Magnetic Electrostatic Plasma Confinement

Review Article: Magnetic Electrostatic
Plasma Confinement
By Thomas James Dolan, Idaho National Engineering Laboratory, EG&G Idaho,
PO Box 1625, Idaho Falls, ID 83415-3880, USA
Author contact information as of April 2009: University

of Illinois, Department of Nuclear, Plasma, and Radio
List of Parameters
logical Engineering, 100G Talbot Lab, MC234, 104 SI Units are used, unless otherwise noted.
South Wright Street, Urbana, IL 61801.
a = halfwidth of the anode gap
[email protected], 217 2447257
A = area
This work was supported in part by the US Department of
Energy, Oce of Energy Research, contract DOEAC07 Aloss = e ective area for plasma loss out of magnetic
76ID01570. cusps
A = magnitude of the magnetic vector potential
Abstract B = magnetic field magnetic induction
Electrostatic plasma confinement and magnetic electro Ba = magnetic field in the anode gap
static plasma confinement MEPC have been studied
for four decades. The multiple potential well hypothesis, Bb = 2μonT + Ti1/2 = boundary magnetic field
postulated to explain high neutron yields from Hirsch’s B = peak magnetic field at the coil
colliding beam experiment, has been supported by sev
eral pieces of evidence, but results were inconclusive. B = point cusp magnetic field
Magnetic shielding of the grid was developed to reduce c = speed of light
the required beam current and to prevent grid overheat
ing. Electrostatic plugging of magnetic cusps evolved to cs = ion acoustic speed
a similar configuration. Due to low budgets, early d = grid wire diameter
MEPC experiments used spindle cusps, which are poor
for plasma confinement. Later experiments used multi Da = electron di usion coe
cient due to collisions
pole cusps or a linear set of ring cusps, which have larger with neutral atoms
volumes of fieldfree plasma. To keep the selfshielding Dei = electron di usion coe
cient due to collisions
voltage drop Φ 100 kV, the electron density s in the with ions
anode gap should be less than about 1019m3. The central
Dmax = maximum di usion coe
cient
plasma density can be an order of magnitude higher. The
ATOLL toroidal quadrupole had anomalous electron Do = total collisional di usion coe
cient
energy transport, but the Jupiter2M linear set of ring
= electronic charge
cusps achieved a transport rate about a factor of two
above the classical rate. With nearclassical transport, a fi = fraction of incident neutrals which are ionized
power gain ratio Q 10 is predicted for a reactor with rp instead of charge exchanged
= 3 m, B = 6 T, and applied voltage ΦA = 400kV. Besides
f = fraction of fusion product alpha energy W’’,
producing electricity and synthetic fuels, MEPC reac
transferred to electrons
tors could be used for heavy ion beams sources and neu
tron generators. The main issues of concern for MEPC fi = fraction of fusion product alpha energy W’’,
reactor development are electron transport, plasma pu transferred to ions
rity and electrode alignment and voltage holding.
F = geometrical factor having dimensions of
reciprocal length
g = spacing between grid wire centers
Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 1

h = width of the anode along the magnetic field ro = virtual electrode radius electrostatic con
direction finement and Penning trap
I = electron current from electrodes into the ro = radius of plasma boundary surface in point
plasma cusp
Idi = electron diusion current rp = approximate plasma radius at magnetic field
boundary
I = electron injection current
R = accelerating grid radius electrostatic
= wave number = 2/ confinement
= Boltzmann constant R = magnetic mirror ratio
= subscript denoting either electrons or ions R = ring cusp anode radius
L = length between ends of a linear cusp device S = plasma surface area
L1 = distance between adjacent ring cusps Si = ionization source
= electron mass S = volumeaveraged neutral input rate
= number of pairs of multipole conductors Te = electron temperature keV
= N /2
T = electron temperature
M = ion mass
Teo = central electron temperature
= plasma electron density
u = ion speed
a = uncompensated electron density at centre of
anode gap uia = ion speed in anode gap
B = Brillouin density = electron velocity
= particle density of species k = electron densitygradient drift velocity

c = central core electron density V = plasma volume
N = total number of electrons = characteristic halfwidth of the electron
distribution in anode region
o = central plasma density
Wa = average energy acquired along the slope of the
o = initial plasma density pulsed electrostatic potential well by ions
confinement
W = electron energy
ia = ion density in the anode gap
Wi = ion energy
= trapped cold electron density in anode region
Wloss = average electron energy loss per incident
N = number of multipole cusp gaps neutral atom due to ionization
p = / =1/2 W = ion energy perpendicular to the radial
PE = electron heating by the edge electric field that direction
aects the central plasma it = ion heating by x = r/R
plasma waves
y = Φi/T
Prad = radiative power loss
y =
Φk/T
P = electron power loss to plasma waves
Po = canonical angular momentum Y = 2R/N
Q = fusion power/input power Z = charge number
q = 2pe/2ce Z = 1 for electrons and 1/2 for ions
r = radius = a parameter depending on Ti/T
ra = point cusp anode radius = fraction of current intercepted by the grid
rb = plasma boundary radius

= the eective halfwidth of the cusp gap for i = ion confinement time
plasma loss
0 = electron collision time
= a function, of Ph P2
trap = time for central plasma free electrons to be
Φ = selfshielding voltage drop in anode gap trapped by edge magnetic field
r = thickness of plasma edge region ΦA = applied voltage
= trapped electron pumpout rate/ Φ = potential barrier for electrons or ions
diusion rate Φ = electron potential barrier
= eective open fraction of wire mesh grid
Φi = ion potential barrier
= geometrical open fraction of wire mesh grid
Φp = plasma potential = Φi + Φ
0 = permittivity of free space
c = cyclotron frequency
= global Brillouin ratio
c = electron cyclotron frequency
In = Coulomb logarithm
ci = ion cyclotron frequency
= wavelength
p = electron plasma frequency
μo = permeability of free space
ei = electronion momentumtransfer collision
frequency
a = electron Larmor radius in anode gap
b = electron Larmor radius at plasma boundary
= electron Larmor radius
o = electron Larmor radius in point cusp
i = ion Larmor radius
1 = inner electron turning radius
2 = outer electron turning radius
<> = DT fusion reaction rate parameter

= electron loss time by diusion across the
magnetic field
= pulse length
k = characteristic collision time for electrons or

ions
eond = time for heat loss by conduction
di = characteristic time for electron diusion
across the magnetic field
= electron lifetime or confinement time

E = electron perpendicular energy confinement
time
E = electron parallel energy confinement time
ei = electronion collision time

eq = electronion equipartition time

1. Introduction High Voltage
Feedthrough
Magnetic electrostatic plasma confinement MEPC

Vacuum
uses electrostatic fields to enhance plasma confinement Chamber
by open magnetic fields. The purpose of this review is to
describe: Source
Grid
i the evolution of MEPC from electrostatic
plasma confinement and from open magnetic
confinement systems Accelerating
Grid
ii MEPC theory
iii experimental observations
iv fusion reactor concepts Virtual
Electrode
v salient issues Hypothetical
Trajectory
MEPC devices have also been called ‘electromagnetic
traps’, ‘magnetoelectrostatic traps’, ‘magnetic inertial
electrostatic confinement’, ‘electrostatically plugged
cusps’, ‘kineticelectricmagnetic plasma confinement
KEMP’, ‘electricmagnetic confinement’, ‘Polywell’ and
‘modified Penning trap.’ Discussions of electrostatic
field applications in tandem mirrors, bumpy tori and Figure 1 — Concentric spherical electrodes used for electrostatic plasma confinement.
tokamak helicity injection are beyond the scope of this
review. Deflections of injected particles towards nearby grid
wires decrease the eective open area of the grid by an
amount 7
1.1 Electrostatic Plasma

2
Confinement
= 1
1

1/2
16 g
1
1.1.1 Virtual Electrode Formation where is the eective open fraction of the grid, g = g
d2/g2 is the geometrical open fraction of the grid, g is the
Lavrent’ev 1 first proposed electrostatic plasma confine
spacing between grid wire centers square mesh and d is
ment on 22 June 1950, and magnetic field enhancement
the grid wire diameter. For example, if g = 4 mm and d =
of electrostatic plasma confinement in March 1951. In
the early 19605 Farnsworth 2 proposed electrostatic 0.2 mm, then g = 0.90 and = 0,865. The circulating
plasma confinement for the production of fusion reac current Ic is related to the current I intercepted by the
tions, based upon his experience with spherical multi grid wires by the equation:
pactor vacuum tubes. Lavrent’ev3 developed a theory of Ie 2
charged particle flow and focusing in plane and spherical = 2
geometries. Budker4 calculated the fusion reaction rate I (1 2 )
that could be attained by trapping ions in the negative
electrostatic charge of an electron beam. Early electro For the example case with = 0.865, the theoretical
static plasma confinement experiments 5 consisted of value of Ic/I = 6.9. Thus, for a highly open grid the circu
concentric spherical wire mesh electrodes, which accel lating current can be about an order of magnitude
erated charged particles inwards to produce a virtual greater than the grid current.
electrode inside the central sphere Figure 1. Charged Lavrent’ev and coworkers8 studied ion generation by
particles of the opposite sign would be attracted into oscillating electrons and ion focusing with spherical
the virtual electrode. For example, if electrons were in grids. Probe measurements inside a 5 cm radius spherical
jected into the sphere, they would produce a virtual cathode indicated a strong focusing of injected ions
cathode6, which would attract positive ions. In order to down to radii < 1 cm. The focused ion current density
minimize grid current and heating, it is desirable to have was 1225 mA cm2 at 4 kV in 0.13 Pa of argon.9 ,10 ,11 A
the grid mesh nearly transparent to the charged particle passing ion density of 1017 m3: was deduced from cur
flows. rent probe data.12

Hirsch13 injected electrons into a spherical wire mesh
anode, and used an electron beam probe to measure the
potential well depth produced by the virtual cathode. Cathode
He found that deep, negative potential wells were easily
produced and maintained. He also observed some oscil
lations of the discharge at frequencies 100 MHz.
For the case of electron injectio to produce a virtual
cathode, there are two great diculties: the very high
currents required14 and grid wire overheating. Better
Virtual
confinement could be obtained for the case of ion injec Anodes
tion into the accelerating grid, producing a virtual an
ode.
Rockney15 simulated the formation of a virtual electrode
in cylindrical geometry by solving Poisson’s equation for
Ion
a case with only ions, represented by 2000 charged rods,
Trajectory
and showed that the virtual electrode is stable for long
periods of time. In a simulation with both electrons and
ions, however, Barnes16 found that electrons trapped in ions
the virtual anode tend to wash out the anode potential, (r)
so that the virtual anode gradually disappears. Only in
cases with very high ion currents did the virtual anode
persist. He concluded that the electrostaticinertial electrons
confinement device, as originally conceived, was unlikely
to confine a fusion plasma adequately. Porter and Kle
vans17 investigated the stability of electrons flowing
among ions within a virtual anode, and concluded that
the electrons are marginally, stable to the twostream n(r)
instability.
1.1.2 Multiple Well Hypothesis

Hirsch18 observed high DT neutron yields up to 2 x 1010
s1 from a steadystate device using ion injection from
six ion guns. These neutron yields exceeded the predic
tions of simple theories. One hypothesis to account for r
the observed neutron yields is the formation of multiple,
concentric spherical potential wells of alternate sign Figure 2 — Multiple virtual electrodes inside the accelerating grid
inside the central sphere, with much higher circulating
currents in the inside layers, as illustrated in Figure 2. The following observations appear to be consistent with
According to such a model, ions trapped in interior lay the multiple potential we hypothesis:
ers could have high energies, circulating currents and i High neutron yield;
densities, and could produce significant fusion energy
yields. Hu and Klevans19 modeled multiple wells for ii Peaks in the curve of neutron yield versus radius
cases with zero angular momentum perfectly radial mo measured by Hirsch18 as shown in Figure 3;
tion. They calculated the attainable current amplifica
tion ratio the ion current inside virtual anode/ion beam
current outside virtual anode, and found that values as
high as 150 were possible if large trapped ion popula
tions could be confined in the outer region.

50 iv A large timedelayed pu of gas after beam in
jection cuto measured by Hirsch18;
v Theoretical studies by Cherrington et al 20 for
cases with injected beams having narrow
40 spreads of energy and angular momentum show
Relative Neutron Intensity
a slight double well, as illustrated in Figure 5;

Y
30
1
20
0.5
10
0
0 -1 -0.5 0 0.5 1
X
-2 -1 0 1 2
Radial Position Figure 5 — Theoretical variation of dimensionless electrostatic potential Y versus dimen-
sionless radius X, for a case with applied voltage = 500 V, current = 125 mA, grid trans-
Figure 3 — The results of a neutron collimation study at -90kV, 20 mA and p = 1 Pa parency = 90%, and very small spreads of energy and angular momentum. A double well
(deuterium)18 is apparent.
iii Peaks in the curve of electron bremsstrahlung vi Electron beam probe studies by Swanson e
yield versus radius measured by Hirsch18 as al 21,22 of a spherical electrostatic confinement
shown in Figure 4; device at low pressures;
1.4 vii Peaks in the curve of electron density versus
radius Figure 6 measured by laser heterodyne
studies in a cylindrical electrostatic confine
1.2 ment device by Meeker and coworkers.23 ,24 This
figure is for the case of electron injection; simi
lar, less pronounced peaks were observed during
1.0
ion injection.
Relative Intensity
0.8
0.6
0.4
0.2
0.0
-12.5 -6 0.0 6 12.5 19
Position With Respect to Cathode Vertical Axis, mm
Figure 4 — Electron bremssstrahlung yield versus radial position, as measured by

Hirsch18.

6 they were unable to reproduce the neutron yields ob
served by Hirsch. Baxter and Stewart30 predicted neu
tron yields close to those observed in the Hirsch ex
periment by analyzing ionization and charge exchange in
5 the ion beam, without invoking the multiple potential
Electron Density 1017 m-3
well hypothesis.
Peterson and Oleson31 studied hemispherical focusing of
4 a lowenergy ion beam in a doublebeam plasma device.
When the ion beam energy exceeded Tc the beam ions
were slowed down by a potential hill virtual anode at
3 the focus. They observed formation of a dense group of
cold electrons 0.8 eV, which appeared to ride along
with the ion beam, in addition to the background
plasma electrons 4.5 eV and primary electrons from
2 filaments. Such cold electrons could tend to wash out
potential wells.
Nadler et al 32 built a collimated solid state proton detec
1 tor to observe protons resulting from DD fusion reac
-10 0 10 tions in a spherical electrostatic confinement device.
Radius, mm The energy shifts produced by aluminum and lead foils
verified that they were detecting 3 MeV protons. At 12
Figure 6 — Electron density versus radius measured by Meeker et al using a laser het-
mA cathode current, 30 kV voltage, in 0.5 Pa of deute
erodyne interferometer during electron injection into a cylindrical grid with radius 40 mm,
at 16A circulating current, p = 1 Pa (deuterium)24.
rium fusion reactions between the focused ion beam and
the background plasma were dominant. In future ex
Other studies, however, have yielded results less indica periments at higher currents, the higher proton count
tive of multiple potential wells. Imel25 and Black and rates should permit improved angular resolution, which
Klevans 26 developed theoretical models of the electro could facilitate the detection of multiple potential wells.
static potential profile and electron and ion density pro
files, using distribution functions tailored to various ex
perimental conditions. They analyzed three cases27
1.1.3 Regimes of Operation
Miley et al33 distinguished three regimes of ion injection
i Ion beam injection at very low pressures the
into an inertial electrostatic confinement IEC device:
Hirsch experiment. They concluded that mul
tiple deep potential wells would probably not 1 A discharge mode. In this mode a self~sus
form under the conditions of the Hirsch ex taining plasma discharge like a glow discharge
periment. generates ions throughout the chamber volume.
Free electrons are produced by ionization and
ii Ion injection at higher pressures, where charge
by secondary emission from grids. This mode
exchange produces a broad spread of ion ener
requires comparatively high gas pressures >1
gies several experiments. They predicted a
Pa, which result in substantial charge exchange
central plasma density close to the value meas
rates.
ured in the Penn State experiment 2 x 1015 m3,
and found that substantial potential wells could 2 An emitter mode, based on a patent by
be formed for p < 0.13 Pa 1 mTorr. Hirsch and Meeks.34 In this mode primary elec
trons from thermionic cathodes orbit in and
iii Electron injection at low pressures the Swan
out of an intermediate grid, ionizing back
son et al and Black and Robinson experiments.
ground gas, at pressures typically 0.11 Pa. Ions
They predicted deep potential wells for high
produced by ionization are then accelerated
perveance cases, as observed by Swanson et al,
into the central cathode. This mode was used in
and shallow potential wells for low perveance
experiments by Meeker and coworkers24,35 to
cases, as observed by Black and Robinson.28
measure density profiles. Using this mode with
They did not report any cases with finite angular mo deuterium gas, Miley et al33 observed steady
mentum and multiple potential wells. state neutron yields of 105 neutrons per second.
Gardner and colleagues29 continued the ion beam ex 3 A beam injection mode. Highenergy ion
periment of Hirsch and used a microwave cavity to beams are focused into the central region, to
measure the number of free electrons in the plasma, but produce a highdensity virtual anode. This

mode, which can operate at the lowest pres
sures, was used by Hirsch to obtain high neu
tron yields. This mode is most suitable for fu
sion reactors, because the low pressure mini
mizes ion energy loss by charge exchange. +
- -
+ +
1.1.4 Reactor Prospects
For steady stat operation with radiative cooling of a
Figure 7 — Cutaway view of a few grid wires with currents flowing out (+) and into (-) the
tungsten grid intercepting 1 of the ion current and plane of the drawing. The ellipses represent magnetic field lines.
focusing the ions to a virtual anode at radius ro = 0.01 R,
Lavrent’ev 36 found that a grid radius R > 10 m would be
needed in order to achieve a net energy output. For the 1.2 Open Magnetic Confinement
case of a pulsed reactor the required grid radius is given
by Systems
MinimumB magnetic confinement systems oer good
r
R > 0.001 p A3 0 / 4d 3 MHD stability, but suer from rapid plasma loss along
R magnetic field lines.38 Confinement times in simple
magnetic mirror cells are typically of the order of ion
where is the fraction of current intercepted by the ion scattering times, and confinement times in magnetic
grid, p is the pulse length, d is the grid wire diameter cusps are tens of transit times, which typically makes
and A is the applied voltage kV. For a case_ with =
0.01, ro/R = 0.01, A = 100 kV, p = 0.1 and d = 0.001 m,
( fusionÿpower )
Q 1
this equation gives R > 2.5 m. (inputÿpower ) 4
It is also possible to focus the injected ions temporally for ordinary mirrors and cusps. In order to attain Q > 10
by increasing the voltage during injection.37 Ions starting without requiring very large size or extremely high mag
out later are accelerated more and catch up with the netic induction, it is desirable to find a means for plug
early ions, so that the ion density at the focus is much ging end losses from magnetic mirrors and cusps. Many
higher than the steadystate value for continuous injec means of Qenhancement for open magnetic confine
tion. For example, starting with background plasma at ment systems have been studied, including tandem mir
radii between R = 1 m and 2 m with density o = 1018 m3, rors, multiple mirrors, field reversed mirrors, Tormac,
the focused density would be = 6 x 1025 m3. The con Surmac, plasma rotation, gas dynamic traps, radio fre
verging ion beams could also be used to compress and quency plugging and electrostatic plugging.39,40 Axi
ignite inertial confinement fusion target pellets. symmetric cusps have been suggested as electrostatic
Although purely electrostatic plasma confinement might end plugs for tandem mirrors.41 ,42,43
be attractive for a fusion reactor with a large radius or Several cusped magnetic fields have been considered for
short pulse length, overheating of the grid wires is a se plasma confinement,44 as illustrated in Figure 8. The
rious concern. One way to reduce the heat load from toroidal multipole shown in Figure 8 has N = 6 cusps
charged particle bombardment of the grid wires is to hexapole. In addition to these configurations, arrays of
shield them magnetically by passing high currents point cusps on the surfaces of polyhedrons have been
through them, as illustrated in Figure 7. If electrons are studied by Keller and Jones45 and by Sadowski.46,47 ,48 ,49,50
injected between the grid wires, very few will contact All of these cusp geometries have been considered for
the grid wires directly. Some electrons will gradually be electrostatically plugged devices.
come trapped in the magnetic field and diuse to the
grid wires, but this process is orders of magnitude
slower than direct bombardment. This magnetic shield
ing of grid wires, here called MEPC, has also evolved
from open magnetic confinement systems.

