Plasma Physics and Controlled Fusion, Vol. 32. h o .

13, pp 1237 to 1248, 1990

Printed in Great Briiain

0741-3335 90 $3.00t

.00

IOP Publishing Ltd. and Pergamon Press plc

EXPERIMENTAL TESTS OF LANGMUIR PROBE

THEORY FOR STRONG MAGNETIC FIELDS
R. A. PITTS
AEA Fusion. Culham Laboratory, UKAEAjEuratom Fusion Association, Abingdon,
Oxon OX14 3DB. U.K.

and

P. C. STANGEBY
Institute for Aerospace Studies, University of Toronto, Canada and JET Joint Undertaking.
Abingdon, Oxon, OX14 3EA. U.K.
(Receiced 23 April 1990 : and in revised f o r m 13 JulF 1990)

Abstract-The suppression of electron current to a single Langmuir probe immersed in a strongly magnetized plasma has been the subject of theoretical analysis since the original work of BOHMet ai.[1949,
in Characteristics ofBlectricalDischa~gesin Magnetic Fields (Edited by A. GUTHRIE
and R. K. WAKERLING).
McGraw-Hill. New York]. Experimentally,studies of the phenomenon have been restricted to a comparison
of measured data with analytic predictions of the I-V characteristic. Both theory and experiment show
that the simple exponential law for electron collection assumed in the derivation of T, from the characteristic
no longer holds for probe potentials above the floating potential. It is assumed, however, that the portion
of the characteristic below floating potential can be used in deriving T,. By monitoring the potential of a
small, electrically floating pin as a function of the voltage applied to a much larger plate located just behind
it and interpreting the pin voltage as a direct measure of the plasma potential near the plate, the validity
of this important assumption has been experimentally confirmed for the first time. Certain theoretical
aspects of net electron collection have also been tested and have been found to be in reasonable agreement
with experiment, indicating that more effort may be justified in attempting to use this region of the
characteristic as a diagnostic tool.

1 INTRODUCTION

DUETO the extensive application of the single Langmuir probe as a Tokamak edge
diagnostic (MANOSand MCCRACKEN,
1986), theoretical study of the probe characteristic from magnetized plasmas has been the subject of a number of papers (BOHM
et al., 1949; COHEX,1978; STANGEBY,
1982). Whilst the ion saturation region of the
characteristic is thought to be reasonably well understood, the same cannot be said
for the region corresponding to electron collection where the net current is observed
to be much less than that which would be obtained in a non-magnetic plasma. Indeed,
the ratio Is;t/I&t
is often observed to be only 10% of that pertaining to the field
free case (i.e.
One theoretical study (STANGEBY,
1982) models the complete Langmuir probe
characteristic and, for probe bias potentials above the floating potential (at which
point there is no net current to the probe), finds a marked deviation from the simple
exponential law applicable in the field free case. A reduction factor, r , is defined
describing the depression of electron current in electron saturation over the B = 0
case [see equation (2)]. This factor contains a number of parameters which are
generally unknown to the experimentalist attempting to apply the theory to Langmuir
probe data (ion temperature, T,,electron cross-field diffusion coefficient, D y,impurity

6).

