Characterization of a spheromak plasma gun: The effect of refractory

electrode coatings
M. FL Brown,) A. D. Bailey Ill,b) and P. M. Bellan
California Institute of Technology, Pasadena, California 9112.5
(Received 27 September 1990; accepted for publication 23 January 1991)
In order to investigate the proposition that high-Z impurities are responsible for the
anomalously short lifetime of the Caltech spheromak, the center electrode of the spheromak
plasma gun has been coated with a variety of metals (bare steel, copper, nickel, chromium,
rhodium, and tungsten). Visible light (230-890 nm) emitted directly from the plasma in the
gun breech was monitored for each of the coated electrodes. Plasma density and temperature
and spheromak lifetime were compared for each electrode. Results indicate little difference in
gun performance or macroscopic plasma parameters. The chromium and tungsten electrodes
performed marginally better in that a previously reported helicity injection effect [ Phys. Rev.
Lett. 64,214 ( 1990)] is only observed in discharges using these electrode coatings. There are
subtle differences in the detailed line emission spectra from the different electrodes, but the
spectra are remarkably similar. The fact that ( 1) contrary to expectations, attempts to reduce
high-Z impurities had only marginal effect on the spheromak lifetime coupled with (2) an
estimate of Z,, < 2 based on a O-D model suggests that it is not impurities but some other
mechanism that limits the lifetime of small, cold spheromaks. We will discuss the general
characteristics of the spheromak gun as well as effects due to the coatings.
I. INTRODUCTION
An important consideration for plasma configurations
generated from high current density electrodes (e.g., sphero-
maks, dc helicity injection discharges,2 Marshall gun,3 and
Mather focus4 plasmas) is contamination from impurities.
At high current densities, high-Zelectrode material (as well
as low-Z adsorbed impurity gases) can be sputtered or
evolved into the discharge thereby raising Z,, and increas-
ing the resistivity and radiated power from the plasma. The
interaction of hot plasma with material limiters in large to-
kamaks and the concomitant increase in Z,, is a related is-
sue.5 Impurities have been a particular concern of sphero-
mak researchers. The Caltech Spheromak Injection
experiment* may be susceptible to this problem since
high discharge currents ( < 150 kA) are generated from
small electrodes ( - 100 cm2). We have noted that ( 1) our
spheromak magnetic lifetimes ( r,ife ) are anomalously short
compared with predictions of the resistive decay time and
that (2) copious impurity line emission is observed through-
out the discharge lifetime. In addition, we have observed
ablation of the center electrode and sputtering of electrode
material onto quartz vacuum fixtures. It has remained un-
clear whether there was a correlation between impurity con-
tent (i.e., Z,, ) and short rlife.
Spheromaks are force-free magnetofluid configurations
with comparable toroidal and poloidal magnetic fluxesi
which are generated largely by currents flowing in the plas-
ma itself. In force-free systems, internal JxB forces nearly
balance so that the magnetic field structure of spheromaks is
approximately governed by the eigenvalue equation
) United States Department of Energy Fusion Energy Postdoctoral Re-
search Fellow.
b, United States Department of Energy Fusion Science Research Fellow.
VX B = R B (with il = constant ). This equation must be
solved subject to boundary conditions; if a perfectly con-
ducting wall is assumed then image currents must flow to
provide equilibrium fields. Because of their high plasma den-
sity and magnetic helicity content compared with tokamaks
(the helicity K = SA *Bd 3x is related to volume integrated
field aligned current), spheromaks have been proposed as
both a tokamak refueling and current drive scheme. 13-15 The
Caltech Spheromak Injection experiment has demonstrated
both tokamak refueling and refluxing (current drive) by
spheromak injection.
Early spheromak researchers determined that impurity
radiation (dependent on Z,, ) was the most important ener-
gy loss mechanism in their discharges.16 From bolometery
measurements it appeared that all the plasma magnetic ener-
gy was radiated by impurities during decay. In the larger,
warm, nonradiation dominated CTX spheromaks, Barnes,
et al. ascertained that enhanced particle loss and particle
replacement power were the primary loss mechanisms.
Particle replacement power is that required to ionize neutral
gas and heat cold electrons to the bulk temperature in a par-
ticle confinement time. More recently, it has been recog-
nized that an important loss mechanism in spheromaks is
magnetic helicity dissipation due to electric fields in the plas-
ma edge. The CTX group has inferred large edge electric
fields due to the spheromak relaxation process. Since helicity
dissipation depends on the electric field, helicity dissipation
is greater at the edge in their discharges.
To understand the anomalous lifetimes in the Caltech
Spheromak Injection Experiment and ascertain whether the
cause is large impurity content, we embarked on a program
ofelectrode coating and spheromak cleanup. In Sec. II, theo-
retical issues concerning spheromak lifetimes, impurity pro-
duction rates, Z,, and other mechanisms affecting r,ife are
6302 J. Appl. Phys. 69 (9), 1 May 1991 0021-8979/91/096302-11$03.00 @ 1991 American Institute of Physics
6302
Downloaded 17 Jul 2003 to 130.58.92.72. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
discussed. In Sec. III, the Caltech spheromak gun and the
adverse role of impurities are described. In Sec. IV, our main
spectroscopic results from the various coated electrodes are
presented. In Sec. V, we close with a discussion of our results
in the context of other experiments and a summary. Gun
performance and operation are characterized in the appen-
dices. Electromagnetic properties of the gun are discussed in
Appendix A and we characterize the role of neutral gas in
our experiment in Appendix B.
II. THEORETICAL ISSUES
A. Spheromak lifetimes
The spheromak resistive lifetime can be viewed as an
L /R decay time where L is the spheromak inductance deter-
mined by the dimensions of the current path in the discharge
and R is the Spitzer resistance of the spheromak plasma.
Turner et aZ.l6 have pointed out that the magnetic diffusion
equation governing spheromak resistive decay can be writ-
ten:
dB
-=
- T,JA 2B
at
PO
(1)
where q is the Spitzer resistivity (7 a Z,, T - 32) and il is
the eigenvalue in the equation for the force-free state
(V X B = /z B). The e-folding time for the spheromak mag-
netic fields is then
r
mag =po/($ 2). (2)
We have noted that the magnetic decay time of our
spheromaks is anomalously short compared to this simple
Spitzer lifetime calculation. We have performed extensive
magnetic measurements of the spheromak equilibrium in-
jected into the empty Encore tokamak vessel and have de-
termined from a fit of the data to a simple equilibrium model
that R = 22 m .- . For a value of T, = 7 eV (measured with a
double Langmuir probe) and assuming Z,, = 1 we calcu-
late rmap = 100 ps while the measured e-folding time
r mag = 10 ps in our spheromak (the total rlife r20 ps).