Spindle Cusp Cusp-Ended Solenoid cusps is V = rp2L1 where rp is the eective plasma radius,
and the loss area Al0ss 4r. For this case
X X X
rp2 L1
point
cusp
point
cusp = 6
X X X X
uR
X
line cusp
ring cusp This equation, which estimates the plasma loss time by
(line cusp) freeflowout ring cusps with an eective loss width 2
in each ring cusp, will be used later to estimate the eec
Toroidal Set of Ring Cusps Toroidal Multipole Cusp tiveness of electrostatic plugging.
X
X
X X
Lavrent’ev and colleagues5,,55 proposed electrostatic
X
X
plugging of cusps to reduce the loss rate of plasma flow
X
ing along magnetic field lines out of the cusp gaps, and
conducted early experimental studies of this technique.
X X
If the plasma potential is highly negative relative to the
walls, the ions will be confined electrostatically. Only
axis all line cusps electrons will be energetically able to pass through the
Spherical Cusp
narrow cusp gaps plus a few ions in the highenergy tail
Linear Arrangement of Ring Cusps
X
of the Maxwellian, so the width of the untrapped elec
X
X
tron stream flowing through the cusp gap may be re
X X X X
X X duced to ~ p. Cathodes outside the cusp gap reflect

point X
cusp X escaping electrons back into the plasma. Such electro

X
axis X
line
X
X
X X X
X statically plugged cusps are equivalent to magnetic
cusps X shielding of the grid wires of an electrostaticinertial
X
confinement device. In electrostatically plugged cusps,
Figure 8 — Magnetic field lines (smooth curves) in some cusped geometries. The symbols
the ions no longer flow freely out of orifices, so the free
x and o represent current into and out of the drawing, respectively.
flow confinement time estimates do not apply. Instead,
the ions are confined electrostatically by the potential
For plasma confinement by magnetic cusp fields, the well of electron space charge, and electron confinement
loss area through a ring cusp is approximately 2R2, is limited by diusion across the magnetic field.
where R is the ring cusp radius and the eective half Moir discussed the use of electrodes to contain warm
width of the cusp gap through which plasma is lost. In a plasma for stabilizing open magnetic confinement sys
spindle cusp, the loss area through the point cusps is tems, such as a toroidal quadrupole cusp.56 The augmen
comparable, so the total loss area Aloss = 8R. The ion tation of toroidal magnetic confinement by strong elec
flux into the loss regions is approximately 0.25nu, where trostatic fields was studied by Daugherty et al57 and by
u is the ion speed, so the confinement time in an un Stix.58 ,59,60 Jones61,62 suggested the use of electron or ion
plugged magnetic spindle cusp device is approximately injection to create radial electrostatic potential varia
tions, instead of using plugging electrodes.
nV V
= 5
0.25nuAloss 2 uR
where V is the plasma volume. This time is typically tens
of bounce times back and forth inside the central field
free region. Spindle cusps have very small plasma vol
umes. If pi then unfeasible large R and B would be
required for a reactor. If ppi1/2 a hybrid gyro
radius, as indicated in some cusp experiments and in a
computer simulation,51,52,53 then a reactor using a ‘picket
fence’ cusp such as the spherical cusp and linear set of
ring cusps in Figure 8 is feasible, but high magnetic
fields and large radii are needed.54
For a linear set of ring cusps Figure 8 with distance L1
between the ring cusps, the plasma volume between

gun injection, by laserpellet heating, by neutral beam
2. Theory injection or by other means. Electrons exiting the cusps
are reflected by the negative voltage of the cathode, but
the exiting ions are not initially confined. Since elec
trons are lost less rapidly than ions, the plasma develops
2.1 Electrostatic Potential a slight negative charge, and the plasma potential be
Magnetic electrostatic plasma confinement, which may comes negative relative to the anodes the full curve of
be accomplished with any magnetic cusp configuration, Figure 9. If the anode gaps are narrow enough that their
is illustrated in Figure 9 for the case of a toroidal quad potential is not entirely shielded out by the plasma in
rupole cusp. Highvoltage electrode rings are placed in the anode regions, then the potential there will be near
each of the cusp gaps, as shown. The anodes are biased the anode potential and higher than the plasma poten
positive, and the cathodes negative. Operation of the tial, forming a potential hill i for the remaining ions.
confinement system depends mainly upon the voltage Only a slight fractional charge imbalance is required to
A applied between the cathode and anode, and little set up a potential hill many kV high. Then only ions
upon the relative location of the ‘ground’ potential. with kinetic energies greater than i can escape along
From Figure 9 it is evident that the magnetic field lines out of the cusps. The two
dimensional potential distribution Figure 10 is saddle
A = e + i + 7 shaped in the anode regions, with the saddle point lying
an amount below the anode potential.
This potential shape is the same as that of a simple tan
dem mirror. Plasma
Cathode
4 mm
Coil Anode Coil
Anode
Magnetic
Field
Anode
Plasma
Cathode
0
-1
4 cm -2 kV
Coil Coil -3
-4
-5

Cathode
i
(x) Ions Figure 10 — Hypothetical equipotential surfaces in the anode region, showing saddle,
A shaped electrostatic potential distribution.
The potential hill for ions approaching the walls at an
e
Electrons ode potential is even higher, so most ions will go out
through the cusp gaps as soon as they acquire enough
energy to overcome the barrier i; no ions acquire
Figure 9 — An electrostatically plugged toroidal quadrupole magnetic cusp and the resul- enough energy collisionally to surmount the barrier i +
tant distribution of electrostatic potential . The broken line represents the potential with and reach the walls. The ions are electrostatically
no plasma present.
confined in a negative electrostatic potential well pro
In a vacuum, the interior of the device will be near the duced by a slight charge imbalance. The ion Larmor ra
anode potential broken curve of Figure 9. Plasma may dius is less significant than usual, because the ions are
be produced by electron beam injection into low electrostatically reflected in the boundary layer. The
pressure gas, by rf radiofrequency heating, by plasma maintenance of a potential dierence along the mag

netic field lines is facilitated by the boundary condition boundary, 2 = passing through anodes and reflected by cathode, 3 = magnetically trapped
in boundary layer, 4 = electrostatically trapped by anode potential well.
that the walls are very close to the plasma only in the
anode gaps. Some energetic ions are replaced by cold
ions as a result of charge exchange with neutral gas.
2.2 Electron Orbits
The central plasma region is uniform in density and
There are several types of electron orbits, illustrated in
temperature, free of electric and magnetic fields, and
Figure 11:
surrounded by a thin edge layer. The outside boundary
of the edge layer is the magnetic surface along which 1 Electrons that stay in the central plasma and are
electron orbits graze the anodes, and the inside bound reflected geometrically from the convex edge
ary is the uniform central plasma region. The boundary layer.
layer thickness at any location with magnetic induction
2 Electrons from the central plasma that can pass
B is found by magnetic flux conservation to be
through the anode gap regions and be reflected
RB by the cathodes.
r a a 8
rB 3 Electrons that are magnetically trapped in the
edge layer. Many of these are mirrorreflected
where a is the anode gap halfwidth, R is the ring cusp away from the cusp gaps.
anode radius and r the radius of the boundary layer. If Ba
4 Cooler electrons that are electrostatically
= 5T, B = 1T, R = 4 m, r = 3 m and a = 1.5 mm then r = 10 trapped by the anode potential. Most of these
mm. This thickness is much less than the plasma size, so electrons are produced by ionization of incident
the plasma has a ‘sharp boundary.’ neutral gas atoms along the slope of the poten
2a tial well. These trapped electrons are deleteri
ous, because they increase the selfshielding
voltage drop .
Cathode Electrons may be produced by thermionic emission

from an electron gun, by secondary emission from the
plugging cathodes and by ionization of neutral gas.
Much of the injected electron beam is reflected by the
plasma and lost to the cathodes. This portion of the in
jected beam does not contribute to plasma density
buildup. Yushmanov63 found that only a small fraction
of the injected beam current is eectively trapped. The
time required for plasma density buildup by electron
beam injection many particle confinement times ex
Anode h ceeds the duration of the flattop magnetic field in some
pulsed experiments.
Many of the electrons produced by ionization will be
magnetically trapped in the edge region. The untrapped
4 electrons will circulate back and forth through the
plasma, occasionally passing out through the anodes and
being reflected by the potential barrier . Coulomb
collisions and electric field fluctuations cause diusion
in velocity space. Trapped electrons can become heated
and detrapped, and free electrons can be deflected and
3 trapped either magnetically or electrostatically.64
2
r There are two main electron loss processes: trapping by
the magnetic field with subsequent diusion to the an
odes, and collisional diusion in velocity space over the
potential barrier with loss to the cathodes. Loss to
1 the cathodes converts electron kinetic energy into elec
tricity, returning energy to the power supply. Recombi
Figure 11 — Electron orbits near a magnetic cusp. Smooth arcs represent magnetic field nation with ions is negligible.
lines. Dots represent ions. Electron trajectories are as follows: 1 = reflected off plasma

Parks and Sleeper 71 derived an equation predicting ion
2.3 Loss Rates over end loss rates in the transition region between the non
Potential Barriers adiabatic region central plasma and the adiabatic re
gion boundary layer. In the nonadiabatic limit. their
The loss rates along the magnetic field by velocityspace equation agrees with the result of Yushmanov 72 and in
diusion over the potential barriers have been calculated the adiabatic limit, their equation is similar to the re
by Cohen et al,65 based on the method of Pastukhov.66 sults of Chernin and Rosenbluth,73 but with loss rates a
For a plasma with singly charged ions, the resulting par factor of 34 higher, due to eects of magnetic field cur
ticl loss rates along the magnetic field are vature.
dnk 4nk Z k exp(yk )I(1 / yk )
= 9
2.4 Self-Shielding
dt 1/2 yk k G(RZ k )
The uncompensated electronic space charge tends to
1 shield out the anode potential by an amount . As
1+ + 1
1 RZ k suming azimuthal symmetry, the electrostatic potential
G(RZ k ) = 1 + ln 10 distribution can be calculated from a twodimensional
RZ k 1
1+ 1 Poisson equation in the anode gap
RZ k
1 r 2 (n ni )e
+ = 13
r r
r
z 2 0

I(1 / yk ) = 1 + 0.5 exp(yk )[1 erf yk ]
yk Due to the potential hill i, only a few ions are passing
11 through the anode gaps, so 1a << , as will be discussed
1 1 3
= 1+ 2 + 3 + ... later. With comb probe measurements in some experi
2yk 4yk 8yk ments it was found that the distribution of electrons in
the anode regions had bellshaped distributions across
where the subscript denotes either electrons or ions, the magnetic field. In order to model a variety of condi
is the particle density, Z = 1 for the electrons and 1/2 tions, a Lorentzianshaped density profile
for ions, R is the magnetic mirror ratio, is the charac
na
teristic collision time for electrons or ions, and y = n= 14
ek/T where is the charge number = 1 for singly (1 + z 2 / w 2 )
charged ions, = 1 for electrons, is the electronic
charge and k is the potential barrier for electrons or is assumed, where a is the electron density at centre of
ions here and e are, taken to be positive quantities. the gap and the parameter characterizes the half
width of the distribution. Assuming this electron den
The energy loss rates along the field are given by sity distribution in the anode region, with a = 1.5 x 1017
m3, a = 2 mm and = 0.2 mm, results in the electro
d(nkTk ) 1 1.5 dnk
1.5 = zk e k + 12 static potential distribution of Figure 10. The depth of
dt
= I(1 / yk ) yk dt
= the saddle point relative to the anodes is called . If
the anode length along the magnetic field h >> a, then a
A similar estimate of loss rates was made by Sizonenko onedimensional equation
and Stepanov.67 These equations give the rate at which
particles or energy are lost by electron or ion diusion in d 2 (n ni )e
= 15
velocity space over their potential barriers e and i. In dz 2 0
view of the expy factor in the particle loss equation,
the confinement time should increase roughly exponen can provide an accurate calculation of the potential dis
tially with y. Computer simulations by McHarg and tribution ϕz across the magnetic field. Assuming ϕa =
Oakes 68,69 show such an exponential increase of parallel ϕa = 0, solution of this equation yields
confinement time with k/T. In the limit of high
plasma density or low temperature, collisions may be so n ewa 2 1 a w a2
frequent that these ‘Pastukhov’ equations lose validity, = a
tan ln 1 +
and the equations of Rognlien and Cutter 70 should be 2 0 w a w 2
used. In MEPC devices the dominant electron losses are 16
by diusion across the magnetic field, but ion confine
ment is almost completely electrostatic. Since highenergy electrons flying through the anode
gap have trajectories extending one or two Larmor radii

on each side of the center, it is assumed that w 2p.
2.5 Electron Density in
The resulting values of ϕ are shown in Figure 12 as a
function of a and B, assuming T = 20 keV and a = 2 Anode Gaps
mm. If a = 1019 m3 and B = 6T, then ϕ = 74 kV. Five phenomena aect the ratio a/: ions in the gap,
magnetic reflection, electrostatic acceleration, cold
300
trapped electrons and diocotron oscillations.
,kV na = 4 x 1019 m-3
2.5.1 Ions in the Gap
200 Assuming a quasicylindrical plasma with radius rp and
4 length L, the rate at which ions are lost is
rp2 Ln
2 = N 2 R2 niauia 20
100 i
1
0.5 where is the central plasma density, i is the ion con
finement time, N is the number of ring cusps, R is the
ring cusp radius, is the eective halfwidth of cusp
2 4 6 8 10 gaps for ions, ia is the density of ions flowing over the
B, Tesla barrier and uia is the speed at which ions pass over the
barrier. Ions just barely leaking over the potential barrier
Figure 12 — Variation of self-shielding potential sag with magnetic induction, for various will have low velocities uia 3 x 104 m s1, and they will
anode gap peak electron densities na, assuming w = 2pe, a = 2 mm, T = 20 keV. be restricted to a narrow corridor of halfwidth .
At equilibrium the ion particle confinement time will
These could be typical parameters for a MEPC reactor.
equal the electron confinement time, which is roughly
A comparable result was found by Ware and Faulkner,74
assuming a triangular density distribution. The quantity equal to the diusion time di across the magnetic field.
in brackets varies from 0.96 at /a = 0.01 to 0.67 at /a Thus, the ion density in the anode gaps is estimated to
= 0.2, so the approximation be
n ewa nia rp2 L1

a 17 = 21
2 0 n 4Rwuia diff
gives a slight overestimate of ϕ. Expressing Te in keV where L1 = distance between ring cusps. For the example
and ϕ in kV, this becomes MEPC reactor parameters of table 1, assuming L1 = 1.5
m, it is estimated that ia/ 0.02.
Tek na a
= 6x10 15 18
B 2.5.2 Magnetic Reflection
Sidorkin and Lavrent’ev75 studied electron reflection by
which yields ϕ = 89 kV 20 high for the above ex
multipole cusp gaps, as illustrated in Figure 13. They
ample case. The maximum electric field occurs at the
showed that electrons with a > cri are reflected by
anode wall
the cusp field, where
d
Emax =
dz z=a sin( crit ) =

(1 + x ) x
2m
0
m1
(1 + x ) x
2m m1
22
19
n ew a
0
= a
arctan

0 w a
m 1
1/2 m
x0 = 23
For typical values of MEPC experiments, the electric m + 1
field energy density 1/20Emax2 << B2/2o.
x = r/R, r is the initial electron radius, R is the cusp gap
radius and m is the number of pairs of multipole con
ductors = N/2. The values of cri are plotted as functions

of initial radii in Figure 14, for various values of . In Table 1 — MEPC Reactor Parameters
multipoles with large i the electrons have smaller critical
angles, so more of them are magnetically reflected. Example
I The plasma parameters were estimated using approxi
I mate scalings of attainable density, temperature, con
y finement time section 3.7.1 and Q section 4.1.
Ring cusp magnetic induction, Ba = 6 T
+
Applied voltage, ϕA = 400 kV
Plasma radius. rp = 3 m
r Ring cusp anode radius, R = 4 m
+ Ring cusp anode gap halfwidth, a = 2 mm

Estimated Parameters Equations
x Central plasma density, 57 and
1020 m3 Jupiter2M
+
Electron and ion temperatures, T, 48, 49, 58
Ti 20 keV
z +
Magnetic field at plasma boundary, after 37
Bb 1.3 T
Figure 13a — Current carrying conductors of a multipole cusp. In this figure N = 8, m = 4.
Electron Larmor radius in anode after 37
gap, pa 8 x 105 m
y, cm
Electron Larmor radius at the after 37
plasma boundary, pb 3.8 x 104 m
Selfshielding voltage drop, after 47

ϕ 80 kW
Anode gap peak electron density, after 16

a = 1.3 x 1019 m3
Energy confinement time, 6 s after 40

R
=
5 Power gain ratio, Q 10 after 60
cm
1 Electrons starting o at r/ R = 0.5 near an octupole cusp
= 4 have xo = 0.938 and cri = 29°. Experimentally
measured electron beam transmission is consistent with
-3 -2 -1 0 1 2 3 x, cm such theoretical predictions. Assuming that the central
plasma electrons have isotropic distributions, the total
Figure 13b — Definition of the electron trajectory angle of incidence relative to the mid-
plane of the cusp gap. In this figure N = 8, m = 4.
loss fraction through N cusps from a given location is
approximately
N
total loss fraction 0.5 1 cos( crit ) j 24
j =1
If p = density of electrons that are able to penetrate

into the magnetic cusps, the ratio p/ can be esti
mated from the volumeaveraged value of this total loss
fraction, but this calculation has not yet been per
formed.

