"arne Actinium AIUlnmum An1CnClum Antimolw

Symbol A, AI Am Sb

"arne Hafnium Helium Holmium Hychogen IndIum Iodine Iridium J= Krypton Lanthanum La....TellClum

Symbol

Name Praseodymium Promethium Protactinium Radium Ibdon Rhenium Rhodium Rubidium Ruthenium Samarium Scandium Selenium Silicon Silver SodIum Strontlum Sulfur Tantalum Technetium Tellurium Tt:rbium Thallium TI!orium TIlulium Tin Titanium Tungsten Uranium Vanadium Xenon Ytterbium Yttrium Zinc Zirconium

Symbol

11

""00 Arse.nlC

..
A,
AI

H. Ho H In
J J,

..

P, Pm P. Rn R. Rh
Rb

Astatine Banum Berkelium Beryllmffi BIsmuth Ilo<oo Bromme Cadmium Calcium Californium Cuboo Cerium CesIum Chlorine Chromium Cobalt Copper Curium Dysprosium Einsteinium Erbium Europium Fermium Fluorine Francium Gadolillium Gallium Germanium

B.
Ilk Bn

F. K,
Lo L<

R" Sm S. " SI A" N. S, S T. T, T. Tb

11

BI B B,

Lnod
Lithium Lutetium
~lagnesium

Pb

c..

Cd

~
~lg ~ln

Cf C C. C, Cl C,
Co

\Ianganese \Iendele\ium \1ercur\'

~loIvbdenum

\Id

Hg
~Io

C.. Cm
Dy

E, E,
Eu

Gold

Fm F F, Cd C. C. A..

"eodymium l\oon l\"ptunium KICkei t\iobium "'Itrogen ;'\obelium Osmium Oxygen Palladmm Phosphorus Platinum PI"tonillln Polonium Potassium

Nd
!'e :\p No

Nb
N
;\0

Th Tm Sn TI
W

0,

Pd P
P, P.. Po K

U
V

X.. Yb Y Zn
z,-

_-=_ ""_ _

~--_'-!Io.

......

..

H'
I., Periodic Table, with the Ouler Electron Configurations of Neutral Atoms in Their Ground Siaies

He'
h'

Li'
0,
Nail

Be'
Z.~2

Mg"

The notation \1\cd to describe the c]t!(:IWll(C o:;onhguraljon of lltorn, ,md ion' i' di~cu'~cd in all textbook~ of introductory alomH; phy,ic\. The letler~ $, p. d, ....\ignify c!edron\ having orhital angular momentum O. I, 2, . . . in unih -lI; the numher to the lell of the letter denute, the pnncipal quantum numher of Ollt' urbit, and the ,uper~cripl to the right denole, the nl1mllcr IIf electroll~ in the orhit

B'
2.,.2211
All]

C'

N'

0"

F'

Ne'o

2~22pl 2s22p~

2s!2p'

2.~221,5 2.~2pl

51"
3s'3p'

p"
:l~np'

S"
:h'3p'

CPl

AT'"

3.,

K"
4,1Rb 'l

'CaN "
4~!

~.23p

3.~23pr,

3s'3p'

Se"
3<1
4.~!

Ti"l

V"
:}rP
4.~l

Cr

'

Mn"

Fe""
3d"

Co'

Nill'

Cu'"

Z,' Ga"
3el l

Ge'"

As]'

Se

Br'5

K,

3d'

S,'
5."
Ba l >

y ..

4." Z,
4ct!

4.,

3d 5

3d'

3d'

4."
Tc"

4,'

4,'

3d"
4.~!

4,

3d'~

4,'
Cd'"

Nb'!

Mol!

Ru"

Rh'l

Pd"

Ag Ol

4s'4p 4s!4p' 4s!4p" 4.r4,,' 4s'4p' 4s'4p' In" S," Sb" Te u I" Xe"

4</

4d'
Ss

5,
Cs~s

'" '"
La
ll

4d' 5.,
W'l

4ci'

5,
RefS

4d' 5,
OSI.

4d' 5,
Jr
ll

4d"
0

5.,

4d '
Au"

4d"
5.~'

5.1:'51' 5s!5p'
Pb"'

5.~!5p ,

5.015,)' 5s!5p'
Po" At"S

5.~!5p'

HI1, 4f"

Ta"l

Pt'

Hg" TI"
5d'~

Bi"l

Rn"

fu
Fr'l

fu'
Ra""

5d fu'
Ac"'

5d' fu'

5cP

5d'
6,~!

fu'
Pr"

5d' 6,'

5d' 6,'
Pm"

5r1"
0

5ci'

5ci'

fu
Eu1

fu

6,'

""6p 6s!6p! 6s!6p" 6s!6p' 6s!6 p S 6s'6p' Oy"

4f'~

7,

7,1.2

7,'

6<1

"

Ce"
4f'
6..2

4f'
6.~!

Nd" 4f'
fu' U"
S!"'

Sm"

Gd
5<1

Tb"
5<1 fu'

41'

41'
fu'
Put<

4f'
68'
Am,l

4/'

4f'

Hol 4f"

E..'
4fl>

Tm" 4f"

Yb7~

Lull

4fH

6.' Np" Sf'

7.-'

Th"
0

Pa'!

"" Cm"
7,' -Sf' 6d

6,1,1

6,'

6,'

fu'

fu'

5d fu'

4/"
Lr'"

Sk"7

C,"

Es"

Fm'" Md'' No"!

Sf' Gel! 6<1 ____c.:.:7.' 7~"

Sf'
7:-

Sf'

6d

7.!

---

l!...

CHARLES KITTEL

Introduction to Solid State Physics

SEVENTH EDITION

John Wiley & Sons, Inc., New York, Chichester, Brisbane, Toronto, Singapore

J 3.1.1

--i'l
/'>4 10 DU:: c
eT

About the Author

Charles Kittel taught solid state physics at Berkeley from 1951 to 1978; earlier he was a member of the solid state group at the Bell Laboratories. His undergraduate work was at M.I.T. and at Cambridge University, followed by graduate work at the University ofWiSl.'Ollsin. He is a member oCtile National Academy of Sdem;e and of the American Academy of Arts and St:iences. His research in solids began with studies of ferromagnetic, antiferromagnetic, and paramagnetic resonance, along with work on magnetic domains, spin waves, and domain boundaries in ferromagnets and ferroelectrics. His work on the single domain strudure of fine particles has Imd broad llpplil~ltion in magnetic recording, geomagnetism, and biomagnetism. Along with collaborators at Berkeley he did the first work on (,.yclotron resonance in semicondudors, which led to the understanding of the band structure of sili(,.'()n, gcrmanium, and indium antimonide, together with the theory of their impurity states. He also worked on the interpretation of magnetoplasma resonance in scmieonductors and of A]fven resonance in eledron-hole drops in gennanium, The frrst ooition of ISSP integrated thc elementary aspeds of solid state physics for study by seniors and beginning graduate shldcnts. Now in its seventh edition, ISSP plays the same part for the current gencmtion of students.

Copyright C 1953,1956, 1966, 197t, 1976, t966, 1996 by John Wiley &; Sorn, Inc.
All righu resened. Published simultaneomly in Canada. Reproduction or translation of any part of th~ work heyond that penniUed by S.,ction~ 107 or lOS of the 1976 United States Copyright Act without the permission of the copyright owner is unta",ful. Reque>;ts for permis~ion or furth"r information shoukl be addressed to the Permissions Department, John Wiley &; Soos, Inc.
Library oj Congress Catalogiag in 1'llbllaltlo" Data:
Charles Kittel. Introduction to solid state physics I Charles Kittel. -7th ed. p. em. Includes index. ISBN 0-411_11181--3 (cloth: ali<. paper) ) Solid state (ilysics. L Title. QCI76,KS 1996 530A'I--dc20 95-18445 elP

Printed in the United States of Amt:rica

15 14 13 12 11

Preface

This book is tile seventh edition of an elementary text on solid state physics for senior a.nd beginning graduate students of physk'3l science and engineering. l1Je book is an update of the sixth edition of 1986 and includes additions, improvements, and corrections made in that edition in 13 successive printingswhich it \\,'35 time 10 pull together-and a number of new topics besides. Signifkant ad\'3Jlces in the field have been added or discussed more fully: thus high temperature supercondudors are treated. and results cl scanning tunneling microscopy are displayed; the treatment of fiber optics is expanded. There are diSl....lssions. among other topics, of nanostrm:t:lIrcs. sllpcrlattices, Dlochf Wannier levels, Zener tunneling. light-emitting diodes, and new magnetic materials. 111e additions have been made within a boundary condittoll intended to keep the text within one volume and at a reasonable price. The theoretical level of the text itself has not been cbanged. There is more discussion of useful materials. TIle treatment of elastic constants and elastic "':aves which was dropped after the fourth edition has now been returned because, as many ha\'e pointed out, the matter is useful and not easily accessible elsewhere. The treatment of superconductors is much more extensive than is Ilsual in a text at this level: either you do it or rOll don't. Solid state physics is concerned with the properties, onen astonishing and often of great utility, that result from the distribution of electToos in metals, semioonductors, and insulators. The book also tells how the excitations and imllerfections of real solids can be understood with simple models whose power and scope are now finnly established. The subjed matter supports a profitable interplay of experiment, applirotiou, and theory. The book, in English and in many translations, has helped give several generations of students a pk1ure of tbe process. Students also find the field attractive because ofthe frequent possibility of working in small groups. Instmctors will use the book as the foundation of a course in their Own way. yet there are two general patterns to the introduction. selection and order of the basic material. If students h~'e a signiBcant preparation in elementary quantum meek. b, they will like to begin with the quantum theory of decii,

)
trons in Qne-dimensional solids, starting with the frcc clcdron gas in Chapter 6 and energy bands in Chapter 7. One will need to treat the rcciprocallattice in three dimensiOns (Chapter 2) before plunging into semiconcludors (Chapter 8) and Fermi smfaces (Chapter 9). Crystal stOlchITCS, crystal binding, and phonons (;ould be considered as recreational reading. In a more gradual approach.
the first eight chapters through the physics ofsemi(:(mdm.. 1 ors are read consel.'Ulively as a One-semester intrududion to the field. What abollt the na-essary statistical mechaniC5? A vague discomfort at the thought of the chemical potential is still charad:cristic of a physics education. This intelledual gap is due to the obs(;l.lrity of the writings of J. Willard Cibbs. who discovered and understood the matter 100 years ago. Herbert KTOemer and I have danned the physics of the chemical potential in the early chapters of our book on thermal physics. Review series give excellent extended treatments of all the subjects treated in this book and many more besides; thus with good conscience I give few references to original papers. In these omissions no lack of honor is intended to those who first set sail on these seas. The crystallographic notation conforms with l....lrrent usage in physics. Important equations are repeated in SI and CCS-Caussian units, where thcse differ. Exceptions arc figurc captions, chapter summaries, some problems, and any long section of text where a single indicated substitution will translate from CCS to SI. Chapter COntcnts pages diseuss cOlwentions adopted to make parallei usage simple. The dual usage in this book has been found useful and al.'Ceptable. Tables are in conventional units. The symbol e denotes the charge on the proton and is positive. The notation (18) refers to Equation (18) of the l.1IITent chapter, but (3.18) refers to Equation 18 of Chapter 3. A caret' over a vector refers to a unit vel.1or. Fl.'\\! of the problems are exactly easy; most were de\<ised to carry forward the subjed of the chapter. With a few exceptions, the problems are those of the original sixth edition. This edition owes much to the advice of Professor Steven C. Louie. For collected corrections, data, and illustrations I am grateful to P. Allen, M. Beasley, D. Chemla, T.-C. Chiang, M. L. Cohen, M. G. Craford, A. E. Curon, D. Eigler, L. M. Falicov, R. B. Frankel, j. Friedel, T. H. Geballe, D. M. Ginsberg, C. Herring, H. F. Hess, N. HolonY.lk, Jr., M. Jllcob, J. Mamin, P. McEuen, J. G. Mullen, J. C. Phillips, D. E. Prober, Marta Puebla, D. S. Rokhsar, L. Takacs, Tingye Li, M. A. Van Hove, E. R. Weber, R. M. Wilite, J. P. Wolfe, and A. Zeui. Of the Wiley staff I have particularly great debts to Clifford Mills for publication supervision, to Cathy Donovan for her ingenuity in proressing the additions between the thirteen successive printings, and to Suzanne Ingrao of Ingrao Associates for her skill and understanding during the editorilll pro<.'ess.

)
Corrections and suggt,'stiolls will be gratefully received and may be addressed to the author at the Department of Physics, University of Califomi a, Berkeley, CA 94720-7300; by email to kittel@uclink4.llerkelcy.cdu;andbyfax to (51O) 643-9473.
C. Kittel
An Instructor's Manual is available for this revision; several problems have been added (to Chapter 3 and Chapter 6); one dropped (from Chapter 4), and several corrections made. Instnlctors who have adopted the text for classroom lise should direct a request on departmental letterhead to John Wiley & SOilS, Inc., 605 Third Avenue, New York, NY 10158-0012. Limited requests for permission to copy figures or other material should be addressed to the Permissions Editor at this address.

Contents

Cllide 10 Tables

u
~

Ceneral References
I CRYSTAL SrRVCruRE

2
3

HEClPROCAL ........rncE

21

CRYSTAL BINDING AND EL\S"nC CONSTANTS PIIONONS I. CRYSTAL VIBRATIONS

53

4 5
6 7 8

.7

PIIONONS 11. THERMAL PROPElffiES

.." nEE ELECfRON FERMI GAS

'"
141
173

ENERGY BANDS
SEMICONDVCfOR CRYSTALS FERMI SURFACES AND METALS PLASMONS, POLARITONS. AND POJ.ARONS OPTICAL PROCESSES AND EXQTONS
SUPERCONDUCClVITY

197

9 10
II

233

26.
3Oi5

12
13

333

DlELECfRICS AND FERROELECTRICS

DlA~fACNI::.IlSM

37.

14 15
16 17 18

AND PARAMAGNETISM

...
441 485

FEtrnOMAGNE"nSM AND ANf]l"ERROMAGNEnSM

MAGNETIC RESONANCE NONCRYSTALLlNE SOUDS

.,.
S39
vi

POlNC DEFYIJS

viii

19 SURFACE AND INTEUFACE PHYSICS

20 DISLOCATIONS

21 ALLOYS

'"

Appendix

A TEMPERATURE DEPENDENCE OF TilE REFLECTION UNES

B EWAlD CALCULATION OF LATTICE SUMS

.,.
638

631

QUANTIZATION OF ELASTIC WAVES: PHONONS

D FEUMI-DIRAC DlSTRlBt1I10N FUNCTION

64' 64' 64'

E DERIVATION OF dkfdt EQUATION

F G

BOLTZMANN TRANSPORT EQUATION VECCOR PO'fENTLU.., FIELD MOMENTUM, AND GAUGE TRANSFORMATIONS

'"
656 638 662

H COOPER PAIRS I
GINSBURG-LANDAU EQUATION ELECI1I.ON-PHONON COLUSIONS

Subject Index Table of SI PrefIXes

6<,
613

Guide to Tables

1.I

1.2 1.3 1.4 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10

Fourteen lattice types in three dimensions CharacteristiCli of cubic Ltticct; Crylital structures of Ihe clement!> Density and atomic concentration of the e1emenb

10 12 23 24

Cohesive cncrgiCli of the elements Melting tcmpcraturCli of the elements Bulk modulii and comprcssibilitics of the clements Properties of inert gas crystals Ionization energies Electron afIinitics of ~tiyc ions Properties of alkali halide crystals Fractional ionic character of bonds Atomic and ionic radii Standard radii of ions 3.11 Elastic stiffness comtants of cubic Cl)'stals at 0 K and 300 K 3.12 ElastK: stiffness constants of cuhic crystals at 300 K

57 58 59 60 61 66 73 76 76 79 91 92 126 133

5.1 5.2 6.1 6.2 6.3 6.4 6.5 8.1 8.2 8.3 6.4 6.5 6.6 6.7

Dcbyc temperature and thennal condncliltity Phonon mean free paths

Free electron Fenni s"rfac:e parameten for metals l':lectron heat capacity of metals Elcctric-.d conductivity aml resistivity of metals Hall coefficients
Lorenz number,,;

ISO 157 160 167 166 WI 214 221 223 224 225 229
;,

Energy gaps in semiconducton; Effectivc masses of holes and electrons Carrier mobilitics at room temper<ltuTc Static dielectric cunstants of semimnducton; Donor ionization cnergies Acceptor ioni7.ation energies Elcctron and I,olc concentrations in scmimelals


lO.] JO.2

10.3
10.4

Ultraviolet transminiOll limits of alkali metals Volume plasmon energies Lattice frequcncte5 Polaron masses and coupliug constants

275 278
292

298

)).1 \1.2 11.3

Exciton binding energies

3\4
322

E1ectrorJ-bole liquid parameters Acronyms of C\l.1l'ef'It experimental methods for band structure st~
Superconducthity par-unden cl the dements SupefOOnductivity of selected compounds Energy gaps in superconductors botope effect in supen:Olldlldon Ccherence length and penetration dept)l

328

\2\
\22

336 336
344

\23 \24
12.5
13.1 13.2 13.3
14.1 14.2

347
353

Electronic polariuhilitics of ;0111 Ferroe1ecl:ric crystals "n!iferroelectric crystals

Magneton numbers of l"'mlhanidc group lorIs Magneton numlx:n of irOIl group ions
eritia] point exponents of fcrromagncls

39\ 396 406 425 426 .45 449 465 489 "'2 545 547 549 556 56"
588

15.1 15.2 15.3

16.1 16.2 18.1 18.2

Ferromagnetic: crystals Antifcrromagllctic crystals Nuclear magnetic resonance data Knight shifts in NMH

18.3
19.1 19.2

Diffusion conslanb and <"lctivatioll energies Activation energy for a positive ion VlIcallCY Experimental F center ab!;orptioll energies
Top layer atomic relaxation Elcctron work functions Elastic limit and shear modulus Dislocation demities Electron/atom ntios of electron l'Ompo'Unds

20.\ 20.'
21.1

598 6\6

Selected General References

Statislirol phyriff I.Jockgroond C. Kine! and H. Kroemer. ThenntJi vhvna. !bid ed., Freeman. 1980. Has a fun. clear discussion of the chemical potential and of scmkorlductor statistics; cited as TP.
Inl~le

tm

J.

M. Ziman. Principles of the theory of pich, Cambridge. 1972.

Adt.'GrlCed te%ll C. Kittcl, Quantum thoory of solidt, 2nd re\iscd printing. Wiley. 1987, v.ith solutions appendix by C. Y. Fang;. cited as QTS. J. CallaW8)', Quantum theory of the solid state. 2nd cd., Academic, 1991.
Applied mlid da'e n. Dah'cn, lntroduetum to applied solid state phybia, 2nd cd., Plenum, 1990. A readable introduction to representative areas in a vast field.
Review smet F. Seitz and OtilCrs, Solid slafc pllysics, ad~(Jn:lI in research OIld applicaflons, Vois. 1-(48), plus supplements. This VllluulJlc continuing series is often cataloged as
a scrifll, as if it were a journal. and is dted I\crc as Solid slate ph)'$iC$.

literoture guide. 1bere are many good databases, library and organizatKma1; this is the way to go for mooograph and journal searche5.

.'

1
CnJstal Structure
PERIODIC ARRAYS OF ATOMS
Lattice translation \'L'<:lors

3
4

UlISis and the cl')'Stal structure Primitive lattice cell FUNDAMENTAL TYPES OF l..AlTICES T...dimL'ftSkmaJ lanke typc5 Three-dirnensiooallattice types
INDEX SYSTEM FOR CRYSTAL PLANES

5 6 8 8 10
12"

SIMPLE CRYSTAL STRUCruRES Sodium chloride stnK'ture Ca-ium chloride StnJdllfC Hexagonal close-packed structure Diamond stnidure Cubic rille sulfide structure

15 15 17 17 19
20

DIRECT IMAGING OF ATOMIC STRUCrURE 20

NONIDEAL CRYSTAL S'fRUCfURES Random stacking and p<M)'typism
CRYSTAl. STRUCTURE DATA
21

22 22
25 25 25 25
25

SUMMARY
PROBLEMS

1. Tetrahedral angles
2. Indices of planes 3. Hcp structure

REFERENCES
UNITS, I

25

A=

1 angstrom - 10-8 em

'=

0.1 nm - 10- 10 m.

(oj

(b)

,.)
.lgun: I Relation of thr otnnal ....m of n,... tab 10 the born or the ekmmtar)' building bIocb_ The building bIod.:s an:: idCtl11C21 in (a) and (b). but d~nt CT)~"'1 Etres arede--eloped. (c) ON," 'ng a CIj-.tal of ro<:l.$aIl.

.
CHAPTER

1:

CRYSTAL STRUCTURE

Solid state physics is largely concerned with crystals and electrons in crystals. TIle study of solid state physics began in the early years of this century following the discovery of x-ray diffraction by crystals and the publication of a series of simple calculations all{l successful predictions of tIle properties of crystals. \Vhen a crystal grows in a constant environment, the fonn develops as if identical building blocks were added continuously (Fig. 1). The building blocks are atoms or groups of atoms, so that a crystal is a three-dimensional periodic array of atoms. This was known in the 18th century when mineralogists discovered that the index numbers of the directions of all faces of a crystal are exact integers. Only the arrangement of identical particles in a periodic array can account for the law of integral indices, I as discussed below. In 1912 a paper entitled "Interference effects with Rontgen rays" was presented to the Bavarian Academy of Sciences in Munich. In tIle first part of the paper, Laue developed an elementary theory of the diffraction of x-rays by a periodic array. In the second part. Friedrich and Knipping reported the first experimental observations of x-ray diffraction by crystals. 2 The work proved decisively that crystals are composed of a periodic array of atoms. With an establisl1ed atomic model of a crystal, physicists now could t1link much further. The studies have been extended to include amorphous or noncrystalline solids, glasses, and liquids. The wider field is known as condensed matter physics, and it is now tIle largest and probably the most vigorous area of physics.

PERiODIC ARRAYS OF ATOMS

An ideal crystal is constructed by the infinite repetition of identical stmetural units in space. III the simplest crystals the stnlcturalunit is a single atom, as in copper, silver, gold. iron, aluminum, and the alkali metals. But the smallest stmctural unit may comprise many atoms or molecules. 111e structure of all crystals can be aescribcd in terms of a lattice. with a group of atoms attaclled to every lattice point. The group of atoms is called the basis; when repeated in space it fOnTIS tIle crystal structure.
'R. J. ~tHUY, Usai ,fune thtom sur la $/n,cture des cristallx, Paris. 1784; TraiM de cristalfographie, Paris, tOOl. 'Forpersonal ac<:ou,," ortheearIY)'<;:at"$ ofx-ray dilfractian sludies eX c')'51.1s, see P. P. E".,.ld, ed. Fifly years of r..-m,

,.

fraction, A. Oosthoek's Uitgc"entnij., Utrecht, t962.


Lattice Trafl.!llation Vectors

The lattice is defined by three fundamental translation vectors a.. a2. R3

such that the atomic arrangement looks the same in every respect when viewed from the point r as when viewed from the point
(1)

are arbitrary integers. The sct of points f ' defined by (1) for all defines a lattice. A lattice is a regular periodic array of points in space. (The analog in two dimensions is called a net, as in Chapter 18.) A lattice is a mathematical abstraction; the crystal structure is formed when a basis ofatoms is attached identically
Ill> ll2, 113
til> 112. u3

where

to every lattice point. The logical relation is

lattice '+ basis"'" crystal structure
(2)

The lattice and the translation vectors 8j, 82, ~ are said to be primitive if any two points r, f ' from which the atomic arrangement looks the same always satisfy (1) with a suitable choice ofthe integers III> liz, 113' With this definition of the primitive translation vectors, there is no cell of smaller volume that can serve as a building block for the crystal structure. We often use primitive translation vectors to define the crystal axes. However, nonprimitive crystal axes are often used when they have a simpler relation to the symmetry of the structure. The crystal axes a), a2, a3 form three adjacent edges of a parallelepiped. If there are lattice points only at the corners, then it is a primitive parallelepiped. A lattice translation operation is defined as the displacement of a crystal by a crystal translation vector
(3)

Any two lattice points are connected by a vector of this form. 1b describe a crystal structure, there are three important questions to answer: What is the lattice? What choice ofaj, a2, a3 do we wish to make? What is the basis? More than one lattice is always possible for a given structure, and more than one set ofaxes is always possible for a given lattice. The basis is identified once these choices have been made. Everything (including the x-ray diffraction pattern) works out correctly in the end provided that (3) has been satisfied. The symmetry operations of a crystal carry the crystal structure into itself. These include the lattice translation operations. Further, there are rotation and reflection operations, called point operations. About lattice points or certain special points within an elementary parallelpiped it may be possible to apply rotatiOns <11,,1 reflections that carry the crystal into itself. Finally, there may exist compound operations made up of combined translation and point operations. Textbooks on crystallography are largely devoted to

Ct'JjIUll Strlld .....;

Figure 2 Portion of a <Yl'~ of an ima(onary potItein ~k, in a twQ-<!imen>;imaI "'Qrkl_ (We pt(ked a protein molet.'tlle bcClIuse it is not likely to ha\'e a special symmetrrclits own.jIl>ealorlllc arran~emellt in the crystal looks Cl<lll-11y the same to an observe.' al r' lIS to an ob.cner at 1', pr(l\1ded that the \'eclOr T ...hkh connts r' and r mar be (',;pressed aJ an intewa' multiple of the \'t'cton ., and .2' In Ihi. illustration. T .. -., + 3a... \'ed0l'5 ., and .,aft' pr;miti'~ tnnslalion ,ecton of thf, rv.'O-dimemicoal Iatlice.

-n.e

([)

"

0
~

0
(;)

Figure 3 SI",ilar 10 fig. 2. but .... ith proleill molule. auoclalCl! in pairs. The ~'r')'l;laI tramlalion \"t.'don are a, alld 82' A rotation of.,. tat!lans about any poinl marked )( will carry the (.~~bI.' ,nlo i1se!( lNs OCC\lrs aUo lOr- C<jun-alent points in other cells. but "'~ ha,~...,.,.1<ed the points )( only ""thin 0I"Il' cell.

the description ofsymmetry operations. 'l1le crystal structure of Fig. 2 is drawn to havc only translational symmetry operations. Thc CT)'stal structure of Fig. 3 allows both translational and point symmetry operations. Basis and tile Crystal Srructure A basis of atoms is attacht.-d 10 every lattice point, with t.'Very basis idcntical in composition. arrangement, and orientation. Figure 4 shows how a crystal structure is formed by adding a basis to every lattice point. The lattice is indi+ catcd by dots ill Figs. 2 and 3, but in Fig. 4c the dots are omitted.

Figure 4 "1'e cr}'~lal structure is rormed by the addition of the b;u;is (b) to e~'ery lalli,'" point orthe lattice (a). By looI<in~al (cJ, you can feC<Jgni7.e the bm;is and then you can abst~t the space lattice. II does not matler "'here the basis is put in rc~1lion 10 a lanke point.

The number of atoms in the basis may be one, or it may be more than one. The pmilion of the center of an atom j of the basis relative to the associated
lattice point is
(4)

We may arrange the origin, which we have called the associated lattice point, sO that 0 :5 Xj, Yj, zJ:S 1.

l'rimitice ulttice Cell

The parallelepiped defined by primitive axes <ll> a2. 33 is called a primitive cell (Fig. 5b). A primitive cell is a type ofrell aT unit cell. (The adjective unit is supcrlluQlls and not needed.) A cell will fill all space by the repetition of suitable crystal translation operations. A primitive cell is a minimum-volume cell. There are many ways of choosing the primitive a.xes and primi.tive cell for a given lattice. The lIumber of atoms in a primitive cell or primitive basis is always the same for a given crystal structure.

CryltDI Shwc'....

"

I
"

-"

"

k::::k::::

"

,,'

fig...-e s.. Uttice point:l of a spa<:<: latl~ in two dinll!'ntionll. All pain of vectors 'I, 's an" tnnslatloo ,ten of the lat~. But ,,"'. lis'" are not primit"~ tnndatlon ,'I!'ctof's because ",-e cannot bm the bl;tice tntuJation T from IIltcognl combinatiom ci ,,'" and's... AD other pain of a, and 's 1M)' be taken as the primiti\..., tr21Sbtion ,~of the bl;tic.-e. "The r-nD~ gnms I, 2, 3 are eq.w in area and..,)' ci thffil muld be t....('n ti the pn""t,,-e cd!. "The parallelogram 4 has tv.ice the area cL. primitive tt1t.

.,

sno.-...

Figure

5b Primith-e.,.,11 of.

l.~

lattic.'O:' in three dimtfuiollS.

Figure 5c SufIPOS':' tht"Sc points w'e identical atoms: sketch in On the figure a set oflallk'O:' points, a

dlOice ci pirnit"'e 'ue, a primiti".. cell, and the basis of atoms auociatcd with a lattitt poinl.

There is ah.,"3)'S one lattice point peT primitive cell. If the primitive cell is a parallelepiped with lattice points at eaciJ of the eight comers, each lattice point is shared among eight cells, so that the total number of lattice points in the cell is one: 8 x "" 1. lbe volume of a pal"'J.lIelcpipcil with axes 8\. 8z, 83 is
VC ""la\a2
11

",I.

(5)

by dcmentary vector analysis. The basis associated with a primitive cell is called a primitive basis. No basis contains fewer atoms than a primitive basis contains.

Figure6 A prim;I;,'''' cell may also be dlOSen follO\<" ill~ this procedure. (1) draw linCll 10 connect a givt"n I"Wee point 10 all nearby lalliCO' points. (2) at the midpoint and normal to these lint'll, draw new lines Or planes. The smallest volum.. endOl;ed in this way is the \\'igner-Seilz pril1lili,-e <:<."11. All spaOf' may be filled by theS<" ~-ells. just as by the cells of l'"ig. 5.

Another way of choosing a primitive cell is shown in Fig. 6. This is known to physicists as a Wigner-Scitz cell.
'UNDAMENTAL T\'PES OF LATIlCES

Crystal lattices can be carried or mapped into themselves by tlle lattice translations T and by various olher symmetry opcmtions. A typical symmetry opel1llion is that of rotation about all axis that passes through a lattice point. Lattices can be found such thai one-, two-, three-. four-, and sixfold rotation axes carry the lattice illto itself, correspollcling to rOtations by 21T, 2m2, 2m3, 21T/4, and 2m6 radians and by integral multiples of these rotations. The rotation ..J by the symbols 1, 2, 3, 4, and 6. axes arc denoh. We cannot find a lattice that goes into itself under other rotations, such as by 21Tn radians Or 2n/5 'ddians. A single mok'cule properl)' clt.:signed can have any clegree of rotational symmetry, but an infinite periodic lattice cannot. We can make a crystal from molecules that individually have a fivefold rotation axis, but \\'C should not exp<.'Cl the lattice to have a fivefold rotation axis. In Fig. 7 we show what happens if we try to construct a periodic lattice having fivefold symmetry: the pentagons do not fit together to fill all space, showing that we cannot combine fivefold point symmetry with the required translational periodicity. By lattice point group we mean the colk'ction of symmetry operations which, applk><l about a lattice point, C'oIrry the lattice into itself. The possible rotations have been listed. \\le can have mirror reflections m about a plane through a lattice point. The inversion operation is composed of a rotation of 1T followed by reflection in a plane normal to the rotation axis; the total effect is to replace r by -r. 'l1le symmetry axes and symmetry planes of a cube are shown in Fig. 8.

Two-DimClIsiOlUd Lattice ly,Jes

1llere is an unlimih. ..J number of possible lattices because there is no natural restriction on the lengths of the lattice translation vectors or on the angle !p between them. TIle lattice in Fig. 5a was drawn for arbitrary 81 and a2' A general lattice such as this is known as an oblique lattice and is invariant only under rotation of 1T and 21T about any lattice point.

CJ1Pftd Stntel"

F"gure 1 A fi~ axis of S)mmetry cannot eo:'st in periodIC lalla beowsc il is not l1oOUibl'" 10 flU the area of II p\arno "ith a
conntct~

array of pt'11t~. Wecan,

hO'\O~

e--er. flU all the area or. plane wit), just two dislinct dcUgns ri "tile.- or .. kmenlar)
poI)llO"I- A quasio)'lital is quasiperio:ldic nonnndoo> auc:mbh'o(rwo typei offigures. Quasicr)stals IITC diK'uslial at the md of

Chapter 2.

"'

"

Figure 8 (-l A pi....., of i)TT1metry po.raIId to the ~ of. cubf,. (b) A diagnnal plane ofs)'n,r"etry in a ""be. eej '11", Ihret' tetrad axes ofa t... ~. (dl TIlt' roll. triad ues of II ellbo.-. (..)The six diad Il.~CS of a (:ubc.

"'

10

But special lattices of the oblique 'rile can be im'ariant under rotation of
2'll'f3. 2'17-/4, or 2.m6. or under mirror reflection. We must impose restrictive

conditions on 3. and 3i if we want to construct a lattice that will be invariant under Olle or more of these new operations. There arc fouT distinct types of restriction, and each leads 10 what we may call a special lattice tn>e. T1lus there are fh'c distinct lattice l}'pes in two dimensions, the oblique I.ttlice and the four

special lattices shown in Fig. 9. Bra,'ais lattice is the common phrase for a
distinct lattice type; we say that there arc five Bravais lattices or nets in two dimensions.

TllreeDimelisicmal Louiee

]ype.

The point symmetry groups in three dimensions J'C(luirc the 14 different lattice types listed in Table l. The general lattice is tricliuic. and there are 13 speciallautces. lbcse are grouped for convenience into systems classified according to se',C11 types of cells. which are tric-lillic, monoclinic, orthorhombic, tetragonal, cubic, trigonal, and hexagonal. The division into systems is expressed in the table ill temu of the axial relations that describe the cells. The cells in Fig. 10 are conventional cells; of these only the sc is a primitive cell. Often a nonprimitive cell has a more obvious relation v.ith the point symmetry operations than has a primithe cell.
Table I nle 14 lattice tn>es in tlnee dimensions
";"mlJer

of
S~'l;lem

Ialllces

l\eslridious on conventional ~'dl a:o:cs ~nd 311/1:Ies

a\;.o!'a2';"1aJ

Tridinic

ayi.pv.y Monoclinic
2
al ;.o!'

a'.!. "" aJ

Ol1!JorllOlllbic
Tctr;\b'Onal

,
2

a=y-OO'v"p
al~a2~a3

a-p-y=OCJO
al '"

a'.!. '"' O.l

""'P"'y=9O"
Cubic
Trigonal Hexagonal

111- 112 '" 03

a-{J"'y=9O"
o,=a2=o3

a - p - y < 120". '"

0,-02,",03

9(f

a-p-9(f y - 12Cr

Crpl.al Sfr.-:,urc

II

f]

(aJ Sq...." lattn

f,l loaG ., .. llO"

r~EJ

\\

"

"
~

al~;..{; .. -

,<I)
c-n~
ilU'S

"d<.

reMallcuI.-L.ltic:c. lor hoth the pnnlitM:: n and b the

are
t",:tan&War uuit ~n for

"hidlla,r .. la.I,..
Fig,,", 9

90"

Flgure to TIle cubic .pace

lalt~.

The

~1I h",,"u

arc lhe COlwenllonll! cells.

"
Table 2
Olaracleri.'itics of cubic I:dl~
SIiJ'pIc
Boc!."-<'II:ntered
F~I~

Vol moe, collventional C('II

L"'1ttkc poiuls pcr c:cll Volume, prhniti\'c cdl !..atlice points per unit "olurnc

f'umber nearest Ilcighbo.5' 1'\earc:;I.neighlxw distance Number of second neighbors Sc<:onu neighbor di~l:1nce
J>ackin~ fra<.1ion l,

or

" I " ,,0' 6 ,

12
2 1120

0'
2

"
,0' 410'
12
al2lf! .. O-7C17a

10' ,,0'
8
31/la/2" 0.66& 6

6 ,

I.
-0.524

hV3
=0.660

hrv2
=0.740

"'T.tJks numben dnrigflbon and dislanCeS in Ie, I.rc. ~. Jx..p, and cbamond ilrut1urel.e gi\U'lon pp. 1007-1009 dJ. IlLnchrdder, C. F. Curti> and It 8. Bird, MokaJ"r thftwy tll'JIMf .,nd liquiih. Wiler. 1964. "l11e packing frllclioo i. tho: ,muimum proportion of lhe ~\'ailahle ,,,Iu'ne that can be filled "~Ih hard .pheres.

111crc arc tl.f(,.'e lattices in the cubic system: the simple l1Jbic esc) lattice, the bod)'-centcred cubic (Lee) lallice, and the face-<:entered cubic (fcc) lattice. The charactcri~1ics of the three cubic lattices are summarized in Table 2. A primitive cell oflhe bee lattice is shown in Fig. 11, and the primitive translation "e('(ors are showlI in Fig. 12. The primitivc translation vectors oftile fcc Lattice are shown in Fig. 13. Primitive cells by definition contain only one laltk-e point, but the conventional bee cell contains two lattk:c points, and the fcc cell contains four lattice points. The position of a point in a cell is sp(,ocificd by (4) in terms of the atomic coordinates x, y. z. Here each coordinate is a fraction ofthe axial length 0 .. 02, 03 ill the dircctiOll of the coonlinate axis, with the origin t<lken at one l'Orner of the cell. Thus the l'OOrdinates ofthc bod)' center ofa cell are UI, and the face ('Cnters include +f0. OAt; 10l. III the hellagonal system the primitivc cell is a right prism based on a rhombus with an included angle of12(f'. Fif,,'Ure 14 shows the rclafionshipofthe rhombic cell to a hexagonal prism.
INDEX S'fSTEM "'OR CRYSTAL PLAIIOES

11le orientation of a crystal plane is detCnllined by three points in the plane, provided they arc not collinear. If each point lay 011 a different crystlll u\is, the plane could be specified by giving the coordinates of the points in tenns of the lalti-l'e constants 0 .. 02> 03'

ergs/ill Strudul1I

'' ft
I ,
.

,,,,,"r--

1 1

---...
FIgUre II Bod)'"C.'ft1tercd cubic \atticr. d>owin~ 8 primil;"'e cdl. The l2!imitive ~U s"--n is 8 rhombo-hedron of edge I V3 D. and the angle betwe.'ll adjact.....1 edges is 109"28'. Fipre 12 Primitive tnnsIation ,...ctorl; Ihe body-c.uI. lered cubie lattice; Ihrse \'oclon <:oone<:1 the Iatlice poinl at the origin 10 Iatlice points 81 the body ocnten. The primilive cell is obtaioed on completing the rhomoohce!ron. 10 terms of the lube edgeD the primillvt lnmslatinn vtdors are

a.-la(i+f-t):

~-la(-i+9+i)

1l3"'llI(i-y+i) .

r l

aj-la(i+S");
sian unit \'eClou.

A"'--~,------jI

/ t I""-----c-'I"'--+---j'"('< , ,

--

I , , , , ,

.,
,

, , , ,

).of--h:-n-- --, /
~"_ ~r

FIIU'"t 13 ~ rhomboh('(!ra! primit,,c ceU of Ihe ie-olened cubic: O)'Iital nit prinliti"e translation ,'mOl'S a" a,z. a" connect the lattioe point at Iht origin y,11h lallice points at lht ra<'C l'tllters. As do.wn. the primiti"e vectors 1Irt':

Fisure 14 ~blion dthe primitive eel in the bc~l system (heavy lines) l< a prism of htu~nal s)'mmetr)'. 11.,.<

at- a

.""3

_.-la(9+1);

_s-I,,(i+i)

'The angles betw.-.en the ~ arc 60". Ilert 1, 9. t arc the Carte-

.
,

"

,
Hgurc: 15 This plane intercepts the a" HI. "3 ...:1 al 311,. 211 2 2a.J. The: rt.'<f.procals of these ""mben; arc I, t I. "n,l: sm"lIl,)! thn-e .ole",," h"\'jng the I;lllnC r1/.tio are 2. 3. 3, and tlull> the indil'CS of the plMe are (23J).

Ho\\"e\ cr, it turns out to be more useful for structure analysis to specify tlle

orientation of a plane by the indices dctctinincd by the follo\\;ng rules (Fig. 15).
Find the intercepts on the axes in tcmlS ofille lattice constants"" 61. 03_ TIle axes may be those of a primitive or nonprimiti\'c cell . Take the reciprocals of these numbers and then reduce to three integers having the same ratio, usually the smallest three integers. nlC fCsult, enclosed in parentheses <TIki), is called the index of the plane.
For the plane whose intercepts urc 4, L, 2, the rcciprOC'olls arc i, 1, and i; the smallest three integers having the same r.ltio arc (142). For an intercept at infinity, the CO!"rcsponding index is zero. l11e indices of somc impot'tant planes in a cubic crystal are illustrated by Fig. 16. Thc indices (hkl) may dcnote a single plane or a sct of parallel planes. If a plane cuts an axis on the negative side of the origin, the corresponding index is negative, indicated by placing a minus sign aboo,'e the index: (11M). -O.c cube races of a cubic crystal are (100), (010). (001), (100). (010), and (001). Planes equivalent by symmetry may be denoted by curly brackets (braces) around indices; the set of cube faces is {l00}. When we speak of the (200) plane we mean a plane par-dlel to (LOO) bul cutting the al axis at la. The indices 11Il'1V) of a dircdion in a crystal are the sct or the smallest integers thai have the ratio of the components of a vector in the desired dif(~c lion, referred to the axes. TIle at axis is the \1001 direction; the - a2 axis is the

C,yflol SIron",,.

15

y'

--./

Ii.

"'"

-----

./'
(Il~

./
(1lI1

,,'"
Figure 16 Indices ~ unportant plancs;o a cubic <.T}-!ital. 11>c pbne (1llO) b pamilcllo (100) and to

(ioo,.

LOIO) direction. In cubic crystals the direction lhkL} is perpcndicular to a planc (hkl) having the same indices. but this is not gcncrally truc in other crystal systcms.

SIMPLE CRYSTAL STRUcrUflES

We discuss simple crystal structures of genenl! interest: the sodium chloride. cesium chloride, hexagonal close-packed, diamond, and cubic zinc sulfide structurcs.
SOtlillm Chloride Sfnlcfllre
'111C sodium chloride. NaCl. structurc is shown in Figs. 17 and 18. 'Ole lattice is face-centered cubic; the basis consists of one Na atom and OIlC CI atom scpar,ltoo by one-IJalfthe body diagonal of a uilit cube. 11lCrc are four units of NaCI in cach unit Ctlbe. \-\;th atoms in the positions

Cl

000

110 ;
001;

III

101 ; 010 ;

011
100

16

l'igun 17 We ma~ mstruLi the sod",m chloride cry$. tal stnK1ure b)' arral,j;ing Nfl and CI i"m "hcmntcly at the lattice points of II simple C\Jbic laltil.~ In the ~'S-- tal <:ad, ion issurroun<k'd b). Ii>: nc:arest neighbon efthe oppmitc charge. Thc siJiI'l'O' laltia: is It-c, and the l>a$Is has one CI ion at 000 and one "a~ iooat 411- The twc ........, one COlwcntional a.llk- .,."U. Thc Im,il' diamclcr1i here IIrt: n:dured '" ..... latioll to the cell in Old..... to writ)
tht: Spollial arTlln~ltl ..nt.

figure 18 Model or sodium chloride. ~ xdlOOl ions are srna)k,r tl.n the cWorine Ions. (CcJurte$)' cf A. N. Hold.." , am! P. Singer.)

Hgurc 20
(Uri"

n.... s~ built: is simple .....b~. and the

1011

n,t: <;'eli"", d,loridc cry>lal strut'a' 000 and one d

ion al

t-';glfre 19 r\alunl co",-ak cllcad ."tllde, PbS, whic:fl has the: KaO '-T).~ Itructu= (l'bdogJ"iph h). B. Burleson.)

haMs bas une Co

til

C~al

Slnod"....

17

Each atom has as nearest neighbors silo; atoms of the opposite kind. Representative crystals having the Nael arrangement include those in thc following table. "Ole cube edge a is given in angstroms; 1 }. - 10-8 em _ 10- 10 m -.9,.j nm.
CTyow

uH
Mg"
~1,nO

1.08 1. 420

~;.cl

-I.n
563

CT>staJ AgBr rbS KCI 1Ul'

5.17 1. 692 629 659

Figure 19 is a photograph of crystals of lead su16de (PbS) from Joplin, Missouri. The Joplin specimens form in beautiful cubes.

Cesium Chloride Structure

111e cesium chloride stTUcture is shown in Fig. 20. There is one molecule per primitive cell, with atoms at the comers 000 and body-ccntcred positions iii of the simple cubic space lattice. Each atom may be viewed as at the ccnter ofa cube of atoms of the opposite kind, so that the number of nearest neighbors or coordination number is eight.

Crystal
BLoCu A1NI

CuZo C/l-bruu)
CuPd
Ag:\tg

2701. 2.88 294 299 3.28

Crystal Lillg NH..CI

no,

esCI

nt

3291. 3.67 3.97 411 420

Hexagonal Closepacked Structure (hcp)

There are an infinite number of ways of arranging identical spheres in a regular array that maximizes the packing fraction (Fig. 2L). One is the facecentered cubic stnlcture; another is the hexagonal dose-packed structure (Fig. 22). The fraction of the total volume occupied by the spheres is 0.74 for both structures. No structure, rCl'(ular or not. has dense.- packing. Spheres are arranged in a single closest-packed layer A by placing cach sphere in contact with six othen. This layer may serve as either the basal plane eX an hcp structure or the (111) plane of the fcc structure. A second similar layer B may be added by placing each sphere orB in contact \\ith three spheres of the bottom layer. as in Fig. 21. A third layer C may be added in two ways. We obtain the fcc structure jfthe spheres ofthe third layer arc arkled over the holes in the fint layer that are not occupied by B. We obtain the hcp structure whcn the spheres in the third layer arc placed directly O\'cr the centcrs of the spheres in the first layer.

IS

"
fo'igure 21 A close-P"K'kcd la)'erol" ~phcres is shO\vn, with (;enters alpoints marh..! A. A s.,coud and identical layer spheres UlO1 be ploct:d on lop of this, above :lnd parallel to the plane of the dmwing, with centers o>'cr the points marh>tl B. TIler" are two choices 'Or .. third tarcr. It can go

or

in over A or over C. If il goes in over A the SCQucn,-e;s ABA8AB. . and the Sindure is hexagonal close-palked. If the thin! J.1)'er gol.'S in O\'CT C the sequence is tlBCABCtlBC . .. and the sIn.. is r~ntcred olLic.

,t",,,

, ,
:
,
,

.. a
"
83

::;,

I'igurc 22 The l,cxagonal cIO!>C-l'acL.:d sinK" lure. Th" atom positions in Ih~ slru<.iurl; do not collstitute a Spalt: lalli..,. 111e space lattice

II II

is .impk hc~agona' with a basis or tw" wenli. ,'Ill aloms associated with cadi Iallicc point. 11.., ["Ilil.., parameters a and c arc indic-~tcd,
where a is in the basa11'lanc and c is the mag-

nitude of the axis

of Fig. 14.

t-'igu~c 23 The primili~" ~dl has u, ~ "2. wilh all il1c1 uded "'11g1e of 12(f. The c axis (or a,) is 110r"'al 10 Ihe plane of a, and R TIle ideal rep sln",I"re has c - 1.633 u. The t",o atollls of one basis arc shown as 501id drcles. One atom of the basis is at Ihe origin; the olhe~ alom is aiIH. whid, meanS allhe posilion ~ '" la, + Ia. + Ill;j.

'I1le hep structure has the primitive cell of the bexagonallattice. but with a basis of two atoms (Fig. 23). The fcc primitive cell has a basis of one atom (Fig. 13). The ratio da (or a:Ja]) for hexagonal closest-packing of spheres has the value (1)112 = 1.633, as in Problem 3. It is usual to refer to crystals as hep even if the actual da ratio departs somewhat from this theoretical value. 'I1le number of nearest-neighbor atoms is 12 for both hep and fcc st~uc lures. If the binding energy (or f~ee energy) deper '~d only on the number of

II

\,

1 CIJPIDl S~'u....

19

____

--------tigure 24 Atomic, po6ltion~ In the cubic t'ell cithe diamond Mrudure projce:ted on ;I. CtlLe lee; lBctions denote height a1Jo,." the b:ue in unib of a cube o:..:Jge. n.c points M 0 and i an: on the r.:c Lottic:t'; tho.ie at 1 and J arr on ;I. "milar lattice di.pb<:t'd;Uong the body daagonal by one-burtl, ci ;t~ With a i:c space 1at1iC'e. the basi.s consnts d' two identical

.orm

Ie.
{Ia

Hguf'C 2S Cry~tlll ~truct"re ci diamond, sho...ing the tetr3hcdral bond armngement.

.
, ,
I

~'~

M 00l);

Uf,

nearest-ncighbor bonds pet" atom, there \\'oukl be no difference in energy between the fcc and hcp structures. CI)')lal lie
Be \Ig

Cia
1.633 1.581 1.623

Cl)~hll

C't)o':;tul

cia
1.594 1.592 1.586

Zn

<AI
Co
y

1.861 !,BS6 I, fiZ2

1.5iO

Z,
Gd La

Ti

f.58G

Diamond S'ructure
The space lattice of diamond is fcc. 111C primitivc basis has Iwo identical atoms at 000; tU associated v.ith eadl point of the fcc lattice, as in Fig. 24. Thus the coll\"cntional unit cube contains eighl atoms. Thcre is no way to choose Ihe primitivc ccU such Ihat the basis of diamond contains only one atom. TI1C tctrahedral bonding charactcristic of thc diamond str'lcturc is shO\!o'n in Fig. 25. Each atom has 4 nc:trcst ncighbors and L2 next nearest neighbors. The diamond structl1re is relatively empty: the mro;imum proportion of the availablc volumc which may be filled by hard spheres is only 0.3-1, which is 46 perccnt ofthe filling ~ 'tor for a c1osest-pucked structure such as fcc or hep. TIle

.
diamond structure is an example of the directional covalent bonding found in column IV of the periodic table of dements. Carbon. silicon, gcnnanium. and tin am crystallize in the diamond structure, with latticec:oostants 0= 3.56, 5.43, 5.65, and 6.46 A, respcctively. Here (I is the edge of the conventional cubic cell.

Cubic Zinc Sulfide

Stnl(:tu~

1hc diamOlld structure may be viewed as Iwo fcc structures displaced from eadl other by one-quartcr of a bod)' diagonal. The cubic zinc sulfide (zinc blcndc) strudure results when 1.11 atoms arc placed on one fcc lattice and S aloms on the other fcc lattice, as in Fig. 26. l11c conventional cell is a cube. The coordinates of the Zn atoms arc 000; OU; ",04; aD; the coordinates of the S aloms arc iH, ~~i; ~U; Iii, '111C lattice is fcc. '11lcrc arc four molecules ofZnS per conventional cell. About each atom there arc four equally distant atoms of thc opposite kind arranged at the comers of a regular tetrahcdron. 111e diamond structure allows a centcr-ofinversion symmetry operation at the midpoint of cvcry linc between nearest-ncighbor atoms. The ilwersion operation carries all atom at r into an atom at -r. The cubicZnS stnlcture docs not h;l.\'c inversion !oymmetry. Ellampies of the cubic zinc sulfide strudure are
Crystal CuF SiC cuel ZnS

a 4.26 A
4.35

C~tal

ZnSc

GaA.<

5.41 5.41
5.45 5.45

AlAs

CdS
IIlSb
AgT

AlP
CaP

a ,.65 5.65 5.66 5.82 6.46

6.47

The dose e<Juality of several pairs, notably (AI,Ca)P and (AI,Ca)As, makes possible the construction of semiconductor helerojullclions (Chapter 19).

DIRECT IMAClNC

or ATOMIC STRUCTURE

Direct images of crystal sb"uctufC have been produced by transmission electron microscopy. Perhaps the most beautiful images are produced by scanning tunneling microsCOll); in STM (Chapter 19) one eJCploits the large varialions in quantum tunneling as a function of the height of a fine metal tip above the surface of a crystoll. The image of Figure 27 was produced in this way; sec also Figures )2.19 and 19.21. An STM method has bccn developed that will assemble single atoms into an organi7.ccllayer nanomcter structure on a crystal substrate: sec the e1ectnm cornd in Figl.we 19.21.

'"'"
\

, ,

,
'""

-f,.
.,

'.

"-

t'iI;Utt 27 A K1Uming tunnclin,; microsrope image d' atoms (:0'1. (III) surl'al..:l d platinum" 4 K. The nearest ncigbbo.- spacin" is 2.78 A. (Photo courtes)'d'D. M. E,gkr. IBM RC5CaI"d1 DiYkion.)

NONIDEAL CRYSTAL STRUCI1.JRES

1bc ideal crystal of classical cryst-.tUographcrs is formed by the periodic repetition of identical units in space. But no general proof has been given that the ideal crystal is the state of minimum energy of identical atoms at absolute zero. At finite temperatures this is nol likely 10 be tOiC-sec the discussion of IattW:e defects in Chapter 18. Further, it is not always possible for a structure to attain the equilibrium slate in a TCasulablc .time-see the discossioo of glasses in Cbapter 17. Many stTUCturcs that occur in nature are not entirely periodic; sec the quasicTystals treated at the end of Chapter 2.. We give some examples here that supplement those in the chapters just cited.

"
Rarulom Stacking aJUI Polytypism
TIle fcc and bcp structures arc made up of close-packed planes of atoms. The structures differ in the slacking sequence of the planes. fcc having the sequence ABCABe ... and hcp having the sequence ABABAB . . Structures are known in which the stacking sequence of close-packed planes is random. This is known as random stacking alld may be thought of as crystalline in two dimensions and noncrystalline or glasslikc in the third. Polytypism is characterized by a stacking sequence with a long repeat unit along the stacking axis. The best k'TlO\VIl e;o;ample is zinc sulfide, ZnS, in which more than 150 polytypes have been identified, with the longest periodicity being 360 layers. Another example is silicon carbide, SiC, which occurs with more than 45 stacking sequences of the close-packed layers. The polytype of SiC known as 393R has a primitive cell with a "" 3.079 A and c "" 989.6 A. The longest primitive cell observed for SiC has a repeat distance of 594 layers. A given sequence is repeated many times within a single crystal. The mechanism that induces such long-range crystallographic order is not a long-range force as such, but is associated with the presence of spiral steps due to dislocations in the growth nucleus (Chapter 20).
CRYSTAL STRUcnJRE DATA

In Table 3 we list the more common crystal structures and lattice structures of the clements. Values of the atomic concentration and the density are given in Table 4. Many c1emellts occur in several crystal structures and transform from one to the other as the tempcratureor pressure is varied. Sometimes two structures coexist at the same temperature and pressure, although one may be slightly more stable. 'ne reader who wishes to look up the crystal structure of a substance may consult the excellent compilation by Wyckoff listed in the references at the end of the chapter. Sin/ell/re Reports and the journals Acta Crystallographica and Zcitschrift lur Kristallographie are valuable aids.

\
\

H' ~ hop
3.75 6.12

T.bk 3 Crystal structures or the elements

1bc data given are at room temperature ror the most common rorm, or at
the stated temperature in deg X. For further descriptions or the clements see Wyekoff, Vol. 1, Chap. 2. Structures labeled complex are described there.
hop 2.27

H.4 2K

hop

Lim B.
boo
3.491

- Na '"

3.59
$I(

,_. ...... ... -...

B C H"" '.66 P

,.83
F
N.~

3.57

3.567

<0,)
5

4.46 CI

f"

Mg hop
3.21
5.2~

4.225

, , ,
5, hop
3.31 5.27

..:~Crystal

structure. a lattice parameter, In A c lattice parameter, in A.

M,
Q1b~

, , ,
NI fop 3.52 Pd f" 3.89
pt

4.05

K"

C.
I"

'" 5.225
Rb"

TI hop
2.95

C,

F.
2.87

'.58 5, fop '.08 B.

5.02

'" 3.03
~

bo'
2.88 Mo
3.15

gnplu

'"

Co hop
2.51 4.07

C.
I" 3.61

l'
hop 2.66 43' Cd hop 2.98 562

4.68

y
hop 3.65 5.73

l,
hop 3.23
5.15

Nb 3.30 T.

'"

5.585

'" '"
W boo

T' hop
2.74

R.
hop 2.71

Rh fop 3.80

Ag 4.09 A. fop 4.08

I"

I,

4AO
R. hop
2.76

4.'"

tett. 3.25 4.95

-,. - -... '" -.... ...... .. -,,, _ . ...., ... -...

AI

(Nt)

......
51
5.430

An,l(
fop
5.31

...

(ell)

G.

G.

A.

5.

Bf

Kr

.K

11Mb.

5.658

(Grl )

fop 5.64

Sn (a) 5b

T.

x
f" 6.13

1(

6.49

(fIIiIl

(I,)

C,,'

La
hu.
3.77 ABAC

Hf
hop 3.19

O.
hop
2.74

M4S
F,

'" '"
R.

'.05
C.

'" 3.30
pt

3.16

4A'
Hd h". 3.66

fop " 3.84

fop
3.92

-.
Oy

Hg

TI hop
3.46

Pb f"
4.95

....,
BI

Po 3.34

At

R,

4.32

"

l. h<p 3.SO '.66

'.52

A,
f" 5.31

Pm

I" 5.16

hu.
3.67

ABAC

..... '"
5m

E.

4.58 Am

Gd hop 3.63
5.78

Tb
hop

3.60
5.70

hop 3.59
5.65

Ho hop 3.58 '.62

E,
hop 3.66 '.59 Fm

Tm hop 3.64 '.66 Md

,...
f"

Yb

Th f" '.08

Po
tett.
3.92 3.24

U
mmplu

-".

Np

CllOllplu

".

Cm

Bk

Cf

E.

No

L,

he,.
3.64 ABAC

-_

-,-~-

--

---

-_.-

~J

,
H"
0.088

Table 4

Density and atomic concentration

He

2~

The data are given at atmospheric pressure and room temperature, or at the stated temperature in dcg K. (Crystal modiflcatlons as for Table 3.)

0.205
It)lItm)

li

78J(

B.
1.82 12.1 2.22

B
2.47 13.0

C
3.516 17.6 1.54

N~

F
1.44

0.542 4.700 3.023 Na

5K

1.03

Ne 41( l.51 4.36 3.16

AT 4K

M.
1.74 4.30 3.20

l.013 2.652 3.659

, , ,
5,
2.99 4.27 3.25

Density in g cm-l (lOlkg m-1) Concentration in lOll em-I (101' m- 1)

Nearest-neighbor distance, in A (lO-IOm)

,
Co
8.93
8.45 2.56

Al
2.70 6.02 2.66

51
2.33 5.00 2.35

CI93K

2.03
2.02

1.77 2.66 3.76

K"
0.910 1.402 4.525

C,
1.53 2.30 3.95

TI
4.51 5.66 2.89

V
6.09 7.22 2.62

C,
7.19 8.33 2.50

M_
7.47 8.18 2.24

F.
7,87 8.50 2.48

..,
C, Rh

NI
8.91 9.14 2.49

Z_
7.13 6.55 2.66

G,
5.91 5.10 2.44

G.
5.32 4.42 2.45

As
5.77 4.65 3.16

S.
4.81 3.67 2.32

BT 1231'>

Kr

4K

8.97 2.50

4.05 2.36

3.09 2.17 4.00

XC 4K

Rb

5K

5,
2.58 1.78 4.30

Y
4.48 3.02 3.55

Z,
6.51 4.29 3.17

Nb
8.58 5.56 2.66

M'
10.22 6.42 2.72

T'
11.50 7.04 2.71

Ro
12.36 7.36 2.65

Pd
12.00 6.80 2.75

Ag
10.50 5.85 2.89

Cd
8.65 4.64 2.98

I7.29 3.83 3.25

S_
5.76 2.91 2.81

Sb
6.69 3.31 2.91 SI 9.80 222 3.07

T.
6.25 2.94 2.86
Po 9.31 2.67 334

I
4.95 2.36 3.54

1.629 1.148 4.837

CS Sl<.

12.42 7.26 2.69

3.78 1.54 4.34

B,
3.59 1.60 4.35

l,
6.17 2.70 3.73

Hf
13.20 4.52 3.13

T,
16.66 5.55 2.66

W
19.25 6.30 2.74

R.
21.03 6.80 2.74

Os
22.58 7.14 2.68

PI
22.55 " 7.06 2.71 21.47 6.62 2.77

Ao
19.28 5.90 2.88

Hg227 TI
14.26 4.26 3.01 11.87 3.50

Pb
11.34 3.30 3.50

Af
-

R_

1.997 0.905 5.235

lo
'.84
3.39 3.43

3A6

F,

R,

A,
10.07 2.66 3.76

I'
C.
6.77 2,91 3.65

P,
6.78 2,92 3.63

Nd
7.00 2.93 3.66

Pm

Sm
7.54

Eo
5.25 2.04 3.96

Gd
7.89 3.02 3.58

Tb
8.27 3.22 3.52

Dy
8.53 3,17 3.51

H,
8.80 3.22 3.49

E,
'.04 3.26 3.47

Tm
9.32 3.32 3.54

Yb
6.97 3.02 3.88

N.
20.45 5.20 2.62

3.03 3.59

Th
11.72 3.04 3.80

P,
15.37 4.01 3.21

U 19.05 4.80 2.75

Po
19.81 4.26 3.1

Am
11.87 2.96 3.61

Cm

Bk
-

Cf
-

Es

Fm
-

Md

N,

l,
-

"

Cryr;lal

SUU('lu~

1.5

SUMMARY

A lattice is an array of points related by the lattice translation operator T Villi + ~z + tJJa3> where VI> ~ tJ3arc integers and ab a2. 113 arc thc crystal

"co.
To form a crystal we attach to every lattice point an identical basis composed of' atoms at the positions rJ := rjll + Yjlz + Zjl3' .....:ithj "" 1. 2, . . s. Here x, y, Z may he selected to have values between 0 and I.
111C

axes Ill' 112, aJarc primitive for the minimum cell volume tal a2 l< aJl for -d from a !Jttice translation operator T and which the crystal can be construch.. a basis at every lattice point.

Problems
I. Telrol.edmlongla. The angles betwecn the tctraJ'cdral bonds of diamond are the

same as the angles between th(' bod)' diagonals of a cube, as in Fig. 12. Use dcm('ntary ..ector analysis to fim the value of the angle.

2. Indiur of plllna. Comi<lcr the planes ...;th indices (100) and (001). the laUice is ~, anclthe indices refer to the OOflventional cubic cell. What are the indicn of these planes whcn referred to the primitive lUes of Fig. 13?

3. IIcp ,'rudure. Show that the do ratio lOr an ideal hexagonal c1ose-polcked stnK1ure is (1)112 = 1.633. If do is s.gtliftcantl)' larger IIlan this value. the crystal structure 100)' bt' thought of as composed of planes of ~Iy packed atoms, the planes being loosely
st.::ked.

Referemes
ELEMENTARY

W. B. f'eanon. C~tal chnn/.slry 11M ~sicf of mttDU lind alloys, Wiley. 1972. II. D Mega...-, Crysttll$lrudlllU; II 100",*.", 11",,"*""'. SaunOC'f'S, 1973.
CM1'STALLOCRAPHY

M. J Buergn-, Introduction to a'Jjntll pomelry, McCMlWHill, 1971. C Bums and A. M. Cla.ser. Space groups for lfIiid Jlllle physici.fts. Amdemic. 1978. f. C. Phillips, An introducllOn to cryslllllogra,"''1. <It I. 1.'<.1.. Wiley. 1971. A good pll1CC to begin. II. J JurctK"ke, Crysllli ph!fSlcs: II\6(;tl>JaIplc phfJria of 4"iroIropic lOlids. BenjamIn, 1914 B K Vains!\!cin. Mcxkrn cryltallogrwphy. Springer, t981. J. F. "ye, I'hfJdoal propr"ws 0/ ~Ial.l'. <Mad. 198.5.
CRYSTAL CROWnt

W C I'f&nn, Zoow melting. Koegn, 1976, t966c_ A. W. Vere, Crystm grot<.th. pnncipla ond progreu. Pknum, 19B7. J. C. 1Ince. Cryslol grou:lh p~., Balstoo. 1986.
S. 11 Uu, "Fradals and theIr llppllca.holl ill colllkll""tl matter physics," SuM.tatc llhyslcs 39, '1fJ7
(1986).

.
Series: JourtUll of Crystal Growth, mcluclcs prtK.'et."t.Iingsoftioc International Conferences On Cry~lal

Crowth.
Springer Scrft:s, Cl')"Ctals-Crowth. Properties:. and APIlItC3tiolis. CLASSICAL TABLES AND IlANDBOOKS

IntrnwztionD/ tabk,[orr-ro!J CflI&'Dllogrophll, K},ooch Prns, 4 volumes, Birmingham. 1952-1974. J. F. K)'e; Ph~ P"'J'f'",ja of Cf1I#uh, rhdr ~tDtlon by trnron ond I'IlIJt~. fui,rd,

P. Villars and L O. ('.al,..ert.. I'nlrfJOol', handborJ< of crysuUIOVflphi.:: deta /or Wtnndallk~. Amer. Soc:. Mehh. 3 voh., 196.5. A. F. Wells, St.......m.rol inorgIank chcmi#ry. 5th ed., Oxilrd Un;"'enity Preu. llllO. 1984c. W. C. \\')-duIIf, C'lIslm ztr'UCl'UJU. 2tld ed., Krieger, 1961.

,....

2
Rec;iprocal Lattice
DIFFRACfION OF WAVES BY CRYSTALS Bragg law SCATfEIU:!:D WAVE AMPLITUDE Fourier analrsis Reciproc:allaltK:e vectors DiITnK'tion conditions Laue equations BRlLLOUIN ZONES Reciprocallnuice 10 sc: Inllice Reciprocal lattice to be<: lattice Redprocal Illttice to fcc lattice FOURJER AAALYStS OF THE BASIS Structure factor of the bee lattice Slrudure factor the fcc: lattice Atomic form factor 29 29

30 30 33
34 36 37
40

4.
41

or

...
42

45 45 45

QUASICRYSTALS

SUMMARY

PROBLEMS
I. lnterplanar separation

.,
"

40

2. Hexagonal space lattice

3. Volume of BrillOtiin zone 4. Width of diffraction maximum Structure factor of diamond 6. Form factOf' of atomic h)'tIrogen 1. Dialomic line

" ., " ..
.2 '2

REFERENCES

,
.........

.,

f'

.s
.,,

"",,,-

r-...

,.

~_iY.O.OlfN

Aloton -cr, keY

",.

Fi~n::

Wavelength '"ft'SUS portio

Elt,on -.Y. lOOeV

cIc eror:1'gy. lOr photOIlS, neutmou, arK! dedmlls.

Figure 2 I.)erivali'm of tl'e Bragg ""Illation 2d li,n 9 - nA, Ii." .. d is the sl'Dcing ri parallel atomi~ planes 1111<12..." is the ,Jifferellal in phase between rcflroions from suC'Ct:ssivc Jllall<-... nl<: rdleClLng planes ha,c nothing to do with tI,e slIIface plall<,'$ bounding the pa'I~"b.r 1pl.'Cimcn.
2S

'IAPTER 2: RECIPROCAL LATIlCE

DlFFRACflOX OF WAVES 8Y CRYSTALS

Bragg Law
We study crystal structure through the diffraction of photons, neutrons, and electrons (Fig. I). The diffraction depends on the crystal structure and on the wavek't1gth. At optical wavelengths such a~ 5000 }.. the superposition of the waves SC'.tttert.-d elastically by the individual atoms of a crystal results in ordinary optiC'd.l rd"raction. When the wavclenbrth of the radiation is comparilhle with or smaller tha'l the lattice constant, we may find dHl'racled be.tms in directions quite diOcrent from the incident dirt..'Ction. \Y. L. Bragg presented a simple explanation of the diffracted beams from a crystal. The Bragg derivation is simple hut is convincing only because it rt.'PrQ. duces the corTt.'C!. rcsult. Suppose that the incident waves are reOected specularly from parallel planes of atoms in the crystal, \\ith each plane rcOcctingonlya very small fraction of the radiation, like a lightly silvered mirror. In specular (mirrorlike) reflection the angle of incidence is equal to tlle angle of reOection. The diJTractcd beams arc found when the reflections from parallel plane5 of atoms interfcre constructively, as in Fig. 2. We treat elastic scattering. in which the energy of the x-ray is not changed on rcllection. Inelastic scattering, with excitation of clastic waves, is discussed in Appendix A. Consider parallel lattice planes spaced d apart. The radiation is incident in the plane of thc paper. Thc path differcnce for rays rcfllXtoo. from adjacent planes is 2d sin 9. where 9 is measured from the plane. Constructive interference of tile radiation from successive planes occurs when the path difference is an integral number fl of wavelengths A. SO that

2d sin 9 ::: riA .

(I)

This is the Bragg law. Bragg refllXtion can occur only for wavelength A :S 2d. This is why we canllot usc visible light. Although the reflection from each plane is specular, for only certain va1ut.'5 of 6 \'.ill the reflections from all parallel plnnes add up in phase to give a strong reflected beam. If each plane were pcrfl..'ctly reflecting, only the first planc of a parallel set would sec the radiation, and any wavelength would be reflected. &t each plane reflects 10-3 to 10-5 of the incident radiation. so that 1(fJ to lOS planes may contribute to the fonnation oftllC Bragg-reflected beam in a perfect crystal. Rcf]IXtion by a single plane of atoms is tTealed in Chapter 19 on surface physics. TIle Bragg law is a consequence of the pcriodicityofthc lattice. Notice that the law docs not refer to the composition of the basis of atoms associated uith Cl--ery lattice point. We shall see. however. that the composition of the basis

"

30

determines the relative intcnsit)' of the various orders of diffraction (denoted by " above) from a given set of parallel planes. Experimental results for Bragg reflection from single crystals are shown in Figs. 3 and 4, for rotation about a fixed ads.
SCATTERED WAVE
A~IPLrrUDE

l1tc Bragg dcrh'alion of the diffi-aetion condition (I) givcs a ncat statement
of the condition for the constructi....e interference of wavC5 scattered from the lattice points. We need a deeper analysis to determine the scattering intensity from the basis of atoms, which means from the spatial distribution of electrons within caciJ cell. From (1.3), a crystal is invariant under any translation of the form T = Ula\ + 1/282 + tiaRa. where u" 112. U3 arc integers and 81> 112. aJ arc the crystal axes. Any local physical property of tile crystal is invariant under T. such as the charge concentration, ek'clron number density, or magnetic moment density. FcJ1lrier A,ltIrysis
What is most important to us here is that the electron number density nCr) is a periodic function of r. with periods al. a2, a3 in tile directions of the three crystal :u:cs. Thus
II(r

+ T) =

nCr) .

(2)

I
I

Sueh periodicity creates an ideal situation for Fourier analysis. nle most interesting properties of erystals are directly related to the Fourier components of the electron density. We consider first a function 11(X) with period a in the direction x. in one dimension. We expand II(X) in a Fourier series of sines and cosines:
n(x) = ''0

.>0

[C p cos(21TpxJa)

+ 5p

sin(21Tpxfa)]

(3)

where the p's are positive integers and Cp Sp are real constants, called the Fourier coefficients ofthe expansion. The factor 2ma in the arguments ensures that n(x) has tile period a:
'I(X + a) = 'lo
= no

+ I:[C" cos(2wpxJa + 217p) + 5 p sin(21lp%!a + 217p) + I:[C" c:os(2wpxJa) + 5" sin(2wpxJa)l = n(x) .

(4)

We say that 27Tp1a is a point in the rcciprocallatticc or Fourier space of the crystal. In one dimension these points lie on a line. The reciprocallatticc points tell us the allowed terms in the Fourier series (4) or (5). A term is allowed if it is consistent with the periodicity of tile crystal, as in Fig. 5; other point~ in the rc..'Ciprocal space are not allowed in the Fourier expRllsion ofa periodic function.

....,.............
UnoJe-iotcd

Figure J Sketch of a mOlloch~or "t..ic:h by Bng reflection.ekds a narrow spectrum rJix~ or nrutron .~'t=kngth, from a broaoJ spectrum incidftlt bnm. The uPI'" part of the r"",re shows the analysis (obtaoned by n:fk.ction &om a ~ crystal) rJithe purity of a 1.16 "bnm rJi rM.'ub'OnI &om caJnum fll-ide (TyStaJ monoc:hrornator. ~ main beam is lhal not reflccted from the scrolld ~-$b.I. (AIIer C. Bacon.)

0'

00"

to'

<0"

&0'

"'"

- " U 100"

110'

'W

,<0"

lIO'

"0'

__

Figure.. X-ray <IiITraclomcter rewrding of po;M-mred 5111001\. showing a COUIlI"r n'COriling of the diR"ractt"d beams. (C<lur\<--sy of W. Parrish.)

rlgUre S "periodic funcCjgn n(r) <i period ... and the terms 21rp1p that
tna)'

spprar III the Fourier transfOrm

"~) .. 1: "r ,,",p(12orpda~ The mag~

nit~ofthe
ROI

lI'Idividuai terms "rare

-------;:~;-_,:;-;:.:_-----G

plotted.

-~-~

"
It is a w~al collveniCIl(.'C to write the series (4) in the l'OmllilCl form
/leX) =

L II,. cxp{i2'1T]lXlal
"

(5)

where the sum is over all inteJ*rs lJ: positive, nCWllivl", and 7ro. Thc C'llCll'icil'llls II" now arc complex numhers. To ('nsure Ilmt n(x) is ;1 real fundion, we require
(6)

for then the sum of the terms in r and -Ii is real. The asterisk 011 11!" denotes the oomplcll: cOlljUWltc of 11_,.. With Ip :: 2wpxJa. the sum oftilc tl.-rms in 11 and -p in (5) can be shown to he real if (6) is satisfied. The sum is
n,,(eos IJ'

+ i sin \ill + n_,,(cos If -

f sin lp)

= (I',. +
which in tum is equal to the real function

"-,.)cos 'Y + i(n"

- n_,Jsin If'

m
(81

2Rt..1,vcoslp -

2Jm{nJsinlp.

if(6) is satisfied. Here Rc{II,,! and Im{",J denote the real and imaginary l)3I1s of ti,l_ Thus the number density /I(x) is a real function, as desired. TIle extension of the Fourier analysis to periodic fUlictions tI(r) in three dimensions is straightfurw.trd. We must find a set of vcd:Ol"li G such that
nCr) =

2: tic exp(iC . r) c

(9)

is invariant under all crystal translations T that leave thc crystal invariant. It will be shovo'l'l below that the set of Fourier coefficients lie determines the x-ray scattering amplitude. Inver-lion of Fourier Serino We no\\" show that the Fourier coefficient "" in the series (5) is given by

n,> = a-I

dx o(x) ex(- i21TpJia)

(101

Substitute (5) in (10) to obtain

lip = a-I; op.f dx exp(i217{p'- ,) )xla]

If,,. 'F p the value of the integral is

(ll)

a
i21'1(p' - p)

(e'2~-pl _ I) = 0

because JJ' - rJ is an integer and exp(i27l'(integcr)J = 1. For the tt.. 'rm J1' = rl the inh.. --grand is exp(iO) = I, and the value of the integral is 6, so that lip = 0-111,,6 = 'l p ' which is an identity, so that (10) is an identity. Similarly, the inversion of (9) givt.'S
lie = V;l

dv nCr) cxp(-iG' r)

(12)

<dO

Here Vc is the volume of a cell of the crystal.

Reciprorol Latlice

V~clors

To proceed further "ith the Fourit. analysis of the electron concentration we must find the vectors G of the Fourit. sum}:nc exp(iG . r) as in (9). There is a powerful. somewhat abstract procedure for doing this. The procedure forms thc theoretical basis for much of solid state physics, where Fourier analysis is the order of the day. We construct the mds vectors bl> b}, 113 of the reciprocal lattice:

--':"O'c.:xc"o' b1 = 217-:: .. ':-_...,;"",...~"~,.,...,;";,

b:z = 211"

", x ",

. " . ' 8:z X 83

n.c factors 2rr are not used by crystallographers but are convenient in solid state physics. If 81, a:z, 8;J arc primitive vectors of the crystallatticc, then b l , ~, b:l are primitive vectors of the rcciprocallatticc. Each vector defined by (13) is orthogonal to two lUis vectors of the crystallauicc. Thus bl> b1 b:l have the property
bi , 8; = 211"611 ' where 6 jj = 1 if; =j and 6y = 0 ifi;,l; j. Points in the reciprocal lattice are mapped by thc set of vcetors (14)

(15)

whcre
vcdor.

VI>

Vi, V3 are integers. A vcclor G of this form is a reciprocal lattice

Every crystal structure has two lattices associated \\ith it, the crystallauicc

and the reciprocallatticc. A difIraction pattern of a crystal is. as we shall show, a map of the reciprocallatlice of the crystal. A microscope image, if it could be
resolved on a fine enough scale. is a map of the
Cf)'Stal

structure in real space.

'The two lattices are related by the definitions (13). Thus when we rotate a
crystal in a holder. we rotate both the direct lattice a.nd the reciprocal lattice. Vectors in the direct lattice have the dimensions of(lengthJ: vectors in the reciprocal lattice have the dimensions of (Illengthl. The reciprocal lattice is " lattice in the Fourier space associated with the crystal. TIle tenn is motivated

below. \Vave\'{.'Ctors arc ahvays drawn in Fourier space, so that every position in Fourier space may have a m(.'aning as a description of a wave, but there is a spt;.'Cial signillcance to the points defined by the set of G's associated with a crystal structure. 1'lle v(.'ctors G in the Fourier serk'S (9) are just the reciprocal lattice vectors (15), for then the Fourier series representation of the ek'Ctl"On density has the desin,-o invariancc under any crystal translation T = u.a. + 112~ + U3a3 as defint;.'Cl by (1.3). Fmm (9), n(r + T) = But exp(iG' T) = I, oc-'Cause exp(iG' T) = exp[i(v1b j

L nc exp(iG' r) exp(iG' T)
C

(16)

+ vzb:z + V3!J:J)'
V2U2

(Ujal +
.

u2aZ

+ U3a3)]

= c'\:p[i21T{vlul +

+ V3U~]

(17)

The argument of the exponential has the fonn 2m timt;.'S an integer, because VIUI + VZUz + tl:lU3 is an integer, being the sum of products of integers. Thus by (9) we have the dt;.'Sired inV3riancc, n(r + T) = n(r). This result proves that the Fourier representation of a function periodic in the cl)'Stallatticc can contain components nc exp(iG . r) only at the reciprocal lattice vectors G as defint..--d by (15).

Diffraction Conditions

Tl,eorem. The set of reciprocal lattice vectors G detennines the possible x"ray reflections. We St;.'C in Fig. 6 that the difference in phase factors is exp[i(k - k')' rl between OC-aillS scattered fmm volume elements r apart. The wavevectors of the incoming and outgoing beams are k and k'. The amplitude of the wave scattert;.'Cl from a volume element is propOl1:ional to the local electron concentration n(r). The total amplitude of the scattered wave in the direction of k' is proportional to the integral over the crystal of n(r) dV timeS the phase factor exp[i(k - k')' rJ. In other words, the amplitude of the electric or magnetic field vectors in the scattert;.'Cl ek'Ctmmagnetic \vave is proportional to the following integral which defint;.'S the quantity F that we call the scattering amplitude.'
F= where k+6k=k'.
(19)

f dv

n(r) exp[i(k - k')' rl =

f dV n(r) exp(-i6k' r)

(18)

Here 6k measures the change in wavevector and is calk'Cl the scattering vector (Fig. 7). We add 6k to k to obtain k', the wavevector of the scattered bt,.'3.m.

RmprOf:<lI Ltd/ice

33

Figure 6 11>e dilTeren,'C in path lell~lh of the ineillent "",,'e k;l! the points 0, r b r sin .." andlhe .:Iifference io "hase anJ.:le is (2.... sin ..,YA, Wllich is eqoal 10 k . r. Por the dilTr:\ClL..! .....,,"c the .:IilTere'K'C in pha.., anj(!e is -k' . r. The total differerK.'C in phare an~e is (k - k') . r. ao.:lthe wa,'e scalier,," from IfV at r has the l,hase fXior "",,,!i(k - k'). r] r...lati,'e to the wave "'-""lie"...! from a "ol"me ...Iement at the o..-i!':i" O.

Figure 7 Definition or the s,,,,lIering vector tlk such that k + tlk '"' k'. In elastic "'lIl1ering Ihe magnitudes satisry k' - k. Further. in Bragg SUllteri"g from a "erio<lic lalli<.'C anyaUQwro tlk must equal some redl'rocallallice \'C<.-tor C.

We introduce into (18) the Fourier components (9) of n(r) to obtain for the scattering amplitude

F=

L: f dV ne exp[i(C C

6k)' r]

(20)

When the scattering vector 6k is equal to a particular reciprocal lattice vector,

(21)

the argument of the exponential vanishes and,.. = Vne. It is a simple exercise (Problem 4) to show that ,.. is negligibly small when 6k differs sib'lliAcantly from any 1"t:.'Ciprocal latlice vector. In elastic scatteting of a photon its energy fUJJ is conserved, so that the frequency w' = ck' of the emergent beam is equal 10 the frequency of the incident beam. Thus the magnitudl..'S k and k' are equal, and J,;'l = k'2, a result that holds also for ek-'CtTOn and neutron beams. From (21) we found 6k = C or

36

k + G = k'. SO tllBt the diffraction condition is written as (k

+ G)2 = k2, or
(22)

This is the central result of the theory of elastic scattering of waves in a periodic lattice. If G is a rt.'Ciprocal lattice vector, so is -G, and with this
substihllioll we call wrile (22) as

(23)

nus particular expression is often used as the condition for diffractioll. Equation (23) is another statement of the Bragg condition (1). The result of Problem 1 is that the spacing d(llkl) between parallel lattice planes that arc nonnallo the dirt.'CIioll G = Ilb l + kbs! + lb:J is d(lIkl) :: 2,""Gl. Thus the result 2L: G "" G 2 may be wriuen as
2(217fA) sin 8 =
27rfd(~kl)

or 2d(hkl) sin

e:: A.

Here 6 is the angle between the incident beam and the

crystal plane. The integers 'lkl that define G are not lU.'Ct..>ssarily identical with the indices of an actual crystal plane. because the hkl may contain a common factor n, whereas in the definition of the indices in Chapter 1 the common factor has oc't.-'Jl eliminated. We thus obtain the Bragg result:

2d sin 6 = nA ,

(24)

where d is the spacing belwt.'Cn adjacent parallel plancs with indices Itln, kin,

Un.
Laue EqllOtionl

Ihe original result (21) of diffraction thoory, namely that .o.k = G, may be in anothcr way to give what are called the Laue (,'quations. These are valuable because of their geometrical representation (see Chapter 19). Take the scalar product of both M; and G successively with a .. a2. 8,]. From (14) and (15) \,"-e get
e~resscd

(25)
These (,'(juatiolls have a simple g(,'Omctrical interprcfation. The first cljuatioll al .o.k = 21TOI tells us that .o.k lies Oil a certain cone about the direction of ai_ The second equation tells us that .o.k lies on a coile about l\i; as well, and the third equation requires that .o.Lr. lies on a cone about 11,;). Thus, at a re(J(,'CuOIl .o.k must satisfy all three equations; it must lie at the common line of intersection of three cones, which is a severc condition that can be salis6t.-d only by systematic swceping or scarching in wavelcngth or crystal orientatioll--<lr else by s))(.'Cr accident.

F"ewe 8 The poiDI~ on I"" ~Ihand.ide ar" raipovcal bllier points nfl1"'Cf}'illll Th" .-ector Ir
is dn...n in I"" diredion oitke in<.idrnl Hay bftom, and tbe origin i. cl-.. such lhal k leoninat"" III any n'Ciprol."al Iallk..., poi"l. W" dr:l... a ~-phrre of radius k .. 211fA aboul the: origi" of Ie. A

dilfr:IL1ro beam will be formed Jfthis sphere inl<'TSrclsany olh.... ,)Qint in the reclproeallatlk-e. 'I"" sphere a> drawn inlcrccpt. a poinl ~'Onnl)('\~ wilh theend ofk by a red(lroc.lIllalli~.., vc<:t<>r G. Inc dilfra<.iro ~-ray bram is in Ihe dircdioo k' - k + C. n.e angl" 8 i. Ihe Ilragg angle of fig. 2. Thi. ronslruction is due 10 P. P. E"<lld.

A beautiful conslrtJction, the Ewald construction, is exhibited in Fig. 8. This helps us visualize the nature of the accident that must OCC'lIr in ordcr to satisfy the diffraction condition in three dimensions. Thc condition in two dimensions (diffrnct.ion from a surface layer) is treated in Chapter 19. Refk'Clion from a single plane of atoms takes place in the directions of the lines of intersection of two cones, for example the com.'S defined by the first two of the Laue t.'quations (25). Now two cones will in general intercept each other provided the wa.C\'cetor of the particles in the incidcnt beam exceeds some threshold .-a.lue determined by the first two Laue equations, No ac<:idental coitri(lence is required, unlike the problem of diffraction in 3D. This matter is of prime importance in the difiTaction of low energy electrons from the surface of a ('I)stal.
BRIUOUIN WNES

Brillouin gave the statement of the diffraction condition that is most widely used in solid state physics, which means in the description of electron energy band theory and of the elementary excitations of other kinds.

38

Flgucc 9a 1'It.'t1Ilroca! IaUit.-e I'Oilll~ "ear lhe point Onl origin of the rl.'tipmeallauk",. 111", l'(''';pro.:allatti<'e \n.-Ior connecti points OC, and ell alnn:ts 00. T...v planes I and 2ll1'f: drawn ...hidl ""' Ihe ~ penclicblar biKdors eX e.-:lnd Co. ~hely. Any ,K1or from the origin to the pbnt I, wdl as k .. wtll salis(y the dilfr;oction <."OO>dilion k,' (ICd" OCrf. An)' \ector from the origin to the plane 2. >Udl as will ntisfy the <!iITr:ac!ion condilioll kt (jel)th~

e.-

sro..""

FigUNl 9h S'lu:,re T'-'l:;pnx:al IaUi: "'ith rec;illrOlI lattice .....><.1or.> ~hO"'n as line bL1Ck lines. 'nle lines in while"", perpendicular bi~1ors of die ..,..

"t.

ripn::lC'll1 L,ulC:'e ..,eton ~ centra! sq,we is the smallN ."Ofume about the Clf"'Igin whidlls bounded elltire!)~ ..."'-lle lines. The ~ is the W-~-Srib. pri.niliH~ cell of the re<."prornJ bt~. II is <:ailed the flJSl BrJlouin WIlt'.

(Ie,,)'.

A Brillouin mile is defined as a Wigner-Seitz primitive cell in the recipro-

cal lattice. (TIle construction in the direct lauice was liho\\n in Fig. 1.6.) The
\'a!ue of the Brillouin zone is that it gives a "h id geomctri<:al interpretation of the diffraction condition 2k. G = Gel of Eq. (23). We divide both sides by 4 to obtain (26) \Ve work in reciprocal space, the space of the k's lind C's. Sc!t:ct a \"(:clor G from the origin to a no'ciprocal lattice point. Construct a plane normal to this vector C at its midpoint. This plane forms a part ofthe zone boundary (Fig. 9a). An x-ray beam in the cryrstal will be diffracted jfits wave~ector l: has the magnitude and direction required by (26). The dif&acted beam will then be in the dir<.'Ction l: - C, as we see from (19) with 6k = -C. Thus the Brillouin construction exhibits all the \vaVt:~'cctors k which can be Bragg.reflectcd by the crystal. The set of planes that arc the perpendicular bist.'Ctors of the reciprocal lattice \ectors is of gencral importance in thc theory of wave propagation in cl')'stals. A wa\'e whose \\'3.~evectordrawn from the origin terminates on any of these planes will satisfy the mndilion lOr diffraction. These plancs dh'ide the Fourier space of the crystal into fragmcnts, as shown in Fig. 91> for a square lalliee. The central square is a primitive cell ofthe rt:ciprocal laUice. It is a Wigncr-Scitz cell of thc ( t'l"Ocal Jaltice.

--~"

Flgu", 10 Construction Qf the Orst Brillouin zone for an oblique lalla in twu di.ncnsiolu. We Rnt draw a number of .'rctQn from 0 to nearby pgints in !he reciprocal btlitt'. Nelli wealNlruct lines ~icubr to these ,.ton; allbnr midP'Jinis. 'The mW.I~t rndosed area is the lint

Brillouin

ZOM.

j".

ft"UpfOOllIor,. '~

.I~

A---

FiSUn: 11 Cl'}'$laI and rKiprocal btlices In ooe di"..,...ion. 'The basiJ; ,"to\'" in the rcci~l bttn .. b. oClcOllh e~ 10 bfD_ The shor1oest rropl"llQ.l bttu ve<.1on from the origin ve b and -b. The ~br bisedof. of these .'ectors bm the boundaries of the Ant Brillouin lOne. The houndaries are at k - :twlc.

The central cell ill the reciprocal lattice is of special importance in the theory of solids, and we call it the first Brillouin zone. "lbc Ilrst Brillouin zone is the smallest volume entirely enc!osl.'<l. by planes that are the perpendicular bisectors of the reciprocal lattice \'(.'Clors drawn (rom the origin. The first Brillouin zone of an oblique lattice in t\\~ dimensions is constructed ill Fig. 10 and of a linear lattice in one dimension in Fig. 11. The zone boundaries oCthe linear lattice are at k = 'If!o. where a is the primitive axis of the crystal lattice. ~

.
Historically. Brillouin zones are 1101 part of the language ofx-ray diffraction analysis of crystal structures, but the zones are an essential part of the analysis of the electronic energy-band structure ofcrystals. The sp<.'Cial utility ofthe first Brillouin zone is de\'c1opcd in Chapter 9.

Reciprocal Lattice to

3C

Lattice

The primith"c translation \'eetors of a simple ruhic lattice may be taken as the sel
8,

= aX ;

302

at ;

83

= at .

(2730)

Here i, y. 1 are orthogonal vectors of unit length. The volume of the cell is 8,' 82 X 83 = tr. The primith"c translation vectors of the reciproca1laUice are found from the standard prescription (13):

b,
21T/a.

(2malY ,

b,

(2mal' .

(21b)

Here the reciprocal lattice is itself a simple cubic laltice, now oflatrice constant The boundaries ofthc 6rst Brillouin zones are the planes normal to the six reciprocal lanke \"ectors :!:b" :!:b:z. :tb-.J at their midpoints:

::t{b, = (1Tla)x ;

{b2 = (1Tla)Y ;

i~

""

(tTfa)i .

The six planes bound a cube of edge 2tr1o and of volume (2tr1o"f; this cube is thc first Brillouin zone of thc sc crystal lattice.

Reciprocal Lottice to bee Lattice

The primitivc translation \ectors of the bee lattice (Fig. 12) are

3\"" ia(-i + 5'+ i);

a2"" lo(i - 5' +i);

a3:!a(i+ 5'-i) , (28)

where a is the side of the COIwentional cube and i, Y, i are orthogonal unit vectors parallel 10 the cubc edges. 1be volume of the primitive cell is

(29)

1bc primitive translations of the reciprocal lattice are dcfUlOO by (13). We have, using (28). b, "" (2tTla)(5' + i) ;

b,

~ (2maX'

+ i)

b,

(2maX'

+ Y)

. (30)

Nole by comparison with Fig. 14 (p. 42) that these are just the primitive vectors of an fcc lattice, so that an fee lattice is the reciprocal lattice of the bee laUice. TIle general reciprocal lattice VL'C!:or is. for integra] VI> Vo./:. tlJ. G = vlb,

~b2

V:J~ "'" (2tr1aX(t:t

+ V:Ji + (VI + tl3l5' + (VI + t:JiJ.

(31)

The shortcst G's are the following 12 vectors. where all choices of sign are independent:

(2r.la)(Y i) ;

(2m'a)(H:t i) ;

(2nla)(i :t y) .

(32)

"'lUre 13 First Brillooin zone c:ithe body......... terecl Figure 12 Plimilh-e basis \'eCIQn orthr body-centcred cubic latlice.

cubic buice. The Ggure b regular rhombic

"""""""'~. The primith"c cell of the reciprocal lattice is the parallelepiped de:.>SCribed by the h" ~. l':) defined by (30). The volume of this cell in reciprocal space is b l ~. b;, = 2/..2,"10'1. The eell contains one reciprocal lattice point. because each of the eight comer points is shared among eight parallelepipeds. Each

p:ar.dlelcpipcd contains one-eighth of each of eight comer points.

In solid state physics we take the central (Wigne.--Seit-J.:) cell oCthe reciprocallaUice as the first Brillouin zone. Each such cell contains one lattice point at the central point of the cell. This zone (for the bee lattice) is bounded by the

planes normal to the 12 vectors of Eq. (32) at their midpoints. -Inc ZOlle is a regular l2-faced solid, a rhombic dodt.'C8.hedron, as shown in Fig. 13. "Ole
\uetors from the OI'igin 10 the center of each face arc

(1Tla)(:!:y i) ;
All choices of sign arc independent, giving 12 v(:cton.

(33)

ReciproctlllAtlice to fcc lAttice

The primitive translation vectors of the fee lattice of Fig. 14 are
3.::::

!a(S' +

i);

lit

=! tl(i + t);

33::::

!a(i + y)

(34)

1bc volume of the primitive cell is

V = 1301 . It" x

a3I :::: ~a3

ht =
(2m'aXi i) .

(35)

1lIc primiti\'e translation vectors of the lattice reciprocal to the fcc lattice
b l = (2'Jl"/aK-i + 5' + i) ;

y + i)

b:J :::: (2'Jl"/tlXi + Y -

(36)

./

r 1
o

Figure 14 Primitive basis ,"eclors of lhe f;M:e<centered "ubic l~ttio.:e.

These are primitive translation vectors of a bee lattice, so Ihal the bee lattice is reciprocal to the fcc lattice. one volume of the primitive cdl of the recipl'OC"<I1 lattice is 4(211"10)3.

The shortest C's are the eight veclors;

(21Tlll)(X 5' i)

(37)

"Inc bounclark-s of the central cell in the reciprocal lattice are detennined for
the most part by the eight planes normal to these vl,.'Clors at their midpoints.

But the corners oCthe octahedron thus funned are cut by the planes that are the
pCI])elldicular bist;.'Ctors of six other n..'ciprocal lattice vectors:

(27T!a)(2x) ;

(21Tla)(:t2y) ;

(21Tla)(22) .

(38)

Note thai (21Tla)(2x) is a R'Ciprocallattice vt,.'clor b(.'cause it is equal to ~ + b:J. TIle first Brillouin zone is the smallest bounded volume about the origin, the truncated octaht;.xlron shown in Fig. 15. The six planes bound a cube of edge 47T/a and (before truncation) of volume (47Tla)3.
'OURIER ANALYSIS 0.' THE BASIS

When the diffraction condition .6.1.: = C of Eq. (21) is satisfied, the scattering amplitude is determined by (IS), which for a crystal of N cells may be written as dVI1(r) exp(-iCr) =NSc (:39) ~" l11e quantity Sc is calk..J the structure factor and is defined as an integral over a Single cell, with r = 0 at one cOlller.
Fe =

NI

Figure 15 Brillouin WIlC1 0( the race-ccntcred cubic I~Ui<:e. cells are in IlxipTQCllI spnce. and Ihe rOOllllJ{.'ll1 lattice Is body <:entered.

nlC

onen it is useful to write the electron concentration n(r) as the superplsi. tion of electron concentration functions nJ associated with each atom j of the ceiL IfrJis the \ector to the centerofatomj, then the function nAr - rJ) defines the contribution of that atom to the electron concentration at To The total electron concentration at r due to all atoms in the cell is the sum n(r) =

L nAr J-'

rJ)

(40)

O\'er the s atoms of the basis. "Ole decomposition of n(T) is not unique, for we

cannot always say how much charge is associah."'<1. \\ith eadl atom. l'his is not an important difficulty. The structure factor defined by (39) may now be written as integrals over the s atoms of a cell:

Sc "'"

L f dv nAr J

TJ)

exp(-iG' r) =

L exp(-iG rJ) f
J

(41) dV nj,p) exp(-iG p)

wherep-r-rJ. We now define the atomic fonn factor as

(42)
intt.'STatcd over all space. If IlJ(P) is an atomic property, fj is an atomic property.

We combine (41) and (42) to ohtain the structure factor of the basis in the form Sc=L!JexP(-iG'r-j) .
J

(43)

The usual form of this result fullows on writing for atom j: r-j =
xjal

+ YjRs! + :r:,a3

(44)
&:I we have

as in (104). 1ben. for the reflection labelled b)'

C' I"j = {vib i

&1> ~

= 27J(VIXj

+ &2b2 + V:lb.:J' (Xjal + YJa2 + zja.:J + Vy,Yj + vJZJ)

(45)

so that (43) becomes

Sc(VIV2VJ} =

L /j eXPr-i27J(vlxj + V2!!j + 1Ja'Zj)J

J

(46)

1be stnJCture factor S need not be real because the scattered intensity will involve SS. where S is the complex conjugate of S so that SS is real. At a zero of Sc the scattered intensity will be zero, even though G is a perfectly good reciprocal lattice vector. WIlat happens if we choose the cell in another \'.'3y. as a conventional cell instead of a primitive cell, for example? 11le basis is changed, but in soch a y.'3y that the physical scattering is unchanged.. Thus for two choices. 1 and 2, it is not hard to satisfy yourself from (39) that

Stt"'Udure Fodor of the bee Lattice

Zl

The bee basis referred to tbe cubic cell has identical atoms at Xl = YI = = 0 and at X2 = !l2 "" 'Z2 = i. Thus (46) becomes S(VltliVJ) = f{l

+ exp(-i'l'l'{vi + ~ + v.:il}}

(47)

where f is the form factor of an atom. The value of S is zero whenever the exponential has the value -1, which is whenever the argument is -j1T X (odd integer). 'l1lUs we have

5=0

when

VI
VI

S = 2f when

+ tlt + OJ = odd integer ; + tli + lJ:I = even integer .

Metallic sodium 11as a bee structure. TIle diffraction pattern does not contain lines such as (100), (300), (Ill). or (221). but line.~ such as (200), (I 10), and (222) will be present; here the indices (VIVy,v.:J are referred to a cubic cell. What is the pll)'sical interpretation of the result that the (100) reflection vanishes? The (100) reflection normally occurs when renections from the planes tbat bound the cubic cell differ in phase by 21T. In the bee lattice there is an intervening plane (Fig. 16) of atoms. labeled the second plane ill the figure, which is

.
figure 16 Explanation oflhe absence or II (tOO) rcl1ection From II body~tmtcrt.>d cubic lattice. 'nle Ilhase difference between ~uc<:e"ive pltnes is 'TT, SClIru.1 the ~nected aIJ1!l!itu6e From two adjacent

IIiMes isl

+.-..... 1-1-0.

equal in scattering power to the other planes. Situated midway between them. it gives a reflection retarded in phase by r. with respect to the Ilrst plane. thereby canceling the contribution from that plane. "nIe cancellation of the (100) reflection occurs in tile bee lattice because the planes are identical in composition. A similar cancellation can easily be found in the bcp structure.

Structure Factor of dte fcc Lalli

The basis of the fcc structure referred to the cubic cell has identical atoms at 000; OM; 401; #0. Thus (46) bec::omes
S(VItl2Vi! :: Itt

+ exp[ -hT(~ + tl3l1 + expI -ir.(v) + ti;)))

+ exp[ -i'l7(vl + ~n

(48)

If all indices are even integc.-s, 5 z: 4[; similarly if all indices are odd integers. But if only one of the integers is even, two of the exponenl~ will he odd multiples of -jr. and S will vanish. If only one of the integers is odd, the same argument applies and S will also vanish. Thus in the fcc lattice no reflections can occur for which the indices are partly even and partly odd. The point is beautifully illustJ1l,too b). Fig. 17: botJl KCI and KBr ha,e an fcc lattice, but KCI simulates an sc lattice because the K+ and CI- ions have equal numbers of electrons.

Atomic Form Factor

In the expression (46) for the structure factor, there occurs the quantity Jj, which is a measure of the scattering power of thejth atom in the unit cell. The value off invokes the number and distribution of atomic electrons, and the wlwelength and angle of scattering of the radiation. We now give a classical calculation of the scattering factor.

.
~---

f"'
.u

...
. i
"r'

I
,
,

i
~

I
I
t"igune 11 Co,ollenson of ._n)' refledioN from KO and KBe l'O",jen. In Kd the nun.t.>en of eltmns d K' and ions.are equal. ll1e satlel'inC &",11;lucks JrK') and fta-) are all'l'lOlt ,,>:at'tly oetlual. so lbat the (l)~t.oI IooL:s 10 "-n,,, as if it werll: monat()m~ simple cubic lattice of lattice constant a!2. Only c"'~n Integers occur in the renn;:linn indices when theSt! .I,'.. ""SOld on a cubic lattice orlaltict! C(lTl-

..
"'"

'"'

'"

..I

. . 't ~h")
('l.J1:-r

'"

scattered radiation from a single alom takes account of interference effects within the alorn. We defined the form factor in (42):

nle

.. '" -"

'''J "
h= f

stanl Q. In KDr the form faLior of Br- is quite differellt than thai of K' . and.ll rdlectioll'l of the ICc lallie..., are ~t. (Cooties)' of R. "an Nordstrand.)

dV II)r) e):p( -jC . r)

(49)

with the integral extended over the electron concentration associated with a single atom. Let r make an angle a with C; then G' T = C,. cos a. If the electron distribution is spherically symmetric about the origin, then

fj .. 21T f d,. ,-2 d(ros a) n,{,.) eXp(-iC" cos a) rr _ e-iCr = 21T f d,. r"n,,{,.) iC,.
after integralion over d(cos a) belween -I and 1. TIlUS the form factor is given by

(50)
If the same tot-al electron density were concentrated at ,. = 0, only C,. = 0 ""auld contribute to the integrand. In this limit (sin C,.YC,. = I, and

Jj

= 4.1T

d,.

nj..r)~

Z<

(51)

" " " ,

f
-U""..-f"". GlkuLor"",

A "0.709,1.

(11I)0

,
,-

"'"
"'" "'" ,....
(331)00

{4~(4Jl

{5l1j

.,

"

"

>inllfA

"

,.,

"

"

"

Figure 18 Abduce eo:periJ1lftltal atomic xattcrin& factnn br metallic aluminum, afteT Bal. lennan. Chipman, and DeMllK'O. Eac.-h obsen-ed rtfWtion Q; labeled. No rdkctioN ocau- bo indb:s partly ~'en and partly odd, as pred>cted b an i:c crystal.

the number of atomic eketrom. Therel"ore f is the ratio of the nll.Uahon amplitude scattered by the actual electron distribution in an atom to that scattered by one electron locali~ at a point. In the forw'",ro direction C = 0, and! reduces again to the value Z. Values of the atomic fonn factor f for atoms may be found in the International tables for x-ray crystallography, Vol. 3. The overall electron distrihution in a solid as seen in x-ray diffraction is fairly close to that of the appropriate free atoms. This statement does not mean that the outermost or valence electrons arc not redistributed in fonning the solid; it means only that the x-ray reflection intensities are represented well by the free atom values of the fonn factors and are not very sensitive to small redistributions of the electrons. As an example. Batterman and co-workers 6nd agreement within 1 percent in a comparison of the x-ray intensities of Bragg rel1ections of metallic iron, copper, and aluminum with the theoretical free atom values from wavefunction ca1culations. "The results for aluminum are shown in Fig. 18. There have been many attempts to obtain direct x-ray evidence about the actual electron distribution in a covalent chemical bond, particularly in crystals having the diamond structure. The question now lies within the limits of what can be explored hy~ ly diffraction methods. In silicon at a point midway

between two nearest-neighlXlr atoms, there is an appreciable increase in elec tron concentration over what is e):pected from the overlap of the electron densities calculated for two free atoms. Scattering from crystal surfaces is treated in Chapter 19. It is shown in Appendix A that thermal motion does not broaden a diffraction line, hut only reduces the intensity. The lost intensity reappears as long, low wings about the position of the diffraction line.
QUASICRYSTALS

In 1984 quasicrystals were first observed;' these are structures which cannot be indexed to any Bravais lattice and "which have symmetries intermediate hetween a crystal and a liquid."They were first observed in grains of size 2 ILm in an alloy of AI with 14 at pet Mn. The smaller Mn atoms are each surrounded by 12 AI atoms arranged at the corners ofan icosahedron. The structure is made up of parallel icosahedra attached at their edges. Crystals cannot e):hibit the fivefold symmetry of an icosahedron, but a crystal can be constructed by nucleation at aeenter cell, followed by outward growth from there. All of the space of a nodule Ql.nnot be filled by repeating the basic unit (see Figures 19 and L 7 for the picture in nllo uimensions), although the "parallel" part of the specification does give a long-range orientational order to the structure. It is perhaps surprising that the ):-ray diffraction pattern of such a structure Ql.n have fivefold symmetry; that is how they were first observed. The known quasicrystals are intermetallic alloys and are very poor electrical COllductors; they are nearly insulators with a somewhat well-defined band ~p (Chapter 7) at the Fermi level. They are of great interest intellectually in expanding the uefinition of crystal lattice. A uistinctly different crystal difli-action pattern results from an almost periodic structure, one that is neither rigorously periodic nor simply amorphous (as for a glass, Chapter 17). An almost periodic structure can he e):pressed in one dimension if we are given the electron charge density wave:
(i.,X) "" LIC" cos[21fn(1

+ r)xlul

(52)

where r is an irrational fraction. TIle terms in 21fnlu by themselves give the usual lattice with translational peTiodicity u. When the terms in T are added, the charge density is almost periodic; that is, the period (1 + r)a is not an integral multiple of the period a, because T is irrational. The period gives a long-range nonrandom oruer to the structure, anu the long-unge order gives a diffraction pattern, which appears split off from the pattern defined by the short-range oruer. 1bis is uominated by the reciprocal lattice points in nlo but
'J). l.c'llllea"d I' J- Steinhardt. I'hys. Rc,.l...,u 53, 24TI(I98-1), Ph)"" Rev. R.'4, 596(1986), D. S. Schechtman and olh"n, 11l)'s. RL'V. L.,tt. 53. 1951 (1964).

Fig\>,.., 19 1\ quasicT}"stal tiling in ","'0 dimCIIsions, after the "'llri:: Qr I'enrose. The kmg-range alCntationlil order and the long-~ ~ orde-r ale shown.

will appear to be clustered and spread out (broadencd). "nle diffraction pattern of a three-dimcnsional quasicrystal is quitc different. however; the pattern is ....'ell defined and can have the fivefold symmetry by which quasicrystals were flnt diSCQvercd. A computcr-generated diffraction pattern with fivefold symmetry is sh~wn in Figure 20.
SUMMARY
Various statements of the Bragg condition:

2dsinO=nA Laue conditions:

.6.1e "" G.;

1be primitive translation vectors of the reciprocal lattice are

Bere a),

82, 83

are the primitive translation vectors of the crystallaUice.

G = vtb\ + V2~ + V:lbJ

A reciprocal lattice vector has the timn

where

VI>

Vi. VJ are integers or zero.

"O-Je scattered amplitude in the direction Ie' "" k

+ .6.k = k + G is propor-

tional to the geometrical structure factor:

so
Sc "i..

t:~p(-irJ' G) = "i:.Jj cxpf -i21T(xJv. + YjV2 + %:iVa}]

where) Tuns over the saloms ortbe basis, alldJj is the atomic form factor (49) of the jill atom of the basis. 1be ~"pression on the right+hand side is "'Titten for a reflection (VIV~. for which G "" v1b. + 1'21>,; + v:>b:J.
Any function inv:.lriant under a lattice translation T may be el'punded in a Fourier series of the fonn

nCr) =

L ttc exp(iC' r)
c

TIle first Brillouin zone is the Wigner-Seit-.t primitive cell of the reciprocal lattice. Only waves whose wan~...ector k drav.ll from the origin tenninates on a surface of the Brillouin zone can be diffracted by the crystal

e'ysiallattice
Simple cubic

"iTst BriUouln wne

Cube

Body-centered cubic
Face-centered cubic

IDKlmbic dodecahedron (Fig. 13)

Truncated octahedron (Fig. 15)

,
~'-

'"
';0'

;.

. ,

.'

. "V
'

"
'

,',
,

'.

, .
~

.'
Figure 20 j'hotograph of the calculatL-a I'ouri"r t..ansrann (diffraction pattern) or an icosahedral ,......icrystal alallg olle or the fi....,rold IUC!I, iUuslraling ll.-fold .ymmclT)". TIle transform is cakul~lcd from. lheoretical roml)ljle,..~nentcd model, b)( !woel JaLVb"

tI,,,

Problems
I. Interplanar sepamtion. Consider a plane hkl in a crystal lattice. (a) Prove that the

reciprocal lattice vector G = hb, + k~ + 1b:l is perpendicular to this plane. (b) Prove that the distance between two adjacent parallel planes of the lattice i$ d(hkl) .. 21d1C1 . (c) Show fOr a simple cobic lattice that d t = al/(h l + Ie'- + P) .
2. H ~ &pO<JI! lat,ia. The primitive translation vectors lattice rna)' be taken as
8\ - (311laf2)i

u the

hexagunaI space
, , - ci .

+ (af2)Y

lit .. -(31/t1ll2)i

+ (aI2)y ;

(a) Show that the mlume of the primitive cell is (31J212)o'c. (b) Show that the primiti\"e translations of the redprocal Iallke are

b:J .. (2mc)1
so that the lattire is its O\\fI1 redprocal, but with a rotalioo cl axes. (el Describe lind sketch the fin! Erollouin zone of the hexagonal space lattil.'e.

3. Volume of BriUouin %QfIe. Show that the volume of the first Brillouin zone is (21r'flVc where Vc is the volume of a crystal primitive cell. Hint: 11le volume of a Brillouin :r.ooe is ec:lual to the \"Olume of the primitive parallelepiped in fourier space. Recall the vector identity (c: " a) " (a " b) = (c' a " b)a . 4. Width of diffrae'im1 marimum. We suppose that in a IinearCf)'StaI there are identical point scattering centers at e"ery lattice point p. = ma, where m is an integer. By analogy "ith (20) the total scattered radiation amplitude "ill be proportioIJal. to F ... I exp{ -ima' 11k]. The sum O\ler AI laUke points is
F = I - C1p( -U.l(a . llk)) I - eq>{ l(a 6k)) ,

by tile

U5C

of the series
M-l

l_rM

r"'.--- .
l-r

...-0

(a) l1le scattered intensity is proportional to

iff. Show that

IFF. PF _

sin !.Al(.' 6k) sinl i{a' 61c)

(b) We know that a diffraction maximum ap~ when a . Ill: 2:rrlz, \\here h is an

integer. We change 11k slightly and define It" in a' 6Ic "" 2:rrlz + It" such that It" gives tlJe position cLthe Hot zero in sin iM(a' bok). 500.., that E . 21rfM. so that the width cl the diffraction muimum is proportional to UM and call be ertremely narrow for macroscopic values of M. The same result holds true lOr a three-dimensional crystal.

5. S'ru<:turefador of diamond. TIle crystlll structure of diamonu i$ described in Chapter 1. The basi~ ,sisls of eight atoms if the cell is taken aJ the l'Onvenlional cube.

..

.
(110

"" .,

..

- 1--....

..,...,.
.J

.."

,-

PRl
-c

\--.,..

0.80"......,;

Counter (1011_ 2f

"leu"' 21

Neutron diffraction pattero b

powdered dWnond. (After G. Baoon.)

(a) Find tlJe structure factor S r:I this wis. (b) Find the zeros of S and show that the allowed reflections of the diamond structure 5lltisfy VI + tJt + 0" - 4n, where all indices are e\'eO Bnd "u any integer, or else all indices are odd (Fig. 21). (Notice that h. k, I may be written for VI. Vi. V:l and this is often dooe.)

6. Form fodor of atomic IluJrogen. For the hydrogen atom in ils ground stale. the number density is nCr) - (~-I exp(-2rlac), where llO is the BOOrradiu$. Show that the form factor is Ie; = 161(4 + c1d;j.

7. Di4tomic line. Consider it linC ofatomsABAB ... AB, with an A-B bond length of lao TIle form factors are fA. IfJ for atoms A, B, respectively. 11lC incident beam or x-rays is perpendicular to the line of atoms. (a) Shol,.... that the interference condition is nA .. a cos 6, where 6 is the angle between the diffracted beam and the line of
atoms. (b) Show that the intensity <llhe diffracted beam is proportional to If ... ror n odd, and to VA + fill' lOt- n t:\-en. (c) EJ<plain what happens irfA - fll

- !IJF

References
c. S. .Barrett and T
X-IIAY DJI>F1IACTION 8- MassalsId, s.roctul"f'o[ mdaU: ~taUogra""i(:mttl.oo., ,,1'incIplu, data. 3rc!. ...I. rev., OxfOrd 1960. ~I guide m the Jl"'I'IicaI solution of relati\'e1)' simple s'nK't"rd. M. J. fI.ue..ger, COfltcm,/(/~a'll cryBlaI/ogrTIph",. McCraw-HilI, 1970. A line introd,K(ion. 8. E. Wwren, X~ra!f diffrodlon. Addi.m-We*y. 1969. C. Janot, Qua.rin'ynal.r: a ,,1"irIv~. Oxford. 1993. D. I'. l)i\'incen7.0 and I'. J. Steinhardt, eels., QU/lricry.l1als, World Scientific, 1991. ft CUmt and T. JllIUCn, MExcit.tions in incommensurate C1j'stal phases, MSoI..l state pft)'Sics 41, 201 (1986).

NF,liTIION DIFFItACTION G. E. Bacon. ....O':utron diifruetion. 31 cd., OAford. 1915. s. W. 1.Qo.~, T/wmy of neutron 6aI11trin8 from aJrMk.-d matter. 2 \-nIs., Oxford, 1965. W. MaDha.lI .nd S. W. Lcwese)'. ThroTTl of "1flnn41I1f:Utrun _tlf:ring, Mood, 1971.

3
Crystal Binding and Elastic Constants
CRYSTAUi 0.' mERT CASES IO"lC CRYSTALS
COVALE,,~r

00 72

CRYSTALS

MfALS

I-IYDROGEl'i BOl\"DS

76
77

ATOMIC RADU
ANALYSIS OF ELASnC STRAIfliS

50

Eus'nc COPtIPLlANCE AND STIFl'NESS CONSTANTS ELASTIC WAVES IN CUBIC CRYSTALS

SUMMARY

83

87
00

PROBLEMS
I. 2. 3. 4. 5. 6. 7. B. 9. 10. II. 12.. 13. 14. Quantum solid Cohcsh'c energy of bee and fcc neon Solid molecular hydrogen Possibility of ionic crystals n+R-, Linear ionic crystal " Cubic 'ZnS struclllre Divalent ionic CI")'$tal5 Young's modulus and Poisson's ratio Longitudinal wave velocity Transverse .....ave velocity Effective shear constant
Determinantal approach

.2 .2 .2
93 93 93 93

.3
.3

Cenernl propagation direction Stability a-iteria

Il< Il< lIS

REFERENCES

lIS

F~re I Th.e princi.-!Iyprs of ery>.rallin<l binding. In (a) neutrsJ Moms ",itl. closed electron shells are bound together ",~.k1y by the 'an de,- Wa;lls foroes aMCldatecl with nue:tuations m the charge distribution!i. In (b) ekdrons are transfermd from the alkali a10lllS to tho: halogcn atoms. and the resulting ions art' held together by altr.K1i'"<l cleCh'O$latic iJn.-es bei"'erfl tho.- po5'lh'e and ncgati,.., ~ In (el the \Unce elecln:>IU are takl:tl 3WlIy 1f000n ~ aIbIi Ill...n to b"m a<'Ornloonel eloctron - . ~ whicb the POSlu".., ions are diopcrsc:d. In (d) the I"IWtral al<Jl llpe&T to be bound by the O\-erbpping pu-ts cL I.hdr clectrotl tlistnootions. \

toge,,,,,,,,

"

CHAPTER

3:

CRYSTAL BINDll"G

In this chapter we are concerned with the/Juestion: What holds a crystal together? TIle attractive electrostatic interoctiph bet\.\een the negati....e charges ofthe electrons and the positive charges ohhe nuclei is entirely responsible for the cohesion of solids. Magnetic forces have only a weak effect on cohesion, and gravitational forces are negligible. Specialized terms categorize distinctive situations: exchange energy, van der wa.1ls forces, and covalent bonds. TIle observed dificrences !Jetv,een the forms of condensed matter are caused in the 611<1.1 analysis by differences in the distribution of tile outermost electrOIlS and the ion cores (Fig. 1). l11e cohesive energy of a crystal is defined a~ the energy that must be added to the crystal to separate its components into neutral free atoms at rest, aI infinite separation, with the same electronic conf'ib'lll11tion. The term lattice energy is used in the discussion of ionic crystals and is defined as the energy that must be added to the crystal to separate its component ions into free ions at rest at infinite separation. Values of the cohesive energy or the crystalline elements are given in Table 1. Notice the wide variation in cohesive energy hetv.een different col umns of the periodic table. The inert gas crystals are weakly bound. with coJle.. sive energies less than a few percent ofthe cohesive energies of the elemcnu in the C, Si, Ce ... column. The alkali metal crystals have intermediate values of the cohesive energy. TIle transition element metals (in the middle columns) are quite strongly bound. TIle melting temperatures (Table 2) and bulk modulii (Table 3) vary roughly as the cohesive energies.
CRYSTAlS OF INERT CASES

The inert gases form the simplest crystals. The electron distribution is very close to that of tile free atoms. TIleir properties at absolute zero are summari7"oo in Table 4. TIle crystals are transparent insulators. weakly bound. with low melting temperatures. TIle atoms have very higll ionization energies (see Table 5). TIle outermost electron shells ofthe atoms are completely filled. and the distribution of electron charge in the free atom is spherically symmetric. [n the crystal the inert gas atoms pack together as closely as possible l : the crystal structures (Fig. 2) are all cubic close-packed (fcc), except He:! and He".
'Ze~poinl mohon ol"lhe atoms {kinetic ~ III abl;oIule zero} is II quanlum dfec:t lhaI pbys a dorn"..nt role in lie" and He4 TI.ey do no! soIid4y aI :u:ro ~"ure e\''''' aI ~ute :It:rn. a\"enge f1~ at 0 K of a He :w:.m from its ~ll;briuln position is of the ....&er ol30 10 40

n.e

pet'c:enl althr r>ealalnngbbor dislance. beao.~r the atom, the less importanl are the zero. c..cts. If we omit l'lCro-p:>int molion, we eaIt'uIlIte a mobr .-oIumr ol9 em" mol-I lOr did helium, lIS C'Ofn~ "'lIh theobsen>::d \...J.-01"27.5 and 36.8 an" 1001-' b liquid lIe4 and liquid He" .-..5pecti\-e1y In the- ........ nd state ofbclium v.-e must w.e ~t oI"thc :zrro.point ,notiool of
point

n.e

the

alOffil.

figure 2 Cubic close-pad,e<! (reel IT)'$I''' Slrudlln: of the inert g:uo;s Ne, Ar, Kr, and Xe_ The bltice Ilar.lIneteu of the cuhiccell$lIre 4.46, 5.31, 5.64. and 6.13 J., respectively. 4 K.

3'

What holds an inert gas crystal together? The electron distribution in the be significantly distorted from the electron distribution around the free atoms. because the cohesive energy of an atom in the crystal is only 1 percent or less of the ioni7.ation energy of an atomic electron. 11IUs not much energy is available to distort the free atom charge distributions. Part of this distortion gives tile van der waals interaction.
cry~tal cannot

Va'l der Waals-London Interaction Consider t\'o'O identical inert gas atoms at a separation R large in compari. son wilh the radii of the atoms. What interactions exist between the two neutral atoms? If the charge distributions on the atoms were rigid, the interaction betv.'Ce!1 atoms "'ould be 7.ero, because the electrostatic potential of a spherical distribution of electronic dJarge is canceled outside a neutral atom by the electrostatic potential of tile charge on the nucleus. Then the inert ga.~ atoms could show no cohesion and could not condense. Bul the atoms induce dipole moments in each other. and the induced moments cause an attractive interaction betv."eeIl the atoms.

Table I
LI

Cohesive energies

158. 1.63
37.7

Bo 320.
3.32 765

Energy required to form separated neutral atoms in their ground eledrOnle stllte from the .solid at 0 K at 1 atm. The data were supplied by Prof. Leo Orewer in units kOO per mole, revised to May4, 1911, ~er LOL Report 3720 Rev.

B 561 'lJ11 134 AI

C
1~;Z

N
474.

,
81.0

No
1.92

711.

251

170.

AE2-". .,6/) 118'4': 0,026

113.4 6003 19.37 0.46

145. 1.113 1.51 2567 34,7

Na 107

Mg

,
Sc

SI

kJ/mol - eVliltOm' - kcaUmol

327. 446. 331. 275. -.." ~.3E ~ 343': ,:?~5~ , 78.1 106,7 79.16 65,75
NI
~28.

}AO
32,2

C1 135

.
K'

ooa6
185
112

774

K 901
0.934 2154

Ca

TI

C,

M"

178, 376 468. 512 395. 282. 1.84' 3.91r 4.85 5';'t- 4.1~ 2:92 425 899 111 8 122.4 945 67.4

" 4.2&
413.

Co
424.

C, 3.4980.4 336.

4.39 Rh

987 R,

4.44 1013 1024

'" '1:35'"
130
31.04

-2.81

Ga 2"

648 243

3.8~:::

G' 372

A. 682

S,
56.7

B'
118.

88.8

2:96 2.46 T.2' OJ;~

2818 268

285.3 237

Rb 822
0.852

S, 166
1.72

"64 39.7 Ba C.
77.6 0.804 1854 183. 1.90 43,7

422 603 730 65B 66' 650 554. 376 4.37 6.25 7.57' 6.82"'" 6.85_ G,74 5.75" 3:89 155.4 1325 898 1008 1442 1745 157.2 158.
La

"

Nb

Mo

To

Pd

Ag 284
2.9.5" 680

Cd t12

1,l& 32'" ~.14 2~~. 2:-,.9.... "'.16 26.73 58.1 724 63.4 50.34 2562 3.80

'"

S" 303

Sb 265.

T,

'"
Po

1 , 07

cD

X. '59

Ht

Ta

R'

O.

I,

Pt

A'

Hg

T1
182.

Pb
196,

BI

At

782, 859. 715. 431. 621 788 670. 564, 368 65 4.47 6,44 8,10__ 8.90: ~03 8.17 6.94__ 5.84. 3.81'" 0.67 103.1 148.4 186.9 205.2 185.2 188.4 160.1 1347 87.96 15,5

210, .J.,88~ ,'.Q:f 2t8 43.4 46.78 50.2

144. 1,50'" 3-4' 5

"

R. 160
1.66

Ao 4'0
4.25

i"

L, 428 L'

0.202

R" "5

466

C.
4t7
4.32.~

P,
357. 3.70 85.3

Nd

Pm
n

382

98

_3AO,
785
U 536.
128,

328.

-A,14

Sm '06

E,
1.86

'79

Gd 400
4t4 95.5

Tb 39'

0,
294,

"

997

493 Np 456
109.

428 Am 63

4.05 93.4

3.q4"
70.2

.3.1~

Ho 30'

E,

317. ~29~ :-2.42+ ),&0 72.3 75.8 55.8 371

Tm 2"

Yb 154

4.~3

1
~.

"

1022

6.201429

Th 598

Pa
~

P, 830

S:S!L .(7j

347. 264. 3,60 '2.73

Bk Cm 385 3.g9 1:""

92,1

Ct

E.
0_
~

'm
""'~

Md

No

B' 453.7 1562

LI

Table 2

Melting points, in K.

S,

(After

n.

H. La.moreaux, LBL Report 4995)

C, 1356
Z,
692.7

B
2365

63.15 54.36 53.48 24.56

N'

N'
371.0

M9 922

A'

51

CI

933.5 1687

w 317 388.4 172.2 83.81 , 863

..
K'

K
336.3

C,

1113 1814

TI V 1940 2202

C, 2133

M'
1520

F,
1811

Co NI 1770 1728

G,

G'

302.9 1211

S, A' 1089 494 T,

B'
265.9

115.8

Rb

y Z, S, Nb 312.6 1042 1801 2128 2750

Mo
2895

T,
2477

R'
2527

Rh
2236

Pd
1827

A9
1235

Cd
594.3
" 429.8

S,
505.1

Sb

,
At

XO

903.9 722.7 386.7 161.4

C,
301.6

B, L, HI T, 1002 1194 2504 3293

W 3695

R' 3459

0, PI 2045 33D6 2720 "

A, TI H9 1338 234.3 577

Pb

BI

Po

R'

600.7 544.6 527

F,

R' 973

A'
1324

"-

C,
1072

P,
1205

Nd
1290

Pm

E, 5m 1340 1091
p, 913

Oy Gd Tb 1587 1632 1684

Ho
1745

E,
1797

L, Tm Yb 1820 1098 1938

Md

"

Th
2031

P, Np U 1848 1406 910

Bk' Am Cm 1449 1613 1562

CI

E,

Fm

No

Lw

Table 3 .lsothennal bull: moduUI lind comprctsibilities at room temperature 0.002 After K. Cschncidner, Jr., Solid state physics 16, 275-426 (1964); several g.M::." data are from F'. Birch. in H(Jndbook of phytlcal cOnsl(Jnts, Ceolo,l1;lcnl Sod ety of America Memoir 97, 107-173 (1966). Original referenccs should be consulted when values are needed for research purposes. Valucs in paren U Be theses are estimates. Letters in parentheses refer to the crystal form. Let 0.116 1.003 ters in br.lCkets refer to the temperature: -&!-

H",

He ttl

IL ~T
N 1.1 0 0.012 ,0,562' :q:-~ 80-= ..1.78

C idl 4.43

z,;

- H.
.o.O@.

[a] - 77 K; {b] .. 273 K; {e] - 1 K; (d]- 4 K; Ie] - 81 K. Bulk modulus in units 10'f dynlcm' or 10" Nfm' Compressibility in unitS 10" cm1/dyn or 10 II m'1N

Mg
0.3~

-l.1..Z. .k!!;t.

F
~~,

He ('1 0.010

~-

JPO"":
Ar III
O.oy~

AJ SI P (Ill S" CI q.72.2 0.986 0,,304 0.17_8_ 1.385 1..012 3.29_ 5.6.2..- ..;..:..

7"'~

,,7". ~ nO; Aill l8l1l ~~ ']As":

Rb
0.031 0.116 32;

C, K Co S< TI V M. F. HI Co 0.032 0.152 0.435 1.051 1.619 1.901 0.596 1.683 1.914 1.86
O~59:4 'tl~2?

C,
1.37

H;l!

A73

l. Ga 1111 G. As So 0.598 0.569 0.772 0.394 0.091 1.61:: .12"'. -,~ b..5!a J.1~0~

8,

....
I

---..,;...

'--'

..X.
R.

Kr lal 0.Dl8

....

...... &'66

S,

HI T. W R. PI 14 Hg leI TI Pb Po 8' 2.783 1.732 0.382 0.359 0.430 0.315 (0.261 0.020 0.103 0.243 1.09 2.00 3.232 3.72 (4.18) 3.55 " -SQ:~ .J.,!il.. },12. _~2.... ,Q,.ii9::- .Q,20~ !J..6~ COill QJ81 g.3Si O~]7 ~.qp- 1.:1.9.- 1..3,L 3~17 (3'&L

Cs

..

- ;l.j;

R, Te Rh Pd Ag I. Cd T. S. '" Sb (2.971 0.833 1.702 2.725 3.208 2.704 1.808 1.007 0.467 0.411 1.11 0.383 O'~~ll )~ ~ Jl.>,g JIO!l "073U o:'S62' _0&53 0.~9'3 2-\." h ...... .Q.2lIf .t::6ii i,3.5
U

Hb

Mo

'"

.,

,
~.

"->-

At

"
Lr

(\!JJ'

'C f' (0.0201 10.lll \0.251

R.
0 ...6
{~T~

F,

Ce 1,1 0.239

4.18;

Hd Pm Sm E, Dy E, Gd Th Ho 0.306 0.327 (0.35) 0.294 0.147 0.383 0.399 0.384 0.397 0,411 .lll:: :...96.- (2.~ J;'Q.; -,S<>. .~ ~ -Z6P~ )..sz,t .2.""-

'"

Tm

Yb

"

0.397 0.133 0.411 2.52J :Z.52... .2.3

i:. ,
~

Th
0.543

L847 1!:3r :Lill (Ql

P, Hp P. U (0.761 0.987 <0.681 0.54

Am

:Pi:

-.,. .. ---- Cm

8k

CI

Es

Fm

Md

Ho

"--

.
- - - . - - - j f - " -1
FigurCl 3 Coocd'llates of the two oscmooton.

As a model we consider two identical linear hannonic oscillators I and 2 separated by R. Each oscillator bears charges t: with separations %1 and X2. as in Fig. 3. The particles oscillate along the r axis. Let PI and 1'2 denote the momenta, The force constant is C. 111el1 the hamiltonian orlhe unperturbed system is

"0"" -PI + lexi 2m

+ -1'2 + iCXi 2m

(1)

Each uncoupled oscillator is assumed to have the frequency Wo of the strongest optical absorption line of the atom. Tbus C :> ~ Let ". be the coulomb interaction energy of the two OSCillators. The geom elry is shown in the figure. TIle internuclear CQOnlinate is R. TIleD
ICCS)

?t, --+ R R + Xl

e2

el

tl tl ------;
-

X2

R + %1

R-

(2)

Xl

in the approximation Ix,l. Irsd ~ R we expand (2) to obtain in lowest order:

1/,

a.: -

~X,XI R3

(3)

Table 4

t'mperties or inert gas crystals

(Extrapolal:cO loOK and zero pressure)

t<;eano:sI-

Experimental
~"e
.~..,

disunc"e,

II.

-"'"
rnA
3.13 3.76 4.01

_ .....
JOfli
=~ ",r~

~"n in

Lennard-Jones potential, Etl. 10

in to- IO erg

Mehong
point, K

aloon,

kJlrool

IIlVlalorr>

.V

'.

u.
in

A.

(liquid at z~m pressure)

N.
AI

X.

'"

4.3.

1.88 7.74 11.2 16.0

0.02 0.060 0.116 0.17

24 84
117 161

24.58 21.56 15.76 14.00 12..13

14

50
167 225 300

2.56 2.74 3.40 3.65 3.96

H
13.595

Table 5

Ioni7,ation energies

LI
5.39 81.01

B.

'.32
M,

,
19.45

The total enerlO' required to remove the first two electrons is the sum of the Rrst and second ionizatiOn potentials. Source: National Bureau of Standartls CilUllar 467.

H.
24.58 7898

27.53

8.30- .1T :26 14.54 T3.61 33.45 35.64 44.14 48.76

F N. PA2 21:-56'
52.40 62.63

N.
5.14 52.43

AI
. Energy 10 remove one electron, In eV .--Energy to remove two electrons. in eV
~

SI

CI

7.64 22.67

'

'5.91C If.lS~ (0.55" 10:3~- J3.p.r 15.76" 24.80 24.49 30.20 340 36,81 43,38

"

K C. S. '4."34"- 6.[f- 6:56

36.15 17.98

n
6.83' 20.46

V 6.74 * 21.39

C,
6.76~

M,
7.4323,07

F.

23,25

7:-~ 7.~ i.!Th- 7.72 ' 9.39';." YOO~ 24.08 24,91 25.78 27.93 27.35 26.51 23,81

Co

NI

Co

Z,

G.

G.

LSS:
S,

30.0

A, B, K, S. 11.84 "i4.OO ~1":' 9.75:31.2 33.4 38.56

Rb
4.18 31.7
r

S,
5,69 16,72

Z,

Nb
5.77
21.22

Mo

6.5-- 6.95 18.9 20.98

7.18 23.25

T.
7.28 22.54

Ro
7.36 24.12

Rh

P,

A.

7.46. i f j J 7.W S:9"'" '5.78..... 7.~ SS4-- 9.QI- 10":'"45" 12)3 25.53 27.75 29.05 25.89 24.64 21.97 25.1 27.6 29.54 33.3

C'

I,

Sb

T.

X.

C,

B.

La
5.61 17.04

HI
22.

T.
24.1

W
7.98 25.7

R.

0,
8.7 __ 26

3,89- 5.21 29,0 15.21

7. : : 7.88

727 24.5

I, ,--

PI
8.96 27.52

Ao

H.

9.22'" IM3 29.7 29,18

TI Ph BI Po 6:TI- UI""" r.29~ 8.43~ 26.53 2244 23.97

F,

R.
5.28 15.42

A.

6.9 .. 19,0

-"

At Lo
L,

R,
Ib.74~

C.
6.9f"

P,
5.76

6:3r
U

N'

Pm
~

Sm
5.6 ......

Eo G' S.6L 6.I6 Am

Tb
5.74---=-

=.;:
Cf

Dy

--Ho E,
~

Tm

Yb

6".~ 5.0::'

1
1-

"

Th

P.

-.

Np

Po

Cm

Bk

'=.:..
-

--- --E, Fm M'

No

"

.
The total hamiltonian wilh the approximate form (3) for 1/1 can be diagonalized by the normal mode transfonnation (4)

or. on solving for

Xl

and

X2.

(5)
TIle subscripts 8 and a denote symmetric and antisymmetric modes of motion. Further. we have the momenta Pa, P. associated with the lY.'O modes: PI'" V2(P.+PJ;
1

(6)

TIle total hamiltonian :Jio + HI after the transformations (5) and (6) is

,,= [_1

p' + 2m-2

-'-(C _""')~] + [_1 p' + -'-(C + ""')~]. R 2m"2 R

3 3 "

(7)

nle two &equendes of the coupled oscillators are found by inspection of (7) to be

with WeI given by (Clm)lI2. In (8) we have expanded the square root. The zero point energy of the system is ilS(w. + w..); because of tile interaction the sum is lowered from the ullcoupled value 2 . ~IiWo by

tJ.U = !h(tJ.w

+ t::.w..) = -Ii,.... . - --. = -R 6. -v B en,l

1(""')'

(9)

111is attractive interaction varies as the minus sixth power of the separation of the m'o oscillators. 'nlis is called the van cler Waals interaction, known also as the London interaction or the induced dipole-dipole interaction. It is the principal attractive interaction in crystals of inert gases and also in crystals of many organic molecules. TIle interaction is a quantum effect. in the sense lhat 6U _ 0 as Ii - O. Thus the 7.ero point energy of the system is lowered by the dipoledipole coupling of Eq. (3). 11le van der Waals interaction does not depend for its existence on any overlap of the change densities of the tv.'O atoms. An approximate value of A for identical atoms is given by Ii. where IiWo is the energy of the strongest optical absorption line and a is the electronic polarizability. Chapter 13.

C..,.tlI1 8ind~

ti3

A-~
Figure" Elcctnmic dlllrg" distributions o\'erlap as nloms llpproach, The solid ci.. c1es lknote the ""e1li.

Repu4ive l,lteraction
As the tv.'O atoms are brought together their charge distributions gradually overlap (Fig. 4), thereby changing the electrostatic energy of the system. At sufficiently close separations the overlap energy is repulsive, in large part because of the Pauli exclusion principle. TIle elementary statement of the princi. pie is that rn'O electrons cannot have all their quantum numbers equal. Wllen the eharge distributions of two atoms overlap, there is a tendency for electrons from atom B to occupy in part states of atom A already occupied by electrons of atom A, and vice 'ersa. The Pauli principle prevents multiple occupancy, and electron distributions of atoms with closed shells can overlap only if accompanied by the partial promotion of electrons to unoccupied high energy states of the atoms. Thus the electron overlap increases the total energy of the system and gives a repulsive contribution to tlle interaction. An extreme example in which the overlap is complete is shown in Fig. 5. We make no attempt here to evaluate the repulsive interaction 2 from first principles. Experimental data on tlle inert gases can be fitted well by an empirical repulSive potential of the fonn WR l2, where 13 is a positive constant, when used together with a long-range attractive potential of the form of (9). 11le <''OI1stants A and 13 are empiriclJl parameters determined from independent measurements made in the ga~ phase; the data used include the virial coefficients and the viscosity. It is usual to write tlle total potential energy of tv.'O atoms at separation II as
U(R)

~ 4'[ (~)" - (~)']

(10)

where ~ and u are the new parameters, with 4wfi. A and 4wl 2 B. TIle potential (10) Is known as the Lennard-Jones potCQtial, Fig. 6. The force between the tv.u atoms is given by -dUldR. Values of ~and ugiven in Table 4 can

sn.e
The 'nathe

erbp energy nat"nIIy depends m the radi~ di~ribullon of ~ about.,.,t. _ . if;osl calculation is alwa)'s oompl.e:ated ",-en if !he ~ dist~ion is kllown.

(oJ

T",01

ek~1"'"

nIftXY. - il:> 9S ~"

O 0 O
II 11 tle

'"

'"

I'igu~ 5 lbc ~ect of I\luli principle on the ~bl\'c energy, in an extreme example, two h)'drogclI aloms are pushed together unlil the protons arc almost in COfltact. n.e energy of the electron S)~em aJonc can be taken from observations on atomic He, ..hich has two electrons. In (a> the flectrons hnc antiparallel "Pills and the r>.uJi principle has no effect: the electrons U'e bound by -78.98 e\'. 111 (b) Inc spin' are plIrallel: the Pauli principle IOrc-es the promotion of an clcclrQI' from a 18 f orbllal of H to II 2s t orbobll of I-Ie. The electrons now Ilre bound by -59.38 eV, leI.s than (a) by 19.00 eN. l1lis is the amount by which the Pauli pcinciple has iI~ the repulsion. We ha..., omitt~ the repuhh-e coulomb energy 01' the tYoV protons, ,,-hidl i$ the "me ill Leth (ll) "leI (b).

, ,
,

, , ,

.. ,, ..
l'l..... _

" "
-0:,

Figure 6 Form of the LeIlnanl-JOOes potential (10) "hkh dcsc'ibc5 the inlcl"Ild:ion oflwo Inert gllS atoms. The minimum ooeun at AlIT _ 2'.... 1_12. Notjc,e how sleep the etJn'C is Inside the miniIT''um, 81M! bow fb,t it is o,,,tsodc the mlmmum. The ,..J.ue of U at the minimum is and U '" 0 at

R"

fT.

be obtained from gas-phase data, so t.hat calculations on propertics of thc solid do not invoh'e disposable parameters. Othcr empirical form.s for the repulsive int.eraction are widely used, in particular t.he exponential fonn A. exp( -Rip), where p is a measure of t.he range of the interaction. This is generally as ea.~y to handle analytically as t.he inverse po"..er law form.
Equilibrium Lallice Constants

Jf we neglect the kinetic energy oflhc inert gas atoms, thc cohcsive encrgy of all incrt gas crystal i.s given by summing t.he Lennard-Joncs pot.entiaJ (10) ovcr all pairs of atoms in the cryst.al. If there arc N atoms in the crystal, the total potential energy is

U_~IN(4')[L' (--"-r -L' (--"-r] J pvR J puR

(11)

",-here pyR is the dist.ance between rcference atom i and any other atom j, expressed in t.erms of t.he nearest neighbor distancc R. The factor t occurs 'with

the N t.o compensat.e for counting twice each pair of at.oms. The summations in (11) have been e"aluated, and for the fcc structure

L'p;12 =
)

12..13188 ;

"" -, = 14.45392 .
.. PIJ
}

(12)

1bcre arc 12 nearest-neighbor sites in the fex: structure; we sec that the series are rapidly converging and l13.\e \'alues not far from 12. The ncarest neighbors contribute most of the interaction energy of inert gas crystals. The corresponding sums for the hcp structure arc 12.13229 and 14.45489. If We take UIOI in (11) as the total ene'l,')' of the crystal, the Cfjuilihrium valuc Ro is given by requiring tI13.1 UIOI be a minimum with respcct to variations
in thc ncarest. neighbor distance fl:

dU,. dR"" =
"..hencc

0 = -2N (12XI2.13) RI3 - (6X I4 .45)"'Rf

u"

if] '

(13)

RJu =

1.09 ,

(14)

the same ror all elements with an fcc structure. The observed \'alllcs of RJu, using the independcntly determined \'allies of u given in Table 4, are:
Nc

1.14

M 1.11

Xc
1.10 1.09

The agreement with (14) is remarkable. TIle slight departure of RJu tOr the Iightcr atoms from the IIlliVeTSal value 1.09 predicted for inert gases can be

66

explained by zero-point quantum effects. from measurements on the gas phase we have predicted the lattice COIlstant of the crystal.

Cohesive Energy
The cohesive energy of inel1 gas crystals at absolute zero and at zero pressure is obtained by substituting (12) and (14) in (11):

U...c(R) =
and. at R =

2lV~[(12.13)(:Y2 -

(l4.45)(;fl

(15)

Ro.
(16)

the same for all inert gases. This is the calculated cohesive energy when the atoms arc at rest. Quantum-mechanical corrections act to reduce the binding b)' 28, 10, 6. and 4 percent of E~1. (16) for Nc, AT, Kr, and Xc., respccth-c1y. The heavier the atom, the smaller the quantum correction. \Ve call understand the origin of the quantum corrt:ction by consideration of a simple model in which an atom is COnfined by fixed boundaries. rf the particle has the quantum wavelength A. where A is ~termined by the boundaries, then the particle has kinetic energy pi/2M = (hlA'f/2lof with the de Broglie relation p = hlA fOr the connection betv.'een the momentum and the w~welength of a particle. On tins model the quantum 7.ero-point correction to the energy is inversely proportional to the ma~s. '111e final calculated cohesive eucrgies agree with the ClCpcrimental value.~ of Table 4 within 7 to 1 percent. One conSC(juence of the quantum kinetic energy is that a crystal of the isotope Nel!D is observed to have a larger lattice constant tllan a crystal of Ne 22 . The higher quantum kinetic energy of the lighter isotope expands the lattice., because the kinetic energy is reduced by expansion. The obscrvcd lattice constant~ (exh~dpolated to absolute zero from 2.5 K) arc NeW, 4.4644 A; Nc 22, 4.4559 A.
IONIC CRYSTALS

Ionic crystals are made up of posith'e and negative iOlls. nle ionic bond results from the electrostatic interaction of oppositely charged iOIl~. Two common crystal structures found for ionic crystals, the sodium chloride and the cesium chloride structures, were shown in Chapter 1. 11te electronic configurations of all ions of a simple ionic crystal OOrTespond to closed electronic shclls. as in the incrt gas atoms. In lithium fluoridc the configuration of the ncutral atoms arc. according to the periodic table in the front endpapers oftbis book, Li:ls~. F:l~2.02ps. '(11e singly charged ions havc the configurations Li+: 1$2. F-: 1s2zs~p6, as for helium and ncon. respeeti\'c1y. Inert gas atoms ha\c closed shells. and the charge distributions are spherically S)'mmetric, We expect that the charge ill.!, utiom on each ion in

Cryslal Dinding

61

)-"ig",.., 1 E::lcclron density dislribulion ill the base plane of NaCl ..ncr x-ny sludies br C. Sehokneehl. 11le "umbers Oil the conlours gin, lhe rcl~th'e eleeIron COflCClll..... lion.

+
c.

5.14 c\'

l""i7.atl""
e"",'1O'

G..!
Figure 8 The e ..ergy per molecule unit of a eryslal of sodium chlOride I, (7.9 - 5.1 + 3.6)" 6.4 eV 10<>\.... lhan Ihe e"ergy of scparaled neutral atoms. 11le !alike energy wilh respect 10 separated jo"s is 1.9 eV per molecule unil. All ,-alues on tl,e rtgure ~re c.-per[me.. laI. Values of tl,., ionization energy are given in Table 5, a"d ,-al""5 of the electro" affi"ity are given in Table 6.

01

-a e.,

G..!

3.61.N

.
""

""
+
e.,

""

''''''"'
~ffinily

l.<}"tal

79.\

C,hn,
rnt.,.~\

an ionic CryStal will have approximately spherical symmetry, with some distor' lion near the region of contact with neighboring atoms. l1lis picture is confinned by x-ray studies of electron distributions (Fig. 7). A qUick estimate suggests that we are not misguided in looking to electrostatic interactions for a large part of the binding energy of an ionic crystal. The distance between a positive ion and the nearest negative ion in crystalline sodium chloride is 2.81 X 10- 8 em, and the attmctive coulomb part of the p0tential eneTb'Y of the h'lO ions by themselves is 5.1 eV. This value mdY be compared (Fig. 8) with the experimental value of7.9 eV per molecular unit for the lattice energy of crystalline NaCI with respect to separated Na+ and C!ions. \Ve now ca' 'ate the energy more closely.

'.

68

TobIe 6

Electron affinities of negathoe ions

The electron affmit)' is positive f~ a stable negoLti",e ion. Source: H. Holop and Lineberger, J. Phys. Chern. Rc Data 4, 539 (1975).
AO~

w. C.

Electron affinity rneTgy

AO~

Electron affinity entrgy

II

0.7542 eV

Si

U C

0.62
1.27 1.46 3.40

0 F

N,
AI

0.55 0'"

.,
CI
I K
~q

P S

LJgeV 0.74 2.08 3.61 3.36 3.06

0.50

Electrostatic or Madelilng Energy

The long-range interaction between ions wHh charge is the electrostatic interaction t/Jr, attractive bctv.ct:1l ions of opposite charge and repulsive between ions of the same charge. The ions arrange themselves in whatever cryst.tl structure gi\les the strongest attractivt> interaction compatible with the repulsive interaction at short distances bct\\:een ion cores. The repulsive interactions between ions with inert ga.~ configurations arc similar 10 those between inert gas aloms. Tbc van del" \Vaals part of the attractive interaction in ionic crystals makes a relatively small contribution to the cohesive Cllergy in ionic crystals, of the order of 1 or 2 percent. The main contribution to the binding energy of ionic crystals is electrostatic and is called the Madclung energy. If Uj} is the inter.lL1:ion cnergy between ions i and), we denne a sum VI which includes all interactions involving thc ion ~:
UI =2:'UU '
J

(J7)

where the summation includes all ions except) = i. \\'e suppose that Ui} rna)' be writtcn as the sum of a central field repulsive potential of the fonn A exp(- rip), where A and p are empirical parameters, and a coulomb potential rllr. nlUs

(ees)

(18)

where the + sign is taken for the like charges and the - sign for unlike charges. In SI units the coulomb iuteractioll is q2/41ror' "'e write this section in CCS units in which the coulomb interaction is q2/r.

3 Cry.'al Binding

69

The repulsive lenn describes the fact that each ion resists overlap with the electron distributions of neighboring ions. We treat the strength Aand range p as constants to be determined from observed values of the lattiee constant and compre.~.sibility; we have used the exponential form of tbe empirical repulsive potential rather than the R-11' form lIsed for the inert gases. The change is made because it may give a better representation of the repulsive interaction. For the ions, we do not have gas-phase data available to permit the independent determination of A and p. We note that p is a measure of the range of the repulsive interaelion: when r == p, the repulsive interactiOll is reduced to e -lor the value at r == O. In the NaCI strueturc the value of U, does not depend on whether the referenee ion i is a positive or a negative ion. The sum in (17) can be arranged to eonverge rapidly, so that its value will nol depend on the site of the referenee ion in the crystal, as long as it is not ncar the surface. We neglect surface effects and write the total lattice energy V,,,, of a erystal comJXlscd of N molecules or 2N ions as V,,,. == NV,. Here N, rather than 2/1,', OCCurs because we must count each pair of interactions only once or each bond only once. The total lattice elwrgy is defined as the energy required to separate the crystal into individual ions at an inHnitc distant.'C apart It is convenient again to introduce quantities Pu such that r,} - PiiR, where R is the nearest-neighbor separation in the erystal. If we indude the repulsive interaction only among nearest neighbors, we have

(ees)

A exp(-RJp) -. R

,('

(nearest neighbors)

(19)

--

1 ,('

P, R
Thus

(otherwise).

(ees)
where
Z

101'

== NU == N(ZAe-WP _ aR ) '

q2

(20)

is the number of nearest neighbors of any ion and

'<', ()
&

L- - - e I\ladelung constant

Vii

(21)

The sum should include the nearest-neighbor contribution, which is just z. TIle (~) sign is discussed just before (25). 'Ibe value of the Made\ung constant is of centr.t1 important."C in the theory of an ionic crystal. Methods for its calculation are disel.lssed below.

70

At the equilibrium separation dUto/dR = 0,

!iO

that

(CCS)
0'

N - = --cxp(-Rlp) dR p

dU,

Nr.A

Nal + -= 0 2
R

(22)

(ces)

~ cxp(-Rdp) = palhA .

(23)

nilS determines the C<luilihrium separation flo if the parameters Po A of the repulsive interaction arc known. For 51. replace rf b>' q2J4moThe total lattice energy of tile crystal orm ions at their C<lllilibrium separation flo may be written, using (20) and (23), as

(ces)

(24)

The term -Nul/Ro is the l\.-!adelungcllcrgy. We sballlind that p is or the order of 0.1 Ro so that the repulsive interaction bas a vcry short range.

Evaluation of the Madelung Constant

'nit: first caiculation of the coulomb energy constant a \Y3$ made by Madclullg. A powerful gcncl",l.l method for lattice sum calculations was developed b)' Ewald and is developed in Appendix B. Computers arc now used for the calculations. The definition of the Madclung constant a is, b)' (21). a =

L'
j

() . PjJ

For (20) to give a stable crystal it is neccli.~ary that a be positive. [r we take the rereren(.'e ion as a negative charge the plus sign will apply to positive ions and the minlls sign to negative ions. An equivalent definitkm is

~=L' ()
R
J TJ

(25)

where TJ is the distance or the jth ion rrom the reference ion and R is the nearest-neighbor distance. It must be emphasized that the value given ror a will depend on whethcr it is defincd in tenns of the nearest-neighbor distance R or in tenns or the lattice parameter a or in tenns or some other relevant length. As an example, we compute the ~Iadeiling constant ror the infinite line or ions or alternating .sign in Fig. 9. Pick a negative iOIl as rcreren(.'C ion, and let R

..

denote the distance between adjacent ions. 11len

: =2[~ - ~ + ~ - 4~ + -".]
0'

a=2[1-"!"+'!"_"!"+ ... J . 2 3 4
the factor 2 occurs because there are two ions. one to the right and one to the len. at e<Jual distances 1j_ We slim this series by the expansion In(1

+ %) =

%- -

C
2

+ - - - + ... _
3 4

r3

%4

Thus the Madelung constant for the one-dimensional chain is a = 2 tn 2. tn three dimensions the series prescnl~ greater difficulty. It is not possible to write down the successive tcrms by a casual in~"pcction_ More important. the series will not coll\-erge unless the successive terms in thc series arc arranged SO that the contributions from the po.<;itivc and nClltive terms lleRrly cancel. Typical values of the Madelung constant are listed below. based on unit charges and referred to the nearest-neighbor distance:
Structur~

Sodium cblor,de, NllCl Cesium Flloridc, !=sCi Zinc f:llendc, cu~ic ZnS

1.147565
t.162675 t.6381

"Ole Madclung and repulsive contributions to the binding of a KCI crystal are shown in Fig. 10. Properties of alb-Ii halide crystals having thc sodium chloride structure are given in Table 7. 1he cak'lllated values of the lattice energy are in exceedingly good agreement with the obscrved values_

" "

, ,
<

~ , -,
-<

, ,
-, -,
-00

<

ElinIO",n

-" -"
rocure
10

COVALENT CUYSTALS

The oovalent bond is the classical electron pair or homopolar Ixmd of chemistry. particularly of organic chemistry. It is a slrong bond: the bond bcto..,.CCIl two earbon atoms in diamond with respect to separdted nelltral atoms is comparable ",ilh the bond strength in ionic crystals. The covalent bond has strong directional properties (Fig. 11). Thus car~ bon. silicon, and germanium have the diamond structure, with alomsjoinoo to four nearest neighoors at tetrahedral angles, c,"en though this aJTangcmcnt gives a Iov.' filling of space. 0.34 orlhe available space. (:ompared wilh 0.74 for a dose-packed structure. The tetrahedral bond allows onl)' four nearest neighbors. whereas a close-packed slnlcluTe has ]2. We should not O\Icrcmphasize the similarity of the bonding of carbon and silicon. Carbon gives biology, bUI silicon gives gcolOi.'Y and semi{:ondudor technology. The covalenl bond is Ilsnally formed from two electrons, one from each atom participating in the bond. The electrons forming the bond tend 10 be partl)' localized in the region between the two atoms joined b)' the bond. The spins of the two eleetrons in lhe bond arc antiparallcl.

Table 7 Properties of alkali halide c11'stnls with the NaCI structure All values (except those in brackets) at room temperature and atmospheric pressure, \\.ith no correction for changes in and U from absolute zero. Values in brackets at absolute zero temperature and 1.ero pressure, from private communication by L. Brewer.
Neareslneighbor separation lie. in A Bulk modulus B. in 10\1 dyn/em 2 or 1010 Nlm 2 6.71 2.98 Repuhlve energy parameter zA, in 10-~ erg 0.296 0.490 0.591 0.599 0.641 1.05 Repulsive range parameter p, in A 0.291 0.330 0.34<J 0.366 0.290 0.321 0.328 0.345 0.298 0.326 0.336 0.348 0.301 0.323 0.336 0.348 Lalliee energy rompared !(l free ions, in kcalImol Experimental 242.3[246.8J 198.9[201.8] 189.8 Calculated 242.2 192.9 181.0 166.1 215.2 178.6 169.2 156.6 189.1 161.6 154.5 144.5 180.4 155.4 148.3 139.6

no

LiF

LiCI
Linr

2.014 2.570 2.751

2.38
(1.71)

Lil
NaF NaCI NaBr Nal

3.000
2.317 2.820 2.989 3.237 2.674 3.147 3.298

In7
214.4[217.9] 182.6[185.3] 173.6[174.3] 163.2[162.3] 189.8[194.5] 165.8!I69.5] 158.5[159.3J 149.9[151.1] 181.4 159.3 152.6 144.9

4.65
2.40

1.99
1.51 3.05 1.74 1.48 1.17 2.62 1.56 1.30

1.33 1.58
1.31

KF KC] KB, KI RbF

nbCI RbBr

2.05
2.30 2.85 1.78 3.19

3.533
2.815 3.291 3.445 3.671

RbI

1.06
physi~s

3.03 3.99
16. 1 (1964).

1
J.
;

"

Data rrom various tables by M. P. TOOl. SoUd nale

Figure II Cakulatcd \'alence elel.1ron ron<:entration in g<.'J"Ulanium. n'e numben 0" thccontour:s give the deC1ron <'"UI<.'L'fltratioll per primiti\'e cdl, with l(,.'r \'lIk'flCe clcdrons per atom (eight ek'Clrons pl.'>" primiti\'e cell). Note the high concentration midway along the (',<,"(',e bood, as we expect fur CO\,.]ent bonding. (After J. R. CllChkowsky and M, l~ Cohen.)

111C binding of molecular hydrogen is a simple example of a covalent bond,

11lc strongest binding (Fig. 12) occurs whell the spins fA the two electrons are
antiparallel. 11le binding depends on the relative spin oricntation not because thcre arc strong magnetic dipole forces between thc spins. but because the Pauli principle modiflCS thc distribution ofcharge according to thc spin oricuta tiou. This spin--dcpcndcnt coulomb cucrgy is called tbe ellchange interaction. 11lc P. ... uli principle givcs a strong repulsive intcraction bctv.'CC1l atoms with lIlIed shclls. If the shclls arc not lIlled, electron overlap can be aC'COmm&dated without excitation of electrons to high encrgy states and the bond will be shorter. Compare the bond length (2 A) of CI2 with the intcratomic distance (3.76 A) of Ar in solid Ar; also compare the oohesive energies given in Table L The difTerencc between CI 2 and Ar2 is that the CI atom has live ek'Clrons in the 3p shell and the Ar atom has six. filling the shell. so that the repulsive interac-tion is stronger in Ar than in CI. The elements C. Si. and Ce lack four electrons with respect to 611cd shells. and thus these dements (for cxample) can havc an attractive interaction associated with chargc overlap. lbe electron configuration of carbon is h:l2l-2p 2. To fOrm a tetrahedral system of covalent bonds the carbon atom must rlfSl be promoted to the electronic configuration 1s~p3. This promotion &om the ground state requires 4 eV. an amount morc than regained when thc bonds arc fornled. Thcre is a oolltinuous r...nge of crystals between the ionie and the covalent limits. It is often important to estimate the extcnt a given bond is ionic or covalent. A scmicmpirical tht..'Ory of the fmctiollal ionic or oov...lcnt char...cter of

'-

Crystal Binmng

7:;

'-'r--rr,---.-,---,---,-----,-----,

E
-~

~ O.21----\.\-----+....:,,~~

~ ~ -O_21-

s(1

+-"~--l.-<::...--I_.j

-OAI---+---+---j-;

-OGoL--j---l,---J,~tl..:....:i,::~:::i5~---~J,
Figure 12 Energy of mokcular hydrogL'Tl (H,.) rcfern,d to >eparated neutral atoms. A ncgati'e energy corresponds to binding. The LU..e N refers to a dassieal OIkulation ....ith free atom dUlTg<: densities; A is the result for pm-allcl electron spins, taking the Pauli e~dusion principle into acCOIJnt, aoo S (the stable stat,,) for antiparallcl spins. 11,edL'Tl,ity of dUlTge i, reprC$<.'Tltcd by contour lines for the "at"" A and S.

a bond in a dielectric crystal has been developed with considemble success by J. C. Phillips, Table 8. p. 76.

METALS

Metals are characterized by high electrical conductivity, and a large number of electrons in a metal are frcc to move about, usually one or two per atom. The electrons available to move about are ealk-d conduction electrons. The valence electrons of the atom become the conduction electrons of the metal. In some metals the interaction of the iOIl cores with the conduction electrons always makes a large contribution to the binding enerb'Y. but the ehamctcristic feature of metallic bindiJlg is the lowering of the enerb'Y of the valence electrons in the metal as compared with the frcc atom. This lowering is exhibited by severnl simnle models treated in Chapters 7 and 9.

"
Table 8 Fractional ionic character
Fradiunal

or bonds in binary crystab

Crystal Fractional ionic chQl'lloelcr

Cl')'$tal

ionic rnara<.1cr

Si SiC

C,
200 20s 205<

0.00 0.18 0.00 0.62 0.62 0.63 0.61 0.71l 0.69 0.70 0.67 0.42 0.36 0.32

C......
CoSb

0.31 0.26 0.85 0.85 0.77 0.... 0.79 0.79 0.92 0.94 0.96

AgCI
AgO, Agi

2oT,
CdO CdS CdSe CdTc

"gO
MgS MgSe

,"p

," .... InSb

After

UP NaCI RhF

J.

C. P1,iDips., Boowlr lind bDouk in ~ o n , Academ;", Preq, 1973, Chap. 2.

The binding energy ~ an alkali metal crystal is considcrabl) less than that of an alkali halide crystal: the bond fonncd by a conduction elect rOil is not very strong. 11lc interatomic distances arc relativel)' large in the alkali metals because the kinetic CIlC'1.'Y of the COIlduction electrons is 100.\'Cf at large interatomic distances. This leads to weak binding. Metals tend to crystalli7.c in relatively close packed structures: hcp. fcc. bee, and some other closely related structures, and not in looscl},.packcd structures such as diamond. In the transition metals there is additional binding from inner electron shells. Trnnsition metals and the metals immediately following them ill the periodic table have large 4-clcctron shells and are characterized by high binding energy.
IIYDROCEN BONDS

Because neutral hydrogen has onl)' one clectron. it should form a covalcot bond with onl)' one other atom. It is \.;nown, however. that under certain conditions an atom ofh)'drogcn is attracted by rather strong forces to two atoms. thus forming a hydrogen bond between them, with a bond energy of the order of 0.1 eV. It is believed that the hydrogen bond is largclyionic in charolcter, being fonned onl)' betv.'ccn the most electronegative atoms, particularl)' F, 0, and N.

3 Cf'Vdl'I Binding

77

Figure 13 l1le h}'drogen difluoride ion HF; iJ lilllb!liZAld by. hydrog.,n bond. The sketch
is of an C1dremc moUel or the bond, extl'(.'ll'le in the 5enJC 11\8.1 the proton iii iiI"""." bare of

electrons.

(oj

In the extrcmc ionic fonn of thc hydrogcn bond, the hydrogcn atom loses its electron to another atom in the molecule; the bare proton forms the h}'dragen bond. The atoms adjacent to the proton are so close that more than two of them would get in each other's "'-':Iy; thus the h)'drogen bond connects onl)' two atoms (Fig. 13). The h)'drogcn bond is an important part of the interaction betwcen H 20 molecules and is responsible together with the electrostatic attraction of the electric dipole moments foc the striking ph)'sical properties of viatcr and icc. It is important in certain ferroclectric crystals (Chapter 13).
ATOMICRADU

Distances between atoms in crystals <:an be measured very accuratel)' by x-ray di.l.li':lction, oRen to 1 part in l~. Can we sa}' that the obser....ed distance between atoms rna)' be assigned partly to atom A and partl)' to atom B? Can a Mmite meaning be assigned to the radius of an atom or an ion, irrespective of the nature and composition of the crystal? Strictly, the answer is no. The charge distribution around an atom is not limited b)' a rigid spherical boundary. Nonethcless, the concept of an atomic radius is fruitful in predicting interatomic spacing. Thc cxistence and probablc lattice constants of phases that have not ret lx.'Cn synthesi7.cd cml be predictt.'<.I from the additive propertics of the atomic radii. Furthcr, thc electronic conHg. uration of the constitucnt atoms oftcn can be infcrred br comparison of measured and predicted values of the lattice constants. To make predictions of lattice constants it is convcnient to assign (Table 9) sets of scif-colilsistent radii to various types of bonds: one set for ionic crystals with the constituent ions &-coordinated in inert gas closed-shell configurations, another set for the ions in tctrnhedraU)-coordinatcd structures, and another set for 12-coordinated metals (close-packed). 1bc predicted sclf-consistcnt radii of the cation Na + and the anion F- as given in Table 9 wouldlcad to 0.97 A + 1.36 A = 2.33 A. for the interatomic scpard.tion in the crystal NaF, as compared with the obser.'Cd 2.32 A.. This agreement is much better than if we assume atomic (neutral) coo6guralions for

H
2.08

Tablc 9 Atomic and ionic radii Values approximate only. Units are 1 "\=10- 10 m. For original r{!rcrences see W. B. Pearson, Crystal chemistry and physics of metals and alloys, Wiley, 1972.

H,

Li

B,
.
0.35 1.06 1.13

B
0.23 0.88 0.98

C
0.15
,0~77

N
1.71 0.70

0
1.40 0.66

F
1.36 0.64

N,
1.58

0.68
1.56

0.92

N'
0.97 1.91

Me
0.65 1.40 1.60

K
1.33 2.38

C,

So

--_ 0.81 1.&4 0.93 1.80

Standard radii ror ions in inert gas (filled shell) configuration _ _ :Radii of at.QIps w!t~~ in t~.tnp.pdral co~lcnt bond~ Radii or ions in 12-coordinatc<! metals

Ti
0.68
1.46

0.99 1.98

Rb
1.48 2.55

S,
1.13 2.15

Y Ao

. " ,..,.. .... - . 1.35 1.28 1.26

Nb
0.67 1.47

C, Mo
1.40
.~

M' -

F,
.
~

Co
~

Ni

C,
8

1.27

1.25

1.25

1.35 1.28

"
Cd

A'
0.50 1.26 1.43

S,
0.41 1.17 1.32

P
2.12 1.10

S
1.84 1.04

C'
1.81 0.99

A,
1,88

G,
0.62 1.26 1.41

G,
0.53 1.22 1.37

A,
2.22 1.18 1.39

S,
1,98 1.14

B'

K,

0.74 1.31 1.39

1.95 2.00 1.1A"

C,

1.67

_.
,
1.75

"

0.80
1.60
-~

. -" -,...;..
Y,
1.47

Yo
.
1.36

R,
1.34

Rh
1.35
"

Pd
1.38

A,
1.26 1.52 1.45

S,
0.71 1.40 1.55

Sb
2.45 1.36 1.59

T,
2,21

,
2.16 1.28

X,
2,17

0.81 0.97 " lA8. ).441.57 1.66

1.34.Po

B,

1.35 2.24

2.73

HI " US
1.88 1.11

R,
1.37

1.58

R,
1.38

0,
1.35 1.36

PI
1.39

A,
1.37

H,
1.10 1.48 1.57

1.41

- Pm
1.81

T'
0.95 1.72

Pb
0.84 1.75

Bi
1.70

Al
~

R,

1.44

1.76 Vb
1 74 30

1'-

C,
1.01
1...71;;.

P,
'~1.83'"

Nd

Sm

'" Yh
0.99
."V'"

1.82U 0.83

- .'

t-r: 1.80
P,

"

2.04 2 LaoJ'

- Cm

Gd

Vb

0,

Ho

1':"80" f.,8

[77">1.. "1.77"
CI

-Fm

"

Tm
~

"

1.76

1.75

t.94' -

P,
0.90

Np
1.56

Am

Bk

1.80 " C63

-- -

T56

"'k 1.64

-1.81 - -

"

Md

" _. - -No

TabLe 10

Use of the standard radii of ions gh'en in TaMe 9

11Je interionic d~tance D is represented I' D N - lie + Rio + lJ.N, for lOIlic cl)'Stals, where N is the coordination number of the calion (positi,'c ion), Rc and Rio are the standard ratlii of the cation and anion, and lJ. N is a correction for coordination number. Room temperature. (After Zacharial;en.)
N

.o..,(A)
-0.50 -0.31 -0.19 -0.11

""<Al
-0.(6
0

.0.".(.0\.)

