Sie sind auf Seite 1von 150

Thesis for the degree of Doctor of Philosophy

Multipactor in Low Pressure Gas


and in Nonuniform RF Field
Structures
Richard Udiljak
Department of Radio and Space Science
Chalmers University of Technology
Goteborg, Sweden, 2007
Multipactor in Low Pressure Gas
and in Nonuniform RF Field Structures
Richard Udiljak
c (Richard Udiljak, 2007
ISBN 978-91-7291-885-6
Doktorsavhandlingar vid Chalmers tekniska hogskola
Ny serie nr 2566
ISSN 0346-718X
Department of Radio and Space Science
Chalmers University of Technology
SE412 96 Goteborg
Sweden
Telephone +46(0)31772 10 00
Cover: Susceptibility chart for multipactor in a waveguide iris for ve
dierent height/length-ratios.
Printed in Sweden by
Reproservice
Chalmers Tekniska Hogskola
Goteborg, Sweden, 2007
Multipactor in Low Pressure Gas
and in Nonuniform RF Field Structures
RICHARD UDILJAK
Department of Radio and Space Science
Chalmers University of Technology
Abstract
Resonant electron multiplication in vacuum, multipactor, is analysed
for several geometries where the RF electric eld is nonuniform. In par-
ticular, it is shown that the multipactor behaviour in a coaxial line is
both qualitatively and quantitatively dierent from that observed with
the conventionally used simple parallel-plate model. Analytical esti-
mates based on an approximate solution of the non-linear dierential
equation of motion for the multipacting electrons are supported by ex-
tensive particle-in-cell simulations. Furthermore, in a microwave iris the
electrons tend to perform a random walk in the axial direction of the
waveguide due to the initial velocity distribution. The eects of this phe-
nomenon on the breakdown threshold are analysed. The study shows
that the threshold is a function of the height-to-length ratio of the iris
and for a xed value of this ratio, the multipactor susceptibility charts
can be generated in the classical engineering units. Using the parallel-
plate concept, the multipactor threshold in low pressure gases has been
analysed using a model for the electron motion that takes into account
three important eects of electron-neutral collisions, viz. the friction
force, electron thermalisation, and impact ionisation. It is found that
all three eects play important roles, but the degree of inuence de-
pends on parameters such as order of resonance and secondary emission
properties. In addition, a new method for detection of multipactor is
presented. By applying a weak amplitude modulation to the input signal
and performing a fast Fourier transform on the detected signal, accurate
and unambiguous measurement results can be obtained. It is demon-
strated how the method can be used in both single and multicarrier
operation.
Keywords: Multipactor, discharge, breakdown, microwave discharge,
nonuniform elds, coax, iris, low pressure gas, detection methods.
iii
iv
Publications
This thesis is based on the work contained in the following papers:
[A] R. Udiljak, D. Anderson, P. Ingvarson, U. Jordan, U. Jostell, G. Li,
M. Lisak, L. Lapierre, J. Puech, and J. Sombrin, New Method
for Detection of Multipaction, IEEE Trans. Plasma Sci., Vol. 31,
No. 3, pp. 396-404 , June 2003.
[B] R. Udiljak, D. Anderson, M. Lisak, V. E. Semenov, and J. Puech,
Multipactor in low pressure gas, Phys. Plasmas, Vol. 10, No. 10,
pp. 4105-4111, Oct. 2003.
[C] R. Udiljak, D. Anderson, M. Lisak, V. E. Semenov, and J. Puech,
Improved model for multipactor in low pressure gas, Phys. Plas-
mas, Vol. 11, No. 11, pp. 5022-5031, Nov. 2004.
[D] R. Udiljak, D. Anderson, M. Lisak, J. Puech, and V. E. Semenov,
Multipactor in a waveguide iris, accepted for publication in IEEE
Trans. Plasma Sci.
[E] R. Udiljak, D. Anderson, M. Lisak, V. E. Semenov, and J. Puech,
Multipactor in a coaxial transmission line, part I: analytical study,
accepted for publication in Phys. Plasmas
[F] V. E. Semenov, N. Zharova, R. Udiljak, D. Anderson, M. Lisak,
and J. Puech, Multipactor in a coaxial transmission line, part
II: Particle-in-Cell simulations, accepted for publication in Phys.
Plasmas
v
Conference contributions by the author (not included in this thesis):
[G] R. Udiljak, G. Li, D. Anderson, P. Ingvarson, U. Jordan, U. Jostell,
A. Kryazhev, M. Lisak, V. E. Semenov, Suppression of Multi-
pactor Breakdown in RF Equipment, RVK 02, June 10-12, 2002,
Stockholm, Sweden.
[H] R. Udiljak, D. Anderson, U. Jostell, M. Lisak, J. Puech, V. E. Se-
menov, Detection of Multicarrier Multipaction using RF Power
Modulation, 4th International Workshop on Multipactor, Corona
and Passive Intermodulation in Space RF Hardware, 8-11 Septem-
ber, 2003, ESTEC, Noordwijk, The Netherlands.
[I] J. Puech, L. Lapierre, J. Sombrin, V. Semenov, A. Sazontov,
N. Vdovicheva, M. Buyanova, U. Jordan, R. Udiljak, D. Anderson,
M. Lisak, Multipactor threshold in waveguides: theory and ex-
periment, NATO Advanced Research Workshop on Quasi-Optical
Control of Intense Microwave Transmission , 17-20 February, 2004,
Nizhny-Novgorod, Russian Federation.
[J] R. Udiljak, D. Anderson, M. Lisak, J. Puech, V. E. Semenov, Mi-
crowave breakdown in the transition region between multipactor
and corona discharge., RVK 05, June 14 - 16 juni, Linkoping
[K] D. Anderson, M. Buyanova, D. Dorozhkina, U. Jordan, M. Lisak,
I. Nefedov, T. Olsson, J. Puech, V. Semenov, I. Shereshevskii,
R. Tomala, and R. Udiljak, Microwave breakdown in RF equip-
ment., RVK 05, June 14 - 16 juni, Linkoping
[L] V. E. Semenov, N. Zharova, R. Udiljak, D. Anderson, M. Lisak,
J. Puech, and L. Lapierre, Multipactor inside a coaxial line,
5th International Workshop on Multipactor, Corona and Passive
Intermodulation in Space RF Hardware, 12-14 September, 2005,
ESTEC, Noordwijk, The Netherlands.
[M] R. Udiljak, D. Anderson, M. Lisak, V. E. Semenov, and J. Puech,
Microwave breakdown in low pressure gas, 5th International
Workshop on Multipactor, Corona and Passive Intermodulation
in Space RF Hardware, 12-14 September, 2005, ESTEC, Noord-
wijk, The Netherlands.
[N] C. Armiens, B. Huang, R. Udiljak, D. Anderson, M. Lisak, U. Jostell,
and P. Ingvarsson, Detection of Multipaction using AM signals,
vi
5th International Workshop on Multipactor, Corona and Passive
Intermodulation in Space RF Hardware, 12-14 September, 2005,
ESTEC, Noordwijk, The Netherlands.
vii
viii
Contents
Publications v
Acknowledgement xi
Acronyms xiii
1 Introduction 1
2 Multipactor in vacuum 5
2.1 Single Carrier . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.1 Basic theory . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Hybrid modes . . . . . . . . . . . . . . . . . . . . . 17
2.1.3 Factors eecting the threshold . . . . . . . . . . . 18
2.1.4 Methods of suppression . . . . . . . . . . . . . . . 20
2.1.5 Eect of random emission delays and initial veloc-
ity spread . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Multicarrier . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3 Design guidelines . . . . . . . . . . . . . . . . . . . . . . . 28
2.3.1 Single carrier . . . . . . . . . . . . . . . . . . . . . 29
2.3.2 Multicarrier . . . . . . . . . . . . . . . . . . . . . . 29
3 Multipactor in low pressure gas 35
3.1 Simple Model . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.1.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.1.2 Multipactor boundaries . . . . . . . . . . . . . . . 37
3.1.3 Main results . . . . . . . . . . . . . . . . . . . . . . 42
3.2 Advanced Model . . . . . . . . . . . . . . . . . . . . . . . 44
3.2.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.2.2 Analytical formulas for argon cross-sections . . . . 47
3.2.3 Multipactor boundaries . . . . . . . . . . . . . . . 49
ix
3.2.4 Key ndings . . . . . . . . . . . . . . . . . . . . . 52
4 Multipactor in irises 59
4.1 Model and approximations . . . . . . . . . . . . . . . . . . 60
4.2 Multipactor regions . . . . . . . . . . . . . . . . . . . . . . 64
4.3 Comparison with experiments . . . . . . . . . . . . . . . . 65
4.4 Main results . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5 Multipactor in coaxial lines 69
5.1 Analytical study . . . . . . . . . . . . . . . . . . . . . . . 70
5.1.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.1.2 Multipactor resonance theory . . . . . . . . . . . . 73
5.1.3 Main ndings . . . . . . . . . . . . . . . . . . . . . 81
5.2 Particle-in-cell simulations . . . . . . . . . . . . . . . . . . 82
5.2.1 Numerical implementation . . . . . . . . . . . . . . 82
5.2.2 Simulations . . . . . . . . . . . . . . . . . . . . . . 83
5.2.3 Comparison with experiments . . . . . . . . . . . . 88
5.2.4 Main conclusions . . . . . . . . . . . . . . . . . . . 91
6 Detection of multipactor 93
6.1 Common Methods of Detection . . . . . . . . . . . . . . . 93
6.1.1 Global methods . . . . . . . . . . . . . . . . . . . . 94
6.1.2 Local methods . . . . . . . . . . . . . . . . . . . . 99
6.2 Detection using RF Power Modulation . . . . . . . . . . . 100
6.2.1 Single carrier . . . . . . . . . . . . . . . . . . . . . 105
6.2.2 Multicarrier . . . . . . . . . . . . . . . . . . . . . . 105
6.2.3 Main achievements . . . . . . . . . . . . . . . . . . 106
7 Conclusions and outlook 111
References 115
Included papers AF 123
x
Acknowledgement
I wish to thank Prof. Dan Anderson and Prof. Mietek Lisak for accept-
ing me as a PhD student and for guidance and support in my daily
work. I also want to thank Prof. Vladimir Semenov at the Institute
of Applied Physics in Nizhny Novgorod, Russia, for fruitful discussions
and for his patience with all my questions. Thank you Prof. Lars Elias-
son, Director at the Institute of Space Research in Kiruna, for providing
both nancial and moral support making my PhD candidate appoint-
ment possible. A very warm thank you also to Jerome Puech for many
interesting discussions about space related microwave problems and to
his employer, Centre National d

Etudes Spatiales, for nancial support.


Thanks to my friends in Toulouse and especially Dr. Omar Houbloss and
Raquel Rodriguez. I thank my fellow members of the National Graduate
School of Space Technology and my colleagues at Chalmers and espe-
cially Dr. Pontus Johannisson, Dr. Ulf Jordan and Dr. Lukasz Wolf for
benecial discussions and lots of support with Linux and LaTex. Many
thanks also to our secretary Monica Hansen for guiding me through the
administrative jungle. I want to thank my dear mother, Monica, for
encouraging and supporting me and my family when 24 hours a day
wasnt enough and my father, brother and sisters for believing in me. I
am also grateful to my unocial mentors: my father-in-law Lars-Goran

Ostling, my friends Jorgen Otback and Anders Wilhelmsson, and my


brother-in-law Nicklas

Ostling. Most of all I thank my wife Malin and
our daughters Janina and Lizette for encouragement and support during
this time.
xi
xii
Acronyms
AM amplitude modulation
DC direct current
DUT device under test
EDDM electron density detection method
ESA european space agency
FFT fast fourier transform
NLSQ non-linear least square
PIC particle-in-cell
PSK phase-shift keying
QPSK quadrature phase-shift keying
RF radio frequency
SEY secondary electron yield
SMA sub miniature version a
TGR twenty gap crossings rule
TEM transverse electric and magnetic eld
TWTA travelling wave tube amplier
UV ultra violet
VSWR voltage standing wave ratio
WCAT worst case assessment tool
xiii
xiv
Chapter 1
Introduction
Resonant secondary electron emission RF discharge or multipactor was
discovered and studied by Philo Taylor Farnsworth in the early 1930s.
The phenomenon was then used as a means to amplify high frequency
signals as well as to serve as a high frequency oscillator. Using his mul-
tipactor tubes, Farnsworth succeeded in developing the rst electronic
television system. The success stimulated other researchers to investi-
gate the phenomenon and one of the rst detailed analyses were done
by Henneberg et al. [1] in the mid 1930s. Gill and von Engel [2] made
an even more detailed study, both theoretical and experimental, where
they, among other things, showed the importance of the secondary elec-
tron yield on the development of the vacuum discharge. In a follow up
paper [3] Francis and von Engel studied not only the initial stage of the
electron multiplication, but also the saturation stage. They showed that
the electron space charge eect could be one of the major causes for the
discontinued electron growth. Other researchers continued the work and
a basic overview of these results can be found in the two review papers
by Gallagher [4] and Vaughan [5].
During the past 20-30 years, multipactor has mainly been studied
due to the adverse eects it can have on microwave systems operating
in a vacuum environment. It can disturb the operation of high power
microwave generators [6] and electron accelerators [7], but, above all, it
can cause severe system degradation and failure of satellites, which are
dicult or impossible to repair after launch [8]. Satellites operate under
vacuum conditions and the most common means of communication with
the Earth is microwave transmission. Microwave frequencies are required
as the ionosphere is not transparent for low frequency radiowaves. In
1
addition, it is dicult to make compact, light weight, and high gain
antennas for low frequency transmission. Many microwave components
are hollow metallic structures that guide the electromagnetic power. A
free electron inside the device will experience a force due to the electric
eld and since there is no gas or other material stopping the electron,
it can accelerate to a very high velocity. Upon impact with one of
the device walls, the energetic electron can knock out other electrons
and under certain circumstances this procedure is repeated continuously
until the electron density is large enough to counter-act the eect of the
applied electric eld and a steady state is achieved. A consequence of
this can be that the incident power is reected instead of transmitted
to the intended load. Since many satellites lack sucient protection
against reected power in order to save weight, such reected power can
cause severe damage to the high power stage of the system.
When a satellite is launched it carries fully charged batteries and
in the beginning, before the solar panels are deployed, they are the
only source of electric power. The capacity of these batteries is usually
low and may only last a couple of days, since a satellite in operation
will normally only lack access to power from the solar panels for a few
hours, at most, and consequently the batteries are made small in order to
save weight. If the solar panels are not deployed before the batteries are
exhausted, the satellite is permanently lost. Thus, a new satellite is often
taken into operation quickly after being put into orbit. A concern then
is that the satellite components may not be completely vented and there
is a risk for ordinary corona breakdown, which is more prone to occur
at intermediate pressures than at high and very low pressures. A corona
discharge is usually much more detrimental than multipactor and for a
certain range of pressures, the breakdown threshold for corona is lower
or much lower than for multipactor. Basic theory for ordinary corona
microwave discharge, when the mean free path of the electrons is smaller
than the characteristic length of the device, is well known [9]. However,
the intermediate range, between very low pressure and vacuum, has
received little attention and therefore one of the main topics of this thesis
is devoted to explaining what happens with the breakdown threshold at
these pressures.
Theoretical studies of the multipactor phenomenon have to a great
extent been performed using a one-dimensional model with a spatially
uniform approximation of the electromagnetic eld. However, many
common RF devices involve structures where the eld is inhomogeneous,
2
where breakdown predictions based on such simple models will not be
reliable. Examples of important geometries in microwave systems where
the eld is inhomogeneous are waveguides, coaxial lines, irises and sep-
tum polarisers. An important eect due to the non-uniform eld that is
not present when the electric eld is spatially uniform, is the so called
ponderomotive or Miller [10] force, which tends to push charged particles
towards regions of low eld amplitude. This can have both a qualitative
and a quantitative eect on the multipactor regions. Analytical study of
resonant multipactor in a non-uniform eld is not a trivial matter and
most researchers have resorted to numerical methods of investigation.
However, in this thesis dierent aspects of multipactor in structures
where the eld is inherently inhomogeneous are investigated using ana-
lytical methods and the results are compared with numerical simulations
as well as with experimental data found in the literature.
Many microwave systems of today operate in multicarrier mode,
which means that several signals at dierent frequencies are transmit-
ted simultaneously. In contrast to the single carrier mode, the electric
eld envelope of the multicarrier system varies constantly. In most cases
this is advantageous from a multipactor point of view, as the changing
amplitude will destroy the resonance condition and thus suppress the
discharge. In systems where the frequency separation is small, however,
there is a risk that the signals will interfere constructively for a large
number of eld cycles and the amplitude will remain fairly constant,
thus allowing a discharge to develop. In such cases, the microwave en-
gineer will have to try to nd the worst case scenario and design the
component with respect to that case or perform tests that guarantee
that the part fulls the requirements. Some attention will be given to
these aspects in this thesis, which can be useful for the engineer when
making multipactor free multicarrier microwave designs.
The thesis is organised as follows. A general introduction to basic
theory of multipactor in vacuum is given in chapter 2 as well as some
guidelines when it comes to multipactor free design. It will serve as a
base when continuing with the analysis of multipactor in low pressure
gas in chapter 3, where rst a simple model is presented, which only
considers the friction force of the gas molecules. It is then followed by
a more advanced model, which includes also the eect of impact ionisa-
tion as well as thermalisation of the electrons. Starting with waveguide
irises in chapter 4, vacuum discharge in structures where the eld is
non-uniform is considered. It is followed by a detailed analytical and
3
numerical study of the phenomenon in a coaxial line in chapter 5. Fi-
nally, chapter 6 is devoted to dierent means of detecting multipactor
with special focus on detection by means of RF power modulation.
4
Chapter 2
Multipactor in vacuum
Multipactor normally occurs at microwave frequencies, i.e. at 300 MHz -
30 GHz. When discovered by Farnsworth in the 1930s, he applied the
technique to amplify an electric current. Others have also tried to nd
useful application of the phenomenon, e.g. in multipactor duplexers and
switches [11] and in electron guns [12, 13]. However, during the last 30
years it has mainly been studied due to its detrimental eects on mi-
crowave components. It has been found to cause electric noise, which
reduces the signal to noise ratio, a very serious problem if it occurs
e.g. in a communication satellite where the signal power is limited and
counter-measures are dicult or impossible to implement. It can also
detune microwave cavities, commonly used as e.g. resonators in lters,
thus reecting the incoming power back to the power amplier. If the
system does not have an appropriate power protection device, the am-
plier may suer permanent damage. Another concern is heating, which
is a result of the power dissipated to the device walls as the multipact-
ing electrons strike the walls. Furthermore, the discharge can also cause
direct physical damage to the component with the risk of permanently
changing the electric properties of the device. However, the risk of such
direct damage seems low, especially for metallic components. In cases
where damage has been reported, it is not certain that it was caused by
multipactor [14, 15]. Multipactor is known to be able to trigger ordinary
gas discharges [1618], either by increasing the outgassing from the com-
ponent or just by starting the breakdown at a pressure and a voltage
where corona is not expected, since gas breakdown can be sustained at a
much lower voltage than what is needed to initiate breakdown directly.
Corona discharges are much more energetic and are known to be able
5
to physically damage microwave components. Many researchers thus
suspect that the observed damage was due to a multipactor induced gas
discharge.
This chapter will present the basic theory of vacuum multipactor
between two metallic parallel plates with an applied homogeneous, har-
monic electric eld. It is divided into two major parts, one describing
the single carrier case and another devoted to multicarrier multipactor.
2.1 Single Carrier
One of the rst communication satellites, Telstar I, operated in single
carrier mode. It had a capacity of 12 simultaneous telephone conversa-
tions [19] and the solar panels provided a power of only 15 watts. Today,
satellites operate in multicarrier mode with powers of several kilowatts
and new satellites are being designed for tens of kilowatts. Thus, the
single carrier mode is seldom found in real applications. Nevertheless,
the single carrier case is important as it has been thoroughly studied
over the years and by making certain assumptions, the multicarrier case
can be approximated by the single carrier state and design and testing
can be done based on the simpler situation.
2.1.1 Basic theory
There are two main kinds of multipactor, the single-surface and the
double-surface types. Single-surface multipactor can occur in structures
with nonuniform eld or with crossed electric and magnetic elds [20],
where the electron, accelerated by the electric eld, returns to the orig-
inal surface due to the circular motion caused by the magnetic eld.
This thesis, however, will focus on double-surface or parallel plate mul-
tipactor, but some attention will be given to single-sided multipactor in
the case of a coaxial line.
A multipactor discharge starts when a free electron inside a mi-
crowave device is accelerated by an electric eld. In a strong eld the
electron will quickly reach a high velocity and upon impact with one of
the device walls, secondary electrons may be emitted from the wall. If
the eld direction reverses at this moment, the newly emitted electrons
will start accelerating towards the opposite wall and, when colliding with
this wall, knock out additional electrons. As this procedure is repeated,
the electron density grows quickly and within fractions of a microsecond
a fully developed multipactor discharge is obtained (see Fig. 2.1).
6
Figure 2.1: Initial stage of parallel plate multipactor, where a free electron
is accelerated by the electric eld and is forced into one of the
plates, where it causes emission of secondary electrons.
The motion of an electron in vacuum with an applied electric eld
can be studied by means of the equation of motion,
m x = eE (2.1)
where m ( 9.1 10
31
kg) and e ( 1.6 10
19
C) are the mass and
charge of the electron, x the direction of motion, and E the electric eld.
Multipactor requires an alternating eld and in the parallel-plate model
a spatially uniform harmonic eld E = E
0
sin t is assumed. Solving
Eq. (2.1) with this eld yields expressions for the velocity, x, and the
position, x,
x =
eE
0
m
cos t +A (2.2)
x =
eE
0
m
2
sin t +At +B (2.3)
where A and B are constants of integration, which will be determined by
the initial conditions. By assuming that an electron is emitted at x = 0
with an initial velocity v
0
when t = /, fully constrained expressions
for the velocity and position are obtained, viz.
x =
eE
0
m
(cos cos t) +v
0
(2.4)
x =
eE
0
m
2

sin sin t + (t ) cos

+
v
0

(t ) (2.5)
For resonant multipactor to occur it is necessary for the electron to
reach the other device wall (x = d) when t = N + , where N is an
7
odd positive integer (N = 1, 3, 5 . . .). Applying this resonance condition
to Eq. (2.5), the following expression is obtained for the amplitude of
the harmonic electric eld.
E
0
=
m(d Nv
0
)
e(N cos + 2 sin )
(2.6)
An important quantity when studying multipactor is the impact ve-
locity, since this determines the secondary electron yield. It can be found
by inserting t = N + in Eq. (2.4), which yields
v
impact
=
2eE
0
m
cos +v
0
(2.7)
Multipactor boundaries
When constructing the multipactor boundaries, i.e. the boundaries of
the regions in parameter space where multipactor can occur, an assump-
tion will have to be made concerning the initial velocity. In reality, the
initial velocity of the emitted electrons will follow some kind of distri-
bution and a common choice is the Maxwellian distribution [21],
f(v) exp

