Download NEGATIVE HYDROGEN ION BEAMS

REVIEW ARTICLE
NEGATIVE HYDROGEN ION BEAMS
A. J. T . HOLMES
AEA Tcchnology, Culham Laboratory, Abingdon, Onon OX14 3DB, U.K
(RmiurrlX July 1991 ; and in wLii.wd/brm 13 Septenrhcr 199 I )
Abstract-A description is presented of the developmeill of H- and D- sourccs for ncutrsl bcam heating
on thc next generation Tokamaks such as NET or ITER. The two basic types of such ion sourccs are
discussed, wherc either thc ion is formed i n the plasma by dissociative attachment or on the surface of a
negatively biased electron of low work function. Also described arc the acceleration techniques to form
high energy and power beams nccded four neutral beam injectors. Finally. a short discussion is presented
of the future develor”n1 of these sources.
I . INTRODUCTION
THE FUTURE SUCCESS of neutral beam heating for next step Tokamaks such as N E T
o r ITER and also for high energy neutral atom beams for SDI applications, depends
on the development of negative ion beams with a high brightness. In both areas it is
the ability to detach the extra electron from H - o r D- ions in a gas cell with a n
optimum conversion efficiency of 60% irrespective of beam energy which causes
negative ion beams to be of interest. However, the low binding energy of this electron
(0.75 eV for H-) which allows easy detachment also makes it ditfcult to form negative
ions initially and this prevented the early exploitation of negative ion beam systems.
By 1980, however, two methods of forming significant fluxes of negative ions in
practical ion sources had been developed and this has led to two radically different
techniques of accelerating and transporting a beam t o distant targets.
The two basic types of ion source depend on very different methods of producing
the ion. The “surface” source forms negative ions by double electron capture o r by
sputtering when a n H i ion strikes a negatively biased surface coated with a cesium
o r barium film. Roughly a fraction of 0. I o r less of the incoming positive flux is reemitted as negative ions but the transverse velocity o f these ions is large. In contrast,
in the “volume” source, the negative ions are formed by dissociative attachment
between vibrationally excited gas molecules and verycold electrons (Tc 1 eV) within
the plasma volume. However, the emitted current density of negative ions formcd in
this way from the plasma is much less than the positive ion current density and t h x e
is also a significant isotope effect. However, the energy of these ions is usually very
low (Ti< T,) leading to a low transverse velocity.
As a result of the different formation processes, the method of ion extraction from
the plasma is not the same for the two systems. In the surface source, the ions have
a n initial high energy (in excess of 100 eV) and a relatively high emittance while ions
formed in the volume source have an almost zero initial energy and a low emittance.
In the following sections the impact of this is discussed along with the eflect this has
on the overall design of the ion beam system for each type of source. This leads to
-
653
654
A . 1. T. HOLMES
the behaviour of the negative ion beam downstream when it is at high energy and
how the heam can he transported over significant distances.
2 . N E G A T I V E ION P R O D U C T I O N
The design of an ion beam accelerator is very strongly influenced by the initial
energy and flux of the ions as they are extracted from the plasma where thcy are
formed. In addition, the accelerator has to he able to handle the beam transverse
energy and space charge downstream as well a s a problem unique to negative ion
beam systems; the electron flux leaving the plasma. The radically different nature o f
the two basic source type has led to very different solutions to the ion extraction
problem so an initial description of the negative ion source is provided. These source
designs are based on H - ions formed in a hydrogen discharge a s this area dominates
the research but, in principal, other negative ions can he formed in this way.
2.1. The surfuce source
The first results o f the surface process in producing negative ions was first reported
by BELCHENKO
er al. (1973) when they added cesium to an existing source and
observed that the output increased from 5 to 20 mA. This source was subsequently
improved to yield 200 m A by operation at high prcssure ( - 30 Pa) with a n effective
current density in P X C ~ S Sof 3 ,A. cm-*. The high dischxye paver deozi? !imi:ei
operation to very short pulses of about 1 ms.
There has heen a considerable series of experiments (BELCHENKO
ef U / . , 1974;
EHLERS
and LEUNG,1980) to undertstand in detail the production process in these
sources, particularly the influence of cesium on the discharge. The dominant production process is now well understood. When a negatively biased surface in the
discharge (including the cathode) i s bombarded by hydrogen ions (or cesium ions) a t
energies in excess of 100 eV, significant reflection of thcsc ions occurs a s well as
desorbed hydrogen atom emission. If the work function is small and has a minimum
for a partial monolayer of cesium then some of these atoms can capture a n electron
from the Fermi level by tunnelling as the atoms leave the surface. If the atom velocity
is large (which depends on thc incoming ion energy) then the atom can retain the
extra electron despite the image forces which are set up and a negative ion is formed.
The transverse negative ion energy appears to be about 5 eV (MASSMAN
er U/., 1981).
Several versions of this source have now been developed. in particular the magnetron source of Belchenko et ul. has now attained an output of I I A of H - in small
pulses (BELCHENKO a n d DIMOV,1983) and a similar source has been built by K o V A R l K
er U / . (1982). However, these sources have not been used with an advanced acceleration
system so this paper will rocus on the surface source devcloped (EHLERSand LEUNG,
1980) a t Lawrence Berkeley Laboratory where considerable attention has been paid
to the issue of heam acceleration and transport.
This latter source is based on the plasma confinement by an array of magnetic
multipoles and a cross-section is shown in Fig. 1. A discharge is formed between the
tungsten filament cathodes and the outer casing which forms the anode. T h e negative
ions are formed on the molybdenum concave converter which is coated with cesium
from the injector jet a n d is biased to about -200 V relative to thc plasma. This results
in negative ions being accelerated to the samecncrgyas they cross thesheath separating
this converter from the plasma. The concave shape of the converter focuses these ions
Negative hydrogen ion beams
Filament cathode
Cathode
converter
Ptosmo probe-.
655
7
Y
I
Collimator ond
Gos
,0T
~
Ground electrode
watt [,"er
Anode Liner
Suppre~sorelectrode
Gradient electrode
Cesium injector
FE. ].-The
Beom -forming electrode
Lawrence Berkeley Laboratory stdace source ( K w m e/ ut., 1986).
to the extraction aperture where they enter a three-grid accelerator. The high kinetic
i-iieigy ofthi- i i e g a i i ~ions allows easy elecii-irn suppression by means or'ihe transverse
field created by the anodic bar magnets. An electrode a t the extraction orifice between
the plasma source anode and the first accelerator electrode is biased a few volts positive
relative to the source anode to collect the residual electrons which leak through the
field. This source has achieved a maximum output current of I .2S A at 80 keV for a
pulse of 30 s (KWANet al., 1986). The electron fraction appears to be 10-12% of the
total beam current.