Spheromaks which are not fully relaxed, and thus have
smaller dimensions, have been measured with values ofR up
to 30 m - and fully relaxed spheromaks late in the decay
phase of the discharge have T, as low as 4 eV. In these ex-
treme cases, we calculate T,,,~~ r50 ps. We conclude that
there is an anomaly factor of 5-10 between the predicted and
measured e-folding times of the magnetic fields of our spher-
omak discharges. Other researchers have noted a discrepan-
cy between measured spheromak lifetimes and those predict-
ed by simple Spitzer resistivity. Barnes et aZ.6 have reported
a resistance anomaly of 1.6-4.0 coupled with a Z,, of 1.3-
2.
It is important for spheromak injection experiments
that the spheromak can traverse the distance from its source
(a coaxial magnetized plasma gun in our case) to the center
of the tokamak before the spheromak resistively decays. An
important experimental figure of merit is the Lundquist
number, Lu = rlirr/rAlf, where rAlf is the AlfvCn transit time
across the spheromak. Since the AlfvCn transit time is just
the time required for the spheromak to move its length, the
Lundquist number is a dimensionless measure of the dis-
tance the spheromak can traverse in its lifetime. In our ex-
periment, Lur 10 so that this condition is marginally met. A
related dimensionless number is the magnetic Reynolds
number, Rm = ,uO rv/~, where r is a typical spheromak di-
mension (its radius) and u is its velocity. A large Rm is
indicative of a high degree of magnetic turbulence. Magnetic
turbulence and the associated magnetic tearing is critical for
helicity injection experiments. It is because magnetic helicity
is conserved even in the presence of turbulent tearing*,19
that spheromak injection into a tokamak can be expected to
drive current.
As noted earlier, it has recently been suggested by the
Los Alamos CTX group that spheromaks might rapidly lose
magnetic helicity (and therefore have anomalous lifetimes)
due to strong electric fields in the edge. The CTX group has
observed peaking of the current profile (parameterized by
J/B) during the decay phase of the discharge (cooler plasma
at the edge causes higher resistivity there) .20 As the sphero-
mak relaxes back to a flat J/B profile consistent with the
force-free Taylor state, electric fields arise in the edge in
order to drive the current. The helicity decay rate is given by:
$== -2
E-BdV.
(3)
If we assume that most of the dissipation occurs in the edge
where the electric fields are highest then the loss rate can be
approximated:
its K /rK s - 2@,,,, E*dl,
(4)
where aedge is the spheromak magnetic flux either on open
field lines or otherwise immersed in the edge volume where
there is a significant proportion of neutral gas. The electric
field becomes clamped at the value given by the Paschen
breakdown condition. Note that the resistivity directly at-
tributable to electron-neutral collisions is small compared to
the Spitzer resistivity since typically veVen /Y,, < 0.1 in our ex-
periment (where Y,,, is the electron-neutral collision fre-
quency).
We can determine the fraction of flux required in the
edge in order for the edge loss mechanism to have a signifi-
cant effect on our experiment. The minimum Paschen vol-
tage for breakdown in hydrogen is about 250 V while the
minimum path for edge breakdown is a poloidal circumfer-
ence, 2~ (0.12 m) = 0.75 m. If we assume that only the outer
5% is the edge then the edge flux becomes 5 X lo- 5 Wb,
where we have used 0.03 T as the average spheromak mag-
netic field. We can now estimate an upper bound on rK due
to this edge loss mechanism by calculating a lower bound to
k in Eq. (4). Using a total helicity content0*17 of about
10 - Wb*, we therefore conservatively estimate rK < 10 ,us
due to this effect in our experiment (with the assumption of
5% edge flux). This is close to our measured rk r 7-a /2 = 5
~CLS so the edge loss mechanism is a plausible candidate for
our anomalously short lifetimes.
6303 J. Appl. Phys., Vol. 69, No. 9,1 May 1991
Brown, Bailey Ill, and Bellan 6303
Downloaded 17 Jul 2003 to 130.58.92.72. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
B. Zero-dimensional impurity model
Based on the simple Spitzer lifetime calculation, our
anomalously short observed lifetimes could be due to high
Z,, caused by a large impurity content. To investigate this
possibility, a O-D rate equation model for impurity charge
state densities was developed similar to that of Barnes et al.
The model includes (i) the loss of neutral hydrogen due to
electron impact ionization (the rate of recombination of pro-
tons and electrons to reform neutral hydrogen is negligible at
10 eV), (ii) the production of electrons due to ionization of
hydrogen and impurities, and (iii) loss of electrons due to an
empirically determined finite confinement. Ionization, re-
combination and finite confinement of impurities are also
included. The equations also include simple, ad hoc impurity
source terms modeling the production of neutral impurities
during the formation period of the spheromak when elec-
trode ablation is most severe. The strength of these source
terms determines the relative impurity densities and the total
impurity percentage for a given program run. The hydrogen
density is determined by charge neutrality:
nH = n, -
z
zq njq *
Z,, is also calculated at each time step:
nH + c z&q
ZefT =
id
(7)
ne
The initial conditions for the model are a neutral hydrogen
density in accord with a single gas puff (there is no time
dependent source of hydrogen), a small seed electron den-
sity, and no impurities. The rate coefficientsi are calculated
assuming a constant, time independent electron tempera-
ture.
Carbon lines (CII, CIII) were the brightest impurity
lines observed in our experiment and no CV lines were ob-
served. Figure 1 shows a program run in which carbon (the
only impurity included here) peaks at 25% of the density.
The qualitative temporal behavior of the impurity densities
resemble the experimentally measured time evolution of im-
purity line emission (see, for example, Fig. 5 ). This run con-
sidered only carbon charge states up to and including CIV.
The electron temperature is assumed to be 10 eV. Even with
these liberal estimates of T, and the impurity level, we find
that Z,, never rises above 1.3. Similar results are obtained
with various fractions of oxygen, aluminum, chromium,
iron, nickel, and copper (the other observed impurity lines).
Thus, this model excludes the possibility that even a substan-
tial impurity fraction will increase Z,, to the values of 5-10
needed to account for our observed anomolously short life-
times. Our anomalous lifetimes must therefore be due to
some other mechanism.
III. CALTECH SPHEROMAK INJECTION EXPERIMENT
The Caltech Spheromak Injection Experiment involves
the formation and injection of small (rO = 4 cm), dense
(n, = lOI cmw3), high current (I,,, < 150 kA) sphero-
maks into the ENCORE tokamak (r = 12 cm, R = 38 cm,
time (~5)
lo1
Protons
b)
/
I
time (ps)
io
FIG. 1. Plot of result of O-D, time dependent impurity calculation (carbon
only), T, = 10 eV, (a) Z,, (left) and fraction of impurities (right), (b)
density of the various ionic components of the plasma, total density 1OB5
cm .
I, = 4 kA, &, = 700 G) (see Fig. 2). The base pressure is in
the low 10m6 Torr range and titanium gettering reduces
pressure to the low IO- range. The injected spheromak de-
posits density and magnetic helicity into the tokamak dis-
charge. The spheromak is formed by a coaxial, magnetized
plasma gun similar to those originally used by AlfvCn.* The
inner electrode of the gun is constructed of cold rolled steel
or Monel (copper/nickel alloy) 2.5 cm in diameter. The
outer electrode is a tube of stainless-steel 5 cm in diameter.