(v v
m
)
2
2v
2
T

(2.8)
where suitable parameters for the mean velocity, v
m
, and for the rms-
value (or thermal spread), v
T
, have to be chosen. When performing
particle-in-cell (PIC) simulations, such a distribution can be used to
more accurately describe the initial velocity of the emitted electrons.
However, for an analytical solution a simpler assumption will have to
be made. There are two common approaches, one which assumes that
the electrons are emitted with a constant initial velocity, v
0
, regardless
of the impact velocity. The other assumes that the ratio between the
impact and initial velocities is equal to a constant, k = v
impact
/v
0
. Both
these approaches will be used and compared in this chapter, but in the
following chapter, which deals with multipactor in low pressure gas, the
constant k approach will be used only for the simple model while the
constant initial velocity approach will be used for the more advanced
model as that assumption is more physically correct. The reason why
the constant k model has been used to such a great extent is the fact
that it can successfully be tted to experimental data. The cause of this
success will be explained in the subsection on hybrid modes below.
8
In addition to fullling the resonance condition, which resulted in
Eqs. (2.6) and (2.7), the secondary electron yield, SEY or
se
, must be
greater than or equal to unity. For most materials the secondary yield as
a function of the impact velocity (or the impact energy, W = mv
2
/(2[e[))
has the same shape (see Fig. 2.2), even though the absolute values vary
very much between dierent materials. The impact energy where the
secondary yield rst reaches unity is called the rst cross-over point and
is denoted W
1
, after that the yield increases and reaches a maximum
at W
max
and the energy at which the yield drops below unity again
is called the second cross-over point, W
2
. Below W
zero
no secondary
yield is obtained [22, 23]. However, some researchers have published
measurements of the SEY, which indicate that it is possible that the
SEY does not drop to zero below a minimum impact velocity [2428].
On the contrary, it can increase after reaching a minimum yield and even
reach a yield close to unity for very low impact velocities. A yield close
to unity implies that the electron does not produce any secondaries, but
rather that the electron bounces o the surface. This could have an
important eect on the multipactor threshold and development, but in
this thesis, the model by Vaughan [22] has been used unless otherwise
specied.
By setting the impact velocity, Eq. (2.7), equal to the rst cross over
point (converted to velocity, v
1
) and taking Eq. (2.6) into account, the
resonant phase, , can be found as a function of d,
tan =
1
2

2d N(v
1
+v
0
)
v
1
v
0

. (2.9)
Using this result, the amplitude can be plotted as a function of d or
fd using Eq. (2.6) (or Eq. (2.7)).
One nal thing that need to be conrmed before drawing the mul-
tipactor charts is the non-returning electron limit. If the secondary
electrons are emitted before the electric eld reverses, the electrons will
be retarded by the eld and if the velocity is low, they are likely to
return to the wall of emission and thus being lost as their energy is too
low to produce new secondaries. The limit can be found by solving the
following system of equations:

x = 0
x = 0
(2.10)
An analytical solution to this system of equations is not possible and in
order to establish the non-returning electron limit, either a numerical
9
0 500 1000 1500 2000 2500
0
0.5
1
1.5
2
2.5

se
=1
W
1
W
max
W
2
Secondary electron yield
Primary electron energy [eV]

s
e

[

]
Figure 2.2: Secondary electron yield as a function of the impact energy. Plot-
ted using the formula for secondary electron yield presented in
Ref. [22]. Parameters used are: W
max
= 400 eV,
se,max
= 2,
and W
zero
= 10 eV.
10
solution or some kind of approximate solution will have to be used. In
Ref. [21] an approximate formula for the non-returning electron limit is
given and re-writing it for the constant v
0
approach yields,

min
=

16v
0
5v
0
+ 3v
impact
(2.11)
Using Eqs. (2.6), (2.9), and (2.11) with v
1
equal to the velocity corre-
sponding to the unity secondary electron yield, the lower multipactor
threshold can be plotted. However, multipactor breakdown is possible
also for impact velocities greater than v
1
, in fact, for all impact velocities
between the rst and second (v
2
) cross-over points, the phenomenon can
occur, i.e. for
v
1
< 2V

cos +v
0
< v
2
, (2.12)
where Eq. (2.7) has been re-written using the oscillatory velocity V

=
eE
0
/m. Thus, in order to construct the complete multipactor bound-
aries, the thresholds for a number of dierent energies between these two
points should be determined and then the envelope of all the thresholds
will be the complete multipactor susceptibility zone (see Fig. 2.3). Fur-
thermore, each order of resonance, N, will have its own zone and, as
shown in Fig. 2.3, the zones become narrower with increasing N. This
type of chart, based on the assumption of constant initial velocity, will
be referred to as a Sombrin chart, since J. Sombrin is one of the major
advocates of this assumption [29].
Using the other approach, with a constant ratio between the impact
and initial velocities, k = v
impact
/v
0
, the formulas for the resonant phase,
Eq. (2.9), the amplitude, Eq. (2.6), and the non-returning electron limit,
Eq. (2.11), will have to be slightly re-written:
tan =
1
k 1

kd
v
1
(k + 1)N

2

(2.13)
E
0
=
m
2
d
e(
k+1
k1
N cos + 2 sin )
(2.14)

min
=

16
8 + 3(k 1)
(2.15)
Using these formulas, multipactor charts similar to the one in Fig. 2.3
can be produced, cf. Fig. 2.4. When this is used, the charts are com-
monly referred to as Hatch and Williams charts, as they were the rst
11
10
0
10
1
10
2
10
3
10
4
Frequency Gap product [GHzmm]
V
o
l
t
a
g
e

[
V
]
N=1
N=3
N=5
N=7
N=9
Figure 2.3: Multipactor susceptibility chart based on the constant initial ve-
locity approach. Parameters used are: W
0
= 3.68 eV, W
1
=
23 eV, W
2
= 1000 eV, and N
max
= 9.
12
who produced charts of this type [30]. Characteristic for the Hatch
and Williams charts are that the multipactor zones are wider than the
Sombrin charts for increasing voltage. This occurs since a constant
v
impact
/v
0
implies that v
0
increases as the impact velocity increases.
When e.g. W
impact
= 3000 eV, this means that for k = 2.5 the initial
energy W
0
= 480 eV, clearly an unrealistic initial velocity.
10
0
10
1
10
2
10
3
10
4
Frequency Gap product [GHzmm]
V
o
l
t
a
g
e

a
m
p
l
i
t
u
d
e

[
V
]
N=1
N=3
N=5
N=7
N=9
Figure 2.4: Multipactor susceptibility chart produced with the constant k
assumption. Parameters used are: k = 2.5 (corresponding to
an initial W
0
= 3.68 eV when W
impact
= W
1
), W
1
= 23 eV,
W
2
= 1000 eV, and N
max
= 9.
By knowing the secondary electron emission characteristics of a ma-
terial as given by the parameters W
1
, W
2
, and W
0
or k, multipactor
charts for that material can be designed. However, Woode and Pe-
tit [31] performed a large series of multipactor experiments during the
1980s and used the Hatch and Williams charts to t the experimental
data. By tuning the k and W
1
parameters for each zone, they were able
to produce multipactor charts that t the experimental data quite well
(cf. Fig. 2.5). The problem with this empirical approach is that it has
to assume dierent values for the rst cross-over point for each zone in
order to obtain good tting. This is clearly an unphysical approach and
will not contribute to an improved understanding of the phenomenon,
13
even though it may be sucient from an engineering point of view. On
the other hand, the higher order modes have a narrower phase-focusing
range (see below), which makes it dicult to compensate for e.g. initial
velocity spread, and thus a secondary yield of unity may not be su-
cient to sustain a discharge. Consequently, an impact energy somewhat
higher than the rst cross-over point will be needed when constructing
the lower multipactor threshold for the higher order modes. This will
be discussed further in the subsection Eect of random emission delays
and initial velocity spread.
10
0
10
1
10
2
10
3
10
4
Frequency Gap product [GHzmm]
V
o
l
t
a
g
e

a
m
p
l
i
t
u
d
e

[
V
]
Figure 2.5: Hatch and Williams charts for aluminium together with measure-
ment data by Woode and Petit [31].
When the microwave engineer assesses the risk of having a multi-
pactor discharge, it is usually not the boundary of the individual break-
down region that is considered. Typically, the lower envelope of the all
the zones is taken as the design threshold (cf. Figs. 2.3 and 2.4). By
setting the phase, , in Eq. (2.7) to zero, the lowest eld amplitude to
achieve a certain v
impact
is obtained. Thus the lower envelope, which
is the same for both the Sombrin Chart and the Hatch and Williams
chart, is given by,
E
0
=
(v
1
v
0
)m
2e
. (2.16)
14
Phase-focusing
In the previous subsection a mechanism called phase-focusing [1, 5, 32]
was mentioned and in multipactor theory this is an important concept.
In order for an electron to be a part of the discharge, it must have
a phase close to the resonant phase as given by Eq. (2.9) or (2.13).
Due to delays between the impact and emission of a new electron or
a spread in the initial velocity, an electron will always acquire a small
phase error. Inside the phase-focusing range, such an error will decrease
as the electron traverses the electrode gap. In other words, the phases
of the electrons will tend to converge towards the resonant phase, thus
keeping all electrons close together. Outside the range of phase-focusing,
the error will grow with each passage and after one or a few transits
the electron will be lost. In order for a discharge to occur under such
circumstances, the impact energy has to be large enough to produce a
secondary yield suciently above unity to compensate for the incurred
losses.
To see in what range the phase focusing mechanism is active, a small
phase error can be introduced in Eq. (2.5) while keeping the amplitude
and phase constant and setting x = d. The ratio between the nal and
initial error is called the stability factor, G [33], and the condition for
stable phase is:
[G[ < 1 (2.17)
By setting [G[ = 1, the phase range within which the phase is stable can
be obtained. An interesting observation here is that even though the
lower multipactor threshold for the constant k theory and the constant
initial velocity model are identical, the range of stable phases varies sub-
stantially [34]. This can be seen clearly from the analytical expressions
for the phase stability limits, which for constant k theory reads,

R
= arctan(
2
N
(
v
impact
v
0
v
impact
+v
0
)) (2.18)

L
= arctan(
2
N
) (2.19)
and for the constant initial velocity approach,

R
= arctan(
2
N
) (2.20)

L
= arctan(
2
N
(
v
impact
+v
0
v
impact
v
0
)) (2.21)
15
where
L
and
R
are the left and right limits respectively. This dierence
is illustrated graphically in Fig. 2.6. However, when v
0
< v
impact
both
approaches yield the same phase stability limits.
10
0
10
1
10
2
10
3
Frequency Gap product [GHzmm]
V
o
l
t
a
g
e

[
V
]
Unstable phase range (constant v
0
)
Stable phase range (constant v
0
)
Unstable phase range (constant k)
Stable phase range (constant k)
N=1
N=3
N=5
Figure 2.6: Lower multipactor thresholds in vacuum for the rst 3 orders
of resonance (N = 1, 3, and 5). The curves for the constant k
model are plotted slightly oset as the curves otherwise overlap.
Parameters used are: W
1
= 23 eV , W
0
= 3.68 eV, and k = 2.5.
Saturation
In order to sustain a multipactor breakdown, the secondary electron
emission yield must be greater than or equal to unity. If the yield is less,
the electron number will quickly decrease and the discharge disappears.
With a
se
greater than unity the electron number will grow rapidly with
each impact and if no saturation mechanism is considered the number
of electrons after a time t, if the eld frequency is f, will be:
N
e
(t) = N
e
(0)(
se
)
2ft
N
(2.22)
The rapid growth of the number of electrons can be illustrated with
an example. Suppose
se
= 1.5 and f = 2 GHz, then the number of
16
electrons after 20 ns for the rst order of resonance with one initial
electron will be more than 10
14
.
In a very short time, the number of electrons will grow to very high
values and it is clear that some kind of saturation mechanism will be-
come active. Two main saturation processes have been described in the
literature. The rst is the space charge eect [3], which is the most
obvious eect. Within the electron bunch the individual electrons will
repel each other causing a change in phase of those electrons and if the
phase error is too large, the probability of losing electrons increases and
eventually the eective secondary yield will be equal to unity and satu-
ration has occurred. The second type of saturation process [35, 36] can
set in if the discharge takes place inside a resonant cavity. Due to a high
Q-value, the electric eld strength is high and thus the risk for a dis-
charge will increase. If a multipactor discharge is started, the electrons
traversing the gap make up an alternating current, which loads the cav-
ity. Loading the cavity means that the Q-value will decrease and thus
also the electric eld strength. It is clear that this is a self-suppressing
eect. As the multipactor current increases, the eld strength decreases
and with it the impact velocity of the electrons leading to a lowered
secondary emission yield. Eventually the secondary yield reaches unity
and saturation has been reached.
2.1.2 Hybrid modes
It may seem somewhat contradictory to assert that the model based on
a constant initial velocity is more physically correct than the constant
k theory, when the latter approach can be better tted to experimen-
tal data. However, as briey mentioned previously, the reason for this
paradox is the hybrid modes. Some of these modes were identied by
Refs. [29, 37, 38] and a general treatment is given in [39]. The modes can
be found by allowing N in the resonance condition for Eq. (2.5) to be
a sequence of odd half-cycles of the electric eld, N
1
, N
2
, N
3
..., where
N
1
= N and the remaining N
n
N for the hybrid modes between the
N
th
and (N + 2)
th
zones. Each such sequence will result in a narrow
multipactor zone located between the main multipactor areas. The low-
est order hybrid mode in the parallel-plate case is the 1, 3 mode, which
means that the transit time in one direction takes 1/2 RF-cycle and the
return transit takes 3/2 RF-cycles. This mode is then also associated
with two dierent resonant phases, viz.
1
= 0 and
2
= /3 [39].
This mode can be found between the two rst classical resonance zones
17
(cf. Fig. 2.7). When taking the envelope of these zones, the dierences
in the right boundaries of the main zones, between the constant initial
velocity and the constant k approaches, become negligible. This can be
seen clearly in Fig. 2.7 [40]. The existence of the hybrid modes requires
Figure 2.7: Vacuum multipactor with the main zones as well as a few hy-
brid zones [40]. With the envelope of the hybrid zones included
the resemblance between the Sombrin chart and the Hatch and
Willams chart (cf. Fig. 2.5) is striking.
phase stability, just as for the classical zones discussed above. However,
the width of each hybrid zone is very small and thus it is very sensitive
to an initial velocity spread. On the other hand, there are many hybrid
zones very close to one other and this spread will result in a mixing or
overlapping of the resonances [39].
2.1.3 Factors eecting the threshold
There are many dierent aspects that need to be considered when de-
termining the multipactor threshold. The most important and most ob-
vious ones are type of material, gap size, and amplitude and frequency
of the electric eld. These are all part of the basic theory as described
above. Apart from these there are other more or less important fac-
tors. The supply of primary electrons does not eect the theoretical
threshold, which can be determined with methods described earlier in
this thesis. Nevertheless, a weak source of seed electrons can result in
an apparent higher threshold during testing. In a typical test setup for
determination of the breakdown amplitude, an electric eld is applied
18
and the eld strength is increased at regular intervals. If no electron
is in an advantageous position, i.e. has a suitable phase from a multi-
pactor point of view, when the right amplitude is set, a discharge will
not occur. As the amplitude is increased further, the impact velocity
of any free electrons, also those that are not in a favourable position,
will be high and the secondary yield will be an additional source of free
electrons. Thus the chances of getting a breakdown increases until it
eventually occurs. For experimental use, a hot lament or a radioactive
source can be used to produce a sucient amount of free electrons to
achieve reliable measurement results [14].
Another factor that can have a signicant eect on the threshold
is contamination. Both the rst cross-over point, W
1
, and the max-
imum secondary yield,
se,max
, can be drastically aected. A lowered
W
1
means that a discharge can occur at a lower voltage and an increased

se,max
can result in a faster growth of the total number of electrons.
In Ref. [31] a detailed analysis of the impact of dierent types of con-
taminants was made. It was noted that the plastic bags, which were
normally used to protect the microwave components from dirt, were the
main source of contamination. A threshold reduction of up to 4 dB
was found. Also dust and ngerprints were a direct source of a low-
ered threshold. In the report [31] it was recommended that cleaned
microwave parts for space use should be handled with cotton gloves and
stored in hard plastic boxes.
Microwave parts which have not been properly vented before power
is applied can also have a threshold that is dierent from the expected
multipactor threshold. If there is too much gas, corona breakdown may
occur, and within a certain range of pressures, close to the minimum of
the so called Paschen curve, the breakdown threshold can be signicantly
lower than in the multipactor case. In the pressure range correspond-
ing to the transition region between corona and multipactor, a higher
threshold can sometimes be expected. More details about this will be
presented in chapter 3.
Other factors that can have an indirect eect on the multipactor
threshold are the voltage standing wave ratio, VSWR, and the temper-
ature. If the VSWR is greater than what was intended with the design,
the peak eld strength in the system will also be greater than expected
and thus a discharge may occur at a lower power level than assumed. An
increased temperature can lead to increased outgassing from the device
walls resulting in concerns similar to those of improper venting.
19
2.1.4 Methods of suppression
Many of the factors mentioned in the previous section that aect the
multipactor threshold can also be utilised to suppress the discharge. The
without doubt easiest method of avoiding a breakdown is to pressurise
the component. The eld strength required to achieve breakdown at
atmospheric pressure is in general much higher than at low pressures or
in vacuum. However, such a method is seldom feasible for components
that will be used e.g. in space, where the external environment is a high
vacuum. A small leakage can lead to slow venting of the component and
thus risking severe corona discharge when the pressure reaches the range
where the minimum breakdown eld occurs.
Another way of suppressing multipactor is to amplitude modulate
the main carrier [21, 41]. If both signals are sinusoidal, the total eld
can be written:
E
tot
= E
1
sin
1
t +E
2
sin
2
t (2.23)
This means that the envelope of the signal will vary according to (see
also Fig. 2.8):
E
env
=

E
2
1
+E
2
2
+ 2E
1
E
2
cos (
1

2
)t (2.24)
When the total eld strength is well above the multipactor threshold
(see Fig. 2.8), the secondary electron yield will increase quickly accord-
ing to Eq. (2.22). However, as soon as the voltage drops below the
threshold again, the electron loss will be large and according to Ref. [21]
all electrons will be lost in just a few RF cycles. However, whether or
not this is true also depends on the secondary yield properties of the
electrode material. For materials with a very high maximum secondary
yield, the number of electrons gained while above the threshold can be
greater than the losses incurred while below. In such a case, no sup-
pression is achieved and in some cases, the discharge may even become
more powerful than before the modulation carrier was added [42] (cf.
Fig. 2.9). Thus in order to successfully suppress a multipactor discharge
using amplitude modulation, it is vital that the material has a low max-
imum secondary yield (preferably less than about 1.5). Due to the risk
of contamination, which can greatly increase the maximum secondary
yield, great care should be taken to assure a high level of cleanliness if
this method of suppression is to be used.
To AM-modulate the carrier is probably not feasible in most cases, as
it would require extra hardware to produce the AM-signal. However, the
20
0 1 2 3 4 5 6 7 8 9 10
0
0.5
1
1.5
Amplitude Modulation, E0/E1=0.4
A
m
p
l
i
t
u
d
e
Time
Main signal
Sum signal
Envelope
Modulation signal
Multipactor threshold
Figure 2.8: Two signals and their sum signal (absolute values). The enve-
lope varies and is partly above and partly below the multipactor
threshold.
typical bandpass ltering of a PSK (Phase-Shift Keying) signal causes
modulations in the time-domain. The QPSK (Quadrature Phase-Shift
Keying) signal in Fig. 2.10 has unity amplitude before ltering. After-
wards, the peaks are higher than the original amplitude, but the troughs
can sometimes go down almost to zero amplitude. Comparing this with
the AM-suppression, it is clear that an electron avalanche that is ini-
tiated during the peak periods, will be extinguished as the amplitude
falls close to zero. However, for a typical PSK-signal, the duration be-
tween phase shifts (which normally coincides with the troughs) is several
hundreds of RF cycles. Thus for most microwave systems, there will be
ample time for a discharge to develop. But, when the amplitude drops
below the threshold, the electron bunch will disappear and when the
amplitude increases above the threshold again, there may not be any
seed electrons present to restart the electron avalanche. Thus, the sys-
tem will have sporadic discharges, which, if they do not occur too often,
may not seriously degrade the signal.
A very common way of suppressing a vacuum discharge is to apply
a coating [2628], a surface treatment [43], or a lm [44] that has a
high rst cross-over point as well as a low maximum secondary electron
21
0 1 2 3 4 5 6 7 8 9 10
x 10
4
85
80
75
70
65
Amplitud Modulation, w2/w1=1.16)
N
o
i
s
e

(
d
B
m
)
0 1 2 3 4 5 6 7 8 9 10
x 10
4
46.5
47
47.5
48
P
o
w
e
r

#
1

(
d
B
m
)
0 1 2 3 4 5 6 7 8 9 10
x 10
4
20
0
20
40
P
o
w
e
r

#
2

(
d
B
m
)
0 1 2 3 4 5 6 7 8 9 10
x 10
4
12
11.5
11
M
a
t
c
h

#
1

(
d
B
)
Time (ms)
Figure 2.9: Multipactor experiment with two carriers with E
2
/E
1
= 0.36
and
2
/
1
= 1.16. Due to a high maximum secondary yield,
multipactor suppression is not possible (the material used in the
experiment was plain aluminium, which can have a
se,max
3).
When the modulation signal is applied, the magnitude of the
multipactor noise increases signicantly.
1 1.5 2 2.5 3 3.5
1
0.5
0
0.5
1
Time [ s]
S
i
g
n
a
l

a
m
p
l
i
t
u
d
e

[

]
Square Root Raised Cosine filtered QPSK signal
Figure 2.10: Example of a QPSK modulated signal after bandpass ltering.
22
emission yield. So far, no practical coating with a
se,max
below unity
has been found. However, alodine is a commonly used surface coating for
space-bound microwave devices made of aluminium. It increases the rst
cross-over point to around 60 eV and reduces
se,max
to about 1.5, even
though the actual values vary much between samples. The concern with
a material with very good anti-multipactor properties is contamination.
A few ngerprints or a very small layer of dust can drastically alter the
properties of the material and make it prone to discharges.
By applying a DC electric or magnetic eld, the electron trajec-
tory can be disturbed and the important resonance condition can be
destroyed, thus making multipactor impossible. Simulations [45] have
shown that an external DC magnetic eld applied in the direction of
wave propagation in a rectangular waveguide can eciently suppress
multipactor. A drawback with the method is the extra components re-
quired to produce the magnetic or electric eld and thus the method
may not be feasible for e.g. space applications, where extra weight is
undesirable.
The most ecient way of avoiding multipactor is to make a design
where the mechanical dimensions are such that a power much higher
than the nominal power is required to start a discharge. However, that
may lead to large and heavy designs, which are to be avoided in space
systems, and thus one may have to resort to one or several of the above
mentioned methods of multipactor suppression.
2.1.5 Eect of random emission delays and initial velocity
spread
In the above analysis of multipacting electrons in a harmonic electric
eld, it was assumed that all secondary electrons were emitted with a
xed initial velocity, v
0
. However, as briey mentioned previously, the
electrons are actually emitted with a distribution of velocities and the
Maxwellian distribution is often used in simulations. Apart from the
spread in initial velocities, there is also a nite time between impact of
the primary electron and emission of the secondaries. Since this time in
most cases is very small compared to the RF period, it was neglected in
the previous analysis. Nevertheless, this time will cause a small phase
error and if the resonant phase is close to the phase stability limits as
given by Eqs. (2.18) - (2.21), the phase error may result in an increased
electron loss.
A detailed analysis of the eect of random secondary delay times and
23
random spread in emission velocities was done by Riyopoulos et al. [33].
They found that by including the eects of these random parameters,
the eective secondary electron yield,

se
, was reduced to a number in
the range
se
/2 <

se
<
se
. This means that the eective secondary
electron yield will be a function not only of the impact velocity, but also
of the resonant phase as well as the phase spread caused by the spread
in initial velocities and secondary delay times. Another study, which
supports this result, investigated the eect on the dierent resonance
zones for dierent values of the maximum SEY due to initial velocity
spread [46]. It was found that, except for the rst order mode, a real-
istic thermal spread of the initial electrons raised the multipactor SEY
requirement from unity to above unity. For the higher order modes a
SEY greater than approximately 1.5 was necessary to compensate for
the losses incurred. In addition, with increasing velocity spread, the
multipactor zones started to overlap. The increased SEY requirement
will result in an increased threshold for the higher order modes and can
explain the success with increasing the rst cross-over point in the Hatch
and Williams charts when tting experimental data (see Fig.2.5).
The importance of the spread in initial velocities can be seen when
constructing multipactor charts for a constant initial velocity without al-
lowing compensation for electron losses outside the phase stability range.
In Fig. 2.11 zones bounded by solid lines indicate the region where mul-
tipactor can take place under this assumption. The dashed lines make
up wider zones that encompass the other zones and are identical to the
zones shown in Fig. 2.3.
By including a higher secondary electron yield and a spread in ini-
tial velocities, the multipactor zones will become wider than the solid
line zones. A
se
greater than unity, which will be the case when the
impact velocity is greater than the rst cross-over point, will compen-
sate for some of the losses incurred due to phase instability. A spread
in initial velocities will widen the range of possible resonant phases (cf.
Eq. (2.9)) and the left and right limits will not be as sharp as indicated
by the solid line multipactor zones in Fig. 2.11. This widening of the
multipactor zones has been taken into account to a certain extent in the
traditional analytical approach, which is indicated by the wider dashed
line zones in Fig. 2.11. However, the widening should not only be to-
wards the left side, but also towards the right [39]. Furthermore, the
sharp lower left corner of each dashed line zone is misleading, as that
indicate a point where the secondary electron emission is unity and the
24
10
0
10
1
10
2
10
3
10
4
Frequency Gap product [GHzmm]
V
o
l
t
a
g
e

[
V
]
Figure 2.11: Multipactor charts based on the same parameters as in Fig. 2.3.
The solid line zones indicate the zones within which phase-
focusing is active. The dashed line zone is produced by including
also unstable phases until the non-returning electron limit.
phase is very unstable, thus making a discharge impossible. A more
correct boundary would be a rounded shape, which starts in the lower
left corner of the solid line zone and smoothly joins the dashed left hand
side [47]. This is conrmed by experiments [30, 48], cf. Fig. 2.12, which
shows measurement data from one of the early multipactor experiments
by Hatch and Williams [48]. A similar rounded shape can also be seen in
numerical simulations and examples of this is shown in chapter 5, which
includes PIC simulations of multipactor in a coaxial line.
2.2 Multicarrier
Modern satellites operate in multicarrier mode, i.e. several signals at dif-
ferent frequencies exist simultaneously in the microwave and electronic
systems. An example of such a system is Sirius 3, which is one of the
Nordic satellite [49]. It has 15 channels in the frequency range 11.7 -
12.5 GHz and each channel has a bandwidth of 33 MHz. Assume that
each channel has a power of 200 W. Then the maximum instantaneous
power of the system, the peak power, is equal to 45 kW. The peak power
increases with the square of the number of carriers. Such a high instan-
25
Figure 2.12: Multipactor experiment [48] showing the expected rounded o
lower left corner of the rst multipactor zone [47].
26
0 5 10 15 20
0
5
10
15
Time [ns]
A
m
p
l
i
t
u
d
e

[
V
]
Figure 2.13: In-phase multicarrier
signal. The signal os-
cillates rapidly, which
makes the signal enve-
lope appear clearly in
this time resolution.
0 5 10 15 20
0
5
10
15
Time [ns]
A
m
p
l
i
t
u
d
e

[
V
]
Figure 2.14: Random phase multi-
carrier signal.
taneous power is very unlikely in a real system since it will occur only
when all the signals are in phase as illustrated by Fig. 2.13. The most
likely scenario, if the carriers are not phase locked, is that the phase of
each carrier is a random number and will result in a signal with much
lower maximum instantaneous power as illustrated in Fig. 2.14.
The signals in Figs 2.13 and 2.14 are characterised by all carriers
having the same amplitude and a constant frequency. Consider a signal
with N carriers, each carrier having the same amplitude E
0
, but dierent
phases
n
and with a frequency spacing f. The period of the envelope
will then be T = 1/f and the envelope is given by
E
env
= E
0

N1

n=0
cos (n2ft +
n
)

2
+

N1

n=0
sin(n2ft +
n
)

2
(2.25)
A more realistic signal would have dierent amplitudes for each car-
rier and the frequency spacing would not be constant. The envelope of
such a signal can be found from
E
env
=

N1

n=0
E
n
cos (k
n

0
t +
n
)

2
+

N1

n=0
E
n
sin(k
n

0
t +
n
)

2
(2.26)
27
where k
n
is a factor determining the frequency spacing
k
n
= f
n
/f
0
1, n = 0, 1, ..., N 1 , (2.27)
f
0
is the lowest carrier frequency and
0
= 2f
0
. When assessing the
worst case scenario from the multipactor point of view, it is important
to study a whole envelope period. For arbitrarily spaced frequencies, the
envelope period, T, can be found by solving the following Diophantine
systems of equations:
T =
n
i
f
i
, n
i
^ i = 1, 2, ..., N 1 (2.28)
where N is the number of carriers, f
0
is the signal with the lowest fre-
quency and f
i
= f
i
f
0
. The envelope period will be the solution
with the smallest possible integers. For equally spaced carriers, the so-
lution becomes n
1
= 1, n
2
= 2,...,n
N1
= N 1, which is implies that
T = 1/f, like before.
When studying multicarrier multipactor it is common to make cer-
tain simplications that will allow using single carrier methodology to
asses also the multicarrier case, e.g. the mean frequency of all the car-
riers is used as the design frequency. Thus, most of what has been said
about single carrier multipactor will then be valid also for the multiple
signals case.
2.3 Design guidelines
From an industrial point of view it is important not only to understand
the physics of multipactor, but also how the theoretical and experimen-
tal results should be applied when making multipactor-free microwave
hardware designs. In Europe, most space hardware designers follow the
standard issued by ESA [50]. This standard includes both the single
and the multicarrier cases, but for the latter it is stated that the de-
sign guidelines are only recommendations. Most research support these
recommendations, but not enough tests have been performed to verify
the theoretical ndings. When using the standard it is important to be
aware of the fact that it is primarily based on the parallel-plate model
with a uniform electric eld. Design with respect to this approach for
other geometries is normally a conservative and safe way. However, in
many common microwave structures, the geometry is such that losses
of electrons is much higher than in the parallel-plate case. Thus the
28
multipactor threshold in geometries such as coaxial lines, waveguides
and irises, can be higher or much higher than that obtained using the
plane-parallel model.
2.3.1 Single carrier
In the ESA standard [50] components are divided into three categories
or types. Type 1 is a well vented component where all RF paths are
metallic and the secondary electron emission properties are well known.
This type of component has the lowest design margins with respect to
multipactor and, depending on the type of test, range from 3-8 dB.
The second type of component may contain dielectrics with established
multipactor properties and the component should be well vented. Also
depending on the type of test, the design margins range from 3-10 dB.
All other components are categorised as type 3 and the design margins
range from 4-12 dB.
When designing with respect to multipactor a complete electric eld
analysis is performed and regions with high voltages and critical gap
sizes are identied. Using the frequency-gap size product, the multi-
pactor threshold can be found in a susceptibility chart for the material
in question. A susceptibility chart in the ESA standard is basically an
envelope of the multipactor zones as shown in e.g. Fig. 2.5. If a mar-
gin larger than the largest design margin, 12 dB, is found, no testing is
required. However, in most cases, the component will have to be tested
and methods for detecting multipactor will be discussed in chapter 6.
2.3.2 Multicarrier
In the multicarrier case, only components of type 1 are covered by the
recommendations given in the ESA standard. Type 2 and type 3 com-
ponents will require further research before they can be included in the
standard. In the single carrier case, the level that is compared with the
multipactor threshold in the susceptibility chart is the amplitude of the
signal and no ambiguities exist. For multicarrier designs, the traditional
way of designing was to set the design margin with respect to the peak
power of in-phase carriers, shown in Fig. 2.13. This design method is
still allowed by the ESA standard and for type 1 components the design
margins range from 0-6 dB depending on the type of testing that will
be performed. However, as previously mentioned, in-phase carriers for
non-phase locked signals is extremely unlikely and thus the standard
29
allows for another design margin, which is set with respect to the so
called P
20
power level. The P
20
level corresponds to the peak power of
the multicarrier waveform whose width at the single carrier multipaction
threshold is equal to the time taken for the electrons to cross the mul-
tipacting region 20 times [50]. This level is illustrated in Fig. 2.15.
Figure 2.15: An example where the in-phase peak power is above the single
carrier threshold, while the P
20
level is more than 4 dB below the
same threshold. The peak voltage is 128.4 V, the single carrier
threshold is 91 V, and the P
20
voltage is 57 V. Signal data: 12
carriers, equally spaced, f
min
= 1.545 GHz, f = 24 MHz and
each carrier amplitude is 10.7 V. Material properties: W
1
=
23 eV and
se,max
= 3.
In the case when a design is made with respect to the P
20
level, the
design margins range from 4-6 dB depending on the type of testing. A
problem with the P
20
level is that it is not a trivial problem to nd the
peak power level for 20 electron gap crossings. This power level is usually
referred to as the worst case scenario, even though it may not always be
the worst case from a multipactor point of view. A number of dierent
30
ways of nding the worst case scenario have been proposed, e.g. using
parabolic or triangular phase distribution in the equally spaced carriers
case (cf. Ref. [51]). Some of the better methods for nding the worst
case scenario will be described in the following subsections after a brief
discussion about the 20 gap crossings rule (TGR).
Twenty gap crossings rule
The TGR was proposed in Ref. [14] in 1997 and in its original version
it reads:
As long as the duration of the multicarrier peak and the mode order
of the gap are such that no more than twenty gap-crossings can occur
during the multicarrier peak, then multipaction-generated noise should
remain well below thermal noise (in a 30 MHz band). [14]
The rule is a result of an analysis of simulated multi-carrier mul-
tipactor discharges. Comparison with experiments showed great devi-
ations, where the simulated noise could be as much as 75 dB greater
than the measured noise level. In the experiments, a minimum of 99
gap-crossings were required before the produced noise was detectable
above the noise oor of -70 dBm. Of course, there may be bit errors
even at lower noise levels, but as the number of electrons grows expo-
nentially with the number of gap crossings (see Eq. (2.22), there is a
huge dierence between 20 and 99 gap-crossings.
However, the TGR is certainly a good rst attempt to lower the
requirements for multi-carrier multipactor. It is a fairly conservative
method and thus the risk of applying it should be quite limited. How-
ever, more appropriate guidelines should be based on an unambiguous
theoretical concept, which can take the material properties into account.
Then, when performing simulations and experiments to verify the idea,
it is of paramount importance to make sure that the actual material
properties of the test samples are well known and that these proper-
ties are also being used in the simulations. Due to the large dierence
in secondary emission properties between dierent materials, it would
seem reasonable that for a material with a low
se,max
one would allow
more gap crossings than, in the opposite case, for a material with a high

se,max
.
31
Boundary function Method
One of the best engineering methods for nding the worst case scenario
for equally spaced carriers, the boundary function method, was origi-
nally designed by Wolk et al. [51]. Unfortunately the used function was
found empirically while studying the worst case scenario with an optimi-
sation tool, and is thus not physically founded. A consequence of this is
that under certain circumstances, the boundary function produces poor
results. It was also limited to work only for equally spaced carriers. How-
ever, as part of the present thesis work, this method has been further
developed, and it has been found that the original boundary function
approximately describes a function that tries to squeeze all the energy
of the multicarrier signal during one envelope period into a specied,
shorter, time period. This works just as well in both the equally spaced
and the non-equally spaced carrier cases and can be summarized by the
following formulas:

F
V
(T
X
) =

T
H
T
X
N

i=1
E
2
i
F
V,max
=
N

i=1
E
i
F
V,min
=

i=1
E
2
i
(2.29)
Here T
X
is the time period of interest, which is often set to T
20
, i.e.
the time it takes the electrons to traverse the gap 20 times. T
H
is the
period of the envelope and E
i
is the voltage amplitude of each carrier.
F
V
(T
X
) is the design voltage and is shown as two symmetric curved lines
in Fig. 2.15. The design voltage can never exceed the in-phase voltage,
given by F
V,max
, and if all power is distributed evenly over the entire
envelope period the voltage amplitude will be F
V,min
, which is indicated
by a dashed line in Fig. 2.15.
The main advantage with the boundary function method is its sim-
plicity. It is also very reliable, although a little conservative and this is
especially true for non-equally spaced carriers, where the P
20
level can
be much lower than F
V
. The method has been implemented as an auxil-
iary method in WCAT, which is a software tool originally developed by
the present author and Genrong Li as part of a Masters Thesis [52] at
32
Saab Ericsson Space. It has since been upgraded with additional func-
tionality by the present author as well as by Mariusz Merecki as part of
a Masters Thesis [40] at Centre National d

Etudes Spatiales, Toulouse,


France. Fig. 2.16 shows the graphical user interface of the present version
of WCAT and an example when the worst case of non-equally spaced
carriers have been assessed using the built in genetic algorithm.
0 5 10 15 20 25 30 35 40
0
50
100
150
200
Fv =129
Threshold =192
Env Threshold =139
Hyb. Threshold =158
Fv =129
Threshold =192
Env Threshold =139
Hyb. Threshold =158
0 5 10 15 20 25 30 35 40
10
0
10
1
10
2
10
3
0 5 10 15 20 25 30 35 40
10
0
10
1
10
2
10
3
Boundary Function
FVmin
InPhase Envelope
Envelope
Figure 2.16: Assessment of worst case scenario for multicarrier multipactor
using WCAT, Worst Case Assessment Tool. The example shows
a case with 10 non-equally spaced carriers with varying ampli-
tude.
Optimisation methods
A more direct method of nding the worst case scenario is to use some
kind of optimisation tool. In WCAT several dierent methods of nding
the worst case scenario are implemented. One uses the non-linear least
square (NLSQ) functionality of Matlab to nd the set of phases that will
t as much energy as possible inside a time period T
X
T
H
, where T
H
is the envelope period. Another method generates a variable number of
sets of random phases and the corresponding envelopes are compared to
nd the worst case. This method is better than the NLSQ optimisation
method when it comes to nding the worst case for non-equally spaced
33
carriers, because the NLSQ method requires a good seed phase in order
to nd a good local minimum. By combining these methods and using
the result of the best random phase approach as a seed phase for the
NLSQ optimisation, the best results are found. The problem with a
random phase approach is that for a large number of carriers, the number
of phase-sets needed to achieve a good result becomes discouragingly
high [52].
By using a genetic algorithm, a great improvement in nding the
worst case scenario in a short time has been achieved, especially for the
non-equally spaced carrier case. This type of optimisation scheme has
the advantage of being able to nd not only a good local minimum as
in the NLSQ case, but the actual global minimum can be found. The
genetic algorithm was implemented by Merecki [40] and in addition to
improving the optimisation part of the software, he also implemented
the threshold of the hybrid modes (see Fig. 2.16) as well as many other
useful functions.
The main problem with multipactor in microwave systems is the
electric noise that is generated, which degrades the signal to noise ratio.
Thus the worst case may not always be the maximum power within the
T
20
period. Depending on the material properties some other case may
produce a substantially larger amount of energetic electrons. Therefore,
WCAT also includes the possibility of nding the phase-set that will
produce the largest amount of electrons.
In addition to assessing the risk for a vacuum discharge, it can also
be of value to investigate if a multipactor free design may have a risk
of corona breakdown if the component is not thoroughly vented when
it is brought into operation. In WCAT this is analysed using the mean
carrier frequency and comparing the corona threshold with the in-phase
peak voltage. If the minimum of the Paschen curve is greater than this
voltage, then the corona margin is displayed in the output window of
WCAT. If the opposite is true, the pressure range within which there is
risk for corona discharge is presented (see Fig. 2.16).
34
Chapter 3
Multipactor in low pressure
gas
Microwave discharges can occur in both gas and vacuum. In vacuum, the
phenomenon is usually called multipaction or multipactor and the theory
for such vacuum microwave breakdown was presented in the previous
chapter. In a gas it is normally called corona discharge or gas breakdown
and it can occur when the electron mean free path between collisions
with molecules is smaller than the characteristic dimensions of the vessel.
An applied microwave electric eld can widen the velocity distribution
of the free electrons and thus make more electrons energetic enough to
ionise the gas. If the production of electrons exceeds the loss through
diusion, attachment, and recombination, the electron density will grow
exponentially and microwave gas breakdown will occur.
When the mean free path between collisions is of the same order as
the device dimensions, classical theory for microwave gas and vacuum
discharge can not be used. Diusion loss can no longer be assumed,
like in the gas breakdown case, since that requires a mean free path
several times shorter than the characteristic length of the component.
Nevertheless, the electrons will meet a resistance due to collisions with
the neutral gas molecules and thus pure vacuum can not be assumed
either.
Low pressure multipactor has received comparatively little attention.
However, a few studies, theoretical as well as experimental, have revealed
some parts of the complicated picture. Vender et al. [17] performed
PIC-simulations to study the electron density development and showed
that at suciently low pressures, the gas discharge is initiated by a
35
multipactor discharge. Using a Monte Carlo algorithm, Gilardini [53]
made quite a general study of the phenomenon and presented breakdown
voltages normalised to the rst cross-over point of the material for a
wide range of dimensionless variables. He also paid special attention
to a particular and realistic case, namely multipactor in low pressure
argon [54]. This was done partly in an eort to compare the simulations
with the experimental results of Hohn et al. [18].
In paper B of this thesis, low pressure multipactor was studied using
an analytical model that takes into account only the friction force due
to collisions between the electrons and the neutral gas particles. The
main theory and results from this study will be presented in the rst
section below. In addition to the friction force, the collisions will also
cause a random velocity spread of the electrons that results in a higher
average impact energy. Furthermore, due to the long distance between
molecules, the electrons are free to accelerate to very high velocities and
upon impact with a gas molecule or atom the energy is sucient to cause
ionisation. In paper C of this thesis a more detailed analysis has been
done, where all these eects have been considered and the used model
as well some highlights from the results are presented in the section
Advanced Model below.
3.1 Simple Model
In a rst attempt to understand the behaviour of multipactor in a low
pressure gas, a simple analytical model was used, which takes only the
friction force of the collisions with neutrals into account. By deriving ex-
plicit expressions for the multipactor threshold, qualitative comparison
with experimental results [18] as well as results from computer simula-
tions [53, 54] could be made.
3.1.1 Model
The dierential equation governing the behaviour of the electrons in
a low pressure gas is given by the equation of motion, Eq. (2.1), but
augmented to include also the eects of collisions:
m x = eE m
c
x (3.1)
where
c
=
c
n
0
v is the collision frequency between the free electrons
and the neutral particles.
c
is the collision cross-section, n
0
the neutral
36
gas density and v the electron velocity. The collision cross-section is
generally a function of the electron velocity, but in order to avoid a
non-linear dierential equation,
c
is assumed to be a constant.
As in the vacuum case, a spatially uniform harmonic eld E =
E
0
sin t is assumed. Solving Eq. (3.1) with the same initial conditions
as for the vacuum case, i.e. that an electron is emitted from x = 0 with
an initial velocity v
0
when t = /, the position and velocity of the
electron can be found:
x =
1

c
(1 e

c
(

t)
)(v
0
+ [ cos
c
sin ])
+

[(sin sin(t)) +
c
(cos cos(t))] (3.2)
x = v
0
e

c
(

t)
+ [e

c
(

t)
( cos
c
sin )
cos(t) +
c
sin(t)] (3.3)
where
=
eE
0
m(
2
+
2
c
)
(3.4)
The resonance condition requires that an electron emitted when t =
/ should reach the other electrode, at x = d, when t = N + ,
where N is an odd positive integer as in the vacuum case. Applying
this condition to Eqs. (3.2) and (3.3) yields expressions for the required
electric eld and the impact velocity:
E
0
=
m
e
(
2
+
2
c
)(d +
v
0

c
(e

N
c

1))
(1 +e

N
c

) sin + ((1 e

N
c

c
+
2
c

) cos
(3.5)
v
impact
= v
0
e

N
c

+ (1 +e

N
c

)( cos
c
sin ) (3.6)
In order to draw the multipactor boundaries, an expression for the
non-returning electron limit is needed. Like in the vacuum case no ex-
plicit analytical expression for this can be found and thus the limit will
be obtained numerically instead. However, as a rough approximation,
the limit given by Eq. (2.11) can used.
3.1.2 Multipactor boundaries
When constructing only the lower multipactor threshold in vacuum,
which depends on the rst cross-over energy, the threshold value un-
der the assumption of a constant initial velocity is the same as with
37
the assumption of a constant ratio k = v
impact
/v
0
between impact and
initial velocities. This is true also in the presence of collisions, when the
above simple model is used and thus the constant k approach will be
used in the following expressions. Combining Eqs. (3.5) and 3.6 under
this assumption yields an expression for the resonant phase, viz.
tan =

