable stuVey detectors are obviously ones that can be

carried around. Instruments are commercially available for
detecting all of the types of radiation produced by an
accelerated beam. Many detectors are calibrated to give
432
Radiological Safety
readings in dose equivalent units such as ferns or sieverts (Sv),
whereas others indicate in absorbed dose units such as rads or
grays (Oy). Other detectors are calibrated in counts per minute
(cpm), which is useful for measuring low levels of radiation

such as induced activities near background. These detectors
cannot, however) directly indicate the type of radiation being
measured. With a little experimentation and knowledge of the

ranges of emitted particles, however. an educated guess can be
made.
Portable detectors are usually calibrated for only a certain

type and energy range of radiation. Various publications give
standards for the calibration and use of detectors, and the
collection of pamphlets published by the National Council on
Radiation Protection (NCRP Detectors Usage and Calibration

Reports) is an exceUent source. If a detector is not routinely
checked and calibrated, it is of limited use and could be illegal
in the United States. Training in the selection and use of a
detector is extremely helpful, if not necessary, and many
institutions have personnel whose responsibility is to assure the
appropriate us.e and calibration of both detectors and radiationgenerating equipment. These are usually the same people who
maintain the dosimeters.
Most X- or y-ray dosimeters use gas ionization for detection

of the radiation. The gas-filled cavity is assumed to operate as a
Bragg-Grey cavity so that the energy absorbed by the gas is in
radiation equilibrium with the walls and the response is
proportional to the absorbed dose in the wall material. If the
wall material is tissue-like, then the response is directly in units
of tissue absorbed dose. Greater sensitivity is possible by using
a pressurized gas cavity. Some common types of photon
detectors are equipped with a moving metal shield covering
part of the wall material. When the shield is removed, a much
thinner wall is exposed, allowing energetic beta particles and
low-energy photons to penetrate into the cavity. The source of
radiation is measured both with the shield in place and with the
shield removed. The difference in the measurements is an
indication of the energy or type of radiation being detected if
the range of the radiation is knov.n. For example, the shield
would absorb low-energy y- or X-rays, but would pass highercnergy gammas.
Area detectors, sometimes called area monitors, are fixedposition detectors that are otherwise similar to portable
detectors and that continuously monitor the radiation at
specified locations. They are commonly used as part of the
accelerator safety interlock system, which is designed to stop
beam production if a preset level of radiation is exceeded. If
your accelerator is capable of beam energies sufficient for
elastic resonance backscattering, then such a system is a very
worthwhile investment. Moreover, a swvey of the area will
indicate whether such detectors are needed. The definitions of
radiological areas, required postings, and the types of protective
systems required in the United States are described in Part 835
The two most common types of detectors used in accelerator

laboratories are neutron and combination X- and gamma-ray
detectors. Most neutron detectors give a response per unit
neutron fluence that is proportional to the dose equivalent. This
is accomplished by surrounding a thermal-neutron-sensitive
element with moderating material. For an appropriate choice of
thickness of moderating material. the response of the
instrument per unit neutron fluence is dose--equivalent for a
considerable range of neutron energy. These detectors are
easily identified by having a -characteristic -large' round or
cylindrical polyethylene moderator that houses the sensing
detector. Because of the neutron moderation process and the
practical limit to the size of the moderator, these detectors are
not very efficient and have a slow response time. Unless a
passive thennal neutron detector is used, they are useful only
when neutrons are continuously generated, that is, they do not
work in the presence of pulsed neutron sources. Some thermalneutron-sensing detectors are
of Chapler IO (Energy) of the Code of Federal Regulations

(IOCFR835, 1993).
- --------------. --_._._---
_.- -.-.-
-=--~.
.~
.-
-------
17.8 CONCLUSIONS
In today's environment, a careful evaluation of the hazards
to personnel and the environment associated with accelerator
operation is a requirement (and there certainly are other hazards
in addition to radiation). Considerations of both sample
activation and prompt radiation must be addressed. Particular
attention should be paid 10 abiding by the prevailing
regulations, orders, and laws. which seem to be changing daily.
The concept of ALARA. or as low as reasonably
achievable, must also be taken into consideration. If the
evaluation (including credible accidents) shows the hazards
from radiation to be of sufficient magnitude, then radiation
interlock systems by themselves might not be adequate, and
some type of access control or shielding might also be
necessary to keep personnel safe. It is good practice to keep a
log book recording beam species, energies, and target
combinations in addition to any prompt radiation or sample
activation that is detected. This can be used as a reference for
future analyses and for indicating when a new beam-sample
Lil crystals, which use the 'Li(n,a) reaction to produce light