Hydrogen gas (0.1 Tort-) is puffed into the annular region
between electrodes and a 10 kV, 6 kJ capacitor bank initiates
the discharge. A radial current sheet forms ( 100 kA) and
accelerates down the gun distending magnetic flux (0.3
mWb) applied at the end of the gun (often referred to as
stuffing flux) and ultimately forming a spheromak.
Impurities entrained in the spheromak discharge play
an adverse role in all aspects of our injection experiments.
First, from the standpoint of tokamak refueling, impurities
are a deleterious addition to the tokamak discharge. For this
reason, it is advantageous to reduce spheromak impurities
even if they do not appreciably affect spheromak lifetime.
Second, to the extent that impurities reduce the helicity de-
cay time of the spheromak (rk =rnBz = r,/2), impurities
can reduce the amount of helicity deposited into the toka-
mak discharge. In other words, impurities can reduce the
Lundquist number, Lu, so that the spheromak magnetic
6304 J. Appt. Phys., Vol. 69, No. 9, 1 May 1991 Brown, Bailey. III, and Bellan 6304
Downloaded 17 Jul 2003 to 130.58.92.72. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
fields decay excessively before reaching the tokamak. Final-
ly, if the spheromak decays substantially before relaxing to
its lowest energy state then it is difficult to fit experimental
data to a simple (nondynamic) model.
IV. ELECTRODE COATINGS
Materials used for coating the center electrode of the
spheromak gun were chosen for a number of characteristics
(see Table I and Refs. 22-24). First, the material should be
refractory (high melting point) to prevent melting and va-
porization during the high current discharge. Second, the
material should have a low sputtering coefficient (where
c s,,ul is the number of electrode atoms evolved per incident
plasma ion). The sputtering coefficients presented in Table I
represent the number of electrode metal atoms liberated
upon impact of a 600-V argon ion. The coefficient for tung-
sten is a factor of two lower than for other materials in the
table so that tungsten is an excellent coating from the stand-
point of sputtering.
Secondary emission plays a role in the avalanche break-
down of gas between coaxial electrodes (see appendix B).
Ion bombardment of the cathode (the inner electrode in our
case) generates secondary electrons which sustain the dis-
charge. Tungsten has about a factor of two higher secondary
emission coefficient than other materials for which we have
data, so again tungsten has desirable coating characteristics.
Because the discharge has a relatively fast rise time
( < 10 ps, see appendix A), the gun current flows in a skin
layer. The skin depth in the cold rolled steel substrate
S,, = Jw, is less than the coating thickness ( Scoating )
because p is large. However, the skin depth in the coating is
large (S,, = = 0.4mm = 0.0 15 > ?jcoatlng for copper),
therefore most of the current flows in the coating for each
electrode in our experiment. So, for a skin depth large com-
pared to the coating thickness, the resistive power deposited
per
unit volume is given
by
I R /volume = I T/( 27r&,,t,ng ) . Note that if
s
r-3 < Llttng
(as it is for nickel), then the power per unit
volume for a given current becomes independent of the resis-
tivity 7.
The thermal properties of the coating are governed by
the thermal conductivity, /cth, and the thermal mass of the
material (the product of the density and specific heat). A
thermal skin depth can be defined:25 S,, = 2,/w
where At is the time the heat pulse is applied. In each case,
the thermal skin depth (for a 10~s heat pulse) is comparable
to the thickness of the coating. S,, varies from 1.2 mil for
nickel to 2.7 mil for copper while the electroplated coatings
were l-2 mil thick. Ideally, we would prefer to have the
coating thickness large compared to St,, so that the coating
would act as a thermal reservoir to conduct heat away from
the surface. Unfortunately, it is difficult to electroplate a
high quality coating thicker than a few thousandths of an
inch and thick plasma-coated surfaces tend to flake off of the
substrate (due to internal stresses developed during the coat-
ing process).
The steel center electrode of the spheromak gun was
electroplated with copper, nickel, chromium, and rhodium
(0.001 in. to 0.002 in. thickness) and was plasma coated
with tungsten (0.006 in. thick). After plating, the electrodes
were thoroughly cleaned (methanol in an ultrasonic
cleaner) and baked to remove residual electrochemicals.
Typical base pressures with the coated electrodes were
2 X 10 - 6 cm - with residual gas analysis showing predomi-
nantly hydrogen and nitrogen. Each coated electrode was
tested under identical conditions ( V,,, = 4.5 kV, I,,,, = 70
kA, and Qgun = 0.3 mWb) and the visible emission spec-
trum from the spheromak was measured ( 18 shots are re-
quired to measure the entire visible spectrum from 230 to
890 nm in roughly 40 nm increments). Typical plasma den-
sities were lOI cm - 3 and electron temperatures were 8 eV in
the gun discharge.
The visible spectrum (230-890 nm) in the annular re-
gion between inner and outer electrodes is measured with an
optical multichannel analyser (OMA) (Princeton Instru-
ments Model 120, with a 0.22 meter SPEX monochrometer,
Model 1681). A l-mm sealed quartz tube is inserted in the
middle of the electrode gap with the tip exposed a few mm at
the back of the gun (see Fig. 2). A quartz optical fiber is
inserted into the quartz tube such that emission can be
viewed axially along the length of the gun. The line of sight is
midway between the inner and outer electrodes. Attenuation
TABLE I. Comparison of refractory electrode coating characteristics.
Material T
In&
c
l pu rl
hardness
P Ktll cc.
Steel 1550C 1.3 0.061 10 160 7.86 0.52 0.12
Copper 1084 C 2.3 0.050 1.7 80 8.96 4.01 0.38
Nickel 1453 C 1.5 0.053 6.8 140 8.90 0.91 0.44
Chromium 1860 C 1.3 - 13 220 7.19 0.94 0.45
Rhodium 1965 C 1.5 - 4.6 260 12.9 1.50 0.24
Tungsten 3450C 0.6 0.096 5.6 440 19.3 1.74 0.13
Notes: Sputtering coefficient is for dc yield from bombardment of 600-V argon ions, secondary emission coefficient (1: ) is from bombardment of slow
hydrogen ions, electrical resistivity (7) is in units of 10 - 0 m, hardness is in units of kg/mm*, density (p) is in units ofg/cm , thermal conductivity (K,,, )
is in units of Wcm K at room temperature and specific heat (C,.) is in units of Jg K .
References: CRCHandbook ofApplied EngineeringScience, R. E. Bolz and G. L. Tuve, Eds. ( 1970), Metals Reference Book, C. J. Smithells, Ed. ( 1967) and
Glou Discharge Processes, B. Chapman ( 1980).
6305 J.Appl.Phys.,Vol.69,No.9,1 May 1991
Brown,Baiiey III, and Bellan 6305
Downloaded 17 Jul 2003 to 130.58.92.72. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
Fiber
12opF.lOkV
along the length of the optical fiber and OMA dynamic
range limit our measurements to 230 nm at the short wave-
lengths and 890 nm at the longer. The OMA was gated ( 10
,us) so that light from the gun breech was monitored during
only the first half-cycle of the discharge.
