12.12.2012 Views

Multipactor in Low Pressure Gas and in ... - of Richard Udiljak

Multipactor in Low Pressure Gas and in ... - of Richard Udiljak

Multipactor in Low Pressure Gas and in ... - of Richard Udiljak

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Thesis for the degree <strong>of</strong> Doctor <strong>of</strong> Philosophy<br />

<strong>Multipactor</strong> <strong>in</strong> <strong>Low</strong> <strong>Pressure</strong> <strong>Gas</strong><br />

<strong>and</strong> <strong>in</strong> Nonuniform RF Field<br />

Structures<br />

<strong>Richard</strong> <strong>Udiljak</strong><br />

Department <strong>of</strong> Radio <strong>and</strong> Space Science<br />

Chalmers University <strong>of</strong> Technology<br />

Göteborg, Sweden, 2007


<strong>Multipactor</strong> <strong>in</strong> <strong>Low</strong> <strong>Pressure</strong> <strong>Gas</strong><br />

<strong>and</strong> <strong>in</strong> Nonuniform RF Field Structures<br />

<strong>Richard</strong> <strong>Udiljak</strong><br />

c○<strong>Richard</strong> <strong>Udiljak</strong>, 2007<br />

ISBN 978-91-7291-885-6<br />

Doktorsavh<strong>and</strong>l<strong>in</strong>gar vid Chalmers tekniska högskola<br />

Ny serie nr 2566<br />

ISSN 0346-718X<br />

Department <strong>of</strong> Radio <strong>and</strong> Space Science<br />

Chalmers University <strong>of</strong> Technology<br />

SE–412 96 Göteborg<br />

Sweden<br />

Telephone +46–(0)31–772 10 00<br />

Cover: Susceptibility chart for multipactor <strong>in</strong> a waveguide iris for five<br />

different height/length-ratios.<br />

Pr<strong>in</strong>ted <strong>in</strong> Sweden by<br />

Reproservice<br />

Chalmers Tekniska Högskola<br />

Göteborg, Sweden, 2007


<strong>Multipactor</strong> <strong>in</strong> <strong>Low</strong> <strong>Pressure</strong> <strong>Gas</strong><br />

<strong>and</strong> <strong>in</strong> Nonuniform RF Field Structures<br />

RICHARD UDILJAK<br />

Department <strong>of</strong> Radio <strong>and</strong> Space Science<br />

Chalmers University <strong>of</strong> Technology<br />

Abstract<br />

Resonant electron multiplication <strong>in</strong> vacuum, multipactor, is analysed<br />

for several geometries where the RF electric field is nonuniform. In particular,<br />

it is shown that the multipactor behaviour <strong>in</strong> a coaxial l<strong>in</strong>e is<br />

both qualitatively <strong>and</strong> quantitatively different from that observed with<br />

the conventionally used simple parallel-plate model. Analytical estimates<br />

based on an approximate solution <strong>of</strong> the non-l<strong>in</strong>ear differential<br />

equation <strong>of</strong> motion for the multipact<strong>in</strong>g electrons are supported by extensive<br />

particle-<strong>in</strong>-cell simulations. Furthermore, <strong>in</strong> a microwave iris the<br />

electrons tend to perform a r<strong>and</strong>om walk <strong>in</strong> the axial direction <strong>of</strong> the<br />

waveguide due to the <strong>in</strong>itial velocity distribution. The effects <strong>of</strong> this phenomenon<br />

on the breakdown threshold are analysed. The study shows<br />

that the threshold is a function <strong>of</strong> the height-to-length ratio <strong>of</strong> the iris<br />

<strong>and</strong> for a fixed value <strong>of</strong> this ratio, the multipactor susceptibility charts<br />

can be generated <strong>in</strong> the classical eng<strong>in</strong>eer<strong>in</strong>g units. Us<strong>in</strong>g the parallelplate<br />

concept, the multipactor threshold <strong>in</strong> low pressure gases has been<br />

analysed us<strong>in</strong>g a model for the electron motion that takes <strong>in</strong>to account<br />

three important effects <strong>of</strong> electron-neutral collisions, viz. the friction<br />

force, electron thermalisation, <strong>and</strong> impact ionisation. It is found that<br />

all three effects play important roles, but the degree <strong>of</strong> <strong>in</strong>fluence depends<br />

on parameters such as order <strong>of</strong> resonance <strong>and</strong> secondary emission<br />

properties. In addition, a new method for detection <strong>of</strong> multipactor is<br />

presented. By apply<strong>in</strong>g a weak amplitude modulation to the <strong>in</strong>put signal<br />

<strong>and</strong> perform<strong>in</strong>g a fast Fourier transform on the detected signal, accurate<br />

<strong>and</strong> unambiguous measurement results can be obta<strong>in</strong>ed. It is demonstrated<br />

how the method can be used <strong>in</strong> both s<strong>in</strong>gle <strong>and</strong> multicarrier<br />

operation.<br />

Keywords: <strong>Multipactor</strong>, discharge, breakdown, microwave discharge,<br />

nonuniform fields, coax, iris, low pressure gas, detection methods.<br />

iii


Publications<br />

This thesis is based on the work conta<strong>in</strong>ed <strong>in</strong> the follow<strong>in</strong>g papers:<br />

[A] R. <strong>Udiljak</strong>, D. Anderson, P. Ingvarson, U. Jordan, U. Jostell, G. Li,<br />

M. Lisak, L. Lapierre, J. Puech, <strong>and</strong> J. Sombr<strong>in</strong>, “New Method<br />

for Detection <strong>of</strong> Multipaction”, IEEE Trans. Plasma Sci., Vol. 31,<br />

No. 3, pp. 396-404 , June 2003.<br />

[B] R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, V. E. Semenov, <strong>and</strong> J. Puech,<br />

“<strong>Multipactor</strong> <strong>in</strong> low pressure gas”, Phys. Plasmas, Vol. 10, No. 10,<br />

pp. 4105-4111, Oct. 2003.<br />

[C] R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, V. E. Semenov, <strong>and</strong> J. Puech,<br />

“Improved model for multipactor <strong>in</strong> low pressure gas”, Phys. Plasmas,<br />

Vol. 11, No. 11, pp. 5022-5031, Nov. 2004.<br />

[D] R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, J. Puech, <strong>and</strong> V. E. Semenov,<br />

“<strong>Multipactor</strong> <strong>in</strong> a waveguide iris”, accepted for publication <strong>in</strong> IEEE<br />

Trans. Plasma Sci.<br />

[E] R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, V. E. Semenov, <strong>and</strong> J. Puech,<br />

“<strong>Multipactor</strong> <strong>in</strong> a coaxial transmission l<strong>in</strong>e, part I: analytical study”,<br />

accepted for publication <strong>in</strong> Phys. Plasmas<br />

[F] V. E. Semenov, N. Zharova, R. <strong>Udiljak</strong>, D. Anderson, M. Lisak,<br />

<strong>and</strong> J. Puech, “<strong>Multipactor</strong> <strong>in</strong> a coaxial transmission l<strong>in</strong>e, part<br />

II: Particle-<strong>in</strong>-Cell simulations”, accepted for publication <strong>in</strong> Phys.<br />

Plasmas<br />

v


Conference contributions by the author (not <strong>in</strong>cluded <strong>in</strong> this thesis):<br />

vi<br />

[G] R. <strong>Udiljak</strong>, G. Li, D. Anderson, P. Ingvarson, U. Jordan, U. Jostell,<br />

A. Kryazhev, M. Lisak, V. E. Semenov, “Suppression <strong>of</strong> <strong>Multipactor</strong><br />

Breakdown <strong>in</strong> RF Equipment”, RVK 02, June 10-12, 2002,<br />

Stockholm, Sweden.<br />

[H] R. <strong>Udiljak</strong>, D. Anderson, U. Jostell, M. Lisak, J. Puech, V. E. Semenov,<br />

“Detection <strong>of</strong> Multicarrier Multipaction us<strong>in</strong>g RF Power<br />

Modulation”, 4th International Workshop on <strong>Multipactor</strong>, Corona<br />

<strong>and</strong> Passive Intermodulation <strong>in</strong> Space RF Hardware, 8-11 September,<br />

2003, ESTEC, Noordwijk, The Netherl<strong>and</strong>s.<br />

[I] J. Puech, L. Lapierre, J. Sombr<strong>in</strong>, V. Semenov, A. Sazontov,<br />

N. Vdovicheva, M. Buyanova, U. Jordan, R. <strong>Udiljak</strong>, D. Anderson,<br />

M. Lisak, “<strong>Multipactor</strong> threshold <strong>in</strong> waveguides: theory <strong>and</strong> experiment”,<br />

NATO Advanced Research Workshop on Quasi-Optical<br />

Control <strong>of</strong> Intense Microwave Transmission , 17-20 February, 2004,<br />

Nizhny-Novgorod, Russian Federation.<br />

[J] R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, J. Puech, V. E. Semenov, “Microwave<br />

breakdown <strong>in</strong> the transition region between multipactor<br />

<strong>and</strong> corona discharge.”, RVK 05, June 14 - 16 juni, L<strong>in</strong>köp<strong>in</strong>g<br />

[K] D. Anderson, M. Buyanova, D. Dorozhk<strong>in</strong>a, U. Jordan, M. Lisak,<br />

I. Nefedov, T. Olsson, J. Puech, V. Semenov, I. Shereshevskii,<br />

R. Tomala, <strong>and</strong> R. <strong>Udiljak</strong>, “Microwave breakdown <strong>in</strong> RF equipment.”,<br />

RVK 05, June 14 - 16 juni, L<strong>in</strong>köp<strong>in</strong>g<br />

[L] V. E. Semenov, N. Zharova, R. <strong>Udiljak</strong>, D. Anderson, M. Lisak,<br />

J. Puech, <strong>and</strong> L. Lapierre, “<strong>Multipactor</strong> <strong>in</strong>side a coaxial l<strong>in</strong>e”,<br />

5th International Workshop on <strong>Multipactor</strong>, Corona <strong>and</strong> Passive<br />

Intermodulation <strong>in</strong> Space RF Hardware, 12-14 September, 2005,<br />

ESTEC, Noordwijk, The Netherl<strong>and</strong>s.<br />

[M] R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, V. E. Semenov, <strong>and</strong> J. Puech,<br />

“Microwave breakdown <strong>in</strong> low pressure gas”, 5th International<br />

Workshop on <strong>Multipactor</strong>, Corona <strong>and</strong> Passive Intermodulation<br />

<strong>in</strong> Space RF Hardware, 12-14 September, 2005, ESTEC, Noordwijk,<br />

The Netherl<strong>and</strong>s.<br />

[N] C. Armiens, B. Huang, R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, U. Jostell,<br />

<strong>and</strong> P. Ingvarsson, “Detection <strong>of</strong> Multipaction us<strong>in</strong>g AM signals”,


5th International Workshop on <strong>Multipactor</strong>, Corona <strong>and</strong> Passive<br />

Intermodulation <strong>in</strong> Space RF Hardware, 12-14 September, 2005,<br />

ESTEC, Noordwijk, The Netherl<strong>and</strong>s.<br />

vii


viii


Contents<br />

Publications v<br />

Acknowledgement xi<br />

Acronyms xiii<br />

1 Introduction 1<br />

2 <strong>Multipactor</strong> <strong>in</strong> vacuum 5<br />

2.1 S<strong>in</strong>gle Carrier . . . . . . . . . . . . . . . . . . . . . . . . . 6<br />

2.1.1 Basic theory . . . . . . . . . . . . . . . . . . . . . 6<br />

2.1.2 Hybrid modes . . . . . . . . . . . . . . . . . . . . . 17<br />

2.1.3 Factors effect<strong>in</strong>g the threshold . . . . . . . . . . . 18<br />

2.1.4 Methods <strong>of</strong> suppression . . . . . . . . . . . . . . . 20<br />

2.1.5 Effect <strong>of</strong> r<strong>and</strong>om emission delays <strong>and</strong> <strong>in</strong>itial velocity<br />

spread . . . . . . . . . . . . . . . . . . . . . . . 23<br />

2.2 Multicarrier . . . . . . . . . . . . . . . . . . . . . . . . . . 25<br />

2.3 Design guidel<strong>in</strong>es . . . . . . . . . . . . . . . . . . . . . . . 28<br />

2.3.1 S<strong>in</strong>gle carrier . . . . . . . . . . . . . . . . . . . . . 29<br />

2.3.2 Multicarrier . . . . . . . . . . . . . . . . . . . . . . 29<br />

3 <strong>Multipactor</strong> <strong>in</strong> low pressure gas 35<br />

3.1 Simple Model . . . . . . . . . . . . . . . . . . . . . . . . . 36<br />

3.1.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . 36<br />

3.1.2 <strong>Multipactor</strong> boundaries . . . . . . . . . . . . . . . 37<br />

3.1.3 Ma<strong>in</strong> results . . . . . . . . . . . . . . . . . . . . . . 42<br />

3.2 Advanced Model . . . . . . . . . . . . . . . . . . . . . . . 44<br />

3.2.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . 44<br />

3.2.2 Analytical formulas for argon cross-sections . . . . 47<br />

3.2.3 <strong>Multipactor</strong> boundaries . . . . . . . . . . . . . . . 49<br />

ix


3.2.4 Key f<strong>in</strong>d<strong>in</strong>gs . . . . . . . . . . . . . . . . . . . . . 52<br />

4 <strong>Multipactor</strong> <strong>in</strong> irises 59<br />

4.1 Model <strong>and</strong> approximations . . . . . . . . . . . . . . . . . . 60<br />

4.2 <strong>Multipactor</strong> regions . . . . . . . . . . . . . . . . . . . . . . 64<br />

4.3 Comparison with experiments . . . . . . . . . . . . . . . . 65<br />

4.4 Ma<strong>in</strong> results . . . . . . . . . . . . . . . . . . . . . . . . . . 67<br />

5 <strong>Multipactor</strong> <strong>in</strong> coaxial l<strong>in</strong>es 69<br />

5.1 Analytical study . . . . . . . . . . . . . . . . . . . . . . . 70<br />

5.1.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . 70<br />

5.1.2 <strong>Multipactor</strong> resonance theory . . . . . . . . . . . . 73<br />

5.1.3 Ma<strong>in</strong> f<strong>in</strong>d<strong>in</strong>gs . . . . . . . . . . . . . . . . . . . . . 81<br />

5.2 Particle-<strong>in</strong>-cell simulations . . . . . . . . . . . . . . . . . . 82<br />

5.2.1 Numerical implementation . . . . . . . . . . . . . . 82<br />

5.2.2 Simulations . . . . . . . . . . . . . . . . . . . . . . 83<br />

5.2.3 Comparison with experiments . . . . . . . . . . . . 88<br />

5.2.4 Ma<strong>in</strong> conclusions . . . . . . . . . . . . . . . . . . . 91<br />

6 Detection <strong>of</strong> multipactor 93<br />

6.1 Common Methods <strong>of</strong> Detection . . . . . . . . . . . . . . . 93<br />

6.1.1 Global methods . . . . . . . . . . . . . . . . . . . . 94<br />

6.1.2 Local methods . . . . . . . . . . . . . . . . . . . . 99<br />

6.2 Detection us<strong>in</strong>g RF Power Modulation . . . . . . . . . . . 100<br />

6.2.1 S<strong>in</strong>gle carrier . . . . . . . . . . . . . . . . . . . . . 105<br />

6.2.2 Multicarrier . . . . . . . . . . . . . . . . . . . . . . 105<br />

6.2.3 Ma<strong>in</strong> achievements . . . . . . . . . . . . . . . . . . 106<br />

7 Conclusions <strong>and</strong> outlook 111<br />

References 115<br />

Included papers A–F 123<br />

x


Acknowledgement<br />

I wish to thank Pr<strong>of</strong>. Dan Anderson <strong>and</strong> Pr<strong>of</strong>. Mietek Lisak for accept<strong>in</strong>g<br />

me as a PhD student <strong>and</strong> for guidance <strong>and</strong> support <strong>in</strong> my daily<br />

work. I also want to thank Pr<strong>of</strong>. Vladimir Semenov at the Institute<br />

<strong>of</strong> Applied Physics <strong>in</strong> Nizhny Novgorod, Russia, for fruitful discussions<br />

<strong>and</strong> for his patience with all my questions. Thank you Pr<strong>of</strong>. Lars Eliasson,<br />

Director at the Institute <strong>of</strong> Space Research <strong>in</strong> Kiruna, for provid<strong>in</strong>g<br />

both f<strong>in</strong>ancial <strong>and</strong> moral support mak<strong>in</strong>g my PhD c<strong>and</strong>idate appo<strong>in</strong>tment<br />

possible. A very warm thank you also to Jerome Puech for many<br />

<strong>in</strong>terest<strong>in</strong>g discussions about space related microwave problems <strong>and</strong> to<br />

his employer, Centre National d’ Études Spatiales, for f<strong>in</strong>ancial support.<br />

Thanks to my friends <strong>in</strong> Toulouse <strong>and</strong> especially Dr. Omar Houbloss <strong>and</strong><br />

Raquel Rodriguez. I thank my fellow members <strong>of</strong> the National Graduate<br />

School <strong>of</strong> Space Technology <strong>and</strong> my colleagues at Chalmers <strong>and</strong> especially<br />

Dr. Pontus Johannisson, Dr. Ulf Jordan <strong>and</strong> Dr. Lukasz Wolf for<br />

beneficial discussions <strong>and</strong> lots <strong>of</strong> support with L<strong>in</strong>ux <strong>and</strong> LaTex. Many<br />

thanks also to our secretary Monica Hansen for guid<strong>in</strong>g me through the<br />

adm<strong>in</strong>istrative jungle. I want to thank my dear mother, Monica, for<br />

encourag<strong>in</strong>g <strong>and</strong> support<strong>in</strong>g me <strong>and</strong> my family when 24 hours a day<br />

wasn’t enough <strong>and</strong> my father, brother <strong>and</strong> sisters for believ<strong>in</strong>g <strong>in</strong> me. I<br />

am also grateful to my un<strong>of</strong>ficial mentors: my father-<strong>in</strong>-law Lars-Göran<br />

Östl<strong>in</strong>g, my friends Jörgen Otbäck <strong>and</strong> Anders Wilhelmsson, <strong>and</strong> my<br />

brother-<strong>in</strong>-law Nicklas Östl<strong>in</strong>g. Most <strong>of</strong> all I thank my wife Mal<strong>in</strong> <strong>and</strong><br />

our daughters Jan<strong>in</strong>a <strong>and</strong> Lizette for encouragement <strong>and</strong> support dur<strong>in</strong>g<br />

this time.<br />

xi


xii


Acronyms<br />

AM amplitude modulation<br />

DC direct current<br />

DUT device under test<br />

EDDM electron density detection method<br />

ESA european space agency<br />

FFT fast fourier transform<br />

NLSQ non-l<strong>in</strong>ear least square<br />

PIC particle-<strong>in</strong>-cell<br />

PSK phase-shift key<strong>in</strong>g<br />

QPSK quadrature phase-shift key<strong>in</strong>g<br />

RF radio frequency<br />

SEY secondary electron yield<br />

SMA sub m<strong>in</strong>iature version a<br />

TGR twenty gap cross<strong>in</strong>gs rule<br />

TEM transverse electric <strong>and</strong> magnetic field<br />

TWTA travell<strong>in</strong>g wave tube amplifier<br />

UV ultra violet<br />

VSWR voltage st<strong>and</strong><strong>in</strong>g wave ratio<br />

WCAT worst case assessment tool<br />

xiii


xiv


Chapter 1<br />

Introduction<br />

Resonant secondary electron emission RF discharge or multipactor was<br />

discovered <strong>and</strong> studied by Philo Taylor Farnsworth <strong>in</strong> the early 1930’s.<br />

The phenomenon was then used as a means to amplify high frequency<br />

signals as well as to serve as a high frequency oscillator. Us<strong>in</strong>g his multipactor<br />

tubes, Farnsworth succeeded <strong>in</strong> develop<strong>in</strong>g the first electronic<br />

television system. The success stimulated other researchers to <strong>in</strong>vestigate<br />

the phenomenon <strong>and</strong> one <strong>of</strong> the first detailed analyses were done<br />

by Henneberg et al. [1] <strong>in</strong> the mid 1930’s. Gill <strong>and</strong> von Engel [2] made<br />

an even more detailed study, both theoretical <strong>and</strong> experimental, where<br />

they, among other th<strong>in</strong>gs, showed the importance <strong>of</strong> the secondary electron<br />

yield on the development <strong>of</strong> the vacuum discharge. In a follow up<br />

paper [3] Francis <strong>and</strong> von Engel studied not only the <strong>in</strong>itial stage <strong>of</strong> the<br />

electron multiplication, but also the saturation stage. They showed that<br />

the electron space charge effect could be one <strong>of</strong> the major causes for the<br />

discont<strong>in</strong>ued electron growth. Other researchers cont<strong>in</strong>ued the work <strong>and</strong><br />

a basic overview <strong>of</strong> these results can be found <strong>in</strong> the two review papers<br />

by Gallagher [4] <strong>and</strong> Vaughan [5].<br />

Dur<strong>in</strong>g the past 20-30 years, multipactor has ma<strong>in</strong>ly been studied<br />

due to the adverse effects it can have on microwave systems operat<strong>in</strong>g<br />

<strong>in</strong> a vacuum environment. It can disturb the operation <strong>of</strong> high power<br />

microwave generators [6] <strong>and</strong> electron accelerators [7], but, above all, it<br />

can cause severe system degradation <strong>and</strong> failure <strong>of</strong> satellites, which are<br />

difficult or impossible to repair after launch [8]. Satellites operate under<br />

vacuum conditions <strong>and</strong> the most common means <strong>of</strong> communication with<br />

the Earth is microwave transmission. Microwave frequencies are required<br />

as the ionosphere is not transparent for low frequency radiowaves. In<br />

1


addition, it is difficult to make compact, light weight, <strong>and</strong> high ga<strong>in</strong><br />

antennas for low frequency transmission. Many microwave components<br />

are hollow metallic structures that guide the electromagnetic power. A<br />

free electron <strong>in</strong>side the device will experience a force due to the electric<br />

field <strong>and</strong> s<strong>in</strong>ce there is no gas or other material stopp<strong>in</strong>g the electron,<br />

it can accelerate to a very high velocity. Upon impact with one <strong>of</strong><br />

the device walls, the energetic electron can knock out other electrons<br />

<strong>and</strong> under certa<strong>in</strong> circumstances this procedure is repeated cont<strong>in</strong>uously<br />

until the electron density is large enough to counter-act the effect <strong>of</strong> the<br />

applied electric field <strong>and</strong> a steady state is achieved. A consequence <strong>of</strong><br />

this can be that the <strong>in</strong>cident power is reflected <strong>in</strong>stead <strong>of</strong> transmitted<br />

to the <strong>in</strong>tended load. S<strong>in</strong>ce many satellites lack sufficient protection<br />

aga<strong>in</strong>st reflected power <strong>in</strong> order to save weight, such reflected power can<br />

cause severe damage to the high power stage <strong>of</strong> the system.<br />

When a satellite is launched it carries fully charged batteries <strong>and</strong><br />

<strong>in</strong> the beg<strong>in</strong>n<strong>in</strong>g, before the solar panels are deployed, they are the<br />

only source <strong>of</strong> electric power. The capacity <strong>of</strong> these batteries is usually<br />

low <strong>and</strong> may only last a couple <strong>of</strong> days, s<strong>in</strong>ce a satellite <strong>in</strong> operation<br />

will normally only lack access to power from the solar panels for a few<br />

hours, at most, <strong>and</strong> consequently the batteries are made small <strong>in</strong> order to<br />

save weight. If the solar panels are not deployed before the batteries are<br />

exhausted, the satellite is permanently lost. Thus, a new satellite is <strong>of</strong>ten<br />

taken <strong>in</strong>to operation quickly after be<strong>in</strong>g put <strong>in</strong>to orbit. A concern then<br />

is that the satellite components may not be completely vented <strong>and</strong> there<br />

is a risk for ord<strong>in</strong>ary corona breakdown, which is more prone to occur<br />

at <strong>in</strong>termediate pressures than at high <strong>and</strong> very low pressures. A corona<br />

discharge is usually much more detrimental than multipactor <strong>and</strong> for a<br />

certa<strong>in</strong> range <strong>of</strong> pressures, the breakdown threshold for corona is lower<br />

or much lower than for multipactor. Basic theory for ord<strong>in</strong>ary corona<br />

microwave discharge, when the mean free path <strong>of</strong> the electrons is smaller<br />

than the characteristic length <strong>of</strong> the device, is well known [9]. However,<br />

the <strong>in</strong>termediate range, between very low pressure <strong>and</strong> vacuum, has<br />

received little attention <strong>and</strong> therefore one <strong>of</strong> the ma<strong>in</strong> topics <strong>of</strong> this thesis<br />

is devoted to expla<strong>in</strong><strong>in</strong>g what happens with the breakdown threshold at<br />

these pressures.<br />

Theoretical studies <strong>of</strong> the multipactor phenomenon have to a great<br />

extent been performed us<strong>in</strong>g a one-dimensional model with a spatially<br />

uniform approximation <strong>of</strong> the electromagnetic field. However, many<br />

common RF devices <strong>in</strong>volve structures where the field is <strong>in</strong>homogeneous,<br />

2


where breakdown predictions based on such simple models will not be<br />

reliable. Examples <strong>of</strong> important geometries <strong>in</strong> microwave systems where<br />

the field is <strong>in</strong>homogeneous are waveguides, coaxial l<strong>in</strong>es, irises <strong>and</strong> septum<br />

polarisers. An important effect due to the non-uniform field that is<br />

not present when the electric field is spatially uniform, is the so called<br />

ponderomotive or Miller [10] force, which tends to push charged particles<br />

towards regions <strong>of</strong> low field amplitude. This can have both a qualitative<br />

<strong>and</strong> a quantitative effect on the multipactor regions. Analytical study <strong>of</strong><br />

resonant multipactor <strong>in</strong> a non-uniform field is not a trivial matter <strong>and</strong><br />

most researchers have resorted to numerical methods <strong>of</strong> <strong>in</strong>vestigation.<br />

However, <strong>in</strong> this thesis different aspects <strong>of</strong> multipactor <strong>in</strong> structures<br />

where the field is <strong>in</strong>herently <strong>in</strong>homogeneous are <strong>in</strong>vestigated us<strong>in</strong>g analytical<br />

methods <strong>and</strong> the results are compared with numerical simulations<br />

as well as with experimental data found <strong>in</strong> the literature.<br />

Many microwave systems <strong>of</strong> today operate <strong>in</strong> multicarrier mode,<br />

which means that several signals at different frequencies are transmitted<br />

simultaneously. In contrast to the s<strong>in</strong>gle carrier mode, the electric<br />

field envelope <strong>of</strong> the multicarrier system varies constantly. In most cases<br />

this is advantageous from a multipactor po<strong>in</strong>t <strong>of</strong> view, as the chang<strong>in</strong>g<br />

amplitude will destroy the resonance condition <strong>and</strong> thus suppress the<br />

discharge. In systems where the frequency separation is small, however,<br />

there is a risk that the signals will <strong>in</strong>terfere constructively for a large<br />

number <strong>of</strong> field cycles <strong>and</strong> the amplitude will rema<strong>in</strong> fairly constant,<br />

thus allow<strong>in</strong>g a discharge to develop. In such cases, the microwave eng<strong>in</strong>eer<br />

will have to try to f<strong>in</strong>d the worst case scenario <strong>and</strong> design the<br />

component with respect to that case or perform tests that guarantee<br />

that the part fulfils the requirements. Some attention will be given to<br />

these aspects <strong>in</strong> this thesis, which can be useful for the eng<strong>in</strong>eer when<br />

mak<strong>in</strong>g multipactor free multicarrier microwave designs.<br />

The thesis is organised as follows. A general <strong>in</strong>troduction to basic<br />

theory <strong>of</strong> multipactor <strong>in</strong> vacuum is given <strong>in</strong> chapter 2 as well as some<br />

guidel<strong>in</strong>es when it comes to multipactor free design. It will serve as a<br />

base when cont<strong>in</strong>u<strong>in</strong>g with the analysis <strong>of</strong> multipactor <strong>in</strong> low pressure<br />

gas <strong>in</strong> chapter 3, where first a simple model is presented, which only<br />

considers the friction force <strong>of</strong> the gas molecules. It is then followed by<br />

a more advanced model, which <strong>in</strong>cludes also the effect <strong>of</strong> impact ionisation<br />

as well as thermalisation <strong>of</strong> the electrons. Start<strong>in</strong>g with waveguide<br />

irises <strong>in</strong> chapter 4, vacuum discharge <strong>in</strong> structures where the field is<br />

non-uniform is considered. It is followed by a detailed analytical <strong>and</strong><br />

3


numerical study <strong>of</strong> the phenomenon <strong>in</strong> a coaxial l<strong>in</strong>e <strong>in</strong> chapter 5. F<strong>in</strong>ally,<br />

chapter 6 is devoted to different means <strong>of</strong> detect<strong>in</strong>g multipactor<br />

with special focus on detection by means <strong>of</strong> RF power modulation.<br />

4


Chapter 2<br />

<strong>Multipactor</strong> <strong>in</strong> vacuum<br />

<strong>Multipactor</strong> normally occurs at microwave frequencies, i.e. at 300 MHz -<br />

30 GHz. When discovered by Farnsworth <strong>in</strong> the 1930’s, he applied the<br />

technique to amplify an electric current. Others have also tried to f<strong>in</strong>d<br />

useful application <strong>of</strong> the phenomenon, e.g. <strong>in</strong> multipactor duplexers <strong>and</strong><br />

switches [11] <strong>and</strong> <strong>in</strong> electron guns [12,13]. However, dur<strong>in</strong>g the last 30<br />

years it has ma<strong>in</strong>ly been studied due to its detrimental effects on microwave<br />

components. It has been found to cause electric noise, which<br />

reduces the signal to noise ratio, a very serious problem if it occurs<br />

e.g. <strong>in</strong> a communication satellite where the signal power is limited <strong>and</strong><br />

counter-measures are difficult or impossible to implement. It can also<br />

detune microwave cavities, commonly used as e.g. resonators <strong>in</strong> filters,<br />

thus reflect<strong>in</strong>g the <strong>in</strong>com<strong>in</strong>g power back to the power amplifier. If the<br />

system does not have an appropriate power protection device, the amplifier<br />

may suffer permanent damage. Another concern is heat<strong>in</strong>g, which<br />

is a result <strong>of</strong> the power dissipated to the device walls as the multipact<strong>in</strong>g<br />

electrons strike the walls. Furthermore, the discharge can also cause<br />

direct physical damage to the component with the risk <strong>of</strong> permanently<br />

chang<strong>in</strong>g the electric properties <strong>of</strong> the device. However, the risk <strong>of</strong> such<br />

direct damage seems low, especially for metallic components. In cases<br />

where damage has been reported, it is not certa<strong>in</strong> that it was caused by<br />

multipactor [14,15]. <strong>Multipactor</strong> is known to be able to trigger ord<strong>in</strong>ary<br />

gas discharges [16–18], either by <strong>in</strong>creas<strong>in</strong>g the outgass<strong>in</strong>g from the component<br />

or just by start<strong>in</strong>g the breakdown at a pressure <strong>and</strong> a voltage<br />

where corona is not expected, s<strong>in</strong>ce gas breakdown can be susta<strong>in</strong>ed at a<br />

much lower voltage than what is needed to <strong>in</strong>itiate breakdown directly.<br />

Corona discharges are much more energetic <strong>and</strong> are known to be able<br />

5


to physically damage microwave components. Many researchers thus<br />

suspect that the observed damage was due to a multipactor <strong>in</strong>duced gas<br />

discharge.<br />

This chapter will present the basic theory <strong>of</strong> vacuum multipactor<br />

between two metallic parallel plates with an applied homogeneous, harmonic<br />

electric field. It is divided <strong>in</strong>to two major parts, one describ<strong>in</strong>g<br />

the s<strong>in</strong>gle carrier case <strong>and</strong> another devoted to multicarrier multipactor.<br />

2.1 S<strong>in</strong>gle Carrier<br />

One <strong>of</strong> the first communication satellites, Telstar I, operated <strong>in</strong> s<strong>in</strong>gle<br />

carrier mode. It had a capacity <strong>of</strong> 12 simultaneous telephone conversations<br />

[19] <strong>and</strong> the solar panels provided a power <strong>of</strong> only 15 watts. Today,<br />

satellites operate <strong>in</strong> multicarrier mode with powers <strong>of</strong> several kilowatts<br />

<strong>and</strong> new satellites are be<strong>in</strong>g designed for tens <strong>of</strong> kilowatts. Thus, the<br />

s<strong>in</strong>gle carrier mode is seldom found <strong>in</strong> real applications. Nevertheless,<br />

the s<strong>in</strong>gle carrier case is important as it has been thoroughly studied<br />

over the years <strong>and</strong> by mak<strong>in</strong>g certa<strong>in</strong> assumptions, the multicarrier case<br />

can be approximated by the s<strong>in</strong>gle carrier state <strong>and</strong> design <strong>and</strong> test<strong>in</strong>g<br />

can be done based on the simpler situation.<br />

2.1.1 Basic theory<br />

There are two ma<strong>in</strong> k<strong>in</strong>ds <strong>of</strong> multipactor, the s<strong>in</strong>gle-surface <strong>and</strong> the<br />

double-surface types. S<strong>in</strong>gle-surface multipactor can occur <strong>in</strong> structures<br />

with nonuniform field or with crossed electric <strong>and</strong> magnetic fields [20],<br />

where the electron, accelerated by the electric field, returns to the orig<strong>in</strong>al<br />

surface due to the circular motion caused by the magnetic field.<br />

This thesis, however, will focus on double-surface or parallel plate multipactor,<br />

but some attention will be given to s<strong>in</strong>gle-sided multipactor <strong>in</strong><br />

the case <strong>of</strong> a coaxial l<strong>in</strong>e.<br />

A multipactor discharge starts when a free electron <strong>in</strong>side a microwave<br />

device is accelerated by an electric field. In a strong field the<br />

electron will quickly reach a high velocity <strong>and</strong> upon impact with one <strong>of</strong><br />

the device walls, secondary electrons may be emitted from the wall. If<br />

the field direction reverses at this moment, the newly emitted electrons<br />

will start accelerat<strong>in</strong>g towards the opposite wall <strong>and</strong>, when collid<strong>in</strong>g with<br />

this wall, knock out additional electrons. As this procedure is repeated,<br />

the electron density grows quickly <strong>and</strong> with<strong>in</strong> fractions <strong>of</strong> a microsecond<br />

a fully developed multipactor discharge is obta<strong>in</strong>ed (see Fig. 2.1).<br />

6


Figure 2.1: Initial stage <strong>of</strong> parallel plate multipactor, where a free electron<br />

is accelerated by the electric field <strong>and</strong> is forced <strong>in</strong>to one <strong>of</strong> the<br />

plates, where it causes emission <strong>of</strong> secondary electrons.<br />

The motion <strong>of</strong> an electron <strong>in</strong> vacuum with an applied electric field<br />

can be studied by means <strong>of</strong> the equation <strong>of</strong> motion,<br />

m¨x = eE (2.1)<br />

where m (≈ 9.1 × 10 −31 kg) <strong>and</strong> e (≈ −1.6 × 10 −19 C) are the mass <strong>and</strong><br />

charge <strong>of</strong> the electron, x the direction <strong>of</strong> motion, <strong>and</strong> E the electric field.<br />

<strong>Multipactor</strong> requires an alternat<strong>in</strong>g field <strong>and</strong> <strong>in</strong> the parallel-plate model<br />

a spatially uniform harmonic field E = E0 s<strong>in</strong> ωt is assumed. Solv<strong>in</strong>g<br />

Eq. (2.1) with this field yields expressions for the velocity, ˙x, <strong>and</strong> the<br />

position, x,<br />

˙x = − eE0<br />

cos ωt + A (2.2)<br />

mω<br />

x = − eE0<br />

s<strong>in</strong> ωt + At + B (2.3)<br />

mω2 where A <strong>and</strong> B are constants <strong>of</strong> <strong>in</strong>tegration, which will be determ<strong>in</strong>ed by<br />

the <strong>in</strong>itial conditions. By assum<strong>in</strong>g that an electron is emitted at x = 0<br />

with an <strong>in</strong>itial velocity v0 when t = α/ω, fully constra<strong>in</strong>ed expressions<br />

for the velocity <strong>and</strong> position are obta<strong>in</strong>ed, viz.<br />

˙x = eE0<br />

(cos α − cos ωt) + v0<br />

(2.4)<br />

mω<br />

x = eE0<br />

mω2 � � v0<br />

s<strong>in</strong>α − s<strong>in</strong> ωt + (ωt − α)cos α + (ωt − α) (2.5)<br />

ω<br />

For resonant multipactor to occur it is necessary for the electron to<br />

reach the other device wall (x = d) when ωt = Nπ + α, where N is an<br />

7


odd positive <strong>in</strong>teger (N = 1,3,5...). Apply<strong>in</strong>g this resonance condition<br />

to Eq. (2.5), the follow<strong>in</strong>g expression is obta<strong>in</strong>ed for the amplitude <strong>of</strong><br />

the harmonic electric field.<br />

E0 =<br />

mω(ωd − Nπv0)<br />

e(Nπ cos α + 2s<strong>in</strong> α)<br />

(2.6)<br />

An important quantity when study<strong>in</strong>g multipactor is the impact velocity,<br />

s<strong>in</strong>ce this determ<strong>in</strong>es the secondary electron yield. It can be found<br />

by <strong>in</strong>sert<strong>in</strong>g ωt = Nπ + α <strong>in</strong> Eq. (2.4), which yields<br />

