ECR, R. Geller

·1
1
\
!
~ ~
..
~
" ,_
11
o
~
.. ..
tt
~
• , • • • • • • • • • o
o
•
G G
~ ~ ~ ~
/'11 II
1"'1
i!IIH
1 iild,~ i H
'
..JIIII ':IIU"'lc--1l~
;"iJ..1,~,'~i~.lltlt,.J,pn. ~!:*'4
,,13.,.111+1,
In
,-
il,n';il
_.-_ ."_
-----------_,_._,---
Electron Cyclotron Resonance Ion Sources and ECR Plasmas
:!J
1*\
i,J
,;h
'.'
t,·
!.').
(t)
~
(J)
o
e
C!)
Iii
"
liiilill
,I
,,',
IIIII
"'I
------- ---~
Electron Cyclotron Resonance Ion Sources and ECR Plasmas
R Geller
Institut des Sciences Nucleaires, Grenoble

(formerly Senior Physicist at Commissariat
l'Energie Atomique)
Institute of Physics Publishing Bristol and Philadelphia
© lOP Publishing Ltd 1996

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic. mechanical, photocopying, recording or otherwise, without the prior permission of the publisher. Multiple copying is permitted in accordance with the terms of licences issued by the Copyright Licensing Agency under the terms of its agreement with the Committee of Vice-Chancellors and Principals. British Library Cataloguing-in-Publication Data
A catalogue record for this book is available from the British Library. ISBN 0 7503 0107 4. Library of Congress Cata[oging-in-Publication Data
Geller, R. (Richard) Electron cyclotron resonance ion sources and ECR plasmas I R. Geller. p. cm. Includes bibliographical references and index. ISBN 0-7503-0107-4 (ppc : alk. paper) 1. Electron cyclotron resonance sources. 2. Plasma (Ionized gases) 3. Heavy ions. I. Title. QC702.7.M84G35 1996 539.7'33- -dc20 96-31451 CIP
Published by Institute of Physics Publishing, wholly owned by The Institute of Physics, London Institute of Physics Publishing, Techno House, Redcliffe Way, Bristol BS 1 6NX, UK US Editorial Office: Institute of Physics Publishing, The Public Ledger Building, Suite 1035, 150 South Independence Mall West, Philadelphia, PA 19106, USA Printed in the UK by J W Arrowsmith Ltd, Bristol
(' ,
~ ~
I'll! I
j
,
e
i I ·'II,j,f
1(,1,1
... ..• "1
"
,,,
""
,I
'1"
-,
Contents
Preface Acknowledgments 1 SPECIFIC ELEMENTS OF PLASMA PHYSICS APPLIED TO ELECTRON CYCLOTRON RESONANCE ION SOURCES (ECRIS)
vii xi
1.1 1.2
1.3
1.4 1.5 1.6

1.7
The breakthrough of the ECRIS and its plasma background Preliminary aspects of plasma at the electron cyclotron resonance Collective plasma phenomena Atomic physics background in ECRIS plasma Motions of charged particles in ECRISpiasma Confinement in magnetic mirror fields The difficult modelling of diffusion processes In magnetoplasmas
14 25
60 93 I 15
131
WAVE-PLASMA
2. I
INTERACTIONS
on a
147
t'-t I
2.2
2.3 2.4 3
Basic aspects of small amplitude EM waves incident coiJ magnctoplasma Specific aspects of waves in warm plasma Wave launchers and coupling structures
158
!VU
It;Q
200 STEADY 216

216
THE ECRIS PLASMA STATES-BREAKDOWN, STATE AND AFTERGLOW

3.1
3.2
3.3 3.4 3.5
/
ECRIS breakdown at low pressure in a vacuum cavity ECRIS steady state discharges, a tentative physical analysis Simple magnetic bottle ECRIS; experimental steady state characteristics Min- B ECRIS: steady state electron characteristics ECRIS afterglow regimes (or post-discharge)
222
236 249 256
v
VI
CONTENTS
ION CHARACTERISTICS AND ION PROCESSES IN ECRIS

PLASMA 4.1
4.2
266 266 267 268 271 273 275 276 278 289 303 307
. Ion heating
Improvement of multicharged ion confinement in ES potential traps 4.3 Theoretical ion confinement times with electrostatic potentials at steady state 4.4 Criteria for multiply charged ion production 4.5 The power flux criterion and the importance of the lifetime of energy Ten in ion sources 4.6 The necessity of electron confinement. and low power dissipation 4.7 The neutral gas density criterion for MI production in steady state 4.8 Semi-theoretical analysis of highly charged ion production 4.9 Ion extraction from the ECRIS plasma 4.10 Emittance of a source (generalities) 4.11 ECRIS beam emittances 5 SIMPLE MIRROR AND BUCKET ECRIS FOR LESS HIGHLY CHARGED IONS
5.1 5.2 5.3 5.4 5.5 5.6 6
311 311 325 331 346 351 357
The history of the first ECR ion sources (1965-1973) Modern simple mirror ECRIS Overdense ECRIS at 2.45 GHz for ion beam processing Industrial ECR plasma and ion sources research at Michigan State University ECR plasma cathode (also called microwave plasma cathode) ECRIS for specific weakly charged ions
MIN·B ECRIS FOR HIGHLY CHARGED IONS 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 A brief history of the development of ECRIS for multiply
362
362 365 376 380 391 394 398 409
413 430
charged ions The status of understanding of min- B ECRIS The magnetic structure in modern min-B ECRIS Highly charged metal ions production in min-B ECRIS Specific applications of the min-B ECRIS Scaling rule attempts and practical results Comparative min-B ECRIS Conclusions and prospects of min-B ECRIS
References Index
:)
")'
'~.'
I"
; i .~
'II I
Preface
Let me begin with a few words about the strange genesis of this book. I was hired by the CEA (French Atomic Energy Commission) in 1948 and worked with them until 1993. From 1963, in the group I directed, I always had at least one ECR (electron cyclotron resonance) plasma experiment running and since 1965 at least one ECR ion source. However. when I retired, after so many years of interesting but hectic life I almost felt ready to quit completely my work, my good friends, my dear enemies and my capricious but beloved electrons .... Then, Herve Nifenecker, the well known nuclear physicist. one of my cronies from long ago, asked me to ponder one of his brand new projects: 'Try to transform a 20 keY singly charged ion beam, containing short-lived (r < 1 s) radioactive metallic ions, into a highly charged beam in less than one second. This point is one of the bottlenecks of the PIAFE project where we want to utilize neutron-rich (1 +) ions formed inside the old Grenoble high neutron flux reactor-then transport these ions 300 metres further, multi-ionize them, and inject them into the old Grenoble cyclotron complex ... '. Having always preferred inexpensive original accelerator concepts rather than orthodox high cost routines, I promised to join the venture (where, by the way, I found again J L Belmont and a few other colleagues). So, in spring 1993, I moved into my new office at the Institut des Sciences Nucleaires (ISN) located just a few hundred metres away from my former headquarters. However, such a small distance was enough to radically change my life; my hectic activities, administrative tasks, compulsory immediate achievements were finished ... I was free to breathe, to think, to dream. I was just politely encouraged to find a solution. Nobody disturbed the reverie of the old ECRIS veteran. Many thanks to all the ISN people for their kind savoir-faire, Thus, for the first time in my life, I could contemplate from the windows of my office the mysterious actions of time. The pink buds became green, the swallows came, filled the torrid sky with their twittering and departed; I saw the yellow .leaves drop from the birch trees and whirl in the humid wind of autumn. Then gradually nature became more torpid, frosty and white. Had the time eventually come for me too to hibernate? One day during a snow storm I observed how the flakes arrived gently and silently on the window pane of my office. This image encouraged me
Vii
,. •• 17
liS
Jt,nrrt If'Qa,
.!:~ . !I~
~
..
.
\' III
PREFACE
j
:,J
to ponder how a metallic Ion beam could he slowed down before hilling a solid surface (where the metal particles would stick for a long while). For that thc beam has to collide with SOI11C invisible targets. But what kind of collisions should I consider? Elastic or inelastic collisions with atoms'! Binary Rutherford collisions? Obviously. their cross sections arc toosmall, so I came to the unusual idea promoted 50 years ago by Chandrasekhar of considering the collective collisions with the plasma of an ion source. The ions should be completely slowed down by collective ion-plasma collisions, then bc simply trapped inside the very plasma of a min-B ECRIS, multi-ionized by thc hot electron and extracted ... all of which is possible in a fraction of a second. So ended the PIAFE reverie, because this purely theoretical idea, of course, had to be checked in an unquestionable experimental set up, the preparation of which needed time (in effect the verification was ascertained more than a year later). But what should I do in the mean time? That was the question. I then recalled that, from several sides, I had been asked to write about the secrets of ECR ion sources. But I was not at all decided whether to accept such a difficult task. I also recalled the joke muttered in the eighties by accelerator people: 'EBIS engineers at least understand why their source performs poorly, whereas the poor ECRIS people don't even understand why the ECRIS performs so well'. At that time I should have taken up the challenge. but I was too busy and harassed. In addition anyhow, some accelerator people began to utilize the source which revolutionized the heavy physics done with cyclotrons. Why should I write a book today? Most of the above-quoted accelerator people are now out of the business ... or keep silent where ECR plasmas are invoked. A book is time devouring, and exhausting work. Would it not be better for me to allay the torments of my winter days with the poetry of the psalms or some modern painting (which is my usual solace)? I also asked myself a few pertinent questions as probably every potential author does. For instance: (a) Who needs such a book? (b) What are the fields the book would coyer? Then there are more subjective questions: (c) Am I not too ambitious, is it not pure vanity? (d) Will I have the talent, the energy and the time to finish such a book? (e) How will I treat the nonlinear phenomena with linear models? After a while as the worm was already in the apple, more narcissism invaded my thoughts. (f) Is it not a shame that the new generation of ion sourcerers re-investigate subjects that have already been treated? (g) Why do they forget or omit to quote what has already been done? An African proverb says: 'when a sorcerer dies in the bush ... a whole library burns' . Then a mysterious voice whispered in my ear: Write it even if it is only vanity. Write it before the library burns. No matter if the physics is linear or not, write what you can. Anyhow, it will be the first treatise about ECR ion sources. The book will trigger some feedback with a natural desire for criticism.
II '
IIIII
"~It
I\
',j!!
I,
PREFACE
IX
Maybe a better book will then be written by somebody else. C'est la vie. But in fact, another disquieting feeling made me decide to begin the book. I noticed that I had forgotten the names of some of my colleagues and even of my adversaries. In other words, I partly decided to write this book because it was a unique occasion to go back to my younger years, to revisit once more in my mind my colleagues, to recall not only their names but also their smiles or grimaces, their voices and their arguments. Without the pros and even more without the cons of my colleagues, I would have done very little in the field of ECR ion sources. They pushed me inside the ECRIS adventure, and they are the catalysts for this treatise. Before describing briefly the contents of the book, let me first specify to whom this book is not addressed. It is not addressed to those who believe that elegant mathematical developments are the basic instrument of progress in ion sourcery and ECR plasma physics. In my opinion, the theoretical models of ECR plasma are often too vague for good derivations and computerizing. We probably need better physical visions rather than equations. Consequently, this book cannot be an academic treatise and it is not basically addressed to students. It is intended to guide researchers and users of ECRIS and ECR plasmas, since the field of development of ECR applications has grown dramatically over the last 15 years. The book should serve as a review, and a reference for the field. The background assumed of the reader is physics graduate level. Readers will find in chapter 1 a gentle heuristic introduction to complex nonlinear plasma physics applied to ECR. They will find a few tentative explanations which have never been invoked in other books on plasmas. They will understand that plasma theory is a long term adventure though it already shows many positive returns. But those who confuse particle dynamics added to atomic physics with plasma physics are wrong, because in ECR plasmas the collective effects are determinant and that is what I try to show in this book. In fact, chapter 1 is a recapitulation of some useful atomic physics and classical plasma concepts with emphasis on EeR phenomena. But it aiso contains a few allusions to subjects not usually covered in books on plasmas. I have also dared to add a few, personal non-orthodox concepts on collective plasma 'effervescence'. Finally, if this chapter seems so long (nearly one third of the book) one should find the explanation in my stubborn and firm belief that ail further ECRIS progress depends on a better knowledge of ECR plasma physics. Chapter 2 treats linear and nonlinear ECR effects. Famous plasma theoreticians have already tackled this issue with more or less success. Only the Utopian case of very small amplitude EM waves in a cold plasma (Te = 0) has a solid basis. Therefore, I have tried modestly to show the complexuy of the
PREFACE
J J
issue .with a minimum of equations and a maximum of experimental examples and heuristic proposals. I have quoted as much as possible those pioneers who dared to propose some solutions and those who just like me, remain somewhat perplexed when faced with the realm of nonlinearities. However, I underline forcefully the difference in behaviour between underdense and overdense ECR plasmas (the latter plays, to my mind, a major role in the nonlinearities). In the following chapters I describe ECR plasma states (breakdown, steady state and afterglow) with their different properties. Chapter 3 deals mainly with experimental electron characteristics, which are particularly dependent on the wave interactions and on the magnetic mirror effects, whereas the ion properties seen in chapter 4 depend much more on the neutral gas background and the ion plasma density. I then recall the criteria for multiply charged ion production and the ion extraction from the plasma. I consider the formation and the beam emittances. However, these last chapters are voluntarily condensed since they are treated in all the well known books on ion sources and beam dynamics. Chapter 5 concentrates on ECR plasma sources for low charged ions whose importance is now recognized in many fields. They are more and more utilized in all kinds of ion surface processes and especially in microelectronics. Their great advantage is the absence of filaments and electrodes and their ability to be expanded to large size plasma discharges. However, in order to recall the truth (and satisfy my ego) I could not refrain from writing a chapter on the invention of the ECRIS in 1965 and to comment on the exceptional results of this first generation of sources. Then follows the description of various types of modern ECRIS with an effort of categorization. Note that the so-called microwave sources (where an ECR effect is not clearly established) are not included. 'Chapter 6 is entirely devoted to the highly charged ion production in min-B ECRIS. The term min-B ECRIS (though not very nice) is utilized to designate an ECR source where the particle confinement is obtained in a magnetic structure where the field modulus has a minimum value in the centre. Again I recall their historical development, their plasma characteristics, their scientific criteria and their basic technology. I underline the importance of the fundamental parameters but I also insist on the empirical improvements which piay a consideraoie rote in the progress of the modern prototypes.
J.l1 111~ I~l
T_
.J...~
1,...,... ..
,.,,1-."_"01 __ & .1... -", "-n"'.:"'!"!1 1 ~lIapl~1 VI 111':' 11I,;;al1':'~
T
J.
UI,;;':'~IlUC
,..I~,.,
......
_=L. ...... L._=_ll ........ __ ."' __ ....... •. UlIl,;;"Y \..UlIlCIlIPUl i:1J. Y
r.~nTC"
.c'-1'-1~
I
}~ I t
~.;)." ,
prototypes and their performances. Finally, let me express the hope that this book will encourage further thinking and research, and that the young generation of talented ion sourcerers will be interested in what has been thought and done up to now. I would then be happy and rewarded. I would also appreciate any comments that could improve this book. R Geller
P'
'I
Iii.
Ii
jl,l
IIIII
"'I
iii
.n:
Ii:
Acknowledgments
Very many sources were utilized in compiling the material in this book and many people quoted in the references should be cited here by name. I hope that they will forgive me these omissions. In fact, I have even to admit that-many of my English expressions are not mine since I learned a great deal of my English in the scientific literature about plasma physics. However, firstly I wish to express my gratitude to my colleagues of my former CEA plasma team, who contributed so much, and over so many years, to the success of the ECRIS. Many thanks also to all those, from different horizons, who encouraged or criticized ECR plasmas. In addition, I wish to express my appreciation to Drs B Vignon and J Chauvin, Directors of the Institut des Sciences Nucleaires, for their hospitality over the last three years during which this book was written. My gratitude is also due to Mr Favro, who prepared many diagrams, Mme Langellier, who deciphered some of my manuscripts, and Mme Laidet, who typed all the text. On a more personal note, let me express my thanks to Tom Green, the well known Culham ion sourcerer, who filtered not only my numerous flaws in English but also some faults in thinking; and finally to my wife Annie, who accepted patiently my scriptural whim and became in the meantime an expert ...in botany!
Grenoble, 1 April 1996
Xl
I' ,
III;
II
, 'I"
11111
"I
i1 I
i
1.
SPECIFIC ELEMENTS OF PLASMA PHYSICS APPLIED TO ELECTRON CYCLOTRON RESONANCE ION SOURCES (ECRIS)
\J
1.1 THE BREAKTHROUGH OF THE ECRIS AND ITS PLASMA BACKGROUND
'I
1.1.1 Ion sources-generalities Applications, categories and desirable properties of ion sources.
Ion sources have acquired a wide variety of applications. They are used in a variety of different types of accelerator for nuclear research; have application in the field of fusion research; and are used for ion milling and implantation, in isotope separators, as a means of rocket propulsion, in mass spectrometers, etc. The positive-ion sources may be subdivided into sources specifically designed to generate singly charged ions and those designed to produce very highly charged ions. In both cases, desirable qualities of an ion source are large ion yield, little gas flow (i.e. high ionization efficiency), low ion energy spread, high brightness and low emittance. Practical considerations such as reliability, long source life and ease of changing ion species are sometimes even more important. Under these conditions, a first conclusion can already be drawn. Though utilized for a century, the short-lived filaments and hot cathodes should, if possible, be eliminated and long-lived electron emitters should supplant them. The breakthrough of the ECRIS is surely the result of the elimination of the cathodes. but as we will see there are also other reasons. Ion source concepts and development of ion sources have been described in thousands of different studies. Many of them are reviewed in books. Six important modern books deal with sources and beams. One can find in these volumes a large collection of ion sources and references [1]-[ 6}. Our purpose is to present ECR plasma concepts and ion sources which were not thoroughly analysed in the above books.
J. J. J. J
SPECIFIC ELEMENTS OF PLASMA PHYSICS
Ions are extracted from plasmas which are the usual ion reservoirs. The principal method of positive-ion formation inside the ion reservoir is electron impact. Surface ionization, spark discharge, laser ionization, field ionization etc are other possibilities. Electron impact. A common method of ionizing a gas or vapour is to pass high-velocity electrons through it, with the ions being formed as a result of electron-atom collisions. Electron energies are typically about 100 eV. but in some special sources they have to be as high as 100 keY. An externally applied magnetic field is frequently used to cause the electrons to travel along a helical path and thereby increase the ionization efficiency. Examples of ion sources utilizing this concept are the duoplasmatron and the Penning ion source, but also the electron cyclotron resonance ion sources, often called ECRIS. The principle is based on the presence of neutral atoms Ao in the source which are ionized by electrons e having energies larger than the ionization potentials of Ao. One then obtains ion A + according to the reaction: e+Ao ~ A + +e+e. When numerous ionizations occur in the medium all the charged particles (ions and electrons) interact and generate the so-called source plasma from which the A + ion beams are extracted with the help of a negative extraction potential. which separates ions from the electrons. Thermal ionization. Atoms possessing low ionization potentials can be ionized by allowing them to strike a heated surface having a high work function. Provided that the ionization potential of the atom is less than or about equal to the work function of the surface, there is a high probability that the atom will be thermally des orbed as a positive ion. The method is particularly well suited to producing ions of the alkali metals. Spark discharge. There are several variations of this technique, but basically a spark is induced between two electrodes, one of which, at least, contains the element to be ionized. Generally speaking, the spark consists of a high-density, high-temperature plasma from which ions can be extracted. The spark can be produced by applvinz a high alternatina potential between two fixed electrodes or by mechanically breaking contacting electrodes. Laser ionization. A focused beam from a high-power pulsed laser can be used to produce a small ball of dense plasma from essentially any solid, and positive ions can be extracted from this ;:!u.:')lfia.The high temperature of the plasma results in the forrnatior; 01 many multiply charged ions; however these ions are energetic, a"~ me beam emittance is not always convenient [9]. ill principle, lasers or other strong sources of electromagnetic radiation can be used to produce ions by photoionization. A photon can directly ionize an atom if its energy exceeds the ionization potential of the atom, but the probability of photoionization is low. Field ionization. If an atom passes close to or is absorbed on a very sharp point where the electric field exceeds a few times 1010 V m ", there is a probability that it will be ionized; the phenomenon is known as field ionization. Such large
1.1.1.2 Methods oj ion formation in the plasma reservoir oj ion sources.
.I,j
III
II
lit
11111'11
THE BREAKTHROUGH OF THE ECRIS
electric fields can be achieved in the vicinity of a specially sharpened tungsten needle placed close to an annular electrode. and gas or vapour passing close to the tip of the needle can be ionized. Ion source engineers are faced with the task of understanding and dealing with the issues of generation. extraction and transport of the ion beam. The transport and extraction problems are similar for most types of ion source. Their basic equations are unquestionable and their experimental aspects seem rather well understood. They are treated in great detail in books on ion sources [1], [2], [6]. In fact, they belong to the field of particle dynamics with rigorous mathematical bases. Ion generation is another field, much less understood or developed, located somewhere in plasma physics famed for its complexity and empirical technology. Most studies concentrate on (i) (ii) (iii) (iv) engineering issues, atomic physics for ion production, surface phenomena occurring on the electrodes and walls and properties of electrical arcs and discharges.
1.1.1.3 The particular importance of specific ion source plasmas.
However, it is important to recall that, in fact, plasma phenomena dominate other considerations and that adequate emphasis has not been given to plasma physics. Relationships between ion source performance and plasma performance are rarely underlined, though the ions are extracted from plasmas, which are the usual ion reservoir. The ion beam current is first of all determined by the plasma density, the magnetic field configuration, the electron plasma temperature, the plasma size, the plasma homogeneity etc. The beam extraction and the beam optics may optimize or worsen the beam quality but never change the intrinsic properties of the source. Thus
(i)
the beam composition is determined by the ion composition inside the ion
(ii) the ion charge state is determined by the electron plasma temperature, the ion lifetime and the plasma density of the ion reservoir, (iii) the beam emittance is fixed by the ion temperature, the plasma turbulence and the magnetic field in which the reservoir is located and (iv) the beam properties depend on the percentage ionization of the plasma, whose constituents are electrons, ions and neutrals (non-ionized particles). If the ionization is less than 10% the role of the neutrals is considerable, but in the opposite case (called highly ionized plasma) the physics of the ion reservoir is dominated by collective plasma effects and this is the case of high-performance ion sources. So we come to the following very simple conclusions: specific, bighperformance ion beams are extracted from specific high-performance plasmas.
Even though some plasma concepts remain unsettled. modern engineers in the field of ion sources just cannot ignore basic plasma physics. Let us recall that for instance the ECR plasma concepts which determine the exceptional potential of ECRIS are the result of pure plasma R&D (1963-1979) later applied to ion source design. The aim of this book is to build a bridge between ECRIS engineering and ECR plasma physics.
1.1.2 Positive ion source concepts [20]-[27]
1.1.2.1 Arc sources and confinement sources. The classical concepts utilize arc plasmas but for different reasons these plasmas are difficult to confine. In addition, due to the erosion of the electrodes, they are not very reliable. Therefore, nowaday plasma confinement sources without electrodes are preferred and promoted. Note that both concepts allow light (protons and alpha particles) and heavy-ion production. A heavy ion can be singly ionized (one electron removed), can be fully stripped, as in argon-l 8+ , or can have any intermediate charge state. Singly charged heavy ions are most frequently used in industrial applications in isotope separators, mass spectrographs, ion implantation etc. These are much easier to generate than multiply charged ions, and frequently the experimenter has several source concepts to choose from, depending upon the application and the physical characteristics of the element to be ionized. Among these sources, the overdense ECRIS is now very fashionable and we study this in detail in chapter 5 of this book. On the other hand, multiply charged heavy ions are almost exclusively used in fundamental research in accelerators, particularly cyclotrons, linear accelerators and synchrotrons. A very considerable amount of research has been devoted to developing such sources. Until 1982, the bulk of this was directed to perfecting the Penning ion source (often referred to as PIGIS), and to the laser ion sources, but electron beam ion sources (EBIS) and min- P F(,PIS, which ~re ;,!~~~~(,!1nfin~ITlpnt sources, gave better results. Finally, the ECRIS are now dominating since they yield ion flows which are orders of magnitude larger than those of EBIS (whose particular qualities are appreciated in some specific applications) (figure 1.1.1). 1.1.2.2 Two typical examples of arc sources [7J. [8J. Duoplasmatrons. The development of the duoplasmatron ion source in 1956 marked the beginning of the high-current era for proton sources. The duoplasmatron makes use of an arc discharge which is constricted as it passes into a very strong magnetic field shaped by iron or mild steel inserts in an intermediate electrode and anode. The beam is extracted at the point where the arc has reached a very small diameter and a very high brilliance. Sources of this
,)
I' ,
1.·1
·j-;;;I
iii
Ij
II
,
\
-,
I , I , I
I
I
I I
10 10
9
!
I
108 t\
7
10
ie i
i
NEON F~~R,G~1
4 8
10
I Ie
10
I I
10
OXYGEN CHARGES I I
I I I
I
Figure 1.1.1 Comparative optimum yields of charged particles per second of EBIS (a), ECRIS (b), PIGIS (c) for neon, argon and oxygen (1990).
,t
:t
k
.5 ,)
type have been developed for accelerators and are described and commented on in all the books on ion sources [7]. Penning ion sources. These sources are based on a high-current gaseous
discharse
._,
in a mazneric
0
field with u sas at a relatiyplv . -
lmv .
"rp~C;:lIrf"1 -
(1 n-3 , -
T0!"!'
/-
e e
Ie
~S
Ie
IS
The source (figure 1.1.2(a» consists of a hollow anode chamber, cathodes at each end, a means for introducing the desired element (usually a gas) and electrodes for extracting the ions. The cathode may be heated to emit electrons, which then help to initiate the arc discharge current, creating the plasma in which the atoms are ionized. The discharge column between the cathodes (the plasma) consists of approximately equal numbers of low-energy electrons and positive ions. The electron density is much larger than can be accounted for by the primary electrons from the cathodes. The average energy of plasmaetectrons may range from a few volts to a few tens of volts. Electrons travel parallel to the magnetic field, are reflected from the opposite cathode and make many traverses of the length of the hollow chamber. The electrons confined by the magnetic field and the cathode notential thereby have a high probabiky of making ionizing collisions with any gas present in the chamber [8].