1237

1238

R. A. PITTS and P.

c. STANGEBY

content, ZeE,in the plasma edge). Nevertheless, several authors (LABOMBARD,

1986 ;
BUDNYand MANOS,1984; LAUXet al., 1989) have made such attempts.
As a consequence of the departure from ideal behaviour above the floating potential,
the recommended procedure for use of Langmuir probes in strong magnetic fields is
to ignore this region and extract T, from the characteristic for V, d V, where it is
thought to be less likely that distortions are occurring. Experimental work on JET
(TAGLEet al., 1987) showed that errors in T, occurred if the reduction effect was
ignored and too great a fraction of the characteristic used in the conventional analysis
(i.e. assuming electron collection is simply exponential with applied potential). This
is not, however. confirmation that the ion saturation region gives T, correctly, since
the use of standard theory to derive it assumes that all of the probe bias is taken up
by the sheath. In fact, there is no a priori reason to assume in a magnetized plasma,
where the probe is usually biased with respect to limiter and cross-field diffusion must
often close the circuit over long distances in the plasma. that some of the probe voltage
difference does not appear across magnetic field lines to drive the circuit current. The
question, therefore, is whether T, can be reliably extracted from any portion of the
I-V characteristic for a probe in a strong magnetic field.
In this paper, we present the results of a new approach to the problem of investigating the magnetized Langmuir probe characteristic. By concentrating on the
behaviour of the potential in the region close to a probe as the voltage on it is varied.
we are able to draw useful conclusions concerning the validity of the standard method
described above for analysis of the magnetized probe characteristic. In addition, our
experiments demonstrate how net electron collection can be reasonably well described
by theory and indicate that it may be possible to derive usable information from this
region of the characteristic.
2. T H E POTENTIAL RISE N E A R A BIASED PROBE

In the standard picture (STANGEBY,

1982 and Fig. 1), the probe is assumed floating
( V , = V i ) ,of square shape (width d ) and normal to the magnetic field. A region of
disturbance termed the ambipolar collection length ( L,ambz d2c,/8D";"b.with c, the
ion sound speed, DImbthe ambipolar cross-field diffusion coefficient), develops along
the field into which ions and electrons diffuse across field lines to replenish parallel
particle loss to the solid surface. During net ion collection, the large ratio of ion to
electron mass prevents the drifting ions from losing significant momentum by collision
with the electrons. Thus, although the bias may influence the cross-field transport rate
and hence L:mb [and there is experimental evidence that it does (MATTHEWS
and
STANGEBY,
1989)], conditions should remain much the same as in the floating probe
case. Until now; there has been no direct experimental confirmation of this assumption.
When the probe is biased into net electron collection, the electrons suffer significant
momentum loss by collision as they are drawn to the probe through the comparatively
stationary ion fluid and it is proposed (BOHMet al., 1949; COHEN,1978; STANGEBY,
1982) that an electric field develops along the probe flux tube [now of length L;,
where L; >> Limb(MATTHEWS
and STANGEBY,
1989)] to aid the electrons in overcoming this friction. As a consequence, the ions find themselves in a retarding field
and a density depression appears somewhere in front of the probe (approximately
one electron/ion collision mean-free-path in front). Figure 1 illustrates schematically

Tests of Langmuir probe theory

1239

FLUX TUBE

:r
in1

Preshea t h

--]

I .P

Sheath

FIG.1 .-Qualitative variation of the sheath and presheath potential distribution along the
magnetic collection tube of a probe under floating conditions and biased into net ion and
electron collection.

the behaviour of the potential in the presheath of a probe biased into net electron
collection (but not electron saturation).
Use of the Boltzmann relation (based on TI)for the ions in this presheath retarding
field provides a connection between the density at the crest of the hill potential and
the magnitude of the potential. Combining this with a formulation for the parallel
electron current leads to an expression (STANGEBY,
1982) relating the hill potential to
the probe potential for all V, > Vf,

where yp = V J T , and qH = V,/T, are respectively the normalized probe and hill
potentials ( T , in eV) relative to the distant plasma potential, V,,,,,,, defined as zero,
and z = Tl/Te.The reduction factor is given by (STANGEBY.
1982; TAGLEet al., 1987)

with ;.el the electronjion collision mean-free-path (mfp) and CY = D ; / D ; the ratio of
electron diffusion coefficients across and along field lines (classically. D = ;%,&/4
with C, the Maxwellian mean speed). The transcendental relationship of equation (1)
can be solved for the unknown q H and the result used in an expression for the net
current in order to reproduce the observed characteristic for given r and 5 .
Using the pin and plate probe to be described in the next section, we are able to