<strong>Multipactor</strong> boundaries<br />

vimpact = 2eE0<br />

mω<br />

cos α + v0<br />

(2.7)<br />

When construct<strong>in</strong>g the multipactor boundaries, i.e. the boundaries <strong>of</strong><br />

the regions <strong>in</strong> parameter space where multipactor can occur, an assumption<br />

will have to be made concern<strong>in</strong>g the <strong>in</strong>itial velocity. In reality, the<br />

<strong>in</strong>itial velocity <strong>of</strong> the emitted electrons will follow some k<strong>in</strong>d <strong>of</strong> distribution<br />

<strong>and</strong> a common choice is the Maxwellian distribution [21],<br />

�<br />

(v − vm) 2�<br />

f(v) ∝ exp −<br />

2v 2 T<br />

(2.8)<br />

where suitable parameters for the mean velocity, vm, <strong>and</strong> for the rmsvalue<br />

(or thermal spread), vT, have to be chosen. When perform<strong>in</strong>g<br />

particle-<strong>in</strong>-cell (PIC) simulations, such a distribution can be used to<br />

more accurately describe the <strong>in</strong>itial velocity <strong>of</strong> the emitted electrons.<br />

However, for an analytical solution a simpler assumption will have to<br />

be made. There are two common approaches, one which assumes that<br />

the electrons are emitted with a constant <strong>in</strong>itial velocity, v0, regardless<br />

<strong>of</strong> the impact velocity. The other assumes that the ratio between the<br />

impact <strong>and</strong> <strong>in</strong>itial velocities is equal to a constant, k = vimpact/v0. Both<br />

these approaches will be used <strong>and</strong> compared <strong>in</strong> this chapter, but <strong>in</strong> the<br />

follow<strong>in</strong>g chapter, which deals with multipactor <strong>in</strong> low pressure gas, the<br />

constant k approach will be used only for the simple model while the<br />

constant <strong>in</strong>itial velocity approach will be used for the more advanced<br />

model as that assumption is more physically correct. The reason why<br />

the constant k model has been used to such a great extent is the fact<br />

that it can successfully be fitted to experimental data. The cause <strong>of</strong> this<br />

success will be expla<strong>in</strong>ed <strong>in</strong> the subsection on hybrid modes below.<br />

8


In addition to fulfill<strong>in</strong>g the resonance condition, which resulted <strong>in</strong><br />

Eqs. (2.6) <strong>and</strong> (2.7), the secondary electron yield, SEY or σse, must be<br />

greater than or equal to unity. For most materials the secondary yield as<br />

a function <strong>of</strong> the impact velocity (or the impact energy, W = mv2 /(2|e|))<br />

has the same shape (see Fig. 2.2), even though the absolute values vary<br />

very much between different materials. The impact energy where the<br />

secondary yield first reaches unity is called the first cross-over po<strong>in</strong>t <strong>and</strong><br />

is denoted W1, after that the yield <strong>in</strong>creases <strong>and</strong> reaches a maximum<br />

at Wmax <strong>and</strong> the energy at which the yield drops below unity aga<strong>in</strong><br />

is called the second cross-over po<strong>in</strong>t, W2. Below Wzero no secondary<br />

yield is obta<strong>in</strong>ed [22, 23]. However, some researchers have published<br />

measurements <strong>of</strong> the SEY, which <strong>in</strong>dicate that it is possible that the<br />

SEY does not drop to zero below a m<strong>in</strong>imum impact velocity [24–28].<br />

On the contrary, it can <strong>in</strong>crease after reach<strong>in</strong>g a m<strong>in</strong>imum yield <strong>and</strong> even<br />

reach a yield close to unity for very low impact velocities. A yield close<br />

to unity implies that the electron does not produce any secondaries, but<br />

rather that the electron bounces <strong>of</strong>f the surface. This could have an<br />

important effect on the multipactor threshold <strong>and</strong> development, but <strong>in</strong><br />

this thesis, the model by Vaughan [22] has been used unless otherwise<br />

specified.<br />

By sett<strong>in</strong>g the impact velocity, Eq. (2.7), equal to the first cross over<br />

po<strong>in</strong>t (converted to velocity, v1) <strong>and</strong> tak<strong>in</strong>g Eq. (2.6) <strong>in</strong>to account, the<br />

resonant phase, α, can be found as a function <strong>of</strong> ωd,<br />

tan α = 1<br />

2<br />

� �<br />

2ωd − Nπ(v1 + v0)<br />

. (2.9)<br />

v1 − v0<br />

Us<strong>in</strong>g this result, the amplitude can be plotted as a function <strong>of</strong> ωd or<br />

fd us<strong>in</strong>g Eq. (2.6) (or Eq. (2.7)).<br />

One f<strong>in</strong>al th<strong>in</strong>g that need to be confirmed before draw<strong>in</strong>g the multipactor<br />

charts is the non-return<strong>in</strong>g electron limit. If the secondary<br />

electrons are emitted before the electric field reverses, the electrons will<br />

be retarded by the field <strong>and</strong> if the velocity is low, they are likely to<br />

return to the wall <strong>of</strong> emission <strong>and</strong> thus be<strong>in</strong>g lost as their energy is too<br />

low to produce new secondaries. The limit can be found by solv<strong>in</strong>g the<br />

follow<strong>in</strong>g system <strong>of</strong> equations:<br />

�<br />

˙x = 0<br />

(2.10)<br />

x = 0<br />

An analytical solution to this system <strong>of</strong> equations is not possible <strong>and</strong> <strong>in</strong><br />

order to establish the non-return<strong>in</strong>g electron limit, either a numerical<br />

9


σ se [−]<br />

2.5<br />

2<br />

1.5<br />

1<br />

0.5<br />

W 1<br />

W max<br />

Secondary electron yield<br />

σ se =1<br />

0<br />

0 500 1000 1500 2000 2500<br />

Primary electron energy [eV]<br />

Figure 2.2: Secondary electron yield as a function <strong>of</strong> the impact energy. Plotted<br />

us<strong>in</strong>g the formula for secondary electron yield presented <strong>in</strong><br />

Ref. [22]. Parameters used are: Wmax = 400 eV, σse,max = 2,<br />

<strong>and</strong> Wzero = 10 eV.<br />

10<br />

W 2


solution or some k<strong>in</strong>d <strong>of</strong> approximate solution will have to be used. In<br />

Ref. [21] an approximate formula for the non-return<strong>in</strong>g electron limit is<br />

given <strong>and</strong> re-writ<strong>in</strong>g it for the constant v0 approach yields,<br />

�<br />

16v0<br />

αm<strong>in</strong> = −<br />

(2.11)<br />

5v0 + 3vimpact<br />

Us<strong>in</strong>g Eqs. (2.6), (2.9), <strong>and</strong> (2.11) with v1 equal to the velocity correspond<strong>in</strong>g<br />

to the unity secondary electron yield, the lower multipactor<br />

threshold can be plotted. However, multipactor breakdown is possible<br />

also for impact velocities greater than v1, <strong>in</strong> fact, for all impact velocities<br />

between the first <strong>and</strong> second (v2) cross-over po<strong>in</strong>ts, the phenomenon can<br />

occur, i.e. for<br />

v1 < 2Vω cos α + v0 < v2 , (2.12)<br />

where Eq. (2.7) has been re-written us<strong>in</strong>g the oscillatory velocity Vω =<br />

eE0/mω. Thus, <strong>in</strong> order to construct the complete multipactor boundaries,<br />

the thresholds for a number <strong>of</strong> different energies between these two<br />

po<strong>in</strong>ts should be determ<strong>in</strong>ed <strong>and</strong> then the envelope <strong>of</strong> all the thresholds<br />

will be the complete multipactor susceptibility zone (see Fig. 2.3). Furthermore,<br />

each order <strong>of</strong> resonance, N, will have its own zone <strong>and</strong>, as<br />

shown <strong>in</strong> Fig. 2.3, the zones become narrower with <strong>in</strong>creas<strong>in</strong>g N. This<br />

type <strong>of</strong> chart, based on the assumption <strong>of</strong> constant <strong>in</strong>itial velocity, will<br />

be referred to as a Sombr<strong>in</strong> chart, s<strong>in</strong>ce J. Sombr<strong>in</strong> is one <strong>of</strong> the major<br />

advocates <strong>of</strong> this assumption [29].<br />

Us<strong>in</strong>g the other approach, with a constant ratio between the impact<br />

<strong>and</strong> <strong>in</strong>itial velocities, k = vimpact/v0, the formulas for the resonant phase,<br />

Eq. (2.9), the amplitude, Eq. (2.6), <strong>and</strong> the non-return<strong>in</strong>g electron limit,<br />

Eq. (2.11), will have to be slightly re-written:<br />

tan α = 1<br />

k − 1<br />

E0 =<br />

e( k+1<br />

k−1<br />

αm<strong>in</strong> = −<br />

� kωd<br />

v1<br />

mω 2 d<br />

− (k + 1)N π<br />

2<br />

Nπ cos α + 2s<strong>in</strong> α)<br />

�<br />

16<br />

8 + 3(k − 1)<br />

�<br />

(2.13)<br />

(2.14)<br />

(2.15)<br />

Us<strong>in</strong>g these formulas, multipactor charts similar to the one <strong>in</strong> Fig. 2.3<br />

can be produced, cf. Fig. 2.4. When this is used, the charts are commonly<br />

referred to as Hatch <strong>and</strong> Williams charts, as they were the first<br />

11


Voltage [V]<br />

10 4<br />

10 3<br />

10 2<br />

10 0<br />

N=1<br />

N=3<br />

N=5<br />

N=9<br />

N=7<br />

10 1<br />

Frequency − Gap product [GHz⋅mm]<br />

Figure 2.3: <strong>Multipactor</strong> susceptibility chart based on the constant <strong>in</strong>itial velocity<br />

approach. Parameters used are: W0 = 3.68 eV, W1 =<br />

23 eV, W2 = 1000 eV, <strong>and</strong> Nmax = 9.<br />

12


who produced charts <strong>of</strong> this type [30]. Characteristic for the Hatch<br />

<strong>and</strong> Williams charts are that the multipactor zones are wider than the<br />

Sombr<strong>in</strong> charts for <strong>in</strong>creas<strong>in</strong>g voltage. This occurs s<strong>in</strong>ce a constant<br />

vimpact/v0 implies that v0 <strong>in</strong>creases as the impact velocity <strong>in</strong>creases.<br />

When e.g. Wimpact = 3000 eV, this means that for k = 2.5 the <strong>in</strong>itial<br />

energy W0 = 480 eV, clearly an unrealistic <strong>in</strong>itial velocity.<br />

Voltage amplitude [V]<br />

10 4<br />

10 3<br />

10 2<br />

10 0<br />

N=1<br />

N=3<br />

N=5<br />

N=9<br />

N=7<br />

10 1<br />

Frequency − Gap product [GHz⋅mm]<br />

Figure 2.4: <strong>Multipactor</strong> susceptibility chart produced with the constant k<br />

assumption. Parameters used are: k = 2.5 (correspond<strong>in</strong>g to<br />

an <strong>in</strong>itial W0 = 3.68 eV when Wimpact = W1), W1 = 23 eV,<br />

W2 = 1000 eV, <strong>and</strong> Nmax = 9.<br />

By know<strong>in</strong>g the secondary electron emission characteristics <strong>of</strong> a material<br />

as given by the parameters W1, W2, <strong>and</strong> W0 or k, multipactor<br />

charts for that material can be designed. However, Woode <strong>and</strong> Petit<br />

[31] performed a large series <strong>of</strong> multipactor experiments dur<strong>in</strong>g the<br />

1980’s <strong>and</strong> used the Hatch <strong>and</strong> Williams charts to fit the experimental<br />

data. By tun<strong>in</strong>g the k <strong>and</strong> W1 parameters for each zone, they were able<br />

to produce multipactor charts that fit the experimental data quite well<br />

(cf. Fig. 2.5). The problem with this empirical approach is that it has<br />

to assume different values for the first cross-over po<strong>in</strong>t for each zone <strong>in</strong><br />

order to obta<strong>in</strong> good fitt<strong>in</strong>g. This is clearly an unphysical approach <strong>and</strong><br />

will not contribute to an improved underst<strong>and</strong><strong>in</strong>g <strong>of</strong> the phenomenon,<br />

13


even though it may be sufficient from an eng<strong>in</strong>eer<strong>in</strong>g po<strong>in</strong>t <strong>of</strong> view. On<br />

the other h<strong>and</strong>, the higher order modes have a narrower phase-focus<strong>in</strong>g<br />

range (see below), which makes it difficult to compensate for e.g. <strong>in</strong>itial<br />

velocity spread, <strong>and</strong> thus a secondary yield <strong>of</strong> unity may not be sufficient<br />

to susta<strong>in</strong> a discharge. Consequently, an impact energy somewhat<br />

higher than the first cross-over po<strong>in</strong>t will be needed when construct<strong>in</strong>g<br />

the lower multipactor threshold for the higher order modes. This will<br />

be discussed further <strong>in</strong> the subsection “Effect <strong>of</strong> r<strong>and</strong>om emission delays<br />

<strong>and</strong> <strong>in</strong>itial velocity spread.”<br />

Voltage amplitude [V]<br />

10 4<br />

10 3<br />

10 2<br />

10 0<br />

10 1<br />

Frequency − Gap product [GHz⋅mm]<br />

Figure 2.5: Hatch <strong>and</strong> Williams charts for alum<strong>in</strong>ium together with measurement<br />

data by Woode <strong>and</strong> Petit [31].<br />

When the microwave eng<strong>in</strong>eer assesses the risk <strong>of</strong> hav<strong>in</strong>g a multipactor<br />

discharge, it is usually not the boundary <strong>of</strong> the <strong>in</strong>dividual breakdown<br />

region that is considered. Typically, the lower envelope <strong>of</strong> the all<br />

the zones is taken as the “design threshold” (cf. Figs. 2.3 <strong>and</strong> 2.4). By<br />

sett<strong>in</strong>g the phase, α, <strong>in</strong> Eq. (2.7) to zero, the lowest field amplitude to<br />

achieve a certa<strong>in</strong> vimpact is obta<strong>in</strong>ed. Thus the lower envelope, which<br />

is the same for both the Sombr<strong>in</strong> Chart <strong>and</strong> the Hatch <strong>and</strong> Williams<br />

chart, is given by,<br />

E0 = (v1 − v0)mω<br />

. (2.16)<br />

2e<br />

14


Phase-focus<strong>in</strong>g<br />

In the previous subsection a mechanism called phase-focus<strong>in</strong>g [1,5,32]<br />

was mentioned <strong>and</strong> <strong>in</strong> multipactor theory this is an important concept.<br />

In order for an electron to be a part <strong>of</strong> the discharge, it must have<br />

a phase close to the resonant phase as given by Eq. (2.9) or (2.13).<br />

Due to delays between the impact <strong>and</strong> emission <strong>of</strong> a new electron or<br />

a spread <strong>in</strong> the <strong>in</strong>itial velocity, an electron will always acquire a small<br />

phase error. Inside the phase-focus<strong>in</strong>g range, such an error will decrease<br />

as the electron traverses the electrode gap. In other words, the phases<br />

<strong>of</strong> the electrons will tend to converge towards the resonant phase, thus<br />

keep<strong>in</strong>g all electrons close together. Outside the range <strong>of</strong> phase-focus<strong>in</strong>g,<br />

the error will grow with each passage <strong>and</strong> after one or a few transits<br />

the electron will be lost. In order for a discharge to occur under such<br />

circumstances, the impact energy has to be large enough to produce a<br />

secondary yield sufficiently above unity to compensate for the <strong>in</strong>curred<br />

losses.<br />

To see <strong>in</strong> what range the phase focus<strong>in</strong>g mechanism is active, a small<br />

phase error can be <strong>in</strong>troduced <strong>in</strong> Eq. (2.5) while keep<strong>in</strong>g the amplitude<br />

<strong>and</strong> phase constant <strong>and</strong> sett<strong>in</strong>g x = d. The ratio between the f<strong>in</strong>al <strong>and</strong><br />

<strong>in</strong>itial error is called the stability factor, G [33], <strong>and</strong> the condition for<br />

stable phase is:<br />

|G| < 1 (2.17)<br />

By sett<strong>in</strong>g |G| = 1, the phase range with<strong>in</strong> which the phase is stable can<br />

be obta<strong>in</strong>ed. An <strong>in</strong>terest<strong>in</strong>g observation here is that even though the<br />

lower multipactor threshold for the constant k theory <strong>and</strong> the constant<br />

<strong>in</strong>itial velocity model are identical, the range <strong>of</strong> stable phases varies substantially<br />

[34]. This can be seen clearly from the analytical expressions<br />

for the phase stability limits, which for constant k theory reads,<br />

φR = arctan( 2<br />

πN (vimpact − v0<br />

vimpact + v0<br />

)) (2.18)<br />

φL = − arctan( 2<br />

) (2.19)<br />

πN<br />

<strong>and</strong> for the constant <strong>in</strong>itial velocity approach,<br />

φR = arctan( 2<br />

) (2.20)<br />

πN<br />

φL = − arctan( 2<br />

πN (vimpact + v0<br />

vimpact − v0<br />

)) (2.21)<br />

15


where φL <strong>and</strong> φR are the left <strong>and</strong> right limits respectively. This difference<br />

is illustrated graphically <strong>in</strong> Fig. 2.6. However, when v0 ≪ vimpact both<br />

approaches yield the same phase stability limits.<br />

Voltage [V]<br />

10 3<br />

10 2<br />

10 1<br />

Unstable phase range (constant v 0 )<br />

Stable phase range (constant v 0 )<br />

Unstable phase range (constant k)<br />

Stable phase range (constant k)<br />

N=1<br />

10 0<br />

Frequency − Gap product [GHz⋅mm]<br />

Figure 2.6: <strong>Low</strong>er multipactor thresholds <strong>in</strong> vacuum for the first 3 orders<br />

<strong>of</strong> resonance (N = 1, 3, <strong>and</strong> 5). The curves for the constant k<br />

model are plotted slightly <strong>of</strong>fset as the curves otherwise overlap.<br />

Parameters used are: W1 = 23 eV , W0 = 3.68 eV, <strong>and</strong> k = 2.5.<br />

Saturation<br />

In order to susta<strong>in</strong> a multipactor breakdown, the secondary electron<br />

emission yield must be greater than or equal to unity. If the yield is less,<br />

the electron number will quickly decrease <strong>and</strong> the discharge disappears.<br />

With a σse greater than unity the electron number will grow rapidly with<br />

each impact <strong>and</strong> if no saturation mechanism is considered the number<br />

<strong>of</strong> electrons after a time t, if the field frequency is f, will be:<br />

N=3<br />

N=5<br />

Ne(t) = Ne(0)(σse) 2ft<br />

N (2.22)<br />

The rapid growth <strong>of</strong> the number <strong>of</strong> electrons can be illustrated with<br />

an example. Suppose σse = 1.5 <strong>and</strong> f = 2 GHz, then the number <strong>of</strong><br />

16


electrons after 20 ns for the first order <strong>of</strong> resonance with one <strong>in</strong>itial<br />

electron will be more than 10 14 .<br />

In a very short time, the number <strong>of</strong> electrons will grow to very high<br />

values <strong>and</strong> it is clear that some k<strong>in</strong>d <strong>of</strong> saturation mechanism will become<br />

active. Two ma<strong>in</strong> saturation processes have been described <strong>in</strong> the<br />

literature. The first is the space charge effect [3], which is the most<br />

obvious effect. With<strong>in</strong> the electron bunch the <strong>in</strong>dividual electrons will<br />

repel each other caus<strong>in</strong>g a change <strong>in</strong> phase <strong>of</strong> those electrons <strong>and</strong> if the<br />

phase error is too large, the probability <strong>of</strong> los<strong>in</strong>g electrons <strong>in</strong>creases <strong>and</strong><br />

eventually the effective secondary yield will be equal to unity <strong>and</strong> saturation<br />

has occurred. The second type <strong>of</strong> saturation process [35,36] can<br />

set <strong>in</strong> if the discharge takes place <strong>in</strong>side a resonant cavity. Due to a high<br />

Q-value, the electric field strength is high <strong>and</strong> thus the risk for a discharge<br />

will <strong>in</strong>crease. If a multipactor discharge is started, the electrons<br />

travers<strong>in</strong>g the gap make up an alternat<strong>in</strong>g current, which loads the cavity.<br />

Load<strong>in</strong>g the cavity means that the Q-value will decrease <strong>and</strong> thus<br />

also the electric field strength. It is clear that this is a self-suppress<strong>in</strong>g<br />

effect. As the multipactor current <strong>in</strong>creases, the field strength decreases<br />

<strong>and</strong> with it the impact velocity <strong>of</strong> the electrons lead<strong>in</strong>g to a lowered<br />

secondary emission yield. Eventually the secondary yield reaches unity<br />

<strong>and</strong> saturation has been reached.<br />

2.1.2 Hybrid modes<br />

It may seem somewhat contradictory to assert that the model based on<br />

a constant <strong>in</strong>itial velocity is more physically correct than the constant<br />

k theory, when the latter approach can be better fitted to experimental<br />

data. However, as briefly mentioned previously, the reason for this<br />

paradox is the hybrid modes. Some <strong>of</strong> these modes were identified by<br />

Refs. [29,37,38] <strong>and</strong> a general treatment is given <strong>in</strong> [39]. The modes can<br />

be found by allow<strong>in</strong>g N <strong>in</strong> the resonance condition for Eq. (2.5) to be<br />

a sequence <strong>of</strong> odd half-cycles <strong>of</strong> the electric field, {N1,N2,N3...}, where<br />

N1 = N <strong>and</strong> the rema<strong>in</strong><strong>in</strong>g Nn ≥ N for the hybrid modes between the<br />

N th <strong>and</strong> (N + 2) th zones. Each such sequence will result <strong>in</strong> a narrow<br />

multipactor zone located between the ma<strong>in</strong> multipactor areas. The lowest<br />

order hybrid mode <strong>in</strong> the parallel-plate case is the {1,3} mode, which<br />

means that the transit time <strong>in</strong> one direction takes 1/2 RF-cycle <strong>and</strong> the<br />

return transit takes 3/2 RF-cycles. This mode is then also associated<br />

with two different resonant phases, viz. α1 = 0 <strong>and</strong> α2 = π/3 [39].<br />

This mode can be found between the two first classical resonance zones<br />

17


(cf. Fig. 2.7). When tak<strong>in</strong>g the envelope <strong>of</strong> these zones, the differences<br />

<strong>in</strong> the right boundaries <strong>of</strong> the ma<strong>in</strong> zones, between the constant <strong>in</strong>itial<br />

velocity <strong>and</strong> the constant k approaches, become negligible. This can be<br />

seen clearly <strong>in</strong> Fig. 2.7 [40]. The existence <strong>of</strong> the hybrid modes requires<br />

Figure 2.7: Vacuum multipactor with the ma<strong>in</strong> zones as well as a few hybrid<br />

zones [40]. With the envelope <strong>of</strong> the hybrid zones <strong>in</strong>cluded<br />

the resemblance between the Sombr<strong>in</strong> chart <strong>and</strong> the Hatch <strong>and</strong><br />

Willams chart (cf. Fig. 2.5) is strik<strong>in</strong>g.<br />

phase stability, just as for the classical zones discussed above. However,<br />

the width <strong>of</strong> each hybrid zone is very small <strong>and</strong> thus it is very sensitive<br />

to an <strong>in</strong>itial velocity spread. On the other h<strong>and</strong>, there are many hybrid<br />

zones very close to one other <strong>and</strong> this spread will result <strong>in</strong> a mix<strong>in</strong>g or<br />

overlapp<strong>in</strong>g <strong>of</strong> the resonances [39].<br />

2.1.3 Factors effect<strong>in</strong>g the threshold<br />

There are many different aspects that need to be considered when determ<strong>in</strong><strong>in</strong>g<br />

the multipactor threshold. The most important <strong>and</strong> most obvious<br />

ones are type <strong>of</strong> material, gap size, <strong>and</strong> amplitude <strong>and</strong> frequency<br />

<strong>of</strong> the electric field. These are all part <strong>of</strong> the basic theory as described<br />

above. Apart from these there are other more or less important factors.<br />

The supply <strong>of</strong> primary electrons does not effect the theoretical<br />

threshold, which can be determ<strong>in</strong>ed with methods described earlier <strong>in</strong><br />

this thesis. Nevertheless, a weak source <strong>of</strong> seed electrons can result <strong>in</strong><br />

an apparent higher threshold dur<strong>in</strong>g test<strong>in</strong>g. In a typical test setup for<br />

determ<strong>in</strong>ation <strong>of</strong> the breakdown amplitude, an electric field is applied<br />

18


<strong>and</strong> the field strength is <strong>in</strong>creased at regular <strong>in</strong>tervals. If no electron<br />

is <strong>in</strong> an advantageous position, i.e. has a suitable phase from a multipactor<br />

po<strong>in</strong>t <strong>of</strong> view, when the right amplitude is set, a discharge will<br />

not occur. As the amplitude is <strong>in</strong>creased further, the impact velocity<br />

<strong>of</strong> any free electrons, also those that are not <strong>in</strong> a favourable position,<br />

will be high <strong>and</strong> the secondary yield will be an additional source <strong>of</strong> free<br />

electrons. Thus the chances <strong>of</strong> gett<strong>in</strong>g a breakdown <strong>in</strong>creases until it<br />

eventually occurs. For experimental use, a hot filament or a radioactive<br />

source can be used to produce a sufficient amount <strong>of</strong> free electrons to<br />

achieve reliable measurement results [14].<br />

Another factor that can have a significant effect on the threshold<br />

is contam<strong>in</strong>ation. Both the first cross-over po<strong>in</strong>t, W1, <strong>and</strong> the maximum<br />

secondary yield, σse,max, can be drastically affected. A lowered<br />

W1 means that a discharge can occur at a lower voltage <strong>and</strong> an <strong>in</strong>creased<br />

σse,max can result <strong>in</strong> a faster growth <strong>of</strong> the total number <strong>of</strong> electrons.<br />

In Ref. [31] a detailed analysis <strong>of</strong> the impact <strong>of</strong> different types <strong>of</strong> contam<strong>in</strong>ants<br />

was made. It was noted that the plastic bags, which were<br />

normally used to protect the microwave components from dirt, were the<br />

ma<strong>in</strong> source <strong>of</strong> contam<strong>in</strong>ation. A threshold reduction <strong>of</strong> up to 4 dB<br />

was found. Also dust <strong>and</strong> f<strong>in</strong>gerpr<strong>in</strong>ts were a direct source <strong>of</strong> a lowered<br />

threshold. In the report [31] it was recommended that cleaned<br />

microwave parts for space use should be h<strong>and</strong>led with cotton gloves <strong>and</strong><br />

stored <strong>in</strong> hard plastic boxes.<br />

Microwave parts which have not been properly vented before power<br />

is applied can also have a threshold that is different from the expected<br />

multipactor threshold. If there is too much gas, corona breakdown may<br />

occur, <strong>and</strong> with<strong>in</strong> a certa<strong>in</strong> range <strong>of</strong> pressures, close to the m<strong>in</strong>imum <strong>of</strong><br />

the so called Paschen curve, the breakdown threshold can be significantly<br />

lower than <strong>in</strong> the multipactor case. In the pressure range correspond<strong>in</strong>g<br />

to the transition region between corona <strong>and</strong> multipactor, a higher<br />

threshold can sometimes be expected. More details about this will be<br />

presented <strong>in</strong> chapter 3.<br />

Other factors that can have an <strong>in</strong>direct effect on the multipactor<br />

threshold are the voltage st<strong>and</strong><strong>in</strong>g wave ratio, VSWR, <strong>and</strong> the temperature.<br />

If the VSWR is greater than what was <strong>in</strong>tended with the design,<br />

the peak field strength <strong>in</strong> the system will also be greater than expected<br />

<strong>and</strong> thus a discharge may occur at a lower power level than assumed. An<br />

<strong>in</strong>creased temperature can lead to <strong>in</strong>creased outgass<strong>in</strong>g from the device<br />

walls result<strong>in</strong>g <strong>in</strong> concerns similar to those <strong>of</strong> improper vent<strong>in</strong>g.<br />

19


2.1.4 Methods <strong>of</strong> suppression<br />

Many <strong>of</strong> the factors mentioned <strong>in</strong> the previous section that affect the<br />

multipactor threshold can also be utilised to suppress the discharge. The<br />

without doubt easiest method <strong>of</strong> avoid<strong>in</strong>g a breakdown is to pressurise<br />

the component. The field strength required to achieve breakdown at<br />

atmospheric pressure is <strong>in</strong> general much higher than at low pressures or<br />

<strong>in</strong> vacuum. However, such a method is seldom feasible for components<br />

that will be used e.g. <strong>in</strong> space, where the external environment is a high<br />

vacuum. A small leakage can lead to slow vent<strong>in</strong>g <strong>of</strong> the component <strong>and</strong><br />

thus risk<strong>in</strong>g severe corona discharge when the pressure reaches the range<br />

where the m<strong>in</strong>imum breakdown field occurs.<br />

Another way <strong>of</strong> suppress<strong>in</strong>g multipactor is to amplitude modulate<br />

the ma<strong>in</strong> carrier [21, 41]. If both signals are s<strong>in</strong>usoidal, the total field<br />

can be written:<br />

Etot = E1 s<strong>in</strong>ω1t + E2 s<strong>in</strong> ω2t (2.23)<br />

This means that the envelope <strong>of</strong> the signal will vary accord<strong>in</strong>g to (see<br />

also Fig. 2.8):<br />

�<br />

Eenv = E2 1 + E2 2 + 2E1E2 cos (ω1 − ω2)t (2.24)<br />

When the total field strength is well above the multipactor threshold<br />

(see Fig. 2.8), the secondary electron yield will <strong>in</strong>crease quickly accord<strong>in</strong>g<br />

to Eq. (2.22). However, as soon as the voltage drops below the<br />

threshold aga<strong>in</strong>, the electron loss will be large <strong>and</strong> accord<strong>in</strong>g to Ref. [21]<br />

all electrons will be lost <strong>in</strong> just a few RF cycles. However, whether or<br />

not this is true also depends on the secondary yield properties <strong>of</strong> the<br />

electrode material. For materials with a very high maximum secondary<br />

yield, the number <strong>of</strong> electrons ga<strong>in</strong>ed while above the threshold can be<br />

greater than the losses <strong>in</strong>curred while below. In such a case, no suppression<br />

is achieved <strong>and</strong> <strong>in</strong> some cases, the discharge may even become<br />

more powerful than before the modulation carrier was added [42] (cf.<br />

Fig. 2.9). Thus <strong>in</strong> order to successfully suppress a multipactor discharge<br />

us<strong>in</strong>g amplitude modulation, it is vital that the material has a low maximum<br />

secondary yield (preferably less than about 1.5). Due to the risk<br />

<strong>of</strong> contam<strong>in</strong>ation, which can greatly <strong>in</strong>crease the maximum secondary<br />

yield, great care should be taken to assure a high level <strong>of</strong> cleanl<strong>in</strong>ess if<br />

this method <strong>of</strong> suppression is to be used.<br />

To AM-modulate the carrier is probably not feasible <strong>in</strong> most cases, as<br />

it would require extra hardware to produce the AM-signal. However, the<br />

20


Amplitude<br />

1.5<br />

1<br />

0.5<br />

<strong>Multipactor</strong> threshold<br />

Amplitude Modulation, E0/E1=0.4<br />

Ma<strong>in</strong> signal<br />

Modulation signal<br />

Sum signal<br />

Envelope<br />

0<br />

0 1 2 3 4 5<br />

Time<br />

6 7 8 9 10<br />

Figure 2.8: Two signals <strong>and</strong> their sum signal (absolute values). The envelope<br />

varies <strong>and</strong> is partly above <strong>and</strong> partly below the multipactor<br />

threshold.<br />

typical b<strong>and</strong>pass filter<strong>in</strong>g <strong>of</strong> a PSK (Phase-Shift Key<strong>in</strong>g) signal causes<br />

modulations <strong>in</strong> the time-doma<strong>in</strong>. The QPSK (Quadrature Phase-Shift<br />

Key<strong>in</strong>g) signal <strong>in</strong> Fig. 2.10 has unity amplitude before filter<strong>in</strong>g. Afterwards,<br />

the peaks are higher than the orig<strong>in</strong>al amplitude, but the troughs<br />

can sometimes go down almost to zero amplitude. Compar<strong>in</strong>g this with<br />

the AM-suppression, it is clear that an electron avalanche that is <strong>in</strong>itiated<br />

dur<strong>in</strong>g the peak periods, will be ext<strong>in</strong>guished as the amplitude<br />

falls close to zero. However, for a typical PSK-signal, the duration between<br />

phase shifts (which normally co<strong>in</strong>cides with the troughs) is several<br />

hundreds <strong>of</strong> RF cycles. Thus for most microwave systems, there will be<br />

ample time for a discharge to develop. But, when the amplitude drops<br />

below the threshold, the electron bunch will disappear <strong>and</strong> when the<br />

amplitude <strong>in</strong>creases above the threshold aga<strong>in</strong>, there may not be any<br />

seed electrons present to restart the electron avalanche. Thus, the system<br />

will have sporadic discharges, which, if they do not occur too <strong>of</strong>ten,<br />

may not seriously degrade the signal.<br />

A very common way <strong>of</strong> suppress<strong>in</strong>g a vacuum discharge is to apply<br />

a coat<strong>in</strong>g [26–28], a surface treatment [43], or a film [44] that has a<br />

high first cross-over po<strong>in</strong>t as well as a low maximum secondary electron<br />

21


Noise (dBm)<br />

Power #1 (dBm)<br />

Match #1 (dB)<br />

Power #2 (dBm)<br />

−65<br />

−70<br />

−75<br />

−80<br />

−85<br />

Amplitud Modulation, w2/w1=1.16)<br />

0 1 2 3 4 5 6 7 8 9 10<br />

x 10 4<br />

48<br />

47.5<br />

47<br />

46.5<br />

0 1 2 3 4 5 6 7 8 9 10<br />

x 10 4<br />

40<br />

20<br />

0<br />

0 1 2 3 4 5 6 7 8 9 10<br />

x 10 4<br />

−20<br />

−11<br />

−11.5<br />

−12<br />

0 1 2 3 4 5 6 7 8 9 10<br />

x 10 4<br />

Time (ms)<br />

Figure 2.9: <strong>Multipactor</strong> experiment with two carriers with E2/E1 = 0.36<br />

<strong>and</strong> ω2/ω1 = 1.16. Due to a high maximum secondary yield,<br />

multipactor suppression is not possible (the material used <strong>in</strong> the<br />

experiment was pla<strong>in</strong> alum<strong>in</strong>ium, which can have a σse,max ≈ 3).<br />

When the modulation signal is applied, the magnitude <strong>of</strong> the<br />

multipactor noise <strong>in</strong>creases significantly.<br />

Signal amplitude [−]<br />

1<br />

0.5<br />

0<br />

−0.5<br />

−1<br />

Square Root Raised Cos<strong>in</strong>e filtered QPSK signal<br />

1 1.5 2 2.5<br />

Time [µ s]<br />

3 3.5<br />

Figure 2.10: Example <strong>of</strong> a QPSK modulated signal after b<strong>and</strong>pass filter<strong>in</strong>g.<br />

22


emission yield. So far, no practical coat<strong>in</strong>g with a σse,max below unity<br />

has been found. However, alod<strong>in</strong>e is a commonly used surface coat<strong>in</strong>g for<br />

space-bound microwave devices made <strong>of</strong> alum<strong>in</strong>ium. It <strong>in</strong>creases the first<br />

cross-over po<strong>in</strong>t to around 60 eV <strong>and</strong> reduces σse,max to about 1.5, even<br />

though the actual values vary much between samples. The concern with<br />

a material with very good anti-multipactor properties is contam<strong>in</strong>ation.<br />

A few f<strong>in</strong>gerpr<strong>in</strong>ts or a very small layer <strong>of</strong> dust can drastically alter the<br />

properties <strong>of</strong> the material <strong>and</strong> make it prone to discharges.<br />

By apply<strong>in</strong>g a DC electric or magnetic field, the electron trajectory<br />

can be disturbed <strong>and</strong> the important resonance condition can be<br />

destroyed, thus mak<strong>in</strong>g multipactor impossible. Simulations [45] have<br />

shown that an external DC magnetic field applied <strong>in</strong> the direction <strong>of</strong><br />

wave propagation <strong>in</strong> a rectangular waveguide can efficiently suppress<br />

multipactor. A drawback with the method is the extra components required<br />

to produce the magnetic or electric field <strong>and</strong> thus the method<br />

may not be feasible for e.g. space applications, where extra weight is<br />

undesirable.<br />

The most efficient way <strong>of</strong> avoid<strong>in</strong>g multipactor is to make a design<br />

where the mechanical dimensions are such that a power much higher<br />

than the nom<strong>in</strong>al power is required to start a discharge. However, that<br />

may lead to large <strong>and</strong> heavy designs, which are to be avoided <strong>in</strong> space<br />

systems, <strong>and</strong> thus one may have to resort to one or several <strong>of</strong> the above<br />

mentioned methods <strong>of</strong> multipactor suppression.<br />

2.1.5 Effect <strong>of</strong> r<strong>and</strong>om emission delays <strong>and</strong> <strong>in</strong>itial velocity<br />

spread<br />

In the above analysis <strong>of</strong> multipact<strong>in</strong>g electrons <strong>in</strong> a harmonic electric<br />

field, it was assumed that all secondary electrons were emitted with a<br />

fixed <strong>in</strong>itial velocity, v0. However, as briefly mentioned previously, the<br />

electrons are actually emitted with a distribution <strong>of</strong> velocities <strong>and</strong> the<br />