pulses observed by a photomultiplier tube and counting system;
boron-trifluoride-filled proportional counters, which use the
lo:a(n,a.) reaction to generate charge pulses; and
Au foil, which uses the 197Au(n,y)'98 Au reaction to produce

activity that is subsequently measured by conventional y-ray
analysis.
For the active detectors, neutron dose equivalent rates Dear 02
flSV h- I can be detennined.
433
---'--'Chapter 17
combination might need to be monitored for possible radiation

production.
REFERENCES
lOCFR20 (Code of Federal Regulations, Title 10, Part 20)
(1993), United States Nuclear Regulatory Commission Rules
and Regulations of Staodards for Protection Against Radiation,
Office of the Federal Register, National Archives and Records
Administration, Wasbington, DC.
lOCFR835 (Code of Federal Regulations, Title 10, Part 835)
(1993), United States Department of Energy' Rules and
Regulations of Occupational Radiation Protection, Office of the
Federal Register, National. Archives and Records
Administration, Washington, DC.
Alsntiller, R,O., Barish, J.R, and Cbilds, RL. (1981), Part.
Accel. 11, 131-141.
Andersen, H.H., and Ziegler, J.F, (1977), Hydrogen: Stopping
Powers and Ranges in All Elements, The Stopping and Ranges
ofIons in Matter, Pergamon, New York, vol. 3.
DOE 5480.25 (1992), Safety of Accelerator Facilities, U.S.
Department of Energy, Office of Energy Research,
Washington, DC.
Doyle, B.L., Walsh, D,S" and Lee, S.R. (1991), Nuel. instrum.
Methods B54, 244.
Drosg, M. (1992), Computer Code DROSG-87, Technical
Report IAEA-NDS-87, Institute for Experimental Physics,
University of Wein, Vienna, Austria.
ICRP (I974~ Report of the Task Group on Riference Mm, ICRP
Publication i3, International Commission on Radiological
Protection, Pergamon, Oxford, UK.
ICRU (1992), Quantities and Units in Radiation Protection

Dosimetry, Report 51, International Commission on Radiation
Units and Measurements, Bethesda, :MD.
MCNPX (2008), Monte
http://mcnpx.lanLgov/.
Carlo
N-Particle
eXtended,
Moritz, L.E. (2001), Radiat. Prot. Dosim. 96 (4), 297.

NCRP (198~ Recommendations on limits far &posure 10 Ionizing
Radiation, Report No. 91, National COUDcil on Radiation
Protection and Measurements, Bethesda, MD.
NCRP (1993), Limitation of Exposure to ionizing Radiation,
Report No. 116, National Council on Radiation Protection and

Measurements, Bethesda, MD.
NCRP (1996), Screening Modelsfor Releases ofRadionuelides
to Atmosphere, Surface Water, and Ground, Report No. 123,
National Council on Radiation Protection and Measurements,
Bethesda, MD.
NCRP Detectors
Usage and Calibration Reports,
Environmental Radiation Measurements (Report No. 050,
1976), instromentation and Monitoring Methods for Radiation
Protection (Report No. 057, 1978), A Handbook of
Radioactivity Measurements Procedures r' edition (Report
No. 058, 1985), Calibration of Survey instruments Used in
Radiation Protection for the Assessment of Ionizing Radiation
Fields and Radioactive Suiface Contamination (Report No.
112, 1991), Use of Personal Monitors to Estimate Effective
Dose Equivalent and Effective Dose to Workers For External
xjxJsure to Low-LET Radiation (Report No. 122, 1995),

National Council on Radiation Protection and Measurements,
Bethesda, MD.
SCALE (2009), A Modular Code System for Performing

Standardized Computer Analyses for Licensing Evaluation,
ICRP (1990), 1990 Recommendations of the intemational

Commission on Radiological Protection, ICRP Publication 60,
International Commission on Radiological Protection, Elsevier,
Oxford, UK.
Report ORNUTM-2005/39, version 6,0, vols. I-Ill (January

2009), available from Radiation Safety Infonnanon
Computational Center at Oak Ridge National Laboratory, Oak
Ridge, TN, as docwnent C00750.
ICRU (1980), Radiation Quanfifies and Units, Report 33,

International Commission on Radiation Units and
Measurements, Bethesda, MD.
Stapleton, G,B., O'Brien, K., and Thomas, RH. (1994), Part.