5 6 7 8

10

+0.04 +0.08

II I'

+0.11 +0.14 +0.17 +0.19

Na and F, for this would lead to 2.58 A for the intcratomic separation in the crystal. The lalter value is Hn.n. distance in metallic Na + interatomic distance in gaseous F~. 11lc interatomic distance bclWcefl C atoms in diamond is 1.54 A.; onehalf of this is 0.77 A.. In siHcon, which has the same crystal structure, one-half the interatomic distance is 1.17 A. In SiC each atom is surrounded by four atoms of the opposite kind. If wc add thc C and Si radii just given, we predict 1.04 A. for the Icngth of the C-Si bond, in fair agreemcnt with thc 1.89 Aobserved for the bond length. This is the kind of agreement (a fco.\' percent) that wc shall find in using tables of atomic radii. a
lonk: Crystal Radii
In Table 9 we include a set of ionic crystal radii in inert gas configurations. 11m ionic radii arc to be used in conjunction \\ith Table 10. Let us consider BaTi03 in Fig. 13.10, with a lattice constant of 4.004 A at room temperature. Each Ba...... ion has 12 ncan.. 'St 0-- ions, SO that the coordinatiol1numbcr is 12 and the CQrrection A I2 of Table 10 applies. If we suppose that the structure is determined by the Ba-O contacts. we ha,'e D 12 "'" 1.35 + lAO + 0.19 = 2.94 "" or a = 4.16 A; if the TiO contact detcrmines the structure. we have D6 = 0.68 + lAO = 2.08 or a ,. 4.16 A. 1be actual lattice constant is somewhat smaller than the estimates and rna)' perhaps suggest that the bonding is not purely iOllic, but is partly covalent.
3For rtfcrtrJa:s 01\ 1It00nic and ionic radIi, see Pearson. Tahle 9, p. 76, L. Pauling, TM nature olille cMm:tJl bond, 3rd ed., Cornell, 1960; J. C. Slater, J. Chern. Phys 1, 3199 (1964); 8. J. Austill and V. Ileine, J. arm. Pbys. 4S, 928 (l966); It C. PlInons and v. F. WcissbJpl, Zeits. r.

f'bywl; 2Ol, 492 (1967), S. Geller, Z. Kristallugnplue liS, I (1967): It D. Shann<ln. Acta Cryst. K.. oatdes and fIuoridet. R. D. St.nnoo and C. T. PW\O.,tt, Acta CrysI:. 825, 9'25 (1969). An early dud)" is: by W. L. Bragg. "1lIe arnngt:ment of atoms in ~ - Plv.1os. M. .

AJ2. 751 (I1176); and.

{~4O, 2Sl (192O).

80

ANALYSIS OF ELASTIC STRAINS

We consider the clastic properties of a crystal viewed as a homogeneous continuous medium rather than as a periodic array of atoms. The continuum approximation is usually valid for clastic \\'llVCS of wavelengths>. longer than 10- 6 em, which means for frequencies below 1011 or 10 12 Hz. Some of the material belov.' looks complicated because of the unavoidable multiplicity of subscripts on the s)'mbols. The basic physical ideas arc simple: we usc Hooke's law and Newton's second law. Hooke's Jaw states lhat in 3n clastic solid the strain is directh' proportional to the stress. The law applies to small strains onl)'_ \Vc say that we arc in the nonlinear region when the strains arc so large that Hooke's law is no longer satisfied. \Vc specify the strain in terms of the components en. e~/I' e=.. e%/I' ell~' eu which arc defined below. We treat infinitesimal strains only. We shall not distinguish in our notation betv.'ccn isothermal (constant temperature) and adiabatic (constant entropy) defonnations. The small differences between the is0thermal and adiabatic clastic constants arc not oRen of importance at room temperature and below. We imagine that three orthogonal vectors i, 5', i of unit length arc embedded securely in the unstr.lined solid, as shown in Fig. 14. Aftcr a small unifonn defonnation ofthe solid has takell place the axes arc distorted in orientation alld in length. In a uniform defonnation eaeh primitive cell of the crystal is defonned in the same \\'a)'. The nev.'axes 11:', y', z' may be written in tenns of the old axes:
11:'

= (1

+ Eo-)i + f:IY + f:~ ;

y' = f:lI'x

+ (1 + f:w>Y + E~i
(1

(26)

z' = Ez.,i + E;:,S +

f:~Ji

The coefficients Eo(3 dcline the deformation; the)' arc dimcnsionless and have values ~ 1 ifthe strain is small. 111e originalaxcs were of unit 1enb1.h, but the new axes \\ill not necessarily be of unit length. For clffi.mple,
X"II:'

= 1 + 2f:..,+ ~+ ~y+

c.:..

whence:x'lil I + Ell' + "'. nc fractional changes of length of the i. y, and i a.~es arc E....., f:W> E=. respectively, to the lirst order. What is the effect of thc deformation (26) on ail atom originally at r = xi + !/)' + z1? Thc origin is taken at some other atom, If thc deformation is ulliform, then after defonnation the point will be at the position" r' = xx' + yr' + '0.', 11lC displacement R of thc defonnation is defined by
R -c' - c= x(x' -i) + y(y' - y)+z(z' -i).
~This

(Z7)

is ob;.lQUSly C'OITed if ......, cbooM: the i

aril;

such Ihlll r - n, lhen r' - n' by ddWtion

cL:r'.

,
,~

,hI

Figure 14 Cotwdinal:e ..<eli ror the description of the slate of strain; the orthogJlnaI unit ues ill lhe unslnined lUte (1,1 an: deformed in the strained state ~.

0<,

from (26),

R(r) (:U:u + yEll" + U'u)X

+ (Uxy + !lEW + ZEy)y + (XE.n +

yE....

+ u..Ji

(28)

TIlis rna)' be written ill a more gencml form by introducing II, v, w such that

the displacement is givCll by

I R(r) "" u(r)i + t;(r)y + w(r}Z . I

(29)

If the deformation is nonuniform we must relate II, v, w to the local ~traiJls. We

take the origin of r close to the region of interest; then comparison of (28) and (29) gives, by Taylor series expansion of R using n(O) = o.
X"

"'u

liI;X-'

d.

ax'

etc.

(30)

It is usual to work with coefficients eo(1 rdthcr than E(lfJ' \Ve define the strain components e...... cW ' Cu by the relations

-t< = 1/11 iJy

d.

e ~ .'" -u = o;r;'

dW

(31)

using (30). The other strain components e~, cl/1> e.... arc defined in tcnns of the changes in angle bch\ucn the axes: using (26) we rna}' dcllnc

82

e'lI "'X'r
,

I'

~Ex+t;
II

= - +ilx ily = -+ QZ ily

=

{lIlOV

.y',z':!!!:

+E

iJv

ilu;

\I'

(32)

e="" z' x

3! E

+ Eu

-+ilx ilx

au

aw

We may replace the 3! signs by = si~ns if we neglect terms of order 1';2, The six dimensionless coefficients eo ,J..=e(30) completely defmc the strain. Dilation

The fractional increase of volume associated with a deformation is called the dilation. The dilation is I1c~tivc for hydrostatic pressure. The unit cube of edges x, y, z has a volume after dcfonnation of
V' = x' y')( z' ,
(33)

by virtue of a wcll-knovm result for the volume ofa parallelepiped having edges x', y', z'. From (26) we have

x"y'xz'=

',.
l)

(34)

Products of two strain components have been neglected. The dilation

given by

is then

(35)
Slress Compmlnt8
The force acting on a unit area in the solid is defined as the stress. There arc nine stress componcnts: Xr Xv' X Yr. Y\!. Y Z... Z\I' Z. Thc C'J.pitallcttcr indic-J.tcs the direction ofthe force. and the subscript indicates the nonnal to the plane to which thc forcc is applied. In Fig. 15 thc stress component X" represents a force applied in the x direction to a unit area of a plane whosc normal lies in the x direction; the stress component X\I represents a force applicd in thc x dircction to a unit area of a plane whose normal Iks in thc y direction. Thc number of independent stress com[Xlnents is reduced from ninc to six by appl}'ing to an e1emcntary cube (as in Fig. 16) the condition that thc angular acceleration vanish,:> and ht:nce that the total tOrtluC must be 7.cro. It
~Tl.is doc; not mean we emmol tl"<'31 problems In wh,d,lhere Is fTlt:ans II...t we can use the sLJtie .,tu/ltion to .IeRne the clastic ron\

,gular acceleralion; it jt"t

Figure 15 Stre" compo"e"! X. IS a f'oree appliet.! It}

the'" directlOflto It unit area oflt plane ",IIOW IVJt~ ... 1 lies In the 1: din."CIi(ln; X. is .pploed in the 1: di..,. .':'<'\ to a unit area cI. plane ",hose uOI"rnallies in tloo: ,j

din:dion.

,
x,--

l' L---+t-.
Y.

~monstralion th"l fur a body in stat~ ccl"i1ibnum Y... X.. l1lC sum cI the lOccn In the '" direction is ~. "The sum or the ~ in the II dll'tll.1ion is abo rem. The tobI ""-'e \'aI1ishes. "The tol2l ton~ about the origin is

Figure 16

abo u:tO if Y, .. Xy

follO\\'S that
1'",=ZIl;
z.,=X~;

XII "",,

(36)

The six indcpendent stress components IDa) be taken as Xn 1/' z.,. 1'~, Z. XvStress components have the dimensions of force per unit area or cncrgy per unit volume. 11le strain components arc ratios of lengths and arc dimensionless.
ELASTIC COMPUANCE AND STIFFNESS CONSTAI'\TS

Hooke's law states that for sufficiently smaIl deformations tllC strain is directly proportional to the stress, so that the strain componcnts arc linear fUllCtions of the st.. ~ ..s componcnts:

..
enc
= SIIX.

ew =
eu = e,a =

eu =
e.. =

+ St2r" + Sl~. + 5 14 Y.. + Sa:;Z.. + Sl~, 521 X. + 522Y, + S~.. + 52-1 Y + Sts? + S~1I 531 X. + 5 32YI/ + SxJ:. + 53-/0Y" + SJ + S~II S.uX. + S.u}'1/ + S,,~ + 5...Y. + S.:;z,. + S"c;.X" ; SlnX. + SS2}', + S~ + 554 Y. + S.ssZ. + 5:wXv ; 561 X. + 562f,. + S~ + 564 Y. + S~ + S~, .

(37)

x" = Cllen + Cl~w y ... = Cue.:.: + C~w

+ C.;,ezz + Cue", + Gtst!u + Cu;e.. ; + Cdu + Ct#w:t + C~ + CttJ?-. ; Z. = C31 e"" + C3-<>ew + C3:I!" + C36e.. + C3."#u + C:wflall ; y. = C41~ + Cuew + Cd.,;. + C...e.. + C~ + C4f.I!,.., ; Z" = CSlen + C~1/lI + Cs:#u. + C,54e p + C~ + Cs#.. ; Xv = eme.:.: + C~w + Croe~ + C64 e... + C6.u + Cf#...

(38)

The quantities Sil. S 12 are called clastic compliance constants or elas~ tic constants; the quantities Gil. C 12, . . arc called the clastic stiffness constants or moduli of elasticity. 1bc 5's ha\c the dimensions of [arcaV(forccl or {volumclflcncrgy]. The C's have the dimensions oflforcc]l(arca] Or lcnergyY [volume).

EMslie E'lergy Density

The 36 constants in (37) or in (38) may be reduced in number by severnl considerations. The clastic energy density U is a quadratic function of thc strnins, in thc approximation of Ilookc's law (recall the expression for the energy of a stretched spring). Thus we may write
(39)

v.ilere the indices I through 6 arc defined as:

I - xx ; 2 & yy; 3 -:z.z;; 4 _ yz ; 5 - zx; 6. xy.
(40)

The

C's are related to the C's of (38), as we see in (42) below.

1be stress components are found from the derhllti\'e of U with respect to

the associated strain component. This result fOllows from the definition of p0tential energy. Consider the strcss X, applied to one face of a unit cube, the opposite face being held at rest:

ElMI/c Conof...,to

)
Notc lhat only the combination i{Cop + CpJ CTltCrs thc stress-strain relations. It follows that thc clastic stiffness constants are symmetrical: Croll

= i{C<>j:I + C~ = CfJr>

(4')

11J.us lhc thirt)'-six clastic stiffness constants are reduced to twcnty-one.

Elastic Stiffneu Constants

of Cubic CryslQU

Thc number of independent elastic stiffness constants is reduced further if thc crystal possesses symmetry elcments. Wc now show that in cubic crystals there are only three independent stiffness constants. We assert that the elastic cnergy density of a cubic crystal is
U" JCII(e:.,

+ t!." + c~ + iC..c<c~ + e:.. + c~ + CiJ.ewe~ + evJ!1B + eueaJ. (43)

(ene.l' + ...);
(e~&:r

and that no other quadratic tcnns occur; that is

+ ...)

do not occur.
11le minimum symmetry requirement for a cubic structure is the existence offour three-fold rotation axes. The a;tCS are in the {HI) and equivalcnt direclions (Fig. 17). 11le cffect of a rOtation of 27rl3 about thesc four axes is to interchange the :t, y. ~ axcs according to thc schcmes
-x_~_-y_-x

-x- y_

~_

-:t ,

(45)

according to the axis chosen. Undcr the first of thesc schemes, for example.

e;. + e:v + ~-e:l' + e~. +~. ,

,4 3
CD ,

,, ,, , ,,

t1gu1'e 17 Rotation by 21rlJ :o.lKKrt the axis marilee! 3 rlw,ges ~_!(, "'_ ~ and::r:_:r.

86

and similarly for the other terms in parenthcsC5 in (43). Thus (43) is invariant under the operations considered. But each ohhe terms exhibited in (44) is odd in one or more indices. A rotation in the sct (45) can be found \\ hich will change the sign of the tenn. because eq = -e>(-v)o for example. Thus the terms (44) arc not invariant under the required operations. It remains to verify that the numerical factors in (43) arc (.'Orrect. By (41)
aUl~ = XII: = ellen

+ C1it.. + e&J

(46)

1bc appearance ofC11t;.. agrees v,.ith (38). On further comparison, we sec that
(47)

Further. from (43),

aUlae"" ""
on compartson ",ith (38) we have
C fil = C 62 =

X,I =

Cue"" ;
::II

(48)

e 63 =

eM

eM

eM =

C-H .

(49)

Thus from (43) we find that the array of values of the clastic stiffness constants is reduced for a cubic crystal to the matrix

X. Y,

z.
Y. Z. X,

C" C" C" C" C" C" C" C" C"

0 0 0 0 0 0 0 0 0

0 0 0

0 0 0 0

0 0 0 0 0 (SO)

C..
0 0

C..
0

C..

For cubic crystals the stiffness and compliance constants arc related by
C 44 = 11544
;

Gil -

eli::::

(5 11

S,v- 1

Gil

+ 2C l2 = (5 11 + 2S I :v- 1
mahi.~

(51)

These relations follow on evaluating the inverse

Bulk Modulus and Compressibility

to (50).

Consider the unifonn dilation en "" w the energy density (43) of a cubic Cl)'stal is
U = i(C 11

:I:

u "" la. For this defonnation

.

+ 2C Ir)lf

(52)

We may define the bulk modulus B by the relation

U~IBI?

(53)

In.ti.: COIut4n"

81

Fi#,nl 18 eu"," of ~"(}I",ne 6x 4y lJ.it acted on by a stress -X.(:t) 011 the face at :t, and

Volume Ax!Jy IJ.=.

XJ:t + 1J.x);II XJ:t) + face at :t + IJ.%.

ax. IJ.x on a;TI,e IIet

tIIe paraIleI force

ill

(a~.

1J.x) lJ.y IJ..:.. Other forces in the:t direc-

tion arise from the ....mation across the eul>e<f the stresscs X~ and X.. which are not show". 1"e net :t ~'I)mpollcnt of the toree on the cube is

"-,_

ax. =ax +-ax.) 4x,-,y!:>z.. (~+ ax ay ait

11,e toree equals the mllSs of the cube tinle< the wmpollcnt of the acceleration in the x directioll. 11,e mass is plJ.% lI.y luand the act",leration is ;;uJat~.

which is C<luivalent to the definition

-v dpldV. For a cubic Cl)'stal

(54)

B = l(C I1 + 2Cd .

The compressibility K is defined as K"'" liB. Values of Band K arc given in Table 3.
ELASTIC WAVES IN CUBfC CRYSTAlS

By (.'Onsidering as in Figs. 18 and 19 the forces acting on an clement of volume in the cl)'stal we obtain the C<luation of motion in the x direction
p-~-+

ilu

at2

ax~

aXil

ax

ay

aXiL +-az

(55)

here p is the density and u is the displacement in the x direction. Thcre arc similar equations for the y and z directions. From (38) and (50) it follows that for a cubic cl)'stal

au aen (~aeu) c (aeit' +ae=) p-,-=C +-- + --; I1 --+C I2

2

at

ax

ax

ax

ay

a:z

(56)

here the x, y, z directions are parallel to the cube edges. Using the definitions (31) and (32) of the strain components we have
p

a2 u at2

Gil

a2 u (a2u ifu) + (C 12 + C4 J ( ax a2 v a2 w ) 44 iJ!l + ac a:c + G iJy + iJx iJz

~

where u, v. w are

rcomponents of the displacement R as defined by (29).

(57.)

"

Figure 19 If spri"gJiA and B are stretched cqm,lly, the blod< betw"en th"", expcnno net force. '1,is illustrates the fact that" uniform SU"<l55 X. in" solid d"",s "o1 gi''e II. net force on" volume dement. If the spring at B is stretched more than the $pring at A, the block between t~1TI will be accelerated by the force X.(B) - X.{A).
en<.'e!i

nlC corresponding equations of motion for iPvliJF and iPwlaf arc found directly from (57a) by symmelry:
P-----:2=C11---::2+CH ----:"i"+----y or fJy ax iJz

a2v
2

a2v

(ilv

02V )

au a + (C 12 + C H ) ( +-- ;
ax iJy ay iJz

w)
2

(5Th)
p af = Gil

aw

iJ2w

iJc

+ C~4 ~ +

(iJ W
2

(l2W )

iJr! + (C 12 + c)

ax 3z + ay iJz

(l2 u

av

)
(57<)

We now look for simple special solutions of these equations.

Waoos in t'le (IOOj Direction

One solution of (57a) is given by a longitudinal wave

11

= Uo exp [i(Kx - wi) ,

(58)

where 11 is the x component of the particle displacement. Both the v.."a\'Cvector and the particle motion arc along the x eube edge. Here K = 2m}" is the wavcvector and w = 21TV is the angular frequency. If we substitute (58) into (57a) we find
w'lp = cllJ(l ;

(59)

thus the velocity wlK of a longitudinal wave in the [100] direction is

Vs = II}" = edK = (C lI /p)lf2 .

(60)

Consider a transverse or shear wave with the waveveetor along the x cube edge and with the partide displacement v in the y direction:

v = Vo exp [i(Kx - wt)] .

On substitution in (57b) this gives the dispersion relation
w 2p =

(61)

c44J(l

(62)

thus the velocity ,.xK of a transverse wavc in the [100] direction is

Vs =

{C";p)112 .

(53)

The identical velocity is obtained if the partide displacement is in the z direc-

,
( tion. Thus for K pQI'U.llci to [100) the two independent shear ","aves have equal velocities. This is not true for K in a general direction ill the crystal.

Waul in the (lIO/ Direction There is a spocial interest in waves that prollagate in a face diagonal direction of a cubic crystal, because the three clastic constants can be found simply

from the three propagation velocitics in this direction. Consider a shear wave that propagates in the xy plane with particle displacement w in the z direction
w = Wo cxp [i(~.r

+ ~fJ -

wt)] ,

(64)

whence (32c) givcs

w'p ~ C~(I<: + KiJ ~

c~K'

(65)

independent of propagation direction in the plane. Comider other wavcs that propagate in the ry plane with particle motion in tile XlI plane: let
1/ c

"0

cxp (i(K.x

+ 1YJ -

wt)] ;

v = ~ cxp U(K.x

+ ~y -

wi)]

(66)
From (57a) and (57b).

Jpu = (Cll~ + C.u~ + (C12 + C...JK"Kyv ; c.rpv "" (CII~ + C44~ + (Cll~ + C..JIV;, .

(67)

TIlls pair of equations has a particularly simple solution for a \\"3\'C in the [llO} direction, for which Ks = ~ = KJVi The condition for a solution is that the

dctenninant of the coefficients of II and v in (67) should C(lual zero:

-',lp + J{C II + C",JI(l i(C 12 + c"JIf

i{C I2 + C",Jk.-2 -w2 p+i(C II +C",,)If =0.

(68)

111is equation has thc roots

urp = i(C II + C I2 + 2C..Jk,'1

urp =

i(C n - el2l~ .

(69)

The first root describes a longitudinal '\\"3.VC; thc second root describes a shear wa\"c. How do wc dctcnninc thc direction of particle displacement? Thc flnt root wben substirulcd into the upper equation of (67) gives i(C n

+ C I2 + 2C",JIfIl = i(e n + c"JIf" + i(C 12 + C"4)~V

(70)

whcncc the displacement components satisfy II = V. TIlUS the particle displaccment is along [110} and parallclto th~ K vector (Fig. 20). The sc<:oncl root of(44) when substituted into the upper e<luution of (67) givcs i(C n - CdIfu'" i(C l l

+ C..JI(lu + i{C12 + C"",)k,'1v, (71)

...

00

/L

/K

Jr- k

'
I.

I
"
.,

c-----T
\Va'", in [Ill] ."rectioll I.: (Cll + Xlt + 4C4~)

\K
!

\Va'", in [1001 diredion

L:CIl
T: C44

WK'" in IJ 10] direction L:~(Cll+CI2+2C44)

Tl'C"

r. J(Cll-CI2+C44)

Ti: llCII-C1V
Figure 20 ElTective elastic constants fOf the three mocks of elastic ""aVeS in the principal propagation directio"s in cubic crystals. The two frdllsvcrse mooes lin:' degt:.,era(t, for prop.agation in the l100J and (1111 directions.

whence 11 = -v. The particle displm:emcnt is along [lI0] and perpendicular to the K vector. Selccted values of the adiabatic clastic stilfncss constants of cubic crystals at low temperatures and at room temperature arc given in Table ll. Notice the general tendency for the clastic constants to decrease as the temperature is increased. Further values at room temperature alone arc given in Table 12. There arc three normal modes of wave motion in a crystal for a given magnitude and direction of the v.''llvevector K. In general, the polarizations (directions of particle displacement) of these modes arc not exactly parallel or perpendicular to K. In the special propagation directions I tOO}, [Ill}, and [lIO} of a cubic crystal tv.u of the threc modes for a given K are such that the particle motion is exactly transverse to K and in the third mode the motion is exactly longitudinal (parallcl to K). The analysis is much simpler in these special directions than in general directions.

SUMMARY

Crystals of inert gas atoms are bound by the van der Waals interaction (induced dipole-dipole intcmction), and this varies with distance as I1R 6 The repulsive interaction behveen atoms arises generally from the electro static repulsion of overlapping charge distributions and the Pauli principle, which compels o\'erlapping electrons of parallel spin to enter orbitals of higher energy. IOllic crystals are bound by the electrostatic attraez
~

of charged ions of

a-tic eon.r......

"

Table II

Adiahatic elastic ~Iirrneu ronstant5 of cubic cryitals at low temperature and at room t<!ffip<!raturc

The values given al 0 K ,,"'eTC obtained by extrapolation of measurt:fficnb carried out down 10 4 K. The table was compilL-d with the assistance ti Profc5sor Charles S. Smith.
in 10" dpldan" (10" N/ml )
St~ t:(lnJbnb.

""....
IV

Cn
5.326 5.233

C..
~1>I9 ~1>15

C~

Temperalu"" K

Dcn$ity, rJcms

1.631 1.607 0.874 0.818 0.818 0.754 0.511 0.461 0.<54

0.420

0 JOO 0 300 0 300 0 300 0 300 0 300

19.317 16.006 9.018 10.635

19.488

TO

U63
~609

1.582 1.574
1.249

C"

1.762 1.684 1.315 1.240 2.016 1.923

1.214 0.973 0.937 1.697 1.631 0.619 0.607 0.0341 0.0314 0.454 0.423 1.500 1.500 1.761 1.761

Ag A"
Al
K

t.l43
1.068 0.0416 0.0370 0.555 0.495 2.612 2.500 2.341 2.271

0.316 0.282 0._ 0.0188 0.194 0.149

2.733

295 0 300 0 300 0 300

11.599

I'b
Ni

I.m
0.712 0.717

1.311

.968
12.132

Pd

opposite sign. The electrostatic energy of a structure of 2:.'Y killS of charge

q is

(CCS)

U ~ -Na- = -N

q2

()(l 2. -.
YiJ

where a is the Madclung constant and R is the distance between nearest neighbors. Metals arc bound by the reduction in the kinetic energy of the valence electrons in the metal as compared with the frcc atom.

.
Table 12 Adiabatic elastic stiffnen oolllilanl$ of several cuhic crystals at room temperature or 300 K.
Sl~

in IOu dyneIcm' or 10" N/m l

"'(lnstants,

c"
Diamond

c..
1.25 0.062 O.1l4

N,

s;

Ce

CoSb 'oSb

10.76 0.073 0.135 1.2B5 1.66 0.885 0.672

~86

5.76
0.042 0.088

0.483
0.639 0.... 0.367 0.81 0.124

0.680 0.796
0.<33 0.302 1.48 0.126

MgO

f'aCl

0.487

A covalent bond is characterized by the overlap or charge distributions or antiparallel electron spin. 1bc Pauli contribution to the repulsion illi reduced for antipardUel spins, and this makes possible a greater degree or overlap. The o\'crlapping electrons bind their associated ion cores by c1cctrostatic attraction.

Probkrrni
I. Quantum toIid. In a quantum lOIid the dominant repulsive energy is the zcro.point

energy of the atom5. Consider a crude ont....<Jim..:ns.ional modd d cry.o;taUiIiC He4 with each He atonl confined to a linc segment of It:ngth L In the ground slate the wave fUlidioll within each segment is taken as a h;M Wil\dcllb'th of a free particle. Find the :rero-point kinetic cnc'l.'Y pcr particle.

2. CoM-roe ~ ofboc ondf~ nftm. U5[ng the Lcnnard-Joncti pott.. ...tial. calculate the ratio dlhc cohesi\'e energ.ie5 r:L Deoll ill the bee and rcc 5trueturcs (Aru. 0.958). The lattice 511ms for the bee 5tructurcs are:

L' p;12""9.1l4IB;
J

I' pii6-12.253.'3.
J

3. Solid molecuWrhydrogllB. For Il r one finds from mt..'il$Un:mCnt5 on the gas that the

Lennard-Jones paramctLTI are f - 50 X 10- 16 erg and u "" 2.96 A. Find the colled\'e energy [n kJ pl.'!" mole of H r : do the calcl.llatioll for all fcc 5trueture. Treat each H r mokcule :as a 5pht..>re. 11te omcn"(..-d \':alue of the colu:a:ive energy is 0.751 kJI mol, milch 1(.'5s than we cakuiatt..-d, so that quantum corrections must be \'cry important.

F.L".'''' CONlant.

93

4. Pvssibilityo[ionic crystn13 n+n-. Imagine a Cl)'stalthat exploits for binding thc coulumb attraction ofthc positivc and nl..'g:ltive ions of the same atum or mokcule H. This i~ hc!iL...L'tI tu OCCur with certain organic mok>f:ules, but it is 1I0t found .....hen It is a single atum. Use the data in Tables 5 and 6 to cvaluate the stability of such a form of Na in the NaCI structure relative to normal melallic sodium. Evaluate the eneJl,'Yat the observed interatumie distance in melallic sodium, and usc 0.78 cVas the electron affinity of Na. 5. Linoor ionic crystlJI. Considcr a Iinc of 2N ions of alternating charge q with a repulsive potential energy AIR" bctween nearest neighbors. (a) ShO'-'o that at tile equilibrium separation

(ces)

U(1Io) = _

2N~n 2 (1

_;)
ork done hcre

(h) Lei the crystal be CQmprcs~ed ~o that Ro -+ Ro(I - 6). Show that thc in compn.'Ssing a unit ICIII,'th of the cl)'stal has the Icading tcrm jC~,

(ces)
10 obtain thc result~ in 51, replace q2 by q2/4w&J. Note; We should not expect to obtain tllis result from thc cxpn.'Ssion for U(Ro), but we muM use the complete cxpres~ion for U(R).

6. Cubic ZnS strllCture. Using Aand p frum Tablc 7 and thc Madclung constants given
in the text, calculate the CQllI..'Sive encrgy of KCI in the cubic Zns structurc dcscrihed in Chapter 1. Compare with thc value calculall..'<! for KCI in the NaCI structurc.
7. Divalent ionic crystals. Barium oxide has thc NaCI structure. Estimatc thc cohe sive cncrgies per molecule of the hypothetical crystal~ Ba+O- and Ba++O-- refcrred to separnted neutral atums. l1lc obscrved ncarcst-neighbor internuclear distancc is Ru = 2.76 A; the first and SI.."COnd ionizatiun potcntials uf Ba are 5. 19 and 9.96 eV; and the elt:ctrun affinitics uf the first and seCOnd cll..>f:tTOns added to the neutral oxygl..'Il atom arc L.5 and -9.0 eV. "Ine first ckdroll affinity of the ncutral oxygcll atom is the cncJl,'Y released in thc rcaction 0 + e -+ 0-. l1lc second ek-ctron affinity i~ thc ene'1,'Y released in the rcaction 0- + c -+ 0--. 'Vhich valence statc do yuu predict will occur? Assume Ro is the same for both furms, and Ileglt:ct the repulsive L>nergy.
8. Young's morn.lus and Poisson's raliv. A cubic crystal is subjt..'Cl to tcnsiun in thc 1100) dirLdioo. Find expression~ in teom uf the clastic stiffncsscs for Young's modulu~ and Puis~on's ratio as dcfined in Fig. 21.

9. umgitudinalwave velocity. Show that the velocity of a longitudinal wave in the [111] dirtdiOlI of a cubic crystal is given by v ., Ii(C II + 2C 12 + 4C44 )1PPII!. Hint: For such a wave II = V = w. Let " = uoe l K(X+I/+a)lV3e-t..... and use Eq. (57a).

.
T~_

~
r
L

'+,61_1
,

UnderO"~bQdY7

.1. __ . \

L,

, , , , I , ,
L
J

.-.,...., 21 Young's moclulus is ddiJM,d as streW dram lOr a temile ltreP acting in (Inc direction. "'ith the sides r# specimt.'tl M:I\ n.-c. PoWem'$ 1':01>0 is dt... fl, ..., ] all (&du:)f(Wl) b this situation.

I'lgure 22 This debmation is compounded from the: two shalrs,,~ - -II!....

10. Tran.wene lrove velocity. Show that the ~'e[ocily of transverse wa\"es in the [llli direction of a cubic crystal is given by v. - [j{C'1 - C llt + C. 4)1p]I/ll. Hill!: See I'roblcm 9.
11. ffcdive shea,. comtllol. Show that the shear constant j{C II - G,:z) in a cubic CT)'5' tal is defined by setting en co -ew "" k and all other strains equal to 7.ero, as in Fig. 22.. Hint; Consider the energy deruoity (43); look for a C' such that U - +e'e'. 12. lhterminantal approtJeh. It is known thai an R-dimensional. square matriJ: ....id, all demenu equal to unity has roots RandO, \\;th the R occurring once and the zero occurring R - I limes- If ail elements kwc the \'8lue P. then the roots are R" and O. (a) Show thai if the dja~ma1 elements are (I and all other elements are 1'. then there is one root equal to (R - l)p + (, and R - I roots equal to q - p. (b) Show &om the elastic equation (57) for a wa.~e in the [Ill) direction of a crJbk crystal that the determinantal equation which gives ",} as a function of K is

(/ - W-p
lJ

,J
q-6}p
I'

Jl

.. 0 ,

q- cJp

where q i~CII + 2C~ and " . iK'{c 11 + C.J. This exp~ the oonditm that three linea.- homogeneous a1geb.-aK:equations lOr the three displacement U)fflponents U, 11, W ha~'e a SOIUUO'l. Use the resllit of part (a) to find the three roots of

cJ; check

with dle results gh-en lOr Problems 9 and 10.

13. Genual propagation dirlion. (a) By suhstilutKln in (57) find tile dctcrminantal equaUon whKfl expres!ieS the <:a:ldition dlat the displacement
R(r) ~ [Uoi

+ CoY + u.oiJ

exp Il(K' r - wtll

be a solutiml of the clastic wa."e equations in a cubic crystal. (b) TIle sum of the roots of a dctcrmin3nlal c(jllation is e(jllallo tIle sum of tile diap"'"a1 clements Qu. Show

Ia.ric

Com,,,,,,.

from part (a) tlmt the slim of the squares of ti,e three elastic w~we velocities in any dirl,'C!:ion in a cubic crystal is equal to (C II + 2( 44 )1P. Reeallthat v~ = J-/KZ.

14. Stability crilerio. TIle criterion that a cubic crystal with olle atom in the primith-e t'ell be stable against small homogeneous deformations is that the energy denSity (43) be positive for all combinations of strolln components. \Vlmt restrictions are tllereby imposed on tile elastic stiffness constants? ([n matliematical language the problem is to find tIle conditions that a real symmetric quadratic form should be posili~-e definite. TIle solution is gi~-en in books on algebra; see also Kom and Kom, Mathemalit;nl Handbook, McCraw-Hili, 1961, Sec. 13.5-6.)Ans. C 44 > 0, CII > 0, C~, - ~t > O. and C ,I + 2C l t > 0. For an example or the instability whiell results when CII ~ C,t see L. R. Testaroi ct al., I'hys. Re~-, Letters 15, 250 (1965).

References
M. Bam a"d K, Iluang, Dynamical theory of cryslal latti('!, o,.,ford, 1954. The c1aSlk work on

ionic crystals. G. Leihfried, in Encvelo. of physics 7/1,1955. Supcrh rt'Vkw oflallke properties. M. P. TO!;i, "Cohesion of ionic ...,Iitls in the &1m model, ~ Solid stale physics 16, I (1964), L Pauling, Natl1~ oftl", du:mir.al bond, 3rd L'<I., Cor""U, 1960. J. Ilirschfcldcr. C. F. Curti., and R. H. Hin!, Molecular theo'1l ofgcul'S and liquids, Wiky. 1964. G. K. Ilorton, ~ldL'a1 rare gas C1'),.lals," ArneI'. J. Phy. 36, 93 (1968). A good nwicw of the Iauice prop"rtics of crystals of the int'l'!. gases. T. K, Caylon!, ''1<.-''0001' dt'SCription of physical propcrht,s of cry.tals," Amt.'I'. J. Phys. 43, 8tH (t915). M. D. 'oeten :md L. Schaad, llydrogen bonding. Dekker, 1914. W. B. Pearson, C'1Istal d.emi.>ll'1l and physic! of melals and alfoy!, Wiky-Intcr.;cicncc, 1912. F. A. Cotton aod C. Wilk,nson, Adx;attced Inorganic chemi.>ltry, 4lh L'<I., Wiley, 1980. G. Maitl:md el aI., Intenooleculor forC'.$: tl"dr origin attd de''.rn,inatWn, OKford, 1981. II. J. 'u",brlJI<e, Crystal physic': nwcroS<;Olnc I,hysia of anirotropic !oIidS, Benjamin, 1914. R A. Cuyer, "Physic~ of qua"tum crystal.," Solid stale physics 23, 413, (1969). J. F. Nyc, PhysiClJI pr'OfNlrlies of C'1Islals: thdr rq>rt'sentatkm by lemot'S and matrices, o,.,ford, 1951. Il. H. Hunlington. "Elastic conSlanls of <:rystals," Solid stale physks 7, 213 (1958). A. E. II. Love, A Iroofi# on the mathernaticaltheo'1l of elasticity, Dover p"pt ...back, 1944. W. P. Mason, Physlall aCOu!11c! aI"lthe pr'01",rI;e! oflldids, Van Nostrand, 19J8. W. P. Mason, .,d. "'"pical acol1stlcr, in St.'VCral volumes, Academic Pr...., \'01- I, 1964. C. Zt.'1Il. . .., fl<l$fidty and anelasticfty of Wltals. Univt'1'Sity of Chicago PrL'SS, 1948. T. A. Bak, L'<I., l'lwnom anti phonon interoclions', Benjamin, 1964.

4
Phonons 1. Crystal Vibrations
VIBRATIONS OF CRYSTALS WITH MONATOMlC BASIS 99 First Brillouin :wne 100
Group velocity 102

Long wu,,-elengtll limit Derivation of r~ conslanL~ from cl<Jlflriment

TWO ATOMS PER PRIMITIVE BASIS

103 103
104

QUANTIZATION OF ELASTIC WAVES

PHONON MOMENTUM

107
108

INELASTIC SCAITERING BY PHONONS

110

SUMMARY PROBLEMS
I. Monatomic linear lattice 2. Continuulll wa,,-e equillion

112 112
112
112

3. 4. 5. 6. 7.

of two unlike atoms Kohn anomaly Diatomic chain Atomic vibrations in a metal Soft pllOnon modes

Ba~is

112 113 113 113 113 114

REFERENCES

Chapter 5 treats the thermal pmp<:rtics of phQl\Ons. Chapler 10 treats the interaction of phollons and photons, and tIm interaction of pllOllons and electrons.

I
~

........
,
'

x.....

t",ld

....

EII.........,..,.. ..ca-~
~

-,fVII'-

"-

----ill----ill1r-

........

......
Mnt""" ......,

~ ..

......

\~iuI_ .."'"

---

....
and C. A. Sbdr:. Am.

"""""'

I:ltt'UI -+- do>hc dtb1oIaI....

~

..... ...

r'gJ.IrI' I Importall~ elementary O<CItalWln5 in 16I1ds. 11lc origins d the COIK'eplS and the nameI of the eKital;olls lift diKt.lucd by C. T. W:dker
J.
Pb)'I. 38, 1380 (1970).

, , I, , , , , ,

,
I Y
~.

?
~ ,
I , , 'f , ,

1 , l
Of ,

, '"

"

'f

"i : , 1 I
~~

,\ ,
~. ,

, , , , , ,
""s

I'
I

I , ', i' , ,

: ,

.- ,
~

{o .... _ a - - l

Y , , , ? ,

7 ,

!,,
, , .,.
, ,
,

: , ,

Figure 2 (Dluhed lin"s) Plana; of al<">ms

9 ,

s+l

.+2

'"

l' ...
~ I"

when in l.'(jullibr;um. (Solid lincs) Planes

of a10ms Wht.~l di~placed as for a longitudinal wa,'c. The coc:mlinale II mt'asures the diliplacemt;nl of the planes.

...

II.

tI".

-,
Fogure 3 PI;mes of atoms as dilipbc;-ed during

of.

IrlI.n5\.,

~.

CHAPTER 4: PHONONS I. CRYSTAL VIBRATIONS

VIBRATIONS OF CRYSTAl.5 WITIl

MONATO~UC BASIS

Consider the elastie vibrations of a crystal with one atom in the primitive cell. We want to find the frequency of an elastic \rove in tenns of the wa~'evcc tor that describes the wave and in tenus of the clastic constants. The mathematical solution is simplest in the (1001. [1101. and [1111 propagation directMlnS in cubic crystals. 10ese arc the diredions of the cube edge, face diagonal. and body diagonal. When a wave propagates along One of these directions. entire planes of atoms move in phase with displacements either parallel or perpendicular to the dired:ion of the wavevector. We can describe \\ith a single coordinate u~ the displacement of the plane 8 from its equilibrium position. The problem is then one dimensional. For each ,\.'aVCVl,.'Ctor there arc three modes. one of longitudinal polarization (Fig. 2) and tv.-o of transverse polarization (Fig. 3). '''e assume that the elastic response of the crystal is a linear function of the forces. 111at is equivalent to the assumption that the clastic energy is a quadratic fun1,.1ion of the relative displacement of any two points in the crystal. Tenns in the energy that are linear will vanish in e<lllilibrium-sec the minimum in Fig. 3.6. Cubie and higher order terms may he neglected for sufficiently small c1astie deformations, but play some role at high temperatures. as we see in Chapter 5. We a.~sulile accordingly that the lorce on the plane $ eaused by the dispbcemcnt of the plane s + p is proportional to the difference lI.t" - u.oflheir displa(.'cments. For brevity we consider only nearesl-neighbor interaelions, so that p = 1. The total force on s comes from planes s J:

F. =

C(U~"'I - U.)

+ C(U.-I

- uJ .

( I)

This expression is linear in the displacemenls and is of the form of Hooke's law. The constant C is the force constant between nearest-neighbor planes and will diJfcr for longitudinal and transverse waves. It is convenient hereafter 10 regard C as defined for one atom of the plane. so that F. is the force on one alom in the plane $. 1he eqoation of motion of the plane $ is
M d/ = C(U.H

d'u

+ U._I

- 211.)

(2)

....herc AI is the

In,

of an atom. We look for solutions with all displacements

"

'00

having the time dependence cXp(-iwt). Then trIll/fir = -fJfu. and (2) becomes
-Mt;}lu. "" quo+!

+ 11._1

2uJ

(3)

This is a difference equation in the displacements Il and has traveling wave solutions d the form;

Un:1 = uexp(uKa)exp(iKa)

(4)

where a is the spacing between planes and K is the wavcvcetor. The value to usc for a will depend on the direction of K. With (4), we have from (3):

-tJlMu exp(isKa)=Cu{CXP[i(3'+ 1)Ka]+cxP(i($-1)Ka]-2 exp(isKa)}.

We cancel
II

(5)

exp(isKa) from bolh sides. to leave

~M

'" -C[cxp(iKa) + cxp(-jKa) - 2]

(6)

With the identity 2 cos KD = cxp(iKa) + cxp(-iKa), we have the dispersion relation connecting w and K:

or = (2CIMXI

- " " Ko) .

m that the slope of w versus K is zero at the zone boundary:

lIurIclK = (2Cll1M) sin Ka = 0

The boundary of the 6rst BrilJouin zooc lies at K =

mo. We show from

(8)

at K = "ITla, for- here sin Ka = sin (1T) = O. The special significance of phonon wavcveaors that lie 00 the zone boundary is developed in (12) below. may be written as By a trigonomctric identity

cJl :: (4C1M) sin 2 1Kn ;

A plot of w versus K is givcn in Fig. 4.

w:: (4CIM)U!jsin IKal

(9)

Fjrst Brillouin Zone

What range of K is physically signifw::ant for- elastic \.\-aves? Only those in the first Brillouin 7.onc. From (4) the ratio of the displacements of two successive planes is given by - - ::
11 6
116 +1 U

exp[i(8+ I)Ku] ..J :: exl..... lKo) . 11 exp(isKa)

(10)

The range -1T to +1T for the phase Kn covers all independent values cL the exponential. There is absolutely no point in saying that tv.-o adjacent atoms are out of phase by more than 71: a relative phase of I.21T is physically identical with a

<I

I'honON' I. Crv-fal vib,..,rioru

'"

",---------.------,-------,

..,
(.em)'"0.6

--.

>-I-~- ........... _----.I

uurn apPrOl'imation; hcre '" is directly pn'.>pOl1'onal to

"';gure 4 Plot of", versus K The reglon of K <II 1111 or A .. II ccrrelJlonds to the CQI>I;U' K

F1gure 5 lbe wave represcnt~'t! hy thc ~idC\lr"e conve)'S no informat",,> nol W..,,:n 11)'lhe dashed
C\1~.

Only w2"ck:ngths lonSU lhan 211 are rowded 10 represenl tile moliotl.

relativc phase of -O.8'IT, and a rclath'C phase of 4.27T is klcntical with O.27T. Wc need both positi\'c and ncgath'e valucs of K becausc waves can propagate to thc right or to the left. Thc range of independent valucs of K is specified by
0'

- ~ < K:S;~.

This range is the first Brillouin "I.one of the linear lattice. as defined in Chapter 2. The extremc values are K...-:: mao Thcre is a real difference here from an clastic continuum: in the continuum limit a -+ 0 and K.r- -+ o:;. Valucs of K outside of the first Brillooin zonc (Fig. 5) merely reproduce lattice motions described by values within thc limits
:!: 'lTJa.

Wc may treat a valuc of K outsidc thesc limits by subtracting thc integral multipic of2'ITJa that will givc a wavC\'cctor inskle these limits. Supposc K Iics

10.

outside the first wne, but a. related wavC\'ooor K' defined by K' - K - 2mtfa lies within the first zone, where JI is an integCT. Then the displacement ratio (LO) becomes
lJ.+lu. = cxp(iKa). cxp(i2:ll'Tl) exP{i(Ka - 21T11)) cxp(iK'a).

(ll)

because exp{i21Tn) = L Thus the displacement can always be described by a wavcvcetor within the first wne. We note that 2Tm!a is a reciprocal lattice vector because 2mo. is a reciprocal lattice vector. Thus by subtraction of an appropriate reciprocal lattice vector from K, we always obtain 3n equivalent wavcvcctor in the first wne. At the boundaries K...- = ~ of the Brillouin zone the soluHon u, = u cxp{isKa) docs not represent a traveling wave, but a standing """ave. At the 7.onc boundaries s'-"n-a = ~S1T. whencc

ma

II,

= u exp(:;!:is'n) = u (-1)" .

(12)

11lis is a standing wavc: alternate atoms oscillate in opposite phases. because u. = 1 according to whether.! is an C\'cn or an odd integer. The wavc moves neither to t}le right nor to the len. This situation is equivalent to Bragg reflection of x-rays: when the Bragg condition is satisfied a traveling wave cannot propagate in a lattice. but through succes.~ive reflections back and forth, a standing wave is set up. The critical value Kmu, = wla lOund here satisfies the Bragg condition 211 sin 8 = IIA: we have 8 = tw, cl = a, K = 2TT1A. II = 1, so that A = 2n. \Vilh x-rays it is possible to have" Cl:IUal to other integers besides unity because the amplitude of the electromagnetic \\lave has a meaning in the space between atoms, but the displacement amplitude of an elastic wave usually has a meaning only at the atoms themselves.

Croup Velocity
The transmis.~ion velocity of a wave packet is the group velocity, given as
vll'

= dwldK

0'
vll'

= grad" w(K)

(13)

the gradient of the frequency with respect to K. This is the velocity of energy propagation in the medium. With the particular dispersion rclatioo (9). the group velocity (Fig. 6) is
vI! = (Ca2 /M)l/l

cos iKa .

(14)

This is zero at the edge of the zone where K = wlo. llere the wave is a standing wave, as in (12), and we expect zero net transmission velocity for a standing wave.

,.,r-....:::-----,------,

"
Fipe 6 Croup \'dQcil) ~. VU'>''''' K. for mo::wkI m rlg. <I. AI the ZOI>e boundary the groop ,'ekJal) is

,,:.-------------Oi.~

=
Long Wllvelengll. Umit

When Ka 41: I we expand cos Kn

tion (7)
bCl.'OInC$

iii

I - !(Kaf, so thai the dispersion rela-

(15)
The result that the frequency is dircct:ly proportional to the wavcvcetor in the long wavelength limit is cqulvalenllo the statement that the velocity of sound is independent of frequency in this limit. Thus v.: wlK. exactly a.~ in thccontin uum theory of elastic waves-in the continuum limit a ::::: 0 and thus KG = O. Derivation

0/ Force COllstantl from

Erperinumt

In metals the effective forces may be of quite long range, carried from ion to ion through the conduction electro.l sea (Chapter 10). Interactions have been found between planes of atoms separated by as many as 20 planes. We can make a statement about the rallgc of the forces from the observed dispersion relation for w. 111C generalization of the dispersion relation (7) to p nearest planes is easily round to be
w2 = (21M)
,,>0

L.

C,Al -

(.'OS

pKu)

(l6a)

We solve for the interplanar rorce COnstants C" by multiplying both sides

by cos rKo, where r is an integer, and integrating over the range or indepen.
dent values or K;

L::

dK

wI: cos rKa

= 2 =

p~ C
::c

L:,:

dK (l - cos pKa) cos rKa

-2m:/a .
r. Thus

(1Gb)

The integral vanishes except ror p

MaI~' G" = - 2 dK
7T - ..In

wl COS 1'Ko

(11)

gives the rOrce OOj

nt at range 1m, lOr a structure with a monatomic basis.

,0<

n'\.'o

ATOMS PER PRL\tlllVE 8 ....SIS

The phonon dispe~ion relation shows new features in crystals with two oc more atoms per primiti\'c basis. Consider, for example. the NaCI or diamond structures, with twoalDms in the primiti\"c cell. For each polarization mode in a given propagation direction the dispcnion relation w versus K develops two branches, known as the acoustical and optical branches. We ha\'c longitudinal LA and lransven;e aooustical TA modes. and longitudinal LO and transverse optical TO modes, as in Fig. 7. If there are p atoms in the primitive cell. there arc 3p branches to the dispersion relation: 3 acoustical branches and 3p - 3 optical branches. Thus gcnnanium (Fig. Sa) and KBT {Fig. Bb}, each with two aloms in a primitive cell, have six branches: one LA, one LO, two TA. and two TO. The numeroiogy of the branches fOllows from the number of degrees of freedom of the atoms. With p atoms in the primitive cell and N primitive cells. there arc pN atoms. EadJ atom has three degrees of freedom. one for each of the 1', y, z directions, making a total of3pN degrees offreooom for the el)stal. The number of allowed K values in a single braneh is just N for one Brillouin zone. I Thus the LA and the two TA branches have a total of3N modes, thereh) accounting for 3N of the total degrees of freedom. The remaining (3p - 3)N degrees of freedom are accommodated b)' the optical branches. 'Ve consider aeubiecrystal where atoms of ma.u M, lie on one sct ofplanes and atoms of mass M2 lie on planes interleaved between those of the first set (Fig. 9). It is not essential thai the masses be different, but either the force constants or the masses will be different if the two atoms of the basis are in llonequivalent sites. Let a denote the repeat distancc of the latticc in the din~c'" tion nonnal to the lauil..'C planes considcred. 'Ve treat waves that propagate in a symmetry direction for whieh a single plane contains only a single type of ion; such directions arc (Ill] in the Nael structure and \100) in the CsCI structure. We writc thc cquations of motion under the assumption that each plane interacts only with its nearest-neighbor planes and that thc force constants are identical between all pairs of nearest-ncighbor planes. We refer to Fig. 9 to obtain d 2 11 = qv, + 0,_1 - 211,) ; 1

M d/
d'o,

(18)

M2 df = qu,+ I + II, - 20,)

IWe sho;ow in Chapter :I by applOcal.ioI1 of periodK" boundary c:ond'hons 10 lhe moda oflhe erystal of \olume V thalthere 1$ one K value itllhe volume {2.j'1V In FOllr;.,r- space. The volumed a BriUouin zone is (2.,.f1V, ..t.ere V, is lhe volume of. errst.l prim'tive cell. Thus lhe number of alJoo".-ed K values in a Brillouin zone is WV,...... och IS just N, the number of pnm'll\'e cells in the crystal.

-I

PhonoN I. C'1/Ifol VibrotioN

lOS

- - - :(2(:1"1,)""

: .(2CIIoI,J'"
, , , ,

! , ,
Fit;urC' 7 o.)Iica1 and _lical bn.nehes of lhe dililC'TSiorl rel.ollOO for I d,atomic linear IiIltice, showJng lhe hrnlting frequcnc.ics II K - 0 100 K - " - - ",fo. The btlicC' <U15l2Jlt IS fI,

" - --------i-"- - <

,,'.---------------,

,
6

,
= '1.
6 ,
0

0.0

TO
0

J,

Of

J.
TA

>A
0 0 0

TA

o.t

06

It.6

KIK-, .. 1111 r d.retUan

.
""',60,.

"-'t>-, .. till] dlft'dion

Fit;ure 8b D.spC'rsion l'UM.'S In II~ [ II tI d....-ctoon in KBr II 90 K. afler A. D. B. Woods. B :'\ BrocI<houS'C, R A. eo.,""r.
alld W. C<x-hran The cxll'l'll'0btion 10 K" 0 of Ih(' TO. LO branches art'l;llkod Ihc-se an: diso:usscd In O!apter 10-

Figure Sa Phonon dlspt'nion rcbtions In llw: 11111 direclion in ~,um It 80 K The Iwo TA phonon ~ arC' ~ It tile lIOOC' boulKWy posilion. Ie-.(2",loXIUl The LO ami TO branches coincide II K - O.

11"10 also " a t"OtUCqu~ of the crystalsymmelry cfCc'.

"11>e results "C'I"e nbtau>ed ....ith

ing by G, N:ll.soo IOd G. "'din

IlCtItrorI i ~ k"IItC'T-

We look fOl" a solution in the form or a tr-a\'c1ing waH', no\\' with dif1Crent ampliludes II, l' 011 alternate plancs:
..~ =
1/

cxp(isKa) exp(-iWl) ;

v~:: V

cxp(isK6)

t"~p(-iwt)

(19)

Recall the definition of a in Fig. 9 as thc distance between ncarest identical planes, not 11(."aTcst-ncighhor plancs,

"Hi

"._1

-".-.

".

"

...

1_ _

-o----J

.'igure 9 A diatomic Ct')"'lal stru<-'turc with mfl."CS M" M~ ",,"occtro by fOR'e con~tanl C betwccn

ailjao..cnt planes. The disl'l~mcnu of atoms M, arC dCllotcd by fl._I. II 11"" ... ,alld of aloms M. by "._1. V" v ,. The repeat di'tan~'C is tI in the dir.1;01l of the ",wc\'CCtor K. The aloms are
shown in their ur><lislll;\t.'eI! positions.

On substitution of (19) in (18) we have

-w2 M 11I = evIl

+ cxp(-iKa)]- 2Cu -ufMzv = Cu[cxP(iKt!) + l} - 2Cv .

(20)

The homogeneous linear equations have a solution only if the determinant of the cocflicicnts of the unknowns II, v vanishes:

2C - M j w2 -Cli + cxp(iKa)]
0'

(21)

\\le can solve this equation exactly for w2 hut it is simpler to examine the limiting cases Kn 4 1 and Ka = 7T at the zone boundary. For small Ka we have cos Ka 2: 1 - !K 2a2 + ... , and the two roots arc
w22!!

2C(_I- +
M.

_1_) !If'}.

(optical branch) (acoustical branch) .

(23)

w2

=='

-,.,~I~C'-,-,.- K 2a2
!If) +M2

(24)
is the repeat

The extent of the first Brillouin zone is - 'TrIa $ K $ distance of the lattice. At K,na>< ::: 'TrIa the roots are

mo., where

0.

uJl = 2CIM. ;

(25)

The dependence of w on K is shown in Fig. 7 for !If. > Al 2 The particle displacements in the transverse acoustical (TA) and transverse optical (fO) hranches are shown in Fig. 10. For the optical branch at K = 0 we find, on substitution of (23) in (20),
~= _ M 2 V M1

(26)

-I

l'hollOm I.

Cr~lol VibrolioM

,,,-;

Fi&ure 10 Transverw: optical and If2l"llO-

.........,.,.

...;r.u in a diatomic bnear bttlC't, illu,tnloo by th., partid., displacements for- die two modes al the same
,~.couJ.tical

11lc atoms vibrate against each other, but their center of mass is rued. If the two atoms carry opposite charges. as in Fig. 10, we may excite a motion of this type with the electric field of a light wave, so that the lxanch is called the Optical branch. At a general K the ratio !Jlv will he complex, a5 follows from either eX the equations (20). Another solution for the amplitude ratio at small K is u = v, obtained as the K = 0 limit of (24). The atoms (and their center of mass) move togclher, as in long ,...avclcngth acoustical vibrations, whence the term acoustical branch. Wavelike solutions do not exist for certain frcquencies, hcre between (2CIM ,)In and (2CIM.JII'l. This is a characteristic featurc of ela.~lic w<'!ves in polyatomic lattices. There is a frequency gap at the houndary K, ...... = 1rIa of the first Brillouin ..n ne. If we look for solutions ill the gnp with w real. then the wavevcctor K will hc complex, so that the wavc is dmnped in Sp..1CC.
QUANTIZATION OF ELASTtC WAVES

The cnergy of a lattice vibration is quantized. The qu.1ntulll of energy is called a phonon in analogy with the photon of the electromagnetic wave. Ela."tic waves in crystals arc made up of phonons. Thermal vibrations in crystals are thermally excited phonons, like the thermally excited photons of black-body electromagnetic radiation in a cavity. The energy of an clastic mode of angular fn.-'C!uency w is

= (n +i)lzw

(27)

when the mode is excited to quantum number II; that is, whl."tl the mode is occupied by n phonons. The term llzw is the zero point energy of the mode. It OCCUrs for both phonons and photons as a consequence ofthcir equivalence to a quantum harmonic oscillator of frequcncy w, for which the energy eigcn* values are also (n + !)Izw. The quantum theory of phooons is developed in Appendix C.

'08

,
We can easily quantize the mean square phonon amplitude. Consider the standing wave mode of amplitude
tJ :=

tJo cos K:r cos

wt .

Here II is thc displacement ofa volume element from its equilibrium position at
%

in the crystal.

The energy in the mode, as in any harmonic oscillatOf", is half kinetic cnergy and half potential energy, when averaged O\Icr time. n.e kinetic energy density is i/i..iJuJatf. where p is the mass density. In a crystal of volume V, the volume integral of the kinetic energy is lpVcJu~ 5in~ wi. The lime average kinetic energy is
'pvtJl'u~ =

!Cn + i)flw
+ i)hlpVw .

(28)

and the square of the amplitude is

u~ = 4("

(29)

This relates the displacement in a given mode to the phonon occupancy n oflhe mode. What is the sign of w? The equations of motion such as (2) arc equations for w2 , and if this is positive then w can have either sign, + or -. But the energy of a phonon must he positive. so it is conventional and suitable to view w as positivE". (For circularly polarized waves both signs are often used, to distingUish one sense of rotation from the other.) If the crystal structure is unstable, or becomes unstable through an unusual tcmpcraturedcpcndence ofthc force constants (Chapter 13), then tJf will be negative and w will be il1lagiru\ry. A mode with w imaginary will be unstable, at least if the real part of w is negathe. The cr)'stal will transform spontancously to a more stable structure. An optical mode with w close to zero is called a saO mode. and these arc often involved in phase transitious, as in ferroelectric crystals.

PHONON MOMEI'\TUM

A phonon of wavcvcctor K will interact with particlcs such as photons. neutrons, aud electrons as if it had a momentum 11K. However, a phonon docs not carry physical momentum. The reason that phonons on a lattice do not carry momentum is that a phOilon coordinate (except for K "" 0) involves rclath-e coordinates ofthe atoms. Thus in all H 2 molecule the internuclear vibrational coordinate r. - r2 is a relati....c coordinate and docs not carry linear momentum; the center of mass coordinate i(rl + rz) corresponds to thc uniform mode K"" 0 and can call')' linear momentum.

The physical momentum of a crystal is

p = M(dJdt) L u~
(30)

When the crystal eames a phonon K,

p = M(dIJdt)

L cxp(isKa) =

N-'

M(du/dt)(1 - cxP(iNKo)I/ II - cxp(iKa)l

(31)

where

funs over the N atoms. We have used the series

,-0

2: ,. = (I -

>!")/(1 - ,) .

(32)

In the next chapter we enumerate the discrete values of K compatihle with the.. boundar)' conditions. to find that K = 2'7TrlNll, where,. is an integer. Thus cxp(iNKo) = cxp(i27Tr) = 1, and from (31) it follows that the crystal
momentum is zero:

p = M(diJdt)

L cxp{isKa) =

0 .

(33)

lbe only exception to (33) is the unifOnn mooe K = O. for whicb all u" equal u, so that p = NM(duldt). This mode represents a unifOrm translation oftne crystal as a whole. and such a translation does carry momentum. All the same, for most practical purposes a phonon acts as if its momentum were 11K. sometimes called the crystal momentum. In Cf}'stals Ihert' exist W3vcvcctor selection rules for allowed transitoos between quantum states. We saw that the elastic scattering ofan x-ray photon by a crystal is governed by the wavevector selection rule

k' =k+G,

(34)

where G is a vector in the reciprocallatticc; k is the wavevector of the incident photon, and k' is the wavevector of the scattered photon. In the reflection proc'Css the crystal as a whole will recoil with momentum -fiG, but this uniform mode momentum is rarely considered explicitly. Equation (34) is an example of the rule that the total wavevcctor of interacting waves is consen'(.'(\ in a periodic lattice, with the possible addition of a reciprocallatticc vcctor G. The true momentum of the whole system always is rigorously conserved. If the scattering of the photon is inelastic, with the creation of a phonon of wavcvcctOf" K, then the wave\'cctor selection rule becomes

k'+K==k+C.
If a phonon K is absorbed in the Pr'OC'CSS, v.'C have instead the relation

(35)

k'=k+K+C.
Relations (35) and (36) are the natural cxtensions cL (34).

(36)

..
We exhibit the mathematics involved in the wavcvcctor selection rtJlc. Suppose two phonons K,. X2 interact through cubic terms in the clastic energy to create a third phonon 1<.3. The probability of the collision will involve the product of the three phonon wave amplitudes, summed over all lattice sites: (phonon K I in)(phonon K2 in)(phonon
~

out) c:c

"

"

This sum in the limit of a large number of lattice sites approaches zero unless ](3 = X. + K2 or K.J = K, + K 2 + C. If either of these conditions is satis6ed, of which the first is merely a special case of the second, the sum is equal to the number of lattice sites N. A similar sum was ronsidered in Problem 2.4.

INELASTIC SCATTERINC BY PIIONONS

Phonon dispersion relations c..o(K) are most often determined by the inclas~ tic scattering of neutrons with the emission or absorption ofa phonon. Further. the angular width of the scattered neulron beam gives information bearing on the lifetime of phonons. A neutron sees the crystal lattice chiefly by interaction with the nuclei of the atoms. The kinematics of the scattering of a neutron beam by a crystal lattice are described by the general wavevcctor selection rule: Ie+G=k':tK,
(37)

and by the requirement ofconservation of energy. Here K is the W3vcvcctor of the phonon ereated (+) or absorbed (-) in the process, and G is any reciprocal L1.ttice vector. For a phonon we choose G such that K lies in the first Brillouin 'lOne. The kinetie energy of the incident neutron is ,r12M.., where M.. is the mass of the neutron. The momentum p is given by Ak, where Ie is the w3vevcctor of the neutron. Thus f1 2/Cl/2M.. is the kinetic energy of the incident neutron. Ifk' is the wavevcctor of the scattered neutron, the energy of the scattered neutron is !l2k,212A1... The statement of conservation of energy is !l2k2 Il 2k'2 --=--Aw (38) 2M.. 2M" where Aw is the energy of the phonon created (+) or absorbed (-) in the process. To determine the dispersion relation using (37) and (38) it is necessary in the experiment to lind the energy gain or loss of the ~rauered neutrons as a function of the scattering direction k - k'. Results for { manium and KBr are

.,

PIto"o.lt I. Crydal vibrations

111

"001

",
,,

0 0
~

" ...........,
oTr..... _

,_S-' o'

"

'"
0 0 0

[1111

o'
0

.. .
0

.."

110 0

\\.....

~ _

in _ .. ia.

'"

F"ogure II Thedispl'fSioo cmvesofsodium b-phonoos ~ing in ll~ 10011. (1I0J. and [1111 d.1tOrls 90 K. ddermincd by inebstic scattering of neutroru., by Woods, BrockhouSt. Mardr lind Booo.es.

Figure 12 A ITillle axis neutron spectrometer al

Broo~h"'en.

(Cnt,,'leS)' of B. H. Gricr.)

given ill Fig. 8; results for sodium are given in Fig. II. A spectrometer used for phonon studies is shown in Fig. 12. Recently the conccpt of "mirror symmetry" has heen introduccd into the study of alkali-halide ion dynamics. 2 The idea is to consider the crystal that would be fonned if the signs of the ions in A+B- were reversed. Now A-B+ docs not ex-ist, hut a nearest mass pair that docs exist will have. it turns out experimentally. a phonon dispersion relation remarkably similar to that of A ~B-. Thus KF is the approximate mirror image of Nael. for K+ is isoclcctronic with CI- and F- is isoelcctronic with Na+.
11.. 1.. FokI,' and P "'egall, Ph)",. &V. B25, 1200 (l96'2h 1.. 1.. FoId~ and T. A. \\'ltlCfl, Solid State Commufl. ';n. 7( lBl).

'"
SUMMARY

The quantum unit of a crystal vibration is a phonon. If the angular frequency is w. the energy of the phonon is liw.

When a phonon of wavevector K is created by the inelastic scattering of a photon or neulron from wavcvectOf" k to k'. the waVC\ector selection rule thai governs the process is

k=k'+K+G
where C is a reciproc:allattice vector. All elastic waves can be clt"'SCribcd b)' wavcvectors thai lie within the first Brillouin zone in reciprocal space.
If there arc p atoms in the primitive cell, thc phonon dispersion relation will have 3 3l.'Oustical phonon branches and 3p - 3 optical phonon branches.

Problems
I. Mmwlomic linMr Jolliee. Consider a longitudinal ....'aVe
u. "" u ca;(fat - sKo)

willeh propagates in a monatomic linear lattice of atoms of mass AI, spacing a. and llearest-neighlxlr illternction C. (a) Show Ula! t},e total energy of tJle wave is

where' nms over all atoms.

(1)) By slll>5titlltion of 1<, ill Ihis expression, sllow that the time-average total cner'l,'Y per atom is

wllere in the I.a:l;t step we have used the dispeJ"$iocl relation (9) for this problem.
2. Conlim.'l<m wave Cf/I<o'ioo. Show that fur long \V3."e1engths the e(luation of motion (2) reduces to the continuum elastic wa,'e e(luution

....'here v is the velocity of sound.

3_ Basis a/tlco umikeo'olM. For the problem treatal by (18) 10 (26), find the amplitude ratios wv for the two branches at K.._ - '11la. show tI'M at this value of II. tile two lattices act as if d(."('()upled: Olle bttia remains at rest wllile the odler lattice moves.

4. Kohn GrIonudy. We suppose that die interplanar ~ constant C~ between planes 5 and 5 + P is of the fOrm
C ,,"A sin

Pkoa

".

where A aod ko are constants and p runs over all integers. Sud'I a form is expected in metals. Usc tllis and Ell. (16a) to find an expression for c.l and also for aW'/ilK. Prove that iJe-.!laK is infinite when K .. 4 Thus a plot of J versus K or of w versus K has a vertical tangent at ko: there is a kink at k() in tile phonon JiSpersioo relation w(K).

5. Diatomic chain. Consider the nonnal modes of a linear chain in whicll the force constants between nearest~neighboratoms are Illtemately C and lOCo Lei the n1ll!;$eS be C<lual, and lei the nearest-neighbor separation be 412. Find w(K) at K - 0 and
K '" r./a. Sketch in the dispenion relation by e,'c. TIlis problem simulates a crystal of diatomic molecules such as Ill!.. 6. Atom: cibrationl in a metal. Consider point ions ITIaS:!i M and charge III immersed in unilorm 5Ca or conduction electrons. TIle ions are imagined 10 be ill stable equilibrium when at regular lattice points. Lr one ion is displacW a sman distanoe r rom its e<luilibrium position. the restoring force is largely due to the electric charge within the spl,ere of rarJius r centered at the equilibrium position. Take the numbet" densit), orions (or of oond..ction electrons) as 3f4nR', which defines R. (a) Show that the frequency of a single ion set into O$Cillation is w - (f!I IMR")IIt. (b) Estimate the value of this frequeflC)' for sodium, roughly. (e) From (a). (b). and some common sense. estimate ti,e orner of magnitude of the velocit) of sound in tile metal. --T. Soft phonon modes. Consider a line of ions of e<j,La1 mass but alternating in cI,arge. witJl cp '" e(-l)" as the cllarge on the pth ion. 111C interatomic potential is the Slim of two contributions: (I) a short~range interaction of force constant C,a - l' that acts between nearest neighhors only, and (2) a coulomb interaction betWtlen all ions. (a) SI,ow lIlal the contribution of the coulomb interaction to the atomic force constanh is C"c - 2(-1)" c 2/ p 3dJ, where II is the cljuilibrillm nearest.neighbor llistance. (b) From (IGa) show that ti,e dispersion relation may be written as

or

..1I~ -

sinl iKa + u

,-, (-If (I -

2:

cos pKa)p-3 ,

where ~. 4yM and q - t!'/-yciJ. (e) Show tlJat J is negath'e (nnnable mode) at the 7.()IIe boundary Ka., n if u > 0.475 or 411{(3) ..... here' is a Riemann )';eta function. Show further Illat the speed of sound at sman Ka is imaginary if q > (2 In 2)-1 _ 0.721. TIllis Wi goes to zero and IlIC lattice is unsl1lhle Q some \'8lne of Kg in 1l1l;~ inten'al (0.11') if 0.475 < q < 0.721. Notice that ti,e phonon spectrorn is not that of a diatomic lattice because the interaction of any ion wjtJI ibi neighbors is tile same as that of an)' other ion.

--

II'

References
II. nih and W. ~ Phonon dispenlon rdot~ '" 1nn<11l'oN, Springorr, 1979. An atlas.. G. A. Horton and A. A. M.,..Iudin, eds.., ~ pnlJlffl>a oflOlidr. North-Ilolbnd. A continuing Oftit's 1974-(1990}. W. P. ~1&5On, ed, ocowtic:r. A.cademic Preu. A coniinuing series, .-ok. 1-(20). w. Cocbnm, DII"mllla 0{ atamr in CJ1PfDU, Cra.... Russak, N.Y., 1913. A good led. J. A. newbnd. Physia of p#wJnoo18. Wiley, 1973. II. Bottger. I'~ of 1M thwy of "'nice Jynomk:r. Wtlnhe,m: Ph)'Sik \'e<"bg. 1983. P. Broach. PhtmolV, ,hJry Gnd ~u. Spri~. 19!:17

""!pm

5
Phonons II. Thermal Properties
PIIQKON HEAT CAPACITY Planck distribution Normal mode enumeration Demity of stales in one dimension Demity of slales in three dimensions Ileb)"e model for density of slales Debyc T 3 law Einstein model of the density cl states General result for D(w)
ANllAR~IONIC

117 117
118

119 120 122 124 124

127

CRYSTAL II''TERACTIONS Thermal expansion

129 130
131 133
135 137

THERMAL CONDUCfIVITY Thermal resistivity of phonon gas Umklapp PI;'OCeS~S lmpclfct:tions

PROBLEMS
I. Singularity in density of slales 2. Rnu thermal dilation of crystal cdl 3. Zero point lattice disiliaecmcllt and strain 4. Ileal capacity of la}-er lattice

130
139 139 139 139
139

5. Crunciscn ronslllni

nEFERENCES

140


<.>
3

/'/
,/,--

,
0

, /

,,

'/

----,' /

/ /,

/ /

f"lgIIre I floc of Pbncl< diJlribulioo runctioo. At hi$:h lempn2tu1"e:5 the occul*'CY of:l ~e is approo.'malel) 1 . - in the temper.>lure.- TIle lUnctioo (n) + t ......hich is not plotted. apprwches the da.shed line as a.srmplote at high lempcnllur~. 11,., dashed line is the dusi<:allimil.

'"

CtiAPTEf

PHONONS II. THERMAL PROPERTIES

We discu~s the heat capacity of a phonon gas and then tile effects of anhannonic (aUke interactions on the phonons and on the crystal. Thermal properties of metals are treated in Cilapter 6, superconductors in Cilapter 12, magnetic materials in Chapters 14 and 15, and noncrystalline solids in Chapter 17.

PHONON HEAT CAPACITY

By heat capacity we shall usually mean the heat capacity at constant volume, which is more fundamental than the heat capacity at constant pressure, which is what the experimenh determine. 1 The heat capacity at constant volume is defined as C" 55 (aUla1')\,where U is the energy and T the temperature. The contribution of the phonons to the heat capacity of a crystal is called the lattice heat capacity and is denoted hy C~"" The total energy ofthe phonons at a temperature 'l(E kBT) in a crystal may be written as the sum of the energies over all phonon modes, here indexed by the wavevector K and IXJIari....ation index 1': (I) where (nK.,,) is tile tllermal equilibrium occupancy of phonons of wavevector K and polarization p. The form of (nK,,) is given by the Planck distribution function:
(") ~

--;:--':-:-----:exp(Ii"'7)
1

(')

where tile (...) denotes the average in themlal eqUilibrium. A graph of (n) is given in Fig. ].

pltlflck Distribution

Consider a set of identical hannonic oscillators in thermal equilibrium. The ratio of the number ofoscillators in their (n + l)th quantum state of excitation to the number in the nth quantum ~tate is
(3)

'A thermodynamic reblion gi\'es C.. - C,. '" 9o"B\'1', whe .... a is Ihe lemp<:ralurc eodIicit;nl of lh,e:u" expansion, \' lh" 'ol"me, a"d B the bulk modulus. 1be fractional dilfcrt:nce betwctm C,. and C\' is nsually small and oft,,'fl may be n"'gk-ct,,,,1. As T --. 0 we sel;" lhal C........ C," pro,'id",,! a and B are comtanl.
H7

"'
by use of the Boltzmann faclor. 111m the fraction oCthe total number of oscillaIOrs in the nth quantum state is
----~

N.

cxp(-nliwlT)

(4)

exp(-sliwl1')

\Ve sec that the a\'cngc excitation quantum number of an oscillator is

L s exp(-s/ftd'T)
(n>
~ ~''---

L cxP(-sliwlT)

The summations in (5) are

1 2:%'=-, 1- x

(5)

(6)

with

= cxp(-liwl7). Thus we may rewrite (5) as the Planck distribution:

x
Normal Mode Enumeration

(n) = -\---x = C.-': ,,)----'1 T x,"'>(h,w/7

(7)

"Inc energy of a collection of oscillators of frequencies equilibrium is found from (1) and (2):

wKp

in thermal

u ~ 2: 2:
I<.
J>

hWK.,

(8)

cxp(liwl<.JiJ - I

It is usually convenient to replace tllc summation over K by an integral. Suppose that the crystal has D,,(w)dw modes of a given polarization p in the frequency runge w to w + dw. Then the energy is

u=

J
p

dw D';'w) CXP(/in;ij

(9)

The lattice heat capacity is found by differentiation with respect to temperature. Let x = liwh = IiwlkBT; thcn aU/aT gi\'CS

Cbt = k B L

'" f

dw Dr,(w) (

Xi exp x I)' exp x

(10)

The central problcm is to find D(w). the nUIll~ of modes per unit frequency range. This function is called thc dcnsity of C ies or. more often.

S 1'Jv:-:,m If. TNnnaJ Propntia

I ....

,"0 I l1. ,"10 Fogure 2 Elasti<" line or N + I atom., ",ith N - 10. for boundary cOlld.lioN that the end atonas t - 0 and t - 10 are fUed. n,.., putode dis~ in the normal mo<ks lOr riIhu Ioogiludinai 01" m..w.'ll'tSe l&p~rno:nts ar", of the f"onn. ta." sin sKa. TIns ~ is alltomalicall) zero at the atom al the "'00 - 0, and "'"e ~ K to rn.U... the dispba.-emc:nl t"'ro .1 I"'" md. _ 10.

F~_l

. . . :l~ . r-1--1-.~
.

o .....!.....!!-

... ,..
.

... ,.

FIgUre 3 t:-ndary t"OIIdition sin tKa - 0 lOr - 10 Qlfl be salisfiaxl by ~u.,; K.10., bflOcr.... 9m'10.. ",here lOll ., tt>... length L d tt>... line. "J1Mo praenl figure is in K space. The doIs an: not alomi but are the a1'--ed ,.w..s d K. or lhe N + 1 partides OIl 11M. line. OIlly N - I ar", aIJc,..ed 10 JllO\"e, and !heir most gcraaI moIion <Sn be cqwa~ in !eons of I"'" N - I a\So.r.-ed ,-.lues or K. This quantization d K '-s nothing to do ",i1h quarllurn mox:hanics bt>1 1OIJa,.'S dowically from the boondary t"OIIditions IhaI: the .:nd atOlTlS be rolled.

n...

density of states. The best practical way to obtain the densily of states is to measure the dispersion relation w versus K in selected crystal directions by inelastic neutron scattering and thcn to make a theoretical analytic fit to gi,'e the dispersion relation in a general direction, from which D(w) may bc ca1cu~ !ated.

Density of States in One Dimension

Consider the boundary value problem for vibrations of a one-dimensional line (Fig. 2) of length L carrying IV + 1 particles at separation tl. We suppose that the particles s = 0 and s = N at the ends of the line arc held fixed. Each nonnal vibrational mode of polarization ,J has the form of a standing wave, where II. is the (lisplacemcnt of thc particle s:

u. = u(O) exp(-fwK,,t) sin sKa ,

(11)

where WK." is related 10 K by the appropriatc dispersion relation. As in Fig. 3, the W3vevcctor K is restricted by the fixed-end boundary COnditions to the values
K=L'
~ 2~

311" L ,....

(N - 1)11" L

(12)

The solution for K = 7rlL has

u. sin (S1TalL)

(13)

and vanishes ftw s = 0 and s = N as required. The solution for K = N'TTIL = ma = K......~ has u. sin S'tt; this permits no motion of any atom, because sin 817' \1lJ1ishes at each atom. 111us there are N - I allowed inclP'lCndent values of K in (12). nlis number is equal to the

.
number of particles allowed to move. Each allo,"'ed value orK is associated with a standing wave. For the one-dimensional line there is one mooe for each interval 6.K == ?rIL, so that the number of modes per unit range of K is Urr for K:!i mo, and 0 for K> 71'0. There are three polarizations p for COlen value of K: in one dimension two of these are transverse and one longitudinal. In three dimensions the polarizations arc this simple only for wavevectors in certain special crystal directions. Another dc\"ice for enumerating modes is often used that is equally valid. \Ve consider the medium as unbounded. bul require thai the solutions be periodic over a large distance L, so that u(sa) = (sa + L). The method of peri. odic boundary conditions (Figs. 4 and 5) docs not change the physics of the problem in any essential respect for a large ~1'stem. In the running wave solution u. = u(O) exp[i(sKa - WKI)] the allowed ,'alues of K are K=O,
+ + -T' -T'

2'17

4'17

~L'

6'17

N7r L

(14)

This method of enumcration givcs the same number of modes (onc per mobile atom) as given b) (12), but we ha"e now both plus and minus values of K, with the intcrvall1K = 27TfL bctwecn successivc values of K. For periodic boundary <.'Onditions the number of modes per unit range of K is U2'17 for -'I7/tl :5 K :5 'I7la, and 0 otherwisc. Thc situation in a two-dimensional lattice is portrared in Fig. 6. \\'c nced to know D(w), thc numbcr of modes per unit frequcnt'}. range_ The number of modes D(w) dw in dw at w is given in one dimcnsion by
L dK L dw D(w)dw=--dw=---- . '17 dw '17 ckIdK
(15)

We can obtain the group velocity dwldK from the dispersion relaliOfl w,'crsus K. There is a singularity in D(w) whencver the dispcrsion relation w(K) is hori:r.ontal; that is, whencver thc group velocity is zero,

Denrily of StalQ ill Three Dimnlrions

We apply periodic boundary conditions over N 3 primitive cells ",ithin a cube of sidc L. so IIlal K is dctcnnincd by the condition exp[i(K,.x + whence
-

KvJ +

K,:Z cl'p{i[K.(r

+ L) + KJ.y + L) + K..{z + LlJ}

'
~

(16)

+-.

2w
L

4w
L

Nw
L

(17)

Thercfore there is one allowed ,'3.I.ue of K per volume (2TTfL'f in K space, or

(2~r = 8~

(18)

S Phol'lQtU II. T'--'I Propnfia

Fcure 4 CulSilkr N particles ron.'I'tnirre<J to slide un aarcuw- ring. n.., partidits can otc:ilble lrC'OnII<.'<:ted by eWtic springs. In a normal mode: lhe dispbc.-rmn'11 ..,of alom J' ..iU be or~ fOrm sin .KilI 01" ~ sK4. Ihne llTIt indcpo-'olltont modes. B)' lhe gc<lmIttrical periodictt) of the ring the boundarv condition ill that .." - ..,lOr all., SO that NIVl must be an inlegral m~ltiple or 2... For N - tl the aJlowe.l independenl values or K are 0, 2..I8d, 471'1&, 6../&, and 8m'8<l. TIll' '"aIUlt K - 0 i. meaningless ror the .ine fOrm, becaw:e sin .oc. .. O. ""e ,-.Jut." 8../& hal a meaning onl)' lOr lhe rosine rom" in (.I'8..aI&) .. lin $7T _ O. TIll' three other \"alucs or K are allowed lOr both the sine and COline mudes, giving a lotal or eight .lkoweo.I model ror Ihe eight particles. TI",. the perindic boundary rornlilion leads 10 one .llov,ed mode pt;r partide. eXllCtlyas ror Ihe l"ix(.,<!(.'Ild boundary condition or Fig. 3. H we hall t.ken Ihe modex in the <."(1m plex lOr'" exp(/sKa). the periodic boundary conditio" woukJlead to lhe (.;ghl moo(.'S with K - 0, -:r.2mNa, %.4 ..fNa, ~&rrfNa, and 8mNa, as in Eq. (14).

121

.-.
.. ,

.-,

.-,

boca,_

". ,. --.:

.
,

tt

;.

,,. . ,.

tt

.,

KF~re 5 "lJoy,1:d ,.,.]ua of ..;a,~,ecIor K lOr- periodOc boundary conditions applied to a linear Iauicc of pcriDdicil)' N .. 8 atoms On a line of length L. The K - 0 solution iI the unibrm mode. The special points -:r.NmL ~I only a singk l<lIutioo hecauM: exp(in) is identic:al 10 exp( -1m): dlUS there ~ eight ano..1:d modes...itJ, drs~mentsof the .rlh alOm proportional 10 I, up(-:r.ltr.4). exp(-:r.lmI2), e:xP(-:r.i3mI'4), exp(lm).

F~re 6 Allowed valurs in FOlLI'ier spaceo(~ phonon ..;a,~"tt'Ior K a squan" Iauiee oflalhelt comtant CI, ..ilh periodic boundary ron.Jitiom appIi~ o\u" a square of side L" 10.. "'IC unibm mode is rnarb.od ..-ith a cross ..,__ is One aBo-d _-alue of K per am (2mI00)" - ({J:"'I.'!, so that "ithin the cirdc of area ..K!' the Imootbal number of aJp'Cd points is ~w'f.

ror

..
allowed "alues of K per unit volume of K space. for each polarization and for each branch. The volume of the specimen is V = L3 . The total number of modes with \\13\'Cvcctor less than K is fOund from (18) to be (I.J2rr'jJ times the volume of a sphere of radius K. TIlUS
N = (I.J21T)3(41TK 3/3)

(19)

for each polarization type. The density of slales for each polarization is
D(w) = dNldw = (VK 212tr')(dKldw)
(20)

Debye Model for Demit!! of Stores

In the Dcbyc approximation the velocity of sound is taken as constant for each polari7.alion type. as it would be for a classical clastic continuum. The
dispersion relation is written as
w== vK

(21)

wilh v the constant velocity of sound. 10e density of states (20) becomes
D(w) = V,.rI2r,2 0 3
(22)

If there are N primitive cells in the specimen, the lotal number of acoustic phonon modes is N. A cutoff frequency Wv is dctcnnined by (19) as
(23)

To this frequency there corresponds a cutoff WlWe\'ector in K space;

Kv = w,jv = (67tZNN)113 .

(24)

On the Debye model we do not allow modes of ....oa\'e\'edor larger than Kf). The number of modes with K S K v exhausts the number of degrees ofrreedom of a monatomic lattice. The thermal energy (9) is given by

U:

dw D(wXn(whw -

LW" dw ( V~',) (,~.w) o 211'0 e -1

\25)
(26)

ror each polarizaHon type. For bre\'ity we assume that the phonon velocity is independent of the polarization, so that we multipl) by the factor 3 to obtain
_ 3V" U - 2r03 0

L"'b dw e"""r ~

_ 3Vk,!T~ L"'' x 1 - 2rv3A3 0 dx e. _ I '

3

where x hwlT

;00

hwlksT and

Xv - hwdkBT - ()IT (27) This delines the Deb)'e temperature 8 in terms of Wv delined by (23). We may express 8 as

8=

~. (&n'N)'"
ks
V

(28)

5 Phonmu II. Thnnwl f'ropertie.

123

, ,
FIgUre 1 lleat "apa<.~ty C,. of a SQlid, according to the Dehe approximation. The \"ertical scale is in J 1001-' K-'. The horizontal S<::Ile is the temperature IIormaliZl.od k> the Debyc lem!X'rat"rc 6. 'The m gioll of the Till'" is l>clow 0,16. TI,e asymptotic ,..Iue at high ,a1"cs ofTf6 is 24.943 J mol-' <leg-'.

I
I I
0,2 0.4 0.6

'J ,
0

.,-

0.8

1,0

,.,

IA

16

, ,
,

C,rm.",ium

l
SilK.-on

lIeat capacit)' of s~icon aoo gcnnaIlium. Note th" dt.'<"I"ease at low temperatu...... TOl'Olwcrt a "alue in caVmol-K to J/mol-K. multiply by 4.186.

.'Ig""" 8

/ / // : , //
'00

, ,

/
/

------

I
I

I
I

so that the total phonon energy is

U = 9Nk T(T-)' B 8

LXV dx ~ 0 eX_I

(29)

",..here N is the number of atoms in the specimen and Xv = BIT. The heat capacity is found most easily by differentiating the middle expressiOn of (26) with respect to temperature. Then
C '"
v

I"b dw 27f-ti'kBT2 0
3Vh
2

W..!

e"wlT
If

(e"wlT

= 9Nk

BOO

(T)' LX" dx

x eX (eX _ 1)2
(30)

The Ocbye heat capacity is plotted in Fig. 7. At T J> 0 the heat capacity approaches the classical value of 3Nk B Measured values for silk'On and germaig. 8. nium are plotted i

..
Debye T 3 LAw
At very low temperatures we may approximate (29) by letting the upper limit go to infinity. We have

" L

'"' - ' - =

tr - 1

, L"
5!5

dx

%3

_I

- exp(-n) = 6 L 1' L 4" = ~ 15

I 8

(31)

where the sum over ,-. is found in standard tables. Thus U a 311"'Nk sT 415li'fOl" 'f <t 8. and

Cv

l2n' Nk -5s

(T)' 8

ill

234 Nk s

(T)' "8

(32)

which is the Debye '[3 approximation. Experimental results for argon are plotted in Fig. 9. At sufficiently low temperatures the '[3 approximation is quite good; that is, when only long wavelalgth aCOtlst.ic modes are thennally excited. These are just the modes that may be treated as an elastic continuum with macroscopic elastic constants. The energy of the short wavelength modes (for which this approximation fails) is too high for them to be populated significantly at low temperatures. We understand the '[3 result by a simple argument (Fig. 10). Only those lattice modes having flw < ksT will be excited to any appreciable extent at a low temperature T. The excitation these modes will be approximately classical, each with an energy close to kBT, according to Fig. 1. of the allowed volume in K space, the fraction occupied by the excited modes is of the order of (t.wo'f or (K.,IKo'f. where K r is a "thermal" waveveetor defmed such that frvK r "'" ksT and K o is the Debye cutoff Yo'a\'cvector. Thus the fraction occupied is (TIO'f cl the total volume in K space.. l"ere are cl the order of 3N(TIO'f excited modes, each having energy ksT. The energy is -3Nk s T('flO'jJ, and the heat capacity is -l'};Nk s (TIO'j3. For actual crystals the temperatures at v.nich the T 3 approximation holds are quite low. 1t may be necessary to be beklw T::It G'5O to get reasonably pure T 3 behavior. Selected values of 8 are given in Table 1. Note. for example, in the alkali metals that the heavier atoms have the lowest (/s, because the velocity ofsound decreases as the density increases.

or

Eimtein Madel of the Demity of Sfatn

Consider N oscillators of the same frequency % and in one dimension. The Einstein density of slales is D(w) = N&"w - %), where the delta function is centered at Woo l"e thermal energy of the system is
U = N(n)/iw =

N"'" l ' e".J'r

(33)

Plumo,.. II. Thennal Propertie.

)25

11.18

/
./

, ,

"0 13.33

..

4,44

,,

\i"

V
'"
,go
7"',
~el,'in>

'"

'"

'"

'.00

Figure 9 Low kmperaturc h""t C"dpacit)" of solid argon, plnued against 1'~. In lhis lemperature rt,gion the <:>:perimental re."It< a~ in .,xedlenl agr...... ment with thc Deb)"c T"law"ill, 9 '" 92.0 K. (CourtcSl' of L. Finegold and N. E. Phillip. )

fl<.I.,. h~K". ks6 ht:K r ksT

Figllre 10 To oLtain a qualitali"c explanalion of the DeLyc T 3 law, WC sOllflOSC lhat all phonon modesof "",...,welor b Ihan K,h,,,"c the dassicaJ thermal efl<.'1"gl' k8Tand ll,al mode.l>ctw~"Cn K.... and the DeLye cutoff Ko are not excited 3t all. Of Ihe 3N possible modes, lhe rrdction exciled is (K-,lKDt' .. (T/Uf'. l>t:cause lhis is the ratio of lhe "olume of the inner sphere 10 ll,e oul.,r sph.,rc. The energy is U ... kilT' 3N(TI9'f', and Ihe heal capacily is C., .. aU/aT'" 12Xk8(T/9)'.

e.
2.00

Table 1 Debye temperature and thermal conductivity"

344- '<40
0.85

- - e
C N
0.27
22301.29

N.
75

N.
158

Mg

1,41

_400::
1.56

-So
0.16

._;r.o~v-t~P.~~tofll, in Kelvin Thermal conductiVity-at 300 K. in W cm-IK-I

--

AI
428 2.37

SI

645~

1.48

CI

A,

S.
90

92

K
91

C.

Ti
0.22

V
3BO

C,
0.94

Mo
0,08

F.

Co

NI
'50

Co
343 4.01

'0
327 1.16

G,

G.

A,
282 0.50

e,

K,
72

230...... 360-: 1 420"

630"": "41O"'"

1.02

0.31

470- 445 0,80 1.00

0.91

320 - 374 0,41 0.60

002

Rb

S,

56: 0,58
C,
~3:

14.:.

2_80. 0.17 0.23

" 291

Nb
2'5 .,.;.

Mo

To
,

Ro
-600~

Rh
480

Pd
27. 0.72

Ag
225
4.29

Cd
2Q9~

'0
lOS
0.82

So
200

Sb
211
0,24

T.
153

]50
1.38

a.

-,

X.
64

0.54

0,51

1.17

1.50

0.97

0.67

0.02

e,

c.p Hf
0,14 0.23

T,
0.58

R.
... 4~O 0.48

0'

PI

Ao
165
3.17

Hg
7(9

TI
18.5 0,46

Pb
105
0.35

el
119
0.08

Po

Al

Ro

_____

110 _ ~14;_

0.36

lsi'" ~ 1:00

soo
0,88

"

420

1.74

1.47

240' 0.72

Yb
120

lo
2,10 0.16 0.35

F'

R,

Ao

"

C.
0.11

P,
0.12

Nd
0.16

Pm

- - Th
l~.

Sm
0.13

Eo

Gd
200

Tb
0.11

0, 210
0.11

Ho
0.16

E,
0,14

0.11

Tm
0.17

P,

0.54

.lJl7_
0.28

0.06

- - -0.07
~

N,

Po

Am

Cm

ek
-~
~

Cf
r
~

E,
'"""",-

- .-

Fm Md -..,.,.....

No
w

L'

~to~t of the 6 values were ~upplicd by N, Pearlman. referen~s an' lZivcll thc A I,P, Hnm/lXlOk. 3ro etl. thc tberlll..l t'oIlJntti, ,ty V.,lIll' ...TII l"'lll II. W Powcll "'\Ill Y. S. ToulouK'J.n, !>clence 181. 999 (1973),

ph~

11, T'-mal E'rfllJUfin

IZT

,
,

,.

- --

~-

, , ,
/
o
0.1

,.
./

.1/
0.3
o.~

0.2

05

06

0.1

o.tI

011

10

n"
Figure II Comparison of .,xpt:rimcntal.....Jue. of Ihe h~t ca\*:ity of dilltnOnd ",Ih Vll!ue. calc... Iak.JOflthe L'nrhe.1 qu;lllium (EinsteIn) motlc:l, 'lloing tlw chillllcterislic lemperature 81'. - hlAlll<" 1320 K. To COllVer! III JllllOI-dc.og, multiply by 4,186.

with w now written in place of WOo for convenience. The he:lt capacity of the oscillators is
Cv =

aU) v = Nk" (~w)' <1-" (aT -:;:- (1-" _ L)2

(34)

as plotted. in Fig. II, This expresses the Einstein (1907) result for the contribution of N identical oscillators to the heat capacity of a solid, In three dimensions N is replaced by3N. The high temperature limit ofC v becomes 3Nk". which is known as the Dulong and Petit value, At low temperntures (34) decreases as exp(-Ii~T). whereas the experi. mental form of the phonon contribution is known to be T 3 as 3CCOlInted for by the Debye model treated above, The Einstein model is often used to approxi. mate the optical phonon part of the phonon spectrum,

Ceneral Result for D( w) We want to find a general expression for O(w), the number of states per unit frequency range. given the phonon dispersion relation <t.(K), The number of allo.....ed values of K for which the phonon frequency is heN'een wand w + cluJ is D(w) dw =

(~)'J
21f

d'K.

(35)

.hell

where the integral is extended over the volume of the shell in K space bounded by the 1:\\'0 surfaces on which the phonon frequency is constant, one surface on which the frequency is wand the other on which the frequency is w + dw. The real problem is to evaluate the volume of this shell, We let tIS.. denote an element of area (Fig. 12) on the surface in K space of the selected constant

...

FiI;un 12 Ekmrnl c# area 115.. 00 constant &c!querq ~1rliM.-e In K 'P"C"'. 'I1lc \'OIume be1",_.. two surfacrs of constant lrltqu<ency at <II

IlIId

<II

+ d", 15 ~"al

to

f dS.. d/ollfV~,

frequency w. The element of volume between the constant frequency surfaces wand w + dw is a right cylinder of base dS", and altitude dK.L. so that

J ....

d 3K =

dS..dK.L .

(36)

Here dK.L is the perpendicular distance (Fig. 13) betv.<een the surface w constant and the surfuce w + dw constant. The value of dK.L will vary from one point to another on the surface. The gradient of w. wIlich is V..w, is also nonnal to the surface w constant, and the quantity

IV"w! dKJ.

dw

is the difference in frequency bern-ecn the two surfaces connected by dKJ.._ 11lUs the element of the volume is

dS,- dKJ. = dS...

IV.. w1

dw

dw = dS", vI:

where vlt = IV..w1 is the magnitude ofthe group velocity of a phonon. For (35)
we have

D(w) dw

~ (!e-)'J
217
= L3

<IS. dw

v,

We divide both sides by dw :IIld write V result for the density of stales is

for the volume of the crystal: the

v D(w)=-(211'")3

JdS. -.
vII:

(37)

5 " " , - II. TMnrwd P~.

l29

Figure 13 The 'luantlly dK, is the pcrpelldit1.lLar di5t~t>ce be-

I......." '" 1"'0 <:Ollllallt fn:'luellcy 5urfa~'1 ill K spa;, olle at freo 'IUC<tCy I and the other at f't'<llrellt")' '" + d",.

Su.-b<:e '" '"' """,tolll

IJ(w)

IJ(w)

(a)
siTubr points.

(h)

r-optre 140 Dr:n1it} rI stale. :as a. function rI &equrncy b (.., the ~'e solid and (b) an actual Ct)'m.! structure. ~ spedrnm ...... theCf}'Stalstartli as.Jror small .... but discontinuities de.-dop"I

The integral is takl.ll over the area of the surface w constant, in K space. The result refers to a single branch of the dhversion relation. We can use this result also in electron band theory. l'nere is a special interest in the contribution to D(w) from points at which the group velocity is zero. Such critical points produce singularities (known as Van Hove singularities) in the distribution function (Fig. 14).

\
ANIIARMOJ'\IC CRYSTAL IJ'\TERACTIONS

The theory of lattice vibrations discussed tlluS far has been limited in the potential energy to terms quadratic ill the interatomic displacements. This is the harmonic theor)"; among its consequences are: T\\'O lattice waves do not interact; a single wave does not decay or change form with time. 1'Oere is no thermal expansion. AdiabaHc and isothermal elastic constants are equal. The elastic constants are indt..'J)eT'Ident of pressure and temperature. o The heat capacity becomes constant at high temperatures T> 8.

130

In rcal crystals nOne of these consequences is satisfied a(,'Curately. The deviations may be attributed to the neglect of anharmonic (higher than quadratic) terms in the interatomic displacements. We discuss some ofthe simpler aspects of anharmonic effects. Beautiful demonstrations ofanharmonic effccts are the experiments on the interaction of two phonons to produce a third phonon at a frequency w;; ;: WI +~. Shiren described an experiment in which a beam of longitudinal phonons of frequency 9.20 Cllz interacts in an MgO cI)'stal with a parallel beam of longitudinal phonons at 9.18 Cllz. The interaction the two beams produced a third beam of longitudinal phollons at 9.20 + 9.18 = 18.38 GHz. Three-phonon processes are roused by third-order terms in tbe lattice po~ntial energy. A typic-oil term might be U3 = Ae...ewec, where the e's are stram components and A is a constant. The A's have the same dimensions as elastic stiffness constants but may have values perhaps an order of magnitude larger. The physics of the phonon interaction can be stated simply; the presence of One phonon cuuses a periodic clastic strain which (through the anharmonic interaction) modulates in space and time the elastic constant of the crystul. A second phonon perceives the mooulution of the e1ustic constant and thereupon is scuttered 10 produce a third phonon, just as from a moving threedimensional grating.

or

Thermal Expansion
We may understand thermal expansion by considering for a classical oscillator the effect of anharmonic tenns in the potential energy on the mean separnliOn of a pair of atoms at a temperature T. \\'e take the potential energy of the atoms at a displacement x from their equilibrium sepamtion at absolute zero as
(38)

with c, g, and f all positive. lbe term in ~? represents the asymmetry of the mutual repulsion of theatoms and the tenn in x4 represents the softening of the vibration at large amplitudes. 111e minimum at x = 0 is not an absolute minimum, but for small Oscillations the form is an adequate representation of an interatomic potential. We calculate the average displacement by using the Boltzmann distribution function, which weights the poSSible values of x according to their thennodynamic probability:

(r) ~ -'=;--'0-----

f" d:c

x exp[-I3U(x)]

L:

dx exp[ -I3U(x)]

1'''-oN II. T'--'l Prcrprrlia

'"

.., I ". .,.

'j
~

'"

T....

...,11
,;

,,,,,---.,.--rl&'1f'e IS J...atlft constant oholid argon a. a lU..,..

lion of tempnlu""

. .. .. .
T~.oturr.

/
K

-'

.I

'" ,,,

,
Q

u.

'" ,...

with p. UkoT. For displacements such that the anharmonic terms in the en~ ergy are small in comparison wilh kaT, we may expand the integrands as

f dx x ..p( -{JU) '" f dx [e.p( -Ilcr~l(x + pgx' + /3fi/') ~ (3"P'/4Y.1!Jc""1P'''' , f dx e,",,-pU) '" f dx e.p(-Ilcr~: ("'flc)'~ (39)
whence the thermal expansion is
(40)

in the classical region. Note that in (39) we have left 0: 2 in the exponential, but we have expanded exP(Pb>x 3 + f3ft4) a 1 + pgx3 + 13ft + .. " Measurements of the lattice constant of solid argon are shown in Fig. 15. The slope of the curve is proportional to the thermal expansion coefficient. lne expansion coefficient vanishes a5 T - 0, as we expect from Problem 5. In lowest order the thermal expansion does not involve the symmetric term fx4 in U(x), but only the anlisymmetric term gx3.
THERMAL CONDVCnVI1Y

The thermal conductivity coefficient K of a solid is defined with respect to the steady.state flow of heat do....'Il a long rod with a temperature gradient

<trldx,
iu=-K

<tr ' dx

(41)

where Ju is the llull: of thermal energy, or the energy transmitted across unit area per unit time.

132

This form implies that the process of thermal energy transfer is a random process. 11le energy does not simply enter one end of the specimen and proceed directly in a straight path to the other end. but diffuses through the sped~ men. suffering frequent collisions. If the energy were propagated directly through the specimen without deflection, then the expression for the thermal flux would not depend on the temper.llure gradient, but only on the difference in temper.llurc liT between the ends the specimen, regardless of the length of the specimen. The random nature !Sf the conductivity process brings the temperature gradient and. as we shaD see.a mean free path into the expressiOn for the thermal nUll:. From the kinetic theory of gases we find below in a certain approximation the following expression for the thermal conductivity;

or

(42)
where C is the heat capacity per unit volume, tl is the average particle ,,-:eIOCity, and t is the mean free path of a particle between collisions. This result was applied first by Dehye to describe thermal conductivity in dielectric solids, with C as the heat capacity of the phonons, v the phonon velocity, and t the phonon mean free path. Several representative values of the mean free path are given in Table 2. We give the elementary kinetic theory which leads to (42). 1be flux of particles in the 1" direction is ;n(lvJ), where n is the concentration of mo1ectJlcs; in C<luilibrium there is a flux of equal magnitude in the opposite direction. The (... ) denote average value. If c is the heat capacity of a particle, then in moving from a region at local temperature T + !:J.T to a region at local temperature T a particle will give up energy c !:J.T. Now!:J.T between the ends of a free path of the particle is given by
<lI' !:J.T=-t
:-"T

<lI'

<h'<h~'

where., is the average time bety,'een collisions. The net flux of energy (from both senses of the particle flux) is therefure
ju = -n(,,~C'T d; = -in(vi)

dT

C'T

d;"

<lI'

(43)

If, as ror phonons, v is constant, we may write (43) as

iu = -iCvfJ;
with

<lI'

(44)

l>'T

and C nco Thus K '" iCti'.

Ii

phOftmU II. ThnmtJ l'r-opvtiu

133

Table 2 Phonon mean free paths

(Q.1culall.'<l from (44), taking I! = 5 )( lW em/sce as a representative sound velocity. The

(s obtained in this way reler to umklapp pl"OCenes.]

c.yu.I I
T. "C
C, in

J an- 3 deg-'
~oo

----~~~ t, In A
0.13

Quartz""1
r\aCl

0 -190 0 -190

0.55 1.88 1.00

0.50 0.07
0.27

40 540 23

100

~Id 10 optic axis.

Thermal Resistivity of Phonon Gaa

The phonon mean free path e is determined principally by two processes, gl..'OmetriC'.I1 sC'.Ittering and scattering by other phonons. If the forces between atoms were purely harmonic, there v.lluld be no mechanism for collisions between different phonons, and the mean free path would be limited solely by collisions of a phonon with the crysbl boundary, and by lattice imperfections. There are situations where these effects are dominant. With anharmonic lattice interactions. there is a coupling between different phonons which limits the value of the mean free path. The exact states of the anharmonic system are no longer like PUTC phonons. The theory of the effect of anharmonic coupling on thermal resistivity predicts that t is proportional to 1fT at high temperatures, 2 in agreement with many experiments. We can understand this dependence in terms of the numberofphonons with which a given phonon can interact: at high temperature the total number of excited phonons is proportional to T. The collision frequency of a given phonon should be proportional to the number of phonons with which it can collide, whence t tIC liT. To define a thermal conductivity there must exist mechanisms in the crystal whereby the distribution of phonons may be brought locally into thermal equilibrium. Without such mechanisms we may not speak of the phonons at one end of the CT}'Stal as being in thermal equilibrium at a temperature T2 and those at the other end in equilibrium at T I lt is not sufficient to have only a way of limiting the mean free path, but there must also be a way of eSbblishing a local thermal eqUilibrium distribution of phonons. Phonon collisions with a static imperfection or a crystal boundary
15 J. M. 'l..iman. lmrtlfl$ and pllOnOOlS, Odord. 1960, R Berman. ~Heat a:xKIuct.ivily m nOll-metallic <.Y)'Itlls,- Contemp. 1'11)"1 14, 101 (1973).

'"