Maxwellian distribution is <strong>of</strong>ten used <strong>in</strong> simulations. Apart from the<br />

spread <strong>in</strong> <strong>in</strong>itial velocities, there is also a f<strong>in</strong>ite time between impact <strong>of</strong><br />

the primary electron <strong>and</strong> emission <strong>of</strong> the secondaries. S<strong>in</strong>ce this time <strong>in</strong><br />

most cases is very small compared to the RF period, it was neglected <strong>in</strong><br />

the previous analysis. Nevertheless, this time will cause a small phase<br />

error <strong>and</strong> if the resonant phase is close to the phase stability limits as<br />

given by Eqs. (2.18) - (2.21), the phase error may result <strong>in</strong> an <strong>in</strong>creased<br />

electron loss.<br />

A detailed analysis <strong>of</strong> the effect <strong>of</strong> r<strong>and</strong>om secondary delay times <strong>and</strong><br />

23


<strong>and</strong>om spread <strong>in</strong> emission velocities was done by Riyopoulos et al. [33].<br />

They found that by <strong>in</strong>clud<strong>in</strong>g the effects <strong>of</strong> these r<strong>and</strong>om parameters,<br />

the effective secondary electron yield, σ ∗ se, was reduced to a number <strong>in</strong><br />

the range σse/2 < σ ∗ se < σse. This means that the effective secondary<br />

electron yield will be a function not only <strong>of</strong> the impact velocity, but also<br />

<strong>of</strong> the resonant phase as well as the phase spread caused by the spread<br />

<strong>in</strong> <strong>in</strong>itial velocities <strong>and</strong> secondary delay times. Another study, which<br />

supports this result, <strong>in</strong>vestigated the effect on the different resonance<br />

zones for different values <strong>of</strong> the maximum SEY due to <strong>in</strong>itial velocity<br />

spread [46]. It was found that, except for the first order mode, a realistic<br />

thermal spread <strong>of</strong> the <strong>in</strong>itial electrons raised the multipactor SEY<br />

requirement from unity to above unity. For the higher order modes a<br />

SEY greater than approximately 1.5 was necessary to compensate for<br />

the losses <strong>in</strong>curred. In addition, with <strong>in</strong>creas<strong>in</strong>g velocity spread, the<br />

multipactor zones started to overlap. The <strong>in</strong>creased SEY requirement<br />

will result <strong>in</strong> an <strong>in</strong>creased threshold for the higher order modes <strong>and</strong> can<br />

expla<strong>in</strong> the success with <strong>in</strong>creas<strong>in</strong>g the first cross-over po<strong>in</strong>t <strong>in</strong> the Hatch<br />

<strong>and</strong> Williams charts when fitt<strong>in</strong>g experimental data (see Fig.2.5).<br />

The importance <strong>of</strong> the spread <strong>in</strong> <strong>in</strong>itial velocities can be seen when<br />

construct<strong>in</strong>g multipactor charts for a constant <strong>in</strong>itial velocity without allow<strong>in</strong>g<br />

compensation for electron losses outside the phase stability range.<br />

In Fig. 2.11 zones bounded by solid l<strong>in</strong>es <strong>in</strong>dicate the region where multipactor<br />

can take place under this assumption. The dashed l<strong>in</strong>es make<br />

up wider zones that encompass the other zones <strong>and</strong> are identical to the<br />

zones shown <strong>in</strong> Fig. 2.3.<br />

By <strong>in</strong>clud<strong>in</strong>g a higher secondary electron yield <strong>and</strong> a spread <strong>in</strong> <strong>in</strong>itial<br />

velocities, the multipactor zones will become wider than the solid<br />

l<strong>in</strong>e zones. A σse greater than unity, which will be the case when the<br />

impact velocity is greater than the first cross-over po<strong>in</strong>t, will compensate<br />

for some <strong>of</strong> the losses <strong>in</strong>curred due to phase <strong>in</strong>stability. A spread<br />

<strong>in</strong> <strong>in</strong>itial velocities will widen the range <strong>of</strong> possible resonant phases (cf.<br />

Eq. (2.9)) <strong>and</strong> the left <strong>and</strong> right limits will not be as sharp as <strong>in</strong>dicated<br />

by the solid l<strong>in</strong>e multipactor zones <strong>in</strong> Fig. 2.11. This widen<strong>in</strong>g <strong>of</strong> the<br />

multipactor zones has been taken <strong>in</strong>to account to a certa<strong>in</strong> extent <strong>in</strong> the<br />

traditional analytical approach, which is <strong>in</strong>dicated by the wider dashed<br />

l<strong>in</strong>e zones <strong>in</strong> Fig. 2.11. However, the widen<strong>in</strong>g should not only be towards<br />

the left side, but also towards the right [39]. Furthermore, the<br />

sharp lower left corner <strong>of</strong> each dashed l<strong>in</strong>e zone is mislead<strong>in</strong>g, as that<br />

<strong>in</strong>dicate a po<strong>in</strong>t where the secondary electron emission is unity <strong>and</strong> the<br />

24


Voltage [V]<br />

10 4<br />

10 3<br />

10 2<br />

10 0<br />

10 1<br />

Frequency − Gap product [GHz⋅mm]<br />

Figure 2.11: <strong>Multipactor</strong> charts based on the same parameters as <strong>in</strong> Fig. 2.3.<br />

The solid l<strong>in</strong>e zones <strong>in</strong>dicate the zones with<strong>in</strong> which phasefocus<strong>in</strong>g<br />

is active. The dashed l<strong>in</strong>e zone is produced by <strong>in</strong>clud<strong>in</strong>g<br />

also unstable phases until the non-return<strong>in</strong>g electron limit.<br />

phase is very unstable, thus mak<strong>in</strong>g a discharge impossible. A more<br />

correct boundary would be a rounded shape, which starts <strong>in</strong> the lower<br />

left corner <strong>of</strong> the solid l<strong>in</strong>e zone <strong>and</strong> smoothly jo<strong>in</strong>s the dashed left h<strong>and</strong><br />

side [47]. This is confirmed by experiments [30,48], cf. Fig. 2.12, which<br />

shows measurement data from one <strong>of</strong> the early multipactor experiments<br />

by Hatch <strong>and</strong> Williams [48]. A similar rounded shape can also be seen <strong>in</strong><br />

numerical simulations <strong>and</strong> examples <strong>of</strong> this is shown <strong>in</strong> chapter 5, which<br />

<strong>in</strong>cludes PIC simulations <strong>of</strong> multipactor <strong>in</strong> a coaxial l<strong>in</strong>e.<br />

2.2 Multicarrier<br />

Modern satellites operate <strong>in</strong> multicarrier mode, i.e. several signals at different<br />

frequencies exist simultaneously <strong>in</strong> the microwave <strong>and</strong> electronic<br />

systems. An example <strong>of</strong> such a system is Sirius 3, which is one <strong>of</strong> the<br />

Nordic satellite [49]. It has 15 channels <strong>in</strong> the frequency range 11.7 -<br />

12.5 GHz <strong>and</strong> each channel has a b<strong>and</strong>width <strong>of</strong> 33 MHz. Assume that<br />

each channel has a power <strong>of</strong> 200 W. Then the maximum <strong>in</strong>stantaneous<br />

power <strong>of</strong> the system, the peak power, is equal to 45 kW. The peak power<br />

<strong>in</strong>creases with the square <strong>of</strong> the number <strong>of</strong> carriers. Such a high <strong>in</strong>stan-<br />

25


Figure 2.12: <strong>Multipactor</strong> experiment [48] show<strong>in</strong>g the expected rounded <strong>of</strong>f<br />

lower left corner <strong>of</strong> the first multipactor zone [47].<br />

26


Amplitude [V]<br />

15<br />

10<br />

5<br />

0<br />

0 5 10 15 20<br />

Time [ns]<br />

Figure 2.13: In-phase multicarrier<br />

signal. The signal oscillates<br />

rapidly, which<br />

makes the signal envelope<br />

appear clearly <strong>in</strong><br />

this time resolution.<br />

Amplitude [V]<br />

15<br />

10<br />

5<br />

0<br />

0 5 10 15 20<br />

Time [ns]<br />

Figure 2.14: R<strong>and</strong>om phase multicarrier<br />

signal.<br />

taneous power is very unlikely <strong>in</strong> a real system s<strong>in</strong>ce it will occur only<br />

when all the signals are <strong>in</strong> phase as illustrated by Fig. 2.13. The most<br />

likely scenario, if the carriers are not phase locked, is that the phase <strong>of</strong><br />

each carrier is a r<strong>and</strong>om number <strong>and</strong> will result <strong>in</strong> a signal with much<br />

lower maximum <strong>in</strong>stantaneous power as illustrated <strong>in</strong> Fig. 2.14.<br />

The signals <strong>in</strong> Figs 2.13 <strong>and</strong> 2.14 are characterised by all carriers<br />

hav<strong>in</strong>g the same amplitude <strong>and</strong> a constant frequency. Consider a signal<br />

with N carriers, each carrier hav<strong>in</strong>g the same amplitude E0, but different<br />

phases φn <strong>and</strong> with a frequency spac<strong>in</strong>g ∆f. The period <strong>of</strong> the envelope<br />

will then be T = 1/∆f <strong>and</strong> the envelope is given by<br />

Eenv = E0<br />

�<br />

��<br />

� N−1<br />

�<br />

�<br />

cos (n2π∆ft + φn)<br />

n=0<br />

�2<br />

+<br />

� N−1<br />

�<br />

n=0<br />

s<strong>in</strong>(n2π∆ft + φn)<br />

(2.25)<br />

A more realistic signal would have different amplitudes for each carrier<br />

<strong>and</strong> the frequency spac<strong>in</strong>g would not be constant. The envelope <strong>of</strong><br />

such a signal can be found from<br />

�<br />

��<br />

� N−1<br />

Eenv =<br />

�<br />

�<br />

En cos (knω0t + φn)<br />

n=0<br />

�2<br />

+<br />

� N−1<br />

�<br />

n=0<br />

En s<strong>in</strong>(knω0t + φn)<br />

�2<br />

�2<br />

(2.26)<br />

27


where kn is a factor determ<strong>in</strong><strong>in</strong>g the frequency spac<strong>in</strong>g<br />

kn = fn/f0 − 1, n = 0,1,...,N − 1 , (2.27)<br />

f0 is the lowest carrier frequency <strong>and</strong> ω0 = 2πf0. When assess<strong>in</strong>g the<br />

worst case scenario from the multipactor po<strong>in</strong>t <strong>of</strong> view, it is important<br />

to study a whole envelope period. For arbitrarily spaced frequencies, the<br />

envelope period, T, can be found by solv<strong>in</strong>g the follow<strong>in</strong>g Diophant<strong>in</strong>e<br />

systems <strong>of</strong> equations:<br />

T = ni<br />

, ni ∈ N i = 1,2,...,N − 1 (2.28)<br />

∆fi<br />

where N is the number <strong>of</strong> carriers, f0 is the signal with the lowest frequency<br />

<strong>and</strong> ∆fi = fi − f0. The envelope period will be the solution<br />

with the smallest possible <strong>in</strong>tegers. For equally spaced carriers, the solution<br />

becomes n1 = 1, n2 = 2,...,nN−1 = N − 1, which is implies that<br />

T = 1/∆f, like before.<br />

When study<strong>in</strong>g multicarrier multipactor it is common to make certa<strong>in</strong><br />

simplifications that will allow us<strong>in</strong>g s<strong>in</strong>gle carrier methodology to<br />

asses also the multicarrier case, e.g. the mean frequency <strong>of</strong> all the carriers<br />

is used as the design frequency. Thus, most <strong>of</strong> what has been said<br />

about s<strong>in</strong>gle carrier multipactor will then be valid also for the multiple<br />

signals case.<br />

2.3 Design guidel<strong>in</strong>es<br />

From an <strong>in</strong>dustrial po<strong>in</strong>t <strong>of</strong> view it is important not only to underst<strong>and</strong><br />

the physics <strong>of</strong> multipactor, but also how the theoretical <strong>and</strong> experimental<br />

results should be applied when mak<strong>in</strong>g multipactor-free microwave<br />

hardware designs. In Europe, most space hardware designers follow the<br />

st<strong>and</strong>ard issued by ESA [50]. This st<strong>and</strong>ard <strong>in</strong>cludes both the s<strong>in</strong>gle<br />

<strong>and</strong> the multicarrier cases, but for the latter it is stated that the design<br />

guidel<strong>in</strong>es are only recommendations. Most research support these<br />

recommendations, but not enough tests have been performed to verify<br />

the theoretical f<strong>in</strong>d<strong>in</strong>gs. When us<strong>in</strong>g the st<strong>and</strong>ard it is important to be<br />

aware <strong>of</strong> the fact that it is primarily based on the parallel-plate model<br />

with a uniform electric field. Design with respect to this approach for<br />

other geometries is normally a conservative <strong>and</strong> safe way. However, <strong>in</strong><br />

many common microwave structures, the geometry is such that losses<br />

<strong>of</strong> electrons is much higher than <strong>in</strong> the parallel-plate case. Thus the<br />

28


multipactor threshold <strong>in</strong> geometries such as coaxial l<strong>in</strong>es, waveguides<br />

<strong>and</strong> irises, can be higher or much higher than that obta<strong>in</strong>ed us<strong>in</strong>g the<br />

plane-parallel model.<br />

2.3.1 S<strong>in</strong>gle carrier<br />

In the ESA st<strong>and</strong>ard [50] components are divided <strong>in</strong>to three categories<br />

or types. Type 1 is a well vented component where all RF paths are<br />

metallic <strong>and</strong> the secondary electron emission properties are well known.<br />

This type <strong>of</strong> component has the lowest design marg<strong>in</strong>s with respect to<br />

multipactor <strong>and</strong>, depend<strong>in</strong>g on the type <strong>of</strong> test, range from 3-8 dB.<br />

The second type <strong>of</strong> component may conta<strong>in</strong> dielectrics with established<br />

multipactor properties <strong>and</strong> the component should be well vented. Also<br />

depend<strong>in</strong>g on the type <strong>of</strong> test, the design marg<strong>in</strong>s range from 3-10 dB.<br />

All other components are categorised as type 3 <strong>and</strong> the design marg<strong>in</strong>s<br />

range from 4-12 dB.<br />

When design<strong>in</strong>g with respect to multipactor a complete electric field<br />

analysis is performed <strong>and</strong> regions with high voltages <strong>and</strong> critical gap<br />

sizes are identified. Us<strong>in</strong>g the frequency-gap size product, the multipactor<br />

threshold can be found <strong>in</strong> a susceptibility chart for the material<br />

<strong>in</strong> question. A susceptibility chart <strong>in</strong> the ESA st<strong>and</strong>ard is basically an<br />

envelope <strong>of</strong> the multipactor zones as shown <strong>in</strong> e.g. Fig. 2.5. If a marg<strong>in</strong><br />

larger than the largest design marg<strong>in</strong>, 12 dB, is found, no test<strong>in</strong>g is<br />

required. However, <strong>in</strong> most cases, the component will have to be tested<br />

<strong>and</strong> methods for detect<strong>in</strong>g multipactor will be discussed <strong>in</strong> chapter 6.<br />

2.3.2 Multicarrier<br />

In the multicarrier case, only components <strong>of</strong> type 1 are covered by the<br />

recommendations given <strong>in</strong> the ESA st<strong>and</strong>ard. Type 2 <strong>and</strong> type 3 components<br />

will require further research before they can be <strong>in</strong>cluded <strong>in</strong> the<br />

st<strong>and</strong>ard. In the s<strong>in</strong>gle carrier case, the level that is compared with the<br />

multipactor threshold <strong>in</strong> the susceptibility chart is the amplitude <strong>of</strong> the<br />

signal <strong>and</strong> no ambiguities exist. For multicarrier designs, the traditional<br />

way <strong>of</strong> design<strong>in</strong>g was to set the design marg<strong>in</strong> with respect to the peak<br />

power <strong>of</strong> <strong>in</strong>-phase carriers, shown <strong>in</strong> Fig. 2.13. This design method is<br />

still allowed by the ESA st<strong>and</strong>ard <strong>and</strong> for type 1 components the design<br />

marg<strong>in</strong>s range from 0-6 dB depend<strong>in</strong>g on the type <strong>of</strong> test<strong>in</strong>g that will<br />

be performed. However, as previously mentioned, <strong>in</strong>-phase carriers for<br />

non-phase locked signals is extremely unlikely <strong>and</strong> thus the st<strong>and</strong>ard<br />

29


allows for another design marg<strong>in</strong>, which is set with respect to the so<br />

called P20 power level. The P20 level corresponds to the “peak power <strong>of</strong><br />

the multicarrier waveform whose width at the s<strong>in</strong>gle carrier multipaction<br />

threshold is equal to the time taken for the electrons to cross the multipact<strong>in</strong>g<br />

region 20 times” [50]. This level is illustrated <strong>in</strong> Fig. 2.15.<br />

Figure 2.15: An example where the <strong>in</strong>-phase peak power is above the s<strong>in</strong>gle<br />

carrier threshold, while the P20 level is more than 4 dB below the<br />

same threshold. The peak voltage is 128.4 V, the s<strong>in</strong>gle carrier<br />

threshold is 91 V, <strong>and</strong> the P20 voltage is 57 V. Signal data: 12<br />

carriers, equally spaced, fm<strong>in</strong> = 1.545 GHz, ∆f = 24 MHz <strong>and</strong><br />

each carrier amplitude is 10.7 V. Material properties: W1 =<br />

23 eV <strong>and</strong> σse,max = 3.<br />

In the case when a design is made with respect to the P20 level, the<br />

design marg<strong>in</strong>s range from 4-6 dB depend<strong>in</strong>g on the type <strong>of</strong> test<strong>in</strong>g. A<br />

problem with the P20 level is that it is not a trivial problem to f<strong>in</strong>d the<br />

peak power level for 20 electron gap cross<strong>in</strong>gs. This power level is usually<br />

referred to as the worst case scenario, even though it may not always be<br />

the worst case from a multipactor po<strong>in</strong>t <strong>of</strong> view. A number <strong>of</strong> different<br />

30


ways <strong>of</strong> f<strong>in</strong>d<strong>in</strong>g the worst case scenario have been proposed, e.g. us<strong>in</strong>g<br />

parabolic or triangular phase distribution <strong>in</strong> the equally spaced carriers<br />

case (cf. Ref. [51]). Some <strong>of</strong> the better methods for f<strong>in</strong>d<strong>in</strong>g the worst<br />

case scenario will be described <strong>in</strong> the follow<strong>in</strong>g subsections after a brief<br />

discussion about the 20 gap cross<strong>in</strong>gs rule (TGR).<br />

Twenty gap cross<strong>in</strong>gs rule<br />

The TGR was proposed <strong>in</strong> Ref. [14] <strong>in</strong> 1997 <strong>and</strong> <strong>in</strong> its orig<strong>in</strong>al version<br />

it reads:<br />

“As long as the duration <strong>of</strong> the multicarrier peak <strong>and</strong> the mode order<br />

<strong>of</strong> the gap are such that no more than twenty gap-cross<strong>in</strong>gs can occur<br />

dur<strong>in</strong>g the multicarrier peak, then multipaction-generated noise should<br />

rema<strong>in</strong> well below thermal noise (<strong>in</strong> a 30 MHz b<strong>and</strong>).” [14]<br />

The rule is a result <strong>of</strong> an analysis <strong>of</strong> simulated multi-carrier multipactor<br />

discharges. Comparison with experiments showed great deviations,<br />

where the simulated noise could be as much as 75 dB greater<br />

than the measured noise level. In the experiments, a m<strong>in</strong>imum <strong>of</strong> 99<br />

gap-cross<strong>in</strong>gs were required before the produced noise was detectable<br />

above the noise floor <strong>of</strong> -70 dBm. Of course, there may be bit errors<br />

even at lower noise levels, but as the number <strong>of</strong> electrons grows exponentially<br />

with the number <strong>of</strong> gap cross<strong>in</strong>gs (see Eq. (2.22), there is a<br />

huge difference between 20 <strong>and</strong> 99 gap-cross<strong>in</strong>gs.<br />

However, the TGR is certa<strong>in</strong>ly a good first attempt to lower the<br />

requirements for multi-carrier multipactor. It is a fairly conservative<br />

method <strong>and</strong> thus the risk <strong>of</strong> apply<strong>in</strong>g it should be quite limited. However,<br />

more appropriate guidel<strong>in</strong>es should be based on an unambiguous<br />

theoretical concept, which can take the material properties <strong>in</strong>to account.<br />

Then, when perform<strong>in</strong>g simulations <strong>and</strong> experiments to verify the idea,<br />

it is <strong>of</strong> paramount importance to make sure that the actual material<br />

properties <strong>of</strong> the test samples are well known <strong>and</strong> that these properties<br />

are also be<strong>in</strong>g used <strong>in</strong> the simulations. Due to the large difference<br />

<strong>in</strong> secondary emission properties between different materials, it would<br />

seem reasonable that for a material with a low σse,max one would allow<br />

more gap cross<strong>in</strong>gs than, <strong>in</strong> the opposite case, for a material with a high<br />

σse,max.<br />

31


Boundary function Method<br />

One <strong>of</strong> the best eng<strong>in</strong>eer<strong>in</strong>g methods for f<strong>in</strong>d<strong>in</strong>g the worst case scenario<br />

for equally spaced carriers, the boundary function method, was orig<strong>in</strong>ally<br />

designed by Wolk et al. [51]. Unfortunately the used function was<br />

found empirically while study<strong>in</strong>g the worst case scenario with an optimisation<br />

tool, <strong>and</strong> is thus not physically founded. A consequence <strong>of</strong> this is<br />

that under certa<strong>in</strong> circumstances, the boundary function produces poor<br />

results. It was also limited to work only for equally spaced carriers. However,<br />

as part <strong>of</strong> the present thesis work, this method has been further<br />

developed, <strong>and</strong> it has been found that the orig<strong>in</strong>al boundary function<br />

approximately describes a function that tries to squeeze all the energy<br />

<strong>of</strong> the multicarrier signal dur<strong>in</strong>g one envelope period <strong>in</strong>to a specified,<br />

shorter, time period. This works just as well <strong>in</strong> both the equally spaced<br />

<strong>and</strong> the non-equally spaced carrier cases <strong>and</strong> can be summarized by the<br />

follow<strong>in</strong>g formulas:<br />

⎧ �<br />

FV (TX) =<br />

⎪⎨<br />

FV,max = N�<br />

⎪⎩ FV,m<strong>in</strong> =<br />

TH<br />

TX<br />

Ei<br />

i=1<br />

� N�<br />

E<br />

i=1<br />

2 i<br />

N�<br />

E<br />

i=1<br />

2 i<br />

(2.29)<br />

Here TX is the time period <strong>of</strong> <strong>in</strong>terest, which is <strong>of</strong>ten set to T20, i.e.<br />

the time it takes the electrons to traverse the gap 20 times. TH is the<br />

period <strong>of</strong> the envelope <strong>and</strong> Ei is the voltage amplitude <strong>of</strong> each carrier.<br />

FV (TX) is the design voltage <strong>and</strong> is shown as two symmetric curved l<strong>in</strong>es<br />

<strong>in</strong> Fig. 2.15. The design voltage can never exceed the <strong>in</strong>-phase voltage,<br />

given by FV,max, <strong>and</strong> if all power is distributed evenly over the entire<br />

envelope period the voltage amplitude will be FV,m<strong>in</strong>, which is <strong>in</strong>dicated<br />

by a dashed l<strong>in</strong>e <strong>in</strong> Fig. 2.15.<br />

The ma<strong>in</strong> advantage with the boundary function method is its simplicity.<br />

It is also very reliable, although a little conservative <strong>and</strong> this is<br />

especially true for non-equally spaced carriers, where the P20 level can<br />

be much lower than FV . The method has been implemented as an auxiliary<br />

method <strong>in</strong> WCAT, which is a s<strong>of</strong>tware tool orig<strong>in</strong>ally developed by<br />

the present author <strong>and</strong> Genrong Li as part <strong>of</strong> a Master’s Thesis [52] at<br />

32


Saab Ericsson Space. It has s<strong>in</strong>ce been upgraded with additional functionality<br />

by the present author as well as by Mariusz Merecki as part <strong>of</strong><br />

a Master’s Thesis [40] at Centre National d’ Études Spatiales, Toulouse,<br />

France. Fig. 2.16 shows the graphical user <strong>in</strong>terface <strong>of</strong> the present version<br />

<strong>of</strong> WCAT <strong>and</strong> an example when the worst case <strong>of</strong> non-equally spaced<br />

carriers have been assessed us<strong>in</strong>g the built <strong>in</strong> genetic algorithm.<br />

200<br />

150<br />

100<br />

50<br />

Threshold =192<br />

Hyb. Threshold =158<br />

Env Threshold =139<br />

Boundary Function<br />

FVm<strong>in</strong><br />

In−Phase Envelope<br />

Envelope<br />

← Fv =129<br />

10 3<br />

10 3<br />

10 2<br />

10 2<br />

10 1<br />

10 1<br />

0<br />

0 5 10 15 20 25 30 35 40 100<br />

0 5 10 15 20 25 30 35 40 100<br />

Figure 2.16: Assessment <strong>of</strong> worst case scenario for multicarrier multipactor<br />

us<strong>in</strong>g WCAT, Worst Case Assessment Tool. The example shows<br />

a case with 10 non-equally spaced carriers with vary<strong>in</strong>g amplitude.<br />

Optimisation methods<br />

A more direct method <strong>of</strong> f<strong>in</strong>d<strong>in</strong>g the worst case scenario is to use some<br />

k<strong>in</strong>d <strong>of</strong> optimisation tool. In WCAT several different methods <strong>of</strong> f<strong>in</strong>d<strong>in</strong>g<br />

the worst case scenario are implemented. One uses the non-l<strong>in</strong>ear least<br />

square (NLSQ) functionality <strong>of</strong> Matlab to f<strong>in</strong>d the set <strong>of</strong> phases that will<br />

fit as much energy as possible <strong>in</strong>side a time period TX ≤ TH, where TH<br />

is the envelope period. Another method generates a variable number <strong>of</strong><br />

sets <strong>of</strong> r<strong>and</strong>om phases <strong>and</strong> the correspond<strong>in</strong>g envelopes are compared to<br />

f<strong>in</strong>d the worst case. This method is better than the NLSQ optimisation<br />

method when it comes to f<strong>in</strong>d<strong>in</strong>g the worst case for non-equally spaced<br />

33


carriers, because the NLSQ method requires a good seed phase <strong>in</strong> order<br />

to f<strong>in</strong>d a good local m<strong>in</strong>imum. By comb<strong>in</strong><strong>in</strong>g these methods <strong>and</strong> us<strong>in</strong>g<br />

the result <strong>of</strong> the best r<strong>and</strong>om phase approach as a seed phase for the<br />

NLSQ optimisation, the best results are found. The problem with a<br />

r<strong>and</strong>om phase approach is that for a large number <strong>of</strong> carriers, the number<br />

<strong>of</strong> phase-sets needed to achieve a good result becomes discourag<strong>in</strong>gly<br />

high [52].<br />

By us<strong>in</strong>g a genetic algorithm, a great improvement <strong>in</strong> f<strong>in</strong>d<strong>in</strong>g the<br />

worst case scenario <strong>in</strong> a short time has been achieved, especially for the<br />

non-equally spaced carrier case. This type <strong>of</strong> optimisation scheme has<br />

the advantage <strong>of</strong> be<strong>in</strong>g able to f<strong>in</strong>d not only a good local m<strong>in</strong>imum as<br />

<strong>in</strong> the NLSQ case, but the actual global m<strong>in</strong>imum can be found. The<br />

genetic algorithm was implemented by Merecki [40] <strong>and</strong> <strong>in</strong> addition to<br />

improv<strong>in</strong>g the optimisation part <strong>of</strong> the s<strong>of</strong>tware, he also implemented<br />

the threshold <strong>of</strong> the hybrid modes (see Fig. 2.16) as well as many other<br />

useful functions.<br />

The ma<strong>in</strong> problem with multipactor <strong>in</strong> microwave systems is the<br />

electric noise that is generated, which degrades the signal to noise ratio.<br />

Thus the worst case may not always be the maximum power with<strong>in</strong> the<br />

T20 period. Depend<strong>in</strong>g on the material properties some other case may<br />

produce a substantially larger amount <strong>of</strong> energetic electrons. Therefore,<br />

WCAT also <strong>in</strong>cludes the possibility <strong>of</strong> f<strong>in</strong>d<strong>in</strong>g the phase-set that will<br />

produce the largest amount <strong>of</strong> electrons.<br />

In addition to assess<strong>in</strong>g the risk for a vacuum discharge, it can also<br />

be <strong>of</strong> value to <strong>in</strong>vestigate if a multipactor free design may have a risk<br />

<strong>of</strong> corona breakdown if the component is not thoroughly vented when<br />

it is brought <strong>in</strong>to operation. In WCAT this is analysed us<strong>in</strong>g the mean<br />

carrier frequency <strong>and</strong> compar<strong>in</strong>g the corona threshold with the <strong>in</strong>-phase<br />

peak voltage. If the m<strong>in</strong>imum <strong>of</strong> the Paschen curve is greater than this<br />

voltage, then the corona marg<strong>in</strong> is displayed <strong>in</strong> the output w<strong>in</strong>dow <strong>of</strong><br />

WCAT. If the opposite is true, the pressure range with<strong>in</strong> which there is<br />

risk for corona discharge is presented (see Fig. 2.16).<br />

34


Chapter 3<br />

<strong>Multipactor</strong> <strong>in</strong> low pressure<br />

gas<br />

Microwave discharges can occur <strong>in</strong> both gas <strong>and</strong> vacuum. In vacuum, the<br />

phenomenon is usually called multipaction or multipactor <strong>and</strong> the theory<br />

for such vacuum microwave breakdown was presented <strong>in</strong> the previous<br />

chapter. In a gas it is normally called corona discharge or gas breakdown<br />

<strong>and</strong> it can occur when the electron mean free path between collisions<br />

with molecules is smaller than the characteristic dimensions <strong>of</strong> the vessel.<br />

An applied microwave electric field can widen the velocity distribution<br />

<strong>of</strong> the free electrons <strong>and</strong> thus make more electrons energetic enough to<br />

ionise the gas. If the production <strong>of</strong> electrons exceeds the loss through<br />

diffusion, attachment, <strong>and</strong> recomb<strong>in</strong>ation, the electron density will grow<br />

exponentially <strong>and</strong> microwave gas breakdown will occur.<br />

When the mean free path between collisions is <strong>of</strong> the same order as<br />

the device dimensions, classical theory for microwave gas <strong>and</strong> vacuum<br />

discharge can not be used. Diffusion loss can no longer be assumed,<br />

like <strong>in</strong> the gas breakdown case, s<strong>in</strong>ce that requires a mean free path<br />

several times shorter than the characteristic length <strong>of</strong> the component.<br />

Nevertheless, the electrons will meet a resistance due to collisions with<br />

the neutral gas molecules <strong>and</strong> thus pure vacuum can not be assumed<br />

either.<br />

<strong>Low</strong> pressure multipactor has received comparatively little attention.<br />

However, a few studies, theoretical as well as experimental, have revealed<br />

some parts <strong>of</strong> the complicated picture. Vender et al. [17] performed<br />

PIC-simulations to study the electron density development <strong>and</strong> showed<br />

that at sufficiently low pressures, the gas discharge is <strong>in</strong>itiated by a<br />

35


multipactor discharge. Us<strong>in</strong>g a Monte Carlo algorithm, Gilard<strong>in</strong>i [53]<br />

made quite a general study <strong>of</strong> the phenomenon <strong>and</strong> presented breakdown<br />

voltages normalised to the first cross-over po<strong>in</strong>t <strong>of</strong> the material for a<br />

wide range <strong>of</strong> dimensionless variables. He also paid special attention<br />

to a particular <strong>and</strong> realistic case, namely multipactor <strong>in</strong> low pressure<br />

argon [54]. This was done partly <strong>in</strong> an effort to compare the simulations<br />

with the experimental results <strong>of</strong> Höhn et al. [18].<br />

In paper B <strong>of</strong> this thesis, low pressure multipactor was studied us<strong>in</strong>g<br />

an analytical model that takes <strong>in</strong>to account only the friction force due<br />

to collisions between the electrons <strong>and</strong> the neutral gas particles. The<br />

ma<strong>in</strong> theory <strong>and</strong> results from this study will be presented <strong>in</strong> the first<br />

section below. In addition to the friction force, the collisions will also<br />

cause a r<strong>and</strong>om velocity spread <strong>of</strong> the electrons that results <strong>in</strong> a higher<br />

average impact energy. Furthermore, due to the long distance between<br />

molecules, the electrons are free to accelerate to very high velocities <strong>and</strong><br />

upon impact with a gas molecule or atom the energy is sufficient to cause<br />

ionisation. In paper C <strong>of</strong> this thesis a more detailed analysis has been<br />

done, where all these effects have been considered <strong>and</strong> the used model<br />

as well some highlights from the results are presented <strong>in</strong> the section<br />

“Advanced Model” below.<br />

3.1 Simple Model<br />

In a first attempt to underst<strong>and</strong> the behaviour <strong>of</strong> multipactor <strong>in</strong> a low<br />

pressure gas, a simple analytical model was used, which takes only the<br />

friction force <strong>of</strong> the collisions with neutrals <strong>in</strong>to account. By deriv<strong>in</strong>g explicit<br />

expressions for the multipactor threshold, qualitative comparison<br />

with experimental results [18] as well as results from computer simulations<br />

[53,54] could be made.<br />

3.1.1 Model<br />

The differential equation govern<strong>in</strong>g the behaviour <strong>of</strong> the electrons <strong>in</strong><br />

a low pressure gas is given by the equation <strong>of</strong> motion, Eq. (2.1), but<br />

augmented to <strong>in</strong>clude also the effects <strong>of</strong> collisions:<br />

m¨x = eE − mνc ˙x (3.1)<br />

where νc = σcn0v is the collision frequency between the free electrons<br />

<strong>and</strong> the neutral particles. σc is the collision cross-section, n0 the neutral<br />

36


gas density <strong>and</strong> v the electron velocity. The collision cross-section is<br />

generally a function <strong>of</strong> the electron velocity, but <strong>in</strong> order to avoid a<br />

non-l<strong>in</strong>ear differential equation, σc is assumed to be a constant.<br />

As <strong>in</strong> the vacuum case, a spatially uniform harmonic field E =<br />

E0 s<strong>in</strong> ωt is assumed. Solv<strong>in</strong>g Eq. (3.1) with the same <strong>in</strong>itial conditions<br />

as for the vacuum case, i.e. that an electron is emitted from x = 0 with<br />

an <strong>in</strong>itial velocity v0 when t = α/ω, the position <strong>and</strong> velocity <strong>of</strong> the<br />

electron can be found:<br />

where<br />

x = 1<br />

νc<br />

α<br />

νc(<br />

(1 − e ω −t) )(v0 + Λ[ω cos α − νc s<strong>in</strong> α])<br />

+ Λ<br />

ω [ω(s<strong>in</strong> α − s<strong>in</strong>(ωt)) + νc(cos α − cos(ωt))] (3.2)<br />

α<br />

νc(<br />

˙x = v0e ω −t) α<br />

νc(<br />

+ Λ[e ω −t) (ω cos α − νc s<strong>in</strong> α)<br />

− ω cos(ωt) + νc s<strong>in</strong>(ωt)] (3.3)<br />

Λ =<br />

eE0<br />

m(ω 2 + ν 2 c )<br />

(3.4)<br />

The resonance condition requires that an electron emitted when t =<br />

α/ω should reach the other electrode, at x = d, when ωt = Nπ + α,<br />

where N is an odd positive <strong>in</strong>teger as <strong>in</strong> the vacuum case. Apply<strong>in</strong>g<br />

this condition to Eqs. (3.2) <strong>and</strong> (3.3) yields expressions for the required<br />

electric field <strong>and</strong> the impact velocity:<br />

E0 =<br />

m<br />

e (ω2 + ν 2 c)(d + v0<br />

νc<br />

(e− Nπνc<br />

ω − 1))<br />

Nπνc<br />

Nπνc<br />

− − (1 + e ω )s<strong>in</strong> α + ((1 − e ω ) ω<br />

νc<br />

2νc + ω )cos α<br />

(3.5)<br />

Nπνc<br />

Nπνc<br />

− −<br />

vimpact = v0e ω + Λ(1 + e ω )(ω cos α − νc s<strong>in</strong> α) (3.6)<br />

In order to draw the multipactor boundaries, an expression for the<br />

non-return<strong>in</strong>g electron limit is needed. Like <strong>in</strong> the vacuum case no explicit<br />

analytical expression for this can be found <strong>and</strong> thus the limit will<br />

be obta<strong>in</strong>ed numerically <strong>in</strong>stead. However, as a rough approximation,<br />

the limit given by Eq. (2.11) can used.<br />

3.1.2 <strong>Multipactor</strong> boundaries<br />

When construct<strong>in</strong>g only the lower multipactor threshold <strong>in</strong> vacuum,<br />

which depends on the first cross-over energy, the threshold value under<br />

the assumption <strong>of</strong> a constant <strong>in</strong>itial velocity is the same as with<br />