Cathode
compression
Beam
Drift lube s ,., SC solenoid L He cooled panel--'
·l~t, ~ I
I I
1 Gas
• •
Zone'> of
ECR
1 Gas
••
I, I
(
• • • •
Electron beam collector--i
( EBIS)
(0)
ECR Plasma
·/ /,1 .~ -•
Ions
., •
•
Arc plasma -Anode A
--Hexapole K Cathode Salenoids--··
.lif ..
( PIGIS)
(e)
Ions
Ions
(ECRIS)
( b)
Figure 1.1.2 Main elements of the three IS capable of delivering highly charged ion beams (EBIS. [10]. [11], [97], PIGIS, [6], [8] and ECRIS). The net result of all the processes in the arc plasma is that some partially stripped atoms diffuse perpendicular tv the magnetic field out of the arc towards the extraction system. High yields of charge state 1+-8+ (and for heavier elements, perhaps up to 12+) have been obtained for many elements of the periodic table. In figure 1.1.3, we see PIGIS multiply charged ion performance compared to that of ECRIS. Five drawbacks have pratically eliminated the PIGIS from the race towards high-performance sources for multiply charged ions. The yield of multiply charged heavy ions from Penning sources cannot be significantly improved, largely as a result of (i) the short ion-containment time, (ii) the relatively high source pressure (which increases ion recombination), (iii) too short a lifetime of the hot cathode (20 h), (iv) too Iowan electron energy inside the arc and (v) the poor reliability since the performance changes when the electrodes erode during its operation. In short: though still very popular for less highly charged ions, PIG ion sources are not reliable for highly charged ion production. A large collection of different arc sources may be found in specific books llJ-lbj and in the proceedings of conferences UII ion SUUl~CS [2.1j-[25J.
1.1.2.3 Examples of confinement sources (figure 1.1.2(a) and (b)) [10J-[ 12J, [14J, [15J.
The minimum-B ECRIS. Promising sources designed to overcome the limitations of the PIGIS are the ECRIS, based on magnetic confinement, and the EBIS, invented by Donets [10], whose principle is based on electrostatic confinement. However, contemporary EBIS are expensive high-performance sources which are not suited for many routine operations, whereas the min-B ECRIS are much more robust, versatile, practical and rapidly operative. Such a source is simply made of an empty metallic box filled with very low-pressure vapours, microwaves and specific magnetic B fields for plasma confinement (note that none of these three elements is degradable). The plasma confinement is
I~
"I'
"
IIIII
III
THE BREAKTHROUGH
or THE
ECRIS
~C~\5 __ __
25 -
20
15
PIG'S
~Ne
co
S Ca Ni Kr Xe Ar Ti Ge Mo Sm
I-tJ Bi
U FbTh
10 20 30 40 50 60 70 80 90
Figure 1.1.3 Typical values of the ion charge q produced by PIGIS and min-B ECRIS suited to accelerator applications for the elements up to atomic number Z = 92.
generally obtained by superimposing solenoidal and multipolar fields such that the modulus of the magnetic field is minimum in the centre of the box and maximum near the walls. In between there exists an ECR surface which is a closed magnetic surface where the Larmor frequency of the electrons is equal to the frequency of the injected microwaves. Electrons crossing this surface are then energized by the electron cyclotron resonance. When they pass many times through the resonance. according to their random phase, they acquire a global ECR heating, yielding energies of tens of kiloelectron volts and thus exceeding the ionization potentials of many highly charged ions, but as the ion charges are obtained step by step through successive electron-ion collisions (see section 1.4.9), the ions need long exposure times to many electron impacts (i.e. long plasma confinement). The source has generally two stages in series: the first one creates, at ECR, a cold plasma at relatively high gas pressure (> 10-4 Torr) which replaces the cathodes. The plasma electrons diffuse towards a very low-pressure second stage where ECR2 energizes the electrons inside the confined plasma. Due to the absence of cathodes and any kind of electrode, the longevity of the source is unlimited. The magnetic fields are obtained with ordinary or superconducting coils, permanent magnets or with a mixture of these. The sources can operate continuously or be pulsed. Continuous beams of several microamperes of C6+, NH, 08 and Ne1o+ have been routinely extracted

r--
< E
IOOr----
o DU0PiG"lron o Eeh
I
'0
11>
..,
()
til 20 I-
~ ~
o <:
/
200
,/
-:
)
300 400
00
100
Extracled
Beam
Current (rnA)
Figure 1.1.4 Beam current accelerated by RFQ as a function of beam currea extracted from both a duoPIGatron and single-mirror ECRIS [13]. from these sources and large currents of highly charged heavy ions are available. The ECRIS eliminates all the drawbacks of th- PIGIS: ions are confined. hot electron emitters are eliminated (no filaments or cathodes), gas pressure is very low and the energy of the electron can be adjusted by tuning of the ECR parameters. The operation is very reliable and the source can operate for weeks or months without dismantling. Nevertheless throughout this book we will study the limitations of the ECR plasma in order to improve the performmce even more. ECRIS development is still a challenge [29]-[39]. The simple mirror ECRIS (see chapter 5). In these sources only soienoidal E fields are provided. Thus the confinement becomes poorer and corrcarable tc the confinement of PIG sources. Subsequently, the purpose of these sources is to produce less highly charged or even singly charged ions. However, ever for accelerators, these ECRIS also promise to supplant arc discharge; as highcurrent injectors. since they seem to allow better beam transport. Fer instance figure 1.1.4 shows the proton current accelerated by an RFQ (radio frequencquadrupole) SySl~11l as a function of the beam currents ex.~a.::t.::d ;;-~:..._ ~c~': ::.: ECRIS and the duoPIGatron (which is the name of a very high performance PIGlS). The ECRISproduces a slightly higher accelerated beam cirrent wi,:;" less than half of the extracted beam current of the duoPIGatron. Firtherrnore. the hydrogen mass flow rate is reduced by a factor of five [13]. 1.1.3 The leading role played by ECRIS in fundamental heavy ions [16] research on
\,
)'
1. 1.3.1 Applications with acceLerators. The so-called PIGIS were rsed unt, quite recently in heavy ion accelerators so multiply charged ions .ould onibe obtained by incorporating. if possible, some additional strippers :0 remove
J
,
,I
electrons (see chapter 6). However, strippers introduce many difficulties and reduce the overall reliability. ECRIS now dominate as they produce directly highly charged ions. Figure 1.1.5 shows that an immediate increase in performance is obtained by replacing a PIGIS with an ECRIS. This can be illustrated by considering the situations for cyclotrons and synchrotrons. The particle energy (in megaelectron volts per nucleon) delivered by a cyclotron is given by W = K (q / M)2 where q and M denote the charge and the mass of the accelerated ion respectively, and K is an efficiency factor that depends on the size of the machine and its magnetic field strength. K generally lies between 10 for smaller cyclotrons and 800 for the most powerful. By increasing the ion charges by a factor of two to four one can increase the particle energy by fourfold to 16-fold without altering anything else. This represents a bargain because the cost of such a change (i.e. replacing a PIGISby an_ECRIS) is orders of magnitude-Jess that the cost of increasing K by the same amount. (For instance, in order to increase the value of K by 16, the size of the accelerator must be considerably enlarged and the mass of iron and copper of the new cyclotron should be roughly increased by 1000.) Note, however, that ECRIS are external ion sources and cannot be placed between the 'dees' of a cyclotron like small PIGIS. Therefore, specific injection lines have been realized [17] and are now utilized in most cases (Belmont'S Inflector). The situation is somewhat different for existing synchrotrons. Synchrotrons, like cyclotrons, can be adjusted to cope with all values of e = q / m but their injector systems (including linear accelerators) do not have the same flexibility. However, if the entire synchrotron complex is capable of accelerating alpha particles (q / M = 0.5), as most of them are, then they can immediately accept a variety of completely stripped ions such as 12C6+, 14N7+, 1608+, 20NelO+, 32S16+ and 4OCa20+. For instance, CERN's super proton synchrotron (SPS) providing up to 200 GeV per nucleon has therefore been able to achieve beam energies of 3.2 TeV with 1608+ ions and, more recently, 6.4 TeV with 32S16+ particles, thus permitting an investigation of quark-gluon matter to start without building a special heavy-ion tevatron. In CERN's SPS ring, an ECR ion source is coupled through a radio frequency quadrupole to an existing linear accelerator capable of accelerating 1606+ and 32S12+ (s ~ 0.37) to 12 MeV per nucleon. After passing through a foil that strips away electrons (see below), these particles are transformed into 1608+and 32S16+ and then accelerated without further modification by the whole CERN complex comprising a booster ring, the proton synchrotron (PS) with 12 GeV per nucleon and the SPS (figure 1.1.6) [28]. In autumn 1994 with a new heavy-ion injector at CERN a lead beam of 2.9 x 107 ions per pulse was accelerated in the SPS up to an energy of 157 GeVlfJ. The ion source was in operation almost continuously over a period of about I year and proved to be very reliable. It produces a current of more than 100 /-LA of Pb27+ (after the first spectrometer) during the afterglow of the pulsed discharge.
"---1
10
;,
o o
r:)
CERN (199'»)
L. ,
Superdense Supe~ hot..1~Uer' ~
c!)
:, I'
i:)
,!)
I!)
o
Figure 1.1.5 The impact of ECRIS on heay-ion physics carried out using some well known accelerators, The doubly hatched areas are the regimes covered by the original accelerators; the singly hatched areas are those available to some well known accelerators equipped with ECRIS,
r:)
The impact of ECRIS and their capacity to open up additional fields of research in heavy-ion physics is illustrated in figure 1.1.5 for existing accelerators. The doubly cross-hatched regions define the energies and atomic numbers that could be handled by these accelerators until 1983. The large, singly hatched regions outline the regimes that are accessible to these accelerators equipped with ECRIS systems in 1995. Finally, somewhat over-simplifying the situation, we can say that, to obtain higher particle energy, accelerator people have the choice either to utilize ECRIS on existing high-cost systems or to construct new giant-budget accelerators. Let us also emphasize that presently more than a thousand specialists in highenergy physics utilize accelerators equipped with ECRIS, but that ECRIS are also mounted on heavy-ion synchrotrons for medical research [18].
il
pil
Iii
lili
1111;1
Illi
,'I
III
1'' '1
/
THE BREAKTHROUGH OF THE ECRIS 11
FOIL STRIPPER
L1NAC 1
ECRIS
S120r 06•
12 MeV/u
5 KeV/u_/
Simply by replacing a classical arc source with a min-B ECRIS (. in the above figure) one could increase the beam energy from r- 200 GeV (protons) to 3.2 TeV for 0 ions, 6.4 TeV for S ions and later to 34 TeV for Pb ions (with an improved linac) [28]. Without the ECRIS, the cost of these improvements would have been many orders of magnitude higher than the cost of the ECRIS.
Figure 1.1.6 The CERN accelerator complex.
1.1.3.2 Applications in atomic physics research. Atomic physics research can be performed with accelerators at high and medium energy. As the types of available ions, with different charge states, are now extremely diversified a huge field of research is ready for investigations. On the other hand, low-energy physics with highly charged ions was given impetus by the pioneering studies initially proposed by Bliman in 1968. In fact, the fusion research revealed that highly charged ion impurities were present in all machines and studies were necessary to understand their behaviour in the plasma. The time had come to start considering what was the most needed collision programme for fusion. Clearly, the choice was in favour of low-energy ions but electron ionization of ions was initiated using crossed beam techniques (although unsuccessfully). It appears that the fact that ultra-high vacuum is a prerequisite to perform this type of measurement has been totally ignored. Meanwhile, Bliman observed that ion stripping in ion-atom collision had cross-sections as important as those one was trying to measure. This caused the reorientation to charge-changing ion-atom collisions, which appeared to be the most obvious thing to do. Cross-sections were measured, rate coefficients calculated and it was suggested to the fusion physicists that these data should be accounted for in modelling impurities. Further the spectroscopic observation of these collisions would give some insight into the collision dynamics but also make possiele identifications of unknown lines, which constitutes the basis of a diagnostic
12
• .,
,3
~
R.I.B.E. (Reactive Ion Beam Etching) (Chemical As\i\tl'd Ion Beam Eldllng)
ECRIS L- __
: lOr.; IIr AM
", •• '[
,,-'--4~
(I mplantation)
ECAIS
: ION BEAM
..[
f,
f,
'~
:t
:)
1.8.S.0.
(Ion Beam Sputtering Deposition)
:)
I.A.O.
(Ion Assisted Deposition)
,
'!)
:;,
0
Ion Beam Spullcring Deposition with Ion Assistance
•
EGAIS1 :
•
"
,
,;;;)
:)
· ·
•
: ECAIS2
",
) ,)
Figure 1.1.7 Some applications of ECRIS in industrial ion processing. technique in fusion machines (temperatures and impurity distributions). At the same time, the study of the interaction of slow highly charged ions with surfaces was started. This was a first step to the observation of Coulomb explosion a: the surface and of multiply excited ions and atoms (hence, the concept of the 'hollow atom' by J Briand) [45J. This boosted other basic studies [46]. Many 0: these early experiments were performed at the AGRIPA laboratory in Grenoble which was the first facility equipped with an ECRIS (1980). It used gaseou. elements up to fully stripped ones. All these ions had a large Coulomb potentia which induced various interactions on atoms and surfaces with small ion kineti. energy. As this field of research is vast, many other facilities were equippe; with ECRIS after 1983 [40], [44), [448]. Thus, ECRIS have opened up new areas of research in the low-energ:
I
°i
') )
/i
[L,
I~
"
11'1"'1
illl
ill
II
7--·-
13
range, i.e. without an accelerator, by offering many kinds of highly charged ion collisions. The provision of facilities for several hundred researchers in the fields of atomic and surface science has allowed a renaissance of this discipline all over the world. It is out of the scope of this book to cite the thousand original papers dealing with this topic, but let us cite the proceedings of the conference on highly ionized atom physics [40]-[43] and to salute on this occasion a few physicists who boosted this research with ECRIS: S Bliman, F W Meyer, F de Heer, H P Winter, E Salzborn etc.
1.1.4 Applications of ECRIS in industrial ion processing (figure 1.1.7)

Even ECRIS for less highly charged ions are now on the verge of having a significant impact on a number of other fields. The principal attractions are enhanced reliability compared to arc discharge ion sources with short-lived cathodes, and, in many instances, higher currents of the desired ion (such as reactive ions). Thus different manufacturers of high-current ion implanters already depend on ECRIS (see chapter 5). Most of the ECRIS applications in the microelectronics processing arena have been driven by a large effort from Japanese plasma system manufacturers, especially when it was believed that ECR systems could provide high-resolution etching of silicon to better than 0.5 J.Lm dimensions (high anisotropy of etching and high selectivity are also expected). The role of the ECRIS parameters, the efficiency of the source and the potential uses are only now beginning to be investigated. In 1989, Asmussen reviewed the issues associated with ECR plasmas with particular reference to their use as plasma processing sources [47]. Another review was presented in [19]. Nearly 20% of the papers of the last international conference on ion sources dealt with these topics (often called microwave plasma sources). Many of the investigations are performed in Japan. In one of the review papers the merits and demerits of these sources with respect to classical ion sources are compared [48]. 1.1.5
Conclusiccc of section 1.1.
During the last decade, electron cyclotron resonance ion sources (ECRIS) have transformed heavy ion physics. Because ECRIS can generate substantial currents of very high-charge-state ions, the size and the cost of new heavy-ion accelerators have been reduced considerably and the energy of old machines has been increased dramatically. The unsurpassed reliability of ECRIS is also attractive in an industrial environment. ECR ion sources operate comfortably for hundreds of hours on feed gases that ruin arc discharge ion sources. In some ion sources the hot cathode is now replaced by a small ECR plasma which works as a nondestructive electron donor. Thus ECR plasma supplants little by little the arc plasma in many fields. The development of ECRIS is still continuing. Alternative methods of introducing microwaves, and gas mixtures, are still being investigated; different
14
magnetic field configurations are being evaluated. Other extraction geometries

are being considered. Together. these promise higher beam current and ion charged states. lower emittance. increased microwave power efficiency. improved feed gas utilization. simplified tuning etc. However. these are mainly technological aspects. None of the above-cited modifications should be undertaken without a better understanding of the behaviour of the ion reservoir. In other words. the central issue remains the study of the important parameters of the ECR plasma in its magnetic configuration. All the exceptional properties of ECRIS depend on ECR plasma concepts which are not easy to approach and still leave many questions open. These concepts are lacking in all books on ion sources probably because the studies are complicated and still continuing. Only some scattered knowledge of the ECR plasma has been published in compilations of journal articles dealing with plasma or fusion reseach. Other material exists in proceedings of ECRIS workshops and ion source conferences. Our purpose is to gather these dispersed elements in a comprehensive manner and add some pieces to the puzzle which is far from being completed. To illus.rate the ECRIS breakthrough, let me recall that the first ECRIS and the first ECR cathode worked in 1965 [53]. At the first international conference on ion sources in 1969 [20], only one single paper dealt with ECRIS. At the second conference in 1972, there were two contributions and at the sixth conference in 1995 [25] 70 papers out of some 220 contributions dealt with ECR.
'3
1.2 PRELIMINARY ASPECTS OF PLASMA AT THE ELECTRON CYCLOTRON RESONANCE
1.2.1
Ignition of laboratory plasma [49]
Four states of matter exist on earth: solid, liquid. gas and plasma. A substance may exist in one or more of these states. depending upon certain values of quantities such as temperature and pressure. In the laboratory, if we start to tackle a solid substance. we may change it to a liquid by adding energy to it. The substance melts when the 'heat of fusion' of the substance has been added at the melting temperature. If the temperature . is raised enough, the liquid evaporates when the 'heat of vaporization' has been added and a gas is formed. A solid substance directly becomes gaseous when the heat of sublimation has been added. If more energy is added to the gas. a point is reached when the gas ionizes. This occurs according to Saha's equation when the heat of ionization is added. However we have already seen in section 1.1.1.2 that electron impact is a more general means of ion formation {han thermal ionization and ECRIS are based on electron impact with electrons
l·ii;1
I III: Ii
1:1'.
Illil
I I',
.. -
PRELIMINARY
ASPECfS OF PLASMA
15
accelerated by high-frequency EM fields. We have now created a collection of ions and electrons. which we call a plasma. The ionizing process has divided the molecules and atoms of the gas into separate electrons and ions. A simple definition of a plasma might be a large collection of approximately equal numbers of ions and electrons, meaning that a plasma tends to be electrically neutral, but in section 1.3 we will define the plasma state better through the concept of Oebye length. Note that to effect ionization of a neutral atom or molecule enough energy must be added to it to allow one or more of the bound electrons to escape from the atom. The energy is then shared by means of collisional and collective processes between plasma particles. As already mentioned, since the ions and electrons making up the plasma are electrically charged they may be accelerated by means of applied combinations of dc or ac electric, magnetic and/or electromagnetic fields. This acceleration can be sufficient to increase the energy of the plasma particles until they cause more ionization of neutral atoms by collision. Note also that there are, always a few charged particles present even in 'un-ionized' gas, so that this process may be used to 'ignite' a plasma (breakdown) (see section 3.1).
1.2.2 Self-destruction of laboratory plasma

On earth solid, liquid and gaseous matter is long lived under normal conditions; in contrast plasma matter is very short lived and it disappears immediately if one does not try to conserve the electrical particles. The best particle lifetimes obtained in the laboratory are of the order of 10-2 s; generally they are less than 10-6 s. All the difficulties of laboratory plasmas stem from their tendency to self-destruction due to the process opposite to ionization called recombination. In this effect, electrons and ions come back together to form neutral atoms. This situation, if it were to exist by itself, would eventually result in the complete disappearance of the plasma state. However, if ionization is also occurring at the same time, the plasma can continue to exist. In the plasma bulk the main recombination process is charge exchange but the most efficient giobai destruction process is recombination on the walls which occurs when ions and electrons move towards the walls of the plasma container where they just disappear after recombination. Thus the laboratory plasma must be kept isolated from the walls and III a concentrated form. In short there are two vital plasma issues. The first problem is the choice of a suitable means of production of the plasma. through electron impact ionizations achieved by accelerating the electrons at low gas pressures. (ii) The second problem is the storage of the plasma isolated from the walls (also called plasma confinement). This is the more difficult issue. (i)
16
,
')
1.2.3 The need for plasma confinemcnt-magnctoplasma

To achieve confinement one utilizes the electrical properties of plasma particles. Once ionized. the plasma particles are affected by electric and magnetic fields. If a scheme could be made which confined a man-made plasma so that it did not interact with the walls and kept it concentrated for a long time, the biggest problem would have been solved. Until now however systems have mainly used externally applied electric and magnetic fields to attempt to produce, if not adequate, at least reasonable, confinement. When the plasma is contained in a magnetic field we deal with magnetoplasma, which is a medium where involved electron and ion motion is expected. The particles then exhibit different kinds of helical trajectory due to their gyromotion inside the magrietic field (see section 1.5). These complicated particle trajectories hinder the leakage of particles directly to the walls and thus contribute to the confinement by considerably increasing the path length to the walls and also the time needed to traverse this path. The particle transport process is called diffusion. In the absence of a magnetic confinement field the equations of particle motion are well known with a well defined diffusion coefficient. In a magnetoplasma the diffusion coefficient D becomes a tensor, indicating in short the existence of coefficients DU and D 1.. Du describes the diffusion of the plasma particles along the B field lines and D 1. the diffusion across the B field lines. which should be considerably reduced (see section 1.7).
The-confinement principle. 1.2.3.2 The failure of a general confinement theory of the magnetoplasma (figure 1.2.1). According to classical theories, diffusion across the B lines should decrease drastically with increasing magnetic field CD1. ex B-2). In J .2.3.1
fact it does not and anomalous diffusion occurs, which is proportional to B-1. The problem of anomalous diffusion in magnetoplasmas has been a continuing subject of investigation over the last 50 years. In section 1.7 we trv to tackle the Issue but our efforts will mainly emphasize the difficulties of an adequate description of the collisions which are the driving processes of diffusion. We will then consider collective coiiisions and turbulence and we will be obliged to recognize that our theoretical approach to turbulent plasma is still insufficient and our short review of diffusion phenomena in magnetoplasmas will just show that no satisfactory theory of diffusion exists to date (see section 1.7).
1.2.3.3 The struggle for the right orders of magnitude in the magnetoplasma.
"
Finally, we recognize that particles are more or less confined in a magnetoplasma but not as well as expected. The confinement is capricious and unpredictable and so is the matter of magnetoplasma. As the global confinement times are orders of magnitude shorter than foreseen. it follows that other transport coefficients such as electrical conductivity or thermal conductivity are in the same situation.
II'
li.IH,
If
iii
III
'I"
"1"
"I
III II
"I
." I
PRELIMINARY ASPECTS OF PLASMA
17
Anomalous
Plasma
Diffusion
B
B
Figure 1.2.1 Plasma trapped in a simple magnetic mirror, exhibiting anomalous diffusion losses across the field lines. Under these conditions modelling of a magnetoplasma becomes untrustworthy. Thus the equations of balance of particle creation and loss become unreliable due to the badly determined losses. It is pretentious and vain, according to some theoretical concepts, to predict accurate values of plasma density, particle temperature, electrical or thermal conductivity. Only theories supported by limited scaling laws based on experiments are trustworthy to a certain extent. Generally when, in a modelled magnetoplasma, the predicted parameters are of the same order of magnitude as the experimental ones, one considers that the agreement is good and that the theoretical approach is acceptable. Only the exponent is significant; one number before the exponent is plenty, two numbers are no better (section 1.7.5). However, even though the particle motions are not predictable and are questionable, the electrons, between two ill defined 'collisions, are frozen on the B field lines and gyrate around them. Therefore let us briefly introduce three principal characteristics of the particle gyromotion in magnetoplasma, which are unquestionable.
1.2.4 Particle gyromotion in magnetoplasma

The equation of motion for a charged particle in an electrlcrand magnetic field is given by the Lorentz force
mr = F = e(E
+r
x B).
18
SPECIFIC ELEME~'TS OF PLASMA PHYSICS can be
Much useful insight into the confinement properties of magnetoplasmas
obtained from a study of the motion of individual non-interacting particles in electric and magnetic fields. The basic equation predicts helical trajectories with a gyrofrequency and gyroradius (see section 1.5) during rather chaotic motion. 1.2.4.1 Gyroradius (also called cyclotron radius or Larmor radius). In a simple case where the electric field E = 0 a particle with charge e in motion with velocity V.l, perpendicular to a magnetic field of strength B, experiences a force on it that is perpendicular to both the particle velocity and the magnetic field. This force is given by
F = eVl.B = e(B x v).

Because the force is transverse both to the velocity and to the field, the particle moves in a circular orbit, as if 'tied' to a magnetic field line. The radius of this gyro-orbit is given by p-mv.l
eB .
This is called the cyclotron radius, gyro radius or Larmor radius. The mean gyroradius of the magnetoplasma electrons of average energy We (eV) is
= Pe = 0.0035 --
v'We
B
i!)
cm
where the magnetic field B is given in kilogauss. Similarly, the mean ion radius for ions of charge state z. mass number A and average energy Wi (eV) can be written as
p+ = Pi = 0.16
v'AW;
zB
em
(B in kilogauss). For the same energy and field strength, the electron cyclotron
radius is smaller than the ion cyclotron radius by the square root of the mass ratio. In usual ECR plasmas for energies from a few electron volts to some 10 keV and magnetic fields of a few kilogauss p; has values between 0.1 mm and 1 cm. 1.2.4.2 Helical trajectories and sense of the gyromotion. In the previous expressions, the particle energies are referred to the energy components in the direction perpendicular to the magnetic field. The particle velocity parallel to the magnetic field is unaffected, and the resultant particle motion is a helix centred about the field line to which the particle is tied. Because the electrons and ions carry opposite electric charges, the sense of their gyromotion is opposite; electrons rotate in a right-handed sense with respect to the direction of the magnetic field, and ions are left handed. Thus for efficient wave-electron interaction, the electric wave field should rotate in the same sense as the
.~.'.I
,~
II'
I,
III II
"I
iii
,d I
iii
PRELIMINARY ASPECfS OF PLASMA
19
z
(_) Po (__)
±k ±k
P+
x,
Figure 1.2.2 Geometry of right- and left-handed circularly polarized waves propagating along Bo' with electron and ion Larmor radii and corresponding senses of gyromotion.
electrons. Therefore mainly right-hand polarized waves (R waves) are useful ECR (figure 1.2.2).
at
1.2.4.3 Gyrofrequency (or cyclotron frequency). The particle cyclotron motion occurs at a very well defined angular frequency, called the ion or electron cyclotron frequency, which is given by
We
eB = - = 21ffe.
This can be expressed conveniently as fei and
1.52zB/A
(MHz).
i: -= 2 8B
where B is again in kilogauss. For magnetic fieids of a few kiiogauss the eiectron cyclotron frequency is a few to a few tens of gigahertz and the ion cyclotron frequency is typically a few hundred kilohertz to a few megahertz.
1.2.5 Global characteristics of plasmas at the electron cyclotron resonance

1.2.5.1 A selective energy transfer from the waves to the electrons. As ECRIS are based on ECR plasmas let us very briefly describe the properties of the taller before studying them in more detail throughout this book.
20
In contrast to other magnetoplasrnas, where the electrons are energized by

externally applied continuous electric fields. in ECR plasmas. the electrons are energized by the combined effects of EM wave fields and static B fields. which is generally" conducive to a high ionization efficiency. This useful property. resulting from reduced charged particle losses to the wall. is only one of the many advantages of microwave-generated magnetoplasmasvothers being. for example, an efficient microwave power transfer from the feed line to the discharge, even at very low gas pressures, and the absence of electrodes in contact with the plasma. When the frequency of the EM wave is equal to the gyrofrequency of the electrons the waves can transfer their energy to the electrons in a selective manner called electron cyclotron resonance (ECR) and one obtains a particularly efficient coupling of microwaves to the magnetoplasma. To sustain efficiently an ECR plasma with electromagnetic energy one has first to determine the mechanisms of wave absorption. Even though electron-atom collisions seem essential to the process sustaining the plasma, at low gas pressures they contribute only weakly to the electromagnetic energy absorption. In such a case. one must operate under conditions where collisionless absorption is possible (see section 2.2). Collisionless ECR absorption results from a resonant wave-particle interaction which causes damping of the wave. Such a mechanism occurs when electrons rotating at an angular frequency Wee around Bo field lines are travelling with a velocity Vz along the field lines. When the wave has a right-hand circularly polarized electric field component in the plane perpendicular to Bo, these electrons see this component at a Doppler-shifted frequency Wi = w±,BVz where ,B is the component of the wave vector along B. For W = Wee, the electrons are submitted to a constant electric field and gain energy from the wave (see section 2.2). The corresponding absorption process is called cvcLotron absorption. It is accompanied by :'10 increase of the electron velocity component in the direction perpendicular to Bo. This energy increase is subsequently shared within the whole plasma volume through some rare genuine collisions and through nonlinear processes. These mechanisms constitute the basic ingredients of ECR plasmas. Basic studies of wave-plasma interactions teach us that as already mentioned, for ECR absorption, we need a wave with an adequate polarization, namely a right-hand wave which is a pure transverse wave. In practice, however, the presence of boundaries does not allow the propagation of pure transverse electromagnetic waves, thus the wave electric field always possesses an axial component which does not directly contribute to ECR absorption. Fortunately, it happens that the E field of a wave in the plane perpendicular to Bo can always be decomposed into right-hand and lefthand circularly polarized components.' Only the right-hand wave component is absorbed through ECR absorption and the remaining fraction (left-hand polarized
1.2.5.2 Wave coupling to collisionless ECR plasma.
H'
'"
'I"
11111
HI
, 1'''
ii'
PRELIMINARY ASPECfS OF PLASMA
21
..
1
and axial component) of the wave power is either absorbed after some conversion processes or reflected without being absorbed. According to the wave-plasma theories we also know that (i) the amplitude of the electric wave field tends towards a maximum when a wave enters a resonant lone, thus the ECR zone is a privileged location for the energization of the electrons and (ii) that the (R) waves can in most cases propagate inside the magnetoplasma towards the ECR zone without great difficulty (see section 2.1). Based on theories and experiments. we now know many things about the ECR, but some aspects remain to be ascertained and the research is continuing. This is particularly true (i) for the extraordinary wave component (X wave) which feels a special resonance (upper hybrid resonance) whose effects are often included in the global aspects of the ECR and (ii) when nonlinear wave--electron interactions take place, which occurs when substantial RF power is introduced in the plasma. 1.2.5.3 Typical wave frequencies, wave power and magnetic fields far ECRIS. Until now, the utilized wave frequencies have been in the range 2.45-28 GHz . Klystrons and magnetrons, which are the. usual wave generators, operate at fixed frequencies, typically, 2.45, 5, 6.3, 8, 10, 14, 16, 18 and 28 GHz. They are commonly called RF (radio frequency), HF (high-frequency) or J1.W (microwave) power generators: their powers generally are between 50 Wand 5 kW (according to the need). The static magnetic fields required for ECR are given by
..
'"
[1
.'.
e e
n
g d
I B = fee/2.8
" z
d
n
I
f
where B is in kilogauss and fee in gigahertz. Thus typical magnetic fields in the ECR zones normally lie between 0.875 kG and IO kG. 1.2.5.4 Usefulness of mirrors and uselessness of toroidal configuration ill ECRIS. The two basic functions of the magnetic field are to keep charged particles as far as possible from the walls and to allow an efficient transfer of the electromagnetic energy to the plasma. Because charged particle transport occurs mostly along magnetic field lines (in spite of anomalous cross-B motion) to some extent the geometry and extension of the region filled with plasma can be optimized by a convenient choice of the spatial configuration of Bo, the static magnetic field. Varying the magnetic field configuration intensity makes it possible to move the position of the resonance zone and thereby optimize the tuning. Thus, ECR plasmas can be generated under various static magnetic field configurations. Let us immediately discard toroidal configurations since they are closed pJasl'tla structures. In such a case the confinement is improved but it seems very difficult to extract ions from them in an efficient way. Thus they are useless for ion sources. There remain two basic confinement systems: (i) the magnetic
e y e s a a e
~
:1
c e e
[-
s d
'I). '
\;;
22 .
field IS directed essentially parallel to the plasma axis and, at some point along the vessel, the field lines are forced to converge. A so-called magnetic mirror is then created. Such configurations are generally obtained simply by using a set" of electromagnet coils around the plasma vessel. The required Bo configuration is controlled by the spacing between coils and/or by the current (direction and intensity) flowing through them. Unfortunately, these simple mirror systems yield only mediocre confinement: (ii) multipolar magnetic fields are used with min-B configurations which are the highest-performance systems but their production is more sophisticated. The confinement systems are treated in section 1.6 and will not be discussed further here. In both types of static B field configurations ECR zones are provided where the energy transfer from the waves to the electrons is supposed to be optimized and in both types the magnetic field lines conduct the plasma to an 'open end' from which ions are extractable. 1.2.6 Distribution function and quasi-electron temperatures in ECR nlasmas Although ECR plasmas like most magnetoplasmas never achieve a Maxwellian distribution of velocities, they may approach it and it is convenient to assume that the plasma is described by one or more Maxwellian velocity distributions. However we have already mentioned that the collisions are very rare and that they are not binary as in an ideal Boltzmann type gas. Due to the combined effects of EM fields and static magnetic fields some electron motions with respect to the magnetic field lines are enhanced and others impeded. Subsequently the velocity and energy distribution functions might be different when one considers directions parallel to the field lines or perpendicular. In addition due to the ECR effect two electron populations are often observed: the classical bulk population and a superthermal, energetic, beamlike population (see figure 1.2.3). Thus the plasma is obviously non-thermal and non-isotropic. Under these conditions a description or the bel{ plasma should give the location and veiocity of each plasma particle as a function of time but this is impossible to obtain. Therefore, in spite of its inaccuracy it is customary to use the distribution function f to describe an ECR plasma. From the Boltzmann theorem it is known that under the action of 'binary' collisions an ideal gas relaxes to a Maxwellian distribution of velocities izf(v)
= iz (
m
:)
~
''3
:!)
:l)
:) .:)
211 KT
)3/2 exp(-mv /2KT)

2
where iz = N / V, with N the number of particles of a certain type (e.g. ion or electrons) in the system and V the volume of the system. In the Maxwellian distribution, the parameter T, which defines the distribution function, is the temperature of the plasma.
I"
II'
",
'I'
I'
,:
PRELIMINARY ASPECT'S OF PLASMA

t
{
)
23
1,0
s
s i
1
s
o
:::> u..
z
z
~
0,6
0,4
§
~ 0,2
t;
C
0,0
0,2 0,4 0,6 0;8 1,0 1,2 1,4 1;6 1,8 2,0 2,2 2,4 v VELOCITY ~
2KT
2,6
Figure 1.2.3 (a) A Maxwell-Boltzmann distribution function for an isotropic plasma. (b) Distribution function of a plasma with a non-thermal feature (drifting ECR electrons in the tail of a Maxwell-Boltzmann distribution).
I
t
3 -nKT 2
1 =2
iimv2
dv
I
I '
Having no better choice, though ECR plasmas are poorly described by a Maxwellian, one generally relates the experimental distribution function to the Maxwellian distribution, by describing the system as a plasma at temperature T, with T defined by the above formula. In fact, T, is a quasi-temperature proportional to a kind of average electron energy.
1.2.7 Self-generated waves in ECR plasmas

In addition to the externally launched EM waves (aimed for energy transfer to the electrons in the resonance), there are a large number of possible oscillations in a system of coupled oscillators with many degrees of freedom such as an ECR plasma. Most of them are self-generated or somehow excited wavelike disturbances which will propagate in tbe ~lasma. ECR plasma can propagate linear and nonlinear waves. Linear refers generally to very small-amplitude waves whereas nonlinear refers towgeamplitude waves. Among the latter one observes different electrostaticwaves (also called plasma waves since they cannot propagate outside the plasrnajvThey
1.2.7.1 ES plasma waves {50], {51].
24
k~
j
:j
'!)
:;:) :;)
.r-
~ :J
propagate if there is a distribution of electron velocities or an average electron
velocity in the plasma. The most famous are the Langmuir waves aroused by an initially displaced clump of electrons oscillating at frequency wI" We will see that these oscillations are possibly everywhere and always nascent in the plasma (see section 1.3) and wp increases with plasma density. They might cause ES shock waves (solitons), waves propagatingtransverse to the magnetic field lines (Bernstein waves), harmonic and subharmonic waves with beat frequencies etc. They are probably the agents of plasma turbulence and will be invoked again and again in the following chapters.
J .2.7.2
EM waves; plasma radiation [50 j.
When analysing the EM noise
issuing from an ECR plasma, one observes in addition to the injected waves some wave frequencies linked to possible plasma instabilities, driven by discrete
modes. One also measures frequencies linked to plasma radiation, emitted either from radiating atoms (recombinations and excitation free-bound transitions (see section 1.4) or from accelerated charges). This latter case occurs mainly when an electron encounters another charged particle and makes a sudden change in direction; it then radiates in the so-called free-free transition yielding Bremsstrahlung, but energetic electrons also hit the solid walls and create the usual x-ray radiation (see section 3.2). Finally, electrons moving in circular orbits radiate at the cyclotron frequency and its harmonics. In particular if the electrons are energetic (relativistic), which often occurs in ECR plasmas, there is more energy radiation from all the harmonics than from the fundamental. The main interest in ECR plasma radiation comes from the fact that the radiation properties can be used for plasma diagnostics helping to measure the electron temperature, electron distribution functions, densities etc. Note that the useful information has to be extracted from a very noisy spectrum where the presence of the externally launched EM waves is overwhelming (which makes the diagnostics difficult). The propagation properties of these large-amplitude incident EM waves in the ECR plasma are analysed in section 2.2. j.2.7.3 jon sound waves. An ECR piasma aiso propagates low-frequency ion waves, where the ions form zones of compression and rarefaction just as ordinary sound waves do. In this process, the charge separation electric field provides the coupling between electrons and ions. Even standing ion sound waves are observed in ECR plasma when the electron temperature T, is much higher than the ion temperature T;; the electron pressure provides the restoring force and the heavy ions provide the inertial process. The speed of propagation is given by Tonks and Langmuir [52]
VI -
_ (YeKTe
+ YiKTi)1/2
M,
ex
[KTe]'!2
M,
I"
I' .
I· ,II
i:,I,f
;1
III
Ii
jil
li·I"
.1
Iii II III
II
II I
Ii
COLLECTIVE PLASMA PHENOMENA
25
where Yr and Yi are the usual ratios of specific heat for electrons or ions
respectively. M, the ion mass and K the Boltzmann constant. In ECR plasmas the ion sound waves are often generated together with Langmuir waves when comes 'close to (VRF (see section 1.3.9). They do not harm the ECR plasma but they are harbingers of critical conditions which can trigger strong plasma losses when some experimental parameters of the magnetoplasma are then changed. Note that the magnetic field does not hinder the ion oscillations, but other types of magnetosonic wave are possibly excited in the ECR plasma and many ill defined waves are certainly present. Their roles are not clearly analysed and they seem to be of secondary importance.
w"
1.2.8 Preliminary conclusions

We have briefly emphasized the avantages of ECR plasmas without hiding the complexity of the issue. In the next chapters we will propose more involved parameters in order to obtain a better description. However most of our knowledge is based on linear approaches in 'cold' plasma. If many aspects still remain unexplained this is mainly due to nonlinear processes in 'warm' magnetoplasmas with anomalous diffusion losses. These effects are not yet 'ascribable' by simple theoretical models. In the meantime, semi-empirical approaches are often successful and challenging. Anyhow, studying ECR plasma requires more than simply considering particle losses. One must also take into account the complex properties of the wave sustaining the discharge, for example, its dispersion and field configuration, and determine how the wave energy is transferred to the discharge. Thus ECR plasma has become a quasi-separate field of experimental plasma research, with many favourable results. For ion source engineers the interest in electron cyclotron resonance plasmas essentially arises from their efficient wave coupling, robustness, versatility, flexibility and also the fact that, in spite of their low pressure, these plasmas can be used to generate substantial concentrations of charged particles by achieving high degrees of ionization with little input power. 1\11 tnese expenmentai auvamages are now recogruzeu anu corueroarance the theoretical difficulties inherent to plasma physics and its instabilities. In any case, all other ion sources dealing with rnagnetoplasmas also encounter troublesome situations.
. 1~ .l ., ~ • • ~
_!!
1.3 COLLECTIVE PLASMA PHENOMENA
1.3.1 Macroscopic electrical neutrality and microscopic deviations from neutrality

A basic property of a plasma, which is a consequence of collective interactions among the charged particles. is the tendency toward electrical neutrality. If, over
26
a given volume of the plasma, the density of electrons should differ appreciably from the positive ion density. large electrostatic forces will come into play. As a result. the charged particles will move rapidly in such a manner as to approach a condition" of charge equality. The order of magnitude of the electrostatic fields that would result from a departure from electrical neutrality over an appreciable volume may be obtained by considering, for instance, a typical homogeneous ECR plasma in which the density of both ions and electrons is 1012 particlescm ". Suppose that in some manner all the electrons present in a sphere of plasma of radius r em were suddenly removed; the strength of the resulting electrostatic field E, as derived from Gauss's law, would be E=---
Q1
r24rreo
(MKSA)
where Q is the value of the charge removed. If n, is the electron number density of the plasma, then
43 Q = -7rr nee
3
when e is the electric charge. Upon substituting this result into the above equation, it follows that
If the radius of the sphere is taken as 1 em, then it follows that E

5
'..
6 X 10 V em:" so the field strength has an enormous value. A departure of only 1% from charge equality would give rise to a field of 6000 V crn" near a sphere of radius 1 cm. Although on a macroscopic scale the distribution of positive and negative charges in a plasma must be the same, there are microscopic deviations from neutrality. As a result of the thermal motion of the charged particles there will sometimes be an excess of positive charges and sometimes an excess of negative charges. The reverse situation will, of course, exist in a volume near to an electron. The difference between the positive and negative charge densities will obviously be greater in the immediate vicinity of any given charged particle and it will fall off with increasing distance. This leads to the conclusion that every charged particle may be regarded as being surrounded by a cloud having a net charge of opposite sign. The uniform distribution of these microscopic 'clouds' throughout the- plasma then leads to the macroscopic neutrality. As for the length of the microscopic non-neutral plasma, it is called the screening or shielding distance(or Oebye length). For instance, it is a measure of the distance tram an ion
1"
III
If
I I, I, I
II
;~
I'l'
I II II
"I
jill
COLLECflVE
PLASMA PHENOMENA
27
beyond which the cloud, with a net negative charge, screens off the Coulomb field of that ion from the field of another ion moving nearby. It is for this reason that the term 'shielding distance' is used. DUe to their attractions and repulsions the electrons may be forced into oscillatory motions. This phenomenon, called electron plasma oscillation, is characterized by the electron plasma frequency. Since the number of charged particles in the shielding volume is quite large, a particle can interact with many others in traversing a distance equal to the shielding distance. For this reason, as will be shown below, the effect of so-called long-range interactions is more important than that of short-range encounters in producing large-angle scattering of any given charged particle as it passes through the plasma. The combined effect of all the particles within the shielding volume is called collective interactions.
1.3.2 The relative importance of the collective parameters

For describing classical collective effects in a: plasma, there are three principal characteristics: the Debye length AD, the plasma frequency Wp and the longrange- 90° particle scatterings (also called Spitzer collisions) which are due to many small Coulomb deflections of the plasma particle and which are different from binary collisions. In a relatively quiescent plasma these collective phenomena are well understood, and quantitatively determined. We will develop these concepts and comment on their importance in ECR plasma. Unfortunately not all types of ECR plasma are quiescent, and many of them are turbulent plasmas where nonlinear processes are caused by the generation of non-controlled internal waves and different instabilities. The origins of the turbulence, which is characterized by internal fluctuating fields, are 'multiple and it is out of the scope of this book to study the state of plasma turbulence which will only be briefly described in section 1.3.8. However, it is important to inform our reader that, for instance, the thermalization of the plasma particles, i.e. the randomization of the particle drifts, can equally well be obtained through 90° Spitzer collisions (quiescent plasma) as through fluctuating random fields (turbulent plasma). In the latter case the Spitzer- collisions are obscured by the turbulence and become negligible. Similar ambiguities exist for the often debated plasma diffusion through the magnetic field. The classical collisional diffusion cannot explain the anomalous plasma losses. Therefore it is not worthwhile to follow heavy mathematical developments in order to improve a formula by a factor of two or three when sometimes several orders of magnitude separate theory and experiment. However, even if sometimes the Spitzer collisions are obscured by turbulence (which is also a collective plasma effect ::nlt not yet quantitatively describable) the Debye length and the plasma frequency concepts remain unquestionable and their effects have to be considered in ECRIS plasma.
28
1.3.3 General properties and relations between the Debye length AD and
the plasma frequency w,)
The Debye length 1\ D {53 J. A macroscopically neutral collection of charged particles has certain basic properties that are fundamental to its analysis as a plasma. Perhaps the most important of these are the electrostatic shielding of the Coulomb interaction force between charged particles and the spontaneously nascent plasma oscillation. We have seen that, although a plasma is macroscopically neutral, it is locally non-neutral on some sufficently small microscopic scale called the Debye length AD, and its volume is the Debye sphere. inside the Debye sphere, i.e. over a distance of A D, electric fields (even very strong fields) are allowed to exist, but beyond AD the fields are completely screened as we will see in the next paragraph. Among the electric fields, inside a Debye spherc.electromagnetic wave fields with a wavelength A < A D are plausible. Moreover electrostatic oscillation fields due to the spontaneously nascent plasma oscillations at wp are present. The plasma frequency wp is such that we can write the following relation:
J .3.3. J
where Vth is an average thermal velocity of the oscillating electrons inside the Debye sphere. The above relation stipulates that the maximum elongation of the electronic oscillations cannot become larger than a Debye length. behaves in some situations as a system of coupled oscillators. One basic oscillator frequency of the plasma state is the plasma frequency wI" defined as
1.3.3.2 The electron plasma frequency
wp. A
",I!I.,
.... il
plasma
Wp
= 2Jl'fp =
4Jl'ne
m ...
2) 1/2
/
'~
rad
S-I
(CSG)
with fp ~ 104 In Hz, where n is the number of particles per cubic centimetre, and m is the electron mass. It is customary to call wI' the plasma 'frequency', even though it has units of radians per second. In MKS units, one obtains
wp -
") "'t
_ (~)1/2
com
or
I wI'
= 56.4Jn rad
S-I
·1
(M K SA)
'
II
1,1
'
'I'
"'l'
1,111
I'I
I'
II
I,
COLLELilVE
PLASMA PHENOMENA
"f
(Il...
29
",eop
1012
T T
!~
1'1
l'!
P=-'I\ 2ft ,
....
A.S8C
,..J
1
J
-j
10
,,,I,
1019
,j!,",
1020
n(m·3)~
Figure 1.3.1 Plasma frequency against electron density.
~~.":~;,,~ . ~ :"
..:.' ,::~,;
.:' _::;.:.'.'>
'
"_,
..., ..,.::.+:' ,
. >" ..
•: »,
,'+.
.. '"
-----x
Figure 1.3.2 Plasma slab at t
= O.
In figure 1.3.1 we plot wp versus the plasma density. This graph is very useful because the plasma frequency is often used as a means of specifying the electron density in a plasma.
1.3.4
Particular properties of the plasma frequency
Quasi-certitudes (QC) and subsequent assumptions (SA). For an unbiased critique of the properties let us consider as (QC) those assumptions which are no longer debated in well known books of plasma physics, and as (SA) those which are either still debated or ignored. 1.3.4.1 wp and the plasma (or Langmuir) oscillations (QC). If electrons en a two-component plasma are displaced slightly from their equilibrium position Xo, they will experience a force that seeks to return them to Xo. When they arrive at the equilibrium position, they will have a kinetic energy equal to the potential
h
,
v r:~
30
r
.~
~)
;)
energy of their intitial displacement, and will continue past Xo until they have
reconverted their kinetic energy back to potential energy. The frequency of this simple period harmonic motion will be at lVl" the plasma frequency. This phenomenon is known as an electron plasma oscillation or Langmuir oscillation. In all plasmas. quiescent or turbulent. there is a natural trend for Langmuir oscillations which are always nascent somewhere in the plasma bulk (figure
:)
e
1:1)
<::1)
0
!)
I!!)
1.3.2).
/.3.4.2 wI' and EM wave propagation in plasmas (cut-offfrequency) (QC). If
we should attempt to propagate a plane monochromatic wave through the plasma
e
::)
medium, the solutions to the wave equation will be of the form k is the wave number 2rrf =!:!... = k.
VI' VI' VI'
ei(kz-wt)
where
0
I;)
is the velocity of propagation in the plasma defined as 1
0
:)
'~
ti)
As long as e = £0(1 - w~/(2) is positive, vI' will be real and plane electromagnetic waves can propagate. If e is negative, then vI' and k are imaginary and propagating solutions do not exist. Since w~ is proportional to plasma density, the result is that, as the plasma density increases, a point will be reached where W~/W2 > 1 and then propagation is cut off. Experimentally, this point is often observed in ECRIS and the plasma may very sharply cut off the transmission of EM waves through it as its density builds up. However, due to the presence of magnetic field, some exceptional propagations are still allowed in a magnetoplasma (see section 2.1). The cutoff property of the plasma oscillation is of paramount importance in ECRIS because the injected waves have to reach the ECR zone inside the plasma. If wI' > WRF the RF waves can be reflected (i.e. cut oft) but they can also be absorbed by the plasma through nonlinear processes (SA). In this case we will see in section 2.2 that a turbulent plasma is possibly generated (SA). If one wants to avoid unknown processes and allow the microwaves to penetrate into the ECR plasma, one has to limit the plasma density somehow. In such a case the RF power, the magnetic field and the gas pressure are voluntarily matched to obtain adequate wave penetration. Only then is it possible to transfer the wave energy efficiently to the electron in the ECR (QC). Otherwise ill defined transfer mechanisms occur (QC). When wp > WRF one generally stipulates that the plasma is overdense.
WI'
'-i,
~!)
':!l
~)
o
:)
~)
1.3.4.3
The importance of
in ECRIS plasma; overdense plasma.
:~) II)
'.
l:t
f:) ':)
';)
'~'
f~) Y~)
fI) .._
11) ~)
!"
I"
I
"i
Ii
II,
ji
'.
Iii;
Iii Ii
III
I'
II I
31
1.3.4.4 The basic mechanism of the 'always nascent plasma oscillations (QC).
To introduce the concept of plasma frequency. consider a uniform plasma slab. Assume that at t = 0 all the electrons in the interval XI < X < Xo are displaced to the" left of XI. Further assume that ions are fixed. The excess charge to the left XI is ;ioe(xo - Xl). From Gauss's law. this produces a field at XI in the -x direction of magnitude
E(xd
hoe = -(xo
eo
- xd
XI.
(figure 1.3.2). This field exerts a force on the electrons at motion for which is
the equation of
In terms of the relative displacement.
This equation may be written
This is the harmonic osciIlator equation. with solution

~(t) = A exp(iwpet)
+ B exp( -iwpet)
e 2 no
) 1/2
where
Wp =Wpe ( -mecO
is the electron
nln~ma
treauencv .
~ ..-
A similar
definition for the Ion plasma
frequency is
where z is the ion charge state. Thus we see that the plasma frequency is a natural frequency of oscillation. for each species in the plasma. As is well known from the theory of harmonic oscillators. the osciIlations can be excited in response to an external stimulus with frequency less than or equal to the natural frequency. Thus each 'plasma species can respond to a perturbation with frequency W < wp' Because
32·
electrons are able to respond to much higher-frequency perturbations (i.c.

microwaves) than are ions. Therefore in ECRIS plasmas only are basically important.
(V"r
oscillations
1.3.4.5
Amplitudes of the oscillating electric field at plasma frequency (SA).

and, according to the definition of AD, over a
We know that AD ~ vlh/2rrwl'
distance equal to f... D strong oscillating fields arc allowed. Beyond, due to
the screening effect, these fields are completely damped and only a global temperature T, associated with a potential Uth = KT [e, remains observable on a macroscopic scale. Hence we can write that the local oscillating field at wI' integrated over a distance I\. D yields this macroscopic potential [54].
An
e, dl
o; = -e-'
KT
Replacing AD by it: value we find the handy formula
The formula shows that the amplitude of the oscillating field increases with ,Jii (i.e. with wI') and Jr. In figure 1.3.3 we represent some values of Ep for interesting ECRIS electron densities and temperatures and emphasize that fields of the order of kilovolts per centimetre are present inside the ECR plasma but only over microscopic distances of the order of the Oebye length. However, for an electron the Oebye sphere is an infinite universe and inside this volume the oscillating Ep fields can be in competition with the other driving forces of the electron, namely the local electric fields, the magnetic fields and the ECR forces [55].
have seen that the amplitude of the (VI' oscillations can reach high values. It is ouite understandable that such oscillations eenerate some nronazatinc waves. Electrons moving into the oscillation zone with their thermal velocity will carry information about the oscillating fields. The existence of plasma oscillations has been known since the studies of Langmuir in the 1920s; however, it was not until 30 years later that a detailed theory was published by Bohm and Gross [56] explaining how waves linked to wp would propagate and how they could be excited. The waves are sometimes called Bohm-Gross waves, but in general they are termed Langmuir waves (LM waves). It is the thermal motion of the plasma electrons which causes their propagation. The frequency of the wave (VLM is given by
-:I --...... 1 'U ....,
III"
"I
II
il I
COLLECfIVE
PLASMA PHENOMENA
33
~ > ~ o ..J
W
U.
~:~r
10 0,1 .___"""""" ........................... ...___.__.........__,u..u..u...__.__
109 1010 1011 ............. u....u.......___._-l-L.L.>...L ..... 1012 1013
a:
fU
W ..J
S:2
J11
PLASMA DENSITY (cm-3)
Figure 1.3.3 Ep oscillating electric field amplitude as a function of plasma density at

some values of the electron temperature. Typical conditions for various sources: singly charged positive ion ECRIS; 2, multiply charged ion ECRIS [54].
Vth
I,
is the thermal electron velocity: V;h = 3KTe/me and k is the propagation vector of these waves (see section 2.2). Langmuir waves are not EM waves, but ES waves. They cannot propagate outside the plasma. The k vector is oriented parallel to the electric wave field and its value acts upon the wave frequency. Thus the frequency is linked to the local plasma density and temperature. We can replace Vth by other plasma characteristics, for instance wp and AD; one would then find another relation
w'iM = w;[l
+ 3k2 A b]·
As the wI' oscillations are always and everywhere spontaneously nascent in the plasma, LM waves are also everywhere spontaneously nascent (SA); they play therefore ~ ll"~~in~ rolp T-Inwpvpr, in l'\ fl1::.gnf"t0phsma thf" electrons are orbiting on their Larmor radii PL and gyrating at We while moving into the oscillation zone. The waves thus generated according to Bernstein take into account this information [57] with m = 1,2,3, .... w~N
= m2w~ [1 + (wp/we)2 + Fm(k2p~)]
where Fm (k2 depends on a Bessel function and K is the wave number of the mth Bernstein mode. 1.3.4.7 Wave potential trapping; breakdown of linear models in large-amplitude waves ((le) [58J. When we consider an electron in a large amplitude wave field we can no longer write that
F = mx = eEosinwt
pI>
--
- 0.3.1)
34
since even when the wavelength is much less than L (where L is a characteristic
length of the plasma) Eo can no longer be quasi-uniform (i.e. Eo presents necessarily some gradients in space). In the simplest, unidimensional case, grad E is a linear function of X
'\l E ex dX X.
dE
At point X the electron is submitted to the new force F and we determine X X=
= mx.. = e [E 0 + x
dEo dX ]. sin on
(1.3.2)
f[
. eEo(. Eo+X- dE] smwt = --2 smwt) dX mta
which we put into (1.3.2) and find F
= leEosinwtl +
e2 --Eomw2
dEo. 2 sin wt dX
(1.3.3)
An average force due to grad E appears when one integrates over one period T
(F) =
IT eEowt
o
dt
e + --2
mea
Eo -
dE
dX
jT eEo sin 2 tot dt.

0
(1.3.4)
o becomes
Since
f sin2 tot =
1/2 and the first term of (1.3.4) is zero the average force e2
\ r J = mw2
L 2dX J
(F)
r dE~ 1
=
4 mw2
I e2
d(Eo)2
dX .
\.l.J
11
.., 1:'\
..J)
--
I: -,-.\~p. ,1;' •.•

v _"""
_.
",l\II>H/I:: nprivPI:: ". -J!oJ
-_& .., -~
frnm -~ 1"- .._ ........ 11, _... C;:l1hl::pnl1pntlv nntpnti~1 _. r ~nti _A.a" J
A.&
"'--~-'1--......
= e gradlj! =
e
4mw
2
where gradlj! with
V'(£6)
',~
I lj! =
e£2
0 4mw2
or
lj! '" 0.4 x 10"
__Q
£2
w2
(MKSA)
( 1.3.6)
I,
II,
II· I
I I~ I
II
III
II
II

Electron moving slower than wave first goes to the left and then to the right, etc ...
35
Potential wave
.:
Electron moving faster than wave first goes to the right and then to the left, etc...
Crest move at phase velocity
v • 00Ik
Figure 1.3.4
A potential wave showing trapping of charged particles, The particle motion described is relative to the wave-phase velocitycu/ k (E; <eE2/4mcu2). The trapping is conducive to nonlinear electron-wave interactions [49].
Thus in large-amplitude spatially non-uniform .wave fields, we find pseudopotentials 1/1 (also called 'Gaponov and Miller' potentials) which might trap electrons if their energy is smaller than 1/1. Then the electron-wave interaction becomes necessarily nonlinear since the electron not only follows the oscillating field but also rebounces in a moving potential trap. The breakdown of linear theory occurs clearly when the bounce frequency Wb becomes larger than co. Hence the interactions of electrons with all large-amplitude waves become nonlinear (figure 1.3.4). This can occur in the resonance zones where largeamplitude EM waves are converted into large-amplitude ES waves (Langmuir and Bernstein waves-see section 2.2.4).
1.3.4.8 The ponderomotive forces and the
wp
oscillations (QC).
In a general
case, for a single electron one can write (1.3.5) as follows
I'nonlinear
1 E2
----2
4mw
-~., V~-.
{l.J.I)
If one introduces the electron density (i.e. wp) one 'finds (MKSA) (1.3.8)
which is also called the ponderomotive force (just another word for nonlinear behaviour). Note that this force holds not only for Langmuir waves but also for EM waves [59], that it is proportional to the gradient of (£2) and since the drift for each electron is supposed to be the same [59] the force is proportioaal to w~/w2 i.e. to the plasma density (SA). The higher the amplitude of the waves and the ratio of w~/w2
1 the stronger the nonlinearities (SA). For w~/w!
36
linear behaviour is plausible: for
(V;J(V2 »
I nonlinear behaviour is probable
(SA). Hence the time scaling of (v". We and w will playa determining role in the collective plasma behaviour (SA). When the plasma waves become nonlinear the ponderomotivc forces push the plasma. causing local density depressions called cavitons (see section 1.3.8.2) where ES waves are trapped (not to be confused with electron trapping). These waves form an isolated entity called also envelope solitons whose solutions are possibly describable by the nonlinear Schrodinger equation or by the Kortewegde Vries equation [59]. Note that the cavities are dug by the ponderomotive
""\ 1.'"
force, in three dimensions, and the cavities now trap t:,e energy of the waves which according to (1.3.8) increases the ponderomotive effect even more and enforces the density depressions etc, leading to a caviton collapse which liberates
the energy and then contributes to the electron heating (see section 2.2) (SA).
1.3.5
AD, Debye length or screening distance (QC)
I~l" .'"
1.3.5.1 Debye length. Debye spheres and the existence of a true plasma. Ly definition AD is the distance it takes for a plasma to shield itself from an applied continuous electric field. In other words it represents a length within the plasma over which these electric fields are excluded: thus A D is also a minimum dimension of a neutral plasma and subsequently one admits that a uniform homogeneous neutral plasma is quasi-unipotential. Only when an inhomogeneous plasma contains parts with different densities, particle velocities or temperatures linked together by gradients is it possible to imagine differences of potentials. Electric fields are therefore mainly located between the plasma and walls, in 'sheaths'. which are not homogeneous plasmas. We can also say that A D provides a measure of the distance over which the influence of the electric field of an individual charged particle is felt by the other charged particles inside the plasma. The moving particles arrange themselves in such a way as to effectively shield any electrostatic fields within a distance of the order of the DebYl- length and this shielding of electrostatic nelds is a consequence of the collective effects of the plasma particles. Let us compare the potentials of a charged particle in it vacuum and inside a plasma. The electrostatic potential of an isolated ion of charge q is
¢=f}_. r
,""
,
L
1 ,
;
'flt:fJ
"') ""'i_
I
!
'...1·
!'
-/
",~j'
..
, "".',( '" ~~--
In a plasma, electrons are attracted to the vtcrmty of the ion and shield its electrostatic field from the rest of the plasma. Similarly, an electron at rest repels other electrons and attracts ions. This effect alters the potential in the vicinity of a charged particle. The potential of a charge at rest in a plasma is given by
¢ = - exp(
r
-r / A/J)
"
I
I I.
III
I ~I
II
1,,11

C
[1
37
1\ D 1\ ~m
10
)
0,1
O'OO'i-t-r
0,01
l
I
I>
n cm-3 Debye length against n and T~.
Figure 1.3.5
where A D is the Debye length. It can be shown that

AD =
( ne2
EoKT)I/2
(MKSA)
or
AD
= 740 ( -;
T)I/2
where n is the density of electrons (or ions) (cm-3), T is the temperature in (eV), A D is in centimetres. The Debye length, which is a measure of the sphere of influence of a given charge in a plasma, is directly proportional to the square root of the temperature T and inversely proportional to the square root of the electron number density,
n.,
In figure 1.3.5 we plot the Debye length versus plasma density and electron temperature for possible ECRIS application. The relatic- l' = (q/r)exp(-rjl\D) describes a Coulomb potentia! a! S~~!! r (r « AD) but decreases much more rapidly than a Coulomb potential for r ~ AD. Thus the electrostatic potential arising from a microscopic nonuniformity in density-for example, the location of a charged particle-is shielded by a cloud of other charged particles within a distance AD. However, in order for this argument to be valid, the number of particles inside a sphere of radius 1\ D must be substantial. We may compute the number of particles inside a sphere of radius AD to be
nAo
-]'(ADno
= -]'( 3
(£oKT
ne
)3/2 no· ,
It is required that this number be not too small, or the shielding effects predicted by the theory will not take place. A simplified condition for good shielding is
38
SPECIFIC ELEMENTS OF PLASMA PHYSICS.
that the average interparticle spacing be considerably less than the Debye length,
This condition may be expressed as
I) (
n
1/3
« AD'
The conditions for the existence of a 'true' plasma can be summarized as the following inequality:
1) ( ;;
1/3
AD
where L is some characteristic dimension of the plasma. Thus we note again that the Debye length obviously prescribes a lower limit on the macroscopic dimensions (L) of a plasma, by definition. For L < A D the medium would behave as a collection of free charges dominated by mutual two-body interactions, which is in contradiction with the concept of a plasma. . 1.3.5.2 Debye length, individual particle motions and collisional scattering, When the condition of the existence of a true plasma is satisfied, the motion of a charged particle is determined by collective, long-range interactions with many particles, rather than by the sum of many individual two-body interactions. These collective interactions can be represented by fields which can be calculated with electromagnetic theory using the charge and current distributions. The dominance of these long-range collective interactions over local, twobody interactions is fundamental to the analytical characterization study of a . plasma. It allows us to deal with the problem of calculating individual particle motion in electric and magnetic fields, rather than the infinitely more complex problem of simultaneously calculating all the mutual interactions among the '" 1015 particles that constitute the ECRIS plasma. Only when interactions take place in which the particles approach each other closer than A D can these events be treated separately as collisions. Again let us note that plasma oscillations also introduce collision-like scattering when ES fields, randomly generated by the Langmuir oscillation over distances of the order of AD, reach very high values as shown in section 1.3.7. Thus the random fluctuations of plasma turbulence can also play the role of scattering collisions. 1.3.5.3 The Debye length, the sheaths and biased electrodes in ECRIS. When a boundary surface is introduced in a plasma the perturbation produced extends only up to a distance of the order of AD from the surface. In the neighbourhood of any surface inside the plasma there is a layer of width of the order of AD, known as the plasma sheath, inside which the condition of macroscopic electrical neutrality need not be satisfied. Beyond the plasma sheath region there is the plasma region where macroscopic neutrality is maintained. The sheaths are automatically created by the ambipolar diffusion (see secuon 1.7) when a
"')
..
.. ,', '"
II· I
I~ i
Ii
'I'
:1
I'
'i
COLLECTIVE
PLASMA PHENOMENA
39
I
!
,i
plasma is in contact with a solid surface. Since the electrons move faster than the ions. there will normally be a tendency for electrons to leave any given region of the plasma more rapidly than do the ions. At the floating potential of a solid surface in contact with a plasma. there is no net current flow to or from the surface. and so the rates at which positive and negative charges reach the surface must be equal. This can be achieved only if there is a sheath of plasma near the solid surface where the number of electrons exceeds that of the ions. Within this sheath. which is somewhat analogous to the oppositely charged cloud surrounding a charged particle, electrical neutrality is not preserved [145]. The concept of the Debye length may also be useful in considering the effects of various probes or other solid elements that are immersed in a plasma. Potentials may develop or may be applied to these devices. The conditions under which they perturb the plasmas can be obtained by considering that sheaths of the order of a Debye length will develop if the probe is floating, but may not develop if its bias potential is of the order of the local potentials present in the plasma. Bias potentials (dsheath rv AD(V / KTe)1/22) greater than the local potentials often result in thickening of the sheaths and a consequent increase in the perturbing effects inside the unipotential plasma. In ECRIS, electrodes are not necessary. However they are occasionally employed outside the hot plasma zone, on the magnetic axis. If the electrode floats, it becomes negative with respect to the hot plasma and creates an electric field E" B. Thus it reflects electrons leaking along the axis and attracts ions arriving on the plasma edge and forms a Bohm-type sheath [60] with possible oscillations and a special potential profile [145]. When the source works at very low gas pressure and few ionizations occur. a negatively biased electrode saves leaking plasma electrons and its effect is beneficial for plasma maintenance since the impinging ions generate secondary electrons on the electrode which are reinjected in addition into the plasma. This effect is then improved when the electrode is negatively biased. Under these conditions it can replace the first stage of the ECRIS (see section 6). However biased electrodes near the cylindrical wall of the ECRIS may have damaging effects since they favour the ignition of arcs across the magnetic held hnes.
1.3.6 Long-range coulomb scattering (QC)

1.3.6.1 The relative importance of Spitzer collisions {61]-{64J. The longrange encounters (or distant collisions) represent the multiple interactions of a single particle with many other particles such that the net effect is to give a large-angle scattering. i.e. about 90°. In principle. these long-range en<z-tittnters can extend over the whole distance over which the Coulomb forces are effective, i.e. the whole of the plasma. However. in order to make possible the calculation of the cross-section (or equivalent mean free path). it is necessary to choose a distance. given by the Debye shielding length, within which interaction of a
40
Figure 1.3.6 Coulomb scattering.

charged particle with other charged particles may be suppcsed to occur. Beyond this distance, the plasma may be regarded as being electrically neutral. so that the particle under consideration is not affected by Coulomb forces. Spitzer has promoted Chandrasekhar's [61] idea that in strongly ionized plasma (ECRIS plasmas are in this category) the cumulative deflections due to small-angle scattering are actually larger than those due to single large-angle scatterings [62], [63]. These cumulative small-angle scatterings resulting finally in a 90° deflection are then supposed to be the most active scattering mechanism inside the plasma. It must be recognized that this is only true in relatively quiescent plasmas. For instance, when the ECR plasma is inside a stabilizing min-B structure (see section 1.6), Spitzer collisions are probably important, whereas in a plasma with strong fluctuating fields the effects of the turbulence yield more scatterings than the cumulative deflections and the Spitzer collisions become negligible.
1.3.6.2 Expressions for Spitzer collisions in the laboratory system [63J.
I:J
Consider a particle of charge Zie and mass M which passes a scattering centre of charge Z2e at an impact distance b. The particle feels a deflecting force of magnitude Z IZ2e2 / b2 for a time of the order of 2b / v. Hence, it suffers a momentum deflection of
l:l(Mv)
Z Z e2 2b I2 _
2z
Z e2 I2
b2
hv
(CGS)
( 1.3.9)
,)
where v is the particle velocity (figure 1.3.6). The corresponding angular deflection is
(l.3.10)
This deflection angle is in the plane defined by the incident and scattered directions. If the particle passes through a plasma of density n, these deflections are in random directions and hence l:l8 = O. However, (l:l8)2 is not zero and there is a random walk, in angle, away from the original entry direction. The total square deflection in a distance A is directly proportional to the number of scatterings and is given by the expression [63J
11·1
I il
il·
ill;
1·1
iiti
II
II I
41
h"".
(1.3.11)
b.. ,.
The upper limit on b recognizes the fact that electrostatic shielding screens the effect of charged particles beyond the Debye radius and has the value
out
bmax
= y~ ~
=AD.
(1.3.12)
defined
The lower limit is taken to be the classical distance of closest approach by
----bmin
-."
or
bmin =
2e2
--2'
Mv
(1.3.13)
through an angle
It follows that the mean free path for a random walk scattering of magnitude
n 12, i.e. (!:::..(/) '"

Awo
1, is given by
= ----------------------- min)' 8rrn(Zt Z2e21Mv2)2In(ADlb

by means
(1.3.14)
of multiple
Now an 'effective cross-section' for a 90° deflection collisions can be defined in the usual manner
AW,
1 = --.
na90"
(1.3.15)
Hence
aW0
= 8rr (
ZtZ2e
22)2 Mv
In
AD ( --. )
bm1n
(CGS).
(1.3.16)
For most ECRIS plasmas In(ADlbmin) has a value between 10 and 20 and is quasi-insensitive to plasma parameters. It is also called the Coulomb logarithm and is often expressed by In A [62], [64]. However the reader should be cautioned that this is not a cross-section in the usual sense since there is not a linear relation between deflection angle and number of scattering centres but rather a square-root dependence. The quantity
42
A90
is a measure of the depth of penetration required for a multiple scattering
through an angle of magnitude 1T /2. If a particle passes through a thickness which is 100 times greater. it does not suffer 100 deflections through 90° but only about'ten of these. Nevertheless. U9()" is a useful quantity in comparing the effects of Coulomb scattering and other collisions. The cross-section for single Coulomb scattering is obtained from the usual
Rutherford formula. It is
(cas).
Hence, the multiple scattering exceeds single scattering by the factor
and the deflection through large angles via multiple small-angle collisions is about two orders of-magnitude more probable than deflection via a single-angle collision.
8In(bmax/bmin)
1.3.6.3 Spitzer collision times and frequencies ill the centre of mass (eM) system. The characteristic time required for a 90° deflection in the CM system by multiple small-angle collisions depends on the relative velocity between particles V' = VI - \)2 and the reduced mass M' = mlm2/ml + mz. Then, instead of (1.3.16) one obtains
,
u90 and we can write
( coM'U'
ZlZ2e
2 )2
2
-In 2rr
(MKSA)
(1.3.17)
with
\)90"
__
nu~o U'
where lJQ{\, is the collision freauencv. Note that until now we assumed that the scattered particle has an arbitrary mass M and we ignored the electron mass m 1 and ion mass m-, Only the charges Zl and Z2 were specified. This is not important for ee and ei collisions because the deflection in the laboratory system is comparable to the deflection in the CM system when m 1 < m-: Thus for electron-electron scattering (ee) and for the scattering of electrons on ions (ei) the above expressions are also good estimates of the characteristic time for 90° deflection in the lab system and one finds that v~ '" v9b" and 1:~c '" 1:~. Calling these characteristics Spitzer frequencies V.I'P and times 1:,11" good handy formulas then give for, ee, ei and ii collisions [51):
/v ... J
,)
';il
"II
"~
II I
I ~;
II
COLLECfIVE PLASMA PHENOMENA
43
't
..
t
s
10
' 5
2 f---
-,I re~'0'~
""r--.
!\.. T.-l05~
"-,
Sl'-..
2
10·1 5
-,
r-,
"'r--. -,
I'\.
-,
---
~- --- ----
-,
l-r.
re-l03ev
i\.
r-,
-,
"~
10
2"
10-2 5 2 10..Ji'.. 5 2 10
13
I~
b..
~
~r-..
" "t.-40~ <,

<,
<,
"-
-,
'"
-,
-,
102
2
5
~10e~
r-,
103
2
<,
2 5
r-,
!"l
5 10'2 2
109
10
10 2
5 10" 2
nCcm..J)
Figure 1.3.7 Self-collision time as a function of electron energy and density. The dark zone is for high-charged ion ECRIS.
(1.3.18)
II
In figure l.3.7 we plot the Spitzer ee collision frequencies for typical ECRIS plasma parameters.
We have, for ions on electrons, 6.0ie "" (me/mi)t10. Thus the test particle must travel approximatively m, [m, times the distance (A90o) required for a 90° deflection in the eM system before a 90° deflection occurs in the laboratory system. Consequently the characteristic time for a 90° deflection in the lab is about m, / me times the characteristic time for a 90° deflection in the CM system. Thus
Tie .
~9(}' "-'
1.3.6.4 Spitzer collisions for ion-electron scatterings.
"-'
_ (mi)me
rei
90 .
44
,I
1,;1
IJ
, II
"II
1.3.6.5 Momcutun: lrmlsjer ill the Spit~cr collisions
/65/.
The energy : ,}
transferred from particle I to particle 2 in a collision can be found from the collision kinematics (conservation of energy and momentum). For an initial energy of E't. for particle 1, the energy transferred, 6. E, is
J
,
: It "liI
Multiple small-angle collisions that produced a 900 deflection in the eM system would cause a change in energy that can be estimated from the above equation with Be 90
_rv
6.£
2mlm2
Eo -
(m 1
+ m2)2
Thus like-particle collisions result in the transfer of about half of the initial energy in a 90° deflection time. For electrons scattering on ions or ions scatteri,ng on electrons. the fractional energy transfer in a 90° deflection is only about me/mi, Hence the characteristic time for energy transfer, rm. is related to the 90° deflection time by
The above formulas are valid for cold plasma (z = 1; T, '" ~). For ECRIS, one has to take into account the values of z and (Te / T, )3/2 in the Spitzer times for 90° deflection.
1.3.6.6 Consequences of the Spitzer collisions in ECRIS: cold ions, hot electrons and collisionless plasma. We see that electron momentum transfer times equal
the Spitzer times, thus electron heating (energy equipartition) through ee Spitzer collisions seems much easier and faster than ion heating through ei Spitzer
colhsions \ t~'
~... • I .• :
A number of important conclusions follow immediately. The electrons in a ECRIS plasma exchange energy with each other and can reach an equilibrium distribution on a rather short time scale. Let us now consider the hierarchy of characteristic times in usual ECRIS plasma: we note that Vee and Vei are much smaller than Wee and WRF (which are generally between 2.5 and 20 GHz). Hence they do not impede the electron gyromotion. Therefore ECR plasmas are termed collision less plasma, and this remains true in most cases even when inelastic and elastic collisions are added to the Spitzer collisions (see section IA).
t96).
... :
II
Iii
I I II I
III
I"
1,1
"1'
'"
III
"II
;11,,1
II Iii
\
45
The electrons transfer energy to ions. or vice versa, on a time scale that is (111;/111,.) times longer than the time required for the electrons to equilibrate with themselves. For example. in an ECRIS plasma with a density of IOllcm-3 at T, = 10 eV. 6 Va lies between 105 and 10 S-I. Thus electron thermal equilibrium is reached on a very short scale, a few microseconds (which is in general shorter than the electron lifetime) whereas ions would be heated by the same electrons on a time scale of about milliseconds. If their lifetime is shorter than these values they will never be heated at all by ei collisions and will remain cold. The contrast increases for more energetic electrons: when Te '" 1 keY one finds Vee between 102 and 103 s": Only if the electron lifetime could match these values would it be possible to obtain an electron thermalization through ee collisions. As for the ions, their lifetime is absolutely too short to be heated by the electrons. We now understand why in most ECRIS plasmas hot electrons are mixed with cold ions. The electrons which are energized in the ECR do not have the time to heat the ions. Thanks to this cold-ion property the ECRIS may provide ions beams with small emittances (see section 4.1). However, there is also another interesting propeny in plasmas with energetic electrons and cold ions: as the ee collisions between energetic electrons are rare, their collisional diffusion will be weak and their confinement will be possible, whereas the ii collision rate, which at a first glance looks much smaller than v~" can become the main scattering agent because the ion temperature is very low (v~" is proportional to ~-3/2 and to ZIZ2) and highly charged ions might increase the scattering [66]. Of course, all the above properties hold only if no other collision phenomena prevail in the plasma. Thus, when one ignores all the turbulent fluctuations, mainly the 90° collisions provide the randomizing effect on particle motion, and when one considers the case of electrons drifting under the influence of an external RF field these collisions tend to disorder the directed drift motion. In the field direction the equation of motion is
where the last term describes the rate of dissipation of ordered momentum due to collisions. Hence the motion of the electrons will only be impeded in ECRIS plasma when Vei '" WRF which might occur at very low RF frequency and/or when cold electrons populate the plasma (Te « I eV). These conditions are quite marginal in the usual ECRIS for highly charged ions, which seems 'collisionless' but nevertheless they are satisfied in some specific cases. 'Ilten and only then, when the binary collisions are dominated by the Spitzer collisions. and no strong turbulence prevails, the so-called Fokker-Planck equation can replace the ordinary Boltzmann equation for plasma equilibrium [66}, [12J1.
46
1.3.6.7
The slowing
dOWIl
of superthcrma! ions
ill
all EeR plasma through
Spitzer collisions.
The slowing down and the thermalization of fast electric test particles entering into a thermal plasma has been studied theoretically
by different authors [621. [66].
They proposed a collisional damping of
the incident particles through momentum transfer collisions with the thermal electrons and ions of the plasma whose velocities are respectively Vlham and vItam' For fast (super-thermal) incident ions it was shown that they can be thermalized through two kinds of plasma collision: when the velocity ratio Vi~cidenl / VItam ~ 10 the ion-ion collisions are the guiding mechanism. For higher values of Vi~cident/ vItam the damping efficiency drops rapidly until the ratio V;!cidenl/ Vlherm reaches values of the order of unity. In this last case, collisions between the super-thermal ions and the thermal electrons also become frequent and are able to absorb the ion energy of the incident ions. However, this is true in huge astrophysical plasmas, but not in small laboratory ECR plasmas (n 1012, Te > 100 eV, T; 3 eV), where only the long-range (90°) ion-jon collisional damping with V;~cidenl of the order of 104 m S-I (i.e. some 10 eV of incident ion energy) can be efficient since the thermalized ion energy is of some electronvolts (see section 1.3.6.6). The slowing down is then obtained over distances of some 10 ern. Under these conditions the super-thermal ions, after a few ion-ion plasma collisions, are thermalized (in some 10-4 s) and trapped among the ions of the support gas; hence they become cold and Maxwellian and are confined like the other ions. This property is utilized for the trapping of radioactive ion beams inside an ECRIS plasma (see section 6.5.2.2) where the effects of ii collisions dominate the other ion scattering effects (since the slow ions do not respond to high-frequency electric fields). A mathematical expression for this slowing down is given by Delcroix and Bers [66].
f'V f'V
1.3.7
Collective electron scattering caused by plasma oscillations and corresponding collision frequency VW1•
Introduction to the concept of the collective collision frequency
VW
1.3.7.1
Classical collisions based on atomic physics and plasma physics processes are insufficient to describe all the collision-like processes inside an ECR plasma and especially when such a plasma is turbulent (for instance they cannot explain the anomalous particle losses in simple magnetic bottles). For these reasons, we propose a new global type of electron collision frequency vw" related to the always nascent wp oscillations. Such a global collision frequency can explain the ill defined diffusion coefficient V.1 when one considers that V.1 is proportional to Vw (see section 1.7). In " fact, we propose non-classical collisions in order to 'cover' the unknown collective effects due to various instabilities or fluctuations which influence the electron motion. As it is impossible to track separately all these fluctuations, we will try to globalize them. For this let us emphasize that all the observed
l'
(SA).
I·'
-II
II
"il
II
"II
II
ill
47
Debye Spheres
with nascent
oscillations
Scattered e-
Figure 1.3.8 Electron vwp collisions in strongly effervescent plasma.

represent Debye spheres inside the ECRIS plasma.
The circles
instabilities depend in theory somehow on wp (and on the confinement system, i.e. wee). Thus instead of detailing the build-up of collective instability effects, let us rather identify the parameters which act on the wp oscillations which seem to be the basic trigger of these effects and which behave like (collective) collisions. We know that the growing wave instabilities cannot describe 'quantitatively' the plasma equilibrium but if we could evaluate an order of magnitude of the global collective collision rate (without detailing the instabilities) we would be able to predict a global plasma behaviour by writing balance equations. 1.3.7.2 Oscillating fields in the Debye spheres (SA). Let us consider the possible transit time of an electron passing through a Debye sphere where wp oscillations are localized (figure 1.3.8). By definition AD Vth!: = Vth Tp where Tp is the period of the wp oscillation. We have seen that the amplitude of the oscillating field Ep reaches large values (figure 1.3.3). Subsequently, Debye spheres with oscillating fields are plausible electron scatterers. As the transit time of electrons through AD is AD/Vth Tp, the electrons of the velocity distribution with Ve < Vth, will 'see' the oscillations and produce Langmuir waves but a group of swift electrons whose velocity is greater than vi«, will traverse the Debye sphere in less than one period of wp, thus Grey will not even 'see' one entire oscillation, but only random fields oriented in all directions, and they will be subjected to random kicks. After having crossed several successive Debye spheres (with nascent Ep fields) their motion becorae
""V ""V ""V
48
stochastic and not oscillatory. In ECR plasma, the electrons are also submitted
to drifts, whose drift velocity is added to the thermal velocity (Ill> ex E/IJ). Under these conditions the stochasticity is enhanced. In short, one can consider that swift electrons suffer collision type deflections inside the successive Debye spheres they traverse and we assume that \I(t)p is the corresponding collision frequency. Hence the conjugate concepts of Dcbyc spheres and wI' oscillations
allow the hypothesis of

before.
lJ(UI' ,
a type of collision frequency never mentioned
Following our assumption to evaluate VWI' we have to ask what is the corresponding mean free path Awl' (i.e. how far from one oscillating Debye sphere will an electron drift before crossing another Debye sphere with oscillations knowing that the oscillations are not nascent simultaneously everywhere). We can only presume that Awl' is larger than A D and therefore the time interval !:1t is larger than Tf1 and consequently vWp will be smaller than wp; this can be written with
:F ~ 1.
1.3.7.3 Plasma effervescence :F (SA).
:F expresses a kind of effervescence
of the plasma depending on the number and the size of the oscillating Debye spheres (F will increase with the ratio of oscillating spheres with respect to the inert spheres). Thus we have logically fixed the upper limit of the effervescence for vWp tending to wI" but now we must estimate the other possible values of F (table 1.3.1). For the lower limit of F we can utilize experimental data such as measurements of the loss cone flow. For instance measurements in quite noiseless min-B ECRIS proved that for given values of wI' (related to underdense plasma) the loss cone depends exclusively on electron-ion Spitzer collisions [67]. Thus all the other collisions are negligible and in particular Vei > V("n' Consequently F < Vei 2]'(jwp and knowing Vei and wI" one deduces F ;S 10-5. Conversely in underdense simple tandem mirror ECRIS (without min-B) the loss cone seems totally disconnected from the ei Spitzer collisions, and other measurements show that the lifetime of the electrons is '" I 00 times smaller than in the previous example. Hence in this simple mirror configuration one would find F '" 10-3, and we stipulate that vWp is now the dominating collision frequency since we cannot propose other classical collisions (table 1.3.1). Finally in small overdense tandem mirror ECRIS, the measured electron lifetimes are once more 100 times shorter (leading to F '" 10-1) (table 1.3.1).
Relations between the effervescence and the stimulation of plasma oscillations. Now, how can we link this range of F to physical processes
J .3.7.4
in ECRIS? For this, as already specified, it is well known from the theory
'Ii
II
,11111
,11,11
I"
, "111
II
"
"I"
Iii
iii,
II
·,11
II.
i"i
Table 1.3.1
Empirical calibration of the effervescence :F in different ECRIS configurations with some typical characteristics. ECRIS plasma assumed stimuli for wp spontaneous (local) time scaling ECR plasma properties classically
«:
ion charge
lower limit < 10-5 10-5 10-4

<
min-B large quiescent

min-B
Vcoll
«
vwp
Wee
= WRF
ve; Ve;
large, underdense
min-B
idem 0 + microinstabilities (f) idem (f) + gradients (drift instabilities) (£) idem (£) + nonlinear wave conversions micro + macro-instable (bad B curvature) (([) idem (if) + nonlinear wave conversions idem
wp
<
Ve; -+
loss cone >I high

q
+
E Vcoll
Ve;
"V
Vol/V"
small, dense
min-B
>
VeN
cold + hot e: mirroring less performant than 0~<u nonclassical

v
(J
10-3
excessive PRF, overdense simple mirror large. underdcnse simple mirror large, overdense simple mirror
>
Wee
= WRF
vWp
< ,.,
low
q.+
10 .\
E Vcoll
vWp
'"
U.l/v"
Vwp -+
>I
r :>
en
9 -0
0 C ..
10-2
>
VWp
Ve;
>
Wee
VeN
>
vwp
;S Wbounce
"V
loss cone cold + hot e: ± mirroring nonclassical cold e: only off resonance no mirroring Vol/V" I
"V
3::
0 + gradients
(j)
... -0
:>
tTl
E Vcoll
v"'p
v"'p
10-1 upper limit
small, overdense very small, very overdense
>
v"'p
VeN
>
Wee
Ve;
single
q+
idem (j) + strong gradients
>
vWp
>
Wbounce
1.0
..
en Z :>
3::
50
~
STABLE
\_
Magnetic
field lines UNSTABLE
Figure 1.3.9 Schematic diagrams of magnetically confined plasma in regions where the plasma is hydromagnetically (left) stable and (right) unstable. of harmonic oscillators, that the oscillations at wp can be excited in response to a given stimulus with frequencies of wp/2rr or less. Therefore let us envisage a few plausible stimuli which are capable of causing 'forced' electron displacements (in addition to the spontaneous electron-ion separation of section 1.3.4 which is valid for minimum effervescence). Among the stimuli we note for instance (i) ~ fields due to instabilities (see section 1.3.9), (ii) E fields due to internal electron waves, namely those generated by wave conversions near or above the upper hybrid resonance (see section 2.3), (iii) grad E processes due to gradients of density and temperature (see section 1.5), (iv) grad B with Jlgrad B forces and (v) curvature drifts in ordinary mirror configuration. (iii) and (v) apply opposite forces on the electrons and ions and seem to be particularly strong stimuli; (i) and (ii) are high-frequency waves stimulating only electrons. As soon as oscillations at wp are excited, these oscillations may trigger instabilities whose growth rates, frequencies and thresholds depend on the values of wp. Among the instabilities 'micro-instabilities' are observed in all ECRIS. They are mainly due to perturbations of the electron velocity distribution, and occur when 0.1 < w~/w~F < 20, i.e. in underdense and overdense ECRIS [223]. Bad curvature of the B lines yields MHD instabilities (due to magneto hydrodynamics). 'Macro-instabilities' are violent MHO loss processes occurring when nKT /f.-LoB2 > 0.01 (i.e. when W~/W~F ~ 1) but they are triggered mainly in simple magnetic mirrors and not in min-B traps (figure 1.3.9) (see sections 1.6.3 and 3.2.7). On the other hand, magnetic and electric gradients privilege electron-ion
,..,
Iii
II
"I'
"
IIIIII
-II
,II!
III
II,
. 51
separations. Thus the smaller the plasma size the larger the gradients and the
larger the effervescence. Note also that for overdense plasma nonlinear wave conversions are always observed and in addition in overdense plasmas the amplitude of the wp oscillations grows, delivering stronger electron deflections and a better efficiency of the VWI' collisions. Thus we emphasize the leading role of overdense plasma. Probably, the cumulative effects of the invoked stimuli are conducive to variations of :F over five orders of magnitude (see table 1.3.1). Moreover :F decreases with increasing electron energy since relativistic electrons pass through a Debye sphere in less than 10-10 s and then the wp oscillations cannot act on them, whereas in overdense cold electron plasma (T, < 10eV) :F becomes maximum (especially in nonstabilized magnetic structures). Thus, like the Spitzer collisions, VWI' increases globally with ne and decreases with Te. Although the above arguments are only speculative, they are Iinked to ECRIS experiments. They allow one to establish a hierarchy of the agents of effervescence in ECR plasma and yield realistic orders of magnitude for these non-classical collective collision frequencies (inside non-relativistic ECR plasma) expressed by
where:F is given in the table and ne is in cm-3. 1.3.8 Fluctuating potentials and turbulence (SA) 1.3.8.1 Generalities. We always assume that a true thermal plasma is quasiunipotential (section 1.3.4.5) in contrast to plasma sheaths where steady-state potential drops are localized. Let us now consider that fluctuating potentials are somehow generated in the quasi-unipotential plasma. What the linear approach specifies is that the fluctuating potentials should not exceed a given thermal potential U,h = KT/e obtained by stating that the kinetic electron energy Wkin = K T, is balanced by the potential energy Wpo, = eU,h. As in ECR plasmas the electron 'temperatures' are rather high, one can imagine the existence of large fluctuating potentials. The questions now are: where are they located? What are their frequencies and what are their amplitudes? Following Golovanisvsky's assumption the fluctuating potentials are necessarily related to some characteristic plasma lengths and times [54 J. The typical lengths are then chosen to be A D and the times are those of w;l. Under these assumptions the fluctuating potentials must be strongly
52
related to the Langmuir oscillations.
This assumption leads to the initial
location of the fluctuating flelds inside Debyc spheres. with field magnitudes Ep (V ern) fl~(t '" 10-3 j K T (eV)n( cm=") as shown in section 1.3.4.5, for the Langmuir 'oscillation. However, the transition from the harmonic Langmuir oscillations to the sporadic fluctuations is not yet mathematically derived. What we know is that it introduces necessarily nonlinear wave transformations i.e. nonlinear plasma physics. In section 1.3.7 we tried to circumvent this difficulty through the concept of a
global frequency vWt' based on wp oscillations and empirical calibrations. Let us just recall that theoretical studies of plasma turbulence involving very complex nonlinear physics emphasize the leading role played by the Langmuir oscillations in the triggering of plasma turbulence. It is assumed that during the so-called
parametric decay (see section 2.2) the fluctuations may occur at frequencies much lower than wp. According to these assumptions the Langmuir waves are transformed into plasmons, phonons, cavitons etc before establishing a state of turbulence which is precisely characterized by strong fluctuating potentials with a broad frequency spectrum in a macroscopically quasi-unipotential plasma. Now we will briefly invoke some of these modern theoretical concepts. 1.3.8.2 Various tentative descriptions of plasma turbulence turbulence is still the realm of many ill defined concepts. (i) (SA). Plasma
A first school of thought assumes that turbulence is the final state of the
evolution of plasma instabilities which are the initial cause: therefore no

turbulence occurs when instabilities are absent from the plasma. (ii) Other physicists think that all waves inside the plasma, and not only unstable waves, contribute to the state of turbulence. The plasma being a system of numerous coupled oscillators, favours wave-wave interactions, which open the way to nonlinear transitions. Thus a given largeamplitude wave can decay by generating harmonics which interact with the fundamental wave and with the other harmonics they form other waves at beat frequencies which interact with all other waves etc until a broad frequency spectrum is reached. The transition is often called parametric decay. In this case turbulence is possible without instabilities. (iii) In section 1.3.4.8, we noted that in ECRIS plasma large-amplitude waves are created or injected into the system and this might trigger nonlinear phenomena. In addition (see section 2.2) these waves convert into different internal plasma waves, which might interact with the already present selfgenerated waves. In any case, the assumption of coupled oscillators leads to great complexity as shown for instance by Chen [68]. Moreover the difficulties are not only mathematical but also semantic since we have no clear definitions of the interactions and words replace numbers. (iv) For an experimentalist, plasma turbulence is simply the opposite of plasma quiescence. In the latter, no conspicuous noise is observed when diagnostics
I'
II
i,1
II
1·1
f
•
II
j
k
53
l
~
!r
It ,-.
•I
s x s
i
s
f
.
~
are probing the plasma. but below what level the plasma remains nonturbulent now depends on the sensitivity of the probe. (v) These objections become marginal when one invokes the modern semantics proposed by the Russian school of thought which involves new quasiparticles. strongly related to the always nascent Langmuir waves (and other waves); hence all plasmas are supposed to be somehow turbulent. One assumes that the kinetic energy of a clump of electrons inside the field of an LM wave is Wkin = ~ nm (V2) while the wave potential energy (Wpm) is proportional to Here 11 is the electron density, (V) the oscillatory electron velocity and E the amplitude of the electric field of the LM wave. We can always state that (Wpm) = (Wkin) but now the plasma is not only a mixture of electrons and ions. It is supposed to be a mixture of electrons, ions and 'plasmons' which are the carriers of (W poI) whereas the usual plasma particles are only the carriers of (Wkin). As for instabilities they are a temporary result of (Wpot) :f:. (Wkin).
£\'"
In a general case, Golovanivsky even proposed that the stochastic fluctuations a-e of pure thermal origin and thus are unavoidable. For this purpose he simply assumed that [69] In most models, the LM waves seem somehow responsible for the stochastic plasma fluctuations, but other plasma waves carried by other quasi-particles may contribute to the turbulence. 1.3.8.3 Plasmons, phonons and cavitons (SA). Nonlinear LM waves were predicted theoretically in 1972 and since that time numerous theoretical works have dealt with this topic [70]-[74]. In the physics of nonlinear waves in plas~as the waves themselve~ are considered as a collection of quanta with energy hco and linear momentum hk of each quantum. In the case of LM waves the quanta are called 'plasmons' while in the case ot the ion-acousuc waves they are called 'pnonons'. The intensity of a plasma wave then is characterized by the plasmon or phonon densities. A homogeneous plasma wave field appears then as a plasmon or phonon gas of homogeneous density. The main characteristic of such a plasmon gas is that it is unstable and undergoes a cascade of nonlinear transformation if some specific conditions are satisfied. The first transformation, equivalent to parametric decay, occurs when
(1.3.19)
Each plasmon decays to another plasmon with a smaller frequency WL2 and a smaller wave vector kL2 and in addition to a phonon with frequency W,I' and
54
wave vector kJ so that

WLI
= WL2
+W
(I J.20) (1.3.21)
kLJ = ku
+ k,
where (ULI and kLI are the frequency and the wave vector of the initial plasmon. The above condition being easily satisfied, the plasmons immediately
generate phonons which continuously reduce the plasmon energies (frequencies).

Since the frequency of an LM wave WLM = w~ ~k2vlh decreases whereas the plasma parameters wp and Vth remain constant, the wave vector diminishes considerably. Subsequently the plasmons begin to 'condense'. Thus the plasma is filled by a plasmon condensate and a phonon gas. At this stage we can consider that the frequencies of the plasmon gas and the phonon gas already exhibit a broad spectrum. The energy accumulated in the phonon gas is small compared with the en~rgy of the plasmon condensate because hWL »hws• Therefore the phonon gas does not affect the plasma stability and ion sound waves are easily observable even in quiescent plasma. On the contrary, the plasmon condensate is basically unstable and leads to a second cascade of nonlinear transformation [69] which has collective aspects (in contrast to the instability of each individual plasmon during the parametric decay stage). This instability, which is referred to as the 'modulation instability', is such that the initially homogeneous LM wave background violently fragments. Thus the energy of the electrostatic oscillations is no longer uniformly distributed over the plasma but is rather concentrated in a plurality of localized bubbles. The plasma density inside the bubbles decreases automatically in such a way that the sum of the trapped oscillating electrical field pressure and of the kinetic plasma pressure is equal to the ambient plasma pressure. The initially homogeneous plasma density is thus spatially modulated by the bubbles. As a result of this second nonlinear evolution inside the plasma the wave energy is trapped in density cavities, called 'cavitons' which move freely in the plasma. The E field of the cavitons may possibly scatter the plasma electrons in a sporadic way and (i1US account for (he turbulent behaviour. Note that the ponderomotive force (section 1.3.4.8) leads to a similar description of
\"'aVIlUl1~.
................
_-"
1.3.8.4 Plausible caviton characteristics in turbulent ECR1S plasma (SA). Following Golovanivsky [69] to estimate the characteristic time of 'cavitonization' one can use the expression for the increment of the modulation instability:
(1.3.22)
where w is the energy density of the wave. In the case of an argon plasma me/3M = 4.5 X 10-6 so that in this example Imod = :2 X 107 s-I and the caviton birth time is about ( 1.3.23) tmod = 1/ Imod = 50 ns
, I'
i'"
,+I' "
""
'"
II'
"I"
"
"I'
"
III ~ I
'I
"
1i
:11
COLLECfIVE PLASMA PHENOMENA
55
which is short with respect to most plasma times in ECRIS. Hence, as the
Langmuir waves are nascent everywhere in the plasma. the cavitons are also present everywhere in the plasma. Assuming that the above-given model of cavitonization is really what happens in a turbulent ECRIS plasma, in order to evaluate some other values for cavitons one can use the formulas given by Nezlin [69]. For a 10 GHz multiply charged ion ECRIS: 11 = 3 X 1011 cm', T~ = 500eV, PRF = 500 W, one then finds that the density drop in a caviton with respect of the ambient plasma is
Bnln
The caviton diameter is
= ibnk'T,
,...., 0.1.
(1.3.24)
D = 5.5AD(nkT~/w).
(1.3.25)
The velocity of a caviton is
(1.3.26)
According to these formulas the cavitons in a 10 GHz ECRIS contain about 10% less density than the ambient density, are several millimetres in diameter and move with velocity less than 4 x 106 cm S-I in the case of argon plasma. The electrical field amplitude Ecav inside a caviton can be evaluated from the pressure equilibrium condition
(1.3.27)
The field amplitude reaches values comparable to the amplitude of the ES oscillation at wp inside the Debye sphere. At this stage we can say that large oscillating fields, at frequencies lower than the initial wp, but generated originally by the wp oscillations, are concentrated inside quasi-particles called cavitons, which move inside the plasma. The cavitons, which contain the initial energy of the LM waves, constitute large energy reservoirs. They are also capable of heating the electrons after imploding, and thus provide a heating model. Therefore we coasider that the modern theory of plasma turbulence proposes new approaches for many ECRIS problems but the model needs absolutely experimental verificatioe before being aecepted (today it is still considered as an outsider though new nonlinear approaches are urgently required). Note that the other nonlinear models. involving solitons, lead also towards cavitons. They remain mathematical approaches without providing better quantitative evaluations. After all, in the world of particle physics, since 1960, many new particles with odd properties have been invented in order to describe the matter and today nobody discards the models involving quarks, gloons, bozons etc. As the classical particles do not describe plasma turbulence wdl, why not consider odd
"~
56
J
J l
plasma particles if they prove to be better models? However, this is not yet
ascertained since the new quasiparticles seem not to depend on the magnetic fields whereas ECRIS turbulence seems linked experimentally to (Vcr/wPI" In this case. plasmons associated with Bernstein waves (instead of Langmuir waves) or damping of these waves through the Sagdeev-Shapiro mechanism have been considered by Golovanivsky [731.
",
~~
1.3.8.5
Turbulence and instabilities ill ECR plasma.
Many plasma physicists
still assume that instabilities and turbulence are interconnected. Hence a chapter on collective plasma phenomena should incorporate studies of instabilities. These studies however, constitute a vast field of plasma physics (predicting instabilities which are not always observed and ignoring instabilities which were
not predicted). The study of instabilities is tedious-and constitutes a separate discipline which is out of the scope of this book. Nevertheless, we cannot just ignore instabilities and leave them entirely to specialists. In sections 3.2.7 and 3.5.5 we will try to summarize some general aspects and point out a few particular cases related to ECRIS plasma, where one observes macro-instabilities (often called MHD instabilities) and micro-instabilities. In addition RF and drift instabilities might be excited. In short, like turbulence, instabilities are often unavoidable in ECRIS and one has to live with them. C' est la vie ... However, we will see that, if one wants to minimize them, one should utilize min-B structures, avoid overdense plasma, keep the ratio of RF power/source volume as small as possible and maintain a convenient plasma size in order to weaken plasma gradients.
~,
~~
:)
~D
"
:D :I :D
<o/ItiI
,'"'
':t :D
Jl>
::l) '!) !)
1.3.9 An early experiment demonstrating some collective plasma effects involving wP' Wee and AD [78]-[80]
:D 0
!!)
1.3.9.1 Resonant RF plasmoids [76]. In 1959, I performed a series of amazing experiments with RF plasmoids which iliustrate some enigmatic aspects of this chapter and which I briefly describe in the following lines. An RF generator with a frequency range 8-40 MHz and an RF voltage amplitude variable from 0 to 2000 V was coupled to a pumped glass vessel with an RF loop or two metallic electrodes inside the vacuum (figure 1.3.10(a )-(d). Langmuir probes measured the electron densities "1' and electron temperatures T, against RF power, gas pressure and RF frequencies (3 x 106 < n, < 3 X 107 cm >': 5 x 104 < T, < 2 x 105 K). When the gas pressure was decreased from 10-3 to 10-5 Torr a plasmoid with strongly limited contours was built up in the vessel. Its size decreased with decreasing pressure (i.e. the distance from the plasmoid towards the wall increased). This was already known previously
0
0
0 0 0
~
e
,
'0
C)
I ~
,l'
e
~)
e
II
II I
I"
'I"
I.
lit"
,"
111'"1
Ii
ji
,I
I,ll
57
(a)
(b)
(c )
(d)
Figure 1.3.10 Different plasmoids obtained when
WRF
'"
wp.
a,
RF generator; P,
plasmoid; E, metallic electrodes; S, Langmuir probe; W, Wilson seal; V, vacuum pump; g, grid; A.D, Debye length.
[75], [77]. The plasmoid was supposed to be generated by a multifactor effect [77]. Let me now summarize the new discoveries. (i) When the plasmoid was formed, it was possible to reduce VRF to practically o V whilst the plasmoid was maintained unaffected (as if a much stronger VRF was applied) (figure 1.3.10(a),(b),(c». The plasmoid took more or less the shapes of the container walls (several plasmoids in several glass vessels with one RF generator (figure 1.3.10(d~). The probe measurements yielded n, and T, values such that the corresponding Debye length AD ex .JTTii was always smaller than the distance from the plasmoid to the wall (i.e. smaller than the sheath (see section 1.3.5.3». The plasmoid diameter decreased to a minimum roughly equal to AD and then exploded (i.e. min plasma size A D (see section 1.3.5.1 ». The plasma density of the plasmoid was roughly equal to nrr defined by wI' WRr (see section 1.3.3.2 and figure 1.3.11).
r-
(ii) (iii)
(iv) (v)
,~
58
~
..
'"~ 2
-+".-~ __
Alii
P_LA_S_M-,-O_ID_~
~ V5
w
em';
.,~
I-
Gl •
z o
:,~
,)
,,~
w 1,5
a:
:J t-
-c a: w
~
a.
t-
:.~ :.,~
PRESSURE TORR
Figure 1.3.11 nand T against pressure in the plasmoids.
106
" "1
~.
~.
~D
,;~
~~
,)
:E
:I:
40
30 20 10
u.
cr
~, ,
~~
:,
,.D ':1
Figure 1.3.12 n against WRF in the plasrnoids. (vi) The value of plasma density increased with increasing WRF so that :':"1 = WRF was automatically maintained, as long as the plasmoid was fcr::ned (see figure 1.3.12). I came to the conclusion that the plasmoid was resonant at wpe = WRF an: mat the real applied RF field was a kind of internal field as proposed in section 1.:A.5 (Ep ex ..;n;t;) and not the externally applied RF field (ERF '" 0 V cm'-: Let me recall that E is only possible over distances of the order of a Debye srnere, i.e. over the plasmoid in the present case (section 1.3.4.5). The existence )f a field larger than E R F in the resonance was later experimentally dernonstraed in a thesis [78]. [81].
~D
:,
:J)
,D
::>
:D
!)
'.~
:,
~
j) ,;1)
:~
:>
I·i
II
~" I ~
Iii
'I Ii
I'
I I,
r'
il
iii
COLLECI1VE PLASMA PHENOMENA
59
Figure 1.3.13 Mirror plasma is only obtained whca 1.3.9.2 Ion waves at plasma resonance [BOJ.
Wee>
Wp-
(vii) Inside the plasmoid one could easily detect hybrid ion waves (probably the first experimental evidence of pure ion sound waves) which in addition were standing waves, i.e. permanently excited. They obeyed the TonksLangmuir theory [52] with a phase velocity VI/I ()( (KTe/ M+)1/2 (see section 1.2.7.3). The values of Vt/> could be verified by measuring T, with probes and changing the ion masses of the gases (H2, N2, 02, Ne, Hg) from two to 200 [80]. We will emphasize the possibility of excitation of ion waves when Langmuir waves are created, i.e. when WRF wpe (see section 2.4.4).
r-.J
1.3.9.3 Conditions for electron mirroring [79 J. Another very interesting pmpprty is evidenced when the plasrnoid is immersed in a variable magnetic field B obtained with externally supplied coils (figure 1.3.13). (viii)When WRF ex wpe > Wee the plasmoid does not feel the presence of B; the plasma resonance persists. (ix) When Wee = WRF = wpe (i.e. when B rv (m/e)wpe) the plasmoid disappears for weak values of ERF or diffuses in the container for high values of ERF. The plasma resonance is then dominated by the OCR effects. (x) Only when Wee > wpe' does a plasma appear, which is trapped in the 6 mirror field (figure 1.3.13). Similar results are observed in cusped mirrors. This last experiment presumes that magnetic electron mirroring is only possible when Jl = W 1./ B is invariant and is not perturbed by plasma oscillations during its spiralling motion (see section 1.6.3). This point has to be emphasized because
60
it is never clearly stated in books on plasma physics: mirroring of electrons seems only possible when
W(f
>
wp(' .
Thus, overdense ECR plasmas (wpe > wc(' = WRF) are not well confined in a magnetic trap. This is particularly true for small, overdense plasmas in mirrors where VWI' '" wpe > Wee (see section 1.3.7 and 1.3.8.5). In this case Fp = eEp > Fmagn = ev x B, evidencing a lower B limit for confinement (see figure 1.6.10, later). A similar conclusion was drawn by Golovanivsky in 1994 [54] but it is based on a more theoretical rationale including stochastic collective
oscillations inside a thermalized plasma.

1.4 ATOMIC PHYSICS BACKGROUND IN ECRIS PLASMA
.•
.)
.J
1.4.1 Collision times and characteristic plasma times
.J
Collision processes are conveniently described by the concept of a cross-section which is the usual way to express the collision probability. However, cross-sections are not practical for immediate use in ECRIS. The most useful concepts related to the cross-section are the mean free path and the collision frequency. If the collision cross-section is a, the incoming particle travels a distance A
1.4.1.1 Collision parameters.
.:'
.,) .;)
,l'
'n
= Im!p = na
(ern)
(104.1)
known as the mean free path, before it suffers a collision. travel one mean free path in a time the collision frequency
Te
-I
If the incoming
Te.
The reciprocal of this time interval is
Ve
= nav.
( 1.4.2)
,J
In most cases the cross-section is a function of the velocity v, and the velocities of the target particles are thermally distributed. The effective collision rate per particle is the average value of the product of cross-section and velocity, i.e. (1.4.3)
\1 i,if
,J
For many incident particles, the total rate of events per cubic centimetre, per second is
,
,I
II,
;ii I'
iii
ATOMIC PHYSICS BACKGROUND IN ECRIS PLASMA

au
cm2
61
<Ji
au
A\ (Jrec
cm3 S-1
Eeau Figure 1.4.1 The general behaviour of ionization cross-section 0'; (crrr') } ionizati?n ~atecoeffic~ents, (ern' S-I) recombination coefficient CXm (em' S-:-I)
versus electron energy Ee.
R contains (au) also called the ionization rate coefficient (figure 1.4.1).
when a =
CTion
(1A-A)
The variations of S, with the electron energy are much softer than those of CTion and over a small range of electron energy. S, is often considered as a q:asiconstant value. Comparisons of S, with Clree which is the recombination ratio coefficient, gives an immediate idea of the dominant process (figure lA_1 _ 1.4_1.2 The hierarchy of times in ECRIS [82J. The collision frequency v is the most convenient collision parameter in ECRIS because its value CZ1. be compared to typical plasma frequencies such as the cyclotron frequency We, the Spitzer frequencies Vei, Vee and Vii the plasma frequency wP' the b.cnce frequency Wb in a mirror system and all kinds of wave frequency W introzuced into or self-generated in the medium. In this chapter, among the collision frequencies, we will consider those due to the presence of neutral particles inside the plasma, namely the ionizznon, recombination and momentum transfer frequencies. To evaluate their respective importance let us son out the most important frequency scalings=-yielccg a hierarchy of times in ECRIS. We have already seen in section 1.3 that wren
the EM waves do not propagate normally in the plasma, which is then te__ed n overdense. For particle collisions to have the time to occur, the lifetime of the pa-.:de in the plasma must exceed the time between two collisions
.. ~
62
"t
"~
When we consider other plasma times, for instance the passage of an electron
through the ECR which has a duration of
TJ:('R
and we find that
",
"t
,,~
then the resonance will be impeded by collisions. If the opposite occurs the ECR process is collisionless. If we consider the electron helical path in a magnetic mirror and we have collisions between two mirror reflections (bounce time) i.e.
[Tbounce
,>
,)
":~
"
= Vbounce < Vcollis;on

the helical path-will be disturbed by
In general,
,)
then the mirror confinement-and
,J
,)
collisions; if the opposite occurs the mirroring is quasi-collisionless. whatever the type of collisions, when we have
Veollision
>
..~
Wee
.)
',,~
the behaviour of an ECR plasma will be strongly modified and requires a detailed analysis.
"
1.4.2 Binary electron collision frequencies and free paths

In an experimental plasma study, the various binary collision processes among plasma constituents must always be taken into consideration so that their true effects on plasma behaviour can be determined. The binary collisions deal with encounters of one electron with one target particle (atom, ion). The large number of possible collision processes and their varying degrees of importance under different plasma conditions make it quite difficult to account comprehensively for all collisions. As many experimental and/or theoretical cross-sections are given with large error margins, we preferred to consider averaged values coming from different sources. Thus the main purpose is to assemble credible data for some of the more important electron collision processes in ECR experiments and to present these data in a form that facilitates studies. That is to say, the data are plotted as 'collision frequency' and 'coiiision mean free path' versus electron energy. This presentation permits the direct comparison of the collision frequency with other characteristic frequencies of the ECR plasma. On the other hand direct comparison of the mean free path with the other characteristic dimensions, Debye length, Larmor radius, plasma dimensions, wavelengths of applied signals and interaction region dimensions etc, is desirable. We present only collision data for electrons which may have an important influence on the basic behaviour of the plasma. Therefore, total, elastic, momentum transfer, ionization and excitation processes are considered for a wide range of electron energy, in commonly used support gases sud. as He, H2, N2 and Ar. These examples yield typical orders of magnitude, which allow one
.) ,)
.)
,,~ ,) ,) :}
,,) ,) ,.,[1)
:.11
,
.•
,)
,)
"
I
II il I ,Ii I
-a
;
,J!
,,~
,)
...
..)~
ATOMIC PHYSICS BACKGROUND
IN ECRIS PLASMA
63
to consider more specific gases or vapours by comparing their specific crosssection with those of the common support gases.
1.4.3 Total electron collision frequency (figure 1.4.2) [82]-[85]

At the beginning, let us consider a low-power plasma with many neutral atoms and rather low percentage ionization. We start our presentation of electron collisions with the total collision frequency which is the total number of collisions per second per electron for all encounters involving either a strong change in direction or energy, or both, for the incident electrons. That is to say, if a beam of electrons with uniform velocity v (em S-I) covers a distance x (ern), the beam density n (cm ") at the velocity v changes from the original density no to n = no exp( -NO' x)
= no
exp( - PCPox) exp( -vc) exp( -x/A)
= no
= no
where each identity represents an alternative way of characterizing the collision probability, with N (cm=') the density of target particle (3.54 x 1016 at 1 Torr and 273 K), 0' (em") the total collision cross-section, Pc (cm ") the total probability of collision at 1 Torr and 273 K, v (S-I) the total collision frequency, A(ern) the total collision mean free path, t (s) = x/lvl = x l» and Po the pressure (Torr) 273/T K. The relations between the parameters of the above equations are
0'
= 2.83
10-7 P;
=
=
vpo
1.0/ PCPo'
Figure 1.4.2 gives the total electron-neutral collision frequency versus electron energy for He, H2, N2 and Ar. They are derived from well known cross-section data. Figures 1.4.3 and 1.4.4 show some specific total cross-sections for helium, hydrogen, alkali metal vapours and rare gases, allowing comparisons. It should be noted that, far from obeying the 'billiard ball theory', O'd varies a great deal with incident energy except in the case of He and H2 at very low energies. Nevertheless, atoms from the same column in the periodic table behave, on the whole, in the same manner. Thus the curves are very similar for the alkali metals on the one hand, and for the rare gases on the other. However, the effective cross-section of the alkali metals is a great deal higher than that of rare gases (because their electron clouds are more diffuse). Coming back to the usual support gases, some mean free paths are illustrated in figure 1.4.5. All the data are given for gas pressure equal to 1 Torr and can be immediately reconverted for lower pressures.
~I\ !If
64
a: a:
SPECIFIC ELEMENTS OF PLASMA PHYSICS o ~

CJ)
J
.~
V [101 -----------r----·----·--------1------·---------r------l
o
~
<,
UJ
0« -z ;>~O
......1-
ciw
1010 ---_------
---'---rT"
II"\. I.V
- =====~,
f--
.,
.,~
"~
oct
wz
:::>0
OUJ UJ:::> a:..J
zO
H2/
l1.<
...Ja:
o )(
z>
!act
..JO
He/
108
,v v A -: ~/ V ""~ N2 VV
I
,'... ...
,,~
r-r-
---
.... ... '

I"-,
"t
"~
0t:.
Z
oU
UJ W ..J
...J
g:.J
0. II
N\J
~:J
"~
~,
"t
~t
:~
1.0 10 ELECTRON ENERGY (eV) 100
1000
"t
)
,,~
Figure 1.4.2 The total electron-neutral collision frequency versus electron energy for
H2, He,
N2
and Ar at 1 Torr [82].
:~
The general behaviour demonstrated in figure 1.4.2 is for collisions to increase with electron energy at lower energies and then to more or less saturate at electron energies above about 10 eV. However, Ar exhibits 'resonance' phenomena in the lower-energy range which are ascribable to the Ramsauer effect [85]. Globally we observe that for usual ECRIS gas pressures (10-810-2 Torr) the mean free path AT is not only very long with respect to all the other characteristic lengths of the plasma (Larrnor radius, Debye length, wavelength, etc) but also even longer than the plasma dimensions. Thus one has to invoke the effects of the magnetic field resulting in spiralling electron motion with mirroring i.e. rebouncing electrons, which allows the eiectron to move over long distances before colliding with another particle.
,.
"~
"~
,,~
.,
,It
.~
1.4.4 Elastic electron-neutral collisions [82]-[84]

J .4.4. J Angular scattering. We now single out elastic electron-neutral collisions from the total collisions. An elastic collision is defined as one which does not alter the internal energy state of the colliding particles (projectiles and target). In the case of an elastic electron-neutral collision, there is only a slight transfer of energy 6. E from the kinetic motion of the electron (mass m,
I·
"
I,
II· I ~
iii
il
Ii
II
I'
1,1
"I'
11111
III
,ii
'~
cr'o'
600
t 10- ~ cm
6
Cs
65
500
400
300
200
100
VEe
0
~O
VeV
Figure 1.4.3 Total effective cross-sections versus the square root of the electron energy [87] for alkali metal vapours.
alOI
40
10-16 cm2
30
20
10
_j__..L--L--L--L_..J...___7
10
Figure 1.4.4 Total effective cross-sections versus the square rool of the electron energy [87] for helium, hydrogen and rare gases.
66
~(),(cm)
10 ELECTRON ENERGY
100
-7
Ee (eV)
Figure 1.4.5 Total electron-neutral mean free path versus electron energy for He, H2, N2 and Ar at p = 1 Torr [82]. velocity v) to that of the neutral particle (mass M, velocity ~ 0) 2m am /).E ~ - -(1/2mv M 0'0
2
( 1.4.5)
where 0'0 is the elastic electron-neutral cross-section and am is the momentum transfer cross-section for the electron-neutral collisions. Since inelastic electron-neutral collisions are only important for electron energies above several electronvolts (see section 1.4.5), elastic electron-neutral collision frequencies for He, H2, N2 and Ar are identical to the total frequencies of figure 1.4.2 in the low-energy range. The equivalence is obscured for higher electron energies by the influence of the inelastic processes. It appears that elastic coiiisions are approximately equal to the total values for electron energies up to about 50 eV. At higher energies, the ionization collisions Vi become dominant (see figure 1.4.8). The most important consequence of an elastic collision is the possible deflection of the electron. This effect is described in terms of a 'differential' cross-section 10(8) dQ = 10sin 8 d8 d¢ for elastic scattering into the solid angle dQ (using a polar coordinate system with the Z-axis in the initial direction of the electron and assuming no dependence of the scattered electron distribution on the azimuth angle ¢). 10(8) is proportional to the probability of an electron being scattered by an elastic collision into the gi yen element of solid angle dQ;
I,
I, r: I II; I r
Iii
II
I ~I I
II
'I'
1,1
litl
'
,II
Ili<
IN ECRIS PLASMA
67
\
3
10 eV
480eV
610 eV
O~--~~~--~~~~~~~
120° 90° 60° 30°
800eV 30° 60° 90° 120°
> e scattering angle
0°
Figure 1.4.6 Forward scattering probability a(9) for electron-molecule different values for the kinetic energy of the electron [83].
collisions with
i.e. it is a measure of the intensity of scattered electrons as a function of the scattering angle. The differential cross-section is related to 0'0 by
0'0
ff
TC
2rc
10(B) sin B dB d</>.
(1.4.6)
Angular scattering distributions were thoroughly discussed by Massey and Burhop [83] and these authors find that, aside from the diffraction maxima and minima, 10 is usually only slightly dependent on B for electron energies below a few electronvolts but they note that it becomes more and more peaked with increasing electron energy in the higher-energy range. In this latter case the elastic collisions do not strongly scatter the electrons (figure 1.4.6). Finally, for energies less than 1 eV, the de Broglie wavelength is much larger than the target. It is assumed that scattering is then isotropic and independent of the energy (a independent of </». As for higher energies, the wavelength is small in comparison with the dimensions of the target, which tends to give a well defined shadow; only very small angles of diffraction occur.
1.4.4.2 Electron-neutral momentum transfer and electron-ion momentum transfer f82/. The frequency of pure e-N momentum transfer (at low electron
energy) is shown in figure 1A.7. When the incident electron suffers an angular deviation in an elastic collision with a neutral particle, it loses a fraction (1 - cos B) of its forward linear momentum; the momentum transfer can be described statistically in terms of a momentum transfer cross-section
68
, ~,
"
"~
"~
~~
"'~
:)
,,~
Figure 1.4.7 Electron-neutral momentum transfer frequency versus electron energy [82] (p = 1 Torr).
~,
:,~
~~
Cfm
ff
o0
x 211'
(1 - cos 0)/00 sinO dOd4>= (1 -
cosO)Cfo).
(1.4})
~I :D
represents the cross-section for a hypothetical collision in which the total forward momentum of the electron is lost, i.e. Vm = N am V is the number of mv units of forward momentum lost per second per electron and is denoted as the momentum transfer collision frequency. This characterization is useful in the study of diffusion in plasmas with low percentage ionization where the e-N collisions are frequent. Cfm can be calculated directly from (1.4.7), beam measurements of IoCB) being used. As discussed earlier, 10 is essentially isotropic for lower electron energies, in which case equation (1.4.7) gives Cfm ~ CfO (diffraction maxima and minima have little effect on the average of (1.4.7). At higher electron energies, Cfm should fall further and further below CfO with increasing energy due to the peaking of 10 about 0 = O. In any case the electron-neutral momentum transfer collision at low electron energies plays a similar role in the weakly ionized plasma as the 90° electron-ion ."nit7PT c(ll1i(;i()n~ in the stronzlv ionized nlasma. In low-electron-temperature plasma (T, < 10 eV), working at relatively high pressure (c- 10-3 Torr), provided that fluctuating fields are negligible, the e-N collisions are responsible for randomization processes. In figure 1.4.8, we compare e-i Spitzer collisions and e-N momentum transfer collisions, at different hydrogen pressures for a relatively cold electron plasma, with density 108 < n < 1014• We see that above T, '" 1 eV the Spitzer frequencies can become smaller than VeN for gas pressures greater than 10-5 Torr emphasizing the important role played by e-N collisions in weakly ionized plasma.
Cfm
1 _ • ~ _
~D
':1
,~
.,
~
.~ .,
.~
~
, .,
I
•
!)
.~
1.4.5
Binary inelastic electron collisions [82], [86], [87]
When an electron of sufficient energy interacts with a target atom, molecule or ion, a significant amount of the electron's kinetic energy can be transferred
! I I
.
e
0
:>
:0
~)
II
;
II ilti
II I ~ .
o
'.'~
'I
iii
.11
Iii
IN ECRIS PLASMA
p0= TORR 1 ... -
69
----l
j
1 10-1 TORR / ,----~----~~~ I
'
v;\ - ;o~roAR ~
, ,....-r-!
,
I
.JO-1 TORR --.._
CJ
I
'-j
I
r--__
--
10-2
-_
10-3
100
101
102
ENERGY FOR V ELECTRON { TEMPERATURE FOR
_}-. <V> V
(eV)
Figure 1.4.8 Electron-ion momentum transfer frequency Vei versus electron energy or temperature for singly charged ions with electron density (108n « 1014 cm") as a parameter. VM, electron-neutral momentum transfer, and Vion, ionization frequency, are given for comparison (gas H2) [82]. to the internal motion of the target particle. This type of collision is termed inelastic and produces some form of excitation or ionization. Two inelastic collision processes are presented here:
(i)
{1I)
total ionization from the ground state and totai excitation from the ground state.
These cases include the contributions from all levels of ionization and excitation respectively for electrons moving in a ground-state gas. The angular dependence of inelastically scattered electrons is strikingly similar to that of the elastic case, the most important resemblance being that an electron undergoes an ever decreasing deflection with increasing electron energy. For ECRIS plasma with energetic electrons this is an important point. It means that during the ECR acceleration and in the mirroring trajectories the electron will not be strongly scattered either by elastic or by inelastic collisions with the neutral atoms. Thus only the cold population of the ECR plasma will be randomized by these collisions.
1.4.5.1 Angular scattering.
70
J .4.5.2
"t
,~
Reaction threshold /87/.
All these inelastic collisions, characterized
hy a variation ~U in the internal energy, can only occur if an amount
~u of
,,~
the kinetic energy of the two particles involved can be converted into internal energy. This is called the reaction threshold. Now there is an upper limit to the
kinetic energy, ~ W, available for such a conversion. In effect, since the energy of the centre of mass must remain constant, the sum of the separate 'relative' kinetic energies can be transformed into internal energy. It can easily be seen that this sum is equal to the kinetic energy of the particle with reduced mass in the corresponding system of coordinates:
,.~
"~
",
"
"
,j
,,~
,,)
where v is the relative velocity of the two particles. Thus, when the two particles have widely different masses and the lighter of them is at rest, the energy available for conversion to internal energy is only a fraction (~ light masslheavy mass) of the kinetic energy of the incident particle; the rest is attached to the centre of mass which is displaced with almost the same velocity as the heavy particle. If the two particles have similar masses, the available kinetic energy reaches half the initial energy of the particles involved. Diagrams showing effective cross-sections generally have kinetic energy E (or velocity) of one particle as abscissa, the other being assumed to be initially
For E < Es (E; is the threshold energy), the inelastic collision in question cannot take place. E, ~ ~u only when the incident particle is a great at rest.
.~
"
",
"~
.,~
,,~
"t
r
deal lighter than the other. In general:
"~
(1.4.9)
,~
,,~
if
m2
is stationary. This is the case for inelastic collisions involving electronencounters.
,,~
neutral and electron-ion
1.4.5.3
Ionization of atoms by electrons.
The basic reaction equation is

(1.4.10)
If the target is at rest and the incident electron has a kinetic energy E, the
reaction can only take place if

(l.4.11)
~,
e Vi (lP) being the ionization potential.
I'
II'
I· j·1
1,1,1 f
'III!
II
Ii.
'I
1·1
iltl
'I I
71
"
", ",
-: a
o Ia:
.
---~.-~~
.......'
"
Ar N2'
.... .:
~.
'"
.......-:-~-
--
~~.- -...
--.-.--.- ..".---~.-.--
".
··· .... H2
".
..•.. . ... BETHE BORN THEORY
".
". ".
He·····...
L....1-_...J..._
---'-- __
..
---'--
____.J
_j
103 ELECTRON ENERGY (eV)
Figure 1.4.9 Ionization frequencies versus electron energy [82]. As can be seen from figure 1.4.9, the effective ionizing frequency increases rapidly once the threshold energy has been passed, as the conservation conditions are easily satisfied due to the ejected electron. The maximum values of a, are of the order of 10-16 ern? and correspond, in general, to electron energies of the order of 100 eV (about 20 eV for alkali metals). After the collision one slow electron and one fast electron, which carries off the major part of the kinetic energy not used in the reaction, are generally present. The slow electrons have an approximately uniform angular distribution while the fast electrons tend to conserve the direction of the initial electron (see secuon i,4.j.1). selections of total ionization frequencies Vi versus electron energy are shown in figure 1.4.9. Experimental evaluations found by different authors are combined to give Vi for electron energies up to 2 X 104 eV in He, H2, N2 and Ar. Some of the experimental results have been extended to higher energies through the use of the Bethe-Born theoretical expression:
1.4.5.4
OUf
1 q.I ex: -£ In E
or
Vi
q;
1
£1/2
In
£.
(1.4.12)
The ionization frequencies of figure 1.4.9 are total values which include the production rates of all possible charged species. In a unique collision singly
72
I
l
I I
/ /
/ _____ L__ / / / /
/BETHE BORN // EXTENSION /

/
100~~--~~+---~~~4-------~:/--+---------~
102
103 ELECTRON ENERGY (eV)
Figure 1.4.10 Ionization mean free path versus electron energy [82]. charged ions are in general produced much more often than multiply chirged ions. Therefore the singly charged ion is the basic ingredient of the BCRIS plasma. Mean free paths are shown in figure 1.4.10. The effect of a Maxwellian distribution on Vi is shown in figure 1.4.11. For that we have integrated the values of Vi for Ar and H2 over a Maxwellian electron distribution to give the curves of Vi
ni(Te)
f
o
00
!(V)Vi(V)
41fV2 dv.
0.1.13)
demonstrates the importance of ionizing collisions for Maxwellian phsrnas even when the electron temperature (eV) is well below the ionization thresnold.
Vi
.4.5.5
Total excitation frequencies.
The basic reaction is

(].1.14)
if the target is at rest and the incident electron has a kinetic energy E. the reaction can only occur if
(l.!..15)
However, the probability of excitation remains very low close to the threshold energy E.\., since it is difficult to fulfil the condition of conservation of argular 'ltJmentum.
Ii,
IIIII
I,
Iii
II
I'
'ii'
II"
1'1
'I I
:1
IN ECRIS PLASMA
73
a: a:
o I-
,/ Vi ,/ ,/ , // ,Vi
'- ,
101
102 ENERGY -ELECTRON { TEMPERATURE -::-_ _ (eV)
103
Figure 1.4.11 Vi, ionization collision frequencies in H2 and Ar versus electron energy; Vi, ionization collision frequencies in H2 and Ar versus electron temperature [82]. Electron excitation of ground state atoms and molecules is a very difficult phenomenon to analyse due to the multiplicity of possible final excited states, secondary transitions, the presence of metastable states and the occurrence of other inelastic processes. Thus, it is not surprising to find that the experimental data on excitation are rather limited. However, we are interested in the total excitation frequency over an electron energy range large enough to permit a meaningful evaluation of the relative importance of excitation in practical FeR plasma experiments. Figure 1.4.12 presents a compilation of excitation frequencies versus electron energy for He. We have chosen to present excitation measurements for He since they give one a feeling for the orders of magnitude of the phenomena [99]. The combined excitation curves are compared with Vtot and Vi in figure 1.4.12. Elastic collisions obtained by subtracting Vi and Vexc from Vt are also presented. Vi and Vexc are found to be quite important in the medium-electron-energy range where elastic collisions decrease substantially. Even with the uncertainty in the absolute values of VI and Vexc, figure 1.4.12 shows that Vi is dominant but suggests that "exc is at least as important as VeN for electron energies above about 200 eV . This suggestion indicates the limits of influence of excitation collisions. They do not playa leading role in the plasma behaviour of ECRIS but their presence
74
v~
!
1010 0 f~ ~ ~
109 a:: a::
I
-
>
>o
w :::> 0 w
Z
a::
u,
108
Vexc
Vi
10 ELECTRON ENERGY leV)
100
Figure 1.4.12 Comparisons of collision frequencies in helium: and Vexcitat;on against electron energy [82].
VlOta/. Vt'last;c.
V;olli:';:tln
explains some RF power absorption and the presence of optical radiations issuing from ECR plasma.
1.4.6 Momentum transfer collisions and runaway electrons [82], [88]

In addition to the binary collisions, one has to consider the collective long-rang! Spitzer collisions which are expressed by (1 • 4 . ,._' 1 '; where In A ,...., 5; n, is the electron density (cm-3) and Te is in electronvolts: .: 1 is the average ion charge (see section 1.3). We are now in a position to expand our discussion to the combined effecs of all the collision processes treated herein. In particular, we shall consider tw: important points: (i) momentum transfer due to all electron collisions and (ii) ce relative importance of electron collisions with respect to the other characteris;c frequencies of a plasma. The effective momentum transfer collision frequency Vm for a plasma can :t! deduced from the presentation of the combined binary collision frequencies an: the electron-ion Spitzer frequencies versus electron energy as illustrated for r:.: in figure 1.4.8. When electron energies are below a few tens of electronvoin. the momentum transfer from the forward motion of the electrons is entire. determined by electron-neutral and electron-ion collisions.
,.J
I
I ii
I
II
I
i4
II
1·11'
I~
II
·1 ~I
'I
II
"I'
"i
I III
IN ECRIS PLASMA
75
The momentum transfer collision frequency is equated simply to the sum V",+Vei (for a monokinetic electron beam) or (vm)+{Vei) (Maxwellian electrons) evaluated by inserting measured values of electron (or ion) density, neutral pressure and electron energy or temperature directly into the appropriate figures. At higher electron energies, we must consider the momentum transfer ascribable to both elastic and inelastic collisions. Inelastic collision frequencies become comparable to the total collision frequencies at higher electron energies, signalling a marked decrease in elastic collisions. In addition, with increasing electron energy, the 'forward scattering' diminishes the momentum transfer collision frequency relative to both elastic and inelastic collision frequencies. This is true even though the ionizing collision contributes to the momentum loss of a fast electron without necessarily producing an angular deflection, because this loss is small with respect to momentum loss at higher electron energies. Therefore, we are led to conclude that the effective electron momentum transfer frequency will decrease with electron energy at the higher energies and that 'electron runaway' can occur to some extent for all ECR plasma conditions under which fast electrons are present [88]. The consequence of electron runaway is that, without the strong friction due to the momentum transfer collisions, a family of fast electrons can be decoupled from the bulk of the 'colder Maxwellian electrons of the plasma. This situation leads then to a two-electron-population plasma which in tum can generate the well known beam/plasma instability (see section 3.2.7.7). As, due to the ECR, the fast electrons are permanently produced, a permanent state of instability can be maintained. Note that for a heavy-ion plasma under good confinement conditions, the fast-electron family can lose some of its energy in multiply charged ion production. These inelastic binary collisions between electrons and ions can take away a part of the electron energy in a single encounter. However, as we will see later, their collision probabilities are not as high as those presented in this chapter and thus they do not change the basic plasma collisions. Finally, we must accept the idea of the presence of decoupled high-energy, runaway electrons in ECRIS plasma as soon as the gas pressure is lower than .. ,.... "1..... ~-. _. lU ~ rorr. i neir number increases wun uecreasing pressure ano increasmg Kl" power. They are the generators of x-rays in the source, and as they do not feel the effects of collisions their diffusion is weak and their lifetimes arc long. Their contribution to the plasma equilibrium is not always clearly established. An important question therefore remains: are the fast electrons in the tail of a Maxwellian type velocity distribution, in which case they are already the byproduct of the beam plasma instability, or are they really a separated group of electrons? The last suggestion is the most probable (see figure 1.2.3).
"_!' t·· ~.~
I
I
l:
.~
II
1.4.7
Quasi-colllsionless confinement and anisotropy [82]
To demonstrate the utility of the figures of Vc and Ac, let us now apply them to a simple magnetic bottle ECRIS plasma obtained at 10 GHz and ascertain the
76
relative importance of collisions. The plasma is produced through the cyclotron interaction of electromagnetic waves with electrons in a static magnetic mirror. The plasma electrons are given a transverse energy and they then spiral along the gradients of the magnetic bottle. Typical operating conditions for less highly charged ion ECRIS are for instance an RF power of '" 100 W at 10 GHz in a hydrogen gas at pressure Po '" 10-3_10-4 Torr, and a static magnetic field Bo which varies in the active region of the plasma from '" 5 x 103 to '" 2 X 103 G; WRF and Wee corresponding to Bo are in the region of 1010 S-I. They are always higher than any collision frequency. Thus this plasma seems 'quasi-collision less ' . The cyclotron acceleration of electrons occurs over a small axial distance in the ECR zone where WRF :::: Wee' Downstream from this interaction region, the energized electrons are magnetically confined and can ionize the background gas. From figure 1.4.8, for H2, we find that Vi ~ 2x 106 s-I for Po ~ 10-3 Torr, corresponding to a mean free path of ionization of Ai ~ 300 ern (figure 1.4.10). Thus v;/(w/2Jr) = 10-4 and ionization should practically never occur during the ECR and have very little effect in the interaction region. However, due and only due to the magnetic confinement fields the spiralling path of the mirroring electron can exceed the ionization mean free path. Then only, we would expect a background plasma to accumulate, which, in fact, is the case. We can also say that, due to the magnetic confinement, the electron lifetime in the plasma becomes longer than (Vi) -I = t, '" 5 x 10-7 s. Thus ionization can occur and a plasma can be created. For lower RF frequencies and higher gas pressures the conditions are easier. But it then follows that for gas pressures X times lower than 10-3 Torr the electron confinement has to be X times longer. This illustrates the necessity of better confinement, and ultimately more sophisticated magnetic systems for very low-pressure ECRIS (such as the multiply charged ion sources) are required. Let us reconsider the previous plasma at p = 4 x 10-4 Torr, with the following typical characteristics at steady state: 1l '" l O!' cm-3 and T, '" 10 eV. According to figure 1.4.8, the momentum transfer frequency Vm ;S 106. As Vee = Wee 12]"( we find Vm 1 Vee < 10-"'. The electrons will gyrate around the magnetic field lines '" 1000 times before making a momentum transfer with an atom, and '" i0000 times before suffering a Spitzer coiiision. Thus during a few microseconds the electron conserves the energy acquired in the ECR, and subsequently its velocity remains perpendicular to the B field lines. If the electron lifetime is only a few microseconds the electron trajectories will not be randomized (as expected in a classical plasma) and a strong anisotropy (v1.lvll » 1) will be observed. According to figure 1.4.8. for decreasing gas pressures and increasing electron energy, the anisotropy will even be stronger and thermalization of the electrons will not be reached, and this is what one observes experimentally. However, in some cases one also observes a threshold above which the anisotropy disappears (see section 1.3.7). The resultant isotropy (v 1. '" vII) is accompanied by plasma noise and strong localized fluctuations
.~
I· II· I
II
~ I
ii,
IIIII
III
,I I
IN ECRIS PLASMA
77
(Ll V) in the plasma. Then 1.1 V. due to some nonlinear processes. replaces the momentum transfer collisions. and rapidly thermalizcs the electrons. Thus figure t .4.8 serves only as a minimum indication of real momentum transfer since nonlinear effects are neglected whereas one knows that anisotropic velocity distributions are conducive to instabilities. Only in very stable multimirror ECRIS should the momentum transfer collisions intervene in the plasma balance. In such a case the Spitzer collisions are always efficient for cold electrons. For warm electrons when v.rp « WC(' wp and WRF, the collisions do not impede the ECR process. However, when (v.rp]-I is smaller than the electron confinement time inside the magnetic mirror, these collisions cancel the confinement (as we will see in"section 1.6) since they break the invariance of the magnetic moment.
1.4.8 Inelastic ion collisions

We have seen the influence of various electron collisions on the plasma behaviour; for the ions only those binary collisions which lead to ion destructions seem important in the balance equations of the ECR plasma (ion-ion Spitzer collisions only intervene in transport equations). In ECRIS plasma the ions are cold (T, < 10 eV). In an encounter with even colder electrons (near the wall for instance) electron-ion recombinations are possible but inside the plasma they are not probable. However, the cold ions may attach some orbiting electrons of a neutral atom when the ions collide with neutrals. This charge transfer process has large cross-sections and its importance must be emphasized.
I
~
II
lon-electron collisions and recombination coefficient <Xc [89]. Ions and electrons are attracted to one another but the attractive force is not always sufficient to cause them to combine. As in the case of excitation and ionization, the probability of recombination depends on the ratio tiT where t is the duration of the interaction <that is alu. the radius of target particle divided by the velocity of test particle), and T is the mean time necessary for the reaction to take place, T obviously depends on the particular process of recombination. In fact, there are several such processes, each of which has a different value of a. The probability of recombination depends on the type of particle involved since, whatever the type, neither a nor T varies very much. On the other hand, a does vary considerably with the level of excitation of the atoms or molecules; a is clearly higher for excited particles. When describing this phenomenon, instead of using the concept of effective recombination cross-section a., that of the coefficient of recombination, a, is often applied, a being defined as the rate of decrease of positive (n+) and negative (n-) charges (1.4.17)
J .4.8. J
78 where

I
II
1l-
represents electrons.
As the values of
II
arc expressed in cm -J, ex
is expressed in cm' s-I like Sj the ionization coefficient. The lifetime of ions. which. by definition. is equal to the inverse of the collisional frequency will therefore be given by r = I /ct.n-. Comparing with the formula defining the collisional frequency. it can be seen that
,
,
11 ,~
f
o
00
v 0", (v f(v)
dv
= (u, O",(v)).
(1.4.18)
J
,J
Thus a :::::: va,(v) where v is the relative velocity of particles involved. Fortunately, a, the recombination coefficient, decreases rapidly as the relative velocities of the particles increase. For all recombination processes, a has maximum values (between 10-10 and 10-8 crrr' S-I) for electron temperatures less than 0.5 eV. Thus recombinations are not important for steady-state ECRIS plasma but only for the plasma breakdown and the afterglow when the electrons are cold. 1.4.8.2 Direct ion-electron recombination types. We shall briefly consider four types of recombination with increasing probability. Radiative recombination. e-+A+~A+hv ~ A* +hv.
, II>
....
(1.4.19)
In the first reaction a cold electron is captured into the ground level and a photon is emitted. The electron may also be captured into an excited state. Recombination with double excitation.
-----*
A.
+ h l' + h v'.
(1.4.20)
During the first stage, the incident electron is drawn into an excited state of the atom and the energy available raises another electron of the same atom to another excited level (cf Auger effect or autoexcitation). After a certain time has elapsed, one of these electrons is ejected leaving the ion in its initial state (first possibility), or the two electrons fall to the ground state emitting two photons (second possibility). Figure 1.4.13 shows plausible values of radiative recombination rates a, for aluminium ions with different charge states from Z = 1 to Z = 13, compared with corresponding ionization coefficients Si. Recombination in the presence of a third particle. (1.4.21)
1I
~.
I ilill
1'1,1
II
I"
,I
;i
79
Si
OCrcm3 s-l
Figure 1.4.13 Successive ionization rate coefficient S; and radiative recomnaation

coefficient am for aluminium from Z+ = 1 to Z+ = 13 versus electron temperature
(T~).
cm3 S-l
OC
I
',,<
I
10 100
1000
(<>crec +<>C(ec) rad dielectr
(OCrec)
rad
Figure 1.4.14 collisions.
Recombination coefficients in oxygen
r 1061.
The effect of dieectronic
80
SPECIFIC ELEMENTS OF PLASMA PHYSICS exchange recombinations
j,'
't
1.4.8.3, [oil-neutral collisions-charge
/891, /90/.
J
If an electron encounters a particle X in the neighbourhood of an ion A 1, it may communicate its energy to this particle and slow down sufficiently to recombine with A +. Instead of being emitted in the form of photons the recombination energy. E, + E;. of this reaction is used to accelerate the third particle X. The probability of this occurring obviously depends on the nature and density of the X particles. The most considered process is dielectronic recombination (with X being a second electron). The corresponding recombination coefficient
IS
ad. Figure 1.4.14 shows a,
+ ad against
(Te) for oxygen.
(1.4.22)
A captures an electron and B loses one. Thus, in general An+
III
,~
+ Bm+
= A (n-I)+
+ B(m+I)+.
(1.4.23)
"~
Here, A and B can be either atoms or molecules. Energy balance. Taking the above reaction as an example, the energy balance equation may be written as follows: total initial kinetic energy
+ ionization
energy of A energy of B.
= total
final kinetic energy
+ ionization
(1.4.24 )
In general, the ionization energies (E;) differ little; as a result, the total kinetic energy varies very little in the course of the reaction. Furthermore, there is little exchange of impulse in the laboratory coordinate system, each particle more or less conserving its initial velocity. The condition defines a threshold energy
(1.4.25) When (E,) B > (E,) A, the reaction can only take place above a certain minimum value of the relative velocity of the two particles. This is the case, for example, with H+ in molecular hydrogen H2 (figure 1.4.15). The effective cross-section for neutralization (capture of an electron), designated by aIO or a+o. thus increases to a maximum. which is usually fairly high (10-15 ern? or more). and then decreases as the relative velocity of the particles increases. This type of charge exchange is particularly important for accelerated ion beams in a poor vacuum but it is not important inside an ECR plasma since the ions there are cold.
")'.,
"
,I
II I
1,1
II
I'
.Ii
i.1
.....
IN ECRIS PLASMA
81
(em/sec)
Figure 1.4.15 Non-resonant charge exchange cross-sections for H+ + H2 (experimental) (ions + molecules) versus ion velocity; resonant charge exchange cross-sections (theoretical) for He+ + He, 0+ + 0, Ne+ + Ne, Ar" + Ar, H+ + H, K+ + K (ions + atoms) versus ion velocity [90] (ao = 5.3 x 10-9 ern). In ECR plasma another type of charge exchange occurs when (1.4.26) In this case the energy threshold Es is zero, and under these conditions the cross-section 0"10 is very large even for an ion with kinetic energy Ek = O. This is true for instance for He+ in all gases (since the ionization potential of He is greater than that of other substances), and for any atomic ion in its atomic gas (E, = 0) which is the general case of ECRIS plasma. This charge exchange is called resonant charge exchange (figure 1.4.15) [90]. In this case, the variation in internal energy during the reaction is 6.E = Es. The time for this transition is T ~ hiEs. If a is the mean radius of the particles and v is their relative velocity, then a maximum probability for the reaction (that is, a maximum effective cross-section) would be expected at t l t = aEslvh ~ 1. (1.4.27)
82
Experiments show that the relative velocity corresponding to amlH does vary significantly with Ef; in particular. when E\ ~ (), this velocity is zero and then the colliding particles have all their time to make a charge transfer and the orbiting electron of the atom can jump to the ion and neutralize it. In conclusion. direct recombination of electrons and ions in ECRIS plasma is very rare. but the high cross-sections associated with charge transfer make this process one of the most important in ionized gases. One of the experimental difficulties arising due to this phenomenon is that, when heavy particles have been neutralized by charge transfer, they fly out to the walls since they can no longer be controlled by electric or magnetic fields. In the last example the loss is not dramatic because a cold A + ion is replaced by an even colder B+ ion and if A + and B+ are of the same gas the situation is not changed radically. However, when one considers a highly charged ion Az+ colliding with a neutral B, it loses its high charge and therefore the charge exchange becomes a crucial issue in multiply charged ECRIS where the highly charged ions are the ultimate goal. This point will be developed in more detail in the next section.
1.4.9
Collisions of multiply charged ions with electrons and atoms
1.4.9.1 Ionizing collisions. In ECRIS plasma, multiply charged ions (MI) are obtained with the help of two types of ionization. They can be produced by electron-ion collisions, where in each collision one electron is removed from the outer shells (the multiple-impact process, also called step by step ionization). In this case the impact energy has to be higher than the binding energy of the last electron to be removed. The other possibility is that a high-energy electron removes an electron from an inner shell, in which case a multiply charged ion can be produced by Auger processes and electron shake-off which are singleimpact processes. (IP) Ionization potentiaL. An atom is composed of a central nucleus of positive charge surrounded by a cloud of electrons. These electrons are grouped in consecutive shells called the K shell (two electrons), L shell (eight), M (16), N (32) and so on. Each shell is divided into subshells. An electron belonging to an external shell is only weakly bound to the positive nucleus, owing to the screening effect of the internal shells, and a low energy is sufficient to eject it into the continuum and to transform it into a 'free electron'. For such an electron the corresponding IP is of the order of a few electron volts, IP are higher for bound electrons belonging to internal shells, and increase progressively from external shells to the K-shell. In the same shell, IP values of the different electrons are close together; from one shell to another (the external shell being empty), discontinuities appear in the IP scale. In a collision an incoming electron transfers energy only to one bound electron, and if the transferred energy exceeds the binding energy W, ionization occurs. Wand the ionization potential (IP) = W [e are functions of the nuclear charge (eZA)
I,
III
1,1
II
'I"
,;1 I
I,
ATOMIC PHYSICS BACKGROUND J.P. eV
IN ECRIS PLASMA
83
50
100 Z
Figure 1.4.16 Ionization potentials of elements [91]. The nuclear charge is plotted on the X axis and the ionization potential is plotted on the Y axis. The numbers on the curves indicate the ion charge multiplicity. and of the quantum state of the electron. For example, the energy W of an electron of a hydrogen-like ion is eV ~ I3.6Z~. It varies within wide limits from 13.6 eV for hydrogen to 115 keV for uranium. The ionization potential for an outer electron is, according to Bohr's theory, (1.4.28)
,
!
where R is Rydberg's constant and a is the screening parameter of the nuclear field by the electron frame, which is a function of nand e, whose values are known. n, the principal quantum number (1, 2, 3, a positive integer), roughly defines the total energy of the electron and its mean distance from the nucleus. e is the azimuthal quantum number: 0, 1, 2, ... , n - e roughly defines the orbit eccentricity. Formula (1.4.28) allows the outer-electron ionization potentials of ali eiements or me periodic system to be calculated. A famous example of such a calculation is shown in figure 1.4.16. It was derived in the I960s by Carlson et al at Oak Ridge [91] and it gives the IP values of all elements, and charge states from 1 to 100. Discontinuities between successive shells of an atom can easily be seen as well as the discontinuities between rare gases and alkalis. We may also remark that an ionization in the K shell necessitates an energy of about 900 eV for Ne, 3.2 keY for Ar, 14.3 keY for Kr and 35 keY for Xe. 1.4.9.2 The partial ionization cross-section. The partial cross-sections characterize the different possible ionization channels. Let us consider collisions with z" -tirnes-ionized atoms A z+. In addition to the classical single ionization model another possible mechanism of single-ionization--excitation and subsequent decay of the auto-ionization state-proceeds in two steps,
R4
o
N
!>p
FINAL VACANCIES
5s 4d 4p
Xe20+
~M
(/l
...J ...J
3d
3p
2p
L
INITIAL VACANCY IN KSHELL
Figure 1.4.17 A conceptual Auger cascade leading to Xe20+ after ejection of a K-shell
electron ([92]). Al.+
+ e"
--+ (Al.+)*
+ e"
(Az+)* --+ A<z+l)+
+ e",
(1.4.29(a»
In this process, an incoming electron excites one of the inner electrons to a vacant level with higher n, leaving an unfilled 'hole' at a level with lower n. The decay of this strongly excited (auto-ionization) state is accompanied by ejection of one of the electrons. Another process yields globally a z-times-ionized atom in one collision with an electron
(1.4.29(b) )
Two such channels are known; Auger ionization and shake-off ionization. The former process is similar to excitation and decay of an autoionization state, except that, with Auger ionization, an inner electron is not transferred to one of the higher bound states but is detached. Excitation energy is removed by transition of one of the outer electrons to an inner vacancy. The energy that is liberated here is expended in detachment of one or more additional bound electrons. For instance in an Auger process one inner-shell vacancy produces two vacancies in the higher shell. Each new vacancy will in its turn be filled by transitions from still higher shells until all vacancies reach the outer shell. Such a vacancy cascade results in a multiply charged heavy ion [92]. A conceptual Auger cascade is shown in figure 1.4.17. An estimate of the contribution of the Auger process to stripping of argon ions by an electron beam with an energy of 20 keV made in [93] shows that this channel reduces the time required to attain an Arz+ by a factor of four to five. However, this case seems rather exceptional.
",ill,
'<.t'
.)
:)
'I"
iii
Ii,
IN ECRIS PLASMA
85
An additional helpful process is 'electron shake-off in which outer-shell electrons are excited to the continuum due to a sudden change in the central
potential as, for instance, caused by inner-shell vacancies [95]. Not much is known about the cross-section a, (Ee) for the production of such an ion by an electron with kinetic energy E, in a single-impact collisions. Some crosssections have been measured by Schram et at and van der Wiel and calculations have been performed by van der Wiel et at and van der Woude [94]. They show that the relative abundance of the highly charged ions decreases rapidly with increasing charge. A K-shell vacancy becomes less effective for producing multiply charged heavy ions if we go to higher z. Finally, the hope for efficient ionization rates with single electron impact vanished after 1972. A review on K-shell ionization cross-section is given in [95]. 1.4.9.3 Ionization cross-section.tl Csr for successive collisions. Suppose that we are able to maintain an ion in a plasma system for a long time r where it is bombarded continuously by electrons of energy E; of a few' hundred of electronvolts. The ion suffers several successive ionizing collisions, each of them ejecting one peripheral electron, A z+
'I
+ e"
-+ A (z+l)+
+ 2e-
(1.4.30)
a
!.
h
)
I
i
If
e
J,
,.
I
j fl
oe
y
It
;s
y h
11
!e
)f
n
I.
where z is the ion charge state. As a result of this process, the ion charge is increased by one after each collision. Thus, despite a weaker electron energy Ee, very high z may be obtained; all electrons having a binding energy less than Ee can be ejected. Early experiments performed by Redhead [96] have demonstrated the possibility of producing ions XelO+, CsIO+ and BaIO+ with an electron beam of only 250 eV energy. Later Donets et at gave accurate ICS for N, C and Ar around E, = 2.2 keY [97], [98]. It then became clear that the successive collision IeS is much greater than the single-collision ICS (see figure 1.4.18) and therefore successive collisions are the basic ingredient for multiply charged ion (MI) production. Many theoretical works have been devoted to the computation of cross-sections a (Ee) for successive ionization. Recently, experimental data on electron impact ionization cross-sections for atoms and ions ranging from hydrogen to uranium were compiled and given, in graphical form, as a function of the electron impact energy by Tawara and Kato [99]. Some selected theoretical data were also included. Thus hundreds of data are now available. A comprehensive review of the subject of electron impact ionization was published recently [100]. To handle the ionization processes more easily, a number of empirical formulas for calculating the ionization cross-sections have been proposed by different authors. In particular the formula derived by Lotz for removal of a single electron in successive ionizations is now widely used [102]. Younger and Mark in [IOOJhave reviewed empirical formulas of ionization cross-sections. Other convenient formulas are proposed by Muller et al [10 1).
86
0-1
•
i
;_
i
i
0-2 4-5
I
I I
I
I
!
5-6 0-3
i
I
7-8 8-9 110-11
6-7 9 10 1- 11 12 1--
"
r r
0·4
!
I
0-5 0-6 0-7
I,
E eKeV
(/
0,5 1
..... ..-
Figure 1.4.18 Single-impactionization cross-section auz compared to successive-impact ionization cross section Zz-I.z against electron energy (argon). It has recently become recognized that on increasing the ionic charge of the ions, other ionization channels tend to play a certain role. Among them, the excitation auto-ionization and resonant recombination double auto-ionization processes contribute significantly to the total ionization of heavy alkali-like and alkali -earth-like ions. In such cases the choice of the dominant ionization channel is difficult; otherwise the step by step ionization is overwhelming. i.4.9.4 The muitiply charged ion (MI) prouuction cricerion and the dominant ionization channel [103 J. As seen, each of the ionization channels (seen in 1 A () '" ~_..l A 93) can be ~ <u l characterizedU by' ~ ~ross - section uZ')i2'Ye)' \ ~u .I-.~ •• lHG _(x) ro. ~~ lHal I-.~ 1.'"t."."- £111U 1.'"t.. ~UVII transition of an ion from charge state ZI to charge state Z2 through a given channel requires, on average, a time (1.4.31) where Ve and n, are the electron velocity and density and x indicates the type of ionization channel. ECRIS electrons in the simplest case have a Maxwell velocity distribution, and, since the reaction cross-section is a function of Ve, to obtain the reaction time for a given channel, formula (1.4.31) should be averaged over the Maxwell distribution with electron temperature 'L, As a
.)
,,_J
,.~
"~ .... #
!i
.,
I'II
II.'
·'1
'I
III

ECR, R. Geller

Uploaded by

Copyright:

Available Formats

ECR, R. Geller

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

ECR, R. Geller

Uploaded by

Copyright:

Available Formats

·1

Electron Cyclotron Resonance Ion Sources and ECR Plasmas

Electron Cyclotron Resonance Ion Sources and ECR Plasmas

Institut des Sciences Nucleaires, Grenoble

Institute of Physics Publishing Bristol and Philadelphia

© lOP Publishing Ltd 1996

... ..• "1

1.4 1.5 1.6

200 STEADY 216

THE ECRIS PLASMA STATES-BREAKDOWN, STATE AND AFTERGLOW

ION CHARACTERISTICS AND ION PROCESSES IN ECRIS

311 311 325 331 346 351 357

,.,,1-."_"01 __ & .1... -", "-n"'.:"'!"!1 1 ~lIapl~1 VI 111':' 11I,;;al1':'~

_=L. ...... L._=_ll ........ __ ."' __ ....... •. UlIl,;;"Y \..UlIlCIlIPUl i:1J. Y

Grenoble, 1 April 1996

1.1 THE BREAKTHROUGH OF THE ECRIS AND ITS PLASMA BACKGROUND

SPECIFIC ELEMENTS OF PLASMA PHYSICS

1.1.1.2 Methods oj ion formation in the plasma reservoir oj ion sources.

THE BREAKTHROUGH OF THE ECRIS

1.1.1.3 The particular importance of specific ion source plasmas.

SPECIFIC ELEMENTS OF PLASMA PHYSICS

1.1.2 Positive ion source concepts [20]-[27]

THE BREAKTHROUGH OF THE ECRIS

field with u sas at a relatiyplv . -

SPECIFIC ELEMENTS OF PLASMA PHYSICS

Drift lube s ,., SC solenoid L He cooled panel--'

Electron beam collector--i

Arc plasma -Anode A

--Hexapole K Cathode Salenoids--··

SPECIFIC ELEMENTS OF PLASMA PHYSICS

THE BREAKTHROUGH OF THE ECRIS

SPECIFIC ELEMENTS OF PLASMA PHYSICS

Superdense Supe~ hot..1~Uer' ~

Figure 1.1.6 The CERN accelerator complex.

SPECIFIC ELEMENTS OF PLASMA PHYSICS

Ion Beam Spullcring Deposition with Ion Assistance

THE BREAKTHROUGH OF THE ECRIS

1.1.4 Applications of ECRIS in industrial ion processing (figure 1.1.7)

Conclusiccc of section 1.1.

SPECIFIC ELEMENTS OF PLASMA PHYSICS

magnetic field configurations are being evaluated. Other extraction geometries

1.2 PRELIMINARY ASPECTS OF PLASMA AT THE ELECTRON CYCLOTRON RESONANCE

Ignition of laboratory plasma [49]

1.2.2 Self-destruction of laboratory plasma

SPECIFIC ELEMENTS OF PLASMA PHYSICS

1.2.3 The need for plasma confinemcnt-magnctoplasma

PRELIMINARY ASPECTS OF PLASMA

1.2.4 Particle gyromotion in magnetoplasma

SPECIFIC ELEME~'TS OF PLASMA PHYSICS can be

Much useful insight into the confinement properties of magnetoplasmas

F = eVl.B = e(B x v).

PRELIMINARY ASPECfS OF PLASMA

This can be expressed conveniently as fei and

1.2.5 Global characteristics of plasmas at the electron cyclotron resonance

SPECIFIC ELEMENTS OF PLASMA PHYSICS

In contrast to other magnetoplasrnas, where the electrons are energized by

PRELIMINARY ASPECfS OF PLASMA

SPECIFIC ELEMENTS OF PLASMA PHYSICS

)3/2 exp(-mv /2KT)

PRELIMINARY ASPECT'S OF PLASMA

_=L. ...... L._=_ll ........ ."' ....... •. UlIl,;;"Y \..UlIlCIlIPUl i:1J. Y