". .,
g.

~~~~~~",:M

"'"

Figure 16,. Flow <:X gas molecules in a slalc of driftIng equ,librium down II long open 'u"" with frictionless walls. Elasticcollision p~ses IImong the I?U mol"""l.: . do not change the momentum or energy nu~ of the g.~' because in cach ('(lilision the velocity of the center of mass oflhe colliding particle. and !.heir Cll("lt)' remain unchanged. Thus energy is lmmporlcn from left to ri~l "'lhot,1 !>dng driven by IIlcmpcrlllurc gradient. Therefore the thermal resistivity is zero and tile thermal conductivity is infinite.

Figure t6b The usual definition of thermal conducti"ily In II ~ C" to uituatiOfl wJ.e.reno mass now is permitloo. Here the lube is closed at both .,nds, preventing the C"SCl'Ipc or cntnOCf: of molecules. With II temperature gradient the collidinJ.( paiu with abovc"vcragc center of m~s velocities wiJltcnd to be directed 10 the r.ght, tho"" w,th klow-a,'cragc "coolies willl""d 10 be <!irectro to the left. A .Jight ronnlmlion gradient. high On Ihe right, will be ~et up to enable lhe net ma$s transport 10 be ttlV "hile allowing .. net energy t .... nsport from the 00110 the cold end.

,..,r

will not by themselvcs cstablish thermal equilibrium, because such l'OlIisions do not change the energy of individual phonons: the frequency W2 of the scattered phonon is equal to the frequcnt)' WI of thc incidcnt phonon. It is rather rcmarkable also that a three--phonon collision process
(45)
will not establish equilibrium. but for a subtle rcason: the total momentum of the phonon gas is not changed by such a l'OlIision. An equilibrium distribution of phonons at a temperature T can movc down the crystal with a drift velocity which is not disturbed by three-phonon collisions of the form (4:5). For such collisions the phonon momentum

(46)
is conserved. because on collision thc change in J is Ka - K2 rlK is the number of phonons having wavevcdor K.
-

K l == O. Here

S Phonu II. T""'"-J Propmia

IJ,."

Net n- of pM...."

..... ...

Fiplre 1& In .~tal ......, may arranse 10 crute phonoN midi)' al oOleend.:at by iDuminalmgthe

Ief\ nld ....i lh. b.mp.. From tt. end there .... 111 be. nt'l Om of phma'll t_wi! the ri!!Jlt mtl olthe "f)'SIai. 1 only N ~ (K, ... K1 - K,,) OC'eUr. t .... phonon nUl< is unchanr;ed in momenlum 00 roIl,sion and IOf1lC IlIlOrlOO nUll ""iii perllsl down Ihe length of the "f)slll\. On arrival Ii phonoos at
the right end we can arrange in prindple 10 comerl rnmt of their energy 10 nldi8lion. thereby eresling a sink for the phonollll. Jost as in (a) the Ihennal resistivity is :r.ctll.

Figure Hid In U t""0<:t:SSl:S there is large net ehngc in phonon momentum in each collisioo evcnI. An initi;al net phonon Om ....~11 Bf,lidly dl:C&y as ....'e move to lhe righl. The ends may ad as SQ.Jn;ft lind $in13. N.,t CIlCIg)' tnmporl under Icm~tuu grad..... l (lC'CUn as in (b).

For a distribution with J " O. collisions such as (45) arc incapable of establishing complete thermal equilibrium because they leave J unchanged. If we start a distribution ofhat phonons do\\--n a rod with J ;E. 0, the distribution will propagate down the rod \\ith J unchanged. ThcrcfOt'"e thcre is 110 thermal rcsis tan<.'C. The problem ali illustrated in Fig. 16 is likc that of the collisions between mol.ccules of a gas in a straight tube \\;th frictionless walts.
UmkIapp l'rocet/Ses

Thc important three-phonon pnx.'esscs that cause thcrmal resistivity are not of the form K t + ~ = ~ in which K is conservctl. but are of the form
(47)

where G is a reciprocal lattice vcctor (Fig. 11). 'I1lese processes. discovered by Peierls, are called umklapp processes. \Ve recall that G may oc'Cur in all momentum conservation laws in crystals. In all allowed processes of the form of (46) and (47), energy is COflscr"ed. We have seen examples of wave interaction processes in crystals for which the total \Vavcvcctpr change need not be zero, but may be a reciprocal lattice

'36

" _\

..
w,

'"

"
<

'"

...

rJgllrt: 17 (a) ~ormaI '" + K., - li.3 .00 (h) umkbpp K, + 1(. - ~ + G phonon coIIioion pro~ in. tu-wmemional >quare lattice. l1lc ~ in each tcure reprrsents the rIOt Brilkluin zone in the phonoo K spaoe; this woe rontallli all the possible independent ""JtIO ..,(Ihe phonon "'a\.'C\'edor. \'edon K with arrowhcJlds. IN- centn of the zone r~lI'UClll phonons absorbed in the collWor! P"O"f'U, tho.c with arrowtlc.ds away rrom the center ofthc zone reprnrnl phonons emitted in t1k"c:ollisim. We sec in (h) that in the: um~ pn:lC'ftS the dirc:dion of the r-componrnl of tbe phonon flux hat bcetl revt:'TKld. The rcci(ll'"OC'a.llatltce \'eC-ta C shown is "length 2m'... ",1)e't'Co is Ihr lattice constant ofthc:cryst.J IallJce. and is 1~110 tile K" ...il. F"orall proeesscs, N or U, energy must be ronse,...,ed, so 1M' + .... - toIJ

vector. Such processes are always possible in periodic lattices. The argument is particularly strong for phonons: the only meaningful phonon K's lie in the first Blillouin Wile, so that any longer K produced in a (.'ollision must be brought back into the first zone by addition of a C. A collision of two phonons both with a negative value of K" (,',lIl by an umklapp prot.'Css (C '" 0) create a phol1on with positive K,. Umklapp processes arc also called U processes.

Collisions in which G = 0 arc called normal pnx:esses or N processes. At high tempemtures T> 0 all phOilon modes arc excited occause kfjT > f!w,na>.' A substantial proportion of all phonon t'OlIisions will then be U processes, with the attendant high momentum change in the t'OlIision. In this regimc wc can cstimatc the thennal rcsistivity without p.1rlicular distinction bctwccn Nand U pnx.'Csses; by the earlicr argumcnt ahout nonlinear cffccts we cxpect to find a lattice thcrmal resistivity IX T at high temperatures. The cnergy of phonons K 1 , K2 suitable for umklapp to occur is ofthc order or ikfjO, because each of the phonons I and 2 must ha\'e wa\'c\'cctors or thc order of iG in order for the collision (47) to be possiblc. Ifboth phollons have low K, and therefore low energy, there is no ....':1y to get from their collision a phonon of "':1\'c\'cctor outside the first ZOILC. 'rne umkbpll process must conserve energy, just as for thc normal process. At low temperdturcs the number of suitable phonons of the high energy ikfj8 required may be expected to vary roughly as cxp(-812T), according to the Bolt-.nnann factor. 111e cxJXlnential IOnn is in good agrccment with experiment In summary, the phonon mean

PItor.ctN II. T""-l Propmin

137

..
, '00

'!

00

Fiplr~

18 n.c:rm...1 conductivit)' of a h;ghly purifocd crystal of :!Odium fluoride, aflCt'

,
',,L-:---:---f.,-;;;---,J;-~ 2 S 10 20 SO 100
T~..." ..

H. E. Jackson, C. T. Wallru, ....d T. F.

MeNcD)'.

free path which enters (42) is the mean free path for umklapp collisions between phonons and not for all collisions between Ilhonons.
Imperfediort3

Geometrical effects may also be important in limiting the mean free path. We must consider scattering by crystal boundaries. the distribution of isotopic masses in nalural chemical clements, chemical impurities, lattice imperfections, and amorphous structures. When at low temperatures the mean frcc path t becomes comparable with the width of Ihe tesl specimen. the value of t is limited by tbe \\;dth, and the tbennal conductivity be<."Omes a function of the dimensions of the specimen. This effect was discovered by de Haas and Bicrmasz. The abrullt decrease in tbcrmal conductivity of pure crystals at low temperatures is caused by the size effect. At low temperatures Ibe umklapp process becomes ineffective illlimiling tbe tbermal conductivity, and the size effect becomes dominant, as shown in Fig. 18. One would expect then that the phonon mean free path would be constant and of the order of the diameler D of the specimen. so that
K-CvD.
(48)

1hc only temperature-dcpcndent term on the right is C. the heat capacity. whicb varies as T 3 at low temperatures. ,Ve expect Ihe thennal conductivity to "ary as T 3 at low temperatures. The size effect enters whenever the phol\on mean free path becomes comparable with tbe diameter of the specimen.

1
136

wo
K - O.06T' /

....
Normal

~nntlll..J

&7

+--

10

If.
( /

1"L'" J\

, ,
1

if , 1/
o

II

I'"

, ,

III

20

50

100

200

500

Tt"'IKr.llll''', K

figure 19 Isotope dk'Cl on thermal conduction in genm",ium, amounting to a f""lor of three al the ronducli\'ily fTklxirnum. The enriched specimen is OOIlCr<.'Cnt Gc~'; ""turdl germanium is 20 perrenl (;em, 'l:1 pcfC'Cnt Cc711 8 \>CITenl Ce7J , ~ percent Gel<, ""d 8 pe.-cenl Gc7tl Below 5 K

the enriched specimen has K ., 0.060 T3, which agR'CS well with Casimir's theory for thermal r"s~tliJ1ce Clu,;ed by boundary scattering. (After T. H. Cerolle and c. w null.)

Dielectric crystals may have thennal l-'Onductivitics as high as metals. Synthetic sapphire (AlzO;J has one of the highest values of the conductivity: nearly 200 W em-I K- 1 at 30 K. The maximum of the thermal conductivity in sapphire is greater than the maximum of 100 W ern-I K- 1 in copper. Metallic gallium, howcvcr, ha.~ a conductivity of 845 \V cm- I K- 1 at 1.8 K. Thc elcctronic (.'Olltribution to thc thcrmal (.'Onductivity of metals is treated in Chaptcr 6. In an othcrwise perfect crystal, the distribution of isotopes of the chemical elements orten prO\idcs an im[Xlrtant mechanism for phonon scattcring. The random distribution isotopic mass disturbs thc periodicity of the density as seen by an clastic wavc. In some substances scattering of phonons by isotopes is (.'Omparable in importaJK'C to scattering by other phonons. Results for gennanium arc shown in Fig. 19. Enhan(.'Cd thennal (.unductivity has been observed also in isotopically purc silicon and diamond; the latter has devk'C importance as a heat sink for laser SOUR'CS.

or

5 Phonon' II.

Th~r",ul

Propu,;".

139

Problems
I. Singularity in cUnsity 0/ ,totd. <a> From the dispersion relation dcrh'oo in Chaptcr 4 fa- a monatomic lincar latlice of NatOlI's ....ith nearest ncighbor interactions, show that thc dcruity of modes is
2.\'
D(w) - - . .

7(~-'--'-.?)--"'

where w", is the maximum frc<juency. (ll) Suppose that an optical phonon brand, has tile form "'.K) - /.I.tI- AK 2 , near K - 0 in three dimensions. Show ll~'lt D(w) _ (U2'1r'f(2rrfAmx/.l.tl- W)112 for W < /.I.tIand D{w) - 0 for w > /.I.tI. Ilere the den5ity of modes is discontinuous.
2. 11:,", t~mol dilalion 0/ crystal cell. (a) Estimate for 300 K the root mean square thermal dilation t1V1V'for a primith'e cdl of Odium. Take the bulk modulus as 7 x to LO erg el.1l-3. Note that the Deb)'t" temperature 158 K is less than JOO K, llO that the themw energy is of the on:Ier of k.T. (b) Usc this result to estimate lbe root mean square thernlal nuctuatioo .1aJa of the lattice parameter.

3. Zero point lattice rlit]J/aCfmlcnl atld stroin. <a) In lhe Oebre approximation, show thal the mcao square displacement of an atom at allsolute 7.(.'1"0 is (R~ 3liw1f8~pv3, wllCre v is the velocity of sound. Start from thc result (4.29) summed O\Icr the indcpcndenl lattice modes: (R~ "" jIil2pV)}:w-I. We 11l\\'c includL,(\ a factor of f to go from mean square amplitude to mean ~uare displaet'ffieot. (b) Show that It..-I and (R~ dh'crge fOl' a Qrle-dimcnsiooallatticc, bul that tht.. 1llC8.l1 square strain is flnitc. Comidcr aRlaxf) = t}:K lIr.4 as thc mean square strain, and show that it is equal to tiwtU4l>fNv3 lOr a lioe of N atoms each of mass M, counting longitudinal modes only. 1be divergence of Rt i~ not significant lOr any physK.-a1 measurcment. 4. Hw. cupac;ity o/wyer lallice. (a) Consider a dielectric crystal made up oflayen of atoOls, with rigid coupling betv.'t'Cn la)'Cl"S so that the motion of the atoms is restric!t't! to the plane of the layer. Show that thc phonon IH.'at capadty in the Ocbye appro..imation in the low temperature limit ill proportional to T i . (h) Suppose in~tcad, as in many layer structUl'C5, that adjaceflt layers are \'cry weakly bound to each other. What form would you expect the phonon heat eapacity to approac::h at extrcmely low tcmperatures?
S. Cruneiten amstant. (a) Show thal thc rrce eoergy of. phonon mode offrequeocy w is kilT In 12 sin!l (lfcJ2Jrs TJ). It ill nCC'CSBry to retain the ttro-point energy itiw to obtain this result. (b) If ~ is the fractional volumc change, then the free cocrgy of the crystal may be written as
F{A, T) = fBa
This: pmblcm
u.t

, + ksT J " In [2 sinh (lfwl,flksT)1

IS _

diffinllt.

'"
where B is the hulk modulu.. Assume tllat ti,e volume dependence of W)( is liwlw = -yl., where yis known as the Griinciscn constant. If'Y is taken as independent of the mode K, SIIOW that F is a minimum with respect to l!. wilen BIJ. .. y:titiw (.uth (tiwl2k lJT), and show Ihal this may be written in terms of ti,e thermal energy density as !:J. = ."u(1')/B. (c) Show that on the Debyc model y"" -a In fJliJ In V. Note: Many approximations arc involvc<1 in this theory: the result (a) is valid only if w is independent oflempcralure; ymay Ix: {Jllite different for different rnodC$.

References
R A. Cowley, Anhurmonic cry.tals, Repl. Prog. Phys. 31, pI. I, 123-166 (1968). C. Leibrrit:d and W. Ludwig, 'I1\COTY or anharmonic effects in crystals, Solid slate physics 12, 276--444 (1961). T. Ristc, ed., Anharmonic latllces, strtlclurol traMitlons (Inti melting. NoordholJ, 1964. R. S. Krishnan, Thennal np....sion of crystols, Plenum, 1980. A. D. Broce and H. A. Cowley, Sfrucfurol pilau fronslflollS, Taylor and Francis, 1981. M. Toda, Theory of nonlincar lattil:es, Springer, 1981.
M M

THERMAL CONDUCnVITY
E. Parrott and A. D. StllckC$, TI,enlWl cvndllcfh,i1v of solids. AC'lidcmic Pre.s, 1975. P. C. Klemens, "Thermal conducti\ily and lalliee vibration modes," Solid slatc physics 7,1-98 (1958); we also E"cydo. of ,>hysit;8 14, 198 (1956). C. Y. Ho, R W. Pov.'ell and P. E. Liley, Then/wi co"dllctivitlJ of the eUllIumfs, A comprehensioo review. J. of J'h~,. and a'em. Rd. Data, \'01- 3, SUllPlemcnt I. It P. T~'C, cd., Thermal comillctivily, AC'lidemic Pre", 1969. J. M. Ziman, Electrons and pho..ons, Oxford, 1960, C1\llptcr 8. It Bcnnan, Thermal co"dllctit:Jn In solids, Oxford, U176.

J-

C. M. Bhandari and D. M. Howe, ThenlWl conductio" if' 6emlco,u1uctOfll, Wiley, 1958.

6
Free Electron Fermi Gas
ENERGY LEVELS IN ONE DIMENSION EFFECT OF TEMPERATIJRE ON TIlE FERMI-DIRAC DISTRIBtnlON
FREE ELECfRON GAS IN THREE DIMENSIONS HEAT CAPACITY OF TIlE ELECl1\ON GAS

144
146
146 151

E:Ilperimental heat capacity of metals

Heavy fermions ELECffiICAL CONDUCI1vrrr AND DUM'S LAW Experimental electrical resistivity of metals

1M
156 J56
159

Umklapp scattering
MOTION IN MAGNETIC FlEWS

162
163 164

HalJ

e(f~t

TIIERMAL CONDVC11VI'IY Of METALS Ratio of thermal to dectrK-al rooductivity

NANOSTRUcrURES PROBLEMS

166 166
HiS

Joo
16&

2. 3. 4. 5. 6. 7. 8. 9. 10.

~ of e1ed.roo ps Pressure and bulk modulus of an electron gas Chemical potential in two dimensions Fermi gases in astrophysics Uquid lie' Frequency dependence c:J the electrical condUdhity o,.namk magnetocooouctivity tensor for free eledl'"(llls Cohesi~'e energy of (ree eled:ron Fermi gas Sialic magnetooonductivity tensor Maximum surface resistance n. Small metal spl~res 12. Density of states-nanomelric wire 13. Quantiultion d condudance

t. Kinetic

169 169 1GB 170 170 170 170 111 171 171 172 172
172

REFERENCES

i"igure I Sch~matic model rJ a crystal of sodium meW. 1hc alomic cord AI"(! Na+ loons; they are imm".,;cd In a sea of ('Oflduction dCdrons. The conduction electrons arc derived from the 3s ,aIc"ct! electrons of the free aloms. The alomlc COR'S contain 10 c1cdrous in the COflI1j;lurolion l~pa In an alkali metal the alomic ~""'C1 llC.'l'\lP)'. relali\"dy small pIIrt (-15 percent) of the total \'Olumc of the cO)..tal, but in a nobk mdal (Cu. Ag. Au) the a10mic core> arc rclRth'c1y larger and mIl\" lJc in contact "';Ih cadi othrr. The mmmCll l.T)"Stal struc.1u ' room Icrnpemlure iJ bl-c .... the alkali met<ab .00 kc b- the nobk meta.b.

'''-

CHAPTER

6:

FREE ELECTRON FERMI GAS

lilliS'

In a 'heory lOhicll 110s gilifin rUlllb lik: ,lleSe, there cerlai,Jy be a gJTa' deal of tr"tI. H. A. Lorcllt-....

We can understand many physical propcrti~ of mctals, and not only oftllc simple metals, in tcrms ofthc free electron model. According to this model, the valenee electrons of the constitucnt atoms become conduction electrons aue! move about freely through the \'Olumc of the metal. E\'en in metals for which the free e1C(.1.ron model works best, the charge distribution of thc couduction electrons rcllects the strong elcctrostatic potcntial ofthc ion cores. Thc utility ofthc free electron model is greatest ror propertics that dcpend csscntially on the kinctic propcrties of thc (;onduction e1ectrous. The interaction oflhc conduction elcd:rons with thc ions ofthc lallicc is treated in Chaptcr 7. The simplcst metals are the alkali metals-lithium, sodium, pot ssium, cesium, and rubidium. In a free atom of sodium thc valctll.'C e1ectroll is in a 3s statc; in the mctal this electron becomcs a conduction electron. We speak ofthc 3s conduction band. A monovalent crystal whidJ contains N atoms will ha\c N collcluetion electrons and N positivc ion cor-es. TIle 1 a t ion COI'"C COntains 10 elcctrons that occupy thc b, 2$, and 2p shells ofthc free ion, with a spatial distribution that is esscntmlly the samc whcn in the mctJl as in the free ion. The ion cores 611 only about 15 percent of thc \ulumc of a sodium crystal, as in Fig. 1. The radius of thc free Na+ ion is 0.98 A, whcreas ollc-half of the ncarest-11cighbor distance of the mctal is 1.83 A. Thc intcrpretalion of mctallic propcrtics in tcrms of the motion of frcc electrons was dcvclopcd long before thc invention of quantum mechanics. l1le c1a~sical theory had several conspicuous su(.'cesses, Ilotably the derivation of the form of Ohm's law and the rclation between thc electrical and thcnnal condue-thity. The classical thcory fuils to explain the heat capacity and the magnetic susceptibility of the conduction electrons. (These arc not failures of the free electron model, but failures ci tile Maxwell distribution function.) 1here is a furthcr difficulty. From many types of experiments it is clear that a conduction electron in a metal can move freely in a straight path O\'er many atomic distances, undel1cctcd by collisions with other conduction electrons or by collisions with the atom cores. III a very pure specimen at low temperatures the mean free path lIlay be as long as lOS interatomic spacings (lllOt"c than 1 cm). \Vhy is colldcllsc<! mattcr so transparent to cOllduction electrons? l1lc answer to the (J. ion contains two parts: (a) A conduction electron is not
lU

...

(
ddlected by ion cores arranged on a IH:riodic lattice because maUer waves propagate freel)' in a periodic structure. (b) A conduction electron is scattered only infrequently b)' other conduction electrons. This property is a conseo. quence of the Pauli exclusion principle. Bya free clect.-on Fermi gas. we mean a g.1S of free electrons subject to the Pauli principle.

ENERCY LEVELS IN 01'\ DlMENSJON

Consider a free electron gas in one dimension, taking account of quantum theory and of the Pauli principle. An electron of mass m is confined to a !tmgth L by infinite b..,rriers (Fig. 2). The wavefunction 1jJ..(x) of the electron is a solution of the Schrooinger equation:HI/! = .y, with the neglect of potential energy we have H = ,,'lJ<bn, where I' is the momentum. In quantum theory p may be represented by -ill didx, so thai
(1)

where E.. is the energy of the electron in the OI"bital. We use the term orbital to denote a solution of the wave equation for a system of only one electron. The term alm-s us to distinguish between an exact quantum state of the wave equation of a system of N electrons and an approxi~ matc quantum slatc which we construct by assigning the N electrons to N different orbitals. whcrc each Orbital is a solution of a wave equation for one electron. 1"11e orbital model is elCact only if there are no interactions between electrons. TIle boundary conditions are l/J..(O) = 0; l/J..(L) = 0, as imposed by the infinite potential energy barriers. The)' are satisfied if the wuvefunction is sinelike with an integral number" of half-wavelengths between 0 and L:

fJI.. =A sin (211

A.

x) ;

inA.. = L

(2)

where A is a constant. We see that (2) is a solution of (I), because

d~" -_ - A - - "" - - r . cb: L L'

whence the energy
f:..

('m) (nw)
f:..

(l'Zl/J" - = -A - dr' L

(.m)' "n . (nw - - r)

L
(3)

is given by

= 2m

.' (nw)' L

We want to accommodate N e1cctrons on the line. According to the Pauli exclusion principle no t....,o electrons can have all their quantum numbers iden-

---EMrgy Ie-"'h
A __ 2 L

--\\'"".,r'''odKlrlS.
.cl:oli,,,, ><:ale

fipn 2 fin!: three ftlerfP' and _''eofundKlns of. K'I' ekrtron or mass ... mnfined to line oflmglh L. 11M: l':I'lCTg)' b-ds ~ Iabdcd llIX'Ording 10 the quantum number n ...h~h gi'"Cli the numbcT of hal."",..,k'flgd.. in lhe wa,..,runction. TIlc wavelengths are indicated On Ihe Wll\erunctions. TIle cnc.-rgy ~"dlhe len,oI or lluantum nnmber n is c.-qnaJ to (1I~I2mXnI2L)"

In..,.

-::=,.-/---_

.~

-----7 26

,
,

, ,

,,~----------:,

A-2/.

tical. TImt is, each orbital can be occupied by at most one electron. This applies to electrons in atoms, molecules, or solids. In a linear solid the quantum numbers of a conduction e1e<:tron orbital are "and til.. where 11 is any poSitive integer ond the magnetic quantum number m, "" :d-, according to spin orientation. A pair of orbitals labeled by the quantum numba- n can accommodate two electrons. one "'ith spin up and one with spin down, If there are six ele<:trons, then in the ground statc of the system the filled orbitals are those given in the table:
EIfron Electron

"

m.

=',""",

"
3 3

m.

~,

..

",

, ,

1 1

t t
1

, ,

t
I

1 1 0 0

More than one orbital may have the same energy. The number of orbitals with the salTlC energy is called the degeneracy, Let "F dcnote the topmost filled energy level, where we start filling the levels from the bottom (0 "" 1) and continue filling highcr levels with electrons unlil all N electrons are accommodated. It is convenient to suppose that N is an e,'cn number. The condition 211F "" N determines I1F' the value of 11 for the uppennost filled level. Thc Fermi energy F is defined as the energy of the topmost filled level in the ground state of the N electron system. By (3) with" "" OF we have in Olle dimensiorl:

EF "" ~("",)2 ""

2m
L

fl

2m

2(NW)'
2L

(4)

..,
EFFECT 01" TEMPERATURE ON THE FERMI-DIRAC DISfRIBUTlON

The ground slale is tbe state of the N cllX,tron system al ,lbsolute zero. What happens <15 the temperature is increased? This i~ a standard problem in elementary statistical mechaniCS, and tIle solution is given by the Fermi-Dirac distribution function (Appendix D and TP, Chapter 7). The kinetic energy of the electran g.ls increases a~ the temperature is increased: some energy levels are occupied which were vacant at absolute zero, :'md some levels are vacant which were occupied at absolute :t.ero (Fig. 3). 111e Fenni-Dirac distribution gives the probability that an orbital at energy E will be occupied in an ide,ll electron gas in thermal equilibrium:

(5)
nle quantity J.I.. is a function of the temperature; IJ. is to be chosen for till;: particular problem in such a way that the total number of partides in the system comes out rorrectly-that is, equal to N. At absolute 7-ero ~ = t:~., because ill the limit T_ 0 the function f(t:) changes discontinuously from the value 1 (filled) to the value 0 (empty) at t: =' t:F =~. At (Ill tempcrrltllrcsf(t:) is ec/tllll to .. wilen I:: =' p" for then the denominator of (5) has the value 2. The quantity p, is the chemical potential (rp. Chapter 5), and we see that at .Ib~olute :t.ero the chemical potential is equal to the Fermi energy. defined as the energy of the topmost filled orbitrJ at absolute 7-era. The high ellergy tail ofthe distribution is that part for which e - p. J> kpT; here the exponential term is dominant in the dcnominator of(5). so thatf(e) ~ c'\plV-.t - e)JkpT]. This limit is called the Boltl.rnann or Maxwell distribution.

FREE ELECTRON CAS IN nmEE DIMENSIONS

The free-particle Schrodinger equation in three dimensions is

_ ,,,'1.

2m

(~+ ~ +~) th(r) = t:~ ~(r) o~ o!f oz?

(6)

If the electrons are ronfined to a cube of edge L. the wavefunction is the standing wave
(7)

where II" n~, are positive integers. TIle origin is at one corner of the cube. It is ronvenient to introduce wavefunctions that sati~fy periodic boundary conditions. as we did for phonons in Chapter 5. \Ve n( require the wavcfuuc

"=

"-r helron

F~...;

c...

14.

,
o

y<

10' A:

.,

~~

,."
,

--

""

, ~ l'--10' "
4

, I--,s rQ' , ,o
"j'~

A:

I
,

--

f\-\ ' " ~

,

"\

~ l:---. ,
di.. m un;..... 10' l.:

...,
,

,I'--- -I'---,

n.., results apply to a lC" io three dimensions. n.e total num1K-r or partock'S i$ ronstanl. indcpme1mt oft.,n,pcr...re. 11M: .,ho:,nicaI pot.,..tiallU.,..h temperature 'nay be react off the ~ &I the
<--nergy at whid, f = 0 5

Figure 3 F.,nni-l)jr.lC distnb"tion rUlIction at "arious tcmpo.llltur<:$. fer T~ .f.Jk~ - 50.000 K.

lions to be periodic in x.

I}.

z with perioo L. 111us

tKx

+ L,

I}.

z) = (x. !I. z)

(8)

and similarly for the !I and .z: roordinates. Wnvefunctions s,ltisfying the free particle SchrOdinger equation and the periodicity condition are of the fonn of a tra\'eling plane \Vil\'e:

I",(,) .
L

e<p

Uk . ,)

1
. ...

(9)

provided that the components of the wavevector k satisfy

+-'
2~

+-

4" - L .

(10)

and similarly for k ll and k~. Any component ofk is of the form 2IlTT/L. where 11 is a positive or negative integer. The components of k are the quantum numbers of the problcm. aloog \\ith the quantum number m. for the spin direction. We confirm that these values of k. satisfy (8). for
exp[ik,.{x + L)J = exp(i2I11r{x

+ L)lL}
(11)

..,. cxp(i2mn/L) exp(i2l1'lT) = cxp(i211TTX1L) = c\ll{ik.x)

.
On substituting (9) in (6) we have the energy vector Ie:
k =

of the orbital with wave-

--k'

h2 2m

h2 --(1..-: 2m

+ III + ~ .
II'

(12)

The magnitude of the waveveclor is related to the wavelength A by k = 27T1A. The linear momentum p may be represented in quantum mechanics by the operator p = -iIiV. whence for the orbital (9)
(13)

so thai the plane wave"'" is an eigenfunction of the linear momentum "ith the eigenvalue Ilk. The particle velocity in the orbital Ie is given by v = Mum. In the ground stale of a system of N free electrons the occupk.d orbitals may be represented as points inside a sphere in k space. 11k: energy at the
surface of the sphere is the Fenni energ)'; the W3\'C"edors al the Fenni surface have a magnitude k" such that (Fig. 4):
1'

= --/&, .

2m

,,'

(14)

From (10) we see that there is one al~-ed wavevector-that is, one distinct triplet ofquantum numbers kr kll" k~-fo .. lhe volume element (2mLr1 of

k space. Thus in the sphere of volume

2
4'1TJ(J,,13

47TKJ~

the total number of orbitals is

(15)

(27T1L'f

---J(J""N
3~
F

where the factor 2 on the left {.'Omes from the t\\"O allowed values of m., the spin quantum number, for each allowed value of L:. Then

kp

""

(3~N)I~

(16)

whieh depends only on the particle concentration.

Using (14),

(17)

TIlis relates the Fermi energy to the electron concentration NIV. The electron velocity &F at the Fermi surface is
(18)

Calculated values of k"" VT, and E,.are given in Table 1 for selected metals; also given are values of the quantity T F whieh i.s defined as E~kB' (The quantity r,. has nothing to do with the temperature of the electron gas!)

Fru: Electron Fer"'i Ca.

149

F..-mi IU"""", "''''''''l>Y

'.

Figure 4 In the ground slate of a system of N fn'C e1ecIrons the occupied orbitals of the S)'litcm IlII II. sphere of radim k". where ~ . - h 21.-f../2m is the energy of an ck-ctwll
having a wa,'C\'cctor k._

1,

".)

1 ,
FigureS DCllsity of single-particle slalt.'5 ,<5 a function of ..- ncrg)'. fur II. fre.. ..k'dron gas in three dimensions. The d"sh..-d curve represents the dt.'1lsily f(~,T)D(~} of filk'll orbitab at a fluile Icmpcl1lturc, bul such Ihal kilT is small in 0:III1P8ri5011 with EF- The sh"d.. -d area reprt.~ ..,Ills the filled orbitals at absolule zero. The ave,rage energy is increased when the temperature is ;llcrt,&ccl from 0 to T, rur electrons are IhcnnaJly excited from region I to region 2.

\
\

'--=-----"---);--.::~---

'.

,"

We now find an expression for the number of orbitals pcr unit energy range, D(I:), called the density of states. l We use (J 7) for the total number of orbitals of energy s:
N=3'7f
V

(2m')'" !F
,~2

(19)

so that the density of slates (Fig. 5) is

V . (2m)~' D() e _ _ = __ __ . 112

dN
dE

2'7f

(20)

'Slrictly, D(~) is lhe dcnsily of ollc-particle slatcs. Or dcn.itl' of orbitals. Oftcn onc .s<.1s lhc "01",,,,, V ~ 1 in c~prcssions for D(~).

Table 1 Calculated free electron Fermi surface parameters for metals at room temperature (Except for Na, K, Rb, Cs at 5 K and Li at 78 K)
Fermi temperature T,. (,Ik", In uel: K

Valency

Metal

Electron concentration, In em- 3

lUdius" parameter

Fermi
wa~e~edor,

FermI In
~Iodt)',
('In

"
3.25 3.00 '.86 '.20 5.63 2.67 3.02 3.01 1.88 2.65 3.27 3.56 3.69 2.31 2.59
2.07 2.19 2.41

In MTl- 1

._1

Fermi energy, In eV

LI
~a

4.70 X 1(j3' '.65

lAO
1.15

,
~

Rb C, Cu A. Au Bo
C, S.

0.91 8.45

'.85
'.90

Lll X lOS 0.92 0.75 0.70 0.64 1.36 1.20 1.20

1.29 X lOS 1.07 0.86 0.81 0.75 l..57 L39 1.39

4.72 323 2.12

1.85
1.58 7.00 5.48 5.51
14.14 7.13 4.65 3.95 3.65 9.39 7.16

,5.48 X 3.75 '.46 2.15 1.83 8.12 6.36 6.39 16.41 8.27 5.43 '.56

lit

M.

24.'

B, Z, Cd AI C,
In

8.60 '.60 3.56 3.20 13.10 9.28

18.06 15.30 11.49 13.20 14.48

1.00 1.37 1.11 1.02 0.08 1.57 1.4<)

1.75 1.65 1.50

'.23 1.56 1.28 U8 1.13 1.82 1.62

2.02 1.91 1.74 1.82 1.88

10.90
8.88
13.49 12.01 9.08

'.24

11.63
10.35 8.60
9.37 10.03

Pb
Sn(w)

'.36

~ d'men.l<mle....aI,\lul pa... mele.ls dellncd .... r. - rdGH, where

..

2.23

1.57 1.62

10.87
11.64

"H Is the lint Bohr radius a"d ro Illhe radlul of a Iphere thaI 011,,111,," o"e e~lron.

"-ue Ekctron "-umi C...

lSI

This result may be obtained and expressed more simply by writing (19) as
III

N=

3 '2 In E + constant

.IN
N

=-'-

3 2

d.

whence
D(E).-~-

.IN d.

3N
2<

(21)

Within a factor of the order of unity. the number of orbitals per unit energy range at the Fenni energy is the total number of conducUon electrons divided by the Fenni energy, just as we would expect_

HEAT CAPACITY Qt- THE ELECl"RON CAS

The question that caused the greatest difficulty in the early de\'elopment of the electron theory of metals concerns the heat capacity of the conduction electrons. Classical statistical mechanics predicts that a free particle should have a heat capacity of Jk'h where ka is the Bolt....mann constant. If tV atoms each give one valence electron to the electron gas. and the electrons are freely mobile. then the electronic contribution to the heat capacity should be JNka , just as for the atoms of a monatomic gas. But the observed electronic contribution at room temperature is usually less than 0.01 of this value. 111is discrepancy distracted the early .....o rkers. such as Lorentz: how can the electrons p..1rticipate in electrical conduction processes as if they were mobile, while not contributing to the heat capacity? 111e question was nnswered only upon the discovery ohhe Pauli exclusion principle and the Fermi distribution function. Fenni found the l.'Drrect equation. and he wrote, "One ret:ognizes that the specific heat vanishes at absolute zero and that at low temperatures it is proportional to the absolute temperature:' \-"hen we heat the specimen from absolute zero not every electron gains an energy -kaT as expected classically, but only those electrons in orbitals within an energy range kaT of the Femli level are excited thennally: these electrons gain an energy which is itself of the order of kaT. as in Fig. 5. This gives an immediate qualitative solution to the problem of the heat capacity of the conduction electron gas. IfN is tile total number of dcctrons, only a fraction of the order of TIT~ can be excited thennally at temperature T. because only these lie within an energy range of the order of kaT of the top of the energy distribution. Each of these NTITF electrons has a thennal energy of the order of kBT. 1be total electronic thennal kinetic energy U is of the order of
(22)

IS'

The electronic heat capacity is given by

C~I = aUlaT- NkH(TIT,,)

(23)

and is directly proportional to T, in agreement wilh the experimental results discussed in the following section. At room temperature C J is smaller than the classical value iNk" by a lOCtor oflhe order orO.OJ or less, for TF - 5 X l~ K. We now derive a quantitative expression fOr the electronic heal capacity valid at low temperatures k"r ~ f:Fo 1be increase l:i.U"" V(T) - U(O) in the lotal energy (Fig. 5) of a system of N electrons when heated from 0 to T is
liU =

L-

dE dJ(E)f(E) -

f'
1" ,

dE dJ(E) .

(24)

Here I(E) is the Fermi-Dirac function and D(t) is the number of oro{tals per unit energy range. We multiply the identity 1

by

El'

to obtain

1 ,U~r + L)
N=

tlE D(E) f(e)

,IE D(E)

(25)

dE Ed(E)D(E) =

r'

dE fiFDCE)

(26)

We use (26) to re\\-nle (24) as

flU =

J"

dE(E - EI')f(E)D(E)

+ L#:r df(EF - E)[l - !(E)]D(E)

(27)

TIle first integral on the right-hand sKJe of (21) gives the energy needed to take electrons from f:f' to the orbitals ofenergy f: > Ef , nnd the second integral gives the energy nee<led to bring the electrons to f:" from orbitals below E". Both contributions to the energy are positive. The product f(f:)D(e)clE in the first intcgr<l.l is the number of electrons elevated to orbitals in the energy range ,IE at an energy f:. The factor (I - f(f:)] in the second integral is the probability that nn electron has bet!n removed from an orbital f:. The function AU is plotted in Fig. 6. In Fig. 3 we ploned the Fermi-Dirac distribution function versus E fOr six values of the temperature. 1lle electron COIlCelltration of the Fermi gas was takell such that f:f./k ll = SO,eM)) K, characteristic of the conduction e1cdrons in a metal. The heat capacity of the electron gas is found on differelltiating flU \\;th respect to T. The only temperature-dependent term in (27) isf(E), whellce ....~ can group terms to obtain

eft = -<lu elT

L-

dE(E -

f:f')-/T D(E)

<If ,

(28)

Al the temperatures of interest in metals T/f:f ' < 0.01, and we see from Fig. 3 that (E - e,...) dfldThas large positive pellks at energies near f:f" It is a good

F,.,,, Elulron Fermi Cw

153

f , s '

Figu~e 6 Tempernture dependence of the energy of a noninteracting fermion gas in three dimeusions_ The ene~gy j" plotted in lIorm... li~.cd fo~m as tJ.UINEF. where N i' the -n"mller of d<.'Ctrolls. The temperature is plott<.-d as k~TIE,-.

/ ,, . /
--......

v
0.4

V
0.6

"

..,

Region of d<'g<-"",r~'~ quanlum go>

,, ,
,,~

-:, ,

J"lIi<>fi of <"1

1..

1:"'_

~
............

.~...

,,

......

,
k~T_

Figure 7 I'lot of the chemical pot<''fltial Ii \'ersll'< temperature k~T ro~ a gas of nOllinteractillg fermion" ill three dimensions, Fo~ coI"'ell;cn"" in "Iutting. the unit' of Ii and k,;r arc O.763E,..

approximation to evaluate the dcnsity of states D(E) at E~- and take it outside of the integral:

(29)
Examination of the graphs in Figs. 7 and 8 of thc variation of Il with T suggests that when kRT <C Er we ignore thc temperature dependcnce of the chemical potential Jl. in the Ferlni-Di~acdistribution function and replace Jl. by

..

.
''''

,..t--~:::::::=.. k'-----".~,c-----",~" ,

--"

l-1p.lre 8 \'ari:lIion with lempef2lure oflhe ~ieaI potmtial Il, lOr free elcdroo Fcnni gases in 0"'" and throe dimensions. In common metal, d#~" 0.01 at room Icmpualure, so that lot is dO$cly etjual to #,. 1bese curves were calo:ubtcd from Rrid cJ"pansions of the intcgral lOr the number of
jlllr1idc. in t}"" $)"$Iem.

the constant

Er.

We have then, with

T -

kfjT,

(30)
We set
(31)

and it follows from (29) and (30) trull

Cd =

'?aT D(E,.)

I"

c' (Ix 7? (. I~ -~,/r e + ,

(32)

We may safely replace the lower limit by -Q(> because the factor ex in the integrand is already negligible al:r = -"tiT if we are concerned with low temperdtures such thai 100 or more. 'Ole integr<ll2 becomes

",h -

I"

_" drr' (c'+

.. 11-3' ".
..

(33)

~ inley-al is not elementary.

" LlU,..+I-.24"
r

but may I,r, l"'a!u:ued from the

IllO<'e

&mlliu result

on

dur~tiatiDn ofboth ~idcs

with resprd. 10 the puame1er D

"r-~---------------,
..

crr-2.<l'l+2..51~~ _ _. - .

"<,' . , -

"

n..--'m '"_.

_ ..
-"..-

-.~

.~

.- - . _.

.!:

8"

~.fII~'"

"ic------,l,-------./;,------"J o Ql Q2 Q3
1", in K 2

Figure 9 F..xperimental heal opacity values for potassium, plotted as crrversu5 T'l. (After W. H. Lien and N. E. Phillips.)

whence the heat capacity of an electron gas is

Cd = !~D(F)kiT .
From (21) we have
D(F) = 3NfM,.. = 3N/2kB T F

(34)

(35)

fur a free electron gas with kBTF

EiI

F'

Thus (34) becomes

(36)

Cd = iw2Nk sTITF .

Recall that although T F is called the Fermi temperature, it is not an actual temperature. hut only a convenient refe.-cnce notation.

Experimental Heat Capacity of Metab

At temperatures much below both the Ochye temperature and the Fermi temperature, the heat capacity of metals may be written as the sum of electron and phonon contributions: C = yT + AT3 , where l'and A are constants characteristic of the material. The electronic term is linear in T and is dominant at sufficiently low temperatures. It is convenient to exhibit the expelimental values of C as a plot of CIT versus T 2 ; CIT = 'Y+AT2,
(37)

fOr then the points should lie on a straight line with slope A and inteccept 'Y. Such a plot for potassium is shown in Fig. 9. Observed values of 'Y, called the Sommerfeld parameter, are given in Table 2. The observed values of the coefficient 'Yare of the expected magnitude, but often do not agree very closely with the value calculated for free electrons of mass m by use of (34). It is common practice to expre.ss the ratio of the Observed to the free ele<:tron values of the electronic heat capacity as a ratio of

'56

a thermal effective mass the rehltion

fnlh

to the electron mass m, where mlh is defined by ;{obscrved) ')'(free)

(38)

This form arises in a naturAl way because f' is inversely proportional to the mass of the electron, whence 'Yam. Values of the ratio are gi\'cn in Table 2. The departure from unity ill\'olves three separate effects:
The interaction of the couduction electrons with the periOOic potelltial of the rigid crystal lattice. The effective mass of an electron in this potential is called

the band effecti,'c mass and is treated later. The inler.tction of the conduction electrons with phonons. An elcctron lends to polarize or distort the lattice in its neighborhood. so that the moving electron tries to dnlg nearby ions along, thereby increasing the effective mass of the electron. 111c interaction of the conduction electrons with themselves. A moving electron causes an inertial reaction in the surrounding electron gas, thereby increasiug the effective mass of the electron.

Heavy Fermiom. Several metallic compounds have been discovered that have enonnous values, two or three orders of magnitude higher than Wiual, of the electronic heal capacity. The heavy fenRion compounds include UBe I30 CeAlJ, and CeCUjtSi2 It has been suggested that f electrons in these compounds may have inertial masses as high as 1000 m, because of the weak 0\"Crlap of wavefunctions off electrons on neighboring ions (see Chapter 9, "tight binding"). References are given by Z. Fisk, 1- L. Smith, and H. R. OU, Physics Today, 38. S-20 Oanuary. 1985). The heavy fennion compounds forrn a class of superconductors h-nown as "exotic superconductors."
ELECTRICAL CONDUCTM'Il' AND Onr..fS LAW

The momentum of a free electron is related to the wavevector by mv = "k. In an electric field E and magnetic field B the foree F on an electron of charge -e is -elE + (Uc~')( B], so that Nev."lon's seoond law of motion becomes

(ces)

F=

d" k = mdi =d hdi

') -e ( E + 7")( B .

(39)

In the absence of collisions the Fenni sphere (Fig. 10) in L: space is displacetl at a uniform rate by a coostaut applied electric field. We integrate with B = 0 to obtain (40) L:(t) - k(O) = -cEll" .

r
LI
163 0.749

B.

Table 2

E:lperimenlal and free electron values of electronie heat capacity const3nl 'Y of metals

2.18

a.5OO a.34
M.
13

0.17

(From compilatioll~ kindly fllrni,hed by N. PhilliJ1S and N. Pearlman. The ,hermll! efTcdive mass i~ Jdlned by Eq. (38).

N.
1.38 1.26
K

1.0gif 15.92
1.3

y
10,2
~

Observed'f in mJ mol-' K-'. Ca~iilcd Free

electron:i:1i;':fn:rmOF"k'
F. Co Ni
7,02
-~

m"lm -

(observed y)/(I'ee electror"l y).

_r - AI SI
1.35

- - ~ .~

P
~

0:91'2
1.'"

A'
0.19

C.
2'

So
10.7

Ti
~

'.08

tUB'

flIT

~~

3.35

1,25

I'

Rb
2.41

S,
3.6

,ao
HI

Z,

-9.26

C,

1.40

--- .- -9.20 4.98

~

Mn(y)

C,
0.695

4.73

1.911 1.26

,a
B.
2.7 1.4

1.799.

C,
3.20 1.43

L.

Nb

Mo

To

7,79
~

,a
~

,-

_.R,
3.3

0.596 o:s~g' 1i7~ -IT2S 1.38 0.85 a.58

Z, a.64
Cd'

G.

G.

.;:::
Sn
(_1

..,.......

Rh
4'
~.

Pd
9.42
~

A.
0.646

Sb

0.688
_Q:.~

" 1.69

Q.64j_
1.00

1~33

0,73

1.78 0.11 1.4~Q. ~

1.37

1.26

T.
"tv._ _

W
1.3
'.":lII'"

R.
2.3

0,

2"238" -1"93T :J.s~

la.

~-

2.16

5.'

-...,..,.

2.4

-.

I,
3.1

pt

A,

-.-,...-

6.8

." . 1.14

@,.
0.729

1.79

Hg(o) TI 1.47

Pb
TS'09~ 1.97

Bi

O":"5~

J.'i9"
1.14

'.98

a.OO8 --...-

1.88

.....-

f
p

15'

. .

. . "

+-+-+---t-'

Figure 10 (a) The Fenni sphere eodoRs tile occupied elcdron orbitals in k spao:e in the ground stale of the electron ps. The net momentum III Un), ~ W e"~ orbital k th<= 15 an OttUpicd orbital al -I.. (bl Under the influence of a constant ':1I'~ t" acting for a time inten-al t e-.cry orbital has its k '"ector increased by at - FtJ". l1"s is cq,,.,..1cnl to a dU:pbremenl d the whole Fum; sphe~ by Ok. The total momentum is filM"', if there are N cl.cdroos flreS<"l1. The application of the iOrce increa.!CS the energy u the I)"stem by N(MAittlm.

If the field is applied at time t = 0 10 an electron gas that fills the Fermi sphere centered at the origin ark space, then at a later time t the sphere ....ill be displaced to a new center al (41) Ok = -eEl/A Notice that the Fenni sphere is displaced as a whole. Because of collisions of electrons wilh impurities, lattice imperfections. and phonons. the displaced sphere may be maintained in a steady state in an electric field. If collision time is T, the displltccrnenl of the Fcnni sphere in the steady state is given by (41) with l "" T. 111C incremental velocity is v == -cET/m. Irin a constant electric lleld E there are n electrons of charge q = -e per unit volume. the electric cUlTent dtmsity is j = nqv = ne2TElm . (42) This is Ohm's law. The electrical conductivity u is defined by j "" uE, so that
(43)

The electrical resistivity p is defined as the reciprocal of the cooouctivity, so that

p = mJruflT .

(44)

Free J-;fednm H:rm'

c...

159

Values of the electrical conductivity and resistivity of the elements are given in Table 3. It is useful to remember that q in Gaussian units has the dimensions of frequency. It is easy to understand the result (43) for the conductivity. We expect the charge trdnsported to be proportional to the charge density ne; the factor elm enters bee-dwe the acceleration in a given electric field is proportional to e and inversely proportional to the mass m. The time 7 describes the free time during which the field acts on the c<irrier. Closely the same result for the electrical conductivity is ohtained for a classical (Maxwelli<in) gas of electrons, <is realized <it low carrier concentration in many semiconductor pfllblems. The matheillatics of this similarity is developed in the section on transport theory in TP, Chapter 14. It is possible to obtain crystals of copper so pure that their conductivity at liquid helium temperatures (4 K) is nearly 10"5 times that <it room temperature; for these conditions 7"'" 2 X 10- 9 s <it 4 K. The mean free path of a conduc.tion electron is defined as

(45)

where VF is the velocity at the Fenni sUiface, because all collisions involve only electrons near the Fermi surface. From T<ible 1 we have Vf- == 1.57 x lOS cm S-I for CUi thus the mean free path is l(4K)""'O.3cm;
((300 K) = 3 X 10- 6 cm .

Mean free paths as long as 10 cm have been observed in very pure metals in the liquid helium temperature range. For electron-electron collisions, see Eq. (10.63).

EqJerimental Electricm Resistivity of Metals

The electrical resistivity of most metals is dominated <it room temperature (300 K) by collisions of the conduction electrons with lattice phonons and at liquid helium temperature (4 K) by collisions with impurity atoms and mechanical imperfections in the lattice (Fig. 11). The rates of these collisions are often independent to a good approxiulation, SO that if the electric field were switched off the momentum distribution ""Quid relax back to its ground state with the net relaxation time

-:-+-,
7 7L 7,

(46)

where 7L and 7"1 are the collision times for scattering by phonons and by imperfections. respectively. The net resistivity is given by
P""PL+fJI

(47)


Table :l Electrical conductivity and
resi~tivity

of melals at 295 K

LI
1.01 9.32

B.

3:00.
3.25

(Resistivity value~ as given by G. T. Meadcn, Electrical resistance of metals, Plenum, 1965; residual resistivities have been subtracted.)

_ .. _F. Co

N,
2.11. 4.75

Mg
2,33 430

CQn..Quell'{Lty In units 01 W~ (o!iiri:ri!!):!. ~ Resistivity In units of 10-. ohmem.

C,

K
y",
7,19

C,

So

TI
0~3 43.1

C,

M,

NI

_. ------.. ....... - -- - -
C N

N.

AI

SI

C'

n~::.
2.74

Z,

G,

~7!L 021 36 46.8

jl.50 19.9

Q1lr 0i2 ToT' T7~ 1A3; S:-BS-- 169_'" 0.61::: 14.85 12.9 139. 9.8 5.8 7.0 1.70 5.92

Rb
080 12,5

S,

Z,
0.24 42.4

Nb
6-6~~

Mo
'L89 5.3

To
_-0.7
-14.

R,

Rh

Pd

Ag

Cd
"

0:;;;-' O~i'-_ 58,5 21.5

l.3'
7.4

14.5

2,00:. 1l.9S: ~.2.G. 1.3-g'- 1:n; "D:"ll': 727 8.75 41.3 4B 10.5 1.61 11.0

C,
O-SO

B,
0.26 39

L,
013
79.

Hf
0.33 306

T,
13.1

R.

0,

I,
1.% 5.1

PI 1l:W
10,4

A,
4:S~

Hg 'flI.

TI
16.4

..... --- - -- - - x.-O-:m. - - -~(wJ

G.

A,

S.

B,

Sn

Sb

T.

-K, R,

Pb

BI

Po

0-76~ ~1.89

ZOO F,

5.3

0.54..... nO' 186 91

bT1: b.6J:.
95.9

-o.~

-0.096_
116,

2,20

21.0

o,22.... 46

R,

Ao

-At L,
53.

"

C.
81

P,
57.

Nd
0.1'59.

Pm
~

{rt:c '0.(5'

Gd " 0]0- O:W

99.
89.

Sm

Tb
111.

Oy

Ho
77.7

"

~ 1l1i9O "TIl
134.

90.0

~ ~

"

Tm
62.

Yb
26.4

81.

-o:l6 'O"~ -lrn"'"

Th
O.~

P,

U
Qj 25.7

Np
0.085 118.

P,
0.070. 143.

Am
~

15.2

- ~

Cm

Bk

Fm Md " - - - Cf

- No L'
~

--::--""

FrTf' t;lf"," Fb-m' COl

161

Flgurf I L Electrical rl:si,ti"ily in mOSI metal, ari5ell from roI1isklm of flc.>ctrons "ith irrrularilic5 in t)w bttioe, 15 in Ca) by phonons and in (b) by Impuritlel: and , - . 1 bttft "t~.

Fi",re 12 Rnis~ or potauiom Lei"".. 20 K, 15 mcasoroo on two spel:irnens by

D. MacDonald and K. Mellddssolm 'Ibc dif ferenl inlc~pls at 0 K arc attributed 10 difICcu.I roncenlr,lliom cl impwitic:s and Iblle im~'dionl in ~ two spK'imcns.

,,k,----i'c---~,,;--,,",--."':,_
T.....,....~.K

where PI.. is the r~istivity caused by the thennal phonons, and p, is the resistiv> ity caused by scauering of the electron \\~dVeS by st-dtic defects that disturb the periodicity of the lattice. Often PI.. is independent of the number of defects when their concentrdtion is small, and orten P, is independent of temperature. This empirical observation e.>.:1)resse.~ Matthiessen's rule. which is convenient in analyzing experimental data (Fig. 12). 'lllC residual resistivity, p,(O), is the extrapolatt:d resistivity at 0 K because PI.. vanishes as T _ O. lbe lattice resistivity, P - p,(O), is the same for different specimens of a metal. even though p~O) Inay itself vary widely. The

p,.cn ""

'"
rC5isthity ratiO of a specimen is usually defined as the ratio of its resistivity at room temperature to its residual resistivity. It is a convenient approximate indicator of sample purity: for many materials an impurity in solid solution creates a residual resistivity of about I p.ohm-cm (I x 10- 6 ohm-em) per atomic percent of impurity. A copper specimen with a resistivity r-a.tio of WOO

will have a residual resisti\ity of L7 X 10- 3 p.ohm-cm, corresponding to an impurity concentration of about 20 ppm. In exceptionally pure specimens the
resistivity rJtio may be as high as toe, whereas in some alloys (e.g., manganin) it is us low as 1.1. The temperature-dependent part of the electrical resistivity is proportional to the rate at which an electron collides with thermal phonolls and Iher mal electrons (Chapter 10). The coUision rate with phonons is proportional to the concentration of thermal phonons. One simple limit is attemperdtures over the Debye temperdture (J; here the phonon concentration is proportional to the temperdture T, so that pIX T for T> (). A sketch of the theory is given in Appendix J.

Umklnpp Scattering
Umklapp scattering of electrons by phonons (Chapter 5) accounts for most of the e1ectriC'dl resistivity of metals at low temperatures. These are e1ectronphonon scattering processes in which a reciprocal lattice \'ector G is iO\'olvoo. so that electron momentum change in the process may be much larger than in a normal electron-phonon sC'dttering process at low temperatures, (In an umklapp process the wavevector of one particle may be "flipped O\'er. ") Consider a section perpendicular to [100] through two adjacent Brillouin mnes in bee potassium, with the equivalent Fermi spheres inscribed within each (Fig. 13). The lower half of the figure shows the normal electron-phonon collision k' == k + q, while the upper half shows a possible scallering process k' = k + q + G involving the same phonon and terminating outside the first Brillouin 7.one, at the point A. This point is exactly equivalent to the point A' inside the original zone. where A.\' is a reciprocal lattice vector G. This scattering is an umklapp process, in analogy to phonons. Such collisions are strong sC'dllerers because the scattering angle can be close to 7T, and a single collision can practically restore the electron to its ground orbital. When the Fermi surface does not intersect the zone boundary, there is some minimum phonon \Iia\'evector qo for umklapp scattering. At low enough temperatures the number of phonons available for umklapp sC'dttering falls as exp(- (}vln, where (}v is a chamcteristic temperature calculable from the geometry of the Fermi surface inside the Brillouin zone. For a spherical Fermi surface with one electron orbital per atom inside the bee Brillouin zone. one am show oy geometry that qo = 0.267 k t .

Frl'e Ele.:lron Fermi Ca.

163

l-'igurc 13 Two Fermi ~pherc:s in aclf'>c<'nt mnes, a wllstrodion to show the role of phonon umklapp processes in e1cdrical rcsi.livity.

The experimental data for potassium have the expected exponential form with 8u = 23 K compared with tbe Debye 8 "" 91 K. At the very lowest temperdtures (below about 2 K in potassium) the number of umklapp processes is negligible and the lattice resistivity is then caused only by small angle SC'dttering, which is the normal scattering. Bloch obtained an analytic result for the nonnal scattering, with Pt, 0: TS/(ji at very low temperatures. This is a classic limiting result. These nonnal prol'esses contribute to the resistivity in all metals, but they have not yet been clearly isolated for any metal bec-duse of the large competing effects of imperfection SC'dltering, electron-electron SC'dttering, and umklapp scattering.
MOTION IN MAGNETIC FIELDS

By the argument of (39) and (41) we are led to the equation of motion for the displacement 8kofa Fermi sphere of particles acted on by a force F and by fril'Iion as represented by collisions:

'(~+~)6k=F elt 7'

(48)

11Je free particle acceleration term is (ftclJelt) 8k and the -effect of collisions (the friction) is represented by M>kh, where 7' is the collision time. Consider now the motion of the system in a unifonn magnetic field B. The Lorentz force on an electron is (CGS)
(51)

F=-e(E+ ~VXB)
F=

(49)

-e(E +vx BJ

If mv = h8k, then the equation of motion is (CGS)

m(1 +~)v= -e(E+: vXB)

(SO)

164

An important situation is tile following: let a static magnetic field B lie along the z axis. Then the equations of motion arc
(CCS)

m(:t +-;)v~= -e(Er+ ~ v~) m (.i. + ..1-) v ~ -, (E - .!'-v.) d1 . d 1) m( -;ji+-;

T"
1/

(51)

~=-eE~.

The results in $1 arc obtained by replacing c hy l. In the steady stale in a static electric field the time derivatives arc zero. so that the (hift velocity is

vz=--E z
m

<T.

(52)

where We eBlmc is the C)'C1olron frequency, as discussed in Chapter 8 for cyclotron resonance in semiconductors.

11011 Effect
l1lc Hall field is the electric field developed across two faces of a conductor, in the direction j .. B, w'hen a current j flows across a magnetic fidd B. Consider a rod-shaped specimen in a longitudinal electric field E. anda trans verse magnetic field, as in Fig. 14. Ifcurrent cannot flow out of tile rod in the !I dircdion we must have 8v, = O. From (52) this is possible only if there is a transverse electric field
(CCS)

E = -wTE = - - - E
1/

,BT

~"

me"

(53)

(SI)

The quantity dei'i,.ed by

I"H~ I
E, J.B
is called the Hall coeffICient. To C\."3.luate it on
OUT

(54) simple model ",..e use (55)

'IeC

j" = ,w'TE/m and obtain

(CCS)

RII = -

eB'TE/mc ne'l'TE"Blm
I

:z: - -

(Sij

R,,= - -

'"

(;

,,'rec Elecfron Fermi C4t

165

I I iL.:. -fC-,--;;~'~" ~J'

,r"j'
.
+

co

: ~

,.,:::'~
fotuu;

I
+ + +

-"'!-_~..12<"_'. lJ'

-~

.
.

}.

Figure 14 Jnc slandard gc<lmetry for the 11311 clI'e<:t: a .od-5hapcd specimeu of redangul:u- crOUis plllC't.'<.I in .. magnetic fielll 8 .. a5 in (a). An elt.'Cl.ril' Oekl E. applioo 1It'1'(W the elill c1ednxlcs ""lL'CS Ill' electric currenl lIcnsily J, to flow <Io<oTl the nxI. TIle lIrift ~"elocil)' of the nepti,'el~ e1eehulS immediatelf after the elt'Ctric 6cld iii applied as sbown in (b). The lIdIeetion ill the V direc:tioll is caaed by the '1\a@l'ICtiefreld. Electrons ar:ewnulate 00 one fxot: of the rod and a posit;,..., ion ellceH is C$ubIil;hed or, the opposite until. ali in (e), the trarlS'"er'sre e~ril' Geld (liall frek!) j"st eaneeb the Lo..,,,17. folft: 1I.re 10 the magnet;" freld.
~ion

race

This is negative for free electrons, for e is positive by definition. 11le lower the carrier concentration, the greater the magnitude of the HaU cocfBcient. Measuring HII is an important way of measuring the carrier concentration. 11le symbol HI( denotes the Hall cocfHcient (54), but it is sometimes llSed v.itl1 a different meaning, that of Hall resistance in two.dimcruiional problems. When we treat such problems in Chapter 19, we shall instead let

P"

:=

BHIt

:=

E,jis

(55a)

denote the Hall resistance, where}. is the surface current density. TIle simple result (55) follows from the assumption that all rewation times are equal, independent of the "elocity of the electron. A numerical factor of order unity enters ifthe relaxation time is a function of tile velocity. The expression becomes somewhat morc complicated if both electrons and holes contribute to thc conductivity. The theory of the Hall effect again becomes simple in high magnetic fields such that WeT):o 1. where We is the cyclotron frequency and l' the re1axation time. (See QTS. pp. 241-244.)

'"
In Table 4 observed values ofthc Hall coefficient arc compared with values calculated from the carrier concentration. The most accurate measurements are made by the method of helicon resonance which is treated as a problem in Chapter 10. In the table "conv." stands for "conventional." The accurate values for sodium and potassium arc in excellent agreement with values calculated for onc conduction electron per atom, using (55). Notice, however, the experimental values for the trivalent clements aluminum and indium; these agr~ \\ith values calculated for onc positive charge carrier per atom and thus disagree in magnitude and sign with values calculated fOT the expected three negative charge calTiers. The problem of an apparent positive sign for the charge (:arricrs arises also for Be and As. as seen in the table. The anomaly of the sign was expbined by Peierls (1928). nle motion of carriers of apparent positive sign, which Heisenberg later called "holes:' cannot be ci'l"plainoo by a free electron gas. !luI Gnds a l1<ltural explanation in terms of the energy b,md theory developed in Chaplers 7-9. Band theoly also accounts for the QCCurrellee of very large values of the lIall coeffICient, as for As. Sb. anti Ui.

THERMAL CONDUC'TlVITY OF METALS

In Chapter 5 we found an expression K "" iC"..(' for the thermal concluctiv ity of particles of velocity v. heat capacity C per unit volume, and mean free path f. 'f11e thermal conductivity of a Fermi gas follows from (36) for the heat capacity, and with EF "" imui:

(56)
Here f = VF1"; the electron concentration is n. and 1" is the collision time. Do the electrons or the phonons carry the greater part of the heat current in a metal? In pure metals the electronic contribution is dominant at all temperatures. In impure metals or in disordered alloys, the electron mean free path is reduced by collision with impurities. and the phonon contribution may be comparable with the electronic contribution.

Ratio of Thermal to Electrical Conductivity

The ''\Tiedemann-Franz law states that for metals at not too low temperatures the ratio of the thenna! conductivity to the electrical conductivity is directly proportional to the temperature. with the value of the constant of proportionality independent of the particular metal. This result was important in the

Table 4 Comparison of obsen'ed Han coefficients "ith fr elt'dTon theory

[The ellpt'rlmental v,tlucs of RH as obtained by conventional methods are $umnlaTiud from data at room lempera!ure pusentoo in the Landolt-Bomstein tables. TIle values obtained by the helicon wave method at 4 K are by J. M. Goodman. The values of the carrier concentration n are from Table IA except for Na, K, AI, In, where C.oodman"s values are used. To convert the value of RH In CCS units to tile value in volt-cmfampgauss, multiply by 9 x 1011; to con\ert Rn in CCS to m3/cou:omb, multiply by 9 x

10 13.)

I::xpcruTM:fltal
Meial Melhod
in JO-~ CCS units

As.umcd

ClIlculated
in IO-~

R .

"".~

"'"""

-Un, CCS unils

conv. helicon oonv. helicon

",",v.

-1.89 -2.619 -2.3

-4.9-16 -4.7

I e1ectroo 1 eltttroo
1 electl'Oll
I c1e<:troo

-IA8

N.
K

-2.603 -4.944 -6.04 -0.82

Rb
Cu Ag Au
Be

cony. cony.
",",v. ",",v. ",",v.

-5.6 -0.6 -1.0 -0.8

+~7

1 elt.'Ciron

1 electron 1 electron

-1.19
-1.18

"g
AI

conY. Ilelicon l1elicon cony.

","v. ",",v.

-0.92

+1.136 +1.174

1 110le

+1.135 +1.780

'" Sb
B;

'"

1 hole

+50.
-22.

-6000.

'"
Tllble 5
LX Ilf' _ t..d ..nI~
Metal

Experimental Lorem: numbers

L )( 10" watt-ohmldeg'-

<J'C
2.31 2.35
2.42

100'(;

Metal

<J'C
2.47 2.51

100'(;

Ag Ao Cd Co

2.37 2.40

Pb
PI

2.56 2.60
2.49

2.23
2.61

Mo

2.<3 2.33 2.79

S.
W

Z.

252 3." 2.31

3.20 2.33

history of the theory of metals, for it supported the picture of an electron gas., It can be explained by using (43) for u and (56) for K:

! =
q

w2kjTnTl3m = .",. (k T1C'TIm 3 e

B )2T

(51)

The Lorenz number L is defined as

L
~

KJcfl'

(58)

and according 10 (51) should have the value

L =

3"" '" (k:

)' = 2.72 X 10-

13

(crg/esu - dcgt

=2.45 X 10- 6 watl-ohmldetr . \ (59)

111i5 remarkable rcsult involves neither n nor m. It docs not involve 'T if the relaxation times are identical for electrical and thermal processes. Experimen. tal values of L at O"C and at lOO"C as given in Table 5 are in good agreement with (59). At low temperatures (T q;: f1) the Lorenz numher lends to decrease; sec the book hy Ziman.

NANOSTRucrURES
"niC

term nanostrueture denotes a condensed malter structure having a minimum dimension approximately between 1 nm (10 A) and 10 nm (100 A). These structures may be fine particles, fine wires, or thin films. Fine particles typically contain between 10 and 1,000 atoms. Semiconductor technology (scc seetion supcrlattices" in Chapter 8) has made it possiblc to fabricatc small pools of electrons called in various ,"rays: Single-electron transistors, quantum dots, artificial atoms, Coulomb islands, or quantum COTnIls (Chapter 19). nlC unusual physical properties of nanostructures compared with bulk solids arc attributed to severnl factors involving properties trcated abovc and in latcr chaptcrs:
U

6 t-ree Itctron Fermi CII'

169

Thc ratio of the number of atoms on thc surface to the number of atoms in thc intcrior may be of the order of unit)'_ Thc ratio of surface energy to total cncrgy may be of the ordcr of unity. lbe conduction or valence electrons are confined to a small length or volume. so that the quantum wavelength of the Iov.'est electronic statc is constricted and COflScquently the minimum ...."3\'e1ength is shortcr than in the bulk solid. A wavelength or boundary condition shift will change the optiaU absorption spectrum (Chaptcr 11). Assemblies of nanoclusters of metals may have great hardness and )'ield strength, because it is difficult to create and to move dislocations (Chapter 20) in spatially confined regions. Magnetic monolayers, as of alternating films of ferromagnetic iron and of paramagnetic chromium. may have the magnetization (Chapter 15) of dIe iron films coupled by tunneling of the magnetization through the chromium barriCTS.

Problems
1. Kiflelic energy of elcctron gas. SI,ow that the kinetic energy of a thrt:e-4:1imcnsional gas of N free c1t.><:trons at 0 K is (60)

2. Preaure and bulk modulm cf an ekdron gas. (a) Deri\'t! a relation connecting the pressure and volume of an electron gas at 0 K. Ii int: Use the resulr af Problem 1 and the rebtion betv.-een f,. and electron concenmuion. The result may be ",Tittcn as p -I(UJV). (b) SIIOW that the bulk modulus 8 ~ -l'(ap/aY) of an ek.'dron gas at OJ( is 8 = 5p13 - lOUol9V. (c) Estimate for pot-.usium. USing Table I, the value of the electron gas contribution to 8.

3. Cl.emicul polenliuE in

fICO dimensiom. Show that tile chemical potential of a }'cnni gas in two dimcmions is givcn by:

(61)
for" electrons per unit area. Note: TIle denSity of orbitals of a free electron WIS in two dimensions is il1dellCl1denr ofenergy: D(.:) - mJdt pcr unit area ofspecimen.
4. Fomi gtlkS in asl~. (a) Ch'ell M0 = 2 x

g for the mass of the Sun, estimate the n\lmLer of electrons in the Sun. In a white dwarf .star thU number of electrons may be ionized and contained in a spllcre of radius 2 x l<f em; fmd the }'cnni energy of the electrons in electron volts. (b) TIlc energy of an electron in the relativistic limit f :B> fII~ is related to ti,e wavevector as Ii ill pc - like. SI,ow that the Fenni energy in this limit is ~.... Iic{NIVl I13 , roughly. (c) If the above number of electrons were contained within a pulsar of rndiuJ 10 km, show that the Fennl energy y./ould be -Iff eV. TIus value explains why pulsars are believed to be com-

to=n

170

posee! largely of neutrons mlher tlum of protons and c1t.'Clrons, for the energy release in the reaction II .... 1) + e- is only, 0.8 )( Iff cV, which is not large cnougll to enable many c!t.'drons to form a Fermi sea. TIle neutron decay proc~'(ls only until the electron roncentration builds up enougll to create a Fermi level orO.8 x 111 eV, al which point the neutron, proton, and electron concentrations are in Ctluilibrium.
5. Liquid Ife'. 1lJe atom lIe3 has spin land is a fennion. The density oflKimel lIe3 is 0.081 g em-a near absolute 'lcro. Calculate the Fermi energy for and the J,'enni temperature TF.

6. Frequency dependence o/i/te ekctricol amdUC:livit". Use the equation m{dvldJ. + ,iT) - -e lQr the electron drill Yelocity v to show thai the conductivity at fre'luency CoO is
u(w) "" 0(0) 1 + (wif

l+'~)

'

(62)

where 0(0) - ncTfm.

7. Dynamic Inag1l(!JowndUClil:ity 'crtlOr fur free dedronll. A metal with a l;"(Inct.'Olralion n of free electrons of charge -e is in a .stat~ magnetic field 8t. TI,e electric current den5ity in the X!/ plane is reL"\ted to the electric Held by
j, .. 0",..

+ O"..,E,.

Assume that the frequency Co) ~ w., and Co) ~ U'I'", where "'I: t-Blmc lUld 'I'" is the collision time. (a) Solve the drill velocity equation (51) to find the components of the magnetoconductivity tcnsor;
U"" ""

O"w - i~411W ;

0"., .. -q.. ,. W~~47TJ-

whcre 4'1T71c/m. (b) Notc from a Maxwell equation that the dicledric function tcnsor of thc medium is rclatt'<l to the concludivity tensor as ~ .. 1 + i(4mw)u. Consider an elcdromngnctlc ....'avc Witll wavevt"Ctor k - /d. SI.ow thnt tile dispersion relation for this wave In tile mCt"lium is

w: -

(63)
At a given frcquen<y there are t...."o modes of propagation with different wllve,'eC'lors and cHffcrent velocities. The two modes.correspood to cirn>larIy polarb-.ed waves. Because a linearly polariwd wll\'e C2I1 be dec."OlTlposed into ",""0 circular-Iy polarized w;!,\'es, it follows that the plane of polarization of. linearly polarized ....-a'e ....ill be rotated by lhe magnetic fiekt.

-8. Co/,t-.ive energy offret- t-kc;tron Fermi~. We def'me the dimensionleu length r. as rr/uH, where ro is the ....dius of a sphere that contains one electron, and aH is the Bohr .. .willS lit/em. (a) Show that the average Idnetic energy pcr c1t'CIrOn in a free
electron Fermi gas at 0 K Is 2.2Ur:- where the energy is expressed in ry<lbcrgs,
"1,il IlroLlcm is som.,whut difficult.

with I fiy - 1TlfJ412Ji'l. (b) Show that the coulomb energy of a point po!Iiti\<e charge e interacting with the uniform electron distribution of one deuTOn in the volume of radius 1"0 is -3l!'/21"o, or -3/1". in rydbergs. (c) SIIO'W that the coulomb self-energy of the dectron distribution in the sphere is 3ClSro , or &'5r. ill l)ubergs. (d) 11le sum of{b) and (c) gi.es -I.BCYI". for tile total coulombene:rg)' per electron. Show that the C(luilibrium .-a.luc d r. is 2.45. Will such mebl be stable Voitll resprd to sepan!lted II atoms?

9. Static magnnoconducdvily tensor. for tI,e driI'l: velocity theory of (51). show that
the static current density em be written in matm rorm as

In the high magnetk field limit

),
&

.) =

U,

('

1+(6Jcif

talc:'f"

-",T

0 I, show that
-00.,.

(&l)

or ~'f"~

Up

= nedB ..,

(65)

In this limit un ""' 0, loonier 116J&7. The Iluantity Uv- ls called the HaU condudi.ity.

10, MIlri",um fUr/ace resistance. Consider a ,,;quare sheet c:A side L, thickness d, and electrical resistivity p. The resistance measured between opposite edges of the shet.'t is Clllled the surface resistanre: Roq = plJul .., pld, .vf1ich is independent of the area L'l of the sheet. (Roq is called the reistance per St]Ullro and is elCprC!iSed in ohms per SlJuare, became p1d has the dimensions ofohms.) If we Cll]lress p by (44), then Roq '"" mfndr'f". Suppose now that the minimum value of the collision lime is (Iclermined by scattering from the surfaces oflhe slJeet, Kllhat 1'''' divl" whcre VI" is the fermi velocity. Th"s Ihe maximum SUrf.1Ce resistivity is ROIl - mtl~n,(ltf. Show for a monatomic metal sheet one atom in thickness d,at liM! ... Mtf ... 4.1 kfI, where 1 kO is ld' ohms. 'II, Small m4ltol spheres. Consider free electrons in a sphcriCll! square well potential of radius a, with an infinitely high boundary. (a) SI,OW that the Vo'llve fUllction of an orbital of angular momentum l and projection m has ti,e form

(66)
where the radial wave function has the form

aoo Y is a spllClical harmonic. Here J is a Bessel function of 111M integral order and satlsflt.'5 the boundary cond.ition h .,J,ka) ,.. O. TI,e roots give the energy eigenvalues II! of the levels above the bottom ofthe well, Vo~lere l! - h'/('!l2m . (b) Silow that the order of the levels above the ground orbital is
Lt, Ip, Id. 21. If, 2p, Ig, 2d, Ih, 3.f, 2f,

where " p, d,

f (

g, h denote f -- 0, I, 2, 3, 4. 5.

172

12. Den.rily of .tates-nanomdric rein', (a) Coruider _ nanometric ..ire in the Conn d a rectangular parallelepiped, with two sides L" - Lr - 1 nm and the long axis l~ 1 em. TIle single particle eigenstatcs of the system may be written <I.'l

l/J - sin (n"wxlLJ sin ("",""It,,) elPii2mvt.]

The energy of the eigenstate is, ..ith n - n.. 'T
('1, rl) - f{n) = (27r1lN'f/2lll "" Aftl "" Mlmt1

where v is the elmron velOCity along the ~l\.l;is. Here A - (2wfi'f12m and N .., \I[(E ~ EtA). Then ~ "" 2ANliN, show that die density of states V. at fixed 'I, with llCCOunt of the two spin orientations and the two:!: values of N, is

D~(E) - 46N1M: - VAN = Wf(E

(b) Slim over thc ", ..Ines of 'I for which
O!: f

f"yAJ - Vwfiv

n to show tllat

where G(r) is the HeaYiside unit function, zero lOr r

< 0 and

unity M r> O.

13. Quanli%ation of co"ducla1lCe. TI1C currellt in the nallometric wire of Problem 12 is 1 - (N+ - N-)ev" where N ~ - N_ - D(E,ftN, where V is the bias voltage. SIlO'" that the current may he written as

where '1_ is the number of occupied states n" n~, Ilcnce the quantized conductance is (2e2/"1rli~n.....,.

References
O. N. l.angenberg. -ResoufCf; leiter OEPM-I on tile onlill8l')'
cl~

properties d metals,

Amer. J. l'h)'5. 36, 777 (1968), An e>ellent earl)' bibliography on IrulSport effects, anomalous skin ell"cd, A7.bcl-Kaner cyclotron res<>n"1C'C. Irlas.:netoplaslna W3\"CI, ~ize effects, conduction elcdron spill reson.nc:e, optic.-aI speetl'll wlcl photocmissiou, quantum oseiliMtions, magnetic:
b~'ll, ullrasonic: effecb, .nd tI~ Kolu, efed. C. T. Meaden, Ehdriccll rmstoncc of metal", Plenum, New York, 196.5C. M. lIurd, TI~ lieD rff~ in .~ ond aIloy., Plenum, 1912.. C. L. Chien uK.! C. R. Wt'Slgate, 1s. Jloll effect Gnd ils Ilpplic<ltions, t'lenum, 19f1O. C. M. Ilurel. Dt'Cfr-ons In .nt'ltIz" Wiley, 1975. S. Kagoshim. et aI., On,,-dimeMklnol wnducr~, Springer, 1988. J. Ziman, EleI:rronr and "I",nonl, 0xf0rcI, 1960.

7
Energy Bands
NEARLY FREE ELECTR01"\ MODEL Origin of the energy gap Magnitude of the energy gap
BLOCH FUNcnO/,\S

176 177 179

1711

KRONIG-PENNEY MODEL WAVE EQUATION OF EI.t:CTltON IN A PERIODIC POTENTiAL Restatement of the Bloch tllC()rem Crystal momentum of an electron Solution of the central equation Kronig-Penney model in reciprocal space Empty lattice approximation Approximate solution near a zone boundary NUMBER OF ORBITAlS IN A BAND Metals and insulators

180 183 18.S 186 186 187 188

189

192 193

SUMMARY
PROBLEMS
I. Square lattice, free ell'Ctron ellergics 2. Free electron energies in reduced ZOJle

195
195

195 195

3. KronigPenney model
4. Potential energy in the diamond structure

195
195

5. Complex
REFERENCES

wa~'tttors

In the energy gap

196
196

6. SQuare lattice

196

--

D
Q

o
IN

D D

D
U

Insulator

\1~",1

~ur" t Schemat;,:, ck...-tmn (K'(."I,pancy of allowed energy bands ror an insulator. mda!. scmi"""tal. and sconieonductor. nl<: ,"ertical Volent d the boxes Indicates the allowed energy n.-gions; the <h:oded art.oas indicate the regions filled ...~th elect"",,\$. In a snnimdal {such as bWnuthl one band is a1~t &lied aM -.otI..,.. bomd is nearly empty aI. absolute ~ro,. but a pun:' 5Ql\icxlndudor' (such as sillcQol) ~., inp,lator at absolute zero. The loft of the t"'05Cntiowducton d.own is .t a finite lempenlure...ith e:-ri~ ellriled Ihtnnally. The ot~.,- "eonductm is ekdron-dt6dent henUK' oJ impurilie$. {

CHAYrER

7:

ENERGY BANDS

W',e" 'starlell to thi"k aboe,t it,lfclt tllat tile Illain

problem 100. to n-plUi" 11010 tire electro", cotll,lsueok by oil ti,e ion. ill a metal. ... By straigllt "o..rier a""lys;, I fOlmd to my deligltt tho' the loot.'e differed from tile rJia'1fI lOO&e offree electro~ orliy by 0 periodic morl.. fation.

F. Bloch

The free electron model of metals givcs us good insight into the heat capacity, thennal condudi\ity, electrical conducti\ity, magnetic susceptibility, and electrodynamics of metals. But the model fails to help liS with othcr largc questions; the distinction between metals, semimetals, semiconductors, and insulators; the occurrence of positive values of the IIall coefficient; thc relation of conduction e1cctrons in the mctal to the valence electrons of free atoms; and many transport pl'Operties, particulllrly magnctotransport. Wc need a less na'ivc theory, and fortunately it tUniS out tlLat almffilt any simplc attempt to improve upon thc free electron model is enormously profitable. The difference between a good conductor and a good insulator is striking. 1bc electrical resistivity of a pure metal may be as low as to- IO ohm.-cm at a temperature of 1 K, apart from the possibility of supcrconductivity. The resistivity of a good insulator may be as high as Ie? ohm-em. 11\is range of Ian may be the widest of any common physical property of solids. E\'ery solid contains electrons. nlC important question for electrical conducti\ity is how thc electrons respond to an applied clcct.rie field. We shall sec that electrons in crystals are ammgcd in energy bands (Fig. I) separated by regions in energy for which no wavelike electron orbitals cxist. Such forbidden regions arc called energy gaps or bund gaps, and result from the interaction of the conduction electron waves with the ion corcs of thc crystal. Thc crystal behavcs as an insulator if thc allowed cnergy bands arc cithcr filled or empty, for thcn no electrons can movc in an electric Held. Thc crystal behavcs as a metal ifonc or more bands arc partly filled, say between 10 and 90 percent filled. Thc crystal is a semiconductor or a scmimctal ifonc or two bands are slightly filled or slightly empty. To undcrstand the difference between insulators and conductors, we must extend the free electron model to take account of the periodic lattice of the solid. The possibility of a band gap is thc most important new propcrty tl13t emerges. We shall encounter other quite rcmarkable properties of electrons in crystals. For cxample, they rcspond to applied electric or magnetic fields as if the electrons wcre endowed with an effectivc mass m, which may be larger or smaller than thp frec electron mass, or may even be negative. Electrons in crystals respolh J applied fields as if (mdowed with ncgati\'c or positive

116

charges. -e or +e. and herein lies the explanation of the negative and positive values of the Hall coefficient.

NEARLY FREE EU:GmON MODEL

On the frcc c1crtron model the allowed energy \'alucs arc distributed essentially continuously from zero to infinity. We saw ill Chapter 6 that
Ifl.:=-(~+~+~

h'

2m

(I)

where. for periodic boundary renditions over a cube of side L,

+-

2w
L

+4w _.

....

(2)

TIle frcc electron \wvcftmctiOllS arc of the form

""'-(r) = cxp(ik . r) ;

(3)

they represent running wavcs and carr')' momentllm p "" hk. TIle band structure of a crystal can often be explained by the nearly frcc electron model for which the band electrons arc treated as perturbed only weakly by the periodic potential of the ion cores. This model answers almost all the qUlllitativc questions about the bclwvior of c1cctrom in metals. We know that Bragg rcf1cetion is a dlaractcristic feature of wave propagation in crystals. Bragg rcncction of electron wavcs in crystals is the cause of energy gaps. (At Bragg reflection wa\'clike solutions of the SchrOdingcr equation do not exist, as in Fig. 2.) These energy gaps are ofdccisi\'e significance in determining whether a solid is an insulator or a conductor. We explain physically thc origin of energy gaps in the simplc problcm of a linear solid of latticc constant a. Thc low energy portions of the band stTucturc are shown qualitath'e1y in Fig. 2, in (a) for entirely frcc electrons and in (b) for electrons that are nearly free, but with an energy gap at k = TrIa. The Bragg condition (I.: + Gf = Jil for diffraction ofa wave ofwavevcctor I.: becomes in one dimension

k= iC= WrrlD,

(4)

where G = 2Trnla is a reciprocallattkc \"Cd:or and n is an integer. The first rencctions and the first energy gap occur at k "" ::!: ma. The region in I.: space between -TrIa and TrIa is the fint Brillouin wile of this lattice. Other energy gaps occur for other values of the integer 11. The wavefullctiOlls at k::; -rria arc not the travcling wavcs exp(im:/a) or cxp(-hrxla) offrce electrons. At these special values of k the wavefunetions arc

1 ..", ......

Mllo-o,<1
_l~~
~'urbidd. n

,-,
boon,!

..... ......
...51

"
:' ,
,
__ t;.

", ,

, , ,
,

I',

,.,

..-

t'igure 2 (a) !'lot of energy ~ ,'crsus wa,..,vt:dor k for 1I f,ee d ......'lron. (h) 1'101 of energy "'<-onu. "'ll,..,,-edor i:lr an c\ectron in a monatomk linear lattice ofbltk.., ronslant II. nle enet19' W'p E" shcM'n is ;USOC'iated wit1l the first Bragg rdlectim "t k .. :!: mtl; other g3pS an, blndat ::!:r1.1:ltJ, lOrintegral ,0lIUCS or It.

made lip of equal parts of wavcs traveling to the right and 10 the left. When the Bragg reflection condition k = ,"Ia is satisfied by the wavcvector, a wave

traveling to the right is Bragg-reflected to travel to the left, and vice versa. Each subscqUCflt Bragg reflection will reverse the direction of tra\d of the
\\'3\'e. A wave that tra\cls neither to the right nor to the left is a standing wave: it doesn't go anrwhcrc. The time-independent state is reprcscnted by standing waves. We can fOim two different standing waves from the two traveling waves cxp( imla).
namely

l/I(+) = cxp(i1rX/a)

+ CXP(-im111) =

2 COS (rrna) ;

l,!.{-) = cxp(im1a) - cxp(-im1a) = 2i sin (r.:tIQ) .

(5)

TIle standing waxC'i arc labeled (+) or (-) according to whether or not thcy change sign when -x is suhstitllted for x. Both standing waves arc composed of equal parts of right- and left-directed traveling wavcs.

Origin of lite Energy Cap

The two standing wavc!" l,!.{+) and l,!.{-) pilc up electrons at different rcand therefore the two waves have different values d the potential energy. This is the origin of the energy gap. The probabilit)' density p of a particle is I/J*l/J = 1l/Jl2 For a pure traveling wave cxP(ikx), wc havc p = cxP{-ikx) cxP(ikx) = 1, so that the charge density is constant. "111e charge density is not constant for linear combinations of plane wavcs. Consider the standing wave l,!.{+) in (5); for this wc havc
giollS,

A.+) = 1l,!.{+H2 oc cos 2 wxla .

This function piles up clectrons (ncgati\'c charge) on the positive ions centered at x = 0, G, 20,. in Fig. 3, where the potential encrgy is lowest.

".
Figure 3a pictures the variation of the electrostatic potential Cflcrgy c:1 a ronductiml electron in the rlCld of the positive ion cores. The ion cores bear a nct positi\c charge because the atoms arc ionized in the metal. with the valence electrons taken off to form the conduction band. The potential ellergy of an electron in the Held of a positive ion is ncgutivc, so that the force bclw(,>C1l them is attractive. For the other standing wave IJ.{-) the probability density is

p(-) = 1~_~2 ex sin 2 mJa

which concentrates electrons away from the ion cores. In Fig. 3b we show the electron concentration for tllC standing waves 1/1(+), !JI(-), alld for a traveling
\\';1\"C.

When we calculate the average or expectation values of the potential energy over these thrcc charge distributions, we Gnd that t)le potential energy of

,,)
Figure 3 (a) Variation of potential energy of a conduction clt,<:tron in the neld of the ion cores of a linear Ialtia:. (b) Di5hi!.>utioo of probability density p In the lattice for lll(-H~ '" 5in" =fo; 11l<+Jf" '" ros~ miG. and rnr a tra"eling ""..,'t. Th.e ,........Junction 11(+) piles up elcd:ronk charge 00 the Cll1"n of the.- posit;,,, ionJ. tht.'reby Joy,,,nng the p;oteflti:al energy in comparison "ith the ..-enge poteolial o.'I'"IeI"KY seen br a lra,ding "",,,. n,., ".;.krunction \1(-) piles up ~ ,n the repoo ~ the i005. thereby raising the potefltial energy In comparison with that seen by a Int\-eling "~. This figun: 15 thl- h:y to un<kntand.ing the ongin of the enol g;ip.

7 Energy

&~

17:

p(+) is lower than that of the traveling wa\'e, whereas the potential energy of p(-) is higber than the trn\"eling wave. We ha"e an energy gap ofwidth 1: if~he encrg;es of p(-) and P(+) differ by Ell' Just below the energy gap at points A in Fig. 2 the wavcfunction is lj.o(+). and just abo\'C the gap at points B the wa"cfunction is lj.o(-).

Magnitude of the Energy Gap

Thc wavcfllnctions at the Brillouin zonc boundary k == 7T/a are 'lrX/lJ and V2 sill 'TTXia, normalized ovcr unit Icngth oflinc. We writc the potcntial encrgy of an electron in the crystal at point x as

v'2 cos

U(2') = U cos 2'TTXia .

nle 6rst-order energy differCllCC bctwCCl1 thc

E.

tv."O

standing wavc states is

-1' ,
= 2

dx U(x)

11,,+X' - I"- i~

dx U ros(2m/o) (cos2 'TTX!a - sin 2 m:/a) = U

Wc sec that the gap is equal to the FOllrier componcnt of the crystal potential.

BLOCH FUNcnONS

F. Bloch proved the import:'lnt theorem that thc soluHons of tile Schrodinger equation for a periodic potcntial mllst be of a special form:

IlfIJ.(r) = 1I~(r) cxp(ik . r) _I

wherc Il~(r) has thc period of thc crystal lattice with result (7) expresses thc Bloch theorem:
ll~(r)

(7)
=
1I~(r

+ 1'). Thc

The eigenfunctions of the wavc equation for a periodic potential are thc product of a pIanc wa\'c exp(ik r) times a function utJr) willI the periodicity of the crystal lattice. is called a Bloch function and A one-electron w"J.'efunction of thc form can be decomposed into a sum of tra"c!ing wavcs, as wc sec later. Bloch functions can be asscmbk-d into localized wave packets to represent electrons that propagate frecly through the potcntial field of the ion cores. Wc give now a restricted proof of the Bloch theorem, valid whcn 1/1", is nondcgcnerate. TImt is, whcn therc is no othcr wavefunetion with the samc energy and wa) X1.or as 1/1",. nlc gcneral casc \\;11 be treated latcr. We con-

"'I
sider N identical lattice points 011 a ring of Icnb'th Na. The potential energy is periodic in a, with U(1") ~ U(x + sa), where s is an integer. We arc l.,'llidcd by tile symmetry of the ring to look for solutions ofthcw3.\'c equation such that
~x

+ a) =

C,p(x)
OI'lCC

(8)

where C is a constant. Then. on going

I/J(x

around the ring,

,

+ Nu) =

I/!{x) ""

eN l]i(x)

because l/J(x) must be single-valued. II follows that C is one of the N roots of unity. or

c=
We sec thai

cxp(i2mtN) ;

:J

0, 1, 2, ... N - 1 .

(9)

tP(x) = llJJX)

cJ(ri.i21T~/NlJ)

(10)

satisfies (8), provided that Uk(X) has the periodicity a, so that u/,;(x) :::: uiJx + a). With k = 2'fU1Na. we have the Bloch result (7). For another derivation, sec

(29).

KRONIG-PENNEY MODEL

A periodic potential fOl'" which the ....1WC equation can be solved in terms of clclT\Clltary functions is the square-well array of Fig. 4. The wave equation is

h 2 cFIjI
- 2m

(lXl +

U(x}1;l = f:1jJ

(II)

where U(x) is the potcntial energy and t" is the energy eigenvaluc. In thc region 0 < x < a in which U "" 0, the cigcnfunction is a linear combination,
'" = A~ + Be-IK>.
(12)

of plane wavcs traveling to the right and to the left, with energy
t" ""

fl'lK'l12m .

(13)

In thc region -b <

< 0 within the barrier the solution is of thc form

"':: CeQ>.

+ De-Q>.

(14)

with
Va t"

= h'lfJ'l12rn .

(15)

181

I~
, ,_ _ u.
Figure" ~ I periodic potmtial as inlroduced b). Kronig and Pmnq'.

-{,o

+ "l -b 0

.... + "

,-

We want the complete solutioll to have the Bloch form (7). 11ms the solution in the region Il < x < (l + b must be related to the solution (14) in the region - b < x < 0 by the Bloch theorem:
ljJ(a<;r<a+b)""-b<::t<O)~Ho).

(16)

which servcs to define the W3vcvedor k used as an index to Iabclthc solution. The constants A, B, C, D arc chosen so that l/J and dl/Hdx aTe continuous at x:: 0 and x '" a. These arc the usual quantum mechanical boundary conditions in problems that involve square potential wells. At x IE 0,

A+8=C+D;
IK(A - B) = Q(C -D).

(17)

(18)

At:r::: a, with the usc of (16) for !M,a) under the barrier in tenus of ljJ(-b).
AtKa

+ Be- INJ :: (Ce- Qb + DeQ"J el 1(4+1

s::

(19)
(20)

iK(A~ - Be-I":")

Q(Ce- Qb

DeQ

'1 e'l(4+b)

The four equations (17) to (20) ha...c a solution only if the determinant of

the coefficients of A, B, C. D vanishes. or if

leQ! -

k~2QKl sinh

Qb sin Ko + cosh Qb cos Ko "" cos k(a + b) .(21a)

It is rather tedious to obtain this cquation. The result is simplified if we represent the potential by the periodie delta function obtained when we pass to the limit b '" 0 and Uo = 00 in such a way that Q200/2 = p. a finite quantity. In this limil Q J> K and Qb <s I. Then (2la)
~~W
I

(PIKn)sin Ko + CCfi Ka "" cos kn .

(21h)

The I1Ulges of K fOf which this cquation has solutions arc plotted in Fig. 5, for the case P ... 37(/2. The corresponding values of the energy are plolted in Fig. 6. Note the energy ~ps al the zone bo\mdarics. "Inc wavcvcctor k of the Bloeh functioo is the important index. not the K in (12). which is related to the energy by (13). A treatment of lhis problem in \\'3.\C\'cetor space is given later in this chaplcr,

".

Figure S Plot of the function (PlK.o) sin K4 + COl Ktz. ro.. P - 3wf2,. The ..IJo,o.~ ..JU6 of the energy f an gi\~ b ~ ranges of K.J - (2m.Ift~lII" lOr wtuch the lUnction lies bet..~ :!:.l. For otJw:r \UJn of the energy there ~ no l",,"ding .......1l or I3kdl-lib soNlions 10 the ....,..,"e equatiool. so lhal fOrbidden lPP in the mergy $pedrum lin: b'rnnl

"'.--~-~-~----,

~
0

"

> ii " ., ,

Figure 6 Plot or L11Crgy "$. wa\'l,."Il11mber

the energy gaps at kit - 'IT, 271', 3'/1' . . .

lOt the

Kronlg.rcnllL' potential, with I' _ 3.../2. Notice

Ennp~

163

WAVE EQUATION OF LECffiON IN A PERIODIC POTENTIAL

\\'c considcred in Fig. 3 thc approximate fonn wc expcd: for the solution of the Sctuooinger equation if the W3\'CVector is at a zone boundary, as at k = .,,{o. Wc treat in dctail thc wavc equation for a !,'CTIcral potcntial, at gencral values of k. Let V(%) dcnote thc potential energy ofan clectron in a linear lattice oflattice constant a. We know that the potential energy is invariant under a crystal lattice tmnslation: U(%) = U(% + a). A function invariant undcr a crystallatticc translation may be exp.1.nded as a FOllrier serIes in the rcciprocallattice vectors G. We write the Fouricr series for the potential energy as
U(x) ~ ~ Uc c

to> .

(22)

The values of the coefficients Uc for actual crystal potentials tcnd to dccrcasc rapidly with increasing magnitudc of G. For a barc coulomb potential Uc decreases as )jGt. We ",,-ant thc potential encrgy V(%) to be a real function:
V(%)
=

2:
c>o

VcJ.efCz + e- JC"),.. 22: V c cos G% .

c>o

(23)

For CO!wcnicncc wc ha\"C assumed that thc crystal is symmetric about % "'" 0 and that Vo = o. The wave equation ofan clectron in the crystal is:Jfl/J = El/J, whcre:Jf is the hamiltonian and E is the energy eigenvalue. The .solutions l/J arc called cigenfunctions or orbitals or Bloch functions. Explicitly, the wavc equation is

(-'-,)2 + U(%)) 1/1(%) = (_I_p2 + 2:

2m 2m c

Uc e 1G2) 1/1(%) = E!Ji(%)

(24)

Equation (24) is written in the onc-ck'CtrQn approximation in which thc orbital l/!{x) describes the motion ofonc electron in thc potential ofthc ion lo'{)(CS and in the avcragc potcntial of the othcr conduction electrons. The wavefunction 1/1(%) may be cxpressed as a Fourier series summcd over all values of thc wavc\"(x:tor pcnnittcd by the boundary conditions. so that
(25)

where k is rcal. (We could equally well writc the index k as a subscript Oil C. as in C".) 1bc set of values of k has the fonn 2mJL, because thesc values satisfy periodic boundary conditions over )cnt.'lh L Herc n is any integer. positivc or negativc. We do .. _f assulllC, nor is it gcncrnlly true. that lJi(x) itself is periodic

...
in the fundamentallatttce translation o. The translational properties of ~%) are detcnnined by the Bloch theorem (7). Not all wavevectors of the set 21rnlL enter the Fourier c~pansion of any one Bloch function. If one particular wavevector k is contained in a lb. then all other wavevectors in the Fourier expansion of this I/J will have the Conn k + G, where G is any reciprocal lattice vector. We prove this result in (29) beLo\\'. We can label a wavefunction '" that contains a component k as !/II: or. equally well, as 1jJ,,+c, because if k enten the Fourier expansion then k + G may enter. 'OIC waveved:ors k + G running over G arc a restricted subset of the set 2mJL, as shown in Fig. 7. We shall usually choose as a label br the Bloch fWlCtion that k ",'hich lies within the fin.t Brillouin zone. \Vhen other conventions are used. we shall say so. This situation differs from the phonon problem, where there are no components ofthe Ion motion outside the first zone. 'nle electron problem is like thc x-ray dif&actjon problem because the electromagnetic field exists everywhere within the crystal and not only at the ions. To soh<e the ",..ave equation. substitute (25) in (24) to obtain a set of linear algebraic equations for the Fourier coefficients. The kinetic energy term is
1 d)2 oIi.,) ~ -p'oJ-{r) ~ - ' ( -'h2m 2m dx

l/f ~ -L fit -h-dk'C(k)'" ;

2m dr 2m
It

and the potential energy term is

Ue "e,) "x) = L L Ue e""C(k) ...' (L celt

11Jc wave equation is obtained as the sum:

f,' L -k'c(k) ... + L L UeC(k) ",He), =.L It 2m c It It

Each Fourier component must equation. Thus
ha\'C

C(k) ".. .

(26)

the same coelflCient on both skies of the

I
\\.-ith the notation

(AI.- - E}C(k)

Ur;C(k - G);; ()

(27)

(28)

Equation (27) is a useful form of the "1I.\'e equation in a periodic lattice. although unfamiliar becausc a sct of algebraic equations has taken the pl:ace of

...,rgy 8.lIIw

185

.",,_ 2..

...
-10 0 10

.J;" + ~ ..

-30

f....1 f.f .T.-...... f.....ff...f

-20

30

I. 100 enih. 2r I,
Fig,,",l The kl\\'Cr poinb ~ t ,-aJoeo. dthc "1I'C'cdor I; .. 2m>JL allowed by the periodic 00t".dary ron<!;tion On the WlI.,"functioll on:r. nn~ of circumference L romj,os(:d of20 primit;"c (..,II~. TI", allowed "allll'l COf,tlnlle to ~ .... The "Pi,.,r poinu reprclCnt the flut few ",a"e\'(~cto1"S whieh nl:l)' enter into the t'OI,ri"r C1Jlomlon of a wa...:fundtOIl tJ!(x), ~tarllllg from a llarticular ...a.e.'edor k - ko .. -8(2ttIL). The shortest reciprocal lattltt "ector ll; 2../" '" 20(2...11.).

the usual differential equation. The set appears unpleasant and formidable because there are, in principle, an infinite number of C(k - C) to be determined. In practice a small number will often suffice, perhaps t",'O or four. It takes some experience to appreciate the practical advantages of the algebraic approach.
Restatemmt of tlte Bloclt Tlteorem
Once we determine the C's from (27), the Wllvefunctioo (25) is given as

01>(,) ~ which may be rearranged as

L C(k C

G)

"..-c.

(29)

with the definition ut<%)

L C(k C

C) e-IC~ .

Because 1IJJX) is a Fourier series o'\'Cr the reciprocal lattice vectors, it is invariant under a crystal lattice translation T, so that tlJ;(%) = 1IJ:(% + We verify this directly by evaluating 1I1{% +

n:

n.

UI{1' +

n=

}:

C(k - C)

e-IC(~+1") ;::: e-ICT[~

C(k - C) e-K:Xj =

e- ICT

1/J:{1')

Because exp(-iGT) "" I by (2.17), it follows that u~% + T) = Il~(X), therebyestablishing the periodicity of UJ:. TIlis is an alternate and exact proof of the Bloch theorem and is \'3lid even when the l/J~ are degenerate.

'86
Crystal Momentum of an Electron
What is the significance of the wavevector k used to label the Bloch function? It bas several properties:

Under a crystal lattice translation which carries r to r

+T

we have
(30)

l/Jk(r

+ 1')

= elk ' T

e 1k.r

uldT + 1') =

tfI"T rh(r)

because u.Jr + 1') = llk(r). Thus exp(ik . 1') is the phase factor l by YJhich a Bloch function is multiplied when we make a crystal lattice translation T.

If the lattice potential vanishes, the central equation (27) reduces to (A" - t"}C(k) = 0, so that all C(k - G) are zero except C(k), and thus tliJr)
is constant. We have l/Jk(r) =

eft'.r.

just as for a free electron. (This assumes

we have had the foresight to pick the "right" k as the label. For many purposes other choices of k, differing by a reciprocal lattice vector, will be
more convenient.) The quantity k enters in the conservation laws that govern collision processes in crystals. (The conservation laws are really selection rules for transitions.) Thus fill; is called the crystal momentum of an electron. If an electron k absorbs in a collision a phonon of wavevector q, the selection rule is Il; + q = k' + G. In this process the electron is scattered from a state k to a state Il;', with G a reciprocal lattice vector. Any arbitrariness in labeling the Bloch functions can be absorbed in the G without changing the physics of the process.

Solution of the Central Equation

Equation (27) may be called the central equation:
(A, - .)C(k)

+L
c

UGC(k - G) ~ 0

(31)

represents a set of simultaneous linear equations that connect the coefficients C(k - G) for all reciprocal lattice vectors G. It is a set because there are as many equations as there are coefficients C. 'nlese equations are consistent if the determinant of the coefficients vanishes. Let us write out the equations for an explicit problem. We let g denote the shortest C. We suppose that the potential energy U(x) contains only a single
'We may also say that exp(l1< T) i~ the eig"'l\'alue oflhe~laltramlalionoperation Y, and >h is the ..~genvector. That is, Yth(x) = th(:r + T) .. e>l1'(l1<' T}oh(.t), so Ihat 1< il; uuilable 1abe1 for the eigenvalue. Here we h""e used the Bloch theorem.

Fourier component Uc = U- co denoted by U. TIlen a block of the determinant of the coclncients is given by:
),. .... ~ E

o o o

U ),.._,; U 0 0

0 U
),..-E

0
U
),.1<+11 - E

0
0
(32)
E

U
0

U
),..... 2g

Th see this. write out five successive equations of the set (31). The determinant

in principle is infinite in extent, but it will oftcn be sufficicnt to set equal to 7.ero the portion we have shown. At a given k. each root E or EI< lies on a different energy band, c.xcept in case of coincidence. 111C solution of the detenninant (32) gives a set of encrgy eigenvalues En", whcre n is an inde~ for ordering the energies and k is the \\o"a,vevector that labels C". Most often k will be taken in the first zone, to reduce possible confusion in the labeling. If we chose a k different from the original by some reciprocal lattice vedor. we y.-ould have obtained the same set of equations in a different order-hut having the same energy sped.rom.
Kronig.Penney Mockl in Reciprocal space'

As an example of the use of the centTlll equation (31) for a problem that is exactly solvable, we use the Kronig-Penney model of a periodic delta-function potential: U(x) = 2

c>o

Uc cos Gx =

AaL 8(x -

sa)

(33)

where A is a constant and a the lattice spacing. The sum is over all integers s between 0 and l/a. The boundary conditions are periodic over a ring of unit length. which means over I/a atoms. l1\Us the Fourier coefficients ofthc potentia! arc Uc

dx U(x) cos Gx

= Aa~

dx 6(1" - sa) cos Gx

(34)

= AaL cos Gsa = A

SoJ'hi.

tn::at~1 ,...u s~ed by SUl'jil Singh, Am.

J.

PIIys. SI, 179 (1963).

'88

We write the central equation with k as the Bloch index. Thus (31) becomes
(A, - .)e(k)

+ A2: "

C(k - .=010) = 0 ,
fl.

(35)

where Ale - f1 2 k 2121n and the sum is over all integers for (k). We define

We want to solve (35)

I(k) - 2: C(k - .=010) ,

(36)

"
so that (35) becomes

(k)
C

=_ ('mAlh')f(k)

k' _ ('booh~

(37)

Because the sum (36) is

O\<eT

all coefficients C. we ha\'e. for any n,

(38)

I(k) = I(k - ,=01,) .

'nlis relation lets us write

C(k - ,=01,) = -('mAlh~I(k)[(k - 'mVo'f - ('m"h~r'

(39)

We sum both sides over all n to obtain. using (36) and canoellingf(k) from both sides.

(h'l2mA) = -2: I(k -

'=o1'f - ('m"h~J-'

(40)

"
The sum can be carried out with the help of the standard relation

ctnx=L

" "11+%

(41)

After trigonometric manipulations in which \','e use relations for the difference of hvo cotangents and the product of two sines, the sum in (40) becomes

tf sin KD 4&(005 lea - cos KG)

where we write I('l "" 2meflt. as in (13). The final result for (40) is

(4')

(mA a t.12h 2 XKa)-1 sin KD + cos KD = cos ko

(43)

which agrees with the Kronig-Penney result ('lIb) with Pwritten for mAtfl2ll2
Empty
Latti~

Approrimation

Actual band structures are usually exhibited as plots of energy versus waveved.or in the first Brillouin zone. When wavevectors happen to be given outside the first zone. they are carried bad into the Hrst zone by subtracting a

7 Energy Band,

lli9

suitable reciprocal lattice vector. Seh a translation can always be found. 1be operation is helpful in visualization and economical of graph paper. When band energies are approximated fairly 'well by free electron energies ~ :: f1 t kt /2JIl, it is advisable to start a calculation by carrying the free electron ellcrgies back into the first zone. 111e procedure is simple enough once you get tIle hang of it. We look for a C such that a k' in the 6rst zone satisfies

k'+C=k.
where k is unrestricted and is the troe free electron w:wevector in the empty
Iatlk:c. (Once the plane wave is modulated by the lattice, there is no single "true~ W3\'C\'ector fur the state l)/.) If we drop the prime on k' as unnecessary baggage. the free electron

energy can always be written as

E(k..,kv,k~) = W"l2m)(k + C)t

= W"l2m) (k... +

GJ' + (kl/ + GI/'f + (k". + G~'f1

with k in the first zone and C allowed to ron over the appropriate redprocal lattice points. We consider as an example the low-lying free electron bands of a simple cubic lattice. Suppose we want to exhibit the energy as a function ofk in the [100) direction. For convenience, choose units sudl tbat 11 212t1l = I. We show severallow-Iylng bands in this empty lattice approximation with their energies E(OOO) at k ;;;;; 0 and (k)X) along the k.. axis in the 6rst zone:
Bond
1

CoJ2.

.1000)
0 (2mo)1 (2#/0)1 21.,2mof 2(2#10)2 2(2#10)1

011:,00)

2,3 4,5,6,7 8,9,10,11 12,13,14,15 16,17,18.19

000 100,100 0I0,OIO,~1,00! 110,101,110,101 llO,IOI,TIO,"iOI OII,OII,OlI,OU

.'

(1:, :t: 21110'f

+ (21110'f (k, + 2,",0)1 + (2#/0)1 (k, - 2111a)1 + (2#lo'f I.{ + 2(2mo)1

1::

These free electron bands are plotted in Fig. 8. It is a good exercise to plot the same bands for k parallel to the (JlI) direction ofW3ve\'cctor space.

Approximtlte Solutiotl Near a 7.one Boundary

We suppose that the Fourier coInponents Uc of the potential ener!,')' are sllJall in comparison with the kinetic energy of a free electron at the zone boundary. We first consider a wave\'ector exactly al the ZOlle boundary at iG, that is. at TrIo. Here
(k - G)' - (IG - G)' = (lG)'

".

Figure 8 Low-1ring (ree eleclron mergy bands of the.""pty K !altice, as InmIortned 10 the flrSl BriI louin zone and j>lolled \'5. (k.ool. 1hI: free dertrou enngy u A'(Ie + Cjlrlm, "ile the C's are ghen in the JoeaJnd column c;( the t2bk. The bo&d cur-.-e an:

'

--'''''_-;,-_''''--__,!
0 ~_

'i"

in the fin! Brillouin zone. "ilh -#fa'" I:" so mo.. Enrrgy bands d ........ in this "'<I)' are 5llld 10 be in the ,educed zone adleme.

so that at the zone boundary the Idnetic energy of the two component v.'3ves k:::: :t: Ie arc equal. Jr cue) is an important coefficient in the orbital (29) at the zone boundary. then C(-IG) is also an important coefficient. This result also follows from the discussion of(5). We retain only those equations in the central equation that contain both coefficients C(iG) and C(-iG). and neglect all other coefficients. One equatton of (31) becomes, \\"ith k :::: iG and A. - !l~jG)2/2m.

(A - .)C(IC) + UC(-IC) = 0
Another equation of (31) becomes

(44)

(A - .)C(-IC) + UC(IC) = 0

(45)

TIlese two equations have nontrivial solutions for the two coefficients if the energy E satisfies
U [_ 0
), E

(46)

Energy Bond.

1111

whence,
=

A::t U= _(!G)2 V. 2m

h'

(47)

The energy has two roots, one lower than the free electron kinetic energy by V. and one higher by V. Thus the potential energy 2V cos Gx has created an energy gap 2V at the zone boundary. The ratio of the C's may be found from either (44) or (45): C(-IG) C(IG)

<- A -U- = 1 ,

(48)

where the last step uses (47). nlUS the Fourier expansion of IjJ(x) at the zone boundary has the two solutions
t{J(x) = exp(iGx/2)::t exp(-iG112) .

These orbitals are identical to (5). One solution gives the wavefunction at the bottom of the eoergy gap; the other gives the wavefunction at the top of the gap. Which solution has the lower energy depends on the sign of V. We now solve for orbitals with wavcvector k near the zone boundary iG. We use the same two-eomponent approximation, now with a wa,efunction of the form IjJ(x) = C(k) e lh + C(k - G) ei(k-G)~ (49) As directed by the central equation (31), we solve the pair of equations
(A, - <)C(k)

(Ak-G 2

+ UC(k - G) = 0 )C(k - G) + UC(k) =

with Ak defined as h J(l/2m. These equations have a solution if the energy satisfies

Ak -

C -

-0

whence ~ - (Ak-G + A.tJ + Ak_CAk - V 2 = The energy has two roots:

o.
(50)

and each root describes an energy band, plotted in Fig. 9. It is convenient to expand the energy in terms of a quantity j( (the mark over the j( is called a tilde), which measures the difference k !!!! k - iC in wavevector between k and the zone boundary:
k"'(fi2/2m)(lG 2 +

i<2'j ::t [4A(fi 2k 212m) + lPl'f2

"., (h 2/2m)(lG 2 +

k 2)::t

V(1

+ 2(AllP)(1i2j( 2/2m)]

(51)

in the region fi2Cltf2m 4

lvi.

Here A = (fi 2/2mHiCt as before.

".
,

"

.-

~ <.' , ,
(to,)

, , ,

,.,

"
n-.e

Figllre 9 SoIulioos of(:w:J) in lhe paiodlC lOne Idw:me. in lilt' region neat. boundary oflh~ fIrSt Brillouin zone. The units 8", such lhal U - -0.45; C - 2, and ARfm - I. frft eleclron euooe is drAwn IOroompariJon. The energy 8"P alcho: wile boundary 1s0.90. The ...a1"eolU lla!! deli~ralely 1><....... o:hoJen large fOl'" this lIIuSIr.Mion, too large for t1~ ~Ienn a''1>ro>;imalion 10 be ."""rate.

Writing the two zone boundary roots of(47) as E(), we may write (5J) as

E;J) = f()

+ ~ I:!: U

"K'( 2A) .

(52)

These are the roots for the energy when the \\'8vevcctor is very close to the zone boundary at iG. Note the quadratic dependence of the energy on the waveveetor K. For U negative, the solution f(-) corresponds to the upper of the two bands, and E(+) to the lower oflhe tv.u bands. 11lC '''''u c's are plotted in Fig. 10.

NUMBER 01" ORBITALS IN A BAND

Consider a linear crystal constrncted of an even number N of primitive cells of lattice cOIulant a. In order to count states we apply periodic boundary conditions to the wavefunctions over the length of the crystal. The allowed \'Slues of the electron wavevector k in the first Brillouin zone arc given by (2):

k=O;

+-

2w L

+_.

4w
L

...

Nw
L

(53)

We cui the series off at Nft/L ~ mao for this is the zone boundary. The point -N1tIL = -ma is not 10 be counted lU an independent point because it is

1 ""'D' Borttb

193

"C----------t , - - - - -

....
Fin!

,
nNr

FISUft 10 Ratio olthe metlicients in ~ - C{i) ap(ib) i" C(k - C) ~i(k- C)rjasoakubted the boulllCbry ol the finl Brillouin Wne.. One o::ornponent dominales as mo.,e ...'a)' From die

...-

...
iC

-=

.........",.

oonnected by a rcciprocallattice veclor with m(J. l1le total number of points is exactly N, the number of primitive cells. Each primitive cell contributes exactly one independent value of k to each energy band. This result carries over into three dimensions. With account taken of the t.....o Independent orientations of the electron spin, there are 2N independent orbitals in each energy band. If there is a single alom of valence one in each primitive cell, the band can be half filled with electrons. If each atom contributes two valence electrons to the band, the band can be exactly filled. If there are two atoms of valence one In each primitive cell, the ~nd can also be exactly filled.

Metab and

',~ulaton

(f the valence electrons exactly fill one or more bands, leaving others empty, the crystal will be an insulator. An external electric field will not cause current flow in an insulator. (We suppose that the electric field is not strong enough to disrupt the electrotlic structure.) fro,.ided that a filled band is separated by an energy gap from the next higher band. there is no continuous way to change the total momentum of the electrons if every accessible state is filled. Nothing changes when the field is applied. 111ls is quite unlil--e the situation mr free electrons for which L: increases uniformly in a field (Chapter 6).

'94

,__

11: jj

k_
(b)

'i"

k_
(c)

iI

If

{a}

F'Uft II Occupied WU~ and b-nd ~ructures l'J"i,,& (a) an insw.lQI". (bl metal 01' semimeta1 brouse rlb.nd O'-..,Np. and (ela metal beau.se 01ftedron ~tl1ltioo. tn (bl the o.... ,rbp nn'd not occur IIIong tbe_ di~ in the Brillouin 'O)nl,!. 1ft'" O\"er!ap iJ ~mI1I. "ith relati~'dy lew lUles in~'tJt,.-ed, """ .peak of. semimetlll.

A crystal can be an insulator only if the number of valence electrons in a primitive cell of the crystal is an even integer. (An exception must be made for electrons in tightly bound inner shells which cannot be treated by band theory.) If a crystal has an even number of valence electrons per primitive cell, it is necessary to consider whether or not the bands overlap in energy. IT the bands overlap in energy. then instead of one filled band giving an insulator. we can have tv.a partly filllirbands giving a metal (Fig. 11). The alkali metals and the noblc metals have one valence electron per primitive cell. so that they have to be metals. The alkaline earth metals have two valence electrons per primitive cell; they could be insulators, but the bands overlap in energy to give metals, but not very good metals. Diamond, silicon, and gennanium each have two atoms of valence four, so that there are eight valence electrons per primitive eell; the bands do not overlap, and the pure crystals are insulators at absolute zero.

7 Energy Bml/t.

195

SVMMARY ne sohltions of the wave equation in a periooic lattice arc of the Bloch form l/Jk{r) = c ite . r u~{r), where uJr) is invariant undcr a crystal lattice translation. nerc are regions of energy for which no Bloch function solutions of the wave equation exist. These energies form forbidden regions in which the wave-functions arc damped in space and the values of the k's arc complex, as pictured in Fig. 12. 111e existence of forbidden regions of energy is prerequisite to the existence of insulators. Energy bands may often be approximated by one or tv.'O plane waves; for example, t/JJ,x) E!i C(k)rJ h + C(k - C)el{k-G)x ncar the wne boundary at iC. The number of orbitals in a band is 2N, where N is the number of primitive cells in the specimen.

Problems
I. Square lalike. free electron elwrgiet. (a) Show lOr a simple SQuare lattice (two di-

mensions) that the kinetic energy d" a free electron at a <."Orner of the 6rst zone is higher than that of an electron at midpoint of a side face of the wne by a factor of2. (b) what is the (.'Orresponding factor for a simple cubic lattice (three dimensions)? (c) What bearing might the result of(b) have on the conductivity of divalent metals?

2. Free electron energies in reduced wne. Consider the free electIOn energy bnnds of an fcc crystallallice in the approximation of an empty lattice, but in the reduced zone scheme in whidl all Ie's are transformed to lie in tile 6rst Brillouin 7.one. Plot roughly in the {Ill] direction the energies of all bands up to six times the lowest band energy at the zone boundary at k = (2-m'a)(I,l,J). Let this be the unit of energy. TIlis problem shows why band edges need not necessarily be at the zone centcr. SC\'cml of the degenel'3cies (band crossings) will be remowd when account is taken of the crystal potential.
3. Kronig.Pemtey mooel. (a) For the deltafunction potential and with P ~ 1. find at k "" 0 the energy of the Jov,'cst energy band. (b) 'or the same problem find the band gap at k "" 7fla.

4. Potenti4l energy in the dianwnd .!tructure. (a) show that for the diamond structure the 'OlIrier component U<; of the cry:o;tal potential seen by all electron is equal to zcro for G = 2A, where A is a basis \ector in the reciprocal lattice referred to the corl\"entional cubic cell. (b) show thai in the usual fint--order approximation to the solutions oftbe W",)\'e equation in a periodic lattice the energy gap \'lllli~hes at the zone boundary plane normal to the end of the vedor A.

'06

; 025

'"~

----1 1 ~ _----+-tiii-:-----~)
......,,\
_ 0

0.1Ill.5

0.01

lmopwy .,.n d tIG

... '=-~O::::=-~~----~----_,',~---~=_ 0.48 Q.f9 o.SO 0.51 052

IIatI par1 <i ItJC
FIpI", 12 In the energy pp the exist solutions oflhe _
wa''e'-..cI.or-. At the boundary mtbr fil"$t ZIOnt'
Ilfl

~ion

b- compleJ< VllIues d

thr;

the

~ putofthe~tttorb

ie. The imaginary

put 011: in the gi1p is pkltted in the approll.imalioll 01 two plane _ a , or U - 0.01 A"G~I'Jm. III infinite unbounded crystal Ii., .... _"eVt'dOr must boo neal. 01' else the amplitude will increase without limit. But on. ~ or at ajuJldioo there can mst soIutiolll .... ith complell ~l:ctur.

S. Complex lCGt:ll!vrctor. in 1M mugy gop. Find an expreuion for lhe imaginary part of the mwe"edor in the energy gap at the boundary of the first Brillouin :rone. in the approximation that led to Eq. (46). Ci,'e the result for the Im(le) at the center of the energy gap. The result for small Im(k) is
(h~f2rn)(Im(k)J2

_ 2mU'IfI'C I

The fonn as plotted in Fig. 12 is of importan<:e in the Ihool")' of Zener tunneling from one band 10 another in the presence of a strong electric neld.

6. Square lattice. Consider a s(luare lattice in Iwo dimensions with the crystal potential
U(x,y) - -4U cos(21tt/a) cos(2'11TJla) .

Apply the central equation to find approximately the energy gap at thc comer point <'TrIa, 'TrIa) of the Brillouin wne. It will sulftee to 50h'e a 2 x 2 determinantal equation.

References
J.
M. liman, Principlu of,he ,IKorv of wlid,. Cambndgt:, 2nd ai,. J97'2. N. F. Moll and H. Jonte'"<. T~,..", of'~ prope:rtiu of md~ and allOtfl. OxfOrd, 1936. (Dm...,r

pllpeI"baek

reprint.)
ml

(Further nlerencn

band theory an- glwn II lhe end of Chapter 9.)

8
Semiconductor Crystals
BAND GAP EQUATJOJ'llS OF MOTION PhYIk:a1 derivation of iiI: ::::: F
Hm~

199 203 205

200
209 210
it:;!

Effecth-e m&S5 Ph)'Sicai interp~talion or the effective mllSl Effective ITlllSSeS in semiconductors Silicon and germanium

214

INTRINSIC CARRIER CONCENTRATION Intrinsic mobility

216 220

IMPURn'Y COl'"DUCI1VlTY Donor stale$

Acceptor slates l'hermal iooizlltion of dootln and acceptors

ttl
222-

224 226
ttT

THERMOELECfRIC EFFECTS
SEMIMEl'ALS

228

SUPERUTIlCES
SUMMARY

228
231

PROBLEMS
I. Impurity orbits

231
231 231 231 232 232

2. 3. 4_ S.

lonJulion or donors Hall effed with two carrier types Cyclotron re50flance Magneloresislam:e with tv.'U carrier types

REFEREI'\CES

232

NOTE: 1be discussion of carrier orbits in applied fields is continued in Chapter 9. Amorphous semiconductors are treated in Chapter ]1. Junctions and barrierS are healed in Chapter 19.

,or'

,or'

-10'"

.......
a,

A>

-,.,.
,.' , :i , , -,0"
, "

Sb

I ,

10" ~

{
-l(t'-

.i

_10'5

-10"

_L

'''"

rtgUre 1 CarrXreonce-ntn.tion.s b" mdab. JoeI1\i.-tab. an:!~.n.e ~KIuc:tor range may t...,. elliended upward by inD"N$ing the impurity CCJrla:I't ....1, and the nngoe can be e>:lendnJ downward to mngc e\"ffItuaily witll the Insulator ~. \.
,~

CHAPTER

8:

SDflCONDUcrOR CRYSTALS

Carrier concentrations reprcsentative of metals, semimetals. and semiconductors arc shown in Fig. L Semiconductors are generally classified by their electrical rcsistivity at room temperature, ",1th values in the range of 10- 2 to to" ohm-cm, and strongly dependent on temperature. At absolute zero a pure, perfect crystal of most semiconductors will be an insulator, if we arbitrarily define an insulator as having a resistivity above lOt4 ohm-cm. Devices based on semiconductors include transiston. switches, diodes, photovoltak cells, detectors, and thennistors. 11lesc may be used as single circuit elements or as components of integrated circuits. We discuss in this chapter the central physical features of the classical semiconductor crystals, particularly silicon, gennanium, and gallium arsenide. Some useful nomenclature: the semiconductor compounds of chemical formula AB, where A is a trivalent clement and B is a pentavalent element, arc called III-V (three-five) compounds. Examples arc indium antimonide and gallium arsenide. Where A is divalent and B is hexavalent, the compound is calk-d a II-VI compound; examples are zinc sulfide and cadmium sulfide. Silicon and gcnnanium arc sometimes called diamond-type semiconducton;, because they have the crystal structure of diamond. Diamond ilsclf is more an insulator rather than a semiconductor. Silicon carbide SiC is a I\'.IV compound. A highly purified semiconductor exhibits intrinsic conductivity, as distinguished from the impurity conductivity ofless pure spocimens. In the intrinsic temperature range the electrical properties of a semiconductor arc not essentially modified by impurities in the crystal. An electronic band scheme leading to intrinsic conductivity is indicated in Fig. 2. Ibe conduction band is vacant at absolute zero and is separated by an encrgy gap E g from thc filled valence band. The band gap is the difference in energy between the lowest point of the conduction band and the highest point of the valence band. TIle lowest pointin the conduction band is called the conduction band edge; the highest point in the valence band is called the valence band edge. As the temperature is increased, electrons arc thennally excited from the valence band to the conduction hand (Fig. 3). Both the electrons in the conduclion band and the vacant orbitals or holes Iefl: behind in the valence band contribute to the electrical conductiVity.
BAND GAP

The intrinsic conductivity and intrinsic carrier concentrations arc largely controlled by E/kBT, the ratio of the band gap to the temperature. When this ratio is large, the concentration of intrinsic carriers will be low and the conductivity will be low. Band gaps of rcpresentativc semiconductors are given in Table I. 11Ie alues of the band gap are obtained by optical ahsorption.

be,

.e,

Figure 2 Balld scheme for intrinsic conductivity in a scmiconduct.... AI 0 K the oollductl\;I)' is zero 11'IX'3us., all stales ill the valeTK.'e hand are filled and all stalt;S in the conduction band are vacant. As lhe lcmpcrnlure is increased, electrons arc thermally c><cited from till: vll1erl bind to the conduction band, where they become mobile.

" "
(')

""

,/

"
"

V
"

"

/'

1/

/
"

1/

1/

."

/
-

/
~/
-

."

"
(.)

,.)

Fig...... 3 Intrinsic electron <x>ncelltralion as a function of temperAture for (a) gennanlum and (b) .'Ikon. Under intrinsic conditions the hole concentration is equal to the da'lron concentration. ".., inlrinsiccollccnlralioll at agiven temperdlu"" is higher in Gt: than iuS; becaust: thcenc.'rgygap is llarl"(lWl...- in Ce (0.66 eV) than in Si {I. 11 .N}. (After W. C Dunlap.)

~micond ..do"

Cl"Jllldr

20'

(
Table 1 Ene,-gy gap between the valence and conduction bands
(i - indirect gap; d - direct gap)
E~ eV ~ eV

0,""
Diamond 51 Ge oSn JoSh

Gop

OK

300<

Cry>b'

Cop

OK

300K

d d d d d d

,nA.
'nP GaJ'

e..A.

c.sb AISb
SiC(lJex)

Li1 0.744 0.00 0.23 0.43 1.42 U2 1.52 0.81

1.65

1.1I 0.66 0.00 0.11 0.36 1.27 2.25 1.43 0.68

HgTe Pb5 PbSe

PbTe

d d

Cds CdSe CdTe ZOO

ZnS

d d d

SnTe
AsCI AgI

-0.30 0.286 0.165 O.t90 2.582 1.840 1.007 3.436 3.91 0.3

O,J.I-O.37

0.21 0.29
2.42

I."
0.56

Te Zrlsb

3.0 0.33 0.56

1.74 1.44 3.2 3." 0.18 3.2 2."

CliiO
Ti0,:

2.172 3.03

'lIgTe Is a semimd8l. tht: bandl O'o....-Iap.

The threshold ofcontinuous optical absorption at frequency wI( dctcrmint.:s thc hand gap 1: .:: hWI: in Figs. 43 and Sa. In the direct absorption process a photon is absorbed by the crystal ""tlt thc creation of an electron and a hole. In the indirect absorption process in Figs. 4h and 5b the minimum energy gotp of the band structure involves electrons and holes separated by a substantial wavcvcctor ltc. Here a direct photon transition at the energy of the minimum gap cannot satisfy the requirement of conservation of wavevcctor. hecause photon wavevectors are negligible at the energy range of interest. But if a phonon of wavevcetor K and frequency n is created in the process, then we can have

k{photon) = kc + K 13l 0 ;

hw=E,;+hn,

as required by the conservation laws. 111e phonon energy hO will generally be much less Ihan E: a phonon even of high '\I3vevcetor is an easily accessible sourec of crystal momcutum Ix:causc the phonon energies arc charactcristkally small (-0.01 to 0.03 eY) in comparison with the energy gap_ Hthe temperature is high enough that the necessary phonon is already thermally excited in the crystal, it is possible also to ha,-c a photon absorption process in whidl the phonon is absorbed.

".
CllI!>TAI. Wrnt OIRI:cT CAl'

ctl'"!>TAL \\'Inl I:OIREC1" CAP

Ahoorplion

Figure 4. Opticaillbsorplionin pure in"llIaIOf$ _, ab..,l"t" orero. in (3) Iht, Ihrbhold d"'"rmlllt'l ,"" e.Jergy gllP as E. - Aw.. In (hI the: opticoal absorplioo il we;J;e, nar the threshokl: at
h", ~ ." + flO. photon is absorbed "ilh 1M crnIion ollhrce.particles; free declron rift! hole....d. phonon of ~ "n. In (b) rho: energy E..... marks the ~ b- ~ <nation of. free elcctrotl and free bolr...ith no phonon in\vh-..d. Such li-ansiliDII is ndkd \"ertic:al; it is similar to the direct !JanlIition in Ia). "I1>ew: ploo do not ~how absor-ption lines that ~Ddintel are seen lyingjusllo the Io.-energy lide ollhe threshold. such linn"" due 10 the cn"lIIion of. bDuod

electronhole plir, called all e>ociron.

J
filUl"e S In (aJ the ",","I point of the ronduetion band OC'OJrs at ttrsame ,~olk as the highrsl point of the . - . - band. A dlrecl: optialtranlillOn is drawn \"ertical1y with no signi5cant dIange of k. because doe aboorbed pholon ha5 a ''''ry llnall "~e\t'Ctor. 1be thrcshokf frequency .... for ablOrptiOll by the direct lnn.itioo determines the energy gap ... !'i&l., TIle indirt'd tnnsitioo in
(h) j'lVol\'es

ooth a photon and I phonon beuuse the band ed~ of the oonductioll and \'$knce bands arc widely Icparat...d in k IpllCC. l'he threshold ent'.!,')' for thc indi'l'<:l proces. in (b) b greater than thc true band gap. 1hc: absorption threshold for Iho:: Indirect lun.ition betwf.'<en the band ooges is at fi..... E~ + An, ....hton n iii the frequency of an emitted Jlhorloo of wa,e'ector K .. - ..... AI higher lempenlluns phonons an: a1re..dy preknt; if. phooon Is absorbo.-d along""" 1\ pbuton, the th.es.hoId energy Is: fi.. - E~ - An. ....ot.,: The rogure d>ows only the th~hoId tnns:itiom. Tr2Nitions D<'C.'UI" genenily between almost all poinb cl the two b.nd'I b "hictI the ....,..",...non and ftIef'gy (W> be COIlse.,..ed.

,~

..

(j

S.,micond"cfor Crlld""

203

-------------

,, ,

,~

i .
w'
<

J !

w ..

A",ineV ____
rLgUn,6 Optical absorption in pure indium antimonide. InSb. The transition is direct because hath conduction and vll.len~ barn:! edb'CS are at the center or the Brillouin mne. Iol .. O. Notice the

sharp threshold. (After C. W. Cobeli and Ii. Y. Fan.)

nw band gap may also he deduced from the temperature dependence of the conductivity or of the carrier concentration in the intrinsic range. The canier concentration is obtained from measurements of the Hall voltage (Chapter 6), sometimes supplemented by conductivity measurements. Optical measurements determine whether the gap is direct or indirect. nw band edges in Cc and in Si arc connected by indirect transitions; the band edges in InSh arc connected by a direct transition (Fig. 6). The gap in aSn is direct and is exactly ZCro; HgTe and HgSe arc semimetals and have negative gaps-the bands overlap.
EQUATIONS OF MOTION

We derive the equation of motion of an electron in an energy band. We look at the motion of a wave packet in an applied electric field. Suppose that the wave packet is ma" up of wavcfunctions ncar a particular wavevcctor k. nle

group velocity is vI! "" dwldk. 11lc frequency associated wh.. a wBvcfunction of energy E is w = ell, and so
VII

Ii-I

dedk

or

v = /i--I VkE(&l) .

(1)

1be effects of the crystal 00 the electron motion arc contained in the dispcrsioo relation E(k). The work & done 011 the electron by the electric fteld E in the interval
& is
(2)

We observe that
& = (dE/dk)& "" flv, Bk
(3)

using (1). On comparing (2) with (3) we have

&.
c

-(eElA,., ,

(4)

whence "dkJdt = -eE, the same relation as for free electrons. We may write (4) ill terms of the external force F as
(5)

This is an important relation: in a crystal !IdkJdt is equal to the external force on the electron. In &cc space d(mvYdt i5 CClual 10 the force. We have not overthrown Newton's second law of motion: the electron in the crystaJ is subject to forces from the crystal lattice as well as from external sources. 111e force term in (5) also includes the Lorentz force on an electron in a magnetic Held, under ordinary conditions where the magnetic field is nol so strong that it brcaXS down the baud structure. Thus the C<luation of motion of an electron of group velocity V in a constant magnetic field B is
(CCS) . - = --v x 8 . dt c '

dk

(51) A dl

dk

EO

-ev x B

(6)

where the right-hand side is the Lorclltz force on the electron. Wilh the group velocit), flv :::: grad"E, thc rate of change of the wavcvcdor is
(CGS)

dk , = --VkE x B . dt file '

dk , (SI) ----V"txB dl fll

(I)

whCTe now both sides of the equation refCT to the coordinates in k space. Wc sec from the vector cross-product in (7) that in a magnetic field an electron movcs in k space in a direction normal to the direction of the gradient of the energy E, so that the electron moves on a surface of consiant energy.

The value of the 1 Jcction k o of L: on B is constant during the motion. The motion in I.: space is on a plane nonna! to the direction of B. and the orbit is defined by the intersection of lhio; plane with a surface or constant energy.

Pllysical Derivation of ~k = F
We consider the Bloch cigcnfundioll I/!k belonging to the energy eigenvalue ~ and wavevcdor Ie:

"" ~ L c
The expectation value
Pc'

C(1<

+ G) cxp{;(k + G). <I

(8)

or lhe momentum of an electron in the slate 11: is

~ (kl-"Vlk) =

L '(k + C)IC(1< + G)I' = '(1< + L G1C(1< + GII~ c c

(9)

usillg l:!C(k + G~2 = L We examine the transftt of momentum between the electron and the latlice when the stale L: of the electron is changed to It + Ai: by the application of

an external force. We imagine an insulating crystal elcctrostatically neutral Cll:CCpt for a single electron in the slale k of an otherwise empty band. We suppose thai a weak external force is applied for a lime interval such that the total impulse given to the entire crystal system is J = IF dt. If the conduction electron were free (m em). the total momentum imparted to the crystal system by the impulse would appear in the change of momentum of the conduction electron:

J = .6.P,oo

6prl = fuU: .

(10)

The neutral crystal suffers no net interaction with the electric field. either directly or indirectly through the frcc electron. If the conduction electron interacts with the periodic potential of the crystal lattice. we must have

J = .6.1ltoI c
From the result (9) for Pd we have
dPd = .""

+ .6.P,,1

(11)

+L
c

'C[(V,IC(k

+ G)I~ . dk)

(12)

The change 6pw in the lattice momentum resulting from the change of state of the electron may be derived by an elementary physical consideration. An elcctron .rcf1cctcd by the lattice transfers momentum to the lauicc. If an incident electron with plane "''ave component of momentwn 11k is reflected

...
with momentum Ii(k + G), the lattice acquires the momentum -IiC. as rc(Juircd by momentum conservation. The momentum transfer to the lattice
when the state

4'k

goes over 10 rh+l>k is

dp,. = -~l: Cl(V.]C(k +

G

ct -<ll<J

(13)

as the portion

(14)

of each individual component of the initial state is reflected during the stale change 61:. lOe total momentum change is therefore (15)
exactly as for free electrons, E(I_ (10). Thus from the definition of J, we have
fuIkJ dt = F ,
(16)

derived in (5) by a different method. A rigorous derivation of (16) by an entirely dillcrcnl method is given in Appendix E.
Holu

1bc properties of vacant orbitals in an otherwise filled band arc important in semkooductor physics and in solid state c1cdrDntcs. Vacant orbitals in a band are commonly called holes. A hole acts in applied electric and magnetic fields as if it has a positive charge +e. 1bc rcason is given in five steps in the boxes that follow.
L

k" "" -k.,. .

(11)

The total wavevcetor of the e1cetrons in a filled band is zero: :Ik = O. This result rollows &om tlle geometrical symmetry or the Brillouin zone: every rundamental lattice type has symmetry under the inversion operation f _ -r about any lattice point; it rollows that the Brillouin zone or the lattice also has inversion symmetry. U the band is filled all pairs ororbitals k and -k are filled, and the total wavcvcetor is 7.Cro. lr an electron is missing rrom an orbital or wavevcctor k,., the total wavcvector of the system is - k.,. and is attributed to the hole. This result is surpl;sing: thec1cctron is missing rrom k., and the position of the hole is usually indicated graphically as situated at k.., as in Fig. 7. But the true wavcvector Iq, or the hole is -I.:.,., whieh is the wavevector of the point G ir the hole is at E. The wavcvcctor -k.. enters into selection rules ror photon absorption.


Q C""duclion band

.
F.

Figure 7 Absorption a phuton of ..mcrgy fUlllld negligible waveH'<:l<l. lakes 1'" electron f.ol11 E in the 1111ed ,'plcll! hamllo Q in the conduction h~JKl. Irk.. was the ";l"ev...,<:l~ of tile e1"droll lit f;, it hC<:<Jmt."$ the wan.'V,:clor orthc ek"Clron at O. The total wa\"c"ector oflhe valence band lifter l~ absorption is -1l.. and this is the W;''''ft'edor ~ mUlil ascri~ to the hole if we tincri!>e the valence 00nd as ot'eUpied by one hole. "Thus k. - -Ie.; the: "'';we....:dor g{ the hole is the $aIIW! as the _"e\~rorlhrdectron which ",mains al C. For the entiRo s)1ilem the total ...,~ector after the Uorption ..,rIM pOOtOll is k.. + Ir., - 0, so I" the l<bl .....evmor q unchang! by the abs0rption 01 the photon mil the creation lfff dedron and ~ hok.

or.

The hole is an alternate description of a band with one missing electron, and we either say thai the hole has wave"ector -k,.or that the band
with one missing electron has total wavcvcclor -k.,.
2.

(18)

Let the zero of energy of the valence band be al the top of the band. The lower in the band the missing electron lies. the higher the energy of the system. l1le energy of the hole is opposite in sign to the energy of the missing electron, because it takes lIlore work to remove an electron from a low orbital than from a high ex-bital. 111l1s if the band is symmetric, I A) = ~-k,,) = -t).{-1cJ = -t",(k,,). We construct in Fig. 8 a band scheme to represent the propa-Ucs of a hole. This'hole band is a helpful representation because it appears right side up.

'Bands aTe ah-.. ys symmetric under the im'Cnion k_ -I< iflhe spin..crbit inla-adion Is ne~ E,~n WIth spin-orbit interaction, bands aTe always symmetric If the crystal "nJdure permit.. the innTSion operation. Without aa:ntcrors~mmctry, wt with spin..,rbit interaction. the

b;.nds are symmetric if we compare subbalids fOl" ..f1ich the spin direction is <evened: 41, t 1 - d-Ic. ! 1(" e OTS. Chapt~... 9.

..
I~ ~JJ<.I-

lwlIl-.-.l ..'\til ......

....... "' ......

- .. (kJ.

It>

-"and

::-----.

Figure 8 l1>e upptt half aliI(' r~ shows the hole band that simulates the d~'nantics of. hole. cOllstruded by j,wersion allHo \unce t.nd in the origin. The .....ev!:or.nd ~ oCtile hoIf, an: equal, but oppos:itc in sign. 10 thr and ~y 0( the empty elechtlll orLit',", In the _-dcne:., woo. w.. do JK)I dlOW the d"po$ilion mill(' electro" Il:'oo''e<! frum the \-a1c1K't" h.<nd at 1<.,.

_....,.<ector

3. From Fig. 8 we see that

4.

V" '"' v"

10
Jnh

(19)

The velocity of the hole is equal

Vlf/,(~) = h./..~).

the velocity of the missing electron. so that v,,{kJ,) ::: vC<~).

(20)

-m.. .

We show below that the effective mass is inversely proportional to the curvature cJ2f1dJ!l. and for the hole band this has tht: opposite sign to that for an electron in the valencc band. Ncar the top of the valence band m.. is negative. so that In... is positive.

5.

~--= e(E

dk,
dt

I + -v" l" B)
c

(21)

This comes from the equation of motion (CCS)

" dl = -.JE "\ +...!.-" C"

l'l

B)

(22)

that applies to the missing electron when we substitute -Iq. for k" and VII for V r . The equation of motion for a hole is that of a partide of positive charge c. The positive charge is COllsistCllt with the electric current carried by the valence band of Fig. 9: the current is carried by the unpaired electron in tnc orbital G:
j ~ (-,)v{C) = (-,)I-v(E) = ",(E)
(23)

which is just the current of a posilive charge moving with the velocity ascriLed to the missing electron at E. 'nle current is shown in Fig. 10.

(
--,

-----:>~'=-- '.

'.

--C

...

~'..,---I

-~

"

8
A

<

fOCL,,-e 9 (.t) AI t,. 0 aU stales are filled aorpt F -' the lop of the band. the ,~Iocil)' 0. Is zero at F ~ dfld1r.. - O. (b) An drdri<.' field E. is applifti in lhe "r din'clion. 11>e forl.'e on the electrons is in !he -Ie. dirtion..d all eIaVons ma1<e lnlmitions logethcr in lhe -.l:..direclipn, moving the hdc 10 the llaIo: E. (el Aftn a ""the inl~ the electrons 11lO'l'e archer along in I:
spK'e

and tM hole is now at D.

Figur~

10 MllIion of dcctrons in the conduction band and holes

in Ih" valenl't: blllnd in the "I(uric fIeld E. 1bc hole and electron

drlfl "clocitit'l art' in QfJ'P'ite dirret ..;.... h"t their electrk- cu.... r('~ll~ .re in the ~me di..... ' dl(ln, the directIon alh e electric: field.

Effective Mass

When we look at the cncrgy-wavcvcctor relation "" (h2I2m)~ for free electrons, we sec that the coefficient of P determines the curvature of E versus k. Tume<! about, we can say that 11m, the reciprocal mass, determines the curvature. For electrons in a bam! there can be regions of unusually high curvature ncar the band gap a! the zone boundary. as we see from the solutions oCthe wave C<Juation ncar tbe zone boundary. If the energy gap is small in comparison with tbe free electron energy A at the boundary, the curvature Is enbanccd by the factor MElt. and the reciprocal mass is enhanced by the same factor. In semiconductors the band width, which is like the free electron energy. is of the order of 20 eV, while the band gap Is of the order of 0.2 to 2 eV. Thus the reciprocal mass is enhanced b)' a factor 10 to 100. and the effective mass is reduced to 0.1-0.01 of the free electron mass. These values apply near tbe band gap; as we go away from the gap the curvatures arc likely to approach those of free electrons.

".
To summarize the solutions of Chapter 7 for U positive. an electron ncar the lower edge of the second band has an energy thai may be wrltlcn as

mJm = U[(IWU) -

1J .

(24)

Ilerc K is the waVC\lcc!or measured from the zone boundary, and m.,. denotes the effective massofthc c1ectrorl ocar the cdge of the second band. An electron ncar the top cJ the first band has the energy
TIlt/m = lI[(2AlU)

IJ .

(25)

The curvature and hence the mass will be negative ncar the top of the first band, but we have introduced a minus sign Into (25) in order that the symbol m" for the hole mass wiD have a positive value-see (20) above. The crystal docs not weigh any less if the effective mass rJ a carrier is less than the frcc electron mass, nor is Newton's second law violated for the crystal taken as a wIlDie. ions plus carriers. The important point is that an electron in
a periodic potential is accelerated relative to the lattice in an applied electric or magnetic field as if the mass of the electron were equal to an effective mass which we now define. We differentiate the result (l) for the group \--clocity to obtain

do! dt

= ,,_I
2

cPt:
dk dt

= ,,_,(cPt:

dk)

(26)

dk'- dt
11 2 do -"d'<1dk' dl

We know from (5) that dkldt = FIfI, whence

do~
dt

(2.... d ) F .
11 2

dJ.2

'

0'

F=

(27)

If we identify h2 /(d 2ddJcZ} as a mass, then (21) assumes the fonn of Newton's second law. We define the effecti\-'C mau m" by

J d 2E

~=flJk2

(28)

It is euy to generalize this to take account rJ an anisotropic energy surface, as for electrons in 51 or Ceo We introduce the components rJ the reciprocal effective mass tensor

1
~,

-dt C

do. (1)
-.-

(29)

where IL. "are Cartesian coordinates.

Ph/pica1 Interpretation of the EIJectiue Mast

How can an electron of mass m when put into a crystal respond to applied HeMs as if the mass were m"'? It is helpful to thin' the process of Bragg reflection of electron waves in a lattice. Consider the weak interaction approxi-

S~iconduCfor

C,....rot.

21'

C)-.tal"\

rend , , ,
I I I I
I I

f"':!<bt .,kt-/nlq
..,;;;

\'-0

(a) the energy g( the e~ron

F"lgUre II ~Ioo d negati'-e e6ect;''t! maun"itichOllCUr new. Brillouin zone boundary. In beam inddml on thin ~1aI ;. lilighd)' 100 low 10 salw)' the rondilion lOr B~ rel'lection and the be.m is t.....smitted through lhe nylibl. The appliollion g( a PI'Ilill roItage llcrau the grid rn.&)', as in (b). ('llUSe the Br.lgg COlll.!ilioa to Ix: oatid'ed. and lhe eIe<:troo be.m wiD then be ref\ed:ed from the appropriate 5t:t of nylihll planes.

mation treated in Chapter 7. Ncar thc bottom of thc lower band the orbital is represented quite adequately by a plane wave exp(ikx) with momentum lik; the wave component exp[i(k - G)xJ wilh momentum Ii(k - G) is small and increases only slowly as k is increased, and in this region m+ .... m. An increase in thcrcf1ccted component exp(i(k - C);r} as k is increased represents momentum transfer to the electron from the lattice. Near the boundary the reflected component is quite large; at the boundary it becomes equal in amplitude to the forward component, al which point the eigenfunctions are standing waves, rather than running waves. Here the momentum component 1t(-iG) canccls the momentum component Ii(iG). It is not surprising to find negative valucs for m just below a wne boundary. A Single electron in an energy band may have positive or negative effective mass; the states of positive effective mass occur ncar the boltom of a band because positive effective mass means that the band has upward curvature (tfl1dfl is positive). States of'negative effective mass occur ncar the top of the band. A negative effective mass means that on going from state k to state k + t:.k, the momentum transfer to the lattice from thc electron is larger than the momeAtum transfer from the applied force to lhe electron. Although k is increased by t:.k by the applied c1cc::tric field. the approach to Bragg reflection can @.ve an ovcrall decrease in the forward momentum d the electron; when this happens the effective mass is ncgati,c (Fig. 11). As we proceed in the scrond band away from the boundary, the amplitude of cxP(i(k - G)]"J decreases rapidly and m+ assumes a small posilive value. Here the increase in electron velocity resulting from a given external impulse is larger than that which a free electron would experiencc. The lattice makes up the difference t' ,ugb the reduced recoil it experiences when the amplitude of exp[i(k - G)x]. .Iiminished.

212

If the energy in a band depends only slightly on k, then the effective mass will be very large. Thai is, m*hn J> 1 when d'lEldf<2 is very small. The tightbinding approximation discussed in Chapter 9 gives quick insight into the formation of narrow bands. If the wavdunctions centered on neighboring atoms

overlap very little, then the overlap integral will be small; the width of the balld narrow; and the effective mass large. The overlap of wBverunctions centered on neighboring atoms is small for the inner or (;ore electrons. The 4/e1cctrons of the rare earth metals. for example, overlap very liUle. TIle overlap integral determines the rate of quantum tunneling of an elcctron from one ion to another. "'hen the effective mass is large, the electron tunnels slowly from onc ion to an adjacent ion in the lattice.

Effective Masses in Semiconductors

In many semiconductors it has been possible to.determine by cyclotron resonance the fOnTI of the energy surfaces of the conduction and valt:nce ballds ncar the band edges. The determination of the energy surface is equivalent to a determination of the effective mass tensor (29). Cyclotron resonance in a semiconductor is carried out with centimeter wave or millimeter wave radiation at low carrier concentration. The current carriers are accclerated in helical orbits about the axis of a static magnetic field. The angular rotation frequency We is
(GGS)

we ""--

eB

m*c

~I)

(30)

where m* is the appropriate cyclotron effective mass. Resonant absorption of energy from an rf electric field perpendicular to the static magnetic field (Fig. 12) occurs when the rf fruJuen{.')' is equal to the cyclotron frequency. Holes and electrons rotate in opposite senses in a magnetic field. \Ve consider the experiment for m*/m = 0.1. At Ie = 24 CHz, or we = 1.5 X 10'1 S-I, we have B ""' B60 C at resonance. The line width is determined by the collision relaxation time 7, and to obtain a distinctive resonance it is necessary that W e7 2: 1. The mean free path must be long enough to permit the average carrier to get one radian around a circle bctv.:een collisions. The requirements are met with the use of higher frequency radiation and higher magnetic fields, with high purity crystals in liquid helium. In direct-gap semiconductors with band edges at the center of the Brillouin zone, the bands have the structure shown in Fig. 13. The conduction band edge is spherical with the effective mass ,~: t"c = Ell

+ /i2J..!J.l2m"

(31)

referred to the valence band edge. The valence bands arc characteristically

F~ It
IOI!'ln;l!'

Arnngcrnellt ofRdds ill' C)"dotron rf'S(llW~ expmlTlnlt ill a 5'"mit'OndUl<1Ol"."The of lht' dt('\lblion is opposile for electrolls ,lid jlllleS.

--=~;4,<::""==----,

Figure 13 Simplifk-d vi.:.... of the ()alld ....Ige structure

or a dlrL'(.'l glIfl semkvnductor.

".
Table .2 EfTecth'c masses of deelrons and holes in direct gap semiconductors
Electron
CI'}"1'b1
Hc::n')'

hole

l...ight hole

mJm
0.015 0.0"2 0.073 0.047 0.066 0.99

mu/m
0.39
0041

SpI,I-off

me/m
0.021 0.025 (0.078) 0.06 0.082 0.58

m..Jm
(0.11) 0.08 (0.15) (0.14) 0.17 0.69

"'*

S(lin.oo..u. 0., eV

InSb

low
InP CaSb CaA.
CUIO

0.4 0.3 0.'

0.82 0.43 0.11 0.80 034 0.13

threefold ncar the edge, with the heavy hole 1111 and light hole III bands degenerate at the center. and a band soh split orr by the spin-orbit splitting A:

E,;(Ilh):S!! -1i 2f2l2rnlJo;

~(so1l)
.6!!

~(lh) a
2 2

-1i21!12m,1o

-!J. - fi k f2m.,.

(32)

Values of the mass parameters arc given in Table 2. TIle fonns (32) arc only approximate, because even close to k "" 0 the heavy and light hole bands are not spherical-see the discussion below for Cc and SiThe perturbation theory of band edRcs (Problem 9.8) suggests that the electron effective mass should be proportional to the band Jo:3P, approximately, for a direct gap crystal. We usc Tables I and 2 to find thc fairly constant valucs m.l{mEJ = 0.063, 0.060, and 0.051 in {eV)-1 for the scrics [nSb, [nAs, and Inr. in agreemcnt with this suggcstion.

Silicon a'id Cermanium TIlC conduction and valence bands of germanium arc shown in FiK- 14, based on a combination of theoretical and experimental results. TIle valence band edge in both crystals is at k = 0 and is derived from P3I2 and Pm slates of the free atoms, as is clear from the tight-binding approximation (Chapter 9) to the wavcfunctions. TIle P3I2 level is fourfold degenerate as in the atom; the four states correspond to mJ vaJues i and t. The PII2 level is doubly degenerate, with mJ = i. 11Je P3I2 states are higher in encrgy than the PII2 slates; the energy difference t::. is a measure of the spin-oroit interaction. TIle valence band edges arc not simple. lloles ncar the band edge are characterh:ed by two effective masses, light and heavy. llJese arise from thc two bands fol'TllCd from the PM Ic\ c1 of tbe atom. lllet"c is also a band fonned from tlle P J12 level, split offfrom tlle P3.'lt level by tllC spin-orbit interaction. The energy surfaces aTe not spherical. but warped (Q1'$, p 271):
~)

"" AI.:'.l. (B 2k' +

C'(I.~k;

k2;~

+ kz --i))112

(33)

, , , ,

5.4 eV

,s.,mkunductor C'l"lob

'IS

. -o.g<:-V
U.

-, t -,
"

.-.
@

-,

-, -,

-,
-,
-w
-u

-"
-" k. !!'.(_'_I.!.)
" 2 2 2 [Ill]

,-,
I
(100)

k ... ~IIlO)

F"~e 14 Calcu1ated band structure cl germUlillm, after C. Y. Fong. The g<.'tl(...,.,.) Ie.turn are in 9lOd llG"'""""nl ",~th experiment. The WT ~'a1cncc bauds are shown in gray. The flne structure or the ~*tlCe Ind edge u cau$ed by .pin-orbit splitting. TIle energy gap i$ indirect; the ronduetioo band ed~ is al the point (2.,.{oXiH). 'l"heconsUnl CrlC.--rgy surfat-es aroond!hill point are ellipsoidal.

."
The choice of sil,'T1 distin~ishes the t\o\'O rna'i.~es. The split-off band has d.k) co -/1 + AJ.:2. The experiments give. in units II'1/2m,

Si:
Cc:

A = -4.29 ;
A = -13.38 ;

181 ~ 0.68 181 ~ 8.48

ICI = 4.87 , ICI = \3.\5 ;

A = 0.044 eV A = 0.29 eV

Rou~hly. the light and heavy holes in ,",TCrmanimn have ma~scs 0.043111 and

0.34 m; in silicon 0.16 111 and 0.52 m; in diamond 0.7 m and 2.12 m. The conduction band edges in Ge are at the equivalent points L the Brillouin zone, Fig. 15a. Each band edRC has a ~1Jheroidal energy surface oriented along a (HI) crystal axis, with a longitudinal mass J1lf = 1.59111 and a transverse ma.~s 1I1t = 0.082 m. For a static magnetic lleld at an angle with the longitudinal axis of a spheroid, the effective cyclotron mass fflc is

of

1 coile sin 2 8 -=--+-111~ m~

mjln,

(34)

Results for Gc are shown in Fig. 16. In silicon the conduction band edges arc spherokls oriented along the equivalent (100) directions in the Brillouin zone, with ma."iS parameters m, = 0.92 111 and 111/ = 0.19 m, a.~ in Fi~. 17a. The band edges lie along the lines labeled 6 in the zonc of Fig. 15a, a little \vay in from the boundary points X. In eaAs we have A :: -6.98. B ~ -4.5, lei == 6.2, .6. :::: 0.341 eV. The band structure is shown in Fig. 171>.
INTRINSIC CARRIER CONCEl\'TRATION

We \vant the concentration of intrinsic carriers in terms of the band g.lp. We do the calculation for simple parabolic band edges. We first ca1culate in terms of the chemical potcntial p. the number of electrons excited to the conduction band at temperature T. In semiconductor physics po. is called the Fermi level. At the temperatures of interest \yc may suppose fOr the conduction band of a semiconductor that E - P. kBT, and tile Fernli-Dirac distribution fullC."tion nxluces to

= expl(p. - E)lkn'fJ .

(35)

This is the probability that a conduction electron orbital is occupied, in an approximation valid when f~ <Il I. l11c energy of an electron ill the conduction band is
Ek ::::

E" + "2J...2l2m,,

(36)

where E" is the energy at the conduction band edge, as in Fi~. 18. Here m~ is the effective mass of an electron. TIllIS from (6.20) the density of states at E is

D.,(E) :::: -

2"r

\ (2m.)M - - (E ,,2

EJ1I2

(37)

B Se""lfIduc'.,.. Cry.tal$

217

Wi

, , ,

"

I><'C

!.

1 ,
{I
,

: ,_L
,h
K

, , "

1'.;--;..---- -"-1---..

I, "

-:f~~~-------

, ,1,.. ~D

.j
Figure 15 Standard labels oflhc symmetry points and a.<<;$ofthc Brillouin woes oflhclCc and bee laUiccll. TI,e wile cent<:n are r. In (a) the boundary point at (27rla)(JOO) i. X; the bourldarypoinlll.t (2.,.fll)(ll i);5 L; the line!J. runS !.>elwc<;n r arid X. III (b) the corrcsp:mding 5>11,001. are H. P and tJ..
0.<0

.,,
'" ,
II
1.1

'"

11

/
II
I~

Figure 16 Eff<.'CIive <.;dotron mass of dc<1rons in g<....ma. nium at 4 K for magnetic field din:cbons in a 110 plane. There are lOur indepelldt,."fl! mass sph<.'ll)id. ill CO'. o"c along each III a,is, but vit,....,-.:d in the 110 plane tWQ spheroids alwa~'S appear l,."luival"nt. (Aller Oresselha"., Kip, and Killel.)

" If lc' " I'> ,. "" ,

0'

r-.

t::
'=
,~
4(}

-w "

10 20 30
In

",m

Angle in degree>

lIO pL".., frum 1001]

"''''

00 '00
a>J'

".
..-------~~~~~--~

'.

.L--"
..

'.

-. -,
_.L_~L-_.....L

_ _-:>

.J
Figure 17a Comlanl ~lIerg.... ellipsoids for decIrons in silicon, d....wn lOr ill/lit, - 5.

.J

Figure 17b Band slmClu"", of CW, after S. C. Louie.

TIle concentration of electrons in the conduction band is

which integrates to give

Il =

2( ~::):W cxp{(,.L -

E.,)lksT]

(39)

The pl'Oblem is solved for '1 when I.L is known. It is useful to calculate the equilibrium concentration of holes p. 111c distribution function fit for holes is related to the electron distribution function Ie by flo = 1 - Ie. because a hole is the absence c:I an electron. We have

fit

= 1-

cxp[(t:

p,)lksT]

L =

--,,-_I~:=7''''''' cxpL(jt - E)/ksT] + 1

('0)

e cxp((E - l.L)/kJjT] ,

provided (j.L - f') keTo If the holes near the top of the valence band behave as particles with dTerti\'C mass mn. the density of hole states is given by
D,,(t") =

2~ (2~" )312(" _

E)U2 ,

('1)

~
f--~""""""-',""II:.,,~

I "
,.

-------------- = __=1
Fipr~ 18 EIIttJO' MlIl!I' b 5tatistical cakubliIon$. The Fermi didributiQn function is .oo..T1
(lIl

the tame 1Cl11', b a IemperatUl'"c l ..T <C E... The ."~rmi Je-.-el,... is Ial<I'II to lie ",,,n within thr bMd gap, lOll b ... intrin"" ~ic(lndudOl" If
t -,..

Iht.'I\f- I

\\here Eo> is the energy at the valence band edge. Proceeding as in (38) we obtain (42) for the amcentration p of holes in the valence band. We multiply t~ether the expressions for n and p to obtain the equilibrium relation, with the energy gap E8 = Ec - Eo.
(43)

This useful result docs not involve the Fenni level JA.. It is an cxprcssion of the law of mass action. \Ve have nowhere assumed in the derivation that the matClial is inhinsic: the rcsult holds in the presence of impurities as well. l11e only assumption made is that the distance of the Fenni level from the ed!;e of Loth bands is large in comparison ".ith kBT. At 300 K the value of np is 2.10 x 10 111 cm-~, 2.89 x r0Z6 cm-6 , and 6.55 x 1012 cm- 6 for the actual band structures ofSi, Ge, alld CaAs, respectively. A simple kinetic argument shows \\1lY the product rl" is constant at a !;iven tcrnperature. Suppose that the C(juilibrium population of electrons and holes is maintained by black-body photon radiation at temperature T. The photOflS gencraie electron-hoI", oai~ at a rnte A(T), while B(l1np is die rate of the rerombi-

nation reaction e

+"

'* photon, Then

d"ldt = A{T) - B(T)np = dpldt

;0

(44)

In equilibrium dnldt 0; dpldt = 0, whence np = A(1')B(J'). Because the pnx!uct of the electron and hole concentrations is a con.stant independcnt of impurity concentration at a given temperature. the introdliC'tion of a small prolXll1ion of a suitablc impurity to incrcase n, say, must decrease p, This r~u1t is important in pradice-\\IC can reduce the total canier concentration n + p in an impure crystal. sometimes enonnously, by the C'OrltroJ(e<1 intl'()duction of sUitable impurities. Sudl a reduction is called compensation. In an intrinsic semiconductor tJle number of electrons is equal to the number ofholcs, because the thcrmal excitation man electron leaves behind a hole in the valence band, l110S from (43) we have. leUing the subscript I denote intrinsic and Ell = Ec - E.,.
lIj = V, "'" 2 2~2

( kT)'" (m..m,l/~ exp{-E/2kBT)

(45)

The intrinsic carrier depends exponentially on E,I2k BT. where Ell is the energy gap. We sct (39) cqual to (42) to obtain. for tile Fermi Icvel as measured from tJle top of the valence band. exr{2J.llklJT) = (mlo'm,,'f'l eKp{E,/kBT);
po = lEI:

(46)
(47)

+ !kgT In {m,/mJ

If1ll/o = tn... then po = lEI and the Fenni Icvel is in the middle of the fOrbidden gap. A thorough treatment of the statistical physics of semiconducton- is givcn in TP. Chaptcr 13,
Inlrjr18ic Mobility

The mobility is thc magnitude of the drift velocity per unit electric field:

,,- WE

(48)

The mobility is dclined to be positive fOr l>otJl electrons and holes. a1thou~ their drift vclocities are opposite. By writing PO.. or p." for the clectron or holc mobility we can avoid any confusion !.>et,,:een p. a" thc chemical potential and as the mobility. l1Je electrical conductivity is the sum of the electron and hole contributions:
u = (nep...

+ pep.,,) ,

(49)

where n and V arc the concentrations of electrons and holes, In ChaptcT 6 the drift velocity of a ehalXC q was found to be tl = qTFJm. whcnce
~ "" eT;m.. ;

p." = e-r,/m/o .

(SO)

,
(
Table 3

Se.nkond"~lOr

Crypfo/$

221

Carrier mobilities at room temperature, in cm1fV_s

Holes Crystal

Ccyslal
Diamond

Electrons

Electrons

Holes

.600

.200

CaAs

Si Ce InSb loA>
Inl' AlAs

1350
3600

480
'600

GaSb

8000 5000

PhS
PbSc PbTc Agel
K8r (HJO K)

BOO
30000

4500
280 900

450 450 .00 400

2500

."'"

550

300
()()()

600 930

1()()()

AISb

SiC

50 '00 '00

10-20

The mobilities depend on temperature as a modest power law. The temperature dependence of tIle conductivity in the intlinsic region will be dominated by the exponential dependence exp(-EJ2k B T) of the runicr concentration,

EIJ. (45).
Table 3 gives experimental values of the mobility at room temperature. 111c mobility in Sl unit~ is expressed in rn 2 N_s and is 10- 4 of the mobility in practical units. For most substances the values quoted are limited by tllc scattering ofcarriers by tllcrmal phonons. nlc hole mobilities typically arc smaller than the electron mobilities oc"Cause of the occurrence of band degenerdC}' at the valence band ed~e at the zone center, thereby making possible interhand scallering pnx.'csses tImt reduce the mobility considerably. In some crystals, particularly in ionic crystals, the holes are essentially immobile and get about only by tIlermally-activated hopping from ion to ion. 111e principal cause of this "self-tl'llpping" is the lattice distortion a~sociated with the Jahn-Tellel' effect of degenerate states (Chapter 14). TIle necessary orbital degenel'llcy is much more frequent for IlOles than for electrons. There is a tendency for crystals with small energy gaps at direct band edges to have high values of the electron mobility. By (9.41) small gaps lead to light effective ma~ses, which by (50) favor high nJObilities. 11Je highest mobility observed in a semiconductor is 5 X l<? cm 2N-s in PbTe at 4 K, where tIle gap is 0.19 eV.
IMPURITY COl"iDUCTJVITY

Certain impurities and imperfections drastically affect the electrical properties of a l;Cmiconduetor. 11Je addition of boron to silicon in the proportion of I boron atom to 10" silicon atoms increases the conductivity of pure silicon by a factor of 103 at room temperature. In a compound semiconductor a stoichiometric deficiency of one constituent will act as an impurity; such semiconductors

...
are known as c:leflcit semiconductors. nle deliberate addition of impurities to a semiconductor is called doping. We consider tlle effect of impurities in silicon and gennanium. These elements aystallize in the diamond structure. Each atom forms four covalent bonds. one with each of its nearest neighbors. corresponding to the chemical valence four. If an impurity atom of valence five, such as phosphorus, arsenic. or antimony, is substituted in the lattice in place of a nonnal atom, there will be one valence electron from the impurity atom lert over after the four covalent bonds are established with the nearest llCi~hbors, that is, after the impurity atom has been accommodated in the structure \\;Ih as little disturbance as possible. Donor SJafn. The structure in Fig. 19 has a positive dlarge on the impurity atom (which has lost one electron). Lattice constant studies have veriGed that the pentavalent impurities enter the lattice by substitution for f\OI'1l1<I.l atoms, and not in interstitial positions. Impurity atoms that can give up an electron are called donon, The crystal as a whole remains neutral because the electron remains in the crystal. The electron moves in the coulomb potential e'Er of the impurity ion. where E in a covalent crystal is the static dielectric constant of the medium. The factor UE takes account of the reduction in the coulomb force between charges caused by the electronic polari"altion of the medium. This treatment is valid lOr orbits large in comparison with the distance between atoms, and for slO\... motions of the electron such that the orbital frequency i.~ low in comparison with the frequency w, corresponding to the energy WlP, These conditions arc satisfied quite well in Ge and 5i by the donor electron of P, As, or Sb. We estimate the ionization energy of the donm' impurity, The Bohr theory of the hydrogen atom may be modified to tuke into account the dielectric constant of the medium and the effective mass of an electron in the periodic potential of the crystal. 'Ole ionization energy of atomic hydrogen is _e 4m12h 2 in CGS and -e 4 mf2(47TEofl)2 in SI. In the semiconductor we replace bye-'E and m by tile effective mass file to obtain

e-

(51)

as the donor ionization energy of the semiconductor.

The Bohr radius of the ground state of hydrogen is h 2fmt! in CGS or 47T2fnufl in 51 Thus the Bohr radius of the donor is (CGS)
aJ

= --: -- "
~

Eh'

(0.53<)
m/m

(52)

t..

_. .
n.-. bound bel

",

f"1@;...e 19 ~ aQOCi.ated with an anenic irnPJrily a10rn in .iIicon. Ancnic hal; fi,-e \'lIJenc:e ekctrons. but .iicon t - only four .-ale,K'e dcrb'ClnI. ThU5 four dedrons on the U1aIic lbnn letr2hednJ cm-alcnt bonch ~milas- 10 W1icon. and the IifU1 electfOn is ...... Iab1c b conduction. 11M: llr$rf\ic ~ iii calkd a donor because when ionizlld il donaIcs an dectron lutlle cond~ion band.

1be application 10 germanium and silicon is complicated by the anisotropic effective mass of the conduction electrons. But the dielectric constant has the more important effect Clfl the donor energy because it enters as the square, whereas the effective mal;S enters only as the first power. To obtain a general impression of the impurity levels \Ye usc 111., - O. L In for electrons in gennanium and m. "'= 0.2 m in silicon. TIle stalic dielectric constant is given in Table 4. The ionization energy of tile free hyd~cn alom is 13.6 eV. For germanium the dOllor ionization energy Ed on our model is 5 meV, reduced with respect to hydrogen by the factor mJml "" 4 x 10- 4 , l1lC colTCsponding result for silicon is 20 meV. Calculations using the correct anisotropic mass tensor predict 9.05 meV for gemlanlum and 29.8 meV for silicon. Observed values of donor ionization energies in Si and Gc are given in Table 5. Recall that I meV ... 10- 3 eV. In eaAs donors have Ed - 6 meV. 11le radius of the first Bohr orbit is increa.~ed by ~mjlll. over the value
Table"
Crysbl

Static nlative dielectric COIl.'ltant of semiconductors

5.5 11.7 15.8 17.68

14.55 12.37

CI')'sllil

15.69 13.13 10.1 10.3 10.2 7.1

Diamond

C.Sh
CaA.

5,

Cc

AIM
A)Sb SiC CUlIO

."'"
I,P

(nSb

Table 5 Donor ionization energies Ed of pcnta"alcnt impurities in gennanium and silicon, in meV
p

56
'9.
1~7

5;

C,

45.
I~O

39. 9.

the free hydrogen atom. The corresponding radius is 80 A in germanium and (60XO.53) '"' 30 A in silicon. These are large radii. so that donor orbits o"erlap at relatively low donor concentrations. compared to the number ofhost atoms. With appreciable overlap. an "impurity band" is fonned from the donor states: see the discussion ofthe metal-insulator transition in Chapter 10. The semiconductor can conduct in the impurity band by electrons hopping from donor to donor. The process of impurity band conduction sets in at lower donor concentration levels if there are also some acceptor atoms present, so that some of the donors are always ionized. It is easier for a donor electron to hop to an ioni7.ed (unoccupied) donor than to an occupted donor atom, so that two electrons will not have to occupy the same site during charge transport. Acceptor StCltes. A hole may be bound to a trivalent impurity in gennanium or silicon (Fig. 20), just as an electron is bound to a penta,'3lent impurity. Trivalent impurities such as 8, AI, Ca, and In are called acceptors because they accept electrons from the valence band in order to complete the CO\falent bonds with neighbor atoms. leaving holes in the band. When an acceptor is ioni7.ed a hole is freed. which requires an input of energy. On the usual energy band diagram. an electron rises when it gains energy, whereas a hole sinks in gaining energy. Experimental ioni7.ation energies of acceptors in germanium ami silicon are given in Table 6. The Bohr model applies qualitatively for holes jU6t as for electrons, but the degeneracy at the top of the valence band complicates the effective mass problem. The tables show that donor and acceptor ioni7.ation energies in Si are compantble with kBT at room temperature (26 meV), so that the thermal ioni7.ation of donors and acceptors is important in the electrical conductivit) of silicon at room temperature. If donor atoms are present in considerably greater numbers than aoceptors, the thermal ioni7.ation of donors will release electrons into the conduction band. The conductivity of the specimen then will be controlled by electrons (negative charges). and the material is said to be n type. lf aoceptors are dominant, holes will be released into the valence band and the conductivity will be controlled by holes (J.mitive charges): the material is p
(I60XO.s:3)
:::>t

0.53

for

II

s,.",iC()nil"rlor CryslaU

225

Table 6

Aa:eptor ionization encrgies . or tri,aJcnt impurities in gCTJJlllnium and silicon, in meV

B

I,

51 C<

45.
lOA

57.
10.2

65. 10.8

157. 11.2

.,..-...
'-"''-' --------T ,...

'"

Figure 20 Boron has 0111)' thn:o: ,... knee .,J~rons; il C1In complele its tctr:oho:dral bonds only by taking ... dedron rrom. SI-Si bond. leaving behind a hole in lhe silicon'~b.>d. n.e ~iti"e bole is lhen a''llilable rCl'" conduction. n.e boron a10m iscolk<! an lICptor because ..hen ioniud il IIC.'l:'epIs ., ",ledron rrom lhe ,~band. AI 0 Ii: I"" hole is bound n:mnnber lhat hoks Iloat.

type. sign of the Hall vollage is a rough test for n or p type. Another handy laboratory test is the sign of the thermoelectric potential, discussed below. 'The numbers orhoIes and electrons are equal in the intrinsic regime. The intrinsic electron concentration n/ at 300 K is 1.7 X 10 13 cm -3 in germanium and 4.6 x 1()'1 cm- 3 in silicon. The electrical resistivity of intrinsic material is 43 ohm-em for germanium and 2.6 X 10-" ohm-cm for silicon. Cennanium has 4,42 x lQ22 atoms per cm3 . The purificatioo of Ge has been carried further than any other element. 2; The concentralion of the wnlmon electrically acli,'e impurities-the shallow donor and acceptor impuritieshas been reduced below 1 impurity atom in 1011 Ce atoms (Fig. 21), For example, the concentration of Pin Ce can be reduced beloo.y 4 X to lO cm- 3 . Thc experimental sensitivity for detection of these impurities is 107 cm- 3 by the method of photothermal ionization spectroscopy.3 1bere are impurities (1-1, O.
tEo E. Haller, W. L Hansen, and F. S. Cuulding. "l'hysics ur ullra-pun: g"rmanium," All.. Ph)'S 30.93-138 (1961). "s M. Kogan and T M ufrhils, t'bys. SbhlS Sulidi (a) 39, II (1977).

n.e

..
10"

,O"o~-----"\;o---c",;;----,,;;----,,:;\;----~,,

...'"

Figure 21 Ternpeldlure d"P"ndcnce oflhe free ('lIm"r roncelliration in ultrapure Ce, llfter R. N. Hall. The od l'OnCt.'fllralion of electrically active irnpurities;s ;2 )( 10'" cm- 3 , as dr-tennino:<! by Hall C'Ol'ff..:ic:n1 meuuren"~nls. 'Ille npid onset cl;ntrinsic .. ~ibtion is evident allow wlnCll cliffl1le CllTOC'l" <:UIlCefllrlllion p closely comlan' klw1:<'n 20 K and 200 K.

Si.

q whose
cm-3

1014

concentrations in Ce cannot usually be reduced below 1012_ but these do not affect electrical measurements and therefore may

be hard to detect.
Thf!OlWl loniuJtion of Donor, and Acceptor,

TIle calculation of the equilibrium concentration of conduction electrons from iOI\I7.ed donors is identical with the standard calculation in statistical mechanics ofthe thermal ioniz,,'l.tion of hydrogen atoms (Tr, p. 369). Ifthere are no acceptors present, the result in the low temperature limit kBT 4Ji Ed is

(53)
with no. 2J..rn); BT/27rli 2>f2; here Nd is the concentration of donors. To obtain (53) we apply the laws or chemical equilibria to the concentration ratio le)(N;Y {N.J. and then set {N.tJ = {el = n. Identical results hold for acceptors, under the assumption of no donor atoms. If the donor and acceptor concentrations are comparable, affairs are complicated and the equations are solved by numerical methods. However, the law of mass action (43) requires the np product to be constant at a given temperature. All excess of donors will increase the electron concentration and decrease the hole concentration; the sum n + p will increase. The conductivity will increase as n + p if the mobilities are equal, as in Fir "2.
\

S..miconduclor Cryslal$

227

10" 'i

10" ~

Figure 22 Electrical <;onducti"ity and hole concentration p calculated &S a functio" of dectron concentralion II for" semiconductor at" tempernlure wch IMI lip '" ufo cm- e. 11,e conducti"ity LHyrn....."trical about n = 1010 cm- 3. for II > 10'0 the specimen is n Iype; for II < 10'0, it is P type. We m.ve taken II-. ... P.A. for lhe mooilitie<.

THERMOELECTRIC EFFECTS

Consider a semk'Onductor maintained at a constant temperature while an electric field drives through it an electric current density jq. If the current is carried only by electrons, the charge flux is
(54)

where Me is the electron mobility. The average energy transported by an electron is, referred to the Fermi level p..
(Ec
-

p.)

+ tk 8 T

where Ec is the energy at the conduction band edge. We refer the energy to the Fermi level because different conductors in contact have the same Fer[JIi le\'el. The energy flux that acco[JIpanies the charge flux is
ju = n(E c - p.

+ ik8 T}(- p...)E

(55)

The Peltier coefficient n is defined by the relation ju = energy carried per unit charge. For electrons,

nj...; it

is the
(56)

n.. = -(Ec -

p.

+ ik8 'f)/e

and is negative because the energy flux is opposite to the charge flllx. For holes
jq::::: ,>ep."E; ju::::: ,J(p - EO'

+ ikBT)p."E

(51)

where E" is the energy at the valt:uce band edge. Thus

11" = (p. - E" + ik B 7)le

(58)

and is positive. Equations (56) and (58) are the result ofour simple drift velocity theory; a treatment by the &lt7.JAAnn transport equation gives minor numeri cal differences." The absolute thermoeleetric power Q is defined from the open circuit . electric field created b)' a temperature gradient:
E~Qgr.WT.

(59)

The Peltier coefficient

n is

related to the thermoelectric power Q by

(60) s This is the famous Keh'in relatton ofi~'ersible thermodynamics. A measure ment of the sign of the voltage across a scmiconc.luctor specimen, one end of which is heated, is a rough and ready .....-ay to tell if the specimen is n type or p type (Fig. 23).
SEMIMETALS

Il=Q"f.

In semimetals the ronduction band edge is very slightly lower in energy than the valence band edge. A small overlap in energy of the conduction and valence bands leads to small concentration of holes in the valence band and of electrons in the conduction band O'able 7). Three of the semimetals, arsenic, antimony. and bismuth, are in group V of the periodic table. Their atoms associate in pairs in the crystallauice, with two ions and ten valence electrons per primitive cell. Ibe even number of valence electrons would allow these elements to be insulators. Like semiconductors, the semi metals may be dopec:l .....ith suitable impurities to vary the relative numbers of holes and electrons. Their concentrations may also be \1lrioo with pressure, for the band edge o\'erlap varies with pressure.
SUPERLATIlCES

Consider a multila)'Cr tt)'stal of alternating thin layers of different compositions. Coherent layers on a nanometer thickness scale may be deposited by
.oR. A. Smilh, ~ucfQt"f. 2ndl.;Cambridge, 1978; H. Fritzsef1"" Solid S~eCornmun. 9, 1813 (1911). A ~irnple discussion cl BolIzrnann tran!pUrt theory is &h'Ul in Appendi~ F. H. B. Callom, T~, 2nd ~.; Wiley, 1!Ii5.

...: -0

, ..d ""'\ ,, I , , \ , "'1"

..

-.
-, 8
Table 7

-,..

~~n'::1--

14'

.. ,., .., ... .. .. ..

,.,
Tempcntur<-, in
I(

~ I

r~ 23 Peh~, <.'Udftciml of n and p tilioon ., fundic.ln of lcfnpentlure. At-'e 600 K the specimen5 8d 1.$ inlnnw: SftIIiC'Ooducton. The "'....'eI are ealc:tJbloo and tho: points are eo:pcrilllC'lltal. (After T. H. Cdlle and C. W. Hull.)

Electron and hole ('OO('Cntrations in semimetals

..... in cm- s

no., mem

AI"S(,.'1IK:
AntimOfl)'

(2.12 0.01) x 10(5.54 O.CE) x 10 111

2.88 X 1017
2.72 X 1018

Bismulh

Graphite

(2.12 0.01) x lOW Ci.49 o.en) x 10'11 3.00 X 1011 2.04 X lOiS

molecular-~am epitaxy or metal-organic vapor deposition, thus building up a superperiooic structure on a large scale. Systems of alternate layers of eaAs and GaA1As have been studied to 50 periods or more, with lattice spacing A of perhaps 5 om (SO A.). A supeTperiodic crystal potential arises from the superperiodic structure and acts on the conduction electrons and holes to create new (small) Brillouin 7.ones and mini energy bands superposed on the band structures of the constituellt Ia)'ers. Here we treat the motion of an electron in a superlattice in an applied electric field; magnetic fields are treated in Chapter 19.

Bloch Oteill6lor
Consider a collisionlcss electron in a periodic lattice in one dimension. with motion nonnalto the planes of the superlattice. The equation of motion in

a constant electric field pamllel to k is hdkJdt -eE or, for motion across a Brillouin mne with reciprocal lattic-e vector C :: 27T1A, we have hC = fl27T1A = eET. The Bloch frequency of the motion is w8 27rIT = eEAlh. The electron a(,'CClerates from k = 0 toward the mne boundary; when it reaches k = 7TlA it reappears (as by an Umklapp process) at the mne boundary at the identical point -7TlA, using the argument of Chapter 2. We (,'Onsider the motion in a model system in real space. We suppose that the electron lies in a simple energy band of width 0:
=0 =0

= 0(1 - cos kA) The velocity in k-space (momentum space) is

v = ft-ldslfk = (AoIh) sin kA

(61)

(62)

and the position of the electron in real space, with the initial (,'Ondition z = 0 at t = 0, is given by
z = fv dt = fdk v(kXdtldk) = (Aq/h) [llk (-liIeE) sin kA

= (-c!eEX(,'Os kA - 1) = (-EJeEXcos(-eE..\IIfI) - 1) .

(63)

This conflrllls that the oscillation frequency in real space is w8 = eEAlh. The motion in the periodiC lattice is quite different from the motion in free space, for which the acceleration is constant. Wunnier Lndder The electrostatic energy of the system is increased by eEA if the initial position of the electron is displaced by one lattice constant from z = 0 to Z = A. For a displacement of the initial position by n lattice constants, the energy eigenvalues are
n

= neEA

+ 0(1

cos kA) .

(64)

This set of solutions defines the Wannier Ladder of energy levels, with adjacent level separation.6. = eEA = hW8' The exact solution of the wave equation gives a continuum of states, with density peaked sharply at the eigenvalues (64). The ladders are observed in optical experiments. The Bloch frequency has not been observed directly (as of mid-l993). Estimates. LetE =3 x 104 Vern-I, 0=0.1 eV, and A = IO-6 cm ;then WIl = 4.6 X 1013 S-I and the localization length is c!eE = 3 nm. For narrow bands in strong fields it may be possible to localize the electron wavefunction dominantly on a single atom; changing the index n in n will change the localization site of the electron. Zener Tunneling Thlls far we have considered the effect of the electrostatic potential - eEz (or - eEnA) on one euergy band; the potential tilts ," whole band. Higher
\

8 Snni<:ollll"clor Cryltflb

2,a

bands will also be tilted similarly, creatin~ the possibility of crossing between ladder levels or different bands. The interaction IJetween different band levels at the same energy opens the possibility for an electron in onc band at n to cross to another band at n'. 'This field.im]uced interoond tunneling i~ an example of Zener breakdO\'ln, met most often at a single Junction as in the 7..ener uwe.
SUMMARY

'J11e Illotion of II wav~ 1);lCkct celltcrt..-o at wavcn~ctor k is described lJy F "" fl(Iklclt, where F is the applied force, The motion in real space is olJtained from the group vdocity va: = h-IV.~). The smaller the energy gap, the smaller is

Im1 near the gap.

A crystal with one hole has one empty electron slate in an othenvise filled

band. 'The properties of the hole are those of the N - I electrons: (a) If the electron is missing from the state of wa\cvector ke. then the waVE!'ector of the hole is k,. "" -k., , (b) 'nle rate of change of k,. in an applied field requires the assignment of 11 positive charge to the hole; el> = e = -ee' (c) If v. is the velocity an electron would have in the state k." then the velocity to be ascribed to the hole of wavevector kl> "" -II. is VI> "" V" (d) The energy of the hole referred to 7.ero for a faIled band is positive and is ,,,(1<,,) = -\<.) . (e) 111e effective mass ofa hole is opposite to the effective mass of an electron at the same point on the energy band: mit "" -m"

Problems
1. Impurity orbiU. Indium antimonide has E,l = 0,23 eV; dielectric constant E - 18; electron clfectil'c mass me - 0.015 I1L Calculate (a) the dOllor ionization energy; (b) the radius d"the ground state orbit. (c) At what minimum dolJm concentration will appreci.ablc O'-crlap effccts between the mbi,ts I adjacent impurity IItoms occur?Thls O"'erlap tends to produce :an impurity bm-a band I CfIerg)' I~ds whieh permit conductivity presumably by II hopping medJanism in which electrons IT\(l\'C fmm one impurity site to a neighboring ioniz(.od impurity Slle.
2. loneotion of donon:, In. particular semiconductor there lire IOU dononIan3 '\ith an klniZation CfIcrgy E.J I I meV and an effecthc IIliISS 0.01 m. <a) Estim...tc the COIlct':l1tration of conduction e1cctmlU at 4 K. (b) What is the \"3.luc of tile Hall coefficient? Assume no acceptor atoms Me present lind that 1::1 ~ kilT.
3. llaU effecf lL'ith IteO c~,. typn. Assuming a'lfIC.'entrntions n,p; l-el:uatjon times 1'... 1\; and masses 'n.., m", show thai the Hall coefficient In the drift velocity approximation is

'"
(CGS)
II" ,.. -

ec

. -f':--;-";';,-

,,-nl}

<1' + nb'f

where b = ~Il" h the mobil it) ratio. In the derivation neglect terms of order 8 1 . In 51 we drop the c. 1Ii"1: III the presence of a bagiludinal electric Held, find the transverSe c1ec1ric field such that the transverse current \'8flishe5. The algcbm rna)' seem tedious, but the result is worth the trouble. Usc (6.64), but for two carrier t)'PC5: neglect (WeT'f ill comparison with WeT.

4. Cyclotron

~e

for a 'pMroidal ,mergy A4rfoa.

2111, 2m,

Consider lhe energy surUce

E(k)=!l,(~+li + ~).
where m, is the lrans,"en;c mass parameter and "II is the longitudinal mass parametCl". A surface on ....+.ich ~ is constant \vill be. spheroid. Usc the equalion of mctiorl (6), \lith \' ., fl-IV~ol'. to show that We - eBl(mrm.)lI2c when lhe stalic magnetic fleld B lies in the .t,y plalle. This result agrees with (34) when 6 = Td2. The result is in CGS; to obtain 51, omit the c.
5. Magnt"Grl$lflance with tulo (;Q~" type3. Problem 6.9 shows that in the drin \'CIoo-

il)' appro:Qrnatioo the rnotiorl of charge carrien in electric and magnetic 6c1.h does
not lead to transverse magnctoresistance. The result is different \\lith tv.'O carrier

types. COlIsklcr a collductor With a concentration n of c1t;cfrons of effective lllilSli m~ and relaxation lime T~; and a oollCClIlration P ofholt.>5 of c1Je<:ti\'e mass "'" and rcla;o;a00n time 1), Treat the limit of \'t~ry strong magnetic fiekls. "'~T ~ J. (a) Show in this limit that up .. (n - p}eclB. (b) Show that the Hall field is gi\"U! hr. with Q .. "'~T,

w"-('l-P)(;~ +;)-1.,
which \'a.nW!es if n .. p. (c) Show lhat the efi"ecti'"C conducti\ity in the :a: direction is
u<'lI'cB- 2.

,,[(n p) +(PI-PI.(n -+- + p)-'J .

BQ~Qh Q~Qh

If PI .. II, U 0: If n ~ p. u saturates in slrOflg ficlds; that is, it approaches a limit independent of B as B __ 0:>.

References
K. Seeger, St'J""Q).oductor p/~: un j"tl"Oductfoo" 4th cd., Sprillb'er, 1989. It o.I''eI'I. lnt.od"ctWm '0 llppliul rolid .at", J1h~ria. 2nd ed., Plenum, 1990. It A. Smith. ~0f"I. 2nd ed., er"nbridge, 1978. S. M. S, Sffll.ieorlducfor- dl!L'i.I, phlPia, ond technolorw, Wiley, 1965. M. L. Cohen and J. R. Cheliko",'lIky, Ekctro"ic $truc"J~e llrul oplifAll prvperlil'$ 0/ """ieond"".. 'or$, 2nd ed., Springer, 1989. C. W. Bc:enaHa- and II. van 1I0011en. HQuantum tran~port in iielnlconduc:tor naJloOStrtJd"res,~ Sohel sbl<: ph)'Sic:s 44. I (1991). C. Bastard d ai., HEkdronic sbtes in 5oenliconduc:tor hdrtQitruetures. - Solid .Iate physics 'U, 230 (19(1).

9
Fermi Surfaces and Metals
Reduced
WIle

scheme

235

Periodic zone scheme

CONS1'RUCnON OF FERMI SURFACES Nearly free electrons ELECrRON ORBITS, HOLE OR8ITS, AND OPEN ORBITS
CALCUUt.1l0N OF ENERC\' BAf'DS

131
238 !-40 U2
244

Tight binding method for energy bands

Wigncr-Scil2 method.

24S
~8

Cohcsi,'c energy Pscudopotcntial methods

EXPEIUMEfliTAL METHODS 1('1; FERMI SURFACE STUDIES Quanfuation of orbi~ in nlagnet>c lick! De Haas-\-an Alphen effect &tremal orbits Fcnni surface of alpper EXlln.plc: Fermi surface of gold Magnetic bn::nkdown

249 2:12
W 155

251 200 262 262

264

SUMMARY
PR08LEMS
I. Brillouin

26S
266

2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

U)IlC5 of rectangular lattice Brillouin 7.ClnC. rectangular lattice J-Iaagonal elose-padl:ed structure Brillouin zones of t...-o-dimensional dinlcnt metal Open orbits

266 166 266 266 266

266

Cohesive energy for

fl

square well potential

De Haas-van Alphen period of pohl5sium

Band edge structure on k . p perturbation theory Wannier functions Open orbits and magncton::5istance

266
261 267 267 268 268

LaMau

1~'Cb

REFEREf'CES

--

"'"

FlIllre I F"..eeeledron Fermi fluf~ 0<" fcc rncta1I wilh ~ (Cuj and thrt.'e(AI) '-alcnccelairons per prin"">-e ""II. The Fermi .urfae:,., .hown for copper has been ddOrmcd from .. )pIlC:~ to.grt-'e wilh the experimental re~uhs, TIle second zone of aluminum is nearly half_filled with electrons. (A. R. M.cklnlosh.)

-=--

CHAPTER

9:

FER.\II SURFACES ANI> METALS

Few people u;ouId fl~"e a flU'tal Q "a lO1id U'ilh a Fermi Iwrfou." Tllill llIoy Ilft;uthdeu be 'lit! mOISt mMning(1l1 d~"iIio" of II me'alone C(l" give today; i, reprnellllJ a prof 011"" IIflt'lllll;t i" lire IIndenlamling of U;/ly me/alii belrar;e all ,Iley do, I'he OOltCelll of the Fermi tllIrface. 0$ deo;e1oped by filumllllll pllysicN. Ilroddcli a precise exv'Ulwlio,1 of "Ie maill pJlysical propertin of metal,.

A, R. Mackintosh

TIle Fenni surface is the SUr"1ce of constant energy ,.. in k sp.'lCe. lne Fermi surf.'lCe separotes the unfilled orbitals from the filled orbit:lls, at absolute zero, TIle e1ectriClI properties the metal are determined by the Shtlpe of tile Fenni sur"1ce, because the current is due to changes ill the occuptl.llC)' of states near the Fermi surf."1t:e, TIle shape may be very intricate in a reduced zone scheme and yet have a Simple interpretation when reconstructed to lie near the surface of a spllere. We exhibit in Fig. 1 the free electron Fermi surfaces constructed for two metals that have the face-<entered cubic crystal structure: copper, with one mlence electron, and aluminum, with three. The free electron Fenni surfaces were ~'eloped from spheres of radius k,.. detennined by the valence electron COIIcentration. How do we construct these surfuces from a sphere? TIle (:onstruetions require the redll(:ed and pertodic zone schemes.

or

Reduced Zone Scllemf

It is always possible to select the wavevector index k ofallY Bloch function to lie within the first Brillouin zone. 'llJe procedure is knO\\"11 as mtlpping the band in the reduced wne scheme. 1 we encounter a Bloch function written as I,.(r) = e..r"".(r). \\ith k' outside the first wne. as in Fig. 2. we rna)' always find a suitable reciprocal lattice vector G such that k = k' + G lies within the first Brillouin wile. Then

lbk(r) = c"...lldr) = e'''{e-.c.. rIlL(r))

iii

eilt-rIlL(r) = lbk(r) ,

(I)

where ",,(r) .. e~fC'<lIdr), Both e' .c,. fUld Il".(r) are periodic in the crystallattice, so "L{r) is also, whence 1/IL(r) is of the Bloch form. Even \\ith free electrons II is useful to work in the reduced wne MCheme, as in Fig. 3. Any 'rgy k' for k' outside the first wile is equaJ to an 1< in the

,.

___ -_~~,_-:;_7J\(~"' __ :-_~_~_-_-_""f',-"

c
.L

'igure 2 Fiut Brillouin ~OIlC of asquare luUic(:of ,ide n. The "'ll'c"e<.101"I<' <.1lO beClrried into the lint wne I' forming k' -+ G. 1bc W3"V"L'<:tor at a point A on the Ulfl<' bo.mdary is untied by G to the point A' on the opposite hoondary ofthc, same zoue. Shall we ro,,,,t bolh A and A' as I)'jng in t~ If'S' wnr? Bea.U!i(' Ihq> can be mnneded by. rmprocallattk'e ,-m..... we mUll' tbem as one identicaJ point in the ZOOoe.

"

,,

,,

,,

,, ,

,
\,
,.
_~

0
Flnt

L
tQf'C

cl.

HTi~~

---_i

Figure 3 E.lt'rgr-wa'"e'-edor relation 'l - PJ<2I<a'1 r.... free elm"..., udrawn in the :<luttd zone ...,IW"N'. Thll; construction often gh"eli. w;eful ide.dthe O\'O:wl aj~e oflho: band structure of. ~lIaL 1'he bnndl AC ifdisplaco:d by -2m.. gi,a the uwaI rc..e dertmn nJO-e ro.-toqIIoli.-e l. as li~rd b.- the dasb.:d I:'U....~. 11le branm .... C if ~ ~. 2:nf.. p'u the -.w nil''''' for posot"e I. It. tt}Xa1 pol:~tW U(,,) will introduce band g.>p at lho:mges oflbe zone {as atA and A 'J and al the center oflhe:tOne (as at C). The point C ... hen '"ie\O...,d In the "'.ten<k.'CI. mn., .rile'"" &lis at the ftIgn<lfl1w S<."COfld :tOne. The O\'cnll width arod gross r..'IIturC5 oflhe band structure are often indk-ated pmP"'rly by sudl free ek<:lmn bands in lhe rcd,,...:! ~one scheme.

9 Fermi Surf..ce~ lind Metals

237

first zone, where I.: = k' + C. Thus we need solve for the energy only in the first Brillouin zone, for each band. An energy band is a single branch of the t"k \'crsus k surface. In the reduccd zolle scheme we should Ilot be sUI'pl'ised to find different cnergies at the same value of the wuvevector. Each different energy character b:eS a different band. Two wavefunctions at the same k but ofdifferent energics will be independent of each other: the wavefunctions will bc made up ofdiffcrent combinations of the plane wavc components exp(i(k + C). r] in the expansion of(7.29). Because the values of the coefficients C(k + C) will diller for the different bands we should add a symbol, say n, to the C's to serve as a band index: C,,(L: + C). Thus the Bloch function for a state ofw:\vc\'edor L: in the band n can be written

'"
Ibn." = exp(ik . r)II...J.r) =

L C..(k + C) exp(i(k + C) rl c

Periodic Zone Scheme We can repeat a given Brillouin zone periodically through all of w:wevoctor space. To repeat a zone. we translate the zone by a rcciprocallattice \"(~tOf'" Ifwe can translate a band from other zones into the first zone, we can translate a ba.nd in the first zone into every other zone. III this scheme the energy Ek of a band is a periodic function in the reciprocal lattice:
(2)

J lere t"hC is understood to refer to the same ener'b'Y b:lnd as Ek. The result of this construction is knowll as the periodic zone scheme. llle periodic propert)' of the energy also can be seen easily from the central equ.'1' tion (7.27). Consider for example an energy b.'1nd ofa simple cubic lattice as cakulated in the tightbinding approximation in (13) below:
~k = -a - 2')' (<-'Os kxa

+ cos ktl, + 'cos

k~a) ,

(3)

where a and ')' are constants. A reciprocal l:lttice \'ector of the sc lattice is C = (2r.fa)i; if we add this \'ector to k the onl), change in (3) is cos k"n - cos (k., + 2m'a)a = cos (k"n + 2r.) but this is identically equal to cos k.,u. The enerl,'Y is unchanged when thc wuve\'ector is increased by a reciprocallaltice \'ector, so tlll:.t the energy is a periodic function of the wU\"cvector.

238

FlgU", 4 Th~ ene<g)' bands of '" I;n~'&r bllice pIoued in the exlended (8rillo",n), ~......,d. and periodic 7.01\<: schemes.

TIlrctl different zone schemes are llseful (Fig. 4):

111e extended zone scheme in whk:ll different bnnds are dmwn in different zones in v.:avevcclor space. The reduced zone scheme in which all bands are drawn in the first Brillouin rone. The periodic zone scheme in which every b."lnd is drawn in C\'cry zolle.
CONSTRUCTION OF FER:<.n SVRJo"ACES

We consider in Fig. 5 the armlysi.. for a square lattice. The equation of the zone boundaries is 2k . G + G 2 "" 0 and is Sa1tisfied irk terminates 011 the plane normal to G at the midpoiJl1 ofG. The first Brillouin Zl'-~ nftlie square lattice is

'....nn; Surfacu and Mtltlh

I~
0

c,

c,

;(0

C,
0

0_

'.

, ,

, ,,

'"

.J

Figure 5 (al Consln,ctlon in k space orlhc llrsllhn.-e Bnllouin wnes or 11 "luarc bUice. Th., llm.,<: shortest Gmos orthe ...'ciprocal blli<:e \"l~don are indiCllted 1I5 C,. c;., and C 3. The lin..-s d .."'n are In., perpendicular hlStol'S of these C's. (bl On comlmcting alllinc.'!i C:<lui\"k:nt by s)mmetry to the Ihn:<: Ii".,. in (a) ..~obtain the~. in k.pace ...toch fOnn In., flut three 8rilll)l.lin zon.." I. 11lc n.....,ben denote lhe zone to ,,'hicb the regioN belong; the numhcrs here an onlered acrording to the k . of the ,-..' dor C imu,>:d in the C(lIUtruction of the outer boondar}' oflhe region.

the area encLosed by the perpendicular bisectors ofG. and of the three recipro-cal lattice .'edors equimlent by s)'mmetry to G, in Fig. Sa. These four reciprocal lattice vector.; are (2r.ia)k... and (2m'a)~. lbe second zolle is constructed from ~ and the three lectOrs equivalent to it by symmetry. and similarly for the third zone. 11le p~s of the second rul(\ third zones are drawn in Fig. 5b. To determine the boundaries of some zones we have to consider sets of severnl noncCjuivalent recipn:x:al lattice vectors. lllUs the boundaries of section 3" of the third zone are formed from the perpendicular bisectors of three C's, namel)' (2m'alk,; (4wfal~; and (2ma)(\;... + ~). 1he free electron Fenni surface for an arbitrary electron concentration is shown in Fig. 6. H is inconvenient to have sections of the Fermi surface that belong to the same wile appear detached from one another. The det.JChment can be repaired by a tr:ulsfonnatioll to the reduced zolle scheme. We Ittke the triangle labeled 2" :1Jld move it by a reciproca1latlke \ector C "" -(2m'a)k... such th:l.t the triangle reappears in the area of the first Brillouin zone (Fig. 7). Other reciprocal lattice vectors will shift the triangles 2bo 2.:-, 2.t to other parts of the first zone, completing the mapping of the second zone into the reduced zolle scheme. The parts of the Fermi Sllrfat.' falling in the second Zone are now col""e<:ted, as shown in Fig. 8.

".

"'gure 6 Ilrliiouin :ronCi of a square lallice ,n two dimcm;olll. The ol'Ck mo..T1 is a IU...c..,.., ofron~lanl eneq;y rOO" fl'l'e ek-ctronl, it will be the t~nni surf:ace klr scone partirobr \-alue oflhe electron corocmlDIIOIl. The lotal un oflbe iiJkd ~ IfI k SjIC'e depends onIr on the deron ~lration and ill independent of the inlt--nd1On of the ela:ironJ ...i lh the 1attn:. The shape of the .'b11l' sun.c., deprnds on the lallice inlt.'nctioo, ;and the m.pe will not be an e:uod arde in an w:tual b1lice. The Iabo:Is ,,-jth'n tbe fedions of th" o<>nd and third WOC>s rd'cr 10 ;~. 7 .

,>I......

2IId zorw:

3rd

ZOO\<:

"'gun- 7 Mapl'ingoflh e first, 100d. and third Brillouin ~.onCll in the n:duc.,d Wn., sc:hcme, n.., ...,rtionl of the ~~"'OJld 1.0"" in F~ 6 an: pUI 1~t11er into a SQuare hI' translation lhro~ an appropriate n.'riproctlliaiticc >-...:tor. A dllTcn:nt G is nct:dcd b .......h piC<)<: of a zone.

1'-.\

//

,\
'/1
3rd.,...

1/

VI
J...
2nd........

\-----

\'

tlgu..., 6 'J1le rreo" dt.'dron ....' qI'Ii surface of Fig. 6. as viewed in the r('(!....,oo ~ scbeme. The lohaded art.-as represml ocropo.,d dedron lilatn Paru of the F"enni IIJ~ fall In the IIeC'OOd, third, -.d fourth nlOeli The .... rth :r.ono: ill not dl<w.n. The fir$t WOO' is entirdy oc.=pied..

A third zOlle is n.'semhled into n square in Fig. 8, but the pnrls or the Fermi .~Ulf.'l.ce still appear dis(:oflnecled. Wheu we look at it ill the periodic wne sdleme (Fig. 9), the Fermi surfa<.'e forms a lattice of rosettes. Nearly Free Electro,1t lIowdo we go from Fermi surfaces for free electrons 10 Fenni surfaces for Ile.lrly free eledrons? We c.m make approximate constructions freehand by the use of four f.K:ls:

--trt---*---~-N~ --ttl---~---m-, , ,
I

1
I

1
,

tlJ;ure 9 The .'cnni surfao::.., in third ZOIIC 8$ dnwn in the pl'riodic >:one~. 'The Ggure was C'Onlitructed by repeating tJw, tlurd zone of Fig. 8.

ti,,.,

--m---m----m-, ,

, ,

, ,

G
"",~JgUI'"e 10 Qual,tsl;,"e impresMOll ""rface of tOg. 8. At one pot,,! on

of the dTect of a .....,u ".,nod... (.T}~a1 potential on the Fermi each F"'.mi ",n.:c ~ ha,'e Shc:Mll th!.- '""Ctor ~L. In the K'<:Ofld <tOOC the energy i1ICfCllSC.'II",,'3lthc interior ofth" figure, and in the third zone the ,,"crgy i~ rowaro the exterior. The shaded rt,gtOos are filled "itll dt:t"lrons and arc lower in energy tlwl the unihadcd regiolU. We dJalI s thai a ."cnnj wrf3ce like thai of the third UIllC is electronId<e, when:as one like lballll the -=ond wne is hoIelil.e. It is saKllhat clcctrons sink and hoks

lIooo

11,e interaction of the electron with the periodic potenti...I of the cT)'stal causes energy gaps at the zone boundaries. Almost always the Fermi surface \~;II intersed zone bound:uies perpendicularl)' (see below). TIle crystal potential will round out sharp corners in the Fenni surfuees. 1bc total volume cncio:sed by the Fermi surface depend'> on1)' OIllhe electron conccntration and is independent of the details of the bttice interaction. We cannot make qutlntit.1.Ii\'e statements without calculation, but qualitativel)' we eq:>ed. the Fermi surfaces in the second and third 7.ones or Fig. 8 to be cn..1.nged as shown in Fig. 10.

..,
Freeh.'U1d impressions of the Fermi surfaces derived from free electron
5urf.'lCeS are useful. Fermi surfaces for free electrons aft: constructed by n pro-

cedure credited to Harrison, Fig. Ll. TIle reciprocal lattice points are determined, and a free-electmn sphere of radius appropriate to the electron con<''elltmHon is drown around each point. An)' point in k space that lies within alleast olle sphere COfTesponds 10 an OCC\Ipicd state in the first zone. Points within at least two spheres oorrespond to occupied siaies in the second ZOIlC, and similarly for points in three or more spheres. We s..'IkI. earlier that the alkali metals are the simplest metals, \\;Ih weal; interactions between the conduction electrons and the laltK:e. Because the alkalis h:H'c only one VIDence electron per atom, the first Brillouin zone boundaries are distant from the approximately spherical Fermi surface that fills onehalf of the volume of the zone. It is kno\l:n by calculation and experiment that the Fermi surface ofNa is closely spherical, and that for Cs the Fermi surface is deformed by perhaps to percent from a sphere. 'nle divalent metals He and l\Ig also have \veak lattice interactions and nearly spherical Fermi surf.'l(.'cs. Hut because the)' havc t\\'O valence electl'Ons each, the Fenni surface encloses twke the volume in k space as fOr the alkalis. nwt is, the volume enclosed by thc Fenni surf.'"lCC is exactly equal to that of a zone, but because the surface is spherical it extends out of the first zone and into the scwnd zone.
ELECTRON ORBITS, IIOLE ORBITS, AND OPEN ORBITS

We snw in Eq. (8.7) that electrons in a stalic magnetic field move on a curve of constant energy 011 a plane norm.,j to B. An electron on the Fermi surf.'lcc w:ill move in a curve on the Fermi surface, because this is a surface of constant energy. Three types oforbits in ..'I magnetic field are shown in Fig. 12. TIle closed orbits of (3) and (b) arc traversed in opposite senses. Because p..wticles of opposite charge circulate in a magnetic field in opposite senses, we S<1y that one orbit is electronlikc and the other orbit is holelike. Eledrons in holelike orbits move in a magnetic field as if endOWt:d \\;th a positive charge. 11lis is consistent \\;th the trc.'ltment of holes in Chaptcr 8. In (c) the OI"bit is not closed: the particlc on reaching the zone boundary at A is instantly umklapped back to 8, where 8 is equivalent to 8' because they are connected by a reciproca1lauice vector. Such an orbit is called an open orbit. Open OI"bits M\e an important effect on the magnetoresistance. Voc:lnt orbitals near the top ofan otherv.ise filled band givc rise to holelike orbits, as ill Fig<;. 13 nlld 14. A view of a possible energy surf."Ce in three dimensions is given in Fig. 15. Orbits that enclose filled states are electron orbits. Orbits that enclose empty slates are hole orbits. Orbits tlmt move from Zll'" to zone witllOut doSing are open orbits. \

J.'er",i Surfoce, lind Melah

213

rlgUre II Hanion Cl)n$tnx:tioa ~ fn.'C' dcrtron Fermi s..rfaca on the second, lhird, and birth 1XIJlelI lOr a $Quare lattoce n.. Fenni surfac-e encloses the entilr finl zone, which therero...c is filled ,,ill! clcrlrolUi.

II ,1 uf paper

'"

o
.

~
)

rq;ure 12 Motion in a m:'lgneIic field of tile: wa'e>Ioo- of an c~ronon the F"nni surfK.c, in (a) and (b) lOr t-rnni sutra.ti topologicall)' eq..h'&lent to those, ~t~ 10. In (a) the wa'"t:\ectOO" JnO\eli around the orbit in a ~ dio-ection; in (h) the wa'e>"CdOO- mO':eli around the mbit in a countrr_ dCJcko,o,,SC din:c:tion. The direction in (bj Is ....... we e>;ped for. he ekrtron of d1lugt: ~t: the srnalkr " ,....ues ha,'C' the lower energy, li(l that the IlIled ck'rtron sllItes lie IOsidc the Fennl slIr1iIce. W" call thc orbit in {bl dedronlil.e. TIle SCnse ~the motion in. magnetic lleld is opposite in (alto th.t in (b), li(l that we :fer to the orbit ,n Ca) as hoIelil<r. A bole nlO\"" lI.Ii a particle ~ positive charge e. In eel for a n.....tan!,'lJlar zone we sl""" the motion on an opt.'Tl orbit In the periodic zonc scheme. ThIs is topologically intcrmeuiate between a hole orbIt and an dooron orbit.

1'''1Int! 13 (aJ Vacant "bltrs at the romen of an aLnwt-llllcd band. dnown In the ~ 7.()Ile odle.~ (hI In the periodc J't)ne Kbc-me the '-arious parts or the Fenn, wrfact: are ('(WIne.:too. Each eircl", forms a hoklike orbit. Th'" diffi,renl cim..s ....c cntircly l~lui.-aIt.'Tlt to l'aCh other, and the denslly of states is that of a single cird.. CfI,,, orbits """,d not be tr"", ci~k"S: for the lauicc shown It Is only <equin...I that the OrbitJ; have rouri>k1 lifmmetry.)

$------4>~---~-, I I , I , 'I

..

"'.

l.J--~
F"1fUn: l<t Vla,,1 stat~ near the top of an almost filled band in a t\OlO-dimens.onal cryJW. ~ figun~ is equ"'"t to Fig. l2a. l.

l.

",
Figure 15 Constant "'IC'1:>'Y surfuce in the Brillouin :roll" <:l a simple cubic bttice. fur the lOS.umed energy band e, - -a - 2}(<.'OIIi k,.. + IIi k"a + 00:1 k~). (,I) Constant energy SUmt-e e .. -cr. The fllled \dume ./Olltains one eleroon pcI' primiti\'C cd. (b11be same su~ exhibited in the periodic tone 5dlcme. The ron,Mrl...ity ci the orbits is dearly Can you find electron, !wk. and open orb.... b rnoI:ion in ~ic field 8t? (A. Somme&kl and II. A. Bethe.)

"-.-n.

CALCULATION OF ENERGY BANDS

Few masters of energy band calculation learned their methods enlirei}' from books. Band calculation is a craft learned by experience. oncn developed in groups, and needing access to computers. Wigner and SCit7.. who perlormed the first seriOllS band calculations in 1933, refer to afternoons spent on the manual desk cakulators of those days, using one afiemoon for a trial wavcfullolion. Modem computers hav(' eased the pain. However, the formulation of the problem re<luires great care. and the computer programs are not tri\"ial. Here we limit ourselves to three methods useful 10 beginners; the tightbinding method, useful for interpolation; the Wigner-Seitz method, useful for the visuali7..ation and understanding of the alkali metals; and the pseudo(Xltelltial method, utilizing the general theory of Chapter 7, which shows the simplk'it)' man)' problems. Reviews of thcse and other methods arc cited at the end this chaptcr.

or

or

.)
Hgurt' 16 (II) Schema';.: dr:1wing I w""d"unctioos of eledrons Of> lwo h)'drogcn atoms al large ~ration. (b) Cround state "",efull,.1;on at cIost:r Sepllration. (e) Ex<.;l<:d slale w,:<\"cfunclioll.

Tight Binding Method few

Ener~'Y

Bandt

U:ll us start with neutral sl-parated atoms and watch the changes in the atomic energy levels as the charge distributions of adjacent atoms o\'erlap when the atoms arc brought together to fann a crystal. Consider two hydrogen atoms, each with an elcetroo in the Is ground state. The wavefunctions A. "'8 Of! the separated atoms aTc li~m in Fig. 16a. As the atoms are brought together. their v.'3,vcfunctions overlap. We consider the tv.'O combinations I/JA I l/JIJ' Each l'Ombination shares an c1edron with tile two protons, but an electron in the slate l/JII. + I/JIJ will have a somewhat lower energy than in the 51ale !/Ill - 1/18' In lJ'1l + I/JIJ the electron spends part of the time in the region midway between the two protons, and in this region it is in the attractive potential of both protons at once, thereby illCTC3Sing the binding energy. In 1/1", - I/IB the probability densit) vanishes midway between the nudei; an extra binding docs not appear. As two atoms are brought together. two separated energy Ic\'els are IOrmcd for each level of the isolated atom. For N atoms, N orbitals are formed for each orbital of the isolated atom (Fig. )7). As frec atoms are brought together. the coulomb interaction between the atom l.'Ores and the electron splits the energy levels, spreading them inlO bands. Each slate of given quantum number of the free atom is spread in the crystal into a band of energies. 11lc width of the band is proportional to the strength of the overlap interaction bet....'CCn neighboring atoms. There will be bands fonned from P. d. ... stales (I = J ,2, ... ) of the free atoms. States degenerate in the free atom will form different bands. Each will not have the same energy as any other band over any substantial range of the waveve<.1or. Bands rna)' l.'Oincide in energy at certain values ofk in the Brillouin woe.

'"
"

t
-

,.,
-0.2
-1.0

~ -18

-2.6

-4.2
, ! ! ,

I! 3 4 t'>ean:d neighbor dist>Jlc:e. in Bohr nodi;

Figure 17 The Is band uf a ring ufOO h}'drugen alOlll$; the Qne-dcdrOI1 energy .,...Jcul31ed ill the tight-binding approximation with the nearcst-ncighb<r overlap inlegral of Eq. (9).

11lC approximation that starts out from the wavcfunctions of the free atoms is known as the tight binding approximation or the LCAO Oincar C(lmbination ofatomic orbitals) approximation. The approximation is quite good for the inner electrons of atoms, but it is nol often a good description of the l..'Onduclion electrons themselves. It is used to dcS(.,-ibc approximately the d bands of the