R. A. PITTSand P. C. STAKGEBY

1240

measure the magnitude of the potential hill, V,, directly and hence use equation (1)
to check the validity of the model. It is important to note; however, that the theory
incorporates a number of debatable assumptions. For example, part of the electron
cross-field transport is attributed to anomalous cross-field mobility, p y , which is
assumed to satisfy the traditional Einstein relation between mobility and diffusion,
D i p = - kT,je. Since, for classical, ambipolar cross-field transport in a fully ionized
plasma, there is no cross-field mobility (CHEN,1984), the concept of an anomalous,
net electron 1.1- is clearly questionable. In the highly turbulent edge plasma, however,
the existence of mobility-like cross-field transport processes driven by radial electric
fields may play an important role. Indeed, in a recent proposal to protect edge
structures such as ICRH antennas through the creation of an electrostatic-barrier
scrape-off layer using an externally imposed positive bias, LABOMBARD
(1990) includes
cross-field mobility in a non-ambipolar, two-fluid transport model of the boundary
plasma.
3 . EXPERIMENT

Measurements have been made on the DITE Tokamak using the pin and plate
probe (Fig. 2), also described in an earlier paper (STAKGEBY
et al., 1984). A small
tungsten plate of dimension 5 x 10 mm and a cylindrical pin length 5 mm and diameter
1 mm are housed within a graphite probe head which is electrically connected to a
stainless steel probe tube. The pin and plate are rigidly fixed into a machineable-glass
ceramic base and are thus isolated from each other and the probe tube. Their separation is 2.5 mm, measured from the front face of the plate to the pin axis. Bias
voltages may be applied and currents to each element measured simultaneously with
all bias potentials referenced to the torus. In all cases, the plate is used to determine
the local T, and n, since uncertainties in the projected area of the pin during ion
collection can introduce errors in the calculated density (STANGEBY
et al., 1984).
The probe was inserted from the top of DITE through a standard drive assembly

GRAPHITE

CERAMIC

PROBE TUBE

FIG.2.-Cut-away

isometric of the pin and plate probe head

1241

Tests of Langmuir probe theory

along a vertical chord displaced 4 cm outwards of the magnetic axis (R = 1.20 m).
Experiments were performed in deuterium for I, = 100 kA and B, = 2.0 T with the
pin and plate at relatively large minor radii (rprobe 28-29 cm) facing the ion drift
direction. Under these conditions, the shortest connection length to the plate,
L,,, -v 3.9 m, is from the DITE pumped limiter located 180" away and centred on the
outside midplane defining a plasma minor radius of a = 0.24 m. This should be
compared with the collection length defined by the probe housing, L:mb 3.0 m, for
the edge conditions pertinent to this study.

4 RESULTS A N D DISCUSSION

4.1. General remarks

Figure 3 gives an example of the typically observed behaviour. The data have been
obtained during the steady-state phase of the discharge with 2, = 2.2 x 10" m- 3,
r'probe 2: 28 cm. A ramped voltage with period 50 ms is applied to the plate so that the
usual Langmuir characteristic is obtained (Fig. 3a). At the same time, the voltage
appearing on the pin is monitored when the pin itself is floating (Fig. 3b), or the pin
ion saturation current (Fig. 3c) is measured for a large negative bias applied to the
pin (the data in Figs 3b and c were obtained in consecutive identical discharges and
have been smoothed slightly for clarity). Included in Fig. 3a is a list of the standard
parameters derived from the normal non-linear least-squares fit to the characteristic
using only data for V,,,,, d Vglate.The fitted line is also shown. As an illustration of
the point made earlier, if the characteristic in Fig. 3a is re-analysed including all data
up to the point just before electron saturation is reached, a much poorer fit results
and the fitted temperature at T,= 13.8 f2.S eV is more than a factor of two greater.
Note that the plate voltage scan reaches positive values high enough to push the
plate well into electron saturation and that the ratio Z G ~ / Z & 8 is much smaller than
the field free value (i.e. 60 for a pure D - plasma). The variation of VLln with V,,,,,
is a striking illustration of the marked difference between the physics of net ion and
electron collection. If the pin voltage really is a direct indication of the plasma potential
near the plate, then the theory predicts little change in its value for V,,,,, d
Beyond this point, net electron collection occurs and the behaviour of V,,,,,, should
be governed by equation (1) if the theory is correct. Inspection of Fig. 3b shows that
the pin voltage is indeed constant for most of the negative part of the voltage sweep.
This observation rules out any possibility of a (constant) resistive path for leakage
current between pin and plate. In addition, for these conditions, the local Debye
length, A,, = (EoTe/n,e)' * z 25 pm and, since the sheath potential fall occurs over
a distance of order IO& (ALLEN,1974), the piniplate separation (= 2 mm) should be
sufficient to ensure that the pin potential is not directly affected by the plate sheath.

4.2. Net ion collection

Concentrating on the region of plate voltage V,,,,, < V,f,,,,, the pin ion saturation
current remains relatively constant and the ratio I,',, (plate)/Z& (pin) 2: 3.2 is much
lower than the ratio of the projected areas ( 2 lo), in agreement with the findings and
conclusions of STANGEBY
et al. (1984) regarding the ambiguity of the effective ion
collection area of small objects. Also, the pin floats at a fixed value of a few volts
negative with respect to the plate floating potential throughout most of the net ion

1242

R. A. PITTSand P. C. STANGEBY

0.24 -

80 60 40 20 0 20 40 60 80

Plate Voltage (VI

FIG.3.-Typical examples of the experimental data obtained with the probe; (a) plate I-V
characteristic, (b) behaviour of pin floating potential in response to plate bias, (c) behaviour
of pin ion saturation current with plate potential. Plasma conditions as in text.

collection region of the plate characteristic. This is direct experimental confirmation

that most of the bias voltage change applied to the plate is indeed taken up by the
plate sheath. Noting from Fig. 3a that VLlate= -2.3 V, it is clear: however, that
Vi,,,begins the steep increase evident in Fig. 3b before Vplate= Vfplateis attained.
To illustrate this, Fig. 4 concentrates on the behaviour of Vil,, in the small region
of Fig. 3b near the plate floating potential and shows that the increase begins near

1243

Tests of Langmuir probe theory

Y-

'6

>

Vplate (V)
FIG.4.-Expanded

region of V;," from Fig. 3b near the plate floating potential.

V,,,,,
- 15 V at which point VL,,, has decreased to a value of
-2.5 V relative to
its value when V,,,,, = VLlare.Note that VLlateis also marked on this curve and that it
is merely coincidental that Vbln N 0 at this point. It is generally the case, however,
that in our data Viln reaches a constant magnitude some 2-3 V below its value when
VPI,,, = VLlare.The increase in VLln occurs just in the voltage range most important
for the extraction of T,, i.e. - 3kT,/e < V,,,,, < Vilate(for Boltzmann electrons). In
simple probe theory, such a difference would not be expected since, during net ion
collection, the plasma potential is assumed to remain unchanged from its value at

Vpldte

f
Vplare.

There are two possible explanations for this observation :

1) A small fraction of the bias voltage appears not across the plate sheath, but
across magnetic field lines completing the biasing curcuit. If this is the case, Fig.
4 implies that the maximum error incurred in using data for VPidre< VgId,,to
extract T, will, in this example, be 2.5115, that is, an error of 17%. A survey
of all available data showed this result to remain approximately true although
our measurements encompass a T, variation of only a few electron volts.
As
the plate voltage decreases below VLlare,electrons are repelled in increasing
(2)
numbers from the plate and some will be collected by the nearby pin. Since, by
definition, the floating pin must collect zero net current, its potential must
decrease to compensate for the excess negative current and the decrease will
continue until V,,,,, = - 3kT,/e when the maximum possible number of electrons are being repelled.

Of course, if the effect is due to the second process, our results indicate that there will
be a relatively small error in the electron temperatures deduced from the characteristic

R. A. PITTSand P. C. STAXGEBY

1244

using data in this region. However, since the nature of cross-field diffusion is not
understood, we cannot rule out the possibility that a small fraction of the bias voltage
is appearing elsewhere.
4.3. Net electron collection
Above the plate floating potential, VLin increases monotonically with V,,,,, (Fig.
3b), indicating the formation of a potential hill in front of the plate as predicted.
Recalling that the crest of the hill is expected to form -A,, away from the plate
(STANGEBY,
1982) it is clear that, since Ael is in the range 10-30 cm for these conditions
(depending on choice of ZeE,T, etc.), the pin will not be located at this position.
Provided, however, that the piniplate separation is small in comparison with the
distance between the hill crest and the plate, the pin should be a measure of the
potential rise. This is illustrated in Fig. 5, where the potential distribution in the
presheath has been approximated as linear for convenience. In the simple fluid theory
of plasma flow to a surface (STANGEBY,
1986), a total presheath drop of -0.5kTJe
exists to accelerate the ions to sonic speed by the time they reach the sheath edge. For
a stable sheath to form in net electron collection, the same -0.5kTJe must also exist
from the top of the potential hill to the plate sheath. Thus, as shown by Fig. 5, the
magnitude of V , should be correctly monitored by the pin.
When the plate is biased into electron saturation, the sheath no longer exists and
the same conclusions regarding the presheath fall do not apply. Indeed, the theory
(STANGEBY,
1982) applies only up to the point where electron saturation begins.
Beyond this point, the plasma is unshielded from potentials applied to the plate and
is presumably strongly perturbed. It is evident from Fig. 3b, however, that no sharp
break is observed in the behaviour of VLln at the value of V,,,,, corresponding to the
onset of electron saturation. Such a change in gradient might be expected if. in the
absence of a sheath, the plasma potential simply followed that applied to the plate,

VpIasma

/
Location of pin

'i

-cl

Sheath

FIG.5.--Schematic

of the potential variation in the presheath showing how V;)"should be

a direct measurement of V,.

1245

Tests of Laiigmuir probe theory

although this portion of the characteristic is difficult to interpret owing to the substantial plasma disturbance caused by drawing large electron currents.
Proceeding with the assumption that Vkln gives V , directly [at least until I;, (plate)
is reached], it is instructive to compare the experimental data with the variation of
VH predicted by equation (1). To do this requires only input values of r and T, since
T, is determined from the plate Langmuir characteristic. Once TI has been chosen, r
may be calculated from equation (2) if Re, and D are known and D Iis assumed
classical.
Retarding field analyser measurements (PITTSet al., 1989) in deuterium at the same
poloidal. toroidal and radial location and for similar global operating conditions as
those relevant to these experiments, indicate values of TI in the range 20-30 eV.
Choosing T, = 25 eV gives 7 4.5 for the example in Fig. 3. The electronlion collision
mfp is given by (WESSON,1987)

13 for
where T, is in eV, n, in m P 3 and with log, A the Coulomb logarithm
conditions in these experiments). Mass-spectrometric measurements of the charge
state distribution in the DITE boundary (MATTHEWS,
1989) indicate the effective
charge to be 3 in deuterium discharges. The density may be calculated in the normal
way (STAKGEBY,
1986) using the value of Is;, from the characteristic in Fig. 3a
( w

where c, = ([e(T,+ T,)]/TN~)'~~

and Aplateis the plate area. This gives n, = 3.9 x 10''
m-3, leading to i,,
=,28.5 cm and 0 ; = 1.12 x 10' m 2 S K I . Finally, our estimate of r
requires knowledge of D I
-an experimentally unknown quantity. Previous measurements of Dymbin the DITE edge (MATTHEWS
et al.: 1987) indicated values comparable
to the empirical Bohm expression (BOHMet al., 1949), i.e. DImbe 0.06TJB 0.17
m 2 s- ' for our example. Setting D ; = Dymb= 0.2 m 2 s-I gives r = 0.60.
A second estimate for the reduction factor can be made directly from the experimental data using an equation from the theory (STANGEBY,
1982) giving the expected
electron saturation current in a strong magnetic field
N

Using the experimental values of I;,, T, and from Fig. 3a and the calculated density,
equation ( 5 ) gives r = 0.80. We note that the relatively close agreement between these
two estimates of r is perhaps fortuitous given the assumptions involved in deriving
the theory (Section 2).

1246

R. A. PITTSand P. C. STANGEBY

Theory and experiment are compared in Fig. 6, where data from Fig. 3b have again
been used. this time concentrating on the behaviour of VLln for V,,,,,
3 VLlate.The
theoretical prediction for T, = 25 eV, T, = 5.5 eV and the two values for the reduction
factor are shown as dotted curves. The theoretical curves are generated by using the
experimental values of Vfilnto calculate V,,,,,
for the given Y and z.Since equation (1)
is only valid for the case of electron collection, the point where V,,,,, = VLiatehas been
taken as the reference point for all voltages.
An evident feature of this comparison is the relatively close agreement between the
theory and experiment in terms of the shape of the dependence on plate voltage. That
the experimental data behave in approximately the predicted manner is encouraging
and indicates again that the increase in Viin is not due to some simple instrumental
effect such as leakage currents. On the basis of scale lengths, one would certainly
expect collisional behaviour since the computed value of iel
is very much lower than
the connection or collection lengths given in Section 3. This helps to justify the use of
equation (1) for comparison, since the theory assumes the potential fall along the flux
tube to develop as a consequence of momentum transfer collisions between the drifting
electrons and the quasi-stationary ions. Moreover, since the theory leading to equation
(1) is derived on the assumption that the electron drift velocity toward the probe
greatly exceeds that of the ions, it will not strictly apply in the voltage region very
near VLldtewhere U; just begins to exceed E D . If instead, the theory and experiment
were not really comparable until V,,,,, is kT,/e above Vglatewhenthe drift velocities
begin to differ significantly, one could argue that shifting the theoretical curves up to
N

- 5 0 -LO -3G -20 -10

10

20 30

LO

50

50 70 80

" w e (")

FIG.6.-Comparison of experimental data with the predictions of equation (1) for fixed
T, = 25 eV, Z,, = 3.0 using two values of r calculated in the text and for a single value of r
using equation (6).

Tests of Langmuir probe theory

1247

Vplareaxis by this amount would approximately allow for the effect. This would
improve the agreement between experiment and theory.
For sufficiently large values of VHIT,, the logarithmic term in equation (1) is
dominated by the exponential and the equation becomes

so that the theoretical curve in this approximation is simply linear with slope determined solely by the ratio z and with an intercept on the abcissa governed by T, and
r . Using once more the experimental values of V , to calculate Vplateand choosing
r = 0.80, T, = 25 eV, equation (6) has been used to generate the third theoretical
curve in Fig. 6. The result is clearly a reasonable approximation throughout most of
the plate voltage range. Equation (6) also demonstrates how future experiments
encompassing a wider range of parameter variation might hope to observe significantly
different behaviour of VH. For example, measurements of T, and T, in DITE (PITTS
et al., 1989) have shown variations in z greater than a factor 2 for a 5-fold increase
in 5,.
Finally, we note that the behaviour of I,+,,(pin)(Fig. 3c) during net electron collection is not predicted theoretically. The abrupt increase when V,,,,, = VLiateis reproducible and is in contrast to the density depression (STANGEBY,
1982) expected in this
range of plate voltage. Given the negative bias on the pin ( - - 90 V), it may be that
the large potential differences between pin and plate are responsible for complex twodimensional effects occurring there (e.g. the deflection of ion trajectories from plate
to pin) and the simple theory becomes invalid. Note, however, that IGt (pin) does
begin to decrease around V,,,,, = 25 V corresponding to the onset of electron saturation when the space potential is reached, the sheath disappears and the plate begins
to repel ions.
5 . CONCLUSIONS