Despite these successes, the use of cesium in a source of this type has proved
difficult. The major problems have been a lack of reliability because of breakdown
along insulators contaminated with cesium and also discharge control through having
too little o r too much cesium on the convertor surfaces. Recently a new idea has
emerged where the ccsium is replaced by barium (VANOs et al., 19x8). This idea has
now been extended to the construction of a true surface source (VAN Os er al., 1990)
which is very similar in design to the earlier cesiated version. A t present a peak current
of 140 mA of D- has been observed through a 7 c m diameter extraction orifice.
However, the transverse temperature of this beam is about 11 eV, which is higher
than t h e 5 eV seen in the cesiated version. This is probably caused by the very high
converter voltage bias, typically 300-500 V which also causes severe sputter erosion
of the barium surface.
2.2. The iiolumc source
T h e first results of significant negative ion production in the plasma volume were
first reported by BACALer d.(1978) who used a thin cylindrical Langmuir probe to
measurc their presence by the rcduction in the plasma electron signal. However, the
first true volume plasma sources did not appear for several years (HOLMESel al.,
1982; YORKeta!., 1984) until the production process was understood. This production
proccss is unlike that of positive ions in that the dominant formation channel takes
656
A. J . T. HllLMES
place in two Stages (HISKES,1987). The first step is the formation of vibrationally
excited molecules H,(u) via the radiative decay of the B or C excited molecule state:
efi,5,+H2(X'Z;i)= 0) + H:(B'X:C'n,)
+ H2(X'Z$1;j+er,,\,
The electron energy threshold for this process is around 13 eV and the cross-section
is maximum at about 25 eV.
Once the vibrational molecular state is formed it is long lived unless a second fast
electron collision occurs leading usually to dissociation or if it collides with the wall
and jumps to a lowcr vibrational quantum level. If a slow electron collides with this
vibrational molccule dissociative attachment occurs forming an H- ion via thc shortlived H; state
H,(o)+e,l,,+H;
-.H-+H.
WADEHRA
(1984) has calculated the cross-section for this process and has shown that
to IO-"' m2 as 1; changes from 0 to 6.It then saturates at
it increases from about
4x
m2 for H, up to the highest vibrational level. The electron energy for a
maximum rate coelfiicient is about 1 eV for all vibrational levels.
However, the low binding energy of H- (about 0.75 eV) means that the extra
electron is easily detached. There are several channels for this to occur:
e+HH'fW
+ H+2e
+H
fH
H + W +2Hte
H , + W +H,fH+e.
The first process has a very large cross-section for electron energies in excess of 3 eV.
To overcome the dominant destruction channel, the plasma source designs have
sought to divide the plasma chamber into two distinct regions; a driver region which
has a high fast electron density to maximize H2(1;) formation and a cold electron
region of about I eV whcre the negative ions are formed and fast electrons are
excluded. This is achieved by the introduction of the so-called "magnetic filter" field
as shown in Fig. 2. This can be created in several ways; either by rods carrying small
bar magnets passing through the plasma (LEUNGrtul., 1982) or by long range external
;?no& (Hn!.Mrs c! d,!9X3) as
fields crcz!pd by !he mu!!ipo!e !nagnc!s on the ~;mrrc
shown in Fig. 2. A detailed comparison of the two techniques has been made by
H A N A i ) A f i U!. (1990) who found that the latter is superior, probably because of the
reduction in plasma losses due to the absence of the rods.
The negative ion yield of a large source of this type (55 x 30 x 2 I cm) is shown in
Fig. 3 for deuterium operation. The tendency for the yield to saturilte at high discharge
power is common to all volume sources and ariscs from the non-linear nature of the
two step process. This can be seen more clearly by turning the production and loss of
H,(u) and H - into particle balance equations (GREEN
el al., 1987) :
Negativc hydrogen ion bcams
657
I
Filamenls
Bwm fwming
electrode
Hte-H+tete
FIG2.-Concrptual
diagr:im shoning "filtcr" field in magnetic multipole volume
SOU~CCS
where
ti, =
N*
Cas! electron density
= vibrational moleculc
n. = U : r t n m den&,,
,."
..
density
I."
,.....ll" ...
collision time
N = H, molecule density
S,, = averaged (over Icvels) Vibrational moleculc production rate
S,, = vibrational molecule dissociation rate by fast electron impact
S,,, = vibrational molecule dissociation rate by atomic H collision.
I* = wall
This equation can be simplified by usc of the cxpressions:
j + = an,NS,
nh = ljj+
where a is a c~&ciefi! dependen! on the gourre geometry, N i p the gas density and s,
is the ioniiation rate. As the production o f a t o m i c hydrogen and ionization to produce
protons are closely linked, nh only depends on the wall recombination rate through
and the current density,j+. Substitution gives
fl
This expression shows a saturation in N * for large valucs ofj+ with a n absolute upper
bound for very large sourccs (where T* becomes large) of S , / a ~ S , S , , .
We consider the following H - production and loss processes
e,l,>w+H2(u) H - + H
+
dissaciativc allachmcnl riitc = S,,,
H t H - -2HS-e
dclachmcnl riilc by alomic collisions = S,.
H++W
+ 2H
recamhinalion rate
e + H-
+
=
S,
H +2e
clearon detachment ratc = S,
The effect of this on the negative ion density, n _ (and,j_)can he seen in the appropriatc
Negative hydrogen ion h e m s
659
particle balance equation (GKEEN
el al., 1987) where it has been assumed that the
plasma is sufficiently cold to make electron detachment negligible:
Nn-S, +n+n_S+
fnhn_Sh- = n,N*SoA
(5)
(6)
n, = n- + n ,
where the symbols have their usual meanings
Rearrangement yields :
..
,
.
However, nh/n+ IS a constant From equation
(ji a n d ( n + - n - ) / n + is ais0 near unity
in value. Hence equation (7) can be written in the form after combining with equation
(3):
B C N DN
-I --_A +--+--+-+E.
i- i,
N
i:
.I+
An experimental indication of this scaling law is seen in Fig. 4. Usually D and E are
ignored as they are thought to be small. Figure 4 also shows thc relative proportion
p = 12 mTorr
I/j+km2 mA-'l
Rr;. 4.--Dlol o f l h c reciprocal of j . wilh reciprocal of I , showing the isolope eflecl and t h e
inlinilc dirchargc current limit, B/M+ E i n cqualion (8).
A. J . 1. H o . b r i m
660
of D to H for cqual discharge conditions. I t varies from about 80% Tor low intensity
discharges to below 50% for high power discharges.
Recent experiments have shown a significant enhancement of negativc ion yield
when ccsium is added to the discharge (LBIJNGP I d.1988; KOJIMA
(’I o.. 1990). The
cause o f this clfect is not yet well understood hut is clearly related t o thc cesium
coverage o f the plasmn grid (the unmagnetizcd surface of the discharge chamber
r,u :
n g thc accelerator). This suggests that a surface process is involved. possibly
electron c:ipturc hy sputtered H atoms proposed by Y . OKUML;KA
(private communication). Similar. hut smaller in scale. effects have been observed with the addition
of barium and aluminium coatinas on the discharge surfaces.
3 . SURF4CE SOURCE
ACCELEKATORS
The larae stii-hce source developed at 1.awrence Bcrkeky Laboratory (KWAN et
U / . . 1086) hiis been operated with a n 80 keV accclcrator a s shown in Fig. I . The hasic
concept is very similar to the positive ion systems apart from a reversal o r t h e elcclrode
potentials and the additional digital clectrodc located at the plasma grid t o suppress
electrons.
K W A NC I d.(1986) have examined this accelerator geometry with the aid of the
W O L F code and the output is shown in Fig. 5 . The current density in Fig. 5 is 70 A
m ’ i l l the lirst electrode aperture hut the source h a s been operatcd a t currents up to
100 A m~ o f H (1.25 A). This is close to the upper limit set by the Lsngmuir
Blodgett equation for a flat emitter:
’
Acceleration of ion beaman computed by WOLF Code
(825kV / 0 92 A I
FK 5.-Outpiit
from the WOLF codc showing ihc i o n Int~iIJcclomsin thc stirface sourcc
iicccleiator ( K w n ~C I til.. 19x6).
66 I
Negative hydrogen ion beams
V'!2
j- = 1.71 x 10- 7 dl
=
(9)
118Am-*
when d i s 38 mm (measured from the equipotential lines). The extraction aperture is
a slit 50 x 250 mm in area. The accelerator is designed to accept beams from a concave
converter as this gives a higher beam current from a larger converter as seen in Fig.
5 . This reduccs the accelerator current limit as the emitter is no longer flat but this is
compensated for by the high initial beam energy (equal to V,,,,,,,,,
100 eV) which
has the opposite effect.
The third electrode is biased at a small positive voltage to reflect positive ions in the
downstream beam plasma and prevent them from being accelerated backwards to thc
plasma source. T'ne smaii field in ihe gap docs not usuaiiy infiuence ihe beam optics
as the beam normally becomes spacc charge neutralized downstream of the third
electrode.
The optical quality of the beam from this accelerator can be measured'by its
emittance which is the phase space area measured in the divergence-beam radius plane.
Figure 6 shows the emittance diagram which has a 4-rms emittance (unnormalized) of
107 mm mrad. In the more usual normalized units this becomes 1.1471 mm mrad
which corresponds to an ion temperature of 7.8 eV. This agrees closely with the
reported high transverse ion velocities measured in converter experiments (KWAN f f
al., 1986). The need for a collimated output beam and the existence of a final diverging
electrostatic lens in the accelerator led to beam contraction during the accelerator
process. In this design a factor of two in radius contraction exists which increases the
beam temperature by a factor of four through adiabatic compression. The final beam
-
0
leml
FIG. h.-Measured e m i l l " diagram of B 0.98 A H- beam from the surface source.
Thc ims normalizcd emittilncc is 1 . 1 4 ~mm mrad which coircsponds Io an ion tempxNule
oF7.8 eV (KWANet nl.. 1986).
A. J . T. HOLMES
662
divergence is approximately ( T i / ? Vh)”*
*
where y is the compression factor (typically
0.5) and Vh is the final beam energy.
Electron suppression is achieved by the combination of the transverse magnetic
field of about 0.07 T formed by the bar magnets and a positive bias on the digital
electrode upstream. The electrons become pinned to the field lines which intersect the
positively biased clectrode. Only about 5% of these electrons cnter the accelerator
when the beam pervcance is optimum. Negative ions have a much higher cnergy a n d
are not affected by the small positive potential. However, ions with extremely large
transverse velocities are removed by this elcctrode which hence acts a s a collimator.
3. I . The hrrriuni sourcc uccebrnror
T h e recent development o f a barium convcrter source to replace the cesium source
described above had prompted a re-cnamination of the accelerator design. Most of
the basic Features ofrhe cesium source developed by KWANer ai. (i986j are unchanged
in the barium version. However, the evolving design o f neutral beam injectors has
caused some re-design of the overall system and has led to the abandonment of slit
optics and its replacement by large aperture circular apertures. The design for the
ITER injcctor based o n this barium source proposes a diode accelerator operating a t
100 keV which acts as a pre-accelerator for the electrostatic quadruple accelerator
(ESQ) which increases the beam energy to i .3 ivieV. This iaiicr acczicimioi~ia ~csLI.;;~;
more fully in a later section.
A t present no experimental test of a high energy beam from a barium source has
been made although an experiment is planned in the l i a r future. However, the beam
profile of a 10 kcV D- beam has been measured from a 10 cm diameter barium
converter through a 7 cm extractor apcrture. This has led lo a n estimate of a n I I eV
D- temperature, sorncwhat higher than the earlier H- beam from a Cs coated
convcrter, even though the current density is small (3.5 m A
a t the plane of the
lirst electrode of the accelerator.
4. V O L U M E S O U R C E A C C E L E R A T O R S
In this section we examine the extraction a n d collimation of negative ions to form
a heam. This involves solving two major problems simultaneously; (lie suppression
of the electrons within the plasma and the collimation o f the ion trajectories subject
to their own space charge forces. The electron suppression effect and beam collimation
are linked through the beam extraction process and this is discussed in the following
sections.
A I
_.I.
FI ,,,.
fwnY
LllLLll,,,,
...,”
n-,,,...
.’MI’,”C,’.,L
;,,.,
l,,,
iM
,,,, I
..-,,
111 “ , , ’ U , I I L
,,,, ,.,,
I
v
. l l l M l LL.I
Electron suppression techniques in surface sources are not directly transferable to
the volumc source as here the electrons and negative ions have very similar energies
(about I eV). The electron to negative ion flux ratio in the plasma is in excess o f 40
For hydrogen and rises t o 68 for dcuterium bccause o f the heavier ion mass. This
could reduce the accelerator power efficiency to below 2% if the electrons remain
unsuppressed. HAAS
and HOLMES(1990) and MCADAMS
et 01. (1990) have described
a method for achieving electron suppression by introducing a short rangc intense
magnetic ficld adjacent to the extraction aperture. If the field lines terminate o n a
Ncgalive hydrogen ion beams
bh3
Permanent mognct ekctro“ suppresro,
Source
,
Accetkotoi
I
Section of extraction aperture
FIG.7.-lllusiraIion
o f thc CICCIICIIIsuppressor sliuclurc ~n the plasma grid
surface which can be positively biased a s shown illustratively in Fig. 7, then it has
been shown that (HAAS
and HOLMES,1990)
1, = rcO
exp ( - a F e x p ( - ‘ ~ T J T Y )
(10)
where
Ic0 = err,u,/4
a = (2e/nmc)”*/(8e)
and I, is the extracted electron current. F is the imposed magnetic flux, Tc is the
electron temperature and ‘p is the plasma to biased surface potential (note ‘p is always
greater than zero). This exponential behaviour has been observed over a large range
in I,as seen in Fig. 8, where the magnetic field was generated using a flattened solenoid.
This exponential decay of the extracted electron flux ariscs from the pinning of
plasma electrons to the magnetic field lines ncai the extraction aperture with the resuh
..
20A
100 v
Vfnr -13.5V
vo,c
a
* 0.2TtS
r
0
04TIs-1
0.6Tlrl
+ OBTIr-l
40
80
I20
b,W
FIG,x.-Exfraclcd
clc~lroncurrrnl from a hydrogen discharge shown
field at the suppressor.
a6
a funclion of the
A. 1. T. HOLMES
664
that they develop a large transverse drift velocity and a much reduced axial velocity
towards the aperture. If the electrons can leave the system by striking the outer metal
structure with onlya small (or zero) potential barrier when i t is biased positively then
a large attenuation of the electron flux occurs (HAASand HOLMES,
1990). There is a
minimum field to "magnetize" the electrons against randomizing collisions of about
3 mT but above this critical field the electron flux is reduced according to equation
(12).
The presence ofthe inner exponen!ia!; exp (-rp/TJI i n equation (!& alsn gives a
powerful and simple method of attenuating the electron nux. The potential rp is
controlled by the bias of the plasma electrode to the discharge anode. Typically if left
floating, the potential is T, volts negative to the anode and the anode is about 2T,
volts ncgative to the plasma so that the floating potential is around 3T, volts ncgative
to the plasma, which is the value expected from sheath theory in hydrogen. The
presence of fast electrons in the extracted region can be readily detected by a more
negative floating potential of the plasma electrode.
The effects of biasing the plasma grid positively (i.e. reducing q) in hydrogen and
deuterium can be seen in Figs 9 and IO where a rapid reduction in electron current is
observed while the negative ion yield remains virtually constant. The relative dilficulty
in electron suppression in deuterium can be seen clearly as well as the change in
negative ion yield. The slight increase in negative ion current for low values of
bias voltage is attributed mainly to the increase in accelerator perveance foilowing
suppression of the electron space charge.
It has, however, been observed that very large magnetic fluxes or small values of q~
can cause loss of the negative ion current (MCADAMS
et al., 1990; LEUNCet ul., 1983).
There appears to be an upper limit to the applied magnetic flux of about 0.25 T mm
for good electron suppression and minimal loss of the negative ion current. Values as
low as e/H- = 0.5 and e/D- = 3 have been reported (LEAet ul., 1990).
4.2. Electron rrupping unii heum,focusing
The reduction or the ratio of electron to negative ion flux to about unity is not
sufficiently low to have a high overall power efficiency as defined above. However. if
-800
3025-
-
--6M
-
20 -
a.
4
E
<15-
8
-4w
I
I
E
.,
V,(VI
FIG.Y.-Plot or extracted electrons and H- currents in an 80 keV accelkrator as a funclion
of the bias potential oflhe electron ~ ~ l l e ~(plasma
l o r grid) relalivc lo the discharge anode.
Negative hydrogen ion beams
665
DCYtWiYm
Lc*l€QA
d;
%.,-1OOV
Vm.80kV
Vmt-12kV
I
' 'BJ.
\
' 1.
\
lloo0
0.0.56TIs~I
0-16mm
I"""
M"SIV1
FIG.IO.--Plot ofcxrracted electrons and D- currents in an XO keV accclerator as a function
ofthc hias potential of the electron collector (plasma gnd) relative to the discharge anodc.
the extracted electron current can be trapped at a low energy while the main negative
ion beam is accelerated to the full required energy ivirhour degrading the beam
emittance o r brightness, then the accelerator efficiency can be enhanced. The overall
eficiency can be expressed by :
'1 = I _
v,i(r- v,+r,v,)
(12)
where V , <c V,.
There are, however, considerable restraints on the value of V , and V , / V b .The first
of these is simple and is given by the Langmuir-Blodgett equation which can be
written in the form (HOLMES
and THOMPSON,
1981)
This is a more generalized version of equation (9) where o is the aperlure radius and
0 is the edge trajectory angle to the axis and is negative ifconvergent, The extraction
potential, V , , has a minimum value equal to the first g a p voltage and hcncc d is the
effective (electrostatic) length of this gap. Normally a collimated output beam is
required from this part of the accelerator and it is possible to show, using linear optics
modelling, that (HOLMES
and THOMPSON,
1981):
o=
-u/4d
(14)
Confirmation of the value of 0 can be obtained from the curve of the extracted beam
current versus first g a p potential. For low vaiuCS of V I ,prior to the Saturation in
current, which arises from a production limit set by the discharge, the value of ak/d
is:
A. J.
666
T. HOLMUS
where k = ( I +0.8dO/u). Figure I I gives thevalueofrl/ukversus theequivalent current
and prior to saturation d/uk attains a value of 1.2 when d/rr is 0.91. Hence, k equals
0.76 and
0 = (0.76- I)/(O.W/u)
=
-0.33.
This is similar to the theoretical value of -0.28 given by equation (15).
Experimentally (SURREY
a n d HOLMES,1990) it has been shown that u/d must he
much smaller and cannot exceed about 0.4 if third-order aberrations are to be kept
small, so we find that if we use this upper limit then:
I _ < 1.75 x IO-" V:/'
(15)
irrespective of the sizc of the cxtraction aperture. There is also a n upper limit set by
voltage brcakdown to the value of V , but this limit is rarely reached.
A much more restrictive limit to the upper value of V , is generated by the trapping
of the residual electrons which pass through the magnetic fields of the plasma electrode.
The lowest potential at which these electrons can be trapped is that of the extraction
electrode, V , , where the magnetic field a t the upstream edge of this electrode is
sufficient for them to bc dcflcctcd by about 90" into a recess in the electrode. The
negative ions are also deflected by about 2" for H- and 1.4" for D-. T h e valuc of V ,
must be small for high electrical eficiency and is further limited by the power dcnsity
of the dumped electron beam. An approximate figure of merit for the upper limit of
VI for equal electron and negative ion currents (a difficult target to attain) set by a
power density limit of 2 MW m-'is
'-
40
20
40
30
M
€a
1.0
F I G I I.-Plot
af the cffeclivc aspect riilio of Ihc ~ ~ C C C I C ~ ~ I Idluk,
O C . versus the normalized
enlraclcd current using the cxlr~clionvollsgc, V , . as the variable pi!ramelcr. Thc constancy
of dlok a m bc used lo derive the curva1urc angle o f Ihc pkdSIIYd boundary.
Negative hydrogen ion beams
2x
v:'2
667
IO"nu2
1.75x 10-8
or
VI i6.4 x 1 0 ' ~ "!.
(16)
Substitution into equation (16) yields:
I-
< 9 . 7 ~ ' [A].
(17)
Higher values of I - than this limit can be obtained by either increasing the second
electrode power density or by increasing aldwhich leads to aberration.
4.3. BE<* dloergence andfbc!?:?!.ng
As described above, a criterion for focusing is beam collimation on exit from the
first accelerating gap. However, usually the beam energy has to be much higher than
the first gap potential which is largely set by electron power handling requirements
so additional gaps are required. The next simplest accelerator is a two gap systcm
and this has been examined extensively a t Culham Laboratory by HOLMESand
NIGHTINGALE
(1986) and also at JAERI by WATANBE
PI al. (1990).
In the two gap accelerator, the first gap criterion is still approximately maintained
hut the beam must be truely collimated on exit from the second gap. The long drift
distance through the thick second electrode complicates the analysis of the properties
of the accelerator geometry and it is more appropriate to use a numerical ion trajectory
code with space charge to get a more exact answer. One such code is AXCEL which
assumes an ion current density and starts the trajectory calculation in a plasma of
electron temperature T,. Allowance for a tinite ion temperature is also included.
Essentially such a two gap accelerator consists of threc electrostatic lenses, as shown
schematically in Fig. 12, whose focal length is given by the expression
View of H-occelerator
100 kV
85 kV
0 kV
A . .I. T. HOLMES
h6X
.l.l "
I
Fic. 13.
I,
Ui;igr;im 01- the focusing cuwc 01.ii two gap acccler;klor whcrc d , = I X mm and
d, = 3 3 m m . A I w qhow'n is 1hc prediction hy ihc AXC'EL code.
,/ =4Clh;(EI-E,)
(18)
where uh is the local beam energy and E l and E , are the downstream and upstream
axial electric fields. The first of these lenses is located a t the upstream face of the
second clectrode a s shown in Fig. 14 and has a fixed diverging focal length or 3 d
(HOI.MFSand NI(;HTINciAI.F. 1986). The sccond lens is focusing and has a variahle
focal length o f 4 V , d J ( V h - V , ) while the last lens a t the third electrode position is
diverging and has an almost constant value of 4Vhd2/(Vb-VI). This shows that the
dominant parameter controlling focusing is V I and this can hc seen in Fig. I3 wherc
the beam divergence is plotted as a function of V , and compared with the predictions
of the AXCEL code. Agreement is quite good except for very defocused beams wherc
F I G 14.-Diagram of an I . ?MeV accelcriitur 1-or NET shouiog a series 01 liw
accelerating clrctrodes and the clectrosLalic collimator near the source.
niiiin
Negative hydrogen ion bealms
Ob9
A X C E L does not agree with experiment. Under thesc conditions severe vignetting
(i.e. beam scrape-off) occurs with the majority o f the beam ions striking the second
electrode.
Despite these successes, code prediction is still relatively simple a t present, a s the
A X C E L code and the other versions such as W O L F (used at Lawrence Berkeley
Laboratory) o r the SLAC code d o not model negative ion systems. These codes can
only simulate two component (or one component in the case of the SLAC code)
plasmas. No account is taken of the positive piasma potentiai at the piasmajbeam
interface, nor is there any simulation of the two particles (ions and electrons) which
are extracted. Very recently PAMELA
(1990) incorporated the electron suppression
effect which introduces a diffusion coefficient caused by the electron suppressor field
to simulate the excess electron space charge near the plasma boundary which has the
effect of reducing the apparent accelerator perveance even if the extracted electron
current is small. However, an increased cxtraction potential can overcome this effect.
5 . A C C E L E R A T I O N T O M E G A V O L T BEAM E N E R G I E S
T w o major techniques have arisen which can accelerate a n ion beam to energies in
excess of I MeV using d.c. potentials. In this paper we exclude discussion of radio
frequency accelerators such as linacs o r radio frequency quadrupole accelerators such
as the !?FQ or MEQP.LAC. The first of thcse is essentia!!y an extension of the
technique described in Section 4.3, called an electrostatic (ES) accelerator while the
latter is an electrostatic strong-focusing quadrupole accelcrator (ESQ) which has been
primarily developed a t Lawrencc Berkeley Laboratory by PURGALISand COOPER
(privatc communication). As these two techniques are very different in concept they
are described separately in the following scctions.
5. I . The electrusturic occelerutor
T h e electrostatic gaps shown in Fig. 12 can be continued to raise the bcam energy
in stages to any desired level. However, three effects have to he taken into account;
the repulsive effect of the beam space charge, control of secondary electrons and the
effects of stored electrical energy. The first elrect requires a periodic focusing lens to
retain a collimated beam while the second requires stray electrons to be deposited on
a suitable dump without allowing a significant build up in electron energy. In addition,
the stored energy strongly influences electrical breakdown and also the size of the
accelerator.
T h e beam can he collimated using a two gap accelerator as described in Section
4.3. However, this has a maximum upper energy of typically 8-10 fold V , where V ,
is thc maximum value of V , given by equation (16). Normally this is relatively small,
typically less than 200 keV. Higher beam energies require further gaps and a typical
example of a n ES accelerator is shown in Fig. 14 which was designed for the NET
neutral beam injector (HOLMES,1990). Here the first two gaps collimate the beam and
thereafter there is a uniform field of about 2 MV ni-' broken up into a series of
approximately equal gaps. There is a weak focusing lens at 200 kV and a very weak
diverging lens of 1.3 MeV. These two lenses together provide suficient focusing to
oppose the space charge force o i the I6 mA cm-' Li- beam.