trans:tion metals and thc valence bands of diamondlikc and inert gas crystals. Suppose thai the ground stale ofan c1cdnm moving in the potential U(r) of
an isolated atom is 'J'{r), an s state. 11m treatment of bands arising from degenerate (p,d, ... )atomic levels is morecomplicaled. If the influence ofone atom on another is small, we obtain an approximate wavefunction for one electron in the whole crystal by taking

,,(,) ~ L ,

C,,,,, -,,)
rj)

(4)

where the sum is over alliatticc points. We assume the primitive basis contains one atom. This function is of the Bloch form (J.7) if Cig =: N- I12 e1kT!, which gives, for a crystal of N atoms, lJIk(r) = N- 1I2

L exp(ik rJ).cp(r J

(5)

We prove (5) is of the Bloch form. Consider a tr.mslation T (.'Onnecting two lattice points;

9 Ferm, S"rfaca and Metah

241

l/Jk(r

+ T) == N- I12
=

L exp(ik . r)q:{r + T J
J

rJ) T)]'P[r - (rJ - T) (6)

exp(ik . T) N- I12

L exp{ik . {rJ -

== exp(ik . T)l/Jk(r)

exactly the Bloch condition. We find the first-order energy by (''akulating the diagonal matrix clements of the hamiltonian of the crystal:

(killik) ~ N-'
where 'P,,,
!'$

L L c,p('k . ('J J '"

,,,,)] (.".1111."

q:{r - r"J. Writing Pm = '",n - rJ'

(killik)

~L ... c,p(-'k' P.J

dV .'(, - p".)lIo1.')

(8)

We now neglect all integrals in (8) except those on tllC same atom and those between nearest neighbors (.'()nnccted by p. We write

f dv

'P(r)lf~r) "" ~a ;

dV 'P(r - p)H!p{r) "" -)' ;

(9)

and we have the first-order energy. provided (klk) = 1: (kII-flk) "" -a -)'

L ... exp(-ik' P"J = E~

(10)

The dependence of the overlap energy), on the interatomic separation p can be evaluated explicitly for two hydrogen atoms in Is states. In rydberg energy units, Ry = me412Jr.'J-, we have
~Ry) ~

2(1 + ria,) c,p(-pl",,)

(II)

where U() = fr.'J-lnur. The overlap energy decreases exponentially with the separation. For a simple cubic structure the nearest-neighbor atoms arc at
P'" = (a,O,O); (0, u,O); (0,0, u) ,
(12)

so that (10) lxx:omes

E~ =

-a - 2)'(cos k,.u

+ (.'Os

kya

+ cos

k~a)

(13)

Thus the energies are confined to a band of width 12)'. The weaker the overlap, the narrower is the energy band. A constant energy surfuee is shown in Fig. 15. For ka 4: 1, El:'" -a - 6)' + )'k2u2 . The effective mass is m = ,,-2/2"(a2 . \\'hen

'"
the overlap integral 'Y is small, the band is nalTow and the effective mass is high. We L'Oosidercd one orbital of each free alom and obtained onc band e". The numl~r of orbitals in the hand that corresponds to a nonclegcncn\te atomic level is 2N, for N atoms. We sec this directl)': values of Ie within the first Brillouin zone define independent wavcfunetions. TIle simple cubic zone has -.,,(0. < k A < 'fda, etc. 11lc 7.o0C \'olumc is &riJ/a3. The number of orbitals (counting both spin orientations) per unit \'Olumc ark space is V/4r. sotJmt the number of orbitals is 2VltP. Here V is the volume oflhc crystal. and lItP is the number of aloms per unit volurnc. Thus there are 2N orbitals. For the bee structure with eight nearesl neig.hbors.

(14)
For the fcc structure wilh 12 nearest neighbors.

e"

-Q -

4y(cos ik~ cos ik~a

+ L'05 !k::;ll

(:os

ik..a +
cos ik..a
(,'05

lkya)

(15)

A constant energy surface is shown in Fig. 18. WignerSeitz Method Wigner and Seitz showed that at least for the alkali metals there is no inconsistency between thc electron wavcfuoctions orrroo atoms and thc nearly free electron model of thc band structure of a Cfystal. Over most of a band the encrgy may dCpcJld on thc wav~ector nearly as fOf a free electron. J-1owc\,er. thc Bkx:h wavefunction, unlike a plane wavc, wiII pile up chargc 011 thc positivc ion cores as in the atomic wavcfunction. A Bloch function satislles thc wave e<luation
(16)

With p -jh grad, we have

p eil<-ru...<r) = hk ~htJ...<r) + ell...pu~(r) 2 p rllaJr) = (hk)2 e ...Il .,(r) + F(2hk p)u...<r) + e.....p2uIiJr)

thus (16) rna)' be written as an equation for

1I~:

(~ (p + ilk? + U(r) lI...<r) = f"~II.(r)

(17)

At k "" 0 wc have l/Jo "" uo(r), where uo(r) ha.~ the periodicit)' of the latlicc. sees the ion cores, and ncar them will look like the wavefunction of the frec atom. It is much easier to lind a solution at k "" 0 than for a general k bct."3use at k = 0 R nondcgenerate solution will have the full symmetry of the I..TYStal. We

Fl'rmf Surfoon ..nd /lfl'tal,

'.
\

I'-igul'<: 18 A mlul~1l1 enc:'ltr suroce of an K'C

Cflo~ul ~n>c:tun:. tlilhe ~l-nelghbor11f6i1'

biooUl'; aprlfOIlimaliort. The

~--c>+2M

~tltUce

sho.-n bas

can thcn use llo{r) to construct the approximate solution

l1Jl,. "" exp{ik . r)UQ(r) .

(18)

111is is of the Bloch fonn, but 110 is not an exact solution of (17): it is a solution only if we drop the term ill k . p. Often this tenn is treated as a perturhaHolI, as in Problem 8. l1lc k . p perturbation theory developt:d there is especially useful in finding the effective mass m at a band edge. Ihx:ause it takes account of the ion l.'()rc potential the funl.'tion (18) is a much better approximation than a plane wave to the correct \\'3.vcfunction. l1le ene'l.'Y of the approximate solution depends on k as (ftk'fl2m, exactly as for the pl.lllC wavc, even though thc modulation represclltcd"by uo(r) may be very strong. Because Ilo is a solution of
( _I_ p2 +

2m

U(r) 1I,,(r) "" oUo(r) ,

(19)

the function (18) has the energy expectation \alue Eo + (fr..2 f.!l12m). 1be functiou 1I00r) often will gi\'c us a good Ilicture of the charge t.\istribution witl,in a cell. Wigner and Seitz developed a simple and fairly accurate method of caku Iatillg lIoCr). Figure 19 shows the Wigner~Stdtz wavefunction for k "" 0 in the 35 conduction band of metallic sodium. The function is practically constant over 0.9 ol"the atomic volume. To the extent that the solutions for higher It may be approximated by cxp(ik . r)uo(r), the wavefunt.1ions in the conduction band will be similar 10 plane waves ()\Ier most of the atomic volumc. but inl.Tease markedly and oscillJte within the ion COTC. Cohaive Energy. The stability of the simple metals with respect to free atoms is caused hy the lowering of the energy of the Bloch orbital with k = 0 in Ihe crystal compared to the ground valcnce orbital of the free atom. The effect is illustrated in Fig. 19 for sodium and in Fig. 20 for a linear periodic potential

----.-. -.-.-

1.~a1.

A.. 0

----.
-

/ v
o
F. SciIL)

klal. A Brillouin -...oj

""".....,

Flgure 19 Radi3l ..'3\..,fUnctions b the 3s- orbital oCfree S<ldium alom IInll b 3J lVlldlKiioll haf~ in sodium metal. 1be " ....-efunctUui, ",-hom :or., not .-maIized here, are bond by inl~ing the SduMingn equation b an electron io the p<lt'mtialweU of lIII I'a' ion cor",. For the ~a1om the ..-a,'d"Wldioo, is integrate.:! su~ 10 the us.w Sd.rtldinger boundary condition 9<r) ..... 0 as r ..... ""; the energy e.gem.-alue is -5.15 .,V. 1be ....'..,(unction be- ...a~ k "" 0 in the metal is subject 10 the Wigner-Sei17. boundary rond.itif.w, that d+tdr"" 0 wnen r is meN")' bec..-een ~~ atoms; the ~orthisorbililis -8.2 eV. oonsidrrably Iov.-er than uthe!rft atom. 1beorbitaJs lII. the ~., boundary are not lifled in sodium; their energy is -+2.7 .,V. (Aftn E. WIgI"Iet and

.,"

-0.45 -0.6

1
~

-0.8 -

-1.0

...--

Figure 2(l Ground orbital (k" 0) energy (or an eJedron in a periodic iquanl ~Il potmtial d' depth IUG!- 'iJotfJ. 1be energy i$ lo...ered as the ""lis "ODIe ~ lose1her. Ikr., G is hekl cantanl and b is~. I..aq;e biG conupoods to separlItcd atoms. (Cour1ny ole. Y. FOfC.)

Fermi Surfacu and

MetcJ~

~.'J

-515'.V~

"-tom

Cround .lal,
(".ol"",;,~

,_

:\lclal

F"':.'l'i ""<:1

......rgy

--J--FIgure 2l Cohcsi~e energy uf sodium m.:tal is the differen<'C I>et"tlCn the average energy of an electron in th" melal (-ti.3 eV) and the Jqound sbte energy (-5.15 eV) of the vaI"n.:e 38 ek'Ctron in the ~ atum, referred to an Na ion I'lu$ fi-t..e "Jed:run at infinite s"parntion.

'3.'..",A,... === nge en"'ltY

-8.2.'/----~ - 0 ....e

ofattmctive square wells. The ground orbital energy is much lower at the actual spacing in the metal than for isolated atoms. A decrease in ground orbital energy will increase the binding. The decrease in ground orbital energy is a consequence of the change in the boundary l.-'Ot1dition on the wavefundion: The Schrticlinger boundary condition for the free atom is I/!{r) _ 0 as I" _ <Xl. In the crystal the k = 0 wavefunclion 1I00r) has the symmetry of the lattice and is symmetric about r = O. To have this, the normal derivative of I/J must vanish across every plane midway between adjaceltt atoms. III a spherical approximation to the shape of the smallest Wigner-Seitz l.-"CII we use the Wigner-Seitz boundar)' condition

(20)
where 1"0 is the radius of a sphere equal in volume to a primitive cell of the lattice. In sodium,'-o = 3.95 Bohr units, or 2.08 A; the half-distance to a nearest neighbor is 1.86 A. The spherical approximation is not bad for fl.-'C and lx:c structures. The boundary l"Ondition allows the ground orbital wavefunction to have much less l"llrvature than the free atom boundary l'Ondition. Much less curvature means much less kinetic energy. In sodium the other filled orbitals in the conduction band can be represcnted in a rough approximation by wavefunctions of the form (18), with
Ek=EO+-- .

,'Ii'
2m

The Fermi energy is 3.1 eV, from Table 6.1. The avemge kinetic energy per electron is O.60fthe Fermi energy, or 1.9 eV. Because EO = -8.2 eV at k = 0, the average electron energy is (E.J = -8.2 + 1.9 = -6.3 eV, compared with -5.15 eV for the valence electron of the free atom, Fig. 21. We therefore estimate that sodium metal is stable by about 1.1 eV wilh respect to the frC'"" <l.tom. This result agrees well with the experimental value

.52

1.13 eV. We have neglected several rorrcd:ions whose overall effect in sodium is small.

Pseudcpotenti.o1 Method.!

Conduction electron wavefunctions arc usually smoothly varying in the region between the ion cores, but have a complicated nodal structure in the region of the cores. This behavior is iIIustratcO b) the ground orbital of sodium, Fig. 19. It is helpful to view the nodcs in the oonductiOfl electron wavefunetion in the core region as created by the requirement that the function be orthogonal to the wavcfunctions of the core electrons. This all comes out of the Schrfi.. dinger equation. but we can see that we need the flexibility of Iwo nodes in the 3s eonduction orbital ofNa in order to be orthogOllal both to the Is core orbital \~ith no nodes and the 2.s core orbital with one node. Outside the <.'ore the potential energy that acts on the conduction electron is relatively weak: the potential energy is only the l'OuJomh potential of the singly-eharged positive ion corcs and is rcd:uced markedly by the elcetrostatic S(:reening of the other condudion electrons, Chapter 10. In this outer region the conduction electron wavefundions are as smoothly varying as plane waves. If the conduction orbitals in this outer region are approximately plane wavcs, the energy must depend on tile WB1.'C\'Cdor approximately lIS E" ;; 1j2f(ll2m as for frcc eledrons. But how do 1.\'C treat the conduction orbitals in the core r(.-gion where the orbitals arc not at all like pla.ne waves? What goes on in the core is largcl)' irrelevant to the dependcnl'e of E 011 k. Recall that we can calculate the enet"J.,'Y by applying the hamiltonian operator to an orhital at any point in space. Applied in the outer region, this operation will give an energy nearly equal to the free electron energy. "J1lis argument leads narutally to the idea that we migllt replace the aliual potential energy (and filled sheJJs) in the core region by an effective potential energyl that gives the same W3vefunctions outsiJe the core as arc given by the actual iOIl cores. It is startling to find that the efTecti1."C potential or pseudopotential that satisfies this requirement is nearly 7.oro. This conclusion about pseudopotenllals is supported by a large amount ofempirical experience as well as by thcoreHcal arguments. The result is referred to lIS the cancellatiOJJ theorem.

'J. C. Philli~ lind 1_ KJei'unall. Ph)'>. f\ey. 116,281 (195B), E. Antoncik, J. Phys. o,em. SoIK1s 10, 314 (1959). TIle generalthcay of pseudoalOtentiais iJ disnJssed by 13. J. Auston. \' Heine. and L. J. Sham, Ph}"S. Rl'\'. 121,216(1962), see also \'01. 2~ ofSoluhtDte phynct. The utililyoCthe emply rore model has been known lOr ffilI,fly yean it goes 1-* to Eo Fenni. N,IO\'O Ciml':nto 2. 157 (1934). II. Hellmann, Ada f'h)'>ioclllmiul URSS I, 913 (1935), and II lIellmann and W Kassatoldkin, J <:hem. I'hrs. 4. 3l!4 (1936)....ho wrole ~Sinre the Geld rllhe: ion dele,mined in tnis

wa, rimS

Ii

",tner lhrl <."QfJ~, it is suHicil:nl in the frnl appnWmation 10 set the:.....Ien> electron in

the latliooo equal to plllfle "'3,-..,. ~

Fermi Sur/ace- and Metal.

The pseudopotential for a problem is not uuique lIor exact, bllt it may be very good. On the Empty Core t-,Iodel (EeM) we can even take the unscreened pseudopotential to be Ulro illside some radius R..:
U(,)
~,

0 ( -(>/r ,

forr<R., ; for r> R...

(21)

This potential should now be screened as described in Chapter 10. Each (.'Omponent U(K) of V(r) is to be divided by the dielectric eonslant E(K) of the electron gas. If, just as an example. we use the Thomas-Fermi diehx:tric function (10.33), we obtain the screened pseudopotential plotted in Fig. 223. The pseudopotenlial as drnwn is mueh weaker than the true potential, but the pseudopotential was adjusted so that the wavefunctiou in the outer region is nearly identical to that for the true potcntial. In the language of Sl.-attering theor)', we adjust the phase shifts of the pseudopotential to match those of the true potential. Calculation of the band strul1.ure depends only on the Fourier (.'Omponents of the pseudopotential at the redprocal lattice vectors. Usually only a few values of the coefficients U(G) are needed to get a good baud strueturc: sec the U(C) in Fig. 22b. 111ese coefficients are sometimes calculated from model pot~ntials, and sometimes they are obtained from Ilts of tentative band structures to the results of optkal measurements. Cood values of V(O) (.-a11 be estimated from Ilrst prineiples; it is shown in (10.43) that for a sercened Coulomb potential U(O) := -!E,.. In the remarkably successful Empirk-al Pscudopotential Method (EPM) the band structure is calculated using a few eocfficiel1ts V(C) deduced from theoretk-aJ fils to measurements of the optical reflectance and absorption of crystals, as discussed in Chapter n. Tables of values of V(C) arc giv~n in the review by M. L. Cohen and v. Heine. Charge density maps can be plotted from the wavefunctions gCllernted by the EPM-see Fig. 3.11. TIle rc.~ults are in cx(.'Cllent agreement with x-ray diffraction determinations; sllch maps give an understanding of the bonding alld have great predictive value for proposed n~... strueturcs and (.'Ompoullds. 111e EPM values of the (.'OCfficients U(C) oftell are additive in the contributions of the several types of ions that are present. 1110s it may be possible to (.'Onslruct the V(C) for entirely new structures, starting from results on known structures. Furlher, the pressure dependence of a band structure may be determined when it is possible to estimate from the form of the U(r) curve the dependence of U(C) on small variations of G. It is often possible to calculate band structures, whesive energy, lattice constants, and bulk moduli from first principles. In such ab initio pseudopotential caleulations the basic inputs arc the crystal !itructure tn>e and tIle atomic number, along with well-tested thcoretk-al approximations to eX(."'hange energy

r, in unit. of Bohr ..01;;_

o.<II-------+------Jf--4'--------j'~------__i

-----~~~:~~-----J
:
y""~

tiro\. __

..v,wt,."t'" .-

.f -,
~

......~;.-:....

:";~..!'~","'''''''''''

,<J,

P"'':';';;';-'''-

~ ,

.'

Ionic potentia!

Figure 22>l Pseudopoltmtial ror metallic sodium, based on the empty <.We model and screened by the Thomas-Fermi dielectric function. The calculations Were made ror an empty COre rod'u. R.,J.66ao, where 00 ill the Bohr rndh.., and ror a >ereeninl': parameter 1:.00 "" 0.79. TIl" dashed cu...'e shows the a~.umcd uoscreened potential, as from (21). The dotted ell......, ill the actual potential of the ion core, other values of Vir) ,...... -50.4, -11.6. and -4.6, for r" 0.15, 0.4, and 0.7, respec" lively. Th". the actual potential of the ion (chosen 10 fit the energy [",~!.I; of the free atom) ill vuy much larger than the pscudopotential, ovcr 200 timC$ larger at r - 0.15.

O~----'c~'----~c~.'--:7''''':::;:::::'''"--

c,

Figure 22b A typical re<:iprocal .~ pseudopotential. Values of U(k) lOr wavevecton equlIl to the reciprocal lattke vectors, G, are indicated by the dnts. For "ery small k the potential lIpproaches (-213) limes the Fermi energy, which is rhe screened-ion limit ror metals. This limit is derived in Chapter In (Arter M. L. Cohen.)

Fermi Suif_ ond Mt:tot.

-"

terms. This is not the ~me as calculating from atomic number alone. but it is the most reasonable basis for a fin-t-principles calculation. The results of M. T. Yin and M. L. Cohen, Phys. Rev. n 26, 5668 (1982), are compared with experilIlent in the table that follows.

........
LatlU

~,-'"

(I.>

.....,

,Wi<
nxxIulus
(~II-f)

(."

Silicon

Calculated
Experimental Cennanium

5.45 5.43 5.'" 5.65

3.60 3.57

4.84 4.63
-4.26

0.98 0.99 0.73 0.77 4.33 HI

Calculah:d
Experimental Diamond

3.85 8.10 7.35

Calclllawd
Experimental

EXPERIMEI\'TAL METHODS IN FERMI SURFACE STUDIES

Powerful experimental methods have been developed fur the determination or Fermi surfaces. The methods include magnetoresistance. anomalous skin effect. cyclotron resonance. magneto-acoustic geometric effects. the 51mbnikow-de Haas effect, and the de Haas-van Alphen effect. Further information on the momentum distribution is given by positron annihilation, Compton scattering. and the Kohn effect. We propose to study one method rather thoroughly. All the methods are useful, but all need detailed theoretical analysis and are not for beginners. We select the de Ilao.s-van Alphen effect because it exhibits very well the characteristic periodicity in liB of the properties of a metal in a uniform ma!,'Jletic field.
Quantiz.alion of Orbiu in a Magnel&c Field

The momentum p ofa particle in a magnetic field is the sum (Appendix C) of two parts, the kinetic momentum PIc... = mv = Ilk and the potential momcotum or Held momentunl Pf1dc:l = qAlc. where q is the charge. The vector potential is related to the magnetic field by B = curl A. The total momentum is (CCS) P = Pk.n + Pr",1d "" Ilk + qAlc (22)

In SI the factor ..-I is omith,..d .

Following the semicla~sical approach of Onsager ane. .fshitz., we assume that the orbits in a magnetic field are quantized by the Bohr-Sommerfeld relation

P . dr "" (n

+ ')')2'l11i

>

(23)

wheu n is an integer and')' is a phase correction that for free electrons has the value t. Then

fp'dr=fhk'dr+ : f A ' dr .

(24)

The equation of motion of a particle of charge q in a magnetic Aeld is

elk q dr ,-=--,8 dt c dt
We integrate with respect to time to give

(253)

f1k",,!L rxB ,
c
apart from an additive COflstant which does not cootribute to the Anal result. Thus one of the path integrals in (24) is

flik.dr=: trxB.dr=-:Bfrxdr=-;$,

(25b)

where $ is the magnetk: nul': contained within the orbit in real space. We have used the geometrical result thai

II

dr

2 X (area enclosed by the orbit) .

The other path integra.! in (24) is

-

qf"J
c
A'dr=-

curIAdu""-

"J
c

Bdu=-cI>

"
c

(25c)

by the Stokes theorem; here du is the area element in real space. The momentum path integral is the sum of (25b) and (25c):

tpdr=-:$=(n+')')21T'f1.

(26)

II follows that the orbit of an electron is quantized in such a way that the nux through it is
. (27) lt 1be nux unit 2Trflde = 4.14 x gauss em or T mlt. In the de Haas-van Alphen effect discussed below we neeclthe area of the orbit in \\-'aveq!(:tor space. We obtained in (27) the nux through the orbit in real space. By (25a) we know that a line element 6,. in the plane normal to 8 is
$ .. = (n

+ ')')(21T'f1de)

10- 1

related to 11k by 11 (hcleB) t:.k, so that the area S" ill k space is related to the area An of the orbit in r space by
(28)
It follows that
In ""

fuo)2 1 21rllc (-;BS" = (n + 1')-,-

+ 1')-B h,
2".
.

(29)

from (27), whence the area of an orbit in k space wiII satisfy

Sn "" (n (30)

In Fermi surface experiments we may be interested in the increment t:.B for which two successive orbits. n and n + J, have the same area in k spal.'(! on the Femli surface. T~e areas are C.lual when
(31)

from (30). We have the important result that equal increments of liB reproduce similar orbits-this periodicity in 1/B is a striking feature of the magnetooscillatory effects in metals at low temp~~ratures: resistivity, susceptibility, heat capacity. The population of orbits 011 or near the Fermi su-rface oscillates a.~ B is varied, causing a wide variety of effects. From the period of the oscillation we re<.'onslruct the Fermi surfuce. The result (30) is indeJX)ndent of the gauge of the vector potential used in the expression (22) for momentum; that is. p is not gauge invariant, but Sn is. Gauge invarianl.'C is discussl.'<i further in Chapter 12 and in Appendix C.

De Haas-wn Alphen Effect

The de llaas-van Alphen effect is the oscillation of the magnetic moment of a metal as a function of the static m3b'T1etic field intensity. The effect can be observed in pure specimens at low temperatures in strong magnetic fields: we do not want the quantiZ<ltion or the electron orbits to be blurred by collisions. and we do not want the population oscillations to be averaged out by themlal population of adjacent orbits. The analysis of the dl-lvA effect is given for absolute zero in Fig. 23. The electron spin is neglected. The treatment is given for a I\llo-dilllensional (2D) System; in 3D we need only multiply the 20 wavefunction by plane wave factors exp(ikzZ). where the magnetic field is parallel to the z axis. The area of an orbit in kx kll space is quantize<! a.~ in (3O). The area between successive or~ bits is
t:.S "" Sn - S,,_l = 2'lreBlhc .
(32)

.-,

"

"

8-0

..

..

,
(.) (1)) fe)

Id

CO"

.'gur", 23 E.:plmation of the de Ilus-\'&Il AI~ dfttt for II. free eltron ps in two di~ in II magnetic field. TIw: filled orbitals of the Fcnnl _ mille ~ or. m.gnetic foeld are shaded in" lind d. The energy le\.ds in. magnebc flt'ld are shown in 11, c. and r. In b the field has. ,.... ue 8, such that the total Cllt'rw of the electroos is the same &$ In the ab5ence of. magnetic fw:ld: lIS many dectrons ha\"e their energy n.~ as lowered by the orbital qU"ntrt.ation ill the magnetic field 8,. When ~ illO"esse Ihe lldd 1(1 St the lobI electron energy is increased. because the upperJrlOl electrons h:,ove th('ir energy r:liald. In e ror Geld BJ tbe energy ;. BglIIl equal 10 thai !Dr the field 8 - O. The lotal energy is a minimum lit points Jtoch as 8 8,.. 8:10 .... and I Il'IVlmum near points such as 8", 8.., . . . .

The area in k space occupied by a single orbital is (21rlL)2, negleclillg spin,

for a square specimen of side L. Using (32) we find that the number of free electron orbitals thai coalesce in a smgle magnetic level is

D
2

= (2-rreBlflc)(U21r)2 = pB ,

(33)

where p = eL /27Tf1c, as in Fig. 24. Such a magnetic level is called a Landau level. Tbe dependence of the Fenni level on B is dramatic. For a system of N electrons at absolute zero the Landau leveL~ are entirely filled up to a magnetic quantum number we identify by s, where s is a positive integer. Orbitals at the next higher level s + 1 will be partly filled to the extent needed to accommodate the electrons. The Fermi level will lie in the Landau level s + 1 if there are electrons in this level; as the magnetic field is increased the electrons move to lower levels. When s + 1 is vacated, the Fermi I~'el moves down abruptly to the next lower level s, The electron transfer to Iov.'er Landau levels can occur because their degeneracy D increases as B is increased, as shown in Fig. 25. As B is increased there occur values of B at which the quantum number f'r the uppemlOst filled

(a) Allowed electron orbitab in ~ dimensions in IIbsence of a m.gnctlc 6dd. Cb)ln a rTW.gIldlc fiekl the points ...him reprc:5l."lllllw: orbitals ofln:e elcrtrons may be ~iewed as remicted to J:ird6 in the onner-Ic);., plane. lhe SUC'CeSS;"-e circles allTrSpOOC! to ~"f: nlues of tm: quantum number n in tbe enngy (n - l)boo... 11M, _ between ~"e cirdes Is

.';guN: 14

lees)

n.e angular positiln of the points has nosigniflC:lU~. n.e numbttofQlbilahoo .cirde il; constant
..Ill is c.'qual to the &rCIl between ~'f: circles times the number of orbilab per unil area in (a). or {2_Bflu:)(IJ<l7if '" LadJI2Trflc, negkcting e1ectroo ~pln.

"

"

4060800 11(II)

23 10018(1))

Figure 2.5 (s) The heavy line gives the numbcrof ()Mliclcs in levels which an: complctelyoccuptcd In a maJ:netlc field 8, for a two-dimensional I)'stc.n with N .. 50 and p" 0.50. The .haded .. rea Rives the number of p;lrticles in '",-cls pnrtially occIIllied. The value of. <knot"S the quantum ""mk. of the highest k'Vd which is <'OInpletely filled. Thus at 8 .. 40 we have, .. 2: the Ie_d. II .. I and n ... 2;1,re lilled and th~ lin 10 l)Qrticle. in the level n .. 3 AlB" SO the le'o'e1 n .. :I is empt)', (b) The periodicity in UB i. C\-'KIcnt when the same polnts "Il! plotted ag;linst UB.

level decreases abruptly by unity. At the critical magnetk fields labeled B. no level is partly occupied at absolute zero, so that
spB. = N .

(34)

The number affilled levels times the degeneracy at 8. must e<lualtl1e number of electrons N.

'"
10 show the periodicity of the energy as B is varied, we use the resull that the energy of the Landau level of magnetic quantum number n is E" = (n - i)liwe where We = eBlm*c is the cyclotron frequency. The result for E" follo\\ls from the analogy between the cyclotron resonance orbits and the simple harmonic oscillator, b\ll now we have found it convenient to starll'Ounting al 11 = 1 instead of at n = O. The lotal energy of the electrons in levels that are fully occupied is

.-.

L DhwJn -

!) ~ IDhw.,'

(35)

where D is the number of electrons ;n each level. The tolal energy of the electrons in the partly occupied level s + 1 is

"wAs + i)(N -

sD)

(36)

where sD is the number or electrons in the lower fillet! levels. The total energy of the N electrons is the sum of (35) and (36), as in Fig. 26. Tne magnetic moment J.J. of a system at absolute zero is given by I-t = -aUlaB. 'Ine moment here is an oscillatory function of liB, Fig.-27. This oscillatory magnetic moment of the Fenni gas at low temperatures is tile de Haasvan Alphen effect. From (31) we see that the oscillations occur at equal intervals of liB such that

6(-'B-) ~ 2 .. f1cS'

(37)

where S is the extremal area (see below) of the Fermi surface normdl to the direction of B. From mea~urements of 6.(118), we deduce the corresponding extremal areas S; thereby much can be inferred about the shape and size of the F elTIli surface. Extremal Orbits. One point in the interpretation of the dHvA effect is subtle. For a Fermi surface of general shape the sections at different values of kB will have different periods. The response will be the sum of contributions from all sections or all orbits. But the dominant response of tile system comes from orbits whose periods are stationary with respect to small cllanges in k ll. Such orbits are called extremal omits. Thus in Fig. 28 the section AA' dominates the observed cyclotron period. The argument can be put in mathematical form, but we do not give the proof here (QTS, p. 223; Ziman, p. 322). Essentially it is a question of pha~e cancellation: the contributions of different nonextremal orbits cancel, but near the extrema the phase varies only slowly and there is a net signal from these orbits. Sharp resonances are ohtaineU even from complicated Fenni surfaces because the experiment selects the extremal orbits.

'"
1

T.... ""'"

~
~

l!

" "

,
lOOiIl-

,
n.e

fiKure 26 "The UPPft"cun~ is the total dectroniIC energy \eTSUS JIB. The oscin.tions in the energy U may be detected by mcaliuremcnt orthc magnetic !noment. gi"cn by -iJUlaB. thermal and transport prur,Joerties of the metal also OK'iUate as .uc::cflSi'"C orbitlllle\-eb cui througt> the Fermi

""-'1.'1 "ben the field i5 i",-~. The ~hadet.l region in the figure givt':$ the contribution to the L"OCflO' rom Ic..eb; that We 001)' portly 1il1eO. -ow, panmrters ro.- the flgllre an the same u lOr F"~. 25, alld ~ h.ve laken tile units of B such that 8 .. fllR~.

-I

F"'U",27 At Ibsolute zero the ml\8l'C\~ 'TlOfl'\Cnt is g"en by -aUlilB. "11Je energy pk>ued III Fog.. 26 kads 10 the magnetic moment 51....." here, m oscillatory function of UB In impure SJlC<.~mrt'li the Cl5Cillalions are smudged oul in part bece."~e tl'e energy lcvel$ are 1\0 IOllga- s~rpl)'
dcllned.

".

.';gufe 28 The orbits In the $Cldion AA' IIrc utrcrml orbits: the cyclotron period is roughly COlUtant O\"er II 'f'llSonable -'lion cf the Fermi surface. Other ~k>flS Juch Q 88' ha\'e orbits that "*"y in period along the
~ion.

Fermi Surface a/Copper. The Fenni 5Urtace of copper is distinctly nonspherical: eight necks make contact with the hexagonal faces of lhe rU'S1 Brillouin zone of the fcc lattice. 111e electron concentration in a monovalent metal with an fcc structure is " :::: 4/0 3, there are fouf electrons in a cube of \lolume if. The radius of a free electron Fermi sphere is
(38)
and the diameter is 9.8Q1a. 11Je shortest distance across the Brillouin zone (the distance between hexagonal faces) is (21'1fa)(3)1n "" 10.8810, somewhallarger lhan the diameter of the free electron sphere. The sphere does not touch the zone boundary, but we know that the presence of a zone boundary tends to lower the band energy near the boundary. Thus it is plausible that the Fermi surface should neck out 10 meet the closest (hexagonal) faces of the zone (Figs. 18 and 29). The squve faces of the zone are more distant. with separation 12.57/0. and the Fermi surface does not neck out 10 meet these faces.

EXAMPLE, Fermi Surface of Cold. In gold for (Iulte a wide range of Held din.'CCions 5110enberg Hnds the magnetic momenl has a perioll of 2 x 10-' gauss-I. This period l'Ofn.-sponds to an e:lllremal orbit of area

S
:=

2m1Ac
6(118)
ill"

7 9.55 x 10 2 x 10-'

ill"

8 16 -t 4. x 10 cm .

From Table 6. 1. we ha"e k F '"' 1.2 x lOS em-I ~rafree electron Fermi sphere forgold. or all edremal area of 4.5 x 10 111 em-to in genera.! agreemenl y,ith the experimental \mue. 1be actual periods reported by Shoen~ arc 2.05 x 10-8 gauss-I for the orbit B lll of Fig. 28 and 1.95 x 10-g gauss-I lOr B u In thc (1111 W.l"l'CCion in Au a large period of 6 x 10-8 gauss- I is also found; the l'orTCspomling orhilal area is 1.6 x lOIS cm- 2. This is the "neck" orbit N. Another extremal orbit. the "dog's bone." is shown ;0 Fig. 30; its area ill Au is about 0.4 d thc belly area. upcrimental results are shown in Fig. 31. Todo the cumpJe in 51, drop c from the relation for 5 and use as the period 2 x 10- 8 tesla- I

" ..

Figure 29 f eon; surface of OOl'per, after PiI'Pord. The Brillouin :wne of the n, $tructure is the lrunClItoo geb. hedroo derh"t'd in OlapIer 2, Fermi surface makes COI\bct with the boundar) " lhe tttlier of the heugpnaI faces of the :woe, in the I11I1 directions in k space. 1\0'0 DeIIyR e:drema! orbits are shcMTI. dmoced by 8; lhe extremal "ned;- oottl is denoted by N.

n.e

""',

I'igure 30 Dog's bone orbil of.... electroo 00 the fermi surface of <."Opper or gold if! a ml@:llebe ficld_ This orbil is dauifoed as hoIt:ti.lt: btuuse lhe enerxY increases toward the interior of the

45.0kG

45.5 kG

46.0kG

rlgUrc 31 De Haas-'"" Alphen dJect in go&d ",;th B I (1IOj. The OIdllation is from the dq;"s boo., orbil of Fi&- 30, 'Ole signal is ft;b.ted to the StII'ld deri'....i'e of the m.gnebc- monw:nt "';lh ~ to Reid. The results Wrft; obtained by. field modulation technique in. high.homogeneity "~ng "-"enoid at about 1.2 K. (Courtesy oil. M. Templetoo.)

Figun: 32 Multipl)'-c:oonectcd hole ~urtaee of ,nagnesium in b.nds I and l, a<:oording to L. ~l FaJicoy. (Dnr--i~ br MarU Pucbla.)

-(
-

I)

(D

} (I

(D

Figun: 33 11.",,1.:00.'11 of band structure by. strong n\.lgndic field. llrillouin zone boundaries are the lighllines. The froc ek'drOIl orbils (a) in a .trollg field mange col1lledivity in. wcal.: field (b) to bec:omc oslet1 orbits in the first band and electron ..hiu in the sa:ond ~. Both t-nds are

..

'"

mapped togethn.

The frl.'C ek-'Ctron Fenni sphere of aluminum fills the first 7.onc entirel)' and has a large overlap into the second and third zones. Fig. 1. The third zone Fenni surface is quite complicated. even though it is just made lip of certain pk"'CCS of the surface of the frcc electron sphere. lne free electron model also gives small pockets of holes in the fourth ZOIlC, bllt when the lattice potcntial is taken into account these empt)' out, the electrons being added to tlle third zone. lne general features of the predicted Fermi surface of aluminum are quite weJl verified by experiment. Figure 32 shows part of the free electron Fenni surface of magnesium. Magnetic Breakdou,-n. Electrons in sufficicntl) high magnetic fields will move in fn,"C particle orbits, the circular C)'c1otron orbits of Fig. 33a. Ilere the rnagnetic forces arc dominant, and the lattice potential is a slight perturbation. In this limit the classification of the orbHals into bands mar have little importance. 1I00:ever, we blOW that at low magnetic fields the motion is descrilx.>d

9 I:",.",i Sur/act. and Mdab

by (8.7) 'with the band structure Eic that obtains in the absence of a magnetic field. The C\'entual breakdown of this description as the magnetic field is incrcasw is calk>d magnctic breukdown,2 The passage to stl'Ong magnetic fields rna)' drastically change the conncctivity of the orbits. as In the figure. The onset of magnetic break~\ll will be rC"eaIed by physical properties such as magnctoresistance that depend sensitively on the connectivity, The condition for magnetic breakdov.ll is that IiW..E~ > appl'Oximately, Ucre E~' is the free electron Fenui energy und Eg is the cnergy gap. 1l1is condition is much milder. espcciall)' in metals with small gaps. than the naIve oonditlon that the magnetic splitting liw.. exceed the gap, Small gaps may be bund in hcp metals wl1ere the gap 3CI'05S the hexagonal face of the 7.Qne would be zero except for a small splitting introduced by the spin-orbit interal.'lion. In Mg the splitting is of the order of 10- 3 eV; for this gap and EF - J 0 eV the breakdown condition is Iiw.. > J O-ll eV, or B > 1000 G.

e:.

SUM~lARY

A Fermi smface is the surlace in k space of constant energy equal to E~" The Fenni surface separah.-s filled states from empty states at absolute zero. The form of the Fenni surlace is usually exhibited best in the reduced zone scheme, but the connectivity of the surfaces is clearest in the periodie zone scheme.
An energy band is u single bnlllch of the

E" versus k surlU(.'c,

The cohesion of simple metals i<; accounted for by the lo\vt.>(ing of energy of the k = 0 conduction band orbital when the boundary conditions on the wavefunction are changed from SchrOdinger to Wigncr-Seitz.
1l1c pCliodicity in the de Hans-vall Alphen effet.'t measures the extremaJ

cross-section area S in L: spat.'C of the Fenni surlat.'C, the cross section being taken perpendicular 10 B:

IR \\. Stark and L M. Falicov,

~Magnctic

BreaJ<d.:,wn In MetaIi, ~ in lAw lanprrat..,.., phyI.

a, Vol. v, ,,"orth Holbnd. 1967, pp. 23.5--266.

'"
Problems
l. Brillouin::.onu ofrtangular laltice. Make a plot oflhe flO't two Brillouin ZOIl(.'$

a primith'C ":uangular two-dimensional lattice "'ith a.,es o. b '" 30.

2. Brillouin WIW. rwanglltar lotlK:e. A two-dimcnsklrtal metal has one atom of va lcnt'Y one in a $imll]C n:clanglliar primitive cell a .. 2 A; b - 4 A. (a) Draw the first Brillouin ronc. Gtvc its dimensions. in em-I. (h) Calculate the mJius ofthc free electron Fermi sphere, in em-I. (e) O.-dW Ihis sphere to SC"cdc on a drawing oflhc first Brillooin ZOIlC. Make another .dretch to ~ the fir.>t [(.ow pcl'iolb of the free ck'dmrl band in the periodic ZIDflC scheme, hi' both the first and S(.'COnd t.'ncrgy )J8nos. Assume there is a small energy gap at the zone houndary.
3. IIczogoMI dOflS-ptlcktd rtnlC'lIre. Con~idcr the first Brillouin zone ora crystal with

a simple hcxagonallallice in three dimensions "1lh lattice constants a and c. Let Gc denote the .soorte.st n:eiprQUll lattice \1.'dor panJlci 10 the c a:ds of the cryMaI lattice. (a) Show that tor a hexagonal claoie-paclc:cd ~taJ itructure the Fouricf" component U(G e ) ol'thc crystal potential U(r) is Zt.'TO. (b) is U(2Ge ) also 7.crO? (cJWh)' is it poosible in principle 10 obtain an insulator made up of divoIlcnt atoms at the lattice points of a slmilic hcxagonal lattice? (d) WIlY is it not possible to obtain an insulator made up uf monovalent atoms in a hcxagOllal-elost. ...packL>d structure? 4. Brillovin:tOflft of two-JimerutomJ dit:almt metal. A twod imensklnal metal in the fOrm of a ~uare lattice has two conduction electrons per aloin. In the almost free electrorl lJI'proxilllation, Sketch carefully the c1cd:ron and hole ellCrgy surlact:s. For thc ek'Ctrons rn()()Sc II zone scheme such that the Fenni surface is shown as cl()Sl."(I.

5. Open orbit,. An OIlCn orbit in a monovalent letragon.J metal connL'(.1;s opposite faces c the boundary of a Brillouin 7.OIle. 1be faces are separnted b). G "" 2 X
lOS em-I. A rnagnd.ic field 8
:0

Hi' gau55

:0

10- 1 tesLa is normal to the plane orthe

open orbit. (a) What is the onb of magnitude of the IlCriod of the rootk>n in k space? Take v - lOS em/sec. (b) Descrihe in rcaI space the motion uf an electron on tllis orhit in the IJrescnce c the magnetic field.
6. CoIlni.:e mn-gyfor Cl Mjuare wtU potenmJ. (a) Find an cxpression for the binding t.'f"ICrgy c an electron in one dimension in a sillgle ~uarc well ufdepth Un alld ....;dth a. (This is the standard first problt:m in c1ernt:ntary quantum mcdlanics.) Assume that the solution is s)'mmetric about the midpoint of"the ....'CII. (b) Find a numerical result for tile bil1<ling energy in tenns of Un fur the spL'Cial <:asc luol ... 2f12/mat and compare with the approlJrillte limit of Fig. 20. In this limit of widely sepnrakd wt:lls the band width goes to zero, so the energy for k ... 0 is the same as the cnergy fOr any other k in the lo.....e st energy band. Other bands arc fOrmed from the excited states cJ the well, in this limit.
1. De lfCUlI-l"(/n Alphen pniod of polamum. (a) Calculate the period 1l{1IB) expected for potassium on the free eledron model. (b) What is tllC area in real space of the extremal orbit, for 8 ,., 10 kG - I T? TIlc same pcriod alljJ!ies to oscillations in the ck"Cl:ricaI rt:!iistivity. kllown as the Shubnikow-dc Haas ('f1"--ct.

F........ SurfOCetl and Mdl'll

267

.8. &md. W"ucture on k -" prrluroorion thtory. Consider a nondeg('ncnte orbital 4'"" at k ir the band n of a cubic:: a)'$lal. Use se<'Ol'ld--<lrdcr perturbation tM.'OI'Y to rind the result

"

(I..) ..

(0)

+ fl!P +!!.... L'

2m
'~J

l<nOIk p\jO}J!
f.{O) fAO) ,

(39)
mass at this

where the ~Ilm i~ OVCr all other orbitals point is

4'". lit k ... 0. The effective

(40)
The mass lit ti,e conduction band L'(]ge in a narrow gap st.'miconductor is oRell domina!(.'(] by the effect d'the vak'I'K.'C band L'(]ge, whcnee

(41)
where the Slim is ()\'tt the \oalence bands; Ec is tbe ('ncrgy gap. For gi\'L'J\ m..atrb: clements. small gaps Icad to small rnasst.'S. 9. Wtlnnier fUnd~. 11Je Wannk:T functions of a band arc defined in tenns of the Bloch functions of the ~e band by

u(r - r,,) = N-~

L cxp(-ik' rJ If.\(r)

(42)

wnere r. is II lattice point_ (a) Prove that Wannier functions about difc~nt lattice points ",1'11 are orthogonal n,o\m.

(43)

This orthogonality !'rollerty mllkL'S the (unctions often of greater usc thlln atomic orbitals centerL..J on different lattice sites, hccause the latter arc not generally (.X"thaguna!. (b) The \Vannier functions are peaked around the lallice sites. Show that for I/J. - N- 1f2 tJh uti.x) the \Vannier function is

for N atoms on a line of lattice conslant

tI.

10. Open orbit, tiM magnttorl5Utana. WC considcn.-d the tn.rtli\"Cr$C magnetorL'Sislance offrec clt.'droos in Problem 6.9 and of clce:trons aoo hole!' in Problem 8.$. In some crystals the magnctorcsistance saturatt..s except in special crystal orientatlons.. An open orbit carries current only in a single direction in the plane normal to the magnetk Held; such carriers are not dcfk'CIL't! by the rlCld. In the arrangement of Fig. 6.13, let the open orbits be parallel to k.; in real space these orbits carry

'"
this defines the constant
We",;:, 1 is
Ii.

(
l'urren! parallel to the !J R~b. Let Uno ,., 1Uo be the conductivity of the open orbits;

The IllRgnctut:ondut1ivlty tenSOr in the IliSh field limit

0-- -0-'
Uo

( o

Q~l

" 0

0)
I

O.

tive resbtivity in the r din.:d.ioo is p - (otluJ(~, not saturate, but il'lCl'l:ll5CS as 8 2.

with Q _ <oIc"'. (a) Show that the Hall flCld is E. - -E.JtQ. (1)) Show that the effec+ I). so that the resbtivlty does

-II. Londau Uveh. The V\."Clor potential ua unibm magm.>licfidd Bi is A the l.andau gauge. lbc hamiltonian of a free ck'Ctron \\ithout spin Is

-BlIi in

H = -(fj'12mXatla';

+ a'tv?) + (1I2m){-iliala:r -

cyB/c]' .

We will look lOr an eigenfunction uf the Wll\'C equation H - (41 in the fonn

(a) Show d'UI ,U1I) satisfies the cquatkm

(fj'l2Jll)t1l)(ld,J

+ (t: -

(Iitk~2m) -im~1I -

Jlx -

whe:rc we'" eBirru; and !hi - cAAJeB. (1)) Show that this is the wa,'c ....- quation of a harmonic mcillalOT with freq.....'OCy ~ ",'here

E~

(rl +i)~+ h2~

References
T. E. Fabn. 'nlroduaiQn to II.. theory of liquid Ao:sdemic Press, 1974. W. A. Harmon. Ell:lCtronic "ructu~ lind lhe properllu of IOIUb, Freeman, 1980.

mttm..

J.

M. Zlman, Principles of tlte I/lJry of solids, 2nd ed., Cambnd~, 1972. V. I. Moruni, J. F. JanaI<, I'M A. R. Williams, CDlcu'at~d ~leclronfc propft't~1 of mdllLs, Pergamon, 1978. Comllilatiuo. A. B. Pill~rd, MagnetorubfllllCf III metC/u, Cl\mbridgc, 1969. W. A. de Uccr. W. D. Knighl, M. Y. Chou, lind M. L. Cohen. MEleelronit' shell slrudure and metal clusters, Solid slale physics. 40, 93 (19S7).
M

FERMI SURFACES A. P. Cracknell and K. C. \\long. Fenni ... r[ace. 0xf0r1I. 1973. L. M. FaJicoy, MFcnni surface studies,~ E1mronr In erJISUl1I.iN lViids, IAE.... Vienna, 1973.

I. M. l...diJIil2" M. Ya. AWd. and M. l.

II..

B. Pippanl, ~ of roMudllln GI:lCtnmr, Cordoo and Breacn, 1965P. B. Vlpd,(or IlDd I~ M. Fahaw, ~Fcnni surfatt- properties ~ mctaIs, Pb)'s. Stalus Solidi B s.c, 9 (1972). Index to F~I Alnac.es and band struchllU. M. Springfonl, ed., Electt"OlU 41 the Fmni. I1Irfoct, Cambridge, 1980. D. ShllCn~ Mcgnd.ic NdllatioM in mnab, Cambndge, 19&1. C. R. SlL-waJ1, "llcavy krmion $)stcms,~ Re... s. Mod. l'hys. 56. 755 (19fW). A C. Hfwson, K.ondD problem 10 1_11IJ fennlllN. Cambridge, 1993.
M

11m

~,UclrorUcWoryofrn&Jls, Consultants

Bureau,

l'SElIDOPOTE~TIALS

Solid stnte physics 24 (I97Q) ronlains three articles nn Ilscudopnlclltil\h by V. Heine, M. I~ Cohcn, lind D. W.:sire.

10
P!asnwns, Polm"itons, and Polarons
DIELECTRIC FUNCTION OF THE ELECTRON GAS P1.35rn3 <Iptics Dispersion relation for electromagnetic "'~dl'eS Transverse optical modes in a plasma Transparency vf alkali metals in the ultr.wiold Longitudinal plasma oscillations 271
212

273 274 274 216 278 280 282 283 284 286
287 291

PLASMONS ELECTROSTATIC SCREEI\'ING Screened coulomb poh:ntial Pseudupotential component U(O) Moll metal-insulator transition Screening and phonons in metals
POLARITONS lST relation

ELECTRONELECTRON INTERAGnON
Fermi liquid

294
294

ELECTnONPUONON I"'TERAcnON: POLARONS

PEIERLS INSTABLLITY OF UNEAR METAlS SUMMARY
PR08LEMS

297
300 301
302

I. Surface plasmons 2.. Interface plasmons

3.
Alrl~n

302 302
302 303 303 303 303 303 303 304 304
~

4. 5. 6. 7. 8. 9. 10. 11.

wave; lIelM:on wav~

l'I.asmQll mode of a sphere Magnetoplasma frt:qucncy Photon brand, at low wavcvcetor Plasma freqUL'IlCY and ek'Ctric:al OOIlilucti\ity Uulk modulus of the Fenni gas Response of electron ~ Cap p1asmons and the van der Waals interaction

REFERENCES

NOTE: TIle tcxt and problems of this chapt....... auumc Etcilily in the use of elt."Ctromagndic thl.'Ory al the k,,'c! 8 gO()(I senior COur.iC.

'f-----,-,7;---:;iL----;<;-------:-, ,., , " "

-,

I ,
I I

I , ,
!
drdron Pi' venus frequnw:y in units of the pbsma ~I~ without damping ... .. ben c is pos,ti'".., and..,.J is ncptr."c.

-,
r.equency ...... EkdJ"(ll"Ollgndic
. _'1.71' _

Figure 1 Diekdric function

-<...) of. (tft

"""'0:$

Electl"llrJlagnC.'tie ".....,,, are tot.dly reIko:ic:d (rool the medium wf>

CHAPT.ER 10: PLASMONS. POLARITONS. AND POLARONS

DlELECI1UC F1JNcnON OF THE ELECfRON GAS

The dielectric function ~w,K) of the electron gas, wilh ils strong dependence on frequenc)' and wavtM':doT, has significant consequcnces G>r thc physi4 cal properties of solids. In one limit, ~w,O) describes the collective excitations of the Fenni sca-lhe volume and surmce plasmons. III another limit, If(O,lq describes the electrostatic screening of the electron-electron, clectron-Iattice, and electron-impurity interactions in crystals. Wc usc the dielectric function of an ionic crystal to derive the polariton spectrum. Later we discuss the properties of polarons. nut rlTSt wc are concemt.'C1 with U1C electron gas in metals. Definitioru of the Dielectric Function. nle dielectric constant If of electrostatics is defincd in terms of the electric field E and the polari7..:1tioll P, the dipole moment density:

(eGS)

0 = E

+ 41l'P =

IfE ;

(SI)

=If'oE + I' = E~E

(l)

TIlUS

defillt.'C1, E is also known as the relative permittivity. The introduction of thc disp\a(,'Cfficnt 0 is motivated by the usefulness of this vector related to the external applied charge density P~", in the same way as E is related to the total charge density P = Peot + Pind, where PI,,<.t is the charge dcnsit)' induCt.'C1 in the s)'stem by Put. Thus the divergence relation of lhe electric field is

(eGS)

div 0 = div EE = 41l'p....t div E = 41Tp = 47l'(Pel<! + p,"&

div D = div EEOE = Pc,t

(2) (3)

wv E

::::>

p1Eo ;: (Put

+ PwwJ.Yo

'w.

Parts ofthis chapter will be written in CGS; to obtain results in SI, write 1Ilfo for

,..,

(
We shall need relations between the Foorit.... components u D, E, p, and the clt.'Ctrostatic potential '1'. For hn.'vity we do J1Qt CJlhibit here the frequenc)' dependence. Ocfine E(K) such that
(Jo)

tht.'f1 (3) 1:H.."COmt.'S

div E co div I E(J..1 elK... _ 471 I p(1<) rfK'r and (2) beeomc..-s (3b)

(30)
Each of the (:quations must be satisflt..d term by term; \\'C divide one by the other to

00bin

~K) ... Ao.,(K) '" I _ An.iK)

p(K)

P(K)

(3d)

11lc electrostatic:: potential '1'.... dcfint.'d by -Vip"" ,. 0 satisfies the Poisson l'qualion V2q>.", = -4:ITp-; and the c1cctru>tatic potcntialtp dcGoc-d by - '\Ill' '" E satisr.es vilp ,. -4r.p. The Fourier compont:n1s c:i the potentials musl thcn..forc satisfy (30)
by (3<1). We usc Ihls relation in the treatment utile scn:ent.-d coulomb potential.

Plasma Ollties
The long wavelength dielectric response (w,O) or (w) of an electron gas is obtained from the equation of motion of a free electron in an electric Held:

d'r
m

dt2.

= -el~ .

(4)

If x and E have the time dependence e-l "". then

-cfmx = -eE;

x == eEIrm./- .

(5)

The dipole moment of one electron is -ex '" -,fElflUJf. and the polari7..ation, defined as the dipole moment per unit volume, is

"e' P = -rlex = - -E
fIlW'l

(6)

where

II

is the electron conccntration.

"nle dielectric function at frequency

(CCS)

is
P(w) 411'--' E(w)

f(w)---"'l E(w)
E{W) ==

D(w)

(7)

(51)

D(ti
.oE(w)

== 1 +

pew)
<oE(w)

"nlc dielectric function of the free electron gas follows from (6) and (7): lSI)
E{w) = 1

The plasma frequency wp is defined by the relation (eeS)

ne'
-~

(8)

w: - 41Jfltflm ;

(SI)

....,...~~. n('t/~1I

(9)

A plasma is a medium ....ith equal concentration of positive and negative dlarges. d which at least one charge type is mobile. In a solid the ncgati,'e dlllrgCS cl the conduction electrons are balanced by an equal concentration of

positive charge of the ion cores. We write the dielectric function (8) as
(10)

plotted in Fig. 1. If the positive ion core background has a diel~tric constant labeled E(OO) t.'SSCOtially constant up to frequencies well above Wpo then (8) becomes
(11)

w1u..r e

wp

is defll1(..d as

(l2)

Ditpersiorl Relation for Electromagnetic Waves In a nonmagnetic isotropic medium the electromagnetic wave equation is

(ces) a'DIat'
....'C

c'V'E ,

(13)

We look for a solution with E 0( cxp(-iwt) exp(iK' r) and D = E(w,K)E; then have the dispersion relation for ek"Ctromagnetic waves: (14)

(ces) w,W: C'K' ;

".
This relation tells us a great deal. Consider
E Teal and

> O. For w real. K is real and a transverse eleecromagnctic wave propagates with the phase velocity d;l2. E Teal and < O. FIX w real, K is imaginary and the wave is damped with a
E

characteristic length 1I\KI. complex. For w real. K is complex and the waves are ~ampcd in space. E = co. This means the system has a fmite response in the absence of an applied force; thus the poles of e(w,K) define the frequencies of the free oscillations of the medium. E = O. We shall see that longitudinally polarized waves are possible only at the l'.eros of E.
IJ

Tron.weru Optiazl Modes in

plasnuJ
(15)

The dispersion relation (14) becomes, with (II) lOr e(w),

(CCS)

2 For w < w p we have K < 0, so that K is imaginary. TIle solutions of the wave equation arc of the form exp( -I"lx) in the frequency region 0 < wSw". Walles incident on the medium in this frequency region do not propagate, but will be totally reflected. An electron gas is transparent when w > Wp. for here the dielectric f\lI'lCtion is positive real. The dispersion relation in this region may be written as

(CCS)

w2 =

w: + tflK /E(g;;) ;
2

(16)

this describes transverse electromagnetic waves in a plasma (Fig. 2). Values of the plasma frequency <Up and of the free space wavelength Ap 2mlwp for electron concentrations of interest are given below. A wave will propagate if its free space wavelength is less than lip; otherwise the wave is reflected. n, electrolls/cm3 1010 HI" 1018 1014 13 5.7 x 101.5 5.7 X 1011 5.7 X 10 5.7Xl<f w,,, s 3.3 X 10-.5 3.3 X 10-3 0.33 33 lip, cm

-,

Transparency of Alkali Metau in tm Ultraviolet. From the preceding discussion of the dielectric function we conclude that simple metals should reflect light in the visible region and be transparent to ultraviolet light. TIle effect was discovered by Wood and explained by Zener. A comparison of calculated and observed cutoff wavelengths is given in Table 1. The reflection of light from a metal is entirely similar to the reflection of radio waves from the ionosphere, for the free electrons in the ionosphere make the dielectric constant negative at low frequencies. Experimental results for InSb willI '= 4 x 10"1 cm- 3 are shown in Fig. 3, where the plasma frequency is near . '-'9 cV.

10 I'lasmom. I'olaritoru. and Polaronl

273

-.

-.

Figur<> 2 Dispersion relation ror transverse c1ooromagnetic "'lin's in a plasma. lOco group \"elocity t> ., drJdK is the s\ope t>f the dispersion curw. Although the dickctric (unction is betwi:en zero and ooe, the group .'docily is less than the ,rlO<.'Ily of light in ''a<:uUJn.

F'>gun~

3 Rdlcdance of indium antimonide with

n" 4 )( WI. em-a, (Aner J. N. Hodgson.)

Table I

Ultraviolet transmission limits of alkali metals. in "

u
Ap , calculated, mass .J,. obu:rvoo
Tn

N.

Rb
3220 3<00

c,
3620

1550 1550

2090
2100

2870
3150

'76

Longitudinal Planna OscillaUo,.,

The 7.croS of the dielectric function determine the frequencies of the longitudinal modes of oscillation. That is, the condition

E(w,J = 0

(17)

determines the longitudinal frequency WL ncar K "" O. By the geometry of a longitudinal polarization "..ave there is a depolarization field E CIt -41TP. discussed below. Thus D ... E + 41TP = afar a longitudi_ nal wave in a plasma or more generally in a crystaL In SI units, 0 = EoE +

p=o.
For an electron gas, at the zero of the dielectric function
E(WJ=

I-J,Iwl=o.

(18)

whence filL = wp Thus there is a free longitudinal oscillation mooc (Fig. 4) of an electron gas at the plasma frequency described by (15) as the low-frequency culoff of transverse electromagnetic wa\'cs. A longitudinal plasma oscillation with K = 0 is stxn\lIl in Fig. Sas a unifonn displacement of an electron gas in a thin metallic slab. The electron gas is movcd as a wholc with respect to thc positive ion background. The displacemcnt II of thc clcctron gas creates an electric field E = 4mwu that act.. as a restoring force on the gas. lbe equatiOll of motion ~ a unit volumc of thc electron gas of concentration n is (CGS)
0'

&'.. mn-- = - neE = - 4mJ t e2u

dt'

(19)

(CGS)

d~,

df

W;II=O;

.
:

'''"'') - - on (m

(2lJ)

This is the equation of motion of a simplc harmonic oscillator of frequency w". the plasma frequcney. nle expression for w,> is idcntit'al with (9). which arose in a diffcrent connection. In 51. the displacement II creates the electric field E = netJ4J, whence W p = (nrIEom)IJi. A plasllla oscillation of small \o\'3vcvcctot..has approximately the frequency wi>" Thc \o\'3vevector dcpendence of the dispersion relation for longitudinal oscillations in a Fcrmi gas is givcn by
(21)

whcre

l,;t"

is thc electron velocity at thc Fermi cncrgy.

-- ---

-~-

-~--

---

..

.1
"

SUi.......

~bl;'"

.lo.".lty

a- -"""
[~

f"1gU", 5 In fa) ili ~hcr4.... a thin slab or film of a metal. A. cnm ~>on is shQo..n in (h), with the posilr.c jon <.'(>res indicated b)' + signs and the electron sea indkat.. '<1 by the gray blIckgmund. 11>e slab ;s ek'Ctrically m,utral. In fe) the ncgalin' charge has been displac<..,j ul'",-ard uniformly by a small distance ..., shown exaggerated in the figure. Ju;n (d), this diSI'Ia.m<.:nl establishes a su.uee ~ density -l'Iftl 00 the upper surlace ci the slab and +UU 00 the boo... , . su~. "herr n is lhedectron concentration. Andectril;: ficld .. 4,""", is producrd irWdc the dab Thil6eJd ~5 to res.lore the ekaron sea 10 its equilibrium positJoon (bl. In 51 unib, E - nm/4:;

...

27'
PLASMONS

A plasma oscillation in a metal is a collective longitudinal excitation of the conduction electron gas. A plasmon is a quantum of a plasma oscillation; we

may excite a plasmon by passing an electron through a thin metallic film (Figs.
ti and 1) or by rcnccting an electron or a photon from a film. The charge of the electron couples with the electrostatic field nucluations of the plasma oscillations. 111c renected or transmitted electron will show an energy loss equal 10

integral multiples of the plasmon energy. Experimental excitation spectra for Al and Mg are shown in Fig. 8. A
comparison of observed and calculated values of plasmon energies is given in

Table 2; further data are given in the reviews by Raether and by Daniels. Recall that wp as defined by (12) includes the ion corc effects by lise of (lXl). It is equally possible to excite collective plasma oscillations in dielectric l'iIms; results for several dielectrit.:s are included. The calculated plasma energies of Si, Ce, and InSh are based on four valenl.'C electrons per atom. In a dielectric the plasma oscillation is physically the same as in a metal: the entire valent..-e electron sea oscillates back and forlh with respect to the iou t..vrcs.
Table 2 Volume plasmon energies, in eV
Calculated
Material

O1>5Cl'"ed

'".
8.02 5.95 4.29 10.9 15.8 16.0 16.0 12.0

"w,.
7.96 5.58 3.56

Aletcl/s
Ii

N.
K

Mg
AI Dielectrics

7.12 5.71 3.72 10.6 15.3

16.4-16.9 16.0-16.4 12.0-13.0

51 G.
InSb

I . illo..... "Itctron

'?-",-

J'

........

"'"~

t1plre 6 Creahon of a plasmon in a mdaJ film b)' meb.~ SC'llt Icring of an eJr..ctrnn. 1be incMlcnt ekd rnn t~'J'ic:d'r has ., enngy 1 to 10 ke,v, the pbsmon erterg)' ITlll)' be oflhe- order of 10 eV...... e\U'Ot is abo shuwn;n which 1"-0 pIasmons are c:reated.

-...
6dd

tigt.l~ 7

A spc:<'lrtlmcler wilh ek,<:Iroslalic anal)'Ur

DIlniels et at)

lOr the stud)' of plllSmon udblion II)' ckdrons.

(After

J.

" " , f ., , i ,
'I>

,
j
.~

" " ,
6

'"

GO '00 Elc<:Iron me<X!' I.... le\')

.. .
'"

,
''''
I02030.a~60;o

Eltron nll'lJ)' b. (e\?

11

r'f:"~ 8 Energy Ios:o spcdnl forclcctrons ~e:d &om filmsof(a) a1unlinum and(b) magnesium, n.pn"wy d.,dron cncrg;es of2020 eV. n.e 12 loss pcaksobserved in AI are made upofrombm. tioou of 10.3 and 153.,v Iosoes....JK:n, the 10.3 cV lou is due to ",rbce p1umons and Ihe 15.3 cV "- is due 10 ,'OIurne pIasmons. ",., ten loss peaks observed in ),1& are made up of c:ombl'nallOflSd 1.1 eV sunxe plasmons and 10.6 cV ,oIume pml1lOn$. Surfloe plasmons an: the sub;r:c1 of frob"'m 1 (MILT C. J. foIo,....u and J. B. Swan.)

280

I
ELEcrROSTAllC SCREENINC

The electric field of a positive charge embedded in an electron gas falls off with increasing r faster than lIr, because the electron gas tends to gather around and thus to screen the positive charge. 111c static screening can be
described by the waveveclor dependence of the static dielectric function E(O,K). We consider the response of the electrons 10 an applied external electro-

static field. We start with a uniform g."1S of electrons of charge concentration -noe superimposed on a background of positive charge of concentration nac. Lei the positive charge background be deformed mechanically to produce a
sinusoidal variation of positive charge density in the x direction:

(22)
The term p".t(K) sin Kx gives rise to an electrostatic field Ihat we call the external field applied 10 the electron gas. The electrostatic potential q> or a charge distribution is round rrom the Poisson equation Vlrp == -47fp, by (3) with E == -Vrp. For the positive charge we have
P == p"d(K) sin Kx
(23)

The Poisson equation gives the relation K 2rp.,.t(K) == 41TPr.t(K)

(24)

The electron gas will be derormed by the combined influences or the electrostatic potential rp".t(K) or the positive charge distrihution and or the as yet unknown induced electrostatic potential fJ'ond(K) sin Kx the derormation of the electron gas itselr. The electron charge density is

or

(25)

where Pind(K) is the amplitude or the charge density variation induced in the electron gas. We want to find p"Kl(K) in terms p".t(K). The amplitude or the total electrostatic potential q;{K) == fJ'u,(K) + rp...d(K) or the positive and negative charge distributions is related to the total charge density variation P(K) = Put(K) + Pond(K) by the Poisson equation. Then, as in Eq. (24),

or

(26)
To go rurther we need another equation that relates the electron con(''entration to the electrostatic potential. We develop this connection in what is rolled the Thomas-Fermi approximation. The approximation consists in assuming that a local internal chemiCll potential can be defined as a runction or the electron con(''entration at that point. Now the total chemical potential or the electron gas is constant in equilibrium, independent or position. In a region where there is no electrostatic contribution to the chemiC' .! potential we have

10 """-'. r ~ and l'oWrmu

251

=0;;;=-

'-Jlr,;i<m Ii ~ ..1o<'trm

=-=

C'QnllOOltnlion

Fisu'" 9 In thennal and diffusive cquillllri"m the dlef1\iC'lll poIential is comunl. to ITlainUin il constant we increa5e the electmn ~tnlion in regions 0(.J*'f' "'~ thr poIenliai en~ is low. and we d ~ the concenlration ",-here the potential is high.

(27)
at absolute zero. according to (6.17). In a rcgiOll where the electrostatic poten tiat is q:(x), the total chemical potential (Fig. 9) is constant and equal to
Po = E,.{r) - e(tl(r) a -[3~n(x)J2!3 - eq;(x) a -(3~1l0J2!3 ,
~

Ii' 2m

Ii'

(28)

"'....ere E,.{X) is the 10cal value of the Fermi energy. ne expression (28) is valid for static electrostatic potetltials that vary slowly compared with the wavelength ofan electron at the Fenni level; specifically, the approximation is q ~ kF By a Taylor series expansion of EF. E'l_ (28) may be written as
-/-(n(x) -

d.,

<no

noJ a

eq;(x) .

(29)

From (27) we have dEFldno = 2FI3/1o> whence

3 e(/l(x) n(x) - no iil -110-- .

2 "

(30)

l1le leftIJand side is the induced part of the electron concentration; thus the Fourier components of this equation are

(31)
By (26) this beromes
(32)

By (3d) we have
AO K) ."

~ 1_

""",(K) = 1 + Il'IK' . ,;,K) .'

(33)

here. after some rearnmgement.

(34)

where cJo is the Bohr radius and D(fF) is the density of states for a free electron gas. The approximation (33) lOr ~O.K) is cilled the Thomas-Fenni dielectric function. and Uk~ is the lnomas-Fcrmi screening length. as in (40) below. For copper with no = 8.5 X uP cm- 3 the screening length is 0.55 A. We have derived t".o limiting expressions for the dielectric function of an electron gas:
~O,K)

= 1

+ -' .
K' '

l'

~w,O) = 1 - ~

.I.!

(35)

We nolke that f(O,K) as K_ 0 does not approach the same limit as ~w,O) as w_ O. 111is means that great care must be taken with the dielectrk function near the origin of the w-K plane. The full theory for the general function ~w,K) is due to Lindhard. I Scruned Coulomb Potential. We consider a point charge q placed in a sea of conduction electrons. 10e Poisson C<luation for the unSCf"eened coulomb potential is

V'.. =

-4~,),

(36)

and we know that 'Po = ql,.. Let us write

Nr) = (2'7J)-3 f dK tpo(K) exp(iK r)
(37)

We use in (36) the Fourier representation of the delta function: 6(r) = (2'7Jr 3 f dK exp(iK' r) whence ~tpo(K) :: 4'7Jq. By (30),

(38)

oJ.Ky",K) =

~K)

where rp(K) is the total or screened potential. We use E(K) in the Thomas-Fermi form (33) to find
<p(K)=
41TCf

K2

+ k~

(39)

The screened coulomb potential is the transform of 1f'{K):

4wq ~,.) = ( \3 2'7J1
= '7J,.
2q

L-

dK 2wK' 2 1..2 K + ~

I'

-1

d(cos l1) exP(lKr COS

f1l
(40)

Lao dK Krin Kr

= - exp(-k~,.) ~+J.:,.

'A ~ discusNon of the Lindhard dielectric function is given by J. Ziman. Prindp/es 0/ fM of.oodr. 2nd cd., CamI:Jridg<:. 1972, Chapter 5. The algc:braie steps in the e-),13lion cJ Ziman', equation (5-16) are gi\-en in dtuJ.l by C. Kiud, Solid I~te ph~"S n, 1 (1968), Sed.ion (j.
I~

, ,

,'"r-~--:o=_J'f--------

.,,
i

l:< -0 r

-02

F4,'U,"" lOa Comparison of screenl,.-d and un$<.Tet:,..,'d "",-,lomb potcntials of a Sialic un;t positive charge. The screening Icngth Uk. is set ('flual to unity. TIle static scn'Cnl,.-d into..'l"aCtion h taken in the 1llomas.Ferml approximalion, which holds for low wa"cvt,'Clors q k..; IllOrc complete calculations with aI' WlWf,.VcctOrs indudt.-d f,.'Xhibit spalial oscillations, callcd Frif,.-dd oscillations, in 2J<.. r and are ploued in QTS, p. 114.

-:

as in Fig. lOa. The screening parameter k~ is del'ined by (34). The exponential factor reduces the range of the coulomb potential. nle bare potential lflr is obtained on letting the charge concentration tJo _ 0, for then k. _ O. In the vacuum limit If(K) "" 41TqlK?-. One application of the screened interaction is to the resistivity of certain alloys. 'nle atoms of the series Cu, Zn, Ca, Ce, As have valences 1, 2, 3, 4, 5. An atom of Zn, Ca, Ce, or As added substitutionally to metallic Cu has an excess charge, referred to Cu, of 1,2,3, or4 ifall the valence electrons join the conduction band of the host metal. The foreign atom scatters the conduction electrons, with an interaction given hy the screened coulomb potential. This scattering contributes to the residual electrical resistivity. and calculations by Mott of the resistivity increase are in fair agreement with experiment. Pseudopotential Component U(O). In the legend to Fig. 9.22b we stated a result that is important in pseudopotential theory: "For very small k the potential approaches -j times the Fermi energy." The result, which is known as the sueened ion limit for metals, can be derived from Eq. (39). When converted to the potential energy of an electron of charge e in a metal of valency z with '10 iOns per unit volume, the potential energy component at k "" 0 becomes
U(O) "" -ezf1(#'(O) = -47TZutr/k~
(41)

The result (34) for

k; in this situation reads

k;
=

67TZnrr/EF

(42)

whence
U(O) = -JEF .
(43)

...
Mott Mdlll-lnsulatoT Transition
A crystal composed of ont: hydrogen atom per primitive cell should always

be a metal, according to the independent-electron model, because there will always be a half-filled energy band within which chaTSe transport can lake
place. A crystal 't\1th one hydrogen molecule per primitive cell is a different matter, because the two electrons can 611 a band. Under extreme high pressure. as in the planet Jupiter, it is possible that hydrogen ocrurs in a melaUic Conn. But let us imagine a lattice ofhydrogen atoms al absolute zero: will this be a metal or an insulatOl"? The answer depends on the lattice constant, with small values of a ghing a metal and large val.ues giving an insulator. Motti made an early estimate of the critical value Dc r:1 the lattice constant that separates the metallic state from the insulating state: Oe: = 4.5ao. where Go "" fl2/ is the radius of the first Bohr orbit of a hydrogen atom. On one approach to the problem, we start in the metallic state where a conduction electron sees a scn.>enoo coulomb interaction from each proton:

tnr

U(r) = -(r/r) exp(-k.r) ,

1l3

(44)

where ~ = 3.939no /00. as in (34), where "0 is the electron concentration. At high concentrations k.. is large and the potential has no bound state, so that we must ha""e a metal. 1be potential is knO\\ftl 3 to have a bound state when k. is smaller than 1.19r'ao- With a bound state posSible the electrons may condense about the protons to form an insulator. lhe inequality may be written in tenns of no as

(45)
tic>

With no = 1103 for a simple cubic lattice, we may have an insulator when 2.7&0. which is not far from the Mott result 4.56 0 found in a different _yo The term metal-insulator transition has come to denote situations where tbe electrical conductivity of a material changes from metal to insulator as a function of some external panuneter, which may be composition, pressure, strain, or magnetic field. 1be metallic phase may usually be pictured in terms of an independent-electron model; the insulator phase may suggest important electron-ekctron interactions. Sites randomly occupied introduce new and interesting aspects to the problem, aspects that lie within percolation theory. lbe percolation transition is beyond the scope of our book.

'For an C1l"!'11en11l:\'iew d"the Geld oCmdaJ..in,,,blor trmsitio.... see N. F. Motl. Pro<: R Soc. London It 382, I (1982). 'F. J. &gen. H. C. Cra\.Io$k~. Jr., and O. J. lI'II'Wood. Pby'. Rev. AI. 1sn..(l970).

'.',--,,--,,---,-----,

,
'(
s ,

,.

.. "
In."l.ator

MeW

,,~!-----!---~:----!---~ o 2 .. 6 8
n. in
IO'"cm-~

Figure lob Scmilog plot of ob~ "zcro temlle"'lure" coodudi"ily 0(0) ,"enus donor roncenlfallon n for pnmphoroos doflon; in silkon. (After T. P. Roso:nbaum, "I at)

When a semironduetor is doped with increasing concentrations of donor (or acceptor) atoms, a transition will occur to a conducting metallic phase. Ex ~rimental results for P atoms in silicon are shown in Figure IOh. Here the insulator-metal transition takes place when the concentration is so high that the ground state wa"efunctions of electrons on neighboring impurity atoms overlap signifiC3lltly. 1be observed "alue of the critical concentration in the 5i: P alloy system is 18 em-a, as in the figure. If we take 32 X 10- 8 em as the radius Tic:: 3.74 X 10 Oflhe ground state of a donor in 5i in the spherical approximation, then by the Mott criterion ac "" 1.44 x 10- 6 em. The P atoms are belicved to occupy lattice sites at random, but ifinstcad thcir lattice were simple cubic, the critical ~Iott concentratjon would be
n~ = lJ~ = 0.33
X

10 16 cm- 3

(46)

appreciably less than the obsen'Cd value. It is usual in the semiconductor liter-

...
ature to refer to a hea\"uy-doped semiconductor in the metallic range as a degenerate semiconductor.
Screening and PhOJWJU in Metab An interesting application of our two limiting fonos of the dielectric function is to longitudinal acoustic phonons in metals. For longitudinal modes the total dielectric function, ions plus electrons. must be 7.ero, by (17). p~'ided the sound velocity is less than the Fermi \'ckrity of the electrons. we may use for the electrons the 1Oomas-Fermi dielectric function
EJ(w.K) = 1 + '/~/k"l . (47)

Pro\ided also that the ions are weU-spaced and mO\"e independently, we may use for them the plasmon E(w.O) limit \\ith the approldmate mass M. The total dielectric function, lattice plus electrons. but without the eh.-etronic polariz.ability of the ion cores, is
E{w,K) = I 4'1T11e2 Mw'

I.';
i.,-2

(48)

At low K and w we neglect the term I. At a zero of E(w,K) we h;n."e. with

EF -

lmt.1~.
W

4mle2 4~ EF - - K2 = - - . - - K ' :
Mk; 111
6'l'T'~

In 2 -vJ.K

3M

(49)

0'

w = vK ;

= (m/3M)112 OF .

(SO)

This describes long wavelength acoustic phonons. In the alkali metals the result is in quite good agreement with the obsen'ct1 longitudinal wave velocity. For potassium we calculate t>:::: 1.8 x lOS cm s-\ the ohsen'ed longitudinal sound velocity at4 K in the [100] diredion is 2.2 x lOS cm S-I. There is another zero of E(w,K) for positive ions imbedded in all electron sea. For high Frequencies we use the dielectric contribution -JpIuf of the electron gas:
E(W,O) = I

(51)

and this function has a zero when

-:-+p..

At

(52)

This is the electron plasma frequency, but \~;th the reduced mass correction lOr the motion of the positive ions.

0."",,-.-----';/
0_050
0,,"

--,

~"'W

0<,,,

;.;.;;-i------~=-~-;;-~-~-:-:-=---'i'u"""-"-"-=--=<i-~-~-1 , ,

: ,

, , ,

."'"
0",

., .''

.
~:

i:
01

o ,

: I
0.2 0-3 0.4 O~ AcK, In "V 0.6 0.7 0.8

""igllrc II A plol of the observed energies Ind "'D\'''\'CC!ors of the polarilolls and of the 1.0 phonons in CaP. The t11t:()rellcllf dispersion cu...l:S ~n, show" by th" solid lines. The di~perslon CUIVt:S for the uncuupl~'<I. IlhonolU and photons Ire sho-;"T) by the short, <lashed lines. (Arter C. II. Ilcory and J. J. Iloprtdd.)

POLARITONS

longitudinal optical phonons and transverse optical phonons were dis-

cussed in Chapter 4, but we deferred treatment of the interaction of transverse optical phonons \\;th transverse electromagnetic waves. At resonance the phonoll-photon coopling entirely changes the character of the propagation, and a forbidden band is established for reasons that have nothing to do with the
periodicity of the lattice. By resonance we mean a condition ill which the frequencies and w3,veveetors of both waves are approximately t:<jual, The region of (he crossover of the two dashed curves in Fig, 11 is the resonance region; the two dashed curves are the dispersion relations for photons and transverse optical phonons in the abo sence of any coupling between them. In reality, howe\'er, there always is cou pIing implicit in Maxwell's equations and expressed by the dielectric function, TIle quantum of the coupled phonon.photon transverse wa\'e field is caned a polariton. In this seetion we see how the coupling is responsible for the dispersion relations shown as solid curves in the figure. All takes place at \'ery iow values of the wa\'e\'ector in I..'omparison \\;th a zone boundary, because at crossover C<l(photon) = ck(photon) "" w(phonon)" 10 13 S-I; thus k .. 300 em-I. An early warning: although the symbol WI... will necessarily arise in the theory, the effects do not concern longitudinal optical phonons. Longitudinal phonons do not cou"le to transverse photons in the bulk of a crystal. . (

111e coupling of the electric field E of the photon with the dielectric polarization P of the TO phonon is described b}' the electromagnetic wa,'e equation:
(CGS)

(53)

At low wavevedors the TO phonon frequency CUr is independent of K. 1be polarization is proportional to the displacement of the positive ions relative to the negative ions, so that the equation of motion of the polarization is like that of an oscillator and may he written as, with p"", Nqu,
_wl!p

+ f}P::

(Nt/IM)E ,

(54)

where there are N ion pairs of effective charge q and reduced mass M, per unit volume. For simplicity \\.'e neglect the electronic contribution to the polarization. TIle equations (53) and (54) have a solution when

This gives the polariton dispersion relation, similar to that plotted in Figs. II and 12. At K = 0 there are t"..o roots, Cd :: 0 for the photon and

oJ

04 + ."'-'q'fM

(56)

for the polariton. ] lere Wr is the TO phonon frequency in the absence of coupling with photons. 11le dielectric function obtained from (54) is:
E(Cd) ::

1 + 47TPIE = 1 +

4-rrNq 2/M
2
WI fA)

(57)

If there is bn optical electronic oontribution to the polari7.Btion from the ion cores, this should be included. In the frequency range from zero up through the infrared, \\.'e write
() () ECd=EOO+ 4-rrNql!/M
l!

WI -

Cd

(58)

in acrord with the definition of E(oo) as the optical dielectric constant, obtained as the square of the optical refracth'e index. We set Cd = 0 to obtain the static dielectric function: (59) whid, is combined with (58) to obtain
E(Cd)

in terms of accessible pantmeters:

,
-,
""'",o"lik

-~--"bpilu&1(

.....

....... "a .,...,

....,.,..,

--. -'.

-'. .
'

"-\7.M

'"

",.-1" 10" lid 1*' "~" I " 10" rad .-'

~ r---:;;7L----::::::::::::::::=::======-=-=-~~~o;;;:~.~~;-=-=-

.. _-----------

.,

'.0

"
n.e

Fipare 12 Coupkd. modes of photons.nc!. 'rtnn..,rw: opIiI:aI p/'M.loom in I.n 100"" crysI.t fme hvriz:ontalline represcnb oscillators of &eque~in the amenoe of coupling to the electl'Ol1J:lg" netic 6eld, and the fine line labeled ... - dCl~ connpoocb 10 ~1K' _\'U in tho: Cf)'JlaI, bul uncoupled 10 the lattice oscillators ...... 11w: '-Y)' lines are the clispeuioo. n:lalion$ in the presence of coupling bdween the lattice OKillaton and the dectromagndic wa\'e. One effect ci the CQUpling 1$ to create the rrequell<'y gap b<':t"'<:nI "'r. and "'T' within thi, gaJl the Wll,"e,"ector i.l pure imal':inary of magnitude gh"en by the brokt.-n line in the figure. In the gap the wan,"tlenuales as cKP(-IKp;). and we see from the pl.ot tlw.r; the attenuation is much .<lronger near ..... than ne-r <tOr.chancl:er- of the brano:hes \/aries with K, there i!; I. ro:gion miJed electric-medwUaLI aspects near the nominal CI"OSSO\"fl'. Note, finally. it is Inluiti\..... yolnio..s thai: tho: group ,riocityol'lighl in dlle medium is always <c. bcGusc the slope &lI~K lOr the~"'" dis~ion n:latioos (heavy lines) is C\~ Io$s than the slope c b the uncoupled photon in free s~

n.e

or

E(W) == E(CO)

+ [E(O)

- E(co)l

oJ}
11;

W'r

E(CO)

-W

2 ( w'~ - w')
OFt W

E(w) ::::

w'}O) - oJ.(w) ( . .2 2 == E(cD)

wr-W
WL,

wi 2

oJ)

WT-W-

2'

(00)

1be zero of E(w) defines the frequency

zero gives

as the pole of E(W) defines "'T. 1be

(61)

Waves do not propagate in the frequency region for which If(W) is negative, between its pole at W ., WT and its zero at W :. Wt,., as in Fig. 13. For negative If, W3\-es do not propagate because then K is imaginary for real w, and

...
,
,
3

)
"""C:"; OJ
_It"')

"

! ,

,
,
2

/-,

,
betWeal W -

.
I
, ,
00

r'U", 13& P\ot of ~~ rom (60) lOr e("') ,. 2 and ~ - 3. n., dielectric COIISIIIn! ~ nrpm-e "'T and wt.- (3/2) ....""". that is. ~n the pol.:{in.6nity) of e(w) and the uoo olll(lIt). Incident t~ic.....es of(reqllencles...,. < "" < ..... will not propagate if! the medium. bur ...ill be rdlerted al the boundary.

", ,

-,
-00

, , -,
-

l-

lII-I-

~W~~~~I~lr~~~W~~~~ FI'f'lI......." in Hz

Figure 13b Di.,I~lric (undio" (real put) ofSrF z n'ellSured o"e. a wid., frequc:ncy range, ed>ibiling the dt:~1lSl: of the ionic polariubility al high frequencies. (A. YOU IIiPllCl.)

""

+ + + +

I'
.-,

El

L.
fOph....

Ej
+ + + + + + + + + +

L,
Wpk.a

El

rop.tre" Rebli\~ d;,-pllliCCmcnh cl the IlO"itive and negati.., ions at one: tmtant d" 1m"" lOr. "1l''e In an optical....ooe lran:lillgalong the Luis. TIM: p/anft or nOdes (zcro d.isplacemenl)~ sI1O"".1'or long ",.,.."length phonons the rooo.bl planes are 1Io:par1llro b)" man)'pwK'S of Morns In the lr.lm\'C:ne opticallJOooon mode the l)Olrtick di.pIaco:l~nl is p'rpl"ndioou!al- to the Wll\'e'>'eClor K, the.- macro-

IIlCOllicek'dric field in an inunite rm:diUln will lie only in the :!:lC dire<.1lon for the mQde .hawn. and by the srmmdry of the problem iJE,JiJx. o. It rollo"", that di.' E - 0 ror a TO phonon. In the kmgitudmal "plic~lllhonon mude the IJarlklc diSlllarnenl. and h,,"ce the diclcctrlcpolarizalion P
are IJ8nlIIeIIo the W!l>'evector. The macroscopic elcctr,e field E satisfies l) - E + 4".P co 0 in CCS Of' foE + P '" 0 in 51; by symmetry E and Pare par'l1c1 10 the lC a.,i and a/;r- ~ O. Thu. dh- E JOoIi 0 WI, an LO phOl"on, and ~(...) di ... E I. uoro only if E(...) - 0.

exp(iKx) _ exp(-IKlx), dam~ in space. The 7ro of E(W), by our earlier argument, is the LO frequency at low K, Fig. 14. Just as with the plasma frequellC)' Wp. the frequency WL. has two meanings, one as the l.O frequency at low K and the other as the upper cutoff frequency of the forbidden band for propagation of an electromagnetic ""lwe. The value d WL. is identical at both frequencies.
LST Relation
We write (61) as

(62)
where t:(O) is the static dielectric constant and E(:< is the high-frequency limit of the dielectric function. defined to include the core electron contribution. This result is the Lyddane-Sachs-Teller relation. The derivation assumed a cubic crystal "'ith two atoms per primitive cell. 4 For soft modes with Wr- 0 we see that E(O) _ 00, a characteristic of ferroelcctricity_ Undamped electromagnetiC waves with frequencies within the W1P cannot propagate in a thick crystal. The reflecti\ity of a crystal surface is expected to be high in this frequency region. as in Fig. 15.
4For the ""lendollto.,.,...plcx stNC'luru, ~ A. S. Barl<t:r. Jr., -Infrared dicleetlT beha.iorof In "'l!rrod~Iri(;lt!l. E. F. Weller. ed., Else\;"r, 1967. Barker gi\'es an Inleresting deri..alion of the LST relalion by causality argument.
ft.~"'-driccry5tak-

For films of thickness less than a wavelength the situation is changed. Because for frequencies in the gap the "'ave attenuates as exp(-IKlx), it is possible for the radiation to be transmitted through a film for the small values of IKf near WI..., hut for the large values oflKI near fdT the wave will be reflected. By reflection at nonnonnal incidence the frequency Wl.. of longitudinal optical phonons can be observed, as in l'"ig. 16. Experimental values of E(O), E(oo), and fdT are given in Table 3, with \'011ues of WI... calculated using the LST relation. Eq. (62). We compare values of
Table 3 Lattice frequencies, ehiefly at 300 K

C.,..,
UB UF
UCl

die1ec:tric
CQII!tanl

"'

...

cliekdril:
......"

""' "" ,,-)

3.6 I.B 2.1 3.2 1.1 2.25 2.6 1.5 2.1 21 I.B 2.6 2.6 3.0 5.1 5.4 4.0 4.6 295 8.5
10.9 14A

...,... in

IOU ,-'

1lIr. in 100s

ellJl"rinlental

,-1 LST relation

21.

12.9

II.
5.8 3.6 3.0 4.5 3.1 2.5 3.6
'7

8.B 120
13.2

"1.5
61 1.8 5.0 3.B 6.1 4.0 2. SA I.B 3.1 1.6 3.0 I.B 3A 2.5
14.

UBr N,F Nael NaBr

IT

KCI
Kl

HbF UbI

c.O
C.l

5.1 5.B 6A 5.5 4.85 5.1 6.5 5.5 1.' 5.65

31.9

I.B 2.B

IA
I.B

I.'
1.2
0.61

110 11B,

29.8
12.3

.....'
CaP
InA> JnSb SiC

AgCl

MgO

131 B.8
10.7 12.9 16.1 12.4 14.9 17.7

CaA,

Ca,h InP

B.6
12.3
13.6

I.B 1.5 1.5 6.B 5.1 4.3 5.1 4.1 35

14.9

1.6 5.5 4.6 6.5 4S 3.1

17.9

'i

B.6 5.5
II.7 15.8

8.1 55
11.7 15.8

Ge

25.1 B.B 5.1

25.1 B.B 5.1

Data larg.,ly due to E. BUTSlein.

'wr-----,--==,----,----,----,--,-,

/ ;-/ . . .-<:~

"

j .,
'"
'10

! "

II

,r
I
'"

I ,/ I I ~ f I ! J I I '

---,""

------:zoo K

--lOOK

.,

'00

rip'''' IS fIef1ectanoe c aaystal ofN.et a. ~'er.ol tempentures. .-ersus ~,'dength. ~ nominal ,... lues or ",,-and...,. at room lempemure cn-nespond lowa\'d~ d35and 61 x 10-4 em. respecti"ely. (After A.. Mibuisrn el all
W"" numb..r In "",

,.,

,00>

,

,
~.

~'r"r , , ,
I

" _.1,,-.'

.~

'00

, , , ,

,
,
I

OPI~'"
.Il~

,I

i ,

I Tr.won.....

...
..,...bs:

v ,,,,,,,,," .......

l..ongjIudlnol

, ,
,
I

....

00 " \\'.nknglll ,,, 10-' "'"

3<0

FiSll1"e 16 Re~ \I,nus "'a\<dcnglh ol. LiFfilm backed by sil,'a', for radiation inci!k... t near The longitudinal uptiQII pbooon abiorbs Wongly the ndiation poWiud (p) in the plane normal to the flIm, but han'll)' .II the radiation polarized (.) p.zaIld to lh.. flIm. (After D. W. Ber...,man.)

c
29.

w,}wr obtained by inelastic neutron scattering with experimental values of ((O)/E{oo)1 1f2 obtained by dielectric measurements;

W,}W"I'

NaI 1.44 0.05

KD,

[(0)/(00)JI/2

1.45 0.03

1.39 0.02 1.38 0.03

eaAs 1.07 0.02

1.08

The agreement with the LST relation is excellent.

ELECTRON-ELECTRON INTERACTION

Fermi Liquid. Because of the interaction of the conduction electrons with each other through their electrostatic interaction, the electrons suffer collisions. Further, a moving electron causes an inertial reaction in the surrounding electron gas, thereby increasing the effective mass of the electron. The effects of electron-electron interactions are usually described within the framework of the Landau theory of a Fermi Iiquid. 5 The ohject of the theory is to give a unified account oCthe effect of interactions. A Fermi gas is a system of noninteracting fermions; the same system with interactions is a Fermi liquid. lAndau's theory gives a good account of the low-lying single particle excitations of the system of interacting electrons. These Single particle excitations arc called (Juasiparticles; they have a one-to-one correspondence with the single particle excitations of the free electron gas, A quasiparticle may be thought of as a Single particle accompanied by a distortion cloud in the electron gas. One effect of the coulomb interactions between electrons is to change the effective mass of the electron; in the alkali met:tls the increa.~e is roughly of the order of 25 perccnt. Electron-Electron CoUisionJ, It is an astonishing property of metals that conduction electrons, although crowded together only 2 A apart, travel long distanccs between collisions with each other. l1te mean free paths for c1cctronelectron collisions are longer than 10,1 A at room temperature and longer than WcmatlK. Two factors are responsible for these long mean free paths, without which the free electron model of metals would have little value. 'The most powerful factor is the exclusion principle (Fig, 17), and the second factor is the screening of the coulomb interaction between two electrons. We show how the exclusion principle reduccs the collision frequency of an c1ectron that has a low excitation energy EI outside a filled Fermi sphere (Fig. 18). We estimate the effect of the exclusion principle on the two-body (:ollision 1 + 2 _ 3 + 4 hetween an electron in the excited orbital 1 amI an
~L Landau, '~fhe<lry ur a Fcrmi liquid," Sovict Ph)'llics IETP 3, 920 (1957); L, Landau, "Oscillalioos in a Fermi li'luid," S<;.viel PhY5ic1 IETP 5, IO (19;>1); 5eC alw D. Pines and P. K<niercs, Theory of quontu", liquids, Benjamin. 1966, Vol. I.

10 1'1...."..".,. polarilons. lind polo,..,...

29.5

Figure 17 A collision between tWl> ",Ieruns of

wavc~ctor k, "nd ~. Afler tnc collision the

parlides ha~-.: "'''''\""''''''dc. k3 "nd 1<... lhr: ~ri oeJw.ion principle all.o."S collisions <lilly to final .tales ~ I<.. "tuo;h,,~ ~_I beb", the coll..

""'.

/"~-

,(
\ \

---

.-,
"\

,,

-,
C

,, ,
(0)

.-, ,

.,

,.)
F"lg\Ue 18 In (.:J lhe ",Iertmns in initial urbilab I and 2 collide. If the urbitals 3

""oj" we initWly

~t. lhe electrons I and 2 can OOOJp)' orflitals 3 and " after the ."oll;s>un. Energy and morrwmtum are ~5ef\w. In (b) Ihe eledrons in initial orbi...l, I and 2 ru.ve no \-acant final urllitals .,-aibbk-

thai allow CTlergy to he (''O'1Set"Ved in the collision. Or-hilals soch as 3 "nd 4 would COl>sef\C cnergy ..,d momentum, bUlthcyllre ..Iready lilted wilh other elcct,.,..... In {el we ru.\'edenoted wilh x the _"CVedor of the cenl.". of ....... I and t. All ~irs of orbitals 3 and 11 Jomeroe rnotnmtum ..nd energy if they lie at upposile ench of" diameter oftbe small sph<ore. The smaI1 sphere willi dnwn &om the .,..,.,ter of rl\lWC to ~ Ihrough I and 2. Rul oot all pairs of points 3. "au' aDu..'ed by lhe e..clusion prir'dple, b- both 3. 4 musllie ouhioo the Fenn; spllere. Ihe fraction all"""",, is -I,I"F.

296

electron in the filled orbital 2 in the Fermi sea. It is convenient to refer all energies to the Fermi level p. taken as the zero of energy; thus t"J will be positive and E2 will be negative. Because of the exdusion principle the orbitals 3 and 4 of the electrons arfer collision must lie outside the Fermi sphere. all orbitals within the sphere being already occupied.; thus both energies ('3. ~ must be positive referred to zero on the Fermi sphere. Th(' conservation of ellergy recJuire5 that k21 < E.. for otherwise 3 + ., = EI + E2 could not be positive. This means that collisions are possible only jfthe orhital2 lies within a shell of thickness E) within the fermi surface. as in Fig. 1&. Thus the fraction -EifEl-' of the electrons in filled orhitals provides a suitable target for electron 1. But even if the target electron 2 is in the suitable energy shell, only a small fraction of the final orbitals compatible with conservation of energy and momentum are allowed hy the exclusion principle. This gives a SCL'Olid (actor o( EI'EF. In Fig. 1& we show a small sphere on which all pairs of orbitals 3, " at opposite ends of a diameter satisfy the conservation laws, hUI collisions can occur only ifhoth orbitals 3, " lie outside the Fermi sea. 'Ole product ofthe two fractions is (E,JEF'f. If EI corresponds to I K and EF to 5 x l<r K, we have (E1JEF'f - " X 10- 1 the factor hy which the exclusion principle reduces the collision ratf'. The argument is not changoo for a thermal distribution of electrons at a low temperature slleh that k,/f EF. We replace EI by the thermal energy -kB T, and now the rate at which e1ectron-clechun collisiolls take place is reduced below the classical value by (kDTJEF'f. so that the effective collision cross section 0' is

(63)
where 0'0 is the cros5 section (or the electron-electron interaction. 'Ole interactiOIl ofolle electron with another has a range of the order of the screening length 111.1 as in (34). Numerical calculations give the effective cross section with screening for collisions between electrons 3S of the order of 10- 15 em! or 10).2 in typical metals. The effect of the electron gas background in c1cetron-electron collisions is to reduce 0"0 below the v",lue expected from the Rutherford scatlering equation for the unscreenoo coulomb potential. However. much the greatest reduction in the cross section is causetl by the Pauli factor (kIlTJEFt. At room temperature in a I)'pical metal kllTJEF is _10- 2 so that 4 19 0" - 10- 0"0 - 10em 2 The mean free path for electron-elcctron coIJisioos is f - llnu - 10- em at room temperature. This is longer than the mean free path due to electron-phonon collisions by at least a factor of 10,. so that at room temperature collisions with phonons are likely to be dominant. At liquid heIiUln temperatnres a contribution proportional to r 2 has heen found in the

resistivity ora nllmber of metals, consistent with the fonn ofthe electrun-electron scattering cross section (63). The mean free path of t'lectruns in indium at 2 K is of the order of 30 em, as exp:ied from (63). Thus the Pauli principle explains one of the central questious of the theory of metals: how do the electrollS tra\'ellong distances ,vithout colliding with each other?

ELECTRON-PHo,...ON J.1\o7ERAcnON: POLARONS

loe most common effect of the electron-phonon interaction is seen in the temperature depemlence of the electrical resistivity, which for pure copper is loSS micrOOm.an at O"'C and 2.28 microhm.an at lOO"C. The electrons are scattered by the phonons, and the higher the temperature, the more phonons there arc and hence mOTe scattering. Alxn<e the Deb)<e temperature the number of thermal phonons is roughly proportional to the absolute temperature, arK1 we find that the resistivity increases approximately as the absolute temperature in any reasonably pure metal in this temperature region. A more subtle effect of the electron-phonon interaction is the apparent increase ill electron mass that occurs because the electron drags the heavy ion corcs along with it. In an insulator the combination of the electron and its strain field is known as a polaron, 6 Fig. 19. The effect is Large in ionic crystals because of the strong coulomb interaction between ions and electrons. In covalent crystals the effect is weak because neutral atoms have only a weak interaction with electrOlls. The strength of the electron-lattice interaction is measured by the dimensionless coupling (;onstant a gi\'en by deformation energy 1 -a"" 2 ,,"WI.

(64)

where WI. is the longitudinal optical phonon fl'Cfluency ncar zero wavevector. We view ia as "the number of phonons which surround a slowmoving c1edron in a crystal." Values of a deduced from diverse experiments and theory arc gi\'en in Table 4, after F. C. Brown. The values of a arc high in ionic crystals and low in CO\'alent crystals. The values of the effecth<e mass m!oot of the polaron are from cyclotron resonance eJ(periments. The valucs gi,'ell for the band effectivc mass m were calculated from m!,.,.. The last row in the table gives thc factor m~ 1m by which the band mass is increased by the defonnation of the lattice.
~ QTS. Chapter 7.

.
Table 4 Polaron coupling COlUtanu 0', masses m::.,a, and band masses OIl"' for electrons in the conduction band
KCl Crystal

..
KB, 3.52 0.93
0.43
AgCl Ar)l'
ZnQ

PbS
0.16

InSl>

CaA5

" m~m
mOl/OIl

m;.Jm"'

3.97 1.25 0.5(1 2.5

200
0.51 0.35 1.5

1.69 0.33 0.24

lA

0.85

2.2

0.014 0.014 0.014 1.0

0.00

ma~s

Theory relates the effective mass of the polaron mr,.., to the effective band m" of the electron in the undcformcd lattice hy the relation
m!>ot
;!;>!

mt' ( 1

I - 0.0008a'

'a +

O.OO34~

) ;

(65)

for a .::;g 1 this is approximately m"'(L + iO'). Because the coupling constant a is always positive. the polaron mass is greater than the hare mass, as we expect from the inertia of the ions. It is common to speak oflargc and small polarons. The electron associated with a large polaron moves in a band, bullhc mass is slightly enhanced; these are the polarons we have discussoo above. 'The electron associated with a small polaron spends most of its time trapped on a single ion. At high temperatures the electron moves rrom site to site by thermally activated hopping; at low temperatures the electron tunnels slowly through the crystal, as ir in a hand or large effective mass. Holes or electrons can becume selr-trapped by inducing an asymmetric local deformation or the lattice. This is most likely to occur when the band edge is degenerate and the crystal is polar (such as an alkali halide or silver halide), with strong coupling of the particle to the lallke. The valence band edge is more onen degenerate than the conduction band edge. so that holes are mOre likely to be self-trapped than are elcctrollS. Holes appear to be self-trapped in all the alkali and silver halides. Ionic solids at room temperature generally have very low conductivities for the motion or ions through the crystal, less than 10- 6 (ohm-cmr 1 , but a ramily of <:ompounds has heen reported1 with conductivities of 0.2 (ohm-cmr 1 at 2O""c. The compounds have the composition MA~Is. where M denotes K, Rb, or NH 4 The Ag'~ ions occupy only a fraction of the equivalent lattice sites available, and the ionic conductivity proceeds by the hopping of a silver ion from one site to a nearby vacant site. The crystal structures also have parallel open channels.
7See B. B. {)",cn$ and G. R. Argue. Science 157. 308 (1967); S. Geller, Sciencc 157. 310
(1967).

"

a
F.krlroo

(0

0)
(0

c,-

8
<0

<0

8
'"

8
0)
<0

(0

8
(0
n.e

,.
8
'"

e
0

Figure 19 The fonnlltiun a polaron. (a) A condOC1;on c1L>ctron ;s shuwn in a rigid bttice of an ionic a-yslal, KCI. iOnu on the ion5 adjaC'enllo the electron are ~_ (b) The clmron is shown In an da>lic or debmabIe latticr. ~ dedron plus the usociakd stnIin field is NIIcd polaron. displacement of the toM ina'eases the dier.ti~-e inertia and, hence. the dkdn"c maB ufthc cltron; in KCI thr m&$S is i~ by a l"actorof2.5 "ith ~ to the Iall! 11;0:0')' mllSS in a rlgid lattice. In extreme iltuatium, "len ",;Ih holes, the particle can be<ume self-trapped Ooeali:ted) in the Jatlice. In <;O\...\enl crystab the foll:eS 00 the atoms from the e1eClrOn;lre weaker lhan in Ionic crystals, SO thai pvIaron deformation, are small in covalcut cry>;tals.

or

n.e

\
PEJERLS INSTABlUTY OF UNEAR METALS

Consider a one-dimensional metal with an electron gas filling all conduction band orbitals out to the wavevector at absolute zero of temperature. Pe-ierls suggested that such a linear metal is unstable with respect to a static lattice dcfol'lnatioll of wave-vector C == 2k". Such a deformation creates 311 ellergy gap at the Fermi surface, thereby lowering the energy of electrons below the energy gap, Fig. 20. The deformation proceeds until limited by the Ulcrease of elastic energy: the equilibrium de~rmatioll li. is given by the rool of

k".

d d6. (Ee1ed"",1c + ..10..11.,) ::: 0 .

(66)

Con!>ider the clastic strain Ii cos 2ktt:t. The spatial-average clastic energy per unit len~..t h is E.,1&ot1c = iC.6.2{cos2Qk,.-x) = iClI. 2 , where C is the force COilstant of the linear metal. We next calculate E..ledronk:- Suppose that the ion contribution to the lattice potential seen by a conduction electron is proportional to the deformation: U(x) = 2Ab. cos 2kf'X. Frorn (7.51) we have
EK

= (h 2/21'11)(kJ

+ K2j

:!:

(4(h 2 k1l2mXfl 2 K2/2m)

+ A 2b. 2pl2

(67)

It is convenient to define

XK" A2J(2/rn ;

XF ... h 21!f:lm ;

x - htKkFlrn .

We retain the - sign in (61) and fonn dEK = -A 2b. db. (X"<XK + A 2 b.')112 whence. willl dKhr as the number of orbitals per unit length.

We put it all together. The equilibrium deformation is the root of

iCb. (2;\211U1/'lJh2 k,.) sinh-l(h2k~/mAb.) =

0 .

The root b. that corresponds to the minimum energy is given by

h 2k}/mAb. = sinh(-h2k,mC14mA~
(68)

whence
IAIb.
=z

(2h 2 kf,/m) exp(-h 2 k,."7TC14mA 2)

(69)

if the argument of the sinh ill (68) is :t> 1. We assume k r ... i k _.

Ek<.1'",' ell,,'fjCI

.n.,. Introd...."""
. . <':tIftI)'

F"~
too _
11)'

20 Prierls instability. I::ledroru ..ith ....~

the Fnmi surface have

t'-'if energy lowered

a IattM.-e deformation.

The resull is of the form of the energy gap equation in the Des theory of superconductivity, Chapter 12. The deformation /1 is a collective effect of all the e1cctrollS. If we set W = h'lk~/2m = (.'Onductioll band width; N(O) = 2mJ7rli'kF = density of orbitals at Fermi level; V = 2A'IC = effective electron-clectron interaction energy, then we can write (69) as

1A1/1 "'" 4W exp[ -lJN(O)\']

(70)

which is analogous to the DeS energy gap equation. 8 An example of a Peierls insulator is TaS3.
SUMMAR1'

The dielectric function may be defined as

E(w K) :::

p,,~t(w.K) PUI(W,K) + Pw.d(w,K)

in

t~ms

of the applied and induced charge density components at w, K.

lbe plasma frequency w" = (41T11c2lt(o;I)mprl is the frequency of the uniform collective longitudinal oscillation cl the electron gas against a background of fixed positive ions. It is also the low cutoff for propagation of transverse electromagnetic waves in the plasma. The poles of the dielectric function define lJ>r and the zeroes define w,-.
-A """'iewof ch.arge density "~'<5 and supcrlattires in metallic layered structuns is gi"en by Di Sakoand S. Mahajan. Ad,,,,nres in Pll)'sks 24. 117 (1975); this article contains .\Orne of the most beautiful diffraction photographs In $OM stal(' phrsia. P. B. Littlewood and V. llcine. J Ph~. c. 14.2493(1961) Iu."e d\g,.'TI '-'wthe Frieds dfeo;t Ii modified by the: electron.elrdrtJII intrraction; see alsu J. 1::. Uusd.. Ph)'s. Rev. Letl SI, 296 (1983).

J. A. Wilsun. F. J.

""
In a plasma the CQulomb interaction is screened; it bccQllles (qlr) el<p(-k.r), where the screening length 11k, "" (f).J61TJ1~)IIZ. A metal-insulator transition may occur when the nearest-neighbor separation

a becomcs of the order of 400, where flo is the radius of the IIrsl Bohr orbit ill the insulator. The metallic phase c,,-ists at smaller values of o.
A polariton is a quantum of the coupled TO phOilOIl-pllOlon fields. The coupling is assured by the Maxwell equations. The spl-'Clr;l! re~oll is forbidden to e1ectroma~nclic wave propagation.
nie Lyddanc-S<lchs-Teller relation is
WT

<

<

WI.

wVu?t =

E(O)It:(oo).

Problems
l. SIIr[nce p/(l$rIIO'I'. COllsi<Ier a semi-inl1nite (Jh~m"'l on the positive side of the plane ;:: = O. A solution of L.1place's equiltion V2q> == 0 in the (lbsma is 'Pb:,z) = A cos kx e- h , wlience E" - kA (:os kx e- h ; Ed - kA sin b: e- 4 . (a) Show tliat in the vacuum Ipo(x,z) "" A COli h ~ for z < 0 !h\tidles the boundary cunditioll that the tangential cumporlent ofE be COIltinuous at the houndary. that is, fin(l E. o (b) Note that D, = ~w)E.; Do = E". Show that the boundary condition that the Donnal component ofD be continuous at the oound"lry requires th,lI ~w):: -I, whence from (10) we ha\'e the Stefll-Ferrdl result:
(71)

for the frequency w. of a surface plasma Oscillation.

2. 11lterfoce plasmon:J. We cunsi(!er the plane interf."lCe :t "= 0 between metal 1 at % > 0 an(1 metal 2 at z < O. Metal I has bulk pl:umOrl frequen<.')' W v.' metal 2 has wp2' The (Iielectric constallts in both metals are those of free electron gases. Show
tliat
5Urf.lce

plasmons associated with the intcrUce Iiave tlie frequency

W "" [l(w~1

+ w~,JJII2

3. Al/ven

Consi(ler a solid \\ith all e<Jual conCClltration 11 of e1edrons of mass holes of mass m~. 111is situation may arise in a semimetal or in a compensate<1 scmiC(lllductor. Place the solid in a uniform m"gnetic fieM 8 '" Rz. Introduce the <.:oon!illate ~ =- x + iy appropriate for drcularly polarize<l motion, with ~ lI,wing time (Iependence e -""'. Lt:t w., - eBlm~ all(1 w~ eRlmhc. (a) In CCS units, shuw that ~ - eE+/II~w{w + w~); 6. "" -eE+1mhw(W - w,,) arc the (!ispl"cements oflhe electrons and holes in the electric fldll E+ e-..... = (Er + iE;J /)'-...... (b) Show that tile dielectric polari7.3tion p+ = ne{6 - {.o) ill the regime W <c w~, Wi, liMy be written as p+ = nC(mh + m.)E+/R'l, mill tlie (Iiclectric function ~w). ~ + hrP+'E+ => I + 4m:?pR 2 , where I is tlie (Iiele<:tric C()Ilstant of tlie host lattice anll p _ n(m~ + mh) is the mass lIensity of the carners. If I may be neglected,
lVaVCl!.

m~ and

the dispersion reulioo w'f(w) - cK' be<:omes, for ck!t:trornagllCl.ic w;wcs Ilm),..gating in the:; dircetioll. (2 .. (R 2/41rp)K ll Such ....11\C5 are known as Alfv~11 wa,,', they propagate with the COClstant "e:roty B/(41Tp)l12. If fj ,.. 10 ),;C, II - IOlll em -a, m - 10 S'1 g. the velocity is -lOS em ,-I. Alr,,~ \W\"C$ h.we been observed in JIClTlimeta15 3n(1 in ekdron-hok drops in germanium (Chapter II). 4. lIelioon tOOVU. (a) Emilloy the metl}()(1 of Problem 3 10 treat a specimen with only 00 carrier Iype, say holes in conttntrnlion p, ......tI in the limit Col <:( "'" '" efYllII.,c. Show that t(w) iii 4rrpe'-lm~ .... here O+(w) = E(w)ECtoI). The leon (/ in 0{ has bet>1l nqlectetl. (b) Show further th.11 the diSpcrskln relatiorl becomes

w = (Rd417pf!)K t the helicon (!ispenioo relation, in CCS. For K _ I em I am! B "" 1000 C, estimate the helicon frequency itl sodium metal. (file freqU('ncy is negath'c; ....ith circular-polarized modes the sign d the frequency refers to the sense of the rowtie:m.)

5. E'Wmon mode of (I rpMrfl:. 111 frequency of Ihe uniform {llasmon mode of a sphere is detenllined by the depolariz.ation field E '" -41TP/3 ofa sphere. where the polarization P .. -nero ..... ith r as the average Ilisplacement of the electroflS of conCt:mtra{ion n. Show from F .. rna that tht- resonance frequency of the e1cd:roo gas is ... - 4wnrl3m. Because all elcctrons participate in the oscillation. soch an excitation is called a oollecth'e excItation or collec1ive rmxle of the electron gas.

6. MClgndoplamw frequcnt:lj. Use the metll()(1 of Problem 5 to find the frequency of Ihe uniiOnn plasmon mode of a sphere pJace..1 in a constant uniform rnagnetac rICk! 8. Let 8- be ,dong the .: nis. The solution shook! go to the cyclolfOfl frequency -We - -e8lmc in one limit anll to wo "" (4mlcl3m)112 in another limit. To.lke the motion in the x - !I Illane.

7. Pl!oIon lmmc.h at low~. (a) Find what (56) becomes ..... hen t(...) utalen into account. (b) Show that there is a solutiQr, of (55) which at 10w .....a vevector is
W ..

cKlVf(O). wtlich

is what you elI:pect for a 111l0ton in a cr)'stal of refractive inllex

Ill_f.

8. PIMma frequftlCy and eltrical conductivity. An ?rpnic conductor has recently been fotUKI by optical studies to ha~'e wI' ,.; 1.80 X 101:1 S-I for the plasma. frequency, ami T = 2.83 X 10- 1:1 S for the electron relaxation time at room lempera ture. (al Calculate the electrical condudivity (rom these data.. TIle colrrier mass is 1101 kno.....n and is JJOl oeeded here. Take fE(...) '" I. Corwert the rewlt to units (0 cm)-l. (b) From the crystal aod dJemical structllJ'e, the comillction electron coocefItration is 4.7 x 10"1 cm- 3. Calculate the e1cctTOll effective ,nass me.

9. Bulk moJUfU6 of,he Fermi ga.t. Show that the contribution of the kinetic energy to
Ihe bulk modulusoflhedectmn gas al amolule zero u 8 "" lnrrw}. It iscoovenient to use (6.60). We can use our result for 8 to filKI the ~'eb:ity of 5OUlul, "illch ill a cornpreulble fluid IS v - (Blp)III, "hence v -= (mf3M)I'tCI'. ill agreement wltl, (46). These e51irnates neglect atlractive inlenu.1.ioTls.

10. fle.pon~e of electron go.r. It is sometimes stated erroneously in Looks on deaTO_ magnClism that the static conductivity u, which in gaussian units has thc dilllcnsions of a frequcncy, measllrt.'S the response frequency of a metal to an ek'Ctrlc field suddenly applied. Critici7.e this statement as it might apply to copper at room temperature. 1lle res.isthity is - I .uohm-crT1; the electron col1CCntntion is 8 x tOUcm- 3, the mean free path is -400 A; the Fermi velocity is 1.6 x lOS em S-I. You will not necessarily need .11 these data. Cive the order- or magnitude of the three frequencies u, Wp, and U,. that might be reb'8tlt in the prob&em. Set up and soh-e the problem of the I't'SpOOse of the S)'stem to an electric field (t < 0) - 0; E(t > 0) = l. The syl>tem is a sheet of mpper, the field is applied normal to the meet. Include the damping. Sol,-e the difTerClltial equation by elementary methods. -II. Gop pla.a:mom lind tile I:Gn der Waah irduocfion. Consider ""0 semi-infinite media with plane surfaces t "" O,d. 'Ibe dieled.ric fuoc.tion of the identical media is (w). Show that for surface plasmOllS symmetrical with respect to the gap the frequency n.lliSt satisfy f(W) = -tanh (Kd/2), where K 2 - k! + ~. "Jbe electric potential will have the form '" - I(t) e~p(ik,.x + ily; - lwt) . Look for nonretarded !iOlutions-that is, solutions of the Laplace equation rat!Je.. than of the wave equation. The sum of the rer<>point energy of all gap modes is the nometarded part of the ''''n der Waals attnldion be"'~n the specimens-see N. C. ,-an Kampen, B. R A. Ni;boer. and K. Schram, Physics leiters 2GA, 3lJ7 (1968).

*)

"''0

References
J.
N. Ilodgson, OptiaJ cblcrptiool lInd dilpn"riorl In aoIid$,
ductol'y SUl"\'q'.

a...pm:an and 11aI,

11110. Good intro-

F. Stem. MEk,mentaly theory dthcoptit'al pn:>pcrtiesd mebls, ~ ~ state pbysics IS, 300(1963). II. C. Schuster, cd., One-dimenrlOtlGI condvcfon, Sprinser, 1975.
N. F. Mou, M~IGl-iP\Slllolor InlIl$IUOrUi. Tayklr &: Francis, 1990. L. V. Kekl)'llh, D. A. Kirthnitt, "IHI A. A. Maradooin, Didtdrk fu.naicn of CQnJe,,~d Slis1ems, Ehevlcr, 1989.

SOUD STATE PLASMAS

A. C. lIaynllllm lUld A. D. Uoaonlma." Pl/lS1M Cfflld$ In lmlioondu.cIon': hIllicon lInd II1ft'ln W<lV/l:l, Tllykw &: Francis. UI1l. J. Danleb, C. v. Festenlx-rg. 1/ Racthcr, and K. Zcppenl'cld, "Optical constlUllS d'solids by electron spectrw.-opy, ~ Spnr~r InJeU in modem pilIPla' S4. 78 (1970). E. N. Economou and "- L. Np;. -S~ pbsmaoscilbtions..d related surface cffectJ; in soIids,~ Adt"ClPICU In thmit:al phllfia 27,265 (1974). P. M. f'bt:onan and P. A. WotIf, \\"L'U lInd fnl~ In tdid .stIlte plapnaI, Aadcm'c Press. 1973. Good gencnl mererote'; also mntilins acoount dFerml liquid theory UId l:I<perimcnts. II. Raethu, ucilation of~, Springl'l", 1960; Su.rfoce~. Springe>', 1966. POI_4.RONS

J. J. Oe\'recsc, cd.. PoUzrotlJ rind acitonr, I'tcnum, tlJ64. J. Appel. ~PuIamns, SuI,d .Iate physics 21, 193 (1965).
H

J. Singl"

~Dynamics

of e~cilons," Solid .(ate phys;a 38, 295 (lY64J.

"'is prohlem is wmcwhat diflkult.

11
Optical Processes and Excitons
OPTICAL REFLEcrA~CE KramersKnlllig relations Example: conductnity of collisionless eledrDll gas Electronic interband lrans:itions EXCITONS Frenkel excilons Alkali halides Molecular crystals Weakly bound (Moll-Wannier) excilOtlS EllCiton oondensation into eleclron-hoI.e drops (110)
RAMAN a-FECf IN CRYSTALS ElectrOIl spe<:troscapy with x-rays ENERCY LOSS OF FAST PARTICLES IN A SaUD
SU~f)tARY

307 308 311 312 312 316 318 318 318 319
322 324 326 329 329

PROBLEMS

l. 2. 3. 4.

Cau$3.lity and the response function Dissipation sum rule Renection at normal incidence Conducthit)' sum rule and supen:onducthity 5. Dicledric constant aoo the sem;evnduc;:lor energy gnp 6. Hagen-Rubens relation for infrared reneeti,i')" of melals 7. Davyt!o,' splitting of exciton lines

329
:wi

330 330 330 331 331

~

REFERENCES

--

.,

.
n',K'

,k

JCigure I There are many types of experiment. in ",hich lig.,. interacts w cryst.... Sc\eml prOCCUC$ are illnslr:olcd here

"~th

"''a\'chkc clliCit3tions ,n

'"

CHAPTER 11: OYfICAL PROCESSES AND EXCITONS

The dic!ect.-ic function E(w,K) was introduced in the p.-eccding chapter to describe the response of a crystal to an ekctTOmagnetic field. The dielectric function depends sensiti\c!y on the electronic band stmcture of a crystal, and studies of the dielectric function by optical spectroscopy arc very usdul in the determination of the overall band structure of a crystal. Indeed. in the last decade optical spectroscopy develo(X.'l1 into the single most important experimental tool for band structure detenninalion. In the infrared, visibk. and ultraviolet spectral regions the \\"..lvevector of the radiation is very small compared with the shortest rL"Ciprocallattice vector, and therdore it may lIsually be taken as zero. \Ve are conC<;Tned then with the real E' and imaginary i' p.1rts of the dielectric function E(W) = E'(W) + ii'(w), also written as ](w) + iEz(w). Howe\'er, the diclechic fUllction is not din.:etly accessible experimentally from optical measurements: the directly accessible functions are the renectan(.'C R(w), the refractive index n(w), and the extinction coefficient K(w). Our first objective is to relate the experimentally observable quantities to the real and imagin3ly parts of the dielectric function.
OI'11CAL REFI.ECfANCE

The optical measurements that give the fullest information on the electronic system are measurements of the .-enectivity of light at no.-mal incidence on single crystals. The .-eflectivity coefficient r(w) is a complex function defined at the crystal surface as the ratio of the rcnected electric field E(.-el1) to the incident electric field E(inc): E(.-dl)/E(inc) - r{w)
ill

p(w) exp(i8(w)l ,

(I)

whe.-e we have separated the amplitude p(w) and phase 6(w) components of the .-el1cctivity coefficient. The refractive index n(w) and the extinction coefficient K(w) in tbe crystal are related to the r-ellcctivity at normal incidence by
ll+iK-l r(w) - .c...-,-':::--:~ -1l+iK+l'
(2)

as derived in Prohlem 3 by elementary consideration of the continuitv of the COmponenl~ of E and B parallel to the crystal surface. By definition Il(w) and Iqw) arc related to the dielectric function E(W) by

V f'(w)

- n(w)

+ iK(w)

... N(w) ,

(3)
3/',

".

where N(w) is the complex refractive index. Do not confuse K(w) as used here with a wavevector. If the incident traveling wave has the wavevector k, then the !I component of a wave traveling in the :t direction is
E~illc) = E~

exp[i(kt - Colt)]

(4)

The transmitted wave in the medium is attenuated because. by the dispersion relation for electromagnetic waves, the wavcvcctor in the medium is related to the incident k ill vacuum by (n + iK)k:

EJtrans)

0:

cxp{i[(n

+ iK)b- -

wt)} = cxp( - Kkx) expli(nkx - wt)).

(5)

One quantity measured in experiments is the reflectance R. defined as the ratio ofthe rencctcd intensity to the incident intensity:

R = P(rcn)E(rcflYP(inc)E(inc) = r.r = p'l .

(6)

It is difficult 10 measure the phase O(w) of the reflected wave. but we show below that it can be calculated from the measured reflectance R(w) if this is

known at all frequencies. Once we know both R(w) and 6(w), we can proceed by (2) to obtain n(w) and K(w). We use these in (3) to obtain E(W) = E'(W) + iE'(w), where E'(W) and E'(w) are the real and imaginary pw1:s of the dielectric function. The inversion of (3) gives
E'(w) = 2nK

(1)

We now show how to find the phase 6(w) as an integral over the reflectance R(w); by a similar method we relate the real and imaginary parts of the dielectric function. In this way we can flnd everything from the experimental R(w).

Kramen-Kronig Relations
The Kramers-Kronig relations enable us to flnd the real part of the response ofa linear passive system if we know the imaginary part of the response at all frequencies, and vice versa. They are central to the analysis of optical experiments on solids. The response of any linear passive system can be represented as the superposition of the responses ofa collection of damped harmonic oscillators. Let the response function a(w) = a'(w) + ia"(w) of the collection of oscillators be defined by (8) x... = a(w) F... , wherc the applied force field is the real part ofF... exp(-iwt) and the total displacement % = % is the real part of x... exp(-iwt). From the equalion of motion.

L J


we have the complex response function of the oscillator system:

~ f, _ "" w] - WI + iCMPI o(w)-.<. ,. , . --.<.h ( ' ."'+ OJ ) ...,-CU-IWp, t<f-W} urPJ

(9)

where the constants Jj "" VAt) 3n<\ relaxation frequencies PI are all positive for a passive system. 'Ifle masses are AIr If a(w) is the dielectric polanzability of atoms in concentration n. lhen f has the fonn of an oscillator strength times ne'llm; such a dielectric response function is said to be of the Kramers-Hcisenberg Conn. 1be relations we develop also apply to the electrical condueth'ity 0(",) in Ohm's law, j.. = o{w).,.. We need not assume the specific Conn (9), but we make usc of three properties of the response function viewed as a function of the complex variable w. Any function with the following properties will satisfy the Kramers-Kronig relations (II): (a) The poles of a(w) are all below the real axis. (b) The integral of a(w)/w vanishes when taken around an infinite semicircle in the upper halfof the complex w-plane. It sufficcs that a(w) _ 0 unifonnly as IwI- al. (c) The function Q'(w) is even and a"'(w) is odd with respect to real w. Consider the Cauchy integral in the fonn
I P a(w)=-.
1Tl

I" --d, ,-w

0(,)
_00

(10)

where P denotcs the principal part of the integral. as discussed in tlle mathematicalnote that follows. nle right-hand side is to be compldoo by an integral over the semicircle at infinity in the upper half-plane. but we have seen in (b) that this intejUal vanishes. We (.>(juate thc real parts of (JOj to ohtain

J P a'(w) = -0"(,) - d s = -1 P 1T _... 17"

I" ,-w

[I" "'(')
0

--lIs $-W

I"

o"(v) -dp] _.. 1J-W

In the last integral we substitute, for -1' and use property (c) that a"(-,) = -a"($), this integral th<"'11 becomes

and we have. with

---+--~

I $-W

..
s2_cJl

s+w

310

the result

(lJa)

This is one ofthc Kralllcrs-Kronig relations. equating the imaginary parts of Eq. (10):
a~(w):1

'nlC

other relation follows on

1p -7T

J"

a'(,) 1 P --ds== -7T

_,.$-W

a'(,) I L" ---, a'(,) d] [L" ---,,s-w s+w

whence

a"(w) = - ---;-

2w

f"" P.t

o:'(s)
Sl

'(U

(lib)

These relations are appli(~1 below to the analysis ofoptical rcflectalll:e data; this is their most important application. Let us apply the Krarncrs-Kronig relations to r(w) viewed as a response function between the incident and reflected \va,'cs in (1) and (6). We apply (ll) to

In r(w)

In R 111(11) + i6(w)

(12)

to obtain the phase in terms of the reflectance;

6(w) = - w -P
7T

L" In Res) ell

$2

ds .

(13)

We integrate by parts to obtain a form that gives insight into the (;()nlrihulions to the phase angle:
1 8(w) = - 21T

L" In l>+wldlnR(.) -d,. s-w ds

(14)

Spectral regions in which the rcl1ectance is constant do not contribute to the integral; further. spectral regions $ J> w and s <c w do not contribute much because the function In Ks + w)/($ - Cd~ is small in these regions. MlltMmalical Nole. To obtain tllC Cauchy integral (10) we take the integral J a(s)(s - W)-I d.s over the contour l in Fig. 2. 1bc function o(s) is analytic in the upper half-plane. so that the value of the integral is zero. The contribution of segment 4 to the integral vanishes if the integrand a(s)/s_ 0 faster than 181- 1 a~ 1$1- QQ. for the response function (9) the integrand _ 0 as 1$1-3 ; and for the

's E. T.

'\llituJ;er and C. N. Watson. Mod~rn "...u~. C

~,I935, p. 111.


FiJure 2 ConIOllf lOr the Cauchy principal \"3.I\le
integral.

nlC

conduclh'ity 0(8) the integrand limit as 11 -+ O.

0 as IsI- 2 .

segment 2 contributes. in the

(~

- - ds -+ a(w) s-w

a>

1
or

iu elt/dB .. "" -'lTio(w) lie

to the integral, where s ::: W + 11 e l ', The segnl<.:nts I and 3 are by definition the principal part of the integral lx:tv,.~n -;XI and 00. Because the integral over 1 + 2 + 3 + 4 must vanish.

as in (10).

\I)

I I I- a>
+ P
(3)

- _ - (Is"" Toia{w) _.. s W

(15)

EXAMr(.E: Carnfue'iv1ty o!Collhionlat Electron Ca,. Consider a g;Js of free e1eclrons in the linlit as the collision frequency goes to 7A:1"O. From (9) the response function is. with!"" Urn,

1. - ' - -1 a(w) .. - -1 - lim - - [-' - i7T.s(W) ] mILl p-oO w + Ip mw w

Kronig relation (l1a). by which

(16)

by the Dirac identity. We COfIBrm that the delta function in (16) salis6es the KrameTS'

a(wl--!.{'" ~ds ... __ ,m

r-J

,,J'

(17)

in agreement with (16). We obtain the electrical conducti,'lly o{w) from the dielectric function
(18)

(CCS)

o(w) - (-iw/47r)[f(w) - 1] ,

(19)

for the Maxv.-ell equatk>rl can he written either as \: curl H .. 47ra{w)E - IwE or as c curt H .. -iWf(W)E. We combine (16), (18), and to find the cooducthity or a collisionless eledron gas:

[J9)

dew) + Io'(w)" :1 [1T8(W) +

~]

(20)

For collidonless e1ectroN the real part of the conduct;> ity has a delta functiorl at w =

o.

'"
Electronic Intcrband Transitions
It came as a surprise that optical spectroscopy developed as an important experimental tool for the determination of band structure. First, the absorption and reflection hands of crystals arc broad and apparently featureless functions

of the pholon energy when this is greater than the band gap. Second, din:d interband absorption ofa photon will occur at all points in the Brillouin zone ror which energy is COI1SCT\'l.-d:

"W

ftw =

EJk) - dl.:)

(21)
W

where c is an empty band and v is a filled hand. The total absOl"ption at giV(.'I1 is an integral o\'er all trnnsitions in the zone that satisfy (21).

Three factors unraveled the spectra: The broad hands arc not like a s(X-oral line greatly broadened by damping.
hut the bands COOV.1' much intelligence which emerges when derivatives are taken of the rcflect:uu:e (Fig. 3); derivatives with rcs(X.'Ct to wavelength, dee-tric field, tt:mperature, pressure, or uniaxial stress, for example. The spectroscopy of deri\'3tives is called modulation s(X.'Ctroscopy. ne relalion (21) does not exclude spectral structure in a crystal. because transitions accumulate at frequencies for which the bands c, v are parallclthat is, at frequencies where
(22)

At these critical points in k spa<.'e the joint d<'>flsity of states Dc!.If~ + "w)D,.(If~) is singular. according to the same argument we uS<.-<l in (5.37) to show that the density of phonon modes D(w) is singular when V"w is zero, The pscudopotcntial method for calculating energy bands helps identify the positions in the Brillouin zone of the critical points found in modulation spectra. Band-band energy differences can be calculated with an accuracy a~ good as 0.1 eV, The experimental results can then be fetl back to give improvements in the pscudopotcntial calculations.

EXCITONS

Rdlcctance and absorption spectra oRen show stn.cture for photon energies just below the energy gaP. where we might expect the crystal to be transparent. This stru<:ture is caused by the absorption ofa photon with the creation ora bound electron-hole pair. An electron and a hole may be bound together by their attraclive coulomb interaction, just as an electron is bound to a proton to ronn a neutral hydrogen atom. The bOLlnd electron-hole pair is called an exciton, Fig. 4. An exciton l'fln mo\'e through the M)'stal and transport energy; it does not transport charge

'f---+.';c----;3.~.--"3CA:--C3!;':-----

J<'igurc 3 COllllhlrisoll of(~) rcnectanw, (b)wlI'"'-'length dcr(valin: rcn,,,,lml('C (11<$1 dt";\':lli\'e), "m' (e) dc<.trordkdancc (third Uen,.,.t;"c). of the spc.>ct,...d region in gcnnllniUlH bct"'t,'f:n 3.0 lind 3.6 c\". (After data b)- D. D. Sell, E. O. Kane, ".lid D, t:. Aspnes.)

because it is ek>ctricall)' neutral. It is similar to positronium, which is fanned from an electron and a positron. Excilons can be fonno:! in ~'cry insulating crystal. When the barKl gap is indirect, exciton! ncar the direct gap rna)' be unstable with respect to decay into a free electron and frt.'C hole. All excitons arc unstable with respect to the ultimate rc(:ombination process in which the ck'Ctroll drops into the hole. Ex~ citons can also form complexes, such as a bicxciton from two excitons.

We ha\le seen that a free ckdron and free hole arc created whene"'cr a photon of energy greater than the energy gap is absorbcd in a crystal. The threshold for this process is hw > E~ in a din..c t pnx:css. In the indirect phonon-assisted process of Chapter 8 the threshold is lower by the phonon energy hO:. But in the formation of cwtons the energy is lower(.'{1 with respect to these thresholds by the binding energy of the exciton, whieh may he in the range 1 meV to 1 eV. as in Table 1. Excitolls can be form{..d by photon absorption at any critical point (22). for if VJcE. = V"t"c the group ,,elocities of electron and hole arc equal and the particles may be hound by their coulomb attraction. Transitions leading to the formation of cxcitons below the energy gap are intlicatcd in Figs. 5 anti 6.

3..

F"ogure.. An ~Ign bound dcroo-hole pair, un..ny lYe.: 10 ,noo."e toccthl-r through the CT)'5taJ In some rapeds it is s;miw-Io an atom of positron;u,n, ro.."'c~d from a positron and all Ck:o;IroOl. The eX"'lot' ~hown Is a MOll-Wannicr cM;;lun: it ;s _aldy bound, ....ith an ,"cr.-go: ckdronh.olc l!islarw:e I.a~ in t'(lfnp<U1!oO<1 "11h a blljoe constant

OPe

4b A lightly bound or Fre.,kd <:KcilOn .ro.."11 1oc:aIld 01. atom in an a1h1i halidr ~t:d.. An i<k<aI I-'rellkcl ea;1On ...,11 lra>..:1 as a wa'e Ihrou~oout the cr)lotaI. bUllhe eketron IS at..~'S close to the hole.

F"~

Table 1 Binding alel"g)' or cllcilom. in mcV

Si Ce
CaA,

14.7 4.15

&0

56.
4. (0.4)

Rbel
LiF
AgBr

440.
(1000)

42
3.S

I"P InSb
J(J

ro.

CaP

CdS CdS.

29. 15.

KC' KB,

480. 400. 400.

AgC'

TIC1

"'-

nB,

II.

Dab. assembled by Fn:dcrid: C. B"",Tl and Arnold SclunodL

II

Oplkall'roccun aad

E:r:c;'on~

31...

Candudior, balld ./'

(..lfectiw man .....>

Mit"" lewis

---------TE'Ie11W 1:"1'. F.c

Vale""" band

(eO''di,.,

MIa'S

mAl

~,,-----------l-

Figure 5 1i:X(.~toll Ic,-els in rdaHon to the rundudlun band .,dge, for a ')nlple hantlstrnetun: wIth both conduction anti vak~ b"nd Ige-s al k = O. An exciton ~-an ha'e trnnsbliolldl kineticenerg. Exdton. arc I",staule with rc.p.,ct to rndi.1ti,.., Il..~~,rnb;nalion in which the c11.'<.1ron tiroI" 1nlo lh e hole in the valence bm1<l, acrompanioolw the emission nf a ('holon or phollon.

.-

Cootko:tion band continuum

Valence loand <.'Ol,llnunln

Figure (I Encrg levclsofan e,citon cre.tted in" dired process. 01,lirnl transition. rram Ihe topoC the vak"..., band arc shown b Ihe anoW!i; the longest arrow COrTCspomls to the cn('T:>' W\p. TIw., binding ener~ of tile ex~;ton i. E... referred 10 afn:c ck'dron and fn.~ hole. The k>wcsl frt'fluel>t: absorption line of the ~T)'s1al at absolule zero is not Em Ullt i. E~ - E~_

The binding energy of the exciton can be measured in three ways: In optical transitions from the v.llenee hind, uy the difference betwccn the energy required to create an exciton and the energy to credte a free electron and free hole, Fig. 7. In rec.1lmbin.ttion luminesc.'enc.'e, by comparison of the energy of the free e1ectnm-hole recombination line with the energy of the exciton recombination line. By photo-ioni7~\tion of excitons, to form frcc c<\niers. This experiment requires :l high (:oncentration of exdtons.

'"
"

, "

10

If"

, ,
I
I I
t

06

,
, ,

I , , ,
k,.

"'-,,-l;'"-'I~!c""~~\--,rlk"",,,-l.~:;--,,!cb,-Tl
!'hoI.., ,..... ,lQ '"
Figur~ 7
~\

"n~~ ....ar

Efled of all exciton lc-'CI <}I. th.., uilliclol ~'OI' of a ...micond,l(1or fur photons a the ..... nd IW> ,,;n jl,:l1bun, arsenide al 21 K Th,' 'l'rlK".ol,;ca.1e i. II.... I"I"",,t> abMlrpil;

lion coMfocwnt, .. ",I{:i/ ... , ....ap(-m:j. The energy !tIPalld t"Jlnlon bind~t'l"'1O'aK' do..~
from I~ sl~pe ollhc absorpt>on cu.-...:. 111<" pp E" is I 521 ,,\' .od tbe ~on b."dinlt .. ~. o lJ03.l ..\' (Afl6 M V Slwge)

\\'e discuss excitons in two different limiting approximations, one by Frenkel in which the exciton is small and tightly bound, and the other by Mott and "'annier in which the exciton is weakly bound. wilh an electron-hole separation large in comparison with a lattice constant. Intermediate examples are known.

Frenkel ExcitoRS

In a tightly bound exciton (Fig. 4b) the excitation is localized on or near a single atom: the hole is usually on the same atom as the elcdnm although the pair may be anywhere in the crystal. A Frenkel exciton is essentially an exdtoo state of a single atom, but the excitation can hop from one atom to another by \>irtue of the coupling between neighbor-so The exdtation wa\'e trowels througll the crystal much as the re\"Crsed spin of a m~"llon tr-wels through the crystal. The crystalline inert ~s ha\'e excitons which in their ground states correspond somewl1."lt to the Frenkel model. Atomic krypton h.. "\S its IoY..est strong atomic transition at 9.99 e\~ The corresponding transition in the crystal is closely equal and is at 10.17 e\', Fig. 8. 11lC energy gap in the crystal is 11.7 eV. so the exciton ground state energy is 11.7 - 10.17 = 1.5 eV. referred to a free electron and free hole separ-Ited and at rest in the (:rystal. The translational states of Frenkel excitons have the forlll of plUI);lg.lting waves, like all other excitations in a pcriexlic structure. Consider a crystal of N

II

OpfiNJI p~ and ErciIoou

317

""1'\ ,,
,

"'"
"-,
12.0 fw. r\' II 0

"'00

F;g.. ~., 8 Absor-pliun spcdr\Jm of 00/,,[ "I)"(lton at 20 K (Aflcr G. BaldinI.)

atoms on a line or ring. If UJ is the ground state of atom j, the ground st.'l.te of tile (:rystal is
tJI~::: UIU:z U .." _IU.... ,

(23)

if interactions between the atoms are neglected. If a single atom j is in an excited state tlJ , the system is described by
(24)

This function has the same energy as the function t;, with any other atom 1 excited. Hov.'C\er, the functions If' th.'l.t describe a single excited atom and N - 1 aloms in their ground slate are not the stationary quantum states of the prohlem. If there is ally interaction between an excited atom and a nearby atom in its ground state, the excitation energy will be passed from alOIn to atom. The eigcnstatcs will h:we a wavelike form, as we now :.how. 'Vhen the hamiltonian of the system operates on the fundion t;) with the jth atom excited, w(' oblain

(25)
where is the free atom excitation energy; the interad.ioll T measures the rate oftmnsfcr ohhe ('\citation fromj to its nearest neighbors,j - I andj + 1. 111c solutions of (25) arc Wa\ es of the Bloch form:

tJlk :::

L , e),p(ijka) 'PJ

(26)

318

10 sec this we let

'Jb/Jk
='

J(

opemtc on VJk:

L eijka:JlC{!j = 2: elj~a('PJ + T(<f'J-1

) )

+ ~+l)l

(27)

from (25). We rearrange the right-hand side to obtain

Jflh: =

2: dik"[1f + T(e 1k" + e-il''')l'P

j
j

= (If

+ 2T cos ka)t/Jk

(28)

so that the encrb'Y eigenvalues of the problem are

Ek=If+21'coska,

(29)

as in Fig. 9. TllC application of periodic boundary umdiliolls determines the allowed values of the wavevedor k:
k = 2m/Na;

s = -IN, -iN + I, ... , iN - 1 .

(30)

Al/.:.clli lIaIicles. In alkali halide crystals the lowest-energy excitons are 10elllized on the negative halogen ions, as in Fig. 4b. The negative ions have lower electronic excitation levels than do the positive ions. Pure alkali halide CTystals arc transparent in the visible spectral region, which means that the exciton energies do not lie in the visible, but the crystals show considerable excitonic absoJlltion structure in the vacuum ultraviolet. A doublet structure is particularly evident in sodium bromide, a structure similar to that of the lowest excited state of the krypton atom-which is isoelec.. tronic with the Br- ion ofKBr.. The splitting is caused by the spin.orbit interaction. 111ese exdtons are Frenkel exdtons. Molecular Crystak In molecular crystals the (:ovalent binding within a molecule is strong in r.:omparison with the van der '''aals binding between molecules, so that the excitons arc Frenkel cxcitons. Electronic excitation lines of an individual molecule appear in the crystalline solid :\S an exciton, often with little shift in frequency. At low temperatures the lines in the solid are quite sharp, although there may be more structure to the lines in the solid th:m in the molecule because of the Davydov splitting, as discussed in Problem 7.

Weakly Bound (Motl-Wannier) Ercitons

Consider an electron in the conduction band and a hole in the valelll::e band. The electron and hole attract each other by the coulomb lXltential
(CCS)

(31)

where r is the distance between the particles and E is the :\ppropriate dielectric constant. (111e lattice polari7~ltion contribution to the dielectric constant should not be included if the frequency of motion of the exciton is higher than the optical phonon frCt"luencies.) There will be bound states of the exciton system haVing total energies lower than the bottom of the conduction band.

~-2T----

Figure 9 E".. ~ ,r,.,.,~ ".,.\"c....- ctOl' for;l. Frt. .. ,~d <': 1011, mk"lal t ....;th posili,'c ~I- ightx. tr:msf mlCfllld"'" T.

-ft/"

'111e problem is the hydrogen atom problem if the energy surfaces for the electron ;\lId hole are spherical and nondegenemte. 111e energy le\'els referred to the top of the \'alence band are gi\en by a modified Rydberg cquatoo
(CCS) Here
II

En

=:

Eg

2!l2ln 2

Me'

(32)

is the principal quantum number and p. is the reduced mass:

-=-+P.

m.,

mit'

(33)

fonnet! from the effective masses III... 1111, of the electron and hole. The excitoll ground st.'l.te energy is obtained on setting II =: 1 in (32); this is the ionization energy of the exciton. Studies of the optical absorption lines in I,:uprous oXide. CU20. at low tempemtures give results for the exciton level Spacing in good agreement with the Rydberg Ct:luation (32) except for transitions to the st.'l.te n =: 1. An empirical fit to the lines of Fig. to is obtained w;th the relation t{cm- I) =: 17.508 - (800"11 2). Taking =: 10, we filld p. :>I 0.7 III from the coeffidcnt of Hnt '111e constant term 17,508 em-I corTespondsto an energy gap Ell "'" 2. 17 eV.

Exciton Corlden.sation info EleClron-llole Drop.s (EllD)

A condensed ph.15C of an c1C1.:tron-hole pL"lSm.'l. foons in Ce and Si when maintained at :1 low tcmpemturc and irmdiatcd by light. The following scqucll(.'C of cvents takes ploc'C when an electron-hole drop (EHD) is formed in Ce: The absorptior of a photon of energy ht.JJ > Ell produces a free electron and

'"
-,

'"
.",

.".

" .s

L
J -,

,f~,oo"'---'''c ... ''''---~''c .''''---~''c .. .:''-..J ..

rb"lon
t.....,tg)

in tm-'

Figure 10 ~lthm dIne optic.'llIlrnn~ml~sio" "<.'I"5US photon energy in c"prou~ odue at 77 K ,00...'''11 a M'rit."li 01 c.a:iIOll lina. !'>ole that on the ,-erlk2l ...is lh.: Iog:uithm is pIotlccl okcrcllloing up"''3J"d, Ih... :l pnlc coneponds to abiorptiorl The balK! glIp . iii 2.11 c\'. (AAer P. W. &'umeisltt.)

free hole. with high efficiency. 1'lCse (:ombine mpidly, perhaps in los, to form an exciton. "J1IC exciton may decay with :umihilation of the e-h pair wilh a lifetime of 8 ""S. But if the exciton concentration is sufficiently high-moer 10 13 cm- 3 at 2 K-most of tile cxc:itons wiJl condense into admp. The drop lifetime is 40 JJ.S. hut in strained Ge may be as long as 600 JLS. Within the drop the cxcitoos dissolve into a degenerate Fermi gas of electrons atld holes. with metallic prop-erties: Ihis slate was pn..-dictcd by L. V. Kcldysh. 'lbe binding energy in Cc is L8 mcV with respect to free exdtons, and the concentration II = 11 = 2.57 X 10 17 cm-3
Experimental studies of the condensed EIID phase ha"e utilb.ed the methods of recombination (luminescence) radiation. light scallcring. plasm."\ resonan(:e. Alfvcn W,I\'C resonance. and 11 - n junction noise. Figure J 1 shows the recombination radiation in Ge from free exdtons (714 meV) and from the EIlD phase (709 meV). 'I1lC width of the 714 meV line is auDuntcd for by Doppler broadening. and the width of thc 709 mc\' line is (:omv:.ltible with the kinetic cnergy distribution of electrons and holes in a Fermi gas of concentration 2 x 1017 cm -3. Figure 12 is a photograph of a la'1,'C EIIO.

""

h<.I.

'"
~,

meV

'"

'"

Figure II fu,comb;nalion oollltion ci fret: dcdrool anJ a ckctron-hdc drops in Co: at 3.04 K. 1lw': Fern,; energy In the tlrQ() b f'F ami the mh.,si....: energy of the drop ..;Ih respect III 1I rr~ exciton is ._ (Arlc,.... T. K. Lo.)

PholorPPb ofan cltron-hok drop in it 4 n>m th5k r:A pu,,", p:mW1ium. 'The drop is the intense spot a.I~nl (0 die ~ screw 00 the ~n a the daoIk. lh.. phdognoph is the image dille drop obWned by msing ib ek-ctmn-ldc ~hon lum~ 01'110 thr: ~ of an inrl1lr'l':d~ ..... tn"e \idc:cnn i~ lube. (After J P. Wolfe d aI.)
F~ Ii

'"
Table 2
Cl')stai
(unslrt.'SSt...I)

Electron-hole liquid parameter:;

eo.,,:enlration. "or 1'. in C",-3

Binding energy relative

to free {',.dloll, in me\'

Critical
temperature, K

Ge
Si

18 93
17.5

2.57 x 1017
3.5 x 10 111 8.6 x 10 16

CaP 3C-SiC

6.1 23. 45.

11.

10.

10 111

41
JIII17, Springer, 1984.

Comlcsy of D. Bimbcrg, r.,.- r"rth"r <lata see

1""",loIH30rn~lc"', Vol.

The exciton phase diagram for 1>ilkx)Il is plotted ill the temperatureconcentration plane in Fif.:. 13. l11c exciton gas is insulating at low pressures At high pressurCs (at the right of the diagram) the exciton gas breaks up into a (:onducting plasma of unpaired electrons and holes. 111C transition from excitOIlS to the pl,lSma is .m example of the Molt transition, Chapter to. Further data are given in Table 2.
RAMAN EFFECC IN CRYSTALS

Raman scattering involves two photons-onc in, onc out-and is onc stcp more complex th.\O the onc photon processes treated elulier in this eh.lpter. In the Raman effect a photon is 1cattercU ineldStically by a crystal, with creation or annihilation of a phonon or magnoll (Fig. 14). The process is identical to thc inelastic scattering of x-rays and it is similar to thc inelastic scattering of Ileutrons by a crystal The selection rules for the first-oroer R.llnan effcct are w=w' fI; k=k'K,
(34)

whcre w, k refer to thc incident photon; w', k' refer to the scalteR.(I photon; and fI, K refer to the phonon created or destroyed in the scattering event. In the sccond-order Ibman effcct. two phonons arc involved in the inelastic scattcring of thc photon. The Raman effcct is made possible by the strain-depcndenee of thc eI{'(.~ tronic polari7~\bility. 10 show this, we suppose that the polari:wbility a :\Ssociated with a phollon mude may bc written as a powcr series in tbe phonon amplitude (I:

(351
If u(t) = 1'0 cos fit and the incident electric field is E(t) induced electric dipole moment has a tumponent

= Eo COl> wt,

then the
(36)

aJEOIi Ocos wI cos fit "" iaJEollo[cOS(w + fI)f + tus(w - fI)t].

-------_..
'" '"
~',.

.......

_~

, , , '"
~

"

,
I
,

, ,
I
0

I I

0
0

\(~,

mtc..- ---.1

,
/

, 0

'"
0

.
10"

Uquld

10"

,,,..

10"

10"

F"lgUre 13 ~ rl~t"IJn "'" l,hotoeM'ltcd cloedl"ON; _I ~ in unstnssed WICOn n.e d~l mows.. ro.. cumple. that with an a,~ COI>CC'I'Itrntion near 1()17 c.m-;o at 15 K, .. fn:-e-aaton gas ...itk satur.. tW-p COOl"C1.tratioo, or 10" ('(ll:>Itsts .. lth a (.-arlabkJ ....lurnc: ofliquid droplets,
",,,-3

..acb ""tk ... den.II) of3 X lo'(..,.,,~3. 111C liquid enti,,:;!.! temperature " aboul2J K. l1w.'On'lk"a! and experimental \... Iue$ IQr the rncl~li"!I.,lalor Ir.l"silion e'leilon. arc ..Iso sno..,n. (From J. P Wolfe.)

ro..

.',k

.',

. ,
cptial p'-'oo it; lIl\oI.,c.od. SuniLlc proct"ISCS

.'

M""'.. "'"Io-Slok, F"'lure 14 Raman scalltrmg of. photoll .. ttb ..mission or al.-..orption of II phOllOOl The prGo.'aS;s aIIM Bnlouin iealtering who...., an phonou;1i ;","ol.cd and pobnlOil scattering "ben ..,

a<'\lUS""

0C'C'Ur

w,th 'nagr>ons (lpIlI .....on).

Thus photons at frequencies w + nand w - n C:ln be emitted. ac'("f"l'!lanie<! by absorption or emission of .l phonon of fre<lucncy n. The photon at w - n is called the Stokes line and that at w + n is the anti-Stokes line. The intensity ofthe Stllh-cS line in\lolves the matrix clement for phonon creation, which is ju~t the matrix clement for the harmonic oscillator. as in Appendix C:
J(w - fi)
Q(

Krill. + 11...lu,,)12

Q(

II"

+I

(31)

where 11" is the initial I)Opulation of phonon mode K.

'"
The anti-Stokes line involves phonon annihilation, proportional to

(
\VHI,

a photon intensity
(38)

If the phonon population is initially in thermal equilibrium at temperatuTC

T, the intensity ratio of the Iwo lines is

lew

+ 0)

l(w -

OJ

(39)

with (nK) given by the Planck distribution function lI[cxp(ft!lIk ll 1J - 1]. We sec that the relative intensity of the anti-Stokes lines vanishes as T _ 0, because here there are no thermal phollons available to be annihilated. Observations on tlle K = 0 optical phonon in silicon arc shown in Hgs. 15
and 16. Silicon has two identical atoms in the primitive cell, and there is no electric dipole moment a~sociated with tile primitive cell in the absence of defomlation by phonons. Hut O'ltl does not vanish for silicon at K "" O. so that we can obscr-ve the mode by first-order Raman scattering of light. The second-order Raman effect arises from the term 02112 in the polari7.ability. Inelastic scattering of light in this order is accompanied by the creation of two phonons, or the absorption of two phonons, or the creation of one and the absorption of another phonon. The phoilons may have different frequencies. TIle intensity distribution in the scattered photon spectrum may be quite complicated if there are several atoms in the primitive cell because of the corresponding number of optical phonon modes. Second-order Haman spectra have been obsclVed and analyzed in numerous crystals. Measurements on gallium phosphide (CaP) are shown in Fig. 17.

Electron SpectrosC01Jy willi X-Rays

The next degree of complexity in optical processes involves a photon in and an electron out of the solid, a~ in Fig. L TIle important techniques of x-ray photocmission from solids (XPS) and ultraviolet photoemission (UPS) have recently been developed. In solid state physics they are lISoo in band structure studies and surface physics, including catalysis and adsorption. XPS and UPS spectra can be compared directly with valence band densities of states D(f:). The specimen is irradiated with highly monochromatized x-rays or ultraviolet photons. "I11e photon is absorbed, with the emission of a photoelectron whose kinetic energy is equal to the energy of the incident photon minus the binding energy of the electron in the solid. The electrons come from a thin layer near the surface, typically 50 A in depth. The resolution of ESCA spectrometer systems is of the order of 0.6 eV, which permits refined studies of band structure, particularly of core electrons. ESCA stands for ek'Ctron spectroscopy for chemical analysis.

JI

Opt""" PrOOl'N and bao",

." ".... .."

~
00'

:;~

..... .... ....

."

.! ,

00'

!
Fit:~ 16 Intcrn;;ly .-alio anliSto!<eo 10 Stokes I",f,"" a (unction oftempcraturc, for thcohscr,'alions of Fig. 15011 the cpcK<lJ n-'" of 'The ~o:d lempentu", d<,:rndence lS in good ~menl WlIh the prediction of Eq. (39): the solid curo-e is a plot d the function eIlp(-ftOlk.TJ.

-500

+500
I

IU.-.n.n.....

"'"

Figu.-e IS Fin;I-onltt Ran'lall spectn utile K - 0 opI;. csI moxie of a siliron <Y}'$I;U ~ at three !eITlpcPlures. The incident pho(on has a wll"denglh ofSI45 ". The optical phonon frequency is eqU2l to the frequency dUll. it depends dighll)' on the temperature. (AflerT. R. Hart, R L. Aggarwal, B Lu.)

,.!>on.

aoo

An

,. " " "

"

II II

" ..

II

h
-fO)

-1'00

-600

-SOU

-.fOO

300

-200

-100

1'UmM, .I"fl, ..."

Figun: 17 Rmv.n spectrum of GaP al20 II: two h'.at pew..-e the first-orner Raman hIleS auocialcd wilh the eocilalion of an 1.0 p/lono<> at 40-,\ em-I and a TO phonon al36l) ern '. All the other peU$ in,"OIve Iwo phonon. (1\l'tN M. V. HoLden and J P. RuSS('lI)

n.e

The valenl'e band ~tl1Jctlire of silver is shown by FiR. 18, ,vith thc zero of energy set at the Fenni 1~'e1. t::lc<:trons in thc first 3 cV below thc Fermi Icvel cmne from the 5s condlldion band. 111C strung peak with slnlcturc below 3 cV is fmlll the 4tl "al('ncc electrons. Excitations are also seen from deeper 1c\'e1s, oftl.'Il accompanied by excitalion of pl:\Smous. For example, in silicon the 2p elcccruTi with a binding energy close to 99.2 eV is obscrvl-'(I in replica at 117 cV with single plasmon excitation and at 134.7 eV with two plasmon excitation. nIl" plasmon energy is 18 eV.

ENJmc)' LOSS OF FAST PAR11CLES IN A soun

So mr we have used photons as probes of the electronic slnldure of solids. \Ve can also use electron beams for the same purpose. 'nle rt.'Sults also involve lhe diek-Clric fundion, Il()\.\' through the imaginary part of UE(w). TIlC (lielcetric function enten as Im{E(w)} into the energy loss by an c1cccromagnetic "'"ave in a wlid, but as -Im{lf(w)} into the energy loss by a charged particle that penetnltes a solid. Consider this difference. TIle gl.'Ocml result from c1cctromaglletic thCOf)' for the power dissipation density by diclectric losses is (CCS)
I @I = 411 E' (dD/elt) ,

(40)

per ullit volume. \\lith a tnmsverse e1cc1romagnctic wa"e Ee-''''' in the crystal, wc have dDttl, =' -jw E{w)Ee- 1w ,. whence the tillle-a\'crage power is

'!J> ~ _l_(ne{"~e-'W'}lk-{-i~w)Ee-jW'}) 4w
~

411

wE!' (E" cos wi -

~ Sill wt)cos wt) = -

(41)

811

tK(w)'i

proportional to t"(w). The tangential component of E is conlinuous a(:ross the boundary of (hc solid. If a particle of ch.trge e ami velocity venteI'!> a crystal, the dielectric displacement is (CCS)
0(',/) ~ -g",d

Ir -

vI

(42)

because by the Poisson equation it is D, and not E, that is related to the free charge. In an isotropic mcdium the Fouril'r component E(w,k) is related to the Fourier (:omponent D(w,k) of V(r,t) by E(w,k) =' D(w,kV~(w,k).

~ V "
o9

Il....,hng Cl'l"'lO'. c"

-nlC time--Rvemgc power dissipation ass<x.;aIL-'d with

nellt is

this Fourier compo-

1 @(w,k} = (l\c{~-I(w.k)D{w.k)e-""'}Rc{-iw D(w,Ic}e-'4) 4r. :::

whence

4~

w0 (w,k)([(:)' ooswt+(~r S;\lwt]t-SinwrJ)

2

9'(w,") = - 8'l'T W

( 1 )-;-

1 t:"'(eu,L:} IJ2(w,k)::: 811' w ItF IJ2(w,k).

(43)

The result is the motivRtion for the introductWJIl of the energy los5 function - hn{l/E(w,kH and it is also a motivation for experiments on energy losseS by last ck'Ctrons ;1\ thin films. Jr the dielectric function is independent of k. the power loss is
~(w)

2 = -- '1T

.r

fu;

hn{UE(w)}\n(kovIw) ,

(44)

where rlko is the maximum possible momentum transfer from the primary partide to an ck"Ctron of the cry1>tal. Figure 19 shows the cx(:cllent ('xpcrimcntal agreement between values of ~(w) deduced from optical reflectivity measurements with ,'alues deduced from electron eneq,')' loss measurements. Table 3 lists the acmllYms ror some or the principal experiml.'fllal methods used in studies or cnerl,')' band structure. particularly of the gross aspects or the structure. -ntcse mcthods diffcr rrom those IIsed in rcrmi SUrf:.K'C studies, Chapter 9.

,,. , , ,
,
",
"~-----

c.

,
,
,
3

'"

,
"

\
"

,
.y--o

" i
"

., "

'"

"

" .,'

'"

r'SUr"'e 19 f"{",) for Cu M>d Au; !he bold Iinft &Al from ~ ~ nlHSU~rl1ffib by J. Daniels, and the otl~ linn "'eJ'e nJcuIalcd from opticallTl<':'aSUremomlli by D. 8eagkhok and L R. Canfield ct .1.

Table 3

Acronym~

of currcnll.-.:perimenlal methods

ror blllld s"lructure studiCJ

AES EL5
ER~1

Auger electron spectroscopy Electron (energ)') loss spectroscopy

Evaporation rate I1lOnitori'lg

ESCA EXAFS
!'OMS 1"1:1.1 lAP INS
IH~[

Electron lipectroscopy Ktf .-he1l'i..,1 lI.'l3l)1'is ExtenJcJ i(-ray absorption fine structure ..1ash (lcsorption mass spectrometry FielJ joniza.tion microscopy Im~ng atOOl ~ Inelastic neutron scattering
1011

rate montloring

I'm
LEED RIIEED
SAr>.t SEM SES
SI~IS

ISS

SLEED

TOMS
TDI

THEED
UPS XPS

Ion scallering spectrometry Isothermal desorptKlIl spectrometry Low energy electron diffiactiOll Renooion high ene'1,'Y electron Jirrrnction Scanning Auger microscopy Scanning elecl.ron microscopy Secondary eled rorl spectroscopy Secondary ion mass spectrometry Spin-polarixed low ellergy electron diffraction l11ennal (!esorplion mms ~peclrometTY Transmi~~ion c1ectron mkroscopy Transmission high energy electrOn diffn1lction Ultra\'iolcl photoelectron SpectTOSCOPY X-ray pllotodectron ~'f.I'Ir()S(.'OPY

SUMMAR\'

Thc KmmersKronig relations connect the real and imaginary parts of a response function:

a (w) "" - - p
7T

Ow

L- r-Ja'(.)

'"

nle colf1plcx refractivc index N(w) "" n(w) + iK(w), where n is the refractive index and K is the extinction coefficient; further, (w) = N'f,w), whence '(w) = n2 - k"Z and ~(w) = 2nK.

ne reflectance at nonna} inddence is

R _ ';:(":"'-,..-7.1).. ' :...+-OK'"

(n + 1)2 + I('l

The energy loss function -lm{UE(w)} gi\'es the energy loss by a charged
particle m<Wing in a solid.

Problems
J. CoUlO1ity cmd' lhe rupon,e [ullction. TIle Krdmers-Kronig relations lIrC consistent

with the principle that an effect not PreceUc ibcause. Consider a delta-function force applied at lime I ~ 0;
F(I) = 6(1) = - I

211" _..

L-

1:- 1...

(lw

whence F.. - 1127T. (a) Sho-v by l\irect integration or by use of the KK relations that the osdllatN roponse function
a(fal) "" (~-

J -

ifalp)

gi\'e~ zero di~p1acement. x{t) "" O. for t < 0 undcr the alxwe force. for 1 < 0 tbe contour integral may he completed by a scmicircle in the IIp[X'r half-planf'. (h) Evalnate x(J) for I > O. Note that a(w) has pol~ at z(~ - fjl)ll2 - Hp. hotll in the lower half-plane.

2. tn.ipoJion .um rule. By comparison of C)'(fal) from (9) &nO from (Ita) in the limit w_ ce. show that the fOllowing ~lIm rule for the oscillator ~trengths must hok!.

3. Rtl1ectiOll at Ilornwl iud/felice. Comitler'tll dt.'CIrom.lgnetic wave in \':'lI,:tlllnl, with

fl('1t1

('OlIlpOllent~

of the fonu

Let the WOlve be incidenl upon a mediulll of dielectric constant E antI permeability
po - 1 IhJt fllls thc h.tlf-!>p.'Ce x> O. Show that the reflectivity coeflidellt roI) as

definlod b)' E{ren) = roI)E(inc) is

giVCIl

by

,,+ iK -1 1'(01)= n+iK+1

where n + iK
1 E .'!-,

with

II

:uKi K. reiI. Show further that the rdle<131lC't' R(oI _(n-IY'+K' ()-(1l+1Y'+k"l

15

"4. COIl/luetici." '"m rule fllltl.It,lercont!ueticity. \\'e write thc el..:c1rioll Wlldllclh'ity .IS 1J'(oI) EO u'(w) + iu"(w), whnc u, rI' are r('ii!. (a) Show by ... KI;lll1crs.Kronig rela tion th.lt
~~I~
(oj

o"(w) ""' -:;; 2

L" u(r) tis .

111is result is usetl in the theory of superconductivity. If at very high rrt.."I!uencies (sllch lU' x-r.ty fretlllC1lCleS) 11"01) is idl"lltical lOr the supenxmduding alld lIormal stales, then we must llave

o''(wl dw

o',(w) dw .

But at frt..-quencies 0 < w < w" within the superconducting energy gap tile real part of the COlKiuctivity of a superClIlKiuctor \'anishes, Ii() that in this region the integral on the left-halKi side is lower by .. 0'""',,. There must be an additiollall."Ollltributionto <r. to IxllanC(' this deficiency. (II) Show that if tI.(w < w.J < o',,(w < woJ, as is observed experimentally, tllen u.(w) C',1I111lWC a delta fimction contribution at w O. and from tile delta function tllcre i~ n ('Qntribution U=(w)'" u.,w~/(oI. The delta function corresponds 10 infinite conductivity at zero frequl'1ICY. (c) By clementolT)' conskleration or the chbsical motion of couduction eledrons .tt very hi,,;h frequencies, ~how that

(ees)

L'" u(w)

d(ol':

mlrf2,n

a rcwlt bund by Ferrell and elo\cr.

5. Dielectric oom'CUl' and 'he remicontluctor energy gap. The elTed 00 ("01) of an
energy g.l1l (ole in a semit:ooductor may be approximated very roughly by sumhtuting Ui(w - we> for .5(",) in the response function (16); that is, we take ("(w) '" (21me21mw}m5(w - w,J. 'nlb i~ C1"ude because it puts all the absorption OIt the gap

frequency. The l:'\dor U2 enten; as KlOIl as we Ulove the delta function lM-ay from the origin, because the integral in the slim rule of Probkm 2 star15 at the origin. Sho.... that tile real p.u1 of the diekctric constant on this model is
('(w) .. 1 + ?'(w:

- JJ,

It folb.V5that the smticdielcctriccoustant is ('(0) = 1 + ..,I,Iw:. widely used as a rule of thumb.

6.. llagen-Ruool$ relation lor irt/rared I?jlectidty index II + iK of ;\ metal for W'f <C 1 15 gi"eu hy

0/ inc/aU.

TIle compleJO n>fmcti\c

(ees)

((w) ("

+ 1"1- 1+ 41ri<.rolw

where Uo is the conductivity iOr static f.ekls. We assume hene that iutraband currents are dominant; interband transitions are neglected. Using the result of Problem 3 for tlte reflection coefficient at nonnal iocKlence, show that

(ees)
l)t'O\'tc.Icd that un" w. n,is i~ the lIagetl-RuLetls rcbtion. For sodium at room ternpernture, (fo "" 2.1 X 10 17 s I in CCS and T 30 3.1 X W-I~ s, as deduced from T '"

of 10 ~m has w - 1.88 X 1011 s-1, so that the lIagcll-Ruben~ n:sult ~houkl apply: Ii - 0.976. 1bt- result calculated from expnimental "a1ues of II and K is 0.987. Hlllt: If Uo .. w, then "t ... l11is ~impllnes the algebra.

uo'nlrtC~. Hadiation

,,/2.

1. Davyrlov Iplittillg 0/ exciton lintll. 'nle Frenkel exciton hand of Fig. 9 is doubled when there are two atoms A. B in a primiti"e ttl!' Extend til(' theory of Eqs. (25) to (29) to a linear <:I)'stal A8.AB.AB.AO. with tmtlsfer integrals T I "tween AB i\OO T2 betvo-een B.A. Find an C1lltation for the two bands as fllnctions oftile wave\ector. TIle splitting het"..een the bands lit k .. 0 is allied the O;I\')'clov splitting.

Refe,.ences
M. Balkans. a)., Opticcll f1rvpn"tia of -.irorIdudort. North-lioUand, 1980.
N B~mt-gen, NorJiNIl,. opIia. Be1jamon. 1963. D. 11.. Shiriey. at., ltttn:m tpf'CI~. Northllo1land, 1972. M. Cardona.. Alodulotil.lrl ",roroow. Jl.cll&mic, 1ll8l. C. D. Jeffrioo:I and L. \', Kek!)v.. ~li. lIn:mItol, droplet.. III . .,loonduCIOn. North-HoI1and,

C. Phillips, ~Fundamenb1 optical ,pert... eX sOOth. ~ SoIll1lCate ('h)'W('li 18, 56 (1966). R. loudon, Quall'um t1KcTy of /;gJli, 2nd ed., om.nI, 1963. E. I. &shbo and M. D. S~, eds. &dt0l'll. North-Ilol.bond, 1982. D, ~,dds aOO T. C. CoIIu,", Erotoru. ,lid,. pl"Opn'Ha.nd ..... Academic, 1961. K.. 0.0. .u:iIotlI, Springer. 1979. W. M. Yen and P. M. Sdzrr, edli. l../UIIr.tpfr'OK'<lflr of sclIdr. Spnnger, 1966; ,-01.. II. 1969. Y. R. Sloen, I"rfllCipla of non/itl(l,. ofll'ic:I, Wile)'. 1984. F. BtO\Oo.... ~mlr.il.vioId ~' of dido with use (/ liYnchrotron radialion, ~ SoIKl. stale plJ)'liics 29. I (1974). P. O. l\'iluo'ln. ~Oplical propoertieli of n>daJs aud aIIo)'li, Solid ~c ph)'liks 29, 139 (1974). T. M. Rice, MEJectron-hoie liquid in semirondueton: theotelical aspeclli. ~ Solid slllle pt.)'liics, 32, I (19Ti). M, Cardona. ed. Liy)It -'teriRf: &r -alidl. 3 ,'Ok., Srwiuge.., 196:2-83.

J.

'''''.

c.

c.

12
Superconductivity
EXPt-:RlMENTAL SURVEl' 33:S Occurrence of ~Ix-rronductivily 337 Dcstr~lion of superconductivity by magnetic foelds 338 Meissner eITect 338 Heal capacity 342 Energy 'lip 344 Microwave and infrared properties 345 Isotope effect 346
THEORETICAL SURVEY Tllcrmooynamics of the superoonducting transition London equation 346 347
349
~2

Coherence length Des theory of superoonclucthity ncs ground stale

F1Wl quanliw.tion in a superoonducling ring l>ul"lltioo of penbtent culTcl.Is Tn>e U superconductors "ortc~ slale

3S4
~

3:l6

359 360 361

362

Esl;matioo of lid and II", Single particle tunneling Josephson superronductor tunneling Dc Josephwn effect Ac Josephson effect MacrO!iCOVic quantum intcrrerellce HJCHn:MPI:..'RATURE SUPERCONDUCTORS CriticMl f.ekls and critical currents
H311 number Fullc,'cncs

364 300 366

368 369

371
372
373

373 374

SUMMARY

NOTATION: In tilis eMpler B. dcnQtcs the applied magnetic.- lick!. Ln the CGS s)'Slem the critical value R"" of the applied field will be denoted by the ~yrnbol H" in lIca)rtIance with the custom of workers in supenxmoucti\'ity. Values of B.., lire gi"ell in gauss in CC5 ullits and ill te~1as in 51 uuits. with IT - I~ C. In 51 we have B.., = 1Ao1i.,.

PROBLE~IS

375

I. ~Iagl:lclic field penetr.J1iCWl in

2. 3. 4. 5. 6. 1.

II plate Critical field of thin films T"o-nuid mood of II superconductor Structure of a "oncx London penetration depth Diffraction eff~t of JQ!;eph)l)n junc:tion Meissner eITed in sllilerc

37' 37.
376

37.
377 377 377

REFEREro;CES

377

APP..: NDlCES RELEVANT TO SUPEIICONDUCTIVITY JI Cooper Pairs 656 658 1 Cinmurg-undau Equation J Electron-Phonon Collisions 662

< ..
./

,.
d

,/

,
' ,;,

,,, ,
.'r.
',~

, ,
\

'emf'C'nhme. llus Pot by

Fipore l Rel~ In ohms oIa ~~ 01 mertury \'ftWS ahso!, I(.metlingh O'lDeS matloed lhe d,,,onry ol supenv"dudwily

CHAPTER 12: SUPERCONDUCOVITY

The elec:triC'o1l resistivity of HUlny metals and o1l1o)'s drops suddenly to zero

"11(.'1\ the specimen is cooled to a sufficientl)' low temperature, often a temperature in the licluid hcliUln range. This phenomenon, called supcrronducthity, ....'as obser.,ed first b)' Kamerlingh Onnes' in Lcidcn in 191 L. three years after he first liquified helium. At a critioll temperature Tc the spl.-cimen undergoes a phase transition from a state of normal ek'ctrical resistivity to a supcrcom:lucting state, Fig. 1. Superconductivity is now ,er)' well U1Klerstood. It is a field with mall)' prachcal and theoretical aspects. The length of this chapter ,U1d the rclc,-ant appendkes reflect the richness and subtleties of the fiekl.
EXPERl),IENTAL SURVEY

In the sllperconducting state the dc electrical resistivity is zero. or so close to zero that persistent electrical currents have been observed to flew.' without attenuation in supercolldueting rings for more than a year, until at last the experimentalist wearied of tile experiment. The det'Ry of supen;urrents in a solenoid mlS studied by File and Mills! USitlg precision nnclear magnetic rcsonam:c methods to measure the magnetic field associah.'1.1 with the supcrcurrellt. 111ey condllClt.'(lthat the decay time of the supercurrcnt is not less thall 100,000 years. We e~timate the ,It.'Cay tillie below. In some supen."ollduetillg materials, particularly those used for SUpcrcoll<Iueting magnets, finite decay times are obscrv(.'(1 bC(."AuSC of an irreversible r<.'(!istribution of magnetic !lux in the material. The magnetic properties exhibit<.xl by super<:onductors arc as dramatic as their ck'drh... "l.! properties. The magnetic properties <."Annol be a<''COUllted for by Ihe assumption that a superconductor is a normal <''Ollduclor with zero ekctrical resiSlivity.
'H. K:unerlingh Onlles, Abd. van \\'elenschaAw.n (Am~lenJam) 14. 113, 818 (1911). '-I1,e vatue of Ihe me........ry res;~lance u~ was 172. 7 oI,ms in ,he "quid ~'(ItI(htlon 31 0"<;; t'xlrapolahon from lhe melting point 10 O"C by means uf 'he lempenture coefficielll of solid men:ury Jtj\es reistanrf: ron-espondi"g 10 lhis of 39.1 ohms in dlt: solid stale.....1 4.3 " tlll' hd , .."k 10 0 O&l ohms, thai IS. 100 0021 limes the resistance "t.icl, the wild IT'IC1"nlry ,,'OU1d lla"" at O"'C. At:'l K I~ resl$bnCe "'0.5 found 10 have fallen below 3 X 10 ooml, Ihat I' to (lrl(' Icnm,UlOClth olll... _1II..e which It ~k1 ha'-eal O"C. As the lemperature sankfurtherlo I 5 I( this '1Ilue remained Ihe ul'PCr !unit oflhe res~.R Historical rd'cnncei are Ill,..,n by C. J. Coc1er. Rt. ... Mod 1'1.)5. J6, I (l964.). -J FiIe.ld R. .I.n.. "")'5. Rev_ Lett. 10.93 (1963).

B'
0.026

Table 1 SUllercomluctivily parameter. oi tiIC demenls

~
B C
_J..;.-

An asterisk denotes an clement superconducting only in thin films or under high pres~ure in a crystal modification not normally stable. Data courtesy of " B. T. Matthias, revised hy T, Geblllle.

F
.0

N,

M.
Transition temperature in K Critical magnetic field at absolute zero in gauss (10 ' tesla)

'P'

- A,

N,

AI
lX40 i05-

Si

..
K C,
5,

S'

CI

- B, K'
~

Ti
0.39

. 5.38
1420

C,'

M,

F,

T,

Co

Ni

C,

Z,

100

Rb
n

5,

V'

Z,
0.546
47

Nb
9.SO 19SO
To

Mo

R,
0.51
70

Rh
.0003
.049 I, 0.14

Pd

-0.92- 1.77 95 l410

---- -- ....
-~

Cs'

B,'

la fcc

HI
0.12

.,

-A.

G'

G,'

0.875 ~1.09f 51 53

- - - Sn (..) Sh'
309

Cd
30

I,
393
(~)

056" 340> ~'m

Hg

Os
0.655 65

Pt
~ .~

6~OO

4.483

0.012
1.07

1.4

.
Gd

A,

1100

830

198

19

Tb Bk

TI
2.3.9~

Pb
).193 S03

4.15). 412
Dy

171

F,

R'

A'

"C,'

- -- ., - - .- T,'
1

As'

5,'

X,

Bi

Po

At

-~
~

"
Th

J~
1,62

- - -- -- -- - - ........ - -:.. - i-"'-

Nd Pm Sm
.

-P,
.

"

P,

U(a)

Np

P,

-- - -Am Cm CI
~

- -- "
~

Ho

~ ~ ~

~ ~.

"

Tm

Vb

L,
~

- Md

Fm

--

No

L'

fi&\'.... 2 Mriwler effec.-I in a lupnnmduding sphere rookd in. constant applied ~ rock!, on pusing below tho: tran~ition temperatu.... the htlel of induction B are ejected from the Iphere.
It is an experimental fact that a bulk superconductor in a weak magnetic field will act as a perfect dianlaf.,rnet, with zero magnetic induction in the interior. When a specimen is placed in a ma~nelic field and is then cooled through the transition temperature for superconductivity. the magnetic Oux originally present is ejected from the specimen. TIlis is called the Mewner effect. TIle sec:luence of events is shown in Fig. 2. The unique magnetic properties of superconductors are central to the characterization of the superconducting state. The superconducting state is an ordered slate of the conduction electrons of the metal. TIle order is in the IOnnatioll of loosely associated pairs of electrons. The electrons are ordered at temperatures below the transition temperature, and they arc disordert.-d above the transition temperature. The nature and origin oftbe ordering was explained by Bardecn, Cooper, and Schrieffer. 3 In the present chapter we develop as far as we can in an elementary way the physics of the supcrconducting state. \Ve shall also discuss the hasic physics of the materials used for superconducting magnets, but not their tcchnology. Appendit:es I [ and I give deeper treatments of the S11pefCOndueling state.

Occurrence of SllpercOJlductivity Superoonductivity OCC\lrs in many metallic elements of the periodic systelll and also ill alloys. interlllctallic compounds, and doped semiconductorI'. The range of transition temperatures best confimled at present extends from 90.0 K for the oompound YBa:,CU306.9 to below 0.001 K for the clement nh. Several f-band superconductOl'S, also known as "exotic superconductors," arc listed in Chapter 6. Several materials become sllperconductinp; only under high pressure; for e.umple. Si has a superronducting form at 165 kbar. with T.... 8.3 K. 11lC elements known to be supercondu<:ting are displayed in Table I, for zero pressure.

338

will every nonmagnetic metallic element become a superconductor al sufficiently low temperatures? We do 110t know. In experimentaJ searches for
superconductors with Illtralow transition temperatures it is important to eliminate from the specimen even trace quantities of foreign paromagnetic elements, because they C311lowcr the transition temperature severely. One part of Fe in 104 will de:.troy the superconductivity of Mo, which when pure has Tc = 0.92 K; and 1 at. percent of gadolinium lowers the transition temperature of lanthanum frOm 5.6 K to 0.6 K. Nonmagnetic impurities have no very marked effect on the transition temperature. The transition temperatures of a numher

of inIeresting snpcrconductillg compounds are listed in Table 2. Several organic colllpOunds show superconductivity at fairly low temperatures.

Destruction

0/ SlIpercOflductivity by Magnetic Fields

A sufficiently strong magnetic field will destroy superconductivity. The threshold or critical value of the applied magnetic field for the destruction of superconductivity is denoted by He('f) and is a function of the temperature. At the critical temperature the critical field is zero: He(Te } = O. 11Je variation of the critical field with temperature for several slJpercomlucting elements is shown in Fig. 3. The threshold curves separate the superconducting state in the lower left of the figure from the nonnal state in the upper right. Note: We ~hould dcnote the critical value of the applied magnetic field as RQC' but this is nOI common pl1l<:tice among workers in superconductivity. In the CGS system we shall always understand that lIe e Bae and in the 51 we have He!!!l BocJJLo. The symbol Bg denotes the applied magnetic field.
Mei.~sner

Effect

Meissner and Ochsenfeld (1933) found that if a superconductor is cooled in a magnetic field to below the lransition temperature. then at the transition the lines of induction B are pushed out (Fig. 2). TIle Meissner effect shows that a bulk superconductor behaves a~ if inside the specimen B = O.
Table 2 Superconductivity of selected compowlds

T..
Componnd
in K

T..
CompolJlId
in K

Nb:JSn
N~Ge

18.05
23.2

V3Ga V3Si
YUa2CUaO~9

Nb,A1 NbN ,,"C,O

17_5
16.0 19.2

Rb.:2CsCw

""'"

16.5 17.1 900 31.3 10.4

.oooj---+--',,*:---+---j

"

HfI;:ure 3 ExperinM,'nlru th",shold ""n ...5 of the CtitiNlllcklll..(D ''''nus tcmtlCrnturc for ...".,tIIJ sUJI"reOndUdun. A .pcdm':11 is suprreonductong IX'klw the ~"\l""e and nona.al .t.v. ... the ""nc.

We obtain a particularly useful (or1l1 of this result if we limit ourselves to long thin specimens ..'lith long axis p.valld to B.. ; 00'0'1 the demagnetizing field contribution (see Chapter 13) 10 B \\ill be negligible. whence;4

(ees)
(SI)

B=B.+47tM=O;
B=B.+Jlol\f:Q;

0'

M 1 --= - - 4w B. M 1 = -- = B. 1'0

(1)

0'

-.<If

The result B = 0 cannot be derived from the characterization of a )upc:rconductor as a medium of zero resistivity. From Ohm's law, E = pj, we see that if the resistivity p goes to zero while j is held finite. then E IIllist be zero. Bya Maxwell equation dBldt is proportional to curl E. so that zero resisti..ily implies dB/tit::::;; o. TIlis argument is not entirely tl~lnsparcnl. but the result predicts that the flux through the metal cannot change on cooling through the transition. "Ole Meissner effect contradicts this result and suggests that perfect diamagnetism is an essential property of the superconducting ~tate. We expect another difference between a superconductor and a perfect conductor, defined as a conductor in which the electrons h.we <tn infinite mean free path. When the problem is solved in detail. it turns out that a perfCl.1 oonduetor when placed in a magnetic field cannot pnxluce a permanent eddy current screen: the field will penetrate ahout 1 em in an hour. oS
Olamagnet...". the magnctiution .\1, and tile 1l'\,lftTlcl;., ~uKqJtlbility arc definetl in Chap:cr 14. The magnitude oCt"" appRIfi1t diamagnetic: suscq>l,bolityoCbulk .upawndurton iI '''Cry mum brger dllm in typtcaI dia~t;" subltances. In (I). 1<1 IS the ma.gn<':lil.lltion equ"a!cnt to Ihe SlIpCrco"duding ~rrcnts ill the specimen. sA. 8 Pipl'anl. I' w,,~ of a",dlJCflon ,Itrom, Cordon llnd Bread!. 191.

n.
Tn'" I
<

, , , ., " , ,

, ,
,
II...

If,.

Awl..'" """CI"d'" lid.! 8 . l'iIlure" (a) "lll~lIctization \"eI'!>US apPIird magnetic flekl CUI II bulk liUpo.."'Oflductor .,,,hlui/intl. compkte Mei"ucr dfed (pe&rt diamagnt1lsm). A liupereooduc.1.....ilh mrs bella'v is calk-d II I>ve I liupo."n:'OIl(h.cror. AI.-.., die critical 6dd II, the:: speamcn is II non1lal conductor" and tho, m~nctiurion is 100 small to I>e M'eJl On 1"1$ sao'" Not.. thal ntimn; " ...M is pIottoo QII the ,..,rtical 1iC'llle. the nq:aU"c value of M (.'OI'WIponds 10 d':ImII~t;sm. (b) Supcmmducting mllfltldizalion "","vI' of II In'" II .u..... r(.'()rlll,,~1or. The fl,,( Marls 10 1""H:lrnlc ti,e speciow" III II lleld II" luwcr than the thermoxh'namic ttilical Geld If... TI1c spedmcn i. in II ''<>rIC) dill<' IJetween II,. and lid. and it has supcft.'Ol>ducting .,lc-d.ric:al pl"OplYtlCS lip 10 II... AllO'.., lid II,.. lilJeCirnm 's II I'>lIf'flW cuoductol'" in o:YC'l')' rcspc:d, c:>:cq)l for poUiLk surfac.e eITu. For g;''l::I'\ 1/. the al'", under the ,~fLJh(J11 ain't' is the pme lOr II type II supen'OOductor as for a type I (CCS units in all p"rb or this IlI;Uno.)

.,

If.

nlC fll,ognctization curve expected for a superconductor under the condiHons of the Mcissner-Och~nfcldexperiment is sketched in Fig. 4a. 1l1is applies (Juantitath'ely to a spedmen in the (01"111 of a long solid cylinder placed in a longitudinal magnetic field. Pure spcdmcns of many materials cxhibit this behavior; they arc cAlled type J supcreonductors or, fonnerly, soft superconductors. TIle "alues o( 1-1e- arc a1wd)'SIOO low (or t}l>C I supcroonductors to have any uscful tedmical appli<:lll1on in coils for- superconclu,,:ting magncts. Other materials e;\hibit a maf.,'lletization cun e o( the form of Fi~. 4b and are known .IS type JI superconductors. 'nlCY tend to be alloys ('L~ in Fig. 5a) or transition metals with high values of the electrical TCsistl"it)' in the normal sl.1.te: that is, the electronic mean free path in the nonnal state is short. We sh.1.11 see later wb)' the mean free path is inHlhl,.'<l in the "magneti7.ntion" of sllpcrcondl1(.101"5. Type II superoolldul.'tors havc l>upcrconduc..tiug e1cctri<:.\\ properties up to a field denoted by I-1 d . I3ctween the lower critical field 1-1<1 and the upper critical field 1/<.'2 the flux density n ~ 0 and the MeiUllcr erJCct is said to be incomplete. 111e value of Jld. may be 100 times or morc hiWlCr (Fig. 5il) than the value o( the lTitiml field Il r calculated from the therlllOdynamks of the transitioll. In the region hetween lie. and lid the superconductor is threaded by flux lines and is said to be in tllC \ortex state A field H d of 410 kG (41 t~las) has been attdilled in an alloy of Nb. AI, and Ge at the boiling point ofhehllm. and 540 kC (54 tcs/as) has been rCfX1rtecl for l'b~IOt;SIl'

12 S"p"rmduC/i"i,y

341

,,

, ""

'"

Nb-Ti

,
I;ig"re sb

Ten']"'.... t""'. I:. St..:mgcl' magn~'fic fields than any now ('OOtcmplat~'<I III JX"'cli{,lI s"lx'n:onduding

"

"

"

"

dc' ice. are "ililin the capal.>;lily of ceria;" Type II materials. '11lesc material. l':UlIK,l1 he explOIted, howcvt'r, until their critical current density elln he rdlSt,d and un!ilthe)' c<U1 1)( r.. l>ri<;atro as finely djvi<kd t<:m<!lldors. (MolgTlctic field, cl'more than aboul 20 le.la. can L... gencrdtcd onl)" in pulws,

and so portion. of the curves sllOWl1 as I.>rokcn l,nc' were mea,ured in that way.)

""

1
E
10

,,!f:::"'_,,~,~-~,,"~.,;::::-_",;,---,,:..,----,C""--"';2-Enltop)' S of ahnllinum in the IlOITIlol.l and ... ~K1uet;ng states as a rul'lCtioo of tloe tell1peralure. l1Ko entropy i~ \ov.u- in the supel"C'Olldllctin~ st<tle be.:",uSt' the electrons a~ tnOR' OJdered hen.- than in tbe normal stalo:. AI any temperature below the m'raJ lell1perlllun: T< lfM.~mcn ~ lie put in the normal st_loy appoo.tion ua 'I\.glletic f..,1et stronger lhan lho: crilical

.'gun: 6

T~onpent"'~.

......

Commercial solenoids wound ,\;th a hard superconductor produce high steady fields over 100 kG. A "hard superconductor" is a trlle II superconductor with a large magnetic hysteresis, usually induced by mechanical treatment.
Such materials have an important medical application in magnetic resonance imaging (MHI).
f1ellt CllJXJcity

In all superconductors the entropy decreases markedly on cooling below the critiall temperature Tc:' Measurements for 1I1uminum arc plotted in Fig. 6. TIle decrease in entropy between the normal state and the 5uperconclucting statc tclls us that the supercooducting state is morc ordered than thc nonnal sbte. for the entropy is a measure ofthe disordcrofa llystem. Some or all ofthe electrons thennally excited in the normal state are ordered in the superronducting state. The change in entropy is small, in aluminum of the order of 10-4 k B per atom. 111C sm.'ll] entropy change mu:st mean that only a Sln..-ill fraction (of the order of 10- 4 ) of t.he conduction c1cctron:s participate in t.he transition to the ordered superconducting state. The free energies of nonnal and superronducting states are compared in Fig. 7, 'l11e heat capadty of g.ll!ium is plotted in Fig. 8: (a) compares the nonna! and superconducting states; (b) shows that lite c1ectrc' ',;: <-'Ontriblltion to the he'dt capacity in the sllpcrromillding state is an cxponc.. .al form with an argu.4

III

S"".,rcorulucti"if"

,
-OJ

-" ... ,
-o.~

1-0.5 , ,

'.

"

p -tti ! -0.8

,.,

-09

-" -"
-!>

,
T.... ,[><....

..,
t""'. K

"
n.e

Figure 7 E~IXrinIC"I(l1 \'alut"S mtl,.. fll.'t' cnc'l,'Y.' a r'lndlon oflCmpcrnlu,c for nluminum in the supcn:oududing slate ~rltl in th", nonnal s",te. Below the Ir.msition tcmpcrdt"re or, - 1.180 K the frn' cnt:~ ;510...."r in the .uvcroonductinllstalc. The Iwo CUrvCll n'Cf}1;C "tlhe Inns,lilm lemperaIurl', so Ih:w the plwse transition is ""Jud order (there Is no latenl heat of Inmillon at T<~ all'V'e 1'.15 tn<:a'iur! In:zero magnetic 6Md, and F", is mc~red in a n>agnetic:" llckl."ff"lCk-nllo ",.t the specimen in die lIornw lMtc. (Courtny of N. E. PIlilU.,..)

J.~-

.1I.-200C
_II. _0

c.o..

,
:.l

1.0-

iL--:..:.'--;-...... :;:.::::......~--.
,~j

: ICIT_05ll6+00i568:r"

A~I:.

, .,I
I

1
i I
I

."
1.5

o,ooIL-,;----1,-;---I;--"'-/; I 2:l 5 6
~"
(b)

~~
(a)

i"igurc 8 (a) The IIt'lill capllC\ty <;If ~dl1ium In t1,,~ ",mna) and sUllcrrond"cting slalcs. Th... normal ..turcs) nuclear S!ale (...tlich is rt.':ilon..d I,y. 200 C fleld) h:15 ek'dfOnk, Iall;tt, and (at low lcmper ,!,lIl(lrupok contributions. In (b)lhe electronic palt C", of the hc-at capacity III Iht' stlpt'rcondoctmg Slate Is pIotkd on a ~ ~~ T)T, II,,: ""ponenlial depcnderK"e on 1fT is t""dent_ II.,,,, ., - 0 60 mJ mol ' l (Afte.- N. E. pt,iUips.)

11.'"

llIent proportional to -liT, suggesth'c of e\:citation of ek,,- liS across an Cn~ ergy gap. An energy gap (Fig. 9) is a eharaderistic. but not universal. feature of the supcrt.'onclueting state. TIle ~p is accounted for by the Ilardccn-CooperSchneffer (DCS) theory of supcrconclucthit)' (sec Appendix H).

Energy Gap
The energy gap of supcrroncluetors is of entircl)' different origin and nature than the energy gap of insulalors. In an insulator the energy gap is Ccl.uscd bv the elcdronlalticc inleractio:l. Ch..lpter 7. This interaction ties the electrons to the lattice. In a superconductor Ihe important interdcHon is the electron-electron intenlCtion which orders the electrons in k space with respect to the Fernli gas of electrOns. J1le argument of the CKponential f.dor in the electronic heat (''apacity of a superconductor is found to be -E,.I2ksTand not -E~/ksT. This h<1S been lcaml from comparison \vith optical and e1edron tunneling detenninations of lhe ~p E". Values of the gap in sC\'eraJ supcrcondlldOrs arc given in Table 3. The transition in zero magnetic field from the supcrcondueting state to the nO",lal state is observed to be a second-oruer phase transition. At a sccondorder transition there is no latent heat. bul there is a discontinuity in the heat e<lpacity, evident in Fig. Sa. Further, the eTlergy gap decreases continuously to zero as the temperature is increased 10 the transition lempcrnturc T~, as in Fig. lO. A first-ordcr tmnsitioTl would be characterized by a hltent heat and by a dismntinuity in the energy gap.

ToLle 3

Energy gaps in supcrcondudon, III T = 0

f,IO) 10

AI
3.4 3.3

Si

In

eV.

f~O)lk.lJr~.

S<

r;

C,

M.

Fo

Co

Ni

Cu

2.
2.4 3.2

G.
33 3.'

Go

16. 3.4
Y

2,

Nb

Mo
2.7 34
W

T,

Ru

Rh

Pd

Ag

Cd

I.
10.5 3.

SIl (..)

30.'
3.80

I.,
3.2

11.5 3.'

l.
19.

HI

T.
14. 3.60

Ro

0,

I,

PI

Au

Hg (.)
16.5 4.

TI
73' 3.57

Pb
27.3 4.38

3.7

J2

S .. p<!r'cOllducli"il~

"

Ncnn.I

Supcn'Orlduotoo-

Figl.,ril' 9 (a) Conduction IIllIKI in the nonnal ~tate, (b) encll,'Y gap at the FermI le"e1 in Ihe Sllp'-'rroIKhJding state. Ek'C.'1rons in excitL-d states atJovc I...... 8"P beha"e as normal ek'C.'1rons in rf fiel~: they cause "",btar>c.-e, III <Ie Ihey arc dlO'led out by the wpe,cooductinK dt.'drOlll. The gap f. is eaaggcr.lted in the flg\lK': t)lJia/1y . - IO~' ft.

" -- .,
Jlc:duaod \'a!un d ...... obseo-o:d energy gap F.J,TY,IP) as a function of the ....... duocd tcmpcr.lturc Trr .ner li:M"scnd and SuttO!>. 't1,c wlid curve Ili drnwn fur the BCS
FlIlI:urll' 10
p

" '" '\[]

8CS """.,. Tin o T""IaI"m
N""~wn

I\"

, , ,.

0.2

03

O,l'i

06

0.7

08

0.11

1.0

Iht."Or)'.

Tff,

Microwave and Illfrored Propertie.

"Inc existence of an energy ~p means that photons of energy Jess than the gap energy are not absnrbcd. Nearly all tbe photons incident are reflected as for any metal because of the impedance mismatch at the boundary between \"lll-UUlIl and metal, but for a very tbin (-20 A.) mill more photons arc transmitted in the supcrconducting state than in the nonnal state. For photon energies less th:ul the energy gap, the resistivity of a superconductor vanishes at absolute zero. At T <t '1'.. the I'csistance in the SUPCf<.'Qlldu(;tin~ state has a sharp threshold at thc gap energy, Photons or lower energy see a resistanceless surface. Photons of highel' energy see a resistance that

>"
approaches that of the nonnal slate because such photons cause transitions to unoccupied nonna] energy le\'els above the gap. As the temperature is increased not only does the gap decrease in ellerg}'. but the resisti\<ity for photons ",;Ih energy bdow the gap energy no longer vanishes, except at zero frequency. AI zero frequency the supercollducting electrons short-circuit any nonllal electrons that ha\"c been thcnnally excited above the gap. At finite frequencies the inertia of the supcrconducting ek'CIrons prevents them from completely scrCI..>ning the ck-ctric field, so that thermally cxl'ill... >d 11011na\ electrons now call absorb cncrrw (problem 3).

bOIOfJe Effect
It has been observed that the critical temperature of superconductors \'aries witll isotopic mass. In IIlCfCUI)' T~ \'aries [TOm 4.185 K to 4.146 K as the average atomic mass M varies &om 199.5 to 203.4 atomic mASS units. 11lC transition temperature changes smoothly when we mix different isotopes of the same clement. The experimental resulls \\ithin each series of isotopes may be fitted by a relation of the fonn

More:::: <:onstant .

(2)

Observe<] values of a arc gin:n in '1:1ble 4. From :he dependence of T e on the isotopic mass we leam thut lutticc vibrations and hence electron-lattice interactions are deeply involved in superconductidty. '111is was a fulldamelltal discovery: there is no other Tl~ason for the supcrcondueting transition temperature to depend on the nwnber of neutrons in the nucleus. 111e original BCS model gave the resull To; ex 8Dro-1'" ex M- m" so that ex :::: i in (2). but the inclusion of coulomb interactions between the electrons changes the relution. Nothillg is sacl'l..>d about ex :::: t. 111e absence of an isotope effect in Ru and Zr ha.~ been accountl..'C!. for in terms of the cIL'Ctron band stntdUTC of these nll..'tuls,

THEORI:.""TICAL SURVEY

A theoretical understanding of the phenomena 3S5OC'iated with supcrconductkity has been reached in SC\~ral ways. Certain results follow directly from thermodynamics, ~lany important results can be describt.>d by phenomenological equations: the London t.'(]uations and the Landau-Ginzburg equations (AP'" pendix 1). A successful quantum tll(.'ory of superconducti\'ity was gi\'en by Hardeen, Cooper, und Schrieffcr, and has pro\ idl..'C!. the basis for subst.'<]uent work. Josephson and Anderson discoveroo the importance of the phase of the super conducting W3\'efunction.

It SIl~WfiriIV

3H

Table 4 Isotope effect in supel'C.'Ollductors

Experimental \'ltlm~s of (I in Al"T~ "" coruitalll, where M Sub.tance
'$

the isotopic man.

O.~:!:

S"mtarvx

O.OO:!: 0.15 ~ 0.33 0.08;t O.OO;t 0.05 0.05 0.02 0.05

lOn Cd

S"
IIg Pb

0,32:!: 0047 :!: 0.50:!: 0.49:!:

0.05 0.07 0.02 0.03 0.02

0>

"" M.
N~n

lO,

l'llenfl()(IYflmllic~ of 11le

Sllpercollflllclillg Trausitioll

lllc tnmsition bctwccn the nornM\ and supcrcondllcting states is thcrmodynalllk.;ally rC\'crsiblc, just :'IS the transition bctwccn liquid and mpor phases of a substance is rcvcrsible. Thus wc I1My apl,ly thcnnoo}1mmics to the transition, and we thcreby obtain an expression for thc entropy differcnce betW(.'C1l nonnal and supercondllcting slates in tenus of thc critical field curvc He versus T. TIlis is analQKous to the v<lpor prC5Sure L"quation for the liquid.gas coexistence CUr\]! (TP, Chapter 10). \\'e treat a type I superconductor with a complete ~Iei5Sner dfL'ct, so that 8 = 0 insidc the superconductor. We shall 5(..'e that the critical field He is a quantitative measurc of the free cnergy difference between thc slIpcrconduct. ing and normal St.ltCS at constant tcml'cl-aturc. TIle symbollf r will a1wars refer to a bulk specimen, nC\'cr to a thin film. For type II Slllx:rconductors. J1~ is umlel"~tood to bc the thermodynamic critiClI field rclated to the stnbilil'..ation frL'C ener'b'Y. '111c stabilization free ene'"'b'Y of the superconductillg state with resf*Ct to the normal state can be determined by calorimetric or magnctic measure"(l as a function ments. In the calOl'imetrie method the heat capacity is measufI.. or temperature for the superconductor and for the normal condul'lor. which means the supcrconductor in a magnetic field larger than Hr. From the difference or the heal capacities we can compute the frL'C energy differencc. which is thc stabilization free energy or the superconducting state. III the magnetic method the stabilil'.ation rree encrgy is round rrom thc value or the applied magnctic field that will destroy the superconducting state. at constant temperature. The argument follows. Consider the work done (Fig. 11) on a superconductor when it is brought ...,crsibly at constant temper.lture from a position at infinity (wherc the applied flCld is zcro) to a position r in the field or a pennancnt magnet:
\V= -

L"" 1\1. dB...

(3)

'"

C
~I

Supcre<l"dud(lr pi"""
No<mal pha<e (coe>:,>I1l111 in

<><1"il;1><,,'n 1)

Ca} A superconductor in which the Meissner elJect is oompkle h"" B - O. as if the magnetization wer.. hi .. -Bj47r, in CCS 1111,15_ (b) w}..." the appliftlllcW reaches the "a1ue B"". Ih(' normal state can CO('xist in .,quilihrium with the superoonductmg state. I" e<:><:><is!encc the fr~-i ellcrb')' densities are equal: "',AT, 8....) = Fs(T, 8...,). Figure II

per unit volume of specimen. This work appears in the energy of the magnetic field. l11c thermodynamic identity for the process is
(fF

= -1\01' dB"

(4)

as in 1P, Chapter 8. For a superconductor with 1"1 rclateclto B" by (I) we have
(CGS)

I df\ = 41r B" dB" ;

(5)

(SI)

dF s = - B" dB"
The increase in the free cncrb'Y density of the superconductor is

...

(CGS)

(6)

(SI)
on being brought from a position where the applil..'<1 field is zero to a position

where the applied field is R".

J2

SUJ'l'('OOlducticilll

34'

Now consi a nonnal nonmagnetic metal. Ifwe ncglect the small suscep>rgy of the normal tibility; a metal in the nonna] state. then M :: 0 and the ent.. metal is independent of field. At the critical field we h"vc

or

F....{D....) :: F ....(0) .

(7)

or the supcrcond\lcting state at absolute zero.

(ces)
(51)

The results (6) and (7) are all we need todetennine the stabilization energy At the critical value D"" of the applied magnetic field the energies are t.'<lual in the normal and supt.>rconducting statt..'S:
(8)

In SI units lie - D...JP-ih whereas in CCS units He - nne' 111e specimen is stable in either state when the applit.'<1 field is C<lual to the critical field. Now by (7) it follows that

(ces)

!:J.F. F,,{O) - Fs(O) ::

n:J8TT

(9)

where !:J.1" is the stabilization free t..nergy density of the superconducting state. For aluminum. BM at absolute zero is HE gauss. so that at absolute zero!:J.F :: (lOSfI8TT:: 439 erg cm- 3 in excellent agreement with the result of thennal llll.. ' asuremcnts. 430 erg cm- 3 . At a finite temperature the normal and supcn:onducting phases arc in equilibrium when the magnetic field is such that their frcc energies f:: (I 1'S are equal. "111e frcc energies the two ph.l"l.."S arc sketclu...d in Fig. 12 as a f.ulction of the magnetic field. Experimental curves of the frcc enl."Tgies of the two phases for aluminum are slKl\\ll in Fig. 7. lk."C8use the slopes dFldT arc equal at the transition temrwrature. there is no latent heat at T c-

or

London Equation We saw that the Meissner effect implies a magnetic susceptibility X = -1/47f ill CCS ill the supcreond\lcting state or. in SI, X "" -I. <111is sweeping assumption tends to cut off further discussion, and it does not account for the flux penetration obscrvt.'(\ in thin films. Can we modify a constitutive equation of electrodynamics (such as Ohm's law) in some way to obtain the Meissner

GThi:s is an IMkqual'" U'iumption lOr l}~ I su~ In l)"IlI' II 5U~lidOIs in high ",,"Ids lhe- many in spin paramagnetism of tllf" rondud:ion dK1rons b."f'n ,..., ~ J 1M nur,na1 phaw signirlCalll1). tn SUInt", but nut an, t)"j>C II su~rWllducton d"" UPP""" critical field i>; Hmlled by this ,,,IT<<t. Clogstoll has sUi:&'"Sted thatll,.(mu} .. 18,400 T,. ",t...,re 11 4 b; in gIlU~ and Tc in K. s..e A. M. ClogstOll. P\,~s. ~. Lo::Il. 9. 266 (1962): B. OJan<lf1lSd:h;r,r. ApPI 1'b~'S ktt.

J. 7 (I962l.

>,

'Figu~ 12 lM, I1cco ffIeIJ:Y &ffi~ily F~ <:I/non~ic ... wrnai ....."'1 ... i~ apprcmn.atcly iooCp"ndent ulllM' inleu\,ly of I~ applied m;I~IC field 8. At.l ~mPf'r2luu' T < T< 1M n...LLl '" a MJper<.... wluctor In a:ro n~,~ic flCld, 00 Ihal FI(T, 01" Ic...-cr lhlln F..(T. 0) An "'l.Jlcd magnet", r...1d iocrN'<.-s F. by 8:ts.r. in CCS u'llls. SO dl:ll "'s(T, 8a' - F,(T, Ol + B"',.nr If 8... ~ lhan theenl...,.l fK'W lI_lh.' rr~ ec'J:Y tkMty" Jo-...., .ln the nnnnal ~tal:c than in th.. supt:lt.'OI1<!uclonlt Male. a....1 now tho- normal Jutei:511wo stabIc ~1J.le. TIle origin oI"thlo '''11;.,... K1IIe In die dnwing is at F~(T, 0). ~ Iigure rq.uJI}' applies 10 Us and U", a' T - O.

errt."CI? We do not want to modify the ~1axwcll equations tIJelllsc!\,(.'S. Electric'"l conduction in the normal slate ora metal is described by Ohm's law j = {TE. '\'e ne<.od to modify this drastically to describe I..'Onduclion and the t\lcissncr efft.'Ct in the supcrconducting state. Let us make a poslllf(Jte and sec what happens. \VC postulate that in the supcrcondueting state the currenl density is directly proportional to the \'L'<:tor potclltial A of the local magnetic ficld, where B:: cuI'! A. The gauge of A will be spt.'Cified. In CGS units we write the constant of proportionality as -d4'1lAr. for reasons that \\ ill become clear. llere c is the spL't.d of light and AL is n constant with the dimensions of length. In 51 units wc write -lip-oAf.. 1110S (CGS) j:: - - - 2 A ; 41l"A L

(51)

(10)

'nlis is the London L"()t13tion. \Ve f'\press it another way by taking the curl of both ~ides to obtain
(CGS) curl J = - - - , B 4'1lA,_

{sn

curl j ::

(11)

The London equation (10) is understood to be written with the n:ctoc potclltial in the London gauge in which di\' A = O. and A.. = 0 on any edemal

c .

'11111
" .. 'j()()

....igure 13 Pl"uetration of an "PP1il:d magnetic rK'kl into a semi.i"llnite ,ul>l"n:uoductor. ~)leOl' tl1llion depth A " ,,,.,,......,I.os the di>L'lnce in which the r""ld decreases hy the f.'lctor <!I-I Typically,

A.

in a

jJUrt'

wpcm:>nc:!uclor.

surface thrOugh which no extcn'lal current is fed. The subscript n denotes the component normal to the surf3CC. 11ms div j = 0 <Indj.. = 0, the actual phy~ical bOllndary l..'onJitions. 111C form (10) applies to a simply colllu.'Ctcd Supt~rcon(llle tor; additiOl\tl.1 terms may be present in d ring or cylinder. but (II) holds lrue il"Klcpendent of gCOInetry. First we show that the London equation leads to the Meissner t'fTcct. Bya ~l.l'l:well c<luation we know that
4w (CCS) curIB:::-j

(St) curl B "" .uoi

(12)

under static conditious. \Ve lake the curl of both sides to obtain
(CGS)
(SI)

cllrl curl B = - V28 "" -

4w

curl j

curl curl B "" -V2B = 11-0 curl j ,

which may be combilled with the london equation (II) to give for a supcrcomluctor
(13)

This etl\lation is scen to account for the Meissner effect because it docs not allow a solution uniform in space, so that a uniform magnetic field cannot exist in a superconductor. 111at is, 8(r) :: 50 "" constant is not a solution of (13) unless the ronstant field Bo is idcotically zero, The rcsuh follows because vtBu is always zero, but Bo'AL is not zero unless 50 is zero. Note further th.\! (12) enSUtcS that j "" 0 in a region where B "" O. In the pure supcroonducting state the only field allowed is exponelltially dampt.'<1 as we go in from an external surface. Let a semi-infinite supen:undllctor occupy the space on the posith'e side of the x axis. as in Fig. 13. If n(O) is the field at the plane boundary. then the fJeId inside is
8(x) "" 8{O) cxp(-x/AtJ

(\4)

'52

for this is a solution of (13). In this example the magnetic fT is assumed to be parallel to the boundary. 'nms we see At, measures the depth of penetration of the magnetic field; it is known as the London penetration depth. Actual penetration depths arc not described pn..'Cisely by At. alone, for the London l."Cluation is now known to be sOlUC\vhat o\'crsimplilk'<!, II is shown by comparison of (22) with (II) that

(CGS) AL = {l'l'lc2/4r.nq~lIt ;

'"

(14a)

for particles of charge q and mass m in concentration n. Values are gh'en in Table 5. An applied magnetic field 8. will penetrate a thin (11m fairly unifonnly if the tMckness is much less than AL ; thus in a thin film the ~leissner effect is not complete. In a thin film the induced fielcl is much less than B". and there is little cffl.'Ct of B" on the Cl)crb'Y density of the stlpNconducting state, so that (6) docs not apply. It follows that the critical field fie of thin films in p..1mllel magnetic fields will be \'cry high.

Collere"ce Le"gtll
111C London penetration d<'1)th At. is a fundamental length that characterizes a supen::onductor. An ind<.>pcndent length is the oobercnce lengtb ~. 1ne coherence length is a measure of the distance within which the superconducting ekoctron concentration cannot change drastically in a spatially-\mying mag~ netic ficld. 111e London equation is a local equation: it relates the CUrTent density at a point r to the \ector potcntial at the same point. So long as j(r) is gh'en as a constant thllcs A(r), the current is reqUired to follow exactly any variation in the vector potential. But the coherence length ~ is a measure of the range o\"er which we should average A to obtain j. It is also a measure of the minimum spatial eKtent of a transition layer bctw<.'Cn normal and superconductor. The coherence length is best introduced into the thoory thwugh the LaIld.1uGinzburg <''quations, AppendiK I. Now we give a plaUSibility argument for the energy rCfIUircd to modulate the superconducling eJt..octl'On concentration. Any spatial \'ariation in the state of an electronic system rCfluires eKtra kinetic energy. A mOflulation of an eigenfunction incn.>ases the kinetic energy because the modulation will increase the intcgrd.l of It is reasonable to restrict the spatial \'ariation ofj(r) in such a way that the eKtra energy is less than the stabilization enCl"g)' of the supercondueting state. We compare the plane wave l/I(x) =. eih with the strongly modulated \\'a\'efunction:

cr"'dr.

(ISa)

J2

S",H'rl'ot,d"clidly

353

Tar

Calcullltcd inlrinsic .,:ohcrence length and

London penctration depth. at absolule wro

lnlrin~ic

Pipp81d

Loodoo.
pcnetrlliion
in 10~ nn

<,'ohe....."ce
k-ngth ". in 10-'1.,."

'depth A".

S"
Al

Ph Cd

2J. 160. 8.3

3.4
I .

0.16 0.010
0.45

"
After R.
MCSft'V~

78. 3.8
and B B. Schwartz.

3.7 11.0 3.9

0.14 1.02

l1lC prooobility density associat(,'(\ with the plane wtwe is unifonn in space: ~l/J = e I~' flh = I, wlleTt~as Ip*Ip is lIlodulat(,'(\ with the W3y(,"Vcctor q:
fP.fP = He-d+qj.>-

+ e-ik'Xe'll+ q)< + r)

= H2

+ c'Y + e- jq') = I + cos qx .

I

(ISb)

The kinctic encrgy of the wa\'c Ij/(x) is = fl2/.?/2m; the kinetic energy of the modulat<xl density distribution is higher, for

Jdx~ (-~ J2)

2m dJ!l

fP =

-.!.-(.!!...-) {{k + q'f + ~I ~k'- + ~ kq 2 2m 2m 2m

where we neglect q2 for if <0: k. lhc increase of energy re<luiretl to modulatc is 1l 2k(112m. If this increase cxceeds thc energy gap EI!.' supcrcooollCChity v.ill be destroyed. 111c critkaJ value qo of tile modulation W3Ve\'edor is gkcn by
--k~qo= ~.

2m

"

(160)

We define an intrinsic concrence length tion by to = lIqf}- We have

related to the criticalmodula (16b)

where I)~. is thc ekctrOiI velocity at the FCl11li surface. On the BCS tht..'OT)' a similar result is found (QTS. p. 173):

(17)
Calculat(,"(] values of f"o from (17) are given in Table 5. TIle intrinsic mhcrcnce length f"o is characteristic of 3 pure superconductor.

In impure materials and in alloys the coherence ICIlJ.,'th 'is :r.horter than fo. This may be understood qualitatively: in impure 1ll.lterial tile electron eigenfunctions already ha,e wiggles ill them: we can construct a given localize<l \'8riation of currCll1 density with less energy from \\'8\cfunctiolls with wiggles than from smooth w~l\'cfunctions. 'n,e coherence length 6rsl appeared in the Llndau-Cinzhurg equalions; these equations also follow from the BeS t1'M.."OI). l11ey dt.-scribe the stmcture of the transition layer between normal and supen:onducting phases in contact. TIle coherence length and tile actual penetration depth Adept..1ld 011 tIle mean free path l of the electrOnS measuretl in the oormal:r.tate; the relatioll:r.hips are illdicated in Fig. 14. Whcn thc supercomluctor is very impure. witll a \ery small t. then'"'" (fof)'12 and A ... A,. (Wt}ll2, so that AI, ... Adt. l1lis is the "dirty superconductor" limit. The ratio is denoted by 1(.

M,

BCS Theory of Sllperconlluctjljj,y

T1,e OOsis of a quantum theory of supcrconducti\'ity WdS bid by the classic 1957 papers of Banlcen. Cooper, alld Schricffer. 7 There is a "Bes lheory of superconductivity" with a n::ry wide range of applicahility. from lIe3 atoms in their COIKlensed phase. to type I amll)'C 11 metallic supefC(Jf)(luctors. and to high-temperature superconductors based on plant.'S of Cliprate ions. Further, there is a "ues \\~.tvefunetion" compol>C<! ofparticle pairs I.: f and -I.:! , which. when treated by the ues theory, givt.'S the r.ulliliar electronic superconducth'ity observed in metals and el(hibits the t.'l'ell.'Y gnps of Table 3. TIlis pairing is known as s-wave pairing. There are other funns of particle pailinK possible with the Bes theory, but we do 1l0t have to consider otliel' than the DeS wavcfunction here at this time. In t1Ji~ ch.'lpter we tn.'Ut the specific accomplishments of lies theory with a BeS wavefunction, \\hich incllllle:
l. An attractive intemction between electrons am lend 10 a ground slate separated from el(cited states by an energy g..'lp. "J1le critical field, the thermal properties, and HlOSt of the eh..-etromagnetic properties nrc COnSe<]llellt.'CS of the energy gap, 2. The electron-lattioc>..c1ectnm intemction lends to an enerb'Y gap of the observed magnitude. TIle indirt.'Cl interaction pn>Ct.'Cds when one ek"'<:tron interacts with the lattice and deforms it, a S(."COnd c1t.'Ctl'On St.'CS the ell'formed lattice and adj\lSts itself to take advantage of the deformatioll to lower its enefb'Y Thus the second electron interacts with the first electron \'ia the lattice defonnation. 3. 1'he penetration depth and lhe coherence length emerge as natural consequences of the UCS theory. TIle London eqnation is obtaint.'1..I for magnetic fields that vary slowl)' in space. TIllis the central phenolllt.non in superconduliivlty, the Meissner effect. is obtained in a natural way,
1J. 8artIefll, L N Cooper. and J. R Schnt'flff, Ph), ~ tOO.
(I!li7), 108, 1175 (I!bl)-

.-

Jo'igure'4

l'cr.ct....liot, lkpli, .\

,w!. the

c:ober~

k'flglb f

.at

5,onctlOlU of tIM' mean

r......, path t

...
.3

01'11>.. <.'0.,,1<.11011 el....1m"'; In the ""rmal SI~I<-. All 1"l\gth. are on units of fa. th.. intrinsic rollff.. ~ Itnglh. 1l>f' l .........-n lU"e ~kt'k:b<.'tl 10.\.. For shott melm rrec p.otlu tiM' C'Oherenct: Ienlt1h l>t'COm<:s >harter anti IJot"

:0

for" -

.2

0.1 -

pcnt.1.r:>I.iun

tl.:pth

br.rof,lel;

loo,ger TI'e Inc_ ,n I.... ntio >('\'/f 6n"On type II s"pc.'1"nlIloueth~t).

~.--~~--~1---~---2~-

~-

4. 1'he criterton for Ille transition temperature of all clement or allo)' in"olvt."S the ekctron density of orbitals D{E~,) of one spin at the Fermi levelllnd the electron-Ialtice interaction U. which enn be estimated from the electrical n:sisli\"ity because the resisthity at room temperature is a measure of the el~ Iron-phonon interaclioo. for UD(Ep) ~ 1 the UCS theory prt."(licts
T e = 1.149 exp(-IIUIJ(EF)}
(18)

where 8 is the Deb)'e teml>eratUTc and U is an attractive interaction. 'nle result for Te is satislit."<! at least qualitatively by thc eXl>erirnental data. 'nlere is an interesting apparent P.'\r.IUOX: the higher the resistiVity at room temperature the higher is U. and thus the more likely it is lhat a mctal will be a Superconductor when cooled. 5, Magnctic flux through a supercolldllcting ring is qU:lIltized and the effective lI11it of charge is 2e ralher than e. 11\(> BCS ground state im'oh't.>s pairs of dcetrons; thus nux quantizalion in tt.>rms of the pair charge 2e is a consequence of the theory,
BCS Cround Slate

The filled Fermi sea is the ground state of a Fermi gns of non interacting ek>ctrons, 'nlis slnte allows .rrbitmril)' small excit.ations-we can form an exdtt."<! stale by taking an ek"Clroll from the Fermi surface and rdising it jll!>1 abm'e Ihe Fenni surface_ 11le ncs theory shows Ihal with an appropriate atlractin~ interaction between elt.'Ctrons the new grollud state is slIpcrconducling and is SI..1larat<."(\ by a finite energy E from its lowest cxcitt.-'d stale. 'nle fonnalion c:L the BCS ground state is suggested" b)' Fig. 15. The BCS state in (b) contahls ru:lmixturt.'S of one-elcctron orbitals from above the Fermi energy ~" AI first sight the BCS slnte appt."'I"S 10 hu\'e a higher energy than the Fermi state: the comparison of (b) 'vith (a) sh(1WS that the kinetic energy of the BCS state is hifL r than thdt of the Fermi stale. Bill the atlracti\'e potential

energy of the BCS state, although not reprcscnh,>d in tI gurc, acts to lower the total energy the BCS state with respect to the Fenni state. When the BCS ground state of a man)'-clectron s),stem is described in terms of the occupanC)' of one-particle orbitals, those nl.>ar F arc filled somewhat like a Fermi-Di!!iC distribution for some finite temperature. 'Ine central feature of the BCS state is that the one-particle orbitals are occupied in pairs: if an orbilal with wa\'c\"cctor k and spin up is occupied, then the orbital with wan:\"ector -k and spin downjs also occupied. Uk l t --t! \-acant, then -kit is also vacant. The pairs arc called Cooper pairs, treated in Appendix H. 'JOey ha\'e spin zero and ha\'C man) attributes of \>osons. The boson condensation temperature (TP. Chapter 1) calculated fOr metallic concentrations is of the order.of" the Fenni temperature (Ja' - Itf K). The supercondueting transition temperature is much lo\.\'Cr and takes place when the electron pairs break up into "'''''0 fennions. Thc model of a superconductor as composed of noninteracting bosom cannot be taken absolutel)' litcrally, for there are about Hf electrons in the \"olume occupied b)' a single Cooper pair.

or

Flux Quantization in a Supercondudillg Ring We prove that thc total magnetic nux that passl.'S through a supcrronduct+ ing ring rna)' assume only quantized "alues, integral multiples of the llu.\ quantum 27rhdq, where by experiment q = 2e, the charge or an electron pair. Flux quantization is a beautiful example of a long-range qu:mtum dfect in whieh the coherence of the supcrconducting state extends over a ring or solenoid. Let us first consider the ck'Ctromagnetie field as an example of a similar hoson field. The electric field intensity E(r) acts qualitatively as a probability field amplitmle. When the total number of photons is large, the energy density mal' be written as
P(r)E(r)/4w ill n(r)hw

where n(r) is the number density ofphotons offrcquency w. 'J11en wc ma~' write the electric field in a semiclassical approximlltion as
E(r) a; (4whw)If2 f1 (r)I/2 dtJ:r) E*(r)
ill

(4whw)ll2 n(r)lf2

e-Hl,r)

where 6(r) is the phase of the field. A similar probahilit)' amplitude describes Cooper pairs. 'fne arguments that follow apply to a boson gas with a large number of bosons in the same orbital. We then can tn.'8t the boson probability amplitude as a classical quantity. just as the electromagnetic field is USl.'<! for photons. Both amplitude and phase are then meaningful and observable. The arguments do not apply to a metal in the nonnal state because an ek'Ctron in the normal state acts as a single unpaired fennion that cannot be treated classically. We first show that a charged boson gas abe)'S the London equation. let

.,
,'--"------4,1----

r-,--..,,'"

-II-"
I I I I I I
I I

.-

"

.-

, "