37


the assumption <strong>of</strong> a constant ratio k = vimpact/v0 between impact <strong>and</strong><br />

<strong>in</strong>itial velocities. This is true also <strong>in</strong> the presence <strong>of</strong> collisions, when the<br />

above simple model is used <strong>and</strong> thus the constant k approach will be<br />

used <strong>in</strong> the follow<strong>in</strong>g expressions. Comb<strong>in</strong><strong>in</strong>g Eqs. (3.5) <strong>and</strong> 3.6 under<br />

this assumption yields an expression for the resonant phase, viz.<br />

tan α = ω2 [βΦ + γ] + 2ν 2 cvimpact(Φ − k)<br />

ξω(Φ + 1)<br />

(3.7)<br />

where Φ = exp(−νcNπ/ω), β = kdνc + (k + 1)vimpact, γ = kdνc −<br />

vimpact(k + 1), <strong>and</strong> ξ = [kdνc + (k − 1)vimpact]νc.<br />

Equation (3.7) can be used together with Eq. (3.5) to plot the multipactor<br />

threshold <strong>in</strong> a low pressure gas as a function <strong>of</strong> gap size, pressure,<br />

or frequency. However, by multiply<strong>in</strong>g the expression for the amplitude<br />

<strong>of</strong> the electric field with the gap size, d, an expression for the voltage as a<br />

function <strong>of</strong> the frequency-gap size <strong>and</strong> the pressure-gap size products can<br />

be obta<strong>in</strong>ed. This approach will be used <strong>in</strong> a subsequent section, where<br />

a more advanced model is used to analyse the phenomenon. Figure 3.1<br />

shows the lower multipactor threshold <strong>in</strong> low pressure air for different<br />

pressures. The graphs are based on Eqs. (3.5) <strong>and</strong> (3.7) only <strong>and</strong> do not<br />

consider the non-return<strong>in</strong>g electron limit nor the phase stability limits.<br />

From Fig. 3.1 it is clear that the multipactor threshold <strong>in</strong>creases<br />

with <strong>in</strong>creas<strong>in</strong>g pressure. By sweep<strong>in</strong>g the pressure <strong>in</strong>stead <strong>of</strong> the frequency<br />

<strong>and</strong> compar<strong>in</strong>g the chang<strong>in</strong>g threshold with the corona breakdown<br />

threshold, an underst<strong>and</strong><strong>in</strong>g can be obta<strong>in</strong>ed <strong>of</strong> how the transition<br />

between these two types <strong>of</strong> discharges can occur. Fig. 3.2 shows that<br />

the multipactor threshold first <strong>in</strong>creases until a certa<strong>in</strong> po<strong>in</strong>t, where it<br />

<strong>in</strong>tersects the curve correspond<strong>in</strong>g to the corona threshold. It will then<br />

follow this curve towards the m<strong>in</strong>imum <strong>of</strong> the Paschen curve. This is<br />

just a qualitative picture <strong>and</strong> the sharp <strong>in</strong>tersection would <strong>in</strong> reality be<br />

a smooth transition.<br />

Figure 3.1 does not consider the non-return<strong>in</strong>g electron limit, nor<br />

the phase stability limits. As expla<strong>in</strong>ed <strong>in</strong> the previous chapter, phase<br />

focus<strong>in</strong>g is needed to ma<strong>in</strong>ta<strong>in</strong> the generated electrons <strong>in</strong> a close bunch,<br />

s<strong>in</strong>ce an electron with a too large phase error will be lost. By <strong>in</strong>troduc<strong>in</strong>g<br />

a small phase error <strong>in</strong> Eq. (3.2) <strong>and</strong> keep<strong>in</strong>g the amplitude <strong>and</strong> phase<br />

constant while sett<strong>in</strong>g x = d, the error after the passage can be found.<br />

The ratio between the f<strong>in</strong>al <strong>and</strong> <strong>in</strong>itial error is the stability factor, G,<br />

<strong>and</strong> when the absolute value <strong>of</strong> this factor is less than one, the phase<br />

focus<strong>in</strong>g effect is active. With the present model, the expression for the<br />

38


Voltage (V)<br />

10 3<br />

10 2<br />

10 1<br />

Multipaction threshold curves <strong>in</strong> low pressure air (d=0.1 m)<br />

p=0.1 Pa<br />

p=1 Pa<br />

p=10 Pa<br />

Vacuum<br />

10 −3<br />

Frequency (GHz)<br />

Figure 3.1: <strong>Multipactor</strong> chart show<strong>in</strong>g the lower multipactor threshold at<br />

different pressures. i.e. each curve is based on an impact velocity<br />

correspond<strong>in</strong>g to W1 = 23 eV. Phase stability <strong>and</strong> the<br />

non-return<strong>in</strong>g electron limit are not considered, only resonance.<br />

Parameters used are: σc = 6.9 × 10 −20 m 2 , k = 2.5, N = 1 <strong>and</strong><br />

d = 0.1 m.<br />

10 −2<br />

39


Voltage (V)<br />

10 3<br />

10 2<br />

10 1<br />

10 −2<br />

Threshold curves <strong>in</strong> low pressure air (fd=1GHz ⋅ mm)<br />

<strong>Multipactor</strong> <strong>in</strong> air<br />

<strong>Multipactor</strong> <strong>in</strong> vacuum<br />

Corona <strong>in</strong> air<br />

10 −1<br />

10 0<br />

<strong>Pressure</strong> (Pa)<br />

Figure 3.2: Thresholds for multipactor <strong>and</strong> corona discharges <strong>in</strong> air as functions<br />

<strong>of</strong> pressure together with the multipactor vacuum threshold<br />

as a reference level. Parameters used are: σc = 6.9 × 10 −20 m 2 ,<br />

W1 = 23 eV, k = 2.5, N = 1, f × d =1 GHz·mm <strong>and</strong> d = 0.1 m.<br />

40<br />

10 1<br />

10 2


stability factor becomes:<br />

G = (νc<br />

ω<br />

ω<br />

(CΦ − 1) + C (Φ − 1))tan α + 1 − C<br />

νc<br />

ω<br />

νc<br />

(1 − CΦ)tan α − (CΦ + 1)<br />

(3.8)<br />

where C = (k + 1)/(k − Φ). In Fig. 3.3 the phase limits, where |G| = 1,<br />

have been plotted together with the non-return<strong>in</strong>g electron limit. Both<br />

the positive <strong>and</strong> negative phase error limits tend to decrease with <strong>in</strong>creas<strong>in</strong>g<br />

pressure. However, the limit for non-return<strong>in</strong>g electrons <strong>in</strong>creases,<br />

which is a more important limit when the electron impact energy exceeds<br />

the first cross-over energy, thus reduc<strong>in</strong>g the width <strong>of</strong> the multipactor<br />

zone.<br />

Phase limit (degrees)<br />

40<br />

20<br />

0<br />

−20<br />

−40<br />

−60<br />

−80<br />

10 −3<br />

−100<br />

Postive phase error<br />

Negative phase error<br />

Non−ret. el. <strong>in</strong> vacuum (Semenov et. al.)<br />

Numerical non−return<strong>in</strong>g electron limit<br />

10 −2<br />

Phase Limits<br />

10 −1<br />

<strong>Pressure</strong> (Pa)<br />

Figure 3.3: Phase limits <strong>in</strong> a low pressure gas based on the simple analytical<br />

model. The dashed l<strong>in</strong>e <strong>and</strong> the solid l<strong>in</strong>e (dashed at the end)<br />

show the upper <strong>and</strong> lower phase limits beyond which a phase<br />

error will start grow<strong>in</strong>g. The dash-dot l<strong>in</strong>e is the phase below<br />

which emitted electrons will not be able to escape from the wall<br />

<strong>of</strong> emission. The dotted l<strong>in</strong>e is the phase limit obta<strong>in</strong>ed us<strong>in</strong>g<br />

Eq. (2.15) <strong>and</strong> is an approximation <strong>of</strong> the dash-dot l<strong>in</strong>e <strong>in</strong> the<br />

vacuum case. Parameters used are: σc = 6.9 × 10 −19 m 2 , N = 1,<br />

k = 7.6, d = 0.1 m, <strong>and</strong> W1 = 23 eV.<br />

10 0<br />

10 1<br />

41


3.1.3 Ma<strong>in</strong> results<br />

The ma<strong>in</strong> result found <strong>in</strong> paper B is that a higher microwave power is<br />

required to <strong>in</strong>itiate breakdown <strong>in</strong> a low pressure gas, s<strong>in</strong>ce the collisions<br />

tend to slow down the electrons. By comb<strong>in</strong><strong>in</strong>g the low pressure multipactor<br />

graph with the corona threshold curve, it was concluded that<br />

with <strong>in</strong>creas<strong>in</strong>g pressure, the required threshold will first <strong>in</strong>crease <strong>and</strong>,<br />

after reach<strong>in</strong>g a plateau, it will make a smooth transition to the low<br />

pressure branch <strong>of</strong> the Paschen curve. This behaviour is confirmed by<br />

the <strong>in</strong>vestigations made by Gilard<strong>in</strong>i [53] for materials with a low first<br />

cross-over po<strong>in</strong>t, close to the ionisation energy <strong>of</strong> the gas, <strong>and</strong> for N = 1,<br />

i.e. for the first order <strong>of</strong> multipactor.<br />

For materials with a higher first cross-over energy <strong>and</strong> for higher<br />

order multipactor, Gilard<strong>in</strong>i found no <strong>in</strong>itial <strong>in</strong>crease <strong>in</strong> the multipactor<br />

threshold, <strong>in</strong>stead a monotonically decreas<strong>in</strong>g breakdown voltage was<br />

seen. A possible explanation for the differences between the result obta<strong>in</strong>ed<br />

by the simple analytical model <strong>and</strong> the result <strong>of</strong> Gilard<strong>in</strong>i is also<br />

presented <strong>in</strong> paper B <strong>and</strong> it is suggested that the reason is that for materials<br />

with a higher W1 <strong>and</strong> for N > 1 the contribution <strong>of</strong> electrons from<br />

impact ionisation decreased the required W1 (see Fig. 3.4). However,<br />

as will be seen <strong>in</strong> the next section, electron contribution from impact<br />

ionisation is not the only reason for this behaviour. The collisions will<br />

also cause an electron velocity spread, which will result <strong>in</strong> a larger total<br />

impact velocity <strong>and</strong> thus a lower voltage is needed to achieve the<br />

necessary first cross-over energy.<br />

A comparison was made with experiments by Höhn et al. [18] <strong>in</strong> low<br />

pressure argon as well as with PIC-simulations by Gilard<strong>in</strong>i [54] <strong>in</strong> the<br />

same gas (see Fig. 7 <strong>in</strong> paper B). However, the fd-product chosen by<br />

these authors was located <strong>in</strong> the middle <strong>of</strong> the right boundary <strong>of</strong> the<br />

first multipactor zone, an area dom<strong>in</strong>ated by the hybrid modes [39].<br />

The simple model used <strong>in</strong> the presented analytical approach is not applicable<br />

to these modes <strong>and</strong> consequently the behaviour found <strong>in</strong> the<br />

experiments <strong>and</strong> simulations could not be confirmed. Furthermore, the<br />

impact energy <strong>of</strong> the electrons at this fd-product is several times higher<br />

than the ionisation energy <strong>of</strong> argon <strong>and</strong> thus a significant contribution<br />

<strong>of</strong> electrons from collisional ionisation would be expected. In addition,<br />

the required W1 would be reduced due to the electron velocity spread<br />

<strong>and</strong> thus a behaviour similar to curves (b) or (c) <strong>in</strong> Fig. 3.4 should be<br />

expected <strong>and</strong> it is also what is found.<br />

To further analyse multipactor <strong>in</strong> a low pressure gas, a better ana-<br />

42


Voltage (V)<br />

Threshold curves <strong>in</strong> a low pressure gas<br />

<strong>Multipactor</strong> <strong>in</strong> a low pressure gas without ionisation<br />

<strong>Multipactor</strong> <strong>in</strong> vacuum<br />

Corona<br />

a<br />

b<br />

c<br />

<strong>Pressure</strong> (Pa)<br />

Figure 3.4: Qualitative form <strong>of</strong> the dependence <strong>of</strong> breakdown threshold with<br />

pressure <strong>in</strong> the region between which multipactor <strong>and</strong> corona,<br />

respectively, dom<strong>in</strong>ate the breakdown process: (a) the friction<br />

force due to collisions with neutrals dom<strong>in</strong>ates, (b) electron velocity<br />

spread reduces the required W1 <strong>and</strong> collisional ionisation<br />

contributes significantly to the total number <strong>of</strong> electrons, (c) <strong>in</strong>termediate<br />

situation.<br />

43


lytical model is obviously needed. The model must, <strong>in</strong> addition to the<br />

friction force, be able to take collisional ionisation <strong>in</strong>to account as well<br />

as collision <strong>in</strong>duced velocity spread <strong>of</strong> the electrons. In the next section<br />

a more advanced model that <strong>in</strong>cludes all these effects will be presented.<br />

3.2 Advanced Model<br />

The simple model used <strong>in</strong> the previous section provided important qualitative<br />

underst<strong>and</strong><strong>in</strong>g <strong>of</strong> the multipactor threshold behaviour <strong>in</strong> a low<br />

pressure gas. However, due to the <strong>in</strong>herent limitations <strong>of</strong> the model,<br />

some <strong>of</strong> the results found by other researchers could not be confirmed.<br />

This section will present an improved model for multipactor <strong>in</strong> a low<br />

pressure gas <strong>and</strong> it is based on paper C <strong>of</strong> this thesis. As a representative<br />

gas, the noble gas argon will be used <strong>in</strong> the <strong>in</strong>cluded examples.<br />

3.2.1 Model<br />

Just as <strong>in</strong> the simple model, the basic geometric configuration is electron<br />

motion between two parallel plates perpendicular to the x-direction.<br />

Dur<strong>in</strong>g the passage, no electron loss, only generation through collisional<br />

ionisation, will occur. Us<strong>in</strong>g the differential equations for the total electron<br />

momentum <strong>and</strong> for the change <strong>in</strong> the number <strong>of</strong> electrons, one can<br />

derive the follow<strong>in</strong>g equation for the electron drift acceleration:<br />

du<br />

dt<br />

= eE<br />

m − u(νc + νiz). (3.9)<br />

where u is the drift velocity <strong>and</strong> νiz the ionisation frequency. In general,<br />

the collision <strong>and</strong> ionisation frequencies are functions <strong>of</strong> the electron velocity.<br />

However, by assum<strong>in</strong>g that νc <strong>and</strong> νiz are constants, Eq. (3.9)<br />

becomes a first order l<strong>in</strong>ear differential equation. <strong>Multipactor</strong> requires<br />

an alternat<strong>in</strong>g driv<strong>in</strong>g electric field <strong>and</strong> as <strong>in</strong> the previous model a harmonic<br />

field E = ˆxE0 s<strong>in</strong> ωt is used, where ˆx is the unit vector, ω the<br />

angular frequency, <strong>and</strong> t the time. Assum<strong>in</strong>g the electric field to be<br />

homogeneous, the drift velocity will be parallel to the field, u = ˆxu,<br />

<strong>and</strong> the vector notation for E <strong>and</strong> u can be dropped <strong>in</strong> the follow<strong>in</strong>g<br />

analysis. By sett<strong>in</strong>g ν = νc + νiz, u = ˙x, <strong>and</strong> du/dt = ¨x, Eq. (3.9) can<br />

be written,<br />

¨x = eE<br />

− ˙xν. (3.10)<br />

m<br />

44


S<strong>in</strong>ce the equation has the same form as Eq. (3.1), it will also have the<br />

same solutions <strong>and</strong> as the <strong>in</strong>itial conditions are identical, the formulas<br />

for the resonant field amplitude <strong>and</strong> the impact velocity will be identical.<br />

However, it should be noted that <strong>in</strong>stead <strong>of</strong> νc one will have ν <strong>and</strong> v0<br />

should be replaced by u0. An important difference is that the velocity<br />

<strong>in</strong> the previous model was only directed <strong>in</strong> the x-direction, but now<br />

there is also a thermal velocity component, vt, i.e. the total velocity<br />

is v = u + vt. With these new designations, the expressions for the<br />

resonant field amplitude <strong>and</strong> the impact velocity become:<br />

E0 =<br />

m<br />

e (ω2 + ν2 )(d + u0<br />

ν (Φ − 1))<br />

(1 + Φ)s<strong>in</strong> α + ((1 − Φ) ω<br />

ν<br />

2ν + ω )cos α<br />

(3.11)<br />

uimpact = u0Φ + Λ(1 + Φ)(ω cos α − ν s<strong>in</strong>α) (3.12)<br />

where Φ = exp (−Nπν/ω) has been <strong>in</strong>troduced for simplicity. Λ is given<br />

by Eq. (3.4) as before, but νc should be replaced by ν.<br />

In order to construct the multipactor boundaries, the same approach<br />

as <strong>in</strong> the simple model case is taken <strong>and</strong> an expression for the resonant<br />

phase is obta<strong>in</strong>ed by comb<strong>in</strong><strong>in</strong>g Eqs. (3.11) <strong>and</strong> (3.12), which yields,<br />

tan α = ω2 [ρΦ + χ] + 2ν 2 (Φu0 − uimpact)<br />

(dν + uimpact − u0)νω(1 + Φ)<br />

(3.13)<br />

where ρ = dν + uimpact + u0 <strong>and</strong> χ = dν − uimpact − u0 have been<br />

used for convenience. The reason why the expression looks somewhat<br />

different from Eq. (3.7) is that the constant <strong>in</strong>itial velocity approach has<br />

been used <strong>in</strong>stead <strong>of</strong> the assumption <strong>of</strong> a constant ratio between impact<br />

<strong>and</strong> <strong>in</strong>itial velocities. This will also affect the expression for the phase<br />

stability factor, which <strong>in</strong> this case becomes<br />

G = (Φ − 1)(ν2 + ω 2 )s<strong>in</strong> αΛ − Φνu0<br />

ν((1 + Φ)(ν s<strong>in</strong> α − ω cos α)Λ − Φu0)<br />

(3.14)<br />

So far, the differences between the simple <strong>and</strong> the more advanced<br />

model are fairly trivial. However, the parameters used (νc <strong>and</strong> νiz)<br />

are not constants, they depend to a great extent on the total electron<br />

velocity, which is the vector sum <strong>of</strong> the drift <strong>and</strong> thermal velocities. The<br />

thermal velocity will have a r<strong>and</strong>om direction <strong>and</strong> therefore the average<br />

total velocity will be equal to the drift velocity. However, the total<br />

(average) energy, ɛ, will still depend on both velocities <strong>and</strong> it becomes,<br />

ɛ = mv2<br />

2<br />

= m<br />

2 (u2 + 〈vt 2 〉) (3.15)<br />

45


where 〈vt 2 〉 represents the average <strong>of</strong> the square <strong>of</strong> the magnitude <strong>of</strong><br />

the thermal velocity. In paper C, a differential equation for the thermal<br />

velocity is derived, viz.<br />

d〈vt 2 〉<br />

dt + (νcδ + νiz)〈vt 2 〉 = u 2 (νc(2 − δ) + νiz) (3.16)<br />

where δ is the energy loss coefficient. By assum<strong>in</strong>g that νc, νiz <strong>and</strong> δ are<br />

constants, like before, Eq. (3.16) can be solved explicitly <strong>and</strong> with the<br />

<strong>in</strong>itial condition 〈vt 2 (t = α/ω)〉 = 0, the thermal impact velocity, when<br />

ωt = Nπ + α, can be found <strong>and</strong> thus the total impact velocity can be<br />

determ<strong>in</strong>ed. However, the expression is very complicated <strong>and</strong> will not<br />

be reproduced here.<br />

The total impact velocity will determ<strong>in</strong>e the secondary electron emission<br />

yield. For vacuum multipactor as well as <strong>in</strong> the previous simple<br />

model for low pressure multipactor, the impact velocity was perpendicular<br />

to the electrodes. In such a case, the secondary yield depends only<br />

on the impact velocity. However, for angular <strong>in</strong>cidence, which will be<br />

the case now with the r<strong>and</strong>om three dimensional thermal velocity component,<br />

the yield will be a function not only <strong>of</strong> the impact energy but<br />

also <strong>of</strong> the angle <strong>of</strong> <strong>in</strong>cidence. To account for the angular <strong>in</strong>cidence the<br />

expressions given <strong>in</strong> Ref. [22] have been used <strong>and</strong> for ease <strong>of</strong> reference<br />

they are reproduced here,<br />

ɛmax(θ) = ɛmax(0)(1 + θ 2 /π) (3.17)<br />

σse,max(θ) = σse,max(0)(1 + θ 2 /2π) (3.18)<br />

η = ɛimpact − ɛ0<br />

ɛmax(θ) − ɛ0<br />

σse = σse,max(θ)(η exp 1 − η) k<br />

(3.19)<br />

(3.20)<br />

where θ is the impact angle with respect to the surface normal. ɛmax is<br />

the impact energy when the secondary emission reaches its maximum,<br />

σse,max. ɛimpact is the total impact energy <strong>and</strong> ɛ0 is the energy limit<br />

for non-zero σse. The formulas are valid for “a typical dull surface”,<br />

accord<strong>in</strong>g to Ref. [22]. The coefficient k is given by k = 0.62 for η < 1<br />

<strong>and</strong> k = 0.25 for η > 1.<br />

In vacuum multipactor, the only source <strong>of</strong> new electrons is secondary<br />

yield from each impact. When the phenomenon takes place <strong>in</strong> a gas,<br />

another potential source <strong>of</strong> new electrons is impact ionisation <strong>of</strong> the<br />

gas molecules. The ionisation threshold <strong>of</strong> most gases <strong>of</strong> <strong>in</strong>terest is<br />

46


<strong>in</strong> the range 10-20 eV. This is well below the first cross-over po<strong>in</strong>t <strong>of</strong><br />

most materials <strong>and</strong> thus when the electron energy is sufficient to <strong>in</strong>itiate<br />

multipactor, it is also enough to ionise the gas molecules. In this model,<br />

the contribution from impact ionisation is <strong>in</strong>cluded by modify<strong>in</strong>g the<br />

breakdown condition from σse = 1 to<br />

σse + 〈νiz〉Nπ<br />

µ = 1 (3.21)<br />

ω<br />

where 〈νiz〉 is the average ionisation frequency <strong>and</strong> µ is an ionisation<br />

factor, which ranges from 0 − 1 <strong>and</strong> <strong>in</strong>dicates the fraction <strong>of</strong> the electrons<br />

from ionisation that is able to become a part <strong>of</strong> the multipact<strong>in</strong>g<br />

bunch. Determ<strong>in</strong>ation <strong>of</strong> the correct value <strong>of</strong> µ is not a trivial problem<br />

<strong>and</strong> for simplicity a constant µ = 0.75 is used. This is quite a rough<br />

approximation, but for materials with a low first cross-over po<strong>in</strong>t, it will<br />

be shown that the ionisation contribution is fairly small <strong>and</strong> the exact<br />

value for µ is not so important. On the other h<strong>and</strong>, for materials with<br />

a high first cross-over energy, the importance <strong>of</strong> µ can not be neglected<br />

<strong>and</strong> thus a detailed <strong>in</strong>vestigation <strong>of</strong> µ should be performed, but due to<br />

the complexity, it will be left as future work.<br />

Apart from µ, there are other parameters, which need to be determ<strong>in</strong>ed<br />

with good accuracy <strong>in</strong> order to obta<strong>in</strong> useful quantitative results.<br />

The energy loss coefficient, δ, which is used <strong>in</strong> Eq. (3.16), is a small<br />

quantity <strong>and</strong> for pure elastic collisions, the value is equal to 2m/M,<br />

where m is the electron mass <strong>and</strong> M is the mass <strong>of</strong> the argon atom [55].<br />

It is also a function <strong>of</strong> the electron energy <strong>and</strong> for <strong>in</strong>elastic collisions,<br />

which will occur when the electron energy is greater than a few eV, the<br />

value is about 10 to 100 times larger than the elastic value [56]. However,<br />

the value is still quite small <strong>and</strong> will not have major effect on the<br />

low pressure multipactor threshold <strong>and</strong> for simplicity a value 10 −3 will<br />

be used, which is about 37 times greater than the elastic value. The<br />

rema<strong>in</strong><strong>in</strong>g parameters, νc <strong>and</strong> νiz, are very important for the threshold<br />

<strong>and</strong> a detailed description <strong>of</strong> these values will be given <strong>in</strong> the next<br />

section.<br />

3.2.2 Analytical formulas for argon cross-sections<br />

For most materials <strong>of</strong> <strong>in</strong>terest, the first cross-over energy is <strong>in</strong> the range<br />

20-70 eV <strong>and</strong> it is this value that determ<strong>in</strong>es the lower multipactor<br />

threshold. Thus it would be <strong>of</strong> value to have expressions for the collision<br />

<strong>and</strong> ionisation cross-sections that give an accurate description <strong>of</strong><br />

47


these quantities <strong>in</strong> the range from 0 eV to about 100 eV. For the electronargon<br />

collision cross section, the data given <strong>in</strong> Ref. [57] is used <strong>and</strong> it<br />

covers the range up to 20 eV. An analytical formula has been devised,<br />

which approximates the given data quite well <strong>in</strong> the measurement region<br />

(cf. Fig. 3.5). Outside the measurement po<strong>in</strong>ts, the cross-section<br />

for very low energy electrons has been set to converge towards the geometrical<br />

cross-section <strong>of</strong> the argon atom. For high energy electrons, the<br />

cross-section is set to fall <strong>of</strong>f with the same rate as for the last few eVs.<br />

The analytical formula is given by the expression,<br />

σc = (<br />

1.68<br />

+<br />

1 + (8ɛ) 3<br />

ɛ<br />

1 + (0.07ɛ) 2 )1.5 × 1.15 × 10 −20 [m 2 ] (3.22)<br />

where ɛ is the total electron energy, given by Eq. (3.15).<br />

Total collision cross section [m 2 ]<br />

10 −18<br />

10 −19<br />

10 −20<br />

10 −3<br />

10 −21<br />

Buckman <strong>and</strong> Lohman<br />

Analytical approximation<br />

10 −2<br />

10 −1<br />

10 0<br />

Electron energy [eV]<br />

Figure 3.5: Absolute total collision cross-section for electrons scattered from<br />

argon. The stars <strong>in</strong>dicate measurement data by Buckman <strong>and</strong><br />

Lohmann [57] <strong>and</strong> the solid l<strong>in</strong>e is the analytical approximation<br />

given by Eq. (3.22).<br />

The ionisation cross-section <strong>in</strong>creases rapidly for electron energies<br />

slightly above the ionisation threshold <strong>and</strong> thus it is <strong>of</strong> importance to<br />

have an accurate description <strong>in</strong> this range, especially s<strong>in</strong>ce this is close<br />

to the first cross-over energy <strong>of</strong> many materials. A simple function that<br />

48<br />

10 1<br />

10 2


accurately describes the cross-section for the entire measurement range<br />

can be given by (cf. Fig. 3.6)<br />

σiz = q1<br />

ln ɛ/ɛi<br />

�<br />

ɛ/ɛi + 0.1(ɛ/ɛi) 2<br />

[m 2 ] (3.23)<br />

where ɛi is the ionisation threshold <strong>of</strong> argon <strong>and</strong> q1 = 4.8 × 10 −20 m 2 .<br />

Ionisation Cross Section [m 2 ]<br />

10 −19<br />

10 −20<br />

10 −21<br />

10 −22<br />

Ionisation cross section for Argon<br />

10 2<br />

Electron energy [eV]<br />

S.C.Brown<br />

H.C.Straub<br />

Analytical approximation<br />

Figure 3.6: Ionisation cross-section for electron-argon collisions. The circles<br />

<strong>and</strong> stars <strong>in</strong>dicate measurement data by S. C. Brown [58] <strong>and</strong><br />

Straub et al. [59] respectively <strong>and</strong> the solid l<strong>in</strong>e is an analytical<br />

approximation given by Eq. (3.23).<br />

3.2.3 <strong>Multipactor</strong> boundaries<br />

In the follow<strong>in</strong>g section the above model will be used to determ<strong>in</strong>e the<br />

multipactor boundaries. The best accuracy is atta<strong>in</strong>ed when solv<strong>in</strong>g<br />

the basic differential equations numerically while us<strong>in</strong>g good approximate<br />

formulas for the different parameters. However, such computation<br />

takes very long time, s<strong>in</strong>ce both the <strong>in</strong>itial <strong>and</strong> the f<strong>in</strong>al multipactor<br />

conditions have to be fulfilled. Faster computation can be achieved<br />

by us<strong>in</strong>g different approximations, e.g. constant parameters as shown<br />

<strong>in</strong> Eqs. (3.10)-(3.12), Eq. (3.13), <strong>and</strong> Eq. (3.14). Two different implementations<br />

are used <strong>in</strong> paper C, one purely numerical <strong>and</strong> one semi-<br />

10 3<br />

49


analytical. Attempts were made to f<strong>in</strong>d a purely analytical implementation<br />

as well, but due to the strong non-l<strong>in</strong>earities <strong>in</strong> the functions for the<br />

cross-sections, no accurate such implementation could be found. Details<br />

concern<strong>in</strong>g the two implementations are presented <strong>in</strong> paper C <strong>and</strong> will<br />

not be reproduced here.<br />

As mentioned <strong>in</strong> chapter 2, when construct<strong>in</strong>g the complete multipactor<br />

zones, the multipactor thresholds correspond<strong>in</strong>g to impact velocities<br />

between the first <strong>and</strong> second cross-over po<strong>in</strong>ts are determ<strong>in</strong>ed<br />

for a specific order <strong>of</strong> resonance with<strong>in</strong> the phase range from the nonreturn<strong>in</strong>g<br />

electron limit to the upper phase stability limit. The zone<br />

for that order <strong>of</strong> resonance is then the envelope <strong>of</strong> all these curves (cf.<br />

Fig. 2.3). However, to explore the basic effects on the multipactor phenomenon,<br />

it is sufficient to study the threshold correspond<strong>in</strong>g to unity<br />

SEY. Thus, <strong>in</strong> most <strong>of</strong> the follow<strong>in</strong>g charts, only the lower multipactor<br />

threshold will be considered. However, <strong>in</strong> keep<strong>in</strong>g with the multipactor<br />

tradition, the complete zones will be presented as well.<br />

One concern that appears when mak<strong>in</strong>g low pressure multipactor<br />

charts is the parameters which should be used on the chart axes. Classical<br />

vacuum multipactor charts use eng<strong>in</strong>eer<strong>in</strong>g units with voltage as a<br />

function <strong>of</strong> the frequency-gap size product, like <strong>in</strong> Fig. 2.3. By multiply<strong>in</strong>g<br />

Eq. (3.11) by the gap size, d, to get the voltage <strong>and</strong> rearrang<strong>in</strong>g<br />

Eqs. (3.12), (3.13), <strong>and</strong> (3.14), these expressions can all be written as<br />

functions <strong>of</strong> two natural parameters, viz. fd <strong>and</strong> pd, i.e. the frequencygap<br />

size <strong>and</strong> the pressure-gap size products. Thus, for a given pd the<br />

multipactor zones can be constructed <strong>in</strong> the classical eng<strong>in</strong>eer<strong>in</strong>g units<br />

as shown <strong>in</strong> Fig. 3.7. Note that three different pd-values are used, one<br />

for each zone. The chosen values are close to the limit <strong>of</strong> stability <strong>of</strong> the<br />

numerical implementation for each zone.<br />

In Fig. 3.7 both the analytical (semi-analytical) <strong>and</strong> the numerical<br />

implementations are used to plot the thresholds. Very good agreement<br />

between the two implementations is found <strong>and</strong> therefore the faster analytical<br />

version is used to produce all other figures. The most strik<strong>in</strong>g<br />

first impression <strong>of</strong> the graphs <strong>in</strong> Fig. 3.7 is the difference <strong>in</strong> behaviour<br />

between the first <strong>and</strong> the higher order modes. The first order mode<br />

shows an <strong>in</strong>creased threshold, which is a consequence <strong>of</strong> the friction<br />

force experienced by the electrons due to collisions with neutrals. This<br />

is <strong>in</strong> agreement with the model presented <strong>in</strong> paper B, which only considered<br />

the friction force. However, for higher order modes, the result<br />

is the opposite. Instead <strong>of</strong> an <strong>in</strong>creased threshold as <strong>in</strong> the friction<br />

50


Voltage [V]<br />

10 2<br />

Numeric stable phase<br />

Numeric unstable phase<br />

Analytic stable phase<br />

Analytic unstable phase<br />

pd=15 Pa mm<br />

10 0<br />

pd=7 Pa mm<br />

Frequency − Gap product [GHz⋅mm]<br />

pd=5 Pa mm<br />

Figure 3.7: <strong>Low</strong>er multipactor thresholds <strong>in</strong> low pressure argon for three different<br />

fixed pd-values, one for each zone. The dotted l<strong>in</strong>es represent<br />

the multipactor zones for vacuum multipactor. Parameters<br />

used are: W1 = 23 eV , W0 = 3.68 eV, σse,max(0) = 3, ɛ0 = 0.<br />

51


only model, the <strong>in</strong>clusion <strong>of</strong> ionisation <strong>and</strong> thermal spread leads to a<br />

decreased threshold with <strong>in</strong>creas<strong>in</strong>g pressure.<br />

Thresholds as <strong>in</strong> Fig. 3.7 can be found for all impact velocities between<br />

W1 <strong>and</strong> W2 <strong>and</strong> by construct<strong>in</strong>g the envelope <strong>of</strong> all these curves<br />

for each order <strong>of</strong> resonance, the complete, classical multipactor zones can<br />

be found for a given pd-product for each zone. This is done <strong>in</strong> Fig. 3.8,<br />

which shows the complete zones for three different pd-values. The drawback<br />

with this chart is that the model does not account for the hybrid<br />

zones <strong>and</strong> thus the right boundary <strong>of</strong> each zone will not accurately reflect<br />

the true multipactor threshold for those fd-values. The model can,<br />

however, be extended to <strong>in</strong>clude also the hybrid modes, but due to the<br />

<strong>in</strong>creased complexity, this is left as future work<br />

Voltage [V]<br />

10 3<br />

10 2<br />

pd=10 Pa mm<br />

10 0<br />

pd=4 Pa mm<br />

Frequency−Gap product [GHz⋅mm]<br />

pd=2 Pa mm<br />

Figure 3.8: <strong>Multipactor</strong> susceptibility zones <strong>in</strong> low pressure argon (solid l<strong>in</strong>es)<br />

together with vacuum zones (dotted l<strong>in</strong>es) for comparison. Parameters<br />

used are: W1 = 23 eV, W2 = 1000 eV, W0 = 3.68 eV,<br />

σse,max(0) = 3 <strong>and</strong> ɛ0 = 0.<br />

3.2.4 Key f<strong>in</strong>d<strong>in</strong>gs<br />

Among the ma<strong>in</strong> results is that the friction force dom<strong>in</strong>ates the low<br />

pressure multipactor threshold for materials with a low first cross-over<br />

52<br />

10 1


energy for the first order <strong>of</strong> resonance, as <strong>in</strong>dicated by Fig. 3.7. However,<br />

this figure only shows the behaviour for a given pd-product <strong>and</strong> <strong>in</strong> order<br />

to see what happens when the gas density <strong>in</strong>creases, the threshold can be<br />

plotted as a function <strong>of</strong> the pd-product. This has been done <strong>in</strong> Fig. 3.9<br />

for a material with a low first cross-over energy <strong>and</strong>, as expected, the<br />

threshold <strong>in</strong>creases with <strong>in</strong>creas<strong>in</strong>g pressure for the lowest order mode,<br />

N = 1, <strong>and</strong> after reach<strong>in</strong>g a maximum, the threshold starts to decrease<br />

aga<strong>in</strong>. For higher order modes, the threshold decreases monotonically<br />

as the gas becomes dense enough to affect the multipact<strong>in</strong>g electrons.<br />

This behaviour is identical to that found by Gilard<strong>in</strong>i [53] <strong>in</strong> his Monte<br />

Carlo simulation <strong>of</strong> low pressure multipactor. He also observed that for<br />

materials with a higher first cross-over po<strong>in</strong>t, the threshold does not<br />

<strong>in</strong>crease with <strong>in</strong>creas<strong>in</strong>g pd, <strong>in</strong>stead it falls <strong>of</strong>f monotonically, which is<br />

the behaviour shown <strong>in</strong> Fig. 3.10. The ma<strong>in</strong> reason for this difference<br />