Acce1. 44 (I), 1-15.
434
.... ..:...;.,
INDEX
A
experimental geometries, 49
heavy ion, 71
high energy, 48
kinematic factor, 50-51
:mass resolution, 46-47, 70
medium energy backscattering, 71
non-Rutherford cross sections, 47-49
resonance depth profiling, 151
Rutherford cross sections, 47
sample analysis
by deficiency method, 65-67
by elastic resonance, 58
by peak integration method. 54-57
by peak widths, 60-64
by surface spectrwn heights, 64
of bulk materials, 64
of multilayer fIlms, 57
of thin fihns, 54
stoichiometry of thin films, 57-58
with overlapping peaks, 58-60
screened cross sections, 85
Beam contamination. See Accelerators
Bobr theory. See Energy straggling
Bobr velocity, 9,
Bragg's rule, 16-17, 115, 155, 156
Breit-W iilner fonnula, ISO, 151, 182'--..:::cc:.:-.-.c.-'--__'--''--': ._~ .
Bremsstrahlung
background, 234-235
atomic, 238
quasi-free electron, 238
:secondary electron, 238
Absorber foils
for deuterium-induced reactions, 139
for ERD, 90-93
Accelerators
beam contamination
causes of, 388-389
determination of, 355
in hydrogen NRA, 179
beam energy scanning, 160
beam viewers, 390-391
energy calibration
absolute, 389-390
generating voltmeter, 388

ion sources for, 387
nitrogen beams, 152
pelletron,J88
single-ended, 388
tandem, 179,387-388
Van de Graaff; 212, 289, 388
Activation analysis. See Charged particle activation
Alignment procedure. See Channeling
Angular scans. See Channeling
Areal density
in backscatteritig spectrometry, 45-46, 52 ---
in energy loss, 8
inERD,103
in hydrogen analysis, 185
in nuclear theory, 49
Axial half-angle. See Channeling
Background reactions
in hydrogen NRA, 178-179
Calibration
accelerator energy, 356
detector, 153-154
electronic gain, 360-361
energy, 153,356
for hydrogen NRA, 185
solid angle, 358
system geometry in ERD, 112
Calibration sources. See Properties
Center ofmass system, 38, 192
Channeling, 318
alignment, 268
angular scans, 276-277
beam collimation, 430
blocking, 269
Backscattering spectrometry
areal density, 70
Bragg's rule, 49-50
cautions, 74
comparison of techniques, 70-71 .
computer simulation, 70
depth profiling, 61-68, 16, 78
depth resolution, 51, 70, 77-78
depth scale, 50-51
elastic resonance, 78- 79
energy loss, 49-50
energy loss factor, 51
experimental considerations, 70
435
dislocations, 254
high T, supercooductors, 375
lattice vibrations, 254
point defects, 266
minimum yield, 258. See also Channeling:
normalized yield
axial channels, 260
by backscattering, 273
by charged particle activation, 226
by nuclear reactions, 273
by particle induced x-ray emission, 249

calculations of parameters, 249
channel, choice of, 273
channeling dip, 251
characteristic angle, 251

computer simulation, 26&, 270
Moliere screening function, 250

normalized yield, 251. See also Channeling:
minimum yield
axial channel, 260
displaced atoms, 118-119,254
double alignment, 268, 271
continuwn potential, 253

critical angle_ See Channeling: characteristic angle
crystal orientation by x-rays, 276
damage, 256
data
channeling microscopy, 296
lattice site determination, 259-264

notation, 278
of compound epitaxiallaycra, 271
channeling contrast microscopy. 293 .