FCU~ 15 (.) PnJbIt,ilily l' that IU1 orbi:ta1 clkinetie energy It" i. OCCllpicd in the ground dateofthe noninlcmcting fen,,; gas; (h) the BCS ground .Me dilT..,n from the "'errn, Slale in a region of width of the order of the cnerg)" gap f,.. 80th ewve$ are for absolute zero.

l/!{r) be the particle probability amplitude. We suppose that the pair concentration n = ~I/I = constant. At absolute zero n is one-half of the concentration of electrons in the conduction band, for II refers to pairs. -nlen we may write
(19)

TIle phase O(r) is important for what folktws. In 5l units, set C = I in the equations thai fOllow. 1lJc \dooty of a particle is, (rom the Ilamihon (.'qualions of mechanics,
(eGS)

The particle nux is given by

!/J*vljl = -;;;-

" ( 'IV 0 - --; q) A

(I

(20)

so that the electric current density is

We may take the curl of botl1 sides to obtain the London equation:

(CCS)

.= -"of curl J 8 , me

(22)

wilh use or the fact that the curl or the gradient or a scalar is ioc.>ntically zero. "l1Je constant that multiplies B agrees with (14a). We recall that the t\lcissner effect is a consequcnce or the London equation, which we have hcre derived. Quantization the magnetic 11m: through a ring is a dramatic COIlSf.."qucnce or Eq. (21). Lct.us take a closed path C through the interior orthe supercon-

or

~1ll\

lin",

!'iKure 16 Path of ;Illc~ion C thro"~" lhe interior of a slljX'rrondud;ng ring. TIle through tl,e rinK is the SUm of the nUll: >..... frotn eo;trnwd SOUI'tt'li -.x111,e nUll: >.. l'roen Ihe lUpt"I'IIOnducting <:'UI'TerIt'J "hid! Row in lhe .uriion: Ihe n~ 41 - 41.., + >... The flu.'C > is 'luantiM. Then: is norm.llly no quantization l.'Ondition On the flux fm'n cxtcmal ""ur,:es. so Ihat 01>... ll1ust adjust ibl'lf "ppl1'priatcly in "rtletthat 01> :l.5JUlne a qu.IlIltizc:c1 ,alu",

nm

or

dueting material, well away from the surface (Fig. j6). The Jl,leissncreffect tells us that 8 and j arc zero in the interior. Now (21) is zero if
ftcVO = qA . We fonn

(23)

VO dt =

lk. -

01

for the change of phase on going once around the ring. The probability amplitude l/1 is measurable in the classical approl(imation, so that l/J must be single-valued and

8zwhere
$

01=2m,

(24)

is an integer. B), the Stokes theorem,

Adl=

1
c

(curIA)du=

1
c

Bdu=$

(25)

where du is an dement of area on a surface bounded by the curve C, and <l> is the magnetic flux through C. From (23), (24), and (25) we have 2r.fJes '" q<ll, or
.. =

('modq .

(2)

Thus the flux through the ring is quantized in integral multiples of 2wfu:1q. By experiment q = -2e as appropriate for electron pairs, so that the quantum of flux in a superconductor is

(ees)
(Sl)

41 0 = 2r.hd'li!" 2.0678 x 10-1 gauss cm 2

lo =

2,""'2" 2.0678 X 10-15 tesla m"2

This unit of flux is called a fluxoid or fluxon. TIle flux through the ring is the sum ofthc flux 41..., from extcrnal sourceS and the llul( It<" from the superconducting currents which flow in the surface of the ring: <l> = ~d + ....,. The nux 41 is quantized. 'nlere is normally no <]uanti-

zation condition on the flux from extemal sources, so that ~>C must adjust itself appropriately in order that ~ assume a quantiz.t."l1 value. Suppose a magnetic monopole of strength g is siluatt.'tl just below the center of a superumducting ring. The magnetic flux through the ring is (w,-2)(2.~ "" 2.wg, and by (27) this must equal an inl<.>gral multiple of .,mde. I1lUS the minimum pcnnissible valuc of g in CCS is 'lc12e, the famous Dirac result.
Duralion af Persislerd Currenu

Consider a (JCrsistent current that flows in a ring of a t)'('lC I superconductor of wire of length L and cross-sectional area A. '111e persistent current maintains a nux through the ring of some integral number of Iluxoids (27). A Iluxoid cannot leak out of the ring and thereby recluce the persistent current unless by a thennal fluctuation a minimum \'olume of the superconducting rillg is momentarily in thc normal state. Thc probability l'lCr unit timc that a fluxokl willicak out is the product
p = (attempt frequency)(activation barricr factor) . (w)

The activation barrier factor is eJqJ{-~FlkBT), where the free energy of the barrier is
Ii.F .. (minimum volumeXexccss free energy density of normal state) .

The minimum volume of the ring that must tum normal to allow a nuxoid to escape is of the order of Rf, where of is the coherem;e length of the superconductor and J{ the wire thickness. "n\(.' excess free energy density of the normal state is H':!8'fT, whence the barrier free energy is
(29)

let thc wire thickness be )0-" em, the coherence length = )0-4 em, and He = l(il C; then li.f .. 10-7 erg. As \\e approach the transition lempcmture from below, Ii.F will dl..'Crca5c t()Y,'ard zero, but the value gi\'en is a fair estimate between absolute 7..cro and 0.8 Teo Thus the activation barrier factor is
exp(-Ii.FlkBT) - exp(-IOS) - 10
(O-t><IO') .

'111e characteristic frl.."'<]uel1cy \"ith which the minimum \'olume call attcmpt to change its state must be of order of E,jfl. If Eg = 10- 15 erg, the attempt frcquency is - 10- 15/10- 27 - 1012 5- 1. The Icakage probability (28) becomes
P _ 10'2)0-..... 34><10' 5- 1 """

10- 4 :w><ui'

S-I .

The reciprocal of this is a measure of the time required klr a f1uxoid to leak out, T = Jlp= U)"3ol><lo' s. The age of the universc is only 1OIOJ s, so that a lluxoid will never leal. out

in the age of the univcrse, under our assumed conditioru ccordingly, the current is maintaim,.'{1.. 'J11erc arc two circumstances in which thc activatinn energy is much lower and a lluxoid can be obscrn,,,J to leak out of a ring-either \ery close 10 the critical temperature, where H t is \'cry small, or when the matcrial of thc ring is a type II supc..>rtlOnductOl" and already has "maids embedded in it. These special situations arc discussed in the literature under the subject of fluctuations in superconductors.

Type II Sllpcrcondudors
'nlcrc is no difference ill the mechanism of supcrcolxlucti\'it)' in I)'PC 1 and t)'pC II superconductors. Both types ha\'e similar thcnnal properties at the superconductOf'-normal transition in zero magnetic field. But the ~'leissner effect is entirely different (Fig. 5). A good type I superconductor excludes a magnetic field until superconductivity is destroyerl suddenly, and then thc field penetrates completely. A good type U superconductor cxdudl.'S the field completely up to a fiekl H.:+ AI)()\'e H tl the field is partiall)' excluded, but the specimen remains electrically supcrconducting. At a mudJ highcr ficld, H et, the flux penetrates compJetely and superconductivity vanishes. (An outer surface layer of the specimen may remain superoonducting up to a still higher field HOj.) An important diflcrence in a type 1 and a type II superconductor is in the mean frl."C path of the conduction ek'Ctrons in the nonnal state. If the coherence length f is longer than the pcneiratlon depth A, the superconductor will be type I. Most pure metals are type I, with I( < I (see Table 5 on p. 353). But, when the mean free path is short, the cohcrence length is short and the penetration dcpth is great (Fig. 14). TIlis is the situation when I( = Mf> ], and the superconductor will be typc II. We can change some metals from type I to type II by a modest additioo of an alloying element. In Figure 5 the addition of 2 wt. pcn:cnt of indium changes lead from typc I to type II, although the transitX>1l temperature is scarcel)' challb'Cd at all. Nothing fundamental has been done to the electronic stntcture of lead by this amount of aIlo~illg. but the magnetic bcha\'wr as a superconductor has changcd drastically. The thL'Ory of type II supert.'Onductors was dc\,c1opL'(\ by Ginzburg, Landau, Ahrikosov, and Gorka\'. Later Kunzler and co-workers observed that Nll:JSn wirL'S can carry large supercurrcnts ill fields approaching 100 kG; this led to thc a>mmcrcial dc\'Clopmcnt of strong field superconducting magnets. Consider the interface between a region in the supcroonducting statc and a region in the normal statc. 11lc interface has a surface cncrgy that may be positive or negativc and that decrCases as the applied magnetic field is in-

J2 S"f1UmJuelldty

361

1111111, 'IUl111111111'
I z
i z
figure 17 (a) Magnetic Geld penetr.ttion into a thin mm of Ihidmeu flIua1lo thr penetration depth A. TIle AI'J'(IWS indicate the intCf\$il)' of the magnetic f>dd. (b) ~l~ r.e&d.pencl..-;o" in a homogeneous bulk Itructure In the mwd or \'OI1ex :ItlIle, "ith altcrnale b)~ in DI)fm:a1a..J ~upcrcoocludjng .tales. "The superroodnctir-.,; byen are thin in coonparioon with A. the laminar str\lcture is .hown for ronv<,nk'Jlcc; the actual struclt,.... <'On."ts of rods of the normal $tatc .ur I'O<mdcd by the .upcrron<'ucllllg .tate. ('The N region. in the vorlu state are not e~actly normal, bot are d...,robed by I....... wloes of the stabiliutioo energy density.)

creased. A superconductor is type I if the surface energy is always positi\'e as the magnetic field is increas(:d, and type II if the surface energy lx.'COmes ncgati\'e as the magnetic field is increast..-d. 'l1te sign of the surface energy has no importance (or the transition temperature. T11C frcc energy of a bulk superconductor is increased when the magnetic field is expelled. J lowever, a parnllel field can penetrate a vcry thin mm nearly uniformly (Fig. 17), only a part of the nux is expclk--d, and the energy of the superconducting mm will incrcl1Se only slowly as the external magnetic field is increased. nis causes a large increase in the field intensity required for the dcstruction of superconductivity. The Olm has the usual energy gap and will be n:sistancck>ss. A thin mm is not a type II superconductor, but the mm results show that under suitable conditions superconductivity <''1IJI exist in high magnetic fields. Vorlex Stale. The results for thin mms suggcst the question: Arc there stable confib'lJrntions of a superconductor in a magnetic field with regions (in the form of thin rods or plates) in the nonnal state, (.>3ch normal region surrounded hy a superconducting region? In such a mlxL-d state, called the vortex state, the external magnetic field will penetrate the thin normal regions uniforml)', and the field will also peoctTatc somewh.at inlo the sunuunding superconducting material, as in Fig. 18. 'Ibc term vortex state describes the circulation of sllperconducting currents in vortices throughout the bulk sfH.:dmcn, as in Fig. 19 below. There is no chemical or crystallographic diflcrencc belwccn the nonnal and the superconducting regions in the \'Ortex state. 1bc vortex state is stable when the pcnetra-

1\1"" r

,,,""<>n<!....1ur

..

,-

...,
,

-,

~l'" II "'P'...-duttor

..

Ncm:lLll

'.

B.

+ 8.

1---,
~,}

:"
-

'I

,."

Figure 18 \'ariIIlion oflbe magnetic field and e~rgy PI' ~er~) at ~ mlni":K'e of ~upeyroadlrl fng.ond normal regillns. lOr Iype t and type II supcl"conductors. 'nle energy gap parameter is '" meru;u", or the stabilization encrgy dcnslty of the supclU'lOdueling .late.

tion of the applk--d field into the supcrconducting material causes the surface energy to become ncgative. A type II superconductor is characteriu.--d by a \1)rtex state stable o\'er a certain range of magul..>tic field strength; namely, bem~n H~I and Ht:?. Ertimatiofl of lId and lId' What is the condition ror the onset ofthe \'ortex state as the applied magnetic field is increased? We estimate H~l from t.he penetration depth A. 111e field in the normal core of the fluxoid will be He! when the applied field is H el The field will exlend oul from the normal core a distance Ainto Ihe superconducting environment. '111e flux thus associated with a Single corc is 7rA 2 I /~" and Ihis must be equal 10 the flux quantum 4'0 defint..'t! by (27), -J1ms
He. _lfJr/7rA'l. ,

(30)

This is Ihe field lOr nuclealion of a single fluxoid.

l'igtlrc 19 Flu~ lallke In "bSe, at J.(l(X1 gllU at 0.2)(, ;IS ~icwc<' "ilh a scanning t"nneling rn>eroscope. The photo ~hOW1< the <kosity ofstat"" ,I! the Fermi kwel, lIS ill Figun' 23. The vortex cores ha"e' high density of Sll.l!es and are shaded "hite; t.be supercon<lUding..egions ~ tim, WIth no state~ Q.t the t'-enni Ie,'d. The amplitude Mel spatial edent of these states is o.Ie~mlilloo by. potential well 'mned by Allr} ;as in Figure 18 lOr a T)I>e JJ wpenvn<!uctor. The poIl.'IIW ",,11 confines the rore Sbte ""...-d"unctions in the image hen,. The stu Wpe is. 6...... r..-.t"re, a ~uJt special W ,,~ ~ lhe PdoId dish..banc:e ~ the d.Iuge dC<uity at the Fenlli wrf.oer Photo courtesy of II. F. Ilea. ATliT Bell l...abontones.

Al I-I c2 the Ilu:>;oids arc packed together as tightly as possible, consistent with the prt.'Servatioll of the slipercondlicting slate. This me.lllS as denscly as the coherence lenb>th ~ will allow. TIle cKtenlal field penetrates the specimen almoSI uniformly, with small ripples on the scale of the fluxoid lattice. Each core is responsible for carr:ying a flUll: of the order of 1'T~ H d , which also is quantized to o- Thus
(31)

gives Ihe upper critical field. The Illrger the mHo JtI~, the larger is the ratio of 11<:2 to Hel It remains to find a relation between these critical fle1ds and the thermodynamic critical field Ht: Ihat measures the stabili7.3.tion energy densil)' of lhe superconducting slate, which is known by (9) to be IP.J81'T. In a type II super~ cooc:Iuctor \\-'C can dclennine He only indirectly by calorimetric measurement of

the stabilization energy. To estimate H el in tenns of H",I _'Onsider the stability ohhe vortex slatc at absolute zero in the impure limit ~ < Ai here I( > ) and thc coherence length is short in comparison with the penetration depth. \Ve estimatc in the \'or1cx state the stabilization energy of a fluxoid core viewed as a normal metal cylinder which canies an a\cragc magnetic field BaThc radius is of the order of the coherence length, the thickness of the boundar')' between N and S phases. The energy of the nonnal core referred to the cnergy of a pure superconductor is gi\'en by the product of the stabilization enc'b'Y timl.'S thc arca of thc core;
(CCS)
pet"

(32)

unit length. But thcre is also a decreasc in magnctic energy because of the penetration or the applied field B.. into the supercondueting matl."Jial around the core:
(CCS)

f,''''X'' --B~ 8w

X 7TA

(33)

For a single f1uxoid we add these two contributions to obtain

(CCS)
(34)

Thc core is stahle iff < O. The threshold field for a stable f1uxokJ is at! = 0, or, with He' writtcn for B",

HellH e

QA .

(35)

'nle threshold field di\'ides the region of posith'e surfuce energy fmm the region of negath-e surface energy. We can wmhine (30) and (35) to obtain a relation for He:
7T~AHe

.. <llo .

(36)

We can combine (30), (31), and (35) to obtain

(lIe,H,dII2 .. He ,
(37a)

'nd
(31b)

Single Particle Tlmnelillg

Consider two mctals scparall.>tI by an insulator, as in Fig. 20. The insulalor normally acts as a barrier to thc flow of conduction electrons from one metal to the other. If the barrier is sufficiently thin (less than lO or 20 A) there is a Significant probability Ihal an electron which impinges on the barrier will pass from one mctal to the other: this is callal tunneling. In many experiments the

Figure 20 Two metnls, A 1l1\(1 B, liCplIr<lI..>d by a thin Iaycr uf an Insulator C.

Figure 21 Prcparation of an AVAl~n ,.,dwich. (a) GlaM ~lidc \\ilh indium coot-:ts. (bl An aluminum strip I mm ...ide and 1000 1o:tOOl) Athick has beet! dqJosiletl aetoliS thecootaets. (c)l1>e aluminum strip ' - been oUlized to om, an AlA 1.)'0" 10 10 20 Ain thodcnas. (d) A tin film ' been dqlosited kIOII5 thealuntinum film, formi,,&an A1fAltO:/Sn sarxh.i,h. l1>e oh':malleMk are Corlnected 10 t})o, indium oont-cts., two oonlactli ~ UliCtI for the currenl m ~ 1 and lwo lOr the \"01"'Je measuremcnt. (Afler GiaL...~r MId Me~c"'e.)

..... "
"J

Figure 22 (al Unr:nr C\lfre"",.'{)ltdgc r..J'\lion rOl Junctitm of normal m..lal~ .cpo.\Tdtcd I oxidc larer; (b) .."rTcnt-\'u!lnR" rdation with one ",elal nomlnl anel the otl,c'r mclal ~"pcl"('O",l"dilJ~.

insulaling larcr is simply a thin oxidc laycr formed on onc of two c\'aporatcd metal films. as in Fig. 21. Whcn both metals are normal conductors, the current-voltage relation of the sandwich or tunneling junction is ohmic at low \'oltagcs, with the current directly proportional to Ihe applied voltage. Ciac\'er (1960) discovcred Ihal if one of the metals be(;omcs superconducting the eurrent-voltagc characteristic changes from Ihe straighl line of Fig. 223 10 Ihe curve shown in Fig. 22h. Fib'Ure 23a contrasts the ek"CIron density oforbitals in the superconductor with lhat in the oonnal metal. In the supcrconductor thcre is an energy gap centcred at the Fcnni k,,e1. At absolutc U-'fO no current can flow until the applied voltage is V:: EJ2 = 6fe.

~,

------

T>O

F"1f;U1Y 2J 1be dnuih'clorhtl.-)s and lhc~"'f"-... I-\'Oll:l# c"a"'1N~>Ch"alunnclinl(jund;oll. In {.a}thc: <."flO:'fJ:)' 1$ pk>tted 011 the .nk alld tho: d.,....I)'oI...hilaoo Ill(' hOJuout.<1 ocaIc. One

,..,rt..,....

metal" In th., nom>al st.Ih, and or'" in II... ~l'P<'n""""ud"'K)talc (b)' 't'f".m \'. thtd:......... indicale the n;p<.'C1..-d b'.ok al T" O. (Aft.". Cw..-vtt ..1<1 M"'K".Ie.)

l11c g;.tp /1( L'OrrCspOoos to the hreak-lIt> of a pair of dt.'Ctrons in the superconducting stille, with the formation of two ck-clruns. or an ell"Clrun and a hole, in (he normal state. The current starts when eV '" 6. Al finite Icmpcrnhlrcs there is a small (:uTrenl flow even al lnw \,oltagt:s, because of ck"Clrons ill the supcn:ondul-Ior 111..1.1 arc thermally cxcitceJ across the ellerg)-' g;.lp.
jos'!J>llSon SU,Jefcoruluclor TWlrlcling

Under suilable conditions we ohserve remarkahle eITttts assuciah..~1 with lhe tunneling of SUpcfl'onduclillg electron j)"1.irs from a supcn.'Ollductur through a lavcr of all insulator into al\othL"f superCOmhK.1or. Such a junction is l.-alkd a weak link. The effects of pair tunncling include:
Dc Josephson effect. A de l1Jrrenl flows across the junction in the absence of allY electric or magnctie field. Ae Josephson effect. A dc voltage appliL'll across tllC junl1ioll causes rf cmrent osdll.ltions across Ihe jUIlCtiOI\. This clfect has bel'n utilized in a precision determination of the vdlue of Me. Further, all If voltage applied with the de voltage can then cause a dc CllITent across the junction. Macroscopic long-range quantum intcrferCllCC. A dl' umglll.:tic field applied thrOllglt a supe:onducting drcuit l'Olllaining two junctions ClUlSL'S the ma.'CimullI supercurrent to show interferenl'C effects as a function of magnetic field intensity. 111i5 effect can be utilized in scnsitive magnetometers.

Dc Joscphson Effect. Our discu~sion of Josephsoll jUl1Clion phenomena follows Ihe uiscussioll nfnuJ[ <jllantizatklll. Let "'I IJC lhe probnbility .unplitude of electron pairs on one side of ajunl.1ion. anti k'il/J21Jc the amplitude 011 the other side. For simplicity. let both SlLllCTt:onductors bc identical. For the present '\e suppose that they nrc bolh.1l zero jX)lL'tltial. 'l1\e (jmL"-CIepenuL'ut Schr(Xlinger C(llIdtK:tn jfril!JIiJt = 1';' applied to the two amplitudl."S gives

iii

al/1l.
Ilt

:::z

I,T.r.
'PI

(38)

lIerc I,'" reprcscnts the dfel"' of lhe c1ectron.p,lir couplingOl'lransfcr inlcr.ll'lion nlTOSS the iIlSul.ltor; T h.\S Ihe dimensions of a late or frequenl')'. It is :l measure oftl\(: leakage of lb. into the legion 2, .md of tP1. iulo the n'gion 1. If the insulator is vcry thick, l' is J;(>ro anti ,llere i~ no ll<\ir tunneling. Lei l]I\ :::z 1I1'l e''' anti t/J1. = 11~12 flO.. 11lcn
(39)

al/11. _ i ---a;We multiply (39) by

-lf2
/12

elO> """dt alii + '. ','. a0:2 . .2---al to obl,lin, with

l)

"Iii eo-,III

-/T,'. .. , ~ - 0,.
c' .
,0;

(.0)

1 itll l ao, -2 - + ill I - = at ilt

We llluitiply (40) by rl~12 c- i91 to obtaill
c'hh d8z. -1 --+ ;1/2-=

-i1'(n

lIJI2
I

at

at

-iT(lIII1JII2 e ,a

(.2)

Now l.'(luatc the real and imaginary parts of (<11) and simiLu'ly of (42):
(-13)

__ I

a.

= - 'f --!.
flr

(n ),n cos
at at

/j ;

- - = -T -

a.,
ilt

(... )"
112

l'O~/j.

ill

If 11,

ill

'l2 as for Klcnlical supcrumductors Land 2, we h"..e from (-1-1) tl~lt

aOaoz. I :From (43)

WC

-(8.l at
aliI ---

01)

::;

(45)

Sl..'t: that
--~

ant at

at

T11C l.'UHent !low from (1) to (2) is propo.-lilJl\.Ilto a,,~iII. or, the sJ.mc thing, -allllat. We thi'rcforc condutle froUl (43) that tllc currcnt i of superconductor p,urs :K.1OSS the junction depcnd~ nn the pll.lSe dilTercl\(;e 5.lS (-17) where in is proportion,,] to the trrrnsfcr il1tcl'ilc.1km 'I'. 11\c
1.11HCUt

in is the

't---,,{ , ,, , ,, ,,

,, , , ,,

,,
CurK'tll-\'Olt:ogc chanctffistic d a flow under zero applietl voltag<' up to acritical CUITt"llt 1,.: this is the.tic JO!'ephSOll clfect. At voltages abo...., V< the junclion ha~ a finite resi~l3IlCe, out the ell".,..,nl hal an oscillatory component Orrrc<IU"'lC)' />J - 2cVlh, Ihi~ is the ac JOj;cplisoll effect. Figure 14

JC!5CphsQo,

junctioll. Dc <.'\IrTelliS

In~imum zcro-voltagc curft.. 'nl thai can be !>"'!SSl.'l! Ily the junction. wnh no applied voltage a de CUITL'Ili will now across the junction (Fig. 24), with a value IJelwcen)o and -}o ::cording to the value or the phasediffercucc ~ - 0. This is the de Josephson effect. Ac)osepll3Oll Effect. Let a de voltage V be applied across the junction. We

C'.I11 (10 lhis because the juncUon is an insulator. An ek'ctroll pair cxperitmccs a polential energy difference qV on passing across the junction. where q .,. -2e. We ('all say that a p..'\ir on Olle side is al potential cnt:rgy -cVarn! a p-,ir on the other side is al eV. The equations of motion th.. 1t replace (38) are

(48)
\-\le pro(:ccd as above to filld in place of (41) the equation
illi l ilO I -1 - + ;111-=

ilt

at

ieVnr,,-1 - iT(lIlnVI

12

e' .

(49)

Ihis Ctluatiun breaks up into the real part

illll/iJl = 2T(nlnJIIt sin l)

(50)

exactl)' as \\;tOOut the voltage

v, and the imaginary part

(51)

ilOI/ill = (eVlII) ~ T(nil'''I)112 cos l) ,

whieh difft:rs from (44) b)' the term eVl"Further, by extension of (12),
iJn2 il02 -1 - + iI1 2 -=

ilt

ill

- j eVtl~zll-1

- iT(IIll1iJlf. e - ,

/6

(52)

12

SUJ)nnIndUC1iritli

F"1pI~ is The arnmgcment for experimenl on ~ 1"ll5CO(lK: quantum ;nlcrfer~. A ~lic >

nm,

ra~

lhn)t.lgh th... interior of the loop.

""'-lnwLotoo- b

whence iln.jat:= -2'f(n 1 n!llfZ sin 8 ;

il().jilt = -(eVIl!) - T(n I /1l2)112
(.'OS

(53)
8 .

(54)

From (51) and (54) wilh n,

'=!!

n2. we have
(55)

il(02 - Ol)!dt = il&ilt = -2eVlh

Wc scc by integration of (55) thai with a dc voltagc across the jum.1ion Ihe relative phase of Ihe probabilily amplitudes vary as

8(t) = 8(0) - (2eVtl/!) .

111e supcn.:onducliug current is given b)' (47) with (56) for the phase,

(56)

I1

:=

10 sin (8(0) -

(2eVt'h))

(57)

The current oscillalcs with frequency w = 2eVlh

(58)

This is the ac Josephson cITcct. A dc voltage of I p.V produ(''(.''S a frC1ll1ellC)' of 483.6 M liz. 'l1lC relation (58) says Ihat a photoll of energ.\ flw = 2cV is emitted or absorbed whell an ck'ClfOtl p.1.ir (.TOSSCS thc b<lrrier. B~ measuring Ihe \oltage and the frequcm1' it is possible to obtain a vcry preciS<" valuel; of ell!. Mocrmoollic Quantum Inler!erence. We saw in (24) and (26) thai the phase diITercll(.'C O 2 - 0 1 around a dllsc<! l"irl'Uit which Clll'Omp.asses a tolal maguctic nux (~ is given by
(59)

11lC nux is Ihe sum of Ihal dllc 10 edcmal fiel(ls and that due to cuneliis ill the circuit itself. We l'Ollsider two Josephson junctions in parallel. as in Fig. 25. No voltage is applied. Let the phase difference bctwcCfl lXlinls I and 2 taken on a p.1.lh through junllion (l be 8". \\Then taken Oil (l path through junclion b, the phase

'70

!
-~

!
- lW

I
~200

I
lllU 0

-JOO

roo

\1_111. rll'io,\ (u,,".l....)

""

..
I

""

Fig"n 26 E'pcm.....~It:illroK:e of J~_ ",n\l~ ""'lo7K'11C: r... k1 ~h.ming i"lcrf~",n~-e a"d .t.rfr:lrli(lII crrc...ls for 1"0 ,'"lctlOllSIr. and 8. lick! pcriodicitv IS 39 5 aud 16 mG lOr ami B, ""1M......"'e1r

nlC

AI'!HU\imall' IIJ:l,'imum L"rn,,,h;are I inA (II) and 0.5 iliA (B). '11"" jUII('tiou sc

r;dion is 3 mill anti

lundlOl'l "ithh 05 "u" lOr h011! caK~ l1lC 1<"fD of.et or A is due to a h...-kgrnuml magnetic lldd. (Artcr R. C. J.d.""ic, J. Larnbe, J. E M.-nct"t'311 and II II ~,h"r.)

diffcrClll'C is 8/,. In the absence of 1I magnetic field these two phases must be l'(lual. Now lellhe nUl( lllaSS through tile interior of the circuit. \VC do this witl, a .\lraighl solenoid normal to the plane of the paper J.lld lying iU'Iidc the c;jrl.'Uil. By (59). 8" - 1>.. ::: (2o'hc}cl1. or

6"

=.

<.otI

--<!> fie .

(00)

TI,e tolal current is Ihe sum ofJ.. illlcllb. TIle current through t:ach junc:tion i.s of the fOrm (47), so lhat

hut;,] ::: Jo {sin (60 + :c <1J) + sin (c5n - :c <1J)} ::: 2(jo sin 60) (."OS
11.(.' C\Jrrcnt varies with

':

<1J

,lIld has maxima when

S7T

e<1>lhc :::

= inli.:ger

(61)

'niC periooidly ofthl.' CUtrellt is ...how in Fit(. 26. The ~h{lrt pcrioc1 variation is pnxluccd by intclferellcc from the two junctiulls, as predicted by (61). 11.l.' IOIl~er I>criod variation is a diffr~tion eJTect and arises from the finitc dilllellsiol1~ e~h jUJX:tioll-this eauscs 4> to depend 011 the 1).1rti<.1JI,lr 1)''1th c:i integration (Problcm 6). '111c diJTr:l(,'tioll effect w.'S not 31l!icipah." in carly work tlMl resohed only single eleetroll tunneling.

or

HlGII-U:J\lPt:nATUHE SUPEHCONDUGfOns

High T c or HTS, dennles superc;onduclhily in materials. chiefl)' oxides, with high transition temperatures, ac<.;ompank'tl by high crilkal currents and magnetic fields. By 1988 Ihe long-standing 23 K cciJjn~ of T c in intcrmetallic <.;ompounds had been elevated to 125 K in bulk supercllnductlng o\ides; these pa5SIX.I the st.mtlard It..'Sts for supcrc:onllm;th ity-the ~lcissncr effect, ac Josephson cJfl,.'Ct, persisknt <:UlTents of 10llg tluratioll, and substalltiall)' zero de resistivity. B)' J994 HTS showed promise in pre-<.;olllmercial applkations, as in tlnn 6Jm devices, and wires were being fahricated. Levitated vchides and long distance power transmission have Jlllt arrived. Memorable steps in Ille advance illdutle: DaPbo 7sDio z;O:I Lal !j5Ba() 15CUO~ YBa2CU:l07 1"2Ba2Ca2Cu:)OIO
T c = 12 K

T c = 36 K Tc = 90 K T c = 120 K

[BPllO] [LBCOI [YBCOI {TIlCOI

A related lIlel'CulY l"Ompolllld may have T" = 140 K. 11le m.. 1.terial with most resean:h behintl it is YBCO. 111e C1ystal stmdures of these materials, except for UrBO, are oxygelldefl..'l,.'t modifkations of tile perovskite structtlre of Figure 13.10, with about one-third of the oxygen positions vacant. The primitive cell is dl,.'Vclopetl from that of a tetragonal pcrovskite tripled along the C a.\is. Per formula unit of YDa2;CuJ,07, the lXJsilive ion valellcies bast;d Oil y+:l, Ba+2, Cu+Z arc 3 x 1 + 2 x 2 + 2 X 3 = 13 :Uld the Ilcgative (oxygen) valency is -2 x 7 = -14. 111e quasitetragonal or OIthodwlIlbic cell dimensions of YUCO are, in nm, a = 0.388, b = 0.38-', c = 1.163, wilh cell volume 0.173 nm3; if there is one CXI,.'Css charge carrier ill this volume the l,.<lnicr concentration in 110 = 5.77 X IO Z1 cm-3. '1l1C excess l,.'Urriel's Iypically arc holes embedded in doped alltiferrolTIHgnetic insulators. Parallel shl,.'Cts ofCu02 arc a structural feature of all HTS with T" > 40 K; thesc are known as Ihe cuprate HTS. In rBCO slwwn in Figure 27 the ClI02 planes m-e pamllel to the plane of tbe ab :l"(es ill Ille orthorhombic primitive cell, Two of these plaTlt.'S, those through the interior of the l'ClI, arc tile dominant l,.olltlucting pathways. -l1lC oxygen 1,."OIltent per celll,.<ln be changed reversihly from 7.0 10 6.0 atoms, simply u}- pumping oxygen in aud out of the p,l.l<lllci chains of CuO that nlTl along the 1) axis of the figure, At the l,."()mposition YDaZCu306 the cr)-slal is all insulator, with anliferromagnetic ortler of the eu spin!>'. An inl,.YCo1.Se in ox}'gen ahove 0 6 5 makes the crystal metJ.lIic and nOI1magnetic;:; tllC crystal is supermllduding alXl\ e Ob 1).1. A dmrge-neutral formula for YBa2Cu301 can be written ,1,,<; YBaz(Cu2+MCu3+) (0 2->, or as YBaiClIz+h(02-MO-j.

" r
..
_,_

r
o

--c<lhY03 O'l,o.. '"

~
:

Cn)--o

01

0:"

~lt:I

Figure 21 'Thr CI)")la\ JtnK1ure of l"Ba,.C'l:J0 "hO"o"itlK primiti,"1,' ll The planes ('Onlalning I1WJ"Iced Cu2, 02, .. w.l 00 the conducting pathwa}-s.

rom.

In the TlBaCaCu compounds, To; inCH.>ast:s \..nlh tlte number of sht.'ets of CuOz stackc..'(] u:mtinuously in a elyslal ('ell: thus fur the st:ries (Cu02J13aOiJ10 2 , (CuOJiCaBaOz110z, (Cl102hC~Ba202TI202 the T<;s arc 82. J09, 122 K. lsostruttural compounds where Hi rt-,placcs 11 and Sr n'1,1al.'CS 8.1 h..l VC Tc :: 10. 90, 110 K for the one, two, and thrcc laycr compounds. All these la)cr compounds arc strongl) anisotropic in electrical properties mcasurccl p.1rallcl Of" pcrpcndil...'lIlar to tile layers. II is unrealistic in :\11)' approximation to Ilt.ogk ..(;t the anistropy of tlte normal or liUpCrt.'OIl<.Iuc:ling properties.

Critical Fieltls aml Critical Currt'tltll

Critil-al field and cum.." !lt values in IITS may be quite uncertain, because in anisotropic media thc measuremcnts arc difficoh and pCI"haps subjed to large rccvah"1.tion for giant flux creep timc effects. We give reasonably acl.'eptct.l values ofthe critical p.1.ramctcrs for comp.1.risOIl with COl1ventional mah.... rials. Iligh T" lmggesls a 11igh stabilization CIlCl'gy alld high enCl'gy gap for tile sUI)(.'f't'Ondul.1ing slalc, from which follow high H" (from Eq. (9 and short coht.'ft.>ncc lengths ~ (from Eq. (I7. Taken in combimltion witl. a normal pCnt.lralion dl.1>th from Eq. (143), these results lead to extrcme type 11 behavior; high Ie in Eel. (371,) and very high f-ld. in Eqs. (31) and (1.25). At the high lelllpel-aturt.'S al.t.:essiblc with ill'S materials, Ihennally activalt.'(! (Tet.,) of

nUXOflS may limit L J.blc l'tlTTent values; thus the achiL'vclllent of high T" does not assure that a given material will IUl\'e useful high CUTrt:nt applications. In the low tcmj:x:rature limit the London penetration depths are )"..1, _ 140 nlTl and A.:: - 700 nm, frolll magnctization measurements; the l'Ohert.. '"Ill'C It:ngth ~b'" 1.5 nm from critk111 ficld measurementS. TIle Cinzburg-Landau pnrameter Kat> "" AaJ:iaI,'" 100, which amply satisfit:S tlte cTiterion K> 1 for a type II superconductor. TIle coherence Icngth nomal to the CU02 planes is ", ... 0.2 ~ 0.6 nm, of the order of lhe intcTplauar spacing itself. III thin mills the c..Titical cuTTent dcnsityat 4 K isjol> = 1.2 X 101 Alero' in the ab plane, andj" =0 4.2 x I~ Alon' in the C dircclioo, suggestive of a quasi-2D supc..'1'COnductor with planes l'Ollllo...'tOO by Josc..'Phsoli tunnding. Criticnl eUTTents in cc..'r.Imic pOwtlers are mudl smaHer, typknlly l~ Alun', From jab we may infer H<;'l(ab) ~ 107 C = 1000 T; values up to 100 T have 1x.'Cn meIL~uTt.:d in YDCO. In n:uTO\V brK.lges (50 nm) mudl higher (:ritk-al current densities have been obst.,'Tvcd. l>cing limited by the dcpairing of Cooper pairs. A bridge is narrow if no nUXOlIS fonn within the volume of the bridge, and the CL parameter IjJ lila)' be taken as const:ult within the bridge. 11lell (1.10) bt.."l'Omes

-a

+ {J1jJ+

ln~

"" 0,

(62)

where the la.~t tenn is the kinetic energy of the Cooper pain;, each uf ma.~s m. TIle l'IIl'r..' lt density is
j:::

2e11/Jl2 v =

(2ell/ll2Iml'~2a -

21311/Jl2 )III!,

(53)

which is rt mUll:illlum with respect to I::: 1lJ.f~~:t when I" ::: Y(2I3). Ilere I/Jo is the CL prtr.uneter at zt.:rtJ CtITTent. nus the maximum (dcpa.iring) current density is

the critical dep<lirillg velOCity Vc - Klm{. For)" - 100 nm amI { ... I have Jc'" 1.3 X lotI AI('fl,2 in good agrcemellt with experiment.

111ll,

we

llallNllmbcr
TIle Ilallllumbcr is de()nt.'ll as the inv(.."TSC Iiall constant llR,,c normalized to the fonnuL'l volume. If onl) holes or e1edrons conduct, as in simple mctals, the Hall number gives the carric..,. (;onccntmtion per fol'milia volume_ In Y8CO the Iiall number is p-t)1)e and J>rofK'l1ionai to temperature for fIDe, for T somewhat above T", and is of the ordeT of unity at 300 K, giving a (:anicr concentratiOn - 7 x liP em -3. FllUercuc, 11lcsc are stable, cagclike molecules th.. 'lt COnstitute the third fonn of pure carbon: the other two are diamond and graphite, ne archetype fullen..'lc is

'7<

Figure 28 Ceo fullenene molecules CT>"'-alliu in a fl,re.cmtcn:d (tiMe stnlC1u...... Court<'Sy r.:J Ste-'e1l louie.

Coo; coldl molCClllc has the fonn of a lrunroloo kos.,hecJron \\ilh 20 hex"II,'OIla1 foc'CS and 12 pentagonal f.'1CCS. like a soccer ball. lhis is one of the 13 Archirncdean solids CharactCriZl."t! by having all their angles CtIUal and all their Eu..'cs
regular polygons. CliO crystallizes in a f.1Ce-<;'(:nh:.'f"(',(! cubic strudurc. as in Figure 28. Alkali-fullcrcnc (.uml>Ounds su<:h as K:J,C1;o arc supcn;omluc.:tillg; Ihis one has T c = 19.2 K. TIle K alOlns (K:CUpy thcoctah(."(\ral sites in tilt: cubic t"ClI. HhCs 2 C OO has T c = 33 K.
SVMMAHY

(In CGS Units)

A superconductor exhibits infinite condudh it)'. A bulk spc.:cimen of metal in the supcl"COlltluctillg st. ..tc exhibits perfect diamagnetism. wilh the m.,gnctic intluctton B = O. 11m is the McisSlll..' r effect. The edernal magnetic field will penetrate the 5urfat'C of the specimen l.'!'cr a distance ddcrminro by the penctration depth A.
"Incrc arc two t}'JlCS of slJl>cn:onductors, I :Uld II. In a bulk spedml.'tl oft}'I>c I SU!>Cf'(;olldul'tor the supcn:onducting state is dt..'Stro)'oo :Uld the normal state is rcston..-tl by application of all cxlcnlal magnetic field ill CXl'CSS of a critkal

\I:lluc lie. A t)'PC II supcn.'tJllductor hn." twol'l'itical fid

fl el

<

lIe

<

Hdl;:l

vortex state t.'Xists in the range betWt.'CIl lid :uJdllc2 , nle stabilraltion l.'tlt.'Tgy uensity of the pure sUllCrconuuding slate is l~rr in both type I and II supereonUUl"Iors. In thc SUI>en:Onuueting state:Ul cncrgy gap, E~" 4kB 1',; separates supen.undueling ckoctl'Ons below from normal dcctnJlls above thc gap, The gap is dctccted in experimcnts on heat l'apadt)" infrart.>t! ahs<wpUon, and tunneling, 111C London c(Ju:ltion
J~---A

41TA~

01'

curl J = - - - , B 41TA,.

Icads to thc t\lciSSllcr clreet through the penctration equation V2 B = BIA~., where Ar_.. (mCI4'1rllC)II2 is thc London pt."netr,ltiOIl tlt..1Jth, l11rt.'C important lengths CfItcr the tht.ory of supcn.Ulluudivity: the Lonoon penetration uqJth At:. the intrinsic coherencc Icn!,>1h 6>; and thc normal electron mean free path t. In the London equation A OT B shoulu he a weighted average (}\o'er the coherence length {. The intrinSic cohert.'11l'C ICllb'lh a = 2Ii&,JttEc' l11C DeS theory aa.uunts for a supcn:omluding state fonned from pairs of electrons I.: t and -I.: t, 111csc pairs ad. as OOsons.
Type II superconductors h.'we {< A, The lTitk.-aI fields aTe rclatC(1 by lId .. (gA)H c anu llc2" (AlfJHc- l11C CillwurgL.,ndau parametcr K is defim.-d as Ai{. Valucs of Il c2 aTe as high as 500 kC = 50 T,

Problems
l.
M(lg",~iic field

penetration in 1I. plote. 111C penctration I,.'<juation may l>c writtcn as AzVzn _ n. WIICI"C A is tile penetration depth. (a) Show that n(x) insidc 3. supercolldueting platc pcrpendicular to tllc x a;(i~ and of thickncss [; is giVCII by
cosh (xlA) B(x) "" B. d)S11 (8I2A)

where B. is the field outsKie the plate and parallel to it; here x ~ 0 is at the l't:llter of the plate. (b) The effective magnetizatjoo M(:I) in the plale is dellncd by n{,) - B .. 4m\l~). show th.,t. in CCS. 4'ITM{x) - -BJ.II8Aty.lf - 0. for /j c:g A. In SI we noplace the 4. by llo-

2. Cri,icol fleld

0/ 'hi" fiInu.

(a) Using the

~ult

of Problem lb, show that the free

energy dCllsity:lt T "" 0 K ...ithin a supt..'rCOndud:ing llIm of thickness /j in an CAtCmal magnetiC llclJ B. is gin~n by. for /j <t A,

(CCS)

In 51 the factor TT is n:plaet.'d br!J.<o. We negled: a kinetic energy contribution to the problem. (b) Show that the magnelic contributiOn to Fs when averaged O'oW the thidrneu or the film is FfJ,&A)'19fnr. (c) Sho....that the entkat field of the thin film is proportional to {A1li)He> when: H~ is tlle bulk critical field. if we consider ally the magnetic contributiOn to Us

3. Tu-f1uid modd of (! wperoonductor. On the two-nuid model of a superconductor "-e assume that at temperatures 0 < T < T~ the ClIrrent density may be Written as the sum of the contribmions of oonnil and superconducting electrons: i - j..., + is. where 1\ - uoE and is is given by the London eqUiiltiOl'l. lIere Uo is an ordinary nonnal conducti\'ily, decn:ased by the rcductiOO in tile number of normal electrons at temperature T as eotnpared to the nonnal state. Neglect Inertial effects 01'1 bothh-' andJs. (a) Show from the Maxv.-eD equations that. the dispersion relation connecting wa\'evedor k and frequency w for electromagnetic waves in the superconductor is
(CCS)

.,

(51)
where Al is given by (148) with " replaced by Os. Recall that C\lrl Cllr! B = -flit (b) If l' is the rel8lliiltion lime of the normal electrons and "N is their concentration,

5110\'0' by use tile expressioo Uo .. ",,T!m that at frequencies IN <t lIT the dispersion relation does not involve the nannal ek'Ctrons in an important way. SO that the motion of the electrons is desc.ribcd by the London equation alone. 11le Supcrctlrrcnt short-circtlits the nonnaJ ek'Ctrons. 11le London e(jualion itselfonly holill tmc if fiw is small in comparison with the energy gap. Note; TIle frequencies of interest are such. that w <t Wp, wl'cre w,. is the plasma ftC(IUCncy. "'4. Structure of (! vortex. (a) Find a solution to the London equation thai has cylindrical symmetry and applies outside a line core. In cylindricel polar coordinates, we Wiillli a solution of

or

that is singular at lhe origin and for WlliclJ the total flux is the nux (juanlllm:
2:7'

dp pB(P) - $ll .

111e equation is in fact valid olll~' outside tile normal core of rndius f. (b) Show that the solution has the limits
B(p) "" ((I'21l'A~ln()Jp) , B(p) "" (4V2'1TA~:7'A12p)1I'lexP(-{iA)

"This problem is somewhat difficult.

..
5. London r.ICndrot'OI. ~Ih. (a) Take the time derivative of tI,e London equation (10) to show Ihat iJj/ilt .., (c/o1?TAi)E. (b) If mdvlrlt - qE, as for free carriers of charge q and mass m, sI,ow that Ai = 1I,cIo1?Tnqt.
~.

. ....

Diffraction rff~ of joupllMm junc'ion, Consider a junction or rectangular cross ~on wilh a magnetic field B applied in the plane of the junction, normal to an edge ol ....idth w. Let the thickness ollhe junction be T. Auume for convenience that the phase difference of the two superconductol"$ is ?TI'l whl..'fl B = O. Show that thc de CUrTl..'Jlt in the prescnce of thc magnetic field is

i-in

sin(wTBellic) (wTB.,/!ic)

7. Mrismo:r rjJect in ~ere, Comidcr a sphere of a type I superconductor with critical flCld Ho:- (a) Show that in the McWner regime the effective mag:neti7.ation M ....i thin the sphere is gi\'en by -8Trl1fl3 - B.. the uniform applkd magndic field. (b) Sho..... that the magnctic field at the surface of the sphcre in the cquatorial plane is 3BJ2. (It follov.'S that the applil..'<! field at which the 1>leissner 1..>ff1..'Cl: starts to break down is 21f/3.) Rcminder: lbe demagnetization field of a uniformly magnctized sphl..TC is
-hTMI3.

References
It D. Parks. ~-d . Supvcot,ducUtJity, Dekker, 1900. Good rol1ection ohevicw articles. R. SchridTer. TI,eory of S1l/.>r:rcvndttCfiviry, ",,-is..d ..-d., Benj"min/Cumming,:, 1983. C. fuclo.y.u:n, TI,w'1l of .UJ!f'rro.ulucdlll'!I, )ntenricnce, 1965. M. 1'1llkh:am, lntrodudion 10 S1lperamduclh;i1I1. Kri<1;er, 1960. M. Cn>an, ThrorJI of "'p"..r ondud;Lwy, World Scic1.tif..,. 1969.

J.

n'PE II SUPERCONOUCJn'ITY
C<:tlllC$. Suprn:onduclivity of ....:tllh Dnd IIflolP. AdditlO<,.Wcdey, 1981l. T. Luhman and I). Dew-Hughes, cds., MttD/lurgy of $U~rro"d!JCIb<g"'D'~ritlh, Academic-, 1979. M. K Wil",n, S"/.>r:n:v"ductlng ""'gMf., OxfOrd. lllllJ. D. Saint-J"'".."" E. D. 'lllOwa>', :and C Sam..... TIIJ.>r: Illullt'rrond"CCivitll. Pc'W'mon. 1969. R. P. lIeubener, MDf/Il!l1t;f/ml Urnduru ill S1lJ!f'rctlMUclou, Springer. 1979.

P. G de

HIGIl n:MPERATURE SUPEROONDUcrORS C. Bums. l1igh-u:n1p.o:rrltun ~rro"dU(1lci1y, An.demi<-, 1991. M. Tinlh:am aod C. J. LollI" ~M1ysicl1 properties oflhf: new suprn:ond"cton, Solid stale ph)'sics
R

R. D. R. Huffman, "Solid Cwo" Ph)s;cs Today, No. II, 22 (1991). Summary of php;iaol properties of 1I~ buckminSI"rful!c",n..' S.

,I,

<12, 90 (1980). Curl and It E. Smalley, w"'uUerct..,.:' Sci. Am. 1'\0. <I. p. 54 (1991).

JOSEPHSON EHECTh

Phricr ."ul "PVlit:r:lioru (Jf tile J~n ~:t, w.ley, 19liZ. E. L Wolf. Princi,lu of dteI,..". tunndirJ SJXdroa:ow, Odord. 1965. J. Callop, SQUIDS, flleJ~hMm rffeee., Dnd w~rwnducl;ng dfrona, Hilgeo-, 1991. J o"rkt' and It II. Koch, Rlmpact ofh,gh-tempenlturc superconductwity on SQUID rnagnl!tomekrs, Sciencc 242, 217 (lOBS).
A

Barone and G. Paterno,

c.

13
Dielectrics and F erroelectrics
M:U:\\-'ell equations 381

l'olariu.tion MACROSCOPIC ELECTRIC FIELD Depolarization fJeld. . LOCAl. ELECTRIC FIELD AT AN ATOM Lorentz fteld, E,; Field of dipoles inside ca\ity. ~
DIELECTRIC CONSTANT AND POLA.RlZABILITY Electronic poIarizabilily

381 3liZ

3M
366

388
388 388 390

STHUcrURAL PHASE TRANSITIONS

n:m{QELECfRIC CIIYSTAI,S Classification of ferroelectric cr)'stals

393
393 394

DlSI'LACIVE TRANSITIONS optical phonoos Landau throry of the phase transition Sond-onier transition First-order transition ArItiferroelectricity Ferroeledric dornail15

Son

398 400 400 402 402

404

406
408 410 410

Piezoe~ricity

Fenoelasticity Optical ceramics

SUMMARY
l'1\081.M5 1. 2. 3. 4. 5. 6. 7.

410
411

Atorok hydrogen Conducting spllcre Effect of air gap Interfacial polariulion Polarization of sphere Ferroelectric critt...Tion Saturation polarization 8. Dielectric coostant below transilton
9.

411 412 412 412 412 412 412

413 413

Sort modes

10. Ferroelectric linear array

I'\OTATION: (CCS)
41 - JOl/4 m"

414

D ... E

+ 47fi' .. a

- (I + 4'11'X)E

a = ,JJo<.j ;
a=p/lo<o.l ;

(51)
t(;es =
t'SI ;

('igu~ 1 Tho: I'Cnnancnl dipole mome.." t of a molecule of wat..,.. has the magnitude 1.9)( IO- la eiu-.;m ."d is directed rrom 0-- ion the midpoint olhc Une con-

lo\.o-aro

ntttiRl the II' ionl. (To C'OII\'m to SI units. mohiPY b) I x 10".)

, -<'------------.-.

Figure 2 Elt'Ctrostatic potential and rlCld compo....',!, in CGS at po$itioll r, 9 lOr a dipole p diTed<.-d alon! the % lUis. 'or 8 - O. we ha,'c E. - E. - 0 and E. - 2pI,;s; for 9 _ 'fff2 we have E.. - E. - 0 and E... -plr. 1b COIwerl 10 SI. repI:;oce p by p/4fft;>. (After E M. Pureel!.)

CHAPTER

13:

DIELECTRICS AND FERROELECTRICS

First we relate the applied electric field to the internal electric field in a dielectric crystal. nle study of the electric field within dielectric matter arises when we ask: "'hal is lht' relation in the matt'rial between the dielectric polarb.ation P and the macroscopic electric field E in the Maxwell equations? What is the relation between the dielectric polari7..atioo and the local electric field which acts at the site ofall atom in the lattice? The local field determines the dipole moment of the atom.

Maxwell Equations
(CC5)
4r. curl H = - j

(5I)

(;

+ - -(E + 4r.P)
c at

curl I-I :: J curl E =

. + -lEoE a + P)
at

curl E:: - - -

au

au
at

, at

di\' E

= 4r.p

diVEoE

divB=O:

~ p; cliv B ,.. 0 .

PoiariUltion
The polarization P is defined as the dipolc momcnt per unit volume, averaged over the \'olume of a cell. 111(' total dipole moment is defined as (I) \\hcre r .. is the position vector of the charge q ... nle value of the sum will be independent of the origin chosen for the position vecton, provided that the !>)'stcm is nt'lltraL Thc dipole moment of a watcr mok'Cllle is shown in Fig. 1. The electric field at a point r from a dipole moment p isgivcn by a standard result of elementary electrostatics:

(CC5)

E(,):

3(0,~- r'p ;

(51) E(,) -

3(0';':; r'p. 4.

(2)

-n1e lines of force of a dipole pointing along thc z axis arc shown in Fig. 2.
38'

MACROSCOPIC EL.ECfRIC fiElD

One contribution to the electric field inside a body is that of the applied electric field, defined as

Eo fit'ld produced by fixed charges extt-mal to the

bO<i';J

(3)

1lle other contribution to the electric freld is the sum of the fields of all chargt'S that constitute the body. If the body is neutral. the contribution to the average field may be expressed in terms of the sum of the fields of atomic dipoles. We define the average electric field E{ro) as the average field over the \'olume of the crystal cell that contains the lattice point ro:
E(ro) = -'-

v,

dv c(r)

(4)

\,here e(r) is the microscopic electric field at tlle point r. nle field E is a much smoother quantity than the microscopic field e. We could well have written the dipole field (2) as e(r) because it is a microscopic unsmoothed freld. We call E the macroscopic electric field. It is adequate for all problems in the electrodynamics of crystals provided that we kllOw the connection between E, the polari7.ation P, and the current density j, alld provided that the wavelengths of intert.~;t are long in comparison with the lattice spacing. I To find the contribution of the polarization to the macroscopic Aeld, we can simplify the sum o,'er all the dipoles in the specimen. By 11; famous theorem of c1ectrostaticr the macroscopic electric field caused by a unifonn polarization is equal 10 the electric field in vacuum of a flct:itious surface charge density

IA ddaik-d dl.'Ti,,,,lioll of Ihe Mll-'<well l.oqllal;ons for the IlllOCroompk fields E and 8, stal'ling from the MaxweU l.oquations in Icnns mlhe min'o5ropk "e1ds e 1lI,d h,;s giH'n by E. ~1. P"rcdl. /1001'1 'JIId IlttIgJtdis..... 2nd cd. McCraw-HIll. 1985, IJbe ckdRbtOltic pottlllia1 in CGS unil$ of II dipole P is: <i') ... p' for. ,'OIume di5tnbution of ~ P we lw,-e

grad(I''''.

Wllich by a "lI!Clor ident;l)' b<.'<'Omes

~r)" dV(--;'d;vP+d;v~)
If P is conRant, then dr.. P - Oll',d by the eaus. theorem "" ba,-e

<f{r)-fdS~"-fdS:. whcr-e JS i$ an dement of area on the wnace of the body, lllis rompletes Ihe proof.

Q.
U

u ... t'

tr -I'

''''

(hJ

FiP'ft 3 (a) II uniformly ~ ffide.ctrir, ~, ",ilh the poWiution vector- P normal to the plane of the ,ab,. (b) II pair or unibmly char-grd p;Jn1...1 plat"" \\hllCh gl\e rise to the identic.'al dl'lhic 6dd E, .. in (a). Th.: upper plat" has the wrface cb;o~ denslt" u - +P, lind the 1ov.-er .>hte 1_ u - -Po
:::: Ii, P on the surface of the body. Ilere Ii is the unit normal to the surface. drawn outward from the polarized matter. We apply the result to a thin dielectric slab (Fig. 3a) with a unilOnn \'olume polari7..ation P. TIle electric field E,(r) produced by the poLvization is equal to the fidd proouced by the 6ctitious surface charge density u :::: Ii' P on the surfiKc ofthc sL<tb. On the upper boundary the unit vector Ii hi directed upward and on the lower boundary Ii is directed downward. The upper boundary bears the fictitious charge u :::: Ii' P = Ppcr unit area. and the lower boundary bears - P per unit area. The electric field E. due to these charges has a simple rorm at any point Ix:tween the plates. but comfortably removed from their edges. By Causs's law

(CGS)

1 = -4njul :::: -4r.P ;

(51)

E,

= lui
Eo

__ c

P
'"
J

(4a)

We add E 1 to the applied fidd Eo to obtain the total macroscopic field inside the slab, with z the lInit vector normal to the plane of the slab:
(CC5)
(5)

(51)
We define E1

-, E,
p
011

6e1d of the surface charge density Ii' P

the boundary

(6)

This field is smoothly "arying in space inside and outside the body and satisfies the Maxwell equations as written for the macroscopic field E. The reason E, is a smooth function when viewed on an atomK: scale is that we have replaced the discrete lattice of dipoles Pj WIth the smoothed p:llari7AltiO'll P.

Depolari:wtion Field, E.

The geometry in many of our problems is such that the polari7,.ation is uniform within the body, and then the oilly contributions to the macroscopic field are from Eo and E 1 :
(1)

Here Eo is the applied field and E 1 is the field due to the unifonn polarization. TIle field E, is called the depolarization field, fOT within the body it tends to oppose tbe applied field Eo as in Fig. 4. Specimells in the sllape ofellipsoids, a class that includes spheres, C)'linders, aud discs as limiting forms, ha\'e an advantageous property: a unifonn polari7,.ation produces a unifonn depolarization field inside the body. This is a famous mathematical result demonstrated in classic texts on electricity and magnetism. 3 If p.. , p,. P: are the components of the polariUltion P referred to the principal axes of ellipsoid, then the components of the depolarization field are written

an

(CCS) (SI)

Eh

= -N,P.
N:J>",

1,

= -N,P,
'0

L:. = -N:P: ;
,= =--

(8)

t.

Here N"" Nil' N: ratios of the principal axes of the ellipsoid. The N's are positive and satisfy the sum rule N, + Nil + N: :::: 41T in CCS, and N. + Nfl + N: == I in SI. Values of N parallel to the flgure axis of ellipsoids of revolution are plotted in Fig. 5; additional cases have been calculated by Osborn~ and by Stoner. In limiting cases N has the values:
Shape
~phere

'" are the dcpolan7,.ation factors; their values depend on the

'0

E'II =

_!!lJ.

_ N:Pr.

Axl$

N (CeS)

N (51)

111in. slab 111in slab LOlig cin.'ular cylinder Long circular cylinder

"y normal in plane longiludillal transverse

,.

417/3

1/3
1

0 0 2.

0 0 112

We can reduce the depolarization field to 7..ere in two ways, either by working with a long fine specimen or by making an electrical connection between electrodes deposited on the opposite surfaces of a thin slab.
4J. A. Osborn, Pb~"S. Re\. 67, 351 (1945), E. C. SIOfl('r,
~It Bedeo-. Etf'(:t'OtIwfYtdic~lds /lfId ;"'eracti<ltu, B4bdell, I9&!, 1'(\. 102-107. PhIIosoph.IcaI MllgllZif'M' 36, &l3{19-1.S}

~~

:E+ ~_~=+ + +
-

FIgUre <I 'l1le depolari:r.:ltioll fleld E, it OI,lllO<ilc to P. 'the fletitiou> 5Url'act: charge. are indkstl'd: the field t:i t110e d>iul:;cs Is E, "ilhin Ih" eUipsoid.

"" " "

, . <

"

"

0.'

"

..

o
Figure 5
J)epol"';~1tiol1

d. factor N panollt-l to the figure axis of cllipsoio.b of 1'C\'Oluholl, as II rllnctioll

,!---!---!-,~:,~=+.==.!==J,

..
"'
0

" "<

of the axi;LI ratio do.

A 1I1'liform applied field Eo will illduce unifonn polarization in an ellipsoid. We introduce the dielectric susceptibility X such that the relations
(ees) P = XE ; (51)
p= <OXE

(9)

connect the macroscopic field E inside the ellipsoid with the polarization P. Here XSI "" 4nXccs If Eo is uniform and parallel to a principal axis of the ellipsoid, then

(CCS)

r:

= ~

+ 1 = Eo -

NP ;

NP ,51) E-F-o--

(10)

'"
by (8), whence

(eGs)

p =

~F.o

- :.\'P)

p=

1 + NX

Eo;
Eo.

(II)

(51)

P = x(EoEo - NP)

p=

x<o
1 + NX

1lle value of tIle polarization depends on the dt"p.)larization factor N.

LOCAL ELECTRIC FlEW AT AN ATOM

loe value of the local electric field that acts at the site of an atom is significantly different from the value of Ole macroscopic electric field. We can convince ourselves of this by consideration of the local field at a site with a cubic arrangement of neighbor;' in a Cf)'stal of spherical shape. The macroscopic electric field in a sphere is
ICCS)

E=

Eo + E 1 = Eo - -

4~

(12)

(51)

E=Eo+E1=E:o--1_p

3<0

by (!O). But conside.- the field that acts on the atom at the center of the sphere (this atom is not unrepresentative). If all dipoles are parallel to the z axis and have magnitude p, the z component of the field at the center due to all other dipoles is, from (2), ICCS)

(13)

In SI we replace 1) by lJ141TEo. The x, y. z dirt.'Ctions are equivalent because of the symmetry of the lattice and of the sphere; thus

whence F-'d;~ = o. The correct local field is just ~ual to the applied field, Eloc.l = Eo. for an atom sile with a cubic environment in a spherical specimen. Thus the local field is not the same as the macroscopic average field E. We now de"'elop an expression for the local field at a general lattice site, not necessarily of cubic symmetry. The local field at an atom is the sum of'the electric field Eo from external sources and of the field from the dipoles wiOlin the sIX'Cimen. It is convenient to decompose the dipole field so that part of the summation over dipoles may be replaced by integration. We write
(14)
"Atom siln in a CIIbic Cf}'S1al do not ~ han cubic lymmrir)~ ("'1$ Ihr 0"- iii" in the banum tilanate slructure oll'ig. 10 do not ha.."fl. cubic cnvironflll.'flt. HO\O'eWr, lhe and liles '" the NaCl shuClUrc and the Cs al\<I CI- sHes in Ihe esCl S!rUclun: ha,'c cubic S)"Inmd.ry

N.

a-

13

n~fuia

mid

f~r~I'rin

~,

"guN: 6 iutc",,,1 dc.'d.ric: focld on an atom in a crpta! is the Iwn of the eo.tcmallI(lfll~ rlC'ld of the field dUC' to the ochn- atOll1S in the Cf)')b1. The standanl mcthcld of summing the dipok fK'1ds of the othc'r atoms is fint to lum iooj,dclually O'I-a" a rnodcnlc numbrr of nrighLoring atoms insidC' an imaginary spbC'n ~ ..~th the refcn:ooe atom, this dcfillC'S the flCk! ~ ",hich ~-anilok:s at a ~ sitc ....; th eubic S)'1ll'1IC'tJy. _101M outside thc sphn-c .."..1 be treated as a unibln1r polariu-d didcrtric 1bea- contribution to tbe f",k! at the ~ point ~ E. + Ez. ",hcre E , is the depolarization 6ck!auociatoo ...;th the onte. ... bouncbryancl Et II the fidd associated wilh the surface of the s~ .,...ity.

Eo."

n.e

n.c

Ilere

Eo : field produced by fixed dlarges exlernal to Ihe body;

E t = depolarization field. from a sutface charge density i\' P on Ihe outer sutface of the sp<,'cimen; 2 : Lorentz cavity field: field from polarization charges on inside of a spherical cavity cut (as a mathematical fiction) out of the slx..c imen with the rererence atom as center. as in Fig. 6, E 1 + E 2 is the fidd due to uniform polari7..lltion of the body in which a hole ha.'i been created; 3 : field of atoms inside cavity. 11le contribution E 1 + ~ + E3 10 the local field is the lolal fidd at one atom caused by the dipole momenls of all the other atoms in the specimen:

(ees)

3(Pi . rllt. - .fPI

-7

([5)

and in 51 we replace Pi by p/4~ Oipoles at distances greater than perhaps ten lattice l'Ottstants from the reference site make a smoothly ....arying conlributioll to this sum, a contribution which may be replaced by two sutfac:e integrals. One sutface integral is taken over the outer sutface of the ellipsoidal specimen and defines E lo as in Eq. (6). l1le second sutface integral defines &.z and may be taken over any interior surface that is a suitable distance (say 50 A) from the reference sHe. We count in 3 any dipoles not induded in the volume boundt:d by the inner and outer surmces. It is convenient to let the interior sutface be spherical.

Lorent1. Field.

The field E 2 due 10 the polarization charges on the surface of the fictitious cavity was calculated by Lorentz. Jr (J is the polar angle (Fig. 7) referred to the polarization direction, the surface charge density on the sunace of the cavity is - P cos 8. The electric field al the center of the spherical cavity of radius a is
(CGS)
E2 =

r
o

(u- 2)(211u sin 8)(a clfJ)(P cos 0)(005 0) = -

4n p 3

(16)

(SI)

This is the negative or the depolarization field 1 in a polarized sphere. so that ) + E 2 = 0 for a sphere.
FieW

of Dipoles

Insule Cavity, E 3

The field E 3 due to the dipoles within the spherical cavity is the only term Ihat depends on the crystal structure. We showed for a reference site with cubic surroundings ill a sphere that 3 = 0 if all the atoms may be replaced by point dipoles parallel to each other. -n.e total local field at a cubic site is, from (14) and (16),

(CCS)

EJo<.aI

= Eo +

E1

+ ""3 P = E +""3 P
+ -- P
3'0

471

471

(17)

(SI)

EIO<"al = E

This is the Lorent"L relation: the field acting at an atom in a cubic site is the macroscopic field E of Eq. (J) plus 471P/3 or P/3Eo from the polarization of the other atoms in the specimen. Experimental data for cubic ionic crystals support the Lorentz relation.
DIELECTRIC COX-STANT AND POLARlZABlUTY

The dielectric evnstant E of an isotropic or cubic medium relative to vacuum is defined in terms of the macroscopic field E:

(CCS)

+ 471P
E

=1+471X

(18)

(SI)

EoE + P

.0

"""Ii-X_

Remember that XSI = 471XCCS. by definition, but ESI e ECCS'

..

.,

FIgure 7 Cala,lalrn ol the r>c\d it" formly polarized m~-di"m. liph<:rical ca"i'y in a "ni-

+ +

+ +
C1.... !U. ,... in,:

2lfR _in

eR ,I~ p"" (J

"Ole susceptibility (9) is related to the dielectric constant by

(CGS) X = - = - P E- 1
4~

(S!)

(19)

In a noncubic crystal the dielectric respWlse is described by the components of the susceptibility tensor or of the dielectric constant tL11501'": (CGS)
(SI)

(20)

nle polarizability a of an atom is defined in terms of the local elt:ctlic field at the atom: p ::: aElox.aI , (21) where " is the dipole moment. This definition applies in CCS and in 51, but aSI == 41fEoO'ccs, l1lC polarizability is an atomic property, but the dielectric constant will depend on the manner in which the atoms are asscmbk-d to form a crystal. For a non-spherical atom a \\;1I be a tensor. The polarization of a crystal may be expressed approximately as the product of the polarizabilities of the atoms times the local elt:ctric field:

P=

L Njp) = L NjoJE.....JJ)
j ,

(22)

where Nj is the concentration and aJ the polari:t..ability of atoms j, and EnU) is the local field at atom sites j. We ....-ant to relate the dielectric constant to the polariubilitics; the result will depend on the relation that holds between the macroscopic electric field and the local electric field. We give the derivation in CCS units and state the result in both systems of units.

If the local field is given by the Lorentz relation (17), then

(CGS)

p=(INpJ)(E +

p)

ami we solve for P to find the susceptibility

(CGS)
x~-=

'WC""' .e'c__ --

--'W.

4r. 3

(23)

' ,

By definition

= I
~

+ 41J"X in CCS; we Inay rearrange (23) to obtain

(SQ
(24)

E - 1 (CGS) - E

+2

4r. -I.No}

the Chmsius-Mossotti relntiOIl. 'Inis relates the dielectric constant to the electronic polari:tability, but only for crystal structures for whieh the Lorentz local field (l7) obtains.
Electronic Poiarbability

TIle total polarizability may usually be separated into three parts: eketronic, ionic, and dipolar, as in Fig. 8. The electronic contribution arises from the displacement of the electron shell relative to a nucleus. TIle ionic oontribu+ tion comes from the displacement of a charged ion with respect to other ions. The dipolar polarizability arises from molecules with a pennanent electric dipole moment that can change orientation in an applied electric field. ]n heterogeneous materials there is usually also an interfacial polarization arising from the accumulation of charge at structural interfaCt:s. TIlis is of Iiule fundamental interest, but it is of considerable practical interest because commercial insulating materials are usually heterogeneous. 6 111e dielectric constant at optical frequencies arises almost entirely from the electronic polarizability. The dipolar and ionic contributions are small at high frequencies bewuse of the inertia of the mok-cules and ions. In the optical I1lnge (24) reduces to (CCS)
11 2
II

4'JT

I.Njo)electronic) ;

(25)

here we have used the relation n 2 = E, where" is the refractive index. By applying (25) to large numbers of crystals we dctcnnine in Table I empirical values of the electronic polarizabilities thai are reasonably consistent with the obsen'ed values of the refractive index. 111e scheme ;s not entirely
61"or
rer..,rel\c~s

lit,.,

D. E. AspnL"S. Am.

J. I'hys. SO.

7Q.I

(1982).

--:;.f--------"-lgUre 8 Fn:<ltl.,~ depcnd<--nce of the

M. . ..... ra1

_1:~~

__,- /

coutributiolU 10 the pol"nr.ab'lit\'.

Table 1 Ekdronie polllriz.abililiC5 or ions, in JO-14 cm:l

II.

U'
0.029

....

B"
0.003

C"
0.0013

Pauling

0.201

0.006
),Igh 0.094

JS
Pauling JS-(TKS)

0'3.'"
(2.4)

F1.04
0858

O.fr29

N.
0.390
A<

l'a+

AP'
0.052

S....
0.0165

S'Pauling )5-("K5) Pauling )S-(fKS) Pauling )5-(fKS)

\"aJ~ froon

CI-

0.179 0.200

K'
0.83

10.2 (5.5)

3.66
2.947

1.62

"""
0.47 (1.1)

So"
0.286

nH
0.185
(0.19)

50'10.5 (7.)
Tel~

o.-

1.133

K.
~46

Rb'
1.<0
1.679

4.TI 1.091

S... 0.86
(1.6)

y"
0.55
1.."\"

ZrH
0.37
Ce~

,-

x.
3.99

c,'
2.42

na'
1.55

14.0

(9.1

7.10 6.116

1.01

0.73

2.743

(~5)

L.. Paulifll:, Pnll.'. R. Soc-. Londoo, All., 18t (1927), S. S. Jaswal.neIT. P. Sharma. A. Kahn, and W. Shockley. Ph~"$. Rc". 92. 890(195.'). l1lc TKS p<:>lanZdbilitie5 an: at the fn..q ucncyorthc I) liTles of sodium. The vaI...."J a..e in CGS, to to 51. multiply by , X 10-'''.

Ph~ ....

ale"" SoIids:w. 309 (19731; and J. T.....sm....

roll''''''

sclr-consistent. because the ch.:'Ctronic polarizability or an I ..... depends somel\ hat on the environment in which it is placed. The negative ions are highly polarizable because they are large.

CI08Sicnl TIleory ofElectronic PoiariuWility. An eh..'Ctron bound hannonically to an atom will show resonance absorption at a rrequcncy WI) = (JJIm)I12, where P is the roree constant. The dispw.'ement x ohhe electron occasioned by the application or a field E..... is givcn by

-eEIo

Px =

m~

(26)

so that the static electronic polari7.3.bility is a{electronic) = plEIcc = -eriEIcc = e21,,~ "
(27)

-me electronic polarizability will depend on rn:quency, and it is shO\\Tl in thc rollo\lling example that ror rrequency w

,r'm

EXAMPLE: Frequem:y dependerlce. Find the frc(IU<,ll()' dcpcndeocc of the electronic polarizability of an electron ha\'ing the resonance f".'<I\I("\('')' WOo treating the system as- a simple harmonic oscillator. 'nll'l equation of llIotion in tllC local c1l.'t:lric field 10< sin wt is
d~x

m -, >

,,- + m~ -

-et:k>< sin wi

so lhat, for x

= :to sin wI,

m(-w2 + ~)xo" -r:E"", .

"nlC

dipole moment has the amplitude

from which (28) fol1O\~"

In quantum th(,-'()ry the e):pressioll correspom:!ing to (28) is

(ees)

a(electronic) = -c' '" "--

JIl}wt- w2 '

f;

(29)

where Iii is called the oscillator strength or the electric dipole transition be+ t""(,-en the atomic statcs i and j.

.
~'RVcruRAL

PIIASE TRA"'SJTIOlliS

It is not ullcommon fOT crystals to transform from one CI)'stal structure to another a~ the temperature or pressure is varied. 11lC stable structure A at absolute zero generally ha~ the lowest accessible intemal energy of all the possible structures. Even thi:; selection of a structure A can be varied with application of pressure, because a low atomic volume will &''01" closest-pacK<.-d, or even metallic structures. Ilydrogen and Kenen, for e)."llmplc, become metallic under extreme pressure. Some other structure B may have a sorter or lower frequency phonon spectrum than A. As the temperature is inereased the phonons in B will be more highly excited (higher thenna! average occupancies) than the phonons in A. Because the entropy increases with the OCC\Ipancy, the entropy of B "ill become higher than the entropy of A as the temperature is increased. It is thereby possible for the stable structure to transform from A to 8 as the temperature is increased. The stable structure at a temperature T is deter mined by the minimum of the frt..'e energy F = U - 1'$. 'lnere will be a trami tion from A to B if a temperature T c exists (beJo..v the melting point) such that F...(Tc ) = FJI{Tr } Often several structures have ncarly the s.. .lIne internal encrgy at absolute zero. The phonon dispersion relations for the structures may. howe\'er. be mlher different. The phonon energies are sensiti\'e to the numlx:r and arrangement of nearby atoms; these are the quantities that dlange as the structure is changed. Some structural phase transitions have only small effects on the macroscopic physical properties of the material. However, if the transition is innuena."d by an applied stress, the crystal may yield mechanically quite easily ncar the transition temperature because the relath'e proportions in the two phases will change under stress. Some other structural phase transitions may have spectacular effects on the macroscopic electrical properties. Ferroelectric transitions are a subgroup of structural phase transitions. a subgroup marked by the appearance of a spontaneous dielectric polarization in the crystal Ferroclectrics are oftheordical and technical interest because they often have unusually high and unusually temperature-depl.'ndl.'Ilt values of the dielectric constant, the pie-l.OClectric effect. the pyroelectric effect, and electrooptical efft..'Cts. including optical frequency doubling.

FERROELECTRIC CRYSTALS

A ferroelectric crystal exhibits an electric dipole moment evcn in the absence of an external electric field, In the ferroelectric state the center of posi. tive charge of the crystal does not coindde with the center of negative ch3fge. TIle plot of polarizations versus ek'Ctric field for the ferroeledric state shows a hysteresis loop. A crystal in a normal dielectric state usually docs not

show signifICant hysteresis when the e1cctric field is increased and then reversed, both slowly. In some cl")1;tab the ferroelectric dipole moment is nOl changed by an de(.-tric field of the maximum intensity which it is possible to apply before causmg electrical breakdown. III these CI)'stals we arc often able to omer...e a change in the spontaneous moment when the temper.dure is changed (Fig. 9). Sudl crystals are called pyroelectric. Lithium niobate, LiNb0 3 , is pyroelectric ,,-t room temper.tture. Jt has a high transition temperature (rr = 1480 K) and a high saturation polari",.ation (50 JLClcm~. It can be "poled." "htch means gi\'en a remanent polarization, by an elcrtric field applit.od over 1400 K. Fermelectricity usually disappears above a certain temper,,-ture called the transition temperature. Above the transition the cr~ stal is said to be in a paraelcctric state. TIle term paradectric suggests .In analogy with paramagnetism; similarly, there is ul-tlally a rapid drop in the dielectric constant as the tempclature increases. Classijicalian of Ferroeuctric Crysta13 We lht in Table 2 some of the cl)'stals commonly (:onsidercd to be ferroel(:ctric, a100gwith the tranSition temperature or Curie point Tr at wllicll the crystal chang~ from the low temperature polarized state to !.he high temperature unpolarized state. 1'hermdl motion tends to destroy the ferroelectric order. Some ferroelectric crystals have no Curie point because they melt before leaving tJlC ferroelectric phase. The table also includes values of the spontane om polarization p. Ferroelectric crystals may be classified into n\o main groups, orner disorder or displacive. If in the paraelectric phase the atomic displacements olTe oscillations about a nonpolar site, then after a displncive trnnsition Ihe oscillations are about a p:llar site. If in the paraelcctric phase the displacements are about some double\\~11 or multi-well configuration of sites, then in an o.-der-disorder transition the displacements are about an ordered subsel of these \\ells. 11lCre has been a tendency recently to define the character uf the lransition in terms of the dynamics of the lowest frequency ("\>Off') optical phonon modes. Jf a soft mode can propagate in the CT)'stai at the transition, then the transition is displache. Jf the soft mode is only diffusive (non-propagating) there is really not a phonon at all, but only a large amplitude hopping motion between the wells of the order-disorder ~ystem. Many ferroeleclriC'S have soft modes that fall between these two extremes. TIle order-disorder class offerroelcctrics includes crystals with hydrogen Londs in which the motioll of the protons is related to the ferroek'Ctrtc properties, as in potassium dihydrogen phosphate (KIl 2 PO.J and isomorphous salts.

13

Vifectrin "nd l;ert'Of'fedriJ:.

''''''
~

~~

,"'" " . ,
,

"'"

, ,
100

OJ

T 4-l9'C T, --IW"C

(h)

, " 5
~

03

""'

'00
or

300

'00

..

Figure 9 TIle temperature ~.... iation (a) the dielectric constant e, (b) the p)roelectrie cocffidcnl dl'ln, and M th., specirJc heat c~. of PbTi03 . (After Remeika and Cia... )

Table 2 Ferroelectric crystals To obtain p. in the CGS unit of e!iU em- 2 , multiply the value given in poC em- 2 by 3 x lif.
To> In K

KDP type

KH 2P04 KD 2P04 HLII 21'04 KH 2As04

123
213 147

4.75

97 610
322

4.83 5.6 5.0

[96] [180]

[90]
[78]

TGS type

GeTe Tri-glycine sulfate Tri-glycine selcnate BaTiOa

KNbO,

2.8 3.2
26.0

[29]
[283[

295
408 708 71S

PbTiOa L.TaO a LiNb03

938 1480

30.0 >50 50 7I

[296] (523) (296) [296)

"A compilation of dala On rcrroclCClrk and anlik"'l"QCIL~trk materials is given arao, FerrocketriC'l 5, ffi7 (1973).

by E. C. Sub-

The behavior of cl)'stal~ in which the hydrogen has been replaccd IJy deuterium is interesting:
KHzPO'1

Curie temperature

123 K

KDzPO.j 213 K

The substitution 01" deuterons lor p.otons nearly doubles '1'<:. although the fral.'tional change in the molceulilr weight of the compound is less than 2 peTL"(!nl. 't11is extraordinarily large isotope shift is believed to be a quantum eITect involving the mass-dependencc of the de Broglie wavelength. Neutron diffradion data show that above the Curie tempemture the proton distribution along the hydrogen bond is symmetrically elongated. Below the Curie temperature the distribution is more conccntrated and asymmetric with respect to neighboring ions, so that one end of the hydrogen bond is preferred by the proton over the other end. The displadve class of Icrroclcctric:s includes ionic crystal stnlctures closely related to the peruvskite and ilmenite stTUc:tures. 't11e simplest ferrueledric l.Tystal is GeTe with the sodium chloride stru<.1ure. We shall devote ourselves primarily to crystals with the perovskite stnlcture. Fig. 10.

,- ...

0., ..-J , , 'D 0 ' -r ,

,

on"

o
(;

w
figu~e

10 (a) The CI)"StaI slndu'e of buium tilanate. 'The prototrpr! CI)'Stal is calcium titanale ~ile~ 11M: structure is cubic....itll &1"" ion$ at the cube COI'T>Cn. ~- ions at the f.M:e centers, .nd a Ti'" ion at lhe bod)'rentl'r (b) Below lhe c..ric tempr!ralWllthe struct",e is slightly debmed, with Ba and n<' iom dispbcrd "'bti~-e I<:> the ~- ions. thcreb)' dn-cloping a dipole moment. 11M: uppe~ and lo....er Oll)'P ions may 1t1O\"C down....ard slightlr.

00 K 10

.. /
/

/7 ~

-y ,
If

',1

,
,-

,
,
-

IU""nl.oht,1r,~

"'"
Tctrr,,,,d

\I,.,,,,,link"

-,m .,.,

.,

T.-.P<""'....fC1

.. ..

Figu", II Spontaneom pobuiulioo projt:cted on cube edg.. of barium tit.oat.., as II ruoctioo of temperature. (After w. J. ~'en.)

Consider the order of magnitude of the ferroelectric effects in barium titanate: the observed saturation polarization p. at room temperature (Fig, II) is 8 x 1()4 esu cm- 2 . lnc volume ofa cell is (4 x 10-")3 = 64 X 10- 24 cm3, so that the dipolc moment of a cell is
(CCS)

P
p

S!!

(8 X 10" csu cm-Z){64 X 10- 24 cm3) a 5 x 10- 18 esu em

(SI)

2!:

(3 X 10

Ie m

~(64

X 10- 30 m:l)

i!!l

X 1O-2'l em.

If tile positive iom Ba2+ and TI H were moved by 8 = 0.1 A. with respect to the negative 0 2 - ions, the dipole moment of a cell would he 6e8 5! 3 X 10- 11\ esu em. In LiNh03 the disphlCCments ilrc considerably larger, being 0.9 Aand 0.5 Afor the lithium and uiobum ions n:~pectively.

DISPI.ACIVE TltAi\;SlTIQNS Two viewpoints contribute to an understanding of a ferroelectric displ.lcive transition and by extension to disphlCive transitions in general. \Ve may speak of 1I polarization catastrophe in which for some critic:lIl c:ondilion the polariLation or some Fourier component of the polarization becomes very large. Equally, we may speak of the condensation of a transverse optie.!1 phonon. lIere the word l'Olldensalion is to be understood in the Bose-Ein~tein sense (TP, p. 100) of.! time-independent displacement of finite amplitude. This can oc:cllr when the l'Orresponding TO phonon frequency vllnishC5 at some lXlint in the Brillouin zone. LO phonons always have higher frequenc:ies than the TO phonons of the same wavevector. so we are not t:OIlC:erned with LO phonnn l'Ollllcnsation. In a poillriz.alion catastrophe the loc-al electric field caused hy the ionic displacement is l.\rger than the e1l1\tic restoring force, therehy giving an .Isymmetrkal shift in the positions of the ions. IIigher order restoring forc.-cs will limit the shift to a finite displac.'Cmenl. The oc'Currencc of ferrt)('l~clridty (and antiferroeleclricity) in many pcro\Skite-structure crystals suggests that this strodllre is f,\vor.lhly disposed to a displacive transition. LCH,:al field calculations make dear the reason for the hlVored position of this structure: the 0 2 - ions do not have cubie surroundings, and the 10l'31 field factors tum out to he unusually large. \Ve give first the simple form of the catastrophe theory, SUPIXl~ing that the local field at all atoms is equal to E + 41TP/3 in CCS or E + PI3Eo in Sf. 111e tlwol)' given now leads to a scc..'Ond-order transition; the physical idea.~ l'iln he carried O\'er to a first-order transition. In a secondorder transition there is no latent heat; the order parameter (in this instanl'C. the polari.."ation) is not discontinuous at the tnmsition tcml)Cmture. In a first-order transition there is a I.dent heat; the order parameter changes discontinuously ut the transition temperature, We rewrite (24) for the dielectric.- l'Onstant in the form

+ ""'3 YNIO,
(CCS)
l--Y 3 Na

8w

4w .

(30)

.B.o..5r.. ,TIO,
..(' ..... ~.. ,TIO,

l3~,j67h\lIOlll~

Tl~~
Uer C. RtlPllfttftl
~"d

111("1'

Figure 12 Di.h'tnc roNt2nt ....."".. l/(T - T.) in the pano:lt'Ctne: state (T > T.) of pt'I'O'>~.k..

R. O. Bdl.

where OJ is the cle<.'tronic plus ionic polariz.ability of an ion of type i and Nj i.s the number of Kms i per unit .'Olume. l'he didectrie l"Onstalit becomes infimte and permits a fi"ite polariz.ation in zero applied flcld when
(CGS)

1: N,a. = 3141T .
N(:Zi

(31)

TIlis is the l"Ondition for a polarization catastrophe. The value of E in (30) is sensiti"e to slTlall departures of 1: (:ritkal value 3141T. If we write
(CGS)

from the

(41TI3) 1: N(:Zi = 1 - 3s ,

(32)
(33)

\\here s <C I, the diele<.1.rie (.'onstant in (30) l>ecomes

E "'"

11$ .

Suppose near the critieal temperature s varics linearly with tcmpen\ture:

(34)

whcre f is a const.lnt. Sueh i\ variation of $ or 1: N,a. might rome from normal thermal cKpansion of the lattic'.". 1be dieledric constant has the fonn
E"'"

T _ T., .

(35)

dose to the observed temperature variation in thc paraelcctric state, Fig. 12.

Soft Optical Pllonons

The Lyddane-Sachs-Teller relation (Chapter 10) is

w1-Iwf. =

l(COYl(O)

(36)

'Ole static dielectric collslant increases when the ttallS\'ersc optical phono'll fre<luency decreases. When the slatic dielectric constant l(0) has a high value. such as 100 to 10.000. we find lhat w-r has a low value. When w-r = 0 the CJ")'stalls unstable because there is no effective restoring force. The ferroelectric BaTi~ at 24C has a TO mode at 12 em-I. a low frequency for an optirnl mode. If the transition to a ferroelectric state is first order. we do not find Wr = 0 or l(O) = co at the transition. The LST relation suggcsts onl}' that ~O) extrapolatcs to a singularity at a temperature To bclo.... T e . In disordered ('dirt}'l ferroelectriC'S the stattc dielectric (.'onslant can be larger than suggcsted b)' the LST relations. 7 The association of a high static dic!e(:tric (.'onstant with a low frequene)' optiUl.1 mode is supported by experiments 011 strontium titanate. SrTi03- Ae(;ording to the lST relation, if the reciprocal of the static dielectric constant has a temperature dependence lfl(O) IX ('f - To>. then the MjUare of the optical mode frequency will have a similar temperature dependence: wl IX ('f - To), if wL.. is independent of temperature. The result for tt4 is very well confirmed by Fig. 13. Measurements of wr versus 'f for another rerrocledric crystal, SbSI. are shown in Fig. 14.

Landau Theory of the Phase Transition

A ferrocle<.1;ric with a first-orcler phase transition between the rerroclectric and the paraelcctric state is distinguished by a diS<,ontinuous change or the saturation polarization at the transition temperature. -Ole transition between the normal and supen.und\lcting states is a sctundordcr tr'dnsilion, as is the transition between the ferromaJ.:nctie and pammagnetic states. In these transitions the degree of order goes to zero without a dis<'untinuous change as the temperature is increased. We can obtain a consistent formal thermod}'namic thcory of the behavior of a ferroelectric CI"}'stal by considering the rorm or the expansion of the energy as a function or the polarization P. We assume that the LandauS frcc energy density f. in one dimension rna)' be expanded rormall)' as
t(P;T,E) = -EP

go

+ igzp2 + ig..P~

+!~

+...

(37)

where the cocfflCicnts g., depend on the temperature.

'G Bums and E. Burstein, F"rroekctrial 7. 297 (1974) 'In TP, see pP. 69 and 296 for a di_ssion nf 11M: Lmdau r"r.aion.

"'"
- /

/ /

LA

/
-

,, , ,

"

, ,,

,I'

/1/
'00

...

T."nl'.:r~hlrt. "

Fi!un!; 13 1'101 of the O(ju;an, of the frequmq' dlhe zero wave-.ector 1ran5\"efW" optic mode :ag;ainsl tenlpt'ra1uR', b- Sff~ as ot-r,'oo in neutron ddIi-~ion ""peri,m:nts by Cov.ie). l1le ~ 1"'0: i' II", lecip~ of the diclt'd.ric: constant from the m=~uremcnts of Mib'''i and Westphal.

"

',; .. ;;_---,,".;.;_-,".. ;;--",oo::---c.:.:--c.. !:----: ..:--~ .. ;;--~,

ITT<~

,n"

Figure 14 o..u,'Me of a trans"crse pioonon froqueIlC)' as the Curie temperalure i$ approached from lido,,", ill th,' f"m>elcctric crystal antimony ~ulpholodidc. SbSI. (After Raman SClIltt'nng 0~nlC"b by C. II Pcrry and D. K. Agrawal.)

The series docs not c:ont.'\in lcnns in odd powers of P if the unpolar;zoo crystal has a center of inversion symmetry. but crystals arc known in which odd powers arc important. Power series expansions orlhc frcc energy do not aIW3}'S exist, for nonanHlytic terms arc known 10 Oct:Uf. cspcci:\l1y when VCT}' ncar a tr3rlsition. For cumplc. the transition in KI1 2 PO.. appears to l1ave a logarithmic singularity in the heat capadty at tile transition, which is not dassifiable as either first or sc<:ond ordcr.

""
l11C value of P in thermal e<luilibrium is given by the minimum of t as a function of f; the value of to at Ihis minimum defines the IlcllIlholtz free energy
F(1',E). The equilihriulIl polari:mlioll in an applied cb:trk: field E satisfies the
extremum cundition

(38) In this section we assume that the specimen is a long rod with the external

"pplied field E pumllel

10

the long axis.

10 obtain a fcrroclcc:lric slate we must suppose that the coefficient of the term in p2 in (37) passes through zero al some temperature 10 :
(39)

where l' is taken as a positive (:onslanl and 10 may be cquallo or lo\~cr than the tmnsilion temperature. A small positive value of g2 means that the lattice is "soft" and is dose 10 instability. A negative vallie of g2 means thallhe unpolarized lattke is unstable. The variation of ~ with tempemture is ao:;ounted lor by therlllal expan~ion and other eITcc:ts of anharmonic lattice interactions.

Se(,und-Ortlcr T runsition
If g.. in (37) is positive, nothing new is added by the term in gr., and this may then be neglected. The polaril:ltioll for zero applied electric: field is found from (38):

-y{'1' - 1'o)P$

+ F,4P:

= 0 ,

(40)

so that either p. =: 0 or = (Y!F.4)(1 0 - '1'). For '1' ~ 1'0 the only real rout of (40) is at p. = 0, because l' and gl are positive. Thus '1'0 is tile Curie temperature. For T < 10 the minilllum ol"the Landau free energy in zero applied field is at (41) as plotted in Fig. 15. The phase transition is a second-order transition bec:ausc the polariwtion goes continuously to zero at the tnlnsition temperature. The tnlllsition in LiTa03 is an example (Fig. 16) of a scconcl-order transition.

P;

HrstOrder Transition
Tile transition is first order if g., in (37) is negative. \'Ve must now retain gt; alldlake it positive in order to restroin t from going to minus infinity (Fig. Ii). 1'1Ie equilibrium condition for = 0 is given by (38):

if so tlmt either p. = 0 or

'ro)p, -

19,11') + g,P~

= 0

(42)

(431

-----------_...
PJT)
P,{O)

"
"'

"
00

"

n,

m.-_

"

n,

"

Figure 15 Spunl:lrll'OuS poIari:<aliOJl verSus temp"TallJrc, for a Sf.'<)Tu\-onl cr phase tranSItion.

" "

'.,,
,

,"
'"

.., ,

"

Class.)

I-'igure 16

1hnl"'nIUl'c varilllloll of the polar-axis .Iatlc dIelectric w"st.lnl of l.JTa~. (After

T:> T,

l'i&ure 11 Landau fret': e..ergy functiof,.'ersuJ (pobrizatiooljl' in a r>nl-oroe,.Iram.ilio<!, al ~ """,,Ialh e tL'lTlper.lrura. At T. the L:ancbu fur>ctiool hal L-quaI monilrla al P - 0 and at a finile P ... WKmTl. Few Tbelow Tc the .w.o!ule minimum is at 1~... vaIllC.'J olP,1oS Tpas..es IhltlU~ T.loc-n: is a dISCOntinuous change in the positioo of the aI.oluk' HUmmlWU. l1lC arfO"'!i nark the minima..

At the transition temperaturc Tr: the free energies of thc paracledric and fcrroclc(.1ric phases will be c<lual. 'Olat is, the value of F for p. = 0 will be equal to the valuc of F at the minimum given b)' (43). In Fig. 18 we show the d13mctcrislic variation with tcmpcrature of p. for a first-order ph;lSe transition; l'Ontrast this with the variation shown in Fig. 15 for a SCt.'Ond-ordcr phase transition. The trnnsition in BaTiOJ is first orocr. The dielectric constant is cak'ulatcd from the C<luilibrium pobrization in an applied eledlic field E and is found from (38). In equilibrium at temperatures ovcr the trnnsition the terms in p 4 and p6 may be ncglected; tbus E = -y('f - 1'o)P, or
(eeS)
~(1'

> 'fr;)

= I + 41TPIE = 1 + 41TI-y(T -

'1'0)

(44)

of the fonn of (36). 111e result applies whether the trnnsilion is of thc first or scl'Ond order, but if second order we have 1'0 :::: 1"'; if first order, then 1'0 < 1".:. Equation (39) defines '1'0, but l'r: is the transition temperature.
An .i[erroelectrici'y

A fcrroeledric displacement is not the onl)' t)1>C of instability that may develop in a dielectric crystal. Other defonnations Ol'l'\lr, as in Fig. 19. These deformations, even if the)' do not gi....e a spontaneous polarization, may be accompanied by changes in the dielectric constant. One type of delOnnation is called antiferroeJectric and bas neighboring lines of ions displlKw in opposite senses. The perovstdte structure appears to be slJ,)CCJ)tible to man)' t)'J'CS of deformation, often with little differenl'C in energy between them. The phase

..
u

J3

Did""'r;"" lind hrrOf:lfC'lrk-

40.

,
" ,
T<T,

T<T,

t--.

t'igure 18 Calculated values of the spontaneous pobl"ization Il$ fUOClioo of tempeniuTe, with paramelen as 1M l-ium tllanate. (Aner w. Cochran.)

0:

-m

T-T, .ndt-gC-

-'"

" ,

T:> T,

.....,opulor

l""'I..p-.I_ .... lI:"'UPf

or~lructural

Figure 19 Sch",malic representation offu,KIam",nlal t)'PeS

cenlros)'Ollll<:tric prototn)e. (After LInes and Cla.)

phase

tron~ihrllls

rrol1l a

Table 3

Anfiferroc:lt:dric Cl"Y$tal~
T ....nsit'OIl t"I11I)C .. ~ lure tn Rlltirermck'dric state, ill K

Cryst.ll
w~ ~a~b03

1010

793,911

PbZ.o,

PbllfO:t
l\'~HtP04 ~D4DtP04 ~1I411tM04 t\D4~$04 (~1-14~bJOtl

S06 488

'48
:lO4
254

".

242

From ~ COIn(.ilation by

Walter J.

Men.

diagrams of mixed perovskite systems, such as the PbZr03-PbTiOa system, show lransitions between p.'lr.l., ferrO+. and antiferroelectrie states (Fig. 20). Ordered antifen-oclectric arrangements of permanent ekoctrte dipole moments occur at low temperatures in ammoniunl salts and in hydrogen halides. Several crystals believed to have an ordered nonpolar state are listed in Table 3. Ferroclcctrk Domains ConsKier a ferroelectric crystal (such as b.'l.rium titanate in the tetragoo.'l.l phase) in which the spontaneous polarization may be either up or down the c all:is of the crystal. A ferroelectric l:rystal generally consists of regions called domains within each of which the polaril'..ation is in the same dirc<:tion. but in adjacent domains the polariz.'l.tion is in different directions. In Fig. 21 the polari7..ation is ill opposite directions. The net polaril'..'ltioll depends on the difference in the volumes of the ltpward- and downward-directed domains. The {.'rystal as a whole \\;11 appear to be unpolarized. as measured by the dlarge on electrodes covering the ends, when the volumes ofdomains in opposite senses are equal. The total dipole moment of the crystal may be changed by the movement of the walls bet-ween domains or by the nucleation of new domains. Figure 22 is a series or photomicrographs of a single crystal of bariUUl titanate in an electric Held normal to the plane of the photographs and parallel to the tetragonal axis. The closed curves are boundaries between domains po-lari-red into and out of the plane of the photographs. The domain bound:"lries change size and shape when the intensily of the elc<:tric field is altered. The motion of domain walls in ferroele<:lrics is not

,.,
,w

,
c

,.,

, , J ""
~

t,

,.,

".

'.~

n:,

,~

\101., pte< .... nl 11JliO,

"

00

'00

11"-'0,

Figu", 20 Fcrroclt.octrte "', 1lI.tLft,rroelcrlricA. llnd psr.ockct.;" P p1wcs oCthe lead 1.in,onate-lc:ld titanate .w>!id w1ulion 'litem. 11lc subscript T dcnot'''5 a tdragonal p1_', C a (:ubic phasc; R a .homboh.,dnl plwc:. of "hid, t~ are high temPffilture (111 and low tcmpnalure (L'I) una. N~ Ihe rtlOlllbohcdral-Id~1 phase boundaries one finds \'Cry high pile"D:>ekrlri<- CO"lll,,,g rodrociornts. (ArIa- Ja&l

+
+

j.
+

+
+

I I III

Figure 21 Cal &;hcmatic dnw'ing cC :lIolllic displaccments on <:,Ih..... side of. boundary between domains polarizc.-d in oppmilc dircdions in a rcrroel....iric CI')'S!aI; (h) "K-'W of a domain structl,rc I"""ng ISO" bound>l'ics betw''en ooma;ns polari7.l"(1 in opposite dir."dions.

;u.j\h..... tllll
<tIO

Figure 22 F<-rroelcrt';" domains on the &re d a single crystal of barium titllllllle. The f~ is norm"llo d,e lclragm:al or (: ax;';. TI,., nc.-t polarization or u,e a)'Slal ali judged b)' domain _"llIumes is intT'CllM"d markedly as the d<.dric field intensity panillcllO the axis is increas<.'<1 from 550 volts/ern 10980 Vlern. 'Ille domain bo"ndaries arc made \;~ible by elching lhe crystal in a weak kid 1'l>1,, lion (R. C. MIlIc,.)

simple: it is knO\.vn 9 that in an eledric field a l&::r wall ill BaTi03 appears to move by the repeated nuclealion of steps by thermal nuctualions along Ihe -parent wall. A ~18O" waIl" is the boundary between regions having opposite polarization dircdioos.
Piezoelec'ricily

All crystals in a ferroclet.1ric state arc a.Iso p;ezoclectrtc: a stress Z applied to the cr)'stal will dmnge the e1ectrie polarization (Fig. 23). Similarly, an electric field E applied to the crystal will calise the crystal 10 bewme strained. In schematic one-<1imensionalnotation, the pie'"l..oclcctric equations arc
ICCS)

P=Zd+EX;

c=Z~+Ed,

(45)

where P is the polariL1l.tion, Z the stress, d the piC7.oclectric strain constant, E the electric fjeld, X the dielectric susceptibility, e the clastic strain, and s the elastic compliance constant. To obtain (45) in SI, replace X by EaX. These relations exhibit thc dC'\'clopment of polarization by an applied stress and thc dcvelopment of elastic strain by an applied elc(:tric field. A crystal may be piC7.0Clcclric without being ferroelectric: a sdJcmatic example cl such a slrudure is gi\'(~n in Fig. 24. Quartz is piezoelectric, but not ferroelcctric; barium titanate is both, For order of magnitude, in quartz

"w. J. Men, Phys. Rco.'. 95, 690 (1954); R. undauer, J. Appl Phys. 28. 221 (1957); R C. Miller and A. Savage, 1't"S. R<. .... 115, 1176 (1959). For the theory of tile thickness of dQl'nain walls in fcnod<,'(;l, iI:s, ICC C. Killd, Solid Statl' Comm. 10, 119 (1972) and referenl'CS cit<.,,1 thcre.

1
<

p+.ll.pf

~~

FigtJn: 23 lal Unstressed fCrJ'Ol."lcdn., cr)'!;Ull amI (Ill slren"d fcrrod~trK" crystll. The SlreliS d,ang.,s the polari7.ation by toP, the ind\lc<'d pici1lCkctric pol...-iz:.llion.

w
,..,01..,,..

(.,

Figure: 24 (a) The uIIslrcSS<.'<l L'f)'st;d has a thrCl'fo!d symmelI)' axis. TIle arrOW!> represent dipolc lnanenll;, c""h set ~ three: ~ reprC>A.'lIh a 1Jla,w- group of ioos denoted hy N 8 3 , ,"'it". n 3ion at elldl n,., Sum <)/"the thrnldipok nlomenU at eah ,'<.'<tell i5 zrm. (blTIK,cryst..J ..itetl $lrased dC\'clops a poLoriz:atlOO In the dIrection inlLcated. 111C sum 4th.. dipole moments abot,1 o:och vcrlex is Ill') Iongl:r zero.

d - 10- 7 cm/statvolt and in barium titanate (f ." 10-5 cmlstatvolt. ,11c general
clc6nition iO of tl1c piezoelectric strain constauts is
(46)

where i-x, y, z and k - xx, yy, zz. yz, :tX, xy. To convert to em/stat V from values of dilt gi\'cn in mIV. multiply by 3 x lit. The lead rin:onate-lead titanate system (called the l:ozr system), Fig. 20, is widely used in polycrystallinc (ceramic) fonn with compositions of very high piC'l.oclectric t.'oupling. The synthetic polymer polyvinylidcnfluoride (PVF~ is five limes more strongly piezoelectric than crystalline quartL Thin stretehed films of PVF2 are f1c~ble and are easy to handle as ultrasonic transducers. Being flexible, the films arc ellSY to handle in mc<lical applications to monitor blood pressure and respiration.
""""" p>eZOclectrr 'tmn C'OIlSt.... ts fonn a thirdnulL temoo'. A C'OIlCise tmlllncnt elthe cfU.'d of symmctriC's 00 the ten..,,- pn:lPe1"tie$ of er:-stak i, gi'"CJ\ in the A.I P Iftlndbook, 3rd ed.. w 9-10 to 9-15, sec IIlso II. J. Jun.1sckc, Cryslill pl,ysiu-m"crorropk; "hvo,a ofanisotropic .0li,ls. I3cnjamin, 1974

..,
The response of piezoelectric crystals in transducer applications is characterized b)' the eledromcchanical coupling faclor k, whose square is defined b)'

/...2

= mcchankal energy stored .

(47)

clcdric energy stored

The result can be expressed in terms of the material constants used in (45). The rclacioo is given in works on the applications of piezoelectric crystals. Ferroelaslicity
A Cf)'sta! is fcrroclastic if it has two or more stable orientation stale'S in the absence of mechanical strcss and if it can be reproducibly lransfonncd from one 10 another of these states by application of mechanical stress. Reali.t.:alions are few: gadolinium molybdate Cd:z(MoOili has a combined ferroelectricfcrroclastic transition; lead phosphate Pb:,(PO~ has a pure fcrroelastic transition; SC\"eraJ crystals ha\'C combined fcrroclcctrie-fcrroclastic lrnnsilions.

Optical Ceramics l1lC addition oflanthanum to the PZT system gives a reasonably transpar+ enl product with useful c1cctro-optical projX:rtics. The PLZf system has been used as the basis of optical memories.

SUMMARY (In CCS Units)

The electric field avernged over the volume of the specimen defines the macroscopic e1ectrie field E of the Maxwell equations. The c1edric field that a<.1s at the site rJ of an atom j is the local electric field, E loc It is a sum over all dmrges, grouped in terms as EIoc{rJ) = Eo + E I + E 2 + E:!(rJ)' where only E:l varies rapidly within a <.'(!II. Here: Eo = external e1edric field; E 1 = depolarization field associated with the boundary of the sped men; ~ = field from polariUltion outside a sphere centered about rJ; E:!(rJ ) = field at rj due to all atoms inside the sphere. 11)C macroscopie field E of the Maxwell equations is equal to whidl, in general. is not equal 10 E.",,{rJ).

Eo + E"

The depolarization field in an ellipsoid is Ell' = -N,..J'... where N,... is the depolarization tensor; the polarization P is the dipole moment per unit volume. In a sphere N = 4wf3.

.....

 TIle Lorentz ficld is ~ = 41TPI3. The polarizabllily a of an atom is dcfincd in lerms ofthc local electric field as
p = aEI<>c

The dielectric susceptibility X and diele<.1ric constant nrc defined in terms of the macroscopic electric Held E as 0 "" E + 4"P = E "" (1 + 4nx)E. or X = PIE. In 51. we ha\'e X "" PI. AI atom at a site with robic symmetry has E""" = E the Clnusius-r.lossotti relation (24).

+ (4r.f3)P and satisfies

Problems
I. Polariwbilit!l of otomic ',yJrogen. Consider.a scmicl.assical model or the ground

the hrdrogen .atom in an ek'Ctric field non,,;!\ to the pluoc or the orLit state (Fig. 25), and show th;,t ror this ll"lCK1el Ct = Dt,. wlLere flU is the rnUiu$ of the \lnpcrturLcd orLit. Note: If the applict! field is ill the x din.-ction, then the xcomponenl or thc field or the nucleus at the dislll:\Ct.-d positioll or the c1ectt(m orbit must be (.'(Iual to the applied ficld. The correct ljllantllm-m(.-chanicaJ result Is larger than this Ly the fuctor f. (We arc speaking cJ ao in the c:q>;\llsion a .. 0'0 + atE + ....) We assume x <C (11/. Qlle can also Clkul,lle "'I on this model.

or

,/

T"
\

,;'

PruI.w'0,
Figure 2S An electron ,n. circtlbr orb;t rl radius QH Os dis(l~. distmce:r (lrI ~ppl:i<alion of an electric fldd E In the -x diJ"ecl:1OIl. nil! force or> the elec1ron due to the nude~ is e"fr1" in CCS or rf4~. in 51 The problem Do SU'nei :r <C DH'

,, ,,

,,

2. PolariUlbiJity rf conductirlg ~11eU. Show thai the poIama.l )' of a conducting metallic sphere of radius Q is 0: ... tfJ. This result is most easily obtained by nOting thai E = 0 iMide the sphere and then using the depolarization factor 4wf3 for a sphere (Fig. 26). 1be res"lt gh'C!i values of 0: tbe order of magnitude of the obSCTVed poIarixabilities of atoms.. A lallke ofN conducting sphct-ef per unit ~'Olume has clielectril'constant IE - 1 + 4TTfo,'i? fOf Ntr <ell I. The suggested proportiooa.lity of 0: 10 the cube of the ionic radius is ~lidled quite .....e ll fOl" alkali and halogen ions. To do the problem in Sl. usc, as the depolari:talion b::1:or.

3. E/fecr ofair gap. Discuss the cffuct oJ an air gap (Fig. 27) between conck-OSCr plates and dielectric on the measurement olhigh dielectric COOlitants. What is the high~ apparent dielcctri<' oonslant possible if the air gap thidmcss is 10-3 of the total. thickness? The presence of air gaps can scrioul;;ly distort the m<''llSurement of high
dielectric c:onsl:anu.

4. Intnfacilll polariultion. Show thai II pan.l.lcl-platc capacitor made up of two parntIcllaycn of malerial----()f'lC layer ith dick"Clric: rorutant f, 7.cro ool\(ludr,l!y, and thickness d, and the olhCT larer ilh . , 0 lOr corwcnience. flnite conductivity u.
and thickness qd-behaves as if the space bct"-'et.'f1 the condenser plates were filled with a homogenemu dielcdric with diclt.'dric constant (CGS)

where w is the allgular frequency. Values of t"f:fF as high as 10" or lOS caused largely by this Maxwell-Wagner ffiC(:hanisrn arc sometimes round, but the high values are a1....'a)'S accompanied by large a<: losses.

5. PolariUltion of sphere. A sphere of diek'Ctric constant

t" is plaet.'tl in a uniform external c1t.'Ctric field Eo. (a) What is tIle "olume 3VCI"uge elt.'Clric flcld E in the sphere? (b) Show thut the polari~1lion in the sphere is P .. xf:oI( I + (4wX/3)], where X .. (e - 1)/4'l'1". Hint: You do not nt."(.'tlto calculate E"", ill this problem; in fact it is confUSing to do so, br-cause t" lind X :Ire (lefined so tlmtl' .. XE. \Ve require Eo 10 be unchanged by insertion onhe sphere. We can produce <I fixed Eo by placing positive charges on one thin plale of an insulator lind nl.'glltivc charges on an opposite plate. If the plates arc always far from the sphere, the field of the plates will remain unchangcd when the spht.'re is inserted bct...."(.'(.'Il them. TIle TCsults abovc are in CCS.

6. Ferroeledr: criterion for alonl,f. Consider a system of two neutral atoms separated by a flXCd distance a, each atom haVing a polari7.ability a. Find the relation between a and a lOr such a system to be fen'OClt."Clric. Hint: 'The dipolar fieW is strongest a100g the axis of the dipole. 7. SoturatkJn poIoriuWon at Cum point. In a farst-ordcr transition the t.'quilibrium condition (43) with T set (.'(Iual Tc givcs one equalion for lhe poIariLllion PJ..TJ. A

..

I I I', , ,

Figur~ 26 ll1e lotal f>Cld inside a c(>ndocting, $(Ihe...... ~ro. Ira focld Eo is applied <-"'l"malty. then the fidd , due 1(> ."rfa(." <-~~rgt.'S on the sphere m"~t just <-1lIX'C1 F-o. SO Ib"t Eo + E, - 0 within

the slll'er". But E, C>IIl he s1mubk..! by the dqldari:l.llliou field -4".PI3 or a uuiformly polari,.....! $fllJcrc of p<lbri7~a. P. I'.clat~ P 10 Eo and cabbl" the dipok ftlOIIK>Ilt P of 1M .phcrt'. In Sllhc

depobrizata. field is -

Pr.w...

IJ

Figure 27

,,, " An air pp ofthicknCliS qd is In serio in a C>lpacilor with a diek,ctri<: sbb ofthiclmcu J.

,A::

.. ,

,.

further coodition~llt the Cnrie point is that {:(I'. Tc ) =- 1'(0. Te ). ((I) Combining these two conditions, show thatl>~Tc) = 31g.V4g. (h) Using this result, show thai Tc '" To + ~16l'868. Dielectric constant below 'ransi'ion temperature. In terms of Ihe p..,\,-amelers in the Landau rn:c energy cxpllosioo. show Ihal klr a sccunJordcr phase transition the Jidect:ric constant below Ihe tr.lnSilion temperature is
t:

= I

+ 41T6P/l:: '"'

+ 2w1)(fc - 1)

This result may lx: comporeU wilh (44) above the transitWm. 9. Sc!t modes and lattice 'rans!ormo'iom. Skelch a monalomic linear lattice oflattkc: coostanl a. (a) Add to c:acb six aloms a vector to indicate the din'Ction of Ihe displacement at a gi\'cn lime cawecl by a Iongi}udinal phonon wilh wm.... "''CClor al the 201"<: OOtlndar). (b) Sketch the CI')'sbl structUI"C that results ifthis ZOllC boundary

or

...
phonon I.>ccomes unstable (w--+ 0) as the Cf')'stal is cooled through T~. (e) Sketch on onc graph the essential aspects of the longitudinal phO!lon dispersion rclatiOfl for the monatomic lallice at 'f well aoovc Tc and at T - 'fe. Add to the gmph the SlIme

information for phonons in the new structure at T well below Te_ 10. FerTile/'ric lineu,. array. Consider a line of atoms or poIari:tability a and scporalion Q. Show that the army can polarize spoCltaneously if a ~ a'/4rn- 3, where the sum is O\'er all po$ith'c intt.'gcrs and is givcn in tables as 1.20'2 ...

References
A. D. Bna and R. A. <::owky. SlrIJctuml"m.. tr.... ritionf, Ta)U & Fr.ln<:11., 1!l61. H. Frohlich. Thwy.~diclf"Cfriu. (hford, 1958c. 1986.

E. Fa!Ul'm and W. J. MCI"7.. Flirrot/Iri(;ity. Northllo1],ul<I, 1961. K. A. Mf,lIcr and H. TIlOmas, cds., StnlctlJmll'hau ImllritiO'If. Springer, 1981. M. E. 1.lnes anc! A. ~l. Clout. FflfN')<'ledria mid n:wtfi,l m6leriuls. OdonJ. 1977. B. Jaffe, w. R. Cook, Jr. and It. JnlTt:, Pw...leclrit: Ct'romlcl, AU'demic, 1911. Ftnof'lcclrOc. A Journal. C. A. SaHlura and P. S. Peercy, "SIud)' of soft-mode trdnsitions It highvrc.'SSillC," Solid,*,"Ie physics 36,1 (1981). J.-c. Tolcdano.oo P. To&edallO.l..ond<>lIlNon; of,J-h'amiliom: ""plicf"joo to.trudl'Ul. lnrommenJllt'<lft, ",agnrnc, <RId Ilqufd ertP'ollljrtmtl., \\btldSc.Ynlilic, 1987.

'"

..

14
Diamagnetism and Paramagnetism
LANGEVIN DIAMAGNETISM EQUATION
QUANTUM THEORY OF DIAMAGNETISM OF

417

MONONUCLEAR SYSTEMS
PARAMAGNETISM

419
420

QUANTUM THEORY OF PARAMAGNETISM R<lre earth ions Hund 1'Ulcs lnm group ions Crystal field splitting Quenching of the orbit31 :mgulur momentum Spectroscopic splitting factor Van vtu::k temperature-independent parumagnetism

420 423 424 425 420 426 429 4JO

COOLING BY ISENTROPIC DEMAGNETIZATION

Nuclear demagnetization

431
432

PARAMAGNt.-nc SUSCEPTlBILITY OF CQNDucnON ELEcrnQNS

433

SUMMARY

436

PROBI,EMS
1. 2. 3. 4. 5. 6. 7. 8.

436
436
431

Diamagnetic SU!>Ceptibility of atomic hyd.-ogen Hund 1'Ules Tliplet e>;citcd states lIeat cap3city fn)m internal degrees of freedom Pauli spin susc:eptibility Conduction e1ecll'Oll ferromagnetism Two-le"e1 s)'Slcm Paramagnetism of S = 1 system

437 438 438 438 440

440 440

REFERENCES

NOTATION, In the problems trcalc(! in this chapter the m3gnctic field B is always closely cql131 to the applied field Ba so that we wrlle B for Ba in most instances.

t
+ -------------------------------

------~~~~----

o'-----~=====--== I hula f*UI"IP"'....... (-..J.) TftIIPft'Il"""

Figure I

Charact~rislic

magnelle 5llsccplihilit,e of ,jidlntlgnClle lint.! paramagnetIC

s"h5tan~'C'S_

CIiAPTEI.

4: DIAMAGNETISM AND PARAMAGNETISM

Magnctism is inseparablc from quantum mcchanics, for a strictly c1assic.11 s)'stem in thermal equilibrium can display 110 magnetic moment, c\en in a lnagnctic field. 111e magnetic moment of a free atom has three plindpal sources: the spin with which ck'C!:rons arc endowed; their orbital angul3r momentum about thc Ilucleus, and the change in the orbital moment induced by an applied magnctic field. The first two eITel'ts give p.1ramagnetic contributions to the magnetization, and the third gives a diamagnetic contribution. In tile ground Is state of the hydrogen atpm the orbital moment is zero, and the magnetic moment is that of the e1cctron spin along with a small induced diamagnctic moment. In the 1s2 statc of helium the spin and orbital moments are both zero, and there is only an induL'CiI moment. Atoms with filled electron shells havc zero spin and zero orbital momcnt: these moments arc associated with unfilled shells. '111C magneti7.1tion M is defined as tllc magnetic moment per unit volume. Thc magnetic susceptibility per unit volume is defined as

ICCS)

M X~B'

(SI)

(I)

where B is the macroscopic magnetic field intensity. In both systems of units X is dimcnsionless. \Ve shall somctim{~s for {:ollvenieJl{:c rcler to MIB as the susct:ptibility without specifying the system of unit.s. Quitc frequcntly a susceptibility is dcfincd rcfcrred to unit ma.~ or to a mole of the suhstance. 111C molar susceptibility is writtcn as XM; the magnetic moment per gram is somctimes written as (T. Substances with a Ilegative magnetic SUSC(.l)tibility arc called diamagnetic. SubstanL'(~s with a positive susceptibility arc called paramagnctie, as in Fig. I. Ordered anays of mUJ!;netic mOlllcnts are discussed in Chapter IS, the arrays may be ferromagnetic, ferrimagnetic, antiferromagnetic, helical, or more complex in forll1. Nuclear magnetic moments give rise to nuclear paramagnetism. Magnetic moments of nuclei arc of the order of 10- 3 times smallcr th.m tile magnctic 1I10mcnt of the electron.
LANGEVIN DIAMAGNETISM EQUATION

Diamagnetism is associated with the tendcncy of e1cctlical charges partially to shield the interior of a body from an applied magnetic field. In electromagnetism we arc lamiliar with Lenz's law: when tile llux tluough an electrical circuit is changed, an mduced current is set up in such a direction as to oppose the nux change.

."
In a superconductor or in an elcctron orbit within an atom, the indu('(.-d current persists as long as the field is present. The magnetic field of the induced current is opposite to the applied field, and the magnetic moment 3ssodatcd with the current is a diamagnetic moment. Even in a normal metal there is a

diamagnetic contribution from the conduction electrons, and this diamagnetism is not destroyed by collisions of the ck'Ctrons. The usual treatment of the diamagnetism of atollls and ions employs the Lannar theorem: in a magnetic field the motion of the electrons around a ('Cnlml nucleus is, to the first ordl.'T in B. the same as a possible motion in the absence of B except for the superposition of a precession of the electrons with angular frequency

(ees) "':: eBJ2nl

(SI) w = cBI2J'f1

(2)

If the field is applied slowly, the 1Il0lion in the rotating reference system will be the same as the original motion in the rest system before the applk:ation of the field. If the average c1~tl'On current around the nucleus is zero initially, the appliU'ltion of the magnetic field will cause a finite current around the nudeus. 'J1le current is equivalent to a magnetic moment opposite to the applied field. II is assumed that the L.annor frequell(:y (2) is much lower than the fn. quency of the original motion in the central field. This condition is not satis61.-d in free carrier C)'CIotron resonance. and the cydotmn frequency is twice the frequency (2). TIle Lannor precession ofZ electrons is equivalent to an electric current (SI)

1= (chargeXrevolutions per unit time) "'"

(-Ze)(..!..... ~B ) 217" 2m

(3)

'l1le magnetic moment M of a current loop is given by the product (current) X (area of the loop). '111e area of the loop of radius p is We have

7T,r.

(SI)

'"

(CCS) .. = - - - Ip') 4nu!'

Ze'n

(4)

Here VI> ::: (,;2) + (If> is the mean square of the perpendicular distance of the electron from the field axis through the nucleus, The mean square distance rl the elooroos from the nucleus is (,-2) ::: (,;2) + (if) + (r). For a spherically S)'mmetncal distribution of cllarge we have (,;2) - (If> "" <z2>. so that (,-2) ""

lip')
From (4) the diamagnetic susceptibility per unit "olume is, if N is the number of atoms per ullit volume,

(CCS)

Xc

Np.

NZe 2 = - 6mc2 (..2)

(5)

\
.~-.,--~---

(SI)

This is the classical Langevin result.

The problem of calculating the diamagnetic susceptibility of an isolated

atom is redua...d to the calculation of(,-2) for the electron distribution within the alom. The distribution ron be calculated by quantum mechanics. Experimental values for neutral atoms arc most easily obtained for the inert gases. Typkal experiment,,1 values of the molar susceptibilities arc tile

following:
N.
-7.2
A,

II.

K,
-28.0
-4.').0

-1.9

-19.4

In dielectric solids the diamagnetic contribution of the iOIl corcs is described roughly by the Langevin rcsult. 'lllC contribution of conduction elecIrons is more complicated, as is evident from the de Haas-vlUl Alphen effect
disl..'usSt,"(\ in Chapter 9.
QUArITUM 1llEORY OF
DL~1AGNETISM OF

MOl\'ONUa..EAR SYSTEMS

From (G. 18) the effect of II magnetic field is to add to the hamiltonian the
tcnns
(6)

for (In atomic electron tllese terms Illay usually be treatcd as a small perturbation. If the magnetic field is uniform fIIld in the z dircction, we may write
A.. "" -!yB
A~ ""

ixB

A.. = 0

(7)

and (6) heromes

:JI' = - 2711c
nlC

t ""B (. rB x- - y.) - +- ( r + !/)

dy

in

8m~

(8)

first tcrm on the right is proportional to the orbital fIIlgular mumentum component La if r is mcasured from the nuclcus. In mononuclcar systems this term gives rise only to paramagnctism. 111c scc.'Ond term gives for a spherically symmetric system a contribution

(9)

'"
by Ilrst-order pcnurbaliofllheory. '''e associated magnetil netic: ,menl is diamag_

ill agreement with the classical result.

PARAMAGNETISt.t

Electronic parnmagnetism (positive oontributioo to X) is found in:

1. Atoms, molecules, and lattice defects possessing an odd number of electrons, as here the total spill of the system cannot be zero. E",amples: free sodium atoms; gaseous nitric oxide (NO); organic free mdicals such as triphcnylmetJlyl. c(Ct;H~; I; centers in alkali halides. 2. F fee atoms and ions with a partly filk-d inner shell: transition clements; ions isoelectronic with transition clements; mrc earth and actinide clements. Examples; t.'1n2 +, Gd 3 +, UH . Paramagnetism is exhibited by many of these ions C\<-en when incorporated into solids, bul not invariably. 3. A fe\\1 compounds with an evCfl number of c)C1;'trons, including molecular oxygen and organic biradicals. 4. Metals.

QUANTUM THEORY OF PAI\AMAGNETlSM

'Inc magnetic moment of an atom or ion in free space is given uy

(ll)
where the total angular momentum (tJ is the slim of the oruital hL and spin (IS angular momenta. The constant 'Y is the ratio of the magnetic moment to the angular momentum; ')' is called the gyromagnetic ratio or magnetogyric ratio. For electronic systems a quantity g called the g f.'lClor or the spc<:trosoopic splitting factor is defined by
I!Jl.n -yfl

(12)

For an electron spin g = 2.0023, usually taken as 2.00. For a free atom the g factor is given by the Lande equation

:-+"1,,,1_--,4",',-+,--,,-11 g=l+ ~J(",J_+,--"I)_+_S;;;(7;S

2J(J

1)-'

(13)

.. , ,
14 Du"..t1gud"m ..nd r ......mug'oe'.. m 421

.:;! O.7S

.!

I "~--=====I
Uwcr_

~0.50

"
t'igun: 2 Enc'l.'Y Icvel ~plitti"g for onc I'lectron in II n>agnl'tic flcld B directed along Ihe IlOSiliw.l' ui5. For an eledron Ihe nlAgnctic n1Olllefl1 ,.. i~ OJlIlmllC III .ign to the spin S, SO that ,..-llIo'aS. In the low <'OCrgy .rate the magoc'tic ITlOfJlCnt is IllU1Illei to the magnetic field.

~0-2$

.1.--,.,,--=~:::==~==~. 10 1.5 20
!,rm.~T

Figure 3 Frad;onal popubt.....s of. two.Jc..d in t1ll:1"ma1 equi!lmunl at tl'1nllCratun' T in a magnellC" roeld 8. 1lle magnetic 'lKlrntnl I. JII'OIl<101iona1 to the dilfereooe hetwnm the two
.~~tc:m
Cl' ....'eli.

'Ibe &llr magneton POll is defined as dJ2mc in CGS and eM2m in SI. It is closely equal to the spin magnetiC moment of a free electron. nle energy levels of the system in a magnetic field are
U = -Ii-' B = m)gJ.tIlB ,

(14)

where mJ is the azimuthal quantum number and has the values J, J - I, -j. For a Single spin with no orbital mom('flt we have 1Il):t:: alld g w!J<"'llce U '" ::!:poIlB. lnis splitting is shCl\\n in Fig. 2. If a system has only two levels the Ctluilibl"hlm populations arc, T. kBT, cxp(,J.BIT) N, = N exp(jtBlT) + exp(- P.BlT)

::!:,

.. ,

'" 2,
with
(15)

N, N

exp(-IlBlT)

exp(p.Bh) + cxp(-poB/T)

(16)

here Nt> N2 are the populations of the lower and upper levels, and N '" Nt + N 2 is the total number ofatoms. TIle fractional populations arc plotted in Fig. 3. The projection of the magnetie moment of tbe upper state along the field din..'d.ion is -Po and of the lower state is po. TIle resultant magnetization for N atoms per unit volume is, with ]I; poBlknT,

For]l; <Ii 1, tanh]l;" x. and we ha'e

(18)

In a magnetic field an atom with angular momentum quantum number J has 2J + 1 equally spaced energy levels. The m:'glletization (Fig. 4) is given by
(x.
g]~/lBlkRT)

(19)

...
'00

,,,

'00

V
~
~

, ,-
5.1"

(Cd")

'"

~
~

f
1

,;

"

(Ft"")

"

~ HI)

~
s- ~

~300

I
".

(Q3')

/
0

'. 200

..00

./ ,

l>

'" ,
lOOK

'00'
411 K

1lt,1bnn

_I.. .
"
I " "

,, ,,

Figure 4 Plot of magnetIC rnllr'IIen! ~ BIT lOr ,phcrical samples oC (I) potas,,;utn chromIum alum, (II) lCrric ammonium a.lum, llnd (III) gaddmium 5ulf..te octahydrate. <hoer 99.5'ito magPetic salllrlliioo is acl1ie\~ at 1.3 K .nd ,bout 50,000 ~ss. (51). Aflcr w. E. IIcnry.

wllere the Brillouin function BJ is defined by

Bftx) =

2J + 1 c,nh(21 + 1)%) __'_ Clnh(2-)

2J 2J 2J 2J

(20)

Equation (17) is n special case of (20) for} = I. FOt" x <C 1. we have I z i' ctnhx=-+---+_ x 3 45
and the stlSC't.'ptihility is
-~

(21)

r..'J(J + IllflJ.i
3k s T
IWrobel"

~-

c
T

(22)

Here p is the cffedivc

of Bobr magnctons, defined as

p _

gU(j + 1)1'1'2 .

(23)

....

..

"
-

/
II I

.
,,

1/
""

..,

I'igm'C 5 Plot g( IIX &I' T few- g;dolinjnm ~lt. Cd(c.ll90ch 911.0. "Ole straight line is the Curle law. (After L C. Jacb.m and H. Kamcrli"gh Onnes.)

The constant C is known as the Curie conslant. 111C form (W) is known as the CUlie-Brillouin law, and (22) is known as the Curie I:lw. Results for the paramagnetic ions in a gadolinium salt are sbown in Fig. 5.

Rare Earth

lOllS

The ions of the rarc-cal1h clements have closely similar chcmkal properties, and their chemical separation in tolerably pure form was at.'COmplishcd only long after their discovery. lbcir magnetic properties arc fascinating: tile iOlls exhibit a systematic \'ariety and intelligible complexity. '111C chernic.11 properties oCthe trivalent ions arc similar beea\ISC the oulcrmmt electron shells arc identically in the &25,}6 configuration, like neutral xellon. III I~ntln\n\lm. just before the rare earth group begins. the 4fshell is empty; lit cerium there is one 4f electron, and the number of 4/ eleclrons increases steadily through the group until we have 4p3 at ytterbium and the filled shell4p" at lutecium. The mdii of the trivalent ions contract fairly smoothly as we go through the group from l.lJ Alit cerium to 0.94 Aat ytterbium. This is the ramouJ, "lanthanide COfltradion." What distinguishes the magnetic beha,tior of one ion species from another is the number of 4f electrons compacted in the inner shell with a radius

of perhaps 0.3 A. Evcn in the mctals the 4f corc rC\ s its integrity and its atomic properties: no other group of elements in the periodic tablc is as intcresting. '111c prc(.'eding discussion of paramagnetism Applies to atoms thAt have a (2J + I)-fold degeneratc ground statc, thc degeneracy being lifted by a mag_ netic field. The influence of all higher energy statcs of the systcm is neglcl:ted. 'I1lesc assumptions appear to be satisfied by a number of rare-earth ions, Table I. 'I1lc cakulatcd magneton numbers are obtained with g \,aluC$ from the LanM result (13) and the ground-state level assignment predicted 1>" the Ilund theory of spectral tenns. The discrepancy between the experimental magneton numbers and those calculated on these assumptions is quite marked for Eu3 + and Sm3 + ions. For these ions it is necessary to Insider the influence of the high states of the L-S multiplet, as the intervals between successive states of thc multiplet are not large compared to kljT at room tcmperahlle. A Illultiplet is the set of levcis of differentJ values arising out of a given Land S. The levels of a multiplct arc split by the spin-orbit interaction.
lhllld Rlllu

'111e Hund rules as appli(.'d to electrons in a given shell of an atom affinn that electrons will occupy orbitals in sum a way that the ground state is dlar8Cterizcd b). the follOWing:

1. Ine maximum value of the total spin S allowed by the exclusion principle; 2. TI,e maximum value ofthe orbital angular momentum L consistent \\ith this \'aluc of S; 3. Thc value of the total angular momentumJ is CtluaI to IL - 51 when the shell is Ics.~ than half filII and to L + 5 whcn the shell is more than half full. Whetl the shell is just half full, thc application of the first rule gives L "" 0, so that J "" S.
'J1,e first I'lund rule has its OIigin in the exclusion plinciple and thc l.'OlIIomL repulsion between electrons. The exclusion principle prcvcnts two electrons of tl,c s,'une spin from Ix:ing at thc same plat..'C at the samc time. Thus electrons of the same spin lU'e kept apart, further apart than electrons of opposite spin. Because of the coulomb inter.u:tion the energy of electrons of the same spin is lower-the average potential ene"b'Y is less positive for parallel spin than for antiparallcl spin. A good example is the ion Mn2+. This ion has five electrons in the 3d shell. which is lheref~re half-filled. The spins can all be parallel if each electron entCl'$ a different orbital, and there are exactly five different orbitals available, characterized by the orbital quantulll Ilumbers JIlL = 2, 1.0, -1, -2. These will be occupied by one electron each. We expect 5 "" t. and because :rml. = 0 the only possible vallLe of L is 0, as observed.

.. ,
Table I

Errl

e mapleton munben V for trivalent lanthanide group ions

(Ncar room temperature)

I~

Configuration

8aslc \eo.'CI

dJU +

P(C2k} -

I)I"~

appmo:imat...

"'' ' '

2.4 3.5 3.5 1.5 3.' 8.0 9.5 10.6 10.4 9.5 7.3 '.5

""" Pr"
Pm"

4f'sryti 4rsrye

fl;,"
3fl~
41 110'1

t\d'+

.1'....,.

4P5s'prl
4f~pti

5m 3 + Eu3+

ed"
1b"

'4!'sryti /'50'' '

'f'5N'
./'
4r-ar,:l'

"Y" lI03+

'I, 'H M 'F, 'S1ll 'F,

'I,
'H,
IF 712

E~' Tm'+ Yb3+

4fIOSsV
4f'~ti

6HI~
~ll~

4fllsrpti

....".
,,2.

2.54 3.58 3.62 2.68 0.84 0 7." 9.72 10.63 10.60 9.59 7.57 '.54

The second Hund rule is best approached by model calculations. Pauling and \Vilson, I for example, give a calculation of the spcctr.ll tenns that arise from The third Hund rule is a consequence of the sign of the the configur..Ition spin-orhit interaction: For a single elcclron the energy is lowest when the spin is antiparallcl to the orbital angular momentum. But the low energy pairs mL. ms are progrL'Ssh'c1y u$t.'d up a<; we add electrons to the shcll; by the cxclusion principlc when the shell is morc than haiffullihe state oflowest enel'gy necessarily has the spin parallel to the orbit. Consider two exmnples of the Hund rules: The ion Ce3 + has a single f electron; an f c1o::tron ha.~ I = 3 and $ = i. Ba'!luse the f shell is less than half fuU. the] value by the preceding rule is IL - 51 = L - i = t. 111e ion p~+ has two f c1edrous: one of the rules tells us Ihal the spins add 10 give 5 = I. Both f elCC'trons Cflllnot. have m, ::: 3 without violating the Pauli exclusion principle. so tho,t the maximum L consistent with the Pauli plindple is Ilot 6. but 5. The] value is IL - 51 = 5 - I := 4.
1ro,. Crouv 1011$

Table 2 shows that the expclimcnlal magncton numbers for salts of the iron transition group of the periodic table are in poor agreement with (18). TIle "alues often agree quite \\'<'11 with magneton numbers V = 2(5(5 + 1)}1t2 calcllIL P.uhng lllod E. 8. Wilson. ',,'roductiofl lO qua"',,m nltClwn/Q. McCno.w-lIiIJ. 1935,
PI' 239-2-Ki.

."
Table 2

Effceli,<c magneton numbe" or iron group ions

ConfigI~
C~

....,
!t'\"e1

Ilralion

diU + 1'"
1.55

ple:-k) -

2{S(S

P(nIc) 1.73
~83

+ I)J' I

P(expf

TfH.

y4+

V".

3</'

C~+.

V'H-

3d' 3d'
3d'

'F,
5D

'Vu

4F3.'2

~ln3., C~

Fe:!+. Mnh

3<P
3<~

Feu Cou
N"i1.+
Cu1+

's",

'Do
4

3d' 3d' 3d'

'F.. 'F,
'Vu

1.83 0.77 0 5.92 6.70 8.83 5.59 3:>5

3.87

1.8 '.8 38

'.00
5.9"2

'.00
3.87 283 1.13

5." 5A '.8 3.'

,."

l."

"Representative ....Iues_

laled as if the orbital moment were 001 there at all. We say that the orbital nloments are fJuenched.

Crystal Field SpIittillg

lhe difference in behavior of the rare earth and the iron group salts is that the 4[ shell responsible for paramagnetism in the rare earth ions lies deep inside the ions, within the 51 and 5p shells, whereas in the iron group ions the 3d shell responsible for paramagnetism is the outcnnosl shell. The 3d shell experiences the intense inhomogeneous electric field produced by neighboring ions. This inhomogeneous electric field is called the crystal field. The interaction of the paramagnetic ions with the crystal field has two major effects: the coupling oIL and S ",ectors is largely broken up, so that the statd are no longer specified by their J ",alues; further, the 2L + 1 sublewls belonging to a given L which are degenerate in the free ion may now be split by the cryslnl field, as ill Fig. 6. l1lis splittiug diminishes the contribution of the orbilnl motion to the magnetic moment.

Quenching of Ihe OrVilal Angular Momenttnn

In an electric field dirccted toward a l'bled nucleus, the plane of a classical orbit is fixed ill space, so that all the orbital angular momentum components 4, L". L,. arc constant. In quantum theory one angular momentum componCflt, usually taken as L,., and the square of the tot.'Il orbital anguJaJ" momentum L2 are constant in a centmlfleld. In a nonccntmlfleld the plane of the orbit will mo\'C about; the angular momentum components are no longer constJ.nt and rna)' average to zero. In a crystal L1 wi.ll no longer be a constant of the motion, although to a good apprwdmation L2 rna)' continue to be constant. When ~ averages to zero, the orbital angular momentum C{ d to be quenched. "The
no

..

14 Diamagnet"m and Paramagnet.,m

~.

,
===-P"P.

, --,.
@
(b)

@
(a)

@
(e)
(d)

Fig,,",6 Consid..... all alom wilh orhital angular mo....,ntum L" I pbct.'<I in the uniaxial crystalline elt:dric fIeld of the two positi,'e ions alnng the z axis. In the &ee atom the states m.... :!oj, 0 have idenlical energies-th.." afe degenerate. In the crystal the alom has a lower energy "'m", the e1ectr<m cloud is close to posit;"e ions as in (a) than when it is orieuted midway between them, as in <b) and (c). l1le wav"functions that gi,'e rise to these charge densities are ofthc fonn "f<r). Ilf(r) and !If(r) and arc called the P.. P.. ""orbitals, respect;,..,ly. In an axiaUy symmetric Iiek!, as shown, the p, and Pw orbitals are degenerate. The energy levels referred to the free atom (dotted ]ine) afe shown m (d). If the electric fick! does not ha,'e axial symmetry. all three stat<.-s Will ha\'e d,lTerent ene.gies.

magnetic moment of a state is givcn by thc average \~<lluc of the magnetic moment operntor J./.1J(L + 2S). In a magnetic field along the;;; direction the orhital contribution to thc magnetic moment is proportional to the quantum expectation value of L,,; the orbital magnetic momcnt is quenched if the me-. chanical momcnt L" is qucnched. \Vhen the spin-orbit interaction cncrgy is introduced, the spin may dntg some orbital momcnt along with it. If the sign of the intemction favors parallel orientation of the spin and orbital magnetic moments, the total magnetic moment will bc larger than for the spin alone, and the g value will be larger than 2. The experimental results are in agreement with the known variation of sign of the spin-orbit interaction: g > 2 when the 3d shell is more than half full, g =:: 2 when the shell is half full, and g < 2 when thc shell is less than half full. \Vc consider a single electron with orbital quantum number L =:: I moving about a nueleus, the whole being placed in an inhomogencous cryst..'tlline elechic field. We omit electron spin. In a crystal of orthorhombic symmetry the charges on neighboring ions will produce an electrostatic potential q; about the nuclcus of the form
(24)

where A and B are constants. This cxpression is the lowest degree polynomial in :1:, Y. ;;; which is a solution of the Laplace equation VZq; =:: 0 and compatiblc with the syml '1' of thc crystal.
\

".
In free space the ground state is thT(.'e-fold dc~ ..:ralc, ....ith magnetic quantum numbers "'L = I. O. -I. In a magnetic field these levels are split by energies proportional to the field 8, and it is this ftCld+proportional splitting which is responsible for the normal paramagnetic susccptibilit)' of the ion. In the crystal the pkturc may be different. We take as the three w3.\'efuoctions associated with the unperturbed ground state of the ion

(25)

These wavefunctions are orthogonal, and we assume that they arc normalized. Each of the U's can be shown to have the property

~Uj = L(L

+ l)U. = 2U

(26)

where 9fl is the operator for the square of the orbital angular momentum, in units of fl. The result (26) confinns that the selected wa\efunctions arc in act p functions, having L = 1. We ohser..c now that the U's arc diagonal wilh respect to the perturbation. as by symmetry the nondiagonal clements \'anish:

(U,leolU,) ~

(u.\eolUJ

(uJ.oIuJ
(A +

~ 0 .

(27)

Con5idcr for example

(UJ,oIu,) ~ I

nAMFIAx' + BY' -

B"'J ,Ix dy dz

(28)

the integrand is an odd function of x (and at50 of 1/) and therefore the integral must be zero. The energy levels arc then given by the diagonal matri", elements:
(U,lc~IU.-) =

f [f(r)12{Ax 4 + B!iY' - (A + B)zY} (Ix tty dz

A(I, - Ii) ,

(29)

where In addition,

(UJeolU,)

~ B(I, - 1,) ,

(U,leolUJ = -(A

+ B)(I, - 1,) .

111e three eigenstates in the crystal field are,) functions with their angular lobes dir<.'Cted along each of the x, y, z axes, reSp(.dively. l11C orbital momcnt of each of the Ic\'c1s is zero, because
(UJLJU,) ~

iUJLJU')

~ (U,ILJUJ ~ 0 .

nus dTect is known as qucnehing. lhe level still has a deflnitc total angular momentum. since!l is diagon.al and gives L:c: I. but the spatial components of the angular momentum are not constants of the motion and their time average is zero in the first approximation. 'l1lercfore the components of the ocbirn] magnctie moment also vanish in the same approximation. 11lc role of the crystal

field in the qu, ...:hing process is 10 split the origjually degcnernte levels into nonmagnetic levels separated by cncrgies:f> IlH, so thai thc magnetic field is a small perturbation in <:Qmparison with the crystal field. At a lattice site of cubic symmetry there is no term ill the polenti.1.I of the form (24), that is, quadratic in the electron coordinates. Now the ground slate of 3.1\ ioll with onc p e1ectrou (01' with one hole in a p shell) will bc triply degcncrate. However. the energy of the ion will be lowered if the ion displaces itself with respect to the surroundings, thereby creating a noncubic IJOtential such a~ (24). Such a spontaneous di!>vlacement is known as a Jahn-Teller effecl and is often large and important. llarticularly with the Mn3 + and Co 2 + ion~ and with holes in alkali and silver halides.
Spec~l'"OSoopic

Splitting Fador

\\'e SUPlJOse for convenience that the crystal field constants, A, B are such that V K = if(r) is the orbital wave function of the grollnd state of the atom in the crystAl. For a spin S = i there arc two possible spin states S~ = j: i represented by the spin functions a, P, which in the absence of a magnetic field are degenerate in the zeroth approximation. The problem is 10 t.'lkc into account the spin-orbit interaction cnergy AL S. If the ground state functiou is 1/10 = U~ = xf(r)a in thezcroth approxima tion, then in the first approximation, considering the AL' S interaction hy standard perturbation theory, we have
~ ~ IU. - i{AI'UI,)U,la - i{AI2t>;)UJ3

(30)

wherc AI is tlle energy difference between the V~ and VII states, and ~ is the difference between the V~ and V~ states. The tcon in UJ3 actually has only a second order effect on the result and may be discarded. 1bc expectation \-alue of the orbital angular momentum to the first order is given directly by
(<P!LJ~) ~ -JJ6,

and the magnetic moment of the statc as measured in the z direction is

M,(<P!L. + 2S.I~) ~ [-(JJ6,) + 11M, .
As the separation between the levels
AE
S~ =
j:

i in a field

H is

= gp,8H = 2[1

- (MAI)]llllH ,

the g

\~dl\le

or spectroscopic splitting factor (12) in the z direction is 211 - (JJ6,) . (31)

rerl"n"nres.~ g"'~n

-See 1.. Orgel. lnl.-odudlon to tramitiDfl Ifldcl c:hDrWf~. 2:nd ed., Wiky, 1966; v.lcnSl\~ by M. D. Siurge, 1't1ys. Rev. 140. A680 (l965).

Vtl Vleck 1emperature.hldcpellderlt Paramagnetism

\\'e consider an atomic or molecular system which has no magnetic IllOment in the ground state, by which we mean that the diagonal matrix clement of the magnetic moment operator ~ is zero. Suppose that there is a nondiagonallllatrix clement (sllldO) of tile magnetic moment operator, connecting tile ground state 0 with the excited state $ of energy lJ. "'" E. - Eo above the ground state. lOen by standard perturbation theory the wavefunction of the ground state in a weak field (pzB <:t lJ.) becomes
(32)

and the wavefunction of the excited state becomC!i

(33)

111e Jlerturbed ground state now has a mOlllent

(34)

and the upper state has a moment

(35)

'lnere arc ","'0 interesting cases to consider: Case (a). lJ. <C keTo The surplus population in the ground state over the excited state is approximately equal to NlJ.I2k eT, so that the resultant magncti:-.ntion is
(36)

which gives for the susceptibility

X - NI(~",IO)I'lk.T .
(37)

llere N is the number of molecules per unit volume. This contribution is of tile usual Curie fOrm, although the mechanism ofmagnctiz..1tion here is by polarization of the states of the system, whereas with free spins the mechanism of magnetization is the redistribution of ions among the spin states. We note that the splitting .1. docs not enter ill (37). Case (b) .1. keTo Iferc the population is nearly all in the ground state, so that
'1=

(38)

11lc susceptibility i5
x~

(39)

independent or temperature. This type or contribution is known as Van Vleck paramagnetism.

COOUJ'liC BY ISENTROPIC DE."'lACNETlZATION

The first method ror attaining tcmpcr.t.turcs much below I K was that or isentropic, or adiabatic, demagnet~~1.lion.3 By its use temperatures or 10- 3 K and lower have been reached. The method rC'Sts on the fact that at a fixed temperHture the entmpy or a system or magnetic moments is lowered by the appli<:ntion or a magnetic field. The entropy is a measure or the disonler or a system; the greater the disorder, the higher is the entropy. In the magnetic field the moments will be partly lined up (partly ordered), so that the entropy is lowered by the field. The entropy is also km'Cred ir the temperature is lowered, as more or the moments line up, Irthe magnetic field can then be removed without changing the entropy of the spin system, the order or the spin system will look like a lower temperature than the samc degree or order in the presence orthe field. When the specimen is demagnetized at constant entropy, entropy can flow into the spin system only from the system of lattice vibrations, as in Fig. 7, At the temperatures of interC'St the entropy ofthe lattice vibrations is usually negligible; thus the entropy of the spin system will be essentially constant during adiabatic demagnetization of the specimen. Magnetic cooling is a one-shot operation, not C)dic. We first find an expression fOr the spin entropy of a system or N ions., each of spin S, at a temperature sufftciently high that the spin system is entirely disordered, That is, T is supposed to be much higher than some tenlpcr.iture A which characterizes the energy of the interactions (E1nt Ie kIJA) tending to orient Il,e spins preferentially, Some of thesc interactions arc discussed in Chapter 16, The definition or the entropy u of a s)'stem of G accessible states is u:: k ll In G. At a temperature so high that all of the 2S + l states or each ion arc nearly equally populated, G is the number of ways of arranging N spins in 2S + I states. Thus C :: (25 + l)N, whence the spin entropy Us is: Us :: 1c8 In (25

It' == Nics In (25

+ I) .

(<OJ

This spin entropy is reduced by a magnetic field ir the lower le\'els gain in population when the field separates the 25 + 1 states in energy,
method was suggested by P. Deb)'e, Alln. Physik 81, 1154 (1926); and w. F. Ciallque, Am. Ch..m. 50.-. 49, 1864 (1927). For many purposes the method has been SUllplanktl by the He"-Il.. ,Mutio" remg..... tor " ....ich operates in a continuous <')'1:1.. The lleJ atoms In solution in liqui<llle' play the role of atoms in a gas. ant! cooling is cfft.'Cted hy "'...poriu.tion' of lIe"; Il TP. Chapter 12.

J.

roo

""
,
I I I ,
I
TwI

--L

I
I

,
I I

.....Ii'

Tlmr I' ...hd1

lo ...""" ....

, .... ...... "'"

Nnr~""""'itlm

r ... a1 ........
mogntti<-

....,......rd

fit.k1

Figu~ 7 During ooiahatic dem.ajtllclizalion the lotal entropy of the specimen is mnstanl for ",ITL'<.'li\t: cooling the 'nililll enlropy ofthc lattil-e should he 5m.1I'1l romp.lrlSOl1 with the entropy of

the spin system.

111C steps carried out in the cooling process arc shown in Fig. 8. 111(: field the specimen in good thennal contact with the surroundings. giving the isothcnnal path ab. The specimen is then insulated (l1u "" 0) and the field removed; the specimen follows the constant entropy palh be, ending up at temperature T 2 _ The thenna! contact at T] is pro-vidcd by helium gas, and the thenna! cootact is broken by rcmo..ing the gas
is applied at temperature T l with

with a pump.

Nuclear VemagnetiuItion
'111c population of a magnetic sublevel is a function only of f.l,B/k/jT, hence nf Bff. The spin-system entropy is a function only of the population distribution; hence the spin entropy is a function only or BIT. lf BlI. is the effective fleld that L'orresponds to the local interactions, the final temperature T2 reached in an adiabatic demagnetization experiment is

7,

~ 7,(8.'8)

(41)

where 8 is the initial field and T. the initial tcmperature. Because nuclear magnetic momCllts are weak, nuclear magnetic interactions are much weaker than similar electronic interactions. We expect to ream a temperature 100 times lower with a nuclcarparamagnct than with an electron paramagnet. The initial temperature T] of'the nuclear stage in a nuclear spincooling experiment must be 100IIer th.HI in an ck"Ctron spin-cooling experiment. If we start at B = 50 kC and T l = 0.01 K, thcn pJllk aT. "'" 0.5, and the en-

..

00

10

15 T, ",II:

20

25

30

Figure 8 Enlropy lOr a ~pin I s)"Slem IIIi ',meh'X1 of te<npmlture. _tming .. intrmal random ,~k; Reid 8" of 100 g;ow;s.. spenmen is rnagoctiud isothe...mal!)- along ob. and is then ....ubtoo thenmul)'. The len...! rnaK1~I;'" Rdd is tumal off along b<-_ In <!Iller to kep th.. rLgUJ"c on "n:asonabk JCale the initial temperature T, is 10"'..... than ..'OOk! he used in practice, and SO is the "~ll'nJal "'a.R;netlC field

n...

troP)' decrease on magncthmtioo is over 10 percent of the maximum spin Clltrop)'. This is sufficient to oven~'helm the lattice and from (41) we eslimatc a final tcmperature T 2 - 10-7 ti:. The first 4 nuclear cooling experiment was carried oul on Cu nuclei in the metal, starting from a first stage al about 0.02 K as allainro by electronic cooling. 'foe lowest temperature rea('hed was 1.2 x 10- 6 K. The result.. in Fig. 9 fil a line of the form of (41): T 2 '" T j (3.118) \vith 8 in gauss. so that 8" "' 3.1 gauss. This is the effccth'C intcraction field of the magnetic moments of the Cu nuclei lhe motivation for using nudd in a metal is that conduction elcctrons help ensure rapid Ihconal contact of Ianice aod nuclei at the temperaturc of the first stage. The present recordS for a spin temperature is 2BO pK. in rhodium.
PAHAMACNt:TIC SUSCEPT,BIUTI' OF CONDUCTION EI.CTRONS

We are going to try to show how on the b.lsis of these statistics the fact that many metals are diamagnetic, or oCily w.ally par:lDUIgIlClic, <.'3Jl be brot'J!!N into agreement with tbe existence of a mag.l<:tK: moment of the electrons.
W. Pauli, 1921

Classical free elcctron thoo!"}' gives an unsatisfactory account of the paramagnetic susceptibility of the conduction e1edmns. An electron has associated ",;Ih it a magnetic moment of one Bohr magneton. IJ.B. One might expect that
'P". Kurti, F K II. Robimon. 1'. I:. Simoo, ano.! D. A Spohr, P"ature 118. ~ (I9:;6t; for ",,,ie.., sre K Kurti, Cryoge:ni<:s I. 2 (I9O), Ad". in Cryogenk r....l;lnerring 8, 1(1963). sp.

J-

lIa~ .o nen et at, PIII/S. Rft>. UtI 10,2818 (1993).

Ipot'"

fidd ...C

1.
, ,

, , 't ~ ,

"'r<:-~,----~",----_~",----"301',

i. !
~'~':-~~"".~--'--}-,----i.--'l,
IHlu.d Nr io' 1000CIK
Fi(tUre 9

r-;.,lcar J..m.ogncliulions of coPP" lII",lci i" 1M mdal, il.rti"g frolll 0.012 K mel
(Aft...... \1

v.lnOUS

roe-kH

\' lIobdao and N. Kurti.)

the conduction electrons ",.ould make a CuriCt)1lC paramagnetic contribution (22) to the magnelization or the melal: M "" Np.iBlkBT. Instead it is observed that the magnetization or most normal nonfcrromagnctie mctals is independent of temperature. Pauli showed that the application of tile fermi-Dime distribution (Chaptcr 6) would correct the theory as rcquir~:.'(l. We first givc a qualitative explanation of the situation. The result (18) tells us that the probability an atom will he lined up par,lllel to the field B excc{.-ds the probability of the antiparallcl oriellt,ltion by roughly p.BlknT. for N atoms per unit volume, this gives a net magnetization -N,.,.2Blk B T, the standard result. Most conduction electrons in a metal, however, have no possibility of turning over when a field is applied, because most orbitals in the fermi sea with parallel spin arc already O{.'(:upicd. Only the electrons within a range k,/f of the top of the Fenni distribution have a chance to turn over in the field: thus only the fraction TITI' of the tolaillumber of electrons contribute to the susceptibility. Hence
M~----=--B

N,.,.2B

N,.,.2

knT

T,.

k/jl',.

whjch is indepelxlent of temperature and of the observed order of magnitude. We now calculate the expression for the p....rarn,...gnctic su~ptibilit)' of" free electron gas at T <C T F' We rollow the method of calculation sugg~ted by Fig. 10. An altcnldtc deri"'atioll is the subjed of Problem 5.

-..

Figure 10 Pl>uJi par.unagrlIelism ill .hsolute UfO; the orbitals in t~ dulled regOl~ in (:oj are occupied. TIle numh",rs of elcctru's In the wur~ and ~OOwnM hand "illatlj"sl to mal", the eocrgies eClual at the Fer"'i IeIT!. 1he chem,cal potentlal (Fermi level) Qf the IlIUllIen! up electrons is ""lila! to that of the "1OfTI(',,1 do...n doclrolls. In (h) we show the, t">ll;CSS of m()Inenl up dedroPs '" 'b" magnetic Geld.

The concentration of electrons with magnetic moments pamJlc1 to the magnetic field is
N... ::: 2

If"

_lOa

dE D(E

+ ,ul) a -

Ii"

dE D(E)

+ -".8 D(IfF)
2

wriltt:n for absolute ,.ero. Here iD(E + J.tB) is the density of orbitals of one spin orientation, with allowance for the downward shift of energy by - pB. The approximation is 'written for kaT <C If,.. The conccntmtion of electrons "'ith magnetic moments antiparnllcl to the magnetic field is N_ = -

2 ,.,

If"

dE D(E - p.B)

iI! -

Ii"

dE D(E) - - p.B 2

D(E~')

111c magnetization is given by M = IJ.{N... - N_), so that

M :: ~t. D(EF)B :: - - B
3N~2

2kST F

(42)

with D(E,:) = 3N/2EF '" 3NI2kIl T,. &om Chnptcr 6. 1be result (42) gi\,cs tile Pauli spin magnetization of thc conduction electrons, for kilT <C Er. In deriving the paramagnetic susceptibility. we ha,'c supposed that the spatial motion of the electrons is not affeded by the magnctic field. But the wa\'cfunctions arc modified by the magnetic field; Landau has shown that for

<:"
free ckdrons this causes a diamagnetic moment equal Iv - j of the paramag_ netic moment. Thus the total magnetization of a free electron gas is
M
0 __ IJ

N",' B kBTf'

(43)

Ik.fore comparing (43) with the experiment we must take a<''COUllt of the diamagnetism oflhe ionic cores, of baud effects, and ofelectron-elcctron interactions. III soclium the interaction effects lilt.T ease the spill susceptibility by perhaps 75 percent. The magnetic sUM:eplihility is considernLI)' higher for most transition metals (with unfilled inner electron shells) than for the alkali metals (Fig. II). lbc Iligh values suggest Ihat the density of orbitals is unusually high for transition metals, in agreement with measurements of the electronic heat cap..'tCity. \VC saw ill Chapter 9 how this arises frolll band lhl.-"Ory.
SUMMAR\' (In CCS Units)

The diamagllctic suS(,'Cptibilily of N atoms of atomic uumber Z is X = -Zc 2N(,-2)/6mc. where (,-2) is the mean square atomic radius. (Langevin) Atoms with a pem13nent magnetIC moment p have a paramagnetic susceptibility X = Np.2/3J;lJ'f, for p.B 4 klJ'f. (Curie-Langevin) For a ~ystem of spins S = i. the exact J\'p, tanh(J..tBlklJT). whcre p. = igp,lJ. (Brillouin) magnetization is
!If =

TIle ground state of electrons in the same shell havc the ma.l:imum value ofS allowed by the Pauli principle and the maximum Lconsistcnt with this S. The } value is L + S if the shell is more than half full and jL - SI if the shell is less than half full. A cooling process operates by demagnetiz."Hioll of a paramagnetic salt at (,'011slant ('ntropy. The final temperature reached is of the order of (80.18rrin;"al> where 80. is the effective local Held and 8 is the initial appliL-d magnetic field.

11\(~ !luramaWlctic susceptibility of a Fermi gas

of colldl1ctiOIl electrons is X = 3NJJ-2/ZE F illdcpclldcnt of tcmpcratul'e for kB'r 4 ~" (Paltli)

Problems
I. Diamagnclic WSul,tibiIit!/ 0/ olomic hydro~'l. TIle wave fUT\(Ction of tho: h)Jrugen 3tom in its grollnd stale (Is) is I/J" ('T(l~rllll e,;p(-rlDo), "here 00 = "~/mr 0529 X 10- 8 em, The dw-ge densit) is p(x, !/' z) = -e1I/1l~. according to the statistical interpretation of the wa~'e functiOO. Show !llat for tillS slate (r) - 3ll~ and calculate the molar Jia.magnetic sU5Ceplibilit), of atomic h)'drogen (-2.36 x 10 6 cm'/Ioole).

""

"s

, '\

\
\
To

t.,
"

, ,

.
,

1/

V
\

.--c.

I/(

-1\r"
1/
ro,

f--"

"
\1" To

., "

'T llbi200 400

6(l(I

~-

"r
hOO 1000 1200 1400 ](;01') IliOO 2llOO 2200

()

T. I" K

.iJ:ure II Tcmpcrature dCIlCuJcnre of the Kril'$linl;\II.)

lI\a~lletic

SUSl:ClltiLility of mcl(Us. (C',uurlcli> of C.

J.

2. IIund rules, Appl>' the lIund rub to find the ground state (the basic le\'el in the notation ofTaLie I)of(a) Eu++. in the ronfigul1ltion 4[' s?p6; (b) YIJ3+; (c) Th3 ' The results for (b) and (c) arc in Table 1, but )'0\1 should give the sqxlratc steps in app!) ing the rules.

3. Triplet ~xcitnllitatn. Some org.mic lnolealies have a triplet (5 "" I) eltCited state at
an energy ka6 above a singlet (S = 0) ground state. (a) Find an expression for the magnetic moment (p.) in a Held B. (b) Show that the suscepCi!.>ility for Til> 6 is appl'(lJ[ilOiltel)' independent of 6. (c) With the help of a diagram of energy Ic.e1s \'eniUS field and a rough ~ketch of entl'O{l)' veDUS Held. explain how this s)'Stcm might be cooled b)' adiaLahc magnetization (not detn~IF...lion),

4. Hoot copadlyFom inlenwl (legren offrcedorn. (a) Cunsider a t\'A)-level system with an energ)-' splitting kflii between upper ,md lower states, tile splitting may arise frum a magnetic field or in other ways. Show that the heat capacit) per system is

c ('U)
aT
to

k (li/T'f,fIT . H (1 + ,fIT'f

111e functiun is plutted in Fig. 12. Peaks oftl1is type in tile he-.lt cap.lCit) are often "nuwn as Schott!.." anumillics. The ma'l:imum heat cajJ.lCity is quite high, LlIlt for T ~ li and for T ~ li the heat capacity is luw. (b) SlIuw that for T ~ ii we have C a k ll (A12T)2 + .... TIlI~ hypcrfine internction bdween Iludear and electronic magnetic muments in parnmagnetic salts (and In systems havmg electron spin order) causes splittings with li "" I tu 100 mK. These splittings are uften dek:ctw experimentally by the presence uf a term in W in the heat capacity in the r~ion T P ii. Nudcar electric quadrupole interdClions (see Chapter 16) with crystal fleld~ also cause splittings, as in Fig. 13.
5. l'(mU Sinn susceplibilily. The spin sll<iCCptiuility ofa (:undudion electron gas at allsulute zero may be discussed by anotlu:r method. Let

be the C{)Jlccntratinns of spin-up and spin-(Io"n electrons. (a) Show that in a magnetic field B the total energy of the spin-up band in a free d<.'dron gas is

where Eo = ~N~f" in terms of the Fenni encrgv ~" in zero magnetic field. Find a similar expression fur -. (b) Minimize ,otal = + + - with respect tu {and sulve for the C'()uilihrium value of (; in the approximatiun {~ l. Go un tu sllow that the mdgnetizatinn is At .,. 3f',1p.2BI2~", in agreement with Eq. (42). 6. CUfl(luction electron ferronwgnetism. We approximilte the effect of exchange interactions among the conduction electrons if we assume that electrons with pard.llel spins interact witll eaell uther with energy - V, and \' is positive, while electrons with antiparallel spins do not interact ",ith each uther. (a) Shuw witll the help of Problem 5 that the total energy uf the spin-up band is
+ = ,,(1

+ ({"'J -

jVN~1

+ (f

-INp.B(1

+ () ;

And a similar expression for E-. (b) Minimize the tutal energy <!nd solve for {in tile limit {~ l. Show that the magneti7.<ltiun i
3I\'p.~

AI -

2~"

i VN

B,

so that the exchange illteraction enhances the susceptibility. (c) Shuw tllat ",illl B = 0 the total energy is unstable at (; = 0 when V > 4f;"f3.V. If this is satisfied a ferromagnetic state ({ "" 0) will have a lower enelb'Y tllan the paramagnetic state. Because of tile assumption {<g L, this is a sufficient condition fur ferromagnetism, but it liMY not Le a necessary conditiun. It is knuwn as tl.e Stoner condition.

",-~-~~-~-~--~--,

.;: 03

, ,

-.,--k>d!

_--,-I__ u.. ,

J"
n,

"~--+---;,~_:::;,:::::=.!,,==,,,P=~6
~.

TlA

f'lpare 12 11.....( capoM"I'y of II hoo.... k. ......II~.,m as 1I runo;1ion ofTlIJ.. whcro~ A j, lhe k-\..,I sphtting.. "r1>c: Schutlkv .u'IOI"",ly is a '"cry useful luuI lOr doMennin;"!!: cn..,.gy 1e,..,1 Ipl,ttinp of ion5 in rareNllh and Irmition1t"t"'P n><:tak. c:umpountb 00 a!Io)'$.

''',-~-~-~-~---,

..

"
(e T-~ alld c"uduc.lion ek<-tmn (e Icnl.(X"r.oIurtl. ("~ N E. Plullips_)

uoal

UOOI
1~. ill

0000 K'

Duo

Dol

i'lgune 13 The: _~-i1alf' ~at CIIII_ity of pllium at T<021 K The ,,,1dNr q....rupolc 0: f) c.'odrib"hOlls domm.otc the t.e..t OIl>acity at .~ry k-

410

(
7. Two-level system. The result ofl'roLJem <I is often seen in anot".,r furm. (a) If the two energy levels are at 6 and -6, shuw that the energy and heat cap<lcity are
U = -6 tanh(!:J.IknT) ;
(1) If the system has a random composition such IIlat all values of 6 are equally likely up to some limit !:J.o. shuw that the heat capacity is linearly propurtionalto the tem_ per.ltu.-e. pruvided kllT.o::g 60- This result was applied to the Ileat capacity of dilute magnetic alloys by W. Marshall. PllYS. Rev. U8, 1519 (1960). It is also used in the theury of glasses. Chapter 17.

8. Paramagnetism oIS = 1 system. (a) Find the magnetization as a function of magnetic Ileld and temperature for a system of spins with S = I. mument IJ-, and concentration 2 1'1. (b) Show that in the limit p.B .o::g 1.1' tile result is M ;;a- (21'1J,t f3kT)B.

References
A. Aimlgam and B. Bkanc)', EI""tron parm1l<'lglletU; resorwn(.lj' ()f trall$itiQn iQns, Dover, 1986. H. B. G. Casimir, MagnetISm and ury low te:",peratutu, Don'r. 1961. A classic. M. I. Darby and K. N. R. Taylor, PIIYsic$ of rare 000111 so/ids, Hal.skd, 1972. A. J. Freeman. rIle actinides. electronic structure and related propeo1is. Academic, t974. R. D. Iinmon, Principles alld applit::alions Qf magnetit:: eco/illg, Elsevier, 1972. H. Kno"llfel, Pv/.sed 1llgh magneticf~lds, North-Holland, 1970. O. V. Louna$maa, Experimental principles and mf!lllods below 1 K, Acad..mic P, ...."'. 1974. L. Orgel, Introduction to transition melal cllemi.$tn;, 2nd ....J., W,k-y, 1966. !. II. Van V1l'Ck, 1'he tlleQry of electric and magnetic susceptibil/fies, OxfOrd, 1932. SUlle,b deriva-

tions of basic theorems. C. K. While, ":l'perlmf!ntal techniques in 10", lempcmtur41 physics, 3rd cd., Oxford, 1987. R. M. White, Quantum tlleery of magnetism, Springer. 1982. A. J. F....-crn"n and C. H. Lander, lIandbook 011 tIle physics and dleml$/ry of/he =tiI,wes. North Holland, 1984~1993. M. D. Sturge, "Jahn-Tellt:r errect in solids," Solid stste physics 20. 91 (1967). M.C. M. O'Brien and C. C. Chancey, "ne JahnTdlcr err...'CI: An inlroducti"" and current rev ... " "," Amer. J Physics 61. 688 (1993).

15
Ferromagnetism and Antifer1"Omagnetism
FERROMAGNETIC ORDER Curie point and th~ c~change integral Tcmllerature dependence of the saturlltioo magncliution Saturation magnetization at absolute zero
~~C~ONS

443 443 446 448

4~

Quantization of spin "1/o\'C5 lbcrmal excitation of rnagnons NEUTRON MAGNETIC

SCATI'EI\I~G

454

455 456

FERRJMACNETIC ORDER Curie temperature IUld susceptibility of ferrimPgnels Iron gamets

ANnFERROMAGNt.,'TlC ORDER Susceptibility below Ihe "'eel temperature Antierromagnetic nJugnons FERROMACN... " C DOMA1f'\S
Anisotropy energy

4.58 460 461

462 465 466 468
470

Transition regkxI bet'A'eeJl domains Soliluos

472 473

Origin of domains Cocrcivity and hysteresis

SINGlE DOMAIN I'ARTICLES Geomagnettsrn and biomagnctism Magnetic force micro5COp)'

474 475
471 478 478

MAGNETIC BUBBLE DOMAINS

SUMfo,lARY

480
481

PROBLMS
1. Magnan dispersion relation 2. Heat capacity (If magnons 3. Nee! temperature 4. Magnctoeillstic coupling 5_ Coercive rorce or a small particle 6. Saturation magneti'lation ncar T.. 7. N6e1 wall
REFERENCES

481
481 482 482 492482483 483

483

NOTAnON, (CC5) B "" " + 41T1\f; (51) B - IJ.(/..H + M). We call B" the applied magIletic field ill both systems of units: in CC5 we have B" - H" and in 51 we have B. ~ . 1he ~sceptibilit) is X = MIB. in CC5 and X - Mm. - ~f/B. in 51. One tes.la" I <t gauss.

111111

I I1I I

I \ I \I
4t2

B~ ""-

--='"", '--

=-,
band

II..tM:al.pon array

F~""",.g" . 1 '" elM'')'

CIIAPTER

15:

FERROMAGNETISM AND A.vnFERRO~1ACNI:."ISM

FERRO"IACNEllC ORDER

A ferromagnet has a spontaneous magnetic moment-a magnetic moment evrll in 7.c.:ro applied maglJetic field. The existence of a spont.meous moment suggests that ek'Ctroll spins and magllctic monll:nts are drranged in a regular malineI'. The order need not be simple: all ohhe spin arrangemellts sketched in Fig. I except the simple antiferroll)agnet have a spontaneous magnetic moment, calk"l! the saturation moment.

Curie Poillt u,ItJ tile F-:n:lwnge Integral

Consider a pardmagnet with a conccntration of N ions of spin S. Given an intemal interaction tending to line up the magnetic moments parallel to e3l:h other. we shall hdve a ferromagnet. Let us postulate such an interaction and call it the exchange field. I The orienting effect of the eM:himge field is opposed by thermal a~it:ltion, and at e1emted tempemtures the spin order is destroyed We treat the exchange field as equIvalent to a maguetic fieM Df 111e magnitude of the exchange field may be as high as 107 g;lUSS (lQJ tesla). We assume that Bt: is proportional to the magnetiz..1tion M. 111e magnetiz..1!ion is defined as the mai.'netic moment per unit volunlC; unless otherwise sJX."Cified it is understood to be the value in thermal equilibrium at the temperature 'f. If domains (regions magnetized in different directions) are present, the magneti7.. 1 tioo refers to the value within a domain. In the mean field approximation we assumc each magnetic atom experiences a field proportional to the magncti7A1tion:
DE = AI\1 ,

(I)

where A is a eomtallt, independent of temper,\tme. According to (I), l'ac.:h spin sees the average magneti7.ation of all the othcr spins. In trllth, it may sec oul}' ncar ncighbors, but our siJnplifk:ation is good for a first look at the problem. '!llC Curie temperature Te is the temlx:r.Lture above which the spontaneous magnctiztltioll vanishes; it separate" the disordered paramagnetic phase at T> 're from tIle ordered ferromagnetic phase at I < T.,. We Colli find Ie in terms of A.
I Aho calk<! In.: lOO~ field or the w...~u 6dd. afl..... JYrT~ w........ ho "'0lS the flnl 10 liuch II Gdd. 1lr ~tthange 6eld B, .. mulat~li.reall~>cticfidd in thcUjln'SiOOl'lli ~ the enngy -~. Br &Old the lorque,..)( BoE 001. m~thc ITIOll1oCllt 1Jo. Bul B. Iii not r.."Oy. "",nchc fl<.1d ollld Ihcn:l'ore ~ not mt~r mto tilt' \tU"ul ~u.otlOnli. /i)r eumpk-. lJxono is 110 aun~,1 denliity j n:-Ialcd 10 BI: bJ.' curl II .. 4"yc:. 1lr lIIagnitudr of BI: 15 1)-pK.'ally 10' ~.,. than 1M "'<t.'Rge ~hc flC'" ri IlK- magnrtlC dipolrs of lhe: b~t.
irn;q!.m~

...
Consider the paramagnetic phase: an applied fide 1.>" will cause a finite magnetization and this in tum will cause a finite exchange Geld Bf ;. If Xp is the paramagnetic SLJs(''eptibility,

(CCS)

M = X,.(B" + Bd ;

(SI) 1'<1>1 ~

x,l.B o + BEl

(2)

11'1e magnetization is equal to a constant susceptibility times a field only if the fractional alignment is small: this is where the assumption ellten that the speci_ men is ill the paramaguetic phase. The paramagnetic susceptibility is gh'cn hy the Curie law xp = err, where C is the Curie constant. Substitute (1) in (2); we find AlT = C(B. + AM)

.nd
(CCS)

X= -

AI C = B. (T - CAl

=-''-=-

(3)

'J11C susceptibility (3) has a singularity at 'f = CA. At this temperature (and below) there exists a spoutalloous magneti"'~'ltion, hecause if X is illfillite we can have a finite M for zero Bd From (3) we have the Curie-Weiss Illw

(CCS)

-"0r--'c;T;- ,
<

(4)

11lis expressioll describes fairly well the obsened susceptibility variation in the paramagnetic region above the Curie point. Detailed ca1C\Jlation~ predict

x 0: ("r _ T..)I.33
at temperatures very close to 'f,~ in gencl<ll agrecmcllt with the experimental data summarized in Table I. 'Ole reciprocal susceptibility ofniclc.el is plotted in Fig. 2. From (4) aud tbe UCfillilioll (14.22) of the Curie constant C we may determille the value of the mean field constant A in (I): ICCS)

(5)

For iron Tc - 1000 K. g'" 2, and S ... I; from (5) we have A - 5(X)(), With M. - 1700 we have BE == AM - (SOOOX1700) - 101 C = 103 T. The exchange field in iron is "cry much stronger than the rcal magnetic fieid due 10 the other magnetic toilS in the crystal: a magnetic ion produces a Geld - p.ilt? or about lQ3 G = O. I T at a neighboring lattice poiut.
T.. is gi,-al quite .ca,ralely hr eler - (I), wht....e 9 Tc- St."" the ft.'\"icw by C. Domb ill Magntfls"I, \'01. 2A, C. T. Ratio II.nd H. $11hl, t...:ls., Alwl~mie Pn,~s, 1965
~E"pcrjn\Cnl... i~ apllft.'(:illbly RrCll.ter than the lIClual tl1ll\5ilioo tClllllCllIh"e

n)' the 5Usa.1)tiWity ror T ~


T:l...... I

'7

Critical point

c~ts (or

fen-omagneb

As T -+ Tc from above, the susceptibilil)' X becomes proportional to (T - T~)-7; as T- Tc from below, tile magnetization M. becomes proportional 10 (Tc - T)1l. In tl.e mean field approximation, 'Y ". 1 and fJ "" I.
y

P
0.015 0.04 0.02 0.1 0.02 0.02 0.34
:!:

'r~ in J(

Fe
Co Ni Cd

C.o,
C rBr3

1.33 1.21 1.35 1.3 1.63 1.215

:!:

0.04

10<3

l38B
0.42 0.07 627.2 292.5 386.5
3~56

EuS
EJlj>nimcntal dab. oolIcctoo by II. E. Slanley.

0.368 :!: 0.005 0.33 0.015

16.50

, ,
,
/

./

V
'"'

'"

l/
/
~

'"

f'igu~ 2 Rcriprocal cl the SlIJC'CIllibility pet" J:J'UTl cl nidel in the ~ir;hl:orhoOO cl the Curie lemprntun: ~ "The dt:"slt) isp. 11lc dashed line Q; a lin""'ro~lioo from high lenlpCntuteS. (Aft" P. Weiss lind R. .'orr.,...)

The ~ehange field gives an approximate rcpresentation of the quantummechanical ~change interaction. On certain assumptions it C'.ln be shown3 that the energy of inter.v;:tion of atoms E, j bearing electron spins S" S} contains a term (6) where} is the exchange intcgral and is rel.tled to the o\'erlap of the charge distributions of the atoms i, j. Equation (6) is ('".tiled the Heisenberg model. TIle charge distribution of a ~,.stem of two spins depends on whether the spins are parallel or anti~.J.lel,4 for the Pauli principle excludes two ele<.1:rons of the same spin from being at the same place at the same time. It docs not exclude two electrons of opposite spin. Thus the ele<trostatie energy of a S)'5tern will depend on the relative orientation of the spins: the difference in energy defines the eM:hange energy. The exdlange energy ortv.o electrons may be written in the form -2}sl . ~ as in (6), just as if there were a direct coupling between the directions of tile ,"'0 spins. For many purposes in ferromagnetism it is a good approximation to tTC'at the spins as classical angular momentum vectors. We com establish an approximate connection between the exchange integn,] J and the Curie temper,lture T". Suppose that the <ltom under coosiderdlion bas :t nearest neighbors, each connected with the centr.ll atom by the intemction}. For more dist.mt neighbors \\'e take} 'IS zero. The mean field theory result is
] = 2zS(S + I)

(7)

Dettcr statistic'll tlpproximations give somewlmt different results. For the se, bee, and fcc structures with S "" t Rushbrooke and 'Vood give kI.l7".,1:t} "" 0.28; 0.325; and 0.346, respectively, as comp'lred wilh 0.500 from (7) for all three structures. If iron is represented by the Heisenberg model with S"" 1, then the obselVcd Curie temperature corresponds to} "" 11.9 IneV.

Temperature Dependence of tile Saturation Magnetization

'Ve can also lISC the me<ln fielel approxillliltion below the Curie temperature to find the magneti:wtion as a function or tempemture. 'Ve proceed as
3See most teds on quantl"n theory; also J. II. Van Vkdr. R..:v. Moo. Pll)"'. 11, '1:1 (1945). 11Je origin of~.gr in insulatOf'S i$ r... 'Vieweo.J by P W Ander$Oll m Rado and SuM, Vol. I. 2S (1963). "1 metals by C. Herring in Vol. IV (1966). 4lftv,v spi"s are anliparatlel, the wa,-d'unetions of the tv,o eledf'Ofls must be symmetric, as in the oombi.nation II'r,MrJ + u(r.).;(r,). If the two spint .e \'lI1Je!. the Pauli principle rcqul/d that the orl>ilal p;u1 of the "",vefunchon be anhs)'mmetric:-, as in lI(rJv(rJ - u(r.).;(r,), lOr ~ If we int~ the <.'OlJI'dinatt'S rio rs the wavnund.ion d~ sign. If we $et the posot....... equal >0 t....t 1'1 - r.. then theantisymmctric fund.1Ofl ''IIfliJhes, IbrparaUelspins th.-re isuro probab.l,t~ of finding the "'..., da'lrnns at the same poution.

..
08

-~.,<---"""" b'~05
"" I " I

"
"
o""'----'-_-'---:'_--":--,L----1
o
02 0,4

.06

08

10

12

Figure 3 Crnlll'icoll solutio" of Ell. (9) for the reduced ml\l{uellzalion no "Ill function of lemp<,rature. The rl'dUl.......I mllgJleli:utlon;5 defined liS'" - MIN,... 11,c lcfthand m!c<'iI-:'1. (9) ilplolled as II )lrd,ght line '" with unit ~Iopc The right.haIld i;de '" 1.."10(",/,, and iii 1l1ottt:d vs. '" for three <li/felt'nt \'31""5 ..the ...-dua.~1 taTIpeniture' .. k,.TfNjI."A .. Trr.. ",.. thl'('ec:un-n corTcqlond to the lempnaturcs 'rTr. 7. and O.ST~. 11le "'''''e b- t .. 2 inter>ts the str..w.t line ,>I onl}' at .... 0. .... ~e for the p.&nnugnd:ic r~ (Ibrre is 1M) IDlnnal app\k...J nugnetie fi.Id). Tht CUI"\.'efort .. I (nrT '= 7.1 IS t:anp:nt to the ).lrai!4>t line -' lhe~, this tnnpc.-rature n>a<ks the OIud d'i'eTTomagnetism. ",.. .,.,n~ b I .. 0 S IS in the Je~tit' region and i"lers1s the " ....id,! line '" .II dbout m - 0.94loo'1-'. As , - 0 the i"lc 1l1O\'a Ilil to III .. I, SO that .all magnetic moments <Ire lined lip .., absolute zero.

..,>1

before. but instead or the Curie law .....e lise the complete Brillouin expression 'Or the magnetization, For spin i this is AI = Np. tanh{p.Bfk 8 T), If """e omit the applied magnetic field and replace B by the molecular field Bt:; = AM, then
AI = Np. tallh(p.AMfkDT) ,

(8)

We shall see that solutions of this equation with nunzero M exist in the temper,Iture range between 0 and T e . To solve (8) we write it in terms of the reduced magneti:llltion III MfNp. and the reduced temperature t - k 8 TfNp.2 A. whence
m = tanh(mfr) .
(9)

We then p~tthe right alld left sides of this equ;ltion scJXlroitely as f\lIlcliolls of m, as in Fig. 3. The intercept of the two CUTVCS gives the value of III al the tcmpcr"turc of interest. The critic,ll tempcrahlre is t = 1, or Te = Np. 2 Mk a. The CllTVes of fl.J versus T obtained in this way reproduce roughly the features of the experimental results, as shown in Fig. 4 lOr nickel. As 'f increases the m"gnetization decreases smoothly to zero at T = Te - This behavior classifies the usual ferromagnetidparamaglletic tr.msition as a second order transition.

4-15

"-

!\o

..
o

\
0

,
o

Figure 4 retical curve for S .. Ii 0" thc mcn" ncld theory. Experimenhll values by P. Weiss and R. Forrcr.

"'. Saturation m;,gnctization of nickel as a function oftcmpc....ture, together with the theo-

'.0

The mean field theory does not give a good description of the variation of M at low temperatures. For l' <Si T c the argument of tanh in (9) is larb'C, and tanh
~ l1!!

1 - 2e-2t. ..

To lowest order the magnetization deviation !:J.M - M(O) - M(T) is

(10)
The argument of the exponential is equal 10 -21~n. For T = 0.1 T c we have AM/NIL l1!! 4 x 10-9 . The experimental results show a m\lch more mpid dependence of l:J.M on tempewture at low tempemturcs. At T == O.ITc we have I1MIM 5!!: 2 x 10- 3 from the data of Fig. 5. The leading term in 11M is observed from experiment 10 have the form

(Il) "1\ere the constant A has the experimental v'Llue (7.5 0.2) x 10- deg- m for Ni and (3.<1 0.2) x 10- 6 deg- 3n for Fe. The result (11) finds a natural explanation in terms of spin wave theory. us discussed below.
6

l:J.MIM(O) = A1<312 ,

Saturation MagnetizatiQn at Absolute Zero

Table 2 gives represenlative values of the saturation ma/,,.netization M~. the ferromagnetic C\lrie temper.tture, and the effective magneton I)umber defined by MiO) = naNIL/J' where N is the number offormula units per unit volume. Do not confuse n8 wilh the pammagnetic effective magneton number p delined by (14.23).

. ",

"", .r
",
TI.. , ........"
Mi<klle

'. "

-,
,

" f---

u,""

lUG

r,"

' ..:.

1-.,.

"0 "C

.. ..
"

T. b, to':

.'

"

.,

Figu",S J:kcrease in maglld:wtioo of n!eke! with tempemture, alter Argyle. Olarap, and Pugh. In the I'lot AM - 0 at 4.2 K.

Table 2

F~rromagoetic

CI")'stals

MagnetwtiOll AI., in p"s S"b;lmlCf'

Room tcmpt'r ..tun'

F,
Co
Ni

li07 I"JO 4S5

"
li40

II,.{O tq. Iler ronnul3 unit

Con.
tempcrnture,
in K
1~13

1446 510

Cd

Dr
MnAs MnUi MnSb

""" 2920
670 '''-0 710

SiO
680

2.22 l.i2 0.006 7.63 10.2 3.' 3.52 3.5

~03

138B
6Z7

292

31B

..

,>30

crt>,
MnOFeA F..o()FC2~
NiOFc2~

515 410

400
Z70 135 110

5.0 '.1
~.

587 386 5i3

658 658

CuOFcA
Mb-oFC2~

1.3
I.J

EIlO
Y3F~OI2

J920
130
200

6.B 5.0

728 713 6!J 560

Obscrvc<1 values of nll ;Ire often nOllintegral. There are many possible causes. One is the spin-orbit inter..lction \\hich adds or subtracts some orbital magnetic moment. Another cause in fel1"Omagnetic metals is the conduction electron magnetiz;ltion induced locally ;Ioout u pammugnetic ion core. A third amse is suggested by the drawing in Fig. 1 of the spin arrangement in a fcrrimagnet: if there is one atom of spin projection -S for every two ;ltOITIS +S, the average spin is is. Are there in fact .my simple ferromagnetic instll.lIors. with ;Ill ionic spins parallel in the ground st<lte? '111e few simple ferromagnets known at present include CrBr3. EuO, ,md EuS. A buml or itinerant electron mexlel s accounts for the ferromagnetism of the transition metals Fe. Co, Ni. The approach is ilUlic-.lted in Figs. 6 and 7. The relationship of 4s and 3d bands is shown ill Fig. 6 for copper, which is not ferrOmab'lletic. Ifwe remove one electron from copper, we obtain nickel which has the possibility of <I hole in the 3d b.ll1d. In the b;lm! structure of nickel shown in Fig. 7a for T > T" v..e have taken 2 x 0.27 = 0.54 of an electron away from the 3([ band and 0.46 away fwm the 4s buml, as compared ",ith copper. The band structure of nickel at absolute zero is shown in Fig. 7b. Nickel is ferrom<lgnetic, and <It ab~olute zero nll = 0.60 Bohr magnctons per 'Itom. After ;Illowance for the mab'lletic moment contribution of orbital electronic motion, nickel has lin excess of 0.54 electron per atom having spin preferentially oriented in one direction. The exchunge enhancement of the susceptibility of metals was the subject of Problem 14.6. The properties ofamorphous ferromagnets are treated in Chupter 17.
MAGNONS

A magnon is aqu;mtized spin wave. \Ve lISC a classical argllment, jllSt as we did for phonons, we find the magnon dispersion relation for w versus k. We then quanti7.e the m<lgnon energy and interpret the qtJantizatioll in terms of spi n Ievers<l!. The ground state of a simple ferromagnet has all spins pamllel, as ill Fig. &1. Consider N spins each of lnagnitude S on;l line or a ring, with nearest neighbor spins coupled by the Heisenberg interaction:
N

U'=' -2]

L ,,"'1

Sp'Sp+l

(12)

Here] is the exdl;mge intcgnll and (lSI' is the angular momentum of the spin at site p. If we treut the spins S" as classical, ectors, then in the ground state 5,,' 51'+1 = S2 and the exchange energy of the system is Uo = -2N]S2.
5C. IIcrring in Rado and Suhl, \'01. IV (I966).

11 cV

f, L

.~

......
>I

riel

I
>II
5~.:d ......

follnl-IO drcl.......
(0)

Figure 6a Schematic- rclaliomJup of 4r and 3d Iwlds in 1I1el.;o]W copper TIw: 3rl1.lllnd OOIds 10 dectmns per alOfn and t5 Illkd in <.'Opiler TIoe 'b l-.d can t.:>Id t.-v dcdrons per alOrn; .t i~ ,hown Iw.lr-fillct.l, ... <:ollpet has olle ''ak-ncc d1.ron onbidc the lillrd 3d shell.
Figure 6b 11'0: Filled 3</ bnnd of ~~)I'pcr .hown as two R1",,,tlc Jub-b;"lds of 0l'l'o>itc clcdron spin orientation, L-ach b.~nd holdinK lhe dL~'h1)fl$. W,tI, hoth >ub-I,.Ulds fllle<! ol> ~hown. the net spin {and hence the net m.ag..... - tiUtion) of l10c ,1 h"nd .. ~.

'"

"" , ....

GSQ
4.13

..........

,'=
~

,*""",,,. >I,

"',

>I'

>It

Fipln: 7a n..nd "''''tiooships in nid<1 above tho: ClIne lcrnper.ltu", Ybr net rnagJloe(ic-l'IOIlmf:nt ,s zero, as there ue eq<W numben cl tdcl; in both 3d! ilInd 3d T bands.
3d! sub-bands llre sepamtcd hy an exch.inge Irlt<'nll1ion. The 3d 1 band is filk-d; the 3d! oolld ~nta;tl$ 4.46 electrons llnd 0.54 hole. TIle 4$ baml 'I usuilly thoughlto <.'(lfltaill ajlprCD"",alely equal numhCr$ of electro..s ill botl, Spill ui~ions, and 50 we not troubled to divK1...t I"to sub.b.onds. The nM ~elil: nw:menl 01'0.54 1'. pee atolla ariK's rrorn the neftS populatlOfl oCthe 3d 1 t-.d o-u the 3d 1 t-.d. It IS ofien c:otwenientlo spral< oI'lhe magncta;ahon as ansiug from the 0.54 hole in the 3d! band.

fig,,",1I> Schematic" ..,LationsJ.'llofWnd. in ,tid;e! at absolute Kro. TIle ~'1t"'s oftl"" 3d f and

"-'W:

452

,
What is the energy of the first e,;cHed state? Consider, _excited state with one particu];lr spin revenoo, as in Fig. 8b. We see from (12) that this increases the ener~'Y by 8JS 2, so that VI "" Vo + 8JS2. \Ve can form ,1Il excitation of much lower energy ifwe let all the spins share the reversal, as in Fig. Be. The elementary excitations of a spin system have a w;welike form 'lIld are called magnons (Fig. 9). These are analogous to lattice vibrations or phonons. Spin w.lves are oscillations in the relative orientations of spins on a lattice; lattice vibrations are oscilllltions in the rehltive positions of atoms on a lattice. We now give a classical derivation of the magnon dispersion relation. TIle terms in (12) which involve the pth spin ,Ire

(13)
We write the magnetic moment at site p as P" "" -gp.8Sp. Then (l3) becomes

(14)
which is of the form - P,> H", where the effective magnetic field or exchange field that acts on the pth spin is
(15)

From mechanics the rate of change of the angular momentum "Sp is equal to the torque Ill' x 8,> which acts on the spin: ft (!S,,1dt = PI' x HI" or

dSpMt "" (-{!;tuff!) 5" x 81' = (2JlftXS,> x 5,,_1 + S"x Sp+I).

In Cartesian components

(16)

(lS:J(lt "" (2J"t)(S~(S~_1

+ S:,~ I) -

S;,cS~_1

S~'I)]'

(17)

and similarly for dS/Vdt and dS~hlt. These equlltions involve products of spin components and are nonlinear. If the amplitude of the excitation is small (ifS;, S~ 4i S), we may obtain an approximate set of line,lr eq\lations by taking all S~ "" S and by neglecting terms in the product of S~ and SII which appear in the equation for (!szhlt. The lineali7.ed equations are

dS;hlt = (2JS/")(2SX - S~_I - S~+I) ;

dSr/dt = -(2JSIfl)(2Sr, - S~_I - S~+I)

(1&1)
(J8b)

dS;Jdt = 0

(19)

By a1wlogy with phonon problems we look for traveling 'wave solutions of (18) of the fonn S; = u exp[l(pku - wt ; where u, v lire constants, p is substitution into (J8) we have
3n

Sf, "" v exp[i(pku - wt)) ,

(20)

integer, and u is the lattice constant. On

.-"

111111 111111 ... .. . ..... .,

co::,

..
I

Figure 8 (a) C\a.s)iaI1 picture of the ,;round s~le of a siml>1e ferromagnet, aU spins are p8ra.1.1el. (b) II pouibh: cKCibl.lon. One spin is n~\'Cned. (c) TIle low-I)jog elemental)' CAcltntions art' ~p;n "'lives. TI,e cnds of the SP'" n,l'tors p~s OIl the .um"",. of oones. with sucC'cnivc spins ad.-anced in p'- by a cunstant angle.

F"ogurc 9 A spin W.l''eOn a linc ofspins. (a) TI,e Silins \'iewoo in lJ.enpect'vc. (1)) Silins viewed from w'e, showing 0..., wilvclength. TIle wa\'C i. dluwn throull:h the ends the S\l;n \'eCton.

or

-jW11 = (2)5''')(2 - e- iM

eik1v

= (4}5/(1)(I - ('os

ka)v ;

-iwtl = -(2.JSffl)(2 - e-,k" - e""'}u == -(4J5111)(1 - cos ka)u

1bese equatkns have a solution for u and v if the determinant of the coefficients is equal to zero:

I''

~4JSlfl){l - cos ka)\ == 0

-(4lS/1IKl - cos ka)

(21)

''''
(22)

flW

== 4]5(1 - cos ka) .

This result is plotted in Fig. 10. With this solution we find th;\! v = - ilol. corrc.sponding to circular precession of each spin about the z <lJtis. We see this on taking rC'oI1 parts of (20), with tJ sct equal to -til. Then

S;

:= II

cos{pka - wt) ;

sg ""

II

sin{pka - wi) .

Equation (22) is the'dispersion rclatM>n for spin ",,'aves in one dimension with nearest-neighbor interactions. Precisely the same result is obtained from the q\tantum-mechanic-rt1 solution; see QTS, Chapter 4. At long wavelengths ka <C I, so that (1 - cos ku) 3l i(kuri and

h",

~ (2}Sa')k' .

(23)

1lle frequency is pmportional to 1.1; in the same limit the frequency of a phonon is proportional to k.

00

FIgure 10 Di~lIcrsion rd~t;oll for m.."nOlu ' - intcradlon$.

11\

M fl" ..."nagnc! moue tlimcmior, with "eatl',Hlcigh

The dispersion rel.\tiou for a fClTomagnclic cubic lattice with ne.lreslneighbor interactions

hw - 2JS[%

L ",,(\<. 0)1
('IS"''''

(24)

where the summ,ltion is over the z vectors denoteJ by Ii which join the central atom to its nearest neighbors. For ka <C I,
(,w :

(25)

for all three cubic lallices. where a is the lattice constant The coefficient of J.!l often may be determined accurately by neutron scattering or by spin wave resonance in thin films, Chapler 16. l3y neutron scilHering C. Shir.tne .md coworkers find, in the equation flw "" Dkl, the values 281, 500, and 364 me\' 2 for D at 295 K in Fe, Co, Ni, respectively. Quunti:.ation of Spin WUt."U. 111c quantization of spin v,-aves proceeds ex actly as for photons and phonons. The energy of a mode offrequency Wrt with Ill; mllgnons is given by
fA. = (11k

+i

)flW/r .

(26)

TIle excitation of a magnon corrcspouds !o the reversal of one spin l.

Tllermm Excitation of .UagnofUf

In thennal equilibrium the average "a1ue of the number of magnons excited in the mode k is given by the Planck distribution:G

1 (f1W = exp(hl<.'l!kBT)
The total number of magnons excited at a temper.tture T is

(27)

~ nl.: =

dw D(wXn(w)} ,

(28)

where D(w) is the number of magnon modes per unit frequeucy r.mge. TIle integral is t-.d.:en m,'er the allowed r..mge ofk. which is the Grst Brillouin zone. At sufficiently low temperatures we may carry the integral between 0 and :0 because (n(w -I' 0 exponentially as w -I' 00. Magnons ha\'e a Single polarization f()l" each v-.t.Iue ofk. In three dimensions the 1ll1mber of modes of Wa\'e\'eC'lor less than k is (l/217f(4nP13) per llnit volume, whence the number of magnons D(w)dw with frequency in ltwat w is (l/2nf"(4nl...2)(ltkJdw) dw. In the approximation (25),

Jw = 4]5trk = dk h

2( 2]5u
h

) Ill! wIll!

11lUS the density of modes for m;lb'llOns is

D(w) = - - - - 4~ 2J50 2

1(")" w'o

(29)

so llli1t the totul number of magnom is, from (28),

11le definite integntl is found in tubies and has the v.uue (0.0587)(4w2). 11le numher N of almns pcI' unit volume is Qft?, where Q = 1, 2, 4 for sc, !XC, fcc I"Uices, respectively. Now (1:n"yN5 is equ,,1 to the fraction'll ch.Ulge of magne!i:-.ttion l!MIM(O), whellcc

AM
--~

0.0587 .

'1(0)

(k.T)'"
2JS

SQ

(30)

0'fl1Cl lUKUnlffil is eo:actly as lQr phooonl or phoIom. 1be PIandc diUnbulion rouo..,,; bally problem ....hne the energy It..""k are .dentin! ..,th those J /I harmonic osalbtur or collection of
harl'llOllK-

0K:'i1btors.

This rc!>ult is thc Bloch 1-312 law and has Leen conf'innt..'f:! ~.(perimcntall)'. In neutron scaltcring cxperimcnts spin waves havc Lccn observl.:'(l up to tempera_ turcs ncar the Curie temperature and even above the Curie temperature.
~TEUTnON

MAGNETIC SCA'ITEIUNG

An x-ra)' photon sees the spatial distribution of electronic charge, whether the chargc density is magnetizcd or unmagneti7.ed. A IlCUlron sees two aspects ofa crystal: the distribution nuclei and thc distribution of electronic magnetization. The nt.."lltron diffraction pattern for iron is shown in Fig. 1 J. "nie magnetic momcut of the neutron interacts with the magnetic moment of the electron. The cross section for the neutron-electron interaction is or the same order of magnitude as lOr the neutron-nuclear inh..'faction. Diffraction of neutrons by a magnctic crystal allows the determination of the distribution, direction, and order of the magnetic mOments, The magnctic llloments associated with particular components of alloys may be investig<lle<l by neutron diffraction. Results for the Fe-Co binary alloy systcm (\\ hich is ferromagnctic) are shown in Fig. 12. TIle magnetic lI1oll1ent on the cobalt atom docs not appear to be affected by alloying, bul that on the iron atom increases to about 3P.B as the cobalt concentration increases, TIle maglldizations arc shown in Fig. 13. A neutron can be inelasticall) scattered by the magnetic structure, with the creation or annihilation of a magnon (Fig, 14); such events make possible

or

..
I

(lU~

So:aI....m~ onIle

t"igur., II Kot:hler.)

N"ulr(Hl diffraction .,.,tt"ro for iron. (Aft"r C G. Shull. E O. WollNl, lInd W. C.

"s'

"v-'
,
I
1.0

1
::F.

1;1:t!ure 12 Mommb attnbulabk to 3d...,.

Irons in F e-Co alloys III function

lion, after M. F. Collins and

J.

of COItlpOl.i_ B. Fonyth.

'.

20160_
1~lcobal,-

'00

oS

IL ['.

,.v .,,~

.,~

o
~

.S

I';:'

1'4:
;J
1/;' I

NI-Coo 6 NioCu

..

~b
N~V

1/
/
Mn
7
~

.....
~

.~
.~~ .~

.-

.~~

-"
~

1Io"mrth_l

J
~

I
II

l?' 1'1
~

II

10

II

Eledron

.,.,."..,.,.nolian

.';gOlre 13 A....,mge utomic moments of binary alloys of Ihr. e1"nlflnls in lllfl iron gro"p. {After

I,'igure H Smttering of a neutron by an ordered m.gnrtll: slruc:ture, ..itil creation of a magnon.

".....

<S,
"r--~-~--~-~-,
0(100)
08

1>(1111) c(UJ)

"
Ir". '" A t Figure IS Maglloll enerJ:Y U function of the iiflua.., of th.. wavo:vector, for the ftrr()l11agJlIll "t"t't,.. (After n. Anlonin; and V. J. Minkiewicz.)

'.,-"".. ""-C"... ""-".j; .".,-",j;.",-",~,,, .

the experimental determination ofmaglloll spectra.Hlhe incident neutron has \\'3V(."\I(.'ctor k" and is scattcn:d to k~ with the creation of a magnoll of W3V(.>vCCtOr k, then lIy conservation of crystal momentum k.. :: k~ + k + G, where G is

a reciprocal lattice vector. By consCI"vation of energy

(31)

where fie"" is the energy of the magnon created in the process. The obscn.'ed magnon spectrum for MnPt3 is shown in Fig. 15.
fEIUUMAGNEllC OnDER

In many ferromagnetic crystals the saturation magnelizlltion at T ~ 0 K Jocl; not correspond to parallel alignlllent of the magnetic moment!; of the constituent paramagnetic iom., even in crystals for which the indivklual paramagnetic kms have their normal magnetic momcnts. 111e most familiar example is magnl.'tite. Fe:J04 Or FeO FC20a. From Table 14.2 we sec that ferric (Fe3+) ions are in a state with spin S = t and zero orbital momcnt. Thus each ion should contribute 51ot8 to thc saturation momcnt. 11le ferrous (Fe!") ions ha,'c a spin of2 and should contributc 41ot8. apart from any residual orbital moment contribution. Thus the effectivc numher d Bohr magnctolls per Fe304 fonllula unit should be about 2 x 5 + 4 = 14 if all spins were parallel. The observed value (fable 2) is 4. L The discrepancy is 3CC'Ounted for if the moments of the Fe3-+- ions are 3ntiparallcl to each other; thcn the observed

--.---

SF""

T~ 1111111 ~S.f ~

IIIIIIII II jJll ~s-,

SF""

,.

.-OSUr'e 16 Spin ~mo:nb in magndite. Fa> Fl't03, shvwUlIhow thl' momrntJ oflhe Fr ionl cant:'l'1 out. I....,ing only the II1OITII'I1b fA the Fe" ions.

moment arises onl)' from the Fe2 + ion. as in Fig. 16. Neutron diffradion results agree wilh this model. A systematic discussion fA the consequences of this type of spin oruer was given by L. Neel "ith reference to an impot"tant class of magnetic oxides knO\VIl as fcrrites. TIle usual chemical fonnula ofa ferrite is MO, F~03. where M is a dh'3lcnt cation. oflen Zn. Cd. Fe. Ni. Cu. Co. or Mg. The term ferrimagnelic was coined originall)' to describe the ferrite-type ferromagnetic spin order such as Fig. 16. and by extension the tenn covers almost 311)' compound in which some ions have a moment antiparallel to other ions. Many ferrimagnets are poor conductors of electricity. a quality exploiled in applications such as rf tl'ansformer cores. TIu: cubic fcrrites have the spinel crystal structure shO\yn in Fig. 17. There are eight occupied tetrahedral (or A) sites and 16 occupied octalledral (or B) sites in a unit cube. TIle IaUice constant is aoout 8 A. A remarkable feature of the spinels is that all exchange integrals JM' h.l), and JBn are negative and favor antipamflel alignment of the spins connecle<l by the interaction. But the AB interaction is the strongest, so that the A spins are parallel to each other and the B spins arc parallel to each other. just in Ortler that the A spins may be antiparallcl to the B spins. If J in U =< -2J5. 5) is positive, we say that the exchange integral is ferromagnetic: ifJ is negative. the exchange inte gml is antiferromagnetic. \Ve now prove that three antiferromagnetic interactions can result in fcrrimagnetism. 111e mean exchange fields acting on the A and B spin lattices may be written

(32)
taking all mean Aeld constants A. p.... to be positive. TIle minus sign then corresponds to an antiparallel interaction. 111e interaction energy density is
U
=<

-1(B.,\I\I A + BBoM B) = lAM}

+ ~MAoMIl + lvMi

(33)

'00

"".

i'lpre 17 Crystal dnJd~ orl~ minnal spinel MgAlp.; the- MK" io"s <X'C\Jp)' let~ liites. eIO('h surrounded by lOu~aq'gell ions; !he AP' <X'C\JI~ oxtahcdnl.lles. ek'h liU~ br six ~., iuI>5. ThU is a normal,pMI arrang<-'mftII: the di,'lllmi mew ions OC'CUPf tho: ld.BhMnl lite. In the uwrne spinel arr.ongcmml the tc~ sites ace <X'C\Jpied by trh'lllent ~ ions, whik lhe octal...dr.al. .il"" ~ o<.'<.'UpiOO half'" divalent Ind half by tnV1llent melal ions.

this is lower when M" is antiparallcl to Mu than when M" is parallel to Mll . 1be t.'ncrgy when antiparallel should be compared wilh zel'Q, because a possible solution is Mil. = hl B = O. 11ms when
(34)
the ground stale will have Al" directed opposilcly 10 M n _ (Under certain conditiOIlS there may be nOllcollincar spin arrays or still lower energy.) Cllrie 'lempemture (nul SUSCC11tihility of Joerril/lllgneis 'Vc define s<'j>aTate Curie constants C... and ClI for the ions 011 the A and B sites. For simplicity, let all interactionS be zero except for an antiparallcl interaction bctwt."ell the A and B sites: BA "" -~M'J~ Btl = -~I\fA, where IL is positive. TIle same constant IL is involved in both expressions because of the form of (33). \\le have in the mean field approximaUOIl (CGS)
MAT = CA(B" - ILM B )
;

(35)

where E" is the applied field. ll1ese l.'qualions have a nonzero solution fOr MA and M B in zero applied field if

T
ILC B

"CAl ~0 T '

(36)

so that the fcrrimagnctic Curie tl.'ITlpcratuT"C is givcn b) T c = I{CA C B)J/'2.

46'

,.,

.... v
,/"

, ,
, ,
,
T"

/'

1/

:/
10

""
+

Tcmpen"\ltC: rcJ

""

'"

We solve (35) for MAo and MD

obtain the susceptibility at 'f> Tc :

Gs)T - 2j.l.CA CB

(ees)

(GAo

1"'

(31)

a result more complicat.:.'tl than (4). Experimental values for F6j04 are ploued in Fig. 18. TIle curvature oCthe plot of lIX versus T is a characteristic feature of a fcrrimagnet. We consider below the anlifclTomagnctic limit GAo = CoIron Carnet!. The iron garnets arc cubic fenimagul,."llc insulators with the general Cannula ~hFesOI2. woc-re M is a Irh'3lent metal ion and the Fe is the trivalent ferric ion (5 = t, L = 0). An example is)'tlrium iron garnet Y3Fe~;012. known as nc. Here y3" is diamagnetic. The net magncti7.ation of YIC is duc to the resultant of t\"o oppositcl)' magnetized latticcs of Fc3 + ions. At absolute zero each f...mc ion contributes j:5j.l.B to the magncti7-ation, but in each fonnula unit the three Fe3... ions 011 sih.>S denoh:d as d sites are magnetized in one Sl,nsc and the two Fe3+ ions on a sites are magnetized in the opposite sense, givinga resultant of5JlB per formula unit in good agn..'ement with the measurements of Geller et al. The mean field at all a site due to the ions on the d sites is B. = -(1.5 X Ilt)Mof. The observed Curie temperature 559 K of YIC is due to the it-il interaction. The only magnetic ions in YlC are the ferric ions. Because these are in an L "" 0 state with a spherical charge distribution their interaction with lattice dcfom\ations and phoooos is weak. As a result YIC is eharoctcri7.ed by very narrow Iincwidths in fcrromagnetic resonance c);pcrimenls. In the rare-carth iron garnets the ions M3 + are paramagnelic trivalent rare-earth 101lS. Magneti7.ation curves are given in Fig. 19. TIle rare-earth ions occupy sites labeled c; the magneti7.alion Me of the ions 00 the c lattice is opposite to the net magudi7.ation of the ferric ions on the a + d sites. At low temperatures the combined moments of the three rare-earth ions in a formula

"11-----------------,-----1

1 t loH-l-----------'" .. ---1
: ,

.I

1 ,

".

..
"",t

T, ..

I;

Fi9'"' 19 E>.perimental ....ues of lhol Siltuntion "'_~"eDUS ttmpenlurt of various iron gMTIelJ, after R. Paulhmd. 'Thebrmub unit is M,F~OI..... here M is a tm~n>lI~aJ ion. ~ lempmalun:aI ...tJiclt the InIIIP'dQation ~ zero Is 0II1r.d the rompemation lemper:ilu~ be", the ~iutionci the ~I subbnlce is rqual and oppooile 10 the net INf'J'ctwbon of the fffrit. ion suLoLotticel. fa" fannula thert are 3 F.,3> ioN on tet...tw:dnol sMkos d; 2 Fe3 ' ions on ~ ~tes Ill; alld 3 MH ions On sit" &"011 by c. feme ions oontribule (3 - 2}5Jo'll .. 5fl" per ilnnula unit. The ferric ion coupling il stronga"d delft'minellhe Curie tempcnture. If the MH >0111 are I':Ite-elIrth ions the)' are In~netiZ<...l. ~tc 10 the 1'C3ultani of the Fc-l' ions. The M3'

n.e

comlihutioll.!tOpS r..p;dly wilh increllSing temperature becau$Ol the M-Fe coupling is weak. Me.uurerncnls On sillgle Cf)"tal spc.'CimCl\li lIJ'l' reported by Celler ct III., I'hys. nev. 131, 1034 (1965).

unit ma)' dominate the nel moment of the Fli' ~ ions, Lut because of the wt.-'3k c-a and c41 coupling the rare-earth lattice loses its magncli7.alion rapidly with incrcasing temperature. l1le total moment can pass through zcro and then increase again as the Fe3 + moment starts to be dominant.
Af\Tl.FERROMAGl'IETJC ORDER

A classical exaIllple of magnetic stnleture determination b)' neutrons is shown in Fig. 20 for MnO. whidl has the Naet structure. At 80 K there are extra neutron renections not present at 293 K. The renections at 80 K may be classilled ill terms of a cubic unit cell oflattice constant 8.85 A. At 293 K the renections correspond to an fcc unit cell of lattice constant 4.43 A.

''''
I(Ill)

(31 11

C3f1'

(Sill

-.-88:1A

" '" ,
,
j

"'.
'''I''
MoO
(II~

(Ill)

hA
'f"

<3Jr,L
I
lIt- 4.431.

"

.,,,

..'"
" ,

--~

'"

'"

Figu'e 20 Netotron diff~ patt~ b ~lnO below and ~ the ~g temperature of 120 K. after C. C. Shull, W. A. Strauser. and E. O. \\blbn. The rdlec:tion iodica ~ l:.ed OIl an 8.!35 Accll31 SO K and on. 4.43). ""II at 293 K. At the higher tempmltu", the Mn u ic)fll arc still Il1l11f1:netic. but tJoey an': nO longer ordered.

-'"

...

\.

"'

But the lattice constant determined by x-ray renection is 4.43 A at both temperatures, 80 K and 293 K. We conclude that the chemical unit cell bas the 4.43 Alattice parameter, bUllhat at 80 K the electronic magneUc moments of the r ...ln2+ ions are ordered in some nonferromagnetic alT3ngement. If the QT. dering were ferromagnetic. the chemical and magneltc cells would give the same reflections. TIle spin an-angement shown in Fig. 21 is consistent with the neutron diffraction results and with magnetic llleasurements. The spins in a single 1111) plane are parallel, but adjacent (I11J planes ~e antiparallel. Thus MnO is an antifenomagnet, as in Fig. 22.. In an antiferromagnct the spins are ordered in an antiparallel arrangement with zero Ilet moment at temperatures below the ordering or Nl,'iel temperature (Tahle 3). nle susceptil,ility of an antiferromagnet is not infinite at T = T"" but has a weak cusp, as in Fig. 23. An antiferromagnet is a special case ofa ferrimagnet for which both sublattices A and B have equal saturation magnetil.ations. Thus C.... = Cn in (37), and

'"

Figure 21 Orden.>d arr.l,,gcments of'l'iM of the Mill' In malll:alle>'e oxide. MilO, as dclermined by neutron diffraction. 'l'he 0"- ioru; are not d-oown.

io".

,rJgUre 2! Spin orden"g in rCrrom:Igneu

U > 0) and antiferromag1H'lS U <

0)

1'll~'_1Um

Fc",orll.ill:"",I>m

SU><'-l>1,Lm,y)(

"nll':,rlII1IllJI''".''''

L
0

......~I

~\
I
T,
X'
~.Wc

.,, ,
,
0

I
I

,,I I
I

,.C T

."...

...

C T - T. ... _

--

-.

T,
C

X-T+i
(T>

(T:>

TJ

T"J

Figun: 13 Temprtlllure~~ oflhe: m;;,groetie lU>iC:'ephbihly in ~1eU. remlC~.

"...d antir"nOrn:l$tneb. Below the Nlellernpcrature of a,n illllifernllnagnN the- opms tu;"..e ..,(ipara!Icl orientation., the susceptibilit) alt";", lu ma~irnuln value at T",. ...here t1,CIl.' i. a ..ell-de6ned kink in the cune of X ,..nul T '111e lramitio" is also m~rked b, r-I<:l in the heat capacity lind the

thennal exp:lnSiOll eoclficicnt.

IS

Fl!'fl'OtIWgndiAm

lind

-""t9,errorowrgrtd.....

.65

Table 3 Aotiferroma&odlc er)'$lab

Traailioll lrmper.llure, T,.~ in K

~odie

Subotanee MnO

ion Ialtice
r~

Curie-Weiss 8, in K

T._

AfT,,)

...
I

116

MnS
MnTe

r~

100
3lfI fi1

610 528
fiX)

MoFr FcFr FcCI 2 FeD

helL b)'er be lelr

62
117

5.3 3.3 2.25 1.24

0.62
0.76 0.72 <0.2 O.S

be letr

Coel,
CoO
I'\i~

hex. la}'Cr l<c 00<. larcr l<c hex. b}'et

r~

NiO

C,

b<c

7. 24 198 28 291 50 528 398

48 570 38.1 330 682

-2OOlI

,.
2.0

1.48

1.53

I.J4

-,1.37

the Neel temperature in the Ilu:an field approximation is given b)'

'fN>=p..C,

(38)

where C refers to a single sublauice. The susceptibilil}' in the par.unagnetic region T> T". is obtained from (37);
X

2Cf - 2jJ.C 2

l' - (p.cf

= -;;7:-"+"'--~c;:- = -;;r:-'+"':o:r.-,
2C

2C

2C

(39)

The experimental results at T > T", are of the form

(ees)

x=

T+ 0

(40)

Experimental values of orrN listed in Table 3 often differ substantially from the ,,-alue unity expected from (39). Values of OfT". of the observed. magnitude rna)' be obtaint.d when next-nearest-neighbor interactions are provided for, and when possillie sublaHice arrangements are considered. If a mean field constant -E is introduced to describe interactions within a sublattice, then OfTN = (p.. + E)/(;A. - E).

Smcq>tibility BelOit! the Ntel Temperature

There are two situations: with the applied magnetic field perpendicular to the aJtis of the spins; and with the field parallel to the axis of the spins. At and above the N~I temperature the susCt.'ptibilit)' is nearly independent of the dift.'Ction of the field relative to the spin axi~.

...
For 8 .. pcrpclldiclll.U' 10 the ."'is of the ~pins we can Cdk'ulate the SIlI><:cplibility by elementary (.'On.\ic.leralious. The elll!~' density in the presence (If tl.e Geld is. with At = IM"I = IMJ/I. ~M,,:M8 - ll..-(M" + Mil) !a -1A.M 2(1 - j(Zrp)'} - 2B"Mfj!. (41)

u .:

wlu."TC Zip is the angle the spins lllolke with each other (Fig. 24a). The Ctlcrgy is
a minimulIl when

dUltllp = 0 = 41olMzf/! - 28",\1

(42)

so that

ICCS)

(43)

hl the parallel orientatioll (Fig. 24b) the m,lgnctic energy is 1I0t changed if the spin systems A and B make equal angles with the fickl. 111us the !>IISL'Cptibili~ OIl T = 0 K is zero: XI(O)=O.
(44)

TIle parallel susceptibility increases smoothly with temperature up 10 T,.;. Measurements on ~lnFI arc shown in Fi~ 25. In very strol\~ fields the lopill systems "ill tum disconlilHlOlIsly from the parnllcl orientation to the pcrpcndietl!ar OriL'11talioll where the cncTJOl is lower.

Antifcrnnnognetic Magnons
\\'e obtain the dispcrsion relalion or magnons in a oll<.>..(lilllcnsional .1I1tifc.-.-omagllet by making the appropri.ltc substitutious in the I.-catmcnl (16)-(22) of the fCHoma~netie Huc. Let spins with evcn indit:cs 2p colllpo~e sublatlicc A. that with SpillS lip (S~ "" S); and let spinS with odd indices 21J + I compose sublattke B, that with spins down (S~ "" -S). \\'e consider only nca.-cst-neighbor intcr.1etions. with J negativc. '111cn (18) written fo.- A becomcs, wilh (I ca.-crullook at (17),

dSOId' = 12JSlhX-2SI" - SI,,_. lne coHesponding CtJuations fo.- a spin on B are

dS~.,Jdt = (2jSlflX2SJ.,,+1

st..,) ;

(45)) (451))

dSVdt = -(2jSlhX-2S~ - S2,--1 - S~'+I)

dSL,u1dt =
\\'e rorm

+ Sf,. + St,+:t> -(2jSlhX2S2,.+, + S~ + S2,.+~

(4&)
(4&)

s'"

5~

+ is'':

thcn

dSvdt = (2ijS1flX2S;P

S~_I

d5t,,+Ildt = -(2ijSlflX2S~.1

+ 5;;'+1) + 5~ + 5~.~

(47) (48)

24 Calculati<>n of (a) perpcooicuand (b) pa....llci susceptibilitics at 0 K in the mean field app'OJeimatioll.
Ia~

F;g,,~e

'k

..
150

100

"
.. ",,-;,,,,",-;,,,,;},,.L,;j,,,

o\"-'-c""~","~;j".-'.",,,,,,-;
T. in K

Figu~e 25 Magnet;':' 5u~eplibility c:l manRBnese fluoride. MnF parallel and perpcndicular t<J the tetragonal axis. (....ftcr S. Foner'

Wc look for soilltions of the filrm

S!p = u exp(i2pkt,- iwt);

so that (47) and (48) become, with

S!p+1 =vcxp(i(2p
W ..~ . .

+ l)kCl - iwt] , (49)

(SOa) (SOb)

-4JSfIl = 4lJlslh,

+ r;c- iAa + t;e'kQ) -wv = twe.(2v + ue-iJ<s + lIe'ka)

wu = twc.(2u

Equations (SO) have a solution if

Iw" ~w We. cos ka

thus

We. cos kClI

We.+W

O
'

(51)
(52)

lne dispersion relation lor magnons in an nntifclTomagnct is quite different from (22) for magnons in a ferromngnet. For kCl <ll 1 we see that (52) is linear 7 in k; W '3i we.lkClI. The magnon spectrum of RbMnF3 is shown in Fig. 26, as determined by inelastic neutron scattering cxperiments. There is a large region in which the magnon frequency is linear in the wavcvcctor.
7A physical discuSSllll' oflhe d,fference between the dispersion relations '='r felTOTnagnetj(, and llIll,ferromagnctic magnons is gwen by F. Keffer, II. Kaplan. and Y. Yarer, Am. J- Phys. 21. 250 (1953).

""
w

..
00

, ,

,~ .,
Figure 26 M:lfPlOIl di'lperPon .dation;n the Jimple
00

i"

w .,.ttIO.-l. ,n A-I

"

04

cub;c antil'erromagnel Ilb~lnF3 as dct"nninnl -4.2 K by indiblic '\(.'\Ilron llCI.Ucring. (Aft C. C.

Windsor ......111. W. II. Stt".'CflSOO.)

\Vell-resolved ffiagl10ns have been observed in MnF 2 1\t specimen tcmpcrattires lip to 0.93 of the Ned tcmpcmlmc. Thlls even at high temperatures the magnoll appro"-imation is IlscfuL Further details concerning antifcrromagnetic magnons arc gi\'cn in Q'fS. Chapter 4.
'ERnOMACNJ:.'TIC
J)O~IAINS

At temperatures well below the Curie point the clectronjc magnetic f1lO'menls of a ferromagnet arc essentially parallel when regarded 011 a microscopic scale. Yet, lookiug at a specimen as a whole, the magnetic moment rn.'"I)" be very milch less than the saturation mOffiCut. alld the appHcation of an external magnetic field ma)' be required to sahlratc the specimen. lhc behavior obscr\'ed in pol)'crystallille specimcns is similar to thai in single crystals. Actual spccimcns nrc composed of slllall regions called domnins, withill directions of mngncti1A1each of which the local mngndi1A1tion is saturated. lioll of diOcrcnt domnins Ilccd nol be pamllcl. An arrangement of d()m~\ills with approximately zero resultant magnetic moment is shown in Fig. 27. Domains form also in antifcrromagllctics. ferroelc(;"!rics. antifcrroclcctrics, fCIToclastics, superconductors, and sometimes in mctnls under conditions of a stroug de Haas-,'an Alphen erred. The increase in the gross magnetic moment ~ a ferromagnetic specimen in an applied magnetic field takes place by two independcnt pl"OC'CSSCS:

one

In weak applied fields the volume of domains (Fig. 28) favorabl) orit.uted with n:specl to the field incrcases at the expense of ullfavorably oricuted
domains~

III strong applied fields the domain ll1agncti1.ati011 rotates to\vurd thc dircctioll of the field.

Fip;re!7 Fenvmagndic domain ..tton 01'" slngk ~ pbtclel eX nlcl...I. TIle domain boundaric5 are made \'isiblo: b) tilt Biu.,.. m~ )lO\'ode.. ... pattcm '.,mniq.... 1he direction cl magneti.

zation "ilhin a .JOIn.in i~ dclenninf'd b)' oLs.......ing gro....'h or contraction of the oomai" magnet;c Ileld. (ARcr It W. Dc B\Qis.)

In it

.~--;:;;;;::;;;;;;;;------:::::=----1 I MlIt:I"'tu.a.....
rotoH"',

t"igul'e 2S ~p""Kflt.hH magncUut'oo eu....'e, sl"",ing the dcx"''''''''1 magllct'Dltion prtlI.esws ill the diffcre:t't re[tion~ of the CUI"\"".

.t.opo._ ond....

hon

~l

8_"+-4".\1

'.

__---/~--8.

+-1/

\1.opIrto< SolI "

Figun: 29 The tn~onica1 ,llaJ:....(izal ...." l'Un'C TI,e COI.''''';.";I)' II. is (I"" tC\!~ r..,kl tbal recl,K'n B to zero rcbh.><I corn+. it) fI", ... '<l"ces M or II - " to l:IelO. ~ remanence II r is the ,,,Iuc or B dt II .. 0 '11.e 5at,,~.Mion indud.ion B. oJ; the !lnut of B - II., large II, and tI"" !I;Ilur.dion fnal.,'Tleha lioll M... IJ)4,". I" 51 Ihe \'cr1i<..J ""~ is B - ~1

+ Ml

Technical tenus dcfint..'t1 by the

h~stcrcsis

loop aTC shown in

Fi~.

29. 'l1lC

L'OCrcivity is usually defined as Ihe revcrse field He Ih.1! reduces the induction B to 7..ero, starting from saturatioll. III hi~h cocrcivity materials the cocrcivity lI et is defined as the revcrse field that reduces the magucli:t.atioll ."" to zero. '[1It: domain stnlcfurc of ferromagnetic matcrinls ellters their applications. In a transformer corc we walll low losses in taking the core around a cycle: Ihis means that we wallt low cocrcivity, \\hich goes with a high permedhility. By making pure, homogeneous, ,md well-oriented material we ft\eilitate domain boundary displacement ,1Ild thereby attain high permeability. Relative penneability values as high ,IS 4 x Hi; have been reported. In a permanent magnet we \'r7Int a high coercivily. which we may achieve by suppressing boundary displacement, which is best done by eliminating domain boundaries. This is accolllpli.,hed in very fine particles or crystallites that consist of.\ single domain, without boundaries. By controlled precipitation of a sC<.'Ond metallurgical phase the specimen may be helerogeneOtls Oil a very fine 5<:3olc. Characteristically there is a critical r.ldiliS below which single domain attributes arc Ii::Jlllld.
Ani3Olropy Energy

There is an cncTJO' in a ferromagnetIC crystal which directs the maWictiution .1Iong ecrtain cryst,lllographic a.\(CS called directions of "'"by maglleti.....allon.

J5

Fe~~"'''(Isnel''''' ",,,I

'\"'if..~ro"'tisr ...fi.",

,."

~ :f.71!
,.

[1001

.. .. r"l ..
."
00

11111 (/\, .............

(HOl

~l'l""~

"'"
I

"

"'"

"" "'" " '00

<00,,,
o

/~Plane

B.(gwl

"'" '"'

\!OJ')

4000 6UOO

Fib'llre 30 MagnetlC41tion ('I'''''e~ for single e<)slals "f,ron, n,ekel, ~nd cobalL From the <II,...C~ fot iru" \\" sec th~t the (100] dlr"-~tiolls a,e "-'<l>y dirt"t;tion. nf magnctiz.l.tion and the [Ill] dir<-'Ctions art' I',lnl i1irl.'ctk>ns The applied lick! ,s B. (AHer Hond., "",1 Kay~.)

,.

(h)

Figure 31 Asrmmchr of the overlap of electron .lislnbutiollS on neighboring ions jltO\i<l"s one '''echaniSlll of rnagllCltx'IJ'stalline anisotropy. l.lc<.'au.\.C of spin-orbit Interaction the marge di,tr,I",lion IS spheroidal .....1 not spherical. '111e asymmetry IS tk'<! to the direction of Ihe spin. SO thai .. rot.ltion of the spin <11..><1'ons relative to Ihe cry.tal a.'cs changes the e><ehango: enelll:Y '1ml also <'hange. Ihe electl'O\l,,-lic inteludion enerKY of the dlarge distributions on polin of atn",s. Lloth dr.'ds gi\.., me to an anisotropy energy. The energy of (al is not the same lIS the enCI'gy of (\..

This energy is called the ffiagnetocrystalline nr anisotropy energy. It does not come about from Ihe pure isotropic exchange interaction considcred thus far. Cobalt is a hcxagollal ClystllJ. The hexagonlll axis is the direction of easy ma~nctization lit room temperature, as shown in Fig. 30. One origin of the anisotropy energy is illuslrated by Fig. 31. 1110 magncti.1tion of the crystal sees the crystallllttice through orbitalll\"criap of the electrons: the spin interacts with the orbital motion by lIIeans of Ille spin-orbit cullpling. In coball the anisotropy energy dcnsity is givcn by
(53)

where 8 is thc angle the mllgneti:l.ntion makes with the hexagon.ll axis. At room tempcrature K; = 4.1 x H? erglcm:\ KiI. = 1.0 x Hf erglcm3. Iron is a cllbic crystal, and the cube edgcs are Ihe directions of easy magneti7.&ltion. To represcnl the anisotropy energy of iron magnctiJ:ed in all arbitmry direction wit' Iin..-'Ction cDsincs ah az, a3 rcferred,to thc cube edges, we

m
arc guided by Ctlbic symmetry. 1he expression for the anisotropy energy must

be an

C\'CIl

power of each 0" prO\;<lOO opposite ends ofa crystal axis arc equh-a_

lent magnetically, and it mll1>1 be ill\'arianlundcr interchanges of the at amo-l1K thcmsehcs. The lowest order combination salisf)'in/l1; the symmetr)' rc<juire_ ments is + ~ + a5. but this is identicall} CQualto unity and docs not de scribe anisotropy effects. The ne);1 combinatioll is of the IOl.lrth degree: afai + a't~ + a2~, and then of the 5i"th dewce: atof~ Thus

at

UK =

K.<afai +

~..o!

+ a'fli) +