T h e final beam divergence is about 1.5 mrad a n d is mainly caused by aberrations
a t the plasma boundary and is achieved by balancing the diverging lenses located at
670
A. J .
T. HOLMIS
the leading edge of the second electrode, third electrode and last electrode (at 1.3 MV
in this casc) against thc focusing lcnscs located a t the downstreaiii edge of the second
elecrode and fourth electrode. Divergence caused by space charge is small due to the
high beam encrgy. However, if significantly higher current densities are required than
the 16 mA c n i r 2 in thc above design, then spacc charge elrects would n o longer be
negligible and could cause a considerable problem. This could he overcome by increasing the axial field to shorten the distance over which the radial space chargc field acts.
In the design of HOLMES (IgYO), the electrons are suppressed a t the plasma
grid and any residual electrons are trapped in the thick second electrode. However,
gas stripping of negative ions within the accelerator column creates additional electrons
which must not he allowed to reach high energies. In the ES accelerator dcsign, this
can bc prevented by placing dipolc magnetic ficlds across each orifice in the accelerator
electrode so that the electrons are deflected by a large angle. T h e direction of the field
reverses in each electrode so that the total net integral of field along the accelerator
axis is zero, thus avoiding significant negative ion deflection although there is a
residual clectrostatic deflection arising from the displacement of the beam in the
accelerator aperturcs.
The advantage of the ES accelerator dcsign is that many apertures in parallel can
he used to obtain a large current as shown by WATANABE
et al. (1990) and this
technique can he applied to a megavolt accelerator subject to the current limit of the
supply. Howcvcr as the area, A,, of the grids increases, so does the electrostatic
capacitance and thc cncrgy rclcasc in a breakdown between the electrodes. This
energy, c, is given by
where it is assumed that the interelectrodc gap is the same as that between the shields
whose arca is A , and E is the electric field for a potcntial of V volts across the gap.
As shown in Fig. 15 for a conceptual injector for NET, each gap can be decoupled
from its neighbours by a set of closed Faraday shields so that we can consider each
gap separately. A stored energy of around 40 J per gap is thought lo be Ihc upper
limit (HOLMES,1990; BOTTICLIONIand BUSSAC,1980) before the onset of electrodc
surface damage caused b y electrical breakdown which leads to an inability to attain
reliable operation.
The value of A, is given approximately b y :
A, = 2 J n . I,, J A ,
where I, is thc neutralizer length. Thus
It should bc noted that the electrodes usually have a support structure which has a
weaker electric field and this should he included (with a suitable weight factor) in the
Negativc hydrogen ion bcams
61 I
13 m
LV power supplies
Radiation Shielding
Side elevation
of beomline model
FIG.15.-Diagram of an 8.5 M W injector for NET which operatcs ill 1.3 McV. In this
injector the source is at ground potential and the ncutmlizer is at high voltage.
value of A,. The value of A, is linked to the current I , the current density j - and the
transparency I , yielding
A, = I / t j _
Equation (20) has some interesting properties. Firstly, V i s a suitable fraction of the
beam energy bui is subjeci io praciicai mechanicai constraints; 200 keV is a reasonable
value but it could be larger with fewer accelerating electrodes. Similarly t is constrained
by mechanical limitations; an upper limit (excluding the support structure) is about
0.25 and for large systems with supports it may be less than 0.05. Choosing V = 2 x 10'
V and t = 0.05 with a beam current of 16 A and j - = 12 mA cm-2 gives a maximum
valueforEof2.3MVm-'.
Theaboveequation indicates that theaxial field of2 M V m - 'chosen for thedesign
of the NET injector is near the limit set by stored energy in the accelerating structure
but either a significant increase in V (fewer accelerator electrodes) or I (nominally 16 A
per injector) could lead to ditficulties as the stored energy per gap becomes significantly
greater than the limit with consequent increased risk or alternatively a weaker axial
field, E, must be used for acceleration with consequent beam spacecharge limitations
(i.e. a reduction in J- would be required). Operation of the source at high voltage
would reduce I , by about a factor of three but could lead to reduced gas handling
capability.
5.2. The electrostatic quadrupole accelerator
The basic concept of ihe ESQ accelerator is the use of electrostatic quadrupole.lenses
to oppose the space charge force of the beam. These lenses are significantly stronger
than the aperture lenses described above and SO a much higher beam current can he
transported as a single beam. The need for subdivision into many (typically 1 0 6
A. J . T. HoLms
612
1000) apertures for the ES accelerator concept is no longer necessary. However,
quadrupole lenses only focus in one plane so a series of alternating focusing and
defocusing quadrupoles forms the beam channel. In the orthogonal plane the lens
order is rcversed.
PURCALISand COOPER (private communication) have designed a single channel
accelerator from a series of electrodes with peg-like teeth as shown in Fig. 16. The
four pegs in each gap form the quadrupole, two from the downstream electrode and
two from the upstream one so that the beam is accelerated across the gap (as in the
ES accelerator) and is also focused in one plane and defocused in the other. A series
of these gaps with about 100 keV per stage comprises the accelerator. As the beam is
no longer circularly symmetric, the bcam trajectory calculations are more complex
and Anderson makes use of a beam envelope model code to derive a solution which
yields a collimated and symmetric beam at full energy.
The same problems described in Section 5.1 for the ES accelerator have been solvcd
for the ESQ accelerator but the nature of the solutions is different. The electron
problem is reduced in two ways; firstly the use of barium converter source minimizes
the extracted electron current and secondly the series of quadrupoles can only transport a beam to full energy which has a very restricted initial energy and convergence
angle. This inhibits electron acceleration as the electrons have typically a few electronvolts as an initial energy while the H- ions have the converter energy (typically 30&
500 eV).
The stored energy problem is very similar to the ES accelerator except for the value
of V which is smaller because there are more accelerating stages and the transparency
is smaller as there is usually only one aperture in the accelerator. It is possible to have
more than a single aperture (or channel) in the accelerator as in thc MEQALAC r.f.
300
550
800
keV
I
keV
kW
I
I
1050
keV
I
1300
heV
I
-” rl
I
0
Mete15
Single chonnel 1 A ESQ accelerator
FIG.16.-Diagram
a l the ESQ accelerator geometry.
Negative hydrogen ion beam
673
accelerator (VAN AMERSFOORT
et al., 1986) the electrode structure is then more
complex as the quadrupole lenses are shared between adjacent apertures.
T h e ESQ is hest suited to the converter type of source (with either Cs o r Ba coatings)
as this source can inject a high current into a single large aperature accelerator. The
geometrical curvature of the converter can he matched into the focusing periods of
the accelerator to give good beam transport in the system.
6. NEUTRAL BEAM INJECTOR DESIGN AT H I G H ENERGIES
The source and accelerator are probably the two most critical parts of the total
injector assembly whose design controls the form of the other components of the
injector, such as the neutralizer cell where the D- ions are converted to D oand the
residual ion dump as well as the auxiliary systems including electrical power and gas
handling. T h e design of the injector is also very strongly influenced by two key
decisions; the choicc of the earth point a t the plasma source o r neutralizer cell and
also whether the beam should he divided up into small elements which are separated
by conducting barriers.
In Table I , the four partners oflTER [the European Community. Japan, the U.S.A.
and the C.I.S. (formerly the U.S.S.R.)]have each chosen a separate selection of these
various options to form a conceptual design of a D- injector which produces an 8.3
M W D" beam a t I .3 MeV. The table shows the injcctor designs presented in the ITER
studies in 1990. These designs are continuously evolving and may change in the future.
In the C.I.S., a final dccision on the source type has not been definitely made. In
the following discussion only the European design is described in detail hut references
to the other designs will be made as appropriate.
6 . I.The Euuroprun itljecfor
The injector for ITER o r NET consists a t present of nine separate modules of 8.3
M W each arranged in columns of three units on three separate ports. The primary
objective is current drive which has a value:
I=
aN,[1020m-'] ' p [MW]
R [MI
[MA1
whcre N , a n d R are the plasma density and major radius of the Tokamak. The
parameter a is the current drive elliciency and is typically 0.4 for neutral beam
injection. T h e neutral beam power also provides burn control after plasma fusion
ignition.
Each module is a cylinder 4 m in diameter (this is the same for all designs) and up
TAHLB
I
Source
Acceleriltoi
Earth paint
Subdivision
Use ofcesium or barium
E.C
Japan
U.S.A.
C.I.S.
""IUInC
""IUlnC
surlace
""l""X?
ES
source
ycs
no
ES
ESQ
ES
ncutralirer
neutrslizCr
neulrilllzcr
no
YCS
YCS
yes
YCS
yes
-
A. J. T. HOLMES
h74
to 13 m long. The European design (HOLMES,1990) is shown in Fig. 1 5 and consists
of a volume source at ground potential which is also vacuum immersed and a multigrid accelerator to accelcrate this beam to 1.3 MeV. The grids are extended into
concentric electrostatic shields which totally surround the high voltage neutralizer.
Thc stored energy per gap (including the shields) is about 100 J, a factor of two above
the totally safe limit (BOTTICLIONI
and BUSSAC, 1980), hut it may be possible to limit
the fault current by inserting decoupling resistors between segments of these shields
as the high current path does not pass through the shields. A resistor value as low as
400 Q could be used following the results of SIMONand MICHELIER
(1968). An
additional advantagc of this approach is the avoidance of a single I .3 MeV gap.
The neutralizcr is based on an argon plasma cell which breaks up H - or D- ion
by long range Coulomb interactions with an 85% conversion eficiency. The plasma
is contained by multipole fields even across the beam path where the magnet pattern
is such that there i s zero net flux inrersccted by the beam as shown in Fig. 17. This is
only possiblc if the beam is subdivided into ribbons of circular apertures through a
judicious cxtraction aperture array. A wide orifice into thc plasma would require a
very large and heavy magnet to contain the plasma and is hence not practical. The
discharge power is about 200 kW for the plasma neutralizer and is coupled by two
loop antennae (one of which is seen in Fig. 17) at each end of the neutralizer cell. The
hcr or !ewer frequencies :.re pos.i'-!e
r,f. pnwm is at rypice!!y I00 GHr, l!!hOL!
within a factor of I O and the antennae
part of a resonant circuic. I t is not
necessary to have a high voltage platform for the r.f. oscillator or amplifier as the
power can be transferred to the 1.3 M V level by a serics of resonant r.f. air cored
transformers, one for each shicld potential. Thcse could be placed in coaxial transmission line linking the injcctor to the HV power supply.
TO I 3 MV screen
mognet arrangemm1
SUpPre6Lol
m0g"etS
/
'lhoriiantol plonel
EleclmsloliC resrduol ion
'deflector lhorlmnlal planel
Rr;. 17.-Diagnim
d l h c plasma n ~ u t r a l i ~syslcm
er
for an ITER Injector.
Negative hydrogen ion h e m s
675
Downstream of the neutralizer it is necessary to remove the residual ions from the
main beam and dump them on a highly cooled structure. Here subdivision of the
beam simplifies the system considerably compared with the Japanese approach as
firstly the beam can be deflected by electrostatic means as seen in Fig. 17 using plates
between columns of apertures to create the transversc electric fields. In this way the
beam ions are separated and D + and D- ions (about 8% of the total current of each)
are dumped on alternate plates in the dump. The cooling to the beam dump itselfcan
be increased to handle the total beam during commissioning when there is no plasma
target as then no neutral beam is required and the ion beam cannot be allowed to
leave the injector.
The beam exits the injector through slots in the shields and then passes through the
neutron plug, a series of slots through a neutron absorber which reduces the neutron
flux from the Tokamak. An additional neutron flux is formed by D-D drive in
reactions in the beam dump. A set of cryopumps upstream and downstream of the
neutron plug and gate valves limit the tritium migration into the injector.
6.2. Criticul technology
The critical technology areas for the injector are the high voltage power supply and
connections and the ion source and its accelerator. In parallel with similar programmes in
Japan and elsewhere, in Europe two projects have been started to examine these
issues. At CEN-Caderache in France a design study of a 4 A, I M V supply has started
with a group of industrial companies while a t Culham Laboratory in the U.K. a 4 A,
200 keV D- injector is being constructed which has most of the features of the full
scale system save that i t uses a gas neutralizer cell instead of a plasma cell. The work
at CEN-Cadarache together with additional work at Culham is aimed at creating a
scheme design for a I MV 4 A D - injector. In mid-I992 this design report will be
..I.-
rl
LlDC"
I.C
'
l
a
+La
I-n..:..Tnr .,
L l l L "'IDlJ
1v1 CL
-"..-, %LI.l+,,,"p"a'L1
L V
h . . : I A ,.-A
VUll"
'Lll"
L C I L
+h;"
;..:a,~+nv
L 1 1 1 3 "1,C'L"1.
7 . CONCLUSIONS
In the previous seciions the development of two major concepts of the production
of negative ion beams has been described; the surface production or volume production source. Each of these two techniques has many variants, most of which are
not described as discussion of these would confuse the basic issues.
However, ai present it is not yet clear which of the two methods will emerge as the
superior, indeed for someapplications one o r other technique may have an advantage.
In the field of neutral beam injectors for fusion, however, the volume source approach
may have a distinct advantage as it is very similar in concept to the earlier positive
ion beam units now in use on the JET o r JT60 Tokamaks. There are also clear
advantages in the area of beam divergence as the lower limit to divergence is set by
the ion temperature which is less than 1 eV compared with IO eV for the surrace
source.
At present negative ion sources and beams arejust entering that area ofdevelopment
where high power and voltage beams arc needed, particularly for fusion requirements,
and the next few years should see very exciting steps being made.
REFERENCES
W., SII:I,CNL~~~ F.. THOUAS R. W., WolKE R.. SCIIONBWILLI!
F. G. and
r u m CO&. p. 305. SLAC report 303. Palo Allo.
(1986) Pmc. 1986 Lincirr A
V A N AMEKS~OOKT
P.
1VANOV
S. T.
University of Califhrnin. Rcrkuley.
H n r r ~ x i I:.o ~and
~ BUSSACJ . P. (1Y80) PhTsico 104C. 248.
EHI.EKSK . W. and LEUNGK. N.(19801 RPII.Sci~wrl.hmrunl. 51. 721
HAASI:: A. and
HIIMPS A. J. 'I. ( I Y Y I ) Piusnu P1,j.v. Co~rr.
Fiwion 33, 1197: :dso in Culham Laboratory
Report, CLM-P-888. HM Slationary Office, 1990.
HANAIIAM., lNuuE T..KOJIMA
H., Marsol>n Y.. OHAKAY., OKUMUKA
Y., WATANAHE
K . and SEKIM.
(1990) R m S<:imi. In.ilruni. 61, 499.
HISKESJ.R. (1987) Conintmi.v A l . Mol. 1'h.w. 19, 59.
HOLMES A. I. T. (1990) Reti. Sci<>iir.
lmlrun?.61, 3x9.
HOLMLSA . J. T . , DAMMEKTZ
G. and GKEI~N
T. S . (19x3) Proe. hii. Ion Dlgirzeeriny Conf., Phys. Soc. ill
Japan. Vol. I, p. 71. Kyoto.
I ~ L MA.I J~. T.. GKEI~K
7. S.. INMAN
M . . WI(I.KIIK
A . R . and I-IAMPTOS
N. (1978) P r w 3rd l / ~ w m ~ ~ Grmohlc C m l . on Hroiiny in 7'oroiilul P1o.snw.s. Val. I, p. 95. Commission uf Europcan Communities,
Clencva.
HOLMESA. J . T. and NIGHTINGALE:
M . P. S. (1986) Kcu. S i i o i r . 1n.viruni. 57, 2402.
HOLMI:SA. J . T. ;and THOMPSON
E. (1981) R ~ uScieni.
.
imrrun,. 52, 172.
Y::;:;:*
!.!., ":XM
2%.
OL 7..
M X T S U M v.,
Y.. l,n
,,
Y., nota.+ E., SCU. M . !!Ed
W a r m n u e K . (1990) Er
13111S y p . on Ion Sou
Ion As~i.rrrdT d w o l o g y , Tokyo.
KIIVAKIK
V. J.. HEKSCOVITCH
A. I.and PKELEC
K . (1982) Ren. Scicwi. Insrruni. 53. 819.
K w m J . W., ACKEKMAN
G. D., ANDERSON
0.A.. CHANC. F., C ~ I ~W.ES . I, DEVK~BS
~
C.J.. LIETZKE
A.
I:., S ~ R O K AL. and STEEL^: W. F. (1986) /<a:.
Scicrrt. lnsirun,. 57, 831.
LEA L. M . . HDLMWI A. 1. T., T H O K N T O N M . F.and N n Y L n K G. 0. R. (1990) Reo. Scieni. ItwIrtm?. 61,
409.
LIWNI; K. N., EHLPKSK.W.and B n c M~ . (1982) Lawrrncc Berkcly Laboratory Reporl 14292. University
of California, Berkeley.
.Sci<m/.Imwenz. 54. 56.
LEUNGK . N.. EHI.I:KSK . W. a n d B
LEUNC K . N., HAUCK
C. A., KUN
HER S . R. (1988) Lawrence Berkeley Laboratory
Repor1 No. 25720, University of
MASSMAN P.. BOMMEL P. J . V., G K A N N E M A X E. 1.1. A.. Hnl,Max H.H., Lns 1..SIBHENLIST
F. a n d WIJMMIK
J. N.M . v. (1981) Proc. IOIlz Eurweun Con/: on Conlrolled Fusion uml Plamu Pliwics. Papcr H20.
EPS. (ienev:i.
M c A o n ~ sR., KINGR . F. and N E W M A R AF.
. (1990) R<w Scient. l m r u m . 61, 2176.
\'AN Os C. F. A., LEUNG K . N., LIETZKEA. F., STEAMS J. W. a n d KUNKEI.W. B. (IYYO) Prnc. 17lk
Euurqznn Cor!/.'oil Conrroikvl Fmhn utid Piusmo Hauling, Vol. 3, p. 979. EPS. Geneva.
V A N Os C. li. A.. V A N PINXTIWEN
H . M . a n d Los I. (19881 Proc. 3rd Europcun Workshop on Prorluclion
wzd A p p l k u i i m ofLigA1 Nqoiir:n lom Amcrslborl. The Nclhcrkinds. p. 133. FOM-InsIituLc Ibr Atomic
and Molecular Physics. Amsterdam.
J. (1990) Ccnlrc dEtudes Atomiqoe report, Francc.
PAMELA
SIMOND.J. and MICIIFLIIRR. (1968) Plot. 3rd In/. S w p . on Disclioryc~.sam1 Elecirical In,suLrlinn in
vrrcsur,,, Paris.
SUKKEY
E. and H~ILMES
A . J . T. (1990) Rei,. Scienl. ln.rlruni. 61, 2171
WA!1!3A I. M~ (lYX4) ?I!J?R
K., I.~ANAI,AM., INOUF T.. KOJIMA
H.. Marsuuh Y . , OHAKA
Y . and OKUMUKA
Y . (1990)
WATANAHE
P,-,,c. 13rh svmp. on im S O U ~ orld
C ~ ~
Ion As.si.s!<d ' r d d ~ ~ gTokyo.
y ,
YORK R. L., STWENSR. R., LEWUNGK . N. a n d EHLIXSK . W. (1984) Reo. S<iml,!.vlrunr. 55, 796.