transmission channeling, 293
of epitaxial layers, 271

of heavily damaged regions, 256-257
of strained layers, 270
point defects, 254
dechanneling
defmition of, 254
energy dependence, 255

in backscattering, 255
sample damage, 297

sensitivity to lattice damage, 274
stereographic projection, 274-275
string potential, 263
surface studies
impurities, 268
interfaces, 268
surface peak yield, 268
thermal vibrations, 266-268
theory of, 266-267
thermal vibrations, 266-268
depth of analysis, 273

dislocations, 25;' 271
double alignment, 268, 271
experimental methods
alignment procedure, 276
angular scans, 276
axial scans, 276

channeling dips, 252
goniometer, 275
non-normal channel, 276
Thomas-Fermi screening function, 250
normalized yield, 276

planar scan, 278
random RBS spectrum, 275
sample preparation, 271
flux distribution, 263
goniometer, 275
half-width of channeling dip, 251
high energy, 274
laUiee defects, 253-254
vibrational amplitude, 264-268
wavelength of ion path, 257

Charge exchange, 177,352-353,359,
.Charged particle activation

activation for thin targets, 225
carbon analysis, 217
channeled activation, 220-224
current integration, 214
etches for semiconductors, 214-216
experimental technique, 212-214
irradiation procedures, 214
nitrogen reactions, 219
oxygen analysis, 221
post-irradiation chemistry, 214-215
radiochemical separations, 215-216
range approximations, 211-212
reaction choice, 212
recoil implantation, 214
saturation factor, 214
lattice site determination

analysis methods, 271
characteristic profiles, 260
from normalized yields, 260
gas atoms, 265
high symmetry, 265
large displacement, 266
low symmetry, 265
point defect associated, 266
principles, 259
procedures, 271
stopping power approximations, 211

thin layer activation, 225
Computer programs
random sites, 266

substitutional, 260
Lindhard potential, 251
Anahna, 336
Axsia, 334
measurement of
damage, 253-254
436
Carteo, 343
Dan32,335
Depth, 366
Flatus, 336
Fthie,333
GeoPIXE, 334
GISA,105,310,319,326,337,338,340,
GUPIX,343
MCERD,335-337
Monte Carlo simulation, 333-334
MEIS, 336, 364,
NDF, 319, 326, 329-333,337, 341
RBX,337
RUMP, 337
SIMNRA, 319, 326-329, 337
SPACES, 133,336
SQUEAKIE, 307
WinRNRA, 336
Computer simulation
Concentration. See also Backscattering spectrometry:
sample analysis and Depth profile

from charged particle activation, 209
Conversion factors
energy loss, 8
Convolution of a Gaussian with Lorentzian, 182
Correlated gain and offset, 361
Critical angle. See also Channeling: characteristic angle
in channeling, 251
Cross sections. See also Stopping cross sections
Breit-Wigner formula, 150, 157
deflnition of, 8
detector solid angle corrections, 359
differential, 49, 88
for ERD, 78-79
inner shell ionization, 2
Damage
sample radiation damage, 375-376
sputtering, 376
Data analysis
Bayesian inference, 319
counting statistics, 321
deterministic vs. stochastic (Monte Carlo)
procedure, 333-334
direct calculation method, 307
Elastic Recoil Detection Analysis (ERDA), 307
fitting, 238-239
interactive vs. automated procedure, 311
MEIS,336
NRA,336
NRP,336
PIGE,336
PlXE,336
simulation by successive iterations, 308-310
Dead time. See Electronics and Pitfalls
Dechanneling. See Channeling
Depth profile
in backscattering spectrometry, 67
in ERD, 78, 83
in hydrogen analysis, 182
Depth scale
energy loss relations, 17
in backscattering spectrometry, 50-51
Detectors
BGO (bismuth genninate), 177
channeltron, 392
dosimeters, 432
electron mult.ipliers, 391-392 _::.~ ___ _. ~ __._ .__ . ~ __ ~ _
fim; b"adges, 432
for personnel exposure, 432-433
.
Ge (Li), 397
particle telescope, 399-400
photomultipliers, 300, 392-393
pulse height defect, 396
scintillation, 396
semiconductor, 396
sunace barrier, 257-258
thermoluminescent dosimeters, 432
time-of-flight, 71, 73, Ill, 353, 399
Differential scattering cross section. See Cross sections
Dosimeters. See Detectors
Double alignment, 268. See Channeling
non-Rutherford
for backscattering spectrometry, 47-49, 59
for ERD, 78-79
optical model, 35
. particle-gamma, 149-150
potential scattering, 30
resonant reactions, 176-177
Rutherford, accuracy. 363
Rutherford, deviations from, 48