2
[ +] + 2
2
c
v
impact
( k)
( + 1)
(3.7)
where = exp(
c
N/), = kd
c
+ (k + 1)v
impact
, = kd
c

v
impact
(k + 1), and = [kd
c
+ (k 1)v
impact
]
c
.
Equation (3.7) can be used together with Eq. (3.5) to plot the multi-
pactor threshold in a low pressure gas as a function of gap size, pressure,
or frequency. However, by multiplying the expression for the amplitude
of the electric eld with the gap size, d, an expression for the voltage as a
function of the frequency-gap size and the pressure-gap size products can
be obtained. This approach will be used in a subsequent section, where
a more advanced model is used to analyse the phenomenon. Figure 3.1
shows the lower multipactor threshold in low pressure air for dierent
pressures. The graphs are based on Eqs. (3.5) and (3.7) only and do not
consider the non-returning electron limit nor the phase stability limits.
From Fig. 3.1 it is clear that the multipactor threshold increases
with increasing pressure. By sweeping the pressure instead of the fre-
quency and comparing the changing threshold with the corona break-
down threshold, an understanding can be obtained of how the transition
between these two types of discharges can occur. Fig. 3.2 shows that
the multipactor threshold rst increases until a certain point, where it
intersects the curve corresponding to the corona threshold. It will then
follow this curve towards the minimum of the Paschen curve. This is
just a qualitative picture and the sharp intersection would in reality be
a smooth transition.
Figure 3.1 does not consider the non-returning electron limit, nor
the phase stability limits. As explained in the previous chapter, phase
focusing is needed to maintain the generated electrons in a close bunch,
since an electron with a too large phase error will be lost. By introducing
a small phase error in Eq. (3.2) and keeping the amplitude and phase
constant while setting x = d, the error after the passage can be found.
The ratio between the nal and initial error is the stability factor, G,
and when the absolute value of this factor is less than one, the phase
focusing eect is active. With the present model, the expression for the
38
10
3
10
2
10
1
10
2
10
3
Multipaction threshold curves in low pressure air (d=0.1 m)
V
o
l
t
a
g
e

(
V
)
Frequency (GHz)
p=0.1 Pa
p=1 Pa
p=10 Pa
Vacuum
Figure 3.1: Multipactor chart showing the lower multipactor threshold at
dierent pressures. i.e. each curve is based on an impact ve-
locity corresponding to W
1
= 23 eV. Phase stability and the
non-returning electron limit are not considered, only resonance.
Parameters used are:
c
= 6.9 10
20
m
2
, k = 2.5, N = 1 and
d = 0.1 m.
39
10
2
10
1
10
0
10
1
10
2
10
1
10
2
10
3
Threshold curves in low pressure air (fd=1GHz mm)
V
o
l
t
a
g
e

(
V
)
Pressure (Pa)
Multipactor in air
Multipactor in vacuum
Corona in air
Figure 3.2: Thresholds for multipactor and corona discharges in air as func-
tions of pressure together with the multipactor vacuum threshold
as a reference level. Parameters used are:
c
= 6.9 10
20
m
2
,
W
1
= 23 eV, k = 2.5, N = 1, f d =1 GHzmm and d = 0.1 m.
40
stability factor becomes:
G =
(

(C1) +C

c
( 1)) tan + 1 C

(1 C) tan (C + 1)
(3.8)
where C = (k +1)/(k ). In Fig. 3.3 the phase limits, where [G[ = 1,
have been plotted together with the non-returning electron limit. Both
the positive and negative phase error limits tend to decrease with increas-
ing pressure. However, the limit for non-returning electrons increases,
which is a more important limit when the electron impact energy exceeds
the rst cross-over energy, thus reducing the width of the multipactor
zone.
10
3
10
2
10
1
10
0
10
1
100
80
60
40
20
0
20
40
Phase Limits
P
h
a
s
e

l
i
m
i
t

(
d
e
g
r
e
e
s
)
Pressure (Pa)
Postive phase error
Negative phase error
Nonret. el. in vacuum (Semenov et. al.)
Numerical nonreturning electron limit
Figure 3.3: Phase limits in a low pressure gas based on the simple analytical
model. The dashed line and the solid line (dashed at the end)
show the upper and lower phase limits beyond which a phase
error will start growing. The dash-dot line is the phase below
which emitted electrons will not be able to escape from the wall
of emission. The dotted line is the phase limit obtained using
Eq. (2.15) and is an approximation of the dash-dot line in the
vacuum case. Parameters used are:
c
= 6.9 10
19
m
2
, N = 1,
k = 7.6, d = 0.1 m, and W
1
= 23 eV.
41
3.1.3 Main results
The main result found in paper B is that a higher microwave power is
required to initiate breakdown in a low pressure gas, since the collisions
tend to slow down the electrons. By combining the low pressure mul-
tipactor graph with the corona threshold curve, it was concluded that
with increasing pressure, the required threshold will rst increase and,
after reaching a plateau, it will make a smooth transition to the low
pressure branch of the Paschen curve. This behaviour is conrmed by
the investigations made by Gilardini [53] for materials with a low rst
cross-over point, close to the ionisation energy of the gas, and for N = 1,
i.e. for the rst order of multipactor.
For materials with a higher rst cross-over energy and for higher
order multipactor, Gilardini found no initial increase in the multipactor
threshold, instead a monotonically decreasing breakdown voltage was
seen. A possible explanation for the dierences between the result ob-
tained by the simple analytical model and the result of Gilardini is also
presented in paper B and it is suggested that the reason is that for mate-
rials with a higher W
1
and for N > 1 the contribution of electrons from
impact ionisation decreased the required W
1
(see Fig. 3.4). However,
as will be seen in the next section, electron contribution from impact
ionisation is not the only reason for this behaviour. The collisions will
also cause an electron velocity spread, which will result in a larger to-
tal impact velocity and thus a lower voltage is needed to achieve the
necessary rst cross-over energy.
A comparison was made with experiments by Hohn et al. [18] in low
pressure argon as well as with PIC-simulations by Gilardini [54] in the
same gas (see Fig. 7 in paper B). However, the fd-product chosen by
these authors was located in the middle of the right boundary of the
rst multipactor zone, an area dominated by the hybrid modes [39].
The simple model used in the presented analytical approach is not ap-
plicable to these modes and consequently the behaviour found in the
experiments and simulations could not be conrmed. Furthermore, the
impact energy of the electrons at this fd-product is several times higher
than the ionisation energy of argon and thus a signicant contribution
of electrons from collisional ionisation would be expected. In addition,
the required W
1
would be reduced due to the electron velocity spread
and thus a behaviour similar to curves (b) or (c) in Fig. 3.4 should be
expected and it is also what is found.
To further analyse multipactor in a low pressure gas, a better ana-
42
Threshold curves in a low pressure gas
V
o
l
t
a
g
e

(
V
)
Pressure (Pa)
a
c
b
Multipactor in a low pressure gas without ionisation
Multipactor in vacuum
Corona
Figure 3.4: Qualitative form of the dependence of breakdown threshold with
pressure in the region between which multipactor and corona,
respectively, dominate the breakdown process: (a) the friction
force due to collisions with neutrals dominates, (b) electron ve-
locity spread reduces the required W
1
and collisional ionisation
contributes signicantly to the total number of electrons, (c) in-
termediate situation.
43
lytical model is obviously needed. The model must, in addition to the
friction force, be able to take collisional ionisation into account as well
as collision induced velocity spread of the electrons. In the next section
a more advanced model that includes all these eects will be presented.
3.2 Advanced Model
The simple model used in the previous section provided important qual-
itative understanding of the multipactor threshold behaviour in a low
pressure gas. However, due to the inherent limitations of the model,
some of the results found by other researchers could not be conrmed.
This section will present an improved model for multipactor in a low
pressure gas and it is based on paper C of this thesis. As a representa-
tive gas, the noble gas argon will be used in the included examples.
3.2.1 Model
Just as in the simple model, the basic geometric conguration is elec-
tron motion between two parallel plates perpendicular to the x-direction.
During the passage, no electron loss, only generation through collisional
ionisation, will occur. Using the dierential equations for the total elec-
tron momentum and for the change in the number of electrons, one can
derive the following equation for the electron drift acceleration:
du
dt
=
eE
m
u(
c
+
iz
). (3.9)
where u is the drift velocity and
iz
the ionisation frequency. In general,
the collision and ionisation frequencies are functions of the electron ve-
locity. However, by assuming that
c
and
iz
are constants, Eq. (3.9)
becomes a rst order linear dierential equation. Multipactor requires
an alternating driving electric eld and as in the previous model a har-
monic eld E = xE
0
sin t is used, where x is the unit vector, the
angular frequency, and t the time. Assuming the electric eld to be
homogeneous, the drift velocity will be parallel to the eld, u = xu,
and the vector notation for E and u can be dropped in the following
analysis. By setting =
c
+
iz
, u = x, and du/dt = x, Eq. (3.9) can
be written,
x =
eE
m
x. (3.10)
44
Since the equation has the same form as Eq. (3.1), it will also have the
same solutions and as the initial conditions are identical, the formulas
for the resonant eld amplitude and the impact velocity will be identical.
However, it should be noted that instead of
c
one will have and v
0
should be replaced by u
0
. An important dierence is that the velocity
in the previous model was only directed in the x-direction, but now
there is also a thermal velocity component, v
t
, i.e. the total velocity
is v = u + v
t
. With these new designations, the expressions for the
resonant eld amplitude and the impact velocity become:
E
0
=
m
e
(
2
+
2
)(d +
u
0

( 1))
(1 + ) sin + ((1 )

+
2

) cos
(3.11)
u
impact
= u
0
+ (1 + )( cos sin) (3.12)
where = exp (N/) has been introduced for simplicity. is given
by Eq. (3.4) as before, but
c
should be replaced by .
In order to construct the multipactor boundaries, the same approach
as in the simple model case is taken and an expression for the resonant
phase is obtained by combining Eqs. (3.11) and (3.12), which yields,
tan =

2
[ +] + 2
2
(u
0
u
impact
)
(d +u
impact
u
0
)(1 + )
(3.13)
where = d + u
impact
+ u
0
and = d u
impact
u
0
have been
used for convenience. The reason why the expression looks somewhat
dierent from Eq. (3.7) is that the constant initial velocity approach has
been used instead of the assumption of a constant ratio between impact
and initial velocities. This will also aect the expression for the phase
stability factor, which in this case becomes
G =
( 1)(
2
+
2
) sin u
0
((1 + )( sin cos ) u
0
)
(3.14)
So far, the dierences between the simple and the more advanced
model are fairly trivial. However, the parameters used (
c
and
iz
)
are not constants, they depend to a great extent on the total electron
velocity, which is the vector sum of the drift and thermal velocities. The
thermal velocity will have a random direction and therefore the average
total velocity will be equal to the drift velocity. However, the total
(average) energy, , will still depend on both velocities and it becomes,
=
mv
2
2
=
m
2
(u
2
+'v
t
2
`) (3.15)
45
where 'v
t
2
` represents the average of the square of the magnitude of
the thermal velocity. In paper C, a dierential equation for the thermal
velocity is derived, viz.
d'v
t
2
`
dt
+ (
c
+
iz
)'v
t
2
` = u
2
(
c
(2 ) +
iz
) (3.16)
where is the energy loss coecient. By assuming that
c
,
iz
and are
constants, like before, Eq. (3.16) can be solved explicitly and with the
initial condition 'v
t
2
(t = /)` = 0, the thermal impact velocity, when
t = N + , can be found and thus the total impact velocity can be
determined. However, the expression is very complicated and will not
be reproduced here.
The total impact velocity will determine the secondary electron emis-
sion yield. For vacuum multipactor as well as in the previous simple
model for low pressure multipactor, the impact velocity was perpendic-
ular to the electrodes. In such a case, the secondary yield depends only
on the impact velocity. However, for angular incidence, which will be
the case now with the random three dimensional thermal velocity com-
ponent, the yield will be a function not only of the impact energy but
also of the angle of incidence. To account for the angular incidence the
expressions given in Ref. [22] have been used and for ease of reference
they are reproduced here,

max
() =
max
(0)(1 +
2
/) (3.17)

se,max
() =
se,max
(0)(1 +
2
/2) (3.18)
=

impact

max
()
0
(3.19)

se
=
se,max
()( exp 1 )
k
(3.20)
where is the impact angle with respect to the surface normal.
max
is
the impact energy when the secondary emission reaches its maximum,

se,max
.
impact
is the total impact energy and
0
is the energy limit
for non-zero
se
. The formulas are valid for a typical dull surface,
according to Ref. [22]. The coecient k is given by k = 0.62 for < 1
and k = 0.25 for > 1.
In vacuum multipactor, the only source of new electrons is secondary
yield from each impact. When the phenomenon takes place in a gas,
another potential source of new electrons is impact ionisation of the
gas molecules. The ionisation threshold of most gases of interest is
46
in the range 10-20 eV. This is well below the rst cross-over point of
most materials and thus when the electron energy is sucient to initiate
multipactor, it is also enough to ionise the gas molecules. In this model,
the contribution from impact ionisation is included by modifying the
breakdown condition from
se
= 1 to

se
+
'
iz
`N

= 1 (3.21)
where '
iz
` is the average ionisation frequency and is an ionisation
factor, which ranges from 0 1 and indicates the fraction of the elec-
trons from ionisation that is able to become a part of the multipacting
bunch. Determination of the correct value of is not a trivial problem
and for simplicity a constant = 0.75 is used. This is quite a rough
approximation, but for materials with a low rst cross-over point, it will
be shown that the ionisation contribution is fairly small and the exact
value for is not so important. On the other hand, for materials with
a high rst cross-over energy, the importance of can not be neglected
and thus a detailed investigation of should be performed, but due to
the complexity, it will be left as future work.
Apart from , there are other parameters, which need to be deter-
mined with good accuracy in order to obtain useful quantitative results.
The energy loss coecient, , which is used in Eq. (3.16), is a small
quantity and for pure elastic collisions, the value is equal to 2m/M,
where m is the electron mass and M is the mass of the argon atom [55].
It is also a function of the electron energy and for inelastic collisions,
which will occur when the electron energy is greater than a few eV, the
value is about 10 to 100 times larger than the elastic value [56]. How-
ever, the value is still quite small and will not have major eect on the
low pressure multipactor threshold and for simplicity a value 10
3
will
be used, which is about 37 times greater than the elastic value. The
remaining parameters,
c
and
iz
, are very important for the thresh-
old and a detailed description of these values will be given in the next
section.
3.2.2 Analytical formulas for argon cross-sections
For most materials of interest, the rst cross-over energy is in the range
20-70 eV and it is this value that determines the lower multipactor
threshold. Thus it would be of value to have expressions for the col-
lision and ionisation cross-sections that give an accurate description of
47
these quantities in the range from 0 eV to about 100 eV. For the electron-
argon collision cross section, the data given in Ref. [57] is used and it
covers the range up to 20 eV. An analytical formula has been devised,
which approximates the given data quite well in the measurement re-
gion (cf. Fig. 3.5). Outside the measurement points, the cross-section
for very low energy electrons has been set to converge towards the geo-
metrical cross-section of the argon atom. For high energy electrons, the
cross-section is set to fall o with the same rate as for the last few eVs.
The analytical formula is given by the expression,

c
= (
1.68
1 + (8)
3
+

1 + (0.07)
2
)
1.5
1.15 10
20
[m
2
] (3.22)
where is the total electron energy, given by Eq. (3.15).
10
3
10
2
10
1
10
0
10
1
10
2
10
21
10
20
10
19
10
18
Electron energy [eV]
T
o
t
a
l

c
o
l
l
i
s
i
o
n

c
r
o
s
s

s
e
c
t
i
o
n

[
m
2
]
Buckman and Lohman
Analytical approximation
Figure 3.5: Absolute total collision cross-section for electrons scattered from
argon. The stars indicate measurement data by Buckman and
Lohmann [57] and the solid line is the analytical approximation
given by Eq. (3.22).
The ionisation cross-section increases rapidly for electron energies
slightly above the ionisation threshold and thus it is of importance to
have an accurate description in this range, especially since this is close
to the rst cross-over energy of many materials. A simple function that
48
accurately describes the cross-section for the entire measurement range
can be given by (cf. Fig. 3.6)

iz
= q
1
ln /
i

/
i
+ 0.1(/
i
)
2
[m
2
] (3.23)
where
i
is the ionisation threshold of argon and q
1
= 4.8 10
20
m
2
.
10
2
10
3
10
22
10
21
10
20
10
19
Electron energy [eV]
I
o
n
i
s
a
t
i
o
n

C
r
o
s
s

S
e
c
t
i
o
n

[
m
2
]
Ionisation cross section for Argon
S.C.Brown
H.C.Straub
Analytical approximation
Figure 3.6: Ionisation cross-section for electron-argon collisions. The circles
and stars indicate measurement data by S. C. Brown [58] and
Straub et al. [59] respectively and the solid line is an analytical
approximation given by Eq. (3.23).
3.2.3 Multipactor boundaries
In the following section the above model will be used to determine the
multipactor boundaries. The best accuracy is attained when solving
the basic dierential equations numerically while using good approxi-
mate formulas for the dierent parameters. However, such computation
takes very long time, since both the initial and the nal multipactor
conditions have to be fullled. Faster computation can be achieved
by using dierent approximations, e.g. constant parameters as shown
in Eqs. (3.10)-(3.12), Eq. (3.13), and Eq. (3.14). Two dierent imple-
mentations are used in paper C, one purely numerical and one semi-
49
analytical. Attempts were made to nd a purely analytical implementa-
tion as well, but due to the strong non-linearities in the functions for the
cross-sections, no accurate such implementation could be found. Details
concerning the two implementations are presented in paper C and will
not be reproduced here.
As mentioned in chapter 2, when constructing the complete multi-
pactor zones, the multipactor thresholds corresponding to impact ve-
locities between the rst and second cross-over points are determined
for a specic order of resonance within the phase range from the non-
returning electron limit to the upper phase stability limit. The zone
for that order of resonance is then the envelope of all these curves (cf.
Fig. 2.3). However, to explore the basic eects on the multipactor phe-
nomenon, it is sucient to study the threshold corresponding to unity
SEY. Thus, in most of the following charts, only the lower multipactor
threshold will be considered. However, in keeping with the multipactor
tradition, the complete zones will be presented as well.
One concern that appears when making low pressure multipactor
charts is the parameters which should be used on the chart axes. Clas-
sical vacuum multipactor charts use engineering units with voltage as a
function of the frequency-gap size product, like in Fig. 2.3. By multi-
plying Eq. (3.11) by the gap size, d, to get the voltage and rearranging
Eqs. (3.12), (3.13), and (3.14), these expressions can all be written as
functions of two natural parameters, viz. fd and pd, i.e. the frequency-
gap size and the pressure-gap size products. Thus, for a given pd the
multipactor zones can be constructed in the classical engineering units
as shown in Fig. 3.7. Note that three dierent pd-values are used, one
for each zone. The chosen values are close to the limit of stability of the
numerical implementation for each zone.
In Fig. 3.7 both the analytical (semi-analytical) and the numerical
implementations are used to plot the thresholds. Very good agreement
between the two implementations is found and therefore the faster an-
alytical version is used to produce all other gures. The most striking
rst impression of the graphs in Fig. 3.7 is the dierence in behaviour
between the rst and the higher order modes. The rst order mode
shows an increased threshold, which is a consequence of the friction
force experienced by the electrons due to collisions with neutrals. This
is in agreement with the model presented in paper B, which only con-
sidered the friction force. However, for higher order modes, the result
is the opposite. Instead of an increased threshold as in the friction
50
10
0
10
2
Frequency Gap product [GHzmm]
V
o
l
t
a
g
e

[
V
]
pd=15 Pa mm
pd=7 Pa mm
pd=5 Pa mm
Numeric stable phase
Numeric unstable phase
Analytic stable phase
Analytic unstable phase
Figure 3.7: Lower multipactor thresholds in low pressure argon for three dif-
ferent xed pd-values, one for each zone. The dotted lines repre-
sent the multipactor zones for vacuum multipactor. Parameters
used are: W
1
= 23 eV , W
0
= 3.68 eV,
se,max
(0) = 3,
0
= 0.
51
only model, the inclusion of ionisation and thermal spread leads to a
decreased threshold with increasing pressure.
Thresholds as in Fig. 3.7 can be found for all impact velocities be-
tween W
1
and W
2
and by constructing the envelope of all these curves
for each order of resonance, the complete, classical multipactor zones can
be found for a given pd-product for each zone. This is done in Fig. 3.8,
which shows the complete zones for three dierent pd-values. The draw-
back with this chart is that the model does not account for the hybrid
zones and thus the right boundary of each zone will not accurately re-
ect the true multipactor threshold for those fd-values. The model can,
however, be extended to include also the hybrid modes, but due to the
increased complexity, this is left as future work
10
0
10
1
10
2
10
3
FrequencyGap product [GHzmm]
V
o
l
t
a
g
e

[
V
]
pd=10 Pa mm
pd=4 Pa mm
pd=2 Pa mm
Figure 3.8: Multipactor susceptibility zones in low pressure argon (solid lines)
together with vacuum zones (dotted lines) for comparison. Pa-
rameters used are: W
1
= 23 eV, W
2
= 1000 eV, W
0
= 3.68 eV,

se,max
(0) = 3 and
0
= 0.
3.2.4 Key ndings
Among the main results is that the friction force dominates the low
pressure multipactor threshold for materials with a low rst cross-over
52
energy for the rst order of resonance, as indicated by Fig. 3.7. However,
this gure only shows the behaviour for a given pd-product and in order
to see what happens when the gas density increases, the threshold can be
plotted as a function of the pd-product. This has been done in Fig. 3.9
for a material with a low rst cross-over energy and, as expected, the
threshold increases with increasing pressure for the lowest order mode,
N = 1, and after reaching a maximum, the threshold starts to decrease
again. For higher order modes, the threshold decreases monotonically
as the gas becomes dense enough to aect the multipacting electrons.
This behaviour is identical to that found by Gilardini [53] in his Monte
Carlo simulation of low pressure multipactor. He also observed that for
materials with a higher rst cross-over point, the threshold does not
increase with increasing pd, instead it falls o monotonically, which is
the behaviour shown in Fig. 3.10. The main reason for this dierence
in behaviour is the contribution of electrons from collisional ionisation,
which increases drastically when the electron energy is well above the
ionisation threshold.
10
2
10
1
10
0
10
1
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
N
o
r
m
a
l
i
s
e
d