~~~

(54)

At room temperature K. = 4.2 x IW cryJcm3 and K 2 = 1.5 x lW cryJcm 3 .

Transition Region Between Dornains A Bloch wall in a crystal is the transition layer Ihal separates adjacent regions (domains) magncti7.cd in different dircetiO'lIs. lnc entire change in spin direction between domains does nol Ott",. in one disrontinuous jump across a single atomic plane, bul takes place in a waclual wa) o\'cr many atomic planes (Fig. 32). l1le clIchange energy is lov.ocr whcn thc chtll1gc is distributed o\"cr many spins. This IJcbavior may be Iindcrstood by intcillretin~ the Heiscnberg equation (6) c1assietllly. We replace cos (/) by I - i~; thel11O.... = )5 2 (/)2 is the CJl;change energy between two spins making a small angle (/)with each othcr. Here) is thc exehangc intcgral and S is thc spin quantum uumber; 10... is rcfcrred to the cnergy for parallel spins. If a total change of noccurs in N equal steps, thc anglc betwccn neighboring spins is nlN, and the exchan~e cnergy per I)air of ncighboring atoms is 10... = )S'l(nIN'f. Thc total cxchan~c cl1cf!,'y of a Iil1c of N'I I (ltoms is

Nw". = )S'ln2/N .

(55)

Thc wnll would thickcn without limit wcrc it not for thc nnisotropy cnergy, which acts to limit the width ofthc transitioulayer. 'J1Jc spins contained within thc wnll arc largely dircctcd away I"rom the axes of casy magnetization, so thcrc is an anisotropy energy associatcd with till,! wall, roughly proportiol1al to thc wall thickncss. Consider a ,vall parallel to the cube fare of a simple C\lbic lattice and scp..1rating domains magnctized in opposite dircctiol'ls. \\'e wish to dctcrmillc thc nnmber N of atomic planes contained withul thc wall. 111C cncrgy pcr uui! area of wall is thc sum ofcolltlibutiollS from exchange and anisotropy encrgics: u'" = u .... + U.nis' The CJl;change cnergy is gi\'cn approximatel} b) (55) for each line ofatoms normal to the plane of the w.l1l. lhcreare Ua'l such lines pcr unit area, whcreo is the lattice constant. Thus u.... = Tf'l)5 2/Ntr.

..
,

,
,
I I I
I

------ ----I
I I I
..~."

IN".

~
I' "

IN
'

U '.11 ~
I I
,
I

""I t" ') tInn fo -~~y.;~J1/// Ifll!! '1

I"

,
I ,

figurc 32

n.c structure cllhc Bloch wall s.:va.r~hllg domain~

In ;roll lbe thicL'I.'S1 cl the trami-

tion rt:giou is aboul 300 latHc..., c.vnstanls.

-111c anisotropy cllcrgy is of the order of the anisotropy COllstallt times thc thickness Na, or U.no ... .Na; therefore
U"" "" (7r21S2JNa~ + KNa

(56)

'nlis is a minimum with respect to N when

au,jaN = 0 = -(T.21S'lJf',;'la~
0'

+ Ka
(58)

For order of magnitude, N ... 300 ill iTllll. Thc total wall cncrgy per IInit area on our model is
rT~ = 2'1T(KJS'lJa)V2 ;

(59)

ill iron
givcs

17... ...

rT..,

1 crwcm2, Accurate calculatioll for a lBa' wall in a (100) plane = 2{2K11S'lla)ll'l..

Solitons
Bloch walls are an example of a t1scinaling class ofphysical and mathcmatical problems known tIS solitolls. A rough descriptioll of a classical soliton is a solitary wavc that shows grellt stabilit}' in collision with other solitary wavcs. Bloch walls in thc absence of applied magnetic fields arc static solital)' wavcs, but thc)' mm'e 1miformly ill an applied field and their collisions can be studied,

at least thcoretirnlly. It is somewh.lt Sllrpl ising th.'1t this can be done because the equ.1tions motiou arc llotllinCou. 6

or

Origin of Domoins Landau and U!Shitz showed that OOul.I.ill stmdme is a natulOll consequcnce the various contributions to the ellergy-eKchangc, anisotropy. and magnetic--of 3 ferromagnctic bod). Direct evidence of domain strucltlrc is fUnlishcd by photomicTORraphs of domain boundaries obtaincd by the h..- chniquc of magnetic powder patterns and by opt;caJ studies uSlllg F.mula) rotation. 111e powder pattern method de\c1oped by F. Bitter amsish in placmg 3 drop of a colklidal sllspension of finely divided ferromagnetic material, sUl'h as maJ;lIctile, on the surface of the ferronlilgnetic crystal. The oolloid particles in the suspension concentrate strongly about the boundaries between domains where strong local magnetic fields C'l:ist which ",nTact the magnetic p.\rticlC3i. 111e disam~ry of transparent ferromagnetic oompoullds has encouraged the usc also of optical rotation for domain studies. \\le lIl:l.y understand the origin of domains by oonsidering the structures shown ill Fig. 33. each rel>rescnting a cro~s ~tion tllf()ugh a feTTolTh'1gnetic single crystal. In (a) we have a single oomain; as ll. collscquencc of the magnetie "poles" formed on the surfaces of the crystal this confib'llration will have a high value of the magnetic energy (1/81T) f 2 rIV. lnc magnetic ("nergy density for the configuration shO\'I1I will be of the order of M~ .... 1()6 erglcm3; here .\1, dcnotcs the saturation magnctlzation. anti the units arc eGs. In (b) the magnetic cncrgy is redu(.oed b)' roughly ollc-half by dividhlg the crystal into two domains lll.\glletizcd ill OPIXISitc dit'f'ctiolls. In (e) with N donl.lins the magnetic Cllcrgy is redtlced to .1pproKililately liN of the magnctic cnergy of (a), because of the ff.. -duccd sp.1tial edension of the field. In domain arrangements stich as (d) anti (e) the m.1gnetic cnergy is 7.ero. I(ere the IXlUndarics of the trianRul.1r prism domains \lear the end faces of the cr)'stal llh-J..C equal angk-s (45") \"';th thc magncti7..a tion in the rectanglilar doIIlnins and with the magne!i;r.atiol1 ill the domains of elosure. "l1lC component of magnetjl.a!ion normal to thc boundary is continulliis across the bound.uy and thcrc is no magnctic field .1ssoci.1ted with the maglleti;r.ation. 111e llux circuit is completcd within the crystal-thus giVillg rise to the term domains of closure for slll'facc dOll1olins that complete the llux circuit, as in Fig. 34. Domain stnlcturcs arc often more complicated than our simple cxamples, but domain structure alu:ttys !las its origin ill I!le IJO!JSilnlity of loweri,l!!, t!le energy ofa system by goillgfrom a SlIturatetl cor!figllration wilh lliglllJUlgnetic energy to a domain collfigllration ldt!l a lower energy.

or

"Rca'nt ~phs on soIilon lheon mdu<lc R. K BulJough. Solito.... Spnnl!lf'r. tllfiO, C L L.omb. f1r"lrrIlS of ...JilOft tlof.onJ. w,k)", t9liO. Fur a .11Of1 ;nlru<!urtory artldc. 5 R. K 8<.11 1ouJ:h, 1'1')~a Bullet.... FWnary 1978. pv 7&-62.

.'oJ .. $

r;::::-..

t I

tit

t I

Fill:u~

tI "

33

n.r arigl.n a donIaIs.

'"

..

.,

nJ:urc 34 Doma,n of closure ~t the end of a songle l'f)'Stal iron "t"sker, '11C f.>ce is. (l0D) plane; tl'e ....biJo!,;cr a\is is [0011. (Courtesy of It V. Cole,lWl, C. C. Scott, arid A. bin.)

Cocrdvity UfId llyllterelfm

The cocrdvily is the mngnctic field He rCIJllired to redllce the mngllcti:t.. l'llion or the induction B to ".ero (Fig. 29). ""llIe nfthe cocrdvity ranges mer scven orders of magnitude; it is the most SCllsili\'c property of ferromagnetic materials which is subject to control. The Clx'rcivity may vary from 600 C in a loudspeaker permanent magnet (Alnico V) alld 10,000 C ill a ~pccial high stability magnet (SmCOsJ to 0.5 C in a commercial power transformer (Fe-Si 4 wt. pet.) and 0.002 C in a Imlsc transformer (Supermalloy). Low coercivity is desired in a tmnsf(>rmcr, for this means low hysteresis loss per cycle of operation. Materials with low cocrcivity are called soft; those with hi~h cocrc.:ivity are called hard, although there is not necess.1rily a I: 1 relationship of magnctic hardness with mechanical hardllCS~ (Chaptcr 20). 111e cocrci\"ity decreases as the impurity content decreases and also a~ internal ~trajns arc rcmoved by annealing (slow cooling), Amorphous ferromagnetic alloys may Ila\'c Iowcoelcivity, low hystcresis losses, anc.! hi~h permeability, as treated in Cllapter 17. A1lors th.1t cont..ain a precipitated phase may ha\'c a high cocrci\"ity. as 111 Alnico V (Fig. 35).

'n.e

1;6

Figure 3.'l MicrostructUTe of Alnico V in its optimum .tate as II permanenl magnet. The composi. lion of Alnico V i5, by weight, 8 AI, 14 Ni, 24 Co, 3 Cu, 51 f'e. a permanenl magnet il b; a two-phase S)'Stc"" with line partk'!llS oron.~ phllSe emh.: ddcd in til" other pha<e. 11,,: predp'l.alion i5 CIlrric.1 oul in a magnetic field, lmd the particles are orienlL'(l "ith th"'t long Axis paralld 10 the field l!irL'l'tion. "l1,e w;l!th shown i$ I.l ",m. (Courtesy or f'. E. Luhorsky.)

'U

Soft magnetie materials are used to concentrate and shape magnetie llux, as in motors, generators. transformers. and sensors. CL"l5ses of soft materials include electrical steels (usually alloyed with several percent of silicon to increase electrical resistivity and to decrease anisotropy); various alloys of Fe-eo. Mn, starting with pcrmallo)"s of composition ncar Ni7BFe22, which have near zero anisotropy energy and ncar zero magnetostrktion; f\iZn and ~lnZn ferritcs; and meL-ulie gtasses produced by rapid solidification methods (Ch.'pter 17). A commercial metallie glass (METCLAS 26Q5S.2) with composition Fe-;vB13Si9 has a hysteresis loss per eycle mueh lower than the best grain oriented silicon steel. 'nlC high coercivity of materials composed of very small gmins or fine powders is well understood. A sufficiently small particle. with dironeter less than 10- 5 01' 10-6 em, is always magnelizctl to satUT3lion as a single domain because the formation of a flux-closure configuration is energetically unf:1\'orable. In a single domain particle it is not possible for magneti7~'tion reversal to take place by means clthe process of bound'\/)' displacement. which usually requires rel.,tive)y wcru. fields. Instead the m.1.glletization of the p..vt..icle must rotate as a whole, a process that may require large fields depending on the anisotropy energy of the matcrial and the anisotropy of the shape of the p..lTti-

c1e.
The oocrcivity of fine iron p..'lrt.icles is expected theoretically to be alX)llt 500 gauss on the b.'1Sis of rotation opposed by the crystalline anisotropy energy. ami this is of the order of the obseTVe<1 value. I iigher coerci\'ities havc been reported for elongated iron particles. the rotation here being opposed by the shape anisotropy of the dcmagncti7.Htion energy.

"z

Rare earth me, in alloys ",;th Mn, Fe, Co, and Ni have "ery large cryst.ll anisotropics and corresponding!)' large cocrci"ities, of the order of2KJAl, These a1io>'S are exceptionally good permancot magnets, For example, the he),agonal compound SmCo.:; h.'\S an anisotropy energy 1.1 x loB erg em- 3, equivalent to acocrci\ity 2KJM of290 kG (291). Magnets ofNd2Fel4B hm'e energy products as high <IS 50 MGOe, exceeding all other cammereiaJl)' available magnets. Fricllberg and Pilul 9 de,'e1opcll il theory of the pinning of domain walls in single phase polycryst.tIline materials thal gives a good account of experimentnl results over a wide range of materials, In particular their theory gives a good account of the high cocrdvity ofSmCo5 , predicting He = 9 kG where 10 kG is observed.
SIl\'CLE DOMAIl\' PAJrI1CLES

'nlC dominant industrial and commercial applications of ferromagnetism are in magnetic recording devices, where the magnetic material is in the form of single domain particles or regions, The total value of the production of magnetic de... iccs for rccortling rna)' be comparable with the total ,a1ue of semicon~ ductor dC\icc production and greatl)' exceeds the value of superconducting delice production, the L1.t1er being held back by low critical temperatures. as comparccl ,,;th rn."l&Jletie Curie temperatures. The magnetic recording devices or memories typically are in the form of hard disks in computers and tape cassettes in ,ideo and audio recorders. An ideal single dorn.'lin particle is a fine particle, usually elongated, that has its rn.1gnetie moment directed towan! one end or the other of the particle, 111e altcnmtive orientations may be labclcll as N or S; + or -; in digital recording, as 0 or L To have digital properties a ferromagnetic particle should be fine enough, typically 10-100 nm, so thai olLly one domain is within the particle. 10 lflhe fine I),.'utide is elongated (acicular) or has uniaxial cryslal symmetry, only Iwo values oflhc magnetic moment of the single domain arc permilled, which is what one wants for digital properties, '111e first successful rccording material was 3(.1C\.llar TF'Cf03 with lcnboth-to-width ratio of about 5: J, ('oercivily ncar 200 Oc and a length <lp.m; chromium dioxide crO z is the basis of a hetter material, II in a form highly acicular (20: I) \\ith cocrcivity near 500 Dc. Effective elongation can be atlaine<1 v..ith spheres by making a chain, like a string of beads. An ensemble of such duuns or of elongated single domain p.'U"ticies is s.'\id to cxbibit superparamagnetism if the magnetic moment of a
'R. FriedllnJ/ and D I Pao..tl. Pb)~ Rc.'. 1..ctL 34, 1234 (1975). IOC. Klttd, lhror)' mlhe stn>eb:.--e m ~ ~ in fo]ms and sm",11 pKrticlc,~ Ph" Rev. 10, S6$ (l9-16l, ~Ph)'.."al t1>eor)' of~1C oo..llUns,~~"S. Mod Ph)'S- 21, 541-~'S3 (lSl9).. "R M. Wllllc', ~,tat<'ri:ib aspect of magnetic ...-ronhng. ~ J Mag. and Magnetic Mal. 58, 165--176 (1990)..

unit iseonstant. If Ii- is the magnctic momcnt ill a Ill..'gnct.c field B, tI.en the IIlt magneti7.. "ltion of the enscmblc \\ill follow thc CUrle-Brillouin-Lmgevin Imvof (I-I.19), (14.20) if the partielcs arc embedded in a IiquKl so that tile)' are each free to rot."ltc as a whole. If thc particles are frozen in a solid, there will be a rcmanent Ill.."lgnetization (Figure 29) aftcr removal of all applied field.

Ceomagnetism and Biomagnelism

Single domain ferrom.tgnetic properties are ofspecial geological intere.t in sedimentary Il'X:ks because the rocks tlu\lugh their remanent magnetization carry a memory of the direction of the earth's magnetic field at the time that they were l.-ud down, and thus of the geographical IooJtion of the rocks at that epoch. The m.. "\WIetic re<.'Onl is perhaps the I1lOlIt important basis of the theory of the drift of continents. Annually, layers of ~ediment are deposited in stream beds, layers that may bear some m<lgJlctic particles in "inwe domain form. lnis record persists over at least SOO million )ears of geological time and rnn tell us where on the surface of thc earth the deposit was laid dO\\TI at a gi\'cn time. Lava flows also record m...gnctic fieM dircchons. 'Inc eh.. , nge in magnctiz..'ltioll from layer to laycr gives a superb historical reeord of the drift of the contincntal plates Oil thc earth's ~"Urface. TIle p.."lIcomagnetic record is one basis ofthc bmnch of geology callcJ plate tectonics. nle original interprctution oCthe rccofll was made more difficult, or more C~citillg, by the associated discovery (Brunhcs. 1906) tllo.lt the magnetic field of the earth itself can show revcrsals in direction, an cffttt contained within the struulan.! d}narno theory of the earth's mllgnctislll. Reversals havc wkcn place once e\ery 1 X IQ'I to 25 X loG years. Whcn a re\ersal OC(.'lrs, it is relatively sudden. Fine single domain p..1rtieles, often of magnctitc Fe304. arc even of impor~ tance in biology. A direction-secking eITcct known as magnetotll.xisll!. oftCIl controls. possibly sometimes along with an aslJOllomk-a1 guide system, the motion of bacteria, the migration of birds, amI the movements of homing pigcons and bces. The effect is due to the intcraction of a singlc dumain particle (or cluster of such particles, Figure 36) ill the organbm with the eXlernal magnctic field of the earth.

Maglletie Force Microscopy

The success of the scanning tunneling microscope (STM) stimulated the de\ clopment of related scanning probe devices, of which the scanning magnctic force microscope is one of the Illost cfTccthc. A sharp tip of a m.."'lgJletic material, such as nickel, is mounted on a cantilevcr lever (Figure 37). Ideally.
11TariI in LtoIogy means lhe move,m.:nl ~ a

C't"11 or orpmsm IlJ'lOanl or :aw:ay from .orne

enema! fidd of 1Qr<,e.

--

t-"ure J6 nlln K.'dion ol '" cdl 01. ,~llIOCtcrium .......ing. ",ham.,(OO nm fIolrl.d.,. or f'C30,. Or.ming by ~hll-ta PuebL!. from. photogr.lpft by Il D f'nn"'l ,Ind uehen

Figure:n DlISic COI1Ct1,t of onagnct;e lQrre microscopy. A n1<lllt\clic lip .Ilb::h.....1 to a l1c~iblc Clntl\cver;. used to clCh't'llhe magnetic Iicl<l prooua,d hy the regionli of M]ll'nlallllll m3J:lwlizalion III the pilule of the 511llll'lc. (Alier Gruetler. Ma"'in, ilIa! Ru~r. 1992.)

but not yet, the tip is a single domain particle. Forces (rom the nld~nctic sample net on the tip and cause a change. such as n dclltioli, in the cantilever st..'1.tus, and an im,lgc is formed by scanning the sample reliltivc to the tip. '111e magnetic force microscope (:VI FM) is the only mab'llctic imnging technique that can pro\'ide high resolution (10-100 urn) with little sur.,cc preparatioll. One ron, for example, observe and image the magnetic flux that exits from the surface at the intersection of a Bloch wnll with thc surfUl.'e (fib'Urc 32). An important applicnlion is to tile study of magnctic rea:lrtling media-Figure 38 shows the magnetic signal from a tcst p.'lUem of 2 f.l.m bils Illagnetu;ed in the plane of a Co-alloy dhk; the parallel componcnt of thc field sensed by thc sensor tip is what the photo shO\.\'s.

,0

Figure 38 Test drip maj,'tltl'zation ill the plane of a Co-allo)' disk in 2 pm hils. MFM close "lxl\.: the plane of the di~k. (After flugar d at)

8$

detected hy

MACNETIC BUBBLE DOMAI!\'S

Magnetic rccorclillg discs and tapes oITer the lowest cost form ofmnss data storage or memory. 'l1wir dmwback is that in order to write or read data they require mechanical motion of the head with respect to the magnetic material. 'l1lC mechanical motion involves an electrical power dlive. slow access times, and periodic maintenance. Electrical methods of entry and read-out using semiconductor memories am avoid these difficulties. However, semiconductor memories arc not truly permanent-they lose their data if all pm'.'cr is removed. Magnetic bubble domains in a ferrite or garnet film can be usoo 10 record data. and a sequence of bubbles can be shilled by electromagnetic signals. Bubble memories are nonvolatile, which means tlmt they retain data without need for electrical po\\er. Small cylindrical magnetic domains (bubhle domains) can I>e stabilized in a uniaxial magnetic material by a bias magnetic field normal to the plane of the specimen and parallel to the easy axis of magnetization. 11lC formation of buthies is illustrated in Fig. 39. Magnetic hubhles arc used for high density memory storage, for example in portable personal computers. Below a critical value of the bias field a circular domain as in (b) is unstahle with respect to a serpentine domain p.'\ttem, as in (a). At still higher fields the bubble is unstable with respect to a saturated configuration. The energctk'S arc treated by Bobeck. 13
13A II. Boh<.'Ck, Bell S,sl_ Te<:h.

J.

46, tOOl (1967)

r'llUl"e 39 Fonnalioo of buhlJk domains in garnet film 6 ~ thd (.) unnwgnetulc<l. (bj 110 C (After A. II BoIv:d; and E. 0cl1.l Tone.)

SUMMARY (In CGS Uillts)

The susceptibility ofa ferromagnet above the Curie temperature has the form X - C/(T - TJ in the mean l1eld approximation. In the mean l1e1d "Pproximation the cffective ffio."\gllctie flcld seen by a magnetic moment in a ferromagnet is B o + AI\1, when A = TC!C and Btl is the applied magnetic field. The elementary excit:\lions in a ferrom:\gnet are nmgnons. Their dispersion relation for ka <:{ 1 h:\S the form trw - lira!. in 7.cro extem:\1 m.'lgr1ctie field. The thermal excitation of m:\gnons leads at low temperatures to a heat cnp.'lC'ity :\nd to n fractional m:\gncti7.ation change both proportional to TJ'f.. in an antiferromagnet two spin lattices are equal, but antiparallcl. In a ferrimagnct two lattices arc antiparallcl, but tile magnetic moment ofone is larger th:\Il the magnetic moment of the other. In an 3ntUerromagnet the susccplibilil)' .wo,e the N&:I tempcrnturc has the form X = 2C/(T + 8). 111e m:tgnOlI dispersion rcl:\tion in an antiferromagnet has the form liw'" lka. 111e thcrmal excitation of magnons Icads at low temperatures to a tcrm in T 3 in the heat capacity, in addition to the phonon term in l-:i.
A Bloch wall separates dommns magneti7.cd in different directions. The thick-

ness of a wall is - U1KtJ,,112 lattice constants. and the energy per unit area is ... (Kjla)ll2. where K is the anisotropy energy density.

Problems
1. Magnan JUpenloli relalion. Den\'[' the magnon dispersion relation (24) lOr a spill S on a simple c:ubic laUice, z: - 6. Hint: Show first that (ISa) is repl<la.'d b)

..

....
"2
where the a'ntrol] alom ,s at II and the ~i" nearest ncigllhors arc connco::ted to it hy l>i~ \cc:tors 8. Look for solutions of tl'e l''lualiolls for tlS;Jt/1 awl dSj,lt/t of tile form ('~p(lk . I' - iwt). 2. fIca! ca')flCily of mag'WI'S. Usc the approximate m1\h'l10Ll dispersion relation w = Akt to lind tile leading term in the heat mp..'l.cily of a three-dimenSional ferromagnet at low Icmperalul"("s kfjT <t f. The result is 0.113 k,j,k/jrlliA"f2, pcr llIut volume. The ZC't;l function tllal enters the restllt may Ix e~tillmll;d llllmeri('all~. it is talJlllatctl in )alu,kc-Emde.

3.

N~el

telll,>erCllure. T,Iking tile efft."Ctive fields on the two-sublaUice 1l1oc.!el of all antiferromagnetic as

show tlml

4. Mugneloe1asti<; cou,Jlillg. In a cubic crystaltloe d,lstic encllO' density in terms of

the

'l.~lIal

strain components c'1 is

mnl

till' kadill~

term

III

the

ma~netic

anisotropy energy density is, from ($4),

Coupling between el:c;tic sh...ill .md magnetizatioo llilcction may be taken fonnally into al..'COImt by including in tIle total energy dellsity a term
V c= 8 1 (aie..

+ .}:#w +

~e.J + Biala2l'.v + a2(l'>.j8v~

+ a:Ple...,)

arising from tl,e stl~in depcndence of V/(; hen' BI and 8 2 arc c,lkd magnctodastic coupling cunst,ml.'i. Show that the total cnl..'rgy is a minimum when
(I I" j)

T11is explains tlie OJ'"igin of mllgnetostriction, the change of Icngth on magndization.

5. Coen:jl'eforce ofa ~malllHlrtidp. (a) Consider a small spherical single.domain pmtj c1e of a "'liaxial ferrom~nel. sllO\\ that tIle reverse field along the axis required to n.:verse thc magneti:l.ation is G~ .. 2KJM.. in CGS units. 11lC coercivc force ofsingledomain particles is observed to be of t!lis magnitllde. Take V K = K sin 2 (J as the ,mimtropy cncrgy dcnsity and V" = - G"U cos (J as the intcrar.:tion cncrgy densit)" witll tl,e extcmal ficld; here (J is thc anglc betv.'Ccn B" llnd.\1 !lint: Expand Ihc cnergies for small angles abollt (J = 11", and find tllc v:llue of 8" for whkh V K + V" does not have a minimum ncar (J = ;r. (b) Show IIlat thc magnetic energy uf a salll-

Fi".n: 40 A 810m ..;ill ;and a I\"td ..,.11 in II th," Glm The magnc:ti2:Hlion '" lhe Bk>l:h .."U IS ,oorn\;,o! to !ht ,une "rtbe IiIm "nd add. 10 thc .... ""lel'\!'. a <kmagndizab"" t'TlCrgy ,\1:&1 per uml knfo(th of .....n, "Ilere Ii;s the ,,':1.11 thki.'le5s amI d Ihe film t11.,J'"'eliS. In the Ntocl " .. II the ,n" .. ",eh7.<lI,ull is Jl'lr.,lld to the ~llrface. tI,e addition to the "all ene,Jo(y is Ilegh,. ..hk "lIen d q; 6. Thc a(I.llllOll 10 the N<,:..l ",ul1 ':nergy ,,11<''' d ~ Ii IS tht suhject or Proh1clll 7. (After S '\11,,)'

d..lh(lck)

sphere of ui.tmclcr d is -M~(f. An "ITOI.I.Igclllt'nt ,,;t1l a"pll':Ci,t('l)'I~ ll'ldgl1etic cnC"rgy lias a single wall in an ('(luatorial plane. 'n,e c.Iom.Un "'all cocrgy "ill be 'fTU..rfll-t. where u,. is the wall ent'rg)' PC" unit .lJ'Ca. Estimate for tcobah the critical roJius below "flich the p.lI'ticl~ arc stolhle as single domaill$, bking tbe ulue of }S 2/a olS for iron,
6. Salura/ion maglle/i~"liOlI Iwa,. T~. Show that in til(' mean Reid olpprO\im..tiun tlie satumtiull ma~llcti:l.ution just bdow the Curie tl'lllp<'rature hao; tile (IOlllillallt t,,1\Ipcruturc dcpcndcnet (T~ - Tt~. Assume the spill is i. "1,c rcsult is the s..1me as tl"lt for o.serontl-onler trJJlsltiun in a ferroek'l'tric IT\ st~, <IS discussed in Chapter 13. TI,e capcrimentII t\:lta for fem>rnllgllcls (fable I) suggest thai the cxpnnellt is closer to

rnlw

0.33.
7. /'I'k1 wall. The uirection of magnebl'.:ltiOIl change ill a domain wall gon from tholt cor tl", Bloch wall to that of a :\"&-1 ,,"ll (Fig. 40) in tlllll films of mated.!1 of nq~i~h1e crystalline aTrisotrnpf encrJO'", such Wi Pemlallo). TIle illll'I"CCpt of tile Bloch w.J11 with the surface of lire film eroot..... a ~llrfuce rq~on of high demllWletiLation enclK)'. TI,c Neel ,,,,nil avoiJ.~ tllis intercept contribution, but ..t tlJe ('xpcnsc of a dc'nagndi7~1"oll ~'OIltrib'llion throughout tIle volOlnt' of the wall. The Nc':d ""II belllmcs l'llergt't.lt ,Jlly fa'or.ilile when the becomes sulfJCiclltly thin COllsiclcr, however, the cllergt'tics of tllC Nkl ...all in b"lk material of n~g1igible c~stalline anisotropy t'IlCl")O'. TI,ere is now II demagnetia,lion contribution to the wa.ll ('nergy denSity. By a <1UJ.litative .argument simiLtr to (56), show that u.., _ (:?}sl Sal) + l2;r,1l;Na). Find.\ lOr 'I hidl U" IS a minimum. Estimate the onlcr of magnitude of u,. for 1)'PirnI ,allics of}, M., lind a.
l

mill

References
E. Della Torre and A H. Bob,.'<:k. Magn,tlc bulJhlu. NorthlIolla"", 1914 A II E~"'-llrclder, Magnrl!l; bubble fi'CI"'clogg. Spring...r, 19tH A lIeapm, ThtorW /1.. IImt:"nfm..., mss,'s ulll ...~'I3Irl."s. 1968. F Kdrer. "Spin W;l.,ft, ~ ":ncvdo. of'l/,tJ'i4 1812 (1966).

C. Killel..,d j K Galt. donJ:lins. Sohd slate -&37 (1956). G, T. Rado and 11 SuM. eds, AlapctLfJR. An<k-mlC PreM. ~~ encyclopedIC ""Of\.; K J Standk)'. O~i<k- mag-roc material~. 2m! ed. 0",,"1. 197:l. 5 V \'tIflJO'-d.ii. Ma",diS"Rl. lI:1istcd. 1975. 2. ,d R M Wh.te. Quanfllltl Ihrory of ""'&"du.... 2nd eel Sl'r.. ~ ....... 1983. II. M \\lllte.oo T. B. GdJodk. lAJfII, ra"r:rt onk.- ill 101m. /oc:I<IemOc. 1979 I' r Sinh:o and K Kumar. Inlellld""" m ~tiaJlllJ Of'Ilc>ud lOIidr. <hfonl 19fD U C Matt;". "l"hftwy of mag"dinn, S~, 1965. r G Drw.n. Solita.... OImbndge. 1963 E. r. Wohlfarth, ed., 1It1,wlxJol. 011 mlIgncllr .>oIe"'&:, North-lIo1l&nd, SC\-r.d ,'Ol.Ilnes. joum:U d" M:Jgnct~1 :llld Magnetic M:>tenals l" bllll.:-.. .oo ~ Mn.nl, cds.. PI,jpia and ~"It, /I1,pllcdtions of _pwt...... Spring:r, 1991. j jensen 3nd A. R. M -kmtosh, BaR" ~rll, n"'Wld~, Word. 1991 J Cnngk, Solid .atlre Amokl. 19l11 C. D. ~1C'l".oo E. D. Daniel, MtlgndK' _du~ I",rrdl.oool.. Mc:Cr::ro..~IHU, 1990. it S. wadas. B~iml, Ellis 1101""00"', 1991 Good gencnl re\ic-. oflll<: lTIlICnetlC IISpC'ds III c)fp"isms. II. r.; Bertram:lrJ.d )ian-G3ng :a.." -Fundament.lllTlal;Oetu::.h... IlroC"l:Sia in IIIln-Glm rcoordiqg .nedlll. ~ SolId slate 1*')'Sics 46. 271 (19ll2) r. Crueller-, II. J Mamin,:and 0 Ruf;U, -Magnetic lOroe ,nicrosoopy (MF\I).- ... Sronning til.. fldi,"I: nticrwicoptll1. S1lringer- (Ul9"2)

~FcntlOIl3.I:lld.,:

ph)~J

"""ism.

16
Magnetic Resonance
l\'VCLEAU MAGNE'I1C I\ESOl\'ANCE EQlIlltiol15 of motion U(';E WIDTI! Motional narrQ\\ing 487 400 494 494

IIYPERFlNE spurn!\'c Examples: paramagnetic point defects Knidlt shirt

497 499 500

ro;UCLEAR QUADJWPOLE RESOf'A"'CE FElmO;\lAG",I-:nc RESOf'iA/,\CE Shape effects in FMR Spin wiwe reson31')(:e
Al\'TIFERROMAC"'.... "C IlESONAl\'CE
ELEemON PAI\AMACl'".... ."C I\ESOI\'Al'\'CE El'change narrowing Zero-field splitting

503 50:! 504 506

500
510 510 510

P1U!'iCIPLE OF MASEI\ ACrlOl" TIJn~e-lcvel mascr

nub)' lascr

510 512
512

5UMMAIW

514

1'1I0BLE!\IS

515
515 515 515 515 516
516

1. 2. 3. 4. 5.

Equivalent electrical circuit Rotating coordinate system Hyperfinc effects on ESIt in metals F!\II\ in the anisotropy Held ExdlaJlge frequency r'CSOllal1CC.' 6. nf saturation

REFEREl\'CES

516

,,"OTATIO!\:: In thls chaptcr the symbols B. and Bo reft'T to tilt- applied Geld, and 8 1 1$ the applied GeM plus tile dcm~ctizingfield In particular we "'Til... 8.. = Boi. For ecs n;adcrs il rna) bot simplcT to real! H for 8 Wht.'I1("\cT It 0ttUl'!i in tlris chapter

'"
~

1\
1-

'"
'"

"
0

J
o

V~ ..,

"- ~
",.
aI

,..,
298 K at 2 75 Cllz. after Za-visl",.

Fip,re I

E1.c>ct ...... spin n:,wllal'loe aboorption in M.s<).

".
~.,

CHAPTER 16: MAGNETIC RESONAKCE

In this chapter \\Ie discuss dynamical magnctie effects associatcd with the spin angular momentum of nuclei and of electrons. The principal phenomena are oftcn identified in the literature by their initial lellers, such as NMR: nuclear magnetic resonance NQE\: nuclear quadrupole reson.ance EPR or ESR: electron paramagnetic resonance (Fig. 1) FMR ferromagnetic resonance SWR: spin wave resonance (ferromagnetic films) AF""lt antifcrromagnetic reson.1nce CESH: conduction electron spin resonancc lne information that can be obt:lincd about solids by reSOnancc studies may be eatcgorizcrl: Electronic structure of singlc defects, as rcvealed by the fine structure of the absorption. f.,lotion of the spin or of the surroundin~, as revealed by changes in the line width. Internal magnetic fields sampled by the spin, as revealed by the position of the resonance line (chemical shift; Knight shift). Collceth'e spin excitations.
It is best to discuss NMR as a basis for a brief acoount of the other resonance experiments. A great impact ofN""R has been in organic chemistry and biochemistry, where NMR provides a powerful tool for the identification and the structure determination of complex molccules. This Sll<.'CCss is due to the cxtremely high resolution attainable in di:unagnetic liquids. A major mlXliCll1 appliC'Jtion of Nl\1H is magnetic resonancc imaging (MHI), which allows the resolution ill 30 of abnormal gJ"O\\lths, configurations, and reactions ill the whole body.

NUCLEAR MAGNETIC RESONANCE

\Ve consider a nucleus that possesses a magnetic moment II- and an angular momentum til. The 1\\-'0 quantities are parallel, and we may write
(I)

the magnetogyrie ratio')' is constant. By convention I denotes the nuclear ang,ular momentum mpasurcd in units of fr.

.,

".

.,

.- " .. r
~

,
-

TIle energy of intenlCtiOIl with the applied magnetic ficld is

U=-IJ.-B,,;

(2)

if B. = Bui. then
(3)

lhe allowed values of 1. are rill = I. 1 - 1... -1, and U = -1n/yABo_ In a magnetic field a nucleus with I = ! has two energy levels corresponding to In, = I, as in Fig. 2. rf!r.Wo denotes the energy difference between the two levels, then hWo = l'!lJJ o or
W(l=

yB o .

(4)

This is the funrlamcnlal conrlilion for magnetic resonance absorption. For the proton' y = 2.675 x lcrt S-I gauss-I = 2.675 x 10" S-I lesla- I , so that

tJ{t\tHz) "" 4.258 Bo(kilogauss) == 42.58 Bo(tcsla)

where v is the frequenc)'. One tcsla is precisely 10" gauss. Magnetic data for selected Iludei are given ill Table 1. For the electron spin.

v{Gllz) "" 2.80 BJkilogauss) "" 28.0 Bo(tesla) .

(4b)

"Thr m.gnd1C momentI'. 01 11M: protoo it lAI06 x 10 I:) ng C-' or I 4106 )< 10 J" . and 1'. '4J.,!" ~nudnr ~ 1'. i:ftkfilledQrt.l2MFMId isfl!ualto5.n'i09 x JO .. ~rg C 'or 5.0509 )( IO~" J T " lhus 1'.. - 219:'1 nudear IllllgllCtons.

H'
1/2 99.98 2.792

Table 1

Nudea.. magnctlt reSOnanCe data

H.'
1/2 10-"

For every element the most abundant m3gl'letic isotope is shown. After Varian Associates NMR Table.

-2127

lP

Bo'

3/' _3/2 9257 100. 3.256 -1171

Na:J M g1S
Most abundant isotope with nOl'lzero nuclear S(lIn Nuclear spin: il'1 units of K Natural abundance of IsotOpe. In percent Nuclear magnetic morl1l!nt. in units of eX/2M.c
~.

B" C" 3/2- _1/2

81.17
\.lOS

N"
I 99.64 6.404

0" 5/'
0.04

F"

Nell

'.688
,

0.702
Sj19

-1.ll9l

1)2. 3/' 100. 0.257 2.627 =-Q.662

Cps

312
100 "2.216

5/'
10.05

...........
Zn u

All' 5/'
100. 3.639

P"
1/' 100. 1.131

S"
3/'
0.74

1/'
4.70

3/'
75.4 0.821

0.855 Ca') 7/2"

0.13

0.555

0.643

K"
3/'
93.08

Sc' s
7/' 100 4.749

Ti o1 5/'
7.75 0.787

V"
7/' ':'\00
5.139

Cr Sl

3/2
b.474

0.391
Rb 8s

-1.315
Sr87

'"
572

Mn ss 5/'
100. 3.461

Fe H 1/'
2,245

COS~

0.090

7/2 100 4.639

Ni u 3/'
1.25 0.746

Cu bJ

Ga 69

3/2
69,09

5/'

3/'
60.2 2.011

2.221

4.12 0.874

'I'
Tn

Gel l
7.61

As'S 3/2100. 1.435

BrIO Sen 1/2 r 3/'

7.50 SO.57

Kr 8l

0.877
Sn ll '

0.533
TellS

'.099
jl17
-S12~

'"

11,55

-0.97
Xe11q

'5/2
72.8

Zrql 9/2"- 1/2- 5/7

7.02

y..

Nb ql
972"'l!1

MoqS

y,
r

1.34S

f.oM
Ball'

100 0.137

11.23 1.298

15.78 -b.l44 0.910

100.

CSlll
~2

la ll '
~7TZ"""

HfIH
18.39 0.61

Ta lll
100.

Wl8l
14.28
~0.1l5

Ru lol 5/2
16.98

R"IO)

Pd loS

Agiol

Cd lll
12.86

In lls

Sb l1l

172

-0.9

100. 0-:-088 ~:S1

5]2...- 112""'22.23 51.35

-172- '/2
Hgl qq

:;;2''''

-r72--Po

-OJlJ:
Au l "
'3?2~

95.84 -0.59f 5.507

8.68

-=tOll
Pb 101
m~

57,25 7.03 3.342 70882

Re l81

q Irl'J OslJ

PtI'S

TI105

100. 2.564

'3/2 11.32

0.931

F,
~.".--

.,
~

-'12':'E 7!'Z
'.340

172--- 572
62,93 3.176

;'/2
16,1 0.651

372
61,5 0.17

172"
33.7 0.600

02--r
16.86 0.498

'!Y1'

99' 2.761

100. tUM

70.48~ 21.11

-./,100.
4.039

BPo'

1.612

0.584

A,

"

., --26.24 100. 2.794- -0.773

TI2'-

At

Cel4l~

Prill
100.
3.92

Nd lOl

Pm

7/2 .. ' 5/2...

0.16
Th

7/,
12,20
-l.25

Sm1<7 Eu lSl

Gd lS7

Tb ls '

Dy l63
24.97

H OI6S

Er161

Tm 16'
100.
~0.i6-

Yb lll

lUllS

1.

7/2
15,07 -0.&8

5/'
52.23 1.521

3/'
1564

3/2
100.
1.52

5/2 - 712 - ?J2

-0.53
22.82 100. 3.3C- 0.48

11'

5!?
:'0571'

1)r

16.08 97.40

-0.34

P,

Np

p,

Am

Cm

Bk

Cf

E,
-

Fm

M'
~

No

" L'

- - --

''''
Equations of Motion
The rate ofchange of angular momentum of a system is equal 10 the torque that [lets on the system. The torque 011 a magnetic moment p. in a magnetic field B is p. x n, so thai we have the gyroscopic equation
IidIldt = p.
0'
l(

B" ;

(5)

dpJdt = -rIA-

l(

D.

(6)

111C nuclear magnctiwtioll M is the sum "PI over all the lIuclei in a lInit volume. If only a single isotope is important, we consider only a single value of 'Y, so that
dMJdt = yM x 8" .

(7)

We place the nuclei in a static field Da = Boi. In thermal equilibrium at temperature T the magnetization will be along z:
M~=O

M!I=O;
2

(8)

where t)](' Curie constant C = Np. /3k H . The magnetization of a system of spins with I = ! is related to the populatiOn difference N] - N2 of the lower ami upper Ic\"cls in Fig. 2: !II: = (N 1 - NVIl, where the N's refer to a unit volume. The popolation ratio in thcrmal equilibrium is just ghcn by thc Boltzmann factor for the energy difference 2J.tB o:

(9)
111e equilibrium magnetization is M o = NJ.t tanh(J.tBlkHT). \Vhen the magnetization t'Omponcnt M. is not in thermal equilibrium,. we supposc that it appr0<1ches equilibrium at a rute proportional to the departure from the equilibrium "alue Mo:

Mo-M~

T,

(10)

In the standard notatiOn T] is called the longitudinal relaxation time or the spin-lattice relaxation time. If at t = 0 an un magnetized specimen is placed in a magnetic field BoZ. the magnetization will increase from the initiul value M. = 0 to a final value M~ "" Mo. Before and jll~t afier the specimen is placed in thc field. the population N 1 will be equal to N 2 as appropriate to thennal equilibrium in zcro magnetiC field. it is necessary to reverse some spins to estublish the new equilibrium distribution in the field Bo. On integrating (10): (II)

---------

-----

"

"',--

f'"1t;Urt! 3 AI tome I = Oan umnagoctized sprcirncn M,(O) '" 0 is pba.-d ,ro a sl.llicmagrM"l..,rodd So. TIle n>ag1lelJ4Iior, ina-eases ...;Ih lime and ~ the rocwefluilibrM"" val"" 1010 - xoBo. n.is n:pcriOTlCl1I defi,1CS the lonf:iludlnal relaxation hme T, 'n.c mqneiic energy <!nlSit)' - M 8 dccreaocs as plJl ufthc >pin popubhOO lTIOYCS inlo the Iowa-Ie"d. 1hc M)"fl\*ll'" value at' .. T, is -MnBo- The """"l:Y flows from the SInn S)'5t",n 10 the S) ...em oCbttitt "iLnotlUfw, thus T, is alsu adled the spm ,",thee rcLuatiorl lilllC.

0'

T,

(12)

as in Fig. 3. 111e magnetic energy - M . B" decreases as M& approaches its new Cf"Juilibrium ,alue. Typical processes whereby the magnetization approaches equilibrium are Indicated in Fig. 4. The dominant spin-Ianice inter3clion of paramagnetic ions in (.T)'st3Is is by the phonon modulation of tile crystalline electric field. Helaxation proceeds hy three principal processes (Fig. 4b): direct (emission or absorption of a phonon); Haman (scattering of a phonon); and O..I,,"ch (intervention of a third state). A thorough experimental ll1H1lysis of spin-Iatlk'e relaxation in several rJre-earth salts at helium temperatures has been given by Scott and Jeffries;2 they discuss the evidence for the three processes. Tukillg tlCCOunt of (10), the :l; component of the equation of motion (7) becomes dM& _ ~I ) !Ito - M& -d- - 1\1\11( Bft + (133) t T,
&

where (M o - MJlTI is 3n e;dra term ill the equation of motion, arising from the spin-lattice interactions not included in the magnetic field B... 1111t is, besides precessing about the magnetic field, M will rela-; to the Cf"luilil>r m value 1\10If in a static field Boi the transverse m.1gJlCtization compol' nt M~ is not zero, then MJ will decay to zero, and similarly for M,. The deca)' OCCllrs help. L. Scott and C. D. Jclf"nes. Ph)"S. Rev. t27. 32 (t962).

492

[~.""'tor

Metal

Figure 4a Some imporlaot processes that L'Oolribute 10 100gitudiOlLl magoeti7.3lioo rela.ation io 30 insulalor and io a m..1a1.'ur Ihe illsulatur we shuwa I'huooll .ealten,d illelaslieWly by the spin system. The spin system 010'"'" 10 a lower eoergy st<lte, antllhc emit!l'<! phonon has higher energy by 11"'0 than the absoroc..1 phooon. I<or Ihe melal we show a similar inelastic sOlllering pro::.,ss in which a colI.luetioo e!<."Ctrun is scatlere<l.

Di.".,.

llaman

Or-booeb

ur," T

ur," T' or 1""

ur," up(-Q/k"T)

.igurc 4b Spin rebX>l.tiull frum 2 -. I hy phonon emission, phonon scatteriog, 'Illd .. two-singe ph01l00 proc<."SS. 'Ille temperature <lcpeodence of the loogitudillal rcl:lXlll.ioo 11I1IC T, is shown fur Ihe s<....cralproccss<.'S.

cause in thermal equilibrium the transverse components are :otero. \Vc can provide for transverse rda~atiOn: tIMjdt = y(M x BaJ. - M)T2 dM,jdt = y{M x
Ba)~

(131,)

- M,jl"2 ,

(13<)

where 1"2 is called the transverse relaxation time. 111e magnetic energy -M' Ba does 1.101 change as M. or Mil changes. prOvided that Ba is along z. No energy need flow out of the spin system during relaxation of M~ or My. so thai the conditions thai determine 12. may be less strict than for Tl _ Somelimes the two times are nearly equal, and sometimes l"t T 2 depending on local conditions. "'" The lime 1'2 is a measure ofthe lime during which the individual moments that contribute to M M ll remain in phase with each other. Different local magnetic fields at the different spins will cause them to precess at different frequencies. If initially the spins have a common phase. the phases will bemme random in the course of lime alUlthe values of M., Mil will become zerO. We can think of T 2 as a dephasing lime. The set of equations (13) are called the Bloch equations. They are not symmetrical in x, y, and z because we have biased the system with a static magnetic field along z. In ell:periments an rf magnetic field is usually applied

.
'!l3

(
To of IUJ'fIl)'

foo- ",no...-"" induct,.".,.,

"'" .....

and tit<:\lll

'.'- ~Il"

\:~/

(oW"'

along the i or Yaxes. Our main interest is ill the behavior of the magnetization in the combined rf lllld st.Hie fields. as in Fig. 5. The Bloch equations are p13usiblc, bul 1101 exact; they do not describe all spin phenomena, parliC\.llarly not those in solids. We determine the frequenc)' of free precession of the spin system in a static field 8,,::: llfft. [lIld witll M~ = .\/0 , The Bloch equations reduce 10
dilL --;h = O. We look for damped oscillator)' solutions of the form Mr = m exp(-tlT') COS wi ; M., = -m exp{-tfT') sin wi
(15)

(14)

On substitution in (14) we have for the left-hand equation

-<I)

sin wt -

rI

COS wi = --yIJ o sin wi -

I T; COS wf

(16)

so thut the free precession is characterizcrl by

~='YBo;

(17)

TIle motiOn (15) is similar to Ihnt of a damped harmonic oscillator in two oimensions. The analogy suggests correctly that the spin system will show resonance absorption of energy frOm a drh-ing field ncar the frequency wo =: ')IBo. and the frequency wKlth of the response of the system to the driving flcld will be 6w"'" 1n-2 " Figure 6 shows the resonance of protons in water_ 11,c Bloch equations may be soh"ed to give the power absorptiCHl from a rotating magnetic fieLd of amplitude B 1;

Bx

=:

B 1 cos wt ;

B"

=:

-B l sill wt .

(18)

After a routine C'o\lcuL"ltion onc finds that tht' power absorption is (CCS) lhe half-width
!l'(w) =
w')I.\f z T2 2I + (Wo - w)211 B1
halr-ma~imum

(19)
-

or the resonance at

power is
(20)

(6w)J:2 = liT, _

lion

Fi!ture 6 Prot..., ab.orpIn ....u...... (E I. 1I"',n.,

":1011_..

LINE WIDTH
Tile mJJ.,'Tlctic dipolar inh:racHOIl is Illouall)' the most impllrtant t:.IU~e of line bl"Oddcning in 3 rigid lattice of Illagnetic dipolc~. The Ill.lgllctic field 68 sLoen b)' a ffi,lgncHcdipolc IA-I due to a magnetic dipole #A2 at a point TI2 from the

fir,t dipole is
(CGS)

68 = 3(11-1' "drl! - p.2r~2

ri.

(21)

by a fund11l1cnt.11 result of nl.\gneto~t.lti<:s. 'l1le urJcr ofm.~lIitude of the interaction is, \\ith 8, wl'ittcn for 68,
(CGS) dominant. SO tlMl (CGS)

(22)

The strong dependence on r .'>uggcsis that close neighbor interactions will be

(23)

\\ here a is the scparotion ofnearest neighbors. -r1lis result gives US.l mCa!>U1 C of the width of the spin resonance line. assuming random orientation of Ih(' neighhurs. For protons .tt 2"\ scp..ar<llion. 1.4 X

B, -

1O~;!J
-ZI

G cm3

x 10

em

gaUloS

2 X 1O-~ Icsla

(24)

To C'=press (21), (22). and (23) in Sf. multipl)' the right-hand siues by 14:/47f.
Mo,joflul Ntlrrou:i,lg

The line widll] deerc.l~e:; for lIuclei ill r'lpid relath'e motion. l1le eff",..t 1Il sulids is IJlU~t rated hy Fig. 7. cllfru~ion resclllblc~ .. ranillr w.dk as atoms jllllll'

"

.,.

T~ ....

""

t:

..

Fig>are 7 EITt.'d of d.ffilSlOIl of nuclei OIl tho:- 1~1 N\m line width in rno:Wtir I.thlum AI low It'nol'"rlllu"" ,h., ",KIth ..gr...es ",nh I~ lho,on. ...,,-a! ,...rue kw ng.J Lit.... A~ I..... tcmprr.ll"", ,nc~s t.h., d,lfluloo rate ,~ and lhe Ime ..;dlh dec,t"<>SCS. The alm'l" dol-reasc III line ....dth :otxr..... T = 230 t; OC'<.~lrs when t!le ..),frus,olll hopping IlIne TIx-oo..1CS ~crth:m II')'H, Tbm the cx.,..rimetll ~"es a <I,""", 1lK'tit'A: orthc hopping lime (or an alon.IO ~ 1.>111.." 511.,... (Aft..... II S. (;ulO"~y amI B. f\ McC~r"C)'.)

from olle crystal site to another. All .llom remains ill Olle site for ,).11 a\,('fJ.ge

time

that decreases markt:dly as the temperature increases.

111C motional effects on the line width are even more spcctaClJlar in normal

liquids, bcccause the molecules are hi~hly mohile. The width of the proton rcson.mce line ill watcr is only 10-5 of the willth cKpeclC'n for watcr llIok'\:ules
frozen in position.

The cllcct of nucle.u motion on 1'2 and on the line width is subtle, but can be understood hy an clementJ.ry argument. \VC know fflll1l the Bloch equations that 1'2 is a measure of the time in which an indiviuual spill becomcs ueplJased by one radian because of a local pertulbJ.tion in the m,lgnetic field intensity. I,ct (l!.w)o - yB j dcnote the local frequency dc\ iation duc to a pcrturbation B., 111e IO(.~11 field may be caused by dipolar interactions with other spillS. Irthe atoms are ill nJpid relative motion, the local field BI seen by a given spin will fluctuate rapidly in time, We suppose that the local fidd has a \',llue +B j for an aver,lge time T .md then eh.mgc!> to -8" as in Fig. Sa, Such a mndom change could be eauscd by a change of the J.ngle behvel'll IJ. and r in (21), In the time T the spin will precess hy all extra phase all~le &p = :tyB,T relallve to the phase angle of the !>teadv prccc!>sion in the applied field Ijol1le motional narrowing effect arises for short T such that &p <C J. After n intervals of duration T the mean square dephasing angle in the field 8 0 \\ ill be
(25)

by analogy willi a random walk pi acess: the medII ."iC1ll.l.re displacement from the inilwl POSitiOIl afte~ steps of length e in r.tndom diuxtions is (r~ 0(2.

'"

. ,

,,rt-.,(I)

,-

-H"""

IomIGdd 6, - + I

'-

,'"-""~

,/
/

,i'/'

Figure Ii l'base d a Spill ill a <.VlldalltJoea] rleld, ;urornpared ",ith demas'ng of a spi" "hich after fIXed lime inlcn'als Thop$ al random amonK sites Jill' iug 1oc:a1 flClds :f: I. 1hc dep~iJ' is measured rclatn'c 10 the phase r:l a .pin ill the applil-'<I field Bo-

111C average number of steps IlCCCSSat}' to dcphll.';C a spill by onc radian is n = I/rB;~. (Spins dcphascd by much morc than OIlC mdian do not contribute 10 the ahsorption signal.) 111is number of sleps takes place ill a time

T z = 11'1" = I1yBrT

(26)

quite different from the rigid lallicc result T2 Iii! lIyB,. From (26) we obtaill as the Iinc width for rapid motion with a characteristic lime r.
(27)
0'

(28)

where {6w)o is the line ""idth ill the rigid Jatlice. l11c argument assumes that (.0.6))0'1" <C 1, as otherwise &p will not be <C 1. Thus I1w <l! (6w)o. The shorter is 'T, the narrower is the resonancc line! 111is
remarkable effect is known as motionul narrowing. 3 'nle rotational relaxalion
"'l1le physical ideas are due 10 r..'. Bioembelga E )1. Purwll. and It \' Pound. P~'S, 73. 679 (19-18) 11>e fCSIJlt diffas from the theory d opI>ca1line ..1<I1h CIlmed b, Itrn~ rofhsions bl't-.'-"C'11 atom. (as in I. ps di><,~)....here oholt , g;,a broNllil1c In 11>e nudear spin prohicm lbecollWons are ..-cak. In most optlClll proLlerm IhceolL.sions d alOmS arc Sll"Ollgcllough 10 ",lcrrep( II.". pha:st d Ihe oscillation In l1udcar resonance the ph.o.sc mar ''&I'}' smooth!)- in I. collision. although the li-equell()' rna)' ''I.I)' suddenl)' ffC/<ll one ,"llIue to another ncarb ,.,,]uc.

nc....

16

Magnetic

n...o"",nce

49

lime orv."aler molecules at room tempcrtlture is l..'nown rrom dielectric constant measurements to be or the order or ]0-10 s; jr(Aw)o - )~ S-I. thell (Aw)0710- 5 and!:J.w - (Aw~'T - 1 S-I. 111Us the motion narrows the pl'Oton resonancc line to about JO-.:l of the static width.

IIfPERFIl"E SPUTI11"G

TIle hyperfillc interaction is the magnetic interaction between the magnetic momellt or a nucleus and the magnetic moment of an electron, To an observer stationed on the nucleus, the illter.lCtion is caused by the magnetic field produced by the magnetic moment cf the electron and by the motion of the electron about the nucleus. TIlerc is an electron current about the nucleus if the electron is in a state with orbital angular momentum about the nucleus. But even if the electron is ill a state of zero orbital angular momentum, there is an electron spin current about the nucleus, ami this current gh'cs rise to the contact hyperfine i,JterlJCtion, of particular importance in solids. We can. understand the origin of the cont:JCl: interaction by a qunlitath'e physical argument, given in CCS. The results of the Dime theory of the c1ectron suggest that the magnetic moment IA-B =: diJ21/u; of the electron arises from the cireulation of an e1ectTOn ",ith velocity c in a current klop of radius approximately the electron Compton wa\1~lenglh, k.r=: ftjmL; - 10- 11 em. TIle electric current associated wilh the circulation is

or

1- e x (tums per unil time) - eelK...

and the magnetic Held (Fig. 9) produced by thc current is (CCS)

(29)

B - I1~c - elX; .
011

(30)

11w observer

the Iluc!cus Ims the probability (31)

of finding himself inside the electron, tb.'l.t is, within a sphere of volume ~ about the electron, I-Iere (O) is the \'aloe cf the electron w3vcfunct1oll 011 the nucleus. 11ms the average valuc or the magnctic field seen by the nucleus is
(32) where IA-Il =: ehl2mc =: iel\... is the &hr magneton. TIle contact part of Ihe hyperfme interaction energy is
U=: -p.,'

n... -lA-, , P.1l!tJI(0H" U=al'S

')'fl1A-1l1(OlI2I' S

(33)

where I is the Iludear spin ill units of h. The contact interaction in an atom h.. 1 5 the form (34)

...

Figllfe 9 \Iagnetic rw:1d B prodUC<'<l by a eha~ lTlO>ing in a .im,L.r loop. TIM: roulad part <lIb., h)"lxn.ne interaction ",th a n"clear magnetic 'lloment ar;SC$ from the .... gion within or ",,,, 10 the CtJrrenlloop. l'hc fleld l"'craged O\'cr a spherical shell thlll ,",1CfO$CS the loop gha zcro. Thus ilr an s ck'<.'tron (I.." 0) only the eonlad. part ronlribules to the ;nl...ra(.'lion.

Values of the hypcrGnc

arc:
nucleus
/l in gauss I' ;n ~llIz

COI1~tallt II

for tlle ground states of scveral free atoms

II'

Lj7

Nu23

K~

I
5ffI 112ll

'44
40'

i
310 B86

K" i
85 127

"',

83

The value of a in gauss as seen 0)' an electron spin is defined as aI2~1I' In a strong magnetic field the elLergy level scheme of a frcc atom or iOIl is dominatcO by the Zeeman ellclID' splitting orthe electron levels; the hypernnc hltcraclion gives a,l acMitionai splitting that in strong fields is U' ~ a1flsm" where "'5. 111, arc the magnetic quantum llumberS. For the energy Ic\'c1 diagram of Fig. 10 the two clectronic transitions ha\c thc selection nIles lims = I, lim, "" 0; thc fr('()ucncics arc w = -rHo a/2h. 1ltc nuclear transitions are not marked; the)' ha\c lims = 0, so that w""" = al2Jl. The frequency of the nuclear transition 1 __ 2 is equal 10 wt of 3 ...... -I.

...

.
,

-,

""2 "'-7

, ,

Figure 10 Energy Ic... c1~ Ina m;lgneli<' r.eW of a s)'slcm "'Ih 5 ..- .. J .. I. l10e diagram is dr.o."'T1 "r the stlQllg field appJUllimahOO jA..,/J J> II, ",II(:(C II it the 1')'Pl',;,nc COtlpling ,-"""tanl, l~n 10 be \)lXlli.c n,l' lOur lc\('" are Ialx:kd by the magnclk quantum "umhe" "',. m,. 'Ille s\loog electronic tra'lSIliuns I\;\\'<:

lu","

0.6"'5"- :2:1.

"Inc hypcrfinc inlcractioll in a nlJ.gnctic atom rna) split the ground energy level. TIle splitting in hydrogen is 1400 MHz; and Ihis is the radio frequency line of interstellar atomic hydrogen.

Exam,11es: Paramagnetic Point Ikfec-t&

1'lc hypcrfinc splitting of the electron spin resonance romisllcs \'aluable structural informatioll about paramagnetic point defects. such as the F Cf'lllcrs in alkali halide crystals and the (lonor atoms in semiconductor crystals. ,,' Centers in AlIwli Htllides. An F center is a Ilc~ali\'c ion VaclUlCY \\1lh onc excess electron bound at the vacancy (Fig. 11). l11c wnvcfunction of the trapped electron is shared chiefly among the six alkali ions adjacent to the vacant lattice site, with smaller amplitudes Oilihe 12 halide ions th..1.t form the shell of second Ilearest neighbors. nil.' counting applies to CI)'stals with the NaC! structure. If r,c(r) is dIe \vanfullctiOIl oCthe valence electron on a siog!e alkali ion. then in the first (or LeAD) approximation

(35)
where ill the NaCI structure the six values ofr p mark the alkali ion sitcs that IxJUlld the lattice vacalley. 'ne width of the electron spin rcson:Uice line of all F center is determined C\scntiall)' by the hypcrfine interaction orthe trapped electrOn with the nuclear

Figure II An'; .,.,.,ter is ol-pll,-e ion "ith one e....eu c!C<'lron bound at lhe vacancy. n.e distribution of the ell<.ftS electron i$1argrI), on lhe JlO"il~~ metal ions adjareflilo the \'aCan1 lallice ";11',
''lIC:UlC)'

magnetic moments of the alkali iOlls adjacent 10 lhe vacant L1.lIicc site. The obsen'Cd line width is evidence for the simple picture of the wnvcfunclion of the electron. By line ""iclth \.\'C mean the width of the em-c1ope of the possible hypcrfillc structure components. As au example, consider rul F ccnter in KC!. Natural potassium is 93 percenl K311 with nuclear spin 1 ;; i. TIle total spin of the six pota.~sillm nuclei at the 10' ccnter is 1m "" = 6 xi'" 9. so thai the lIumber of hypcrfinc components is 21 mru + I '" 19: this is the number of possible values of the quantum number m,_ There are (21 + 1)6 = 46 :c 4096 independellt arrangements orthe six spins dishibuloo into the 19 components, as in Fig. J2. Often we observe ollly the envclopc of the absorption linc of an F ccntcr. Dollor Atoms in SWoon. Phosphorus is a donor when present in silicon. Each donor atom has fivc electrons. of which four cntcr diamagnetically into the covalent bond network of the crystal, and the fifth bound electron acts as a paramagnetic center of spin S = l The CJi:pcrimental hyperfine splitting in the strong Geld limit is shown in Fig. 13. When the concentration exceeds about I x lOllS donors cm- 3 , the splil line is replaa.>d by a single narrow line. This is a motional narrowing effect (Eq. 28) of the rapid hopping ofthc donor elt.'Ctrolls amollg mallY donor atoms. Thc mpid hopping averages out the byperfine splilling. The hopping rate inCrCa'iCS at the higher conccntrations as the overlap of the dOllor clectron wavl." functions is incrcasc<l. a view supported by conductivity measurements (Chnpter 10). TIIC donor electron wavcfunction extends 1101 only over the CClltml donor alom but signil1cantly O\'Cr some hundreds of silicon atoms. 1bc Si 29 nuclear spins give additional h)"perfine splittings first studied by Feher with a powerful elcdron nuclear double J'CSOnance technique known as Er\DOR. Knight Shift At a fixed frequcncy the rcsonancc of a nuclear spin is observed at a slightly different m<ll,'llctic field in a metal than in a diamagnetic solid. TIle effeel is known as the Kni;ht shift or metallic shift rulC] is valuable as a tool for the study of conduction ek'Ctrons.

1-i z
"00 , '00

, -

6$ - 7- -4

. , . , . , I ,, , ,
QooMtum .......t- ....

Figure 12 The 4096 arrang<.-mcnt5 cJ the ~_ nodcaT spins of KlOlI as distributed ;nlo 19 hnll"rfine components. Ead, CQmponcnt ....ill he split further into. vel')' large numher of rompo"cnb hy ,'irlllc oflhe residual hYJl('rf'inc ink-raction wllh the 12 neighbor CI nuclei, whid, may be ern (75 percent) Of CIJ7 (25 percent). 'll,e .:""e1ope of the pallern i~ apprO"imately gaussian io fonn.

Ir;dion _

Fiun: 13 Elcctron spin ~ lmes ciP dono.- alOIn> in silicon. AI the I,igher dona roocenlJ>.., ...... tal.imubk)r llOln$ihon a donor ckrlron can hop frem S;II: to site so .. pidl~ thai theh)lll"rlinutmcture iU"~ (/\filTH. C. Fldchn. W. A. Yago. C L~. bld.-. R. Merrill.)

502

The intcraction cnergy of a lIuclcus of spin I and magnetogyric ratio 1'1 is

u = (-y,hB o + a(SJ)l~ ,

(36)

where thc first tcrm is the intcracliOIl witll thc applicd magnetic licld Bo and thc second is thc average h}1Jcrfinc interaction of the Ilucleus with thc conduction dectrons. The avcragc collduction c1c<:tron spin (SJ is related to the Pauli spin susceptibility X. of the conduetioll electrons: M~ = gNJ.t8(SJ = xJ3o, whcnce thc intCIac:tioll may be written as
u

~ (~Ylh +~) Bol~ =

gNP,o

-y,(IB o

(1 +fiB) I, Bo

(37)

The Knight shift is defined as

(38)
and simlilates a fi.. . 'etional chmlge ill thc magnetOb'Yric ratio. By the definition (3-4) of the bypcrfilll: contact cncrgy. thc KlIigllt shift is gh ell approximately h)' K "" X.Il/J(OW-fN; that is, by the Pauli spin sUlI(."Cptihllity illcre~ed in the ratio of thc conduction electron (:oncentmtion at the Iluclcus to thc avcrage (:onduction electron coneclltration. Experiment.11 valucs arc givcn in Tahlc 2. The vdlue of thc hypcrfiue coupling COli stant a is somewhat different in the metal than in the frcc atom becausc the wave fi.lllctions at the nuclcus are difiCrcnl. From the Knight shift of mctallic Li it is deduced that the valuc ofl1jJ(O)1 2 in thc metal is 0-44 oftllc value ill the frcc atom; a calculated value of the ratio usillg theorctiCltI "'>ave functions is 0.49. It is only in rare imtances that the absolutc valuc of the spin contribution X. to the magnetic susceptibility C-lll be determined, usudlly by very careful conduction e1ectrOIl spill resonance experiments. TIle Knight shift has becn of valuc in the study of mctdls, alloys, soft and intennetallie superconductor;, lUld unusual e1ectTollie systcms such as Na,.'VO l. Table 2 Knight Sllilh in NMR in metHlli" clements
(At room tempcr.lturc)
Knight shift in ["ICR"CIII Knight sluR in percent

Nudeus

Nucleus

u7

NaLl

AI.!)"

K"
\,~I

C,~

0.0261 0.112 0162 0265 0.':>80 O.6'J

Cu63 Rb"7 Pd ,m
Pt l 9:5 Au 197

-3.
1.4

0.237 0.653

-3.533
1.41

Pbw;"

J6

,\lagJ""ic Il"'Qnance

e
1"

(!)

j.l. p"po,,}

e
Ii""

.'igure 14 (a) (..m'est energy orientation of 3 nuclear clOOne <]u.K1n'po1c moment (Q > 0) in the Iocdl c1cdrie field oflhe ion. ,how". 'l1,e clccl,ons oflhe ion ,lst,If are not shown (Il) llighe.t encrgy oriclll~ti<.>n. (c) Thc cnc)' Ic\'el 'I,hllilll( for / .. I.