<strong>in</strong> behaviour is the contribution <strong>of</strong> electrons from collisional ionisation,<br />

which <strong>in</strong>creases drastically when the electron energy is well above the<br />

ionisation threshold.<br />

Normalised Voltage<br />

1.1<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

10 −2<br />

10 −1<br />

10 0<br />

<strong>Pressure</strong> × gapsize [Pa⋅mm]<br />

Figure 3.9: Normalised multipactor thresholds for vary<strong>in</strong>g pd. The thresholds<br />

are normalised with respect to the vacuum threshold. Curves for<br />

the three first orders <strong>of</strong> resonance are shown. Parameters used<br />

are: W1 = 23 eV, W0 = 3.68 eV, σse,max(0) = 3, ɛ0 = 0, fdN=1 =<br />

0.6 GHz·mm, fdN=3 = 2.4 GHz·mm, <strong>and</strong> fdN=5 = 4.2 GHz·mm.<br />

For higher order <strong>of</strong> resonance, N > 1, Gilard<strong>in</strong>i found no difference<br />

<strong>in</strong> the basic behaviour regardless <strong>of</strong> material. The threshold falls <strong>of</strong>f<br />

N=1<br />

N=5<br />

N=3<br />

10 1<br />

53


Normalised Voltage<br />

1.1<br />

1.05<br />

1<br />

0.95<br />

0.9<br />

0.85<br />

0.8<br />

0.75<br />

0.7<br />

10 −2<br />

10 −1<br />

10 0<br />

<strong>Pressure</strong> × gapsize [Pa⋅mm]<br />

Figure 3.10: Normalised multipactor thresholds for vary<strong>in</strong>g pd. The thresholds<br />

are normalised with respect to the vacuum threshold.<br />

Curves for the three first orders <strong>of</strong> resonance are shown for<br />

a material with a first cross-over po<strong>in</strong>t more than 7 times<br />

greater than <strong>in</strong> Fig. 3.9. Parameters used are: W1 = 170 eV,<br />

W0 = 4 eV, σse,max(0) = 1.3, ɛ0 = 0, fdN=1 = 1 GHz·mm,<br />

fdN=3 = 3.2 GHz·mm, <strong>and</strong> fdN=5 = 5 GHz·mm.<br />

54<br />

N=5<br />

N=1<br />

N=3<br />

10 1


Voltage [V]<br />

10 2<br />

µ=0<br />

µ=0.75<br />

10 0<br />

Frequency − Gap product [GHz⋅mm]<br />

µ=0.75<br />

µ=0<br />

Vacuum multipactor<br />

Figure 3.11: <strong>Multipactor</strong> thresholds <strong>in</strong> low pressure argon for the two lowest<br />

order modes (N = 1 <strong>and</strong> N = 3) with µ = 0.75 <strong>and</strong> µ = 0<br />

respectively. Parameters used are: W1 = 23 eV, W0 = 3.68 eV,<br />

σse,max(0) = 3, ɛ0 = 0, pd = 15 Pa·mm for N = 1, <strong>and</strong> pd =<br />

7 Pa·mm for N = 3.<br />

directly from the vacuum threshold without show<strong>in</strong>g any maximum <strong>and</strong><br />

the same behaviour is seen <strong>in</strong> Figs. 3.9 <strong>and</strong> 3.10. Even with a low W1,<br />

where the contribution <strong>of</strong> electrons from collisional ionisation is low, a<br />

monotonically decreas<strong>in</strong>g threshold is obta<strong>in</strong>ed. The cause <strong>of</strong> this dist<strong>in</strong>ct<br />

lowered threshold is the partial thermalisation <strong>of</strong> the electrons due<br />

to the collisions. The velocity spread results <strong>in</strong> a total impact velocity,<br />

which is greater than the drift velocity alone <strong>and</strong> thus for the same<br />

secondary electron emission, a lowered impact drift velocity is possible.<br />

Even though the friction force requires a higher voltage to achieve the<br />

same impact drift velocity, the thermalisation effect dom<strong>in</strong>ates, with a<br />

lowered threshold as a result. This becomes clear <strong>in</strong> Fig. 3.11, where it is<br />

apparent that it is not the electrons from ionisation that constitute the<br />

ma<strong>in</strong> reason for a decreased threshold, rather it is a consequence <strong>of</strong> the<br />

partial thermalisation. In the case with a high W1, the thermalisation<br />

effect is also important, but without the contribution from collisional<br />

ionisation, the behaviour would not be the same, which can be seen <strong>in</strong><br />

Fig. 3.12.<br />

To summarise the key f<strong>in</strong>d<strong>in</strong>gs from the more advanced model, it<br />

55


Voltage [V]<br />

10 3<br />

10 0<br />

µ=0<br />

µ=0.75<br />

µ=0.75<br />

Vacuum multipactor<br />

Frequency − Gap product [GHz⋅mm]<br />

Figure 3.12: <strong>Multipactor</strong> thresholds <strong>in</strong> low pressure argon for the two lowest<br />

order modes (N = 1 <strong>and</strong> N = 3) with µ = 0.75 <strong>and</strong> µ = 0<br />

respectively. In this case, the second cross-over energy is about<br />

7 times greater than <strong>in</strong> Fig. 3.11. Parameters used are: W1 =<br />

170 eV , W0 = 4 eV, σse,max(0) = 1.3, ɛ0 = 0, pd = 25 Pa·mm<br />

for N = 1, <strong>and</strong> pd = 15 Pa·mm for N = 3.<br />

56<br />

µ=0


can be said that there are three ma<strong>in</strong> effects that affect the low pressure<br />

multipactor threshold. The friction force tends to <strong>in</strong>crease the threshold<br />

as a higher electric field is needed to reach the necessary impact velocity.<br />

The thermalisation, on the other h<strong>and</strong>, <strong>in</strong>creases the total impact energy<br />

<strong>and</strong> thus a lower electric field is needed to achieve the required impact<br />

velocity. For materials with a low first cross-over po<strong>in</strong>t, the first effect<br />

dom<strong>in</strong>ates for the first order <strong>of</strong> resonance, while for higher order modes,<br />

the latter plays the ma<strong>in</strong> role. In addition to these two effects, the model<br />

also <strong>in</strong>cludes contribution from impact ionisation to the total number<br />

<strong>of</strong> electrons. This addition also tend to lower the multipactor threshold<br />

<strong>and</strong> the effect becomes very prom<strong>in</strong>ent for materials with a high first<br />

cross-over energy as a consequence <strong>of</strong> the concomitant high ionisation<br />

cross-section.<br />

57


Chapter 4<br />

<strong>Multipactor</strong> <strong>in</strong> irises<br />

A common microwave component is the waveguide iris, which is <strong>of</strong>ten<br />

used as a shunt susceptance for the purpose <strong>of</strong> match<strong>in</strong>g a load to the<br />

waveguide. There are many different types <strong>of</strong> irises, but a typical configuration<br />

consists <strong>of</strong> a step-like, short length, reduction <strong>of</strong> the waveguide<br />

height. Similar structures also appear <strong>in</strong> other configurations, e.g. as<br />

apertures <strong>in</strong> array antennas, as coupl<strong>in</strong>g slots <strong>in</strong> directional couplers,<br />

<strong>and</strong> as irises <strong>in</strong> waveguide filters. As the field strength <strong>in</strong> the iris can<br />

be very high <strong>and</strong> the gap height is small, there is a pronounced risk <strong>of</strong><br />

hav<strong>in</strong>g a multipactor discharge.<br />

So far <strong>in</strong> this thesis, all the models considered have been based on<br />

the plane-parallel model with a spatially uniform harmonic electric field.<br />

In general, most theoretical studies <strong>of</strong> the multipactor phenomenon have<br />

been limited to this or similar approximations. However, many RF devices<br />

<strong>in</strong>volve more complicated electric field structures where predictions<br />

based on the parallel-plate model are not applicable. This is true for<br />

e.g. the waveguide iris, where the electric field will be a comb<strong>in</strong>ation<br />

<strong>of</strong> several different electromagnetic modes, most <strong>of</strong> which typically are<br />

evanescent. However, due to the short length <strong>of</strong> the iris, these modes<br />

will be <strong>of</strong> importance. Nevertheless, <strong>in</strong> this analysis, which is described<br />

<strong>in</strong> more detail <strong>in</strong> paper D, it is only the importance <strong>of</strong> the r<strong>and</strong>om drift<br />

due to <strong>in</strong>itial velocity spread <strong>of</strong> the secondary electrons that has been<br />

considered. Thus, as far as the field is concerned, a spatially uniform<br />

harmonic field based on the parallel-plate model is used.<br />

Experiments [60, 61] as well as numerical studies [61, 62] <strong>of</strong> multipactor<br />

<strong>in</strong> an iris have shown that the discharge threshold <strong>in</strong>creases with<br />

decreas<strong>in</strong>g length <strong>of</strong> the iris. It has been suggested that the reason for<br />

59


the <strong>in</strong>creased threshold is losses <strong>of</strong> electrons out <strong>of</strong> the iris region [60]. In<br />

this analysis, we show that one <strong>of</strong> the contribut<strong>in</strong>g factors to this electron<br />

loss is a r<strong>and</strong>om drift due to the axial component <strong>of</strong> the <strong>in</strong>itial velocity<br />

<strong>of</strong> the secondary emitted electrons. Other loss mechanisms, which are<br />

due to the <strong>in</strong>homogeneity <strong>of</strong> the field, tend to further enhance the losses<br />

<strong>and</strong> these effects will be more pronounced for small gap lengths. This<br />

means that by tak<strong>in</strong>g only losses due to the r<strong>and</strong>om drift <strong>in</strong>to account,<br />

a conservative <strong>in</strong>crease <strong>of</strong> the breakdown threshold should be obta<strong>in</strong>ed.<br />

4.1 Model <strong>and</strong> approximations<br />

The geometry used <strong>in</strong> the model is the 2-dimensional structure shown<br />

<strong>in</strong> Fig. 4.1. The iris has a gap height h <strong>in</strong> the y-direction, a length l<br />

<strong>in</strong> the z-direction <strong>and</strong> is assumed to be fitted <strong>in</strong>to a waveguide with a<br />

height that is much greater than h. The harmonic electric field E is<br />

assumed uniform <strong>in</strong> the gap, as a simple approximation <strong>of</strong> the actual<br />

field. There are two ma<strong>in</strong> reasons for choos<strong>in</strong>g a uniform field. Firstly,<br />

the determ<strong>in</strong>istic model developed for the parallel plate case, which is<br />

described <strong>in</strong> chapter 2, can be used to describe the basic behaviour<br />

<strong>of</strong> an electron trajectory <strong>in</strong>side the gap. Secondly, the effect <strong>of</strong> the<br />

<strong>in</strong>itial velocity spread <strong>of</strong> the secondary electrons along the z-axis on the<br />

multipactor threshold can be analysed separately from the drift force due<br />

to <strong>in</strong>homogeneities <strong>in</strong> the electric field. In addition, it gives a convenient<br />

base for compar<strong>in</strong>g the results with those <strong>of</strong> the parallel plate model.<br />

By assum<strong>in</strong>g a uniform E-field <strong>in</strong> the y-direction, the electron motion<br />

along the z-direction is not affected by the field. The motion <strong>in</strong> this<br />

direction, the drift motion, will occur with a fixed velocity vz between<br />

the impacts. Lets assume that a seed electron is emitted <strong>in</strong>side the gap at<br />

the coord<strong>in</strong>ate z0, −l/2 < z0 < l/2, at one <strong>of</strong> the walls. As the electron<br />

traverses the gap <strong>and</strong> hits the opposite side <strong>of</strong> the iris, it has become<br />

displaced a distance ∆z <strong>in</strong> the z-direction. This drift is determ<strong>in</strong>ed by<br />

the velocity <strong>in</strong> the z-direction, vz , together with the transit time, tg,<br />

<strong>and</strong> is given by ∆z = vztg. For a fixed mode order, N, <strong>and</strong> frequency,<br />

f, <strong>of</strong> the field, each transit time is the same <strong>and</strong> is given by,<br />

tg = Nπ<br />

, (4.1)<br />

ω<br />

where ω = 2πf.<br />

The electron trajectory <strong>in</strong> the z-direction will perform a r<strong>and</strong>om<br />

walk with a change <strong>of</strong> velocity, vz, after each impact. When the impact<br />

60


h<br />

l<br />

Figure 4.1: The geometry used <strong>in</strong> the considered model.<br />

coord<strong>in</strong>ate is outside the iris area |z| > l/2, i.e. one <strong>of</strong> the gap edges<br />

has been passed, the electron trajectory is lost. The probability <strong>of</strong> survival,<br />

p(k) (see Fig. 4.2), for the electron trajectory decreases with the<br />

number <strong>of</strong> transits, k, <strong>and</strong> for a general one-dimensional r<strong>and</strong>om walk<br />

problem, with the jump size governed by a cont<strong>in</strong>uous distribution function,<br />

Φk(z), an explicit solution for this, the first passage time problem,<br />

is not always possible. However, <strong>in</strong> paper D it is expla<strong>in</strong>ed that the<br />

asymptotic behaviour <strong>of</strong> p(k) is determ<strong>in</strong>ed by the largest eigenvalue γ0<br />

<strong>of</strong> the expansion <strong>of</strong> Φk(z), i.e.<br />

E<br />

y<br />

p(k) ∝ γ k 0 . (4.2)<br />

A detailed description <strong>of</strong> how to determ<strong>in</strong>e p(k) is given <strong>in</strong> paper D<br />

together with approximate solutions for γ0 when the normalised iris<br />

length, η = l/(vTtg), is either very small or very large. This summary,<br />

however, will focus on the effect the r<strong>and</strong>om electron drift has on the<br />

multipactor susceptibility zones.<br />

Each seed electron <strong>in</strong>side the iris gap will start to multiply with the<br />

successive wall collisions. Due to the stochastic losses, the number <strong>of</strong><br />

electrons will sometimes become large <strong>and</strong> sometimes small. However, if<br />

on average the generation <strong>of</strong> electrons due to wall collisions is larger than<br />

the loss over the gap edges, there is a f<strong>in</strong>ite probability that a sufficiently<br />

z<br />

61


Probability <strong>of</strong> survival<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

l=2 mm<br />

l=8 mm<br />

l=16 mm<br />

50 100 150 200 250 300 350 400 450 500<br />

Number <strong>of</strong> collisions<br />

Figure 4.2: The probability <strong>of</strong> survival, p(k), for an electron emitted <strong>in</strong> the<br />

center <strong>of</strong> the iris gap, z = 0, for three different iris lengths. Parameters<br />

used: f = 1 GHz, N = 1, <strong>and</strong> WT = 2 eV (correspond<strong>in</strong>g<br />

to vT, the rms-velocity <strong>of</strong> the Maxwellian distribution<br />

<strong>of</strong> <strong>in</strong>itial velocity <strong>in</strong> the z-direction).<br />

strong discharge will appear. The generated number <strong>of</strong> electrons over<br />

the <strong>in</strong>itial number <strong>of</strong> electrons after k collisions is given by,<br />

Ne<br />

N0<br />

≡ g(k) = p(k)σ k se. (4.3)<br />

Depend<strong>in</strong>g on the start position <strong>of</strong> the seed electrons, the <strong>in</strong>itial behaviour<br />

<strong>of</strong> Ne can vary. If the start position is close to the iris edge,<br />

the average electron number will first decrease <strong>and</strong> then if σse is large<br />

enough, it will start <strong>in</strong>creas<strong>in</strong>g aga<strong>in</strong>. But if the start position is <strong>in</strong><br />

the center, it may first start to <strong>in</strong>crease, but after a number <strong>of</strong> transits,<br />

it will start decreas<strong>in</strong>g (cf. Fig. 4.3). Eventually, it is the asymptotic<br />

behaviour <strong>of</strong> p(k) that will determ<strong>in</strong>e whether or not there will be a<br />

discharge. Thus from Eq. (4.2) <strong>and</strong> Eq. (4.3) one can conclude that the<br />

asymptotic change <strong>in</strong> the electron number is given by,<br />

g(k) ∝ (σseγ0) k . (4.4)<br />

Thus the average number <strong>of</strong> electrons will grow if<br />

62<br />

σseγ0 > 1 (4.5)


N e /N 0 [−]<br />

3<br />

2.5<br />

2<br />

1.5<br />

1<br />

<strong>Multipactor</strong> <strong>in</strong> iris<br />

σ = 1.159<br />

σ = 1.151<br />

0.5<br />

0 50 100<br />

Number <strong>of</strong> collisions<br />

150 200<br />

Figure 4.3: The growth <strong>in</strong> electron number as a function <strong>of</strong> the number <strong>of</strong><br />

gap cross<strong>in</strong>gs for two different SEY-coefficients. Parameters used:<br />

f = 1 GHz, N = 1, l = 2 mm, <strong>and</strong> WT = 2 eV.<br />

or equivalently<br />

σse > 1/γ0 > 1. (4.6)<br />

This implies that the secondary electron yield must be greater than a<br />

value that is larger than unity (1/γ0 > 1) to have growth <strong>of</strong> the number<br />

<strong>of</strong> electrons. This modified breakdown criterion is the only difference<br />

between the model considered here <strong>and</strong> the conventional resonance theory<br />

<strong>of</strong> multipactor <strong>in</strong>side a plane-parallel gap (where σse > 1 is used<br />

when determ<strong>in</strong><strong>in</strong>g the threshold). The condition for σse, Eq. (4.6), can<br />

be converted <strong>in</strong>to a range <strong>of</strong> impact energies,<br />

W1 < Wm<strong>in</strong> < Wimpact < Wmax < W2, (4.7)<br />

where the impact energies Wm<strong>in</strong> <strong>and</strong> Wmax are determ<strong>in</strong>ed by σse =<br />

1/γ0. Consequently, us<strong>in</strong>g Wm<strong>in</strong> <strong>and</strong> Wmax <strong>in</strong>stead <strong>of</strong> W1 <strong>and</strong> W2, respectively,<br />

<strong>in</strong> the parallel-plate model, multipactor regions that account<br />

for the electron losses due r<strong>and</strong>om drift can be obta<strong>in</strong>ed.<br />

63


4.2 <strong>Multipactor</strong> regions<br />

Us<strong>in</strong>g the above described model <strong>and</strong> by employ<strong>in</strong>g a natural scal<strong>in</strong>g<br />

parameter <strong>of</strong> η, viz. l · f, multipactor charts <strong>in</strong> the traditional eng<strong>in</strong>eer<strong>in</strong>g<br />

units (voltage vs. frequency-gap-size product) can be devised<br />

for a specific value <strong>of</strong> the ratio <strong>of</strong> gap height <strong>and</strong> iris length, h/l. Figure<br />

4.4 shows an example <strong>of</strong> this, where the multipactor regions have<br />

been constructed for 5 different h/l-ratios.<br />

Peak voltage [V]<br />

10 4<br />

10 3<br />

10 2<br />

10 0<br />

0.001<br />

1<br />

3<br />

5.2<br />

6.3<br />

<strong>Multipactor</strong> regions for different height/length ratios<br />

10 1<br />

Frequency gap height product [GHz ⋅ mm]<br />

Figure 4.4: The first four multipactor susceptibility zones for a microwave<br />

iris with five different height/length-ratios. Parameters used are:<br />

W0 = 2 eV (y-direction), WT = 2 eV (z-direction), W1 = 85.6 eV,<br />

<strong>and</strong> σse,max = 1.83 (material properties for alod<strong>in</strong>e [50]).<br />

The transit time decreases with <strong>in</strong>creas<strong>in</strong>g frequency accord<strong>in</strong>g to<br />

Eq. (4.1) <strong>and</strong> thus the distance traversed <strong>in</strong> each step becomes smaller,<br />

which implies that the probability <strong>of</strong> surviv<strong>in</strong>g k steps <strong>in</strong>creases. This<br />

causes the multipactor zones to shr<strong>in</strong>k towards higher frequencies with<br />

<strong>in</strong>creas<strong>in</strong>g h/l as is evident <strong>in</strong> Fig. 4.4. The transit time is also a function<br />

<strong>of</strong> the mode order, which <strong>in</strong>creases the transit time for higher order<br />

modes. This counteracts the decrease due to <strong>in</strong>creas<strong>in</strong>g frequency <strong>and</strong><br />

consequently a behaviour similar to that <strong>of</strong> the first resonance zone can<br />

be observed also for the higher order modes.<br />

For materials with a low maximum SEY, like <strong>in</strong> Fig. 4.4, the ability<br />

64


to compensate for electron losses is not very good <strong>and</strong> thus the zones<br />

will disappear for relatively small values <strong>of</strong> h/l. However, <strong>in</strong> the opposite<br />

case for a material with a large SEY, like e.g. alum<strong>in</strong>ium, multipactor<br />

will be possible also for relatively th<strong>in</strong> irises.<br />

4.3 Comparison with experiments<br />

By compar<strong>in</strong>g the current model with experimental data [61], good qualitative<br />

agreement can be observed (see Fig. 4.5). As the h/l-ratio <strong>in</strong>creases,<br />

the threshold <strong>in</strong>creases until, beyond a certa<strong>in</strong> limit, no multipactor<br />

is possible. S<strong>in</strong>ce only the electron losses due to the r<strong>and</strong>om drift<br />

are accounted for, the model predicts the existence <strong>of</strong> a discharge beyond<br />

the limit found <strong>in</strong> the experiments. Consequently, from an eng<strong>in</strong>eer<strong>in</strong>g<br />

po<strong>in</strong>t <strong>of</strong> view, this is a conservative measure <strong>of</strong> the <strong>in</strong>creased threshold<br />

<strong>and</strong> thus it should be safe to apply it when design<strong>in</strong>g multipactor free<br />

microwave hardware.<br />

Peak Voltage [V]<br />

3500<br />

3000<br />

2500<br />

2000<br />

1500<br />

<strong>Multipactor</strong> <strong>in</strong> iris<br />

1000<br />

Current model a)<br />

Current model b)<br />

Current model c)<br />

Current model d)<br />

Measur. −presentation<br />

Measur. −proceed<strong>in</strong>gs<br />

500<br />

0 1 2 3 4<br />

Height/Width [−]<br />

5 6 7 8<br />

Figure 4.5: The multipactor threshold as a function <strong>of</strong> h/l for different parameters<br />

<strong>of</strong> the current model. For comparison, measurement<br />

data from [51] is <strong>in</strong>cluded. Parameters used <strong>in</strong> a): h = 1.2 mm,<br />

f = 9.56 GHz, WT = 2 eV, W0 = 2 eV, W1 = 59.1 eV, <strong>and</strong><br />

σse,max = 2.22, (i.e. W1 = Wf2 <strong>and</strong> σse,max = αmax <strong>in</strong> table A-6<br />

for silver <strong>in</strong> [50])). Modified parameters <strong>in</strong>: b) W0 = WT = 4 eV,<br />

c) W1 = 40 eV, <strong>and</strong> d) W1 = 80 eV.<br />

65


The step-like behaviour <strong>of</strong> the <strong>in</strong>creas<strong>in</strong>g threshold is due to the fact<br />

that several multipactor zones are <strong>in</strong>volved (f · h ≈ 11.5 GHz·mm).<br />

Start<strong>in</strong>g with mode number N = 7 for ’current model a)’, the lower<br />

threshold for the parallel-plate case is found <strong>and</strong> as the h/l-ratio <strong>in</strong>creases,<br />

the zone correspond<strong>in</strong>g to N = 7 shr<strong>in</strong>ks until, with an almost<br />

sudden voltage step, the next threshold, be<strong>in</strong>g determ<strong>in</strong>ed by the N = 5<br />

zone, is reached. F<strong>in</strong>ally the last N = 3 zone determ<strong>in</strong>es the threshold<br />

before it also vanishes.<br />

In addition to the experimental comparison, Fig. 4.5 br<strong>in</strong>gs forward<br />

the importance <strong>of</strong> different parameters <strong>of</strong> the current model as well as <strong>of</strong><br />

the used model for SEY [22]. By <strong>in</strong>creas<strong>in</strong>g the <strong>in</strong>itial velocity (’current<br />

model b)’), the overall threshold decreases as a lower field strength will<br />

be sufficient to reach the same impact velocity (cf. Eq. (2.7)). The effect<br />

<strong>of</strong> an <strong>in</strong>creased thermal spread, WT, is that the electron losses <strong>in</strong>creases<br />

<strong>and</strong> the threshold starts to <strong>in</strong>crease for lower h/l-values (also shown<br />

<strong>in</strong> ’current model b)’). By lower<strong>in</strong>g the first cross-over po<strong>in</strong>t (’current<br />

model c)’), the parallel-plate threshold decreases, s<strong>in</strong>ce an additional<br />

zone, N = 9, comes <strong>in</strong>to play. However, as it shr<strong>in</strong>ks away, the threshold<br />

<strong>in</strong>creases <strong>in</strong> a sudden step to the same level as <strong>in</strong> case ’a)’ <strong>and</strong> then it<br />

follows ’a)’ except that the steps occur at higher h/l-values. An <strong>in</strong>creased<br />

first cross-over po<strong>in</strong>t, case ’d)’, shows a change <strong>of</strong> behaviour opposite to<br />

’c)’, except for the parallel-plate threshold as it is still the N = 7 zone<br />

that determ<strong>in</strong>es this threshold.<br />

In the current model a uniform electric field has been used. Due to<br />

the geometry <strong>of</strong> the iris, the actual electric field will tend to be curved<br />

outwards at the edges <strong>of</strong> the slot <strong>in</strong>stead <strong>of</strong> be<strong>in</strong>g straight (cf. Fig. 4.1).<br />

S<strong>in</strong>ce the field amplitude is higher <strong>in</strong> the centre <strong>of</strong> the iris than at the<br />

edges, the Miller force [10], which is proportional to the negative gradient<br />

<strong>of</strong> the square <strong>of</strong> the electric field amplitude, will tend to push the<br />

electrons out <strong>of</strong> the iris. This effect is most important for the higher<br />

order resonances, where several RF-cycles are required to cross the gap.<br />

In addition to the Miller force, the curved electric field will have a component<br />

<strong>in</strong> the z-direction, which, <strong>in</strong> particular for the first order mode,<br />

will drive the electrons toward the iris edges. This means that the electron<br />

losses will be greater than <strong>in</strong> the case <strong>of</strong> a uniform field, which<br />

will lead to an even further <strong>in</strong>crease <strong>of</strong> the multipactor threshold. This<br />

effect should be more pronounced for th<strong>in</strong> irises <strong>and</strong> could expla<strong>in</strong> why<br />

the current model predicts the existence <strong>of</strong> a discharge beyond a certa<strong>in</strong><br />

h/l-ratio where experiments cannot detect it (cf. Fig. 4.5).<br />

66


4.4 Ma<strong>in</strong> results<br />

This analysis has shown that the r<strong>and</strong>om electron drift along the iris<br />

length due to the <strong>in</strong>itial velocity <strong>of</strong> the secondary electrons tends to<br />

significantly <strong>in</strong>crease the multipactor threshold <strong>in</strong> a waveguide iris as<br />

compared to predictions based on the classical two parallel-plate model.<br />

Inherent <strong>in</strong> the presented model is the scal<strong>in</strong>g parameter h/l, which<br />

makes it possible to produce useful multipactor susceptibility charts <strong>in</strong><br />

the traditional eng<strong>in</strong>eer<strong>in</strong>g units. An <strong>in</strong>crease <strong>in</strong> the h/l-ratio results <strong>in</strong><br />

a shr<strong>in</strong>kage <strong>of</strong> each multipactor resonance zone. For each zone, the zone<br />

reduction effect is more pronounced for lower frequencies. A consequence<br />

<strong>of</strong> this is that the multipactor free region <strong>in</strong> the parallel plate model at<br />

low frequency-gap-height products grows with <strong>in</strong>creas<strong>in</strong>g h/l-ratio.<br />

67


Chapter 5<br />

<strong>Multipactor</strong> <strong>in</strong> coaxial l<strong>in</strong>es<br />

The coaxial l<strong>in</strong>e is a very common <strong>and</strong> important component <strong>in</strong> microwave<br />

systems. It is a transmission l<strong>in</strong>e that consists <strong>of</strong> an <strong>in</strong>ner<br />

cyl<strong>in</strong>drical conductor <strong>and</strong> a coaxial outer conductor. A constant crosssection<br />

is ma<strong>in</strong>ta<strong>in</strong>ed by means <strong>of</strong> a dielectric medium, which is conta<strong>in</strong>ed<br />

between the conductors. In some space applications, as well as <strong>in</strong><br />

other systems, the dielectric medium has been partly omitted <strong>in</strong> order to<br />

save weight or to reduce the dielectric losses. In a vacuum environment,<br />

the l<strong>in</strong>e may become evacuated, which makes it exposed to the risk <strong>of</strong> a<br />

multipactor discharge.<br />

The electric field <strong>in</strong> a coaxial l<strong>in</strong>e is nonuniform, which makes analytical<br />

analysis difficult, s<strong>in</strong>ce the the equation <strong>of</strong> motion for an electron becomes<br />

a non-l<strong>in</strong>ear differential equation. However, multipactor <strong>in</strong> a coaxial<br />

l<strong>in</strong>e has been studied experimentally [63] <strong>and</strong> numerically [64–67]. In<br />

these studies it was found that two different types <strong>of</strong> resonant multipactor<br />

can occur, namely a two-sided discharge between the outer <strong>and</strong><br />

the <strong>in</strong>ner conductors <strong>and</strong> the one-sided analogue on the outer conductor.<br />

In an attempt to underst<strong>and</strong> the effect <strong>of</strong> vary<strong>in</strong>g the relative <strong>in</strong>ner<br />

radius, i.e. vary<strong>in</strong>g the characteristic impedance <strong>of</strong> the coaxial l<strong>in</strong>e,<br />

different scal<strong>in</strong>g laws were suggested <strong>in</strong> these studies. Another study focused<br />

on the current due to the multipact<strong>in</strong>g electrons <strong>and</strong> treated this<br />

as a radially oriented Hertzian dipole <strong>in</strong> order to determ<strong>in</strong>e the electric<br />

field generated by the multipactor discharge [68].<br />

In paper E resonant multipactor <strong>in</strong> a coaxial l<strong>in</strong>e is analysed by<br />

means <strong>of</strong> an approximate analytical solution <strong>of</strong> the non-l<strong>in</strong>ear differential<br />

equation <strong>of</strong> motion, which <strong>in</strong> a large range <strong>of</strong> microwave frequencies<br />

<strong>and</strong> amplitudes agrees very well with the numerical solution. As<br />

69


support for the qualitative analytical results, paper F presents PICsimulations<br />

<strong>of</strong> the phenomenon <strong>in</strong> the same geometry. The advantage<br />

<strong>of</strong> PIC-simulations is the ability to <strong>in</strong>clude aspects, which are stochastic<br />

<strong>in</strong> nature, like e.g. the <strong>in</strong>itial velocity spread <strong>of</strong> the secondary electrons.<br />

This summary will focus on the results <strong>of</strong> the study <strong>and</strong> for a<br />

more detailed description <strong>of</strong> the model <strong>and</strong> approximations used, see<br />

papers E <strong>and</strong> F.<br />

5.1 Analytical study<br />

By f<strong>in</strong>d<strong>in</strong>g an analytical solution <strong>of</strong> the equation <strong>of</strong> motion, general properties<br />

<strong>of</strong> the multipactor can be found, which may be difficult to identify<br />

when numerically study<strong>in</strong>g the phenomenon. In addition, the time <strong>of</strong><br />

computation can be radically reduced when us<strong>in</strong>g explicit analytical expressions<br />

<strong>in</strong>stead <strong>of</strong> a numerical scheme. In this section an approximate<br />

analytical solution <strong>of</strong> the non-l<strong>in</strong>ear differential equation that governs<br />

the electron motion <strong>in</strong> the nonuniform field between the <strong>in</strong>ner <strong>and</strong> outer<br />

conductor <strong>of</strong> a coaxial l<strong>in</strong>e is found. Us<strong>in</strong>g these expressions, the effect<br />

on the multipactor resonances <strong>and</strong> thresholds is studied. The validity<br />

<strong>of</strong> the expressions is then confirmed by solv<strong>in</strong>g the differential equation<br />

numerically.<br />

5.1.1 Model<br />

The cyl<strong>in</strong>drical coaxial l<strong>in</strong>e has an outer radius Ro <strong>and</strong> an <strong>in</strong>ner radius Ri<br />

(see Fig. 5.1). The applied field is the fundamental TEM-mode, which<br />

means that the electric field, E, is radially directed <strong>and</strong> the amplitude<br />

will be <strong>in</strong>versely proportional to the distance from the centre <strong>of</strong> the l<strong>in</strong>e.<br />

There will be no dependence on the angle around the coaxial axis, which<br />

means that the problem can be studied as a one dimensional problem,<br />

provided that the effect <strong>of</strong> the magnetic field is neglected <strong>and</strong> only a<br />

cross-section <strong>of</strong> the coaxial l<strong>in</strong>e is considered. In vacuum, the equation<br />

<strong>of</strong> motion for an electron <strong>in</strong> an electric field can be written<br />

mr ′′ = −qE (5.1)<br />

where m is the mass <strong>of</strong> the electron, q the unit charge, <strong>and</strong> E the <strong>in</strong>stantaneous<br />

strength <strong>of</strong> the electric field. The radial position <strong>of</strong> the electron<br />

is designated r <strong>and</strong> r ′′ is the second time derivative <strong>of</strong> the position.<br />

Assum<strong>in</strong>g a time harmonic electric field, E = Eo(Ro/r)s<strong>in</strong> (ωt), where<br />

70


Figure 5.1: The geometry used <strong>in</strong> the considered model.<br />

Eo is the field amplitude at the outer conductor, <strong>and</strong> <strong>in</strong>troduc<strong>in</strong>g the<br />

notation Λ = qEoRo/m, Eq. (5.1) can be written:<br />

r ′′ = − Λ<br />

s<strong>in</strong> ωt (5.2)<br />

r<br />

The relation between the field amplitude <strong>and</strong> the voltage amplitude is<br />

given by Uc = EoRo ln (Ro/Ri). S<strong>in</strong>ce the field is <strong>in</strong>homogeneous <strong>and</strong><br />

stronger near the centre conductor, there will be a net average force that<br />

slowly, compared to the fast harmonic oscillations, pushes the electron<br />

towards the outer conductor. This force is called the ponderomotive or<br />

Miller force [10] <strong>and</strong> it tends to push the electrons away from regions with<br />

high amplitudes <strong>of</strong> the RF electric field. By separat<strong>in</strong>g r(t) accord<strong>in</strong>g<br />

to r(t) = x(t) + R(t), where x(t) is the fast oscillat<strong>in</strong>g motion <strong>and</strong> R(t)<br />

the slowly vary<strong>in</strong>g motion (the time averaged position), an approximate<br />

solution <strong>of</strong> Eq. (5.2) can be derived (see paper E) where the position<br />

<strong>and</strong> velocity <strong>of</strong> the electron are given by:<br />

r(t) ≈ Λ<br />

ω2 �<br />

s<strong>in</strong>(ωt)<br />

+<br />

(5.3)<br />

<strong>and</strong><br />

�<br />

C1(t − C2) 2 + Λ2<br />

2ω 2 C1<br />

C1(t − C2) 2 + Λ2<br />

2ω 2 C1<br />

r ′ (t) ≈ 1<br />

�<br />

C1(t − C2) +<br />

R(t)<br />

Λ<br />

�<br />

cos (ωt)<br />

ω<br />

, (5.4)<br />

71


where R(t) is the average position,<br />

�<br />

R(t) =<br />

C1(t − C2) 2 + Λ2<br />

2ω 2 C1<br />

. (5.5)<br />

The constants <strong>of</strong> <strong>in</strong>tegration, C1 <strong>and</strong> C2, are determ<strong>in</strong>ed by the <strong>in</strong>itial<br />

conditions, which for an electron start<strong>in</strong>g at the outer conductor are<br />

r(t = t0) = Ro <strong>and</strong> r ′ (t = t0) = −v0. Us<strong>in</strong>g Eq. (5.3), the position <strong>of</strong> an<br />

electron emitted from the outer conductor with no <strong>in</strong>itial velocity has<br />

been plotted <strong>in</strong> Fig. 5.2. The accuracy <strong>of</strong> the expression is evident from<br />

the comparison with the numerical solution.<br />

Position [mm]<br />

10<br />

9<br />

8<br />

7<br />

6<br />

5<br />

<strong>Multipactor</strong> <strong>in</strong> coax<br />

0 1 2 3 4<br />

Time [ns]<br />

5 6 7 8<br />

Figure 5.2: Motion <strong>of</strong> an electron emitted from the outer conductor <strong>of</strong> a coaxial<br />

l<strong>in</strong>e. The solid l<strong>in</strong>e corresponds to the analytical expression<br />

Eq. (5.3), the dotted l<strong>in</strong>e is a numerical solution <strong>of</strong> the differential<br />

equation Eq. (5.2) (almost covered by the solid l<strong>in</strong>e) <strong>and</strong> the<br />

dashed l<strong>in</strong>e is the average motion accord<strong>in</strong>g to Eq. (5.5). Parameters<br />

used: Vc = 1200 V, f = 3 GHz, W0 = 0 eV (the <strong>in</strong>itial<br />

electron energy), Ro = 10 mm, <strong>and</strong> Ri = 5 mm.<br />

An important result can be obta<strong>in</strong>ed by only look<strong>in</strong>g at the average<br />

position, Eq. (5.5). The m<strong>in</strong>imum <strong>of</strong> this equation, Rm<strong>in</strong>, is the small-<br />