2.5.3 Electrostatic Acceleration 2.5.5 Diocotron oscillations
Electrons that succeed in entering the cusp gaps are ac The electric field due to electron space charge is zero at
celerated by the positive applied voltage there. Moir e the middle of a ring cusp gap, increasing on each side.
al 76 showed that the The resultant E x B drift velocity varies spatially, and the
shear of this velocity gives rise to the diocotron slip
90 stream instability.78,79 Computer simulations by Levy
80 and Hockney 80 showed good agreement with theoretical
instability growth rates. Pankrat’ev et al 81 found that the
70 longwavelength instability > 10, where is the elec
m=2 tron stream halfwidth occurs when the electron drift
60
velocity is near the wave phase velocity, and it is sup
crit 50 pressed by a nearby conducting wall. However, Gordi
degrees enko et al 82 found that the presence of two groups of
40 m=4 electrons trapped and passing in the anode gaps broad
30 ens the unstable range of wavelengths and makes dioco
tron oscillations unstable whenever there is a vacuum
20 m=8 gap between the electron stream and the anode wall.
10 The shortwavelength instability < 0.2a occurs near
the electron cyclotron frequency,
0
0.2 0.4 0.6 0.8 1 q
r/R exp 2 / q
ce 2
26
Figure 14 — Variation of critical reflection angle with radial position from which the elec-
tron trajectory starts, for various multipole orders. Smooth curves are theoretical, and full and has a growth rate where q = p2/ c2. These dio
points are experimental data for m = 4. cotron oscillations play a beneficial role by removing
cold trapped electrons from the anode region, where
electron density in the gaps is lower than the magneti they would broaden the width of the electron density
cally penetrating density p by a factor
distribution and increase ϕ. The instability growth rate
1 3 15 is small if q < 0.2, which is equivalent to the condition
1 + 2 3 + ... that
na 2y 4y 8y
= exp(y)erfc( y ) = na (2 x1018 m 3 ) Ba2
np y 27
25 where Ba is the magnetic induction in the gap T. Thus,
as a decreases well below this limiting value, the short
where y = ϕi/T, and erfc is the complementary error
wavelength diocotron instability turns o. At Ba = 5 T,
function.77 For values of ϕi/T = 3 to 6, this ratio is a/p this limit is a 5 x 1019 m3, which is less restrictive
= 0.29 to 0.22. than the value of a 1.4 x 1019 m3 required to keep ϕ
< 100 kV Figure 12.
2.5.4 Cold Trapped Electrons Yushmanov83 studied the buildup and removal of
Magnetic reflection and electrostatic acceleration tend trapped electrons in the anode regions, and determined
to reduce a/, but electrons produced by ionization of the required anode dimensions. The width h of the an
neutral gas tend to become electrostatically trapped in ode Figure 11 should be in the range
the anode region orbit 4, Figure 11, increasing a/. If
n B
2
classical crossfield diusion were the only loss process, 5 a h hmax 6.7 t a2 a R
then the accumulation of trapped electrons could in N
na
a
Bb
28
crease a and ϕ to unacceptably large values, but the
diocotron instability may help remove cold trapped where 4.7, = trapped electron pumpout rate/
electrons without seriously impairing hot electron con diusion rate 1/3, N is the number of cusp gaps and
finement. /a is the trapped electron density/anode electron
density 1/3. Taking N = 6 in the example case of table 1,
it is found that

hmax 4 cm. Lowfrequency oscillations can selectively and diocotron oscillations, there will probably be about
pump out the trapped electrons without seriously in an order of magnitude density ratio between the anode
creasing the loss rates of the free plasma electrons. region and the central plasma.
If the electrons in the central plasma have a quasi na
Maxwellian distribution, the distribution will be up 0.1
shifted in energy by an amount eϕi as the electrons enter n 29
the anode regions, as illustrated in Figure 15. Electrons The ratio would probably be lowest for cases with a
produced by ionization in the plasma edge region will large volume of fieldfree plasma, narrow anode gaps and
have a spread of energies. The diusion rate is high for low neutral gas pressure; and it could be near unity for
lowenergy electrons, due to interaction with diocotron cases with spindle cusp magnetic field, wide anode gaps
oscillations, and it may be nearly classical D W1/2 and high neutral gas pressure. Values inferred from data
for highenergy electrons. The resulting electron energy in a variety of experimental conditions range from 0.01
distribution in the anode regions is expected to resem to 1.
ble the shifted Maxwellian distribution of Figure 15
middle.
.5
2.6 Plasma Potential
.4 At T > 1 keV and < 1020 m3 recombination is negligi
Central Plasma
.3 ble. Ions are confined electrostatically with energy loss
f(W) by charge exchange and by diusion in velocity space
.2 over the potential barrier. The ion and electron particle
.1 conservation equations may be written:
0
0 2 4 6 8 dni dn
Si i
.5 dt dt
30
.4 Anode Region dn I dn n
.3 Si
f(W) .2 Ionization dt eV
dt 31
.1 where Si is the ionization source, I is the electron cur

0 rent from cathodes into the plasma and is the elec
0 2 4 6 8 tron loss time by diusion across the magnetic field. At
equilibrium the plasma potential will adjust itself so that
.5 these electron and ion loss rates are equal85
.4 Relative Diffusion Rate,
I dn n dni
Anode Region 0
.3 eV dt
dt

f(W) 32
.2
Since
.1
e A i 33
0
0 2 4 6 8 these two equations can be solved simultaneously 86 for
W/kT the two unknowns ϕ and ϕi using equation 9 for the
Figure 15 — Maxwellian distribution of central plasma electrons (top): shifted Maxwellian parallel loss rates. The plasma potential distribution has
distribution in anode region, assuming i = 2Te (middle); relative diffusion rate in anode been computed selfconsistently in some two
region, with effect of diocotcon oscillations on cold electrons (bottom). dimensional models.
Some magnetically trapped electrons have low enough

energies that they can immediately interact with dioco 2.7 Two-Dimensional Models
tron oscillations; but most of them, having been acceler
Brunel et al87 solved a twodimensional r, Poisson
ated by the anodes, must be decelerated by collisions
equation, assuming nearly Maxwellian distribution func
before the diocotron oscillations interact strongly with
tions dependent on the constants of motion. They cal
them. For one example case 84 the trapped electron frac
culated the distribution of potential and particle densi
tion was estimated to be /a 0.24. Considering ions,
ties in an electrostatically plugged spindle cusp corre
magnetic reflection, acceleration, cold trapped electrons
sponding to the KEMP device. They found that the

electron and ion densities were both maximum onaxis. 2 ei aV
The electron density had a secondary peak in the ring diff
cusp anode, and the ion density was very low in the an
Sb a 37
ode regions. Matte and Lafrance 88 developed an im
proved numerical method for solution of the Poisson where ei is the electronion momentumtransfer colli
equation in MEPC devices. sion time, V is the plasma volume, S is the plasma sur
face area, ρb = 2mkT1/2/eBb is the electron Larmor ra
Shoucri et al89 simulated an MEPC plasma with about dius at plasma boundary with Bb = 2μ0nT + Ti 1/2, ρa =
5000 electrons and ions M/ = 16, where M is the ion
2mkT 1/2/eBa is the electron Larmor radius in the anode
mass and the electron mass in a 128 x 50 cylindrical
gap and = a parameter varying from 1.33 at
grid, using a capacitive matrix Poisson solver, with pa
Ti/T = 0 to 4.6 at Ti/T = 1. A similar estimate was de
rameters similar to those of the KEMP II experiment.
The simulation showed the maintenance of a two rived by Dolan et al.92 Assuming ln 16 in a hydro
dimensional electrostatic potential well, with saddle genic plasma, the electronion momentumtransfer col
points in the anode gaps, as expected. Electrons were lision time44,93 may be reduced to the form:
very well confined, but ion confinement was limited by 9.4 x1014 Tek3/ 2
plasma selfshielding in the point cusp gaps, especially ei
during electron injection. The confinement time ob n 38
served in the simulation was scaled up by the square
root of the experiment/simulation density ratio to yield where Te is in keV. For spindle cusps, V/SR 0.010.03,
a predicted confinement time of 5075 μs for the KEMP depending upon dimensions. plasma heating and con
II experiment, which was consistent with experimental finement eectiveness. Lowpressure plasma have
measurements. smaller volumes. This low ratio plus diusion enhance
ment by electron collisions with neutral gas give spindle
cusps poor confinement times.
2.8 Energy Confinement Time Assuming classical diusion in a spindle cusp with radius
The nonradiative crossfield electron energy loss time R and length L, Lavrent’ev94 derived analytical expres
may be written sions for the electron density profile, the plasma volume,
the average electron density and the electron confine
1 1 1 ment time. The electron energy loss rate is:

E cond ( diff trap ) 34 R 4 D0 n0 p
Pe i 2Te (cosh( p ) 1)
where cond is the time for heat loss by conduction 1.51 ra2 L sinh( p ) 39
di, di is the characteristic time for diusion across the
where Do is the collisional classical diusion coe
magnetic field and trap is the time for free electrons in
the central plasma to become trapped by the boundary cient, 0 is the central plasma density, ρ = / 1/2. and
magnetic field. ra is the point cusp anode radius. This power loss is
minimized when ρ ~ 5. Lavrent’ev found that the edge
One estimate of the trapping time is90
electric field can cause significant increases of electron
2 mvBb transport rates in. small devices with substantial neutral
trap 8 0 rp gas pressures. For the parameters of the Jupiter1M de
ne 3 ln 35 vice the electric field and electron mobility should result
in about a 45 increase in the electron crossfield trans
where is the electron velocity and ln is the Coulomb port above the classical diusion rate. In the case of a
logarithm. For reactor conditions trap << di, and reactor with ‘burnout’ of the neutral gas, however, the
0.6 diff electron mobility would be small, and the electric field
36 eect would be diminished.
Nearly classical transport has been observed in some Lavrent’ev 95 showed that angular momentum from E x B
MEPE experiments, while rapid anomalous transport rotation produces an eective potential well s that
has been seen in others. greatly inhibits axial ion losses, even if the point cusp
electrostatic barrier i = 0. A similar eect acts on elec
2.8.1 Classical Transport trons, but is /M times weaker. Near the ring cusp the
electric field eects dominate the ion motion. Dol
Pastukhov 91 estimated the classical diusion time to be: gopolov et al96 computed ion trajectories in an electro
statically plugged spindle cusp, and found that the mag
netic field gradient force and the centrifugal force re

lated to the azimuthal E x B drift tend to keep ions where 0 is the electron collision time. This case is simi
from escaping through the point cusps. This inhibition lar to that of the ATOLL experiment, and the resulting
of point cusp ion losses makes plasma confinement ef confinement time is much shorter than with classical
fective in a linear set of ring cusps. diusion. Anomalous transport would spoil the energy
For a linear set of ring cusps, the volumetosurface area balance in a reactor, resulting in a low fusion power den
ratio V/SR 0.10.3 is an order of magnitude better than sity.
or spindle cusps. Assuming V/SR = 0.02 and T = Ti,
equation 37 becomes 2.9 Energy Balance
ei aR
diff 0.09 As neutral atoms impinge on the plasma some are ion
b a 40 ized in the boundary layer. The resultant ions are accel
erated as they fall down the potential hill into the
For the example MEPC reactor parameters case table 1 plasma region. However, Pastukhov91 found that this ion
it is found that ei = 0.84 ms, the diusion time di 20 s heating process is inecient, and that ion heating by
and E 12 s. Since experiments have attained about electrons is very slow. Ions are also heated by plasma
waves. Ion energy is lost by charge exchange and by con
half the classical value, a value of E 6 s is listed in ta
vection, as heated ions escape over the potential hill i.
ble 1. This nearclassical transport would make an attrac
tive reactor, but anomalous transport could prevent a Since ions with energies Wi > i are lost out the cusps,
reactor from being successful. their distribution is truncated at higher energies, and
the resulting fusion reaction rate is reduced.99
2.8.1 Anomalous Transport Electron energy may be supplied by electron beam injec
tion, neutral beam injection, wave heating and heating
Pastukhov and Il’gisonis 97,,98 studied anomalous electron in the edge electric field. Electron energy is lost by elec
transport in the boundary layer of an MEPC plasma. tron flow over the potential barrier to the cathodes, by
When Ti << T, the ion acoustic instability is dominant, transfer to waves, by thermal conduction and convection
giving rise to rapid electron diusion through the across the magnetic field and by inelastic collisions with
plasma edge layer. The longwavelength ionacoustic ions, atoms and molecules.
instability can occur at kp ~ 1 and w ~ cci0.5 when
Approximate energy balance equations for the central
the electron densitygradient drift velocity n > cs. The plasma electrons and ions may be written
corresponding maximum diusion coecient Dmax ~ csp,
which has Bohmlike scaling. The resulting anomalous d (nTe ) Ie d (nTe )
1.5 d PE 0.25 ni2 * v W f e 1.5
transport tends to broaden the electron density profile dt eV dt "!
to a condition where n ~ cs. The resulting theoretical nTe (T Ti )
electron density profile is given by: 1.5 Prad 1.5 n e SmWloss Pw
E eq
n(r ) n0 exp( ecs A / cTe ) 41 43
where o is the central plasma density and A is the d (nTi ) (T T )

1.5 1.5 n e i fi SmWa Pi
magnitude of the magnetic vector potential. The waves dt eq
are partially absorbed by the ions, and may heat them to
the point where the instability turns o. When Ti ~ T, 0.25 ni2 * v ( f iW 1.8 ei )
the ion acoustic instability becomes a Buneman instabil
dn T
ity, and the instability growth rate decreases by a factor 1.5Ti (1 fi )Sm 1.5 1 i
of about /M1/2. Once electrons diuse outside the dt
44
bounding flux surface the surface which just touches
the anodes, they are magnetically mirrorconfined orbit where PE is that portion of electron heating by the edge
3 of Figure 11 until they scatter into the loss cone and electric field which aects the central plasma, eq is the
travel along the magnetic field to the anodes. For a lin electronion equipartition time, Prad is the radiative
ear quadrupole with a/p < 2M/1/2, the anomalous power loss, Wl0ss is the average electron energy loss per
electron confinement time is incident neutral atom due to ionization, P is the elec
tron power loss to plasma waves, fi is the fraction of in
#% a m '% cident neutrals which are ionized instead of causing
$0.4 0.3 exp 2 " ( 0
%& e
M ! %) 42
charge exchange, f and fi are the fractions of fusion
product alpha energy W which are transferred to elec
trons and ions, S is the volumeaveraged neutral input

rate, Wa is the average energy acquired along the slope core is essentially spatially isotropic, but the electrons
of the potential well by ions, <> is the DT fusion re do not have a Maxwellian velocity distribution. The ion
action rate parameter and Pi is the ion heating by plasma flow is expected to remain spatially anisotropic in the
waves such as ion acoustic. It is assumed that the aver mantic region and to retain its nonequilibrium nearly
age energy of each fuel ion undergoing a fusion reaction monoenergetic nature over typical ion lifetimes.
is about 0.9i. Since the alpha particles are poorly
Startup Startup Beam Injector
confined, f<< 1 and fi << 1. Additional terms would be
needed for neutral beam injection, microwave heating,
Through Point Cusps
etc. These equations are nonlinear, because several pa
rameters depend on the density and temperatures. Simi
lar equations have been solved numerically for several R
cases.84 Typical conditions satisfying the plasma poten
tial and energy balance equations were found to be
e
0.5
A 45
1
0.3
A 46

0.2 Dense Plasma
A 47 Core with
Ti 0.05 A Radius rc
48
Te 0.05 A 49
In order to confine plasmas with fusion reactor parame
ters, applied voltages ΦA 300 kV are desirable.100
High-Beta Plasma Operation
Figure 16 — Electron injection through multiple point cusps in a Polywell device. The full
Karpukhin et al 101 studied particle and energy balance in
curves represent magnetic field lines, which are pushed outward by electron diamagne-
an electromagnetic trap with rp ~ 0.3 m, Ba = 13 T, ΦA = tism during high-beta operation.109
2550 kV, V ~ 0.5 m3, a = 1.5 mm and an electron injection
power of 500 kW. They predicted values of ~ 3 x 1019 The degree of convergence of ions flowing radially to
m3, E ~ 0.10.2 s and Ti ~ 0.51 keV. For an octupole elec the central core is determined by angular momentum
tromagnetic trap they found the optimum electrode gap conservation to be
spacing a = 0.51 mm, which maximizes the injection
rc W
eciency, plasma density and confinement time.102 En
ergy balance calculations will be discussed further in R ei 50
section 4.1. In addition to spindle cusps, multipoles and
ring cusps, two additional MEPC configurations have where W is the ion energy perpendicular to the radial
been studied recently: the Polywell concept and the direction and Φi is the accelerating potential.104 The
modified Penning trap. Polywell concept is based on highbeta electron con
finement by the magnetic field. Watrous et al 105 showed
that the required diamagnetic currents could be
2.10 The Polywell Concept achieved with reasonable electron injection conditions,
Bussard103 proposed electrostatic confinement inside a yielding a relatively fieldfree internal region that en
polyhedral array of magnetic point cusps Figure 16, hances electron confinement. Krall et al 106 also studied
called a Polywell, with electron injection at 10100 kV the density and pressure anisotropy profiles resulting
along point cusp axes to provide a quasispherical nega from various ion sources, and assessed processes which
tive potential well capable of trapping and confining may degrade ion focusing, such as angular deflection and
ions. The ions are to be injected at very low energies < energy upscattering. The concluded that isotropy of the
100 eV from the system boundary r = R, along cusp plasma core inhibits the Weibel instability.
axes or the axes of the polyhedron vertices, The electron
distribution in the ‘mantle’ region r > rc outside the