'"

.,

'"
,,'

NUCLEAR QVADRUI'OLE RESO:\'ANCE

Nuclei of spin J ~ 1 have an electric quadrupole moment. The quadrupole moment Q is a measure of the ellipticity or tile distribution of charge ill the lIudeus. The quantity of interest is defined classically by (39) where p(r) is the dJarge density. An egg-shaped nucleus hJ.s <! positive; a sauecr-shaped nueleus has Q negative. The nucleus when placed in a crrstal will see the electrostatic field of its environment, as in Fig. J4. If the symmetry of this field is lower than cuhic. then the nuelear 4uudrupole moment willlcad to a set of energy levels split by the interaction of the quadrupole moment witb the local electric field. The states tbat are split arc the 2l + J states of a spin I. 111e quadrupole splitlings can often be observed directly because an rf magnetic field of the appropriate frequency can causc transitions between the levels. The tenn nuclear lIuatlrupole resonancc refers to observations of nuclear quatlrupole spliUing.~ in tbe absence ofa static magnetic field. The quadnlpole spliltillgs are particuldrly large in covalently bonded molecules such as C1 2 , 8r2, and 12 ; the splittings are of the order 107 or Iff Hz.
FERROMAGNETIC RESONANCE

Spin resonance at micrOIl>71ve frequencies in fenumagnets is similar in principle to nuclear spin resonance. 11lC total electron magnetic moment of tIle specimen precesses about the direction of the static mag.netic field. and cnefl:,'Y is absorbed strongly from the rf h1Ulsverse field when its frequellcy is equal to the precessional frequency. We may think of the macroscopic vector S representillg the total spin of the ferromagnet as quantized in the static mag'rgy levels sepJ.mled by the usual Zeeman frequencies; the netic field. wilh

'"
magnetic selection rule .6.ms = j: I allows transitions omy bctv.ccn adjacent levels. TIle UIIUSUal fc.,tul'cs of ferromagnetic resonancc include:

The transverse susceptibility l.'omponcnts X' and )(' are \'crT large bcousc the magnetization of a fcrrom:lgnci in a given stalic field is vcry mucn larger than the magnetization of electronic or nuclear parnmagncis in the same
field. TIle shape of the specimen pL'l)'S all imporlanl role. Because the magnetiza-

tion is large, the dcmagnctizalKl11 licld is large. The strong exchange coupling between the ferromagnetic electrons tends 10 suppress the dipolar contribution to the line width. so thai the ferromagnetic resollance lines can be quite sharp 1 G) under favorable conditions. Saturation effects occur at )ow rf power levels. It is not possible, as it is with nuclear spin systems. to dri\"c a ferromagnetic spin s),stem so hard that the magnetization At:=. is reduced 10 zero or reversed. 11le ferromagnetic rcsomUlCC excitation breaks down into spin wave modcs before the magnetization vector can be rotated appreciooly from its initial dirccticn,
SllUpe Effects it! FMR
We treat the effects of spt:cimen shape 011 the resonancc frequency. Consider a specimen of a eubic ferromagnetic insulator in the form of all ellipsoid v.ith principal axes pamllcl to x. y, z axes of a cartesian coordinate s}'stcm. The demagnetization factors Nr N fl' N~ arc idellticnl with the depolarization fuctors dclined in Chapter 13. TIle components of the intel'llal magnetic field B; in the ellipsoid arc related to the appliCt"l field by
8~ =

If: -

N~lr ;

8~ = ~ - Ntl'l l1

8~ =

lf1 -

N=M= .

The Lorent"J". field (4w/3)M <uulthe exch.mge field API-'1 do not contribute to the torque because their vector product with M vanishes identiC',l.lIy. In SI we replace the (.'omponellts of 1\,1 by ~M. with the appropriate reckfjnition of the N's. TIle components of the spin equation of motion ~1 = ')'(M )C BI ) bl'COme. for an applied static field BoZ.

~:.. = ")'(M~~ - M=B~) = '}'[Bo + (Nfl dM

NJM]MfI

(40)

~ ~ ,;M(-N"'l,) - MollJ, - N,M)] ~ -';8.

+ (N. - NJ.If}.\f.

To first order we may sct clM)dt = 0 <Uld AI.. "" M. Solutions of (40) with time dependence exp(-iwt) exist if

;w -';8. + (N. - NJM)

~Bo + (Nfl - NJMI! = 0 . 'w

 'U

" , , , , , " -, , .,

,,

1-1gu..., 1S FMR ill a polidlled sphere or )Urium iron ~lCt at 3.33 Gill: and 300 Joe in' Bo II [III]. The tolallir>e "idth al halr-power is only 0.2 G, (AFicr R. C. LcCraw /lull E. Sp<.'flC<..... )

,L"k-~_m.,,---~~.-~-..!..-..J lIW.6 1129.0 Jtl!9.~ tll!9./i

AlJIllled 'nagrlt:llc rICk! {Ill
ga..

'S'

so that the fcrromagnelic resonance frequency in the applied field

(CGS)

J~

is
(41)

%=

rIB, + (N. - NJ-'Ij[l!o + (N. - NJM] ;

(51)

->! =

""lB.

(N. - NJ"",Ij[B. + (N. -

.)"",1]

111e frequcncy u.u is called the fn."(Juency of t1lC uniform mode, in distinction to the frc<Jucllcies of magllon and olher nonuniform modes. III thc unifoml modc all the moments prCCl.'5S togethcr ill phase \\;th the same amplitudc. For a sphere N~ "" Nfl "" N z , so that Cdo = ')'Bo. A \ery sllarp resonance linc in this gcometry is shown in Fig. 15. For a flat plate with Do perpendicular to the plate N~ = Nfl = 0: N z = 411', whence the fcrromagnetic resonance frequency is

(ees) u.u

= ')'(llo - 411'1\1.)

(51) "'" = ,{B. - ""'I) _

(42)

If 8 0 is p.1.rallcl to_ the plane of the plate, tile xz plane, then N~ = N z = 0; 11.'11 = 411', and

(51)

"'"

= nBo(l!o + "",I)]'.

(43)

111e e"I)Climellts determine y, \.. hich i~ relatro to the spectroscopic splitting factor g Iw -y rJ.LBi!l. "dlues of g for mctallic t-e, Co. Ni at room teml)Craturc arc 2..10. 2.18. and 2.21, respcc.1i.. e1)'.
Spin 'Vace Resonallce

Uniform rf magnetic fields can excite 10llg wa\c1cngth spin WU\es in thin ferromagnetic films if the electron spins on the Slln.\CCS of the film see llifferent anisotropy fields than the ~pins within the films. In dTect, the surface spins may be pinned by ~urfaC'e anisotropy interactions, as shown in Fig. 16. If the rf field is unifonn. it mil C\:cite wa"e'i with an odd number ofhalf-w;welcogths within the thickness of the film. \\'a.. es with all even /lumber ofh.llf-wavelengths IlU\'e 00 nel interaction energy \\ith the field. The condition for spin wa\'e resonance (SWI\) with the applictl magnetic (jell! normal to the film is obtainctl from (42) by aJding 10 the right-hand sil.lc the exchange contribution to the frefluency. 11K: exchange contribution ,TIdY IJe written as D~. where D is the spin w.we exchange constant. The .lSSulIlptiOfl ka <C 1 is valid for the swn experiments. 11ms in an applictl field 8 0 the 'ipill wave rcwllallce frequencies arc: (CCSI
~

= )'(80

-I-;r.\1)

+ DI.!J.

= y(Bo -

4111\1) + D(lImL'f '

(14)

where the wavevector for a mode of II half-""J.velellgthli in a film ofthickncss L is k = IImL. An e"lperimental spectrum is ,,!lown in Fig. 17.

ANTIFERROMACNJo:TlC RI-;SONANCE

'Ve consider a uniw;iill i1lltiferromagnet with spins on two sublaUices, 1 and 2. We suppose thilt the lnagnetiL.Ution 1\-1, on subl,lttice I is directoo along tile +z direelion by an ,lIlisotl'Opy field 8,4:2:: the anisotropy field (Chapter 15) results frolll an anisotropy energy density U,.,{O,) = K sin\! 0,. Ilerc 0, is the angle between M I and the;; a~is, whence B,4 = 2KJM. wilh AI = IMII = 11\1 21. The magncti7.ation MI! is dircctCll along the -z llircetion by an anisotropy field -B,\z. If += is an easy direction ofmagncti7.ation, so is -". If one sublatlice is dircctClI <\lollK +'z, the other will he dil'ccwd along -". The I::xchange interaction bctwCCIl M I and 1\1 2 is trC'J.tCll in the mean Held approximation. 11le exchange fields Me
(45)

where A is positi"e. Ilere 8, is the Geld thai acts on the spins of subl..lttke I, and 8 2 ,lets on sllblaltice 2. fnthe absence of all eJ(ternal magnetic feld the tolal field acting on M I is B I = -AM\! + B,,7~ the lolal field on ~I~ is 8.2 = -AMI - B"z. as in Fi". 18.

16

Ma('IIrlic Ikron"nce

5/'.

---11-----,,----\ tll.,-4>rM

t
\

---11---"-----Figure 16 Spill wowc n.";Ufla'K.'e in a thin film. Th~ plant' of Ihe film i$ normal 10 the "Pplied magn('li(' lfidd B,.. A CnJ$$ .!K'dion of Ih e film ;$ ~h<lwn h",e. Th" internal ,n.1gndi<' Field is B., - 411'\/. The 'pins on Ihe sm-raees ofth<, film are assume,1 to be hdd find in direction I.>y SIU-r.l.:<, '''''wlropy forees. A unifurm.rl' Field ",11 e!'ite .pin "'II"" modes ll:.ving an odd nllmller of hatfwa'c!C1'b'f.hs. Thc w,we shuwn IS fur /I .. 3 halr-w..,elenglhs

'~I'm "",,"" onlt-r numb.,

13

II

, , , ,
135&IC
13-474

13219

t j
It73Z

",,,
12191

Fil:ure 17 Spin wavc rcSOllall('C .peclrum ill a I'cnnaJ1oy (80,\',20Fc) film at 9 Glb. "11111 order ''''IIl~r i> the "'Hnlx,r of halr-".l\"clcngths in Ihe Ilud.ness or the film. (After R Wekr.)

508

/1I,(u).
I I

I
-oW,

I I

\
N,

.
\
\ \
A

~ 11,,(.,.,) .. -AM,

t'igUnil 18 Elieci:h-e fock\,; In ..... i~~ ~ance. 7he magnelinlion til, of ~:"bbtlice I $I fldd -AM. + 8..t; the magnetization M. set:I -A~(, - B"L Both ends of the CI}~

axis arc ~ easy ues ~ of magnctiution.

In what foIlO\\, we set motion arc

Mt = M; MI "" -AI. 'me linearized equations of

clMlfdt = 1"IMf(AM
llMffdt ""

dMdt

+ BAl - M(-AM)) -y(M(-AMi) - M1(AM + BA ,;M~-AM - n,) - (-MX-AMn! ;

(46)

dM'Vdt = ')'(-M)(-A/lfj) - M~{-},Al - IJ A )]

(47)

We define M t = !Iff + iM; Aft = IUt + iflll. Then (46) and (47) become, for lime dependence c.,<p(-iwt),
-iwM I' = -i;{Mun" + AM)

+ M{(AM)J
AM,

-iwM; = i;{Mt(B A

+ AM) + Mt(AM)]
II:

These equations have a solution if, with Bt :

')'{B... + B,.;)- w yB E
~ =

Thus the antifenomagnctic resonance frcq\Jcncy is given by

rB.. (B... + 28,.;) .

n", n

(48)

MnF2 is an extensively .studied antifcrromagnct. 'J1lc structure is shown in Fig. 19. The omcrvoo variation of ~ with tcmpcralurc is shown in Fig. 20. anti t; for MnF2. lie estimatcJ Careful cslimates were maJc by Keffer of BE = 540 kC anJ B", = 8.8 kC at 0 K, whence (2B",B f J'12 = 100 kC. The o}). scrvcJ value is 93 kC.

.-~

M'f'U"'S

Fipre 19 O>emiaol and magnetic st",dure ofMnFa- The indiCllte the direction and arr.lrlgement of the mag_ netic fR(Iments usigm:..:! to the mang;mese alams .

..
\
'00

'\..

'.
Figure 20
Antire~ic resQnarK"C

'"
lreqnc:..c)' b- Mnl', U"uus tcmpt"raIure. (Afler John)Qn

and l'ctbereot.)

Richan1.s has malic a oompilatit'>n nf AFMR lTequcncics as extrapolatoo to 0 K:

Crystal
F're(jutmcy in 10 10 Hz

CoF,
85.5

Ft:F1 158.

MnO

82.8

NiO 109

'"
t::LECfRON PARMIACNETIC Rf;SONANCE

Elcclron spill resonance is a vast area of research. We mention two topics of interest.

&cllarige NarrOlring
\\'C coilliidcr a paramagnet with nil exchange interaction] among nearest neighbor electron spins. The temperature is assumoo 10 be well above any spin-orucring temperature Te< Under these coooitions the width of the spin resonance line is llsually much narrower than cApooeol lOr the dipole-Jipole interaction. The effect is caJlcJ exchange narrowing; there is a close analogy \\;Ih motional narro\\ing. We interpret the exchange frcqucIIC)' w,.... ... ]Ill. as a hopping frequency lIT. Then by generalization of the motional-narrowing result (28) we have for the width of the exchange-narrowed line:

Aw .. (l1wWw~.

(49)

where (dw~ = r<nl1 is the square of the stalic Jipolar willth in the absence of exchange. A useful amI striking c.umplc of exdumgc narrowing is the paramagnetic organic crystal known as the g marker or Drrll, diphcnyl picryl hydrazyl, oftcn usaI for magnetic field calibration. This frcc radical ha.~ a 1.35 C halfwidth of the resonance line at half-power, only a few percent of the pure dipole width.

Zero-"ield S'llitting
A number of paramagnetic ions have crystal field splittings of their magnetic ground state energy levels in the range of 10 10 - 1011 J It:, convenicntly accessible by microwave techniq\lcs. Mueh of the pioneer work is due 10 B. Hleaney and co-workers at Oxford. 'J1lC Mn 2+ ion is popular and has been studiCt.1 in many crystals as an additive impurity. A ground slate splitting in the range 107 - 109 liz is ohscrved, according 10 the environment.

PRINCIPLE OF MASI';R ACTION

Crystals can be IIsal as microwave and light amplifiers amI as sources of coherent radiation. A masel' amplifics microwaves by the .~timulatcd cmission of radiaUon; a laser amplifies light by the same mcthoo. 'Inc principle, due to Townes, may be urklerstood from the t,,-'O-lc\'e1 magnetic systcm of Fig. 21. 'nl(~re arc 11 .. atoms in Ihe upper state :ind III atoms in the lower slate. \\'e immerse the system in raJiation at frequency 41; the amphtutle of the magnetic component of Ihe radiation fieW is Brf 1bc probability per atom per unit time

..

16 Magnetic RnolWlncr
u~ ....., ..L ~ -';:;:;;;';;;;;"-"'---T;)'"

Figu~

21 A two-Ieod S)$km. 10 npbon rnaM'I" oper-

alioo. The popublioas of I~ upPer lIIl<1 ~ stales ar., n" and "t. rc:spc:rt".,1) &eoque~ oflhe .,mil-

n.c

led ...adution
~

is w;

,he combined ,..lth of the loUles ;5

'"' t:-. + &.o.!.

of a transiUon between the upper and lower states is

p~(~B")'_I_

(50)

" Aw' here p. is the magnetic moment. amI !:J.w is the combinet.1 width of the two levels. TIle re~u1t (50) is from a standard result of quantum mechanics, eallel.1 Fermi's golden rule. The nct encrgy emittCt.I from atoms in both upper and lower states is

fP

p.B,!), l1w' I (-h-

fUJ)

(n.. -

II,) ,

(51)

per unit time. Ilerefll.1cnotes the flower out; hw is the energy per photon; anl.1 lin - "/ is the excess of the number of atoms " .. initially able to emit a proton O\'cr the number of atoms 'Ii able to absorb a photon. In them131 equilibrium " .. < '1/. so there is no oct emission of radiation. but in a noncquilibrium condition with II" > III there will be emissioo. If we start with 'I" > ", and reflect thc cmitted radiation back onto the system. we incrca~e B rf and thereby stimulate a higher rate of emissioo. The enhancclI stimulation continues unti! the !>oplliation in the upper state decreases anl.1 becomes eqllal to the population in the lower state. We can buill.1 up the intensity of tile radiation lIetl.1 by placing the crystal in an electromagnetic cavity. This is like multiple reflection fmm the walls of the cavity. 'IlLere will be some pll\\cr loss in the walls of the cavity: the rate of powC1" loss is
n~v w (CCS) 'iI' - - _ . '.- 811" Q

(SI)

(52)

wherc V is the \'fllume and Q is IheQ factor ofthecavity. Wc understaOtl B~to be a volume a\'erage. The condition for maser action i.~ Ihat thc emitted powcr (j} cxcecd the power los-~ flP,.. Bolh quantities involve B~f' The maser comlilion can now be

'"
cxprcsscJ in terms of the population excess in the lIppcl .ate:
(eeS)
II.. II,

> -- ,
8r.p.Q

VllB

r
(SI) L
II.. '1/

V.6B

..

> 214JjJfJ

-,

(53)

where J.l. is the magnetic moment. TIle line width dB is <.Icfincd in terms of the combined line widOl 6w d lhe upper ant.! lower slates as p.6B = li.!:J.w. The central problem of the maser or laser is to obtain a suitable excess population in the upper state. 11Us is accomplished in various wars in various devices.

Three-Level Maser

The

thrce-l~c1

maser system (Fig. 22) is a dcvcr solution to the excess

population problem. Such a system may tlcrivc its energy levels from magnetic ions in a crystal, as Blocmbcrgcn showed. Rf power is applioo at the pump frequency liwp = E:J - E, in sufficient intensity to maintain tlle population of levc13 substanliaIl), equal to the population oflcvcl I. Now consider the rate of change of the population 'l2 of Ic\o'c1 2 owing 10 normal thennal rcla;tation processes. In Icnns of the indicaleJ transition ratcs P,

d,kjdt
113

= -'12P(2 -

1) - IIzp(2- 3)

+ 113PC3- 2) + '1IP(I- 2).

(54)

In the ~1.cady state d,wjdt = 0, and by virtue of the saturation rfpower we have = "I, whence
"2
~

fll

P(3-2) + 1'(1- 2) P(2-1) + P(2-3)

The transition rates are affeeted by many Jetails of the paramagnetic ion and its environmcnt, but onc can hardly fail with this systcm, for either 1'12 > II, and we get maser action between levels 2 and l, or 1'12 < "I = 113 and wC get maser action between le\els 3 and 2. '1l1C enerb'Y levels of the Er3 -t ion are uSCl-l in communication fiber optics amplifiers, as treated at the end of Chapter 17. The ion is optically pumped from level 1 to level 3; thcw is fast nonradiative decay from level 3 to level 2. The signal at a wavelength of 1.55 ~un is amplifial b)' stimulatcxl emission from level 2 to level!. 'Ille wavelength is favorable for long-distance propagation in the optical nbcI'. The bandwidth is of the onlcr of if x JOI2 Hz.

Ruby Laser
The same crystal, nib)', IlS00 in the microv.'llve mascr \.\.'as also the firsl crystal 10 exhibit optical maser action, but a different set of cncrb'Y Icvcls of Cr''t- arc involved (Fig. 23). About 15,000 em-I above thc ground statc there lie a pair of states labcloo 2, spaccJ 29 em-I apart. Abovc 2 lic two broad banJs of states, labelcJ 4F I amI 4F2 Becausc the bands are broad thc)' can be

...
C=;:==..

..... .....

=="

,
Figu", 22 Th~1evd m-=r S)Uem Two pmsibk modes r openlion u./uralion of the nates 3 alld I to obtain OJ - "I'
art'

shown. starting rrom rf

"'r_--"o""""'' ' ' ' "' ' '-

,'"O'""o.'.. e'' -'.o ."'' '

'F.--'.;

.. "
.~ '" ,-

. , ..

" ...[['==:3

-----

29 .... - 1

,LJ,..------"""'"'o--,--......---------'
Figure 23 Ent.W level diagnlTl dcr' in ruh), as U5Cd in Ialer q>cration. lbe initial acilatiotl takes place to the broad ba.>ck; the) ~)' (0 thl- intemx'<iiatc leo.'eh b)' the .:miuion olllhoolOn5, Ilnd the intennooiale levels radiate phulOI1S MS the ion makcs the transition 10 the ground Ie.wl.

populalet.\ efficiently by optical absorptiOil from broatlbantllight sources such as xenon flash lamps.

In operation of a ruby laser both ofthc broac.l "F bands arc populatcJ b)' broadb.'mtl light. Atoms thus excited "ill decay in 10- 7 sec by raJiationlcss processes with the cmiss'on of phor"lOOS to the stalcs 2. Photon emission from the lower of the states 2f to the ground state occurs slowly. in aoout 5 x 10- 3 sec, so thal a large excited population can pile up in 2. For laser action this population mllst exccctl that in the ground stale. The storlXl energy in mby is lfil erg an -3 if 1000 c.-3+ ions cm -3 lire in an excited state. The nIb)' laser can emit at a very high power level if all this storet..l energy comcs out in a short burst. The overnll cfficiency ofcon\'crsion ofa nib)' laser from input electrical energy to output laser light is about one percent.

,I<

"'igure U Foo,rk-...:I
~r S)~"'m.

as ill the

ncoo.I)nllUm gLiu b..,r.

Anotlu:'r popular solid state laser is the neodymiulll glass laser. made ofcalcium tllllg.~latc glass dopeJ with Nd' ions. 'Ibh operates as a fouf le\'c1 system (Fig. 24). Here it i.~ not ncccssal)' to empt), oul the ground state before laser action can occur. Semiconductor junction lasen ,uc Irca.t"..d in CllaplCr 19.
SU~I~IAR\'

(In CCS Vllits)

+- -niC

resonance frequency of a free spin is

Wll

= '}'Bo. where 'Y "" j.JflI is the

mab'l1ctogyric ratio. 111e Bloch equations arc

dM,/dl = ;(M x U)~ - MiT'/.

dM.jdt = ){M)l 8)1/ - M.jT2

dMidt = y(M x D)~

+ (,\#0

- M,JJT,

11,e half-width of the resonancc at half.power is (l!.W)1I2

++-

= IfT2.

Saturation effects at high .. rpower enter when

rn!fT1T2 exceeds

unity.

The lhpolar line width in a rigid lattice is (.6.8)0 - plfil.

If the magnetic moments arc ambulatory, with a c11aracteristic time 'T 4: 1/(6w)o, the line width i~ roouccd by the factor (6W}o7. In thi~ limit UTI""" IIT2 - (6.w~7. With cxchange coupling in d paramagnet the line y.idth b:omcs "",(Aw)3wu

16 !ltagrw!;c Resonance

S'

.. The ferromagnetic re~Onance frcquen~y ill all ellipsoid of demagneti.tation factors N x , Ny, N~ is w~ "= y[B o + (Ny - NJMIlBo + (N. - N.)Mj . .. Tllc antiferromagnetic re~onancc frequency is wi\ "= yBJ\(BJ\ + 21J.t;), in a spherical specimen with zero applilXl field. lIere BJ\ is the anisotropy field and BE is lhe exchange field . .. Thc condition for ma.~cr adion is that IJ" - IJI

> V!:J.B/81Tf.LQ

Problems
1.
Equi~alellt electrical circuit. Cansi,lcr an tmpty <:oil of indud:an<.'C to ill a sl,.>ric~ \\ith a resistance Ro; show if thc coil is complctely 1lI1e.;J with a spin s)'stcm dmr.u.1crize<1 by the susccptibilit), COmpOnL"Tlts x'(w) and x"(w) that the m<luctance at fre<JuelK'Y W becomes L = (I + 47TX'(w)]Lo. in series with an eff<.'dive resistance R '" 47TWx"(W)I.,;. + Ro. In this problem X '" x' + ix" is defined for a linearly polari7.ed rf field. Hillt; Consider the impL,,<lance of the circuit. (CCS unit~.)

2. Rotatillg coordina'e system. We define the \'cctor F(t) '" I''/t)i + I'Jt)y + 1'.(t):;'. Let the coor<lmate s)'stem of 'nc unit \'edors i, y, i rotate with an instantaneous anJ;;ular \"tolocity n, so Ihat diM! '" nyi - nS, etc. (a) Show dmt dF/dt z= (dFld')R + n l( F, where (dFld')R is tile time (lcrivati\'e of F as "iewed in the rotating frame R. (b) Show lhal (7) may be wnti<'n (dMldt)R '" )'1\1 l( (Bd + flJ)'), This is the equation of rnlltion of M in a rotaling cooruin;lle system, TIle tr,uufortnation to a rotating system is extraortlinarily useful; it is e~ploited Widely in till' liter.tture. (c) Let n "" - )'noi; thus in the rotating fr<lme there is no static magnetic fiel<]. ShU in the rotating frame, we now apply a lIe pulse nIx for a thlle t. If the magnetlznlion is initiall)' along z, find an expression for the pulse length t sueh that the ITIaj!;netizntion \\ ill be dir<,.'Ctcd along -:;, at the end of the puht. (Negkd relaxation elfCf..1.s.) (d) Decribe this pulse a.~ "iewe<l from tbe laboratory frame of leference.
3, I1yperfine effects 011 ESn in me'ab. We suppose that tbe electron spin of a COut/uetion electron in a metal sees an elfective magnetic field from the IJn>erfine intcmction of the c1<.'Ctron spill wilh ,he nuclear spin. Let the;:; component of the ReM S('cn by the conlluction electron be writtcn

n
where ,; i~ equally likely to be (nt) = 3(af2N"fN1 for N t> l.

j ..

(~)J~I

,; ,

i (a) Sbow that (nf)

= (al2N)zN. (b) Show Ihat

4. FMR in 'lie ouiso'ropy field. Consider a spherical specimen of a unia.~ial ferromagnetic crystal with an anisotropy energy llensity of the form U" = K sill! (J, where (J is the angle bet....'Cen the magnetization and the;:; axis, \Ve as.~ume that K is positive. Show that tbe ferromagnetic resonance frequency in an e.demal magnl,.'lic fiel,1 Boi is Wo = )(no + BA ), where n A 2K1lIf,.

'16

5. Exchongefnqr.tency mononce. Consider a fCrTlmagnct \\ith .} sublatliccs A and B of magncbutions M" and M B where M B is opposite to hi" whcn the spin S)'Stcm is
at rest. TIle gyromagnetic ratios are YA. 'YB and the molecular flClds arc 0" "" -,\,1\111; 8 11 - -AM.... Show that there is a resonance at

This is caned the CM:b.ulgc frequCIlC)' resonance. 6. Rf Mt"rction. Given, at equilibrium for temperature T. a tv.'O'-I~'C1 spin S)'Stcm in a magnetic rJCklll~ with population.\ 1'0'10 Nt and transition nUes \VI!. W1l1 _ We 811M' an If signal that gi\'CS a tnnsition rate Wof (a) Dcrj,,'e the ccluation lOr dM/dt and show thai in the stead} state

....-here In".

l'Io -1\'(W21 - W.!y(W21

WI! + W,._ It \\ill be hclpfullO \\TileA' - N. + Nt. n = N. - 1'1'2; and + W1:z). We see that lIS k>ng lIS 2W.rTi I the absorptic;wl of energy from the rf Gcld docs oot substantially alter the population dishibutiou from its thermal equilibrium value (b) Using the exprCSliioo for n. write dOWTl the
:z

rate at which energy is absorbed from the rf fJeld. Whal happens as W..- approaches lJ'2T\? This effect is called saturation, and ill; ()Il~t ma)' he used 10 measure T I .

References
INTRODUcnON
C P. Slichtcr, Prindpla of magnetic rUOflam:e, 3rd aI. Sllrinser. 1990. An excellellt introduc-

tion.

NUCLEAR MACN"-iIC RESONA"'CE

Wolf, Spin t"mperature and "ucleor spill ITlaxa/lon, O.r..,rd, 1979. A. Ahragam. Nuclear "",gncf's"" O..sonl. 1961. Definitj~c and oomprehclI,jl.'c A Abraham and M. Cotdman, Nuclear magneli,.,. on/er a",1 lliwnfer. O.ford. 1962. T. I'. Das and E. L Hahn. "",,,cleaT 'l"aJll,pole resonance Spc<:lros<.vllY,' Solid statc .,hy,ie.;, S".,.,lcment I, 1958. D. C. C9dion. NuckorlllogJlt'lic ~"'nonCll ami ill appllCIIllon 10 IIdng SIISICII'lJ', Oxfo..d. 1982. V. 'ac<:8nno, "Nudear TeWI,,,ncc in antiferromll,l\lleU," Mognet/s'~'. II". A. ~l. I'orti, and R H. Li""(I,,;,;t, "Nuclear TellOTI8'I(.'e in ferromagnetic malcriah," !lfognf/lsm IIA,
I).

351. !'ergamoo,. A $Cries. G A. W"bb. ".I, Annual n'porn On NAlIllpeCfrCCCllW'. Academic Prelos, l!l6O. A. L Horowitz. AIm phyliQ for Ihe phyridan, Spring.,r, 1969.
PrtlgrY!U in nudCIIT .1IOgnrtle n:$anollCe J11IeC'lI"ll$COJ1V.

ELECTRON Sl'IN RESONA"'CE

A. Abng;lm Ilnd H. 131eancy, t:kctron poronlllgnetlr reIOllDtIlY of Ironli/1OfI 1onI:, Oxfonl. 1910. \V. Low, "l'aramagndlC reSonance in lIOlMls." SoIid"tatephysiCli, Supp,",mcnl 2. 1960. l( J Slandley IlTId R A Vaughan. Eledron 'P.n rel.uDtitln In lOlidr. Ihlg.,... London, 1900 J. E. W"rtz and J. R Bolton. ElIXIr"" lpin raollDllCI!. Chapman ..,dllan, 1986.

Etl,lm by C. T. Radoand II Suhl, Aca<kmic I're$s.

..
16 Mognt,ic 1U:."""ncc

FERKO- ,,!"to ,TlFUtRO!>IACNI:."-IC RESONAI"CE M Sparb. F~ic r'do%a'lOfl 'hrory, M<Cn.. ~lhl'. I~. S Foner. -Antd"t'mlmagt>Cbc and ICrrimagnehc .nonance.- Mopldilm I. 384. C W. Haas and II. 8. ~Ilcn, "FClTOfll-.gncIIC rel,lntiOlI.-.11 'C'iOlIat>Ce lin.c wldths,~ MGglwtlsm I,

n.

W. Damon, MFen'Omagnchc resonBn"" at high

"".

po""~r:'

Magnetism I, 552.

QUANnlM ELECTRONICS C J F. Troup, Mf!#n lind tOlen, 2rld ftI.. 1I:ahlcd, 1973. A Yari", Quantum ckctnmia. W,ley, 3n1 ed. 1959. J W Orton, D. II P""man. and J. C. WallUl~ Solid sIDle .rnuer. Perg:aroo11. 1970. II. M. "'u:lU!Dzvcig. In'rl)(fllClltm 10 quonh"" optics, (',.(ltdon ant! llr~h, 1973.

oq

......

17
Noncrystalline Solids
DlFFRACflON PATfERN 521

MonatomK- amorphous materials Radial distributiorl roncticm StTuc:ture of "itn:ous silK-a, SiO,. GLASSES
Viscosity and tht: hopping rate

522 523 524 511

528

AMonPllOUS

I~ERnOMAGNETS

52!)
53(l

AMOI\PHQUS SEMICONDUCTORS LOW ENERCY EXCITATIONS IN AMORPIIOVS SOLIDS Heat ~apacity calClllatioo Thermal t"Oflducthity FIBER OPTICS Rayleigh attenuation REFERENCES

532 533
.534

53S 536 537

--

CHAPTER 17: NONCRYSTALLINE SOUDS

It is gencrally agreed that the tcrms amorphous solid, noncrystalline solid, disordcred solid, glass, or liquid have no precise structural meaning beyond tilc dcscription that thc structurc is "not crystalline on any significant scale." The principal structural ordcr prcsent is imposed by the approximately constant scparation of ncarest-ncighbor atoms or molecules. \\le exclude from the present discussion disordercd crystalline alloys (Chapter 21) whcre different atoms randomly occupy the sitcs of a regular crystal lattice. DIFFRACTION PATIERN

The x-ray or neutron diffraction pattern of an amorphous material such a~ a liquid or a glass consists of onc or morc broad diffuse rings, whcn vicwed on thc planc nOI'mal to the incident x-ray beam. The pattern is diffcrent from the diffraction pattern of powdcred crystallinc material which shows a large number of fairly sharp rings. The result tells us that a liquid docs not havc a unit of structure that repeats itself idcnticalIy at periodiC intervals in three dimensions. In a simple monatomic liqUid the positions of the atoms show only a short range structure referred to an origin 011 anyone atom. "'e never find thc centcr of another atom closer than a distance equal to thc atomic diameter, but at roughly this distancc we expect to find about the number of nearest-neighbor atoms that we find in a crystalline form of the material. Although thc x-ray pattcrn of a typical amorphous material is distinctl)' different from that of a typical cl)'stallinc material, there is no sharp division betwecn them. For crystalline powder samples of smaller and smaller partide size, the powder pattelTl lines broadcn continuo\lsl)', and for small enough crystalline particles the pattern becomes similar to the amorphous pattern of a liqUid or a glass. From a typical liquid or glass diffraction pattern, containing three 01' four diffuse rings, the only quantity which can be determined directly is the radial distribution function. This is obtained from a Fourier analysis ofthe experimental x-ray scattering curvc, and gives directly the avernge number of atoms to be found at any distance from a given atom. 'l1\e method of Fourier analysis is C<jually applicable to a liquid, a glass, or a powdered crystalline material. It is convenient to begin the analysis of the diffraction pattern with Eq. (2.43). Instead ofwriting it for the structure factor of the basis, we v.;rite thc slim for all thc atoms in the specimcn. Further, instead of specializing the scattering to the reciprocal lattice vcctors G characteristic of a crystal. we considcr arbitrary scattering vectors 6.k :: k' - k, as in Fig. 2.6. We do this be cause scattering from amorphous materials is not limited to the reciprocallaltiee vectors, which in any event cannot here be defined.

'"

l1lCrefore the scattcrl-d amplitude from an amorphous material is described hy

S(tik) ==

2: f~, exp( - itik . r m)

(1)

with!.;, the atomic form factor of atom, as in Efl. (2.50). 111e sum runs over all atom~ ill the specimen. 1be sc.ttlered intensity at scattering vcctor tiL: is given by
(2)

'" "
ill units referred to the scattering from a single electron. If a denotes the angle bctv.'Cen tik and T'" - r", then
I ==

2: 2: f",f" cxp(iKr,,,,, cos a)

'" "

(3)

where K is the magnihldc of tik and Tm " is the magnitude of r,,, - T... In.m amorphous specimen the vector Tm - r.. may take on all orientations, so ..ve average the phase factor over a sphere:

(e~p(iKr cos a

= -'- 21T 4'1'1"

:

r
_I

J(cosa) exp(iKr"," cos a)

sin Kr_

(4)

K,_

Thus we ha..c the Deb)'e result for the scattered densit)':

I ==

2: 2: (f,J.. sill Kr~I ..)/Kr",,,

,. "

(5)

Monatomic AmoryJllOus Motcriou

For atoms of anI)' one type, we let f,,, == I" ::: I and separate out from the summation (5) the terms with II == m. For a specimen of N atoms.
I =

J\~[ I

L' (sin Kr.....)lKr"",]

'"

(6)

The sum nlns ~ocr all atoms m cxcept the origin atom m = 11. If p(r) is the concentmtion of atoms al distance r from a reference atom ....oc can ....Tite (6) 35

I == Nr[ I + L'~ dr 47Tr2p(rXsin KrYKr]

(1)

where R il. the (very large) radius 01 the speClI.w"........ conccntratiou; thell (7) rna)' be written as
I =

MJ

_~_.

l\:rz{ 1 + L'l. ell' 41Tr:![p(r) -

PoI(sin Kr)/Kr +

CPo!K)

ell' 41fr sin Kr}

(8)

The se('ond integral in (8) ghes the scattering from a uniform concentration and Ina)' be neglected e,cept in the forward region of very small angles: it reduces to a delta function at the origin as R _ <XI.

Radial Distribution Function

It is convenient to introduce tile 1i1luid structure factor defined by

S(K)

l/N{' .

(9)

Note that this is not at all the same as S(Ak) in (I). From (8) we have, nner dropping the delta function contribution,
5(K) :: 1

L" d,' 471T {p(r) 2

Poksin Kr)/Kr .

(10)

We define the radial distribution function g(r) such that

P(,) ~
Then (10) becomes
5(K) = 1

"'dPo .
Kr)/Kr

(11)

+ 41TPo roo c1r{g(r) - l],-2(sin

"" 1
bccau~e

+Pof elr{g(r)-llexp(iKr).

(12)
e~pansion of

(sin K.r)lKr is the spherically symmetric or s term in the

e:o;p(iK' f). By the Fourier integral theorem in three dimensions,

g(r) - 1 =
=

8~Po
2~W

JelK [S(K) -

I} exp(-iK' 1')
(13)

dK 15(1\') - I] K sin Kr .

This rcsult allows us to calculate tbe radial distribution function g(r) (also called the two-atom correlation function) from the measured structure facto" S(K). One of the simplest liquids well suited to x-ray diffraction study is liquid sodium. TIm plot of the radial distribution 41fr 2p(r) vs. r is given in Fig. I, together with the distribution of neighbors in crystalline sodium.

....

Il

r
"

fl
"
<

,
0

~"
.J 1[' -Vi'

"

'

[,
,
"

,.,

Figure I (a) R:odiaI dislzibution WI'Yt! 4~p(r) br liquid $OdIUm. (bl A'~c dclUity cur.'e 4wr"Po- (e) Ouhiootion cl neig.l-l in Cf)'SIalline mum. (Afler Tansov and W.rrt'n)

Structure

of Vitrevus Silica,

5iO,

Vitreous silica (fused quartz) is a simple glass. The x-ray scattering curve is given in Fig. 2. The radial distribution curve 41Tr'P(r) vs. r is given in Fig. 3. Decausc there arc h\'o kinds of atoms, p{r) is actually the superposition of rnu electron concentration C\Irvcs, one about a silicon atom as origin and one about an ox)'gen atom as origin.

The first peak is at 1.62 A, close to the average Si-O dbtance found in crystalline silicates. The x-ray workers conclude from the intensity of the first peak that eadl silk'Oli atom is tctrahcdrall) surrounded by four oxygen atoms. The relalive proportions of 5i and 0 tell us that each 0 alom is bonded to two 5i atoms. From the geometry of a tetrahedron. the 0-0 distance should be 2.65 A, compatible \\ith the distance suggested by the shoulder in Fig. 3. The ,,_ra)' results are consistent with the standard model of an oxide glass. due to Zachariasen. Figure 4 illustrates in two dimensions thc irregular structure of a glass and the regularly TCJX!3ling structure of a Cl)'stal of identical dlCmical composition. The x-ray rcsuhs are completely explained by picturing glasS)' silica as a random nehvork in whidl each silicon is tetrahedrally surrounded by four ox)'gcns, each oxygen bonded to t\\O silicons, the two bonds to an oxygen being roughly diametrically opposite. The orientation of one tetrahedra! group "vith respect to a neighboring group about the connecting Si-O-Si

. "
525

00

I\...
-,""

'"

"

Figure 2 Surttere<1 x-ra~ intensity

\'S

scattering angle 6. lOr vltI'reus SiD:. (After B. E. Warren.)

bond can be practically random. 111crc is a dd"initc structural scheme involved: each alom has a definite number of nearest neighbors al a definite distance, but no unit of structure repeats itself identically at regular intervals in three dimensions, and hence the material is not crystalline. It is not possible to explain tbe x-ray results by assuming that vitroous silica consists of very small crystals of some crystalline form of quartz, such as cristoballitc. Small angle x-ray scattering is not obsclVed, but would be expected from discrete partides with breaks and voids between them. The scheme of bonding in gla'is must be essentially continuous, at ICa'il for the major part of the materia!, although tbe scheme of coordination about each atom is the same in vitreous silica and in crystalline eristoballile. nle low thermal conductivity of glasses at room temperature, as discussed presently. also is ronsistent with the continuous random network lll~el. A comparison of experimental and calculated results for amorphous germanium is shown in Fig. 5. TIle calculations are for a random network model and for a microcrystallite mOOel. 11le Iauer model gives a very poor agreement. The random network model is supported for amorphous silicon by studies of the band gap and spcctroscopic work on the 21' shell.

526

1 ..L ""I ..
IN
\

/
Si.SI

"

"

r..
,.A

"

"

..
or

"

Fig".... :1 Rat!i:ll ,Jislribuhon curve kit" nlreous SiOt. as the Fourier transform Fig. 2. The JlO'llions orthc lleak. give the <listanct$ of aloms from a silicon or an mygcn. From tl,e areas unJo.... the peaks it is IlOSs,ble to c;llculale the number of neighbors at that distance. The verticllllines m<llCale tbe llnllew ~ intcntomx: distances; the heights ollhe lint's an: proportional 10 the peak arcu. (An",. B. E. \\'anftI.)

- f- f- - f- - f- f- -
f-

f--

f--

(.J
F""re 4 Sc:hemalicho..lHl.lTlens)onafa~.sI~lralmll the thfi,rences bftwt,en, (a) the reguL.riY rq:>c"olt,~ slJUdute ~ a ~t:oI and {bl contrnuous randc.n neNoO: ~ a gbu. (After bchariaw:n )

..,
r,
(0)

"
0.0

"

-t
--'

,, ,

~ ,

"

-2.0

1 ,

f , ,

1 I

" "
00

-Ito

..

..
CLASSES

'"

16.0

Figure 5 uunpnr;SQn of experimental (dashed curve) and ."k;ubtcd (5ll1id curve) reduced ;01('1\sity function lOr amorphous t:crmanium. (a) Amorphous germanium compared with microcrYSb.llitc model. (h) Amor:l,l!ous germanium CQlIlIlared w,lh random nctwOI'\; model. (Results by J. Cra.'Z)k

and P. Chaudhan.)

A glass has the random structure of the liquid from which it is derived by cooling below the freezing point. without crystallization. Also, a glass has the elastic properties of an isotropic solid. By general agreement, we say that a liquid on being cooled becomes a glass when the viscosity equals 10 13 poise, where a poise is the CCS unit of viscosity.' This defines the gh~s transition tcmpcmtul'c Tg At temper-limes above T,:: we have a liquid; below TI: we have a glass. The transitkn is not a thermodynamic phase tmnsHion, only a transition for "practical purposes." Relatively few liquids can be cooled fast enough in the bulk to form a gla.~s before crystallization intervenes. Molecules of most substance'i have high
'The SI unit of viscosit) is 1 N$m- 2 so that I poise = 0.1 N~m-2. It i5 quite common to find I'ISCO$itie$ S'ven in CI' Of oentipoi$e. bemg \0-2 poise.

50.

enough mobility in the liquid so that on cooling a li<luid-solio melting lransilkm 13 OCC\lr'S a long time before the viscosity increases to 10 poise or lOIS cpo By contrast. liquid '\later has a viscosity 1.8 cp al the freezing point; the viscosity increases enonnously on freezing. We can often make a glass by depositing a jet of atoms of a substrate cooled to a low temperature. a process which will sometimes prOduce an amorphous layer with g1assJikc properties. Amorphous ribbons of some metal alloys may be produced in this way in indus_ trial quantities. TIle value 10 13 poise used to define TIC is arbitrary, but not unreasonable. If \\.'e bond a slab of glass 1 em thick to two plane parallel "ertical surfaces. the glass witi now perceptibly in one year under its own ",,-eight ""hen the ,'isco:sity

drops below 10 13 poise. (For comparison, the viscosity of the mantle of the earth is of the order of uP poise..)

Viscosity arid die lloppillg Rate

The viscosity of a liquid is related to the rate at which molecules undergo thermal rearrangement on a local scale. as by hopping into a \'acant neighbo.site or by interchange of two neighbor molecules. The physics of the transport process is somewhat different from that of viS(;Qsity in the gas phase. but the gas phase result gi\'es a qualitative luwer limit to the viscosity of the liquid phase, a limit that applies to nearest-neighbor hopping of atoms. The gas result (TP 14.34) is ([4) where 7j is the viscosity, p the density, (; the mean thermal velocity, and e the mean free path. In the liquid l is the order of magnihlde of the intennolecular separation a. With "typical" values p" 2 g cm- 3 , C - lOS em s-\ a = 5 X 10-8 em, we have ~mill) - 0.3 x 10- 2 poise'" 0.3 cp (15) as an estimate of the lower limit of the viscosity of a liquid. (rabies in chemical handbooks only rarely list values below this.) We give now a very simple model of the Viscosity of a liquid. In order to hop successfully, a molecule must sunllount the potential enerb'Y barrier presented by its neighbors in the liquid. The preceding estimate of the minimum viscosity applies when this barrier Illay be neglected. If the barrier is of height E, the molecule will have sufficient thermal energy to pass o\'er the barrier only a fraction
(16)

of the time. Ilere is an appropriate free energy and is called the acti\'atiOll

".

energy for the process that determines thc ratc of hopPin~. It is related to thc activation energy for self-diffusion treated in Chapter 18. lne viscosity will be increased as the probability of successful hopping is decreased. lous
"1/'" Tt,min)lf .... Tt,min) exp(EfkB'/l .

(17)

If "1/ = 10

13

poise at the glass tmnsition, thc order of magnitude off must be

= 0.3

lO- IS

(18)

at the transition. using (IS). The corresponding activation energy is

ElkfJ'r" = -lnf= In(3 x 101~) = 35.6.
13

(19)

If TIC"'"' 2000 K, then kBTg = 2.7 X 10- erg and E = 9.6 X 10- 12 erg6 eV. This is a high potential energy barrier. Classes with lower values of Tif. will have correspondingly lower valnes of E. (Activation energies obtained in this way arc often labelcd as E v ' Matcrials that are glass-fanners are chardCtcrized by activation energies of the order of 1 eV or morc; non-glass-fonners rna)' have activation energies of the order of 0.01 eV. When being pressed into molds or drawn into tubes. glass is used in a range of temperatures at which its Viscosity is 103 to 106 poises. The ,",,'Orking rauge for vitreous silica begins O\'er 2000"C, so high that the practical usefulness of the material is se\'erely limited. In common glass about 25 percent of N~O is added as a network modifier to Si0 2 in ordcr to reduce belo\v I(l()(fC the temperature needed to make the glass fluid enough for the forming operations needed to make electrie lamp bulbs, window glass, and boules.
lIC ')

AMORPnous FERROMACNETS

Amorphous metallic alloys are fonned by very rapid quenching (cooling) of a liquid alloy, commonly by directing a molten stream of the alloy onto the surface of a rapidly rotating drum. lnis process produces a continuous "meltspun" ribbon of amorpholls alloy in industrial quantities. FClTomagnetic amorphous alloys were dcveloped because amol'phous materials have nearly isotropic properties. and isotropic materials should have essentiall)' 'Zero magnetocrystalline anisotropy energy. As discussed in Chapter J5. the absence ofdirections ofhard and easy magnetization should result in low coercivities. low hysteresis losscs. and high permeabilities. Bee-.lOse amorphous alloys are also T'".mdom alloys, their electrical resistivity is high. All these properties ha\'e technological value for application as soft magnetic materials. The trade narne Metglas is attached to several of these. 'nlC tmnsition metal-metalloid (TM-M) alloys arc an important class of

magnetic amorphous al1o)'s. The transition metal componl'nt is usually about 8t1 perrent of Fe. Co. or Ni. with the mclallokl component B. C. Si, p. or AI. 11le presence of the metaJloids lo\\en the melting point. making it possible to quench the aJloy through the glass transition temperature rapklly enough to stahilize the amorphous phase. For example. the composition F~ ("-noo....'1 as Metglas 2605) has Til: = 441 c C. as comjl:lred with the melting temperature 1538'"C of pure iron_ The Curie temperature of this composition in the amorphous phase is 647 K. and the value of the magnl'tization M. at 300 K is 1257, compured with Tc = 1043 K and .\1. = 1707 for pure iron (fable 15.2). The cocrc-i\'ity is 0.04 C. and the maximum value of the pemleabiJity is 3 x lW. Coercivities as low as 0.006 G have been reported for anotller coln(Xl5ition. 2 lIigh coerch:ity materials can be produced by the 5,11Oe melt-spin pTOCe!>S if the spin rate or quench rate is dccrcdSCd to procJuce a fine-grdined c1y~talline phase. which may be of metastable compo~itiou. tfthe grdin Sil is arnmged to match the optimum size for single domains, the coeTCIvity can be quite high. J. L CrOat has reported lin = 7.5 kG for the metast.Jble alloy Ndo~Fe0.6at tile optimum melt-spin velocity 5 m 5- 1 111i5 material may come to hd\'e )lr.lcticaI applil'alion. Figure 6 ~hows the peak ill the coercivity of a Sm-Fe a110) as tile spin velocity is varied.

AMORI'IIOUS SEMICONDUCTORS

Amorphous semiconductors can be prepared as thin films by evaporation or sputtcring, or in SOUle materials as bulk glasses hy supercooling the melt. What happens to the electron energy b.\Ild m<X!el in a solid without regular crystalline OJ-der? The Bloch theorem is 1101 applicable when the stnlcture is not periodic, so that the electron states cannot be dcscrihed by well-defined k values. l1U1s. the momentum st,lecHon rtlle for optical tl'ansitions is rcla'\:ed; hence all infrared and Raman modes contribute to the absorption spectra. 1ne optil"d.! absorption edge is mther featurelcss. Allowed bands and encrgy gaps still ()C(:ur because the form of the dcnsity of states vs. eneTb'Y is determined mmt strongly by local electron bonding configurations. Both electrOns and holes can carry CUITCnt in an amorphous semiconductor. The carrien may be scattere<l strongly hy the disordered structure. SO that the mean free path may sometimes be of the order of the sc-d1e of tile disorder. Anderson PrOposed that the stales ncar band edges may be 1oca.Iized and do not exlend through the solid (Fig. 7). Conductic.n in these slates may take place by a thermally-assisted hopping process, for which the Iiall eITect is allOll'lalous

-nlf: "alues are Wen from 1m, 6Cellent reioYw by F

nels,In

E. P woh1&rlh, ed, Fcr,-om.ogMht

_,<,.v,", ,. I, NOlth-lIo:bnd, 1960

E. LuI-sky. - Am,"1>OOus fio.........,.".g-

.
"

,
I ,

00

I
~.(

......)

Figure 6 Coen""ty;at mom temperature n. melHopin ,'elocity v. ror SrT\o .Fee, ... 1'he ma>.imum
cocrri\ity is 24 lo..C and Ol'CUrs at 1.65 In s ',whil-h is bel~'1:d to OOfTeSpotId to $;n~t: d.onw.n bcl,,"'lor In ead:, tt)sl:llhte. At higtll'T 'illn r.ues 11M: ~rm.ily ~ bec.l\l5e the dcpm,t..-d matenal bo:comes amorphous (1l'QC lJOC.o~). At Iowo:r spin ralft the CI'}-st:I.lI,tes anneal to sizes

U\"e tho: single dollGln feI';"~ dow-.n

boundar'e5 gh~ a IoweY ~ty. (Arter J. L Croat.)

and cannol be used to dctenninc the carrier concentration. The quullhlffi Hall {'fred (Chapler 19) provides experimental evIdence for localized states and mobility edges. Amorphous materials appear to behave almost like intrinsic semiconductors, with the Fermi levcllying near to ,lie centcr of the gap. However, the pinning of the Fermi level is known to arise from the presence of defects. such as dangling bonds and other misfits in the structure. wMeh proc.luce localized st.ltes in the gap. The presence of these states, along with the tendency in covalent .utoys for local valence requirements to be sati56ed, ne<:e!>5itates the uSC of special techniques lOr preparation of doped specimens. T\\'o distinct classcs of amorphous semiconductors arc widely studied; tetrahedndly-bonded amorphous solids such as silicon and germanium, and the chalcogcnidc glasses. The lattcr are multkomponent solids of which one major constituent is a "chalcogen" clement-sulfur, selenium, or tellurium. The tetrahedr.uly-lxmdt..J materials have properties sinlilar to those of their crystalline forms, provided the dangling-bond defects arc compcllsatcJ

."

,
"igure 1 Oensoty c1 clcdron slates as ~1i.,.'C:d to

rogrs E,. ; sepante the rafIF of ~rg)

". ),Iott and E. A. fmiJi.)

non-lom}iLrd In 1I.... ~ttr of lhe

balwl.. LocaIi7.ed

amo..pho.... JOI...b....tift! stat.,. :ore are ~... s~. 'The mobllity band wt~ stal15 an 1ocnIu.ed and non-kn.Iaed. (After
Ol.'('\lf In

~"

\\idl hydrogen. l11cy can be doped with slOall amounts of chemical i.mpurities, and their oonductivitycan be sharply modifk.od b). injection offrce carners from a metallic contact. By contrast, the cbakogenidc glasses arc largel)' insensitive 10 chemical impurities and to free carrier injection. Amorphous hrdrob'enaled silicoll is a candidate material for solar cells. Amorphous silicon is a much Jess e,.-pensive material than single crystal silicon. Attempts at using pure amorphous silicon. hQ\\lCver, failed because ofstnlCtural defects (dangling bonds) which were impOSSible to eliminate. Introduction of hydrogen into amorphous silicon appcarli to remove the undesirable stniClure defects. Relatively large proportions of hydrogen are incorpor.'lted. of the orcler of 10 percent or more.
LOW ENERCY EXCITAllOl\'S IN
A~IORrHOUS

SOLIDS

111ere is a great regularity in the unusual thermal properties of amorphous materials at low temperalures, yet the experimcntal discovery of these properties wus entirely unexpected, completely differcnt from crystalline solids, and not well understood in terms of amorphous structure. The low lempemture heut cupaeity of pun: diclectric crystalline solids is knov.'O (Chapter 5) to follow the Oebye 'J<I law, precisely as expected from the excitation of long wavelength phonons. The same behavior was expected in glasses and other amorphous solids-thc point was SO obvious that it did not encourage experimental investigation. After eady indications by a Berkek')' group. Zeller and Poill discovered that many insulating glasses show an une';p<.'Cted lillcar term in the heat capac. ity below 1 K. Indeed, at 25 mK the obse""ed heat C'.tpacily of vitreous silica exeeeds the Oebye phonon contribution by a factor of 1000. Anomalous Iillcar terms of comparable magnitude are found in all, or nearl)' all, amorphous solids. 111eir presence is believed to be an intrinsic consequence of the amorphous states of malter, but the details of why this is so

"

remain unclear. \York by II unklinger and associates gives strong evidence thai the anomalous properties arise from twQ-level systems and not from multi-k'VcI oscillator srsh.'1ns~ in brief, the evidence is that the systems can be satumted by intense phonon fields, just as a two-le'"cl spin system can be saturated by an

intense rf magnetic field (problem 16.6). Ileal Capacity CalClllatiOlI

Consider an amorphous solid \\;Ih a concentration N of two-level systems at low energies; that is, wilh a level splitting Ii much less than the phonon Ocbre cutoff k n 6. The partition function of one system is, v.ith T"" knT, Z = CXp(tJ2T) + c>;P(-tJ2T) = 2 cmh(1iJ2T)
(20)

n.c thennal

a'"crage energy is
U= -

i.6

tanh(MT)

(21)

and the heat capacity of the single system is C v = kB<,dUIih) = k,,{l!J27)'l sech~MT)
(22)

111ese results are given in detail in Tr. PI>. 62-63. Now suppose that t:. is distributed with uniform probability in the range !:J. '" 0 to !:J. = !:J.o. 111e a\'eragc value of C v is

c" ~
=

(kef4,,!

{do dA (6'/l>ol=h'(AI.-.j

(~'T)

too 27 / l1x

(23)
x2

seeh2x .

11J.c integral cannot be carried out in closed form. Two limits arc of ~-pecial interest. For T <C .60. the SC(h2x term is roughly 1 from x = 0 to X = I, and roughly zero for x > L 111e value of the integral is rOughly U3. whence
(24)

for 'f < I:J.o/k B For., :J> !:J.o, the value of the integml is roughly ~(!:J.oIPJ..B'rr, so that in this limit (25) which approaches zero as T increases. Thus the interesting region is at low temperatures, for here the two-level system contriLutes to the heat capacity a term linear in the tClrlperature. This term, originally introduced by W. Marshall for dilute magnetic impurities in metals. has no connection with the usual conduction electron ~t C"dpacity which is also proportional to '(. The empiriC'dl result appears to be that all disordered solids have

'00

'"

10

I
C..... (SiOd-_/

"

002 0001

capacit) IS rooghly h~

FiI;ure 8 Ileal ClpaCiIJ' or \llreous suica ~nd loOOJo, sillCl KWs ~ ;II functIOn g( tcmptT.llure. nw: Ixa In T below I K. 11lc <bsho.-d line repr<'M:nb I~ calculated Drb)..., t-t

CI~lv of ,"jtIf1lUS ~llica_

N -

10 11 cm- 3 "new type"low energy excitaUons nniformly distributed in the energy interval from 0 10 I K. TIle anomaJotlS specific heat can IIOW be obtained from (24). For '( = 0.1 K and fldk n "" I K.
C\' =fNk 8 (O.I)'" lcrgcm- 3 K- 1

(26)

For comparison, the phonon COntribution at 0.1 K is, from (5.35),

C v ... 234Nk n(TIO)3 ... (234)(2,3 x 102.~(1.3R x 10- 16)(0.11300)3 - 2.8 X 10- 2 erg cm- 3 K I

(27)

much smaller than (26). The experimental results (Fig. 8) for vitreous Si~ arc represented by
C" = cIT

+ C3'(3

(28)

whcrec,

12ergg-

K- 1

andc3= 18ergg-' K-

Thermal Condllctil.:ity

The thermal conductivity of glasses is very low. It is limited dt room temperature and alx)\'c by the scale of the disorder of the slnlcturc. for this scale determines the mean free path of thc dominant thcrmal phonons. At low tCIW perature!>. below 1 K. the conductivity is carried by long wavelength phonons

J ';

'\'''''<:'1fIt(J//i,,~

SoIWh

I 'I
.~~
~pbrfl alom

T
I

9 Short phonon mean ~ path i,,:a dl..loOnle~ ~tructurl' A dlOtl ..a\l'letlltIh phollon Ih"l I.. .os lhmo n, ..,I1<I~ ,oIom R ~. mum mI:IIk-.. dlStanc.-e, bl'cauw 01 the ph;rw C3OC'l'lbtioo orthe lJpp"r..oo Irw....,.. pa~ fnlOn L 10 R. 1bl' d~Ol'nI d R 1$ t ... 1 - 0, so tlUII II", "'3Vl' incido;"t nom L is n,llectl'd .11 R.

and is limited by phonon SCiltlering from the mYSterious twole\e1 systems or tunneling states discussed earlier for their contribution to the h"at c<\pacity of amorphous solids. As in Chapter 5, the expres:.ion for the thenna! condUl.. th ity K has the (orm
(29)

where c is the heat capacity per unit volume, v is an average phonon vdocit),. and f is the phonon mean free path. For \'itreous silica at room temperature.
Ka:l.4xLO- 2 Jcm- l s IK
c'S!:

1.6 J em 3 K- 1 ;

(v)

'S!:

4.2 X lli" em 5- 1

11ms the mean free path f 'S!: 6 X lO-1j em; hy reference to Fig. 3 we see that this is of the order of magnitude of the disorder of the structure. Thi~ value of the phonon mean free path is remarkably short. At room tempel".tture and above (that is, above the Debye temperature) most of the phollons have half-\vaveIenbJf.hs of the order of the intemtomic spacing.. It is through phase cancellation processes, as in Fig. 9, that the mean free path is Iimitel! to several interJ.tomic spacings. No other structure for fused quartz will give a 6 A m{>'J.ll free path.. The normal modes of vibration of the glass structure are utterly unlike plane W:.lVCS. But the modes, as distorte<l J.S tliey lIlay seem, still have quan tized amplitudes and may properly be called phollons. The thermal COnductlv ity belov.' room temperature and p:.uticularly below 1 K is tre:.\ted in detail in the book edited by W. A. Phillips that is cited at the end of this chapter.
FIBER OPTICS

Fibers of silica-base<llightguides carr)' a high proportion of the data and infor.nation transmitted on the surface of the earth and under the ~dS. am.I are the medium of dlOice for the communications industry. The optical fibers consist of a thin core (- 10 p.nt) of high.refractive index glass surrounded by a cladding.

0.2

'1 L

__---.JL_ _.L:.'Ao"-_--'1.2
I~'

, , , , , , , , , ,

~-25 nJ.---j

L-'_.L_ _~c_-1.5 1.6

1.4

I.,

ml\'~kongtb (pm)

Figure 10 11le transmissiOIl chllnocterntic ~ ('l;lmmtlnotion-quality gpeical f1hen, showing the attenuation in units cI decibels per km as I function of the ..1n'ele.lglh <i light, in ,un _111.. 1b)'leigh lCattering tt'gfme is dominant on the left of the cun~. elICept lOr. strong impurity absorptioulille associated "''fth 011 iOlls that accomplll)' SiOr. the i> the _od harmonic ela I,tle.t 2.7 I'm known as the -w;lIler Iia,e. -n.e"1n'dength ,..,Wat 1.31 ...m is used in 1994 tnnsmlssloll lines; it Will be rrpIar:ed by the ..-a,-d."lgth 1.55 I'm ...-ailablo: from Eu"" iOIl laxt ampli!icn. ...hid! are used f:\"CI')' 100 Ian in Iypicalloo,g distance urJderwa applications. 1be PM'" needed 10 pump the amplifies iJ suwlied by copper wires. (Courtay cI nngye u, AT&T Bell Labonlories.)

Ii".,

TIre digital data are carried by light. with a minimum attenuation neat 0.20 db km- 1 at wavelengths ncar 1.55 #Lm, whidl is in the infrared (Figure lO). A range of 100 km corresponds to a loss of 20 db, which is readily supplied hy an E u 3-t- laser amplifier. TIle optic window of high-purity gl-tsses near this wavelength is limited on the low frequency side by phonon absorption bands and on the high frequency side by Rayleigh scattering and. ultimately, hy electronic absorption, In the optic window tile losses are determined hy the Hayleigh scattering intrinsic to static fluctuation.~ in the local dielectric constant ofan inhomogeneous medium, and the attenuation varies as the fourth lXlwer of the frequency. It is fortunate that an excellent source is available for mdiation at j.55 Mm. As shown in Figure 16.24, excited (pumped) erbium E~+ ions can amplify in an erhium-doped .section of fiber.

Rayleigll attenuation
TIle attenuation oflight W3.\CS in glass is dominated at short ....'<lvelengths in the infrared by the same scattering process, called Rayleigh scattering, that is responsihle for the blue light of the sky. TIle extinction coefficient, or attenuation coefficient, II has the dimension of reciprocal length and for light scattered in a

.
gas is given after l\dyleigh by
II = (2JJJ"I3m;"l\ry.(n -

If>.

(30)

where n is the local refractive imlcll: and N is the number of scattering centers per unit volume. 111C energy flux a~ a function of distance has the form exp(-llx). Derivations of (30) are found in good texts on electrodyn.'\ffiics; the structure of the result may be ulldcrstood by a general argument: lne radiant energy scattered from a dipole element Jl is proportional 10 (d,rldr'f. and this accounts for the actor w". The local polari7..ability a enters as el; if there are N
random scattering centers per unit volume, the scattered energy averaged O\'er these random sources will go asN(Aa'f), or (6.rl'f)/N. lous we have the essential factors that appear in (30). As applied to a glass, 6.n should refer to the variations in polari'l,.<'\tion around each group of Si-O bonds, and satislaetory nUlnerical estimates of the attenu..'\tioo may be made in this way.

Problem
I. Metallic Optic Filnu? In Nature 379, 124 (1996) it is speculated that metallic wires can act as optic fibfE$. transmitting light with a long delay appropriate to the high refractive index dlamcteristic of metals. Unfortunately the rcfmctive imlc" of a typical metal is dominalL-d by a frec-dcetron term in jill, w that tile propagation ofa ligllt Wll'o'e is ill fact Iligllly damped in a metal. Show tllat in sodiullI (p. 331) at room temperature a wave of vacuum w3vdellgtl. 10 ~nl will Ilave a damping lengtll of 0.1 ~m. This may be contrasted with the 100 km damping length found for Iigllt in high lluality gIa.ss fibres.

References
S. II. Elliott, Pilysia of (lmorl'llf)tJ$ mal.,rfab. 2nd cd., Wil,~ 1990. C. Keiser, Optiol f.ber comllu.micotk.>n.T, 211d .:<1.. McGr.lVo~l,m, t99l. II.-J. Clillthcrodt and II. lIeck, cds. Me'ollic g/&su, 2 \ds., Springer, 1960. 1963. K M. Man:b, Liquid meI~ conaptI ""d 'hmry, ~ , 1990. W. A_ Pbiftip5. ed. A..-pIwxa MJiids. low kJnperatun ~rtiu. Springer. 1981. K F. Mott ..1<1 E. A. [m1S. Electronic ~ in ~t<lalne matnials. OD:Jrd, 2nd ed. Ilrn J. Ziman. Modcb of doorthr, Cambridge, 1979.. C. o. JonO'. clau, 2nd ed. Chapman Mnd llaU, HJ7I. It zaIlen, PIU;sia of alllorplw14 wlidr, Wile)', 1983. F. Wooten alld D Weaire. "Modelillg let",h.i!m1ly !louded random network..!ly compUl"r," Solid ~blte phY$ics 40, 1 (1967). Journal of l\oll-CI)'Stallille Solids. C. Venialt.... D. Sahoo...1<1 V. Bab~ll.n. 8qoond 1M crystalline IIDtiE, Sprizlg<'r, 19b9. C. S_ Cagill. St,.,dlIre of mnDllic ol1<NJ f!Ja-, Solid ph~ 30, 227 (1975). Funuko l'0IM."U\Oo"ll, wCbsli transition ..1<1 relll..o..at.ion of disonle.-ed :s.trud.lITC'l.- Solid $tate pbysks 45. 179 (1991).

"*

II. MunI.. IIDndbvol<ofoptll:fJlf.benand~. Dc""". 1988.

18
Point Defects
LA'rrIC~

VACANCIES

541

DIFfUSION Metals
COLon CENTEHS f' centers Other center) in alkali halides I'ROBLEMS
I. Frenkel defects 2. Schottky vacancies 3. F center

544 547
548
5-18

548
551

551 551
552

nEFEHENCES

552

D
F"ogure I

I:J

A plane of I pure a1bJi halide crystal, ~howi"g I \_11 posil;'"e ion slle. I ,..cant nq;alj\'e ion sit... and I coupled iWt of \'lICant Slt~ of ~Ie s'gn_

CHAPTER

18:

POINT DEFECTS

The oommon point imperfections ill crystals are chemical impurities, vacant lattice sites, and extra atoms not in regular lattice positions. Linear imperfections are treated under dislocabons, Chapter 20. 1be crystal surface is a plmar imperfection, \\;th surface electron, phonon, and rnagnon slates, Chapter 19. Some important properties rJ crystals are controlled. as much by imperfections a>i by the nature of the host crystal, which may act only as a soh'Cnt or matrix or vehicle lOr the imperfections. cooc.luetivity of some semiconductors is due entirely to trace amounts of chemical impurities. The co'or and luminescence of many crystals arise from impurities or imperfections. Atomic diffusion may be accelerated enormously by impurities or imperfections. ~lechankal and plastic properties are usually oontrolled by imperfections.

n.e

UTIICE VACANCIES

The simplest imperfection is a lattice vaClInC)', which is a mlsslOK atom or ion, also known as a Schottky defl..'Ct. A L'"Ittil..'C vacancy is often indicated in illustrations and in chemical equations hy a square (Fig. I). We create a Schottky defect in a perfect crystal by transferring an atom from a lattice site in the interior to a lattice site on the surface of the crystal. In thermal equilibrium a <:ertain number orJattiee vacancies are always present in an otherwise perfect crystal, because the entropy is increased by the presence of disorder in the structure. In metals with close-packed structures the proportion oriaUice sitcs vacant at tempcrflturcsjust below the melting point is of'the order of10- 3 to 10- 4 . But in some alloys, in particular the very IlarU tmnsition metal carbides such a.~ TiC, thc proportion of vacant sites of one component can be as high a.~ 50 percent. TIle probability that a given site is vacant is proporlional to the Boltzmann factor for thermal equilibrium; P = exp(-E\Jk n1'), wherc E v is tile energy rC<luired to take arl atom fmm a lattice site inside the crystal to a lattice site on the surface. If thcre are N atoms, the eqUilibrium numher n of vacancies is given hy the BoIt7J1lann factor (I) If n <C N, thcn
nfN "" exp(-E\Ik R 11 .
(2)

If E v - 1 eV and 1' ... 1000 K, then nlN _ e- I2

10-.5.

..,

....
'"
The equilibrium concentration of vacancies decreases as the tempernture decreases. The actual concentration of vacancies will he higher dIan the et:luiJihl'iulll value if the crystal is gro~vn at an elevateU temperature and then cooleJ suddenly, thereby freezing in the vacancies (sec the discussion of dilfusion

below).
In ionic crystals it is Ilsually favorable energetically to form roughly equal numbers of positive and negative iou vacancies. The formntion of p.'lirs of vacancies keeps the crystal e!ectrostatk\llIy neutml on a local scale. From a st.ltistical calcul.\tion we obtain (3)

for the numher of pairs, where E" is the energy of fomlulion of a pair. Another vacancy defect is the Frenkel defect (Fig. 2) in which all atom is transferred from lattice site to an inlcr:stitial position, a position not normally occupied hy an atom. The calculation of the equilibrium numher of Frenkel defects proceeds along the lines of Problem 1. If the number 11 of Frenkel defects is much smaller than the number ofL'\ttice sites N ,md the number of interstitial sites N', the result is

,I

(4)
where E, is the enerb'Y necessary to remove an atom from a lattice site to an interstitial position. In pure alkali halides the most common lattice vacancies are Schottky defects; in pure silver halides the most common vacru.lcies are Frenkel defects. Lattice vacancies are present in alkali halides when these (:ontain allditions of divalent elements. If a crystal of KCI is grown With controlled amounts of CaCI 2 , the density varies as if a K+ lattice vacancy were fomled for each Cn2 + ion in the crystal. The Ca2 + enters the lattice in a normal K ~ site nnd the two CI- ions enter two Cl- sites in the KCI crystal (Fig. 3). Demands of charge neutrality result in a vacant metal on site. The experimental results show that the addition ofCaCl z to KCllowerS the density of the crystal. The density would increase if no vacancies were produced, hecause Ca 2 + is a heavier and smaller iOIl than K+. The mechanism of electrical conductivity in alkali and silver halide crystals is usually the motion of ions and not the motion of electrons. This has hecn est-dhlished hy comparing the transport of eharge with the transport of mass as mcnsured hy the material plated out on electrodes in contnct with the crystal. The study of ionic conductivity is an importnnt tool in the investigation of lattice defects. Work on alkali and silver lwlides containing known additions of divalent metal ions shows that at not too high tempemtures the ionicconductivIty is directly proportional to the amount of divalent addition. This is Ilot heL'ause the divalent ions are intrinsically highly mohile, for it is predominantly the monovalent metal ion which deposits at the cathode. TIle lattice vacancies introduced with the divalent ions are responsible for til'" enhanced diffusion

...gure I
Ii

Schottky and Fll'nkd det'e.:b: in iIIl ionOcn)'St:l1. n.e......,..-s indoaate the d ~ t g the >OO1S. In a Schotlk) defed the ion end. up on the ~ur&ce ..... the aylilal. in FrfllI<eI ddect 'I
n:1I1O'I'ed to .,\ inlentilial ~hQn..

I-ig~

3 ProdoctioIl of II buice \'2C:lIICY by the soI"tioll of

iI;

t\('utlll1ily 1I positive lOll \'lIoCllnCy

;n KCl: to f:nSun: e~ inlrod,lC'I'd inlO the L.Uioo ",Ih .:ach di\':llt'tlt cation Ca' . TIoe

eaa.

two CI

ions d CaCl a cnter normal neg:llive ion si\t'S.

0000000 0000000 0000000

OOsgOOO OOQ\)OOO 00 000 0000000 O~OClOOO OOOOOGO 0000000

0000000 0000000
or

0000000 0000000

figure" Three basi<: mechanisms diffusion: (II) 11\1('rchallge by rollll'oll .lx",ta midpoint. ~lore thm, two alOlllS may retate tog..thcr. lb) Miwatioll through ;nleutitilll siles. (c) AtOll1$ ('.<change position "ith vlIcant lattice site. (After Seill-)

.,

0000000 0000000 0000000 000-000 0000000 0000000 0000000

'<'

(Fig. 4c). The diffusion of a vacancy in one direction is equivalent to the diffusion of an 3tom in ti,e opposite direction. When lattice defects arc generated thermally. their energy of formation gives an extra contribution to the heat C<lpacity of the cry.>tal, as shown in Fig. 5. An associated pair of vacancies of opposite sign e'l:hibits an electric dipole moment, \\;th contributions to the dielectric constant and dielectric loss due to the motion of pai"" of vacancies. The dielectric relaxation time is a measure of

"
" , '" j" , '" " """

" >

I I I I
-<~~~~~~~._------~ - - - - c. _ 12 ~ O.OO5T I

I I I
I

<00

I'igurc S lleat ClIpKity ofsih-er bromide elhibilillg the e~ ""at ClipllCily from the fonTlllOOIl1 c:f
lattice deree:u. (After R. W. Christy a"d A. W. LawSOI\.)

the lime required for one of the vacant sites to jump by one atomic position with respect to the other. 1be dipole moment can change at low frequencies, but not at high. In sodium chloride the relaution frequCl'lC) is 1000 5- 1 at

850C.
DiFFUSION

\Vhen there is a roncentration gradient of impurity atoms or vacancies in a solid. there will he a flux of these through the solid. In eqUilibrium the impUrities or vacancies will he distributed uniformly. TIle net flux IN of atoms of Olle species in a solid is related 10 the gradient of the concentratiOfl N of this species by a phenomenological relation called Fk'k's law:

IN= -DgmdN

(5)

Here IN is the number of atoms crossing unit area in unit time; the constant D is the diffusion <.'Omtant or diffusivity and ha.~ the units cm'l/s or 08s. The minus sign means that dilfusion occurs away from regions of high concentration. The form (5) of the law of diffusion is often adequate. hut rigorously the gradient of the chemical polCfltial is the driving force for diffusion and not the concentration gradiCflt alone (fP. p. 406). TIle diffUsion constant is often found to vary with temperature as D = Do exp(-ElkB l1 ;

(6)

here is the acti"ation energy for the process. Experimental results on the diffusion of carbon in alpha iron are shown in Fig. 6. TIle data are represented by = 0.87 e\'. Do = 0.020 cm''1s. Representath:e ,'al.ues of Do and E are given in Table 1.

/'

"
I-

"

"v
"I

/
,
3.0 lUi 2.2

3.8 :1.4

,""""

1.8 I 4 10 0.6

Figure 6 Diffusioo, ~nt of carb<lo, in iroll, dter Wert. 11le Ioprilhm dO is directly pt'OplWtional to Jrf.

Table 1

Diffusion conslants and aclivalkm energiel;

"'....
Cu
Cu

llosl

A<~

cm 2. - '

V.

.v
~l>I

""" "', ...

Si

Alom

~~

Do

.v
3.47

Cu
7~

0.20 0."
0040

AI

A. A. A. A.
U

Ag

CU Au
Pb

N.

N.
U

2 X 10-3

1.2 0.26 0.22 0.24

1.98 1.91 2.00 1.98 1.65 0.45 1.20

51 5' 5' 51 5'

Si

G.
I" As Sb
U

8.0 3. 16.0 0.32 5.

2 X 10- 3

:i51 3.90 3.56 3.94 0.66

1.J3

Gu

Au Ge

I x 10- 3 10.0

3.1

To diffuse. an atom must overoome the potential energy harrier pre5enled hy its nearest neighbors. We treat the diffusion of impurity atoms between interstitial sites. The same argument will apply to the diffusion ofyaamt lattice sites. If the harrier is of height E, the atom will have sufficient thermal ene'rgy to pass over the harrier a fraction exp(-ElkBT) of the time. Quantum tunneling through the barrier is another possible process, hut is usually important only for the lightest nuclei. particularly hydrogen.

....
546

If" i~ a characteristic atomic vibrational frequency, then the probability p Ihal sometime during unit time the atom will have enough thermal energy to pass over the barrier is p""
JJ

cxp(-ElkBT) .

In UIl!! lime the alom makes I' passes al the barrier, with a probability e\I"i.-Elk1J T) of surmounting the rnmicr 011 each tiY. TIle qUJ.ntity J1 is called tile jump frequency. We l'Qllsider Iwo pamllcl planes ofimpurit)' atoms in interstitial sites. The planes are separated by lattice constant u. There are S impurity atoms on one plane and (5 + (I dSfcl.r) on the other. The net number of aloms crossing between the planes in unit lime is - -pa dSfdx. IfN is the total conccntration of impurity atoms, then S "" aN per uni! area of a plane. The diffusion nux may now be wrillcn as
iN -' Jll2

(dNldx)

(8)

On comparison with (5) wc have the rc!Sult

D::

1/(1

exp{-Elk n1')

(9)

of the form (6) with Do :: 1/(12 If the implJrities are char~oo, we may find thc ionic mobility jl and the conductivity u from thc diffusivity by using the Einstein relation 1.n1/l :: lJD from 1'P, p. 406:

jl :: (lJ1/(1 211.n11 exp(-EI1. nl') ; u:: Nqjl ::

(10)

(Nllva 2Ik

aT) exp(-ElknT)

(II)

whcre N is the concentnltion of impurity ions of charge l/. The proportion of vaC'.lncies is independent of temperature in thc range in which the number of vacancies is dctermined by the ,wmber of divalent metal ions. Then the slope of a plot of In 0' versus 1/1.n1' gives E... , the barrier activation cncrgy for thc jumping of positive ion vacancics (Tablc 2). Diffusion is very slow at low temperatures. At room temperature the jump frequcncy i!S of the order of 1 s- J, and at 100 K it is of the order of 1O-z:; S-l , '1be proportion of vacancies in the temperature r..mge in which the l,:onccntration of defects is detcrmined by thermal generation is givcn by (12)
\~hcre Efis thc encrgy offormation of a vacancy pair, aet.:oroing to the theory of

Schottky or Frenkel deft..'Cts. Ilere the slope ofa plot of In u versus Il1.nT will be E... + iEf , aet.:ording to (10) and (12). From measurements in differcnt tem-

18 Point DeJ"t. Table 2 AethOlion energy E. for motion of positi.e ion \Deaney

5'1

II

VaJues ohbe energy oriOrmation of:l

\~ey pair,

given in parentheses for the silver salts lefer

.(eV)

to

intentitW siher ions.

Worken

,. lire also

given. TIle numbrn

NaCI Lit'" LiCI

LiBr

0.86 0.65 0.41 0.31

Lit

0.38
0.89 0.39(0.10) 0.25(0 II)

....'

KCI

AgCl

2.02 2.68 2.12 1.60 1".3-1 2.1-2.4 1.4"

1.1"

Etzel and ~lllurer

J la\'en
Haven
Ib\'Cn

1l,\\'cn Wagner; Kelling and Witt Tcltow Compton

perature rnnges we detennine the ellergy of fonnation of a \'acancy p.1ir EJ and the jump activation energy E+. 'I11e diffusion (.'On~tlllt mil be me:.lsured by radiO;lctive tracer techniques. The diffusion of a known initial distribution of radiooctive iOlls is followed as a function of time or distance. Values of the diffusion constant thus determined may be compared wit II values from ionic conductivities. TIle two sets of values do not usuall) agree within the experimental accurac)', suggesting the presence d a diffusion mechanism that does not involve the transport of charge. Fur example. the diffusion of pain: of positive and negative ion vacancies does not involve the transport of charge,

Metals
Selfdiffusion in monatomic metals most commonly proceeds by laUi(.'C vamncies. Self-diITl,.lsion means the diffusion of atoms of the metal itself, and not of impurities. The activation enel"g)' for self-diffusion in copper is expected to be in the range 2.4 to 2.7 eV foc diffusion through vacancies and 5.1 to 6.4 eV for diffusion through interstitial sites. Observed values of the activation energy are 1.7 to 2.1 eV. Activation energies for diffusion in Li and Na can be determined from measurements d the temperature dependence of the nuclear resonance line ,,,kith. As discussed in Chapter 16. the resonance line width n~rrows when the jump frequency ofan atom between sites 1>ecQmes rapid in comparison with the frequency corresponding to the static line width. The values 0.57 eV and 0.'15 e V were determined by Nto.'IR for Li and Na. Self-c,liffusioll mC'.lsmemellts for sodium also give 0.4 eV.

COLOR CE"'TERS

Pure alkali halide crystals are tmnsptrent throughout the visible region d the spectrum. '111e crystals may be colored. in a number of ways; .. by the introduction of chemica'! impurities; .. by the introduction of an excess d'the metal ion (we may heat the crystal in the vapor of the alkali metal and thell cool it quickly-an Nael crystal heated in the presence of sodium vapor ba'Omes yellow; a KCI crystal heated in potassium vapor bC<.'Ol1les magenta); by x-nly, ')'-my, neutron, and electron bombardment; and by electrolysis. A colo.. center is a lattice dek."Ct that absorbs visible light. An ordinary lattice vacancy dOL'S not color alkali halide crystals, although it af fects the absorption in the ultraviolet.
F CtJ'lten

l1tc name F center- comes from the Cerman word for- color, FUf"be. We usually produce F centers by heating the crystal in excess alkali vapor or by x-il'Tal!iation. 'Ibc central absorption band (F baOO) associated with F centers in several alkali halides are shown in Fig. 7, and the quantum energies are listed in Table 3. Experimental properties of F centers have been investigated in detail, originally by Poill. The F center has been identified by electron spin resonance as an electron bound at a n~tive ion vacancy (Fig. 8), in agreement with a model suggested by de Boer. When excess alkali atoms are added to an alkali halide cryst:J.l, a corresponding number of negativc ion vacancies arc created. The valence electron of the alkali atom is not bound to the atom; the electron migt:.ltes in the crystal and becomes oound to a vacant negative ion site. A negative ion vacancy in a perfect periodic lattice has the effect of an isolated po:.itive charge; it attmcts and binds an electron. We t.'31l simulate the electrostatic effect of a negative ion vaC'"..Incy by adding a positive charge q to the normal charge -(I of an occupied nC@ltive ion site. The ,.. ccnter is the simplest lr3pped-e1cctron center in alkali halide crystals. The opticaJ absorption of an F l'Cnter arises from an electric dipole transition to a bound excited state of the center.

Other Grnten in Alkali lIalida

In the FA center one of the six nearest neighbors of an F centcr has been repbced by a different alkali ion. Fig. 9. More complex trapped-electron centers are formed by groups of F centers, Fig. IO and 11. Thus two adjacent F centers form an AI center. Three adjacent F centers fonn an R eent!'r. Different centers are distinguished by their optical absorption frequencies.

18 I'oinl Chf"

S49

Table 3

xperimental J." center absorption energies. in eV

Liel
N,Q

Ka

m.c,
LiBr

C,Q

3.1 2.7 2.2 2.0 20 27

NaB.. KBr

2.3 2.0

RbOr
LiF Nat"

..8
5.0 3. 2.7

KF

"'"'
UO

...,
NoCI

W.,,,lcoglh In

..., ,
'0

J.

...,

"'"'
I ,

..., ,
c.a

"'"'

"'"

II '\ , ,

'.

,'\

Eengy ... eV

I , ,

, ,

Figure 7 Tho: F bands for ,",'Veral alkali halides: optical absorption ,...,rsus wavelength for crystals that contain F centers.

di~lrihutionoftbc CX$$ c1,~ron

Figure S A.n F cenh.'f is a negali"e I0Il '''''''''Y wilb one ~ss derlll)n bound at tbe:: .-acancy. 1bc is largely on ulr po.;iti,..., metal ions adjactmt to the vacant lattice site.

080 08 080

S50

o
.
Fi!ure 9 An F,. <.'f:nlcr in KCI: one of the sU: K' ionl; "hKh hind an F ~h.'r is rrp1;oca) b.. another a1b1i ....,

ocrt' :-..'

Figure 10 An M CfflIt'Honsists of"''O .djatttli

F centen.

o
00
"'lgure II An R ....,Ilter consists of thn'e .ulj3<'Cnl F ccnters: thai is, a group of three negative iOIl VI.'llncies ,n " [1I1] plane of the NaCI slmeture, "ilh three M!iOCiat~-d d<...:lrOl1li.

Holes lIlay be trapped to fonn color centers, but hole centers arc not usually as simple as elCclroll centers. For example. a hole ill the fllied 1>6 shell of a halogen ion leaves the ion in a /15 configuration. whereas an electron added to the filled. p6 shell of an alkali ion leaves the ion in a ,l'6 (.ollfiguration. The chemistry of the two centers is different: '/'8 acts as a sphericall)' symmetric ion, but acts as an as) mmctric ion and. b)' virtue of the JahllTdlcr effect. will distort its immediate surroundings in the crystal.

,r

"
00 0

Poil1llN{cm

@
a

,.
n

0
n-

,.
0

,
a

cg--~' '"
a

~V"""nl"r

tigu", 12 A VI: <'enter limned wl'en ;I. hole iJ tropp._'Cl L~' a pair of Ilc!'2tl\'c .ems r.,scmbkl II. IlC:'pth-e Mlogen muk.'CtJk-ion, which i. 0. ill KCI. No b.ltia: ''X'ancics 01" ~II":I .tOUIS are in."Ohed in a VIC centet". n:n!('r 1II the 1cfi of the Iigu~ proIabIy is not stx.k: lhe hn::agun rcpre..:nts a hole tnl'P":d ncar. (lOIdt"'e ion '~-y. SU( h. reuteT would be the antimocl'h to an F C<.'Jltcr. lIoies h.,'c a lower .,ncrgy tr.r.pp! ill a VIC <'Cntcr than in an anli-'; "t, r

n.e

"l'he antimorph to the F center is a hole trapped at a positive ion vacancy, but no such center has been identified cxperimcnt.,lIy in alkali halides; in insulating oxides the 0- (called V~) dcfcd is known. The best-known trapped. hole center is the VIC center. Fig. 12. The V"centcr is fanned when a hole is troppec.1 by a halogen ion in an 31kali halide cryslal. Electron spin resonance shows that the center is like a IlCWltivc halogen molecular ion, slich as eli in Kd. The }abn-Teller trapping of free holes is the most effective fonn of selftrapping of charge carriers in perfect crystals.

Problems
L

,..nmJ,;d deled.

Show Ihat the nllmber n of intel"Stiti.:l.1 atoms in eql,ilibnllm witb " lattice vaeancin in a crystal hil\ing:>o' Llll~ points ;U1d N' poss.ble intelSlili:llllOSilions iJ given I))' tile C<luation

, ~ keT In {(\' - nXN' - 1I)/,,2J whence, for n ce N, N', we h:l\'e n

ill (.\'N,)12 c~p(-,f2keT). Here E, is the cllergy necessary to rt>mIDe an alom from a lattice site to an intentilial position.

2. SchOll"" ,:conde. Suppose Ihat the e11crgy reql1irt.'<1 to remo\'e a sodium atom from til(' inside ofa sodium C'l'}st.u to the boundaly is 1 eV, G.lkulate the conCClltralioll of Schottky vacancies al 300 K.

~-

-,
3. f' anler. (a) Treat all F center as a fN."e elL>ctlUlI of mus III mo' ., in the l'icld of a polu! dlargc e in a medium of dick't.'tric constant of. - 'II; wllat is the 1s-211 cllergy
difTcr<~lI<''e of ,: l'Cnters in NaCI? (Ill Compare from TaMe 3 the 1; mlcr excitation energy in NaCI with the 3s-3r <''1lC.'"rgy difference of the free sodium atom.

References
C. P. F1)nn, POINt

Ikfcct. 1I1td diff""", o.bd, 1972W B '-owln, cd.. Pho,piQ of mOT CC't"l'n. Amdemk f'rQ$. 1968. N. B. 1Iar....,. Sclldllo,c chtlttUt'1l. Plenum. 1ll73-1976. .... A. lIu"lIc,.. ed., fAlfia Ik/u In _lc.-f..cton, Insliloft of 1'h)'Da, l.ondo;. I97'S. J. J M.. kharn, F .:.olllen in lllkaJi h.l(Kk ~k Ptns. 1966. A. S. "owid< and J. J. Burloo. eds., 'Diffuaon in ~ilb-r<rall' d~opmenn, Aealemlc J'Tess,
1975.

K K. tu:oo.JC. Impuril'l $JWC11lI of soI.dJ. l'k:num. 1970J II. Schulma" and W. D. Complon, Color C4''''en ill .club, Pergamon. 1962.
A. M Stollt'ham, Uwory of defect, ill wIld" <hford. 1!1i5. W. Hares and A. M. Stoneham. Dtfecll /Iud .Iejeet p~IIU In "llfllnlallic rolidl. Wiley. 1005. H Balian, M. Klemall, and j.-I'. I'oirier, e<ls., Physia vf dcf('(;ts. North-HQlland, 1981.

"I. Lannoo luld J. Bourg<>in, PoIIiI deft'Cl$ i~ ","",ieund"cron. 2 '015., Springer. 1981-83.
Y. "'arge R.1d M .....o nlana. ElcdnJrlk lind vibrnlio"al ,'ro~rlfl:s of lminl defrels i" lmlle cryslllls,

!':orlh-1l01land,I919. rkfcctr in ""1<11, f',orth-lIoiland. It. series. C~1I1 ,",Uice dck."C'ls. It. journal

19
Stlrface and Interface Physics
Reconstruction and relaxatm 55S

SURFACE CIWSTALU>CRAPIIY
RCnCClion high-energy electron (liffraction

5SG
560

SUlU'ACE l:LECTRONIC STRUCTURE work function Thenrriooic:: emission Surface slates Tangential surface transport
MACNETORESISTAJliCE IN A nVO-DlMENSlONAL CIIA",,,,,EL Integral Quanti"LCd Hall effect (IQUE) IQUE in real s)'Itcms Fractional quantized Hall effcct (FQUE)
p-n JUNCTIONS Rcctiflcation Solar cells and photo"ohaic detccton: SchoUI')' barrier

500
500 5G I

562 5G4
5G4 SOO 568 570
570 572 0573 573

H .......nmOSTltucrU)\ES noN hetcrojunction

SE~UCONDUcrORLASERS
UCJlT-EMm"L~C

574 576
577

DIODES

579
580

SCANNlflOC TUNNl:o:L1NC MICROSCOPY

PR08U:MS

561

I. Diffraction from a linear arra), and RJU8rc arrllY 2. Surface subbllnds in electric quantum limit 3. Shuhnikov-dc HaaJi effect ill II surface channel REFERENCES

581 583 583 554

till II
So>Ioo:.o
L.t~.

I,'igu..... 1 Oangling Ix"nds from the (III) s\triaC(' of a "O\'3lently oonuL'<! d,amond cubic structure. (After \f PruHon, Surface "hyde Clarendon, 1975.)

""

CUAPTER

19:

sum'-ACE Ai'iO INTElWACE Pl-I\SICS

RecOllstruction lind Relaxation

The surface of a clystalline solid in vacuum is generally defined as the few, approximately three, outermost atomic layers of the solid that differ si~nifi cantly from the bulk. The surface may be entirely deJ.n or it llIay have foreign atomS deposite.:1 on it or incorporated in it. The bulk of tile cr)'st<\lls called the substrate. If the surface is clean the top layer may be either reconstructed or, sometimes, unreconstructed. In unrcconstrudt."C1 surfaces the atomic arr.mgement is in ref!:istry with that of the bulk except for an interlll}'er spacinf!: change (c:.ll1ed multilayer rela."ation) at the top surface. TIle shrinking of the intcrla~'er distance between the first and second layer of atoms with respect to subscquf'lIt layers in tlIP bulk is a mtller dominant phenomenon. The surface may be thought of as all intermediate between the diatomic molecule and the bulk stl ucture. Bt."Cause the interatOIl,ic distal lees in diatomic molecules are milch \lIlaller than in the bulk, there is ,I ratiunale for the surface relaxation. This may be contrasted witb reconstruction where the rela."Gltion of atoms yields new surface primitive cells, III relaxation the ,ltoms maintain their structure in the surface plane as it \\11S accordin~ to the projection ofthe bulk cell on the surface; only their distance from the hulk ch,lIlges. Experimental V1.1lues of the top-layer relaxation for metals are given in Table I. Note the frequency of occurrence of contmctions in the interlayel' spadng at the surface. Sometimes in metals, but most often in nonmetals, the atoms in the surface layer form superstluctures in which the atoms in the layer arc not in registry with the atoms in l'Orresponding layers ill the substrate. Surf."ce reconstruction mn be a consequence of a rearrangement of broJ...en covalent or ionic bonds at the surface, Under such conditions the atoms at the surface hUllch into rows with alternately larger and smaller spacings than in the hulk. That is, for some crystab held together by valence bonds, creation of a sUlf.lce v.ol1ld leave unsaturated bonds dangling into space (Fig. 1). The ener~ may then be lowered if neighboring atoms approach L"ach other and form bonds with their othelwise unused valence electrons. Atomic displ.lccments l':ln be as large as 0.5 A. Reconstruction does not necessarily re(Juire fonn<\tion of a SIJpcr~tructure. For example, on CaAs (ilO) surlaccs a Ca-As bond rolation (x:curs that Ie-J.ves the point group int.lct. The drivin~ force is electron tmmfcr from Cd to As, which fills the danglin~ bonds on As and depletes them on Ca. Surlaces of planes nominally of high indices Illay be bUilt III' of low il,dex

Table J

Rcl;l\.ation or topmmt interlarer slla(:ing at unrCCQnsh ~.Ied dean metal surrac:c5

Each (>fItry ronsists of: Chemical ~ mho! (iuterlayer spacing dl3.nge rroIU hulk value). (Aftff F. Jona and P. M. Mart"tlS, ill 111e Structure or Surf3l."fi II. - (4. J. F.....an der
\'een anti ~1. :\. Van 1Il)\.'C, Springer-Verlag (Jldllclbt>rg. 196b), p. 90J

hql(OOOI) rcqlll) b(1I0) rcr(IOO) b(IOO) rcqllO)

1lCfl{lo--lO)
oc-q211) rfC(311) b(31O) rcc(33l) rcx(21O) bcr(1I1) lxT(210)

JU. (-5'1), Se (-2<k), n (-2<';), Zr(-IIl) AI (+l'.l), Ag (O'.l), CII (-O.7'l), I't (+I'A:), Rh (O'k) r(" (+0.5%), Na (0'1), V (-0.3'k), W (0'1) Al (roc). CII (-J'l:), Rh (lY.C) Fe (-5'4), Mo (-9.5'4). Ta (-U'l:). \' (-7Ok), W (-S'.t) Al (-B.5'.l), Ag (-6'l:), Cu (-B.5St), l'i (-8.5%), Ph (-16'.1:), HI. (-3'1:) Re (-17%) Fe (-10%), W (-IN) Al (-13<;{), l'i (-16'l:), C.. (-5'.\) Ff'{-IW,) AI (-1:2.'X) AI (-J5.5'1-)
Fe (-179:)
Fe (-U%)

planes separated by steps olle (or two) atoms in height. Such terrace-step arr.mgements are importalll in evaporation and desorptioll because the attachmellt enel"~ of atoms is often low at the steps anJ al kinks in tIle steps. The chemiml acth-ity of such sites may be high. The presence of periodic arrays of steps llIay he detected by Jouble ;\IId triple bc.. ..u ns of JilTraetion in LEED (sC() below) cxperimcnts.
SUHFACE CHYSTAUnCnAl'llY

The stnfa<'c structure i~ ill gelleral diperiodie. This Joes 1I0t necessarily mean thilt all its atoms lie ill a plane, but rath~r that the structure is pcliodic only ill two dimensiulls. 111~ surface structure call be the structure of foreign material deposited on the substrate or it can be the selvage of the pure substrate. In Cilapier I we IISed the term Hra\'ais lattice for the array of equivak'11t points in two or in tlln,:c Jim~nsiollS, that IS, for Jipcriodk or tripcriodic structures. In the physics of surfaces it is common 10 speak of a two-dimensional lattice. Further, the area unit may be callt:d a mesh. We SllOWcd in Fig. 1.9 four of the five nets possible for a dipcriodic stnlC-

J9

SurftlC and InterftJ! Phyri

557

ture; the fifth net i" ,e general oblique net, \\ilh no special symmetry relation betwcen the mesh basis vectors :I" ~. Thus the five distinct nets are the oblique. square, hexagonal, rectangular, and centered rcctanl:,'l.i1ar. The substrate net parallel to the surface is used as thc reference net for the description of the surf.'lce. For example, if the surface of:l cubic substrate crystal is the (Ill) surface, the substrate net is hexagonal (Fig. 1.9b). and the surface net is refern.>d to these axes. lbe vectors CI. ~ that define the mesh of the surface stmetl.lre may be eq>ressl.>d in tenns of the reference net a" a2. b)' a matm operation P:

(O')=p("\=(P" p,,\(a,\.
\c.:z
P.)
P21

prJ

a.)

(1)

Provided that the included angles of the two meshes are equal. the shorthand notatiOn due to E. A. \Vood may be used. In this notation. which is widely uSl.od. the relation of the mesh CI, ~ to the reference mesl! 010 82 is expressed as

(~X~)Ra. 0, 0,

(2)

in terms of the lenl:,'1.hs of the mesh basis ,ectors and the angle a of relative rotation R of the two meshcs. If a:: 0, the angle is omittcd. Examples ofthc Wood notation are given in Fig. 2. The n..'Ciprocal net veeton of the surf..1ce mesh may be writkn as cf, c;, defined by
(3)

llere thc 2-rr (or J) indicates that two conventions are in use. The ddlnitions (3) used in Fig. 3 may be compared with the definitions (2.10) and (2.1 J) for the rcciprocallattice vectors of a triperiodic lattice. Thc reciprocal net points ofa lliperiodic net may be thought of-when we arc in three dimensions-as rods. The rods are infinite in extent and norma] to the surface plane, where they pass through thc reciprocal net points. It rna)' be helpflll to think of thc rods as gcnerated by a triperiodic bttice v.nich is C)[pan(kd \\ithout limit along one of its a>.:cs. Then the reciprocal lattice points along this axis are moved closer together and ill the limit fonn a rod. 1be usefulness of the rod concept comes out with the Ewald sphere construction explained in Fig. 2.8. Diffraction oC(:urs evcrywhere the Ewald sphere intcrcepts a Tl:,'Ciprocalllet rod. Each dilTrncted hcam is labdletl with the indices Ilk of the reciprocal net vector

(4)
fonning the beam. Low energy electron diffraction (LEE D) is illustrated by Fig. 4. The lee-

['(I x I)

p(2 x 2J (\13 x V3)II3!r r<'C(IIII. IIq-.(OOOII

[>(1 x II

.-(2 x 2J
!"t100). Ix~{IOO)

1'(1! x 21

'"

1'1

['(2

p(1! x II kc(110) (.1)

,{I:

I!)

"'-'1101

Fi",'un: 2 Surfa<.'C nets of adsorhed lItoms. TI,e drdes rep...;senl 1Iioms in the luI' IlI}'cr of the sll"'lrdte. I" (a) the <ksig,,~tion fcc(lll) means Ihe llll) face of ~n fcc- slructure. This fu'C Jelermioes 1I refcrcnce ncl. n,e lines represcnt orocred o\'crlll)'Crs, \\iI 1, ad~toms ~t the ioterscdkll<' of two lines. 1ne intersection poiols represent diperio<lit' nels Olltlices ill twu dimellSlOns). n,.: Jcsignlliion p{1 x I) in (~) is 1I primili\'e mesh m,i! for "hid, the blIsis is i<!enlical with the Wsis orthe reference net. In (1. tl,e <:(2 )( 2) mesl, unit is" ~'C"lered mesh wilh blIsis ,"edors twice as long as t1lO'iC of Ihe reference lIet. Alomic adsorption on metals takes pLtce most often into t111).SC surfal-e sites (hollow sites) that maximize the number of nwrest-neighbor ~toms un the suhstrale_ (AfI~'I" \'~n 1I0\e.)

I,

L'-7/" .... ,,': II

t l="~' ------,,

8'110 , 1/1' n...... ,," /<'I 1_ I I ,, , ,l<-f ,

,'<>

/y

,I
I

r
,
I

It,

r-=-"'~I -------;1
J

..
.
00

"

Sur/nu IIIId h,ter/an! Phylici

,
S.,,(n.
~.~

"

.
.

"

"",nl>

"

"'I , ", ,

"t,'
I b,:
I I , I

b I

/ A , , I I

---- "

OSul""..,p

-~"

I , Su,f:oc.e .'''.. tUf~ .....-.h

,~-7--1

I '
,

"

"

".,

'"

.,

..

Figure:l A(3 )( I) su.f.....e ~truClurc. (a) f<llll-spolC'e. and{b} rcdpnx'll1-space ~ .... m5. (Mter E. A. \\Wd.)

(oi)

"')

(01)

FlgUre,j E....'lIIJ sphere conslructioo lOr dm"r.octkm of incident ""'xc k by. ~1U>Ull net.... hen k is paraltello me Ibis of the Dtbh. lhe '--It iOlltcretl beams 'n the plane or the papl'r are k;. k,j. ... 1.:; I>lIfn<.1ed 0Catrrs out or the plane or the pll(JCr wul abo Ql:t\l., such as (12) llJ'I<1 (12). The _"Crt ".t lines are the roW of the riprocal nd.

."

,GO

"-igore:; I..EED pIItl.cm$ fmm a Pt{lll) 0)'HIII ... na..-e b iocick..."t electron cncr}(it.'S clSl and 63 S c\'. ~ c.I;ff~ angie <s gr<:'llter at the b-"r mCfW. {After C. A. Somoojai. Chemi.ftrr in tlCO Jimr.... ion,. surft>u:r.J

tron energy is typically ill the range 10-1000 cV. 'nils is the arrangement with \\hich Davisson and Ccnner in 1927 discovcn.-d the wave nature of the e1ecIrOIl. An experimental paltern is shown in Fig. 5. nef1oction High-Energy Electrotl DifJrtlclion. In the RITEED method a

beam of high-enc'l,.'Y clcclrollS is directed 111'011 a crystal surface at gra"illg

incidence. By adjustment of the angle of inddcllCt: olle can arrange the !lonna! l:omponent of the incoming wa\'c\'l:..'C1:or to be vcry small, which will minimize lhe penetration of the electron bC31H and cnl13ncc the role of ti,e C11'5131 sur-

I:\cc.
The mdius k of the Ewald sphere for 100 kc'V ek'Ctrons willl>e -JrP A-I. which is much longer than the shortest nx:iprocal lattice vector 2ma ... 1 A -I. It follows that the Ewald sphere will be nearly a n.\I surface in the central scnttcring rcgion. '111C intercept of the rods of the n ..'Ci!lI'ocal net with the nearly nat spllerc will be Ilt.'arly a Iille whell the beam is din.:cted at gm.ting incidence. '11\C cKperimelltal armnl:,''Cmcnt is sIJown ill Fig. 6.
SURFACE ELECmOf'liJC S'mucrURE

Work Function
The work function W of the ,,"iform siliface of a metal is defined as the differencc in potential energy of an electron !Jel\n.'en the vacuum le\'e! and the Fermi Ic\el. '111e \acUUIll IC\'cl is the cneTb'Y of an ek'Ctl"On at rest at a point sufficiently far outside the surface so that the ck:ctrostatic image force On the

."
F".,;ure 6 lbr RH Etm mo':thod. In (a) the IIigb energy inudenl electron bram llI. a glancing ..>gIe to the ~I surface is associated with an Ewald $pi,,"", ur~ noliUl. ~ large t1ll1llhe: ,urface is licarly flat ill rdalioll to thescparntion beh'~ adjacent rods oftht' rccipnxal net The IOnnation

'"

or

diffraction h"..",

OIl

a plane oCTccn is

~Io"w"

in (b) (After I'r"Upn.)

electron may be negll.'Ctl.'<!.-more than 100 A [rom the surbce. The Fenlli level is the electrochemical potcl'ltial of the electrons in the metal. Typical valnes of electron work functions arc gi\'cn in Table 2. The orientatioll of the exposed cr)'stal face affects the value of the work limctioll 1)('''c:lllsC the strcugth of the ek>ctric double layer at the surface depends 011 the concentration of surf.lce positive ion cores. 'nle double layer exists because the surface ions are in an asymmetrical Clwironmenl, wilh vacuum (or all ac..Isorbed foreign :110m layer) on one side and the substrate 011 the other side. The work function is equal to the tllreshokl encrl,'Y for photoelectric emission at absolute zero. If fiw is the encrb'Y of an incident photon, thf..'ll the Ein steill equation is fiw = W + T, where T is the kinetic energy of the emitted electron and W is the wor" function. Tlwrmioflic Emission

The rate of emission of thermionic electrons depends exponcntially on the work function. Thc derivation follows. We fiBt filld the electron concentration in vacuum in f.."(luilibrililTI \",ith electrons in a metal at temperatnre ~=k8'I) and chemical potentiallL. We treat
the electrons ill the vacuulll as an ideal W-\s, so that their chemical potential is
(5)

by TP. Chapter 5. Here

'"
Table 2 Electron work fllnclions a
(Values ol,t;,lined by phOlllcmission, except tungsten obtained by I1cld emiSl>ioll.)
ElcmCllt
Surface plane

Wock fundi"n, in eV

Ag
C, C" Co
Ni

(100)
(110)

4.64

452
4.74 2. J.l 4.59 4.48 4.98 4.80

(Il1)
polycrystal

(IOO)
(I HJ)

(JJ J) (Ill)

(JOO) (110)
(111)

5.22
5.).1

(JOO) (lW) (111)

5.35 4.63 525

4.47

(6)
for
~rtick'S

Now from (5),

J4.'~1 - P, =

of spin 112. W, by the definition of the work function W. Thus,

= /lQ exp( - \Vir) .

(7)

llie nux that leaves the metal surface when all electrons are dl11Wn off is equal to the flux incident on the snrface from outside:

ill

by 'l'P(14.95) and (14.121). vaCl.1Um. The ek'Ctric charge flux is ej" or

inc = (7121rm)1/2n Here c is the mean speed of the electrOIlS

=

(8)
ill the

(9)

]11is is calle<1 the Richal'dson-Du.~hmanequation for thermionic emission.

Surface Stoles
At the free surface of a semiconductor there orten exist surfa~;-I){)Il1ld electronic states with energies in the forbidden gap between the valence and conduction h:mds of the bulk scmiconductor. \\'e can obtain a good impression ofthe nature of the surface states hy considering the wflve functions ill the we.lk
\

..

binuillR or lWO-C01l1pol\cnt appfmillJ.ltion of Chapter 7, in OIl(" dillLemioll. (1'\t~ wave functions ill three dilllen~i()ns will have c\.lI<I f.J.<,'lors exp[i(k~y + k~z)1 in the y,z plane (If the SlII'f.lCC.) tfthe vaCUUnl lies in tJle rcJtion x > 0, the potential ellergy of <til d('ctroll in this region can be set equal to zero:
U(x)==o,

x> 0 .

(10)

In the Cf)slal the potential energy has the uSllal periodic for1l1:
U(x) ::

L V c; cxP(iGl')
G

r<O

(11)

In Olle dimension G = nmo, whcre II is any intcger, including zero. In the \'<lcuurn the wa'le func1.)()n of a bound !OuTface state must fall off exponentially:

l/I"... =

C\.Il( -$X) ,

.1>0.

(12)

By the ,,,,"we equation tllt~ enc~ of the st.Ite referred to the 'lacuum Ic'le1 is
E = _r,2$2/2111 .

(13)

'Vithin the crystal the tWl>-<.'Onl(>onelit wa'le functioll of a lX)J]\I{1 !>L1r1ace sbte will ha'le the form, for x < 0,

l/Jln = exp(f/x + rklXC(k) + C(k - G)e),:p(-il-x)) ,

(14)

by analogy with (7.49), but with the RdditiOll of the factor Cxp (fiX) whidl serves to bind the electron to the smf.\ce. We now come to an important consideration that r('~tncts the allov..ed 'lalnes of"the wa\C"'ector k. [hhe st.lte is boUllll tht:re can be 110 current flow in the J" direction, normal to the surf<lCe. 111is condition is .l.\loured in qU.J.lltum mechanics uthe wa'le lilllctioll call he written as 11 rc.J.1 fundion of x, a colldition ",I ready s..'1tisfied by tl.e exterior ,\.IVC function (12). But (1-1) call Le a real function only if k = ie, so that
",n = cxp(qr)(C(iG)exp(iC..-l2)

+ C(-iC)e"l:P(-IG1"2)) ,

(15)

'I1lis is real provided COC) == q-iG). 11ms k. for a surface state does not h3\e a continuum of \.llues, but is Iinllted to di'iCrete ~tatcs associ,lted with Dril\ouin wne bOl.lm1.'lrics. TIle sbte (15) is dalnpetl e:l.pollcl1tinlly ill the Cf)stal. 'nle constants s. q are related by the condition that t/J aud dl/idx arc contillllOus at x = O. '111c billding energy E;s detennincd by solving the two-compollt'nt secular (.'qllation I analogous to (7.46). 'm,. plot of Fig. 7.12 is helpful ill this conllet:tion,
'.'or Jchllb. sce E. T Goodwin, Pruc c...mbrod,l:c l'Ill~~.

s.x.

:JS, 2t.l.:> (1005).

,-1
---------------~

--- ---------------.,

.. .--,

-----<

~-----,

Figure 7 Il.I"d bcotlillj:! ,....,... a scmicor,ductor IUrfacc Ihal ('QII gi\'e;l highly "OIIdueting sura"e .t'):io". (a) 1",cn.1OfI b)... oo an ..-t)'Pl' .... mirondlK'tor. For tloe bendingUtllCMTllhc hok <.VnCC'n-Ir.dion at tI.., .urf:.c:r u far ~ II.... the ek'droa conttnlf1lli<Jol in the intlCf'lCW. (b) Attumulahon b)'\.T on an ..-type 5C.'mioondudor, "ilh an e1cdroo, c:uncentration I' the,...~ lbal is Jar higher th:... III the lIIlcrlo,-

Ttmgcrllial SlIrface Transport \\le luwe seen tl'k'\1 there may exist sllIfaec-oound electronic states with encrb'ies in the forbidden gap Letweeu the \'alcno' and condllCtioll bands of the substrate c~stal. These states llIay l>c O(:CUpil:d or varont; their existence must arrcet the statisticalmechallics of the problt:m. 111is means that the states mod~ ify the local equilibrium concentratiOll of ek'Ctrons and holes, as exprcssl'i:l as a 51,ift of the chemical potelltial relali\'c to the band edges. Bl.'Call5e the c11emic:a1 potential is independent of position in an equilibrium system, the energy bands must be displat'ell or bent, as in Fig. 7. The thickness and carrier conCClltmtiOIl ill the surf:\CC layer may be chang(,'(! by applying an electric field normal to the surf.1cc. The eJfl.'"Ct of an cxtemal field i~ utilized ill the metal-oxide-semiconductor field-effect trallsi~tor (M~F':T). 'I11i~ has a metal elt:.:ctrode just outsitlc the scmicondllctor stlrfuec and insulated from it hy a layer of oxide (Fig. 8). A \'olt:..1gc, the gate voltage Vg , is applk>d between the metal ami the blilk semicomJuclor and modulates tlle amtluctance between any other e1l.'ctrodcs pL1CCd in contact with the elt:.:ctrons of the surface spa(''C charge. The physics of tI}() surface contillctancc channel is trt.'ated in Problems 2: and 3.

MACNETORESISTANCE IN A TIVO-DIME1\'SIONAL CJIANNEL

The static magnetoconducti\'ity tensor in 3D was found in Problem 6.9. Ilere we translate that result to a 2:0 surface conductance channel in the X!I pl;me. Witll the static magnetic fipld ill the z direc-tion, normal to the MOS !arer. \\'e assullle ... = ""II} eleclrons per unit area. The surface condnctance is de-

19

S"ifQC~ GOO Intcifa

""gain

56S

Fi,g"n,8 A mcl;l.! . u:o:idL~5Crni<.'Qnduetor tr.mmtur. The gIIte u)ltage V" i, ~pplicd between the metal and the scmlcolloJud()f, cau,ing all accumulatlQll or charge carriers at the o~ide ;"tcr~, Thili ClitHblisl.C1 a Iworuim<'Miolnal coomu<.tivity dU\rlllclln tile pbo.ne ci the intcrfare. which is the xy pL,nc,

fined as th~ volume conducthity times the la)er thickness.. The surface current density is dc..>fint.>O as the current crossing a line of unit length ill the surfuce. ThIlS, with (6.43) and (6.. 65), the surface tensor conductance compont..'fl15 b<come

CT., =

O'oWe'l"

(16)

where 0'0::E ve'Tfm and We = eB/me in CCS and eBlm in SI. 1be following discussion is written in CCS only, except where ohms are used.. TIlCSC results apply spcciflcally in the relaxation time approximation u5<.."(1 in Chapter 6.. When We'l";;" J, as for strong magnetic field and low tempcmtures, the surfucc couducthity components approach the limits (J7) The limit for O'.y is a general property orrree electrons in crossed electric magnetic fields B. \Ve establish the result that SlIC!J electrons drirt in the x direction with velocity VD = eE';8., Consider the ek'Ctrons from It Lorentz l'''lIne th....,t moves in the x direction with this velocity.. By electromagnetic tl1l.'ory there is ill this frame an electric field E~ = -vv8jc that will cancel the applied field Ell for tile above choice of VI). Viewed in the laboratory fmme, aU electrons drirt ill the x direction \\ith velocity VI) in 3lditioll to any velocity components they had before Ell \\."3$ applit.'tl. nlUsj. = U.yElI = lIrol) '= (.-ecI8)E", so that

Ey

Hill!

0',.

'=

JledB

(18)

as in (17). 111e experiments measure the \oltage V in the y direction and the current 1 in the % direction (Fig.. 9). Here I. = J)lI = (ved8XE,).,,) = (...cd8JV,.. The HaJl resistance is

PII

V,fl.

'=

B/we

(1&)

be infinite. Parodoxically, tilis limit occurs only when

\\'e see thiatJ. can flow with zero E~, so th,'lt the effecti\e conductllnce can Un and U w are zero. Consider the tensor relations

11,.,

ill = u""E. + UwEli . (19) (n the'Hali effect geometry ill = 0, so that Ell = (u..)uvJE., with IF." = -up-

J. =

u:afi.

+ u~vEf1'

(20)
and in the limit u Ja =
U

w = 0 the effective conductance is infinite.

Integral Quantized Hall Effcct (IQflE)

The results of the original measurements! under lIu.mtllm conditions of temperature and magnetic field are shown in Fig. 10. 1he results are remarkable: at certain values of the gate voltage the "oltage drop in the direction of current flow goes essentially to zero, as if the effecthe conductance were infinite. Further, there are plateaus of the Hall voltage near these same values of gate voltage, llnd the values of the Hall resistivity V"Il. at these plateaus are atturately e<Jual to (25,813linteger) ohms, where 25,813 is the value of expressed in ohms. The IQHE voltage minima V pp may be explained on a model that is, however, oversimplified. Later we give a general theory. Apply a strong magnetic field such that the separ.ltion fiw c ~ kfj T. It is meaningful 10 speak of I~ndau levels that are completely filled or completely empty. Let the electron surface concentration (proportional to the gate voltage) be adjusted to any of tile set of values that cause the Fermi level to fall at a L1.ndau level: from (9.33) and (9.34),

hie

seB/l,e =

/I ,

(21)

where s is any intcger and v is the electron sUlfaoc conocntrlltion. Whcn the above conditions are satisfied, the electron collision time is greatly cnhanced. No elastic collisions are possible from one stale to ,'lnother st:1te in the same L.1.mlau level because all possible final states of eqmll energy are occupietl. TIle Pauli principle prohibits .m cl."lStic collision. Inelastic collisions to a vacant Landau level are possible with the :1bsorption of the nccessary energy from a phonon, but there are vel)' few thermal phonons of energy greater than the interlevel spacing by virtue of the a.uumptioll fiwc ~ kBT. The qUllntization of the Hall resistance follows on combining (IBa) and (21):

Pu = hlse'l = 2msca .
where a is the fme structure constant effie a Ul37, llnd s is an integer.
II(. 'on KI,tzing, G.

(22)

Donia. lind M. Pqlller. Ph)'s Rev U:t1 45.494 {19SOj.

19 Stlrfoce and

Inf~rffJCe Physia

"
!--~---=.-,,-------.I

F'' ' . .

9 Allplied 6cld E. w and drift CUlTent 1. in a (l~ltum llaJl drect (IQIlf,.l eo:pcrimcnc.

" "
".

",------------------,

" "
, " "
~

'" 1 , ,
-}

,.
V,.

"

"
,

"

figure 10 In IheQriglnallQ1IE mea,suremenh" ~tic Ilelrl .,(160 l..e (181)pomIS o.Itofll.., p3(lCr. TIle tcm~r:Illlre is 15K. A constant ",,,n'nt of I ,..,. u rVl"C'ed 10 I10w !JI:'t'ecn the SOUfC't: and the dnUn Voltages " .... and are 1l10tll-d ''<:nllS the gate \Ullage- V.. "'hidl '5 1l1'Ul~1OIU1 to the Fermi 1C'\"d. (AftCT K ,on Khtzing. G Donla. and !'-1. Pt'I'Ilt. l

,"

"

,
"H

"

"

'"
fQIIE in Real Systenu

The meastJremcnu (Fig. 10) suggest that the above theory oflhe IQI-IE is
100 good. The Hall resisth-it)' is accurately quanli7..ed at 25,8J3I8 ohms, whether or not the semiconductor is of very high purity and perfection. TIle sharp Landau levels (Fig. lIa) afe broadened in the feal crystal (1-.g. llb), but this does nol affect the Hall resistivity. The occurrence of plateaus ill the Hall resist'lOre, evident in the UII curve of Fig. 10, is not expected in ideal systems because partially filled Landau levels will exist for all gale \foliages except those for which the Fenni level exactly coincides with a Landau level. Yet the experiments show that a range of V" values gives the exact Hall resistance. L."1ughlin 3 interpreted the results for real systems as the expression of the

general principle of gauge invariance. The argument is subtle and somewhat: reminiscent of the flux Quantization in a superconductor in Chapter 12. In L..'lughlius thought-experiment the 2D electron system is bent to fonna cylinder (Fig. 12) whose surface is pierced everywbere by a strong magnetic field B normal to the surface. The ClIrrent 1 (fonner IJ circles the loop. 10e magnetic field B acts on the charge carriers to produce a Hall voltage (former "II) perpendicular to the current and to B-that is. v" is de\'eloped betv.een one edge of the cylinder and the other. The circulating current I is accompanied by a small magnetic flux qJ that threads the current loop. The aim of the thought-experiment is to find the relation between I and V". 'Ve start \vith the electromagnetic relation that relates I to the total energy U of a resistanceless system:

"II

1 iJlp -:::::-VI :::::--

au
at

'

c iJt

l=c

bU

llqJ

(23)

The value of I can now be found from the variation llU of the electronic energy that accompanies a small varintion llcp of the flux. The carrier states divide into two classes: Localized states. which are not continuous around the loop. Extended states. continuous around the loop. Localized and extended states cannot coexist at the same energy. according to our present understanding of localinition. The two classes of states respond diITerently to the application of the flux cp. TIle locali7.ed states are unaffected to first order because they do not enclose any Significant part of qJ. To a locaIi7.ed state a change in qJ looks like a gauge trnnsfonnation. which cannot affect the energy ei the state.
3R 8 Laughlin. Ph)"1;
&,..,

book 01 $C>mCI! lind IJchnoftJgJI. 1964. pp. 20&-214. "

Tsui. Sdenl'e UG. 1241 (1963).

B 23. 5632 (1961), ~ aka hIS article in the r.lcCraw.Ili11 iK.<I.... re-ieY.... gh"'n b) II. L Storma'.oo D. C.

Figure II Density stat~ in Il 20 electron gas in a strong mllgnetie Oeld. (a) Ideal 2D crystal. (b) Heal 20 crystal, with impurities aud imllerfOXt>ous.

"r

FW-I"e 12 Geometry b- Laughlin'l. thought-eqlerimeut. "Ole 20 electron ~)"Stem iJ wrapped around to "nn a cylinder. A stroog m~tie field Jl pierces the cylinder e\eryv..I,ere 1l0nllal to iu surface. A eurrenl f eireles the loop, gh'ing ris,", 10 the Hall voltage V" aud a smallmagnellc flu. II' Ihrough the loop

nle extenclal states enclose tp, and their energy may be changed. However, if the magnetic flux is varied by a flux quantum, &p = hde. all e~tCl1ded orbits are identical to those before the flux quantum was added. TIle argument here is identical to that for the flux quanti7.'ltion in the superconducting ring treated in Chapter 12, but with the 2e of the Cooper pair repl~ by e. If the Fermi level falls ,vilhin the localized states of Fig. lIb, all extended states (umdau levels) below the Fermi level will be filled with electrons both before and after the flux change &po However. during the change an integral number of states, generall). one per Landau le"el, enter the cylinder at one edge and leave it at the opposite edge, The number must be integral because the system is physically i,dentical before and after the nux change. If the transferred state is tmnsferred while OCC\lpied by one electron, it contributes an energy change eVil; if N occupied states are transferred, the energy change is NeVil' '111is electron transfer is the onl)' "..ay the degenerate 20 electron system

'"
can change its ene~. We can understand the effect lly looking al a model !i)'stem mthout di~rder in the umdau gauge for Ihe vector potential:
A = -Byi .

(24)

An increase &\ that corresponds 10 the increase ~ is equivalent to a displarement of an extended state by MlB in the y direction. By the Stokes tJleorem and the definition orthe vector potential we howe &p = L"M. Thus &p causes a motion of the entire electron gas in the y direction. 8y /j{} = NeVil and &p = hele, we ha\'e

nux

I = d...I1UIl1tp) = cNcV,l'lc = (Ncfll)VII

so that the Hall resistance is
Pit = Villi = IJNc

(25)

(26)

as in (22). Frocliorwl Qtwnliud Hall Effect (FQJlE). A quantized Hall effect has been reported. for similar systems at fractional values of the index s. by,",,'Ol'idng
at lower temperatures and higher magnetic fields. In the extreme quantum limit the lowc~t Landau level is only partially occupied, :l.nd the integral QHE tre.lted above should not occur. It h:u been obscr\'ed.~ howen-r, that the Hall re.~istance PlI is quantizetl in units of 3h!c when the oecnpation of the lowest Landau le\'el is 113 and 213, and POT vanishes for these occupations. Similar breaks ha\'e been reported for occupations of 215. 315. 415, and 2t7.
p-n JUr-;CfIONS

A p-n junction is made from a single crystal modifled in two separate regions. Acceptor imllUrity atoms Me incorpomted into one part to produce the 1) region in which tile majority carriers are holes. Donor impurity atoms in the other part produce the n region in which the majority carriers are electrons. 'nle interface region may be less than 10-'\ cm thick. Away from the junction re~iol1 on the p side there are (-) ionized acceptor impurity atoms and an equal concentration of fn~e holes. On the n side there are (+) ioni7.ed donor atoms and an qual concentration of free electrons. 'I1ws the majority carriers are holes on the p side and electrons on the n Side. Fig. l:l. Holes concentrated on the" side would like to diffuse to fill the crystal uniformly. Electrons would like to diffuse from the n side. But diffusion witi upset the local electrical neutrality of the system.

4D. c. T,";, II. L Srorme~, and A. C. Gos.Ard. PfI)J. ~ Lett 48. 1562(196:2), A M chang . Ph)'$. Ra'. Lett. 53, (1984). for a diiCUSSOull uidl~ tke<:lry see R Laughlin ,n C. Bauerct aI., eds.. Tlnl-dttnou*>nd 'VS1t:mr, heI~rostrud"ra. ",od ,u~rl(Jtticu. SpI'1ngc~. 1984.
~l al

19 Surfoce and rnt"r!uce

""!15k.
Cor.een'r.ll.....
." c1.'rl,oo,

5.

c.mc..,l...."""

.....

(.J
figllr" 13 (a) Varialinn uf the hole amI ck ...1ron c<:>nccntrutinns :> across an ""honsed (",em applied "o1ta~) junction. Th" carriers 8,'e 1 ill thennal C<jll,librillm wilh the <leecptur and do,.,.,r impurity \i atoms, so tlMI the prodllct Jln of the hole and electron concclltra- 8., I,,,ns is t'Ollstant throughout the el)'$lal in conformily \<ith lhe law ~ r ur mass action. (h) Electrostat,c poIcnti,,) from acceplor (-j and ] dm,or (+) ions n<->(Ir lite JIlIl<.1ioll. The potential gradient inh,b'ts !: t1,fii,siol, or hules rrom the p .,dc 10 tI,c " side, ill"l ,t inhibits :;; d,fiuSlon ordectm.... rlllm the" side lu Ihe 1/ side. lnc electric field in the JllIlcli.m region is called Ihe buill-in electriC field.
-

i+"------+

--------='tc----+ + +
+

I--------:!_

OJ

A small c1wrge transfer by diffusion leaves behind 011 the p side an ex('t;ss of (-) Ionized acceptors and on the II side an excess of (+) ionized dOllors. This ch.. " rge douhle layer creates an electric field directed from n to p that inhibits diffusion and thereby rnaint<lins the separation of the two carrier types. Becau~e of this double I;lyer tIle e1edrostatic potential in the crystal takes a jump ill passing through the region of the junction. [n thermal equilibrium the chemical potential ofeach carrier type is everywhere collstant in the cl)lsta!. even across the junction. For holes
knT III p(r)

+ eq:(r)

= constant ,

(27a)

where p is the hole concentration and 'P the electrostatic potential. lllUs p is low where 'P is high. For electrons
knT In lI(r) - erp(r) = constant
(27b)

and n will be low where 'P is low. TIle total chemical potential is constant across the cryst..l!, The effect of the concentmtion gradient exactly cancels the elt:(:trostatic potelltial, and the net particle flow of each carrier type is 7.l:'ro. However, even in thermal eqUilibrium there is a small flow of electrons from n to 11 where the electrons end their lives by re<'Omhination with holes. The recombination current Jnr is balanced by 11 current Jill'. of electrons which are ~enenlted thermally in the p region and which are pushed hy the built-in field to the 11 region. TIllIS in zero external applied electric field

1../0)

+ 1",(0) : 0

(28)

572

for otherwise electrons would accumulate indellnitely on ont.. rier.

~lde

or the har.

Rectification
A p-n junction can act as a rectifier. A large current ",..m flow if we apply a volmge across the junction in one direction, but if the voltage is in the opposite direction only a very small current will flow. Ifan altemating voltage is applied across the junction the current will flow chiefly in one direction-the junction has rectified the current (Fig. 14). For back voltage bia.~ a negative voltage is applied to the 11 region and a positive volmge to the II region, thereby increasing the IXltential difference between the t\'o"O regions. Now practically no electrons can climb the potential energy hill from the low side of the bamer to the high side. The recombination current is retluced by the BolbJnann factor:

(29)

The Boltzmann fu.ctor controls the number of electrons \'oith enough energy to get O\'er the hanier. The thennal genemtion current of electrons is not particularly affected by the back voltage because the generation electrons flow downhill (from 'P to n) anyway:
(30)

We saw in (28) thaIJ....(O) = -J..J,0); thus the generation current dominates the recombination current for a back bias. When a fonvar-d voltage is applied, the recombination current increases because the potential energy !>arrier is lowered, there!>y enahling more electrons to flow from the 11 side 10 the 1) side:
/"..(V fofW<ll'd) = /"..(0) exp (elVl/kn'f)
(31)

Again the generation current

i.~

unchanged:
(32)

'111e hole current flowing across the junction behaves similarly to the electron current. The applied \!Ohage which lowers the height or the !>arrier for electrons also lowers it for holes, so thai large numbers of electrons flow from the n region under the same voltage conditions that produce large hole currents in the opposite direction, The electric currents of holes and electrons are addithe, so that the total forward electric current is I '" IJexp(eVlksT) -

II ,

(33)

where l~ is the Sum of the 1;\\'0 generation currents. This equation is well satis-

'00

--n-,

ot:J.~_Uol

,
__ R"w:n"
dlt~"""

figu", 14 Rerti!X:atilln characterisl;" of. pon junct......

in gnmllnlnm. afte.. Shocl:Jey. The ~'Ol(age is lOtlt.'<1

,~rticaJl)'

000

IlI1d lhe currenl horizgnlally.

:/..
om

I........... f---' dftmoo

v -[

,.,

"

'00

fled 01' p"l junctions in germnnium (Fig. 14), but not quite us well in other semiconductors.

Solar Cells and Photooollaic Detectors

Let us shine light on a p-fl junction. one without an extemal bias \oltage. Each absorbed photon creates an electron and a hole. WhCfl these carriers diffuse to the junction, the built-in electric l'ield of the junction separ.'ltes them

at the energy barrier. The separation the carriers produces a forward voltage across the barrier; forward. ba-ausc the electric field of the photoexcited carriers is opposite to the built-in field of the junction. 'I1)C appearance of a fonvard voltage across an illuminated junction is called the photo\'oltnic effect. An illuminated junction can deliver t)O\ver to an extemal circuit. Large area p-n junctions of silicon are used to convert solar
photons to electrical energy.

or

Schoul:" Barrie,.
When a semiconductor is brought into contact with a metal, there is formed in the semirooductor a barrier layer from which charge carriers are severely depleted. 11le barrier layer is also called a depletion I.'lyer or ~",haus4 lion layer. rn Fig. 15 an n-type semiconductor is brought inlo contact with a metal. l1-.e Fermi levels are coincident after the transfer ofeloctrons to the conduction band of the metal. Positively charged donor ions are left behind in this reglon

'"

Md.l

S<'mlco<~\uctQf

(ll) I~ """'""'"

(b)

Ju,t .net"

~-

Figure 1.5 ~Iifying ",,"ier !>ct",-een a metal and a n-l)'pc scmj,co"d"ctOI TIle Fcnni bd is ~Ilown ,\S a broken line.

that is practically stripped <l electrons. Here the Poisson e(luation is

(ees)
by
(CCS)

div D

4;me

(SI)

l1iy D .. nd#.t.1

(34)

where" is the donor concentration. 11le electrostatic potential is detennined

cfl"Jtb:' =

-4.,-ne!E

(35)

which has a solution of the fonn

(CCS)
~ ~ -(2mud.),'

(36)

11lc origin of x has been laken for convenience at the right-hand edge of the lh'lrrier. 11H~ contact is at -x/). and here the potential enefb'Y "dative to the right-hand side is -e(/Jo, whence the thickness of the b:lI"ricr is
(37)

With E" 16; eliPo = 0.5 eV; re = 10 1G cm-3 , we find Xh = O.3ILm. This is a somewhat simplified view of the met.ll-semiconduetor contact.

HETEROSTRUCTURES
Semiconductor heterostnlctures are 1a)'erS of 1\\'0 or more different semiconductors brroWI1 coherently with one COInmon crystal structure. Hetcrostruclures offer extr.a degrees of rreedom in the design or semiconductor junction dC\'iees. because both the impunty doping and the conduction and valence band offsets at the junction can be controlled. This freedom is the basis of the prediction Ihat most de-.ices that utili7.e compound semiconductors \\ill in the

19 Sll~ ",,,,

'''/fir/ace Ph'Jl'in

.,-.,--r.,C"...,

16 "fhrcf! l}"J>C' olband edge olf~:at hctc...... in~. nle bbMlden ~ are sh"...n shaded. 1lle 0JI&ct called .....aI ClC.'aIn, r.....- Ooampk, '" GaA....(A1, Ga}A'. 11>e "1.>rot:eol.pp~ olTset OOC\In III the GaSbflnAs heteroJUnctlOn.

future incorporate heterostructllre5. We treat them here in ord"r to keep ahead of the times. A heterostfilcture may be viewed as a single CfY!>tAl in which the occupancy of tile atomic sites changes at the int~rf.tee. As an example one side of the IIlterface can be Ce and the other side GaA~; both lattice constants are 5.65 A. Olle side has lhe diamond structure and the other side the Cllbic zinc suUide structure. Both structures are built up from tetrahe<lr.\1 covalent bonds and fit t~ether coherently as if they were a ~ingle cl)'Stal. Ihere are a few edge dislocations (Chapter 20) to relieve the strain energy near the interface. TIle band gap.~. however, are diffNent. and this difference is the source of the te'J.1 intere~t in the heterostructure. apart from Ule teclllliC"J.1 virtuosity in rormin~ the stlucture. The band gaps are 0.67 eV for Ge afill 1.43 eV for CaAs. at 300 K. The relative Alignment of the oonductiOll ami valence band edges offers several JXls~iLilities, as shown in Fig. 16. C"J.lculation~!I sugge.~t that the top of tile valence b..1nd E" in Ge should lie about 0.42 eV higher than in GaAs. 111e boltom of the conduction band Ec in Ge should lie about 0.35 eV 1000\ler than in C.\J\s. so that the offsets are cl.\Ssified as normal in the scheme of Fig. 16. Band edge offsets act as potential b.'lrriers in opposite senses on electrons lUld holes. Recall that electrons lower their ellergy by "sinking" on <III enerJO' lmnd diagram. whereas holes lower their energy by flo.1ting 011 the same diagram. For the normal alignment both electrons :Uld holes are pushed by the barrier from the wide-gap to the narrow-gap side of the heterostrllcture. Other important semiconductor pairs used in heterostroctures are AlAs! CaAs. ltu\slCaSh, CaP/Si. and ZnSelGaAs. Good lattice matchinlo: in the range
"SreTalJlc 10..1 In W A. II:arriS<lfl. EIKllonK; strllC:lu~(J.w 1I.e proPf!rliaofMJIidJ. Fre....lWl. 1960. See also II. Krocmet". -rh.. ..,..,. ofhctermtnoc:t,,1'fl &"nllCill r~'lCW. ~ 1::.-.,., ~un .. ner ScI.ool,

.""

FL--~--

,.~

-----f' L.
N

__________1:::

..1,. ,.

L-_-',

,.
t"igure J7 (a) Two semiconductors not in contact; the ab$olulC band edge energies arc labeled Eo fOC" th.., conduction band edge and Eo for the ''lIJc~ band edge. An "absolute energy" means n:'fcrrcd to Infinite dislllnCC. 11le Fermi levels in the tW<J materiah lire determined by the donor conccnlraticlns, ... wdl Q by the band.lnlcturc. (b) 'The lame semiconductors as a helerojunction, so th.1 Inc two pam arC in dilJusj,'e equilibrium, 'ntil; requires that the Fermi Ie-'e! (F.L.) be independent of posit , whidJ is accomplishcd by IraNiIe. of eledr(lm froon the lIo'-side to the n-side cl the inle~ depletion la)'Cf of poIili\'Cly i00i7.al donors is IeIt behind on the N-side.

0.1-1.0 percent is often accomplished by use of alloys ci different clements. which may also adjust energy gaps to meet specific device needs. nN Ueterojlmdion
As a practical example. consider two n-type semiconductors with a large

offset of the two conduction bands. as skctcbcd in Fig. 17a for a semiconductor
pair with a nannal band line-up. -Ole n-type material with the higher conduction band edge is labeled with a capit"dl letter as N-typc, and the junction shown is called an n-N junction. The electron transport properties across the junction are similar to those across a Schottky barrier. Far from the interface the two semicollduetors must be electrically Ilcutral in composition. Howcvcr, thc two Fermi levels, each determined by the doping, must coincide if there is to be zero net elcctron tnmsport in thc abscncc of an cxternal bias voltage. Thcse two considenltions fix the "far-off" conduction band edge energies rclative to the Femli level, as ill Fig. l7b. The combinatiOl1 of Il specified band offict (determined by the host material composition) at the intcrface and the distant hand energies (determined by the Fermi level) can be reconciled only if the bands bend near the interface, as ill the figUre. The lIecessary b..lIId bending is created by space charges consequent to the transfer of electrons from the N-side to the lower "-side. This transfer leaves behind 011 the N-sKle a positive donor splICe charge la)'Cr, which through the Poisson equation of electrostatics is the source of the positive second derivative (upward curvature) in the conduction band edge energy on that side. On the n-side there is now a negative space charge because of the c).ccss of electrons 011 that side. 'lhe layer of negative space charge gives a negative

19

S"rfOc~

and

InI~acr PIIyria

S77

second derivative (c.o .nward curvature) in the conductiOIl band edge enCrg)'. On thc 'I-side the band as a whole hends down toward the junction. This diITers from the usual 7.1-n junction. Thc downward bcnding and the potential step fonn a potential well for electrons. The well is the basis for the new physical phenomena chaJaCteristie of heterostniCture physics. If the doping 011 the 'I-side (low EJ is reduced to a negligible value, there will be very few ionized donors on that side ill the electron-rich layer. 'nle mobility of these electrons is largcly limited only by lattice scattering, which fulls oIT sharply as the temperature is lowered. Low temperdture mobilities as high as 2 x J<f cm2y-l s -1 have been observed ill CaAs/(AI,Ca)As. If nOw the thickness of the N-side semiconductor is reduced below the depletion layer thickness on that side, the N material will be entirely depleted of its low mobility electrons. All of the electrical conduction parallel to the intertce will be carried by the high-mobility electrons on the nside, equal in number to the number of ionized N-side donors, hut spatially separated from them by the IXltential step. Sueh high mobility structures playa large role in solid state studies of 2D electron gases and also in new classes of high speed field effect tTansistors for computcr applications at low temperatures.
SEMICONDUCfOR IJ..SERS

Stimulated emission of radiation can occur in dirt:et-gap semiconductors from the radiation emitted when electrons recombinc with holes. The electron and hole conccntrations created by illumination arc larger than their equilibrium conccntrations. 'nle recombination times for the excess carriers are much longer than the times for the conduction electrons to reach thermal equilibrium wilh each otller in the conduction band, and ror the holes to reach thennal equilibrium with each other in the valence hand. This steady state condition for the electron and hole populations is described by separate Fermi levels JJ.c and P-v for the two bands, called quasi-Fermi levels. With P-r and J.t,; rcl'crred to their band edges. the condition for populatioll inversion is tllat
~>~+~. ~

For laser action the quasi-Fermi lcvels must he separdted by more than the IXUld gap. Population inversion and laser action can he achieved by forward Lias of all ordinary CaAs or JnP junction. but almost all practical injection lasers employ the double heterostructure proposed by II. Kroemer (Fig. 18). Jlere the lasing semiconductor is embedded between two \\ider-gap semiconductor regions of opposite doping. An example is CaAs embedded in (AI,Ca)As. In such a stme-. ture there is a potential barrier that prevents the outflow of electrons to the

578

,
E......""" lluw

---\-.

'----

/If'fV

..

-rI
n-.

...
I
11011.
Ib-o'

'---'-

___1

adl\'il

Figure 18 Dooblc heterostructure injection 1.l<;CJ. EIcctI1l05 from the nght into the opticallyJa,,,.., ",t.erelheyforrn ado:1;cne.ale e1CChllOl gu. 1llI" 1'f1I"nlw harncrprovided b) thewide ClICfltY gap on the p side prt'1'Cllls the electrons from escaping to the left Iioles floy., from the left illio the act;\'f~ lap''', bo'l cannot escape to th.. rililil.

p-type region. and all opposite potential harrier that prevents the ootflo\.... of holes to the n-type region. The value of J.tc in the optically u<.tivc layer lilies up with JI." in the n c(mlaCI; similarly, P-u lilies up with p.." ill the lJ C(lIItact. Invcn;ion C'dll be achieved if we apply a hias voltage larger thai I the voltage equiv.llenl of tl1C active layer energy gap. 'nlC diode wafer pl'Oviucs its own electromagnetic cavity, for the reflectivity at the crystal-air iutclface is high. Crystals are usually poli~hed 10 provide two flat parallel sUlfaees; the radiatiOIl is emitted in the plane of the heterojunctiolls. Crystals with direct band gaps are required normally for junction lasers. Indirect gaps involve phonons as well as pllotons; carriers recombine less efficiently because of competing processes, and no lasc., action hus been observed in indirect gap semiconductors. Gallium arsenide hus been widely studied us the optically active layer. It emits in the near infrared at 8383 A. or 1,48 eV; the exact wavelength dcpellds on temperature. 1bc gap is direct (Chapter 8). In a hcterojuncUoll the system is very efficient: the nl.tio of light energy output to de electrical energy input is near 50 percellt, and the differelltiaJ efficiency for small changes is up to 90 percent. The wavelength can be adjusted over a wide range ill the alloy ~stem

,~

\~I
L------1 IIF---(.J

Fe---_

E.

"u_':_=I:t --v

~o 0 0 0

Ec

--,. [L ..

oooo~

1"'
Fi~re

19 Electron-hole rerombindtion into photons, <\Cross ~ p-n JUnction.

Ca.. hll_~PvAsl_V' SO that we call match the laser wavelength to the dhsorption minimum of optical fibers used as a Inmsmission medium. '111c comhination of double heterostructure lasers with glass fibers forms the oasis of lhe new lightw.wc communication technology that is gradually rcplachl& lrJ.n~missioll of signals over copper lillCs.
UGHT-E"'UTTlf';C DIODES

The efficiency of light-emitting diodes is now at the point of exceeding incandescent lamps. Consider a p-n junction with a voltage source V splitting the two chemical potentials iLn and J.tp by eV, as in Figure 19. Electrons from

the" .side are injected illlo the p side. and holes from the p side arc injected
into the n side. These injected carriers annihilate each other across the junction. thus generating photons if the quantum eff"lCicncy i) unity. The generAtion or recombination process will be much stron~cr in a directgap semiconduetor(Figure 8.5a) than in an indirect gap semiconductor (Figure

B.Sb}. In a direct gap semiconductor such as eaAs, the ban_ .o-band photons are absorbed in a distance ... 114m, which is strong absorption. The direct gap tem3l)' semiconductor GaAs l _ z Pz gives light tuned to shorter wavelengths as the composition variable x is increased. 'I1lis composition was made by J lolonyak into one of the first 11-11 diode lasers and into the first visible-spec.. tnlm (red) LED. Blue-emitting hcterostructures have now bccn made, such as In~Cal_.,N - .o\J..Gal_..N. l1lC pcrformance of LED's has ilLcreased markedly over tlle years, from about O.llumclLs/watt in 1962 to "bout 40 lumcns/watt ill 1994, compllroo \\1th 15 lumcns/watt for a standard whitc unfiltcred incandcscent lamp. To quote Craford and Holonyak, "\\le are entering an entirely new er.l in lighting (illumination) with all ultimate form of lamp-a direct gap III-V alloy p-Il hctcrostructure."
SC,.U.;r>lf\;C TU1"\r-;EUr>C f-UCROSCOPY

Thc original and still the central application of scanning tunneling micr()S.. copy is to the ohservatioll of the atomic structure of crystal surfaces. S'TM is a dream method that sounds much too simple to pennit observations on tJle atomic scale for which it was dcvcloped by Bilmig and Rohrer. '[1le developmellt was all extraordinary accomplishment, for which tlley received the J986 Nobel Prize for Physics. An account of the history is gh'en in their f\obel lecture for that YC3J"-accompanicd by a quotation from the almost traditional letter of rejection of major work from a scientific publication to which they submitted an early paper. Qile takes a sharp metal tip, prefcrably a tip tllat has olle atom that protrudes a bit more than aU tile other atoms of the tip, and briHgs it to witllin 4 J.. to 8 J.. of a crystal surface. A voltage applied between the tip and thc surface causes a tunneling currcnt of electrons to flow. By tunneling wc mean the standard quantum tunneling of a particle through a potentilll barrier. 'llie transmissiOn of electrons through a harrier is exponcntially scnsitive to the width of the barricr. The tnl.llsmissiOIl flux in the experimentaJ situation varics by a factor of about 10 for a I-A change ill thc separation of tile tip from the surface. 'lliis fact is the key to how such a simple device can have a vcry fine spatial resolution. The greatest part of thc tunnel current flows from the surface into the outcrmost atom of the tip, thus gh ing resolution Oil an atomic scale. The most COmmon way to operate an STM is to apply a voltagc between the tip and the surface. and to adjust the height of tJle tip continuously to keep tbc tunncl current constant as the tip is scanned across tJle surface. The signal actually recorded is proportional to the voltage applied to a pie7.O<'Cramic cry'Stal whosc displacement moves thc tip perpendicular to tJle surface. nle theory shows that for small bias voltage tJle trajectory of tlle tip measured in this way

19

Surface and Inlof_ Physa

56l

Figure 20 A "quautum corrllr of mt.'an raJius 71 A was formed by m<Wing 48 Fe atom::. On a eu (Ill) wrfao..e. The Fe aloml $Caller surf:lCl! 5tate e!edtOl'\l., confining them to the interior of the COI'1"aI. 'The rings in lhe cornol arc tilt' de...il)' dWribut>on of II", clcdrons in the three quanlum sl<lles cllhe corT:I! thai lie elme 10 II", Fenni ~'. The alonl~ were imagl:d and mo,'Cd i.,IO position 1' a low temperalu"", ultn.hls:h ''80",m !;Canning lunn"hng microscope. Pboto courtL'S)' of D. M. Eigler, lIlM Kcsearch D",\';I0".

corrcsponds to a contour of constant local dcnsity of states ill the surface, taken at the Fermi level. What you arc secing is whattllC STM "fcels"-a llIap of how the conduction electrons in the surface extend out into the vacuum that sepa~ riltcs the surface from the tip. With STfo.l one can resolve single atoms, as in Figure 1.27, and surface colTUgatiolls as small as one thousandth of an atomic diameter. Tbe tip call be usoo to move atoms around 011 the surface, strikingly shown here in Figure 20,

Problems
1. Di!JrodiQII
j'rQflI a lillear array and 0 squore arrflY, TI,e diffraction pattell] of a linear structure olattit'C COI'ISlaJlt a is cxplainedli ill Fig. 21. Somewhat similar stl"UClures are important in molecular bioJog)'; DNA and mall)' proteins arc linear hclice5. (a) A cylindrical IlIm is e~posed to the dif&action pattern of Fig. 21\>; the axis of the C)'lindcr is coincident \\ith the axis of the Iinenr structure or fibcr. Describe the

'Another \ie\oopOlnl is u,,"ul' Ii:w a lu..,.,.1aIl~e lhe dllf~ p811.-rn is de<Tibcd by lhe ringle Laue equation _. Ak .. 1<:"Q... here q b an intl!gJ'r. 1be Ltll~ IUnlS ",'bid> 1.cd 10 the other Laue equMions 00 not CX'ClIr for II hnear talha,. N__ Ak - <'OR~. is IhcCQ""tion of a plane, Ihus the rt.'Cipl1lCltl1Bllict: b,,'C<Hlles a sel of paralld planes normal to the hue of atoms.

' /__ 0>'

_d .........
p.oIh

b ...... ~.,

The d,fTnction [WlIem from a Single line ufl.. tlit'e co;.,S13nl D In' monodln-atK' ,.,y be.lm perpendicular 10 the hnc. (a) The t'OI'l<lltion for oonstrucli."e inlcrfcrc1K:'C is II ros fJ - n'\', where II is all intcl(cr. (b) "'or j.;h'cn " ti,e ditTrach.'<I rny. of con.lnnl ,\ lie on the surf.wc of a cuoc.

F~ 21

~I

Figure 22 (a) BackwBr<l ~lcrhlg paltcm of 76 cV electrons Ineiden' normally on tbc{lIO) rlllC'eof nkLcl crystal. a model of the surface 15 shown in (I.. (Courtesy cl A U. M.c1t.! 1

of the diffr.ll1ioll Ildtll'1ll On 11Il' 111111, (h) -\ noll pllo~'r.ll'llil' roWlo' I~ placed behiod the filX'r and 110111l.11 to the Il>ddcnt Ix',un, ~t..dch ro\l~hl) Ih.. .J1lP'oUa,nce oflhe dlffractK)11 pallern On the IJl.lte, (r) ,\ wigle plane of <lIOIlU fonl,s ,I "lu.Jrc lattICe of1.llhce con~bllt a, 11le plane is IlOn'/I<lI to the IIICIII"111 \'r,l~ "1:"IIl. Skt." roughl~' Ihl' :l1'I'll'arlII.)CE' nllhe diffr:ll:tion IJatte11l OIl the rhUhlj.,'f",vlllt: pl:l.t<.'. Ihut. The ddfrol:tioll from .. plane of atoms ron be inferred flOill the (l:ltt"nn for h\O perpendicular lines of atoms. (d) Figure 2:! shows thl' t,lecln'l1l tltn'r.ol."l:1OO Iwt!t"nl m the Wd.v.anI direction from tilt" nid::d atOPIS 011 the (1101 )urf.Y'C of.\ md.d lrUI.lI E"pWII the orientation ('/ Ihe ddTrac:tion (l.IUt:nl III R'Wholl 10 the .JtOl1l1C 1)()!;1II01l\ of tltt: surface aloms shown in the model. ASloume tJu.t onl~ the )U1f,lCt: "tOlns "'Ie cff.-ethe ill the refiC<"t)on of Joo,..-encrg)' dectlOns.
lIpt>Caraocc

2. Surfare ,,,bbtJl!tu ill dea,ie qua'lltllrl limit. l"".n~icltr tile ront.w.t pbnc beh\een :In il1.'m!atQr and .. Senlleomlul"lor, as in a mcwl O~I(k, 'iCnl;c.llI.ltlt'lor tr.mS"lOr or \lOlotTT. With. )tronJ; electric field d.(J(llictl..c"~ Ihe ~:>lO!-Si mtCIUce, the IlI>lellt~" energy ,.fa C'OflJIld.H'n ekdron m.ar I... :l{\Ilrosilllatc:ll h\ \'(x) ~ tE,1' ..... X posih\e and 1. \'(.1) '" ae lOr 1: ~\t)\'e. wllere the f1rlgill H(.r j, at tile illt(lfaco.., nil' wa\ efunction is 0 fon ncg.Itive "",d may be sep.lr.ltt'l!.lS 4(xy.::) '" "(I) C':p{l(k..'J + k,;J1. "here I"X) satbfol'$ the differel>lioll ('<llIatjllo
-(,,1"2m)dlll/d~+ V(:r)u - fIt

"''e

With the modd potential for V(x) the cxact ci@.. nfunl1l01l~ ,ltC AIr)' fUllc(itm~. hilt Cdn find ,I r.1.irly good ground state enel'b'Y from 11h.' ',lIi.ltillll.1! hl:ll fillldicm x e~p(-fl.r). (II) Show lhat (f) = (/.1/Z311)fl + 3efJ20, (b) Show tlmt the ('llcr~\' is '" minimum when a = (3efllll2t.~I!;\. (c) Show that (t:),,~. - I ,b9(Il'/2111)lrl (3I,E/2)"3 III the exact solution 101' the ground statc cnergy tile f,1l:101' I btl is repl,lCcd hI 1,7b. ,\5 . i5 mcr.'aSt.'li the extent of the ,.......vefunctioll in the r dirt'ction b dlocrl'ased, 111(' fundionll(r) deflncs.1 Sllrfa<.'C collclultion eh.ll1nd t~l tile ~"llliCOnt'llct(lr side of tile intemlCC, "!11C v.lIiollS eigem"a1uc~ of 1/(%) define wlMt .\le c"lIcll dl'Ctl'ie )UllUamls, Because the eigenfulldion) arc real functions of r the \t.tICS .In IIllt ('.11 I} l'Utn'llt in tllc x dirt'l:tioll, but they do call)' a surf.ll'c .It.llmel l'U,I\'llt ill the y,': pl.me. TllC dependl'nce of the challuel Oil the ek'Ctric fleld in the I' dirt'clioll lll~"CS the dc\ Ice a field cAi.oct transistor.

3. Sllllbt.ilw!Hle Ilacn effecl ill a slIrface c1rarlllcl. The etft'Ct clc5(:rihcs thl' peri"dic variolUOIl with lIB (or II itll c1cctrtJll COIl(.'entration) of the 101lgitudhl:ll dt'('lri!:.ll n'sistance cia specimen in a strong pc~pcnclieular maw,ctle flchl at '" low telllpel.lture, TIleS(' are the conditions in wlliell tl de 1Ia,u\an Alphell df...cl: is ,.,hscncJ. ,I~ di~ eusM.'li in C1lallter tl, l11c effect (".tn lit;' discll.'>SCd )impl) m 20, (.,1 51",w th.lt the dt..~I)it)' of free electron olhitals in 20 is 1lI/2w1i2 with omission ci )(lin .Iud ,;dlc)' degt'11emc}', 11Je sUlface a)llceutrati(m of electrons CUI now be '\I'itten 'IS l' = mf~2TrlI2. (b) Show that tl.e S-<!H O!iCillations h:l\'e the !lCliotl f.(UB) - c1t1l11cfrt!21rllcl'. In 51 set c = I. (e) According to the view of the ill\ersillll I.J} l'r.lS a pIaU:l~ 11lpacuor the surfuee cha~ge COlll'entratiou l' \:ui('S directly as tlie ~tc ,oltagr \'11: acrms the interface. Show Ihat the resistance osl,lIatiuns.lrt' IX nodi< JIIlK'hOllS vftl.c gate \Oh':lge. as III Fig. 23.

"'

Figure 23 Sourcc-dnlin resi$tance or an ,.,-"haunc! MOSl... , - at' .lunctioo oCtile gale ,..:lIlage V... at lransn-nc mllgnelic r>ckls or 33.3 and :5'.1 kG, al 4 K. (ArlCT G. uoo...T.!,r cl al.)

References
II. Vanselow, e<l., Cht'mIsI'1l fmd pllll$l~ of solId ""riflC. Springer, 1979-. C. A. Somorjai and M. A. Van Hove. Adsorbed mmwloyerr Oil &olftf rorjucr: Springer, 1979. T. N. 11,,<1;n and C. Ertl, eds., l\'mu,." oft/If' .ur/au chtm/cfll b<md. Nord, Holland, 1979. M. A, Van lIo\'e and S. Y long, Surface Cfl,I'lall'lflfI111h!l by 1.1';1,;0, Sprirll'cr, 1979. A. A. Maradutlin, II. F. WalHs, and L. JA.>brl:)'Tlski. cds. J/tmdllfHlk of surf(l(%, (md Interfaces. C:lrland STM Press, woo. M. t'rullon, Surface physics, 2nd cd., o"ro,.1, 1963. Good inlrotlu<:tioll. C. A. Somorjai, CIIl'ml$lry in hW dimcllsio'13. 'ur/aus, Camell, lOBI. Finc c)Cperimcntal introdlldl()n, wilh emphasis On adsorptiun or molL"""lcs un surfllCC<, T, AmID, A. D. Fowler, aml F. Stem, "Electronic properties of tw{).. tlimeru;ionalli~.,.tclTl5;' lIev. Mot.!. l'h)"1;. 5-1, 437-672 (100:2). II, C. Que> anti M. B. l'a.nish. IIcu:rosln.dun lours, .... c:atlt:mlc. 1978. II. K. llc"lilCl" SemialndtJdor c<mlllCl.8, Oxl"ord, 1964. Suoface Science- A JOUrnal.
A. z",lg'o"ill, f'1o~fia tit surfn, Cambridge, 1968,

1.

M. Blakely, cd, S"rfa;~ pI'f.'lilQ Of"'tll~riDfs,! vol,,, Academic, 1975 It E. l'nr.ngc and S. M. Cini", e.k, {>lIoOltllm 1/011 tlftc/, 2nd eel , Springer 1990.

M. Lannoo alld P. Friedel, AtOMic /III<! tlfXInmlc $ln.duu uf w.rfocu. Springer, 1~1 S. C. Da,uon and J. D. Levine. MSuof.'tC'C sUI,C' ~ Solid lilate ph)'$K:I!S, 1 (l970) l' Chakr:tbortyand P. Pio.InC:n. Frtlditmol quDnli'''' Half ~lfrCl. SPI';~. 1988. M C. Cr.OOn.! "ud F. M. Steranb, ~ugh'-em'tti"gdoodc. ~ Encyclopedia of applIed physics 8,
455-514 (19G-l).

C Julian Chen. Introduction 10 smnning IlInnrt"lfl mJcro'OJlY, OI.:IOnl, 1993.

20
Dislocations
SHEAR STREfliCTH Ot' SINGLE CRYSTALS 587 ~ip 588
DISLOCAT10l"'iS Burgers vectors Siren ftclds of disloc.tions J..oy,'-JUlgk grain botilKlarics Dislocation densities Didocalion multiplication and slip
smENCTH OF ALLOYS

589
592

593
595

598 600
600

DlSLOCATIO!'\S A!'\D CRYSTAL CROWru \\1liskcrs IIARDfliESS OF MATERIALS PROBLEMS

I. Lines of c1Qsc:st packing 2. DislOClltion pllil's 3. Forw on dislocntion

003 604 606 606

006 606 GOG

REFERENCES

W7

o-e---c;rG-Go
b

"'-'----j

i I~

"

~L__\'-,/-:----,,-,--.,--Di'l,bee,,,,,,,,

, , , , " " ,

i 0 , I O 0 , ,
, , , , , , ,
,

OJ

"
Figure I (:0) Hclah,e shear ,,(Iwo pJ.u>ell daloms (,h"",n in S<'dlon) In" untfOrml) slr.urlcc.l crystal; (bl ,heM ~lres.; as a functIOn of the ... .rati\e dls(lla~mcnl of the plan,'s from the'" cqlllhbnum (lO'ition The heavy llfOkn ]one ,1......."11 .It the in;I~,1 dupe dd'incs the she"" nkldulu. G

CHAPTER 20: DISLOCATIONS

111is chapter is ooncemed with the interpretation of the Jll.~tic lllcch,uucal properties of C'l')stalline soliJ~ in tenlls of the thC'Oly of disloc:.tion~. Plaslic properties arc irre\ersible defom1<ltions; clastic pro{X'rtics al"c rc\ l'.:..ble. '111c case with \\ htdl purc single l.'TyStals defonn pl,lstic;lll) is ~trikillg. 111i~ hltrinsic weakness of Cf)'stdls is exhibited ill \'arious \\'.1.)"'. PUle :.ilvcr chloric.le mdls .It 45S"C. )'et at room lemperature it has it ehccSt:hkc con!>lstenc)' and Com be rolk-d into sheets. Pure aluminum CTysl.l.ls are clastIC (follow l1ooke's law) only to a strain of abont 10 5. aft!;"r whidl the~' dl'furm plastic,llly. TIlcoretical e\tim.1-tes of the clastic limit of IlCrfl.'Ct. cryst:lls Kh'(' \..lIuc1> Iii' or 1~ higher than the lowe.t observed \'alllcs. .l.llhough a factor 10-" i~ more usual. There are few exceptions to the rule that pure c1)"itals arc pl<utic dlld J.re not strong: crystals of gennaniom and silicon arc llot pI.L..tic ,It room telllller;lturc and ('Iii or )"k-Id 0111)' b)' fracture. Class .It room telllpc...'tl1rc fdlls onl)' b)' fracture. but it is not cryst.tIline. 111C fr.l(;tun: of gI.l1>S is caused by :.trl."'Ss concentration at minute crneks.

SHEAR Sll\Ei'\CTII 01: SINCLE CIH'STALS

Frenkel gave a simple method of cstimdtillg the thcorctiClI ..hcar :.trcngth of a pcrfe<:t crystal. We CQusidcr ill Fig. I the force needed to u...J..e i\ ~hear displal'cmeut of two plallCS of i\tOlll~ pllst c.l(."h other. I'm SLllJ!I c1,I... tic strains thc stress a is related to thc displ.lcemellt .t h)'

a= Grid,

(1)

Here d is the illtcrplmHlr spacing, and G dCl1utt:s the aJlP1'tlpriat~ shC'lT modulus. When the displ,ll'Clllcnt is large alld has procc(.'(lcd to the Iltlint thai atom A is directl)' over atom B in the figllf(', the two planes ufatollls n.e in a l"Ouflgnralion of unstable eqUilibrium and the ~tress i~ zero. AS;I IIrst <.lpproximation we represent the strcss-displaccment relation b)'
(T

= (Cal2.1fd)

sin (211.110)

(2)

where a is the inte.-atomic spacing in the diret:tion of she.ll, -111is rdJtion i~ OOilstnlcted to reduce to (I) for small valucs of xla, 111e l'ritical slll'al" stress a~ .It which the lattke becomes unstable is ghcu b)' the ma"imulll v"lue of a. or
a~ =

CtvQml .

(3)

If 0 -- cl, then a~'" G127r. the ideal critical shC"Jr ~trcu is of the order uf ~ of the shear modulus.

(
obsen'dtions in Table 1 shw' the experimental "allle:,> of the elastic limit arc milch smaller than (3) would suggest. The theoretical estimate may be impro'lxl by collsidernlioll of the acl1411 IOml of the intermolecular forces and by collsideration of other- configurations of mcch.1nical stabili!)' 3\'ail3ble 10 the I,tnice as it is sllCared. Mackenzie ha.'I Sho\\1' that these two effects I1la) reduce the theoretical ideal shear strcugth to abOllt GI3O. corresponding to a critical shear straill angle of about 2 degrees. The observed low "allies of the shear strength can be explained onl)' by the presence of impcrfcctjolls that can act as sources of mechanical ,vcaJ.;ness in real crystals. TIle movement of crystal im~ perfections called dislocations is r~pol1!>iblc for slip at ,'cry low applied stresses.
nlC

Table I

Q:nnpariSQI'J 1) shear modulus and e1lU'tic limit

.:IastiC limit 0c,

ill d)'Iv'a,,

SII, sillgle crystal Ag. single crystal AI. single crystdl AI, pure, pol)'tt}'stal
AI,

1.9 x 10"
2.8 X 1011

1.3 X 107

15,000

commercial drawn Duralumin Fe, 5011, pol)'Cl')'Stal IIcat-treat.:.'\! carUoIl steel Nickel--cluoomc sleel
Aftt.'r Moll.

2.5 X lOll 2.5 X 1011 -2.5 X 1011 -2.5 X 1011

7.7 X lOll
-8 X 1011

6Xlotl 4 x lotl
2.6 x 1cjl 9.9 x lcjl

45,000 "',000 900

-8 X 1011

3.6 x 1.5 x 6.5 x 1.2 x

If! IOV

""
70 500 65

lrP
1010

,'"

Stip
Plastic deformation in crystals oCCl1rs by slip, 1111 example of which is shown in Fig. 2. In slip one part of the crystal slides as a I1nit across lin adjacellt 1"1rt. 11lC surface 011 which slip takes place is kno"1) as the slip plane. The direction of motion is known as the slip direction. 'J11C grC:'lt importance oflaUicc properties for plastic strain is indicated by the highl)1 :'lllisotropic natllre of slip. Displacement lakes place along crystallograllhic planes with :'l set of small Miller iudices, sudl as the {lll} planes in fcc metals aud the {lIO}, {1l2}, and {123} planes in 1x.'C metals. 111e slip dircction is in the linc of closest atomic p.'k:king, (llO) in fcc mctals and (Ill) ill bee mctals. To maintain the crystal structnre after slip, the displacemcllt or slip vector mllst eqll,,1 a lattice trans!:ltioll vcctor. l1)C shortest I:lttice translation Vct.-tor- c;o;pressed in tcrms of thc I:lttice constant 0 in a fcc structure is of the fonn (tI12Xi + 5); ill " bee strudllrc it is (012)(2 + Y+ i). But

l'igu", 2 Transiation;rJ slip in

7.il1C

single Cf)"'stah. (E. R. Parker.)

in fcc crystals onc also observes partial displacements whidl upset the regular sequence ABCABC of closcst-padied planes, to produce a stacking fault sl1ch as ABCABABC The result is then a mixtme of fcc and hcp stacking. Ddomlation b)' slip is illhomogeneotls. Large shear displacements occur on a few widely scparnted slip planes, while parts of the crystal lying between slip planes remain essentially undefonned. A property of slip is the Schmkllaw of the critical shear stres.'i: slip t.'1kcs place along a givcn slip plane and direction when the (.'Orrcsponding component of she... stress reaches the critical value. Slip is olle mode of plastic defoml:nion. Another mode, twinning, is obscrved partkularl)' in hcp and bee stmctures. Dming slip a collsidcrnblc dispbcement 0CC't11'S on a few \vidcl)' separated !>liP plancs. DuriJlg twinning a partial dispL"lCCmcnt ~lrs successivcly On each of many ncighboring cf)'5tallographic planes. After twinning, thc defOlmed part of the crystal is a mirror imagc of the I1l1dcfonll(.'t! part Although both slip and twinning arc l"3USed by the motion of dislocations, we shall be concerned primaril)' with slip.