The principal result of this sequence of tests is that the
spheromak parameters and line emission were largely the
same for each electrode (see Table II). There were no sys-
tematic differences in spheromak density, temperature, or
lifetimes with electrode coating. However, the helicity injec-
tion effect was only observed with the tungsten and
chrome electrodes. Small variations in the helicity decay
time have a large effect on the amount of helicity that is
ultimately deposited into the tokamak since the helicity de-
cay time appears in an exponent ( Ksph = K, e - rK). Evi-
dently, 7-k in the case of the dirty electrodes was just small
enough so that Ksph decayed below our threshold of mea-
surement.
Before the present clean up program was undertaken,
we measured lifetimes as low as 15 ,us and never as long as 25
,us, we now routinely measure lifetimes up to 25 ,X and sel-
dom as short as 15 /.Ls. This improvement is not due to a
specific electrode coating but due to a combination of the use
of Macor ceramic insulators, titanium gettering and, most
important, effective discharge cleaning from firing hundreds
of shots. The best results have been with an uncoated Monel
electrode with a Macor insulator after firing s 200 shots (we
observe similar results with a tungsten coated electrode with
Macor insulators). Other researchers have made the obser-
TABLE II. Summary of refractory electrode coating results.
Thickness
Material in. Lines suppressed Sputtering Helicity
Steel - -
yes
ll0
%wr 0.001 576.3 (FeI) Ye no
Nickel 0.002 -
yes
no
Chromium 0.002 426.7 (CII), 504.1 (FeI), no Ye
505.1 (FeI), 576.3 (FeI),
634.8 (FeI), 637.0 (NiI)
Rhodium 0.001 426.7 (CII) yes no
Tungsten 0.006 426.7 (CII), 504.1 (FeI), no Yes
505.1 (FeI), 576.3 (FeI),
634.8 (FeI), 637.0 (NiI)
FIG. 2. Schematic ofexperimental set-
up showing location of quartz capil-
lary used for optical measurements.
vation that firing many shots is the best technique for clean-
ing spheromak guns.a.2b It is therefore likely that the dis-
charge cleaning removes some low-Z impurities (carbon
and oxygen) from our electrode surface.
In Fig. 3 we present spectra from six electrodes (tung-
sten, rhodium, chromium, nickel, copper, and bare steel)
and two typical wavelength ranges (487-533 and 606-652
nm). The main point is that the spectra are remarkably simi-
lar from 230-890 nm. There are several hundred identifiable
lines common to each electrode coating. Next, it is interest-
ing that a few lines are suppressed in the case of the tungsten
and chromium coated electrodes (Fe1 and Nil ) . The sup-
pressed lines that have been identified are summarized in
Table II. Also noted in Table II is an indication whether
sputtering onto the quartz capillary was observed and
whether the helicity injection effect was observed. We
should note that the limited resolution of the OMA (only a
few angstroms) made absolute identification of lines diffi-
cult. The identification of a given line is the most likely can-
didate of several annotated lines.
In addition to monitoring visible light emission in the
breech of the gun, we also scanned the visible spectrum of
emission from the tokamak discharge (l/4 of the tokamak
circumference from the injection point) with and without
spheromak injection. There was much less line emission
from the tokamak plasma than from the gun breech and
there were only - 10 additional lines observed upon sphero-
mak injection. In Fig. 4, we present typical data from the
wavelength range 527-572 nm. Note first that there are
many fewer lines than in the scans presented in Fig. 3. Sec-
ond, note that two new lines appear as a result of spheromak
injection (NiI and FeI). The fact that we see fewer lines in
the tokamak could be due to inefficient transport of high-2
impurities from the ablated electrodes. Other researchers
have noted that heavy impurities (mostly Fe1 in our case)
are not efficiently accelerated by the current sheath in the
gun.
We noticed that after 30-40 shots, the quartz optics,
particularly the l-mm quartz capillary used to view the gun
breech, became coated with metal. The coatings on the
quartz capillaries were analyzed using energy dispersive x-
ray analysis. For the most part, material from each electrode
was detected on the capillary used for the experiment. We
concluded that some (perhaps most) of the sputtering is
from the inner electrode but at least some material is being
ablated from the outer electrode. All of the electrodes sput-
6306 J. Appl. Phys., Vol. 69, No. 9, 1 May 1991 Brown, Bailey III, and Bellan 6306
Downloaded 17 Jul 2003 to 130.58.92.72. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
TUNGSTEN
RHODIUM
CHROMIUM
NICKEL
COPPER
STEEL
487 nm
533 nm
TUNGSTEN
RHODIUM
CHROMIUM
NICKEL
COPPER
STEEL
606 nm
652 nm
FIG. 3. Typical spectral comparison for each electrode coating, emission
from the gun breech, (a) the range from 487 to 533 nm showing the sup-
pression ofsome Fe1 lines in the tungsten and chromium electrodes, (b) the
range from 606 to 652 nm, showing supression of Fe1 and NiI lines in the
tungsten and chromium electrodes.
6307 J. Appl. Phys., Vol. 69, No. 9,l May 1991
I . , I,, , 1, I., ,
TOKAMAK DISCHARGE
546.3 (FeI)
(J
TOKAMAK
WITH SPHEROMAK
527 nm
572 nm
FIG. 4. Typical emission spectrum in the tokamak discharge (top) and
tokamak discharge with spheromak injection (bottom) for the tungsten
coated electrode, range from 527 to 572 nm.
tered some material onto the quartz capillary with the nota-
ble exceptions of the chromium and tungsten coated elec-
trodes.
In Fig. 5, we present the time history of a resonant CIII
line (229.7 nm) from different electrode/insulator systems.
The CIII line is among the brightest impurity lines in our
;o
b
50.
loo-
2
li
Y -
s
'i; -
$
E
E -
!
(b)
I I I I
0 20
time@)
40
FIG. 5. Time histories of 230 nm emission (including CIII at 229.7 nm) in
the tokamak discharge upon spheromak injection for representative elec-
trode/insulator systems, (a) I,,, , (b) 230 nm emission with the Monel-
/Macor system, (c) Monel/Delrin, (d) copper coated/Delrin.
Brown, Bailey III, and Bellan 6307
Downloaded 17 Jul 2003 to 130.58.92.72. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
discharge and we expect emission from carbon lines to be
indicative of ablation from plastic insulators. For this mea-
surement, we view emission through a quartz window and
narrow band filter (FWHM of 9 nm, centered at 230 nm)
with a UV enhanced photodiode (UDT-455UV) sensitive
down to 200 nm. The window is located on the tokamak
vessel, l/4 of a toroidal circumference from the spheromak
injection point.