The behaviour of the plasma potential near a biased element in a Tokamak edge
plasma can be monitored directly by observing the changes in potential on a second,
electrically floating element, located a short distance away. Such measurements provide direct evidence for strong perturbations to the local plasma potential when a
probe is above its floating potential. This implies that beyond this point the collected
current no longer obeys the simple exponential electron collection assumption generally applied to the interpretation of Langmuir probes operating in strong magnetic
fields.
During net ion collection, the plasma potential remains fixed until Vpiprobe
N 3kTJe
below the floating potential and then rises a few volts until Vp/probe = VLrobe.This may
be due either to an extraneous effect or to some of the probe bias voltage appearing
not across the sheath, but elsewhere in the biasing circuit. If this is the case, our data
indicate that an error of 15-20/0 should be associated with any value of T, derived
from analysis of the magnetized probe characteristic using only data for vpiprobe d
VLrobe.This is in addition to any error generated in fitting the characteristic.
The experimental dependence of the local plasma potential on probe bias during

1248

R. A. PITTSand P. C. STANGEBY

electron collection is in reasonable agreement with the predictions of simple theory.

This is despite the assumptions inherent in the model and the lack of any good estimate
of the electron diffusion coefficient across magnetic field lines. It does, however. help
to increase confidence that the basic picture of the physics of electron collection (BOHM
et al., 1949; COHEN,1978; STANGEBY,
1982) is of some value. In turn this suggests
that further related studies might improve our understanding and hence enable the
Langmuir probe to be used as a valuable diagnostic tool not just for T,. but also for
T,, Z,, and D; .
Acknowledgements-The authors would like to thank Drs G . M. MCCRACKEN
and G . F. MATTHEWS
for
helpful comments on this work and the DITE team for their assistance in making these measurements
possible. PCS acknowledges support from the Canadian Fusion Fuels Technology Project.
REFERENCES
ALLENJ. E. (1974) in Plasma PhjJsics (Edited by B. E. Keen). IOP Publishing London.
BOHMD.; DURHOPE. H. S. and MASSEY
H. S. W. (1949) in Characteristics of Electrical Discharges in
Magnetic Fields (Edited by A. GLTHRIE
and R. K. WAKERLING).
McGraw-Hill, New York.
BUDKYR. and MANOS
D. M. (1984) J . ?hicl. Mater. 121, 41.
CHEKF. F. (1984) Introduction to Plasma Physics and Controlled Fusion, 2nd Edn. Vol. 1. p. 188. Plenum
Press, New York.
COHENS. A. (1978) J . Nucl. Mater. 76 & 77, 68.
LABOMBARD
B. (1986) Ph.D. Thesis, Plasma Fusion Centre, MIT, U.S.A.
LABOMBARD
B. (1990) Nucl. Fusion 30, 485.
LAUXM., HERRMANN
A., PECHP. and REIKERH. D. (1989) PTOC.
16th European Con$ on Controlled
Fusion and Plasma Physics, Venice, 13-17 March, Part 111, 883.
MAKOSD . M. and MCCRACKEN
G . M. (1986) Physics ofPlasma-Wall Interactions in Controlled Fusion
(Edited by D. E. POSTand R. BEHRISCH).
Plenum Press, New York.
MATTHEWS
G . F. (1989) J. Nucl. Mater. 162-164. 38.
MATTHEWS
G . F. and STANGEBY
P. C. (1989) J . Phys. D: Appl. Phys. 22, 644.
MATTHEWS
G . F., STAKGEBY
P. C. and SEWELLP. M. (1987) J . Nucl. Mater. 145147, 220.
PITTSR. A . , MCCRACKEN
G. M . and MATTHEWS
G. F. (1989) Proc. 16th European Con$ on Controlled
Fusion and Plasma Physics, Venice, 13-17 March, Part 111, 955.
STANGEBY
P. C. (1982) J . Phys. D: Appl. Phjs. 15, 1007.
STAKGEBY
P. C. (1986) Physics of Plasma-Wall Interactions in Controlled Fusioiz (Edited by D. E. POST
and R. BEHRISCH).
Plenum Press, Kew York.
STASGEBY
P. C.. MCCRACKEN
G. M.. EREKTS
S. K. and MATTHEWS
G. F. (1984) J . Vac. Sei. Technol.
A2, 702.
TAGLEJ. A., STANGEBY
P. C. and ERENTS
S. K. (1987) Plasma Phys. Concr. Fusion 29,297.
WESSON
J. A. (1987) Tokamaks. Clarendon Press, Oxford.