Solving the Poisson equation for spherically convergent densities at the centre in one, two and three
charged particle flows, King and Bussard107 found that, dimensional cases. These oscillations are predicted to be
with a high degree of electron recirculation within the stable at large amplitude.
system with long electron lifetimes, central plasma
By definition, the Brillouin ratio is the ratio of relativis
densities of about 1025 m3 could be attained at high elec
tic electron energy density to magnetic field energy den
tron gun currents. Bussard and King108 distinguished
sity:
two bounding operating regimes: a lowelectronbeta
mirror regime, where magnetic trapping and reflection 2 +0 nmc 2 n
are dominant, and a ‘whieball WB regime’, where a Brillouin ratio
highelectronbeta plasma pushes the magnetic field B2 nB 51
outwards, so that the plasma is situated inside a quasi
where B is the Brillouin density. In an ordinary equilib
spherical magnetic surface with small holes correspond
rium case /B 1, the ‘Brillouin limit’. Turner and Bar
ing to the point cusps. In the WB regime they find that
nes113 found that a class of flows with deviatoric strain
the electron recirculation factor Gj0 representative of
can support charged particle densities with Brillouin
the central plasma electron lifetime increases almost
ratios locally exceeding unity. This configuration does
linearly with crc2, where c is the central electron and
not require largeamplitude plasma oscillations. They
ion density within the core radius rc. They advocated
showed that, in a toroidal geometry with a purely poloi
plasma startup at low magnetic field, in order to avoid
dal magnetic field, local values of the Brillouin ratio /
the mirror regime. which requires higher electron beam
B > 2 could be sustained. The poloidal magnetic field
currents.
could be produced by a toroidal quadrupolar current
Three issues arise with the Polywell concept: configuration outside the plasma, and a toroidal octupo
lar electric field applied outside the plasma could be
1 It may be dicult to achieve a high degree of
used to sustain the required plasma configuration. With
ion focusing small W, due to ion reflection by
a 1/ r term in the toroidal flow velocity, the Brillouin ra
the scalloped electrostatic potential well
tio at the x point the center of the plasma becomes
boundary 109 and ion deflection by waveparticle
infinite. In contrast to the usual Penningtrapconfined
interactions. Larger W would result in a larger nonneutral plasmas, this configuration is not in ther
core radius rc and reduced central plasma den modynamic equilibrium. Large values of the Brillouin
sity c. Further research is needed to clarify the ratio are feasible in a cold plasma when the electron ve
attainable degree of focusing. Even if great fo
locity shear length Ls ~ /p.
cusing were not achieved, the Polywell might
still function as a fusion reactor by using a Barnes et al 114 studied nonneutral plasma confinement
larger radius R. in a uniform magnetic field with an applied electrostatic
quadrupole field a Penning trap. They found a class of
2 Initially monoenergetic electron or ion distri
confinement that does not require a large velocity shear.
butions may be driven toward thermodynamic
The plasma reactivity is proportional to <n2>, where < >
equilibrium approaching Maxwellian distribu
denotes a space and time average. Since the Brillouin
tions by enhanced collisional mechanisms, such
limit applies only in the volumeaveraged sense, the re
as the beamplasma and twostream
activity can be enhanced by inducing large spatial varia
instabilities.110 It would be desirable to prevent
tions of over the plasma volume such as ultrahigh
the accumulation of cold electrons, in order to
avoid the beamplasma instability. at the centre. Lowenergy, lowP canonical angular
momentum charged particles such as electrons or deu
3 The polyhedral set of point cusps has magnetic terons could be injected into this spherical Penning
field nulls on its surface, which might not con trap. The eective parabolic spherical potential well of
fine highpressure plasma well. However, no the applied electromagnetic field would accelerate the
significant particle losses were associated with particles radially inwards to energies ~100 keV. Conver
the null points in the early operation of the gence near the centre r = 0 produces a dense, inertially
spherically convergent ion focus SCIF confined, nonneutral plasma core. Strong focusing oc
experiment.111 curs only for a single chargetomass ratio q/, so a pure
deuterium plasma might be used. Miyamoto el al 115
showed that, with a judicious choice of a rational integer
2.11 Modified Penning Trap determining the electric field shape, the DD fusion re
Barnes and Turner 112 found that a timevarying external action products 3H and 3He could also be focused to
electrostatic quadrupole field in a uniform static mag pass through the central region, enhancing the fusion
netic field could induce largeamplitude, nonlinear oscil power density.
lations of a nonneutral plasma with very high plasma

Another possibility is an initially pure electron plasma, the density varies radially as r aa2/r2, where a is
with deuteriumtritium ions trapped in the virtual cath the radius of spherical bounding surface, and the central
ode. In the electronplasma version low energy electrons core density c Ba2/rc2. Since 2 decreases rapidly for
would be introduced through one of the negative end r > rc, the average of 2 is estimated to be
caps of the electrostatic quadrupole, as illustrated in
Figure 17. In a reference frame rotating at angular fre nc2 rc3 16 02V02
n2 2 3
quency = eB/2, the Poisson equation becomes114 a3 e a rc 54
e This proportionality to Vo/a3rc shows that the reactivity
, 2 - (nB n )
0
52
can be enhanced by high voltages and small radii.
The scaling was confirmed by numerically computing
The applied electric field is produced by electrodes with selfconsistent Vlasov equilibria for an example case
hyperbolic surfaces tangent to the spherical boundary of with a = 3 mm, V0 = 120 kV, electron beam radial injec
the potential well.
tion at W = 50 eV, I = 0.125 A, T = 104 eV and Tr = 5 eV.
B Field The numerical result showed that a density c 4 x 1025
m3 could be confined within rc 0.4 um.114 For compari
Neutral son, the above approximate equations yield rc 0.71 um,
Beam Electron
Plasma Gas Source B 3 x 1018 m3, c Ba2/rc2 5.3 x 1025 m3 and <2>
- 3.7 x 104 m3. Thus, for this example case the approxi
mate scalings yield a core radius and density not far from
the numerically computed values.
A second example case considered the eects of ions
that partially neutralize the electron space charge. As
+ + suming i/ = 0.5 for r < 1 um, i/ = 1106 for 1 um <
r < r0, i/ = 0 for r > ro and I = 0.425 A, similar poten
tial and density profiles were obtained, with rc 0.3 mm
and c 4 x 1025 m3. The neutral background penetrates
most of the plasma, but within a small ‘burnout’ radius
- high ion temperatures may be sustained. The central ion
confinement time is about 0.25 ms, and a neutral gas
Dense Core pressure about 110 mPa would be appropriate. For this
Plasma
example case the predicted central ion temperature was
7.8 keV, and the DT fusion power would be about 50
Electron Electron mW, and the energy gain Q = 103. Calculations indicate
Source
Symmetry Axis Collector that this configuration is stable against the twostream
instability, that Coulomb collisions do not significantly
Figure 17 — Schematic diagram of the electrostatic quadrupole Penning Trap.114
alter the electron beam velocity distribution, and that
Particles with energies <2Vo/3 will be confined, where small angular momenta are attainable.114
the applied voltage Vo = Bea2/4o. Although such a Penning trap would not produce a large
The cold injected electron beam will be reflected near total power output, it could produce useful fusion reac
the centre of the spherical well r = 0. If P = 0, then the tion rates in a laboratory scale device. Higherorder mul
characteristic radius rc of the dense central core could be tipole traps could generate higher power outputs.114,116
comparable to a local Debye length. With finite perpen Tiouririne et al117 studied the use of higherorder multi
dicular injection temperature T the angular momentum pole traps for nonneutral plasma confinement. The to
conservation equation yields tal electrostatic potential Φ is the sum of the externally
applied potential Φapp and the potential Φs generated by
T
rc r0 the plasma space charge. The plasma density, computed
W 53 in terms of Φ using the Poisson equation, was assumed
to be zero outside a closed boundary, and the plasma
where W is the electron injection energy and r0 a/6 is
source potential ΦS was computed assuming that it van
the radius of a virtual anode potential peak in the rotat
ing reference frame. The electron beam velocity varies ishes at large distances. Then the required value of ap
slowly, except near the inner and outer turning points. plied potential Φapp = Φ ΦS. The magnetic field contours,
From the steadystate continuity equation with P = 0, plasma density contours and externally applied potential
contours computed for the magnetic quadrupole l = 2

and magnetic hexapole l = 3 cases in cylindrical geome (c) externally applied
try are shown in Figure 18.
electrostatic potential
l = 2 (magnetic quadrupole) 2
(a) isodensity
1
1
0.5
z 0
z 0
-1
-0.5
-2
2 1 0 1 1
-1 r
1 0.5 0 0.5 1
r Figure 18c — Contours of externally applied electrostatic potential for a magnetic quadru-
pole. 117
Figure 18a — Contours of plasma density for a magnetic quadrupole. 117

l = 3 (magnetic hexapole)
l = 2 (magnetic quadrupole) (d) isodensity
(b) magnetic flux 1
2
0.5
1
z 0
z 0
-0.5
-1
-1
-2 1 0.5 0 0.5 1
2 1 0 1 2 r
r Figure 18d — Contours of plasma density for a magnetic hexapole. 117
Figure 18b — Contours of magnetic flux for a magnetic quadrupole. 117

l = 3 (magnetic hexapole) where p2 p1/p2 + p1, and p1, p2 are the turning
(e) magnetic flux points of the oscillations of the plasma boundary.
2 Although the studies of modified Penning traps are just

beginning, they have shown several interesting features:
i In a nonneutral plasma, densities exceeding the
Brillouin limit can be confined by several
1 means: in an oscillating plasma; in a flow with
deviatoric strain related to velocity shear such
as a torodial magnetic quadrupole with applied
electric field; in a homogeneous magnetic field
z 0 with an applied electrostatic quadrupole and
lowenergy particle injection; in a higherorder
magnetic multipole field with an applied elec
trostatic multipole.
-1
ii A large degree of focusing is possible, with re
sultant very high densities inside the core
radius r. This focusing can result in high values
-2 of reactivity, which is proportional to <2>.
2 1 0 1 1 iii Higherorder magnetic multipoles can have
r global Brillouin ratios approaching 2.
Figure 18e — Contours of magnetic flux for a magnetic hexapole. 117 While these results are not immediately conducive to an
attractive fusion power reactor, further developments
(f) externally applied along these lines could result in newer configurations
electrostatic potential with more attractive properties. The evolution towards
higherorder magnetic multipoles is similar to that of
2 the electromagnetic trap experiments, which started
with spindle cusps, then went to the toroidal quadrupole
and higherorder multipoles, such as a linear set of ring
cusps.
1
z 0
-1
-2
2 1 0 1 1
r
Figure 18f — Contours of externally applied electrostatic potential for a magnetic hexa-
pole. 117
The global Brillouin ratio was found to be = 2l/l + 1,

which approaches 2 for highorder multipoles.
Turner et al 118 calculated the frequency of the non
neutral plasma anharmonic oscillations and found that
/c varies from 21/2 when = 0 to 2/31/2 when = 1,

f(W),
3. MEPC arbitrary
units
15 t=0
Experiments
10 t = 1.5 ms
3.1 Spindle Cusps and Mirrors
Early experimental studies of electrostatic plugging were
conducted by Lavrent’ev5,36 and by Hilton et al.119 Due to 5
financial limitations, most early experiments were
pulsed spindle cusps, although spindle cusps are poor
configurations for electrostatic plugging. Electrons were
0.4 0.6 0.8 1.0 1.2 1.4 1.6
injected from electron guns in the point cusps, with W, keV
plasma production by ionization of neutral gas. Pulsed
Figure 19 — Measured electron energy distribution in the anode region immediately after
magnetic fields have several disadvantages for MEPC
cut-off of electron beam injection, and 1.5 ms later.81
experiments:
i Eddy currents cause magnetic field errors; viii A deep, negative potential well could be sus
tained for many ms, provided that the neutral
ii During magnetic field decay the plasma drift is gas pressure was low.
outwards;
ix Electron orbits and loss,. angles agree with
iii Plasma density buildup and heating may take theoretically predicted values, provided that
longer than the magnetic field lasts.120 the starting point is within an adiabatic
Experiments with spindle cusp electromagnetic traps region.125
like S~1, S3M, S4M, and Jupiter1A showed that: x Diocotron oscillations occurred, as predicted
i The electron confinement time increased by theoretically.
three orders of magnitude when the plugging Experimental observations of diocotron oscillations in
voltage was applied. the ring cusp indicated the following:
ii The electron confinement time was inversely 1 The frequency of the oscillations was propor
proportional to the neutral gas pressure.121 tional to applied voltage and inversely propor
iii The rate of electron loss across the magnetic tional to magnetic field strength, as expected.
field was similar to that expected from classi 2 The growth rate of the instability was propor
cal diusion due to collisions with ions and tional to neutral gas pressure: and thus to the
neutral atoms. density of cold electrons produced by ioniza
iv The plasma density increased linearly with tion.
applied voltage, up to a stability limit. 3 By segmenting the ring cusp cathode, the rela
v The plasma density increased proportional to tive ion currents as a function of azimuthal an
B2.122 ,123 gle were measured. These measurements indi
cated that the longwavelength instability had a
vi The mean ion energy scaled linearly with ap mode number = 1 or two wave periods
plied voltage.124 around the circumference of the ring. Higher
vii The electron distribution function81 shown in mode numbers were apparently more stable.126
Figure 19 is comparable to the theoretical 4 The anode region electron density remained
curve of Figure 15. below the shortwavelength diocotron limit of
equation 27.
Sugisaki 127 showed that the application of electrostatic
fields to a magnetically confined plasma prolonged the
plasma decay time. Strijland128 studied particle orbits in
an electrostatically plugged spindle cusp and the varia
tion of confinement time with diaphragm diameter.
With electrostatic plugging of a magnetic mirror,

Nishida et al 129 observed a reduction of ion losses to 15 1.2
of the unplugged value, and ion heating by E x B rota
tion.
1.0
Vdovin and coworkers 130 studied electron loss processes
from the S4M electromagnetic trap. They derived an
equation for equilibrium plasma density that varied with 08
n(z)
electron beam current, electron energy and neutral gas
pressure in accord with experimental data. Plasma elec na
0.6
tron diusion was dominated by scattering o gas mole
cules and molecular ions. They suggested that anoma
lous electron transport in the ring cusp region may be 0.4
caused by a longwavelength twostream instability. The
injected electrons had an anomalously fast loss rate in
the S4M electromagnetic trap, and neutral gas eects
0.2
kept the plasma cold a few eV temperature.
0.0
According to measurements by Komarov et al131 with a 0.1 0.2 0.3
twowavelength microwave interferometer in the z, mm
Jupiter1A spindle cusp device, the central density was
an order of magnitude higher than the average density. Figure 20 — Experimentally measured electron density distribution (full curve) in the ring
Zalesskii et al 132 found that the the plasma accumulation cusp of the Jupiter 1A device at A = 6 kV, B = 0.82 T, and model profiles n/na = 1/[1 + (z/
w)2] with w = pc (broken curve) and w = 2pc (dotted curve).
rate with microwave heating was much faster than with
electron beam injection, and about 50 of the micro In the KEMP~II spindle cusp experiment Stansfield and
wave power was coupled to the plasma. colleagues136 measured ~ 4 x 1017 m3, and electron en
Stepanenko and Komarov133 measured the Jupiter1A ergies ~ 1 keV. Ionization of gas prolonged the plasma
electron and ion energy spectra during microwave heat density persistence after electron beam cut~o.
ing with 10 power absorption, and determined the
electron energy confinement time to be about 0.9 ms.
The dominant ion energy loss mechanism was charge
3.2 Jupiter-1M
exchange. The Jupiter1M spindle cusp experiment at Kharkov137
had pulsed magnetic point cusp fields 13 T 25 ms half
Electron and ion energy spectra were measured with
period, ring cusp fields 4 T, electron injection voltages
gridded electrostatic energy analyzers at the point cusps.
~4 kV, plugging voltages ΦA ~ 6 kV, electron beam injec
The electron energy distribution is shown in Figure 19.
The ion energy spectra were measured with a drift mass tion current 2 mA, neutral gas pressure 0.11 mPa and
spectrometer combined with an electrostatic ionenergy plasma volume 30 cm3. It attained plasma densities up to
analyzer. The distribution appeared to have two Max 1018 m3, with temperatures ~100 eV and confinement
wellian components. The highertemperature compo times ~ 1 ms. The short duration of the pulsed magnetic
nent was believed to be caused by the initial formation fields did not allow enough time for buildup of higher
of the potential well. The average energy of the ions plasma densities. The plasma losses from the point
increased linearly with applied voltage, <Wi> ~ 0.18ΦA.36 cusps were less than from the ring cusp.138
Komarov and colleagues 139 used gridded analyzers to
Komarov and Stepanenko134 measured the spatial varia
measure the energy distributions of electrons and ions
tion of the total number of electrons N leaving the ring
emerging from the cusp gaps after removal of the plug
cusp gap, and found a narrow distribution with width ~ ging voltage. The mean ion energy observed under those
p, as shown in Figure 20. At an electron energy of 0.55 conditions was about onefifth of the plugging voltage.
keV, p = 0.1 mm. The distribution was narrower at high For a few microseconds after electrostatic plugging was
plasma densities, indicating that electron losses along removed, the remaining ions were accelerated to higher
the magnetic field became important relative to cross energies by the plasma potential.140 The measured po
field diusion losses. The model profile with = p is tential barrier in the ring cusp for ions Φi ~ 3Ti exceeded
roughly consistent with the experimental data, but = 2
the calculated barrier ~ 2Ti. The barrier height in the
p is used here in estimating future device parameters to
point cusp was practically zero, but most of the ion
avoid underestimating Φ. Komarov and coworkers135
losses occurred through the ring cusp. Point cusp ion
also measured the distributions of electrons and ions in losses were probably retarded by the centrifugal and
a point cusp gap, finding a narrow electron stream with magnetic field gradient forces.141
radius <1 mm.

Azovskii and colleagues142 found that the inner bound reached 25 eV, then dropped rapidly after electron beam
ing surface of the plasma in the Jupiter1M experiment injection cuto. Using an 8 keV Ar+ probe beam, they
has a radius ro = 2Rpo1/2 in the point cusp where po is measured the energy spectra of Ar2+ ions emerging from
the electron Larmor radius in the point cusp, as pre the ring cusp and deduced the time variation of the cen
dicted from flux mapping to the ring cusp. The density tral plasma potential. When energetic electrons were
of plasma flowing out of the point cusp was two orders injected into lowpressure hydrogen, the plasma poten
of magnitude lower than the central plasma density, and tial Φi + Φ initially was very large. As ionization built
Φ 0.5 kV was inferred from the shape of the ion en up the plasma density, the plasma potential decreased to
ergy distribution emerging from the point cusp.143 a plateau value on the order of 0.1 ΦA. After the electron
Azovskii and coworkers 144 measured the plasma density beam injection was turned o. the plasma density and
in Jupiter1M simultaneously by a microwave interfer potential decayed to zero in a few milliseconds.150
ometer and by the integral of the ion loss current, and
found that 1018 m3. The lifetimes of electrons in 3.3 Magnetic Multiples
jected from the gun were much shorter than the life
times of plasma electrons produced by ionization. The Nakamura et al151 observed enhancement of ion con
confinement time of the bulk plasma was about a factor finement in a multipole magnetic field by a negative
of five lower than classical, assuming that the electrons electrostatic potential well produced by electron injec
were produced by ionization. The injected fast electrons tion into the plasma. Hershkowitz et al 152 observed simi
left quickly ~0.01 ms, perhaps due to a twostream in lar eects.
stability, and imparted only about 12 of their energy Lavrent’ev and colleagues 153,154 studied the plasma pro
to the plasma. The electron current had 0.11 MHz duced by injection of 100 mA of 1 keV electrons into a
modulations. Probe measurements showed that the elec pulsed multipole trap. They measured 1017 m3, con
tron density profile was not consistent with a theoretical finement time 50 μs and Ti 300 eV. Using the at
profile based on classical transport.145 tenuation of a calcium ion beam during a 30 μs period of
Factors influencing the electron confinement times in electron injection, they observed, a rapid density
cluded the magnetic field geometry, impurities in the build~up to 4 x 1017 m3, followed by a decay after elec
plasma, ionization of neutral gas and electron mobility tron beam cuto. With probe measurements they found
in the electric field. The eect of the centrifugal force a uniform central plasma density dropping o steeply in
on radial electron transport would be less significant in the boundary magnetic field layer.155 Lavrent’ev et al156
large devices. Since the electron mobility is related to found that the plasma lifetime in a multipole electro
the neutral gas density, its contribution was significant magnetic trap depended very strongly on the magnetic
in small devices, but would be reduced in high field in the gaps, decreasing from 1.3 ms to 0.1 ms as the
temperature plasmas, where most of the neutral gas be field was reduced from 0.3 T to 0.15 T.
comes ionized.146 Moir et al157 proposed a toroidal quadrupole MEPC ex
From ion energy distributions measured outside the periment using soft iron to enhance the magnetic field
point cusp and ring cusps of the Jupiter1M device at ΦA shape and to minimize the required coil currents.
= 6 kV, Azovskii and coworkers147 deduced the amount Maei et al158 constructed a spherically convergent ion
of potential sagging in the anode gaps, and found Φ focus SCIF device based on Bussard’s Polywell concept
ring cusp = 0.4 kV, Φ point cusp = 0.6 kV. These a polyhedral set of point cusps, with planned values of
values correspond to uncompensated anode electron B = 0.2 T, Φ = 20 kV, I = 75 A from three electron guns
densities a ~ 1017 m3, while the central electron density and plasma production by electron cyclotron heating.
~ 1018 m3. The electronic space charge in the anode
gaps was partially compensated by ions flowing out the
cusps. Although the ion potential barrier Φi was lower in 3.4 ATOLL
the point cusp, the ion loss rate there was lower than in The ATOLL toroidal quadrupole magnetic cusp device
the ring cusp. From gridded analyzer data they found is shown in Figure 21. Typical parameter values are Ba =
that the ion energy distribution was nearly Maxwellian 0.8T, ΦA = 2kV, p = 0.20.5 mPa hydrogen, impurity frac
with Ti = 4060 eV. Ti was slightly higher at the plasma tion = 12 dominant mass = 28, 3 x 1017 m3, T
edge, where the electric field was large.148 100 eV and average ion energy 100 eV. Ioe et al159
Azovskii and colleagues149 measured the electrostatic checked the alignment of the magnetic field with the
potential and electron temperature in Jupiter1M. At the anodes by observing the flow of electrons from an elec
end of electron beam injection the potential sag. in the tron gun placed in the plasma region. The alignment
accuracy was better than 0.2 mm. Diagnostics included
point cusp cathode Φ ~ 1 kV. The electron temperature
an 8 mm microwave interferometer, currents drawn by

electrodes, a charged particle energy analyzer, Langmuir The electron potential barrier Φ~ 10T, so electron
probes including a tiny probe suspended by a 12 μm losses were predominantly by crossfield diusion. From
tungsten fiber, a charge exchange neutral atom energy the ionization rate Ioe and coworkers 162,163 inferred a
analyzer, a pulsed vacuum gauge and a partial pressure plasma confinement time of 0.18 ms. This was close to
analyzer. the theoretical confinement time, based on the ion
Electron Supports acoustic instability, of 0.12 ms, and much shorter than
Gun Cathodes
the classical confinement time of 50150 ms. Similarly,
Magnet
Coils the measured electron density profile was close to that
predicted above for the ion acoustic instability, with a
gradient scale length on the order of the ion sound Lar
mor radius, and far from the profile based on classical
diusion. Piterskii and coworkers 164,165 observed oscilla
Plasma
tions at 4080 MHz propagating in the electron dia
magnetic drift direction at about 105 m s1, which ap
peared to be consistent with ion acoustic instability the
ory. According to Lavrent’ev166 the crossfield transport
resulting from the observed profiles and oscillations in
ATOLL should be inwards.
48 cm 32
64 cm Later measurements revealed that three instabilities
Figure 21 — The ATOLL toroidal quadrupole cusp device160
were occurring simultaneously in ATOLL diocotron
oscillations in the cusp gaps, a longwavelength ion
In one mode of operation the plasma was sustained by acoustic instability in the plasma interior and a lower
electron injection from a gun plus secondary emission at hybrid drift instability at the plasma edge.
the reflecting cathodes. This mode resembled a Penning
Piterskii et al 167 observed diocotron oscillations in the
discharge with the end walls outside the cusps.
cusp gaps at frequencies of 4070 MHz, with wave
In another mode of operation Ioe and co~workers 160 lengths of 12 cm and peak amplitudes ~100V. The wave
minimized the plasma impurity content by careful wall phase velocity was somewhat lower than the electron E
preparation and by the use of very pure hydrogen. Ioni x B drift velocity in the anode gaps, by a factor of 1.52;
zation of neutral gas made up for particle losses, and the but this dierence could be partly due to simplifying
electrons were heated by the intense electric field in the assumptions used in the analysis, such as the shapes of
boundary layer. At high magnetic fields and low pressure the electron density distribution and electric field dis
the discharge could become self~sustaining without elec tribution in the anode gaps. The theoretical maximum
tron injection or secondary emission from the cathodes. growth rate of the diocotron oscillations occurs at ~
This regime had similar plasma parameters, but less 16a, which is within the range of wavelengths observed.
noise, better ion confinement and higher T than the The instability became stronger with increasing plasma
first regime. This ‘emission~free’ discharge was extin potential and decreasing magnetic field, as expected for
guished by adding 5 of an impurity, such as nitrogen. diocotron oscillations. In the central plasma the ampli
Impurities increased radiation losses and diusion rates, tude of the diocotron oscillations decayed with distance
and reduced the potential well depth. from the anodegaps with a scale length of about 5 cm,
The plasma potential increased linearly with magnetic as expected theoretically.
field and had a nearly parabolic radial distribution. Ions Piterskii and coworkers 168,169 found that the lower hy
produced by ionization at the edge were heated as they brid drift instability is dominant in the boundary region
fell into the potential well, and the average ion energy where the density is low. The theoretical values of most
increased with increasing magnetic field. From the unstable wavelength 1 cm, frequency 12 MHz and
plasma confinement time it was inferred that the poten phase velocity 1.2 x 105 m s1 are consistent with ex
tial barrier Φi impeding ion loss through the cusps was perimental data. This instability causes anomalous elec
very small. Ioe and colleagues 161 found that the electro tron transport, flattening the electron density profile
static potential in the cusp gaps was equal to the poten until the density gradient scale length / exceeds
tial at the plasma centre, to within experimental accu the local ion Larmor radius pi. Thus, the confinement
racy 10. The hypothesis of a narrow ‘neutral corridor’ times in ATOLL were much shorter than classical con
through the cusps with no confining potential barrier Φi finement times, and the anomalous transport was well
= 0 was also consistent with the measured variation of understood in terms of the lower hybrid drift and long
plasma potential with magnetic field. wavelength ion acoustic instabilities.

The classical electron diusional loss rate is172, 173
3.5 Jupiter-2M
The proposed Jupiter2 experiment 170 was to be a linear p
I diff Da 1 0.5 Dei " Fn0 R 2 electrons / s
set of ring cusps with B = 5T point cusps, ΦA = 25 kV, V
Te 0
!
= 0.5 m3, > 1019 m3, Ti ~ 1 keV and E ~ 100 ms. Due to 55
financial limitations, a onethird scale model of Jupiter 2,
called Jupiter2M 171 has been built first Figure 22, with a where Da is the electron diusion coecient due to col
primary goal of studying electron transport. lisions with neutral atoms, Φp = Φi + Φ, Dei is the elec
Field Support Support
Bar Coils
Electrode
Feedthroughs Cusp Gaps
Support tron diusion coecient due to collisions with ions, o
Tapering Plate Plate
Coil
Small Ring
is the central electron density, Teo is the central electron
Mirror
Coil Case
Cusps
temperature, and F is the geometrical factor having di
Point
Vacuum
Chamber mensions of reciprocal length. The Φp term represents
Cusp Electron
Coil Gun electron mobility in the strong edge electric field. The
Cathode
electron confinement time is
R 3n0
Electrode
Support Ne
Frame
diff 1.11
I diff I diff 56
Insulator
Cathode where N is the total number of electrons in the plasma.

Magnet Current
Feedthroughs
Repelling
Electrode
For the geometry of the Jupiter2M experiment F = 900/
Ring Anodes
cm.
Vacuum Pumping Port
1-5 mm gap Preliminary lowfield operation at B ~ 0.5 T ring cusp
Figure 22 — The Jupiter-2M device (left) and electrode region detail (right).171 with a 1.3 kV, 0.12 A electron beam injection into 2 x 106
torr H2 produced a plasma with = 5 x 1017 m3, T = 13
The Jupiter2M device has seven ring cusps with R = eV, > 2 ms. With ΦA ~ 2 kV, the plasma potential was
0.215 m in the central region flanked by one small ring about 300 V relative to the grounded chamber walls.
cusp and point cusp at each end. The axial length be About 80 of the injected electron beam energy ap
tween the point cusps is 1.30 m. The coils are pulsed peared in the plasma. Although the point cusp anode
from a capacitor bank and reach peak fields in about 33 diameters were 25 mm, the ion loss through the point
ms. The .maximum attainable ring cusp and point cusp cusps was only 6 of the total ion loss rate.
magnetic fields are about 1.94 and 3.88 T, respectively,
but lower fields have been used in experiments. The 20
shape of the vacuum magnetic field is illustrated in Fig
ure 23.171
40
0
0
0 Jexp / Jth
r,
cm
- - + + - - +
10
20
-66 +66 -66
-16 +
-16 0
+16 0 +66
-4 +2 +16
+4 +2
-1 +1 + -
0
0 20 40 60
z, cm 0
0 0.2 0.4 0.6 0.8 1.0
Figure 23 — Vacuum magnetic field lines in the Jupiter-2M device.171
ne , 1018 m-3
The plugging electrode positions were adjusted by
Figure 24 — Ratio of experimental to theoretical electron loss currents, as a function of
screws until comb probes at three azimuthal locations plasma density. B = 0.375 T (+), 0.5 T (★), 0.625 T () 174
120° apart showed no asymmetry of the electron loss
currents emerging from the cusps. The total plasma vol Highfrequency oscillations were not observed in
ume is 501, and the volume of the uniform central region Jupiter2M. The large uniform central plasma may serve
is 161. as a buer to damp out surface oscillations. Vdovin e

al,174 found that the ratio of experimental electron loss good containment system for heavy impurities. The
current to theoretical loss current decreases with in presence of electrodes in environments with high elec
creasing plasma density, as shown in Figure 24. At higher tric fields and high particle fluxes may constitute a very
plasma densities, the experimental loss rate was about large source of impurities. This might ultimately be the
twice the classical value. The measured ion loss rate cor greatest problem.”
responds to an ion potential barrier Φi/Ti ~ 4.
Auxiliary electrodes were used to suppress secondary 3.7 Comparison of Experiments

electron emission from the plugging cathodes. Dia
Key parameters of various MEPC experiments are
phragms limiting the plasma boundary in the point
summarized in Table 2. In some cases the eective tem
cusps and between the ring cusps served two functions:
perature is taken to be twothirds of the average particle
i to reduce particle flow onto the anodes, where energy. The early experiments at Kharkov S~1, S3, S~3M
gas emission stimulated by particle bombard and S4M, Quebec, Missouri and Jutphaas were small
ment would tend to fill the gap with cold spindle cusp plasmas with weak fields, and neutral gas
plasma. eects were dominant. Such experiments are of less in
terest now, but they served to demonstrate the princi
ii to study the influence of the diaphragm size on
ples of MEPC. The multiple point cusp SCIF device was
plasma transport.
shut down by budget cuts before reaching its projected
Ring diaphragms with diameters of 21, 24 and 28 cm operating parameters, but it may be restarted at another
corresponded to ring cusp diusion layer widths a = 0.9, laboratory.
1.6 and 2.0 mm.
The Jupiter1A device achieved good ion and electron
The theoretical electron density profile depends little temperatures 0.41.5 keV. It also demonstrated the
on the temperature profile. The electron density profile eectiveness of electron cyclotron heating, although at
is flat in the centre and falls nearly linearly to zero near low powers. The Jupiter1M experiment was not able to
the plasma edge. There is a uniform central plasma with fully utilize its high magnetic fields, because the dura
radius ~5 cm and length ~80 cm. As the diusion layer tion of the peak field was too short to permit accumula
width a is increased from 0.9 to 2.0 mm, the electron tion of high plasma density. Due to their low ratios of
confinement improves, but ion confinement degrades, plasma volume to surface area V/SR 0.03, spindle cusps
due to the increase of Φ. There is also an increased are not worthwhile considering for future experiments.
probability of arcing with large gap widths, due to anode The linear multipole devices demonstrated confinement
bombardment and outgassing. At the optimum width in that geometry and validated theoretical predictions
aop = 1.6 mm the maximum plasma density = 1018 m3 of critical loss angles. Due to end losses, these devices
is attained. The electron confinement time deduced did not achieve high plasma pressures.
from the total number of electrons in the plasma di
vided by the total electron diusion loss current is = 3 The ATOLL totoridal quadrupole was carefully built
ms. Future plans are to increase the magnetic field to 12 and operated, but the electron energy transport was
T, and to inject hot plasma with a plasma gun. anomalous. Detailed studies related the anomalous
transport to the lower hybrid drift instability plasma
edge and to the ion acoustic instability plasma inte
3.6 Impurities rior. ATOLL was shut down in the 1980s.
Impurity eects in MEPC devices have been studied The Jupiter2M experiment has attained the most prom
very little. Lavrent’ev94 showed the theoretical eect of ising results. Its demonstration of nearclassical trans
increasing Ze on the energy confinement time. How port is stimulating interest in MEPC research.
ever, nobody has computed impurity source terms,
transport and buildup.
3.7.1 Density, Temperature
A continuous impurity accumulation was observed in
the ATOLL device during the injection pulse. A typical
and Confinement Time Scalings
impurity level was 12 with the dominant species at The plasma density in spindle cusps was typically 1017 to
mass 28.159 The addition of a few per cent impurity ex 1018 m3. The attainment of higher plasma density has
tinguished the ‘emission free’ regime of ATOLL.160 been a major concern for the reactor prospects of
There have been almost no studies of impurity profiles, MEPC devices. The plasma density appears to scale
impurity transport, radiative power loss or methods of proportional to B2,122 provided that the magnetic field
impurity removal from an MEPC plasma. Impurity stud duration is long enough for density accumulation. The
ies need greater emphasis in the future. As Gormezano39 central plasma density is found to be significantly
stated, “any electrostatic system will constitute a very higher than the electron density a in the anode region,

but more definitive measurements of the ratio a/ are For small spindle cusp devices the confinement times
needed at high plasma temperatures. have been ~ 0.1 to 1 ms. Little distinction has been
The measured electron energy distributions resemble made experimentally between particle confinement and
shifted Maxwellians, as expected. Computer models in energy confinement. For ring cusps the confinement
dicate that, if neutral gas eects are not dominant, the time is assumed to scale roughly in accordance with
plasma temperatures scale proportional to the applied equation 40
voltage. Measured ion temperatures have scaled linearly ei aR
with ΦA. For the plasma density and temperature in diff 0.09
MEPC devices, we assume the following scaling
b a 40
n . B 2 57 This equation predicts di 0.8 ms for Jupiter2M pa

rameters of Table 2. The experimentally measured con
Te Ti 0.05 A 58 finement time was 23 ms. If the central electron tem
perature were higher than the 15 eV reported from a
The scalings were used to estimate the hypothetical re Langmuir probe under the magnet coils, then the theo
actor parameters in Tables 1 and 2. That the tempera retical di would be longer. A more precise computation
tures in Jupiter2M were below this assumed scaling may specific to the Jupiter2M geometry173 predicts classical
be due to the short duration of the peak magnetic field, confinement times about twice the measured value.
the weak heating power and neutral gas eects.
Table 2A — Parameters of some MEPC Experiments. (Italicized values are estimates)

Spindle Cusps Reference R a Be A n (m-3) Te Ti a/pe e ve /R
(cm) (mm) (T) (kV) (keV) (keV)
S1, Kharkov 5 5 1.5 0.4 2.0 3.0 x 1017 0.27 0.16 4.0 x 105 11 0.032 7.8 x 103
S3 and S3M, Kharkov 81 10 1.5 0.6 5.0 4.0 x 1017 0.50 0.34 12 0.038
S4M, Kharkov 130 10 1 0.25 1.0 1.0 x 1017 0.054
KEMPII, University of Quebec 135 17 1 0.5 6.0 4.0 x 1017 1.0 0.30 5.0 x 104 5 0.084 5.5 x 104
University of Missouri 122 16 1.5 0.3 5.0 2.0 x 1016 0.50 0.10 5.0 x 104 6 0.005 4.1 x 104
Jutphaas 127 18 0.13 0.20 1.0 x 104 4.7 x 103
Jupiter1A, Kharkov 131, 138 23 1.5 0.65 6.0 3.0 x 1017 1.50 0.40 3.0 x 104 7 0.054 3.0 x 104
Jupiter1M, Kharkov 137, 147 10 1.5 1.7 8.0 1.0 x 1018 0.10 0.05 5.0 x 104 76 0.002 3.0 x 104

Table 2B — Parameters of some MEPC Experiments. (Italicized values are estimates)
Reference R a Be A n (m-3) Te Ti a/pe e ve /R
(cm) (mm) (T) (kV) (keV) (keV)
Linear multipole, m = 4, Kharkov 75 5 1.5 0.15 1.0
Linear multipole, m = 8, Kharkov 153, 154 6 1 1.2 1.0 x 1017 0.50 0.30 5.0 x 105 16 0.002 1.1 x 104
Toroidal quadrupole, ATOLL 158, 154 20 1 1.5 1.7 x 1017 0.03 0.07 1.8 x 104 81 0.001 5.3 x 103
Multiple point cusps, SCIF 104, 158 92 0.2 10
Linear set of ring cusps, Jupiter 173 22 1.6 0.6 2.5 x 1018 0.015 0.03 3.0 x 103 74 0.005 3.1 x 104
2M, Kharkov
Hypothetical reactor, Linear set 400 2 6.0 400 x 1020 20 20 6 25 4.5 1.7 x 108
of ring cusps
Present experiments attain confinement values of about

3.7.2 Dimensionless Parameters 104 transit times. Without eective plugging, the plasma
Three dimensionless parameter ratios are listed in Table would only be confined for tens of transit times.
2: a/p, c, and /R. The a/p represents the number of
This number of transit times is relevant to a mode of
electron Larmor radii within onehalf of the ring cusp
free plasma flow out cusp gaps. It is less relevant to
anode gap. In MEPC experiments this number has
MEPC devices, because ions are confined electrostati
ranged from 5 to 100, and for a reactor it would proba
cally, and their confinement times are roughly equal to
bly be in the order of 2030. The optimum anode gap
the electron crossfield diusion time. Nevertheless, it is
width is of the order of 1.5 to 2 mm, and the electron
instructive to show that if the ions were freely flowing
Larmor radius is fixed by the magnetic field and electron
out cusp gaps, those gaps would be very narrow. Accord
temperature in a reactor to values of the order of 30100
ing to the freeflow model, the ion confinement time is
μm.
The second dimensionless parameter is the ratio of av rp2 L1
= 6
erage plasma pressure to the peak magnetic field pres uR
sure at the coil where c is the peak magnetic field at the
coil a. For Jupiter2M, R = 0.22m, rp 0.1m, L1 = 0.105 m, u 7.6
x 104 ms1 and 3 ms. Using this equation we estimate
(Te Ti )
Bc 2 +0 nk the eective cusp gap halfwidth 2.1 x 105 m. Since
Bc2 59 the electron Larmor radius pa = 2.2 x 105 m, it is seen
that pa. If free ion flow were occurring, if would be
In magnetic confinement fusion it is more common to through a very narrow gap.
define beta relative to the vacuum magnetic field at the
plasma centre, but that central magnetic field is zero in
magnetic cusps, which would yield an infinite value of 3.7.3 Anomalous Transport
beta. The values of c in past MEPC experiments have The diocotron instability in the ring cusp regions is be
been very low, of the order of 0.001 to 0.1 . Using the lieved to play a beneficial role by removing anode
density and temperature scalings of equations 57 and trapped cold electrons, which tend to screen out the
58 results in c = 4.5 for the hypothetical reactor of anode potential, but other instabilities have deleterious
table 2. For comparison, the estimated value for the eects. The ATOLL toroidal quadrupole cusp experi
ITER tokamak Troyon coecient g = 3, I = 24 mA, mi ment had anomalous transport hundreds of times faster
nor radius = 3 m, B = 6T, Bc = 13 T is c = 0.85. than classical in all operating regimes, associated with
the longwavelength ion acoustic instability and the
The third dimensionless parameter in table 2 is /R, lower hybrid drift instability. The Jupiter2M multiple
which is the ratio of the confinement time to the ap ring cusp experiment, however, has demonstrated nearly
proximate electron transit time through the plasma. classical electron transport. So far, the reason for the

dierence is not understood. There are several hypothe parameters. The confinement in some experiments is
ses, as follows: nearly classical, when neutral gas eects are taken into
account. The diocotron oscillations in the ring cusps
i In Jupiter2M directed electron injection from
have been thoroughly studied and are well understood.
thermoelectric cathodes was used. According to
The plasma waves associated with anomalous transport
Lavrent’ev,1 electron injection in ATOLL by
in ATOLL have been correlated with theoretical predic
secondary emission from the cathodes intro
tions.
duced electrons into the diusional layer rather
than into the central Brillouin region, which Why some devices have anomalous transport and others
could foster the twostream instability. The rate do not is not yet understood. More measurements are
of electron production was two orders of mag needed of the density and temperature profiles, impurity
nitude greater than the electron loss rate by distribution and transport, neutral gas distribution, ra
classical diusion, so classical diusion could diation losses and heat propagation. Twodimensional
not accommodate the required outflow rate of computer simulations are needed to provide a link be
spent electrons from the trap. tween analytic theory and experimental data.
ii ATOLL has twodimensional curvature and A common thread of deficiencies links almost all the
only toroidal symmetry in the anode regions, past MEPC experiments: either weak or shortduration
while Jupiter2M has onedimensional curvature magnetic fields; relatively low voltages <10 kV; weak
and since h >> a approximately two degrees of plasma heating; incomplete diagnostics; inadequate
symmetry. This lack of symmetry in ATOLL computer modeling. These deficiencies have been
might be more conducive to growth of the caused by low budgets, not by failures, of the experi
harmful instabilities, such as the lower hybrid menters.
drift waves; or it might somehow tend to reduce
Table 3 compares the parameters of Jupiter2M with the
the beneficial eects of diocotron oscillations
proposed Jupiter2 and JupiterT devices. If successful,
on cold electrons within the anodes.
JupiterT could demonstrate the attainment of thermo
iii The large volume of fieldfree plasma in Jupiter nuclear parameters in a modestsized device.
2M may serve to damp the growth of instabili
ties.
More study will be required to falsify or confirm these
hypotheses.
3.7.4 Assessment
MEPC experiments typically have had small radii R ~ 10
to 20 cm and plasma volume < 11, except for Jupiter2M
501 and the SCIF device. The anode gap halfwidth
ranges from 1 to 2 mm. A larger gap could permit exces
sive voltage sag Φ, and a smaller gap would decrease
the electron diusion loss time. The magnetic fields
have generally been below 1 T, except for a few pulsed
experiments. The applied voltage has been below 10 kV.
Many excellent measurements have been performed in
MEPC devices. Detailed probe measurements have
shown that the electron density profile in the ring cusp
anode has a halfwidth about one Larmor radius. The
electrostatic potential profile has been mapped out with
pulsed ion beam transmission measurements and grid
ded energy analyzers. A deep negative potential well has
been sustained for many milliseconds. The electron ve
locity distribution resembles a shifted Maxwellian, as
expected. These measurements verify the basic theory
of MEPC.
The scalings of density, temperature and confinement
time have been established within the current ranges of

Various cusp geometries suitable for electrostatic plug
4. Reactor ging were illustrated in Figure 8. Mirror geometries,
such as a yinyang field, are also possible for electrostatic
plugging, but the plasma volume inside the anode
Concepts bounded flux surface is smaller than for cusp systems.
4.1 Reactor Design Studies

Table 3 — Comparison of Jupiter-2M, Jupiter-2, and Jupiter-T experiments.174
Parameter Jupiter-2M Jupiter-2 Jupiter-T

Ring cups radius, R m 0.215 0.67 1.4
Length between ends, L m 1.4 3.9 6.4
Ring cusp magnetic field, Ba T 0.5 2.5 4.0
Point cusp magnetic field, B T 1 5 8
Applied voltage, ΦA kV 2.5 25 100
Electron injection current, I A 0.2 5 30
Plasma density, m3 0.8 x 1018 3 x 1019 1020
Ion temperature, T keV 0.03 1 10
Plasma volume, V m3 0.05 0.5 4
Electron diusion current, Idi A 1.6 14 120
Electron lifetime, e s 0.002 0.36 1.3
Laurent’ev175 considered several reactor geometries for A2 BaV

MEPC reactors: a sphere, a cylinder and a torus. They Q 10 5
would typically operate at Ba = 7 T, Φa = 200 kV, ~ 1020 n
S 1 t
m3, Ti ~ 30 keV, V ~ 800 m3 and power density ~4 MW na 60
m3. They would provide intense sources of neutrons,
which could be used for production of tritium or pluto where ΦA is in kV and is the trapped electron density
nium. in the anode region~ Assuming /a = 0.3 and V/SR
0.3 in the example case of table 1, it is found that Q = 10.
Pastukhov and Yushmanov 176,177 estimated the power
This analysis did not include energy losses due to neutral
gain of an MEPC reactor. At each value of applied po
gas and impurities, or the beneficial eect of the alpha
tential Φa there is a plasma temperature that maximizes
particle emission current helping to reduce the required
Q. The optimum gap width makes Φ/Φa 0.25. When cathode current. Alternatively, using the parameters of
the gap is wider, Φ becomes large, reducing the plug Table 1 with Ze 3 in equations 4D1 and 3F14 of ref
ging potentials Φ and Φi. When the gap is narrower, erence 44, we find Q 13.
electron diusion losses are too great. At the optimum A spherical cusp, shown in Figure 8, would be advanta
gap width, the ratio of fusion power to input power for a geous from a physics standpoint.178 ,179 The sphere has
ring cusp plasma reduces to the best surfacetovolume ratio, and the spherical elec
trode arrangement might provide some spherical focus

ing of charged particles. If eective, such focusing could High
increase the central plasma density and decrease the
required magnetic field in the cusps. However, particle
Voltage
scattering by waves and by Coulomb collisions and re Input
flections o the scalloped boundary would tend to di
minish the degree of focusing. It would be dicult to Voltage
align the electrodes with the magnetic field lines in a Divider Cathode
spherical reactor.
A torodial hexapole reactor design had Ba = 8T, Φa =
300kV, = 102 m3, T = 17 keV and a neutron wall load
ing = 1.6 MW m2.84 The direct capital costs estimated Anodes
for this reactor were comparable to those for tokamak Radiation
and tandem mirror reactors.180 Collimator
From the results with Jupiter2M, it appears that ion
Coil
losses from the point cusps are not a: serious problem,
so a linear set of ring cusps would be feasible for a reac
tor. Such a reactor would be superior to either the
Shield
spherical cusp or toroidal hexapole cusp, because of its
simplicity, lower structural mass and easier maintenance.
A linear MEPC reactor, as illustrated in Figure 25, could
have the parameters of table 1, with Q 10, DT fusion
power density = 3 MW m3, neutron wall loading = Plasma
3.6MW m2, L = 50m, and total fusion power = 4.2GW.
Coils Figure 26 — Radiation collimator and voltage divider for high-voltage electrodes.
R rp L1 4.2 Applications
Besides power production, there are other potential ap
Point plications for MEPC devices. Their good confinement
Magnetic Field Lines Cusp of highZ ions would make them well suited for strip
ping and confining heavy ions for accelerators, as sug
gested by Stix.181 ,182
Figure 25 — One End of a Linear MEPC Reactor
Nadler et al 183 have suggested that small inertial
Radiation collimators and voltage dividers would be electrostatic confinement IEC devices operating in the
used to prevent high voltage breakdown, as illustrated in glow discharge mode at 3050 kV could become useful
Figure 26. lowcost neutron sources producing >106 DD neutrons/s
for a variety of applications, including detector calibra
tion, diagnostics and training. By increasing the voltage,
current and size, and using DT fuel, yields of 1010
neutrons/s might be attainable. Such neutron sources
could be stronger than 252Cf sources 2.6 year halflife,
and they would be simpler and cheaper than accelerator
based neutron sources.
Miley et al184 found that IEC device operation in the
‘star’ discharge mode maximizes the eective grid
transmission factor for ions. An IEC device using D3He
fuel could generate 14.7 MeV protons for production of
positronemitting radioisotopes that are useful for posi
tron emission tomography facilities. Miley et al 185 pro
posed the use of highvoltage collectors in a ‘venetian
blind’ configuration to convert the D3He reaction
product kinetic energy directly into electricity.

Miley et al186 studied hydrogen production by radiolysis
of water with 14.7 MeV protons from an IEC device.
The 3.6 MeV alpha energy would be converted directly
into electricity. The protons would pass through a thin
chamber wall and interact with steam flowing through
the cell, splitting about 13 water molecules per 100 eV of
energy deposited by the protons. The steam would pass
through a turbine to generate electricity, and then the
hydrogen would be removed from the condensate. The
overall system eciency could be about 50.
Bussard187 suggested the use of an electronbeamheated
propellant thruster powered by an electrostatic plasma
device for space propulsion. Miley et al188 suggested the
use of IEC engines with venetian blind energy conver
tors to power five 68N arcjetmagnetoplasmadynamic
thrusters with a specific impulse of 3000 s.
In pulsed operation, an IEC reactor with R = 5.25 m and
mass = 5 t could deliver 12 MW of power to the
thruster.189 Bussard190 proposed the use of a Polywell
IEC reactor for space propulsion.
An MEPC plasma with T = 1020 keV would be a good
target for neutral beam injection at Wi = 150 keV with a
beam ion fusion probability ~ 102.191 Such a device could
serve as a powerful neutron source for isotope produc
tion, medical therapy or materials testing.

tion method, the eects of curvature and symmetry, and
5. Issues of MEPC the eect of a large volume of fieldfree plasma. The
extension of the good confinement to larger devices re
Although some recent confinement results appear prom mains to be tested experimentally. A more powerful ex
ising, there are still many issues to be resolved before periment is being designed to test the scaling relations
MEPC devices can become fusion reactors, as discussed at higher values of plasma density, temperature and con
below. finement time.
5.1 Self-Shielding of 5.5 Power Gain Ratio

Applied Potentials The estimated power gain ratio for an MEPC reactor
with nearly classical fuel ion confinement but no alpha
The potential drop in point cusp anodes is very large, particle confinement, ΦA = 400 kV, Ba = 6 T, R = 4 m, is
but point cusp ion losses can be kept low by the cen Q 10. The power gain ratio could be increased by the
trifugal and magnetic field gradient forces, as demon following means: increasing the applied voltage; increas
strated in Jupiter2M. At B 6 T, values of a 1.5 x 1019 ing the magnetic field; increasing the plasma volume;
m3 are required to keep Φ 100kV in the ring cusps. improving alpha energy confinement.
The potential sagging in the anode means that high ap One means of improving alpha confinement would be to
plied voltages are needed to ensure substantial Φ and Φi. add a modest toroidal field to a toroidal multipole device
similar to the Tormac configuration.193 The toroidal
5.2 Neutral Gas Effects field would help to confine alphaparticle energy, facili
tating ignition, and the resultant magnetic shear might
Neural gas increases cross field transport rates and pro help to stabilize the lower hybrid drift instability in the
motes highvoltage breakdown. The plasma tends to act plasma edge layer.194 Another possibility might be the
as a vacuum pump, and in future large MEPC devices addition of a weak longitudinal magnetic field under
the neutral gas eects should be less dominant than in neath the surface cusp field of a linear set of ring cusps.
small spindle cusps. Nevertheless, there will always be
substantial recycling of hydrogen isotopes and helium
o the walls, so shields should be used to prevent exces 5.6 Plasma Impurities
sive flow of neutral gas to the anode regions, where Since alpha particles are poorly confined, helium ash
buildup of cold trapped electrons would increase Φ. accumulation could be slower than in tokamaks. How
Highspeed vacuum pumping will be required in the ever, sputtered wall atoms which penetrate into the
electrode region. plasma core would be well confined. Impurity ion accu
mulation would dilute the fuel density, cool the electrons
by radiation losses and diminish the potential well
5.3 Electrodes depth. One means for impurity removal might be to use
Optimized plasma confinement by careful alignment of the centrifugal force of their rotation in a weak longitu
each anode was demonstrated in Jupiter2M. Radiation dinal field to expand their orbits so that a pumped lim
shields and voltage dividers will be needed to sustain iter could extract them. Another possibility would be to
voltages ΦA > 200kV. To avoid warping, the anodes expand the point cusp hole at one end of a linear reactor
and to flow in fresh fuel from the other end, so that the
should be actively cooled and made from materials with
weak pressure gradient gradually moved impurities to
low thermal expansion. Although highvoltage holding
the hole. If eective impurity removal methods could
in radiation fields must still be demonstrated in MEPC
not be found, then periodic shutdowns might be re
devices, the PSP2 rotating plasma experiment192 has
quired for impurity removal, followed by restarts with
demonstrated the sustainment of 400 kV applied volt
fresh fuel.
age with hot plasma present.
5.4 Transport 5.7 Materials Issues

The wall activation, afterheat, accidental release of ra
The reason why Jupiter2M had good confinement and
dioactivity and waste disposal issues of MEPC reactors
ATOLL had anomalous transport is not yet understood.
would be comparable to those of tokamaks and other
Transport in MEPC devices has been studied much less
magnetic confinement devices. The reactor design
than transport in tokamaks and tandem mirrors, so this
should provide for rapid replacement of damaged walls
lack of understanding is not surprising. Several hypothe
and electrodes. The desirable first wall neutron power
ses have been suggested: the eect of the electron injec

flux would probably be about 25 MW m2. Higher values
could cause heat removal and wall lifetime problems,
and lower values would raise the cost of electricity. The
anodes should be partially shielded from radiation, and
the electrodes and walls must be actively cooled. Low
activation materials, such as purified vanadium chro
mium alloys and SiC, should be used if feasible.
One possible materials combination for an MEPC reac
tor is vanadium alloy structure with a circulating Flibe
LiFBeF2 molten salt blanket and coolant. Use of the
molten salt would avoid the reliability problems of high
pressure coolants and the fire hazard of lithium. A small
amount of lithium could be used to cool firstwall hot
spots, if needed. The irradiated vanadium alloy could be
cooled for tens of years, then remelted to remove the
volatile longlived isotopes mostly 42Ar and recycled.195
With careful choice of materials, an MEPC reactor
would not require active safety systems to prevent
prompt osite fatalities in the event of a serious acci
dent.

need more research to determine the attain
6. Conclusions and able degree of focusing and to explore related
concepts that may lead to attractive fusion
reactors sections 2.10, 2.11.
Recommendations viii Radiation shields and voltage dividers should
be used to prevent high voltage breakdown.
i Purely electrostatic plasma confinement with Voltage holding at 400 kV has been demon
spatial and temporal focusing may be wroth strated in a magnetic mirror geometry by the
reexamination section 1.1.4. PSP2 rotating plasma experiment section
ii Due to magnetic reflection and electrostatic 5.3.
acceleration eects, the central plasma den ix MEPC devices may find applications as sec
sity in an MEPC device may be an order of tion 4.2: electrical power plants; heavy ion
magnitude higher than the electron density in sources; neutron sources for materials testing,
the ring cusp region. Cold electrons from medical therapy and isotope production.
ionization of neutral gas tend to screen out
the anode potential, but they are probably x The main issues of concern for MEPC reactor
removed from ring cusps by diocotron oscilla development are: electron transport, plasma
tions sections 2.5 and 2.6. purity and electrode alignment and voltage
holding. Future research should concentrate
iii Although anomalous transport was observed on resolving these issues.
in the ATOLL toroidal quadrupole device, the
plasma loss rate in the Jupiter2M linear ring
cusp experiment was only about twice the Acknowledgments
classical rate, improving at higher density. The
reason for the good stability of the Jupiter2M This review is dedicated to Oleg A. Lavrent’ev who has
plasma is not yet understood, but might be led this field of research steadily for four decades, in
related to its method of electron injection, its spite of many diculties. Robert W. Bussard provided
better symmetry, or its large volume of uni information on the Polywell concept, and L. Turner, D.
form fieldfree plasma section 3.7.3. C. Barnes and Richard A. Nebel provided information
on the modified Penning trap research. This work was
iv Based on present experimental results, a linear supported by the US Department of Energy, Oce of
set of ring cusps appears to be the most favor Fusion Energy, under DOE Idaho Field Oce Contract
able configuration for an MEPC reactor. number DOEAC0776ID01570.
v Several decades of small, cheap, pulsed ex
periments have demonstrated the basic prin Publication History
ciples of MEPC, but these experiments have
been marred by weak or pulsed magnetic Received 20 December 1993, in final form, 25 March
fields, low voltages, weak plasma heating, in 1994.
complete diagnostics and inadequate com Reformatted and color illustrations added in April 2009
puter modeling section 3.7.4. Further pro by Mark Duncan.
gress depends upon increasing R, B and ΦA in
a steadystate device, so that , T and E can
approach the thermonuclear regime. A new
experiment is needed to demonstrate the at
tainment of interesting parameters > 3 x
1019 m3, T > 5 keV in a steadystate device
with intermediate size R 2 m, L 10 m and
reasonable applied fields B 5T, ΦA 100 kV.
vi For an MEPC reactor with nearclassical

transport, values of Q 10 are predicted at
voltages ΦA 400kV section 4.1. Q could be
enhanced by partial confinement of alpha par
ticles.
vii The Polywell and modified Penning trap con
cepts are just beginning to be studied. They

15 R. W. Hockney; 1968 Journal Applied Physics 39, 41
Footnotes 6670
1Oleg A. Lavrent’ev; 1992 unpublished letter to Thomas 16C. Barnes; 1973, “Computer simulation of electrostatic
James Dolan, 17 October 1992 inertial confinement,” Stanford University Report
AD·763826
2Philo T. Farnsworth; “Electric Discharge Device for
Producing Interactions Between Nuclei,” U.S. Patent 17G. D. Porter; E. H. Klevans; 1971, Physics Fluids,
Number 3,258,402 June 28, 1966. 1442834
3 Oleg A. Lavrent’ev; 1963 Ukrain. Fiz. Zh. 8, 440445 18Robert L. Hirsch; 1967, “InertialElectrostatic Con
finement of Ionized Fusion Gases,” Journal Applied
4G. I. Budker; 1961, Plasma Physics and the Problem of Physics, 38, 45224534
Controed Thermonuclear Reactions I Elmsford, NY: Per
gamon pp. 295301 19K. M. Hu and E. H. Klevans, “On the Theory of Elec
trostatic Confinement of Plasmas with Ion Injection,”
5Oleg A. Lavrent’ev; 1968, “Investigation of an electro Physics Fluids, 17, 227231 1974
magnetic trap,” Magnitnye Lovushki Vypusk 3 Kiev:
Naukova Dumka pp 77147 translation: AECTR·7002 B. E. Cherrington; J. T. Verdeyen; D. A. Swanson; 1975,
20
Rev Annuals New York Academy Sciences, 25113951

6 Robert L. Hirsch; 1970, US Patent 3,530,036 21 D. A. Swanson; B. E. Cherrington; J. T. Verdeyen; 1973,
Physics Fluids 16, 193945

7Thomas James Dolan; J. T. Verdeyen; B. E. Cherring
ton; D. J. Meeker; 1970, “ElectrostaticInertial Plasma D. A. Swanson; B. E. Cherrington; J. T. Verdeyen; 1973,
22
Confinement,” US Air Force Oce of Scientific Re Applied Physics Letters, 23, 1256
search Report AFOSR701656TR
23 J. T. Verdeyen; B. E. Cherrington; D. A. Swanson; D. J.
8Oleg A. Lavrent’ev; V. A. Sidorkin; B. V. Zaitsev; V. P. Meeker; 1975, Annuals New York Academy Sciences, 251,
Goncharenko; L. I. Ovcharenko; 1972 Ukrain. Fiz. Zh. 12638
42, 1436
24 D. J. Meeker; J. T. Verdeyen; B. E. Cherrington; 1973,
9Oleg A. Lavrent’ev; V. A. Sidorkin; B. V. Zaitsev; Yu S. Journal Applied Physics, 44, 534755
Azovskii; 1972, Ukrain. Flz. Zh. 46, 12824
25G. Imel; 1973, Ph.D. Thesis, Pennsylvania State Uni
10S. Yu; S. Azovskii; B. V. Zaitsev; Oleg A. Lavrent’ev; N. versity
N. Sappa; V. A. Sidorkin; 1976, “Generation and focusing
W. M. Black; E. H. Klevans; 1974, Journal Applied
26
of ions from a source with distributed parameters,” Re
port 76·9 Institute of Physics and Technology, Ukrain Physics. 45 250211
ian Academy of Sciences, Kharkov
27 E. H. Klevans,; 1975, Annuals New York Academy Sci
Oleg A. Lavrent’ev; V. A. Sidorkin; B. V. Zaitsev;

11 ences, 251 190212
Azovskii; S Yu; 1976, Zh. Tekh. Fiz. 46 12824
W. M. Black; J. W. Robinson; 1974, Journal Applied
28
12Oleg A. Lavrent’ev; V. A. Sidorkin; B. V. Zaitsev; N. N. Physics, 45 2497501

Sappa; 1973, “Investigation of gas discharge in a spherical
29 A. L. Gardner; D. M. Hatch; A. I. Y. Chan; R. P.
diode,” Proceedings 11th International Conference on
Phenomena in Ionized Gases Prague, Prague: Czecho Evans; 1975, Annuals New York Academy Sciences,
slovak Academy of Science 25117989
30D. C. Baxter; G. W. Stewart; 1982, Journal Applied
13Robert L. Hirsch; “Experimental studies of a deep,
negative, electrostatic potential well in spherical geome Physics, S34597_601
try,” 1968, Physics Fluids, 11, 248690 31 J. M. Peterson; N. L. Oleson; 1992, Physics Fluids, B 4
14 William C. Elmore, James L. Tuck, and Kenneth M. 888901

Watson; “On the inertialelectrostatic confinement of a 32J. H. Nadler; Y. B. Gu; George H. Miley; 1992, Review
plasma,” 1959 Physics of Fluids, 2, 239246
Scientific Instruments, 63481012

George H. Miley; J. Nadler; T. Hochberg; Y. Gu; O.
33 52 A. Marcus; G. Knorr; G. Joyce; 1978, “Two
Barnouin; J. Lovberg; 1991, Fusion Technology, 19, 8405 dimensional simulation cusp confinement of a plasma,”
Report 7855 University of Iowa Department of Physics
34 Robert L. Hirsch; G. A. Meeks; 1970, US Patent and Astronomy
3,530,497
53 G. Knorr; D. Willis; 1982, Z. Nalurf. 37a, 7805
Thomas James Dolan; J. T. Verdeyen; O. J. Meeker; B.
35
E. Cherrington; 1972, Journal Applied Physics, 43, 1590 54N. Hershkowitz; J. M Dawson; 1976, Nuclear Fusion,
600 16, 63942
36 Oleg A. Lavrent’ev; “Electrostatic and Electromag 55Oleg A. Lavrent’ev; L. I. Ovcharenko, B. G. Safronov;
netic HighTemperature Plasma Traps,” translated by T. V. A. Sidorkin; 1966, Ukrain. Hz. Zh. 11, 98290
J. Dolan, in Electromagnetic Confinement of Plasmas
56Ralph W. Moir; 1968, “A new magnetic well configura
and the Phenomenology of Relativistic Electron Beams,
L. C. Marshall and H. Sahlin, Editors, Annuals New tion, importance of the mirror ratio and electrostatic
York Academy Science, 251, pp. 152178 1975. containment of warm plasma,” Report EurCEAFC496
FontenayAuxRoses, France
37 B. I. Inanov; A. A. Kalmykov; Oleg A. Lavrent’ev;
57J. D. Daugherty; L. Grodzins; G. S. Janes; R. H. Levy;
1976, “On the possibility of initiating thermonuclear
reactions by highcurrent ion beams,” Report KhFTI 1968, Physics Review Letters, 20, 369
768 Institute of Physics and Technology, Ukrainian
58 T. H. Stix; 1970, Physics Reviews Letters, 24, 135
Academy of Sciences, Kharkov
59 T. H. Stix; 1971, Physics Fluids, 14, 692
38D. W. Baldwin; 1977, Review Modern Physics 49, 317
39 60 T. H. Stix; 1971, Physics Fluids, 14, 702
39 C. Gormezano; 1979, Nuclear Fusion, 19, 1085137 61R. Jones; 1989, Journal Indian Institute Science, 69,
40 E. E. Yushmanov; 1985, Nuclear Fusion, 25, 121724 3736
62 R. Jones; 1990, Philippine Journal Science, 119, 17182
B. G. Logan; 1980, Communications Plasma Physics
41
Controlled Fusion, 5, 271 63E. E. Yushmanov; 1978, Soviet Journal Plasma Physics
42R. Jones; 1984, Proceedings 10th Symposium on Fu 4, 1117
sion Engineering, Philadelphia, PA, 69 December 1983 64 K. D. Marx; Thomas James Dolan,; Ralph W. Moir; C.
Piscataway. NJ: IEEE pp 18648
E. McDowell; 1976, Bulletin American Physics Society
43J. Kesner; S. F. Horne; V. P. Pastukhov; 1987, Journal 21, 1044
Fusion Energy, 6, 40111 65R. H. Cohen; M. E. Rensink; T. A. Cutler; A. A. Mirin;
44Thomas James Dolan; 1982, Fusion Research, Elms 1978, Nuclear Fusion 18, 122943
ford, NY: Pergamon 66 V. P. Pastukhov; 1974, Nuclear Fusion 14, 3
45 R. Keller; I. R. Jones; 1966, Z. Naturf. 21, 10859
V. Land Sizonenko; K. N. Stepanov; 1976, Soviet
67
46 M. Sadowski; 1967, Physics Letters, 25a, 695696 PhysicsTechnical Physics 20, 46870
68B. B. McHarg Jr.; M. E. Oakes; 1974, Physics Fluids 17,
47 M. Sadowski; 1968, Physics Letters, 27a, 435436
19234
48 M. Sadowski; 1969, Physics Letters, 28a 626627
69B. B. McHarg Jr.; M. E. Oakes; 1975, Physics Fluids 18,
49M. Sadowski; 1969, Reviews Scientific Instruments, 105965
40, 15459
70 T. D. Rognlien; T. A. Cutler; 1980, Nuclear Fusion 20,
50 M. Sadowski; 1970, Journal Plasma Physics, 4, 116 100311
51 71P. B. Parks; A. M. Sleeper; 1984, Physics Fluids 27,

M. G. Haines; 1977, Nuclear Fusion, 17, 81158
28908

72 E. E. Yushmanov; 1978, Soviet Journal Plasma Physics 90 Thomas James Dolan; 1977, “Review of electrostatic
4, 1117 plugging,” Lawrence Livermore National Laboratory

Report UCID17576
73D. P. Chernin; M. N. Rosenbluth; 1978, Nuclear Fu
sion 18, 4762 91V. P. Paslukhov; 1978, Soviet Journal Plasma Physics
43, 1116
74A. A. Ware; J. E. Faulkner; 1969, Nuclear Fusion 9, 353
61 92 Thomas James Dolan; J. M. Larsen; B. L. Stansfield;
1975, Canadian Journal Physics 53, 23417

75V. A. Sidorkin; Oleg A. Lavrent’ev; 1971, Plasma Phys
ics and the Problem of Controlled Thermonuclear Fu 93K. Miyamoto; 1989, Plasma Physics for Nuclear Fu
sion 2 Kiev: Naukova Dumka pp 24751 sion revised edition, Cambridge, MA: MIT Press
Ralph W. Moir; William L. Barr; Richard F. Post; 1971,

76 94 Oleg A. Lavrent’ev; 1981, Ukrain. Fiz. Zh. 26, 146672
Physics Fluids 14, 25312534
95 Oleg A. Lavrent’ev; 1981, Ukrain. Fiz. Zh. 26. 163641
77M. Abramowitz; I. A. Stegun; 1966, Handbook of
96V. V. Dolgopolov; Oleg A. Lavrent’ev; N. N. Sappa;
Mathematical Functions AMS55 Washington. DC: Na
tional Bureau of Standards 1982, Soviet Journal Plasma Physics 8, 7403
78 97V. P. Pastukhov; 1980, Soviet Journal Plasma Physics 6,

O. Buneman; R. H. Levy; L. M. Linson; 1966, Journal
Applied Physics 37, 320322 54954
79 98V. I. Il’gisonis; V. P. Pastukhov; 1981, “Transverse elec

Yu I. Pankrafev; V. A. Naboka; E. F. Ponomarenko;
Oleg A. Lavrent’ev; 1972, Nuclear Fusion 12, 3913 tron losses in a magnetoelectrostatic trap at low plasma
pressures,” Report IAE3495/6 Kurchatov Institute of
80R. Hand Levy; R. W. Hockney; 1968 Physics Fluids 11, Atomic Energy, Moscow
76671
99Thomas James Dolan; B. L. Stansfield; 1973, Nuclear
81Yu I. Pankrafev; V. A. Naboka; S. A. Vdovin; Oleg A. Fusion 13, 9602
Lavrent’ev; A. A. Kalmykov; 1975, “Plasma Physics and
Thomas James Dolan; 1975, Annuals New York Acad
100
the Problem of Controlled Thermonuclear Reactions 1
3,” Kharkov: Institute of Physics and Technology, emy Science, 251, 35866
Ukrainian Academy of Sciences pp 421 Translation:
V. I. Karpukhin; Oleg A. Lavrent’ev; N. N. Sappa;
101
URCLTrans11142
1978, Ukrain. Fiz. Zh. 23, 1294300
82 I. Ya Gordienko; V. D. Egorenkov; K. N. Stepanov;
102 V. I. Karpukhin; Oleg A. Lavrent’ev; N. N. Sappa;
1992, Soviet Journal Plasma Physics 18, 48
1980 Ukrain. Fiz. Zh. 25, 10828
83 E. E. Yushrnanov; 1981, Nuclear Fusion 21, 3293
Robert W. Bussard; “Some Physics Considerations of
103
84Thomas James Dolan; 1976, “Design study of electro Magnetic Inertial Electrostatic Confinement: a New
statically plugged cusp fusion reactor,” Lawrence Liver Concept for Spherical ConvergingFlow Fusion,” 1991,
more National Laboratory Report URCL52142 Fusion Technology 19, 273293
85 104Nicholas A. Krall; “The Polywell: A Spherically Con

Thomas James Dolan; B. L. Stansfield; J M Larsen;
1975, Physics Fluids 18, 13836 vergent Ion Focus Concept,” 1992, Fusion Technology
22, 4249
D. G. Blondin; Thomas James Dolan; 1976, Journal
86
J. Watrous; Nicholas A. Krall; K. Wong; 1991, Bulletin
105
Applied Physics 47, 29036
American Physics Society 36, 2318
87F. Brunei; G. Lafrance; H. Burkhardt; B. C. Gregory;
106 Nicholas A. Krall; K. Wong; V. Stefan; 1991, Bulletin
1974, Canadian Journal Physics 52, 215763
American Physics Society 36, 2319
J. P. Matte; G. Lafrance; 1977, Journal Computing
88
107Robert W. Bussard and Katherine E. King, “Phe
Physics 23, 8691
nomenological Modelling of Polywell /SCIF MultiCusp
89M. Shoucri; J. P. Matte; B. C. Gregory; 1980, Canadian Inertial Electrostatic Confinement Systems,” Bulletin
Journal Physics 58, 121629 American Physics Society, 36, 2319 1991

108 Katherine E. King and Robert W. Bussard, “EKXL: A 125 Oleg A. Lavrent’ev; L. I. Ovcharenko; B. A. Shevchuk;
Dynamic PoissonSolver for SphericallyConvergent 1975, “Plasma Physics and the Problem of Controlled
InertialElectrostatic Confinement Systems,” paper Thermonuclear Reactions 1 3.” Kharkov: Institute of
2T11, Annual Meeting Division of Plasma Physics, APS, Physics and Technology, Ukrainian Academy of Sciences
Tampa, FL, Nov. 48, 1991, Bulletin American Physics pp 228 .
Society, Volume 36, 2319 1991 EMC2119103
126Yu I. Pankrat’ev; V. A. Naboka; M. G. Nozdrachev; E.
109Thomas James Dolan; 1993, Fusion Technology 24, F. Ponomarenko; Oleg A. Lavrent’ev; B. G. Safronov;
128129 1971, Atomnaya Energiya 31, 27480
110 A. B. Mikhailovskii; 1974, Theory of Plasma Instabili 127K. Sugisaki; 1968, Japan Journal Applied Physics 7,
ties, New York: Consultants Bureau 138796
Robert W. Bussard; 1992, unpublished letter to Tho

111 128 W. Strijland; 1970, Physica 47, 61725
mas James Dolan, 14 July 1992
129Y. Nishida; S. Kawamata; K. Ishii; 1977, Journal Phys
D. C. Barnes; L. Turner; 1992 Physics Fluids B 4,
112 ics Society Japan 43, 136470
38901
130S. A. Vdovin; V. I. Karpukhin; Oleg A. Lavrent’ev; M.
L. Turner; D. C. Barnes; 1993, Physics Review Letter
113 Nand Novikov; V. A. Sidorkin; 1984, Ukrain. Fiz. Zh. 29,
70, 798801 83641
114 D. C. Barnes; Richard A. Nebel; L. Turner; 1993, A. O. Komacov; Oleg A. Lavcenrev; V. A. Potapenko;
131
Physics Fluids B 5, 365160 A. I. Skibenko; I. A. Stepanenko; I. P. Fomin; 1981,

Teplojiz Vysok. Temp. 19, 61418
G. Miyamaoto; G. Iwata; SandInoue K Mori; 1957,
115
Journal Physics Society Japan 12, 438 132Yu G. Zalesskii; A. D. Komarov; Oleg A. Lavrent’ev; V.
A. Naboka; N. I. Nazarov; V. A. Potapenko; Stepanenko,
116D. C. Barnes; Richard A. Nebel; L. Turner; T. N. I. A; 1979, Soviet Journal Plasma Physics 5, 5324
Tiouririne; 1993, Plasma Physics Control. Fusion 35, 929
40 133 I. A. Stepanenko; A. D. Komarov; 1984, Ukrain. Fiz.
Zh. 29, 70410

T. N. Tiouririne; Land Turner; A. W. C. Lau; 1993,
117
Physics Review Letters 72, 12047 A. D. Komarov; I. A. Stepanenko; 1980, Ukrain. Fiz.
134
Zh. 25, 14325

L. Turner; T. N. Tiouririne; A. W. C. Lau; 1993, Journal
118
Mathematical Physics 35, 234956 135 A. D. Komarov; Oleg A. Lavcenrev; V. A. Naboka; V.
A. Potapenko; I. A. Stepanenko; 1980, Ukrain. Fiz. Zh.

J. L. Hilton; C. K. Hinrichs; A. A. Ware; 1968, Plasma
119
25, 77680
Physics 10, 4556
136B. L. Stansfield; J. M. Larsen; B. Bergevin; P. Couture;
Oleg A. Lavrent’ev; V. A. Potapenko; I. A. Ste
120
B. C. Gregory; 1976, Canadian Journal Physics 54, 1856
panenko; 1976, Soviet PhysicalTechnology Physics 2163 61
6
A. V. Georgievskii; V. E. Ziser; Oleg A. Lavcenrev; M.
137
Yu I. Pankrat’ev; N. A. Tulin; E. F. Ponomarenko; V.
121
G Nozdrachev; D. P. Pogozhev; 1974, Soviet Atomic En
A. Naboka; 1973, Atomnaya Energiya 35, 2537 ergy 36, 40810
122 Yu I. Pankrat’ev; M. G. Nozdrachev; Oleg A. 138Yu S. Azovskii; V. I. Karpukhin; Oleg A. Lavrent’ev; V.
Lavrent’ev; B. G. Safronov; V. A. Naboka; E. F. Pono A. Maslov; M. G. Novikov; M. N. Nozdrachev; A. I.
marenko; 1972, Atomnaya Energiya 32, 1316 Skibenko; I. P. Fomin; 1978, “Study of the singlegap
electromagnetic trap,” Jupiter 1M Report KhFTI 782
R. L. Hayward; Dolan Thomas James; 1977, Physics
123
Institute of Physics and Technology, Ukrainian Acad
Fluids 20, 6468 emy of Sciences, Kharkov
124 Oleg A. Lavrent’ev; 1978, Ukrain. Fiz. Zh. 23, 192231 139A. O. Komarov; Oleg A. Lavcenrev; V. A. Potapenko;
I. A. Stepanenko; 1979, Soviet PhysicalTechnology
Physics 24, 10602

140Yu S. Azovskii; Oleg A. Lavrent’ev; V. A. Maslov; M. 157Ralph W Moir; Thomas James Dolan; William L.
G. Nozdrachev; 1985, Ukrain. Fiz. Zh. 30, 166971 Barr; 1977, “Design of an electrostatic endplugged
plasmaconfinement device,” Proceedings 7th Sympo
Yu S. Azovskii; Oleg A. Lavrent’ev; V. A. Maslov; 1986,
141
sium on Engineering Problems of Controlled Thermo
Ukrain. Fiz. Zh. 31, 16916 nuclear Research, Knoxville Piscataway, NJ: IEEE
Yu S. Azovskii; Oleg A. Lavrent’ev; V. A. Maslov; M.

142 158K. C. Maei; J. A. Lovberg; R. A. Jacobsen; 1991, Bul
Nand Novikov; M. G. Nozdrachev; 1980, Ukrain. Fiz. letin American Physics Society 36, 2318
Zh. 25, 151118
159 M. S. Ioe; B. I. Kanaev; V. V. Piterskii; E. E.
Yu S. Azovskii; V. I. Karpukhin; Oleg A. Lavrent’ev; V.
143
Yushrnanov; 1984, Soviet Journal Plasma Physics, 10,
A. Maslov; M. Nand Novikov; M. G. Nozdrachev; 1980, 2617
Soviet Journal Plasma Physics 6, 1426
160M. S. Ioe; G. I. Kanaev; V. P. Pastukhov; V. V. Piter
Yu S. Azvoskii; Oleg A. Lavrent’ev; V. A. Maslov; M.
144
skii; E. E. Yushmanov; 1981, JETP Letters 34, 5703
N. Novikov; M. G. Nozdrachev; A. I. Skibenko; I. P.
Fornin; 1981, Ukrain. Fiz. Zh. 26, 42934 161 M. S. Ioe; B. I. Kanaev; V. V. Piterskii; E. E.
Yushmanov; 1984, Soviet Journal Plasma Physics 10, 268

Yu S. Azovskii; Oleg A. Lavrent’ev; V. A. Maslov; 1986,
145
75
Ukrain. Fiz. Zh. 31, 13505
162M. S. Ioe; B. I. Kanaev; V. P. Pastukhov; V. V. Piter
146 Oleg A. Lavrent’ev; 1988, Ukrain. Fiz. Zh. 33, 134852 skii; R. Rand Chistyakov; E. E. Yushrnanov,; 1981, Pro
ceedings 10th European Conference on Controlled Fu
147
sion and Plasma Physics, Moscow PetitLaney: EPS
G. Nozdrachev; 1981, Fiz. Plazmy 7 96872
163M. S. Ioe; B. I. Kanaev; V. P. Pastukhov; V. V. Piter
148Yu S. Azovskii; Oleg A. Lavrent’ev; V. A. Maslov; M. skii; E. E. Yushmanov; 1981, “Experiments in the Atoll
G. Nozdrachev; 1984, Fiz. Plazmy 10, 131013 device,” Proceedings 10th European Conference on
149
Controlled Fusion and Plasma Physics, Moscow Petit
Laney: EPS
G. Nozdrachev; 1983, Ukrain. Fiz. Zh. 28, 21318
164V. V. Pitserskii; R. R. Chistyakov; E. E. Yushmanov;
150
1983, “A study of plasma oscillations in the Atoll trap,”
G. Nozdrachev; 1984, Soviet Journal Plasma Physics 10,
Zvenigorod Conference, Zvenigorod, USSR
36970
165M. S. Ioe; V. I. Kanaev; V. P. Pastukhov; V. V. Piter
Y. Nakamura; B. H. Quon; A. Y. Wong; 1975, Physics
151
skii; R. Rand Chistyakov; E. E. Yushmanov; 1983, Pro
Letter 53a, 856
ceedings. 11th European Conference on Controlled Fu
N. Hershkowitz; K. Hendricks; R. T. Carpenter; 1982,
152 sion and Plasma Physics. Aachen, Germany Petit
Journal Applied Physics 53, 4105112 Laney: EPS pp 2612
166 Oleg A. Lavrent’ev; 1988, Ukrain. Fiz. Zh. 33, 134852
153Oleg A. Lavrent’ev; V. A. Sidorkin; V. P. Goncharenko;
Yu S. Azovskii; 1974, “Electron injection and plasma 167
V. V. Piterskii; V. P. Pastukhov; E. E. Yushmanov; 1987,
confinement in a multigap electromagnetic trap,” Re
Soviet Journal Plasma Physics 13, 2935
port 743 Institute of Physics and Technology, Ukrain
ian Academy of Sciences, Kharkov 168V. V. Pitserskii; V. P. Pastukhov; M. S. Ioe; B. I.
154 Oleg A. Lavrent’ev; N. Nand Sappa; V. A. Sidorkin;
Kanaev,; E. E. Yushmanov; 1988, “Problems of Atomic
Science and Technology,” Series Thermonuclear Fusion
1976, Ukrain. Fiz. Zh. 21, 6379
2 Moscow: Kurchatov Institute of Atomic Energy pp
Oleg A. Lavrent’ev; N. Nand Sappa; V. A. Sidorkin;
155 502
1977, Ukrain. Fiz. Zh. 22, 174144 169
V. V. Pitserskii; V. P. Pastukhov; M. S. Ioe; B. I.
Oleg A. Lavrent’ev; N. Nand Sappa; B A Shevchuk;
156 Kanaev; E. E. Yushmanov; 1988, Soviet Journal Plasma
1979, Ukrain. Fiz. Zh. 24, 105052 Physics 14, 94100
170 Oleg A. Lavrent’ev; 1979, Ukrain. Fiz. Zh. 24, 101923

S. A. Vdovin; Oleg A. Lavrent’ev; V. A. Maslov; M. G.
171 181 T. H. Stix; 1969, Physics Review Letters 23, 1093
Nozdrachev; V. P. Obozniy; N. N. Sappa; 1988, “Prob
182T. H. Stix; 1972, IEEE Transactions Nuclear Science
lems of Atomic Science and Technology.” Series Ther
monuclear Fusion 3 Moscow: Kurchatov Institute of 19, 1505
Atomic Energy pp 405
J. H. Nadler; George H. Miley; Y Gu; T Hochberg;
183
172S. V. Germanova; Oleg A. Lavrent’ev; V. I. Petrenko; 1992, Fusion Technology 21, 163943
1989, “Transverse electron transport in a multicusp elec
184George H. Miley; J. Javedani; Y. Yamamoto; Richard
tromagnetic trap with axisymmetric magnetic fields,”
Problems of Atomic Science and Technology Series Nebel; J. Nadler; Y. Gu; A Satsangi; P. Heck; 1993,
Thermonuclear Fusion 3 Moscow: Kurchatov Institute “Inertialelectrostatic confinement neutron/proton
of Atomic Energy pp 6972 source,” Proceedings Third International Conference on
Dense ZPinches, London, 1923 April New York:
173S. V. Germanova; Oleg A. Lavrent’ev; V. I. Petrenko; AIP
1991, “Crossfield transport of electrons in a multicusp
George H. Miley; A. J. Satsangi; Y. Yamamoto; H.
185
electromagnetic trap across the end magnetic surfaces,”
Problems Of Atomic Science and Technology Series Nakashima; J. B. Lavedani; 1993, “Conceptual design for
Thermonuclear Fusion 2 Moscow: Kurchatov Institute a D3He IEC pilot plant,” Proceedings 15th IEEE Sym
of Atomic Energy pp 746 posium on Fusion Engineering, Hyannis. MA, 1115 Oc
tober Piscataway; NJ: IEEE
174S. A. Vdovin; S. B. Germanova; Oleg A. Lavrent’ev; V.
186George H Miley; Y. Yamamoto; J. Javedani; Y. Gu; A.
A. Maslov; M. G. Nozdrachev; V. P. Obozniy; V. I.
Petrenko; N. N. Sappa; 1990, “Plasma accumulation and Satsangi, P. Heck, Richard Nebel; L. Turner; Robert W.
confinement in the Jupiter2M multicusp electromag Bussard; M. Ohnishi; K. Yoshikawa; H. Momota; Y. To
netic trap,” Proceedings AllUnion Conference on Open mita; 1993, “Hydrogen generation with an inertial elec
Traps, Moscow: Kurchatov Institute of Atomic Energy trostatic confinement power source,” Proceedings First
pp 7081 International Conference on New Energy Systems and
Conversions, Yokohama, Japan, 2730 June
175Oleg A. Lavrent’ev; 1977, “Engineering and physics
187 Robert W. Bussard; 1990, Jet Propulsion 6, 567
calculation of a powerful thermonuclear system based
on an electromagnetic trap,” Proceedings AllUnion 188
George H. Miley; R. Burton; J. Javedani; Y. Yama
Conference on Engineering Problems of Thermonuclear
moto; A. Satsangi; Y. Gu; P. Heck; Richard Nebel; N.
Reactors 1, Leningrad, 2830 June Leningrad: Efremov
Schulze; J. Christensen; D. Strellis; J. DeMora; S. Coo
Institute of Electrophysical Instruments pp 10613
per; A. Ochoa; J. Fluhrer; 1993, “Inertial electrostatic
176V. P. Pastukhov; E. E. Yushmanov; 1978, “A magnetoe confinement as a power source for electric propulsion,”
lectrostatic trap,” Report IAE3042 Kurchatov Insti Proceedings Vision 21 Conference, Cleveland, OR, 3031
tute of Atomic Energy, Moscow March Washington DC: NASA
189A. J. Satsangi; George H. Miley; J. B. Javedani; H.
177 E. E. Yushmanov; 1980, Nuclear Fusion 20 38
Nakashima; Y. Yamamoto; 1994, “Innovative technology
178Oleg A. Lavrent’ev; A. A. Kalmykov; A. V. Geo for an inertial electrostatic confinement IEC fusion
gievskii; V. Ye Ziser; V. B. Yuferov; 1974, “The ‘Jupiter’ propulsion unit,” Proceedings 11th Symposium on Space
Thermonuclear Reactor Based on Confinement of Nuclear Power and Propulsion. Albuquerque, NM, 913
Plasma by Electrical and Magnetic Fields,” Proceedings January New York: AIP
Joint USSRUSA Seminar on Systems Analysis and Con
190Robert W. Bussard; 1993, “The QED engine system:
struction of Thermonuclear Stations, NllEFA Lenin
grad, 920 December Leningrad: Efremov Institute of direct electric fusionpowered rocket propulsion sys
Electrophysical Instruments tems.” Proceedings 10th Symposium on Space Nuclear
Power and Propulsion, Albuquerque, NM, 1014 Janu
179Oleg A. Lavrent’ev; A. A. Kalmykov; A. V. Geor ary American Institute of Physics Publication 199 pp.
gievskii; V. Ye Ziser; V. B. Yuferov; 1975, Izvest. Akad. 160112
Nauk Energetika 1 Transport 6, 549
Oleg A. Lavrent’ev; V. I. Karpukhin; V. I. Petrenko;
191
180Thomas James Dolan; 1977, “Cost estimate of electro N. N. Sappa; 1982, “Problems of Atomic Science and
statically plugged cusp reactor,” Lawrence Livermore Technology,” Series Thermonuclear Fusion 2 pp 9294
National Laboratory Report UCID17568

192G. F. Abrashitov; A. V. Beloborodov; V. I. Volosov; V.
V. Kubarev; Yu S. Popov; Yu N. Yudin; 1989, “Hot rotat
ing plasma PSP2 experiment,” Report 89109 Novosi
birsk Institute of Nuclear Physics
193I. G. Brown; W. B. Kunkel; M. A. Levine; 1978, Nu
clear Fusion 18, 7618
194 N. T. Gladd; Y. Goren; C. S. Liu; R. C. Davidson;
1977, Physics Fluids, 20, 18769
G. J. Butterworth; 1992, Fusion Technology 21, 1994

195
2000

Magnetic Electrostatic Plasma Confinement

Uploaded by

Copyright:

Available Formats

Magnetic Electrostatic Plasma Confinement

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Magnetic Electrostatic Plasma Confinement

Uploaded by

Copyright:

Available Formats

Review Article: Magnetic Electrostatic

Author contact information as of April 2009: University

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 1

B = Brillouin density  = electron velocity

 = particle density of species k  = electron densitygradient drift velocity

Q = fusion power/input power Z = charge number

q = 2pe/2ce Z = 1 for electrons and 1/2 for ions

r = radius = a parameter depending on Ti/T

ra = point cusp anode radius = fraction of current intercepted by the grid

rb = plasma boundary radius

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 2

r = thickness of plasma edge region ΦA = applied voltage

<> = DT fusion reaction rate parameter

k = characteristic collision time for electrons or

 = electron lifetime or confinement time

ei = electronion collision time

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 3

Magnetic electrostatic plasma confinement MEPC

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 4

1.1.2 Multiple Well Hypothesis

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 5

a slight double well, as illustrated in Figure 5;

Figure 4 — Electron bremssstrahlung yield versus radial position, as measured by

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 6

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 7

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 8

X X duced to ~ p. Cathodes outside the cusp gap reflect

cusp X escaping electrons back into the plasma. Such electro

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 9

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 10

Cathode Electrons may be produced by thermionic emission

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 11

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 12

n ewa nia rp2 L1

 dz z=a sin( crit ) =

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 13

Selfshielding voltage drop, after 47

Anode gap peak electron density, after 16

Energy confinement time, 6 s after 40

If p = density of electrons that are able to penetrate

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 14

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 15

.1 where Si is the ionization source, I is the electron cur

Some magnetically trapped electrons have low enough

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 16

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 17

where o is the central plasma density and A is the d (nTi ) (T T )

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 18

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 19

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 20

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 21

Figure 18a — Contours of plasma density for a magnetic quadrupole. 117

Figure 18b — Contours of magnetic flux for a magnetic quadrupole. 117

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 22

2 Although the studies of modified Penning traps are just

The global Brillouin ratio was found to be = 2l/l + 1,

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 23

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 24

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 25

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 1

B = Brillouin density = electron velocity

= particle density of species k = electron densitygradient drift velocity

Q = fusion power/input power Z = charge number

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 2

<> = DT fusion reaction rate parameter

k = characteristic collision time for electrons or

= electron lifetime or confinement time

ei = electronion collision time

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 9

dz z=a sin( crit ) =

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 13

Selfshielding voltage drop, after 47

Anode gap peak electron density, after 16

Energy confinement time, 6 s after 40

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 14

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 15

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 16

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 17

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 18

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 19

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 20

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 21

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 25

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 26

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 29

n . B 2 57 This equation predicts di 0.8 ms for Jupiter2M pa

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 30

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 31

Length between ends, L m 1.4 3.9 6.4

Ring cusp magnetic field, Ba T 0.5 2.5 4.0

Point cusp magnetic field, B T 1 5 8

Applied voltage, ΦA kV 2.5 25 100

Electron injection current, I A 0.2 5 30

Plasma density, m3 0.8 x 1018 3 x 1019 1020

Ion temperature, T keV 0.03 1 10

Plasma volume, V m3 0.05 0.5 4

Electron diusion current, Idi A 1.6 14 120

Electron lifetime, e s 0.002 0.36 1.3

Laurent’ev175 considered several reactor geometries for A2 BaV

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 33

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 34

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 36

vi For an MEPC reactor with nearclassical

Plasma Physics and Controlled Fusion 36 1994, pp. 15391593, doi:10.1088/07413335/36/10/001 38