Our cleanest system is a bare Monel electrode and a
Macor insulator [Fig. 5 (b) 1. We expect our dirtiest sys-
tem to be the high sputtering coefficient, low melting point
copper coated electrode with the Delrin (a machinable plas-
tic) insulator [Fig. 5 (d) 1. Note however that the emission
traces are remarkably similar in magnitude and shape. In
each trace there is an initial burst of CIII emission when the
spheromak enters the tokamak chamber followed by an-
other burst during the decay phase as the spheromak plasma
expands past the viewing window. We see that the time evo-
lution of this typical line is similar for the various electrodes.
The macroscopic parameters were also the same for each of
these electrode systems. Contrary to expectations, the plas-
tic insulator did not contribute appreciably to the carbon
impurity content of the discharge.
V. DISCUSSION
In summary, we have coated the center steel electrode of
the coaxial, magnetized plasma gun used in our spheromak
injection experiments with various refractory metals (cop-
per, nickel, chromium, rhodium, and tungsten). This was
done in an attempt to reduce the level of sputtered high-Z
impurities into the spheromak discharge so that magnetic
lifetimes might be increased. The chromium and tungsten
coated electrodes performed marginally better than the oth-
ers in that we measured only trace amounts of sputtered
chromium on our quartz capillaries and we were able to ob-
serve the helicity injection effect with these electrodes. In
addition, despite the very similar spectra among electrode
materials, a small number of spectral lines were supressed in
the chromium and tungsten electrodes and not in the others.
However, we found that none of the coatings had a measura-
ble effect on spheromak magnetic lifetime, temperature, or
density. In addition, discharge cleaning had only a marginal
effect on spheromak lifetime (not enough to account for the
factor of 5-10 discrepancy between theory and experiment).
On this experimental evidence coupled with a O-D calcula-
tion that Z,, < 2 in our discharge (Sec. II B) we conclude a
mechanism unrelated to impurities causes our anomalously
short lifetimes.
It is instructive to compare our magnetic lifetime mea-
surements with those of other experimenters. In Fig. 6, we
compare our longest magnetic lifetime to those measured in
several spheromak experiments (data from these experi-
ments are obtained from Refs. 1,6,9,16,28-3 1). We compare
only with other low temperature experiments ( < 50 eV)
since spheromak lifetime in hot spheromaks is governed by
dynamic processes rather than simple Spitzer resistivity
(e.g., current driven instabilities32 and ballooningJR ). The
reported experimental spheromak lifetime, rlife, is plotted as
a function of J Tj2 where r is the flux conserver radius.
6308 J. Appl. Phys., Vol. 69, No. 9,i May 1991
Since ~-,~r~ =puo/(77/2 2, and ~a T -32, the data should be
linear when plotted this way. A straight line with 2T22
scaling and passing through the Caltech measurement is
plotted (i.e., if we made our spheromak bigger and hotter we
would expect to move along this line) as is the Spitzer pre-
diction with Z,, = 1. Note first that spheromak lifetimes in
this low-temperature regime indeed have the predicted
Spitzer-like scaling but with an anomalously low coefficient.
This cannot be explained with a large Z,, since the tempera-
ture is not high enough to make Z,, greater than about two.
It appears from the plot that the 20,~~s lifetime of the Caltech
spheromak is not anomalous in the context of results from
other experiments and that a mechanism unrelated to Z,,
may be a ubiquitous feature limiting lifetimes ofcold sphero-
maks.
We should point out that contrary to this observation,
some experimenters have found spheromak r,ife invariant
with size* and with temperature34 under certain conditions.
However, these results were obtained in the range
12Ty2 = 10 to 100 m2eV32 where other processes keep r,ife
clamped near 1 ms. Note that data points in Fig. 6 at high
?Ta begin to deviate below the empirical fit. Recently,
Wysocki et a1.35 have shown that careful design of the CTX
flux conserver has reduced field errors and reduced edge flux
so that CTX spheromak decay time increases with T,.
We suggested earlier (Sec. II A) that the edge helicity
decay mechanism may play a role in determining rTlife of our
spheromak. A non-negligible component of unionized neu-
tral gas is unavoidable in the spheromak formation process.
Several hundred mTorr of hydrogen must be deposited in
_ x8 vs. r2Te30 for Cold Spheromaks (T, < 50eV)
/
/
1000
c
a
F SCALING
0 BETA (1983)
100
0 Ps.1 (1983)
I
/
@CALTECH (1989)
,
I
10
$ I I111111 1 I,,,,,, I 1,111~
.l 1 10 0
FIG. 6. Plot of T,,~~ vs iT for representative small, cold spheromaks.
Expected empirical scaling and the Spitzer prediction are also plotted. Note
that the measured lifetimes fall below the Spitzer prediction and that data at
larger ?T fall below the empirical prediction.
Brown, Bailey III, and Bellan 6308
Downloaded 17 Jul 2003 to 130.58.92.72. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
the gun breech (see discussion in Appendix B) some of
which is not ionized. In addition, since the spheromak plas-
ma is relatively cool and dense, significant recombination
might be expected in the edge. Finally, recent studies have
shownJ4 that a large concentration of neutrals can exist
within the spheromak far from the edge due to multiple
generation charge exchange. This effectively increases the
thickness of the edge (since neutrals are present far from
the wall) as well as increasing the energy lost due to charge
exchange with the neutrals. Based on the simple estimate
presented in Sec. II A, edge helicity dissipation may be an
important loss mechanism in our spheromak discharge.
APPENDICES: SPHEROMAK GUN
CHARACTERIZATION AND OPERATION
(b)
APPENDIX A: ELECTROMAGNETIC PROPERTIES I I
In addition to the spectroscopic measurements de-
scribed in Sec. V, a number of auxiliary diagnostics have
been used to characterize the spheromak gun and equilibri-
um. Various arrays of magnetic probes have been used to
measure the magnetic structure of the spheromaks and to
determine the magnetic lifetime.17 A typical loop in these
arrays is 0.5 cm in diameter and is electrostatically shielded
with a slit copper foil. Both active and passive integrators
have been used. Double tipped and single tipped Langmuir
probes calibrated with a 3 mm interferometer are used to
monitor plasma density. Calibrated Rogowski loops have
been used to monitor the spheromak gun current and preci-
sion voltage dividers monitor the voltage that appears be-
tween the inner and outer electrodes of the gun.
In Fig. 7, we present typical electromagnetic character-
istics of the gun (treating it as simple circuit). We present
data of the gun current [I,, Fig. 7(a) 1, the gun voltage (the
voltage that develops between the inner and outer electrodes
during formation, [ I,, Fig. 7(b) 1, a typical signal from a
magnetic pickup loop [ Bsph , Fig. 7 (c) 1, and a typical signal
from the saturated double Langmuir probe [n,, Fig. 7(d) 1.
The capacitor voltage was 4.2 kV and the stuffing flux was
0.3 mWb for this sequence. Note that V, is determined by
dynamic gun impedance and is usually significantly lower
than the capacitor voltage. We can determine the point at
which the spheromak breaks away from the gun by the char-
acteristic signatures in the 1, and V, traces at about 4,~s. As
the gun plasma distends the stuffing flux, the current path is
extended thereby rapidly increasing the gun inductance.