The low observed \"alues of the critical shear strcss are explained ill tCllllS of the motion thrOllgh the bttice of a linc imperfection known as a dislocation. The idea that slip propagates b)' the motion of didocations was published in 1934 indepcndcntl)' by Taylor, Orowan, and Polanyi; the concept of dislocations was introduccd somewh:lt earlicr by Prant!tl ami Dehlinger. There arc SC\'cral b:lsic l}1>CS of disl(l(.':ltiolls. We first descrihc an edge dislocation. Figure 3 shows a simplc albic crystal in which slip of onc atom distance ha.'i 0ccurred o\"er the left half of the slip plane but not o\'er the right h.1.If. TIle

'"

-- .---

--

Figu", 3 An edge d..kx::otion fF III the gt.dc pbne ABCD '11", ll~"rc ~ tht: dipped region "8,.- WI "iudllhe atoms 1.""1: been di>~ b) mort: .Ioan haIfa a..tllCe COI1iUnl:llld II", unshppro rq;oo n;CD Ylilh wSplan:mcnl ]"u than half;ll WttlCe OOIlslallL

boundary between the ~lipf>ed and uuslippcd regions h culled the di.~locatiol1. Its po~itioll is m.ukcd by the tcmlinatioll of all Mira vcrtical half-plane of atoms CTo\\uccl into the upper Ihllf of the crystal as shoWJl in FiK. 4. NCM the dislocation the crpt.,1 is highly strained. 111C simple edge dislocation cKtends indefinitel)' in the slip plane in a direction 1I011l1allo the slip direction. In Fig. 5 we show a photograph of a clislocatioll in a two-dimcnsiollal 50:11) bubble raft ob

tained by the method of Bragg and Nyc. The mcch:\Ilism responsible for the mobility of a dislocation is showll in Fig. 6. The motion of an c<lge dislocatioll through a erys!.ll is :lIMlogolls to the passage of a ruck or wrinkle :lcross a rug: the ruck moves more easily than the \\ hole rug. If atoms Oil olle .~ide of the slip planc arc llIovcd with rcspcct to those on the other side, atoms at the slip plane will expclience repulsive forccs from somc neighbors and attractive forces from othcr.'; :l(.-ross thc slip pkmc. Thesc fon.'es c:lnce! to a first approximation. 11lc extemal stress required to 1110\'C a dislocation has been calcnlatcd and is quile small, probably below 1O~ d}'n/cr:n 2 , provided Ihal the bonding fOrces in the crystal arc !lot highly dircctiollal. 'f1l11s dislocations may make a <:rystal "cry plastic. PasS.1ge of a dislocnlion tlu-ough a crystal is equivalcnt to a slip displaccme!lt of onc p.1rt of the crystal. l11C second simple type of dislocation is the sere'w dislocation. sketcht.-d in Figs. 7 and B. A screw dislocation marks the ooundary bch\'CeIl ~Iipped and Iinslipped parts of tile CT}"Slal. 11lC bOIlndary '){Irallcu the slip direcliOlI, instead of I} ing peTJ:lClldicular to it as for the edge dislocation. The screw dislocation may be thought of as produced b}' entling the crystal partway through with a

2Q

DiJfKOIio...

figure"

Stn>l~ur..

uf all

edt...

dn.1oo.,,"iorl.

'11>e ddonn.otion n,a,' be thoogbt of a~ (";llJK'd

by iO$Cl1.ing ~n extra pL'lle of ~tOIl1' on the upper half of lile II Ui,. Atom, ;n the upper l"'lf....ystal ,Ire ro"'lll"'~ by lhe in>crtion, thoie in Ihe lower 1",lf are cd<'II<k-d.

"

"

.,

fIlil'
fip", 5 A dl.>location 111 a N.~imcn,ional bubble raIt. 11>1.' ,hdocahu" i. I'XW h' S<."<'T1 I" tummg the page by 3ft' III Its p1aue and sighlin,ll;;,.1 a low ...ide- (W. \1. wII.er ..It r 1l,~ ....ld N,.....)

Figure 6 Mohon of;l UI.Inc:ohon under a .hev wndmg 10 move the ullp"r iUrf,ICIC of the ~-peamen 10 Ihe ngllt (D. lIull)

"
o,
c

--

, , , , , , , , , , ,~ , "t , , , I , , , ,
~

, , , ,
~

~
~

Figure 7 Ii. 5t'reW dislcc:ation. A jW't ABEI' oI'lhc slip pbne has dillped In the di"d.... par.oIkI to the: dislOC\lioo line EF. A $CR''' dislocation may be "isualizrd :as ;l. helical arrangement of IaUia: p1",,~s, ..,eh lhat we rnan h "" pl:;Uleil on J:Olng romplcteJy around the dlSloc:Mioo line. (A[~r Cot1...,11.)

..
I

, ,
,

I-1pJre 8 Another \iew ~ a KTew dislocation. 1lle bn:Jkn \CI"t.cal Ime that marks d>e dlSloca.tion is JlJlTOUnded by slr-uned maim:..!

knife ani! shearing it parallel to the edge of the a1t by one atom sp."k:ing. A screo.v cli~IOC'ation tr:msfoflns succcssh"c atom planes into the surface of a helix; this accounts for till: ll.1.mC of the dislocation.

Burgers Vectors
Other dislocation [orills may be constructed from segments of e<lgc and screw dislocations. Burgers has shown that the most gCllcr'dl form of a linear dislorntioll pattern ill a Cf")'stal c:l.ll be described as shown in Fig. 9. We <''Ollsidcr any closed cun:e not necessarily planar ,v:it!lin a crystal. or an open OIT\'C tCnllin:tting 011 the Slllface at both ends; (a) Make a cut along all)' simple SIIrf.'K'e bounded b)' the line. (b) Displace the material on One side of this surface b)' a ,'ector b relath'e to the other side; here b is called the Bur-gcrs ",ector. (c) III I'cglons where b is not parallel to the ellt surface. this relative dispL'\CCfllcnt will either produce a gap or ClIIlSC the two halves to overlap. In these cases we imagine Ihal wc either add nldtcrial to 61llhe g.1p or Sllutrnct material to pre,'Cllt o,erlap. (d) ltejoin the matcri~\1 on both sides. We leave the st~i11 displacemelll ill tact at the time of rewcldillg, bllt artcrw:ll'ds we allow thc mediulll

Fiu~e 9 GcnenI method offQnning:a ilisloeal:ioa rinK in:a ~um. ~ medium IS n:praen..,d by ,b.., redaugubr block. The ring is ,"cpuwnkd b) die closed eu ....c in tbe interior in !he bb:k. A rot is made along the surfAc.e bournkd by the curve and indicat..' ti by the contoured an.-a, The materl:,J on One ~itk or the cui is .lisplacetl rdati,-e to that on the olncr by \'\...:1or di~talK1l' b. ",hid, mar 00 oricnlt,<J I,bitrari]y n.:latil'c 10 tile su,.faCt'. Fom:li "'ill be n:quired to CrrL"Ct Ihe di~plll(.'(' menlo nit) metlium is fiU..,,1 in or cut a"'3} so ~ to be ....... ,IllUOl.lS aIter the diiIPlacefl\f:Jll. It is til"", joi~'l! in lhe displ.ced ~Ic and 11..., apphed force. are rela..... -d. lie..: b is Ihe Burg.:n ,,-'dOl' dille dislocation. (Aftel' SeitL)

come to intental equilibrium. 1be resulting strain pattenl is that of the dislocation characterized jointl)' b)' the boundar)' cun'c :l.ncllhe Burgers vector. The Burgers vector mllst be Cljl1allo a lattice vector in orcler that the rewelding process will maintain the CI")'Stallinit)' of the material. The Burgers vector of a screw dislocation (Figs. 7 and 8) is pamllcl to the dislocation lille; that of;U1 edge dislocation (Figs. 3 and 4) is perpendicular to thc dislocation line ami lies in the slip plane.
10

Slrcss Fiel& of Dislocaliolls

The stress lIc1d of a screw dislocatiOIl is particularly simple. figure 10 shows a shell of material SIllTollIlding an axial screw dislocation. The shell of circtlmference 211r has been sheared by an amount b to gh'c a shear strain e = bI2'1ff". The corresponding shear stress in an el-tstic continuum is

Cc = Cb/21Tr ,

(4)

This cxpression cannot hold in thc rcgion immel!iatcl)' around the dislocation linc, as the straills here arc too brge for continuum or linear elasticity theory to appl)'. 11lC c1:lStic ener~w of the shcll is dE~ = iCe 2 dv = (Cb 2/41T) dr/... per unit lCllgth. TIIC total clastic ellerg)' per unit lellgth of a screw disl~tion is found on integration to be
E~=--I,,-

Ch2
411

ro

(5)

1.t
b

~ I

i jY

J1\:.

I I
I
I

I I

-{
rogu~ 10 shcU

...~..

d;~kx"IJOIl "~Ih

of clasti.:",!}y dl5lor1ro eryStaJ slIrroond",g Burgen .L"C't.... b; I.ee aho FIg. 16.

wllt'rc R and TO are appropriate upper and lower limits for the \"aTiable r. A fc:\SOllablc \'a!ue of ro is comparable to the magnitude b orthe Burgers vectoror 10 the ldttK..'e C'OfIrtanl; the \-aluc of n cannot (>xceed the dimensions of the l-ry~tal. l1le \":llne of the ratio IUro is not vcr)' important because it cntcl"5 in a
~'\rithm tcnn.
UH

\\'e now show the form of the energy of an edge dislocation. Lei u"" 0111.1 denote the tClisile stresses in the radial and circumfcrcntiol! diJ'C('lioIlS, and lei Urlj denote the shear stress. In 311 isotropic elastic continuum, u, ~ ;1m] UtJIJ are ploportiollal to (sin (Jyr: we need a futlction (holt falls off as IIr :md that changes sign when !I ;5 replaced b)' -yo "J11C shc.1r stress u,fJ is proportiunal to (cos {fj/r: collsidcring the plalle y = 0 \\Ie sec from Fig. 4 that the shear stress is .m (xld fllllction of x. The constallts of proportionalit)' ill the stress :lre proportiollal to the shear moduills G alld to the Burgers \'edor h of the displacClllellt. Tlle filial result, which is derived in books cited in the references, is
u~~ = l'Too =

-2n(1

c--;G::I=-'--:- .~irl 8
JI)

l'TrlJ

Gb
2n(1 - JI)

(:os
r

(J

(6)

\\hcre the Pois!>Oll ratio JI- 0 3 for most cr)'stdls. The strdin energ)' of a unit length of edge dislocation is
Gb l R E = -.:::''---,- In . ~ 4r.{1 - JI) ro

\\'e want 3n c",pression for the shear stress componcnt Uqr on planes parallel to the slip plane in Fig. 4, From the stress components u,~, UH, dlld u... c\'3Iu:lted on the pl311e 3 distallC(' y .tbo"c the slip plane. we find

u,,""- 2n(1

Gb
JI)

(0'

Figure II (o} r.o..-"lIlclc grain boundarY. after lJu",('n (O' Fledmn 1TI1<'1'Vg1~1'11 "I:a 100>-(\J,l/k t;r."1\ hOl.md.iry In mol) I>do;;num. '(be thn..e disloc:uion. ,n the i"13r:c rad. h... " d,c o;I.lIll~ Il".}(..... , ..."l'tllr R5 in the d......'ug III FUlllr" II". The ,,1nl<: arc.... "'.... ~ dOl' pO:flIKll'S of :l!omK' culumns norm:llio I'" plane oClhe paper E3ch "rr~) l"II.Ie. dclinc. IhI: l~'h ..." nf ,.di,locatlon, ",tlo mllr circln on the lop of each /lrray and tl,ret: c; .. Io.:. hc"'w. Clo:oiure ra,l"re is i.\d'ulctlll)' the u,,,, "I..ch c.ldlilc the llulgcn ,ecton. (Court",,)' of 1\ ClOn.l",)

or

It is shown ill Problem 3 that the force ('.H1sccl br a resolved uniform shC':!r stress u is 1" ::= va per nnit length of dislocation. '111C force that ,'u ct.Ige clisloc\ltion :'It the origin exerts upon a similar Olle :It the IOC,llion (y, 0) is

Cb l
F=
bUry

sin 10

== 217'(1

v)

4y

(9)

per Hllit length. Here F is the compOllcnt of force in the slip dilcttioll.

LollHllIgfe Crain Boundaries

Burgers suggested th<lt 10w-.lIlgle bound<lries between :ldjoining l'r)~t<\l lites or crystal gr:\ins consist of arra)'s of disloc:ltiolls. A simple e1\alllple of the Burgers mOOc1 ofa gmin boumlar)' is shown in Fi~, 11. Th" boundary occupies :1(010) pl<lne in a simple cubic lattice and divides two p.1rts of the cl\'stal that have a lool} axis in common. Such a bound'lI}' is called a pure tilt boundary: the nusoncntation l-:l.1l be dCSlTilx>tl br a small rotation 0 about the common [DOl] axis or one p.'\rt of the crystal rcbti\'e to the other The tilt boundar)' is rcpresented as an array of edge dislocations of spacing D = bIB, "here b is the Burgers "ector of the dislocations. E"p'l imcnb h.\\ e subst,lIlliated this lUodd. Figure 12 shows tile distnbl1tioll ofdlslve,llions along ~llI..lll-clll~lc .,-ain hounda riC'S. as obscrv~' illl all electron microscope. FllItlll'r, Head .tnd Sll(xklc)'

Yogurt: 1.2 Electron microg;r3ph of didocation stnlC"lUrel' in Jo,.~angIe grain boundaJilos in an Al7 percent Mg solid solution. N(lIIoe the lines 0( small dots 011 the right Mag. X 17.000. (It Goodr~h and G. Tbornao.l

derived a thcol)' of the interfacial energy as a fundion of the angle of tilt. with rc~ults in excellent agreement with measurements. Direct "'erilkation of the Burgers model is pro\'ided b)' the quantitative x-ray and optical studies of low-angle boundalies in germanium crystals by Vogel and co-workers. By counting etch pits along the intersection of a lowangle grain boundary with an etched germanium sUl'f."lCC (Fig. J3), they determined the dislocation spacing D, They as.~ullled that each etch pit marked the end of a dislocation. The angle of tilt ca1cuhlted from the relatiOIl 0:: bID agrees well with the angle Inea.~ured diroctly by means of x-rays. 111e interpretation of low-angle boundalics as ar'llIys of dishx.'aliolls is further SUPI>oI"led by the fact that pure tilt boundaries move normal to themselves 011 application of a suitable stress. The molion has been demonstrated in a beautiful experiment. Fig, 14, 111c specimcn is a bicrystal of zinc containing a 2 tilt boundary \\~th dislocations about 30 atomic planes apal1. Oloe side of the crystal was clamped. and a force was applied at a point on the opposite side of the boundary. ~Iotion of the boundary took plaC(' by cooperativc motion of the dislocations in the array, each dislocation moving an equal distance in its own slip plane. The ll1otion \\lIS produced by stresses of the order of magnitude of thc yield stress for zinc crystals. strong C\idencc that ordinary dcfonnatron results from the motion of dislocations. Grain boundaries and dislocations offcr rclativcl)' little resistance to diffusion of atoms in comparison with diffusion in perfect crystals. A dislocatioJl is an

20 Vislomlioll$

50

I.)

1-';gun: 13 DIslocation etch pils ill lo>-'-angle IXlull<lar)"JIl (IOO) face of germanium, the allgle of the boundar)' is 27.5", The boundar)" lies in a (011) plane; the lUll' of lire .Ii.10C\itions ,s [100]. The ilurger< "t'Ctor is lbe shortesllallice tl1lnslatiocl 'ector, or II~ = oIV2 = 4.0 A. (1-'. L Vogel, Jr.)

:...~

_.------

-.

1-'igure J4 Motion aIO\'o'-angk- grailllxlllllda", ",I<k.,. s1n:ssbouoolll) il; the ~Iraill '-ertidll 1ille, amI il is pholographed under \'ertidll Illuminatioll, theICh, makitl~ C''ident the 2" all~lllar change in lhe dca' age surl':lre of lhe ine cr~~lal at the IXlUntlary. 11'e i..egllbr hori:r.ontal hne is a small step in 1M r:k"ll'""&C slll'f:on: which se n as a rt-l"cr't-"I'I('O:' mark. l1M"r:n'$Iall' clamped althe left, at the right it IS ~ubject 10 a ~ noru llo tJxo plane dlhr page. Top. origmal position cJ I:ooomdary. boftom. ~~ '-l0.4 mm 0 \\ashburn and E. R. PMkcr,)

or

nw..

'"
open pa5s.a.gc for Jiffw.ion. Dilfll~ion IS grCfllcr in pJ.~IiC".lll~ dcf(ll Illed nMlcri.d

than in annealed (r)"Stdls. Diffusion a10l.1K gruill boundaries controls the ",lies of
soille precipitation reoc"lions in solic.b; the precipitation of tin from lead-till soluhons JI rOOlll temperature !>!"OCC(.'()s aooul Hi' times fasler th.U! t''l:pccted from diffusion in an idC<l1 lattice.

Dislocation Densities
TIle densit) of dislo(",tions is the number of dislocation linMi that intersect a unit area in the cryst.l1, The dens.l) rangc~ from wclillelow IOZ dislocations! em! in the best I*rmanium and silicon ('f)~tals to 1011 or 10 '2 dislocations/em l in 11ea\ i1y deformed metal cry"Slals. -nil..' methods d\'ailahlc for estimating dislocation densities are COlllpared ill Table 2. 111C .lCtU,I) dblociltion l'on6gurJtions in cast or anne.t.led (slowly cooled) (''')'Sla!s corrcspoml either 10 a group of
low-angle grain boundaries or to a three-dimensional nct\\ork of disloe.iHons

amUlged in cells. as shown in Fig. 15.

lattice \"aciludcs that precipitate along an existing ed~c dislocation \\ ill eat away a por1iOn of the extra hair-plane or atoms and cause the dislocation 10 climb, \\ htl.'h means to ffiO\C "t right angles to thc slip dirl,.'(tion. Ir no dislocations are prL"SCTlt, Ille cr)'stdl will become SlIpcTSi\tnmted \\ ilh lattice \'a<'andc~; their prffipitation in pclilly-shaped \3CanC)' pLltl'S 1ll,Iy be rollO\wd by C'Ollapsc or the plates and rormdtion or llislocation rings t!J:H grow with rurthcr \"dc.uIC)' preciplldtion. as in Fig. 16,

TaMe 2 Methods for c:stinJatillR tlislocalioil

dC:ll~rtic:s

Width
Tccluuquc
S(X'Cimcn lhldult'SS

"l...~imu,"
pr.l<'hC"al

of Image

llens.ty, per em! 10 11 _10 12

l()l-IOS 2 x 101 ,I x lOS
Ht-IO~

E!t.c:tron micrD!i(.'opy X-ray transmIssion X-my rt'fleclion

Dec.:omlion Etch pils

W. C. Johl'\l;ton,
bUm.t

>1000 rt 0.1-1.0 nlm <2pm (mill.) - OOPIll (m."1.l.) -lOplll (depth offoctls) no limit

-100rt Spill
2JJ.1ll

O.Spm O.Splllb

rX resQ!ulion rX etch p.b.

"-ogu~ IS Ct-ll (P R. S....aon.)

~lnJdure

of

Ihrcc-<I~ional

w ..~ of d"Ioc.'ho,,~ i..

d.IOn,.....d :uumi..ulll

rtgtre 16 Electron m~ of dr!.1ot-al1OU loor;lS bmt'tl b). "IQ;I".,....tion ".lid roI1ap5e of \oIl"3n'n AI-.5 pt..",..t \Ip; '11Jt'ndx.od rrorn 55O'C Tbr Ikl""j diskxalionlo Mt:' k>o:ntl L) \.tIC2nC') con&nsahon on.!ICft'W d.slocat... \Iag. x.a3.000 (A. [i'<urn a.d C n"",W>o)
COl."'I

DislOation l'tllllti,"icolian and Slip

Plastic deformation causes a H~ry great incre.'\SC in dislocation density, t)'pall)' from lOS to about lO" dislocatloos/cm' during defonTlation. Ifa dislocation mO\'es completely across its slip plane an offset of one atom spacing is produced. but offsets up to 100 to 1000 atom spacin~ are observed. This means thai dislocatioos IlUlltipl)' during deformollion. Consider a closed circular dislocation loop of radius" surrounding a slipped area having the radius of the loop. Such a loop will be partly edge. parlly screw. and mostly of inlenTlediate character. The strain eocrgy of the loop increases as its circumference. so that the loop \~ill tend to shrink in size. Howe\'Cr. the loop will tend to expand if a shear stress is acting thai fa\'or.; slip. A cammer. feature of all dislocation sources is the bowing ofdislocations. A dislocation segment pinned at each end is called a Frank-Read source, and it can lead (Fig. 11) to the generation of a large number of concentric dislocations 00 a single slip plane (Fig. 18). Relaled types of dislocation multiplication mechanisms account for slip and for the increased density of dislocations during plastic defonnation. Double cross-slip is the most common source.
STREJ'oiGTH OF ALU)\'S

Pure crystals arc \'ery plastic and yield at very low stresses. l1\Cre appear to be four important ways of increasing the yield strength of an alloy so that it \\ill withstand shear stresses as higtl a.~ 10 2 G. 11Ie)' arc mechanical blocking of diJ.localion motion. pinning of dislocations by solute atoms. impeding dislocation motion by short-range order. and increasing the dislocation density so that tangling of dislocations results. All four strengthening mechanisms depend for their succes~ upon impeding dislocation motion. A fifth mechanism, that of removing all dislocations from the crystal. may operat(. for certain fine hairlike crystals (whiskers) that al'C discussed in the section on crystal growth. r-.lechanical blocking of dislocation motion can be produce<l most dircctly by introducing liny particles of a second phase into a crystal laltice. This process is followed in the hardening of steel, where particles of iron carbide are prccipitated into iron, and in hardening aluminum, where particles of AhCu are precipitated. The pinning of a <lislocation by particles is shown in Fig. 19. Tn strengthening by the addition of smalll):lrticles tlLere are t",,'O cases to be considered: either tile particle call be dcfonncd with the matrix, which requires that the particle can be tnwersed by the dislocation, or the particle cannot be traversed by the dislocation. 1 the particle canllol be cut, the stress necessary to force a dislocation between parlicles spaced L apart on a slip plane should be approximately (10) ulG:c blL . '111e smaller the spacing L, the higher is the )'icld stress u. Before parlicles precipitate, L is large and the strength is low. Immedaatcl)' after precipitation is

20

l)i.Ioc",um.

(;01

"

Figure 17 Fronk-Read m~"<:h3ni"n for mnhilllicallon of didocalioll<, showin~ s"c<;cssive slages in the ~('nernljon of" dlslocalion loop b' the Sl'gn"mt Be of" di!ilo,:~ltion line. 'the proress can I.>c rcprnt.:<1 indelillikh

~-II,\

." ~\

,,")1~;)
,

.. f

't~-"-'-rj~""",

"

Figu~ 18 A lrnnk-Rl."ad .!L,IOO'IIOll sourre in sil,con, d('<:<)r3lcd "ith copJll'r p1'eclpitates amI .'ic"t'<1 with infrarc<I illun"nation T"ouJlnplcte di'ihK.'llholl loops arc "isihl,,_ amllhe lhi ...!, ;m",rmost loop is n~>3r ~nmplctll)". (Afler W C Dash.)

L_
Figuno 19 Oiloloc-..hon pinned 1))' polttdes III ~, ...." llIe (F.Iocctro.. mKTtlf!1"ph by

C Tbotnas alld J W4!.bum.)

complete and lIlan}' small particles arc present, L is a minimum and the strength is a maximum. If the alloy is then held at a high temperature, some particles grow at the expense of others. so tlwt L irKTC.\SCs and the strength drops. liard intcnnclallic phases, such as refractory o\idcs. cannot be cui b)' disloc.\tions. 111C strength of dilute solid solutions is bclic\'cd to result from tIle pinning of dislocations by solute atoms. The solubility of a fOI1;,ign atom will be greater in the neighborhood of a dislocation than elsewhere in a crystal. An alom that tends 10 expand the crystal \\ ill dissoh'c preferentially in the expanded region near an edge dislocation. A small atom will lend todissokc preferentially in the contractt><l regioll ncar the dislocation-a dislocation olI"eis both expanded and contractcd rcgions. As a "csult of tllC affinit)' of solutc .\IOIns for dislocalJollS, eacb dislocalJotl will collect a cloud of associatcd solute aloms during cooling. at a limc whcn tIle mobility of solute aloms is high. At still lower tcmt>ernturcs diffusion of solute atoms effective!) ceases, and the solutc atom c1ou<1 becomes 6xed in the cryslal. When a dislocation movcs. le~wing Its solute cloud behind, the energy of the en stal l11ust increase. The increase in energy can only be provided b)' an increased stress acting on the dislocation as it pulls .\\\ay flOm the solute atom cloud. and so the prCl>CIlCC of the cloud strengthens the crystal.

'l11C paSl>agc of a dislOC-dtioll across a slip plJ.nc in pnre (;l)~tals doc~ not alter tbe binding encrgy .\Cross thc pl.me ,Iftcr the di~locatiun is gone, The internal ellell--ry of the crystal rem,lim undlfedcd. The S.lllle is true for random solid solutions, because thc solution is equally I.m<!onl .\CrOl>S a slip pbne after slip. :\1ost solid sO!lltions. hO\\ L'\'er, h.,\e short-runge oruer, AtOlllS of differeot species are not arrangcd at raodom 011 tilt: lallicc sites, !.lut tend to h,,\ e an eAcess or a deficienlO)' of p.1.irs of unlike atoms. 1111IS in ordcred .1IIo)s di)locations tend to mo\c in pairs: the second disloc-.aUoll reorders the 1re.11 disolder left b) the first dislocation. TIle strength of a crystalline material increases with plastic defonnatioll. TIle phenomenon is c.,lIed work-hardening 01" strain-hardening. 11le strength is belic\-ed to increJSe because of the illcre.~ed density of dhlocations and the greater difficulty of mo\ing a given dislOCltion acros) a slip pbne that is threaded by many dislocations. Straiu-hardcning fl"equcntly b cmplo)'ed in the strengthening of materials, but its mefulness is limited to low enonglt temperatures so that annealing docs not occur. An important factol" in strain-hardening is the total densit)' of dislocations. In most metals di)locations tend to form cells (Fig. 15) of dislncation-free ;lfea.\ of dimensions of the oroel" of 1 ~m. But unless wc can get a uniform lugh density of dislocations we cannot strain-ll:rrden a metal to its theoretical strcngth, because of slip in the dislocaHon-free al"cas. A high total density is accomplished by explosive deformation 01" by Special thcl"mal-Inech..mical tn>atmenls, as of martensite in steel. Each of the mechanisms of !>1rcngthening crystals can I.lise the )'ield strength to the range of 10- 3 G to 10- 2 G. All mechanisms begin to !.lrcak down al temperaturcs wherc diffu.~ioll can OCCur at an apprcci.\hle rate. When diffusion is rapid, precipitated particles dissolve; ~olule clouds dlin .llong with dislocations a.~ tlley glide; shOl"t-rangc order repairs itself behind slO\\'ly moving dislocations; and di~location dimb amI annc<lling tend to decl"case thc dislocation density. The resulting time-dcpcndcnt deformation is called creep. 11lis inc\'ersible Illotion precedes the clastic limit. "111C scarch for allo)'s for usc at vcry bigh tempcratul"cs i:s a search for reduced diffusion ratcs, so tbat the four strengthening mechanisms will sUl"\i\"e to high tempcl"atul"cs. But the central problem of strong allo)"s is not strength, hut dU(.1ilit),. fOI" failure is often by fracture.
D1SLQC.-\TIONS AND CRYSTAL GROWTH

In some cases the presence of dislocations may!Jc the eontn)l1ing factor in crystal growth. \\'hcn crystals al"c gtO\\ n in eonditiuns oflow supersaturation. of the Ol"der of 1 pcn:ent, it has been obsen cd that the gro'\.\ th ratc is cnormousl)' fastcr than tb"t mlculated for an ideal cr)"staL '111e actual gro\\1h rate is cxplained in tcrms of thc effect of dislocations on growth.

The theory of growth of ideal crystals predicts that in CI'}'Sl'tll gt'O\\1h from \'3por a supersaturation (pressure/equilibrium \'3por pressure) of tllC order of 10 is required to nudeate new crystals, of the order of5 to form liquid drops, and of 1.5 to form a two-dimensional monolayer of molecules on the face of a perfect crySL1.J. Volmer and Schultze observed gro\~1h of iodine crystals at vapor supersaturations down to less than 1 percent, where the growth ratc should ha\'e becn down by the faclor exp(-3000) from the rate defined a.~ the minimum observable growth. 11lC large disagreement expresses the difficulty of nucleating a new monolayer on a completcd surface 'of all ideal crystal. But if a screw dislocation is present (Fig. 20), it is nevcr necessary to nucleate a new layer: the crystal will grow in spiral fashion at the cilge of the discontinuity shown. I\n atom can be bound to a step more strongly than to a plane, The calculated growth rates for this mechanism are in good agreement with observation. \\'e expect that nearly all crystals in nature grown at low supersaturation will contain dislocations, as otherwise they could not ha\'e grown. Spiral gro\\1h patterns ha\'e been 0bserved on a large number of Cf)"Stals. 1\ beautiful example of the gt'O\\1h pattern from a Single screw dislocation is given in Fig. 21, If the growth rate is independent of direction of the edge in the plane of the surface, the growth l>auem is an Archimedes spiral. r = 08, where a is a constant. nle limiting minimum radius of CUlyature ncar the dislocation is detennincd by the supersaturation. If the radius of CUf\'aturc is too small, atoms on the cuf\'cd edge evaporate until the equilibrium cun'3ture is attained. Awa)' from the origin each part of the step acquires new atoms at a constant rate, SO that drMt = const.
WIii"ken

Fine hairlike crystals 01' whhkcrs have been observed to gl'Ow under conditions of high supersaturation wilhout the necessity for more than perhaps one dislocation. It may be lhat thesc crystals contain a single axial Sl..TCW dislocation that aid~ their essentially one-dirnensionai growth. From the absence of dislocations we would expect these crystal whiskers to have higll yield strengths, of the order of the calculated value GOO discussed earlier in this chapter. A single axial screw dislocation, if present, could not cause yidding, because in bending the eT)'stal tlle dislocation is not subjected to a shear stress parnllel to its Burgers \Wor. That is, the stress is not in a direction that can cause slip. J ICITing and Galt obsef\'ed whiskers of tin of radius -IO-~ em with elastic properties near those expected from theoretically perfect crystals. They obscn<ed yield strains oCthe order of 10-2 , whK::h correspond to shear stresscsoforder 10-2 G, about 1000 times greater than in bulk tin, confinning the earl} estimates of the strength of perfect crystals. lliooretical or ideal elastic properties have been obser..cd for a Ilumber of materials. A single domain whisker of nielrel is s11O\\TI in Fig. 22.

~:=,-.

I I II

"-0";-1'1'1
or

J';gurf' 20 De-.dopmellt 0( a ~piral ~kp prodUl...d by intcl"K'<'ti<Jf1 a SC"'W di.iocalioo, "ilh tbe sulfa<:<: C'1"slal as in Fig. 8 ":Xh ("\,be rE'JlI't'V"tli a mokoallc. (t", C tr:o.nk.)

or ..

Figu~ 21 l'twe rontrast microgr>lph of a he>.agonal spiral grov.th pallenl on a SiC crystal. TIle step height is 165 /+. (A. It \'erma.)

"';gurt 22 A niclzl ",hid.l"1' of doamek. 1000}. bell' in a loop (R. W De alots.)

IIAnONESS OF :\IATEIUALS

Thc hardnc.~s of materials is measured in scvcr:l1 ways, thc simplcst test for nonmetals being the scratch tcst. Substancc A is harder than substance B if A will scratch B but B will not scratch A. A standard scale is lISed for rcpresentative mincrals. with diamond. the hardest, assigned the value 10 and talc, the softest. assigned the value 1:
lO diamond C

9 corundum Al 2 0:J

8 topaz

A~Si04F2

7 quartz Si~ 6 orthoclase KAISi 30 s

5 apalite Cas(P04hF 4 lluoritc CaF 2 3 calcite Cac03 2 b'Ypsulll CaSO" . 211 20 1 talc 3MgO . 4St02 . H 20

111crc is great current interest in the de\'clopmcnt of matcrials o( great hardness, for example as films for use as SCnltchresistant coatings on Icnses. It is widely felt that the scale between diamond and corundum is misleading, because c1i:\mond is much, much harder than cOIundum. It has been suggested that one might assign diamond the hardness IS, with the gap between 9 and 15 to be filled in eventually by synthetic materials, such as compounds ofC and B. Modem scales of hardness, such as the VIIN scale. are based on indenter tests in which an indenter is pressed into the surface of the material and the size of the impression is mca!iurcd. The Vickers Hardness Numbers of selected materials are tabulated below, after conversion by E. R. Weber to units of CPa leN/m 2 ]: Diamond SiC
Si~4

Al20 3 B

we

45.3 20.0 18.5 14.0 13.5 1l.3

BcO Steel (quenched) Cu (annealed) AI (annealed) Pb

7.01 4.59 0.25 0.12 0.032

The data arc from the text below by

J. C. Anderson and others.

Problems
I. Linn ofclosed poeJ.;ng.. Show that the lines of closest atomic packing are (110) in fcc

structures and (111) in bee structures.

2. DiJocaUo,t ptlU$. <a) Find a pair dislocations equivalent to a toW oflattice vacandes; (b) lind a pair ol dislocations equivalent to a row interstitial atoms.

or

3. Force on dislocation. Consider II. crystal ill the iOon of a cube of side L oontaiJliJlg an dislocstion c:f Burgers vectOt' b. If the crystal '-~ subjected to a shenr ~tress u on the upper and lower faces in the direction of slip, show by considering energy balance that the force acting on the dislocation is F - ber per' ., lenb'th.

ooge

Refereflces
ELE~IEN'TAR

D. Hull and D. J Bac:on, Introdu.:Uon 10 dWocatiom, 3rd ed., Pergamon, 1984. C. R. Bandt, W. D. N~ and A. S. Tetelman. PrindpIa of engi...,.m", moItriclU, Pl'nllke 14I1, l!r.:t M. J. Whelan, worktd nQ7Iplt, In dillocotlo,I$, 1ll.Jlitute of Metal.., 1990. S. Amc1incb:, DiI"tCf obscroor1lln of Jl3locatioM. Suppl. 6 to Solid state ph)'sil.'1. "'-_"'le, 196-1. 1. C. .-\ndcrsoo and others. Muteriob Idmce. 3m tdltum, Van Nostrand Reinhold (UK). 19M.

AD\'....:\CED F. R. K ~abarro, Th~ory of uyslut di.slocotlotl$, [)o"..:r, 1981. J. P. l!llth lind Jens Lothe. T1,ro'V of d~tlonf. 2nd ed., Wilc)', 19112G, l1\Qmas lind ~l. J. C<.>rlngc, Tra""ntsBion elearon ",Icroscopy ofrt'Ulterial.!, Wiley, 1979.

21
AUoys
CENEIlAL CONSIOEI\AnONS 611
614 617 618 (;22 622
G24 624
628

SUBS'OTlJnONAL SOUD sou.rnONSIIUME-IlOTHEIn' HULES

OIWEH-DiSORDEI\ 'mANSFORMA'noro; Elementary theory onler

or

PHASE DIAGRAMS E\ltect>cs

TRANsrnON METAL ALLOyS Electrical Conducth-il)'
KONDO EFFECT

PIlOBLEMS
L SUJX:I"latlice lines in C",Au 2. Configunllional heat capacity

630
G30 roo

HEFEIUENCES

630

>----< 0 0 0
0

0 0

0 0

0 0 0

0 0

0
0

...

CHAPTER 21: ALLOYS

CEl'\ERAL

CO~SIDEnATIONS

The theory or the band structure or solids assumes that the crystal has tnlllsiationill illvarianee. But suppose tlmt the cr)'stal is composed or two clements A and B that occupy lit random the l'egnlar lattice sites or the structure, ill proportions x and 1 - x ror the composition A.B I _ r . The translational symmetl")' is no longer pcrrect. Will we lose the consequences or band theory, sueh as thc ('xistcncc or Fermi surfaces and orenergy gaps? WiI! insulators become condoctors bcca.use tile cncrgy gap is gone? We touched on thesc questions in the discussion or amorphous semicondoctors in Chaptcr 17. Experiml.nt illld theory agree that thc consequences orthe destruction or perfect translational symmetry are much less serious (nearly a1wa)'S) than v,"e expect at first sight. 1be \iewpoint of the effective screened potential orChaptcr 9 is helprul in these matters, first because the effective potentials are relath'ely weak in comparisoo with a rree ion potcntial and, second and most important, the differences between the effl.-'ctivc potCJ1tials of the host and the additive atoms may be vel")' weak in C'OInparison with either alonc. Alloys or Si UJld Ce or or ell and Ag are cbssic eXi'lmplcs or what we may l-all the relative ineffl."CIivencss or alloying. In any cvent, a low concentration or impurity atoms cannot havc much the effective potential VCr) that is cffect on the Fourier components Vc responsible ror the band gaps and ror the rorm or the Fermi sunace. (11\is statcment implies that tllC C's exist, which implies that a regular lattice exists. This is not an important assumption bec:lUSC we know t1Mt thermal phonons do not havc drdStic effects, so tbat lattice distortions described as rro7-cn-in phonons should not have drastic elfects. Irthe distortions are more serious, as with amorphous solids, the electronic changes can be signifICant.) It is tlue that an impurity atom will introduce Fourier components or U(r) at wavcvtors that are not reciprocal I:\lticc vectors, but at low impurity concentration such components are ncvcr large in comparison with the Uc , argu. ing rrom the statistics of random potentials. lnc Fourier components at the reciprocal lattice vectors C \\.ill still be large and will give the band gaps, Femli sumccs, and sharp x-ray diffraction lines characteristic d regular lattice. The conse<lllCnces alloying \\,11 be particularly small when the impurity element belongs to the same column of tIle periodic table as thc hmt c1elnent it replaces, because the atomic cores will make rather similar contributions to the effecth c potent

or

or

or

",------,--------,r-----,-----,

"

.Aloonic perttn.

'"

"

""

Figure 2 Resis!;...;!)' d a dbon\"rc<1 hinary alloy Qr COIII)('r and gold. The ...ario1tion of the residual
resi~tivlt) depends on the <X>mpl)Silioll Cu.Aul_. as r(1 - Ill. which is known as Nordheim", Rule ror \I disor<krcd alloy. llcre %(1 - x) is II mt'ASure oCthe dct:reC of tna.~hnum diwrdcr possible for II

gl" .... n value

Qr z.

(JoIJanswn a",l Linde.)

One measure of the effect of alloying is the rcsicllJul electrical resistivity, dcfinc..'<l as the low temperature limit of the resistivity. Here we must distinguish between disordered and ordered alloys. An alloy is disordered if the A and 8 atoms arc nmdomly arranged, which occurs (or a general value of x in the composition A.B'_r- For special \'alucs of :I, such as 1/4, 1/2. and 3/4 for a cubic slructure, it is possible for ordered phases to (onn, phases in which the A and B atoms form an ordered array. The distinction between order and disorder is shown in Fig. 1. TIle effect of order on the electrical rC!iisti\'ity is shown in Figs. 2 and 3. TIle rC!iidual resisti\'ity increases "'ith disorder. as discussed for amorphous materials in Chapter 17. The effect is shown in Fig. 2 for the eu-Au alloy system. When the specimen is cooled slowly from a high tempernture.

21

A/lO!Js

613

"

, "
~

"

I
\
,
I'

\
-. i
'00

'"
Figure 3 Effcd of ordered phases On the resistivity of a binary alloy Cu.Aut_r "1'e allo~s Itere Ita"e been annColled. "ilereas those in Fig. 2 bave beetl quenched (o:>oIed rapidly). The <;omposilions nfJvw residual res!>;ti"ity <;or ....spond tQ the mderoo <;ompositions Cu.,Au and CuAu_ ijohansson and Linde.)

ordered stmcturcs are fOflTIl.. ..Q at CuaAu and CuAu; these structures have a lower residual resistivity by vil1ue of their order, as in Fig. 3. Thus we can usc the residual electrical resistivity to measure the clfect of alloying ill a disordered structure. One atomic percent of copper dissolved in silver (which lies in the same column of the periodic table) increases the )-esidual resistivity by 0.077 p.ohm-cm. 1l1is colTesponds to a geometrical scattering cross section which is only 3 perccnt ofthe naive "pmjcctl.-d area" oftbe,impurity atom, so that the scattering efk-d is very small. In insulators there is no experimental evidence for a signiflcant reduction of band gap caused by the random potential components. For example. silicon and germanium form homogelK'Ous solid solutions. known as substitutional

'"
a110)5, over the entire colllplSition range. but the band edge energies valY continuously with colllposition from the pure Si gllp to the pure Ge gap. It is widely believed. however, Ihat Ihe density of stales ncar the lxlnd edges in amorphous materials is smudged out by the gross ab~cncc of translational symmetry. Some of the new slates thus famled just inside the gap may nol necessarily be current-cafT}'ing slates because they may not c~tend

throughout the crrstal.

SUBSTITUTIONAL SOUO SOLUTIONS-IIUMEROTHERY RULES

We now discuss substitutional solid solutions of one metal A in another metal B of different valence. where A arK! B occupy. at random, equivalent sites in the structure. lIume-Rothc....-y treah.. >d the empirical requirements for the stahility of a solid solution of A and B as a single phase system. One TC()uircment is that the atomic diameters be compatible, which mCllns that they should not dilfer by more than 15 percent. For c.'Wmple, the diametel1l are favorable in the Cll (2..55 t ) - Zn (2.65 A) alloy system: zinc dissolves in copper as an fcc solid solution up to 38 atomic percent 7jnc. The diameters are less favorable in the Cu (2.55 A) - Cd (2. 97 A) s}"Stem, where onh 1.7 atomic percent cadmium is soluble in copper. 111c atomic diameters referred to copper J.re 1.()4 for dne and 1.165 for cadmium. Although the atomic diameters may be favorable, solid solutions will not form when there is II strong chellliclIltcndency for A llnd n to form intermel:.\llie compounds," which are compounds of definite chemical proportions. If A is strongly elcctronegati\'e and B strongly electropositive, compounds such as An and AlB may precipitate from the solid solution. (This is different from the formation of an ordered alloy phASe only by the greatcr chemical bonding strength of the intennetallic compounds.) Altllough tile atomic diameter roltio is favorable tor As in Cu (J .02), only 611tomic percent As is soluble. The diameter ratio is also f.worable for sb iu Mg (1.06), yet the solubility of sh in Mg is vel)' small. The elc..>ctronic structure of alloys can often be described by the a\ erage number of conduction electrons (or ,alence electrttu) per atom, denoted b}' n. In the alloy CuZn the value of n is 1.50; in CoAl, n =' 2.00. Ch..'U1gcs in electron concentration determine structuml changes in mall) alloy systems. The pha.~c diagl'Ulll of the copperzinc system l is ~hown in Fig. 4. The fcc stmcture of pure copper (n =' I) persists on the addition ofzillC(1I ;:: 2) until the electron concentration reaches 1.38. A bee structure OC'CllTS at II minimum elec'The p'-esof Ulterest are usuall) dmok'CI 1') mrtaIIurguts '" CreekdJ.uacters, ,n the CuIn y(((lmp\ex U1bic<.'l'U d~atoms). e-(hcpl and 11 (kp), f" and 11d,Rer con$lt1erably m dtl ratio '111e mellOlng of the dl3l'adcn d~'P"lIds UII the wk')- $)"~k"".
s~.tcm"e ba-e ... (l'cc}.IJ(bttl.

... ",,,,,,, p....nt ",,.,

"00
,."

'"

,
,,+

,w

~
"

J"od
~
p + 1"1",<1

1~,JuI

"

,
-

\~ , , , l 6+3\
I
,

"+1"1,,,<1-

"".
00

.... - --p' p Ill'+ "

,.,. " .,..

I\'~

'" , "

,'

,"i

,
I,

..

'"

'"
00

,'.

W"oW>'

p< frer.t .. ',.,

I-'igurc'\l E'IUlhbriulll d;~g.."n of ph:JM:s III the <.'Opprrzin<: all", 'l'.lcm. TI,e a pha.e isl<:c; fJ and fJ' me be:<:. y IS" oomplex structUTf.'. ~ and 1] are both l... p, but" 1m. a do mHo nc"r I 56 :Illd 1] [or pure Zll) has da '" 1.66 TI,e fJ' pl\<l!;e i.I ord"r",l be:<:. by" hi<:h we ",cantllat "Wit of the C" ~l,.ms

tJC<:Upy SltcS On <lI,e .., s"blat!l.:e am] mosl of the Zll odoms 0l~'U1)~' Sltcs on .. "'-'>Ild '" soblaltKe that Intel'penelrales the first .ublall;~". '11,,, 13 ph"", IS dISordered hee: all)' .,te is <'~luany "',cl~ to be occupiro b)' a eu or Zn atom, alnlO:St irresl'ecli\e of "hat atollls are '" lh" ncighbormg SlIt'S.

tron conccntnllion of about 1048. The y phase exists for the appro'l:imnte rnnge of II between 1.58 and l. 66, and the hql phase E O:IUS near l. 75. The term electron compound denotes an intermediate pha~(;' (such as the f3 phase of CuZn) who~e crystnl strucltue is determined by J fairly well denned electron to atolll ratio. For many alloys the ratio is dose to the I [ume-Holhen' rules: 1.50 for the (3 phase, 1.62 for the")' phdse, and 1.75 for the E phase. Representative c'l:periment,11 values arc collected in Table 1, based on the usual chemical valence of I for eu ,md Ag; 2 for Zn and Cd; 3 for AI and Ca; 4 for h Cc, and Sn The lIume-Rothery mlc~ find a simple e'-pressioll in terms of the bm,d theory of nenrly free eleclrons. The observed Iim,t of the fcc phase occurs close to the e1ectr.Q11 cOllcentration of 1.36 at which an inscribed Fermi ~phere maJ.-es contact with the Brillouin zone boundar)' for the fcc Iattic'. The observed electron concentJ .ltion of the bet: phase is close to the concentration lA8 at which an inscribed Fermi sphere m,lkes contact with the zone boundilf\' for the bee

Table l

Eltttronlatom I"lllios of elcdron compounch

Minimum bee phase ooundary

Alloy

"" """"
1.38 1.41
1.41

boundary

CuZn CU-AI Cu-Ga Cu-Si Cu-Ge CuSn

Ag-Zn

...8
1.48

1.58--1.66 J.63-1.77

1.78-1.87

hg-Cd

Ag-Al

1.42 1.38 1.27 1.38 1.42 1.41

1.49 },49
1.50

1.60-1.63 1.58-1.63 1.59-1.63

1.73-1.75 1.67-1.00 1.65--1.82 J.55--J.80

laUice. Contact of the Fenni sphere with the wne boundary for the 'Y phase is at the conccntration 1.54. Contac't for the hep phase is at the concentration 1.69 for the ideal cltJ ratio. \\'Oy is there a connection between the electron concentralions at which a new phase appears and at which the Fermi surface makes contact ,,,ith the boundary of the Brillouin zone? We naU tllat the energy bands split into two at the region of contact on tile zone boundary (Chapter 9). If we add morc electrons to the alloy at this stage. they will han~ to be accommodated in the upper band Of" in states ofhigh energy near the wne comers of the lower band. Both options are possible. and both involve an increase ofenergy. Ihereforc it may be energetically favorable for the crystal structure to change to one which can contain a Fermi surlaet> oflarger volume (more electrons) lx...forc contact is made with the zone boundmy. In this way H. Jones made plausible the sequence of structures fcc, bee, )I, hep with increasing electron concentration. ~Ieasurements of the lattice parameter of Li-I\'!g alloys are sllOwn in Fig. 5. In the range shown the structure is bec. The lattk.-e contracts during the initial stages of the addition of I\Ig to li. When the lithium content drops belo\\! 50 atomic percent, corresponding to an average electron concentration increasing ~'c 1.5 per atom, the lattice starts to expand. We havc seen that for a sphcrical Fermi surface contact with the zone boundary is establishl.-d at n == lAS ek'C(rons per atom. in a bee lattice. It appears that the expansion of the lattice arisl.-"S from the onset of O'.erlap across the zolle boundary. 1 ht: transformatiOn from ft'C to bcc is illustrated by Fig. 6; this sllOws the number oforbitals per unit energy range as a function of energy, for the fcc and bee structures. As the number of ek'Ctrons is increased, a point is reached where it is easier to accommodate additional electrons in the Brillouin zone c:J the bee lattice rather than in the Brillouin zone of the fcc lattice. The figure is drawn for copper.

21

Alloy"

617

"

, , " , ,
~

3520 -

3510

li <

~
~

"'" ".,

Moo"", JI'""""t lith""

Figure 5 l.attice pardmeter of body-ecnlered cubic magncsilllll"lill,ium alloy$. (Artcr o. w. LevinSOIl.)

<
8.
0.30

1\<

~ .5:
"igore 6 Number of orbitals per unil energy range for thc firsl Brilloo'" ZOnC oflilc 1,:<: and !Jcc lattices, as a function of cncrgy.

t- 0 2l.l i
0.1

~ '" ,/
II
01234 :;

,/
"
E,>crgy in ,V

is

"

"

ORnER-DISORDER TRANSFORMAll0N

The dash(.--d horizontal line in the beta-jlhase (bee) region of the phase diagrnm (Fig. 4) of the Cu-Zn system represents the transition temperalure between the ordered Vow temperature) and disordered (high lempernture) states of the alloy. In the common ordered arrangement of an AD alloy with a bee structure all the nearest-neighbor atoms of a B atom are A atoms. and vice versa. 11lis arrangement results when the dominant interaction among the atoms is an altraction between A and B atOms. (If the AB internclion is weakly attractive or repulsive. a two-phase system is formed in which some nyslnllites are largely A and other crystallites are largely D.)

c
'18

1'lC .llloy is completely ordered at ab'>Olute zero. It becomes Ipss orocwd as the temperature is 1I1crcaseu, until a tranSltiOIl temperature is r{'ached above which the structure h disordered. TIle transition tempcmtUlc marks the disappearance of long.range order, which is order m'er many interatomic distances. but some short-range order or correlation among near neighbors rnay persist abo\'e'tbe transition. The long*rallge order in an AB alloy is shown in Fig. 7a. Long- anJ short-nuge order for an alloy ofcomposition A83 is gi\'en in Fig. 7b. TIle dpgree of order is de6ncd below. If an alloy is cooled rapidly from high temperatures to a temperature below the transilioll. a metastable condition may be ploduced ill which a lloncCluilibrium disorder is (roren in the structure. TIle reverse effect occurs when an ordered specimen is disordered at constant temperature by heavy irradiation with nuclear particles. TIle degree oforder may be investigated experimt>ntally by x~ray diffractiOn. TIle disordered structure in Fig. 8 has diffraction lines at the saine positions as if the lattice points were all occupied by only one type of atom, because the effective scattering power of each plane is equal to the average of the A and B S(.-atlcring powers. The ordered structure has extra diffraction lines IIOt possessed b)' the disordered structure. TIle extra lines are called superstructure lines. TIle use of the tenns ordt>r and disorder in tllis chapter always refers to regular lattk:e sites; it is the occupancy that is randomly A or B. Do not confuse this usage with that of Chapter J7 on noncrystalline solids where there are no regular lattice sites and the structure itself is random. Both possibilities, however different, occur in nature. The structure of the ordered CuZn alloy is the cesium chloride structure (Fig. J.2O). TIle space I...Uice is simple cubic. and the basis has one Cu atom at 000 and one Zn atom at Hi. TIle diffraction structure factor
S(hkl) = JCu I"

+ JZ" I: e

llo'(h+k+l)

(1)

This cannot vanish bcl-ause fcu hn; therefore all reflections of the simple cubic space lattice will occur. In the disordered structure the situatioo is different: the basis is equally likely to have either Zn or Cu at 000 and either Zn or Cu at HI. Then t!'le 3\'erage structure factor is
(S(hkm = (f)

+ (f), .",..,./) .

(2)

where (f) = Hfe.. + hill, Equation (2) is c:\actly the form of the result for the bee lattice; the reflections vanish when h + k + I is odd. We see that the ordered lattit'e has renecHons (the superstructure lines) not present in the disordered laltk-e (Fig. B),

EIJ!merltary Tlurory of Order

We give a simple statistical treatment of the dependence of order 011 temperature for an AB alloy with a bec structure. TIle case A3" liflers from AB, tlte

21

,\lloy~

~ 05

"

T~"'l'.,<.'h,,'-

--

(.

"

,,!------------.,,------'
r~"'.,.."'tu,,

--- ---

",

-_

Figure 7 (a) I...ong-r.lnge order \'cl"Sus lemp<'r<it"re fur all ,\8 alloy. "n,e Irllllsform.olion is scwnd order. (b) L""g-r.mge and s1Mlrt-range order rQr an A8 1 alloy. l1te tr<iIISrormalion ror thi~ OJmpl'lion is lirst order.

(.,

(b) ordered by lI11nealmg "I l'

Figure 8 X-my powder phOlogr<iphs HI AIICll~ allo). (3) Disordered by cluenl-hing from l' > 1~, < r,_ ((:"urtcsy of C. \1- C<>rdon.)

fOnner having a first-order transition marked by a latent h.... and the latter ha\'ing a sccond-order transition marked by a discontinuity in the heat capacity (Fig. 9). We introduce a measure of the long-range order. We call one simple cubic lattice a and the other b: the bee structure is romlX1Scd of the two inter_ penetrating lie lattices, and the nearest neighbors of an atom on one lattice He on the other lattice. If tht.'1"c are N atoms A and N atoms B in the alloy, the long-range order parameter P is defined so that the number of A's on the lattice a is equal to i(l + Ii'1N. 'I1m number of A's on lattice ,) is equal to i(1 - p)N. When P = 1, the order is pCl'fL>ct and each lattice contains only 0111..' type of atom. When P = 0, cadI lattice contains pqual numbers of A and B atoms and there is no long-range order. We consider that part of the intcnml energy associated with the bond energies of AA, AB, and BB nearest-neighbor t='Iirs. nil..' total bond energy is
E = NM,U IIII
ij bond.

+ NSS US8 + NIISUAB ,

(3)

where Nil is the number of nearest-neighbor ij bonds, and UII is the energy ofan ,11e probability that an atom A on lattice a \\ill ha\'e an AA bond is equal to the probability that an A occupies a partictJlar nearest-neighbor site on b, times the number of nearest-neighbor sites, whidl is 8 for the bee structure. We assume that the probabilities are inoc>pclxlent. Ihus, by the preceding expressions for the number of A's on a and I),
NAA : 8/1(1 + I')NJlI(1 - 1')): 2(1 - P')N ; N : 8/1(1 + I?NJlIII - I?J : 2(1 -I")N ; NilS = 8l\'{l{1 + PJ)2 + 8!\1l{1 - p)J2 = 4(1 + p2)N (4)

'f1le energy (3) becomes

E = Eo

+ 2NJ"lU ,
U = 2UAS - UM, - USB

(5)
(6)

where

We now calculate the entropy of this distribution of atoms. 'illerI..' are l{1 + p)NatomsAandl{l - I1NalomsBonlatticea;therearei(l- P)Natoms A and i(l + PJN atoms B on lattice I). The number of arrangements C of these atoms is
G: [ [I(L

N! - I')N)' + P)N)'[I(L

J'

(I)

From the definition of the entropy as S = k B In G, we ha\'e, using Stirling's approximation, S = 2/.;k s In 2 - Nksf(1
2.:.\'k s III 2.

+ PJ In (I + PJ + (I - PJ In (I - PJ) .

(8)

This defines the entropy of mi.. .ing. For P = ::!::I. S = 0; for I> = 0, S =

21

A'loy~

621

,." , " , I "

-~

."

"I-~i~

/It-

11/

.w

.,

"

" ,
Figure 9 Ileat capacity versus tcmpcral.un; of CUZ" alloy ep-bnl$S).

/ /1 -100 200 300

- -- '00 "C
500 600 700

"

l""'!",r'ture In

one equilibrium order is detennined by the requirement that the free energy F "'" E - TS be a minimum with respect to the order parameter P. On differentiating F ""ith respect to P, we have as the condition for the minimum

4NPU + NkBT In - - = 0

I+P I-P

(9)

111e transcendental (.>(juation for P may be solved graphically; we find the smoothly decreasing curve shown in Fig. 7a. Near the transition we may expand (9), to find 4,vPU + 2Nk B TP"'" O. At the transition temperature P = 0, so that
(10)

For a transition to (I(.'Cur, the dk-dive interaction U must be negative. 'Ille short-range order parameter,. is a measure of the fraction of the average number q of nearest-neighbor bonds that are AB bonds. When (.'Qm pletely disordered, an An allo~' lIas an average of four AB bonds about each atom A. 111e total possible is eight. We may ddine
r=!(q-4) ,

(ll)

so that,. = 1 in complete order and,. = 0 in complete disorder. Obsen,e that r is a measure only of the local Order about an atom, whereas the long-range order parameter P refers to the purity of the entire population on a gi\'en sublaltiee. Abo\'e the transition temperature T~ the long-range order is rigorously zero, but the short-range order is not.

."
PHASE DlACRAMS

There is a large amount ofinfonnation in a phase diagram even fora hinary system, as in Fig. 4. The areas enclosed by cunes relate to the equilibrium state ill that region of composition and ternpenlture. TIle curves m.uk the course of phase transitions as plottPd in the T-x plane, where x is the composition parameter.

The equilibrium state is the state of minimum free energy of the binary system at gh'en T, x. Thus the analysis of a phase diag.-am is the subject of thermodynamics. SC\'eral extraordinary results come Out of this analysis, in particular the existence of low-melting-point eutectic l'OmposilioIlS. Because the analysis has been treated in Chapter II ofTP, we only outline the principal results here. Two substances will dissolve ill each other and form a homogeneous mixture if that is the configuration of lowest free energy at'(.'essihle to the l'Omponents. The substances will form a heterogeneous mixture if the combined free energy of the two separate phases side by side is lower than the free energy of the homogeneous mixture. Now we say that the mixture exhibits a solubility gap. In Fig. 4 we see that compositions near Cuo ooZno 40 are in a solubility gap and are mixtures offcc and bec phases ofdifferent structures and compositions. The phase diagram represents the temperature dependence of the solubility gaps. When a small fraction of a homogenl..-'Ous liquid freezes. the composition of the solid that forms is almost always differcnt from Ihat of the liquid. Consider a horizontal section nt'ar the composition CUo ~no 20 in Fig. 4. Let x denote the weight percent of zinc. At a gi\'Cn ternpelature. there are three regions:

x> XL. the equilibrium system is a homogeneous liquid.

Xs

< x < XL.

x

<

there is a solid phase of composition Xs and a liquid phase of composition Xv xs. equilibrium s~tem is a homogencous solid.

The point Xl. traces a curve~alled the liquidus curve, and the point Xs truces the solidus curve. Eutectics. r-.Hxtures with two liquidus branches in their phase diagram are called euteclics, as in Fig. 10 for the Au-Si system. nle minimum solidification temperature is calk><:l the eutectic temperature; here the composition is the eutectic composition. 111e solid at this composition consists of two separate phases. as in the microphotograph of Fig. 11. onere are many binary systems in which the liquid phase persists to temperatures below the lower melting temperature of the l'Ollslituents. Thus AU06~io31 solidifies at 37ff'C as a two-phase heterogeneous mixture, although Au and Si solidify at lQ63C and l404C. respcctivelv. One phase ofthc eutec tic is nearly pure gold; the other nearly pure silicon.

21

MIOYI

,.., ,

..
./

,~r

"'"

K"""

R::::
""
,
Pure Au

...

/'

.-/

--, ((J

:\
,,

7
\.-31

""
60
71)

t<o

lOll PUk S.

CQme together at r~ .. 37ete and x.," 0 31 atomic PC"'Cll! Si. (After Kittel ;mel Kroclll"r, Tl'.)

FiguI'C 10 ""tcdIC phase ui;W<\Jll <i w.>ld-slliClln allovs 1bc cutcdic<-'OIl~lsboltwo b1'llllchc. that

\"

'1

'.11' .11'

, ,~rlfl.~",J,unfl~
"",

,,,'
'II
"6

,\\

"",

10 J.un

Figure II

MicrophotOgrdph "fthe l'h-Sn cutc<:t'c. (Courtesy of J. D. 111.1111 and K. A. J,U:!;s<.>n)

'"
TIle Au-51 eutectic is important in semil'onduetor tL>c)ml.. {,y ix.>cause the eutectic pemlits low temperature welding of gold contact wires 10 silicon de'ices. Lead-lin alloys ha,"c a similar- eutectic of Pbo ~no 14 at IS3"'C. 111is or nearby compositions are used in solder: nearby if a range of mchillg tempera_ tures is desired for ease in handling.
TRANSITIO;-'; METAL ALLOYS

When we add copper to nickel the efft.'Ctivc magnctoll number per alom dt.'Creascs Iint:arly and goes through zero ncar Cuo.60Nio.40. as shown in Fig. 12. Allhis composition the extra electron from the copper has flIk--d the 3d band, or the spin-up and spin-down 3d sub-bands thai were ShOWll in Fig. 15.7b. 'nle situation is shown schematically in Fig. 13. For simplicity the block dr.nvings represent the density of states as uniform in energy. "nlc actual density is known 10 be far from uniform; the result of a modem calculation is shown in Fig. 14 for nickel. TIle width of the 3d band is about 5 eV. At the top, where the magneticefft:cts are dt:tennined, the density of states is particularly high. 11le a\erage density of states is an order of magnitude higher in the 3d band than in the 45 band. This enhanced density of statl.-"S ratio gives a rough indication of the expected enhancement of the ek>ctronic heat capacity and of the paramagnetic susceptibility in the nonferromagnctic transitiOIl metals as compared with the simple monovalent metals. Figure 15 shows the effl.-'Ct of the addition of small amounts of other elt. ... ments to nickel. On tlle band model an alloying metal with % valence electrons Olltstde a filled d shell is expected to decrease the magnetization of nickel by approximately % Bohr magnetons per solute atom. This simple relation holds well for Sn, AI, Zo, and Cu, with.z = 4, 3, 2, and J, ."Spt..,ctively. For Co, Fe, and Mn the locali.. moment model of Friedcl accounts forefl.--'Ctke.z values of -I, -2, and -3, respt..>clively. . ne average atomic magnetic moments of binm)' alloys of the elements in the iron group are plotted in Fig. 16 as a function of the concentration of ek'Ctrons outside the 3p shell. 'Ibis is calk>d a Slater-Pauling plot. nle main set"juence of alloys on the right-hand branch follows the rules discussed in connection with Fig. J5. As the electron concentration is decreased a point is reached at which neither of the 3d sub-bands is entirely filled, and the magnetic moment then decreases toward the left-hand side of the plot. Electrical COI!ductivily. It might be thought that in the transition metals the availability of the 3d band as a path for conduction in parallel with the 45 band would increase the conductivity, but this is not the way it works out. l1le resistivity of the 5 electron path is incrcast..od by colli.sions with the d ek-ctrons, this is a powerful extm scattering mechanism not available wilen the d band is filled.

.oo

21

Al1ov-

it

,, '1"\ 'I ' ,

0.

, . oj ~ .- , , ,,
.no,..

n: ,

1"I'"

'"

,,

060 ,'1tn>oI

"

,,

"

[JJ[I]
._ 0

Figure 12 Bohr mlIgIIctoIl numbo-"D d nickel.oopprT

FiCUn- 13 DUlribuliOll of eledf'Olls in the a\Joy 6OC...wNi. The extra 0.6 elearocl PfO'ided by the lppCr hQ filled the d band elllirel~' and inen-tied slightly the number of clectrom in Ihr ~ b.nd \\ith Tnped 10 fo-ig. 15.Th.

,------------------" ."
-

A:'" \

-00

.....-,

,
0- . 0

.,
-,

"

J t'.

,'-, .'
lanD. and A.. R. Williams.)

fo'iI;unI! 14

Density of slates ;11 "ide!. (\'. L Moruui,

...
,
,, ,
0

...
, ,/ ,--, ,, //F~ ,, ,
/

,
,

A'/ ;/
/

,/

Co

,,

lI,k""" Ck""'''h In .l<lrn I,..,.~nt Figunl IS Saturation m3gneliution of nickcllllw)'S in !loh. rnllgnelom pcr atom as II function ~

:~~ ~ '-'" , , , '-, " " ,, 00 "

~ ,.."

,,-.\In

,/

:::--

----'"

------ -----

--,

'00

the atomIC percent of solute element.

",----r----,---,---,--.-----,--r----r--------,
@ F~V

"I---I--+---+--+-~d--i_-+-_I _ff'- . ~

+ Fr-C,
o Fr.:-i,

~"

J"

I----+--+--+---c.jf~

Ih/~. .,' ~

""

,..co !'>,-Co
u.
N..c.
N'Mn

.0. Non.
'I' ....

o
'0

v " ..v

I / 11 ""- ..---1 , ,,1-+-/-----1,?-+-+-+I--f.} ttf'--+---o-----P!> _ .... ! ll/'~ .... "I----+.L.+-----+---I-----,t-,. / +-+----1'1-1 71 I", i
A c<>-C.

e..-~h,

,L_..L_..L_--l_ _.L--'-...L_---l_ _"---_..L_---""'_ _

OJ

M.

",

Co

FipJre 16 A\li'n4;e al:Qlllic I'nomc:1I1$ ofbo'I:lI')' ..JI<'), of the demen15 in the ....... group (8o>:oIIh.)

We compare the values of tbe e1cctrk.11 resi~livitics of Ni. Pd. and Pt in microhm-em at 18"C with thai oflhc noble metals CU, Ag, and All imlllooi31ely following them in the periodic table:

No

Pd

I"

1.'

10.8

105

1.1

1.6

22

-nlC resistivities ofthc noble metals arc lower than those ofthc transition metals b} a factor orthe order of 115. This shOY."S the cffcdivencss of the s-d scattering mcchal\ism.

ro

oro

"

0"

ol----4c-+-f-----"'<-l-+---\

-"
-2110 3 6
~.,

12

F"u~ 17 Magnetization ofa '""'" t'leclrofl Fnmi p i at T - 0 ifllk'ighborilOOd "h point rnagncti<" IlXMllCflt at the origin r" 0, ac<:<.>rdi"ng to the IIKKY theoJry hoorillOaul axis i> 2k, ... where k.- u tl'e "l1\..,\'edor on the Fenn' sphere. (de ee"neli.)

n.c

KONDO EFFECT

In dilute solid solutions of a magnctk ion in a nonmagnetic melal crystal (such as Mn in Co) the exchange l.'Oopling between the ion and the conduction electrons has important conscqucl1(.'cs. 1110 conduction electron gas is magnetized in the vicinity of the magnetic ion, with the spatial dependence shown in 2 Fig. 17.11115 magncli7A'Itioll l,."aUSCS an [n<lircd exchange intcraction bchvccn two magnetic ions, because a sCl.'Ond ion pcn::civcs the Ilmgncti7~1.lioll induced by the first iOIl. The illteradioll, known as the Friedel or RKKY interaction, also plays a role in the mugllclic spin order of the rare-earth metals, where the spins or tile 4fion cores are coupled together by the magnetization induced in the conduction electron gas. A consequence of the magnetic ion-conduelion electron interaction is the Kondo effect. A minimum in the electrical rcsistivit)-temperaturc curve or dilute magnetic alloys at low temperatures has been observed in alluys of Cu, Ag, Au, Mg, Zn with Cr, Mil. and Fe as impuriUCl>, among others.
fA T'f:\ie\< .,(indired exchange interartion~ in metals is gwen b) C K,tle!. Sohd state ph~1CS 2t, 1 (I96S), " ...",ie"".,(tlle Kondo effect is gh~n 1. J Kondo, ~.,( d,lute ~chc a11oys,~ Sol,d state ph~ic$ 23. 1&4 (1969) and A J Heeger. -Loc:at.~ momenls and non,,\OIO>ent5 ,n rneW~ tI'e Kondo efk.ct. ~ So!;d Wle ph)'sics D. 2-&6 (1900). The notat...... IlKKr slanlb for Iludet"1'lWl, Kittet, Kauya, and Yosidi.

"
"""
"00
AuF"

AII,,~

""

'"" \

""
- 0 19(;

,
'i

'""\
O(J2at'loF"

""
0.190

",. 'i ,
no..
0

, J

0.07&

0006

"'"'
0''''
0

0.074

,
T. t.:

Figure 18 A l"OOlp~rison of experimental ~Ild Ihl'Oretlrnl results for the ill(.~ea';C of elco:;tTkal ,",-,sISti,oily 01 1m, temperatures In d,lule alk>)'" of iron III gold. The resistance minim"'" lies to the right of the figure, for the resistl>itv i"Crt'a5<.>S at high 1"mre"'IUl'" bemuse of 5Callering of d,,<;,tron~ by thermal phonon . 11,,) expcrnncnts are due to D. K. C. MacDonald, W. B. ('carson, and I. M. Templeton, the th..'O,y is by J. Kondo. An exact solution was given by K Wil""...

The oc-currcncc of a resistancc minimuill is connected with tbe existence of localized magnetic Illoments on the impurity atoms. \Vhcrc a resislan<,'C minimum is found, there is inevitably a local moment. Kondo showed thai the anomalously high scaHcrillg probability of magne(il; ions at low temperatures is aconsequence of the dynalllic nature ofthe scilttering by the exchange coupling and of the sharpness of the Femli surtace at low temperatures. The temperature region in which the Kondo effect is important is shown in Fig. 18. The central result is that the spin-dependent contribution to the resistivity

"
P.p,,,=CPM

3'J.] . =cPo-cp ln1, [1+--ln1 ',.

1

(12)

where] is the exchange energy; z the number of nearest neighbors. c the concentration; and PM is a measure of the strength of the exchange scattering.

,so
We sec that the spin rcsisti\"ity increases toward 10\\ temperatures if] is negative. If the phonon contrihution to the electrical resistivity goes as .(i ill the region of interest and if the resistivities arc additive, then the total resistivity

has the form

p = a"l{;

+ CPo -

CPt In 'f

(13)

with a minimum at
(14)
\\hence

(15)

"I11C temperature at which the resistivity is a minimum varies as the one-fifth power orthe concentration orthe magnetic impurity, in agreement with experi-

ment at least for Fe in Cu.

Problems
I. Super-lattice /incs ill Cuy\u. CuaAu alloy (75% Cu, 2'5% Au) lIAS .In ordered slate

below 400"C, in Wllich the gal<l aloms occupy the 000 (Xlsitions :llul the Ct.Jpper atoms the Bo, JOI. aml OU posilil.JnS III a filCC-<;enlcrc{1 cubiC lattice. Give tile indices cf the new x-ray reflections that appocar wllcn tile alloy got:S from the tlisortleredlo the onlcrL't1 state. List all llCw rdlections with indices ;:2.

2. Configm"(l,iollld heul capaci'y. Dt:rivc an ex.pression in tcnus of P(1) for tile hc.!t capacity associale.:l with onlcrltliscnler effecls in an AB alloy. [The entropy (8) is <:allctllhc <XJnfigul1\tional elliropy or entropy of mixing.!

References
U, Ehrenreich and L. M Seh",:..IL. "Electro"ic Stnlcture of Allovs," Solid stale physics 31, ISO
(lfITfi).

V. L. Moruni, J. F. Janak, and A. It Williams. Calculated electronic proper1in of metal$, Perg:unon, 1978. F. RosenbeTger, Fundamentals of cry$tal grou;th I. Springer, 19S1. Discuss...m of phase diagrams. D. de Fontaine, "Configurational Ihermodynan'ics of solid solutions," Solid stale physICS 34, 73
(1979).

Kondo, 'Theory of dilute magnetic alloys," Solid state physil;s 23, 1R3 (1968). C. Klttd, "Indirect exchange IOlcraction in metals:' Snhd slate physico 22, I (1968).

J.

APPEl\'DlX A: TEMI'EHATURE DEI'El\'DENCE OF THE REFLECTIO", LINES

... I callle to fhe C(mclu~ion thaf ti,e ~/lUrpness ofthe interference lines would not l;ujJer bUf fllUt flleit" j,,_ femity S/lOlllrl ,limillisfl IVith increasi"g lmgle of Bcaffering, the more BO the higher the femperafllre.

P. Dcbyc As thc tcmpcrature of thc crystal is incrcased, thc intensity ofthe Braggreflected beams decrcases. but thc angular width of thc reflectcd line docs not changc. Experimcntal intcnsitics for aluminum <lrc shown ill Fig. 1. It is surprising that we can gct a sharp x-ray rcflcction fmm atoms undcrgoing largc amplitudc random thcrmal motion, with instantancous ncarest-neighbor spacings diffcring br to perccnt at room tcmpcraturc. Before thc Lauc cxpcrimcnt

, , " , , ,
,
3

,1'....

....-i:>=..
~.

o.

51Q(l\

1'-_.
"
.~

'"

-'--.,
"

'-"
,
"

(400)

".,
"

"".
"

,..,

~)

,
1'00' "
"

'.

"\-',
'00

(to,OO)

\,
'"
T.m K

""

"

""

Figure I The depend<-'fll"e of iolen"h' un leOlpernlufe for the (/'00) x-roW refle<"lion' of aluminum. Ren<-'ction$ (1100) with h odd are forbIdden for an fee strudllre. (Mtcr R. .\1. '\ickl\>W and R A

Young)

.,
was done, but wilen the proposal \\~\S cliscIIssC(P in a coffee h"sc in Munkh, tile objection was made that the instantaneous positions of the atoms in a (;I")'sial at room tcmpcrahJre are f:\r frOIll a regular periodic array, because of the large

thermal fluctuation. Therefore, the argument went, olle should lIot CKpcct a well-defined eJiffractccl beam. Bul such a beam is found. The rC.lson was given by Dcbyc. Consider the radiation amplitude scattered by a crystaL let the position of the atom nominally al rj contain a Icnn II(!) fluctuating In time: ret) = rJ + u(t). We suppose each atom fluctuates independently about its own equilibrium position. 2 'Olen
tbe thermal average of the struclllrc faelor (2.43) contains tcrms

Jj cxp(-iC' rJ){cxp(-iC' u)} .

where (. ;,

(1)

.) denotes thermal avemge. The series expansion of the exponential (exp(-iC

= 1 - i(G u) - i(C' U)2)

+ ..

(2)

But (C u) = 0, because u is a random thermal displacement unl,."OfTe!<lted with the direction of G. Further,
{(C' ul) = G 2(1I 2)(COS 20) =
!(1I 2

)C 2

The fador i arises as the gcometrical average of cos 2 0 over a sphere. The fUllction

has the same series expansion as (2) for the first two terms shown here. For a harmonic oscillator alltenllS in the series (2) and (3) can be shown to be identical. Then the scattered intensity, which is the square of the amplitude, is
(4)

where 10 is the seattered intensity from the rigid lattice. The exponential factor is the Dcbye-Waller factor. Here (1/2) is the mean square displacement of all atom. The thermal average potential energy (U) of a classical hannonic oscillator in three dimensions is ~k/jT, whence
(5)

where C is the force constanl, M is the ma."s of an atom, and w is the frequency of the oscillalor. \\le have used the result w2 = CIM. TluJS the scattered intensity is
(6)
'I'. I' EWllld. prWllt" t"Omm""irnllon. "I1,i, is the Eimlt'in model solid, it " not a very ~ood at low temperat"res. L"t it wt>fks w{'I1,,! hi~h lcmperdturcs. For a gcll<'rallrc;'tm,,"1 of R'altering Ly Ihermlll nllcl"alio",. Moe QTS. Chapter 20.

or"

"'ood

Appendix

633

whcre 'Ikl arc tht! Jiccs of the rcciprocallaUice \'cdor G. 111is classical result is a good approximation at high temperatures. For quantum oscillators (1/2) docs not vanish evcn at T "" 0; thcre is zeropoint motion. On the independent harmonic oscillatOr model the ....em-point energy is U1W; this is the energy of a three-dimensional quantum harmonic ')' of the same oscillaoscillator in its ground state rcfcrrcd to thc classical enerb tor at rest lIalf of the oscillator energy is potential encrgy. so that in the ground state
(U) "" iMw2(1I 2) "'" thw ;
(7)

whencc, by (4),
1(lIkJ) "" 10 exp(-hG 2/2Mw)
(8)

at absolute zero. If G "" w = 101~ S-I, and M = g, the argument of tile exponential is approximately 0.1, so that III o "'" 0.9. At absolute zero 90 percent of the beam is elastically scattcrcd and JO percent is inelastically scattered. We see from (6) and from Fig. I that the intensity of the diffracted linc decreases, but not catastrophically, as thc temperatufC is increased. Reflections of low G are affectcd less than reflections of high G. Thc intensity we have calculatlxl is that of the l'Oherent diffraction or the elastic scath:ring in the well-defined Bragg directions. The intensity lost from these directions is the inelastic scattering and appe,\fS as a diffuse background. In inelastic scattering the x-ray photon causes the excitation or de-excitation of a Iflttke vibration, and the photon changes direction and energy. At a given temperature the Debye-Waller factor of a diffraction line del;reases with fln increase in the mflgnitude of the rcciprocfll Iflttke vedor G assoclflte<l with the reflection. The larger ICI is, the weaker the reflection at high tempemtures. The temperature dependence of the reflected intensity for thc (1100) reflections of aluminum is shown in Fig. 1. nle theory we have worked out here for x-ray refle<.tioll applies equally well to neutron diffraction and to the l\H:issbaucr effect, thc recoilless emission of gamma rays by nuclei bound in crystals. X-rays QI1I be absorbed in a crystal also by the incl1stic processes of photoionization 9f electrons and Compton scattering. In the photocfTcct tlu: x-ray photon is absorbed and ;\n ele<.1ron is ejected from an atom. In the Compton effect the photon is scallcred inelastically hy an electron; the photon loses cnergy and the electron is ejcdcd from an atom. Thc depth of penetration3 of the x-ray bcam depends on the solid and On the photon energy, but- 1 cm is typical. A diffracted beam in Bragg reflcctiollmay remove the cnergy in a Inuch shorter distance, perllaps 10- 3 cm in an ideal crystal. 109 cm- I , 10- 22

"SC(:

\Y.

II<-itler, Q"OIl/"", II'ron; of radiaUQII. 3rd cd.. O.furd. 1954, p. 223.

Al'IlEf"DIX B: EWALD CAl.CULATION OF LATTICE SUMS

1~ problem is to C31culatc the declfTht,ltiC potcnti:1! C\)JeTK:n.>(! hy one KJII in the prCSCII(:C of all the other ions in the crystal. \\le consider a lattice 1I1..wC up of ions with positi\'e or ncpli\'c ch.ugcs and stwll assume th"ltllC ions are spherical. \\'c (""OllLpule the total!>olcntiaJ ({J = (J'I + lPt .11 .m iOIl .15 thc sum uf two llistincl hut relatcll poklliiais. Ihe potcnti.t! 'PI is Ih.ll of a structure \\ith a Gatl~iall dislnbulioll of ch'lrgc situated at each iOIl sile, with sig,lI$ the same as those ohhe real iOlls. Aceordillg to the definition of the Ma<leilingconstani. thc c1Mrgc dbl. ibnliOIl on the reference point is nol considered to c-'Onlribute ttl tile potcnUal /PI or f/>2 (Fig. l.\). We therefore Collculatc the potential 1/'1 as the
dilfcrcl1(.~

,.

Fij;ure 1 (<I) Ch.uj(e ,Jllolrll"'hO" "".., lOr C'Olllputmg p"t~~,I~1 ~" tI,t: pot~nl .. I~. IS ~~""11Ut('d (,I ",.,1",1<-, tl,~' ,1."IIt'l1 ('tln-~' dt tlu' r..r.u:,K'C 11OI"t), "h,l, "'" 1$ the polenl;..1or Iht; ,b,I...~l nit'" alu,w (I C]",I')' d"tnb"t,on f,'r 1~,(p"li ..1 Of' Th.. rdi'r~'n~.., point IS d<.'lIot...I hy an \

"'

A"pendix

of two potellti,tls, 'Pa being the potential of a continuous series of Gaussi'lIl distributions .md 'Pb being the potential of the single Gaussian dish ibutioll on the referencc point. 111e potential 'P2 is that of a lattk:c of point charges with an additional Gaussian distribution of opposite sign !>uperposcd upon the point charges (Fig. lb). 111e point of splitting the problem into the two paris 'PI and 'P2 is th.d by a suitable choice of the parameter determining the width of each Gaussian peak we ~an get vcr)' good convergence ofboth parts at the same time. The Gaussian distributions drop out completely on taking the sum of the separate ch.lr~e distributions giving rise to 'PI and 'P2, so that the value of the tot"l [XItellti.,! 'P is independent of the width parameter, but the rapidity (If C<l1welgence depends on the value chosen for the par,lIneter. \Ve calculate first the potential 'fIa of a l,;QntillIlOus Gallssi,1n distribution. \Ve expand 'Pa and the charge densit)" P in Fourier series:
'Ptt. =

L ccexp(iC . r) G L A:;exp(iG . r)
G

(I)

p =

(2)

where G is 21r' times a vector in the reciprocallatti(.-e. The Poisson equation is

V2'Pa = -41Tp ,
0'

k C2ccexp(iC' r) = 41T L A:;exp(iG . r) ,

so that

(3)
We suppose in finding Pc tllat there is associated with each lattK..-e point of the Bravais laUke a basis contaiuing ions of charge q, at positions r, relative to the lattk-e point. Each ion point is therefore the Cnter of a Gaussian charge distribution of density

where the factor in front of the exponential cnsur~s that the total charge associated with the ion is q,; the range parameter TJ is to be chosen judiciously to ensure rapid (.'Onvergence of the final result (6), which is in value independent of TJ We would normally e~'aluate Pc by multiplying both sides of (2) by exp(-iG . r) and integrating over the volume t:J. of one cell, so that the charge distrib\ltioll to be considered is that originating on the ion points within the (.1::11

and also that of the tails of the distributions origill:l.ting in a" ..other cells. It is eas)' to see, however, that the integral of the total charge densit)' times C.\I>{ -(iC' r over a single cell is equal to the integral of the charge dellsit)' originating in a Single cell times exp(-(iC' r)J over all space, \\'e have therefore Pc

...

exp(iC' r)eXJ.>(-iC . 1') fIr'" Pc6

'ntis expression is readily evaluated: Pc6 =

"

J~ q,r.:"'1TF
"'

exp(-"7C r - r,1Iexp(-iC' r) rlr

~ q.exp(~iC 1',) (T1'1T-ra

(2: q.exp(-iC 1',) exp(-C 2/411) =

exp[ -(iC . t

+ 11t'~] fIt'

S(C)e1OI>(-C 2/411)

whel'e S(C) =

2: q,exp(-iC' r,) is just the stmcture f.1ctor (Chapter 2) in ap

(4)

propriatc uilits. Using (I) and (3),

At the origin r = 0 we have

111e potential If" at the reference ion point i due to the ccntral Gaussian distribution is

and so

111e potential 'P2 is to be evaluated at the reference point, and it differs from zero because other ions ha\'e the tails oftheir Callssiml distriblltionSoover-

App",llIi;r
(

lapping the refen.. ~e point. 111e potential is due to three mntriblltions frOIll each ion point:

q,[.!.. - .!.. r,rJO

p(r) dr -

loo

p(r) (If] , r,r

where the terms lll"e from the point charge, from the part of the Gaussian distribution lying inside a sphere of nJdius r, about the Ith ion point, and from that ,}'1rt I)'ing outside the sphere, respectively. On substitllting for p(r) and caiTying out e1ementalY manipulations, we have
(5)

where "'(x) = (21w ' r.i!) Finally,

tAi) = 4w

I'" exp(-r) cis

2q;(Tj/w)'f2 +

tJ. (;

2: S{G)C-2 exp(-C 2/4Tj) -

L
I

!l.!.... f(Tjlr.i!r,)
r,

(6)

is the desired tollil potential of the reference ion i in the field of all the other ions in the cf)'stal. In the application of the Ewald method the trick is to choose Tj such that both sums in (6) converge mpi(lIy.

EUJ(lIJ..KonJjeld Me/hod for Lattiee Sums for DilJOW Arrays

KOOlfeld extended the Ewald method to dipolar nnd quadrupolnr 1lI"lUYS. We discuss here the field of a (Iipole llrray at a point which is not a laUice point. According to (4) and (5) the potenlilll lit II point r in n Inltice of positive unit point charges is
rp = (47T1tJ.)

2: S(G)CG

exp[iC . r - C 2/4Tj] +

2: F(v';] r,)/r,
,

[l)

where rJ is the distance from f to the lattice point t. The first term on the right gives the potential of the charge distribution P = (Tj/w'f12 exp(-Tjr 2) about each lattice point. Bya well-known relation in e1ectrostlltics we obtain the potential of an alTay of unit dipoles pointing in the z direction by taking -dMz of the above potential. The term under discussiun contributes

-(47ri/tJ.)

2: S{G)(CjC2) expJiG . r G

C2/411] ,

and the z componl'nt of the electric fidd f!"Om this term is E~ = .-r-lf/ilz2 , or

-(4'"'6.)

L
c

S(C)(G~/C~ exp!iC r -

c 2/"lJl

(8)

Tile second term on the right of (7) afier olle difren"lti,l!ion gives

-L , zil(F(v';jrJ/r?) + (2Jr,zKrp1T)'12 e'l:p(-lJrl)]

.md the:: component of this pMt of the field is

L {Zl2(3F(\h1r/)/t/) + (6Ir/)(r11T)112 exp(-rwll

+ (4Irl)(Tj1/ 1T)1I2 cXI'>(-lJrl~1 - [(F('\.0iI)frp) + (2Jr?KlJ/1T)1I2 e\p(-Tjr,z,)11
(9)

The total ~ is given hy the sum of (8) find (9). The effects of any IIll1nIJcr of Lltlicrs may be added.

APPENDiX C: QUANTIZATION OF EL\Snc WAVES: PUONONS

PhOl1ons were introduced in Chapter 4 as (ju..Ultized el..l~tic waves. Ilowdo we <jwmhzc an elastic wllve? As II simple model of phonons in a cry.'>I..II, consider tbe \ ibrations of a linear lattice of particles connected by ~Ill'ings. We com <jIHlntize the particle motion eXHctly as for 1I hlll1nonic o.'>ciltator or set of coupled harmonic oscillators. To do this we make 1I transformation from p<ll'ticic COIll'dinatcs to phonon coordinates, also called wave COQrdin,ltes because they represcnt a traveling wavc. Let N particles of mass M hc connected by spring.~ of force const.lllt C and length (I. To fix the bouudary conditions, let the plIrticles form a circul,lr ring. We consider the trllnsversc displact'ments of the particles Ollt of the plane of the rin~. The displacement of part ide s is (/. and irs momCl1tl1lJl is P. The Iwmiltonian of the system is
(I)

The IHllniltonian of a harmonic oscilllItor is

1 p2 +_CX 1 2 H= __ 2M 2

(2)

and the energy eigenvalues are, where

II

= 0, J. 2, 3, .

..
(3)

111e eigenvalue problem is also exactly soh'able for a chain "'ith the different hamiltonian (I). To solve (I) we make a Fourier Iransfolloation from the coordinates P., q. to the coordilllltcs Pic. QIc, which are known as phonon coordinates.

Phonon Coordinates
l1re transformation from the particle coordinates q. to the phonon coordinatcs Q" is used in all perioc.lic lattice problems. We let ql = N- 1J2

L , Q" exp(ik$a) L q. exp(-iksa)

(4)

consistent with the i1werse tramfonnation

QIc = N-'I"l.

(5)

Ilerc the N valucs of the wa\'ev(..,<:,tor k allowed by the periodic ooundary condition l/. = q.+N are given

k=2mJ1Na ;n=O,

~J, ~2,"', :!:(~

N-I),+N.

(6)

We need the transformation from the partkle momentum P. to the momentum Pic that is canonically conjugate to the coordinate QIc. 111e transfonnation is

P. = tV- l12

L , Pic exp(

iksa); Pic

::::I

N- I12

L P. exp(iksa)

(7)

This is not quite what one would obtain by the nl'li\'e substitution of p for q and P for Q ill (4) and (5), because k and -Ie have heen interchanged between (4)

and (7).
\Ve verify that our choice of P" and Q" satisfies tire quantum commutation relation for canonical variables. \Ve form the commutator

"" N- 1

L L [q ...p.1 expI -i(kr

k's)a]

(8)

Because the operatoni q, " are conjugate, they s'ltisfy the C'OmmutatiOll relation
(qnp~J

= iM(r,.s) ,

(9)

where cS(r,.s) is the Kroned<er delta symbol. Thus (8) becomes

[Qi,P~'J = N- l ih

L ,

exp[-i(k - k')ral = ihcS(k,k')

(10)

so that Q~, p~ also arc conjllg.'lte variables. Ilcre we have cvaluat<.>(l the summatioll as

L cxp[-i(k

k')ral =

L exp(-i2w{n

JI')rlNI

III)

= N6(n.n') = N6(k,k')

\\here we ha\'c used (6) and a standard rcsult for the finite series in (11). We carry Oil! the transfonnations (7) and (4) on the hamiltonian (I), and make use of the summation (J I):

(12)

x exp(ik'su)lexp(ik'tI) - I)

2: , QkQ_,..{J

- cos ka) .

(13)

Thus the hamiltonian (I) becomes, in phonon coordinates, H =

L [_'P"P -It + CQI,-Q_l(I , 2M

WI;

- cos ka)}

(14)

If we introduce the s)'mbol

Wi

defined by
(15)

(2CIM)"2 (I - cos ka)'12 ,

we have the phonon hamiltonian in the fonn

(Hi)

111e eclualion 01 lllolioll of the phonon (..o onlinalc opcmtor Qt is f(lUnd the slandard prescription of quantum mechaniC':':
iM.h

h~

= [Qt,Hj = ihP~J!M

(17)

with H given by (14). Further. using the commutalor (17).

ihQl = {Ch,I-1] = M-IIP_t,lf! = ihw/Qk, so lhat
('9)

(18)

111is is the equation of motion of a harmollic oscillator with Ihe li"<.'qu('lICy The enerso- eigen....alues of a quantum hamlOlllc oscillator arc Et = where Ihe qua.lllum Ilumber tem of all phonons is
Ilk

Wk.

(lit + ~)hWk'

(20)

= 0, J. 2, .... 111e energy of the entire sys-

(21)

111is result demonstr.tles the quanti7.ation of Ihe en"rgy of e1aslic.' W;j\es on a line.

Creation arlll Annihilation Operators

It is helpful in ad"llnced work 10 tr,msfOl'm Ih" phonon hallli!toni,1Il (16) into the form of a sel of h"rmonic osdllalors:

-'2-) .

(22)

Ilere lit, (ll al'" h"'nnolllc oSL'iIl",tor opel1110rs, ",Iso "",lied crc",tioll and dcstmction opemtors or boson operators. 11u: transform.ttioll is derived m.low. The boson ereation oper",tor n ~ which "creates a phonon" is defined b, the propert)
(23)

when acting on a hannonic oscillator slate ofquantum number tl, and the boson anllihilalion operalor n which "destroys a phonon" is defined by lilt" property
a[tI) =

"I,," _ I)

(24)

(I'

(l11l) = a >u11'l1n - l) = 1I1"}

(25)

~(J IliJ.tln) is an ei~enstatc of the operator (1"0 with the intc~ral eigcll":lluc H, (,tlled t!ll' quantulll numher tlr oa'1rpaney of tIlt' olodllJ.lor. 'Vhl'lI tIll' phonon mooe Ii i., ill the ci~cmt.llt'l.d>ek-dby /1.1:. we nM~' SJ.) that tht>rC' .Ire III phonon~ III the IlltKII'. The ei~cn\'alllesof(22) .Ire C = :r (Ill + U!lwk. in J.gn'1.'IlK'1I1 with (11).

(/(/ 'I,,) =

a{1l + 1)14/1 + I} .., (II

1~1l)
ak ~Iisfics

(26)

the lummulator of the IJOson WJ.\,{' opeutors ai.. J.uu

thl" u:btiun
(27)

We ,nil h,L~e to I'ro\'l" tlMt tIle h,lIniltouiJn (16) l';\ll he e~prcssed ,IS (19) in Il'fI1h (If tIll' pholloJl Opl'llItOrs tit, (ll. Thi~ elln he dOlle lJy Ihe Ir,lrIsforll1;ltioll

ut = (2flJ
111e ..werSt' rcl.tliolls arc

1/2!(MwiJI ~()

k -

i(:\fw~) II~Pd

(28)
(29)

(I~ = (2ft) II2({MwIY'2Q~

i(Mwil I'lp_s.l

Q" = (lU2MwAJI~a" + 0".0

Pi = i(M1wt/2)' ~at

(30) (31)

a i) .
bt."t-'OflWS

B, (I), (5), .10(.1\29) the p.u1iclt, pm-ition opcr.ilor

lf~

L ,

(/112NMwl)II2(Ui e\l,(ikS)

+ lit

l''(I,(-ih

(32)

111is etj!loltion rclJ.tl'~ the pat tide dhplal"elllent npcratOl'lo the phonon tn'alio., lind .1IllHhil.ttion npt'r.it!)rs. To uht,lin (29) from (28), we use the properlit.'~

= Pi

(33)

\\hil"h fellow frum (5) .md (7) b~ lrse of tile (jlrantulIl nJl..'chaniealll'<llIireml..'nt th.ll q, and IJ. he hCI mi{jan 0PCl'dtOrs:

(341
"Then (28) follO\\~ fmlll the IrJ.ll\lonnatiom (1), (5), and (7). We ~erlf\ th,!t tltl.' CUUlmutation rclarion (33) is s,ltisfietJ by the operaturs defilled b) (2H) l\IKI (29): {ol,all = (2h) I(MwllQi.Q 11 -

iIQ",rtl + i(P l,Q tI

+ IP
I

Pd/.\lw,)

(.35)

By lise orIQI",Pd "" ili8{k,k') from (10) we h.wc

IClk,CIA'] "" &,I.;,k')

(36)

It fl.... ll1ains to show th,lt the vcr~iom of (16) ,!l1I1 (22) of the phonoll hnmi!l()ni.lil lIrc idl'ulical. \Ve lIote thdt Wk "" W_l from (15), .lItd \\c form

This c~hibits the e<luivall'nce of the two l'xpn.'s~ioll~ (J.ll .\IId (22) for H. We identify WI, = (2CIM)1I2(1 - cos ka)lfl in (15) with the d.ls~icdlli-clJlIlncvofthc oscIllator Illude of W,WCVCl'tor k.

APPENDIX 0: FERMI-DIRAC DISTRIBUTION I~Ul'<crION

111e Fermi-Dirac dbtrilmlion function' may be derived in ~cveral ~tcps hy lise of a modem ,Ipproach to statistical mechanics. \\'c outline the argul1lent here. The notation is sUl'h that l'ollventional cntmp)' S is rcbted to the hllldillI1ental entropy u b)' S = knu, and the kelvin temperature T is rcluted to the fundamental tellll>erature 'j by 'j = kaT, where k n is lhe BoltZll1anll CQlht,mt wilh the value J.3H06fl X 10- 23 J K-'. Thc leadillJ!; quantities are thc entropy, the tCl11per.lturc, the BOItZl11lllll1 f.1etor, the chemkal potentidl, the Gihbs bctor, lind the distnbution functiom. The enllopy me.\Sures the number of quantum state~ acces~ible to a ~ystem. A dosed system lIliJUlt be in any of these qmmtlllll ~t.ltes and (we assume) with equal prornlbility. The fundameillal assllmptioll is that (jll.llltlllll st,ltes arc either aCl:essihle or inal'ccssihle to the Systl'lll, and the S\stCIII is cljll.\lI)' likely to be ill anyone accessible slate as in any other .Icc(~ssihl(' stat<'. Gil'en ~ ;ll'CCssible states, the cuhopy is definC{1 as u = log g. The entropy thus lkfined will he a fundion of the energy U, the number of parlic1t~s N, J.lld the volullle V of the s)'stem. \Vhcn two wstellls, ('<lell of ~pedfied cnl'r,.:v, ale bro~l!!:ht into tlwrlllal cont;let thev 1i1;IY tr,lIlsfer energy; their tot.ll cnerj:!:y lem,lins constant, hut the constraints on t!leir illdivi{Iual encrgies art~ hfltxl A tr,m~fer of energy ill one din..'Cl:ion, or perhaps in the other, may increase the product gil!:! that 1I1l'.lsures the number of accessihle ~Iatcs of the l'ombilled s)stems. WI1.It we c;11l the fundamental assumption biases the oull'ome ill favor of llldl allol"ltion of the total cner~y thatll1.l\imi,ws the number of.lcccssihie ~tale~: mOrt: IS beticI'. lind more likely. This statement is t!le kernel of the ldw of innease of cntropy,

"n,.. .'pptn,
2n<l
.~I

,,,I1,,,,s d05<:I) til" '"troc.l.... I''''' 10 t' I'ludand II "-"""",,,r, fllC.... ""I'I",',~ . Fr.em"". 19lill.

.. ,
wllien is the general expression of the second law of thcnnOlo,namics, \\'e ha\,C brought two systems into thennal rou(act so that they may transfcr energy. What is the most probable outcome of Ihe CIlI..'ounler? One system will wain energy at the expense oflheother. and meau\\'hile the lolal cntropyof the two systems will increase. Eventually the entropy will re3Ch a maximulIl for the given total energy. II is not difficult to show thai the maximum is attained when the value of (iJufau)l\'Y for one system is equal to the value of the same quantity for the sec.'Ond system. This equality property for two systems in ther. mal contact is the propeTt)' we expect of the temperature. AcrordiJl~I)', the fundamental temperature T is defined by the relalion

+- (:)N.V

(I)

The use of lIT assures thai energy \\ill flow from high ., to low '7; 110 more complicate<! relation is needed. NO\v consider a very Simple example of the Boltzmann factor. Let a small system with only two states, onc at energy 0 and onc at encrgy E, be placed in thennal contact with a largc system that wc call the reservoir. The total energy of thc combined systems is Ug; when thc small system is in the statc of CI1Crgy 0, the reservoir has energy V o and will ha\'e &1U~ slates a"essible to it. Whcn the small systcm is in the state of energy E, the r~rvoir will have encrgy V o - E and will have g(Uo - E) states accessible to it. By the fundamental assumption, the ratio of the probability of finding the small system wilh encrgy E to the probability of finding it with cncl'gy 0 is
P(E) &1Uo - E) exp(a(Uo - E) P{O) = g{Vol = c,p[a(Vol!
(2)

111e rescrvoir entropy a may be cxpandl.'C] in a 1i:lylor serics:

a(Uo - E) "'" a(llo) - E(i)ofaVo) = aWol - EIT ,

(3)

by the definition (I) of the temperature. Higher order terms in the expansion mlly be dropped. Cancellation of the tCl'm expla{llO>I. which occurs in the 1lI11HCI'3tor and denominator of (2) aftcr the substilution of (3), leaves us with
P{E)JP(O) = exp(-Ef7) .
(4)

nlis is Boltzmann's result. To show its IISC, we calculate thc thermal 3\'er age ellergy (E) of thc two-state system in thennal (ontacl with a reservoir at temperature r.
(E)

=L ,

E,l'{E,)

= 0 . P(O) + d'{t) ""

exp{-tlT)

I + exp{ EI7)

(5)

Appendi"

645

where we have imposed the normali".a.tion condition on the sum of the probabilities:
P(O)

+. P(I:) = I

(6)

The argument can be generalized immediately to (Iud the lwemge energy of a harmonic oscillator at tempel~tlll"e '1, as in the Planck Ill\\'. TIle most importllnt extension of the theoly is 10 syslems thllt can transfer p.'111ides as well as energy wilh lhe reselvoir. For two syslems in diffusive and thermal contact, the entropy \\ill be a maximum with respel."l to the transfer of pm1ides as well as to the transfer of energy. Not only must {ilolilU)N, v be equal for the two systems, but (ilalilX)l',v must also be equal, where N refers to lhe number of particles of a given species. '111e new equality conllition is the o.-asion for the introlluclion 2 of the chemical potential J.t.
!'- -;- =

(il<Y)
il,'"
U,I'

(7)

For two systems in thermal and diffusive l.'ontact, '1[ = '12 and J.tl = J.t2. The sign in (7) is chosen to ensure that Ihe direction of particle flow is from high chemical potential to low chemical potential as equilibrium is llpproal.'hed. The Gibbs factor is an extension of the Boltzmann faclor (4) and allows us to treat systems that call tronsfl'r particles. The simplest example is a system with Iwo states, one with 0 particles and 0 energy, and one with I (llirticle and energy E. The system is in contaet with II reservoir at tempemlure '1 and chemical potential J.t' We extend (3) for the reservoir entmpy:
o'{Uo - ;No - I) = u{Uo;No) - l:(iJufaUo) - I . (iJufaNo)

= a(Uo;No) -

E/T

+ IJ/r

(8)

By analogy with (4), we have the Gibbs factor

P{I,I:)JP(O,O) = exp{(J.t - 1:)1'1]
(9)

for the rntio of the probability that the system is occupied by 1 particle al energy E to the prooobility thaI lhe system is unoccupied, with energy O. The result (9) after nOlmalization is relldily expressed as
P(l,E) =

exp {E

J.t)h + 1

(10)

This is the Fermi-Dirac distribution function.

tl'P Ch:ll'tcr 5 has a. l....r..r"l Irl.';llmenl

or the

..hcmil."il.l potenhal

APPENDLX E: DERIVArION OF THE dkIdt EQUATION

The simple and rigorous derivation that follows is due to Kraemer. In quantum mechanics, for any operator A Vie have
d(A)ld' = (dh)([I/,A) (1)

where II is the hamiltonian. See also C. L. Cook, American

(1987).

J.

Phys. 55, 953

We let A he the lattice lranslation operator l' defined by

l'ff.x) = f(x

+ a)

(2)

where a is a basis \ector, here in one dimemion. For a Bloch function

(3)

This result is usually written for one band. but it holds even if l/Jk is a linear combination of Bloch slalcs from any number of bands, but having tile identj,....",1 wavevector k in the reduced .lOne scheme. The crystal h::llnihonian llo commutes with the lattice translation operatQlT, ~o that [llo:n = o. If we add a unifonn extemal force F, then

II = llu - Fx ,
[H.TJ = FaT.
From (I) and (5),
d('l)Id' = (dhXFa)('li

(4)
(5)

(6)

From (6) we fonn (1/*d(J)Itl, = (iFaIA)I(J)1 2

;

('l}d('j")Id, = -(Wwh)l('lil'

On addition.

d!('l)I'Id,

= 0

This is the equation of a circle in the complex plane. The ('()(.Irdinate axes in the plane are the real and iUUlginary parts of the eigenvalue cxp(iku). If (1/ is initially on the unit circle. it will rem,lin on the unit circle. For l/J's t1l.1t salisfy pcriodic boundary conditions, ('1) can lie on the unit circle only if l/Jk is a single Bloch function Or a superposition of Bloch functions from different bands, but with the same reduced k. As 0) moves around the unit circle, the waveveclor k changes exactly al the S,lllle r.lte for the COIilPOIlCllts of l/Jk in alllJ<\nds. \Vith 0) =: exp(ika), we

have from (6) that

jn

dklelt = iFnlh
0

(8)

0'

dkldt = Flh (9) oUl etact n.~ult. This docs nOl mcolll th,il 1I1terh.Uld mixing (sulh as Zeller tunneling) docs nOl occur under the influence of olpplied electnc fields. It just meanS tlut k evohes at a constant rdte for t,.overy component of.\ Wd"C pad"d. TIle result is easily extended 10 three dimensions.

APPENDIX F: BOLTZMANN THANSI'ORT EQUATION

The classical Iheory of transport processes is b<tsed on the Boltzmann tramport equAtion. 'Ve work in the six-dimensiOnal spare ofC.lrtesian coordinates r .md velocity v. '[11e c1as~iC"dl dhtribution function /(r, v) is derlned by the rel,llion f(r,v)llrtlv = number of pmtides in dnlv
(I)

The Boltzmann equatiun is derived by the following argulllent. We consider the effect of a time displacement (ft on the distribution fundion. 'nle Liou"ille theorem of classical mechanics tells us that if we follow a volume element along a flowline the distribution is conserved: I(t

+ llt,r +

dr,,,,

+ dv)

f{r.r,v),

(2)

ill the absence of collisions. With collisions

f{t + dl,r + llr,v + (Iv) - f(t,r,v) = dt(iJf!iJt)"on"i",,'

'J1lUS

(3)

ilt(af!iJt) + l/r . gr:Id. f + dv . grarl~ I = clt(iJf!iJt)""n

Let a denote the acceleration dvldt; then

(4)

af/iJt + v grad. f + a . grad.. f =

(~f/iJtkd.1

(5)

'n,is is the BoIt1..mann transport equation. In mallY problems the collision term (af/at)..,ll mdY be tredled hy Ihe introduction of a relaxation time T.(r.v), derlned by the equation

(af/iJt)... ,ll = -(J - fulJTr

(6)

Here 10 is the distribution function in thermal ecluihbriulIl. I)( t confuse T.. for relaxation time wilh T for temperature. Suppose that a nonccluilibrium distribution of velocities is set up by external for.'S which are suddenly remo,,cd. 'Ine decay the distributiOn towards (.'(Iuilibrium is then obti\ilwd. from (6) as

or

<JJ- f~
elt

_f- f.
T..

m
(8)

if we note that clio/elt = 0 hy definition of the Ct'luihbriulll distrihution. TIlis equatiOn has the solution

It is not excluded that T~ may be a function of rand v. We combine (I), (S). and (6) to obtain the Boltzmann trallsport equatiOn in the relaxation time apPrOximation:

af ai+

(l'grad.. l+

\I'

grad.. I =

(9)

III the steady state elf/elt = 0 by definition.

Particle DifJU8ion
Consider an isothermal system with a gradient of the particle concentration. The steady-state Boltzmann transport equation in the relaxation time apI)roximalion bet..'OlTIes
v..

Pdx ~ -If - f<J1To

(10)

where the nOl1equilibriulll distribution function I varies along the x direction. We may write (10) to first order a~

(II)
where we have replaced oflib: by (lfoldx. We can ilertltc to obtain higher order solutions when desired. Thus the sCt.'OlLd order solutiolL is

12 = 10 - v,TclfNdx = 10 - v,T,dfold;r + c;'T~d2Io1dx2

The iter.atioll may be used in the treatmelLt of nonlinear clrcets.
Classical Distribution

(12)

Let 10 be the distribution function ill the classical limit: 10 = exp{(p. - E)ITJ .

(13)

We are at liberty to take whatc,,'er normalization for the distribution function is

most convenient bo se the troillsport C<lualion is Iin('3r in 1 and If}. \\'e take the norOlali7.<llion as in (13) rather than as in (1). lnen

(:ail

<lI';d,

(dI';d"X,I"Jdx)

(I';,Xd"Jdx) ,

(14)

alld the first order solulion (11) for the nonetluilibrium distributiOtl becomes
I~

I. - (v"J.hX<I"Jd.,)
1"

(15)
(16)

TIle particle flux density in Ihe

direction is

J': .:

vJIJ{fE)(tfE

where IJ(E) is the density of electron slates per unit volume per unit energy . 312 range: D(t-) .: _1_ fEI12 (17)

271 2

(2M) h
2

as in (6.20). Thus

J.... .:

VJoD(t)dE - (flpJd;c)

(v,,2"tcio/T)D(E)c!t

(18)

The fll'5t inlegral vanishes because v, is an odd function and 10 is an even function of v. Thili confinns that the nel particle flux vallishes for the equilib rium distribution 10- The second integral will not vanish. Before l....'a1uating the second integral, we Jlave an oPPOitunity to make usc of what \\"e may know ahout the velocity dcpendcll(;e of the rda.lation timeT.... Only for the sake of example we assume that '1... is constant, independcnt of \'elocity; '1.. may then be taken oul of the integral:

J,,' .: -(dpJdxX'1J'1)
The integral may be written

v. 2/oD<E)dE .

(19)

a~

.rIr}J(E)cfE .:

3~1 Ul\lu~/oIJ(E)efE .: m'/M

J
11

(20)

because the integral is just the kinetic energy delLSily illT of Ihe parlicles. Here f/olJ(E)cfE .: II is the com..'cntraIiOIl. The particle flux density is

because
IJ- .:
T

log

+ l.'Ollstant

(22)

TIle result (21) is of the form of the diffusiOn equation with the diffusi\'ity
D.. .: '1"TlM .: i(VZ)T" . (23)

Another possible assumptioo aboul the rcla.'l:atioll time is thai it IS lll\'erscly proportional 10 the \'e1ocity. as in '1" "" lIu, where the mean free path I is <:onstant. Instead of (19) wc have

J,,' .: -(dpJdxXl/'1)

(u. 2/v)IrJJ(E)clE

(24)

...
alit! now the inh:KT.I! lIlav be writtl'lI

!
where i: is the
,1\l'I,lge

,l.S

vfoD(E)(IE =

~IJC

(25)

speed. Thus
= -

J,:
anti Iht.: diffusivity is

l(Ii:f1lT)(c!,.,Jdx) .. - \1c(dnMc)
D. =

(26)

ire

(27)

Fumi-Dir(/c Distribldio/l
'nlC

distribution fUIIl:hon

j\

iiI = expE

Jot)h']

I
(lfl)hJ,.~.

(28) We ,ugue hl.'1o\\' tll.\l

(29)

To form dfoldx as jll (14) we need the Jeriv.ltive

dfultlp. "" O{E - M) ,

<II low temper.duces T <C p.. l-Icrc 6 is the Dime dcltll (unction, whil.:h hM; the property for a gelleml flllll"lion ':(El that

IX "'(E)li(E -

1L)<lE '"' f(p.) .

(30)

Now consider the integral 10 f'(EXdfoldp.}IfE. At low temperatures llfoldp. is \cry

large for E ... IL and is sl1",1I elsl..'Whcrc. Unless the function F(E) is VCI~ f"pidly v.lI'ying ncar p.. we may I.lke feE) ouhide tll/.' illle~. With the value "'(p.):

F(Xd["Jdl')d. - F{J<)

(d["M")<1 -F{J<)

(df.'d.ld.

(Jl)

: - F(")Ud.) I.

F{J<if"O) .

where we have used tl[oh/p. = ~df;"(IE. We h~\Ve <llso u~cdfo = 0 for E = x. At low temper,ltures [(0) "'" I; thus the ri~ht-hand ~ide of (31) is just F(p.), consi.~I ent with the dell.'l function approximation. nm.~

'/foldx = fJ(E - p.}dJAidx .

ollie particle Au" density is, from (16),

(32)

J,,~

-(ll,ddx).,.c

(;/6(t:
E

/-LlD(EW'E,
= P. of the Felmi
~phcre.

(~

where "'c is the rela"alion lime al the surface integral has the \'alue

11lC (34)

..

by use of D(p,) = 311/2(;J: at absolute zero, \\here (;" (;F on the Fermi ~U1face. Thus

iJl/tJ~ defines thc velocity

],,r

= -(mjm}!/pJdx .

(35)

At ahsolute zero }-leO) = (ft2/2m)(3~IIF. whence

dpJdx = nW/2m)(31T 2 Fln 'fJ }dnldx
(36)

= ~(E"'n)dll/dx ,
so that (33) becollies
= _~Vl-2Tc clnldx .

],,r = -(2TJ3m)E" tlnldx

The diffllsivity is the coefficient of dnldx:

(37)

I)n = ~Vl..:zT"

(38)

closely similar in form to the result (23) for the classical distribution of velocities. In (38) the )-e1axation time is to be taken at the Fermi cnergy. \Vc see we can solve transport pTOblems where the Fenni-LJirac obtlih ution .\pplics, as in metals, as easily a~ where the da~sical approximation applies.
f:/~'Ctrical

Conductivity

"nle isothermal electrical conductivity u follows fTOm the n:sult for the particle dilfusivity when we multiply the particle flux deilSity by thc particle charge q and replace the gradient tlpJdx of the clJemical [Xltential by the grudi ent ljdlf>/dx = -qE z of the extemal potential, where E,. is the x component of the electric field intensity. The electric currcnt den~ity follows from (21):

}" = (nq2Tclm)E;
for a classical gas with relaxation time from (35),
T ...

u =

flq2 T

clm ,

(39)

For the Fermi-Oil-ac distrihution,

APPENDIX G: VECfOR POTENTIAL, FIELD MOMENTUM, AND GAUGE TRANSFORMATIONS

"
This section is included because it is hard to find the m.lgnetic vector potential A discussed thoroughly in one place, and we need the vector potential

in superconducti\ity. It may seem mysterious that the hamiltonian ofa partie-Ie in a magnetic field has the form derived in (18) bcIO\\:
II = 2M

'(P - -;;Q)' A + QIp ,

(I)

where Q is the charge; M is thc mass; A is the vcctor potcntial, and (fI is the electrostatic pokntial. This expn.'Ssion is \'alid in classical mechanics and in quantum mechanics. Because the kinetic energy of a particle is not changed by a static magnetic field, it is perhaps unexpech.>(I that the \ roOT potential of the magnetic field enters the hamiltonian. As we shall see, the key is the observation that the momentum p is thc sum of two parts, the kinetic momentum
Pkin

My

(2)

which is familiar to us. and the potential momentum or field momentum

(3)

The total momentum is

IPalld the kinetic energy is

",'0

Pc.", ~ AI, + ~ A -I

(4)

~ MI!- = -'(Mo"f = -'- (p - ..2. A)' 2 2M 2M c

TIle vector potential I is related to the magnetic field by
B = curl A .

(5)

(6)

\Ve assume that we work in nonmagnetic material so that H alld B are treated as identical.

LAgrangian Eqllations of Motion

To find the hamiltonian, the prescription of classicallllleX'hanics is clear: we mllst first find the Lagrangian. The L:Jgrangian in generalizcd l'oordinates is
L =

..!-. Mt? 2

Qq:{(j) +

2.... i j ' A(ql c

M PurttU, Ekdncdv Q"J mat:

'Foo- an element"') treallJll'rlt .otti.sm. 2nd ed.. McCrawII,11. 1984

or ,"'" \ectoo- potentoal sre E

This is correct beL. ...se it k.uls to the correct equation of molioll of a charge in combined electric and magnelic fields, as we now show. In Cartesian coordinates the Lagrullge equation of motion is

d JL JL -----=0

lit a.t

ax

(8)

and similarly for y and ;:;. From (7) we form

!.!:.. =
Jx

_Q

a/{' + -.2. (i
Jx c

JAr Jr

+ y~ +
ar

i,

aA~)
Jr

(9)

,t Q -=Mx+-A .
Jx c'

(to)

daI~ .. Q dAr .. Q (JAr . JAr . JAr . =Mr+---=Mx+--+, --+y - - + ., aA,_) -- . dt Jx c cit c Jt ax Jy Jz

(11) Thus (8) becomes

,~ Q --+y - - Mx+Q-+Jx

["'<
at

('A< 'A,) +z ('A< - - -'A.)] -Jy ax J;:; Jx

=

=0

(12)

0'

.\1

-1 , ,2

d 2x

QE.

+ - c [v>c
1

Blr

(13)

with E =----, ar cJt B=curiA

JIp

aA,

(14)
(15)

Equation (13) is the Lorentz force equation. This confirms that (7) is correct. We note in (14) that E has one contribution from the Clectrostatic potential/{' aile! another from the time derivativ~ of the magnetic vector potential A.

Derivatioll of tIle Hamiltonian

'nle momentum p is defined in terms of the Lagnmgian as P
ll!

JL =Mq' iJt'1

+.Q. A

c'

(16)

in agreement with (4). The hamiltonian H{p,q) is defined by l1(p,q)

:i

p' (1

- L ,

(17)

0'

2M

'(p-~A Q)' +Qip,

(18)

.IS

in (1).

Field Momentum
The momentum in the electromagnetic field that accomp.lJlie~ a particle moving in a magnetic Geld is given hy the volume integral of the Poynting vector, so that
Pr.ckl = - ' 4~c

(IV E lC 8 .

(19)

\Vc work in the Ilonrclalivislic approximation with v <C C, where v i.s the velocity of the particle. At low values of vic we consider B 10 arise from an cxternal

source alone, but E arises from the charge on the p'1I1idc. for- a charge Q al r',
E = -Vip;

v2 rp =
4~c

-417Q 5(1' - T')

(20)

Thus
P/=

--'-J

dV'\lp X curiA

(21)

By a standard vector relation we ha\'c

dV VIf' x curl A = -

dV [A X curl ('Vip) - A <Ii\' vip - (Vip) dh AJ

(22)

BUI curl (VIP) ::: 0, and we can always choose the gauge such that <liv A '" O. This is the Ir,msvcrse gauge. 11105, we have
P/=

--'-J
41TC

dV A V21p ""

2..c

dVAQ c5{r - r')

A SL c

(23)

This is the interpretation of the field contribution to the total momentum p = My + QAlc.
CAl!CE TRANSFORMATION

Suppose

If'" = "'. where

If =

-'-(p - -.2.A)' c
2M

(24)

Let us m.ll-e a gauge lr:lllSformatioll to A', wlH.'re

A'
Schrodingcr equation becomes
1 ( -~

+ Vx

'

(2.3)

\\ here X is a scaldf. I'\ow B = curl A c curl A', 1>cc.l.u.!>C curl (VX) O. TIle

2M

-A' c

+ -Q)' Vx 4i
c

- f'4i .

(26)

What I/! :..ltisfies

-~

2M

1(

p~-

Q c

A')'l;I ,.

=f''I'

".'

(27)

wilh the s..lffie

f'

as for ljI? Equation (27) is el:luivalcllt to

- 1 ( p~-A~-VX I/I-f'y! 2M c c

Q)'

(25)

We tf)'

4i' = exp(iQx1flC)4i .

(29)

"'ow

pI/!

= c'll.p'iQX1frc)pljl

5L (VX)e"p(iQx1hcl4i c

so that

.md

- 1 ( p ~ - A ~ - Vx 2M c C

Q)'t/I =exp(iQxff1c)I (P - Q)' A 2.\1 c

= exp(iQx/fic)f'ljI .

(30)

l1ms t/I = e"p(iQx/flclljlsatisfies the Schrodmgcr equ81ioll aner the gauge transformation (25). The enerl!)' f' is lIlvariant ullder the tr.Ulsform<ltioli. 111e gaugc b fmsformatioll 011 A mercly changes the loc.,1 phase of-thc wl.\vefulletion.
\\lc
St.C

Ih.lt
(31)

so th.\1 thc ch.uge density is inv.lriant tinder a gauge transformation.

Gauge ill tile London Eqllatioll

Hccaul>e of the equation of continuity in the now af e1eclric charge we rt.'(luire that in a superconductor

divj=O, so that the vector potential in the London equation j = -CA/41TA[ mtlst satisfy div A = O.
(32)

Furtller, there is no current flow til rough a "a(:uuln/superconductor interfa(:<'. TIle nonnal component of tile current across the interfal'e must vanisll: j" = 0, so that the vt."Ctor potenti,11 in tile London cquatiol1 lIlust satisfy
(33)

111e g.'luge of the vector potential in the London equation of superconductivity is to be chosen so tllat (32) and (33) are sntisfied.

APPENDIX II: COOPER PAIRS

For a complete set of slates of a two-electron system that satisfy periodic boundary conditions in a cube of unit volume. we take plane wave product functions
(1)

\Vc assume that the electrons arc of opposite spin. \Ve introduce renter-of-mass and relative c.'tXlnlinates:
R=

K=
so that

Hr, + rJ kl + k2 ;

(2) (3)

(4)

TIllis (I) becomes Ip(K,k;R,r) = exp(iK' R) exp(ik . r) and the kinetic energy of tile two-electron system is
t"K

(5)

+ Ek

= (~l!/m)(iK2

kl!)

(6)

We give special attention to the product functions for which the Ct'nter-ofmass wavev(."Ctor K = 0, so that k l = -k 2 \Vith an intertlCtion HI between tile two electrons. we set up the eigenvalue problem in terms of the expansion x(r) =
r~

exp(ik' r)

(7)

(Jl u

+ II, - ('lxlr)

= 0 = r~

liE" -

()~~

+ lI,l-:s.]l'"p(il..'r!,
f

(~)

wllt'r(' II, i' tIll..' inh'r'll'liolJ tIH.'r~~ of tilt' lwo l'icclnlll". IICf(' \'.1 IlIl_

i\ till' d~('Il'

'\"c t.1Ll' till s<:.lldr pnKlu('-t \\'ltll e\l>(ik rl In ohtain

(I-:~ - f~

r"

(kIJI ,Ik') = 0

IU)

til(' SCI:ul.tr (Cju.alton of tile prohlt'lIl Now tr;ln~ftll"llt til(' SlilIl to an IIlte,;tr.ll:
(0)

.U!tl with Lilletit-

where N(') is til(" Iilllllht'r of two colfi-tmn \t.lte" \1 (th Intdl mlllllcntulll K "" 0 cU('I~Y in clE' at E', Nllw (;ol1~idt, tIl(' matrix dcnH'l1t~ 1/,(,E') = (kIII,lk'l, Studit'" llf tlw.'\( hy Bankt'll sUMest tlmt they art' illlportant whcn the two d ...(.'troll.\ are l"OlI' fined 10 a thin CI1t'r~y sl,cI! near Ih(' fermi "urfact'-wi!!l!n a shcllllfthiekness hw" abo,t E., when' WI' is the IJ<'l)}'t' pltVIlOII eutof( frl'lluent:\_ 'Vl' :I\.\UlIW thai H,(E,E'):. -v (II) for E,E' within thl' shell aJxl zero otllt'rwis~:. I-lert" \' is
a.~1IIn(od

to be pOSilh'l.'.

Thus (10) 1)(,("OO\('s

(E - 'lI.-(E) = \'
with
f ...

= f~

"W,)_

' L "

dE' "E')N(E') = C .

(12)

IIl'rc C is a l'OlIstanl, indl'!x'nclt.nt of J-

from (12) w('

h~lw

geE)
alll!
l = V

=...,t f

(13)

f"
~"

c/E' N(n _,f. f

(14)

'Vith N(E') apl)roximatcly ronstant and ('(111m to Nt- o\,('r Ill(' .small (IIl'r~ r.Ul~e lx.-tween 2.. and 2ft, we take it out of Ih(' intewal to "lll.lln

I = I\t-\

, 'f 2.~
iU~
E"

I clE. , " f..

~":".c----"c = . , 100000--;
f

1\,.'.

(15)

2f"

LeI th(' eigt'Il\;llue

of (15) Ix wriltt.'ll as

=2t"t-!:J.,

(16)

whidl defines the binding ('Ilcrb" tJ. of thl' electulll !l.lir, rdatht, til two (n't.'

658

electrons at the Fermi surface. Then (15) be<-omes

1 = N F '" log
0'

2E", -

2EF

+ 6-

N,.-\I log

(17)

IIN,N = 101;(1 + 2"wvfli.)

(18)

This result for the binding energy of a Cooper p.1.ir may be written as 2hWv 0~ cxp{l/NFV) 1 For V posith'c (attradi\'c interaction) the energy of the system is lowered by excitation of a pair of electrons above the Fermi level. Therefore the Fl.Tmi sea is unstable in an impOt"taflt way. 11lc binding energy (19) is closely reL.'1ted 10 the superconducting energy gap Elf_ 111c BeS calculations show that a high density of Cooper pairs 1fi.1.y Conn in a metal.

APPENDiX 1: Cll\'ZBURC-LANDAU EQUATION

\\'e owe to Cinzburg and Landau an eleg.1l1t the<lry nfthe phenomenology oCthe superconducting slate and oCthe sp.1lial \'8rialion of the order p.1.rametcr in that stale. An extension of the theory by Abrikosov describes the structure of the vortex state which is exploited technologically in superronduc1ing magnets. l11e attraclions orlhe GL theory are the natural introduction of the coherence length aud or the wavdunclion use<l in Ihe theory of the JOH~phson effecls in Chapler L2. We introduce the orde... parameter l/I(..-) \\ilh Ihe property that
l/I*(r)~...)

== 11.5("') ,

(I)

the local concentration of sllpcrroml11cting electronso 'L11C mathematical fonnulation of Ihe i1efinition of thc function ~r) will come out of the BCS theory. Wc first set up a rOnll fOr the fn.'e energy i1ensity Fs(r) in a superconductor as a rUllclion of the order p.1I11mcter. We assumc thai in the general vicinity of the trallsition temperature

with the phenoll1enologicalj)OSitho c constants a. be s."\id. Here:

P. and m,

of which more will

1. FN is the rrcc energy density of the normal state.

2. -aly..j2 + iPltPl-l is a h'pical Landau form (as in 01aptcr 13) for the el.panslo" of the free energy in It.'nus of an order par:lmeter that v:lllishes at a sn'ond-order phase transition. This term nl.lY be viewed as -aIls + U31l~ and b)' itself is a minimum \\ith rcspect to II~ whell risen = alp. 3. TIle term in 19r.K! tJ>Il represents an increase in ene~y caused hy a spatial variation of the order p.1.ramcler. It has the fonn of the kinetic cnergy in quantum Incehanics. I TIle kinetic momcntum - ihV is :IL'COmpauied by the flt'ld momentlllll -<jAll' to ensure the g.1.uge illvari.1.n(;e of the free elwrgy, as in Appcndi1( G. Here <j = -2e for an electron pair. 4. 'I1ll' term - jM . (IBn', with the fictitious magneU:.r..ltion M = (D - D..)/41T, rcpresents the increase in the supercomlucting rree ellcrgy caused by the cxp\llsion of Illa~,'nctic flux from the superconductor.

TIle separate terms ill (2) will be illustrated by examples as we progress further. First let us derive the GL equation (6). We minimize the total free energy jdV Fs(r) with respect to \'ariations in the function I/.(r). We ha\'C 6Fs{r) = (-01/1 + P1iPi 21/1 + (lJ2m)(-ihV - qAlc)I/I' (iliV - qAlc6~ + c.c.
(3)
We integrate by parts to obtain

(4)

if &fI" \'anishes on the bound.1.rics. It follows that

6jdV Fs = jdV W[-av +

J31tPl 2.p + (1I2m)(-ihV

- qAll''fifJ] + c.c.

(5)

This integral is zero if the term in hrackets is zero:

[(1I2mX-ihV - qAle't - a +

P1~21.p =

0 .)

(6)

This is the Ginzburg-Landau equation; it resemhles a Schroc.lingcr equation for .p. Dy minimizing (2) with respt..'Ct to 6A we obtain a b'tl.ugl.....invariant exprt>ssion for the S\lpercurrent flm: js(r) = -(iqfiJ2Jn)(l/?V.p - ",Vl/?) - (t/IrTlC)cJ!*",A . At a fr<."'C surf.1.Ce of the specimen we must dloose the g.1.uge to s,1.tisfy the boundary condition that no current flows out of the superconductor into the \":tCtlllln: n' js ... 0, where n is the surface Ilonnal.

'A ronl"bulion cllhc bm jfMf", ",1... '1'1:: M is d.., mag;r>eli:ralion. ""as illhodun-J by t...K!.1u and Laf$h'l:t "\ ..,,",nl Ihr "'....-hanK<' rnrt~ &n1,t)' in I "",,,'l:lgo~l. scr QTS, P. 6S

o
Coherence Length. TIle intrinstc cohen.~llCe length ~ may be defined from (6), Let A = 0 and suppose Ihat J3hW may be negk'Cled in comparison with 0:. In one dimcllsion the CL equation (6) reduces 10
-2md~=a4J.

"fl If'll/J

(8)

111is has a wavelike solution of the form exp(Wl, wllcre { is defined by

f-

(f1 2/2ma)112 .

(9)

A more interesting special solution is obtained if we retain the nonlinear term Plt/ll2 in (6). Let us look for a solution with l/J :: 0 at x = 0 and with 4'- t/Jo as x-+ 00. 111is situation represents a boundary between normal and superconducting states. Such states call ("()Cxist if there is a magllctk field He in the normal region. For the moment we neglect the penetration oCtile Geld inlo the superronducling region; we lake the field penetration depth A <ll {, which defines an extreme type I supercondu<.'tor.

The solution of
(10)

subject to our boundary conditions, is

IjI{x) = (a/J3)ll2 tttnh(xrV2a

(II)

This Illtt}' be vc...rified by dir(."('t substitution. Deep inside the superconductor we ha,"c I/J., = (cJf3)I/2. as follows from the minimiz.."ltion of the terms _aj1,!lj2 + il3ll/ol 4 in the free energ)'. We see from (11) that! m,ll"ks the cxtent of the coherence of the supen..-onducting wa"cfunction into the nonual region. We h.we sc.'C1l that deep inskle the supc."r(."()!lductor the free energy is a minimum \\'llI~n 1",,~2 = 0.113, so that
(12)

by definition of the thermodynamic critkal field II,. as the stahiliz."Ition free energy density of the slIperrondocting state. it follows th.lt the critical field is rel.lted tn a aod 13 hy
(13)

Consider the penetration depth of a weak mal/;ndic lkld (8 <t fI,.) into a supcrcc.Jl1ductor. We asSUI\1C that Il/oI1. in the superl-onductor is equal to It/1ol 2 the value in the al)sC'llcc of a field. 'nu~o the equation for till' supercorrent (lUA reduces to

Appm.dlr

661

(14)

whidl is just the London equation js(r) == -(d41TA 2)A. with the penetration depth

(15)
'Ole dimensionless ratio K - )Jf of the two characteristic lengths is all import:mt Ih'lrruneter in the thL'Ory of slipercondlllti\'ity. From (9) and (J5) we find
K

== mc qli

(J!...)'~
21T

(16)

We now show that the value K = 1M di\'idL'S type I superconductors (K < I[V2) from type II superconductors (K > 11\12). CalCtllatiOIi ofthe Upper Critical fwld. Supc[cQnducling regions nucleate spontaJ1L'ously within a normal. conductor when the applied magnetic lick! is dLueased below a \-alue denoted by Hc2' At the onsct of superconducti\'ity 140'1 is small and we Iineari7.e the CL equation (6) to obtain
(-ihV - qAlc'flJ1 == a . 2m
TI1C

(17)

magnetic Beld in a superconducting region at the Ollset of supcrcon ductility is just the applied field. so that A == 13(0,1:.0) and (17) bL"'COmes

_..!'-(~ + ..!....)~ + _I_(ih~ + 2m ihfl ar.2 2m ay

qB c

1:)2l/J =

al/J .

(18)

This is of the saml' form a.~ the SchriXlinger L"Quation of a free pal'lide in a magnltic lIeld. We look for a solution in the form e1l:p(i(kll y + k~z)]rp(x) and find
(I/2m)[ -h 2cl 2ldXl

fl2k~

(flk ll - ll13X!C)2]rp = cup.

(19)

this is the equation for ail harmonic oscillator. if we set E "" a (h2!2mX~ + ~ as the eigen\'alue of (1I2m){ -h 2d 2/Jr + (q2B2!c!):Jl - (2f1krlBlc}x]rp = Erp . (20)
TIle tCTllllinear ill x can be transformed away by a shift of the origin from 0 to Xo == hk.IIlBl2mc. so that (20) lx"COmes. wilh X == 1" - xo.

.' dX2 d' + im(qB/rncfX 2 ] Ip = (E + h2kV2m~ , - [ 2m

(21)

The largest value of the magnetic fidd IJ for whieh solutions of(21) C'l:ist is given by the lowest eigcllvalllc, whk-h is
(22)

\\ here

w is the oscillator frequency qnhru:.

B....... H c2
'"

With k= set C<1U.l.1 10 ....erG,

(23)

2amdqf' .

TIlis result may be expressed by (13) and (16) in tl'TlTlS of the thennooynamic critil'al field He an~1 till" CL paramett:r K . JJ{:
1I'i!
e

=- - (4 -rrcJIp) 1/2 = qli

2ame

He

v2-

r;;

me
liq

(7f ~ r;;. Y-:;;;;He; = v2Klfe; 2'17"

(24)

When JJt > 1Iv12. a supe.rCOnthlctor has H('l> lie and is Solid to be of type II. It is helpful to write lid in terms of the flux quantum <1>0 = 2'17"fldq and
~:l = fz'i!l2ma:

(OS)

This tells us that al the upper critical field the flux density II~ in the material is equal to one flux quantum per area 211'f\ collsistent With a fluxoid bUi<:e 5!>a('ing of the order of f

APPENDIX

J:

ELECrRON-PHONON COLUSIONS

Phonons distm-tthe local crystal stnlcture and hl-nce distort the local b'lInl structure. This distortion is senset1 by the ronductlon c1ectrol\s. The important effects of the l-'Qupling of electrons with phonons arc Electrons are scaUerec:l from one state k to :Ul<lther state k'. leading to electrical resistivity.

 Phonons can be absorbetl in the scattering event, leading to the aUenualion of ultrasonic waves. An electron will carry with it a crystal distortion, and the effective m.\ss of the electron is thereby in(:reasell, A cryst..l l disloction associated with one ek'Clron can be sensed by a second electron, thereby causing the electron-electron interaction tl"'ll enters the theory of superconductivity.

Tlw deformation potellti..11 approxiwation is that the electron energy t(k) is coupled to the crystal dilation 6{r) or fractional \'olunle change by
1<.<) = ""')

+ C6)

(I)

whe_rc C is a constant. The approximation is useful for spherical band edges f:(J(1c:) at long pllOlloll wavelengths and low electron concentr<ltiollS. The dilation m.'ly be exprt.'Ssed in tenns of the phollOn operators Qq, of Appendix C by

n:

A(r) = i

L (fJ2.MwJ lq l(llqell:p(iq r) I12

a~ exp(-iq' r)] ,

(2)

as in QTS. p. 23. Here M is the mass of the l'rystal. The result (2) also follows from (C.32) on fonning q. - q._1 in the limit k <t L In the 80m approximation for the scattering we are conccmed with the matrix elewents of CA(r) between the one-electron Bloch statf'S llc:) and [l'), with IJc) = exp(ik r)II~(r), In the wave field representation the matrix element ;,

- a; f tflx u\.II~Ic-Ic'-'ll' ')

where l/i,r) = where cI;,
C\,

L L'.),'PJ,.(r) = L C].Cxp(ik . r)II1r)

(4)

are the fenniOtl creation and annihiL'ltion oper<ltOl"s. The product

1I~.(r)II~(r) involves the periodic p.'lrts of the D10ch functions and is itself periodic

in the lattice; thns the integral in (3) vanishes unless

lc:-k':!:q=!O

vector in the reciprocal lattice.

In semiconductors at low temperatures only til{' possibility uro (. rocesses) may be allowed energetically. Let us limit ourselves to N processes, and for convenience we approxin'k'lte J cf3x 11.'11", h)' unit)'. TIlen the deformation pott'lltial perturbation is
(5)

Jie1axation 'rime, Il) the presence of the electron-phonon interaction the W:l\'cvector ~ is not a <:onstant of the motion for the electron alone, hut the sum of the wave\'ectors of the ek'Ctrol1 and virtual phonon is consel'Vcd. Suppose all electron is initially in the state Ik); how long will it stay in the sanw state? We calculate first the probability w per unit tim<' that the electron ill k will emit a phonon q. If nq is the initi"l.l population of the phonon state.

tt,(k - q;nq

ll~;nq) == (2'7TlfI~{k - q;'1q

+ lIWlk;n...>12B(E. - flw,.l - f.-J .

(6)

by time-dependent perturbation theor),. Here

11le total collision r.lte W of an electron in tht' statc system at absolute zero is, with Ilq == O.

Ik>

with a phonon

w= ~

4'7T,x. _ I

II

d{cos

OJ Lq~ tlq crB(E. -

Ifk- q -

flw.,) ,

(8)

where p is the mass density, 11ll' argument of the delta flll)l'tioll is fl2 fl2 --(2k' q - q~ - flc.cl == --(2k' q - q2 - qqJ 2m 2m (9)

where 'Ie 5 2/ml- c with c. the vclodty of soum1. The minimum value of k for which the argument can he zero is kmll\ == i(q + qe), which for q = 0 re(Iu<:es to k.",n == ~qc = mc,lfl. For this value of k the electron ~roup \'elocity vtl = kml,.lm is equ.'l.! to the velocity of sOllnd, Thus the threshokl for the emission of phonons by electrons in a crystal is that tile electron group velocity should exceed the acoustic velocity, 11lis requirement resemhles the CcrenkO\' threshold for the emission of photons in crystals by fast electrons. The elt.'Clroo energy at the threshold is !'".~ - 10- 27 , 1011 - 10- 16 erg - 1 K. An elC<'tron of energy below this threshold \\;11 not be slowed down in a perfect crystal at absolute zem, e\'en b)' higher order electron-pllOilOn interactions, at least in the hannonic apprm.imation for the phoilOlIS,

For k

qr W,

.1)' neglect the qq,.lerm ill (9). 111(' integrals ill (8) bet'Olilc

-\

d~ J dq q3(2m.'fl,2(1)jj(2kp.

- q) = (Bm*lh 2)

~)

[I

(/p, f,..2.Jl.2 = 8m*J,..2/3h 2

(10) and the pllOllon emission r,de is W(emission)

2C l m*k'l
37fpc,fJ l

(II)

direetly proportiOlJ.l1 10 the ('Icdroll energy t~. TIle los~ of the eompool:nl of w;\vevedor Il<'\rallcl to the ori~inal direction of tile ele<.'tron when a phonon is emitted at an angle 8 to II is Kivell by q cos O. 111e fractional rate of loss of k~ is given by the transition rate inlegrHI Witll Ihe extra factor (qtk) <.~)s 8 in the integrand. Instead of (10), we have

(2111*/h 2 k)

lip.

sl2p.! = 16m*"2l511 2

(12)

so that tile fractional rate of de<:rcasc of k~ is

"

W(k,) = 4C 211l*k2{57TPC.!l2

(13)

This quantity ellters into the electrical resistivity. 111(' aoove results apply 10 ubsolutc zero. At a telllper;ttlire Kill' ~ he,k the integrllted phonon emission rate is W(cmission) =
C~IIi"KKn'r
,2:)

we,ph

(14)

For electrons ill then11;\1 equilihriullI at not too Inw tcmpcnltuH-'s the rcquired inequality is easily satisfied for the nns vOIllic of II. If we take C = 10- 12 crg; Ill'" = 10- 27 g; K = 10' cm-I; c, = 3 X Hi' tm S-I; p = 5 g cm-:\ then \V:l< IOlt S-I. At absolute zero (l3) gives W:l< 5 X 10 1 ".s I with tl]('st, S.1I11(' parameters.

Subject Index

Ablolute thermoe1ed:ric

...... 228
Aa:eptor lOniutiorl eoerg:ies.

Band gap. 175, 199 &nd structure, CaAs. US germaruum. 215

Barium titanate, 'Jf11

Acc:eptor stales, 224 Aa!ustica1 phonon lnnch.

ActMtion energy, 5+f,

, _.328

'"

..

Wigner-Sein, 8, 38 Centipoise, S?:7 Central equation, 186 Cesium chlonde struct\Ire.

A.Ifvtn

43'waves. """"'"" aysbh.

302
Alb1i halide 7:J
Alloys. 611

--

Basis. 4, 5 Binding. erystalliDe. 54 Biomllgnetism, 478 BIoc:h equations. 492 BIoc:h &equrney. 230 Bloch functiorl. 179, 246

Chalmgenide gIasse$. 531

'86

Olarge density waves, 301

~

potmtial, 146, 153.

laNe,

BIoc:h 0IIeIlbt0r. 229 Bloch 1'"' ~. 4S6 Bklch theorem. 179, ISS
Bloch

wall.

0assieaI distnbution. 648 Oopton rdation, 349 a..usi:us-Mc.otti rebboo, :tOO

Coertive

l5<. '69

472

Foree. n:WI particle,

4~

Bohr~.421

.72
Coertivity. 470,

-.gtl>. 600
tnns:Ition metal, 624

Boltzmann transport equation.

6<,

Coberenoe length. 352, 660

intrinsic, 3:i3 Cohesive energy, 57. 66. 80
Fo:rmi gas, roum metal,

Alnico V. 476 Amorphous solids, 521 fetTomagnets. 529

$emiconducton, 530
Anharmonic interactions, 129 Anisotropy energy. 470, 471

Boson operators, 641 Botmdary conditions. periodic, 120 Bragg law, 29

1'ro

Annihilation operators, 641

Antiferroelectrlc CI'yIbh, 404, AntiferTomagnetk: Cl')'$tW.

..,

406

Bravais lattice. 10, 556 Brillouin function, 422 Brillouin scatterirlg, 306 Brillouin zone. 37, 217, 266 first. 39, 100. 176
\'Olume. 51 BuDe moduli, table. 59

2.51 square well potential, 266

Cohesive energies, table. 51 Colfuionless electron guo 311

Color centen, 548 Compreuibillty. 86 Compretsibillties. table, 59

Concentration, table. 24 Conduction band edge, 199

resonance, 506, 50S Anti-Stokes line, 323

Atomic form actor, 4.5 Atomic radii, 77

-,.

magnanl. 466

BuDe modulus, 86
Bulk modulus, electron gas. F c:rmi gas. 303
8urgIers vector, 592

'69

Conductioa e~

Coodu<tion """""" r~m,438

SlHOeptiblity,433 Conductivity. dectricIl. J.:I6, 158. 221 impurity. 222
ionic,

Cauchy Ultegral. 309

Causality, 329

8CS theory, 345, 354 energy PI! equation, 301 Band bending, $76

Cell, pmUtM, 6 urnt,6

131. 166 Conductivity pUll I'U1e. 330

thenna1.

298

...
.97

,""'"
Meets. ~ . 499 Ddicil semiconductors., 222 Degeneraer, 145 Degene""le semiconductor,

Conductor. perfe, 339 Configur.dional !leu capacity. 6:lO Contact hyperfine interaction, Continuum WIIV,", equation, 112 Cooper pairs. 356. 656 Covalent bond. 72 Covalent aystak, 72 Creation opentors, 641
C~ep.

Dos', bone orbit, 263

Domains, origin of, 414 closure, 414

Planck, 101

2B6
Demagnetiz.ation, isentropic, ." Demagnetilation facton, 504 Density, table, 24 Density of modes (lee density of stales), Drns.ity of states, 128, 149 Debye model, 120 Einstein mooel, 124

ferfoelectric. 406
magnetic bubble. 481 Donor ionization energies 224 Doo1or mtes, 222 Doping. 222 Dulong and Petit value. 127

603

Critical curn-nu. 372 Critical field, 339, 372 thin llIllls. 375 Critical magnetic field, 372 C"tical point ~ents. 445 Cntical points, 312 Cntical shear $treSS. 588 Cntical ternpen.ture, 399 Crystal firld, 426
splitt:ing.426

general result. 127

one dimension, 116 singularity, 139 three dimensio~. 119 Depolarization facton. 384 DepolariDtion 6dd. 384 Diamagnetism, .417 DIamond structure, 19, 51,
Oie~ constant,

Edge dislocation, 589 Effective mass. 209, 210. 212 negative. 211 semiconductor" I:lble, 212, 21. tensor, 210
Einstein equation, 561 Ei~tein model, density of states. Elasnc compliance, 83 Elastic limit, Elastic stiffness, 83, 91 Ela.stic strain, 80 ElaJlic waves. 87 Elastic wave quantization.

......... ,..

Crystlll growth, 25 Crysral m(lmf:ntum. 109, 186 CryJtal structure. elemenU.

388 semiconduclOn, 223, 330

'905

23 Crystalline binding, 54 Crystallography, surface, S56 Cubic lattices. 12 Cubic zinc sulfide .tructure.
20 Curie constant. 423 1a...,423 temperature. 443 Cune-Wciu law, 444 Cyclotron frequency, 164 Cyclotron resonance, 213 spheroidal energy surface.

Dielectric function, 307 electron gas. 271 Lindh:l.rd, 282

Thomas-Fermi, 282
Dielectna, 381 Dielectnc wsceptibility, 385

E1e<:tric field, klc:a1, 386

macroscopic, 382

DifIT3ction, Josephson
junctiool, 36Q

Electric qtadrup&e moment.

503
~-e,

Diffraction cood.itiOllS,

231
Dangling bond5, 554 Oavydov splftting. 331 Drformation. 80 De Haas-v;an Alphen effect.

periOd. 266
Ocbye model, density of states, 122Dehye scattering, 522 Debye law, 124 Debye temperature, 122 table. 126 Debye-Wallef" factor, 632

""

-r

34 DifFusion .495, 544 particle, 648 Diffusivity, 544 Dilation. 82 Direct pp 5emiconduetor, 213 Direct photon ab$orption. 201, 202 Dislocation densitie$, :598 Dislocation multiplication. 600 Dislocations. 589 Dispenion relation, phonons, 100, 104 electromagnetic waves, 273 Displadve transition. 394, 398 Dissipation sum nt.Ie, 329 Distribution. classical Fermi-Dine. 130

Electrical conductivity. 156, 651 table, 160

Electrical I"C$UIivity. 159 table, 160 Electromechanical coupling factor. 410 Electron affinities, 68 Electron c:ompound, 615, 616 E~D density distribulioD.. 67 Electron heat C31*'ity, table,

ElectrQI). orbits, 242 Electron spin resonance, 48S Electron work functions. 560 table, 562

"'-,

Elect:ron-elect:roll
204 204

coI.lisicm$,

Electron-e1ect:rc:Jn intenction,

Electron-hole drops, ~ ,321 Electron-lattice interaction,

WI

Electron-phol1on collisions, 662 Electronic poIariooilities, table, 391 Ekctroni<: structure, su~, 560 Electronatic screening, 280 Elementary exdtatiolls, 9S Empirical pse"dopoteutial method, 253 Empt)' core model, 253 Empt)' lattice approximation,

FactOl', atomic iOrm, 43, <IS struchtre, 44 Fermi energy, 145 gas, 141, 146 level, 144 liquid, 294 su~, copper, 262 su~, gold, 262 surface parameters, table,
238 Fermi-Dirac: distrib"tion, 146, 643 l'~ermions, heavy, 156, 337 FcrroelillSticity, 410 FerroelectriCity cnterion, 412 crystals, 381, 393 domains, 3S8 linear array, 414 Ferromagnetism, crystals, 443, 449 amorphous, 529 conduction electron, 438 domains, 468 order, 458 resonance, S03 Fiber optics, 535 1""*'5 law, 5 Fine slruc:ture oonstant, S66 First Brillouin zone, 39, 100, 176 First-order transition, 402 Flux quautilation, 356 Fluxoid, 358, 363 Fourier analysiS, 30, 42, IB3 FractionaJ quantized Hall effect (l'QHE), 570 Frank-Read source, 600 Free-energy, superconductors, 343 Frenkel defect, 542, 551 Frenkel ~on, 314, 316 Friedel osciDations, 2S3 Fullerenes, ':.IT3 J<'used quartz, 524
su~s,

GLAG tlteory, 360 Glass, 5Z1 chaJcogenide, 531 tralllition temperature, 527 Crain boundary, small-angle,
595

1'"

Croup velocity, 102 Criineisen COIlstant, 129 Cyromagnetic ratio, 420 Gyroscopic equation, 490 Hagen-Rubens relation, 335 Hall coefHcient, 164 table. 167 Hall effect, 164, 373 two carrier types, 231 Hall resistance, 165 Hall resistivity, 564 Hardness, 606 He!, liquid, 170 Heat capwSty, COrtfJgurational, 630 electron gas, 151 glasses, 533 metals, 155 phonon, 117 superconductors, 342 Heavy fennlons, 156, 337 Hei$Crtherg model, 446 Helicon waves, 303 Heterojunclion, .576 HeterostruCiures, 574 Hexagonal close-packed structure, 17, 25 Hexagonal lattice, 51 High tempernture supcr<.'onductors. 371 Hole, equalimt of motion,
200,208

ISS
ENDOR,500 Energy band calculation, 244 Energy oonds, 173, 175 Energy gap, 173, 175 superconductors, 344 table, semiconductors. 201 Energy levels, one dimension, 144 Energy loss, fast particles,
326

Energy loss sped ra, electrons, 279 Entropy, superoonductors, Entropy of milting, 620 Equation of motion, hole, 208 Equations of motion, electron, 203 ESCA,324 Eutcclics, 622Ewald construction, ':.IT sphere, 559 sums, 634 E\\'J.Id-Komfeld method, 637 E.u:hange energy, 446 field, 443 frequency resonance, 516 integJlII, 443 interaction, 74 narTowing. 510 Excitons, 312 UcIusion prindple, Pauli, 64 uotic superconductors, 156 Extended zone scheme, 23S Extinction coefficient, 307 Extremal orbits, 260

"'2

Gap, direct, 202 Gap, indirect, 202 Gap pl:umons, 304 Gauge trlll1sfolmation, 654 C!magnetism, 478 Gin7Dul'g-Landau equation, 65S Ginzburg-Landau par.tmeter,

Hole orbits, 242 Hole trapping, 29S Hookes law, 83 HTS, ':.IT1 Hume-Rothery rules, Hund rules, 424 If )'l::Irogen bonds, 76 H)<pemne COrtstant, 498 Hypemnc effects, ESR in metals, 515 Hypemue splitting, 497 Hysteresis, 475 Ilmenite, 396

F centen, 499, 545, 552

354, 161

610

Indn

Impurity conduetl\ity, 222 Impurity orbits, 230 Index s}'slcm, 12Indirect gap, 2.02. Indirect photon absollltion.

201
Inelastic scattering, P~OIlS, 110 Inert gas crystals, table, 60 Injection lASer, 517 Inslllatol'S. 193 Interoond tTllllSitions, 312 Intcrface plasmons, 30'2 Interfacial polarization, 412 Intrinsic carricr concentratKin, 200, 216 Intrinsic coherence length,
353

Kohn aoomaIr, 113 Kondo effect. 628 Kramen-Hei5Pnbcrg dispersion, 309 Kramers-Kromg relations, n,310 Kronig-Pennt'Y model, 180. 195 reciprocal space, 187
LA modes, jO-l

La modes, 104 Local elcctric field. 3S6 London equation, 349, 655 Loradon gauge, 656 London penetration depth, 353, 373, 377 Longitudinal plasma oscillations, 276 Longitudinal I c1axa~Lon time,

490
Lagrangian equations, 652 Landau gauge, 570 Landau level, 268, 278, 566 Lanwu theory, fel'mi liquid, 294 Land..l U theory, phase transition, 400 Landau-Cinzburg equations, 354 Lange\in diamagnetism. 417 Langevin rf:'!ult, 419 Lanthanide group, 425 LanllOr frequency, 418 Larmor theorem. 418 Laser, 510. Sit injection, SiS ruby, 512 scm iconductor, 5T1 Lattice, 4 Bravais, 10 Uitlice constants, equilibrium, 65 Lattice frequencies, table, 292 Lattice momentum, 205 Lattice sums. dipole arra}'s, 637 Lattice translation operator, 647 Lattice types, 8. 10 Lattice vacancies. 541 Lattices, cubic, 12 Laue equation, 36, 581, 631 L.tw of mass action. 219 LCAO approximaUoll. 245 LEED,557,558 Lt:nnard-Jones potential. 63, 64 Lenz's law, 417 lindhard dieLectric function. Long-range order, 6Z0 Lorentz field, 38S Lorellz 1ll1l\\I~r, 168 Low-angle groin houndary, 595 LST relation, 291, 294 Luminescence, 315 Lyddane-Sachs-Teller relation, 291

Intrinsic mobility, 220 Intrinsic scmiconductDl'S, 199 Im"erse spinel, 460 lank: bond, 66 lonk: chancter, table, 76 Ionic conductivity, 298.542 Ionic CI')'Stals, 66 lonk: radii. 79 table. 75 loniution, thermal, 226 Ionization eneTgies, acc:qxor, 225 donor, 224 table, 61 Iron gamets, 461 IrOll group, 425 lsentropk: demagnetialtioll. 431 Isotope effect, supen:onduclol'll, 346, 347 Isotope effect, therrndl conductioll, 138

Madelung constant. 69 Madelung energy, 68 Magnetic breakdown. 268 Magnetic bubble domains,

<SO
Magnetic IOrce microscopy. 478 Magnetic moTlOlJOle. 35S Magoetk: resonance, 487 Magnetic resonance tomography, 487 Magnetic susceptibility, 417 Magnetite, 461 Magnetization, 417 saturation, 446
~'lagnetO(:onductivity.ten~r,

jaJm-Teller effect, 221, 429

tllll'ping, 5:H Josephson tunneling. 367, 369, 377 Jump frequency. 546 Junction, SUper<lOnducting. 364 Junctions, p-n, 570

k p perturbation theory, 007 Keh'in relation, 22S Knight shin, 500, 502

28'
Une v:idth, 4!}.1 Uquid He3 170 Uquidus, 622

170, 271 Magnt:tocrystlllline enel'gy, 471 Magnetoelastie CQuplin~, 479 Magnetogyric ratio. 420 Mllgtleton numben. 425 Magneloplasma frequency. 303 Magnetoresistance, 564 two carrier types, 231 Magnetotaxis, 478 Magnoll5, 450 antirerroma~,.netie, 466 dispersion relation. 481 Maser action, 510 Matthicssen's rule, 161 Maxwell equations. 3SZ Mean field approxlmatKin, 443

I"""
Mean free palla. phonon. 233 MeiS$ner effect, 337. 338 sphere, 3T1 Melting points, table, 58 Melt-spin velocity. 529 M.... 556 Metal-inrulator tnlnsition, 284 Metals, 75, 141 Metal spheres, 171 Metg.las, 530 Miller indices (M'. indeJI s).,tem), 12 Mobility, intnnsic, 220 Mobility edge., 5(i9 Mobility gap. 569 Mobilities, table, 221 Molecular crysU1s, 318 Molecular field, 443 Molecular hydrogen, 13. is Momentum CT)"Ibl. 109, 186 field. 652, 61W lattice. 205 phonon, lOB Monopole, magnetic. 358 Nuclear l>aramagnetiml, 417 Nuclear Quadrupole resonance, 503 inelastic scattering. 110 mean free path, 133 mebU,2B6 modes. soft, 113 momentum. 108 Photovoltaic detectun, 573 Piewelectricity, 408 Planck distribution. 116 Plasmon frequency, 98, 273.
278

'"

MOSFET,564.583,584
Motional narmwtng, 494. 496 Mott exciton, 314 Mott transition. 284, 320 MottWannier excitons. 318

Ohm's law, 156 Open orbits, 242, 266 magnetoresistance, 267 Optical ab!orption, 200 Optical. ceramics, 410 0pti<0I phonon bnn<h. 10< Optical phonons, soft. 400 Orbit, dog'. hone, 263 Orbits. electron, 242 extremal, 260 hole, 242 open, 266 Order, Iong.range, 620 short.range, 621 Order panmeter. 658 Order-disorder transformation, 617 transition, 394 OlCillations, Friedel, 283 Oscillator strength. 392
p-n junctions. 510 Paramagnetic defects, 499 Panunagnetism. 417, 420 conduction electrons. 433 nuclear, 417 Van Vleck, 430 Particle diffusion, 648 Pauli exclusion priociple. 63 Pauli SplIl magnetization, 435,

np p.-oduct, 219 l"anostnJctures, 168 I"early free electron model,

276
I"~I

interface. 302 mode, sphere, 303 optics, 272 o$Cillatmn, 'J:T7 surface, 279, 302 table. 278 Poi3e, 527 Poisson equation, 272 Poisson's ratio, 94 PoLaritom, 2B7 Polarizability, 388, 3B9 conducting sphere, 412 electronic, 390 Polarization. 381 interfacial. 412 satuntioo, 412 polaron. 2f1i. 299 coupling constant, 298 Polytypism, 22 Power absorption. 493 Primitive cell, 6, 193 Pseudopotential, components.

2B3
meta1.lic sodium, 2S4 method. 252, 312 Pyroelectric. 394 PZT system, 409
Quantiz.ation. elastic ...-ave. 107, 638 Hall dfect. 566 orbits in magnetic Geld. 2S5 spin wave., 454 Quantum cor-nU. 581 Quantum interference, 366. 369 Quantum solid. 92 Quantum they, diamagnetism, 419 paramagnetism, 420 .oJ Quasicrystals, 48

temperature, 463, 465,

Nee! wall, 483

Negative effective mass, 211 Neutron diffi-action, 50. Ill, 456 scattenng, 454. 456 spectrometer, 111 NMR tomography, 487 Noncrystalline solid, 519 Nonidea1 structures, 21
Normal mode enumeratiOn,

48'

438, 502
Peierls instability, 300 iRliulator, JOl Peltier coefficient, 228 Penetration depth, Loodon, 353 _ Periodic boundary conditions,

120
Periodic zone scheme. 237 Perovsldte. 396 Persi.!itent currents. 359 phase diagrams. 6U5, 622 Phase tr.uujtions. stT1JCtur.I1. 393 Phonons, 107, 108, 117 coordinates, 639 dispersion relatioru, 127 gas. thermal resistivity, 133 heat capecity, 117

Nonnal processes, 136 "onnal spinel, 460 I"uclear demagnetization, 432 Nuclear magnetic resonance, table, 489 Nuclear magneton, 488

II'

48'

QuasIFenni levels, 'S17 Quasiparticles. 294

oching. orbital angular mflfllcntum, 426

1:::,5&i
ial di~tribuHon function, 521 ,an ('fect, 322 -at!"rlng, 306 dom network, 526 dOln ~taeking. 22 ~ ("'11.11 ions, 423 ' ..: 'h "ttenmltion, 536 , . dllattice p()ints, 3S ,pro.,lI lattice \'t'Cton, 33 nmbination radiation, 321 onstlllction, 554 1I1lc;';tion, 570

Schottky barrier, 573 Schottky defect, 541 Schottky \...ancies, 551 Scn:ened Coulomb potential,

2S2
Scr"Cning. electrostatic, 2SO Screw disklcatiOCl, 590 S~nd-olOO- transition, 402 Sdf-diffusion, 547 Self-trapping,. 22l S"miconductor, d"gcnerate,

2S6
Semiconductor crystals, 174,

199
Semiconductor las"rs, 577 Semiconductors, d"fidt, 222 Semimetals, 174, 2:03, 228 SMar conslant, 94 Shear strength, single

,JC("<1 23:;

WIle ~me,

195.

<rymk,5S7
Shear st:rcm, critical, 5S7 Short-range order. 621 Shubnikmo.de Hll35 "f&:ct, 5S3 Single domain particles, 477 Singularities, Van Hove, 129 Sicgularit}, density of states,

"'d.ll>Ce, 293, JIn. 30S ectirJII, normal incidence,

ectivity coeffICient, 307 index, Y11 LU'Kln, 556 re.:t, 492 rbach, 492 otm:m,492 ,mtion time, ngituJinaI, 490 .in-buice, 490 lancuce, 470 sta"ce, surfllCC, 171 :.tallt~ per square, 171 st.,-ity, electlnl, 158, i59 sth-ity mtio, 162 lOm", c1t'ctl"OlL gas, 304 )OIl~C function, 309 lIm,l!ion, 516
"actiH~

""'.

Stress component, .,i, Structural phase tnmsitions, 393 Stnx:tu~e l'aetor, 42, 635 bee lattice, 44 diamond, 51 fee lattice, 45 Substrate, 555 Superconductivity, table, 336 exotic, 156, 337 type I, 340 type II, 340, 360 Superlauices, 618, 630 Superparamagn"tism, 477 Sunace crystallogrnphy, 5S6 electronic structure, 560 nets, 558 plasmons, 279, 302 resistance, 171 states, 562subbands, 562 U:msport, 564 Susceptibility, dielectric, 3S5 magnetic, 417 Symmetry operation, 4

13.
Slater-Pauling pim:, 627 Slip, 5SS Sodium chloride structure, 15 Sodium meb1, 142 Soft modes, 400, 413 Solar ccns, 573 Solidus, 622 Solitons, 473 Solubility gap, 622 Spectroscopic splitting factor, 429 Spheres, metal, 171 Spinel, 459 Spin-lattice inter-letion, 491 Spin-lattice relaxation time, 400 Spin wa,"e (SI?e also magnon),
450

mD,560
,o,ln-Dushman "'-(I"alion, 562 Y theory, 6/2.S ing coordinate system.
31!)

, laser, 512
ration magnet:izatkm, 446 ration polarization, 412 ration, rf, 516 nillg tunn"ling microscopy, 20

quantization, 454 resonance, 506 Square lattice, Stability critf:ria, 95 Stabiliution free ellCrgy, 347 Stacking auk, 589 SfM,20 Stok~ line, 323 Strain component, 51 Strength of allo}'s, 600

TA modes, 104 Taxis, 4i5 Temperature, Deb}"e, 122 T"mperatur" dependence, reflection lines, 633 Tetrahedral angl", 25 111ermal conducti\"ity, 131, 16. glasses, 534 isotope efrect, 138 metals, 166 table, 126 -nlermal dilation, 139 'nlermal effective mass, 156 -nlCrmai excitation of magnons, 455 Thermal expansion, 130 Thennal iOlliution, 226 llJermal resistivity, phOClOll gas, 133 Thermionic "minion, 561 Thermodynamics, superconducting transition, 347 Thermoelectric effects, 227 11lomas-Fermi approximation,

260
1l>omas-Fermi dielcetri<: function, 252

,,,de>;

673

Ib~ce-Ievcl mase~, 51~

I1ght binding method, 245 TO mocles, 10-1 TomograpllY, m:lglletic ~esonance, 342, 4$7 Transistor, 1'.105, 564 Transition, displadvc, 39-1 Ilrst-order, 40'2 metal-insulator, :2S4 ordcr-disorde~, 394 second-order, 402 Transition metal alloys, 624 Trausition temperature, glass,
527

Two-nuid model, 372, 376 Two-level system, 440 'lYre I supe..condllctors, 340 Type II sllpe.-conductors, 340, 300 Ultraviolet tl'3nsmission limits. 275 Umklapp processes, 135, 162 Unit cell, 6 UPS, 324 Upper critical field, 661 Valence band edge, 199 Vau del' \Valils interaction, 54, 56, 304 Van Hovc singularities. 129 Van Vk'Ck paramagnetism, 430 Vector potential, 651 Vickers ha~dness number, 606 Viscosity, 527 Vitreous silica, 524 Vortex state, 340, 361, 376
\\'Hnnie~ \Vannie~

periodic lattice, 184 periodic potential, 183 Wa\'eve(:to~s, energy gap, 196 Weiss field, 443 Wllisken, 604 \Viedemann-Fl'3l1z law, 166 Wigner-Scih: boundary - C(lndition, 250, 251 Wigner-Seitz l.'CII, 8, 38, 251 \\hgne~-Seitz method, 241, 2lS Work function, 560 Wo~k-hardening, 603 XPS, 32<1

Transitions, displacive, 398 Translation opemtion, 4 Translation "ecto~, 4 Tmmp~lIency, alkali metals, 274 TranspOIt, surface, 564 Transport equation, Bo1t?mann, Transverse optiC'<II modes, plasma, 274 Transverse relnxation time,

Young's modulus, 94 Yttrium iron garnet, 505

b~eakdown, 231 tunneling, 230 Zero-point lattice displacement. 139 Zero-point motion, 55, 62 l.ero-field splitting, 510 ZnS structure, 20

Zener
Zelle~

492
Triplet excited states, 437 Tunneling, 364 )osepIISQll, 366 Tunneling microscope, 5SO Twinning, 589

functions, 267 I:l(lcler, 230 \Vave eqllation, continuum, 112

Table of 51 Prefixes
Fraction

Symbol

10-'5
10- 12

10-11 10..6 10- 3

HI

10'

10' 1012

femto pico nano micro milli kilo mega giga tera

r
p

" ~
m

k
M

Table of Values
Quantity Velocity of light Proton charge Planck's consmnt

Symbol
c

Value

CCS
10'0 em s
IO- IQ esu

51

2.997925 1.60219

to" m 5-' 10-" C

10-3< J S 10-34 J s

4.80325

"
.V

tl ""' h/2r.

6.62620 1.05459
6.02217 x 1023 11101-'

to-%'! erg s

to- 17 erg s

Avogadro's numbc.r Atomic man unit

Electron re~t m3H Proton rest mass Proton massfeleclmn mass Reciprocal fine structure constant tlc/e' Electron radius e'/mc' Electron Compton wavelength Mme Dohr radius h'/me' Dohr magncton eM'2mc Rydberg constant mc"'2Jl 1

:lmu m
.U~

Mplm

1.66053 9.10956 1.67261 1836.1

10-2< g 10-8 g
(0-"

lO- n kg 10-3 ' kg

10-" kg

"0
" '. ".
'.

137.036 '
2.81794 3.86159
10-13 em

10-'3 m
lO- lJ m

10- 11

em

5.29177
9.274lO 2.17991
13.6058 eV

11.. or Ry

10-' em to-" erg G-' 10- 11 erg

10- 11 m 10-11 J T-'

lO-l$

I electron volt

'V

eV/h eV/I,c eVlk B

1.60219 2.4(797 x W" Hz

8.06546 1. t6048 x to" K

10- 12 erg 10' em-I

10-"

10" m- 1

Boltzmann constant Pemllttivity of free space Pemleability of free space

k.
~

1.38062

10- 16 erg K-'

I I

""

lO-t:I J K 107/41Tc' 4rr x 10-

J p},y."-:.!t.nd

SOK~. II 'I T.,-r<><. IV H P~,k~f . nd D 'I u~gtn~,g. R~" 'rod fhy. 41. J7~ (1969) s ....I.., ..11. Co""~ a.,.l B 101 T"y......
C...,,,,,,,.lll~k... ~

0.1. 2(4). 663 (11I73)

J""".