V
o
l
t
a
g
e
Pressure gapsize [Pamm]
N=1
N=3
N=5
Figure 3.9: Normalised multipactor thresholds for varying pd. The thresholds
are normalised with respect to the vacuum threshold. Curves for
the three rst orders of resonance are shown. Parameters used
are: W
1
= 23 eV, W
0
= 3.68 eV,
se,max
(0) = 3,
0
= 0, fd
N=1
=
0.6 GHzmm, fd
N=3
= 2.4 GHzmm, and fd
N=5
= 4.2 GHzmm.
For higher order of resonance, N > 1, Gilardini found no dierence
in the basic behaviour regardless of material. The threshold falls o
53
10
2
10
1
10
0
10
1
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
1.1
N
o
r
m
a
l
i
s
e
d

V
o
l
t
a
g
e
Pressure gapsize [Pamm]
N=1
N=3
N=5
Figure 3.10: Normalised multipactor thresholds for varying pd. The thresh-
olds are normalised with respect to the vacuum threshold.
Curves for the three rst orders of resonance are shown for
a material with a rst cross-over point more than 7 times
greater than in Fig. 3.9. Parameters used are: W
1
= 170 eV,
W
0
= 4 eV,
se,max
(0) = 1.3,
0
= 0, fd
N=1
= 1 GHzmm,
fd
N=3
= 3.2 GHzmm, and fd
N=5
= 5 GHzmm.
54
10
0
10
2
Frequency Gap product [GHzmm]
V
o
l
t
a
g
e

[
V
]
=0
=0.75
=0
=0.75
Vacuum multipactor
Figure 3.11: Multipactor thresholds in low pressure argon for the two lowest
order modes (N = 1 and N = 3) with = 0.75 and = 0
respectively. Parameters used are: W
1
= 23 eV, W
0
= 3.68 eV,

se,max
(0) = 3,
0
= 0, pd = 15 Pamm for N = 1, and pd =
7 Pamm for N = 3.
directly from the vacuum threshold without showing any maximum and
the same behaviour is seen in Figs. 3.9 and 3.10. Even with a low W
1
,
where the contribution of electrons from collisional ionisation is low, a
monotonically decreasing threshold is obtained. The cause of this dis-
tinct lowered threshold is the partial thermalisation of the electrons due
to the collisions. The velocity spread results in a total impact veloc-
ity, which is greater than the drift velocity alone and thus for the same
secondary electron emission, a lowered impact drift velocity is possible.
Even though the friction force requires a higher voltage to achieve the
same impact drift velocity, the thermalisation eect dominates, with a
lowered threshold as a result. This becomes clear in Fig. 3.11, where it is
apparent that it is not the electrons from ionisation that constitute the
main reason for a decreased threshold, rather it is a consequence of the
partial thermalisation. In the case with a high W
1
, the thermalisation
eect is also important, but without the contribution from collisional
ionisation, the behaviour would not be the same, which can be seen in
Fig. 3.12.
To summarise the key ndings from the more advanced model, it
55
10
0
10
3
Frequency Gap product [GHzmm]
V
o
l
t
a
g
e

[
V
]
Vacuum multipactor
=0
=0.75
=0.75
=0
Figure 3.12: Multipactor thresholds in low pressure argon for the two lowest
order modes (N = 1 and N = 3) with = 0.75 and = 0
respectively. In this case, the second cross-over energy is about
7 times greater than in Fig. 3.11. Parameters used are: W
1
=
170 eV , W
0
= 4 eV,
se,max
(0) = 1.3,
0
= 0, pd = 25 Pamm
for N = 1, and pd = 15 Pamm for N = 3.
56
can be said that there are three main eects that aect the low pressure
multipactor threshold. The friction force tends to increase the threshold
as a higher electric eld is needed to reach the necessary impact velocity.
The thermalisation, on the other hand, increases the total impact energy
and thus a lower electric eld is needed to achieve the required impact
velocity. For materials with a low rst cross-over point, the rst eect
dominates for the rst order of resonance, while for higher order modes,
the latter plays the main role. In addition to these two eects, the model
also includes contribution from impact ionisation to the total number
of electrons. This addition also tend to lower the multipactor threshold
and the eect becomes very prominent for materials with a high rst
cross-over energy as a consequence of the concomitant high ionisation
cross-section.
57
58
Chapter 4
Multipactor in irises
A common microwave component is the waveguide iris, which is often
used as a shunt susceptance for the purpose of matching a load to the
waveguide. There are many dierent types of irises, but a typical cong-
uration consists of a step-like, short length, reduction of the waveguide
height. Similar structures also appear in other congurations, e.g. as
apertures in array antennas, as coupling slots in directional couplers,
and as irises in waveguide lters. As the eld strength in the iris can
be very high and the gap height is small, there is a pronounced risk of
having a multipactor discharge.
So far in this thesis, all the models considered have been based on
the plane-parallel model with a spatially uniform harmonic electric eld.
In general, most theoretical studies of the multipactor phenomenon have
been limited to this or similar approximations. However, many RF de-
vices involve more complicated electric eld structures where predictions
based on the parallel-plate model are not applicable. This is true for
e.g. the waveguide iris, where the electric eld will be a combination
of several dierent electromagnetic modes, most of which typically are
evanescent. However, due to the short length of the iris, these modes
will be of importance. Nevertheless, in this analysis, which is described
in more detail in paper D, it is only the importance of the random drift
due to initial velocity spread of the secondary electrons that has been
considered. Thus, as far as the eld is concerned, a spatially uniform
harmonic eld based on the parallel-plate model is used.
Experiments [60, 61] as well as numerical studies [61, 62] of multi-
pactor in an iris have shown that the discharge threshold increases with
decreasing length of the iris. It has been suggested that the reason for
59
the increased threshold is losses of electrons out of the iris region [60]. In
this analysis, we show that one of the contributing factors to this electron
loss is a random drift due to the axial component of the initial velocity
of the secondary emitted electrons. Other loss mechanisms, which are
due to the inhomogeneity of the eld, tend to further enhance the losses
and these eects will be more pronounced for small gap lengths. This
means that by taking only losses due to the random drift into account,
a conservative increase of the breakdown threshold should be obtained.
4.1 Model and approximations
The geometry used in the model is the 2-dimensional structure shown
in Fig. 4.1. The iris has a gap height h in the y-direction, a length l
in the z-direction and is assumed to be tted into a waveguide with a
height that is much greater than h. The harmonic electric eld E is
assumed uniform in the gap, as a simple approximation of the actual
eld. There are two main reasons for choosing a uniform eld. Firstly,
the deterministic model developed for the parallel plate case, which is
described in chapter 2, can be used to describe the basic behaviour
of an electron trajectory inside the gap. Secondly, the eect of the
initial velocity spread of the secondary electrons along the z-axis on the
multipactor threshold can be analysed separately from the drift force due
to inhomogeneities in the electric eld. In addition, it gives a convenient
base for comparing the results with those of the parallel plate model.
By assuming a uniform E-eld in the y-direction, the electron motion
along the z-direction is not aected by the eld. The motion in this
direction, the drift motion, will occur with a xed velocity v
z
between
the impacts. Lets assume that a seed electron is emitted inside the gap at
the coordinate z
0
, l/2 < z
0
< l/2, at one of the walls. As the electron
traverses the gap and hits the opposite side of the iris, it has become
displaced a distance z in the z-direction. This drift is determined by
the velocity in the z-direction, v
z
, together with the transit time, t
g
,
and is given by z = v
z
t
g
. For a xed mode order, N, and frequency,
f, of the eld, each transit time is the same and is given by,
t
g
=
N

, (4.1)
where = 2f.
The electron trajectory in the z-direction will perform a random
walk with a change of velocity, v
z
, after each impact. When the impact
60
z
y
h
l
E
Figure 4.1: The geometry used in the considered model.
coordinate is outside the iris area [z[ > l/2, i.e. one of the gap edges
has been passed, the electron trajectory is lost. The probability of sur-
vival, p(k) (see Fig. 4.2), for the electron trajectory decreases with the
number of transits, k, and for a general one-dimensional random walk
problem, with the jump size governed by a continuous distribution func-
tion,
k
(z), an explicit solution for this, the rst passage time problem,
is not always possible. However, in paper D it is explained that the
asymptotic behaviour of p(k) is determined by the largest eigenvalue
0
of the expansion of
k
(z), i.e.
p(k)
k
0
. (4.2)
A detailed description of how to determine p(k) is given in paper D
together with approximate solutions for
0
when the normalised iris
length, = l/(v
T
t
g
), is either very small or very large. This summary,
however, will focus on the eect the random electron drift has on the
multipactor susceptibility zones.
Each seed electron inside the iris gap will start to multiply with the
successive wall collisions. Due to the stochastic losses, the number of
electrons will sometimes become large and sometimes small. However, if
on average the generation of electrons due to wall collisions is larger than
the loss over the gap edges, there is a nite probability that a suciently
61
50 100 150 200 250 300 350 400 450 500
0
0.2
0.4
0.6
0.8
1
Number of collisions
P
r
o
b
a
b
i
l
i
t
y

o
f

s
u
r
v
i
v
a
l
l=2 mm
l=8 mm
l=16 mm
Figure 4.2: The probability of survival, p(k), for an electron emitted in the
center of the iris gap, z = 0, for three dierent iris lengths. Pa-
rameters used: f = 1 GHz, N = 1, and W
T
= 2 eV (corre-
sponding to v
T
, the rms-velocity of the Maxwellian distribution
of initial velocity in the z-direction).
strong discharge will appear. The generated number of electrons over
the initial number of electrons after k collisions is given by,
N
e
N
0
g(k) = p(k)
k
se
. (4.3)
Depending on the start position of the seed electrons, the initial be-
haviour of N
e
can vary. If the start position is close to the iris edge,
the average electron number will rst decrease and then if
se
is large
enough, it will start increasing again. But if the start position is in
the center, it may rst start to increase, but after a number of transits,
it will start decreasing (cf. Fig. 4.3). Eventually, it is the asymptotic
behaviour of p(k) that will determine whether or not there will be a
discharge. Thus from Eq. (4.2) and Eq. (4.3) one can conclude that the
asymptotic change in the electron number is given by,
g(k) (
se

0
)
k
. (4.4)
Thus the average number of electrons will grow if

se

0
> 1 (4.5)
62
0 50 100 150 200
0.5
1
1.5
2
2.5
3
= 1.151
= 1.159
Number of collisions
N
e
/
N
0

[

]
Multipactor in iris
Figure 4.3: The growth in electron number as a function of the number of
gap crossings for two dierent SEY-coecients. Parameters used:
f = 1 GHz, N = 1, l = 2 mm, and W
T
= 2 eV.
or equivalently

se
> 1/
0
> 1. (4.6)
This implies that the secondary electron yield must be greater than a
value that is larger than unity (1/
0
> 1) to have growth of the number
of electrons. This modied breakdown criterion is the only dierence
between the model considered here and the conventional resonance the-
ory of multipactor inside a plane-parallel gap (where
se
> 1 is used
when determining the threshold). The condition for
se
, Eq. (4.6), can
be converted into a range of impact energies,
W
1
< W
min
< W
impact
< W
max
< W
2
, (4.7)
where the impact energies W
min
and W
max
are determined by
se
=
1/
0
. Consequently, using W
min
and W
max
instead of W
1
and W
2
, re-
spectively, in the parallel-plate model, multipactor regions that account
for the electron losses due random drift can be obtained.
63
4.2 Multipactor regions
Using the above described model and by employing a natural scaling
parameter of , viz. l f, multipactor charts in the traditional engi-
neering units (voltage vs. frequency-gap-size product) can be devised
for a specic value of the ratio of gap height and iris length, h/l. Fig-
ure 4.4 shows an example of this, where the multipactor regions have
been constructed for 5 dierent h/l-ratios.
10
0
10
1
10
2
10
3
10
4
Frequency gap height product [GHz mm]
P
e
a
k

v
o
l
t
a
g
e

[
V
]
Multipactor regions for different height/length ratios
0.001
1
3
5.2
6.3
Figure 4.4: The rst four multipactor susceptibility zones for a microwave
iris with ve dierent height/length-ratios. Parameters used are:
W
0
= 2 eV (y-direction), W
T
= 2 eV (z-direction), W
1
= 85.6 eV,
and
se,max
= 1.83 (material properties for alodine [50]).
The transit time decreases with increasing frequency according to
Eq. (4.1) and thus the distance traversed in each step becomes smaller,
which implies that the probability of surviving k steps increases. This
causes the multipactor zones to shrink towards higher frequencies with
increasing h/l as is evident in Fig. 4.4. The transit time is also a function
of the mode order, which increases the transit time for higher order
modes. This counteracts the decrease due to increasing frequency and
consequently a behaviour similar to that of the rst resonance zone can
be observed also for the higher order modes.
For materials with a low maximum SEY, like in Fig. 4.4, the ability
64
to compensate for electron losses is not very good and thus the zones
will disappear for relatively small values of h/l. However, in the opposite
case for a material with a large SEY, like e.g. aluminium, multipactor
will be possible also for relatively thin irises.
4.3 Comparison with experiments
By comparing the current model with experimental data [61], good qual-
itative agreement can be observed (see Fig. 4.5). As the h/l-ratio in-
creases, the threshold increases until, beyond a certain limit, no multi-
pactor is possible. Since only the electron losses due to the random drift
are accounted for, the model predicts the existence of a discharge beyond
the limit found in the experiments. Consequently, from an engineering
point of view, this is a conservative measure of the increased threshold
and thus it should be safe to apply it when designing multipactor free
microwave hardware.
0 1 2 3 4 5 6 7 8
500
1000
1500
2000
2500
3000
3500
Height/Width []
P
e
a
k

V
o
l
t
a
g
e

[
V
]
Multipactor in iris
Current model a)
Current model b)
Current model c)
Current model d)
Measur. presentation
Measur. proceedings
Figure 4.5: The multipactor threshold as a function of h/l for dierent pa-
rameters of the current model. For comparison, measurement
data from [51] is included. Parameters used in a): h = 1.2 mm,
f = 9.56 GHz, W
T
= 2 eV, W
0
= 2 eV, W
1
= 59.1 eV, and

se,max
= 2.22, (i.e. W
1
= W
f2
and
se,max
=
max
in table A-6
for silver in [50])). Modied parameters in: b) W
0
= W
T
= 4 eV,
c) W
1
= 40 eV, and d) W
1
= 80 eV.
65
The step-like behaviour of the increasing threshold is due to the fact
that several multipactor zones are involved (f h 11.5 GHzmm).
Starting with mode number N = 7 for current model a), the lower
threshold for the parallel-plate case is found and as the h/l-ratio in-
creases, the zone corresponding to N = 7 shrinks until, with an almost
sudden voltage step, the next threshold, being determined by the N = 5
zone, is reached. Finally the last N = 3 zone determines the threshold
before it also vanishes.
In addition to the experimental comparison, Fig. 4.5 brings forward
the importance of dierent parameters of the current model as well as of
the used model for SEY [22]. By increasing the initial velocity (current
model b)), the overall threshold decreases as a lower eld strength will
be sucient to reach the same impact velocity (cf. Eq. (2.7)). The eect
of an increased thermal spread, W
T
, is that the electron losses increases
and the threshold starts to increase for lower h/l-values (also shown
in current model b)). By lowering the rst cross-over point (current
model c)), the parallel-plate threshold decreases, since an additional
zone, N = 9, comes into play. However, as it shrinks away, the threshold
increases in a sudden step to the same level as in case a) and then it
follows a) except that the steps occur at higher h/l-values. An increased
rst cross-over point, case d), shows a change of behaviour opposite to
c), except for the parallel-plate threshold as it is still the N = 7 zone
that determines this threshold.
In the current model a uniform electric eld has been used. Due to
the geometry of the iris, the actual electric eld will tend to be curved
outwards at the edges of the slot instead of being straight (cf. Fig. 4.1).
Since the eld amplitude is higher in the centre of the iris than at the
edges, the Miller force [10], which is proportional to the negative gradi-
ent of the square of the electric eld amplitude, will tend to push the
electrons out of the iris. This eect is most important for the higher
order resonances, where several RF-cycles are required to cross the gap.
In addition to the Miller force, the curved electric eld will have a com-
ponent in the z-direction, which, in particular for the rst order mode,
will drive the electrons toward the iris edges. This means that the elec-
tron losses will be greater than in the case of a uniform eld, which
will lead to an even further increase of the multipactor threshold. This
eect should be more pronounced for thin irises and could explain why
the current model predicts the existence of a discharge beyond a certain
h/l-ratio where experiments cannot detect it (cf. Fig. 4.5).
66
4.4 Main results
This analysis has shown that the random electron drift along the iris
length due to the initial velocity of the secondary electrons tends to
signicantly increase the multipactor threshold in a waveguide iris as
compared to predictions based on the classical two parallel-plate model.
Inherent in the presented model is the scaling parameter h/l, which
makes it possible to produce useful multipactor susceptibility charts in
the traditional engineering units. An increase in the h/l-ratio results in
a shrinkage of each multipactor resonance zone. For each zone, the zone
reduction eect is more pronounced for lower frequencies. A consequence
of this is that the multipactor free region in the parallel plate model at
low frequency-gap-height products grows with increasing h/l-ratio.
67
68
Chapter 5
Multipactor in coaxial lines
The coaxial line is a very common and important component in mi-
crowave systems. It is a transmission line that consists of an inner
cylindrical conductor and a coaxial outer conductor. A constant cross-
section is maintained by means of a dielectric medium, which is con-
tained between the conductors. In some space applications, as well as in
other systems, the dielectric medium has been partly omitted in order to
save weight or to reduce the dielectric losses. In a vacuum environment,
the line may become evacuated, which makes it exposed to the risk of a
multipactor discharge.
The electric eld in a coaxial line is nonuniform, which makes analyt-
ical analysis dicult, since the the equation of motion for an electron be-
comes a non-linear dierential equation. However, multipactor in a coax-
ial line has been studied experimentally [63] and numerically [6467]. In
these studies it was found that two dierent types of resonant multi-
pactor can occur, namely a two-sided discharge between the outer and
the inner conductors and the one-sided analogue on the outer conduc-
tor. In an attempt to understand the eect of varying the relative inner
radius, i.e. varying the characteristic impedance of the coaxial line,
dierent scaling laws were suggested in these studies. Another study fo-
cused on the current due to the multipacting electrons and treated this
as a radially oriented Hertzian dipole in order to determine the electric
eld generated by the multipactor discharge [68].
In paper E resonant multipactor in a coaxial line is analysed by
means of an approximate analytical solution of the non-linear dieren-
tial equation of motion, which in a large range of microwave frequen-
cies and amplitudes agrees very well with the numerical solution. As
69
support for the qualitative analytical results, paper F presents PIC-
simulations of the phenomenon in the same geometry. The advantage
of PIC-simulations is the ability to include aspects, which are stochastic
in nature, like e.g. the initial velocity spread of the secondary elec-
trons. This summary will focus on the results of the study and for a
more detailed description of the model and approximations used, see
papers E and F.
5.1 Analytical study
By nding an analytical solution of the equation of motion, general prop-
erties of the multipactor can be found, which may be dicult to identify
when numerically studying the phenomenon. In addition, the time of
computation can be radically reduced when using explicit analytical ex-
pressions instead of a numerical scheme. In this section an approximate
analytical solution of the non-linear dierential equation that governs
the electron motion in the nonuniform eld between the inner and outer
conductor of a coaxial line is found. Using these expressions, the eect
on the multipactor resonances and thresholds is studied. The validity
of the expressions is then conrmed by solving the dierential equation
numerically.
5.1.1 Model
The cylindrical coaxial line has an outer radius R
o
and an inner radius R
i
(see Fig. 5.1). The applied eld is the fundamental TEM-mode, which
means that the electric eld, E, is radially directed and the amplitude
will be inversely proportional to the distance from the centre of the line.
There will be no dependence on the angle around the coaxial axis, which
means that the problem can be studied as a one dimensional problem,
provided that the eect of the magnetic eld is neglected and only a
cross-section of the coaxial line is considered. In vacuum, the equation
of motion for an electron in an electric eld can be written
mr

= qE (5.1)
where m is the mass of the electron, q the unit charge, and E the instan-
taneous strength of the electric eld. The radial position of the electron
is designated r and r

is the second time derivative of the position.


Assuming a time harmonic electric eld, E = E
o
(R
o
/r) sin (t), where
70
Figure 5.1: The geometry used in the considered model.
E
o
is the eld amplitude at the outer conductor, and introducing the
notation = qE
o
R
o
/m, Eq. (5.1) can be written:
r

r
sin t (5.2)
The relation between the eld amplitude and the voltage amplitude is
given by U
c
= E
o
R
o
ln (R
o
/R
i
). Since the eld is inhomogeneous and
stronger near the centre conductor, there will be a net average force that
slowly, compared to the fast harmonic oscillations, pushes the electron
towards the outer conductor. This force is called the ponderomotive or
Miller force [10] and it tends to push the electrons away from regions with
high amplitudes of the RF electric eld. By separating r(t) according
to r(t) = x(t) +R(t), where x(t) is the fast oscillating motion and R(t)
the slowly varying motion (the time averaged position), an approximate
solution of Eq. (5.2) can be derived (see paper E) where the position
and velocity of the electron are given by:
r(t)

2
sin(t)

C
1
(t C
2
)
2
+

2
2
2
C
1
+

C
1
(t C
2
)
2
+

2
2
2
C
1
(5.3)
and
r

(t)
1
R(t)

C
1
(t C
2
) +

cos (t)

, (5.4)
71
where R(t) is the average position,
R(t) =

C
1
(t C
2
)
2
+

2
2
2
C
1
. (5.5)
The constants of integration, C
1
and C
2
, are determined by the initial
conditions, which for an electron starting at the outer conductor are
r(t = t
0
) = R
o
and r

(t = t
0
) = v
0
. Using Eq. (5.3), the position of an
electron emitted from the outer conductor with no initial velocity has
been plotted in Fig. 5.2. The accuracy of the expression is evident from
the comparison with the numerical solution.
0 1 2 3 4 5 6 7 8
5
6
7
8
9
10
Time [ns]
P
o
s
i
t
i
o
n

[
m
m
]
Multipactor in coax
Figure 5.2: Motion of an electron emitted from the outer conductor of a coax-
ial line. The solid line corresponds to the analytical expression
Eq. (5.3), the dotted line is a numerical solution of the dieren-
tial equation Eq. (5.2) (almost covered by the solid line) and the
dashed line is the average motion according to Eq. (5.5). Param-
eters used: V
c
= 1200 V, f = 3 GHz, W
0
= 0 eV (the initial
electron energy), R
o
= 10 mm, and R
i
= 5 mm.
An important result can be obtained by only looking at the average
position, Eq. (5.5). The minimum of this equation, R
min
, is the small-
72
est achievable radial position for an electron emitted from the outer
conductor, provided that the oscillations are not too large. The expres-
sion for the minimum of R(t) will be a function of the eld amplitude,
the frequency, the initial electron velocity as well as the initial phase,
= t
0
:
R
min
= R
o
(
2
+ 2
2
cos
2
+ 4cos v
0
R
o
+ 2v
2
0

2
R
2
o
)
1/2
(5.6)
However, for v
0
= 0 a much more compact expression, which is indepen-
dent of eld amplitude and frequency, is obtained,
R
min

R
o

1 + 2(cos )
2

R
o

3
. (5.7)
This means that if the radius of the inner conductor, R
i
, is smaller than
58% of the outer radius, R
o
, then two sided multipactor is not possible
when the initial velocity is low and the oscillations are small.
5.1.2 Multipactor resonance theory
In a coaxial line, both double-sided and single-sided multipactor (on the
outer conductor) are possible. First, double-sided discharge will be con-
sidered and typical for this is that the one way transit time corresponds
to an odd integer of half RF eld periods. However, in a coaxial line,
the transit time is normally longer for electrons emitted from the outer
conductor than for electrons emitted from the inner conductor. Thus,
the sum of two transits must be considered and the condition for this is
that it should be an integer number of RF periods. This is the resonance
criterion and in addition to this the phase-focusing eect should be ac-
tive, which for the parallel-plate case is given by Eqs. (2.18)- (2.21). It
is instructive to compare the coaxial case with the parallel-plate case,
since in the limit when the R
i
R
o
the coaxial and parallel-plate mod-
els should give the same results. For the parallel-plate case, when the
initial velocity is neglected (v
0
= 0), the phase stability range is given
by the following inequalities [39],
k < <

(k)
2
+ 4, (5.8)
where k is an odd positive integer. The normalised gap width, , is
dened by
= d/V

= m(d)
2
/eU, (5.9)
73
where d stands for the gap width, V

= eE

/m is the amplitude of
the electron velocity oscillations in the spatially uniform RF eld, E

is
the RF eld amplitude inside the gap, and U is the voltage between the
conductors. In addition, for an electron avalanche to start, the impact
velocity should be between the rst and the second cross-over points
(cf. Eq. (2.12)), which for zero initial velocity is given by,
v
1
< 2V

< v
2
. (5.10)
Due to the asymmetry in the electron motion, the simple analytical
analysis that is feasible in the parallel plate case is not applicable for
the coaxial line, despite the fact that the approximate electron position
and velocity are known explicitly (Eqs. (5.3) and (5.4)). One way of
nding the resonance zones in the parameter space is to compute a series
of successive electron trajectories, searching for the conditions when it
converges to a periodically repeated sequence [69]. In Fig. 5.3 one can
see the result of such a method, where stable resonance zones have been
found numerically. To allow simple comparison with the parallel-plate
case, the normalised gap size has been used and in terms of the coaxial
line it becomes,
=
m(d)
2
eU
c
=
G(R
o
R
i
)
2
R
2
o
ln (R
o
/R
i
)
, (5.11)
which coincides with Eq. (5.9) when R
o
/R
i
is close to unity. The con-
venient parameter, G = R
o
/V
,o
, has been introduced as representing
a normalised outer radius and V
,o
= qE
o
/m.
When R
o
/R
i
is close to unity the parallel-plate and coaxial models
give similar results. When the ratio becomes larger, the zones deviate
from the straight lines predicted by Eq. (5.8) and the deviation is more
pronounced for the higher modes. When the ratio becomes too large,
all two sided resonances disappear. This is a consequence of the Miller
force and for the higher order modes, where the approximate analytical
solution is very accurate, the prediction (Eq. (5.7)) is that the double
sided resonances should disappear roughly at R
o
/R
i
=

3 1.73. Fig-
ure 5.3 shows that this is indeed true. The rst order mode, however, is
not accurately described by the analytical expression and the numerical
calculations show that this mode can exist for values of R
o
/R
i
as high
as 4.
In Fig. 5.4 the double-sided discharge regions have been computed
numerically for R
o
/R
i
= 1.4. The classical multipactor zones have upper
and lower thresholds that satisfy Eq. (5.10) in the following sense: since
74
1 1.5 2 2.5 3 3.5 4
4
6
8
10
12
14
16
18
20
22
R
o
/R
i
[]

Double sided multipactor, numerical data


1
c
2
h
3 3
c
4
5
5
c
6
7
8
Figure 5.3: Normalised gap width according to Eq. (5.11) vs. R
o
/R
i
. The
solid straight lines form the classical zones according to Eq. (5.8).
The dots (blue) indicate stable resonances where the sum of two
transits equal an odd number of RF-cycles. The crosses (red) are
for sums equal to an even number of RF-cycles. The sum of the
transit time in RF-cycles for some distinct zones are indicated.
The lowest order hybrid modes are marked with an h and the
classical resonances with a c. The dashed vertical line indicates
R
o
/R
i
=

3 1.73.
75
there are dierent oscillatory velocities depending on whether the elec-
tron starts on the outer or the inner conductor, an average value must
be computed. By setting V
,o
= qE
o
/m = v
1
/2 the corresponding volt-
age, V
Ro
, can be derived. Similarly, by setting V
,i
= qE
i
/m = v
1
/2
the voltage V
Ri
can be obtained. The lower threshold voltage is then
approximately equal to:
U
th

V
Ro
+V
Ri
2
. (5.12)
The upper threshold is computed in a similar manner, only with v
2
instead of v
1
. The hybrid modes, which in general have a lower average
impact velocity will require a stronger electric eld to reach an energy
equal to the rst cross-over point. Consequently, a threshold higher than
the lower envelope is obtained for these zones (cf. Fig. 5.4).
10
0
10
1
10
2
10
3
10
4
A
m
p
l
i
t
u
d
e

o
f

C
o
n
d
u
c
t
o
r

V
o
l
t
a
g
e

[
V
]
Frequency [GHz]
Double sided multipactor: Z
c
= 20
1
c
2
h
3
c
4
6
7
Figure 5.4: Numerically obtained double sided multipactor chart. The
dashed lines are the approximate lower and upper envelopes given
by Eq. (5.12). The sum of the transit time in RF-cycles for each
zone is indicated. Parameters used: W
1
= 23 eV, W
2
= 2100 eV,
W
0
= 0 eV, R
o
= 10 mm, and R
i
= 7.2 mm.
If the inner radius is suciently small, single-sided multipactor be-
comes the dominating scenario. Single-sided multipactor is less compli-
cated than its double-sided counterpart, as the complicated asymmetry
does not appear in this case. This allows an analytical analysis based
76
on the approximate solution for the electron trajectory, Eq. (5.3). The
analytical expression is accurate when the oscillations are small and a
consequence of this is that accurate stable phase multipactor regions are
only found for N 2, i.e. when the duration of the trajectory is at least
2 RF-periods.
By analysing the resonance and stability conditions, one can show
that single-sided breakdown will have not only one region of stable reso-
nant phase, but rather two stable regions can be found. One somewhat
wider region with resonance close to zero and another, which is reso-
nant close to /4. It can be shown that these regions, in the case when
v
0
= 0, are approximately given by:
0 <
R
<
1
(5.13)
and

2
<
R
<
3
(5.14)
where

1

4
N
(5.15)

2


4

1
N
(5.16)

3


4
+
1
N
(5.17)
For increasing N, the second region converges to
R
= /4, which ac-
cording to Eq. (5.7) corresponds to R
min
R
o
/

2. In Fig. 5.5 the


resonant stable phase,
R
, has been plotted as a function of N. Except
for the lowest order resonance (N = 1), the regions of phase stability for
the numerically and analytically obtained phases agree very well.
To obtain the multipactor threshold, it is necessary to know the
impact velocity, which is given by
v
impact
2V
,o
cos +v
0
(5.18)
The lower boundary shown in Fig. 5.6 is obtained for the maximum
impact velocity for each mode, i.e. v
impact
= 2V
,o
when v
0
= 0. In the
same gure one can also identify a second set of regions with a higher
breakdown threshold. These zones correspond to the second stable phase
region, Eq. (5.14). Since the phases in this region are close to /4,
the impact velocity is v
impact

2V
,o
. This value is also indicated in
Fig. 5.6, but it should not be expected to serve as an exact envelope of the
77
0 5 10 15 20
0
10
20
30
40
50
60

3
N (rfcycles)
P
h
a
s
e

[
d
e
g
r
e
e
s
]
Figure 5.5: Stable resonant phase for single-sided multipactor. Analytically
obtained stable phases are shown as diamonds (red) and the ones
obtained numerically are indicated with dots (blue). Eqs. (5.15) -
(5.17) are shown as solid lines. The dashed line indicates =
/4 .
zones as there are phases that are smaller than /4 (cf. Fig. 5.5), which
will yield a higher impact velocity and consequently a lower threshold.
Furthermore, in the numerical solution of Eq. (5.2) for the rst order
mode, an impact velocity of as much as four times V
,o
can be observed.
This results in a threshold much lower than the envelope. The maximum
impact velocity of the second zone is slightly lower than 2V
,o
, resulting
in a somewhat higher threshold. The following higher order modes then
quickly converge to the analytical limit (cf. Fig. 5.6).
When the initial velocity is zero, the only parameters left to vary are
G and the ratio R
o
/R
i
. Since the characteristic impedance in ohms of
a coaxial line in vacuum is given by Z 60 ln (R
o
/R
i
), it follows that
only two parameters remain to be varied, viz. G and Z. By following
trajectories for dierent values of G and Z, stable phase points were
found in this parameter space and the result is plotted in Fig. 5.7, which
was produced using a numerical solution of the equation of motion (a
version of this gure using the analytical expressions can be found in
paper E).
The straight lines in Fig. 5.7 on the right hand side are regions of
stable single-sided resonances. The fact that these appear as straight
78
10
0
10
1
10
2
10
2
10
3
10
4
10
5
Frequency x R
o
[GHz mm]
A
m
p
l
i
t
u
d
e

o
f

C
o
n
d
u
c
t
o
r

V
o
l
t
a
g
e
/
l
n
(
R
0
/
R
i
)

[
V
]
Single sided Multipactor: Z
c
=100
Figure 5.6: Single-sided multipactor breakdown regions based on the numer-
ical solution of Eq. (5.2). Regions corresponding to N = 1 22
RF-periods are shown. The regions with an initial phase, , close
to zero are produced using dots (blue) and the regions with close
to /4 are indicated by crosses (red). The dash-dot line is the
approximate lower envelope given by V
,o
= v
1
/2 and the dashed
line is given by V
,o
= v
1
/

2. Parameters used: W
1
= 23 eV,
W
2
= 2100 eV, W
0
= 0 eV, and Z = 100 (corresponding to
e.g. R
o
= 10 mm and R
i
= 1.88 mm.)
79
0 0.5 1 1.5 2
0
10
20
30
40
50
60
70
80
90
100
Z/Z
0
(Z
0
=50 )
G
Figure 5.7: The normalised parameter G vs. normalised characteristic
impedance Z. Each mark represents a stable phase solution and
an eort has been made to suppress polyphase modes in order
to clearly show the behaviour of the main resonance modes. The
chart was obtained by numerically solving the equation of mo-
tion. Stars mark (blue): double-sided multipactor, dots (red):
single-sided multipactor with 0 <
R
< 20
o
, and crosses (green):
single-sided multipactor with
R
> 20
o
. The dashed line indi-
cates R
i,min
= R
o
/

3 and the dash-dot line R


i,min
= R
o
/

2.
80
lines indicates that there is no dependence on Z, which implies that a
simple scaling law exists in the single-sided case, viz.
P (R
o
)
4
Z. (5.19)
In Fig. 5.6 this law has been used to normalise the axes, but voltage
is used on the ordinate instead of power. The chosen normalisation
of the axes in Fig. 5.6 is general and using the analytical solution of
Eq. (5.2) presented above, it can be shown that this normalisation is
valid also for non-zero initial velocity. It is important, however, to be
careful when scaling to a dierent radii ratio, since for smaller values
of the characteristic impedance, the single-sided multipactor zones may
not exist at all.
In the double-sided case, it is evident that G is a function of Z.
Consequently a more complicated scaling law should be expected. For
small values of Z, however, the coaxial case becomes similar to the
parallel-plate geometry, where the resonance voltage can be written as
function of the frequency-gap-size product. For the coaxial case, this
scaling law becomes
P ((R
o
R
i
))
4
1
Z
. (5.20)
and for the rst order resonance this scaling law is quite accurate (cf.
Fig. 5.3), but for the higher order modes it quickly loses its validity with
increasing Z.
5.1.3 Main ndings
A qualitative comparison with experiments [63] shows good agreement
with the present analysis. The experimental data shows an increase in
the multipactor threshold for increasing radii-ratio R
o
/R
i
. It was also
found that the rst multipactor zone became narrower for increasing
R
o
/R
i
. These features are in agreement with the results of this study
as shown in Figs. 5.3 and 5.7. By mapping the data of Fig. 5.3 into the
voltage vs. frequency-gap-size space, used in the experiments, a clear
threshold increase compared with the parallel-plate case can be seen as
well. Even though the experiments used quite large values of R
o
/R
i
,
no case of single-sided multipactor was observed. This can be explained
by the fact that a material with a low rst cross-over point was used,
where the initial velocity will play an important role when the applied
voltage is not high enough. More importantly, only the rst order mode
was studied and in this case, as shown in Fig. 5.3, multipactor will be
81
possible for quite large R
o
/R
i
-values. In the simulations by Sakamoto et
al [64], the experimental results were conrmed. In addition, single-sided
multipactor was observed, which conrms the result of this theoretical
study.
Among the more important results of this part of the study are the
following. The analytical approximate solution of the nonlinear dieren-
tial equation of motion for an electron in a coaxial line. The dual regions
of stable phase, Eqs. (5.13) - (5.17), which explain why single-sided mul-
tipactor will be possible also for smaller values of R
o
/R
i
. The scaling
law for single-sided multipactor, Eq. (5.19), simplies the presentation
of multipactor prone regions of the single-sided case. The limit formula
for the transition from double- to single-sided multipactor, Eq. (5.7), is
an interesting feature for future experiments to conrm. The reduced
threshold for the rst order zone of single-sided multipactor, which must
be taken into account when constructing the lower boundary of all the
zones. Finally, this analysis shows that the behaviour of resonant multi-
pactor is signicantly aected by the nonuniform eld and it shows the
benets of analytically studying dierent geometries to understand the
basic behaviours before performing numerical simulations.
5.2 Particle-in-cell simulations
In this part of the coaxial study, which is based on paper F, extensive
PIC-simulations have been performed in order to verify the analytical
results as well as to investigate the importance of initial velocity spread
and dierent maximum secondary emission. One of the advantages with
PIC-simulations is the ability to include parameters that are stochastic
in nature. The stochastic properties of some of the parameters as well
as the actual value of the maximum secondary emission coecient may
have signicant eects on the multipactor threshold as well as on the
existence of a discharge. This has previously been shown in the case
of plane-parallel geometry [46] and it is demonstrated that the eect is
similar also in coaxial geometry.
5.2.1 Numerical implementation
The geometry and eld is described in the analytical section. The code
uses normalised parameters such as G and and the SEY follows the
model by Vaughan [22]. The initial velocities of the secondary electrons
82
are assumed to have a Maxwellian distribution, i.e.
f(v
x
, v
y
, v
z
)
v
n
v
exp

1
2
(
v
v
T
)
2

(5.21)
where v is the absolute value of the initial velocity, v
n
its normal com-
ponent with respect to the surface of emission, and v
T
is the thermal
initial velocity spread. Another of the used parameters related to this
is the normalised spread of initial electron velocity dened as v
T
/v
max
,
where v
max
is the impact velocity for maximum SEY.
Calculations were performed for 2-D arrays of dierent sets of the
normalised parameters (e.g. = V

/v
max
vs. with the other param-
eters xed) and each run corresponds to one particular point in one of
these arrays. Each run was primed with 200 seed electrons, uniformly
distributed over initial phase, and the run was terminated when either
the number of particles exceeded 4500 or when 200 RF-periods had
elapsed. The run was also terminated in case the number of electrons
dropped below 10 before 200 RF-cycles had passed. At the end of each
run, the following parameters were recorded:
Number of RF-periods needed to exceed 4500 particles. If 4500
particles were not attained within 200 RF-cycles, this parameter
was set to 200.
Number of electrons at the end of each run.
Heating asymmetry, i.e. the ratio between the average power de-
posited on the inner conductor and the average power deposited
on the outer conductor.
Average electron growth rate (over the 10 last RF-periods), nor-
malised with respect to the RF-period.
5.2.2 Simulations
To facilitate comparison between the theoretical result presented in
Fig. 5.7 a simulation was made in the same parameter space. Figure 5.8
shows the number of electrons obtained after 200 RF-cycles. As ex-
pected, the lower order resonances (i.e. at lower and leftmost G-values)
indicate high electron numbers, since more impacts with the conductors
will occur during the same number of RF-periods. Due to this fact, a
parameter space was chosen, which did not include any points in the
83
upper right region of the gures, where the electron growth is very slow.
There is good agreement in the general behaviour and the transition
from double-sided to single-sided multipactor occurs at more or less the
same impedances for the dierent zones in both the PIC-simulation and
the theoretical data, since the non-zero initial velocity in the PIC-data
is small relative to the oscillatory velocity.
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
10
20
30
40
50
60
70
80
90
100
Z/Z
0
(Z
0
=50 )
G
500
1000
1500
2000
2500
3000
3500
4000
4500
Figure 5.8: Number of electrons after 200 RF-cycles. The white dots are
points of 2-sided multipactor, the green and yellow dots 1-sided
multipactor, and the white dashed lines correspond to R
i,min
=
R
o
/

2 (left) and R
i,min
= R
o
/

3 (right) - all from Fig. 5.7.


Parameters used:
se,max
= 1.6, W
1
= 50 eV, v
T
/v
max
= 0.01,
= V
,o
/v
max
= 0.5, and f = 1.5 GHz.
The straight lines on the right hand side of Fig. 5.8 reveal that this
should be single-sided multipactor. This is conrmed by looking at the
ratio of power deposited on the inner and the outer conductors, which
directly identies the type of discharge. In Fig. 5.9, the dark blue areas
are regions where most or all power is deposited on the outer conductor,
which implies single-sided multipactor on this conductor. The orange
84
and yellow areas indicate that a similar amount of power is deposited
on both conductors and thus the double-sided scenario dominates.
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
10
20
30
40
50
60
70
80
90
100
Z/Z
0
(Z
0
=50 )
G
2.5
2
1.5
1
0.5
0
0.5
1
Figure 5.9: Ratio of power deposited on the inner and outer conductors due
to electron impacts, logarithmic scale log
10
(P
inner
/P
outer
). The
same parameters are used as in Fig. 5.8.
Since the theoretically obtained points agree in general with the PIC-
simulations, this conrms the validity of the scaling laws, Eqs. (5.19) and
(5.20). The two dierent types of modes for single-sided discharge, one
with a phase 0 and the other with /4 can also be seen. The
former type of mode is discontinued before reaching the rst dashed line
from the right hand side and the latter before the other dashed line in
Fig. 5.7. This is also evident in the PIC-data, especially for values of G
between 30 and 40 in Fig. 5.9 where they are fairly well separated and
only the lower of the paired bands extend into the region between the
dashed lines, as predicted.
In Figs. 5.8 and 5.9 the multipactor threshold can not be identied,
since the oscillatory velocity is kept constant. In Figs. 5.10 and 5.11,
however, the oscillatory velocity has been swept for dierent G-values
while keeping the ratio R
i
/R
o
constant and equal to 0.7, i.e. Z = 21.4
(Z/Z
0
= 0.428). Each gure is produced for a dierent maximum SEY.
When
se,max
is low, the ability to compensate for losses is weak and the
zones are well dened and fairly narrow (cf. Fig. 5.10). With increasing
85

se,max
the zones become wider and zones previously suppressed by the
losses can appear (cf. Fig. 5.11). This behavior is very similar to that
noted for the parallel plate geometry [46]. The lower (left) envelope is
500
1000
1500
2000
2500
3000
3500
4000
4500
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
5
10
15
20
25
30
35
40
45
50
V

/V
max
G
Figure 5.10: Number of electrons after 200 rf-cycles. The vertical straight
lines indicate the lower and upper theoretical envelopes accord-
ing to Eq. (5.12). Parameters used: R
i
/R
o
= 0.7, f = 1.5 GHz,

se,max
= 1.3, v
T
/v
max
= 0.01, and W
1
= 50 eV.
obeyed, but the upper can be exceeded since a non-zero initial phase will
yield a lower impact velocity and a concomitant higher upper threshold.
In Figs. 5.10 and 5.11 the velocity spread is quite low, v
T
/v
max
=
0.01. When increasing this ratio, a greater portion of the electrons
can have a large negative initial phase, which leads to overlapping of
the multipactor regions when
se,max
is large (see Fig. 5.12). On the
other hand, the increased velocity spread also increases the losses, which
especially aects the higher order resonances, since the phase-focusing
eect gets weaker with increasing mode order. If
se,max
is low, the
losses are not suciently compensated for and this leads to suppression
of the higher order modes (cf. Fig. 5.13). This is a result similar to
that found in the parallel plate case [46]. The corresponding behaviour
is seen also for single-sided multipactor (see paper F).
For single-sided multipactor it was noted that for the rst order
mode, the envelope of the breakdown zones V
,o
= v
1
/2 was not obeyed.
This is conrmed by the PIC-simulations, where the rst order mode
86
500
1000
1500
2000
2500
3000
3500
4000
4500
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
5
10
15
20
25
30
35
40
45
50
V

/V
max
G
Figure 5.11: Same as Fig. 5.10 only with
se,max
= 2.0.
500
1000
1500
2000
2500
3000
3500
4000
4500
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
5
10
15
20
25
30
35
40
45
50
V

/V
max
G
Figure 5.12: Number of electrons after 200 rf-cycles. The vertical straight
lines indicate the lower and upper theoretical envelopes accord-
ing to Eq. (5.12). Parameters used: R
i
/R
o
= 0.7, f = 1.5 GHz,

se,max
= 2.0, v
T
/v
max
= 0.1, and W
1
= 50 eV.
87
500
1000
1500
2000
2500
3000
3500
4000
4500
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
5
10
15
20
25
30
35
40
45
50
V

/V
max
G
Figure 5.13: Same as Fig. 5.12 only with
se,max
= 1.3.
clearly passes this limit (see Fig. 5.14).
5.2.3 Comparison with experiments
As mentioned in the introduction to this chapter, PIC-simulations can
include aspects of the multipactor, which are dicult to analyse theoreti-
cally. As a more realistic description is possible, quantitative comparison
with experiments becomes feasible.
In the experimental study by Woo [63] coaxial lines made of copper
were used. When taking values for the secondary electron emission prop-
erties for copper from the ESA standard [50], the lower thresholds of the
PIC-simulations are not in very good agreement with the experimental
data. However, the secondary emission properties can vary a great deal
between dierent samples of the same material and contamination can
reduce the rst cross-over point and increase the maximum SEY. Thus
by slightly lowering the rst cross-over point in the PIC-simulations,
very good agreement is obtained (see Fig. 5.15). In paper F an alter-
native method of obtaining the experimental threshold when using the
secondary emission properties given in the ESA standard [50] is pre-
sented, based on a modication of the used model for the SEY [22].
However, this will not be discussed further in this summary.
In the experiments an increase in the multipactor threshold for de-
88
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
5
10
15
20
25
30
35
40
45
50
V

/V
max
G
0
500
1000
1500
2000
2500
3000
3500
4000
4500
Figure 5.14: Number of electrons after 200 rf-cycles. The vertical straight
lines indicate V
,o
= v
1
/2 and V
,o
= v
2
/2. Parameters used:
R
i
/R
o
= 0.1, f = 1.5 GHz,
se,max
= 2.0, v
T
/v
max
= 0.01, and
W
1
= 50 eV.
l
o
g
1
0
(
A
m
p
l
i
t
u
d
e
)

[
V
]
log
10
(Frequency) [GHz]
1.6 1.4 1.2 1 0.8 0.6
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
3
3.2
500
1000
1500
2000
2500
3000
3500
4000
4500
Figure 5.15: Multipactor breakdown regions for copper electrodes. The re-
gion conned by circles (or white crosses, when inside a dark
region) is from Ref. [63]. The dark regions are obtained by the
PIC-code with: R
o
/R
i
= 2.3 (Z = 50 ),
se,max
= 2.25 (from
table A-6 in [50], W
1
= 27 eV, and v
T
= 3 eV (v
T
/v
max
=
0.111).
89
creasing ratio R
i
/R
o
was observed. It was also found that the rst mul-
tipactor zone became narrower for decreasing R
i
/R
o
. This behaviour
was seen in the analytical analysis and it is evident also in the PIC-
simulations. Figure 5.16 shows the result of a simulation for Z = 174
(R
o
/R
i
= 18.26). The agreement between simulations and experiments
is good. The multipactor zone becomes narrower and the threshold in-
creases. The position of the zone is slightly shifted compared with the
experimental data and the reason for this deviation may partly be ex-
plained by an inaccuracy of the dimensions of the copper electrodes.
The main reason, however, seems to be related to the SEY-model and
the description of the initial velocity of the secondary electrons, but a
more detailed study will be necessary to settle this matter.
l
o
g
1
0
(
A
m
p
l
i
t
u
d
e
)

[
V
]
log
10
(Frequency) [GHz]
1.6 1.4 1.2 1 0.8 0.6
1.8
2
2.2
2.4
2.6
2.8
3
3.2
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
Figure 5.16: Ratio of power deposited on the inner and outer conductors
due to electron impacts (log-scale) for the same parameters and
SEY-formula as in Fig. 5.15 except that R
o
/R
i
= 18.26.
In the section summarizing the main ndings of the analytical part,
it was mentioned that no single-sided multipactor was noted in the ex-
periments. This is in agreement with the PIC-simulations and shows
that no measurements were made at high enough voltage and frequency,
where single sided multipactor would be the main phenomenon (the dark
blue regions in Fig. 5.16).
90
5.2.4 Main conclusions
The analytically obtained results are conrmed by the PIC-simulations.
In addition, the signicance of initial velocity spread as well as dier-
ent maximum secondary electron emission have been highlighted. The
results have shown that the behaviour with respect to these conditions,
in both the single-sided and the double-sided cases, is qualitatively the
same as for the parallel plate multipactor. The results are in good agree-
ment with available experimental data, but the commonly used model
for SEY seems to be inadequate for large R
o
/R
i
-ratios, since a small,
but not negligible, deviation from the experiments was noticed.
91
92
Chapter 6
Detection of multipactor
In the space industry, where multipactor is a well-known problem, great
care is taken to avoid the phenomenon. According to the ESA stan-
dard on multipactor design and test [50] only single carrier parts with a
margin of 8-12 dB (depending on the type of component) in the analysis
stage are exempt from testing. The recommended corresponding margin
for multicarrier parts is 6 dB between the peak power of the multicarrier
signal and the single carrier threshold. By performing unit acceptance
tests, the margins can be reduced to 3-4 dB in the single carrier case
and to 0 dB in the multicarrier case. In order to take advantage of the
smaller margins, accurate and reliable testing is required to verify that
the prescribed margins are actually fullled and successful testing relies
on accurate and unambiguous methods of detection.
This chapter starts by briey reviewing some of the most common
methods of multipactor detection. Then a method of detection is pre-
sented, which can be used in combination with other methods to achieve
accurate and unambiguous test results. Beginning with the single carrier
case, a theory for the underlying mechanism making the method possi-
ble is given together with some supporting data. This is followed by a
discussion outlining how the method can be applied also in the multicar-
rier case. Detection of multipactor using RF power modulation, which
is the main topic of this chapter, was analysed in paper A of this thesis.
6.1 Common Methods of Detection
There are several dierent ways of detecting multipactor and they can be
divided into two fairly distinct categories, viz. global and local methods
93
of detection. The global methods are characterised by being able to
discern whether or not microwave breakdown is taking place somewhere
within the tested system. However, it can not pinpoint the location of
the discharge. For ight hardware it is important to completely avoid
multipactor in the entire microwave system and consequently this type
of testing is preferred. During the product development stage, it may
be of interest to know not only that a discharge is taking place, but also
its exact position within the system. In such a case, a local method can
be useful, since it can be used to monitor a certain area inside a device.
6.1.1 Global methods
When performing systems tests on ight hardware, global methods of
detection are normally used and there are several reasons for this. The
methods can usually be applied without modifying the component, which
is advantageous as modications can aect the electromagnetic proper-
ties and give inadequate measurement results. In cases where the dis-
charge is weak, many local methods are unable to detect the phenomenon
and thus the often more sensitive global ones are a better choice. Fur-
thermore, it is a requirement of the ESA standard [50] that two methods
of detection should be used and at least one of them should be global.
By using two methods of detection, the risk for misinterpretations is re-
duced. The most common global methods of detection will be described
in the following subsections.
Close-to-carrier noise
Multipacting electrons will be accelerated to high velocities by the elec-
tric eld and at regular intervals, 2/N times per eld cycle, the electrons
will hit an electrode or device wall and experience a sudden deceleration.
The radiated power is a function of the electron acceleration and it is
described by Larmors formula,
P =
2
3
e
2
4
0
v
2
c
3
(6.1)
where v is the acceleration,
0
the dielectric constant of vacuum, and c
the speed of light. For an applied sinusoidal electric eld, the electrons
will experience a sinusoidal acceleration, except for the sudden inter-
ruptions when colliding with the electrodes. Fig. 6.1 shows what the
acceleration may look like for rst order multipactor (N = 1), where the
electrons hit the electrodes once every half cycle of the electric eld.
94
1.76 1.78 1.8 1.82 1.84 1.86
x 10
8
1.5
1
0.5
0
0.5
1
1.5
x 10
6
C2CNoise Timedomain
A
m
p
l
i
t
u
d
e

o
f

a
c
c
e
l
e
r
a
t
i
o
n

[
G
m
/
s
2
]
Time (s)
Figure 6.1: A qualitative plot of the electron acceleration for multipacting
electrons with N = 1 (average value for a large number of elec-
trons).
The regular deviation from a pure sinusoid will generate harmonics,
but these will be discussed in the next subsection. Due to variations in
the time between impact and emission of new electrons, in the time it
takes to decelerate the electrons, and also in the number of electrons,
close-to-carrier noise will be generated. In fact, most noise is generated
at the carrier frequency, but this can not be seen in a measured spectrum,
as the signal amplitude masks the noise. By performing a Fast Fourier
Transform (FFT) of a sequence like the one shown in Fig. 6.1, the noise
generated close to the carrier can be seen (see Fig. 6.2). Very close to
the carrier, the noise level is even higher than what can be discerned in
Fig. 6.1 and by using a bandpass lter with high rejection at the carrier
frequency, the noise increase close to the carrier can be detected with a
spectrum analyser. In combination with a low-noise amplier, this can
be a very sensitive method of detection.
This method of detection can be used for both single and multicarrier
signals. Care should be taken, however, when using a pulsed signal,
since such a signal will generate harmonics and if the pulse length and
type is not chosen properly, the harmonics may be generated in the
measurement band [31]. A risk with this method is that other sources of
95
2 4 6 8 10 12 14 16
100
90
80
70
60
50
40
30
20
Multipactor Noise Power
(
d
B
)
Frequency (GHz)
Figure 6.2: Noise spectrum of a multipactor simulation like the one in
Fig. 6.1, but with N = 3. The signal frequency is f
0
= 2 GHz.
The odd harmonics as well as the peaks at odd multiples of f
0
/3
are clearly visible.
noise may be misinterpreted as multipactor noise. Nevertheless, by using
two dierent methods of detection as required by the ESA standard, this
risk is greatly reduced.
Third harmonic
The resonant behaviour of the multipactor discharge and the repetitive
acceleration and sudden deceleration of the electrons will generate noise,
which will have harmonics at the basic frequency, f
0
, and at odd mul-
tiples of this frequency (cf. Fig. 6.2). Higher order multipactor will
have harmonics not only at odd multiples of the basic harmonic, but at
each odd multiple of f
0
/N [14] (cf. Fig. 6.2). In such cases there are
many dierent frequencies available for detection. However, since the
third harmonic usually is the most powerful harmonic and it is always
present, regardless of the order of resonance, it is the best choice for
detection.
Third harmonic detection is a very reliable and fast method of de-
tection. According to Ref. [14], it gives the fastest indication of mul-
tipactor and that makes it the method of choice when studying mul-
96
tipactor events that are short-lived, like e.g. multicarrier multipactor,
where the discharges often are weak and of short duration. However,
the method is also sensitive to other sources of noise in the same way as
close-to-carrier noise and, thus, it is not recommended to use only third
harmonic and close-to-carrier noise detection in a test setup.
Reected power
Mismatches in the transitions between microwave components imply
that some of the input power will be reected. However, in a well-
designed system, very little power is reected. Components containing
high Q-value parts, like e.g. cavity resonators, are only well matched
at certain frequencies and minor changes in the component properties
can lead to detuning of the part. Multipactor is known to be able to
detune high Q-value components [70]. Thus the reected power from
a component can be used as an indication of a multipactor event. To
study the absolute value of the reected power is usually not a good way
of detection, since the input power can vary during a multipactor test
and consequently the reected power will vary as well, as the reected
power is a xed fraction of the input power. This fraction is called the
return loss and is commonly measured in decibel. The return loss will be
a stable value, insensitive to power uctuations, until the component is
detuned. Figure 6.3 shows an example where both close-to-carrier noise
and the return loss are monitored simultaneously during a multipactor
test. Both methods indicate a change around t = 50 s and thus it can
be determined with great condence that a multipactor discharge was
initiated at that time.
The main advantage with this method is that it is quite reliable and
there is little risk that other phenomena will cause a mismatch that can
be confused with multipactor. However, for low Q-value components or
badly matched systems, the sensitivity of the method is low.
Electron monitoring
A new global method of detection was presented at the 4
th
Interna-
tional Workshop on Multipactor, Corona and Passive Intermodulation
in Space RF Hardware [71]. It is called the Electron Density Detec-
tion Method, abbreviated EDDM, and uses a set of tri-axial cables as
a probe and electrons picked up by the probe are then monitored with
a high precision electron meter. The data is collected using a computer
97
0 10 20 30 40 50 60 70 80 90 100
90
80
70
60
N
o
i
s
e

(
d
B
m
)
Time (s)
0 10 20 30 40 50 60 70 80 90 100
42
44
46
I
n
p
u
t

P
o
w
e
r

(
d
B
m
)
Time (s)
0 10 20 30 40 50 60 70 80 90 100
16
14
12
10
8
R
e
t
u
r
n

L
o
s
s

(
d
B
)
Time (s)
Figure 6.3: Multipactor monitored by studying close-to-carrier noise as well
as the return loss. At t 50 s the component starts to experience
multipactor discharge and becomes increasingly mismatched. Ex-
cellent agreement with the second detection methods, close-to-
carrier noise, can be seen.
98
software and the time evolution of the electron density can be studied.
The probe does not have to be located inside or close to the area where
the discharge will take place. Even though it is advantageous for higher
sensitivity to have the probe pointing at the critical gap it is also reli-
able when located outside the device under test and thus it can be used
for waveguides and coaxial transmission lines. It is not completely clear
from the paper, however, how the electrons escape from a completely
conned device like a typical waveguide or coaxial cable and one will
have to assume that there must be some small opening somewhere in
the system, where electrons can leak out.
Furthermore, the method can be used to quantitatively measure the
amount of generated electrons. however, this requires some kind of cali-
bration of each test setup and that may prove to be problematic. When
used in this way, the method can no longer be viewed as a global method,
which can detect a multipactor event anywhere in the system, instead it
has become a local method. Among the main advantages of the method
is the low cost involved, since no expensive microwave instruments are
needed.
Residual mass
A very slow global method of detection is to detect the gas molecules,
which are outgassed from the device walls due to the electron bombard-
ment during a multipactor event. The gas molecules consist of resid-
uals of water, air and contaminants, and using a mass spectrometer,
the dierent molecules can be identied. It has been noted [31] that a
detectable increase in the water spectrum can be seen during a multi-
pactor discharge. The major drawback of this method of detection is
its inability to detect fast multipactor transients (not enough molecules
are released from the walls) and thus it is not a suitable method for
multicarrier multipactor studies. Another disadvantage is that there is
a certain delay between onset of the discharge and indication in the in-
strumentation. However, it can be useful as a diagnostic tool together
with one or two of the other described methods.
6.1.2 Local methods
In cases where it is not sucient to only conrm the existence of a
discharge in the system, but also to determine the exact position, local
methods of detection will have to be used. The two most common local
99
methods are charge detection using a probe and optical monitoring and
these two methods will be described below.
Charge probe
A special case of the method of electron monitoring, which was described
in a previous subsection, is the use of a probe that monitors only a
certain position within the microwave device. A very common approach
used when detecting electrons inside a waveguide is to ush mount an
SMA connector and apply a positive potential to the centre pin. The
negatively charged electrons are attracted to the pin and causes a small,
but detectable, current to ow, which can serve as an indication of the
electron density.
The method is easy to implement and therefore it has been quite
common in many test setups. Unfortunately, it is also quite slow due
to the circuit used to amplify the weak current [14] and thus it is not
feasible for measuring fast multipactor events. In addition, it requires
modication of the component, which makes it useless when testing ight
hard ware. In such a case the EDDM is a better choice.
Optical detection
Optical detection is possible since the electrons that make up the multi-
pactor discharge can excite or ionise either the remaining gas molecules
within the device or the molecules in the device wall. It can be divided
into two groups, viz. photon detection via optical probe and photon
detection via photographs or video camera [72, 73]. Both methods are
common, but the former seems to be more frequently used.
The main advantage with these methods is that they can be used to
pin-point the location of the discharge inside the device. A major dis-
advantage is that they may be impossible to use for studying real parts,
as the devices may not have any suitable openings. This is especially
true for the methods based on photographic techniques.
6.2 Detection using RF Power Modulation
During verication of a test setup for multipactor at Saab Ericsson
Space, Goteborg, Sweden, an odd, spike-like phenomenon was found
in the noise generated by a discharge in a coaxial test sample. An ad-
100
ditional test sample, a resonant cavity, was manufactured and the same
type of spikes were noted again (cf. Fig. 6.4).
92.82 92.84 92.86 92.88 92.9 92.92 92.94 92.96 92.98 93
90
85
80
75
70
65
60
55
Noise Power (92.8 93.0 s)
(
d
B
m
)
Time (s)
Figure 6.4: Periodic spikes, which appeared during a multipactor experiment.
The main periodicity is 100 Hz and emanates from the power
supply of the high power amplier.
A fast Fourier transform revealed that the spikes were periodic and
it was noted that the same periodicity could be found also in the input
signal after it had been amplied by the TWTA (travelling wave tube
amplier). Due to interference from the main power supply, the signal
was amplitude modulated with a main modulation frequency of 100 Hz
and with harmonics at multiples of this frequency. The interference was
very weak and would in most cases be disregarded. In order to see if
the periodicity was present only in conjunction with multipactor events,
a large number of test runs were performed. The results were consis-
tent - the periodic noise only appeared when a discharge was detected.
Figs. 6.5 and 6.6 show one of the test runs where the noise is non-periodic
before onset of multipactor but periodic directly afterwards.
The AM (amplitude modulation) that was present in the input signal
was very weak and not deliberately added. A stronger AM was added to
the signal before the high power amplier, resulting in a more distinct
modulation. This made it possible to study if there was any correla-
tion between the modulation strength and the corresponding peak in
101
34 36 38 40 42 44 46 48 50 52
90
85
80
75
70
65
60
55
Noise Power (32 52 s)
(
d
B
m
)
Time (s)
0 50 100 150 200 250 300 350 400 450 500
160
140
120
100
80
Fourier transform (32 52 s)
(
d
B
m
)
frequency (Hz)
Figure 6.5: The beginning of a multipactor test sequence showing the time
before onset of the discharge. The FFT (fast Fourier transform)
gives no indication of dominant frequency components.
54 56 58 60 62 64 66 68 70 72
90
85
80
75
70
65
60
55
Noise Power (52 72 s)
(
d
B
m
)
Time (s)
0 50 100 150 200 250 300 350 400 450 500
160
140
120
100
80
Fourier transform (52 72 s)
(
d
B
m
)
frequency (Hz)
Figure 6.6: The end of the same test sequence as in Fig. 6.5. The sudden in-
crease in the noise oor indicates onset of multipactor. The FFT
shows dominant frequency components at 100 Hz and multiples
thereof.
102
the FFT. It was concluded that there is a positive correlation between
the modulation depth and the strength of the detected signal at the
corresponding frequency. By taking the time average of one of the test
sequences like in Fig. 6.3 and plotting it using linear scales on both axes,
it was found that there was a more or less linear relationship between
the input power and the resulting multipactor noise power (cf. Fig. 6.7),
which can be described by the following function:
P
noise
= k (P
input
P
th
) [W] P
input
P
th
(6.2)
where k = 5.3 10
11
and P
th
= 25.2 W is the multipactor threshold.
50 55 60 65 70 75 80 85 90 95 100
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Linear Plot of Noisepower vs Input Signal Power
N
o
i
s
e

P
o
w
e
r

[
n
W
]
Time (s)
26 28 30 32 34 36 38 40 42 44
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Input Signal Power [W]
Figure 6.7: Time average of the test sequence shown in Fig. 6.3 with a linear
scale on both axes. The straight line has been added to show the
close to linear relationship between input power and noise power.
Note: The input power is increased every 4 seconds, which is the
reason for the step like behaviour.
The reason why the small modulation was so noticeable in the mul-
tipactor noise (see Fig. 6.4) is that the noise signal is a function of the
dierence between the input power and the multipactor threshold, i.e.
no discharge noise is generated until the multipactor threshold has been
reached. Furthermore, since the decibel scale is a relative scale, the small
increase in absolute numbers becomes very noticeable in relation to the
existing noise oor. As a comparison, the rst small steps in Fig. 6.7,
103
correspond to huge increases in the decibel scale, which can be seen in
Fig. 6.8.
0 10 20 30 40 50 60 70 80 90 100
85
80
75
70
65
60
Noise Power (100 samples noise average)
(
d
B
m
)
Time (s)
Figure 6.8: The same sequence as in Fig. 6.7 (except that in this case the
entire sequence is shown). The small initial steps of Fig. 6.7
become huge steps on the logarithmic scale.
The above examples, which used the new detection method, relied
on close-to-carrier noise measurement data. However, the mechanism
which is utilised requires primarily that the input signal is amplitude
modulated, that the detected signal is proportional to the dierence
between the input power and the multipactor threshold and that the
detected signal responds quickly to changes in the multipactor event.
The two rst conditions are likely to be fullled by all detection methods
for multipactor, but the last will have to be veried for each method.
Results from measurements presented in [14] show that third harmonic
detection is faster than close-to-carrier noise detection, which should
make it excellent for AM detection. In general, probably any method
can be used provided that a suitable AM frequency is chosen and that
the instrument used for detection has a sampling frequency that is more
than two times larger than the modulation frequency in order to full
the Nyquist criterion.
104
6.2.1 Single carrier
When using the AM detection method in the single carrier case, the
test setup implementation is quite straight forward. In practice, the
only dierence between a standard multipactor test setup and one that
uses the AM detection method is that the signal data must be sent to
a computer or some other instrument that can perform a FFT. It is
also important that the signal generator is capable of producing an AM
signal, but that is a standard feature of most signal generators.
The spectrum analyser, which receives the signal, should be set to a
sampling rate that is at least two times larger than the AM signal, i.e. if
the AM signal has a frequency of 1000 Hz, then a minimum sampling rate
of 2000 samples/second should be used. However, in order to study the
shape of the modulation signal, it is better to use a sampling frequency
more than twenty times greater than the frequency of the AM. The signal
data can be stored for future processing, or if a powerful computer is
used, real time FFT can be performed, thus allowing the operator to get
an immediate indication when a discharge takes place.
6.2.2 Multicarrier
AM detection in the multicarrier case [74] is somewhat more dicult and
the method has not yet been experimentally conrmed. When perform-
ing a multicarrier multipactor test, the phase of each carrier will have to
be adjustable in order to enable the test engineer to produce the wanted
shape of the signal envelope. The aim is to produce a signal envelope
corresponding to the assumed worst case. When the phases have been
set to their predened values, the envelope will be periodic. If the enve-
lope exceeds the multipactor threshold for a time long enough to allow a
sucient number of gap crossings, a discharge will occur, provided that
a suitable seed electron is available. When the envelope drops below
the threshold again, the electrons will disappear quickly [21] and there
will normally be no electrons left from the discharge the next time the
envelope exceeds the threshold. If no specic source of seed electrons is
available, a multipactor discharge may not occur every envelope period.
However, if an ecient electron seeding source is used, there should be
an ample amount of electrons available to initiate a discharge each time
the envelope exceeds the threshold for a suciently long time.
The use of AM detection of multipactor requires that the discharge
event is continuous or that it occurs regularly. Single or very sporadi-
105
cally occurring discharges will be dicult to identify in a FFT plot. By
using a source of free electrons in the test setup, e.g. a hot lament
or a UV light [14], seed electrons will be abundant and a multipactor
event is likely to occur each time the envelope exceeds the threshold for
a long enough time. If the envelope is amplitude modulated, the dis-
charge events will vary in strength with the AM frequency. This periodic
variation will appear as a peak in the FFT plot of the detected signal at
the same frequency. By applying a weak, 1-5% depth, synchronised AM
to the input signals, the multipactor threshold can be determined with
high accuracy and without risking ambiguous test results. A 1% AM
corresponds to less than 0.1 dB variation in the input signal and will
have no signicant eect on the measured threshold.
The detected signal must then be processed by a computer or a sim-
ilar tool in order to reveal the periodicity, as in the single carrier case.
In Fig. 6.9 an example of a possible test setup is given. In order to
achieve a good AM in the multicarrier case, all the signals should be
modulated using the same reference signal of modulation. Many signal
generators have a signal reference input, thus allowing the user to syn-
chronise several signal generators. To perform a successful multicarrier
experiment, the phases have to be stable in relation to each other and
thus a common reference signal will have to be used in any case. An-
other way of achieving synchronised modulation could be to modulate
the gain adjustment of the high power amplier.
In Fig. 6.9 it is suggested that the third harmonic should be moni-
tored and that is probably the best choice when studying multicarrier
multipactor, since third harmonic detection is fast and sensitive. Close-
to-carrier noise detection is a possible alternative, but it may not be as
sensitive and thus weak multipactor events may be overlooked.
6.2.3 Main achievements
A method of multipactor detection has been devised, which can be used
to obtain accurate and unambiguous measurement results for both single
and multicarrier multipactor. The method does not aim to replace any of
the existing methods of detection, rather it can serve as a complement
to the other methods to improve accuracy and condence in the test
results.
Close-to-carrier noise and third harmonic detection are two fast and
sensitive methods of multipactor detection. Both methods rely on noise
generation, which makes them prone to non-multipactor generated noise.
106
HPA
Vacuum
Chamber
DUT
Spect.
Analys. LNA
Third harmonic detection
Power
Meter
Power
Meter
Reflected power
dB dB
Forward power
Instrument control
&
Data storage and processing

M
U
X
Phase
control
Reference signal generator
dB
Spect.
Analys.
Wave form monitor
Figure 6.9: Test setup for multicarrier multipactor measurements using RF
power modulation. Each input signal is amplitude modulated and
by phase locking the signals using a signal reference, the entire
signal envelope will be modulated. A computer can be used to
control the instruments and collect the output data, which can
then be real-time Fourier transformed to reveal any periodicity
in the detected signal.
107
In a typical test setup there can be many sources of noise, which could
result in ambiguous test results. On the one hand, the multipactor
threshold could be established at a too low value if non-multipactor
generated noise is misinterpreted as the result of a discharge. On the
other hand, a short-lived multipactor event could be disregarded and
lead to determination of a too high threshold based on a more distinct
indication. The AM detection method resolves this concern by only
signalling for true multipactor events.
Another advantage of the AM detection method is the fact that it
is particularly sensitive close to the multipactor threshold, since it only
responds to the signal dierence between the input signal and the thresh-
old, as can be seen from relation 6.2. A weak amplitude modulation,
as shown in Fig. 6.10 where the single carrier signal has just passed the
multipactor threshold, will produce a very distinct modulated output
signal as indicated in Fig. 6.11. Even though the signal is very noisy,
the periodicity is very distinct. Not always will the periodicity be as
noticeable as in Fig. 6.11, but a FFT will reveal any periodicity in the
measured signal.
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
0
20
40
60
80
100
120
Input signal envelope, single carrier, 1%depth AM
A
m
p
l
i
t
u
d
e

[
A
.
U
.
]
Time
Input signal
Multipactor threshold
Figure 6.10: Example of a single carrier input signal envelope with a 1%
depth AM. The signal has barely exceeded the multipactor
threshold.
One of the shortcomings of the AM detection method is that it re-
108
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
0
0.5
1
1.5
2
2.5
Detected signal
A
m
p
l
i
t
u
d
e

[
A
.
U
.
]
Time
Figure 6.11: Qualitative form of the detected multipactor signal when the
input signal to the DUT is as shown in Fig. 6.10. The same type
of signal can be seen in Fig. 6.4, but in this case, the sampling
frequency is 200 times higher than the frequency of modulation,
which explains why the modulated signal is so prominent.
109
quires a certain time for the FFT. One can estimate that a minimum
of 5-10 times the period of the AM is needed for a reliable FFT. If the
frequency of AM is 1 kHz, then the time needed for the measurement
would be 5-10 ms. In most cases this would be acceptable, but if single
events of multipactor are to be detected, the method will not work.
110
Chapter 7
Conclusions and outlook
This thesis has presented basic theory as well as new developments con-
cerning the phenomenon called multipactor. Its deleterious eects on
microwave systems operating in a vacuum environment have been em-
phasised. For satellites, the discharge can be catastrophic as basically no
means of repair or modication is available for parts in orbit. To avoid
the risks associated with vacuum discharges, lots of eort has been put
into studying the phenomenon and today well established engineering
methods exist, which are used by the microwave engineer when designing
hardware bound for space. However, the available methods are based
on the simple parallel-plate model, which in many instances is the worst
case. Thus, when designing important microwave features, in particular
such involving a nonuniform electromagnetic eld, like e.g. irises and
coaxial lines, the parallel-plate model may not be applicable and there
is a risk of unnecessarily conservative designs. Such designs are often
large and heavy, which is a great disadvantage when it comes to devices
to be used in space. As a rst attempt to establish new methods of
assessing the risk for having a discharge in such structures, a large part
of this thesis has been devoted to multipactor in irises and coaxial lines.
It has been shown that by only considering the random walk of the sec-
ondary electrons in an iris, the threshold can be signicantly increased
compared with the pure parallel-plate case. Many interesting aspects of
the phenomenon in a coaxial line have been found, e.g. the dual sta-
ble regions of single-sided resonance and its unexpectedly low threshold
of the rst order mode. In addition, the increased threshold for high
impedance coaxial lines that was found in experiments, was also found
in this study, both in the theoretical analysis and in the PIC-simulations.
111
At altitudes where most satellites operate, the pressure is very low
and for most purposes it can be approximated as a perfect vacuum.
However, during the launch phase of a satellite and during its rst days
of operation as well as at times when the satellite res its attitude and
altitude control engines, the microwave parts may not be completely
vented and thus it is important to understand what happens with the
multipactor threshold with increasing pressure. This has been one of the
main topics of this thesis. It was found that for materials with a low rst
cross-over energy, for the lowest order resonance, the threshold increases
with increasing pressure until reaching a maximum, after which it starts
to decline. The reason for the increase is the friction force experienced
by the electrons when colliding with the neutral gas particles. In all
other cases, for materials with a high rst cross-over energy and in gen-
eral for the higher order modes, a monotonically decreasing threshold
is noted. This behaviour can be explained by the thermalisation of the
electrons, which leads to a higher total impact energy, as well as the
contribution of electrons from collisional ionisation, which reduces the
necessary secondary emission yield and consequently also the required
impact velocity. Improved quantitative results require a more detailed
investigation of the fraction of the electrons from impact ionisation that
actually contribute to the multipactor bunch, but that is left for a future
study. Furthermore, in order to make the model useful for all frequency-
gap size products, an extension of the model to include also the hybrid
modes is necessary. Due to the complexity of such a study, it was not
included in this rst analysis. However, this may be an interesting topic
for a future investigation.
In addition to designing with respect to the proper thresholds, most
hardware will require some type of testing to ensure compliance with
the standard. Such tests must be of good quality to avoid ambiguities,
which could disqualify a component that is multipactor free. In this
thesis a method that gives unambiguous and highly reliable test results
was presented. It is a method based on a weak amplitude modulation of
the input signal, which becomes very distinct in the multipactor signal,
since the generated noise power is a function of the dierence between
input power and the multipactor threshold. By using this auxiliary
method of detection in connection with two other detection methods,
reliable test results can be obtained.
It is quite satisfying when looking back at the chapter on future work
in my licentiate thesis [75] and realizing that the two main topics men-
112
tioned there were multipactor in nonuniform elds and irises, which were
then successfully studied during the second part of my PhD work. How-
ever, there are still interesting projects to consider, e.g. multipactor in
irises where not only the eect of the random walk is considered, but also
the eect of the actual nonuniform eld in the structure. There are many
other important structures in microwave systems, which have not been
studied with respect to multipactor discharge. Among these are crossed
irises, septum polarisers and ridged waveguides. The 20 gap-crossings
rule, which is part of the ESA standard [50], should be re-investigated
both theoretically and experimentally to make sure that the rule has
a sound theoretical base with good agreement between simulations and
experiment. As a continuation of the coaxial study, the axial dimension
could be included by considering not only a travelling wave signal but
also a standing wave. Since electrons aected by the Miller force will
drift towards positions with low eld amplitude, it may not be feasible
to directly apply the peak amplitude of the standing wave in the coaxial
model presented in this thesis.
113
114
References
[1] W. Henneberg, R. Orthuber, and E. Steudel, Z. Tech. Phys. 17,
pp. 115-120 (1936).
[2] E. W. B. Gill and A. von Engel, Proc. R. Soc. London Ser. A 192,
pp. 446-463 (1948).
[3] G. Francis and A. von Engel, Proc. R. Soc. London Ser. A 246,
pp. 143-180 (1953).
[4] W. J. Gallagher, IEEE Trans. Nucl. Sci. 26, pp. 4280-4282 (1979).
[5] J. R. M. Vaughan, IEEE Trans. Electron Devices 35, pp. 1172-1180
(1988).
[6] N. F. Kovalev, V. E. Nechaev, M. I Petelin, and N. I. Zaitsev, IEEE
Trans. Plasma Sci. 26, pp. 246-51 (1998).
[7] C. Bourata,d J.-M Joly, IEEE Trans. Electr. Insul. 24, pp. 1045-
1048 (1989).
[8] A.D. Woode and J. Petit, Microwave Journal, pp. 142-155 (January
1992).
[9] A. D. MacDonald, Microwave Breakdown in Gases, (Wiley, New
York, 1966).
[10] A. V Gaponov and M. A. Miller, Sov. Phys. JETP 7, pp. 168-
(1958).
[11] M. P. Forrer and C. Milazzo, Duplexing and switching with multi-
pactor discharges, Proceedings of the Institute of Radio Engineers
50, pp. 442-450 (1962).
[12] W. J. Gallagher, Proceedings of the IEEE 57, pp.94-95 (1969).
115
[13] D. L. Liska, Proceedings of the IEEE 59, pp. 1253-1254 (1971).
[14] A. J. Marrison, R. May, J. D. Sanders, A. D. Dyne, A. D. Rawl-
ins and J. Petit, A Study of Multipaction in Multicarrier
RF Components, AEA Technology for ESTEC, AEA Ref. No.
AEA/TYKB/31761/01/RP/05 Issue 1, (Culham, UK, 1997)
[15] N. Rozario, H. F. Lenzig, K. F. Reardon, M. S. Zarro, and
C. G. Baran, IEEE Trans. Microwave Theory Tech. 42, pp. 558-
564 (1994).
[16] F. Kossel and K. Krebs, Zeitschrift f ur Physik 175, pp. 382-390
(1963).
[17] D. Vender, H. B. Smith, and R. W. Boswell, J. Appl. Phys. 80,
pp. 4292-4298 (1996).
[18] F. Hohn, W. Jacob, R. Beckmann, and R. Wilhelm, Phys. Plasmas
4, pp. 940-944 (1997).
[19] P. Fortescue and J. Stark (Editors), Spacecraft Systems Engineer-
ing, (Wiley, West Sussex, England, 1995)
[20] S. Riyopoulos, D. Chernin, and D. Dialetis, Phys. Plasmas 2,
pp. 3194-3213 (1995).
[21] V. Semenov, A. Kryazhev, D. Anderson, and M. Lisak, Phys. Plas-
mas 8, pp. 5034-5039 (2001).
[22] J. R. M. Vaughan, IEEE Trans. Electron Devices 36, pp. 1963-1967
(1989).
[23] Z. J. Ding, X. D. Tang, and R. Shimizu, J. Appl. Phys. 89, pp. 718-
726 (2001).
[24] V. E. Henrich, Rev. Sci. Instr. 44, pp. 456-462 (1973).
[25] N. L. S. Martin and A. von Engel, J. Phys. D: Appl. Phys., 10
pp. 863-868 (1977).
[26] D. Wolk et al, Coatings on Mg Alloys for reduction of Multipactor
Eects in RF Components, Proceedings from the 5
th
International
Workshop on Multipactor, Corona, and Passive Intermodulation
in Space RF Hardware, 12-14 September, 2005, Noordwijk, the
Netherlands.
116
[27] A. Ruiz et al, Ti, V, and Cr Nitride Coatings for Reduction of
Multipactor Eect in RF Components, Proceedings from the 5
th
International Workshop on Multipactor, Corona, and Passive In-
termodulation in Space RF Hardware, 12-14 September, 2005, No-
ordwijk, the Netherlands.
[28] R. Gehring et al, Study of Black-Anodized Coating Inuence
in Multipactor Performance of RF Space Hardware, Proceedings
from the 5
th
International Workshop on Multipactor, Corona, and
Passive Intermodulation in Space RF Hardware, 12-14 September,
2005, Noordwijk, the Netherlands.
[29] J. Sombrin, Eet Multipactor, CNES Technical Report No.
83/DRT/TIT/HY/119/T, (CNES, Toulouse, 1983).
[30] A. J. Hatch and H. B. Williams, The Physical Review, Second Series
112, pp. 681-685 (1958).
[31] A. Woode and J. Petit, Diagnostic Investigations into the Mul-
tipactor Eect, Susceptibility Zone Measurements and Parameters
Aecting a Discharge, Estec working paper No. 1556, (ESTEC, No-
ordwijk, 1989).
[32] A. J. Hatch, J. Appl. Phys. 32, pp. 1086-1092 (1961).
[33] S. Riyopoulos, D. Chernin, and D. Dialetis, IEEE Trans. Electron
Devices 44, pp. 489-497 (1997).
[34] R. Udiljak, D. Anderson, M. Lisak, V. E. Semenov, and J. Puech,
Phys. Plasmas 11, pp. 5022-5031 (2004).
[35] R. A. Kishek, Y. Y. Lau, and D. Chernin, Phys. Plasmas 4, pp. 863-
872 (1997).
[36] R. A. Kishek, Y. Y. Lau, L. K. Ang, A. Valfells, and R. M. Gilgen-
bach, Phys. Plasmas 5, pp. 2120-2126 (1998).
[37] A. L. Gilardini, J. Appl. Phys. 71, pp. 4629-4631 (1992).
[38] A. L. Gilardini, J. Appl. Phys. 78, pp. 783-795 (1995).
[39] A. Kryazhev, M. Buyanova, V. Semenov, D. Anderson, M. Lisak,
J. Puech, L. Lapierre, and J. Sombrin, Phys. Plasmas 9, pp. 4736-
4743 (2002).
117
[40] M. Merecki, Rapport de stage. Etude de leet multipactor dans
les dispositifs hyperfrequences, Masters Thesis, (CNES, Toulouse,
2003)
[41] V. G. Andreev and D. G. Zaidin, Translated from Pribory i Tehnika

Eksperimenta 3, pp. 164-165 (1971)


[42] R. Udiljak, G. Li, D. Anderson, P. Ingvarson, U. Jordan, U. Jostell,
A. Kryazhev, M. Lisak, V. E. Semenov, Suppression of Multipactor
Breakdown in RF Equipment, Conference Proceedings RVK 02,
June 10-12, 2002, Stockholm, Sweden.
[43] D. Ruzic, R. Moore, D. Manos, and S. Cohen, J. Vac. Sci. Tech.
20, pp. 1313-1216 (1982).
[44] N. Diaz, S. Castaneda, J. M. Riplada, I. Montero, D. Raboso,
I. Galan, and F. Rueda, Low secondary electron emission thin
lms to prevent the multipactor eect in high-power RF devices
in space, Proceedings from the workshop on Multipactor, RF and
DC Corona, and Passive Intermodulation in Space RF Hardware,
4-6 September, 2000, Noordwijk, the Netherlands.
[45] R. L. Geng and H. S. Padamsee, Exploring Multipacting Char-
acterstics of a Rectangular Waveguide, Proceedings of the 1999
Particle Accelerator Conference, 27 March - 2 April, 1999, New
York, NY, USA.
[46] A. Sazontov, M. Buyanova, V. Semenov, E. Rakova, N. Vdovicheva,
D. Anderson, M. Lisak, J. Puech, and L. Lapierre, Phys. Plasmas
12, pp. 053102:1-8 (2005).
[47] R. Udiljak, D. Anderson, M. Lisak, V. E. Semenov, and J. Puech,
Phys. Plasmas 10, pp. 4105-4111, (2003).
[48] A. J. Hatch and H. B. Williams, J. Appl. Phys. 25, pp. 417-423
(1954).
[49] Website: http://www.nsab-sirius.com/ visited on April 7. 2004.
[50] ESA for ECSS, Space Engineering, Multipaction design and test,
ECSS-E-20-01A, (ESA Publications Division, Noordwijk, 2003)
118
[51] D. Wolk, D. Schmitt, and T. Schlipf, A novel approach for calculat-
ing the multipaction threshold level in multicarrier operation, Pro-
ceedings from the workshop on Multipactor, RF and DC Corona,
and Passive Intermodulation in Space RF Hardware, 4-6 Septem-
ber, 2000, Noordwijk, the Netherlands.
[52] G. Li and R. Udiljak, Threshold Level Determination of Multicarrier
Multipaction and AM Suppression of Multipaction in Resonant Cav-
ity, Masters Thesis (Chalmers University of Technology, Gothen-
burg, 2001).
[53] A. L. Gilardini, Il Nuovo Cimento 18, pp. 919-929 (1996).
[54] A. L. Gilardini, J. Phys. D: Appl. Phys. 32, pp. 1281-1286 (1999).
[55] H. Shimamori and T. Sunagawa, J. Chem. Phys. 106, pp. 4481-4490
(1997).
[56] Y. P. Raizer, Gas Discharge Physics, (Springer-Verlag, Berlin Hei-
delberg, 1991).
[57] S. J. Buckman and B. Lohmann, J. Phys. B: At. Mol. Phys. 19,
pp. 2547-2564 (1986).
[58] S. C. Brown, Basic Data of Plasma Physics, (The Riverside Press,
USA, 1967).
[59] H. C. Straub, P. Renault, B. G. Lindsay, K. A. Smith, and
R. F. Stebbings, Phys. Rev. A 52, pp. 1115-1124 (1995).
[60] The study of multipactor breakdown in space electronic systems,
Tech. Rep. CR-448, (NASA, Washington D.C., 1966).
[61] D. Wolk, C. Vicente, H. L. Hartnagel, M. Mattes, J. R. Mosig,
and D. Raboso, An investigation of the eect of fringing elds
on multipactor breakdown,Proceedings from the 5
th
International
Workshop on Multipactor, Corona, and Passive Intermodulation
in Space RF Hardware, 12-14 September, 2005, Noordwijk, the
Netherlands.
[62] A. J. Marrison, Final report on the study of multipaction in multi-
carrier systems, Tech. Rep. AEA/TYKB/31868/RP/2, (Culham
Laboratory, Abingdon, England, 1994).
119
[63] R. Woo, J. Appl. Phys. 39, pp. 1528-1533 (1968).
[64] K. Sakamoto, Y. Ikeda, and T. Imai, J. Phys. D: Appl. Phys. 22,
pp. 1840-1847 (1989).
[65] E. Somersalo, P. Yla-Oijala, and D. Proch, Analysis of multipact-
ing in coaxial lines, Proceedings of 1995 Particle Accelerator Con-
ference, 1996, Dallas, Texas.
[66] E. Somersalo, P. Yla-Oijala, D. Proch, and J. Sarvas, Particle Ac-
celerators 59, pp. 107-141 (1998).
[67] E. Chojnacki, Phys. Rev. Special Topics - Accelerators and Beams
3, pp. 032001:1-5 (2000).
[68] S. K. Nagesh, D. Revannasiddiah, and S. V. K. Shastry, Pramana
J. Phys. 64, pp. 95-110 (2005).
[69] V. Semenov, V. Nechaev, E. Rakova, N. Zharova, D. Anderson,
M. Lisak, J. Puech, Phys. Plasmas 12, pp. 073508:1-6 (2005).
[70] R. Kishek, Y. Y. Lau and R. Gilgenbach, Temporal Evolution of
Multipactor Discharge, Proceedings of the 1995 Particle Acceler-
ator Conference, 1-5 May, 1995, Dallas, Texas, USA.
[71] D. Raboso and A. Alsta, A new RF breakdown detection method
based on electron monitoring, Proceedings of the 4
th
International
Workshop on Multipactor, Corona and Passive Intermodulation in
Space RF Hardware, 8-11 September, 2003, ESTEC, Noordwijk,
The Netherlands.
[72] A. Neuber, J. Dickens, D. Hemmert, H. Krompholz, L. L. Hat-
eld and M. Kristiansen, IEEE Trans. Plasma Sci. 26, pp. 296-303,
(1998).
[73] T. Fujii and S. Moriyama, IEEE Trans. Plasma Sci. 29, pp. 318-325,
(2001).
[74] R. Udiljak, D. Anderson, U. Jostell, M. Lisak, J. Puech, and
V. E. Semenov, Detection of Multicarrier Multipaction using RF
Power Modulation, Proceedings of the 4
th
International Workshop
on Multipactor, Corona and Passive Intermodulation in Space RF
Hardware, 8-11 September, 2003, ESTEC, Noordwijk, The Nether-
lands.
120
[75] R. Udiljak, Multipactor in low pressure gas, Licentiate Thesis
(Chalmers University of Technology, Gothenburg, 2004).
121
122
Included papers AF
123
Paper A
R. Udiljak, D. Anderson, P. Ingvarson, U. Jordan, U. Jostell, G. Li,
M. Lisak, L. Lapierre, J. Puech, and J. Sombrin, New Method for
Detection of Multipaction, IEEE Trans. Plasma Sci., Vol. 31, No. 3,
pp. 396-404 , June 2003.
Paper B
R. Udiljak, D. Anderson, M. Lisak, V. E. Semenov, and J. Puech, Mul-
tipactor in low pressure gas, Phys. Plasmas, Vol. 10, No. 10, pp. 4105-
4111, Oct. 2003.
Paper C
R. Udiljak, D. Anderson, M. Lisak, V. E. Semenov, and J. Puech, Im-
proved model for multipactor in low pressure gas, Phys. Plasmas,
Vol. 11, No. 11, pp. 5022-5031, Nov. 2004.
Paper D
R. Udiljak, D. Anderson, M. Lisak, J. Puech, and V. E. Semenov, Mul-
tipactor in a waveguide iris, accepted for publication in IEEE Trans.
Plasma Sci.
Paper E
R. Udiljak, D. Anderson, M. Lisak, V. E. Semenov, and J. Puech, Mul-
tipactor in a coaxial transmission line, part I: analytical study, accepted
for publication in Phys. Plasmas
Paper F
V. E. Semenov, N. Zharova, R. Udiljak, D. Anderson, M. Lisak, and
J. Puech, Multipactor in a coaxial transmission line, part II: Particle-
in-Cell simulations, accepted for publication in Phys. Plasmas

Das könnte Ihnen auch gefallen