Since I, = (d /dt) (LgUnIg ), there is a sharp increase in
] V, 1. When the spheromak tears away from the gun, the gun
discharge restrikes at the back of the gun, rapidly decreasing
the gun inductance. At this point there is a drop in ] V, 1. We
can also measure the spheromak velocity by noting that
there is a 7 ,us delay between spheromak formation and de-
tection of spheromak fields and density 30 cm away; the
spheromak velocity is 4 cm/ps in this case. Finally, we are
able to ascertain the gun stuffing threshold,36 jlgun = pOlg/
Qp,, by noting that the spheromak breaks away from the gun
when I, ~50 kA while a, = 0.3 mWb. We find that Rgun
= 210 m - from this dynamic measurement.
b lb i0
tw
FIG. 7. Typical electromagnetic characteristics of the magnetized plasma
gun, (a) Z, ,,,,, (b) V, ,,,,, Cc) B,,,,, and Cd) n,., Z, ,,, and V,,, aremeasuredat
the gun electrodes, BIlrh and n, are measured in the tokamak vessel.
We have performed scans of rg (at fixed Cp, ) and mea-
sured the spheromak magnetic field, density and resistance
of the gun plasma with both 2 and 12 external cables. Gun
plasma resistance is defined as the ratio of I, at the peak of
gun current (to eliminate inductive effects) to the peak Ig.
The fixed stuffing flux for this run was 0.4 mWb. The switch
from 2 external cables to 12 reduced the total inductance of
our power supply and spheromak gun from 660 to 240 nHy.
This reduced the rise time of the circuit from Q-,,~ = 14 to 8.4
,us and enabled us to make shorter, more localized sphero-
mak plasmas. We were also able to operate at higher peak
current for the same capacitor voltage (since the capacitor
charge is switched faster). In Fig. 8 (a), note that we begin to
see small spheromak fields when I, E 60 kA (below this val-
ue the gun is stuffed). From this measurement we find the
stuffing threshold, Agun = ,~,l,/@, r 190 m - .
Our measured gun threshold, Agun = 190-210 m - , is
6309 J. Appl. Phys., Vol. 69, No. 9, 1 May 1991 Brown, Bailey Ill, and Bellan 6309
Downloaded 17 Jul 2003 to 130.58.92.72. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
I 1
i
l 2 cables
0 12 cables
o (a)
4
Bsph
m
0
0
: 0
. 0
.
.,J
0 -.o
. .$. .
I I I I I I I I I
FIG. 8. I,,,, scan, (a) E,+, (b) n,, and (c) R,,,,, closed (open) circles are
data points obtained with 2 (12) external cables.
somewhat higher than predicted by recent theories.6.37~38
These theories point out that the spheromak gun should be
matched to the lowest energy state admitted in the long cy-
lindrical entrance region which has a characteristic energy
per unit helicity:
$4l wIna* /Ksph = P&/@, = G /rent
where r,,, is the radius of the entrance region. The constant,
Cth, is a number between about 3 and 5 (C,, = 3.11 if the
gun forms a spheromak with m = 1 symmetry, C,, = 3.83
for m = 0 symmetry, C,, ~5.08 according to a sharp
boundary modelI ). In our case, 3.83/r,,, = 100 m -
which is a factor of two lower than the measured value of the
gun threshold. Barnes et ~1.~ point out that in a coaxial
magnetized plasma gun the radial cross section of the flow
narrows to form a dynamic nozzle which accelerates the
flow to the AlfvCn speed. The dynamic gun voltage that de-
velops is consistent with (toroidal) flux leaving the gun
breech at the AlfvCn speed. We find that our measured
threshold is closer to %-/a,, where Sgap is the (narrow) an-
nular gap between inner and outer electrode (about 1.3 cm
in our case) where the nozzle is formed. Another possible
explanation is that driving helicity from the gun to the toka-
mak through the relatively long and narrow entrance region
may be an inefficient process. Fernandez et al.39 found that
the helicity transfer from source to flux conserver was very
inefficient through a small cylindrical entrance region (i.e.,
though the gun threshold was exceeded, the magnetic field
developed in the flux conserver was small).
In Fig. 8 (b), note that the spheromak density rises rap-
idly with increasing current. The gas valve pressure was
fixed for this run at a plenum pressure of 15 psi. We observe
only about 10% increase in spheromak density with a factor
of two increase in plenum pressure. We conclude that gas
adsorbed on the electrode surfaces could be a possible source
of neutrals. Similar observations and conclusions have been
made by other experimenters (see Turner et &.,I6 Fig. 24).
In Fig. 8(c), note that the plasma resistance is lower
when the gun is stuffed and decreases slightly with increas-
ing current. We can estimate the plasma temperature in the
gun breech from the relation Rgun = (v/27&, ) ln(r,/r, )
where I,,, is the length of the gun breech and r, and r, are
the radii of the inner and outer electrodes respectively. Solv-
ing for T,, we find that
T,32=5.2x10-5
Z,,InA ln(r,/r, )
24un 4un
This relationship underestimates the magnitude of T, by a
factor of 10 but it points out that a drop in Rgun from 8 to 5
mR suggests a 25% increase in T,.
APPENDIX B: ROLE OF NEUTRALS
The operation of the spheromak gun also relies on a
large amount of neutral gas delivered to the breech by the
action of a fast gas valve. There are several reasons why a
high throughput valve is important. First, from the stand-
point of tokamak refueling, we would like to generate as high
density spheromak as possible. Therefore, we would like to
initiate the discharge with the maximum neutral pressure.
Second, the Paschen breakdown condition is such that the
neutral pressure in the breech of the gun needs to be rather
high (a few hundred mTorr) otherwise the breakdown does
not occur. Third, if the plasma density is insufficient to sus-
tain the discharge as it progresses, the plasma density will
need to be supplemented. If there is insufficient neutral gas
available, then electrode material will be ablated (this is the
gas starved discharge described below). The gas valve
needs to open rapidly to prevent neutral gas from propagat-
ing ahead of the spheromak before the discharge is initiated.
We have tested several valves and have operated the
spheromak gun with two of those tested. In Fig. 9, we plot
the response of three different valves as measured by a fast
ionization gauge (Schultz-Phelps type4 of our own design)
in a test stand. The gauge consists of two collecting plates 1
cm square separated by 0.5 cm and biased at + 45 V. A
heated tungsten filament runs midway between the plates.
When the gas puff moves across the gauge, the gas is ionized
and we measure the ion current to the - 45 V plate.
6310 J. Appl. Phys., Vol. 69, No. 9, 1 May 1991 Brown, Bailey III, and Bellan 6310
Downloaded 17 Jul 2003 to 130.58.92.72. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
In Fig. 9 (a), we plot the response of a valve designed by
Henins of the Los Alamos CTX group.4Y42 This valve con-
sists of a small plenum (0.6 cm3 in this case) which is pres-
surized through a small hole with the working gas (H, ). The
plenum is seated at one end of an aluminum rod which is
lifted rapidly by magnetic forces from a coaxial solenoid.