Rutherford for ERD, 84
Rutherford for RBS, 46
Rutherford window, 364
screened, 85, 364
stopping, 1. See also Stopping cross sections

Current integration. See Pitfalls and Charged particle
activation
I
I
I
I
I
~.
437
fluctuations additivity, 21
Epitaxial laye,,_ See Channeling
ERD. See Elastic recoil detection
Error analysis. See Statistics
Effective charge, 12
Elastic recoil detection

absorber foil, 83
analysis of silicon nitride films
depth profiles, 92
scaling factors, 133
analytical expressioIl5, 253
cross section, 84-87
depth resolution, lOS-107
equations for analysis, 120
kinematic broadening, 111-112
kinematic factor, 84
range foil. See Absorber foils
ranges for projectiles, 89-90
sensitivity, 114-11S
slab analysis, lOS
thick layer analysis, 103-104
thin Layer analysis, 103
tirne-of-fligh~ 97-100
Famdaycup
design, 3S1, 391
transmission, 351
Forward recoil spectrometry. See Elastic recoil
detection
Fractional channel numbers, 361
H
Heavy-energy backscattering. See Elastic recoil

detection
Heavy-ion backscattering. See Elastic recoil detection
Hydrogen nuclear reaction analysis. See also ERD
background reactions, 178-179
calibration constant. 181
depth resolution, 182
depth scale, 183
high-sensitivity measurements, 183
hydrogen bearing materials, 183
transmission ERD. 101

useful equations, 120
Elastic resonance backscattering. See Backscattering
spectrometry
Elastic scattering cross sections
differential cross sections, 47
Electron screening, 316
Electronics
amplifiers, 406-408
cables and connectors, 410-411
dead time correction, 420
logic levels, 404
multichannel analyzers, 408-410
NlM standard, 404
pin asSignments, NlM standard, 403-404
power connections, amplifiers. 355
preamplifiers, 40S-406
Energy calibration. See Accelerators, Detectors and
Properties
Energy loss. See also Stopping cross sections
Bragg's rule, 16
chemical state effects, 17
computer calculations, 12-13
cores and bonds, 17
depth scale relations, 17
electronic, 11
heavy ion scaling, 11-12
physical state effects. 17
Energy loss factor
for backscattering spectrometry, 49-50
forERD,108
Energy stability. See Accelerators
Energy straggling
Bohr's theory, 19 .
compounds and mixtures, 20
corrections, 19-20
Impulse approx.imation. See Channeling

Insulatiog samples
charging effects, 377
in hydrogen NRA, 178
photon emission, 378
Interstitial lattice sites. See Channeling
Ion sources. See Accelerators
Ionization ~ross sections, inner shell, 236
K
Kinematic factor
for backscattering, SO-51
[orERD,84
L
Lattice defects. See Channeling
M
Mass resolution
in backscattering spectrometry, 46-47
Mean energy approximation

in backscattering, SI
in energy loss, 18
Medium energy backscattering. See Backscattering
spectrometIy
Micro--Pixe, 242-243
Model-Free Analysis, Occam's Razor, 370-371
438
Minimum yield. See Channeling