72


est achievable radial position for an electron emitted from the outer<br />

conductor, provided that the oscillations are not too large. The expression<br />

for the m<strong>in</strong>imum <strong>of</strong> R(t) will be a function <strong>of</strong> the field amplitude,<br />

the frequency, the <strong>in</strong>itial electron velocity as well as the <strong>in</strong>itial phase,<br />

α = ωt0:<br />

Rm<strong>in</strong> = ΛRo(Λ 2 + 2Λ 2 cos α 2 + 4Λcos αv0ωRo + 2v 2 0ω 2 R 2 o) −1/2<br />

(5.6)<br />

However, for v0 = 0 a much more compact expression, which is <strong>in</strong>dependent<br />

<strong>of</strong> field amplitude <strong>and</strong> frequency, is obta<strong>in</strong>ed,<br />

Rm<strong>in</strong> ≈<br />

Ro<br />

� 1 + 2(cos α) 2<br />

≥ Ro<br />

√3 . (5.7)<br />

This means that if the radius <strong>of</strong> the <strong>in</strong>ner conductor, Ri, is smaller than<br />

58% <strong>of</strong> the outer radius, Ro, then two sided multipactor is not possible<br />

when the <strong>in</strong>itial velocity is low <strong>and</strong> the oscillations are small.<br />

5.1.2 <strong>Multipactor</strong> resonance theory<br />

In a coaxial l<strong>in</strong>e, both double-sided <strong>and</strong> s<strong>in</strong>gle-sided multipactor (on the<br />

outer conductor) are possible. First, double-sided discharge will be considered<br />

<strong>and</strong> typical for this is that the one way transit time corresponds<br />

to an odd <strong>in</strong>teger <strong>of</strong> half RF field periods. However, <strong>in</strong> a coaxial l<strong>in</strong>e,<br />

the transit time is normally longer for electrons emitted from the outer<br />

conductor than for electrons emitted from the <strong>in</strong>ner conductor. Thus,<br />

the sum <strong>of</strong> two transits must be considered <strong>and</strong> the condition for this is<br />

that it should be an <strong>in</strong>teger number <strong>of</strong> RF periods. This is the resonance<br />

criterion <strong>and</strong> <strong>in</strong> addition to this the phase-focus<strong>in</strong>g effect should be active,<br />

which for the parallel-plate case is given by Eqs. (2.18)- (2.21). It<br />

is <strong>in</strong>structive to compare the coaxial case with the parallel-plate case,<br />

s<strong>in</strong>ce <strong>in</strong> the limit when the Ri ≈ Ro the coaxial <strong>and</strong> parallel-plate models<br />

should give the same results. For the parallel-plate case, when the<br />

<strong>in</strong>itial velocity is neglected (v0 = 0), the phase stability range is given<br />

by the follow<strong>in</strong>g <strong>in</strong>equalities [39],<br />

πk < λ < � (πk) 2 + 4, (5.8)<br />

where k is an odd positive <strong>in</strong>teger. The normalised gap width, λ, is<br />

def<strong>in</strong>ed by<br />

λ = ωd/Vω = m(ωd) 2 /eU, (5.9)<br />

73


where d st<strong>and</strong>s for the gap width, Vω = eEω/mω is the amplitude <strong>of</strong><br />

the electron velocity oscillations <strong>in</strong> the spatially uniform RF field, Eω is<br />

the RF field amplitude <strong>in</strong>side the gap, <strong>and</strong> U is the voltage between the<br />

conductors. In addition, for an electron avalanche to start, the impact<br />

velocity should be between the first <strong>and</strong> the second cross-over po<strong>in</strong>ts<br />

(cf. Eq. (2.12)), which for zero <strong>in</strong>itial velocity is given by,<br />

v1 < 2Vω < v2 . (5.10)<br />

Due to the asymmetry <strong>in</strong> the electron motion, the simple analytical<br />

analysis that is feasible <strong>in</strong> the parallel plate case is not applicable for<br />

the coaxial l<strong>in</strong>e, despite the fact that the approximate electron position<br />

<strong>and</strong> velocity are known explicitly (Eqs. (5.3) <strong>and</strong> (5.4)). One way <strong>of</strong><br />

f<strong>in</strong>d<strong>in</strong>g the resonance zones <strong>in</strong> the parameter space is to compute a series<br />

<strong>of</strong> successive electron trajectories, search<strong>in</strong>g for the conditions when it<br />

converges to a periodically repeated sequence [69]. In Fig. 5.3 one can<br />

see the result <strong>of</strong> such a method, where stable resonance zones have been<br />

found numerically. To allow simple comparison with the parallel-plate<br />

case, the normalised gap size has been used <strong>and</strong> <strong>in</strong> terms <strong>of</strong> the coaxial<br />

l<strong>in</strong>e it becomes,<br />

λ = m(ωd)2<br />

eUc<br />

= G(Ro − Ri) 2<br />

R2 , (5.11)<br />

o ln (Ro/Ri)<br />

which co<strong>in</strong>cides with Eq. (5.9) when Ro/Ri is close to unity. The convenient<br />

parameter, G = ωRo/Vω,o, has been <strong>in</strong>troduced as represent<strong>in</strong>g<br />

a normalised outer radius <strong>and</strong> Vω,o = qEo/mω.<br />

When Ro/Ri is close to unity the parallel-plate <strong>and</strong> coaxial models<br />

give similar results. When the ratio becomes larger, the zones deviate<br />

from the straight l<strong>in</strong>es predicted by Eq. (5.8) <strong>and</strong> the deviation is more<br />

pronounced for the higher modes. When the ratio becomes too large,<br />

all two sided resonances disappear. This is a consequence <strong>of</strong> the Miller<br />

force <strong>and</strong> for the higher order modes, where the approximate analytical<br />

solution is very accurate, the prediction (Eq. (5.7)) is that the double<br />

sided resonances should disappear roughly at Ro/Ri = √ 3 ≈ 1.73. Figure<br />

5.3 shows that this is <strong>in</strong>deed true. The first order mode, however, is<br />

not accurately described by the analytical expression <strong>and</strong> the numerical<br />

calculations show that this mode can exist for values <strong>of</strong> Ro/Ri as high<br />

as 4.<br />

In Fig. 5.4 the double-sided discharge regions have been computed<br />

numerically for Ro/Ri = 1.4. The classical multipactor zones have upper<br />

<strong>and</strong> lower thresholds that satisfy Eq. (5.10) <strong>in</strong> the follow<strong>in</strong>g sense: s<strong>in</strong>ce<br />

74


λ<br />

22<br />

20<br />

18<br />

16<br />

14<br />

12<br />

10<br />

8<br />

6<br />

4<br />

7 8<br />

5 c<br />

4<br />

3 3 c<br />

6<br />

Double sided multipactor, numerical data<br />

2 h<br />

5<br />

1 c<br />

1 1.5 2 2.5 3 3.5 4<br />

R /R [−]<br />

o i<br />

Figure 5.3: Normalised gap width accord<strong>in</strong>g to Eq. (5.11) vs. Ro/Ri. The<br />

solid straight l<strong>in</strong>es form the classical zones accord<strong>in</strong>g to Eq. (5.8).<br />

The dots (blue) <strong>in</strong>dicate stable resonances where the sum <strong>of</strong> two<br />

transits equal an odd number <strong>of</strong> RF-cycles. The crosses (red) are<br />

for sums equal to an even number <strong>of</strong> RF-cycles. The sum <strong>of</strong> the<br />

transit time <strong>in</strong> RF-cycles for some dist<strong>in</strong>ct zones are <strong>in</strong>dicated.<br />

The lowest order hybrid modes are marked with an ’h’ <strong>and</strong> the<br />

classical resonances with a ’c’. The dashed vertical l<strong>in</strong>e <strong>in</strong>dicates<br />

Ro/Ri = √ 3 ≈ 1.73.<br />

75


there are different oscillatory velocities depend<strong>in</strong>g on whether the electron<br />

starts on the outer or the <strong>in</strong>ner conductor, an average value must<br />

be computed. By sett<strong>in</strong>g Vω,o = qEo/mω = v1/2 the correspond<strong>in</strong>g voltage,<br />

VRo, can be derived. Similarly, by sett<strong>in</strong>g Vω,i = qEi/mω = v1/2<br />

the voltage VRi can be obta<strong>in</strong>ed. The lower threshold voltage is then<br />

approximately equal to:<br />

Uth ≈ VRo + VRi<br />

. (5.12)<br />

2<br />

The upper threshold is computed <strong>in</strong> a similar manner, only with v2<br />

<strong>in</strong>stead <strong>of</strong> v1. The hybrid modes, which <strong>in</strong> general have a lower average<br />

impact velocity will require a stronger electric field to reach an energy<br />

equal to the first cross-over po<strong>in</strong>t. Consequently, a threshold higher than<br />

the lower envelope is obta<strong>in</strong>ed for these zones (cf. Fig. 5.4).<br />

Amplitude <strong>of</strong> Conductor Voltage [V]<br />

10 4<br />

10 3<br />

10 2<br />

1 c<br />

2 h<br />

Double sided multipactor: Z c = 20Ω<br />

3 c<br />

10 0<br />

4<br />

Frequency [GHz]<br />

Figure 5.4: Numerically obta<strong>in</strong>ed double sided multipactor chart. The<br />

dashed l<strong>in</strong>es are the approximate lower <strong>and</strong> upper envelopes given<br />

by Eq. (5.12). The sum <strong>of</strong> the transit time <strong>in</strong> RF-cycles for each<br />

zone is <strong>in</strong>dicated. Parameters used: W1 = 23 eV, W2 = 2100 eV,<br />

W0 = 0 eV, Ro = 10 mm, <strong>and</strong> Ri = 7.2 mm.<br />

If the <strong>in</strong>ner radius is sufficiently small, s<strong>in</strong>gle-sided multipactor becomes<br />

the dom<strong>in</strong>at<strong>in</strong>g scenario. S<strong>in</strong>gle-sided multipactor is less complicated<br />

than its double-sided counterpart, as the complicated asymmetry<br />

does not appear <strong>in</strong> this case. This allows an analytical analysis based<br />

76<br />

7<br />

6<br />

10 1


on the approximate solution for the electron trajectory, Eq. (5.3). The<br />

analytical expression is accurate when the oscillations are small <strong>and</strong> a<br />

consequence <strong>of</strong> this is that accurate stable phase multipactor regions are<br />

only found for N ≥ 2, i.e. when the duration <strong>of</strong> the trajectory is at least<br />

2 RF-periods.<br />

By analys<strong>in</strong>g the resonance <strong>and</strong> stability conditions, one can show<br />

that s<strong>in</strong>gle-sided breakdown will have not only one region <strong>of</strong> stable resonant<br />

phase, but rather two stable regions can be found. One somewhat<br />

wider region with resonance close to zero <strong>and</strong> another, which is resonant<br />

close to π/4. It can be shown that these regions, <strong>in</strong> the case when<br />

v0 = 0, are approximately given by:<br />

<strong>and</strong><br />

0 < αR < α1<br />

α2 < αR < α3<br />

(5.13)<br />

(5.14)<br />

where<br />

α1 ≈ 4<br />

(5.15)<br />

Nπ<br />

α2 ≈ π 1<br />

− (5.16)<br />

4 Nπ<br />

α3 ≈ π 1<br />

+ (5.17)<br />

4 Nπ<br />

For <strong>in</strong>creas<strong>in</strong>g N, the second region converges to αR = π/4, which accord<strong>in</strong>g<br />

to Eq. (5.7) corresponds to Rm<strong>in</strong> ≈ Ro/ √ 2. In Fig. 5.5 the<br />

resonant stable phase, αR, has been plotted as a function <strong>of</strong> N. Except<br />

for the lowest order resonance (N = 1), the regions <strong>of</strong> phase stability for<br />

the numerically <strong>and</strong> analytically obta<strong>in</strong>ed phases agree very well.<br />

To obta<strong>in</strong> the multipactor threshold, it is necessary to know the<br />

impact velocity, which is given by<br />

vimpact ≈ 2Vω,o cos α + v0<br />

(5.18)<br />

The lower boundary shown <strong>in</strong> Fig. 5.6 is obta<strong>in</strong>ed for the maximum<br />

impact velocity for each mode, i.e. vimpact = 2Vω,o when v0 = 0. In the<br />

same figure one can also identify a second set <strong>of</strong> regions with a higher<br />

breakdown threshold. These zones correspond to the second stable phase<br />

region, Eq. (5.14). S<strong>in</strong>ce the phases <strong>in</strong> this region are close to π/4,<br />

the impact velocity is vimpact ≈ √ 2Vω,o. This value is also <strong>in</strong>dicated <strong>in</strong><br />

Fig. 5.6, but it should not be expected to serve as an exact envelope <strong>of</strong> the<br />

77


Phase [degrees]<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

α 3<br />

0<br />

0 5 10<br />

N (rf−cycles)<br />

15 20<br />

α 2<br />

α 1<br />

Figure 5.5: Stable resonant phase for s<strong>in</strong>gle-sided multipactor. Analytically<br />

obta<strong>in</strong>ed stable phases are shown as diamonds (red) <strong>and</strong> the ones<br />

obta<strong>in</strong>ed numerically are <strong>in</strong>dicated with dots (blue). Eqs. (5.15) -<br />

(5.17) are shown as solid l<strong>in</strong>es. The dashed l<strong>in</strong>e <strong>in</strong>dicates α =<br />

π/4 .<br />

zones as there are phases that are smaller than π/4 (cf. Fig. 5.5), which<br />

will yield a higher impact velocity <strong>and</strong> consequently a lower threshold.<br />

Furthermore, <strong>in</strong> the numerical solution <strong>of</strong> Eq. (5.2) for the first order<br />

mode, an impact velocity <strong>of</strong> as much as four times Vω,o can be observed.<br />

This results <strong>in</strong> a threshold much lower than the envelope. The maximum<br />

impact velocity <strong>of</strong> the second zone is slightly lower than 2Vω,o, result<strong>in</strong>g<br />

<strong>in</strong> a somewhat higher threshold. The follow<strong>in</strong>g higher order modes then<br />

quickly converge to the analytical limit (cf. Fig. 5.6).<br />

When the <strong>in</strong>itial velocity is zero, the only parameters left to vary are<br />

G <strong>and</strong> the ratio Ro/Ri. S<strong>in</strong>ce the characteristic impedance <strong>in</strong> ohms <strong>of</strong><br />

a coaxial l<strong>in</strong>e <strong>in</strong> vacuum is given by Z ≈ 60ln (Ro/Ri), it follows that<br />

only two parameters rema<strong>in</strong> to be varied, viz. G <strong>and</strong> Z. By follow<strong>in</strong>g<br />

trajectories for different values <strong>of</strong> G <strong>and</strong> Z, stable phase po<strong>in</strong>ts were<br />

found <strong>in</strong> this parameter space <strong>and</strong> the result is plotted <strong>in</strong> Fig. 5.7, which<br />

was produced us<strong>in</strong>g a numerical solution <strong>of</strong> the equation <strong>of</strong> motion (a<br />

version <strong>of</strong> this figure us<strong>in</strong>g the analytical expressions can be found <strong>in</strong><br />

paper E).<br />

The straight l<strong>in</strong>es <strong>in</strong> Fig. 5.7 on the right h<strong>and</strong> side are regions <strong>of</strong><br />

stable s<strong>in</strong>gle-sided resonances. The fact that these appear as straight<br />

78


Amplitude <strong>of</strong> Conductor Voltage/ln(R 0 /R i ) [V]<br />

10 5<br />

10 4<br />

10 3<br />

10 2<br />

10 0<br />

S<strong>in</strong>gle sided <strong>Multipactor</strong>: Z c =100 Ω<br />

10 1<br />

Frequency x R o [GHz ⋅ mm]<br />

Figure 5.6: S<strong>in</strong>gle-sided multipactor breakdown regions based on the numerical<br />

solution <strong>of</strong> Eq. (5.2). Regions correspond<strong>in</strong>g to N = 1 − 22<br />

RF-periods are shown. The regions with an <strong>in</strong>itial phase, α, close<br />

to zero are produced us<strong>in</strong>g dots (blue) <strong>and</strong> the regions with α close<br />

to π/4 are <strong>in</strong>dicated by crosses (red). The dash-dot l<strong>in</strong>e is the<br />

approximate lower envelope given by Vω,o = v1/2 <strong>and</strong> the dashed<br />

l<strong>in</strong>e is given by Vω,o = v1/ √ 2. Parameters used: W1 = 23 eV,<br />

W2 = 2100 eV, W0 = 0 eV, <strong>and</strong> Z = 100 Ω (correspond<strong>in</strong>g to<br />

e.g. Ro = 10 mm <strong>and</strong> Ri = 1.88 mm.)<br />

10 2<br />

79


G<br />

100<br />

90<br />

80<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

0<br />

0 0.5 1 1.5 2<br />

Z/Z (Z =50 Ω)<br />

0 0<br />

Figure 5.7: The normalised parameter G vs. normalised characteristic<br />

impedance Z. Each mark represents a stable phase solution <strong>and</strong><br />

an effort has been made to suppress polyphase modes <strong>in</strong> order<br />

to clearly show the behaviour <strong>of</strong> the ma<strong>in</strong> resonance modes. The<br />

chart was obta<strong>in</strong>ed by numerically solv<strong>in</strong>g the equation <strong>of</strong> motion.<br />

Stars mark (blue): double-sided multipactor, dots (red):<br />

s<strong>in</strong>gle-sided multipactor with 0 < αR < 20 o , <strong>and</strong> crosses (green):<br />

s<strong>in</strong>gle-sided multipactor with αR > 20 o . The dashed l<strong>in</strong>e <strong>in</strong>dicates<br />

Ri,m<strong>in</strong> = Ro/ √ 3 <strong>and</strong> the dash-dot l<strong>in</strong>e Ri,m<strong>in</strong> = Ro/ √ 2.<br />

80


l<strong>in</strong>es <strong>in</strong>dicates that there is no dependence on Z, which implies that a<br />

simple scal<strong>in</strong>g law exists <strong>in</strong> the s<strong>in</strong>gle-sided case, viz.<br />

P ∝ (ωRo) 4 Z. (5.19)<br />

In Fig. 5.6 this law has been used to normalise the axes, but voltage<br />

is used on the ord<strong>in</strong>ate <strong>in</strong>stead <strong>of</strong> power. The chosen normalisation<br />

<strong>of</strong> the axes <strong>in</strong> Fig. 5.6 is general <strong>and</strong> us<strong>in</strong>g the analytical solution <strong>of</strong><br />

Eq. (5.2) presented above, it can be shown that this normalisation is<br />

valid also for non-zero <strong>in</strong>itial velocity. It is important, however, to be<br />

careful when scal<strong>in</strong>g to a different radii ratio, s<strong>in</strong>ce for smaller values<br />

<strong>of</strong> the characteristic impedance, the s<strong>in</strong>gle-sided multipactor zones may<br />

not exist at all.<br />

In the double-sided case, it is evident that G is a function <strong>of</strong> Z.<br />

Consequently a more complicated scal<strong>in</strong>g law should be expected. For<br />

small values <strong>of</strong> Z, however, the coaxial case becomes similar to the<br />

parallel-plate geometry, where the resonance voltage can be written as<br />

function <strong>of</strong> the frequency-gap-size product. For the coaxial case, this<br />

scal<strong>in</strong>g law becomes<br />

P ∝ (ω(Ro − Ri)) 4 1<br />

. (5.20)<br />

Z<br />

<strong>and</strong> for the first order resonance this scal<strong>in</strong>g law is quite accurate (cf.<br />

Fig. 5.3), but for the higher order modes it quickly loses its validity with<br />

<strong>in</strong>creas<strong>in</strong>g Z.<br />

5.1.3 Ma<strong>in</strong> f<strong>in</strong>d<strong>in</strong>gs<br />

A qualitative comparison with experiments [63] shows good agreement<br />

with the present analysis. The experimental data shows an <strong>in</strong>crease <strong>in</strong><br />

the multipactor threshold for <strong>in</strong>creas<strong>in</strong>g radii-ratio Ro/Ri. It was also<br />

found that the first multipactor zone became narrower for <strong>in</strong>creas<strong>in</strong>g<br />

Ro/Ri. These features are <strong>in</strong> agreement with the results <strong>of</strong> this study<br />

as shown <strong>in</strong> Figs. 5.3 <strong>and</strong> 5.7. By mapp<strong>in</strong>g the data <strong>of</strong> Fig. 5.3 <strong>in</strong>to the<br />

voltage vs. frequency-gap-size space, used <strong>in</strong> the experiments, a clear<br />

threshold <strong>in</strong>crease compared with the parallel-plate case can be seen as<br />

well. Even though the experiments used quite large values <strong>of</strong> Ro/Ri,<br />

no case <strong>of</strong> s<strong>in</strong>gle-sided multipactor was observed. This can be expla<strong>in</strong>ed<br />

by the fact that a material with a low first cross-over po<strong>in</strong>t was used,<br />

where the <strong>in</strong>itial velocity will play an important role when the applied<br />

voltage is not high enough. More importantly, only the first order mode<br />

was studied <strong>and</strong> <strong>in</strong> this case, as shown <strong>in</strong> Fig. 5.3, multipactor will be<br />

81


possible for quite large Ro/Ri-values. In the simulations by Sakamoto et<br />

al [64], the experimental results were confirmed. In addition, s<strong>in</strong>gle-sided<br />

multipactor was observed, which confirms the result <strong>of</strong> this theoretical<br />

study.<br />

Among the more important results <strong>of</strong> this part <strong>of</strong> the study are the<br />

follow<strong>in</strong>g. The analytical approximate solution <strong>of</strong> the nonl<strong>in</strong>ear differential<br />

equation <strong>of</strong> motion for an electron <strong>in</strong> a coaxial l<strong>in</strong>e. The dual regions<br />

<strong>of</strong> stable phase, Eqs. (5.13) - (5.17), which expla<strong>in</strong> why s<strong>in</strong>gle-sided multipactor<br />

will be possible also for smaller values <strong>of</strong> Ro/Ri. The scal<strong>in</strong>g<br />

law for s<strong>in</strong>gle-sided multipactor, Eq. (5.19), simplifies the presentation<br />

<strong>of</strong> multipactor prone regions <strong>of</strong> the s<strong>in</strong>gle-sided case. The limit formula<br />

for the transition from double- to s<strong>in</strong>gle-sided multipactor, Eq. (5.7), is<br />

an <strong>in</strong>terest<strong>in</strong>g feature for future experiments to confirm. The reduced<br />

threshold for the first order zone <strong>of</strong> s<strong>in</strong>gle-sided multipactor, which must<br />

be taken <strong>in</strong>to account when construct<strong>in</strong>g the lower boundary <strong>of</strong> all the<br />

zones. F<strong>in</strong>ally, this analysis shows that the behaviour <strong>of</strong> resonant multipactor<br />

is significantly affected by the nonuniform field <strong>and</strong> it shows the<br />

benefits <strong>of</strong> analytically study<strong>in</strong>g different geometries to underst<strong>and</strong> the<br />

basic behaviours before perform<strong>in</strong>g numerical simulations.<br />

5.2 Particle-<strong>in</strong>-cell simulations<br />

In this part <strong>of</strong> the coaxial study, which is based on paper F, extensive<br />

PIC-simulations have been performed <strong>in</strong> order to verify the analytical<br />

results as well as to <strong>in</strong>vestigate the importance <strong>of</strong> <strong>in</strong>itial velocity spread<br />

<strong>and</strong> different maximum secondary emission. One <strong>of</strong> the advantages with<br />

PIC-simulations is the ability to <strong>in</strong>clude parameters that are stochastic<br />

<strong>in</strong> nature. The stochastic properties <strong>of</strong> some <strong>of</strong> the parameters as well<br />

as the actual value <strong>of</strong> the maximum secondary emission coefficient may<br />

have significant effects on the multipactor threshold as well as on the<br />

existence <strong>of</strong> a discharge. This has previously been shown <strong>in</strong> the case<br />

<strong>of</strong> plane-parallel geometry [46] <strong>and</strong> it is demonstrated that the effect is<br />

similar also <strong>in</strong> coaxial geometry.<br />

5.2.1 Numerical implementation<br />

The geometry <strong>and</strong> field is described <strong>in</strong> the analytical section. The code<br />

uses normalised parameters such as G <strong>and</strong> λ <strong>and</strong> the SEY follows the<br />

model by Vaughan [22]. The <strong>in</strong>itial velocities <strong>of</strong> the secondary electrons<br />

82


are assumed to have a Maxwellian distribution, i.e.<br />

f(vx,vy,vz) ∝ vn<br />

v exp<br />

�<br />

− 1 v<br />

( )<br />

2 vT<br />

2<br />

�<br />

(5.21)<br />

where v is the absolute value <strong>of</strong> the <strong>in</strong>itial velocity, vn its normal component<br />

with respect to the surface <strong>of</strong> emission, <strong>and</strong> vT is the thermal<br />

<strong>in</strong>itial velocity spread. Another <strong>of</strong> the used parameters related to this<br />

is the normalised spread <strong>of</strong> <strong>in</strong>itial electron velocity def<strong>in</strong>ed as vT/vmax,<br />

where vmax is the impact velocity for maximum SEY.<br />

Calculations were performed for 2-D arrays <strong>of</strong> different sets <strong>of</strong> the<br />

normalised parameters (e.g. ρ = Vω/vmax vs. λ with the other parameters<br />

fixed) <strong>and</strong> each run corresponds to one particular po<strong>in</strong>t <strong>in</strong> one <strong>of</strong><br />

these arrays. Each run was primed with 200 seed electrons, uniformly<br />

distributed over <strong>in</strong>itial phase, <strong>and</strong> the run was term<strong>in</strong>ated when either<br />

the number <strong>of</strong> particles exceeded 4500 or when 200 RF-periods had<br />

elapsed. The run was also term<strong>in</strong>ated <strong>in</strong> case the number <strong>of</strong> electrons<br />

dropped below 10 before 200 RF-cycles had passed. At the end <strong>of</strong> each<br />

run, the follow<strong>in</strong>g parameters were recorded:<br />

• Number <strong>of</strong> RF-periods needed to exceed 4500 particles. If 4500<br />

particles were not atta<strong>in</strong>ed with<strong>in</strong> 200 RF-cycles, this parameter<br />

was set to 200.<br />

• Number <strong>of</strong> electrons at the end <strong>of</strong> each run.<br />

• Heat<strong>in</strong>g asymmetry, i.e. the ratio between the average power deposited<br />

on the <strong>in</strong>ner conductor <strong>and</strong> the average power deposited<br />

on the outer conductor.<br />

• Average electron growth rate (over the 10 last RF-periods), normalised<br />

with respect to the RF-period.<br />

5.2.2 Simulations<br />

To facilitate comparison between the theoretical result presented <strong>in</strong><br />

Fig. 5.7 a simulation was made <strong>in</strong> the same parameter space. Figure 5.8<br />

shows the number <strong>of</strong> electrons obta<strong>in</strong>ed after 200 RF-cycles. As expected,<br />

the lower order resonances (i.e. at lower <strong>and</strong> leftmost G-values)<br />

<strong>in</strong>dicate high electron numbers, s<strong>in</strong>ce more impacts with the conductors<br />

will occur dur<strong>in</strong>g the same number <strong>of</strong> RF-periods. Due to this fact, a<br />

parameter space was chosen, which did not <strong>in</strong>clude any po<strong>in</strong>ts <strong>in</strong> the<br />

83


upper right region <strong>of</strong> the figures, where the electron growth is very slow.<br />

There is good agreement <strong>in</strong> the general behaviour <strong>and</strong> the transition<br />

from double-sided to s<strong>in</strong>gle-sided multipactor occurs at more or less the<br />

same impedances for the different zones <strong>in</strong> both the PIC-simulation <strong>and</strong><br />

the theoretical data, s<strong>in</strong>ce the non-zero <strong>in</strong>itial velocity <strong>in</strong> the PIC-data<br />

is small relative to the oscillatory velocity.<br />

G<br />

100<br />

90<br />

80<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6<br />

Z/Z (Z =50 Ω)<br />

0 0<br />

Figure 5.8: Number <strong>of</strong> electrons after 200 RF-cycles. The white dots are<br />

po<strong>in</strong>ts <strong>of</strong> 2-sided multipactor, the green <strong>and</strong> yellow dots 1-sided<br />

multipactor, <strong>and</strong> the white dashed l<strong>in</strong>es correspond to Ri,m<strong>in</strong> =<br />

Ro/ √ 2 (left) <strong>and</strong> Ri,m<strong>in</strong> = Ro/ √ 3 (right) - all from Fig. 5.7.<br />

Parameters used: σse,max = 1.6, W1 = 50 eV, vT/vmax = 0.01,<br />

ρ = Vω,o/vmax = 0.5, <strong>and</strong> f = 1.5 GHz.<br />

The straight l<strong>in</strong>es on the right h<strong>and</strong> side <strong>of</strong> Fig. 5.8 reveal that this<br />

should be s<strong>in</strong>gle-sided multipactor. This is confirmed by look<strong>in</strong>g at the<br />

ratio <strong>of</strong> power deposited on the <strong>in</strong>ner <strong>and</strong> the outer conductors, which<br />

directly identifies the type <strong>of</strong> discharge. In Fig. 5.9, the dark blue areas<br />

are regions where most or all power is deposited on the outer conductor,<br />

which implies s<strong>in</strong>gle-sided multipactor on this conductor. The orange<br />

84<br />

4500<br />

4000<br />

3500<br />

3000<br />

2500<br />

2000<br />

1500<br />

1000<br />

500


<strong>and</strong> yellow areas <strong>in</strong>dicate that a similar amount <strong>of</strong> power is deposited<br />

on both conductors <strong>and</strong> thus the double-sided scenario dom<strong>in</strong>ates.<br />

G<br />

100<br />

90<br />

80<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6<br />

Z/Z (Z =50 Ω)<br />

0 0<br />

Figure 5.9: Ratio <strong>of</strong> power deposited on the <strong>in</strong>ner <strong>and</strong> outer conductors due<br />

to electron impacts, logarithmic scale log 10(P<strong>in</strong>ner/Pouter). The<br />

same parameters are used as <strong>in</strong> Fig. 5.8.<br />

S<strong>in</strong>ce the theoretically obta<strong>in</strong>ed po<strong>in</strong>ts agree <strong>in</strong> general with the PICsimulations,<br />

this confirms the validity <strong>of</strong> the scal<strong>in</strong>g laws, Eqs. (5.19) <strong>and</strong><br />

(5.20). The two different types <strong>of</strong> modes for s<strong>in</strong>gle-sided discharge, one<br />

with a phase α ≈ 0 <strong>and</strong> the other with α ≈ π/4 can also be seen. The<br />

former type <strong>of</strong> mode is discont<strong>in</strong>ued before reach<strong>in</strong>g the first dashed l<strong>in</strong>e<br />

from the right h<strong>and</strong> side <strong>and</strong> the latter before the other dashed l<strong>in</strong>e <strong>in</strong><br />

Fig. 5.7. This is also evident <strong>in</strong> the PIC-data, especially for values <strong>of</strong> G<br />

between 30 <strong>and</strong> 40 <strong>in</strong> Fig. 5.9 where they are fairly well separated <strong>and</strong><br />

only the lower <strong>of</strong> the paired b<strong>and</strong>s extend <strong>in</strong>to the region between the<br />

dashed l<strong>in</strong>es, as predicted.<br />

In Figs. 5.8 <strong>and</strong> 5.9 the multipactor threshold can not be identified,<br />

s<strong>in</strong>ce the oscillatory velocity is kept constant. In Figs. 5.10 <strong>and</strong> 5.11,<br />

however, the oscillatory velocity has been swept for different G-values<br />

while keep<strong>in</strong>g the ratio Ri/Ro constant <strong>and</strong> equal to 0.7, i.e. Z = 21.4 Ω<br />

(Z/Z0 = 0.428). Each figure is produced for a different maximum SEY.<br />

When σse,max is low, the ability to compensate for losses is weak <strong>and</strong> the<br />

zones are well def<strong>in</strong>ed <strong>and</strong> fairly narrow (cf. Fig. 5.10). With <strong>in</strong>creas<strong>in</strong>g<br />

1<br />

0.5<br />

0<br />

−0.5<br />

−1<br />

−1.5<br />

−2<br />

−2.5<br />

85


σse,max the zones become wider <strong>and</strong> zones previously suppressed by the<br />

losses can appear (cf. Fig. 5.11). This behavior is very similar to that<br />

noted for the parallel plate geometry [46]. The lower (left) envelope is<br />

G<br />

50<br />

45<br />

40<br />

35<br />

30<br />

25<br />

20<br />

15<br />

10<br />

5<br />

0.2 0.4 0.6 0.8 1<br />

V /V<br />

ω max<br />

1.2 1.4 1.6 1.8<br />

Figure 5.10: Number <strong>of</strong> electrons after 200 rf-cycles. The vertical straight<br />

l<strong>in</strong>es <strong>in</strong>dicate the lower <strong>and</strong> upper theoretical envelopes accord<strong>in</strong>g<br />

to Eq. (5.12). Parameters used: Ri/Ro = 0.7, f = 1.5 GHz,<br />

σse,max = 1.3, vT/vmax = 0.01, <strong>and</strong> W1 = 50 eV.<br />

obeyed, but the upper can be exceeded s<strong>in</strong>ce a non-zero <strong>in</strong>itial phase will<br />

yield a lower impact velocity <strong>and</strong> a concomitant higher upper threshold.<br />

In Figs. 5.10 <strong>and</strong> 5.11 the velocity spread is quite low, vT/vmax =<br />

0.01. When <strong>in</strong>creas<strong>in</strong>g this ratio, a greater portion <strong>of</strong> the electrons<br />

can have a large negative <strong>in</strong>itial phase, which leads to overlapp<strong>in</strong>g <strong>of</strong><br />

the multipactor regions when σse,max is large (see Fig. 5.12). On the<br />

other h<strong>and</strong>, the <strong>in</strong>creased velocity spread also <strong>in</strong>creases the losses, which<br />

especially affects the higher order resonances, s<strong>in</strong>ce the phase-focus<strong>in</strong>g<br />

effect gets weaker with <strong>in</strong>creas<strong>in</strong>g mode order. If σse,max is low, the<br />

losses are not sufficiently compensated for <strong>and</strong> this leads to suppression<br />

<strong>of</strong> the higher order modes (cf. Fig. 5.13). This is a result similar to<br />

that found <strong>in</strong> the parallel plate case [46]. The correspond<strong>in</strong>g behaviour<br />

is seen also for s<strong>in</strong>gle-sided multipactor (see paper F).<br />

For s<strong>in</strong>gle-sided multipactor it was noted that for the first order<br />

mode, the envelope <strong>of</strong> the breakdown zones Vω,o = v1/2 was not obeyed.<br />

This is confirmed by the PIC-simulations, where the first order mode<br />

86<br />

4500<br />

4000<br />

3500<br />

3000<br />

2500<br />

2000<br />

1500<br />

1000<br />

500


G<br />

50<br />

45<br />

40<br />

35<br />

30<br />

25<br />

20<br />

15<br />

10<br />

5<br />

0.2 0.4 0.6 0.8 1<br />

V /V<br />

ω max<br />

1.2 1.4 1.6 1.8<br />

Figure 5.11: Same as Fig. 5.10 only with σse,max = 2.0.<br />

G<br />

50<br />

45<br />

40<br />

35<br />

30<br />

25<br />

20<br />

15<br />

10<br />

5<br />

0.2 0.4 0.6 0.8 1<br />

V /V<br />

ω max<br />

1.2 1.4 1.6 1.8<br />

Figure 5.12: Number <strong>of</strong> electrons after 200 rf-cycles. The vertical straight<br />

l<strong>in</strong>es <strong>in</strong>dicate the lower <strong>and</strong> upper theoretical envelopes accord<strong>in</strong>g<br />

to Eq. (5.12). Parameters used: Ri/Ro = 0.7, f = 1.5 GHz,<br />

σse,max = 2.0, vT/vmax = 0.1, <strong>and</strong> W1 = 50 eV.<br />

4500<br />

4000<br />

3500<br />

3000<br />

2500<br />

2000<br />

1500<br />

1000<br />

500<br />

4500<br />

4000<br />

3500<br />

3000<br />

2500<br />

2000<br />

1500<br />

1000<br />

500<br />

87


G<br />

50<br />

45<br />

40<br />

35<br />

30<br />

25<br />

20<br />

15<br />

10<br />

5<br />

0.2 0.4 0.6 0.8 1<br />

V /V<br />

ω max<br />

1.2 1.4 1.6 1.8<br />

Figure 5.13: Same as Fig. 5.12 only with σse,max = 1.3.<br />

clearly passes this limit (see Fig. 5.14).<br />

5.2.3 Comparison with experiments<br />

As mentioned <strong>in</strong> the <strong>in</strong>troduction to this chapter, PIC-simulations can<br />

<strong>in</strong>clude aspects <strong>of</strong> the multipactor, which are difficult to analyse theoretically.<br />

As a more realistic description is possible, quantitative comparison<br />

with experiments becomes feasible.<br />

In the experimental study by Woo [63] coaxial l<strong>in</strong>es made <strong>of</strong> copper<br />

were used. When tak<strong>in</strong>g values for the secondary electron emission properties<br />

for copper from the ESA st<strong>and</strong>ard [50], the lower thresholds <strong>of</strong> the<br />

PIC-simulations are not <strong>in</strong> very good agreement with the experimental<br />

data. However, the secondary emission properties can vary a great deal<br />

between different samples <strong>of</strong> the same material <strong>and</strong> contam<strong>in</strong>ation can<br />

reduce the first cross-over po<strong>in</strong>t <strong>and</strong> <strong>in</strong>crease the maximum SEY. Thus<br />

by slightly lower<strong>in</strong>g the first cross-over po<strong>in</strong>t <strong>in</strong> the PIC-simulations,<br />

very good agreement is obta<strong>in</strong>ed (see Fig. 5.15). In paper F an alternative<br />

method <strong>of</strong> obta<strong>in</strong><strong>in</strong>g the experimental threshold when us<strong>in</strong>g the<br />

secondary emission properties given <strong>in</strong> the ESA st<strong>and</strong>ard [50] is presented,<br />

based on a modification <strong>of</strong> the used model for the SEY [22].<br />

However, this will not be discussed further <strong>in</strong> this summary.<br />

In the experiments an <strong>in</strong>crease <strong>in</strong> the multipactor threshold for de-<br />

88<br />

4500<br />

4000<br />

3500<br />

3000<br />

2500<br />

2000<br />

1500<br />

1000<br />

500


G<br />

50<br />

45<br />

40<br />

35<br />

30<br />

25<br />

20<br />

15<br />

10<br />

5<br />

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8<br />

V /V<br />

ω max<br />

Figure 5.14: Number <strong>of</strong> electrons after 200 rf-cycles. The vertical straight<br />

l<strong>in</strong>es <strong>in</strong>dicate Vω,o = v1/2 <strong>and</strong> Vω,o = v2/2. Parameters used:<br />

Ri/Ro = 0.1, f = 1.5 GHz, σse,max = 2.0, vT/vmax = 0.01, <strong>and</strong><br />

W1 = 50 eV.<br />

log 10 (Amplitude) [V]<br />

3.2<br />

3<br />

2.8<br />

2.6<br />

2.4<br />

2.2<br />

2<br />

1.8<br />

1.6<br />

1.4<br />

−1.6 −1.4 −1.2 −1 −0.8 −0.6<br />

log (Frequency) [GHz]<br />

10<br />

Figure 5.15: <strong>Multipactor</strong> breakdown regions for copper electrodes. The region<br />

conf<strong>in</strong>ed by circles (or white crosses, when <strong>in</strong>side a dark<br />

region) is from Ref. [63]. The dark regions are obta<strong>in</strong>ed by the<br />

PIC-code with: Ro/Ri = 2.3 (Z = 50 Ω), σse,max = 2.25 (from<br />

table A-6 <strong>in</strong> [50], W1 = 27 eV, <strong>and</strong> vT = 3 eV (vT/vmax =<br />

0.111).<br />

4500<br />

4000<br />

3500<br />

3000<br />

2500<br />

2000<br />

1500<br />

1000<br />

500<br />

0<br />

4500<br />

4000<br />

3500<br />

3000<br />

2500<br />

2000<br />

1500<br />

1000<br />

500<br />

89


creas<strong>in</strong>g ratio Ri/Ro was observed. It was also found that the first multipactor<br />

zone became narrower for decreas<strong>in</strong>g Ri/Ro. This behaviour<br />

was seen <strong>in</strong> the analytical analysis <strong>and</strong> it is evident also <strong>in</strong> the PICsimulations.<br />

Figure 5.16 shows the result <strong>of</strong> a simulation for Z = 174 Ω<br />

(Ro/Ri = 18.26). The agreement between simulations <strong>and</strong> experiments<br />

is good. The multipactor zone becomes narrower <strong>and</strong> the threshold <strong>in</strong>creases.<br />

The position <strong>of</strong> the zone is slightly shifted compared with the<br />

experimental data <strong>and</strong> the reason for this deviation may partly be expla<strong>in</strong>ed<br />

by an <strong>in</strong>accuracy <strong>of</strong> the dimensions <strong>of</strong> the copper electrodes.<br />

The ma<strong>in</strong> reason, however, seems to be related to the SEY-model <strong>and</strong><br />

the description <strong>of</strong> the <strong>in</strong>itial velocity <strong>of</strong> the secondary electrons, but a<br />

more detailed study will be necessary to settle this matter.<br />

log 10 (Amplitude) [V]<br />

3.2<br />

3<br />

2.8<br />

2.6<br />

2.4<br />

2.2<br />

2<br />

1.8<br />

−1.6 −1.4 −1.2 −1 −0.8 −0.6<br />

log (Frequency) [GHz]<br />

10<br />

Figure 5.16: Ratio <strong>of</strong> power deposited on the <strong>in</strong>ner <strong>and</strong> outer conductors<br />

due to electron impacts (log-scale) for the same parameters <strong>and</strong><br />

SEY-formula as <strong>in</strong> Fig. 5.15 except that Ro/Ri = 18.26.<br />

In the section summariz<strong>in</strong>g the ma<strong>in</strong> f<strong>in</strong>d<strong>in</strong>gs <strong>of</strong> the analytical part,<br />

it was mentioned that no s<strong>in</strong>gle-sided multipactor was noted <strong>in</strong> the experiments.<br />

This is <strong>in</strong> agreement with the PIC-simulations <strong>and</strong> shows<br />

that no measurements were made at high enough voltage <strong>and</strong> frequency,<br />

where s<strong>in</strong>gle sided multipactor would be the ma<strong>in</strong> phenomenon (the dark<br />

blue regions <strong>in</strong> Fig. 5.16).<br />

90<br />

0<br />

−0.5<br />

−1<br />

−1.5<br />

−2<br />

−2.5<br />

−3<br />

−3.5<br />

−4<br />

−4.5


5.2.4 Ma<strong>in</strong> conclusions<br />

The analytically obta<strong>in</strong>ed results are confirmed by the PIC-simulations.<br />

In addition, the significance <strong>of</strong> <strong>in</strong>itial velocity spread as well as different<br />

maximum secondary electron emission have been highlighted. The<br />

results have shown that the behaviour with respect to these conditions,<br />

<strong>in</strong> both the s<strong>in</strong>gle-sided <strong>and</strong> the double-sided cases, is qualitatively the<br />

same as for the parallel plate multipactor. The results are <strong>in</strong> good agreement<br />

with available experimental data, but the commonly used model<br />

for SEY seems to be <strong>in</strong>adequate for large Ro/Ri-ratios, s<strong>in</strong>ce a small,<br />

but not negligible, deviation from the experiments was noticed.<br />

91


Chapter 6<br />

Detection <strong>of</strong> multipactor<br />

In the space <strong>in</strong>dustry, where multipactor is a well-known problem, great<br />

care is taken to avoid the phenomenon. Accord<strong>in</strong>g to the ESA st<strong>and</strong>ard<br />

on multipactor design <strong>and</strong> test [50] only s<strong>in</strong>gle carrier parts with a<br />

marg<strong>in</strong> <strong>of</strong> 8-12 dB (depend<strong>in</strong>g on the type <strong>of</strong> component) <strong>in</strong> the analysis<br />

stage are exempt from test<strong>in</strong>g. The recommended correspond<strong>in</strong>g marg<strong>in</strong><br />

for multicarrier parts is 6 dB between the peak power <strong>of</strong> the multicarrier<br />

signal <strong>and</strong> the s<strong>in</strong>gle carrier threshold. By perform<strong>in</strong>g unit acceptance<br />

tests, the marg<strong>in</strong>s can be reduced to 3-4 dB <strong>in</strong> the s<strong>in</strong>gle carrier case<br />

<strong>and</strong> to 0 dB <strong>in</strong> the multicarrier case. In order to take advantage <strong>of</strong> the<br />

smaller marg<strong>in</strong>s, accurate <strong>and</strong> reliable test<strong>in</strong>g is required to verify that<br />

the prescribed marg<strong>in</strong>s are actually fulfilled <strong>and</strong> successful test<strong>in</strong>g relies<br />

on accurate <strong>and</strong> unambiguous methods <strong>of</strong> detection.<br />

This chapter starts by briefly review<strong>in</strong>g some <strong>of</strong> the most common<br />

methods <strong>of</strong> multipactor detection. Then a method <strong>of</strong> detection is presented,<br />

which can be used <strong>in</strong> comb<strong>in</strong>ation with other methods to achieve<br />

accurate <strong>and</strong> unambiguous test results. Beg<strong>in</strong>n<strong>in</strong>g with the s<strong>in</strong>gle carrier<br />

case, a theory for the underly<strong>in</strong>g mechanism mak<strong>in</strong>g the method possible<br />

is given together with some support<strong>in</strong>g data. This is followed by a<br />

discussion outl<strong>in</strong><strong>in</strong>g how the method can be applied also <strong>in</strong> the multicarrier<br />

case. Detection <strong>of</strong> multipactor us<strong>in</strong>g RF power modulation, which<br />

is the ma<strong>in</strong> topic <strong>of</strong> this chapter, was analysed <strong>in</strong> paper A <strong>of</strong> this thesis.<br />

6.1 Common Methods <strong>of</strong> Detection<br />

There are several different ways <strong>of</strong> detect<strong>in</strong>g multipactor <strong>and</strong> they can be<br />

divided <strong>in</strong>to two fairly dist<strong>in</strong>ct categories, viz. global <strong>and</strong> local methods<br />

93


<strong>of</strong> detection. The global methods are characterised by be<strong>in</strong>g able to<br />

discern whether or not microwave breakdown is tak<strong>in</strong>g place somewhere<br />

with<strong>in</strong> the tested system. However, it can not p<strong>in</strong>po<strong>in</strong>t the location <strong>of</strong><br />

the discharge. For flight hardware it is important to completely avoid<br />

multipactor <strong>in</strong> the entire microwave system <strong>and</strong> consequently this type<br />

<strong>of</strong> test<strong>in</strong>g is preferred. Dur<strong>in</strong>g the product development stage, it may<br />

be <strong>of</strong> <strong>in</strong>terest to know not only that a discharge is tak<strong>in</strong>g place, but also<br />

its exact position with<strong>in</strong> the system. In such a case, a local method can<br />

be useful, s<strong>in</strong>ce it can be used to monitor a certa<strong>in</strong> area <strong>in</strong>side a device.<br />

6.1.1 Global methods<br />

When perform<strong>in</strong>g systems tests on flight hardware, global methods <strong>of</strong><br />

detection are normally used <strong>and</strong> there are several reasons for this. The<br />

methods can usually be applied without modify<strong>in</strong>g the component, which<br />

is advantageous as modifications can affect the electromagnetic properties<br />

<strong>and</strong> give <strong>in</strong>adequate measurement results. In cases where the discharge<br />

is weak, many local methods are unable to detect the phenomenon<br />

<strong>and</strong> thus the <strong>of</strong>ten more sensitive global ones are a better choice. Furthermore,<br />

it is a requirement <strong>of</strong> the ESA st<strong>and</strong>ard [50] that two methods<br />

<strong>of</strong> detection should be used <strong>and</strong> at least one <strong>of</strong> them should be global.<br />

By us<strong>in</strong>g two methods <strong>of</strong> detection, the risk for mis<strong>in</strong>terpretations is reduced.<br />

The most common global methods <strong>of</strong> detection will be described<br />

<strong>in</strong> the follow<strong>in</strong>g subsections.<br />

Close-to-carrier noise<br />

Multipact<strong>in</strong>g electrons will be accelerated to high velocities by the electric<br />

field <strong>and</strong> at regular <strong>in</strong>tervals, 2/N times per field cycle, the electrons<br />

will hit an electrode or device wall <strong>and</strong> experience a sudden deceleration.<br />

The radiated power is a function <strong>of</strong> the electron acceleration <strong>and</strong> it is<br />

described by Larmor’s formula,<br />

P = 2 e<br />

3<br />

2 ˙v<br />

4πɛ0<br />

2<br />

c3 (6.1)<br />

where ˙v is the acceleration, ɛ0 the dielectric constant <strong>of</strong> vacuum, <strong>and</strong> c<br />

the speed <strong>of</strong> light. For an applied s<strong>in</strong>usoidal electric field, the electrons<br />

will experience a s<strong>in</strong>usoidal acceleration, except for the sudden <strong>in</strong>terruptions<br />

when collid<strong>in</strong>g with the electrodes. Fig. 6.1 shows what the<br />

acceleration may look like for first order multipactor (N = 1), where the<br />

electrons hit the electrodes once every half cycle <strong>of</strong> the electric field.<br />

94


Amplitude <strong>of</strong> acceleration [Gm/s 2 ]<br />

x 10<br />

1.5<br />

6 C2C−Noise Timedoma<strong>in</strong><br />

1<br />

0.5<br />

0<br />

−0.5<br />

−1<br />

−1.5<br />

1.76 1.78 1.8 1.82 1.84 1.86<br />

x 10 −8<br />

Time (s)<br />

Figure 6.1: A qualitative plot <strong>of</strong> the electron acceleration for multipact<strong>in</strong>g<br />

electrons with N = 1 (average value for a large number <strong>of</strong> electrons).<br />

The regular deviation from a pure s<strong>in</strong>usoid will generate harmonics,<br />

but these will be discussed <strong>in</strong> the next subsection. Due to variations <strong>in</strong><br />

the time between impact <strong>and</strong> emission <strong>of</strong> new electrons, <strong>in</strong> the time it<br />

takes to decelerate the electrons, <strong>and</strong> also <strong>in</strong> the number <strong>of</strong> electrons,<br />

close-to-carrier noise will be generated. In fact, most noise is generated<br />

at the carrier frequency, but this can not be seen <strong>in</strong> a measured spectrum,<br />

as the signal amplitude masks the noise. By perform<strong>in</strong>g a Fast Fourier<br />

Transform (FFT) <strong>of</strong> a sequence like the one shown <strong>in</strong> Fig. 6.1, the noise<br />

generated close to the carrier can be seen (see Fig. 6.2). Very close to<br />

the carrier, the noise level is even higher than what can be discerned <strong>in</strong><br />

Fig. 6.1 <strong>and</strong> by us<strong>in</strong>g a b<strong>and</strong>pass filter with high rejection at the carrier<br />

frequency, the noise <strong>in</strong>crease close to the carrier can be detected with a<br />

spectrum analyser. In comb<strong>in</strong>ation with a low-noise amplifier, this can<br />

be a very sensitive method <strong>of</strong> detection.<br />

This method <strong>of</strong> detection can be used for both s<strong>in</strong>gle <strong>and</strong> multicarrier<br />

signals. Care should be taken, however, when us<strong>in</strong>g a pulsed signal,<br />

s<strong>in</strong>ce such a signal will generate harmonics <strong>and</strong> if the pulse length <strong>and</strong><br />

type is not chosen properly, the harmonics may be generated <strong>in</strong> the<br />

measurement b<strong>and</strong> [31]. A risk with this method is that other sources <strong>of</strong><br />

95


(dB)<br />

−20<br />

−30<br />

−40<br />

−50<br />

−60<br />

−70<br />

−80<br />

−90<br />

−100<br />

<strong>Multipactor</strong> Noise Power<br />

2 4 6 8 10 12 14 16<br />

Frequency (GHz)<br />

Figure 6.2: Noise spectrum <strong>of</strong> a multipactor simulation like the one <strong>in</strong><br />

Fig. 6.1, but with N = 3. The signal frequency is f0 = 2 GHz.<br />

The odd harmonics as well as the peaks at odd multiples <strong>of</strong> f0/3<br />

are clearly visible.<br />

noise may be mis<strong>in</strong>terpreted as multipactor noise. Nevertheless, by us<strong>in</strong>g<br />

two different methods <strong>of</strong> detection as required by the ESA st<strong>and</strong>ard, this<br />

risk is greatly reduced.<br />

Third harmonic<br />

The resonant behaviour <strong>of</strong> the multipactor discharge <strong>and</strong> the repetitive<br />

acceleration <strong>and</strong> sudden deceleration <strong>of</strong> the electrons will generate noise,<br />

which will have harmonics at the basic frequency, f0, <strong>and</strong> at odd multiples<br />

<strong>of</strong> this frequency (cf. Fig. 6.2). Higher order multipactor will<br />

have harmonics not only at odd multiples <strong>of</strong> the basic harmonic, but at<br />

each odd multiple <strong>of</strong> f0/N [14] (cf. Fig. 6.2). In such cases there are<br />

many different frequencies available for detection. However, s<strong>in</strong>ce the<br />

third harmonic usually is the most powerful harmonic <strong>and</strong> it is always<br />

present, regardless <strong>of</strong> the order <strong>of</strong> resonance, it is the best choice for<br />

detection.<br />

Third harmonic detection is a very reliable <strong>and</strong> fast method <strong>of</strong> detection.<br />

Accord<strong>in</strong>g to Ref. [14], it gives the fastest <strong>in</strong>dication <strong>of</strong> multipactor<br />

<strong>and</strong> that makes it the method <strong>of</strong> choice when study<strong>in</strong>g mul-<br />

96


tipactor events that are short-lived, like e.g. multicarrier multipactor,<br />

where the discharges <strong>of</strong>ten are weak <strong>and</strong> <strong>of</strong> short duration. However,<br />

the method is also sensitive to other sources <strong>of</strong> noise <strong>in</strong> the same way as<br />

close-to-carrier noise <strong>and</strong>, thus, it is not recommended to use only third<br />

harmonic <strong>and</strong> close-to-carrier noise detection <strong>in</strong> a test setup.<br />

Reflected power<br />

Mismatches <strong>in</strong> the transitions between microwave components imply<br />

that some <strong>of</strong> the <strong>in</strong>put power will be reflected. However, <strong>in</strong> a welldesigned<br />

system, very little power is reflected. Components conta<strong>in</strong><strong>in</strong>g<br />

high Q-value parts, like e.g. cavity resonators, are only well matched<br />

at certa<strong>in</strong> frequencies <strong>and</strong> m<strong>in</strong>or changes <strong>in</strong> the component properties<br />

can lead to detun<strong>in</strong>g <strong>of</strong> the part. <strong>Multipactor</strong> is known to be able to<br />

detune high Q-value components [70]. Thus the reflected power from<br />

a component can be used as an <strong>in</strong>dication <strong>of</strong> a multipactor event. To<br />

study the absolute value <strong>of</strong> the reflected power is usually not a good way<br />

<strong>of</strong> detection, s<strong>in</strong>ce the <strong>in</strong>put power can vary dur<strong>in</strong>g a multipactor test<br />

<strong>and</strong> consequently the reflected power will vary as well, as the reflected<br />

power is a fixed fraction <strong>of</strong> the <strong>in</strong>put power. This fraction is called the<br />

return loss <strong>and</strong> is commonly measured <strong>in</strong> decibel. The return loss will be<br />

a stable value, <strong>in</strong>sensitive to power fluctuations, until the component is<br />

detuned. Figure 6.3 shows an example where both close-to-carrier noise<br />

<strong>and</strong> the return loss are monitored simultaneously dur<strong>in</strong>g a multipactor<br />

test. Both methods <strong>in</strong>dicate a change around t = 50 s <strong>and</strong> thus it can<br />

be determ<strong>in</strong>ed with great confidence that a multipactor discharge was<br />

<strong>in</strong>itiated at that time.<br />

The ma<strong>in</strong> advantage with this method is that it is quite reliable <strong>and</strong><br />

there is little risk that other phenomena will cause a mismatch that can<br />

be confused with multipactor. However, for low Q-value components or<br />

badly matched systems, the sensitivity <strong>of</strong> the method is low.<br />

Electron monitor<strong>in</strong>g<br />

A new global method <strong>of</strong> detection was presented at the 4 th International<br />

Workshop on <strong>Multipactor</strong>, Corona <strong>and</strong> Passive Intermodulation<br />

<strong>in</strong> Space RF Hardware [71]. It is called the Electron Density Detection<br />

Method, abbreviated EDDM, <strong>and</strong> uses a set <strong>of</strong> tri-axial cables as<br />

a probe <strong>and</strong> electrons picked up by the probe are then monitored with<br />

a high precision electron meter. The data is collected us<strong>in</strong>g a computer<br />

97


Noise (dBm)<br />

−60<br />

−70<br />

−80<br />

−90<br />

0 10 20 30 40 50<br />

Time (s)<br />

60 70 80 90 100<br />

Input Power (dBm)<br />

Return Loss (dB)<br />

46<br />

44<br />

42<br />

0 10 20 30 40 50<br />

Time (s)<br />

60 70 80 90 100<br />

−8<br />

−10<br />

−12<br />

−14<br />

−16<br />

0 10 20 30 40 50<br />

Time (s)<br />

60 70 80 90 100<br />

Figure 6.3: <strong>Multipactor</strong> monitored by study<strong>in</strong>g close-to-carrier noise as well<br />

as the return loss. At t ≈ 50 s the component starts to experience<br />

multipactor discharge <strong>and</strong> becomes <strong>in</strong>creas<strong>in</strong>gly mismatched. Excellent<br />

agreement with the second detection methods, close-tocarrier<br />

noise, can be seen.<br />

98


s<strong>of</strong>tware <strong>and</strong> the time evolution <strong>of</strong> the electron density can be studied.<br />

The probe does not have to be located <strong>in</strong>side or close to the area where<br />

the discharge will take place. Even though it is advantageous for higher<br />

sensitivity to have the probe po<strong>in</strong>t<strong>in</strong>g at the critical gap it is also reliable<br />

when located outside the device under test <strong>and</strong> thus it can be used<br />

for waveguides <strong>and</strong> coaxial transmission l<strong>in</strong>es. It is not completely clear<br />

from the paper, however, how the electrons escape from a completely<br />

conf<strong>in</strong>ed device like a typical waveguide or coaxial cable <strong>and</strong> one will<br />

have to assume that there must be some small open<strong>in</strong>g somewhere <strong>in</strong><br />

the system, where electrons can leak out.<br />

Furthermore, the method can be used to quantitatively measure the<br />

amount <strong>of</strong> generated electrons. however, this requires some k<strong>in</strong>d <strong>of</strong> calibration<br />

<strong>of</strong> each test setup <strong>and</strong> that may prove to be problematic. When<br />

used <strong>in</strong> this way, the method can no longer be viewed as a global method,<br />

which can detect a multipactor event anywhere <strong>in</strong> the system, <strong>in</strong>stead it<br />

has become a local method. Among the ma<strong>in</strong> advantages <strong>of</strong> the method<br />

is the low cost <strong>in</strong>volved, s<strong>in</strong>ce no expensive microwave <strong>in</strong>struments are<br />

needed.<br />

Residual mass<br />

A very slow global method <strong>of</strong> detection is to detect the gas molecules,<br />

which are outgassed from the device walls due to the electron bombardment<br />

dur<strong>in</strong>g a multipactor event. The gas molecules consist <strong>of</strong> residuals<br />

<strong>of</strong> water, air <strong>and</strong> contam<strong>in</strong>ants, <strong>and</strong> us<strong>in</strong>g a mass spectrometer,<br />

the different molecules can be identified. It has been noted [31] that a<br />

detectable <strong>in</strong>crease <strong>in</strong> the water spectrum can be seen dur<strong>in</strong>g a multipactor<br />

discharge. The major drawback <strong>of</strong> this method <strong>of</strong> detection is<br />

its <strong>in</strong>ability to detect fast multipactor transients (not enough molecules<br />

are released from the walls) <strong>and</strong> thus it is not a suitable method for<br />

multicarrier multipactor studies. Another disadvantage is that there is<br />

a certa<strong>in</strong> delay between onset <strong>of</strong> the discharge <strong>and</strong> <strong>in</strong>dication <strong>in</strong> the <strong>in</strong>strumentation.<br />

However, it can be useful as a diagnostic tool together<br />

with one or two <strong>of</strong> the other described methods.<br />

6.1.2 Local methods<br />

In cases where it is not sufficient to only confirm the existence <strong>of</strong> a<br />

discharge <strong>in</strong> the system, but also to determ<strong>in</strong>e the exact position, local<br />

methods <strong>of</strong> detection will have to be used. The two most common local<br />

99


methods are charge detection us<strong>in</strong>g a probe <strong>and</strong> optical monitor<strong>in</strong>g <strong>and</strong><br />

these two methods will be described below.<br />

Charge probe<br />

A special case <strong>of</strong> the method <strong>of</strong> electron monitor<strong>in</strong>g, which was described<br />

<strong>in</strong> a previous subsection, is the use <strong>of</strong> a probe that monitors only a<br />

certa<strong>in</strong> position with<strong>in</strong> the microwave device. A very common approach<br />

used when detect<strong>in</strong>g electrons <strong>in</strong>side a waveguide is to flush mount an<br />

SMA connector <strong>and</strong> apply a positive potential to the centre p<strong>in</strong>. The<br />

negatively charged electrons are attracted to the p<strong>in</strong> <strong>and</strong> causes a small,<br />

but detectable, current to flow, which can serve as an <strong>in</strong>dication <strong>of</strong> the<br />

electron density.<br />

The method is easy to implement <strong>and</strong> therefore it has been quite<br />

common <strong>in</strong> many test setups. Unfortunately, it is also quite slow due<br />

to the circuit used to amplify the weak current [14] <strong>and</strong> thus it is not<br />

feasible for measur<strong>in</strong>g fast multipactor events. In addition, it requires<br />

modification <strong>of</strong> the component, which makes it useless when test<strong>in</strong>g flight<br />

hard ware. In such a case the EDDM is a better choice.<br />

Optical detection<br />

Optical detection is possible s<strong>in</strong>ce the electrons that make up the multipactor<br />

discharge can excite or ionise either the rema<strong>in</strong><strong>in</strong>g gas molecules<br />

with<strong>in</strong> the device or the molecules <strong>in</strong> the device wall. It can be divided<br />

<strong>in</strong>to two groups, viz. photon detection via optical probe <strong>and</strong> photon<br />

detection via photographs or video camera [72,73]. Both methods are<br />

common, but the former seems to be more frequently used.<br />

The ma<strong>in</strong> advantage with these methods is that they can be used to<br />

p<strong>in</strong>-po<strong>in</strong>t the location <strong>of</strong> the discharge <strong>in</strong>side the device. A major disadvantage<br />

is that they may be impossible to use for study<strong>in</strong>g real parts,<br />

as the devices may not have any suitable open<strong>in</strong>gs. This is especially<br />

true for the methods based on photographic techniques.<br />

6.2 Detection us<strong>in</strong>g RF Power Modulation<br />

Dur<strong>in</strong>g verification <strong>of</strong> a test setup for multipactor at Saab Ericsson<br />

Space, Göteborg, Sweden, an odd, spike-like phenomenon was found<br />

<strong>in</strong> the noise generated by a discharge <strong>in</strong> a coaxial test sample. An ad-<br />

100


ditional test sample, a resonant cavity, was manufactured <strong>and</strong> the same<br />

type <strong>of</strong> spikes were noted aga<strong>in</strong> (cf. Fig. 6.4).<br />

(dBm)<br />

−55<br />

−60<br />

−65<br />

−70<br />

−75<br />

−80<br />

−85<br />

−90<br />

Noise Power (92.8 − 93.0 s)<br />

92.82 92.84 92.86 92.88 92.9<br />

Time (s)<br />

92.92 92.94 92.96 92.98 93<br />

Figure 6.4: Periodic spikes, which appeared dur<strong>in</strong>g a multipactor experiment.<br />

The ma<strong>in</strong> periodicity is 100 Hz <strong>and</strong> emanates from the power<br />

supply <strong>of</strong> the high power amplifier.<br />

A fast Fourier transform revealed that the spikes were periodic <strong>and</strong><br />

it was noted that the same periodicity could be found also <strong>in</strong> the <strong>in</strong>put<br />

signal after it had been amplified by the TWTA (travell<strong>in</strong>g wave tube<br />

amplifier). Due to <strong>in</strong>terference from the ma<strong>in</strong> power supply, the signal<br />

was amplitude modulated with a ma<strong>in</strong> modulation frequency <strong>of</strong> 100 Hz<br />

<strong>and</strong> with harmonics at multiples <strong>of</strong> this frequency. The <strong>in</strong>terference was<br />

very weak <strong>and</strong> would <strong>in</strong> most cases be disregarded. In order to see if<br />

the periodicity was present only <strong>in</strong> conjunction with multipactor events,<br />

a large number <strong>of</strong> test runs were performed. The results were consistent<br />

- the periodic noise only appeared when a discharge was detected.<br />

Figs. 6.5 <strong>and</strong> 6.6 show one <strong>of</strong> the test runs where the noise is non-periodic<br />

before onset <strong>of</strong> multipactor but periodic directly afterwards.<br />

The AM (amplitude modulation) that was present <strong>in</strong> the <strong>in</strong>put signal<br />

was very weak <strong>and</strong> not deliberately added. A stronger AM was added to<br />

the signal before the high power amplifier, result<strong>in</strong>g <strong>in</strong> a more dist<strong>in</strong>ct<br />

modulation. This made it possible to study if there was any correlation<br />

between the modulation strength <strong>and</strong> the correspond<strong>in</strong>g peak <strong>in</strong><br />

101


(dBm)<br />

(dBm)<br />

−55<br />

−60<br />

−65<br />

−70<br />

−75<br />

−80<br />

−85<br />

−90<br />

−80<br />

−100<br />

−120<br />

−140<br />

Noise Power (32 − 52 s)<br />

34 36 38 40 42<br />

Time (s)<br />

44 46 48 50 52<br />

Fourier transform (32 − 52 s)<br />

−160<br />

0 50 100 150 200 250<br />

frequency (Hz)<br />

300 350 400 450 500<br />

Figure 6.5: The beg<strong>in</strong>n<strong>in</strong>g <strong>of</strong> a multipactor test sequence show<strong>in</strong>g the time<br />

before onset <strong>of</strong> the discharge. The FFT (fast Fourier transform)<br />

gives no <strong>in</strong>dication <strong>of</strong> dom<strong>in</strong>ant frequency components.<br />

(dBm)<br />

(dBm)<br />

−55<br />

−60<br />

−65<br />

−70<br />

−75<br />

−80<br />

−85<br />

−90<br />

−80<br />

−100<br />

−120<br />

−140<br />

Noise Power (52 − 72 s)<br />

54 56 58 60 62<br />

Time (s)<br />

64 66 68 70 72<br />

Fourier transform (52 − 72 s)<br />

−160<br />

0 50 100 150 200 250<br />

frequency (Hz)<br />

300 350 400 450 500<br />

Figure 6.6: The end <strong>of</strong> the same test sequence as <strong>in</strong> Fig. 6.5. The sudden <strong>in</strong>crease<br />

<strong>in</strong> the noise floor <strong>in</strong>dicates onset <strong>of</strong> multipactor. The FFT<br />

shows dom<strong>in</strong>ant frequency components at 100 Hz <strong>and</strong> multiples<br />

there<strong>of</strong>.<br />

102


the FFT. It was concluded that there is a positive correlation between<br />

the modulation depth <strong>and</strong> the strength <strong>of</strong> the detected signal at the<br />

correspond<strong>in</strong>g frequency. By tak<strong>in</strong>g the time average <strong>of</strong> one <strong>of</strong> the test<br />

sequences like <strong>in</strong> Fig. 6.3 <strong>and</strong> plott<strong>in</strong>g it us<strong>in</strong>g l<strong>in</strong>ear scales on both axes,<br />

it was found that there was a more or less l<strong>in</strong>ear relationship between<br />

the <strong>in</strong>put power <strong>and</strong> the result<strong>in</strong>g multipactor noise power (cf. Fig. 6.7),<br />

which can be described by the follow<strong>in</strong>g function:<br />

Pnoise = k · (P<strong>in</strong>put − Pth) [W] P<strong>in</strong>put ≥ Pth (6.2)<br />

where k = 5.3 × 10 −11 <strong>and</strong> Pth = 25.2 W is the multipactor threshold.<br />

Noise Power [nW]<br />

1.4<br />

1.2<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

L<strong>in</strong>ear Plot <strong>of</strong> Noisepower vs Input Signal Power<br />

Input Signal Power [W]<br />

26 28 30 32 34 36 38 40 42 44<br />

0<br />

50 55 60 65 70 75<br />

Time (s)<br />

80 85 90 95 100<br />

Figure 6.7: Time average <strong>of</strong> the test sequence shown <strong>in</strong> Fig. 6.3 with a l<strong>in</strong>ear<br />

scale on both axes. The straight l<strong>in</strong>e has been added to show the<br />

close to l<strong>in</strong>ear relationship between <strong>in</strong>put power <strong>and</strong> noise power.<br />

Note: The <strong>in</strong>put power is <strong>in</strong>creased every 4 seconds, which is the<br />

reason for the step like behaviour.<br />

The reason why the small modulation was so noticeable <strong>in</strong> the multipactor<br />

noise (see Fig. 6.4) is that the noise signal is a function <strong>of</strong> the<br />

difference between the <strong>in</strong>put power <strong>and</strong> the multipactor threshold, i.e.<br />

no discharge noise is generated until the multipactor threshold has been<br />

reached. Furthermore, s<strong>in</strong>ce the decibel scale is a relative scale, the small<br />

<strong>in</strong>crease <strong>in</strong> absolute numbers becomes very noticeable <strong>in</strong> relation to the<br />

exist<strong>in</strong>g noise floor. As a comparison, the first small steps <strong>in</strong> Fig. 6.7,<br />

103


correspond to huge <strong>in</strong>creases <strong>in</strong> the decibel scale, which can be seen <strong>in</strong><br />

Fig. 6.8.<br />

(dBm)<br />

−60<br />

−65<br />

−70<br />

−75<br />

−80<br />

−85<br />

Noise Power (100 samples noise average)<br />

0 10 20 30 40 50<br />

Time (s)<br />

60 70 80 90 100<br />

Figure 6.8: The same sequence as <strong>in</strong> Fig. 6.7 (except that <strong>in</strong> this case the<br />

entire sequence is shown). The small <strong>in</strong>itial steps <strong>of</strong> Fig. 6.7<br />

become huge steps on the logarithmic scale.<br />

The above examples, which used the new detection method, relied<br />

on close-to-carrier noise measurement data. However, the mechanism<br />

which is utilised requires primarily that the <strong>in</strong>put signal is amplitude<br />

modulated, that the detected signal is proportional to the difference<br />

between the <strong>in</strong>put power <strong>and</strong> the multipactor threshold <strong>and</strong> that the<br />

detected signal responds quickly to changes <strong>in</strong> the multipactor event.<br />

The two first conditions are likely to be fulfilled by all detection methods<br />

for multipactor, but the last will have to be verified for each method.<br />

Results from measurements presented <strong>in</strong> [14] show that third harmonic<br />

detection is faster than close-to-carrier noise detection, which should<br />

make it excellent for AM detection. In general, probably any method<br />

can be used provided that a suitable AM frequency is chosen <strong>and</strong> that<br />

the <strong>in</strong>strument used for detection has a sampl<strong>in</strong>g frequency that is more<br />

than two times larger than the modulation frequency <strong>in</strong> order to fulfil<br />

the Nyquist criterion.<br />

104


6.2.1 S<strong>in</strong>gle carrier<br />

When us<strong>in</strong>g the AM detection method <strong>in</strong> the s<strong>in</strong>gle carrier case, the<br />

test setup implementation is quite straight forward. In practice, the<br />

only difference between a st<strong>and</strong>ard multipactor test setup <strong>and</strong> one that<br />

uses the AM detection method is that the signal data must be sent to<br />

a computer or some other <strong>in</strong>strument that can perform a FFT. It is<br />

also important that the signal generator is capable <strong>of</strong> produc<strong>in</strong>g an AM<br />

signal, but that is a st<strong>and</strong>ard feature <strong>of</strong> most signal generators.<br />

The spectrum analyser, which receives the signal, should be set to a<br />

sampl<strong>in</strong>g rate that is at least two times larger than the AM signal, i.e. if<br />

the AM signal has a frequency <strong>of</strong> 1000 Hz, then a m<strong>in</strong>imum sampl<strong>in</strong>g rate<br />

<strong>of</strong> 2000 samples/second should be used. However, <strong>in</strong> order to study the<br />

shape <strong>of</strong> the modulation signal, it is better to use a sampl<strong>in</strong>g frequency<br />

more than twenty times greater than the frequency <strong>of</strong> the AM. The signal<br />

data can be stored for future process<strong>in</strong>g, or if a powerful computer is<br />

used, real time FFT can be performed, thus allow<strong>in</strong>g the operator to get<br />

an immediate <strong>in</strong>dication when a discharge takes place.<br />

6.2.2 Multicarrier<br />

AM detection <strong>in</strong> the multicarrier case [74] is somewhat more difficult <strong>and</strong><br />

the method has not yet been experimentally confirmed. When perform<strong>in</strong>g<br />

a multicarrier multipactor test, the phase <strong>of</strong> each carrier will have to<br />

be adjustable <strong>in</strong> order to enable the test eng<strong>in</strong>eer to produce the wanted<br />

shape <strong>of</strong> the signal envelope. The aim is to produce a signal envelope<br />

correspond<strong>in</strong>g to the assumed worst case. When the phases have been<br />

set to their predef<strong>in</strong>ed values, the envelope will be periodic. If the envelope<br />

exceeds the multipactor threshold for a time long enough to allow a<br />

sufficient number <strong>of</strong> gap cross<strong>in</strong>gs, a discharge will occur, provided that<br />

a suitable seed electron is available. When the envelope drops below<br />

the threshold aga<strong>in</strong>, the electrons will disappear quickly [21] <strong>and</strong> there<br />

will normally be no electrons left from the discharge the next time the<br />

envelope exceeds the threshold. If no specific source <strong>of</strong> seed electrons is<br />

available, a multipactor discharge may not occur every envelope period.<br />

However, if an efficient electron seed<strong>in</strong>g source is used, there should be<br />

an ample amount <strong>of</strong> electrons available to <strong>in</strong>itiate a discharge each time<br />

the envelope exceeds the threshold for a sufficiently long time.<br />

The use <strong>of</strong> AM detection <strong>of</strong> multipactor requires that the discharge<br />

event is cont<strong>in</strong>uous or that it occurs regularly. S<strong>in</strong>gle or very sporadi-<br />