Firing the solenoid rapidly dumps the contents of the ple-
num into the system. Note in Fig. 9 (a) the abrupt arrival
(about 200 ps rise time) and the narrow width (about 500
ys) of the pressure pulse. The peak pressure is about 200
mTorr. This valve gave us the best spheromak parameters.
Initially, we used a small, readily available solenoid
valve ( Angar Scientific model 008). This valve had a higher
inductance solenoid so there was a longer delay before the
valve poppet moved. The Angar valve did not have a plenum
arrangement and therefore gas flowed as long as the valve
was open. Because of the high inductance of the coil the
valve remained open for several ms as is evidenced by the
trace in Fig. 9(b). The Angar valve had a sufficiently fast
rise time so that we were able to initiate the spheromak dis-
charge on the leading edge of the gas pulse. However, we
found that the un-ionized gas trailing the spheromak had a
deleterious effect on the spheromak lifetime.
We have tested an automobile fuel injector (GM part
No. 17 111965) as a possible fast valve for future spheromak
gun experiments. The fuel injector has a plenum (about 0.5
cm) which is dumped when the valve is opened. However,
unlike the LANL valve, the plenum is recharged through a
larger sized hole so that after an initial burst, gas continues to
(4
FUEL INJ.
I
Oi
I I I I 1 I I I I
012345678
t(ms)
FIG. 9. Plot of valve response measured with fast ion gauge for (a) Los
Alamos valve, (b) Angar solenoid valve, and (c) fuel injector.
flow as long as the valve is open. The fuel injector has a
substantial return spring and relatively low coil inductance
so that the valve closes after a short time (about 2 ms). In
Fig. 9(c), note the initial burst of gas followed by a brief
period of flow before the valve closes. The peak pressure in
the initial pulse is about 100 mTorr. The fuel injector is the
primary candidate for a new spheromak accelerator at Cal-
tech.
We have also tested piezoelectric valves with little suc-
cess. We found that we were unable to initiate a spheromak
discharge when a piezoelectric valve was used (even when
the valve was overdriven) and the signal on our fast ioni-
zation gauge was small ( < 10 mT). Evidently, the gas flow
through a piezovalve is insufficient for our purposes.
It is interesting to note the role the drift parameter,
J/n,, plays in these discharges (Jis the current density and
n, is the electron particle density). J/n, is a useful parame-
terization of current density in high current discharges. We
usually operate with a relatively high value
J/n, = 1-4X 10 - I4 A - m. First, J/n, is proportional to
the drift velocity of the current carrying electrons, so
J/n,T. I is a measure of the ratio of the electron drift to
thermal velocity. Generally, if udrift /qhcrmal approaches uni-
ty then two-stream instabilities and anomolous resistivity
become important. Second, J/n, is a measure of the ratio of
ohmic dissipation (heating) to radiation loss
( Pohmic = vJ,P,,, a nenimpuri,, ). Experimentally, it ap-
pears that J/n, has an upper bound; the maximum
J/n, = l-5 x 10 - I4 A - m over a wide range of currents
and densities in a number of experiments.738*43V44
Breakdown of neutral gas will occur when primary elec-
tron current flowing from the negative cathode is enhanced
by multiple ionizations in the traversal to the anode. It is
straightforward to show 45 that the multiplication of primary
current between parallel plates is given by:
I e
ad
-=
IO
1 ---(cod- 1)
where a is the electron production per cm, yis the secondary
emission coefficient, and d is the plate separation. Physical-
ly, there is the expected exponential growth of the primary
current due to multiple ionization but if there is sufficient
secondary emission from ion bombardment at the cathode
then an avalanche breakdown occurs when
y( cad - 1) = 1. In a coaxial system, ad is replaced by $:
adr where rl and r, are the inner and outer electrode radii.
Operationally, we have found that running with a nega-
tive center electrode affords more reliable breakdown than
running with a positive center electrode. This could be due to
the higher cathode electric field (due to cylindrical geome-
try) which may facilitate the generation of primary current
or enhance secondary emission at the cathode surface. In
any case, negative center electrode coaxial system has been
shown to have a lower breakdown voltage than a positive
center electrode system. 45 We also found that we had fewer
flashbacks (arcs from the high voltage flange back to the
gas valve housing) when running with a negative center elec-
trode so this mode of operation was safer than running with a
positive center electrode.
6311 J. Appl. Phys., Vol. 69, No. 9,1 May 1991
Brown, Bailey Ill, and Bellan 6311
Downloaded 17 Jul 2003 to 130.58.92.72. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
We found that it was important to fire sufficiently late
into the gas pulse so not to produce gas starved discharges.
Material is typically sputtered from the cathode (the center
electrode in most cases, ours included) due to high voltage
ion bombardment. However, material can also be sputtered
from the outer anode in the case of a gas starved discharge.
This occurs when there is not enough neutral fill gas (hydro-
gen) to provide the ion flow that sustains the current in the
radial current sheet between the inner and outer electrode. A
large sheath develops and the material is electrostatically
pulled from the anode into the plasma.46 In order to prevent
the metal electrode surface from becoming a source of ions, it
might be useful to make the anode a porous, high surface
area material (such as plasma sprayed tungsten or titanium
loaded with hydrogen as proposed by Post and Turner25 ). In
this case, gas adsorbed onto the surface of the anode could
sustain a gas starved discharge. In fact, the plot of n, increas-
ing with 1, depicted in Fig. 8(b) (with fixed gas puff pres-
sure) is suggestive of just such a process.
ACKNOWLEDGMENTS
It is a pleasure to acknowledge the technical assistance
of M & T Plating, Frank Cosso, and Larry Begay, as well as
useful discussions with Dr. Cris Barnes, Dr. Tom Jarboe,
and Dr. Juan Fernlndez of Los Alamos National Laborato-
ry and Dr. Charles Hartman and Dr. Jim Hammer of Law-
rence Livermore National Laboratory. This work was per-
formed under DOE Grant No. DE-FG03-86ER53232.
T. R. Jarboe, I. Henins, H. W. Hoida, R. K. Linford, J. Marshall, D. A.
Platts, and A. R. Sherwood, Phys. Rev. Lett. 45, 1264 (1980).
M. Ono, G. J. Greene, D. S. Darrow, C. Forest, H. K. Park, and T. H.
Stix, Phys. Rev. Lett. 59, 2165 ( 1987).
J. Marshall, Phys. Fluids 3, 134 ( 1960).
5. W. Mather, Phys. Fluids 8, 366 (1965).
5B. C. Stratton, A. T. Ramsey, F. P. Boody, C. E. Bush, R. J. Fonck, R. J.
Groebner, R. A. Hulse, R. K. Richards, and J. Schivell, Nucl. Fusion 27,
1147 ( 1987), and A. T. Ramsey, D. Manos, B.C. Stratton, and M. Ulrick-
son, Bull. Am. Phys. Sot. 34, 1965 (1989).