Moliere's screening function, 250
Mosaic spread. See Channeling
Multiple scattering, 109-110, 316-317
N
Neutrons, thick target yic:lds
Nonnalized yield. See Channeling
Nuclear Microprobe
Beam damage, 289-290
EDRA,292
high current applications, 290-292
Ion Beam Induced Charge (TBIC). 296-297
Ion Beam Induced Luminescence (IBIL). 298-299
Ion Induced Electron Imaging, 300
low current applications. 290-291
NRA,291
PIXE,289
Scanning transmission Ion Microscopy (STIM),
294-296
Single Event Upset (SEU), 298
thne-resolved IBIC (TRIB1C), 297-298
transmission channeling, 293
Nuclear reaction analysis: particle-gamma. See also
Hydrogen NRA
background, 149
beam energy scanning, 153
depth profiling, 159-164
depthresoiution, 157-158
y detectors, 153
particle-induced y-ray emission (pIGE), 164-171
sensitivity,157-158
standards, 156-157
thick target yields, 164 .
Nuclear reaction analysis: particle-particle
depth profiling
non-resonant, 130-l32
resonant, 133-134
depth scale
non-resonant, 130-132
resonant, \J 3
filtering methods
absorber foil, 129
coincidence, 129
electrostatic and magnetic, 129
thin detector, 129
time-of-flight, 129
general principles and equipment, 127-128
geometry, 128-129
Q values, 128
standards, 134
proton-induced, 136-138
Nuclear reactions
angular momentum effects, 32
background reactions. See Background reactions
cross sections
for hydrogen analysis, 178. See also Hydrogen
NRA
theory of, 176
Nuclear resonance reactions. See Nuclear reactions
Nuclear theory. See Chap 3. See also Nuclear reactions
o
Optics
beam energy" spread, 289
chromatic aberration, 289
cylindrical electrostatic lenses, 286
electrostatic quadrupole lenses, 286
electric rigidity, 286
magnetic quadrupole lenses, 286
object aperture. 286
demagnification, 286
collimator aperture, 286
solenoid lenses, 286
spherical aberration, 286
orthomorphic focus. 288
magnetic rigidity, 286
quadrupole optics, 286
P
Pitfalls
charge exchange, 352
charge measurement, 351-352
deadtime, 353-355
electronic noise, 355-356- ....... ~
~~~- .-~=
events, 351-356
ground loops, 355-356
lost beam, 351-356
multiple scattering, 365-366
particle identification, 352-353
Pulse pileup, 353-355
surface roughness, 378
PIXE
accuracy. 241
X-ray emission, 233-234
intensity-concentration relationship, 234
depth distributions, 241-242
ionization cross section, 234
fluorescence yield, 242

limite; of detection, 241
matrix correction, 236-237
invisible elements, 240
mass attenuation coefficient, 234
useful reactions
'He-induced,141-142
alpha-induced,142-143
deuteron-induced,139-141
Plural scattering, 116-117, 316

Properties
calibration sources
439
y-ray detectors, 412413

particle detectors, 414
x-ray detectors, 411
coaxial cables, 410413
coaxial connectors, 417
inorganic scintillators, 393
plastic and liquid scintillator.;, 394
silicon and germanium, 396
surface barrier detectors, 396
Pulse-height defect See I)etectors: non-linearity
Pulse pileup, 318, 353-354
secondary electrons, 98-99, 300

Scattering angle, 362
Spectral ambiguity, 368-370
Sputtering. See Pitfalls
Standards. See also Properties
alpha particle energy, 414
for accelerator energy calibration, 356
for Bi, 54, 359
for charged particle activation, 209-210
for hydrogen, 359
intermediate sample, 235
layered samples, 237
for NRA, 358-359
for Sb, 359-360
thin sample, 236
thick sample, 236237
spectrum fitting, 238-239
x-ray, 412-413
Statistics
__ counting, 419
. -.-~~.-~--,,~-,.
error propagation, 419-420
random coincidences, 420-422
uncorrelated errors, 420
Stereographic projection, 274-275
Stopping cross section factor. See Energy loss factor
Stopping cross scctions
accuracy, 17
calculation for protons., 12-13
definition, 8,9
electronic, 364-365
electronic and nuclear stopping H and He, table of.
IS
nuclear stopping, 12-13
reduced energy, 15-16
reduced nuclear stopping, 11
selected ions in elemental and compound targets,
16-17
Stopping power, 313-314Seealso Stopping cross
sections
StriLggling. See also Energy straggling
Strained layers. See Channeling
Superlattices. See Chamieling
Surface barrier detectors. See Detectors
Surface energy approximation, 18, 50
Surface peak. See Channeling
Surface roughness. See Pitfalls
Q
Qvalu,;,s
definition of, 33
R
Radiation damage. See Damage
Radiation detectors. See Detectors ._~ ..~,.,....... __ .
Radiological safety
absorbed dose, 427-428
accelerator safety order, 426
agreement states, 426
dose equivalen~ 428
hazards
activated samples, 431
charged partjcle beams. 428
kenna, 427
National Council of Radiation Protection, 426
Nuclear Regulatory Commission., 426
quality factor, 427
special nuclear materials (SNM), 426
Range foils. See Absorber foils
Ranges
"calculation and simulation, 21

in compounds, 21
isotopic effects, 21
projected range, 20
RBS. See Backscattering spectrometry
Reduced mass, 29
Resolution, 366
Roughness
interface, 318
sUlface, 318
Resonant reactions. See Nuclear reactions
Rutherford cross Sections, See Cross sections
T
Tandem accelerator. See Accelerators
Thermal vibrations. See Channeling
Thiek layer analysis
by ERD,103104
Thin fIlm production, 418-419
Thin layer analysis
by backscattering spectrometry, 54
S
Safety. See Radiological safety
Sample damage. See Damage
Scattering cross section, 314-315 See also Cross
sections
Scattering kinematics, 365
Screening. See Cross sections
440
by ERD, 103
Thomas-Fermi screening function, 250
Time-of-flight. See Detectors
Transmission ERn. See ERD
Vacuum practice
cleanliness, 418
leak repair, 418
Vibrational amplitude. See Channeling

U
Uncertainty Estimation, 367-368

X-ray
absorbing flIters, 238

detector, 237-238
441

Modern Ion Beam Materials Analyisis

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Modern Ion Beam Materials Analyisis

Hochgeladen von

Copyright:

Verfügbare Formate

Handbook of Modern

Ion Beam Materials Analysis

Y ongqiang Wang and Michael Nastasi

Single article reprints from this publication are available through

Manufactured in the United States of America

Energy Loss and Energy Straggling ................................................................. 5

The Interaction of Charged Particles with Nuclei ......................................... 27

Backscattering Spectrometry .......................................................................... 43

Elastic Recoil Detection: ERD ........................................................................ 81

Nuclear Reaction Analysis: Particle-Particle Reactions ............................. 125

P.rticle-Induced G.mm. Emission: PICE ................................................. 147

Nucle.r Reactions for Hydrogen Analysis ................................................... 175

Hydrogen Analysis by Proton .. Proton Scattering ......................................... 187

Ch.rged p.rticle Activation Analysis .......................................................... 207

11. Particle-Induced X-Ray Emission: PTXE ...................................................... 231

Radiological Safety ......................................................................................... 425

Analysis for providing a superb foundation for this second

Los Alamos, New Mexico, USA, July 2009

0/ Modem Ion Beam Materials

edition of the Handbook

Los Alamos National Laboratory

Paul Scherrer institUle and ETN Zurich

Universitat der Bundeswehr Munchen,

University a/Guelph, Canada

University a/North Carolina

University 0/ Western Ontan'o

Middle Tennessee State University

United Stales Naval Academy

The UniversityandINFN, Padua, Italy'

National University a/Singapore

The second edition of the Handbook of Modern Ion Beam

In addition to the revision of many of the original chapters in

is unavoidable, PIXE and PIGE are spectroscopy techniques that

induced charge (IBlC) microscopy. Thus, a new chapter

The fundamental physics of the interactions bctween incident

Although computer programs to analyze data from RBS and

Chapters 4 through 10. The basic principles of particle-induced

In addition to element/composition analysis and depth

Ion beam analysis relies on 1-5 MY ion accelerators to

elemental analysis techniques (e.g., RBS, NRA, PIXE) with the

by highiy ordered singlecrystalline lattice structures. In fact, ion

Over the past two decades, rapid development in accelerator

current, and target (sample) material, as well as the facility layout

reactions, including those involving elastic resonances. More on

In the analysis universe, there are many surface and thin-film

Analysis, Academic Press, Sydney, Australia.

Bird, J.R., and Williams, J.S. (1989), Ion Beams/or Materials

Tesmer, J.R., and Nastasi, M. (eds.) (1995), lIandbook 0/

AND ENERGY STRAGGLING

2.2.5.2 Accuracy, deviations, and corrections to Bragg's rule ................................... 17

The experimental data on ion energy loss are overwhelming.

Knowledge of the slowing down of ions in traversing matter is

semiempirica1 fit for just one element, aluminum, for ions Be

techniques follows directly from the energy lost by the probing

25,000 data points. Figure 2.2 shows a comparison between

The physics of energy loss phenomena is very complex,

Energy ( keV /amu )

Ion Energy ( keV/amu )

The theoretical treatment of energy loss phenomena has been

reviewed by Bohr (1948), Wbaling (1958), Fana (1963), Jackson

International Commission on Radiation Units & Measurements

The experimental methods have been reviewed and discussed

Energy loss in the relativistic energy region bas been reviewed

2.1.1. Basic concepts and defmitions