105


cally occurr<strong>in</strong>g discharges will be difficult to identify <strong>in</strong> a FFT plot. By<br />

us<strong>in</strong>g a source <strong>of</strong> free electrons <strong>in</strong> the test setup, e.g. a hot filament<br />

or a UV light [14], seed electrons will be abundant <strong>and</strong> a multipactor<br />

event is likely to occur each time the envelope exceeds the threshold for<br />

a long enough time. If the envelope is amplitude modulated, the discharge<br />

events will vary <strong>in</strong> strength with the AM frequency. This periodic<br />

variation will appear as a peak <strong>in</strong> the FFT plot <strong>of</strong> the detected signal at<br />

the same frequency. By apply<strong>in</strong>g a weak, 1-5% depth, synchronised AM<br />

to the <strong>in</strong>put signals, the multipactor threshold can be determ<strong>in</strong>ed with<br />

high accuracy <strong>and</strong> without risk<strong>in</strong>g ambiguous test results. A 1% AM<br />

corresponds to less than ±0.1 dB variation <strong>in</strong> the <strong>in</strong>put signal <strong>and</strong> will<br />

have no significant effect on the measured threshold.<br />

The detected signal must then be processed by a computer or a similar<br />

tool <strong>in</strong> order to reveal the periodicity, as <strong>in</strong> the s<strong>in</strong>gle carrier case.<br />

In Fig. 6.9 an example <strong>of</strong> a possible test setup is given. In order to<br />

achieve a good AM <strong>in</strong> the multicarrier case, all the signals should be<br />

modulated us<strong>in</strong>g the same reference signal <strong>of</strong> modulation. Many signal<br />

generators have a signal reference <strong>in</strong>put, thus allow<strong>in</strong>g the user to synchronise<br />

several signal generators. To perform a successful multicarrier<br />

experiment, the phases have to be stable <strong>in</strong> relation to each other <strong>and</strong><br />

thus a common reference signal will have to be used <strong>in</strong> any case. Another<br />

way <strong>of</strong> achiev<strong>in</strong>g synchronised modulation could be to modulate<br />

the ga<strong>in</strong> adjustment <strong>of</strong> the high power amplifier.<br />

In Fig. 6.9 it is suggested that the third harmonic should be monitored<br />

<strong>and</strong> that is probably the best choice when study<strong>in</strong>g multicarrier<br />

multipactor, s<strong>in</strong>ce third harmonic detection is fast <strong>and</strong> sensitive. Closeto-carrier<br />

noise detection is a possible alternative, but it may not be as<br />

sensitive <strong>and</strong> thus weak multipactor events may be overlooked.<br />

6.2.3 Ma<strong>in</strong> achievements<br />

A method <strong>of</strong> multipactor detection has been devised, which can be used<br />

to obta<strong>in</strong> accurate <strong>and</strong> unambiguous measurement results for both s<strong>in</strong>gle<br />

<strong>and</strong> multicarrier multipactor. The method does not aim to replace any <strong>of</strong><br />

the exist<strong>in</strong>g methods <strong>of</strong> detection, rather it can serve as a complement<br />

to the other methods to improve accuracy <strong>and</strong> confidence <strong>in</strong> the test<br />

results.<br />

Close-to-carrier noise <strong>and</strong> third harmonic detection are two fast <strong>and</strong><br />

sensitive methods <strong>of</strong> multipactor detection. Both methods rely on noise<br />

generation, which makes them prone to non-multipactor generated noise.<br />

106


Reference signal generator<br />

Phase<br />

control<br />

∆φ<br />

∆φ<br />

M<br />

U<br />

X<br />

Instrument control<br />

&<br />

Data storage <strong>and</strong> process<strong>in</strong>g<br />

HPA<br />

dB<br />

Wave form monitor<br />

DUT<br />

Spect.<br />

Analys.<br />

Vacuum<br />

Chamber<br />

Reflected power<br />

Power<br />

Meter<br />

dB<br />

LNA<br />

Forward power<br />

dB<br />

Power<br />

Meter<br />

Third harmonic detection<br />

Figure 6.9: Test setup for multicarrier multipactor measurements us<strong>in</strong>g RF<br />

power modulation. Each <strong>in</strong>put signal is amplitude modulated <strong>and</strong><br />

by phase lock<strong>in</strong>g the signals us<strong>in</strong>g a signal reference, the entire<br />

signal envelope will be modulated. A computer can be used to<br />

control the <strong>in</strong>struments <strong>and</strong> collect the output data, which can<br />

then be real-time Fourier transformed to reveal any periodicity<br />

<strong>in</strong> the detected signal.<br />

Spect.<br />

Analys.<br />

107


In a typical test setup there can be many sources <strong>of</strong> noise, which could<br />

result <strong>in</strong> ambiguous test results. On the one h<strong>and</strong>, the multipactor<br />

threshold could be established at a too low value if non-multipactor<br />

generated noise is mis<strong>in</strong>terpreted as the result <strong>of</strong> a discharge. On the<br />

other h<strong>and</strong>, a short-lived multipactor event could be disregarded <strong>and</strong><br />

lead to determ<strong>in</strong>ation <strong>of</strong> a too high threshold based on a more dist<strong>in</strong>ct<br />

<strong>in</strong>dication. The AM detection method resolves this concern by only<br />

signall<strong>in</strong>g for true multipactor events.<br />

Another advantage <strong>of</strong> the AM detection method is the fact that it<br />

is particularly sensitive close to the multipactor threshold, s<strong>in</strong>ce it only<br />

responds to the signal difference between the <strong>in</strong>put signal <strong>and</strong> the threshold,<br />

as can be seen from relation 6.2. A weak amplitude modulation,<br />

as shown <strong>in</strong> Fig. 6.10 where the s<strong>in</strong>gle carrier signal has just passed the<br />

multipactor threshold, will produce a very dist<strong>in</strong>ct modulated output<br />

signal as <strong>in</strong>dicated <strong>in</strong> Fig. 6.11. Even though the signal is very noisy,<br />

the periodicity is very dist<strong>in</strong>ct. Not always will the periodicity be as<br />

noticeable as <strong>in</strong> Fig. 6.11, but a FFT will reveal any periodicity <strong>in</strong> the<br />

measured signal.<br />

Amplitude [A.U.]<br />

120<br />

100<br />

80<br />

60<br />

40<br />

20<br />

Input signal envelope, s<strong>in</strong>gle carrier, 1%depth AM<br />

Input signal<br />

<strong>Multipactor</strong> threshold<br />

0<br />

0 0.5 1 1.5 2 2.5<br />

Time<br />

3 3.5 4 4.5 5<br />

Figure 6.10: Example <strong>of</strong> a s<strong>in</strong>gle carrier <strong>in</strong>put signal envelope with a 1%<br />

depth AM. The signal has barely exceeded the multipactor<br />

threshold.<br />

108<br />

One <strong>of</strong> the shortcom<strong>in</strong>gs <strong>of</strong> the AM detection method is that it re-


Amplitude [A.U.]<br />

2.5<br />

2<br />

1.5<br />

1<br />

0.5<br />

Detected signal<br />

0<br />

0 0.5 1 1.5 2 2.5<br />

Time<br />

3 3.5 4 4.5 5<br />

Figure 6.11: Qualitative form <strong>of</strong> the detected multipactor signal when the<br />

<strong>in</strong>put signal to the DUT is as shown <strong>in</strong> Fig. 6.10. The same type<br />

<strong>of</strong> signal can be seen <strong>in</strong> Fig. 6.4, but <strong>in</strong> this case, the sampl<strong>in</strong>g<br />

frequency is 200 times higher than the frequency <strong>of</strong> modulation,<br />

which expla<strong>in</strong>s why the modulated signal is so prom<strong>in</strong>ent.<br />

109


quires a certa<strong>in</strong> time for the FFT. One can estimate that a m<strong>in</strong>imum<br />

<strong>of</strong> 5-10 times the period <strong>of</strong> the AM is needed for a reliable FFT. If the<br />

frequency <strong>of</strong> AM is 1 kHz, then the time needed for the measurement<br />

would be 5-10 ms. In most cases this would be acceptable, but if s<strong>in</strong>gle<br />

events <strong>of</strong> multipactor are to be detected, the method will not work.<br />

110


Chapter 7<br />

Conclusions <strong>and</strong> outlook<br />

This thesis has presented basic theory as well as new developments concern<strong>in</strong>g<br />

the phenomenon called multipactor. Its deleterious effects on<br />

microwave systems operat<strong>in</strong>g <strong>in</strong> a vacuum environment have been emphasised.<br />

For satellites, the discharge can be catastrophic as basically no<br />

means <strong>of</strong> repair or modification is available for parts <strong>in</strong> orbit. To avoid<br />

the risks associated with vacuum discharges, lots <strong>of</strong> effort has been put<br />

<strong>in</strong>to study<strong>in</strong>g the phenomenon <strong>and</strong> today well established eng<strong>in</strong>eer<strong>in</strong>g<br />

methods exist, which are used by the microwave eng<strong>in</strong>eer when design<strong>in</strong>g<br />

hardware bound for space. However, the available methods are based<br />

on the simple parallel-plate model, which <strong>in</strong> many <strong>in</strong>stances is the worst<br />

case. Thus, when design<strong>in</strong>g important microwave features, <strong>in</strong> particular<br />

such <strong>in</strong>volv<strong>in</strong>g a nonuniform electromagnetic field, like e.g. irises <strong>and</strong><br />

coaxial l<strong>in</strong>es, the parallel-plate model may not be applicable <strong>and</strong> there<br />

is a risk <strong>of</strong> unnecessarily conservative designs. Such designs are <strong>of</strong>ten<br />

large <strong>and</strong> heavy, which is a great disadvantage when it comes to devices<br />

to be used <strong>in</strong> space. As a first attempt to establish new methods <strong>of</strong><br />

assess<strong>in</strong>g the risk for hav<strong>in</strong>g a discharge <strong>in</strong> such structures, a large part<br />

<strong>of</strong> this thesis has been devoted to multipactor <strong>in</strong> irises <strong>and</strong> coaxial l<strong>in</strong>es.<br />

It has been shown that by only consider<strong>in</strong>g the r<strong>and</strong>om walk <strong>of</strong> the secondary<br />

electrons <strong>in</strong> an iris, the threshold can be significantly <strong>in</strong>creased<br />

compared with the pure parallel-plate case. Many <strong>in</strong>terest<strong>in</strong>g aspects <strong>of</strong><br />

the phenomenon <strong>in</strong> a coaxial l<strong>in</strong>e have been found, e.g. the dual stable<br />

regions <strong>of</strong> s<strong>in</strong>gle-sided resonance <strong>and</strong> its unexpectedly low threshold<br />

<strong>of</strong> the first order mode. In addition, the <strong>in</strong>creased threshold for high<br />

impedance coaxial l<strong>in</strong>es that was found <strong>in</strong> experiments, was also found<br />

<strong>in</strong> this study, both <strong>in</strong> the theoretical analysis <strong>and</strong> <strong>in</strong> the PIC-simulations.<br />

111


At altitudes where most satellites operate, the pressure is very low<br />

<strong>and</strong> for most purposes it can be approximated as a perfect vacuum.<br />

However, dur<strong>in</strong>g the launch phase <strong>of</strong> a satellite <strong>and</strong> dur<strong>in</strong>g its first days<br />

<strong>of</strong> operation as well as at times when the satellite fires its attitude <strong>and</strong><br />

altitude control eng<strong>in</strong>es, the microwave parts may not be completely<br />

vented <strong>and</strong> thus it is important to underst<strong>and</strong> what happens with the<br />

multipactor threshold with <strong>in</strong>creas<strong>in</strong>g pressure. This has been one <strong>of</strong> the<br />

ma<strong>in</strong> topics <strong>of</strong> this thesis. It was found that for materials with a low first<br />

cross-over energy, for the lowest order resonance, the threshold <strong>in</strong>creases<br />

with <strong>in</strong>creas<strong>in</strong>g pressure until reach<strong>in</strong>g a maximum, after which it starts<br />

to decl<strong>in</strong>e. The reason for the <strong>in</strong>crease is the friction force experienced<br />

by the electrons when collid<strong>in</strong>g with the neutral gas particles. In all<br />

other cases, for materials with a high first cross-over energy <strong>and</strong> <strong>in</strong> general<br />

for the higher order modes, a monotonically decreas<strong>in</strong>g threshold<br />

is noted. This behaviour can be expla<strong>in</strong>ed by the thermalisation <strong>of</strong> the<br />

electrons, which leads to a higher total impact energy, as well as the<br />

contribution <strong>of</strong> electrons from collisional ionisation, which reduces the<br />

necessary secondary emission yield <strong>and</strong> consequently also the required<br />

impact velocity. Improved quantitative results require a more detailed<br />

<strong>in</strong>vestigation <strong>of</strong> the fraction <strong>of</strong> the electrons from impact ionisation that<br />

actually contribute to the multipactor bunch, but that is left for a future<br />

study. Furthermore, <strong>in</strong> order to make the model useful for all frequencygap<br />

size products, an extension <strong>of</strong> the model to <strong>in</strong>clude also the hybrid<br />

modes is necessary. Due to the complexity <strong>of</strong> such a study, it was not<br />

<strong>in</strong>cluded <strong>in</strong> this first analysis. However, this may be an <strong>in</strong>terest<strong>in</strong>g topic<br />

for a future <strong>in</strong>vestigation.<br />

In addition to design<strong>in</strong>g with respect to the proper thresholds, most<br />

hardware will require some type <strong>of</strong> test<strong>in</strong>g to ensure compliance with<br />

the st<strong>and</strong>ard. Such tests must be <strong>of</strong> good quality to avoid ambiguities,<br />

which could disqualify a component that is multipactor free. In this<br />

thesis a method that gives unambiguous <strong>and</strong> highly reliable test results<br />

was presented. It is a method based on a weak amplitude modulation <strong>of</strong><br />

the <strong>in</strong>put signal, which becomes very dist<strong>in</strong>ct <strong>in</strong> the multipactor signal,<br />

s<strong>in</strong>ce the generated noise power is a function <strong>of</strong> the difference between<br />

<strong>in</strong>put power <strong>and</strong> the multipactor threshold. By us<strong>in</strong>g this auxiliary<br />

method <strong>of</strong> detection <strong>in</strong> connection with two other detection methods,<br />

reliable test results can be obta<strong>in</strong>ed.<br />

It is quite satisfy<strong>in</strong>g when look<strong>in</strong>g back at the chapter on future work<br />

<strong>in</strong> my licentiate thesis [75] <strong>and</strong> realiz<strong>in</strong>g that the two ma<strong>in</strong> topics men-<br />

112


tioned there were multipactor <strong>in</strong> nonuniform fields <strong>and</strong> irises, which were<br />

then successfully studied dur<strong>in</strong>g the second part <strong>of</strong> my PhD work. However,<br />

there are still <strong>in</strong>terest<strong>in</strong>g projects to consider, e.g. multipactor <strong>in</strong><br />

irises where not only the effect <strong>of</strong> the r<strong>and</strong>om walk is considered, but also<br />

the effect <strong>of</strong> the actual nonuniform field <strong>in</strong> the structure. There are many<br />

other important structures <strong>in</strong> microwave systems, which have not been<br />

studied with respect to multipactor discharge. Among these are crossed<br />

irises, septum polarisers <strong>and</strong> ridged waveguides. The 20 gap-cross<strong>in</strong>gs<br />

rule, which is part <strong>of</strong> the ESA st<strong>and</strong>ard [50], should be re-<strong>in</strong>vestigated<br />

both theoretically <strong>and</strong> experimentally to make sure that the rule has<br />

a sound theoretical base with good agreement between simulations <strong>and</strong><br />

experiment. As a cont<strong>in</strong>uation <strong>of</strong> the coaxial study, the axial dimension<br />

could be <strong>in</strong>cluded by consider<strong>in</strong>g not only a travell<strong>in</strong>g wave signal but<br />

also a st<strong>and</strong><strong>in</strong>g wave. S<strong>in</strong>ce electrons affected by the Miller force will<br />

drift towards positions with low field amplitude, it may not be feasible<br />

to directly apply the peak amplitude <strong>of</strong> the st<strong>and</strong><strong>in</strong>g wave <strong>in</strong> the coaxial<br />

model presented <strong>in</strong> this thesis.<br />

113


114


References<br />

[1] W. Henneberg, R. Orthuber, <strong>and</strong> E. Steudel, Z. Tech. Phys. 17,<br />

pp. 115-120 (1936).<br />

[2] E. W. B. Gill <strong>and</strong> A. von Engel, Proc. R. Soc. London Ser. A 192,<br />

pp. 446-463 (1948).<br />

[3] G. Francis <strong>and</strong> A. von Engel, Proc. R. Soc. London Ser. A 246,<br />

pp. 143-180 (1953).<br />

[4] W. J. Gallagher, IEEE Trans. Nucl. Sci. 26, pp. 4280-4282 (1979).<br />

[5] J. R. M. Vaughan, IEEE Trans. Electron Devices 35, pp. 1172-1180<br />

(1988).<br />

[6] N. F. Kovalev, V. E. Nechaev, M. I Petel<strong>in</strong>, <strong>and</strong> N. I. Zaitsev, IEEE<br />

Trans. Plasma Sci. 26, pp. 246-51 (1998).<br />

[7] C. Bourata,d J.-M Joly, IEEE Trans. Electr. Insul. 24, pp. 1045-<br />

1048 (1989).<br />

[8] A.D. Woode <strong>and</strong> J. Petit, Microwave Journal, pp. 142-155 (January<br />

1992).<br />

[9] A. D. MacDonald, Microwave Breakdown <strong>in</strong> <strong>Gas</strong>es, (Wiley, New<br />

York, 1966).<br />

[10] A. V Gaponov <strong>and</strong> M. A. Miller, Sov. Phys. JETP 7, pp. 168-<br />

(1958).<br />

[11] M. P. Forrer <strong>and</strong> C. Milazzo, “Duplex<strong>in</strong>g <strong>and</strong> switch<strong>in</strong>g with multipactor<br />

discharges”, Proceed<strong>in</strong>gs <strong>of</strong> the Institute <strong>of</strong> Radio Eng<strong>in</strong>eers<br />

50, pp. 442-450 (1962).<br />

[12] W. J. Gallagher, Proceed<strong>in</strong>gs <strong>of</strong> the IEEE 57, pp.94-95 (1969).<br />

115


[13] D. L. Liska, Proceed<strong>in</strong>gs <strong>of</strong> the IEEE 59, pp. 1253-1254 (1971).<br />

[14] A. J. Marrison, R. May, J. D. S<strong>and</strong>ers, A. D. Dyne, A. D. Rawl<strong>in</strong>s<br />

<strong>and</strong> J. Petit, A Study <strong>of</strong> Multipaction <strong>in</strong> Multicarrier<br />

RF Components, AEA Technology for ESTEC, AEA Ref. No.<br />

AEA/TYKB/31761/01/RP/05 Issue 1, (Culham, UK, 1997)<br />

[15] N. Rozario, H. F. Lenzig, K. F. Reardon, M. S. Zarro, <strong>and</strong><br />

C. G. Baran, IEEE Trans. Microwave Theory Tech. 42, pp. 558-<br />

564 (1994).<br />

[16] F. Kossel <strong>and</strong> K. Krebs, Zeitschrift für Physik 175, pp. 382-390<br />

(1963).<br />

[17] D. Vender, H. B. Smith, <strong>and</strong> R. W. Boswell, J. Appl. Phys. 80,<br />

pp. 4292-4298 (1996).<br />

[18] F. Höhn, W. Jacob, R. Beckmann, <strong>and</strong> R. Wilhelm, Phys. Plasmas<br />

4, pp. 940-944 (1997).<br />

[19] P. Fortescue <strong>and</strong> J. Stark (Editors), Spacecraft Systems Eng<strong>in</strong>eer<strong>in</strong>g,<br />

(Wiley, West Sussex, Engl<strong>and</strong>, 1995)<br />

[20] S. Riyopoulos, D. Chern<strong>in</strong>, <strong>and</strong> D. Dialetis, Phys. Plasmas 2,<br />

pp. 3194-3213 (1995).<br />

[21] V. Semenov, A. Kryazhev, D. Anderson, <strong>and</strong> M. Lisak, Phys. Plasmas<br />

8, pp. 5034-5039 (2001).<br />

[22] J. R. M. Vaughan, IEEE Trans. Electron Devices 36, pp. 1963-1967<br />

(1989).<br />

[23] Z. J. D<strong>in</strong>g, X. D. Tang, <strong>and</strong> R. Shimizu, J. Appl. Phys. 89, pp. 718-<br />

726 (2001).<br />

[24] V. E. Henrich, Rev. Sci. Instr. 44, pp. 456-462 (1973).<br />

[25] N. L. S. Mart<strong>in</strong> <strong>and</strong> A. von Engel, J. Phys. D: Appl. Phys., 10<br />

pp. 863-868 (1977).<br />

[26] D. Wolk et al, “Coat<strong>in</strong>gs on Mg Alloys for reduction <strong>of</strong> <strong>Multipactor</strong><br />

Effects <strong>in</strong> RF Components”, Proceed<strong>in</strong>gs from the 5 th International<br />

Workshop on <strong>Multipactor</strong>, Corona, <strong>and</strong> Passive Intermodulation<br />

<strong>in</strong> Space RF Hardware, 12-14 September, 2005, Noordwijk, the<br />

Netherl<strong>and</strong>s.<br />

116


[27] A. Ruiz et al, “Ti, V, <strong>and</strong> Cr Nitride Coat<strong>in</strong>gs for Reduction <strong>of</strong><br />

<strong>Multipactor</strong> Effect <strong>in</strong> RF Components”, Proceed<strong>in</strong>gs from the 5 th<br />

International Workshop on <strong>Multipactor</strong>, Corona, <strong>and</strong> Passive Intermodulation<br />

<strong>in</strong> Space RF Hardware, 12-14 September, 2005, Noordwijk,<br />

the Netherl<strong>and</strong>s.<br />

[28] R. Gehr<strong>in</strong>g et al, “Study <strong>of</strong> Black-Anodized Coat<strong>in</strong>g Influence<br />

<strong>in</strong> <strong>Multipactor</strong> Performance <strong>of</strong> RF Space Hardware”, Proceed<strong>in</strong>gs<br />

from the 5 th International Workshop on <strong>Multipactor</strong>, Corona, <strong>and</strong><br />

Passive Intermodulation <strong>in</strong> Space RF Hardware, 12-14 September,<br />

2005, Noordwijk, the Netherl<strong>and</strong>s.<br />

[29] J. Sombr<strong>in</strong>, Effet <strong>Multipactor</strong>, CNES Technical Report No.<br />

83/DRT/TIT/HY/119/T, (CNES, Toulouse, 1983).<br />

[30] A. J. Hatch <strong>and</strong> H. B. Williams, The Physical Review, Second Series<br />

112, pp. 681-685 (1958).<br />

[31] A. Woode <strong>and</strong> J. Petit, Diagnostic Investigations <strong>in</strong>to the <strong>Multipactor</strong><br />

Effect, Susceptibility Zone Measurements <strong>and</strong> Parameters<br />

Affect<strong>in</strong>g a Discharge, Estec work<strong>in</strong>g paper No. 1556, (ESTEC, Noordwijk,<br />

1989).<br />

[32] A. J. Hatch, J. Appl. Phys. 32, pp. 1086-1092 (1961).<br />

[33] S. Riyopoulos, D. Chern<strong>in</strong>, <strong>and</strong> D. Dialetis, IEEE Trans. Electron<br />

Devices 44, pp. 489-497 (1997).<br />

[34] R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, V. E. Semenov, <strong>and</strong> J. Puech,<br />

Phys. Plasmas 11, pp. 5022-5031 (2004).<br />

[35] R. A. Kishek, Y. Y. Lau, <strong>and</strong> D. Chern<strong>in</strong>, Phys. Plasmas 4, pp. 863-<br />

872 (1997).<br />

[36] R. A. Kishek, Y. Y. Lau, L. K. Ang, A. Valfells, <strong>and</strong> R. M. Gilgenbach,<br />

Phys. Plasmas 5, pp. 2120-2126 (1998).<br />

[37] A. L. Gilard<strong>in</strong>i, J. Appl. Phys. 71, pp. 4629-4631 (1992).<br />

[38] A. L. Gilard<strong>in</strong>i, J. Appl. Phys. 78, pp. 783-795 (1995).<br />

[39] A. Kryazhev, M. Buyanova, V. Semenov, D. Anderson, M. Lisak,<br />

J. Puech, L. Lapierre, <strong>and</strong> J. Sombr<strong>in</strong>, Phys. Plasmas 9, pp. 4736-<br />

4743 (2002).<br />

117


[40] M. Merecki, Rapport de stage. Etude de l’effet multipactor dans<br />

les dispositifs hyperfrequences, Master’s Thesis, (CNES, Toulouse,<br />

2003)<br />

[41] V. G. Andreev <strong>and</strong> D. G. Zaid<strong>in</strong>, Translated from Pribory i Tehnika<br />

Éksperimenta 3, pp. 164-165 (1971)<br />

[42] R. <strong>Udiljak</strong>, G. Li, D. Anderson, P. Ingvarson, U. Jordan, U. Jostell,<br />

A. Kryazhev, M. Lisak, V. E. Semenov, “Suppression <strong>of</strong> <strong>Multipactor</strong><br />

Breakdown <strong>in</strong> RF Equipment”, Conference Proceed<strong>in</strong>gs RVK 02,<br />

June 10-12, 2002, Stockholm, Sweden.<br />

[43] D. Ruzic, R. Moore, D. Manos, <strong>and</strong> S. Cohen, J. Vac. Sci. Tech.<br />

20, pp. 1313-1216 (1982).<br />

[44] N. Diaz, S. Castaneda, J. M. Riplada, I. Montero, D. Raboso,<br />

I. Galan, <strong>and</strong> F. Rueda, “<strong>Low</strong> secondary electron emission th<strong>in</strong><br />

films to prevent the multipactor effect <strong>in</strong> high-power RF devices<br />

<strong>in</strong> space”, Proceed<strong>in</strong>gs from the workshop on <strong>Multipactor</strong>, RF <strong>and</strong><br />

DC Corona, <strong>and</strong> Passive Intermodulation <strong>in</strong> Space RF Hardware,<br />

4-6 September, 2000, Noordwijk, the Netherl<strong>and</strong>s.<br />

[45] R. L. Geng <strong>and</strong> H. S. Padamsee, “Explor<strong>in</strong>g Multipact<strong>in</strong>g Characterstics<br />

<strong>of</strong> a Rectangular Waveguide”, Proceed<strong>in</strong>gs <strong>of</strong> the 1999<br />

Particle Accelerator Conference, 27 March - 2 April, 1999, New<br />

York, NY, USA.<br />

[46] A. Sazontov, M. Buyanova, V. Semenov, E. Rakova, N. Vdovicheva,<br />

D. Anderson, M. Lisak, J. Puech, <strong>and</strong> L. Lapierre, Phys. Plasmas<br />

12, pp. 053102:1-8 (2005).<br />

[47] R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, V. E. Semenov, <strong>and</strong> J. Puech,<br />

Phys. Plasmas 10, pp. 4105-4111, (2003).<br />

[48] A. J. Hatch <strong>and</strong> H. B. Williams, J. Appl. Phys. 25, pp. 417-423<br />

(1954).<br />

[49] Website: http://www.nsab-sirius.com/ visited on April 7. 2004.<br />

[50] ESA for ECSS, Space Eng<strong>in</strong>eer<strong>in</strong>g, Multipaction design <strong>and</strong> test,<br />

ECSS-E-20-01A, (ESA Publications Division, Noordwijk, 2003)<br />

118


[51] D. Wolk, D. Schmitt, <strong>and</strong> T. Schlipf, “A novel approach for calculat<strong>in</strong>g<br />

the multipaction threshold level <strong>in</strong> multicarrier operation”, Proceed<strong>in</strong>gs<br />

from the workshop on <strong>Multipactor</strong>, RF <strong>and</strong> DC Corona,<br />

<strong>and</strong> Passive Intermodulation <strong>in</strong> Space RF Hardware, 4-6 September,<br />

2000, Noordwijk, the Netherl<strong>and</strong>s.<br />

[52] G. Li <strong>and</strong> R. <strong>Udiljak</strong>, Threshold Level Determ<strong>in</strong>ation <strong>of</strong> Multicarrier<br />

Multipaction <strong>and</strong> AM Suppression <strong>of</strong> Multipaction <strong>in</strong> Resonant Cavity,<br />

Master’s Thesis (Chalmers University <strong>of</strong> Technology, Gothenburg,<br />

2001).<br />

[53] A. L. Gilard<strong>in</strong>i, Il Nuovo Cimento 18, pp. 919-929 (1996).<br />

[54] A. L. Gilard<strong>in</strong>i, J. Phys. D: Appl. Phys. 32, pp. 1281-1286 (1999).<br />

[55] H. Shimamori <strong>and</strong> T. Sunagawa, J. Chem. Phys. 106, pp. 4481-4490<br />

(1997).<br />

[56] Y. P. Raizer, <strong>Gas</strong> Discharge Physics, (Spr<strong>in</strong>ger-Verlag, Berl<strong>in</strong> Heidelberg,<br />

1991).<br />

[57] S. J. Buckman <strong>and</strong> B. Lohmann, J. Phys. B: At. Mol. Phys. 19,<br />

pp. 2547-2564 (1986).<br />

[58] S. C. Brown, Basic Data <strong>of</strong> Plasma Physics, (The Riverside Press,<br />

USA, 1967).<br />

[59] H. C. Straub, P. Renault, B. G. L<strong>in</strong>dsay, K. A. Smith, <strong>and</strong><br />

R. F. Stebb<strong>in</strong>gs, Phys. Rev. A 52, pp. 1115-1124 (1995).<br />

[60] “The study <strong>of</strong> multipactor breakdown <strong>in</strong> space electronic systems,”<br />

Tech. Rep. CR-448, (NASA, Wash<strong>in</strong>gton D.C., 1966).<br />

[61] D. Wolk, C. Vicente, H. L. Hartnagel, M. Mattes, J. R. Mosig,<br />

<strong>and</strong> D. Raboso, “An <strong>in</strong>vestigation <strong>of</strong> the effect <strong>of</strong> fr<strong>in</strong>g<strong>in</strong>g fields<br />

on multipactor breakdown,”Proceed<strong>in</strong>gs from the 5 th International<br />

Workshop on <strong>Multipactor</strong>, Corona, <strong>and</strong> Passive Intermodulation<br />

<strong>in</strong> Space RF Hardware, 12-14 September, 2005, Noordwijk, the<br />

Netherl<strong>and</strong>s.<br />

[62] A. J. Marrison, “F<strong>in</strong>al report on the study <strong>of</strong> multipaction <strong>in</strong> multicarrier<br />

systems,” Tech. Rep. AEA/TYKB/31868/RP/2, (Culham<br />

Laboratory, Ab<strong>in</strong>gdon, Engl<strong>and</strong>, 1994).<br />

119


[63] R. Woo, J. Appl. Phys. 39, pp. 1528-1533 (1968).<br />

[64] K. Sakamoto, Y. Ikeda, <strong>and</strong> T. Imai, J. Phys. D: Appl. Phys. 22,<br />

pp. 1840-1847 (1989).<br />

[65] E. Somersalo, P. Yla-Oijala, <strong>and</strong> D. Proch, “Analysis <strong>of</strong> multipact<strong>in</strong>g<br />

<strong>in</strong> coaxial l<strong>in</strong>es,” Proceed<strong>in</strong>gs <strong>of</strong> 1995 Particle Accelerator Conference,<br />

1996, Dallas, Texas.<br />

[66] E. Somersalo, P. Yla-Oijala, D. Proch, <strong>and</strong> J. Sarvas, Particle Accelerators<br />

59, pp. 107-141 (1998).<br />

[67] E. Chojnacki, Phys. Rev. Special Topics - Accelerators <strong>and</strong> Beams<br />

3, pp. 032001:1-5 (2000).<br />

[68] S. K. Nagesh, D. Revannasiddiah, <strong>and</strong> S. V. K. Shastry, Pramana<br />

J. Phys. 64, pp. 95-110 (2005).<br />

[69] V. Semenov, V. Nechaev, E. Rakova, N. Zharova, D. Anderson,<br />

M. Lisak, J. Puech, Phys. Plasmas 12, pp. 073508:1-6 (2005).<br />

[70] R. Kishek, Y. Y. Lau <strong>and</strong> R. Gilgenbach, “Temporal Evolution <strong>of</strong><br />

<strong>Multipactor</strong> Discharge”, Proceed<strong>in</strong>gs <strong>of</strong> the 1995 Particle Accelerator<br />

Conference, 1-5 May, 1995, Dallas, Texas, USA.<br />

[71] D. Raboso <strong>and</strong> A. Alstaff, “A new RF breakdown detection method<br />

based on electron monitor<strong>in</strong>g”, Proceed<strong>in</strong>gs <strong>of</strong> the 4 th International<br />

Workshop on <strong>Multipactor</strong>, Corona <strong>and</strong> Passive Intermodulation <strong>in</strong><br />

Space RF Hardware, 8-11 September, 2003, ESTEC, Noordwijk,<br />

The Netherl<strong>and</strong>s.<br />

[72] A. Neuber, J. Dickens, D. Hemmert, H. Krompholz, L. L. Hatfield<br />

<strong>and</strong> M. Kristiansen, IEEE Trans. Plasma Sci. 26, pp. 296-303,<br />

(1998).<br />

[73] T. Fujii <strong>and</strong> S. Moriyama, IEEE Trans. Plasma Sci. 29, pp. 318-325,<br />

(2001).<br />

[74] R. <strong>Udiljak</strong>, D. Anderson, U. Jostell, M. Lisak, J. Puech, <strong>and</strong><br />

V. E. Semenov, “Detection <strong>of</strong> Multicarrier Multipaction us<strong>in</strong>g RF<br />

Power Modulation”, Proceed<strong>in</strong>gs <strong>of</strong> the 4 th International Workshop<br />

on <strong>Multipactor</strong>, Corona <strong>and</strong> Passive Intermodulation <strong>in</strong> Space RF<br />

Hardware, 8-11 September, 2003, ESTEC, Noordwijk, The Netherl<strong>and</strong>s.<br />

120


[75] R. <strong>Udiljak</strong>, <strong>Multipactor</strong> <strong>in</strong> low pressure gas, Licentiate Thesis<br />

(Chalmers University <strong>of</strong> Technology, Gothenburg, 2004).<br />

121


122


Included papers A–F<br />

123


Paper A<br />

R. <strong>Udiljak</strong>, D. Anderson, P. Ingvarson, U. Jordan, U. Jostell, G. Li,<br />

M. Lisak, L. Lapierre, J. Puech, <strong>and</strong> J. Sombr<strong>in</strong>, “New Method for<br />

Detection <strong>of</strong> Multipaction”, IEEE Trans. Plasma Sci., Vol. 31, No. 3,<br />

pp. 396-404 , June 2003.


Paper B<br />

R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, V. E. Semenov, <strong>and</strong> J. Puech, “<strong>Multipactor</strong><br />

<strong>in</strong> low pressure gas”, Phys. Plasmas, Vol. 10, No. 10, pp. 4105-<br />

4111, Oct. 2003.


Paper C<br />

R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, V. E. Semenov, <strong>and</strong> J. Puech, “Improved<br />

model for multipactor <strong>in</strong> low pressure gas”, Phys. Plasmas,<br />

Vol. 11, No. 11, pp. 5022-5031, Nov. 2004.


Paper D<br />

R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, J. Puech, <strong>and</strong> V. E. Semenov, “<strong>Multipactor</strong><br />

<strong>in</strong> a waveguide iris”, accepted for publication <strong>in</strong> IEEE Trans.<br />

Plasma Sci.


Paper E<br />

R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, V. E. Semenov, <strong>and</strong> J. Puech, “<strong>Multipactor</strong><br />

<strong>in</strong> a coaxial transmission l<strong>in</strong>e, part I: analytical study”, accepted<br />

for publication <strong>in</strong> Phys. Plasmas


Paper F<br />

V. E. Semenov, N. Zharova, R. <strong>Udiljak</strong>, D. Anderson, M. Lisak, <strong>and</strong><br />

J. Puech, “<strong>Multipactor</strong> <strong>in</strong> a coaxial transmission l<strong>in</strong>e, part II: Particle<strong>in</strong>-Cell<br />

simulations”, accepted for publication <strong>in</strong> Phys. Plasmas

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!