C. W. Barnes, T. R. Jarboe, I. Henins, A. R. Sherwood, S. 0. Knox, R.
Gribble, H. W. Hoida, P. L. Klinger, C. G. Lilliequist, R. K. Linford, D.
A. Platts, R. L. Spencer, and M. Tuszewski, Nucl. Fusion 24,267 ( 1984).
C. W. Barnes, T. R. Jarboe, H. W. Hoida, B. L. Wright, R. A. Hulse, and
D. E. Post, Nucl. Fusion 25, 1657 (1985).
J. C. Fernandez, C. W. Barnes, T. R. Jarboe, I. Henins, H. W. Hoida, P. L.
Klinger, S. 0. Knox, G. J. Marklin, and B. L. Wright, Nucl. Fusion 28,
1555 (1988).
9N. Satomi, S. Goto, Y. Honda, Y. Kato, M. Nagata, M. Nishikawa, A.
Ozaki, T. Takaishi, T. Uyama, K. Watanabe, T. Sato, and T. Hayashi,
International Atomic Energy Agency, Eleventh International Confer-
ence on Plasma Physics and Controlled Nuclear Fusion Research, Kyoto,
Japan (1986).
M. R. Brown and P. M. Bellan, Phys. Rev. Lett. 64, 2144 ( 1990).
M. R. Brown and P. M. Bellan, Phys. Fluids B 2, 1306 (1990).
M. N. Rosenbluth and M. N. Bussac, Nucl. Fusion 19,489 ( 1979).
lJC. W. Hartman and J. H. Hammer, Phys. Rev. Lett. 48,929 ( 1982).
14P. B. Parks, Phys. Rev. Lett. 61, 1364 ( 1988).
s L. J. Perkins, S. K. Ho, and J. H. Hammer, Nucl. Fusion 28,1365 ( 1988).
W. C. Turner, G. C. Goldenbaum, E. H. A. Granneman, J. H. Hammer,
C. W. Hartman, D. S. Prono, and J. Taska, Phys. Fluids 26, 1965 ( 1983).
M. R. Brown, D. M. Cutrer, and P. M. Bellan, Phys. Fluids B (in press).
J. B. Taylor, Phys. Rev. Lett. 33, 1139 ( 1974).
19J. B. Taylor, Rev. Mod. Phys. 58, 741 (1986).
OS. 0. Knox, C. W. Barnes, G. J. Marklin, T. R. Jarboe, I. Henins, H. W.
Hoida, and B. L. Wright, Phys. Rev. Lett. 56, 842 (1986).
* H. Alfven, L. Lindberg, and P. Mitlid, J. Nucl. Energy, Part C: Plasma
Phys. 1, 116 (1960).
R. E. Bolz and G. L. Tuve, Ms., CRC Handbook ofApplied Engineering
Science (CRC, Cleveland, OH, 1970).
23C. J. Smithells, Ed., Metals Reference Book (Plenum, New York, 1967).
14B. Chapman, Glow Discharge Processes (Wiley, New York, 1980).
25 R. F. Post and W. C. Turner, Proceedings 4th Compact TorusSymposium.
(Los Alamos National Laboratory, 1981), p. 153.
C. W. Barnes, I. Henins. H. W. Hoida, T. R. Jarboe, R. K. Linford, J.
Marshall, A. R. Sherwood, and M. Tuszewski, Proceedings 4th Compact
Torus Symposium (Los Alamos National Laboratory, 198 1 ), p. 125.
*H W Hoida, C. W. Barnes, I. Henins, T. R. Jarboe, S. 0. Knox, J. Mar- .
shall, D. A. Platts, and A. R. Sherwood, Proceedings 5th Compact Torus
Symposium (Mathematical Sciences Northwest, 1982), p. 97.
*A. Janos, Phys. Fluids 29, 3342 ( 1986).
H. Bruhns, C. Chin-Fatt, Y. P. Chong, A. W. DeSilva, G. C. Golden-
baum, H. R. Griem, G. W. Hart, R. A. Hess, J. H. Irby, and R. S. Shaw,
Phys. Fluids 26, 1616 (1983).
C. Munson, A. Janos, F. Wysocki, and M. Yamada, Phys. Fluids28.1525
(1985).
G. C. Goldenbaum, J. H. Irby, Y. P. Chong, and G. W. Hart, Phys. Rev.
Lett. 44, 393 ( 1980).
Y. Ono, R. A. Ellis, Jr., A. C. Janos, F. M. Levinton, R. M. Mayo, R. W.
Motley, Y. Ueda, and M. Yamada, Phys. Rev. Lett. 61,2847 (1988).
F. J. Wysocki, J. C. Fernandez, I. Henins, T. R. Jarboe, and G. J. Marklin,
Phys. Rev. Lett. 61, 2457 (1988).
R. M. Mayo, C. K. Choi, F. M. Levinton, A. C. Janos, and M. Yamada,
Phys. Fluids B 2, 115 (1990).
. F. J. Wysocki, J. C. Fernandez, I. Henins, T. R. Jarboe, and G. J. Marklin,
Phys. Rev. Lett. 65,40 ( 1990).
.IbC. W. Barnes, J. C. Fernindez, I. Henins, H. W. Hoida, T. R. Jarboe, S. 0.
Knox, G. J. Marklin, and K. F. McKenna, Phys. Fluids 29,3415 ( 1986).
M. J. Schaffer, Phys. Fluids 30, 160 (1987).
*C. W. Barnes, T. R. Jarboe, G. J. Marklin, S. 0. Knox, and I. Henins,
Phys. Fluids B 2, 1871 (1990).
J C Fernandez, B. L. Wright, G. J. Marklin, D. A. Platts, and T. R. . .
Jarboe, Phys. Fluids B 1, 1254 (1989).
40G. J. Schultz and A. V. Phelps, Rev. Sci. Instrum. 28, 1051 (1957).
4 Kindly provided by I. Henins, LANL.
421. Henins and J. Marshall, Rev. Sci. Instrum. 40, 875 ( 1969).
4 T R Jarboe, Proceedings 7th Compact Torus Symposium (Los Alamos
National Laboratory, 1985) p. 34.
44H. W. Hoida, C. W. Barnes, I. Henins, T. R. Jarboe, C. J. Buchenauer,
and G. Marklin, Proceedings 7th Compact Torus Symposium (Los Ala-
mos National Laboratory, 1985), p. 49.
45A. Von Engel, Ionized Gases (Clarendon, Oxford, 1955).
4o D. A. Platts, T. R. Jarboe, H. W. Hoida, I. Henins, A. R. Sherwood, R. K.
Linford, and S. Shimamora, Proceedings 6th Compact Torus Symposium
(Princeton Plasma Physics Laboratory, 1984) p. 72.
6312 J. Appl. Phys., Vol. 69, No. 9,1 May 1991 Brown, Bailey III, and Bellan 6312
Downloaded 17 Jul 2003 to 130.58.92.72. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp