Home

Search

Collections

Journals

About

Contact us

My IOPscience

The plasmasheath boundary region

This content has been downloaded from IOPscience. Please scroll down to see the full text.
2003 J. Phys. D: Appl. Phys. 36 R309
(http://iopscience.iop.org/0022-3727/36/22/R01)
View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 147.46.38.48
This content was downloaded on 28/04/2015 at 14:54

Please note that terms and conditions apply.

INSTITUTE OF PHYSICS PUBLISHING

JOURNAL OF PHYSICS D: APPLIED PHYSICS

J. Phys. D: Appl. Phys. 36 (2003) R309R320

PII: S0022-3727(03)61082-X

TOPICAL REVIEW

The plasmasheath boundary region

R N Franklin
Oxford Research Unit, Open University, Boars Hill, Oxford OX1 5HR, UK

Received 19 March 2003

Published 30 October 2003
Online at stacks.iop.org/JPhysD/36/R309
Abstract
In this review an attempt is made to give a broad coverage of the problem of
joining plasma and sheath over a wide range of physical conditions. We go
back to the earliest works quoting them, where appropriate, to understand
what those who introduced the various terms associated with the structure of
the plasmasheath had in mind. We try to bring out the essence of the
insights that have been gained subsequently, by quoting from the literature
selectively, indicating how misunderstandings have arisen. In order to make
it accessible to the generality of those currently working in low temperature
plasmas we have sought to avoid mathematical complexity but retain
physical insight, quoting from published work where appropriate.
Nevertheless, in clarifying my own ideas I have found it necessary to do
additional original work in order to give a consistent picture. In this way I
have sought to bring together work in the late 1920s, the 1960s, and now
mindful of the commercial importance of plasma processing, work over the
past 15 years that adds to the general understanding.

1. Motivation

2. Historical introduction

This review has been stimulated by controversy following the

work of Godyak and Sternberg (1990a, b, 2002) and by a
recent publication by Slemrod (2002) showing that the results
obtained by Riemann (1997) in his treatment of a collisional
generationless plasmacollisionless sheath represent a solution
to that problem that in a mathematical sense has well-defined
properties. It follows by analogy that the solutions given by
Su and Lam (1963), Lam (1965), and Franklin and Ockendon
(1970) for similar related problems have the same properties.
Thus, it is opportune to survey the large body of work that
there has been since Langmuir introduced the terms sheath
(1923) and plasma (1928).
In what follows it is necessary to define some terms
by generationless we mean that the creation rate of charged
particles is negligibly small, by collisionless we mean that
the motion of the ions is essentially inertial. These or their
opposites combined with plasma and sheath constitute the
territory we seek to describe.
For simplicity we will, in general, deal here with the onedimensional situation, but we note that for a probe immersed
in a plasma it is perfectly possible to envisage a situation that
is both collisionless and generationless.

We begin by quoting relevant passages from Langmuir, giving

the original two definitions.

0022-3727/03/220309+12$30.00

2003 IOP Publishing Ltd

(1) Let us now assume that the plane electrode be charged to

a negative potential of 100 volts. Electrons will therefore
be prevented from approaching close to the electrode,
whereas positive ions will be drawn towards it. There
will therefore be a layer of gas near the electrode where
there are positive ions but no electrons, and in this region
there will therefore be a positive ion space charge. The
outer edge of this sheath of ions will have a potential of 1
and the positive ions pass through this outer edge with a
velocity corresponding to 2 volts, Langmuir (1923).
We note that 100 V is large compared to kTe /e for a
typical gas discharge, and the implication that the sheath is
a region of effectively zero electron density, i.e. it contains
only positive ions.
(2) Except near the electrodes where there are sheaths
containing very few electrons, the ionized gas contains
ions and electrons in about equal numbers so that the
resultant space charge is very small. We shall use the
name plasma to describe this region containing balanced
charges of ions and electrons Langmuir (1928).

Printed in the UK

R309

Topical Review

Thus, there is an apparent significant difference in

electron density in plasma and sheath. However where
they meet electron and ion densities must be comparable.
The next significant development is usually described as
the criterion attributed to Bohm (1949), namely that the ions
leaving the plasma and entering the sheath are required to
have a speed v
cs where cs2 = kTe /M, Te is the electron
temperature and M the ion mass. But the basic concept is
there in the passage quoted above, and Langmuir (1932) was
even more explicit. The value of the wall potential used by
Bohm, no doubt influenced by earlier work, was also 100 V.
In their groundbreaking paper Tonks and Langmuir (1929)
introduced two new concepts the sheath edge and the plasma
balance equation. The first represented the point where
plasma ended and sheath began, and they identified it with
a point inside the singularity for the plasma solution where for
the full plasmasheath solution the difference in electron and
ion densities reached a certain fractional part. The second
represents the fact that in the steady state the volume generation
rate of charged particles must equal the wall loss rate in the
absence of volume recombination. For their sheath edge it is
clear from their figure 6 that they had in mind 1% as being a
certain fractional part.
The plasma equations obtained by setting electron and
ion densities equal have infinite spatial derivatives in all the
potential, field and charged particle densities at a certain point,
near where Tonks and Langmuir had put their sheath edge.
Whereas the collisionless generationless sheath equations have
the property that at large distances from the wall, all physical
quantities have zero spatial derivatives when the criterion is
satisfied.
The scale lengths of plasma and sheath are fundamentally
different, and this was made explicit by Caruso and Cavaliere
(1962), but they did not take the further step of considering
how to join plasma and sheath for finite plasma size L and
finite Debye length D .
The mathematical tool that allowed a suitable joining of
plasma and sheath, was the method of matched asymptotic
expansions, which had been developed primarily to deal with
the structure of boundary layers in fluid mechanics and in
aerodynamics (Van Dyke 1964). The first application of it
to plasma boundaries was by Su and Lam (1963), and Cohen
(1963). This was followed rapidly by Lam (1965, 1967), and
Su (1967)all concerned with electrostatic probes in plasmas
of varying degrees of collisionality. Blank (1968) dealt with
the positive column in a collisional active plasma, and Franklin
and Ockendon (1970) gave the corresponding treatment for a
collisionless active plasma, in both the fluid approximation and
the free-fall approximation.
What emerged was that there was a fundamental difference
in the structure depending on whether the spacecharge sheath
was collisionless or collisional. In the collision-dominated
case, the plasma and collisional sheath joined smoothly
without the need for a transition layer. But in all cases
when the sheath was collisionless, there was the structure of
plasmatransition layercollisionless sheath, with the scale of
the transition layer varying with the nature of the model used
to describe the plasma.
This has been confirmed by subsequent work summarized
by Riemann (1991), and more recently in Riemann (1997),
R310

Slemrod and Sternberg (2001), Riemann (2000), Benilov and

Franklin (2002), and Riemann (2003). There is now effectively
a full coverage of parameter space in terms of the ordering of
the quantities, distance from the plasma centre to the wall, L,
ion mean free path, i , and Debye length, D , using the method.
Of course D  L, for one to be able to talk of a plasma, but
i ranges from the largest to the smallest length, and so in
terms of the method, which depends on expansions in terms
of a small parameter, that varies from treatment to treatment
quoted above, as they range from i  L to i  D .
In recent years, considerable effort has gone into
generalizing the Bohm criterion to take into account the
velocity distributions of both electrons and ions, the effect of
several species of both positive and negative particles, and the
inclusion of ion temperatures. These we shall review below.
The increase in computing power has led to other
treatments solving the full plasmasheath equations again over
a wide range of parameters (Franklin and Snell 2000, 2001).
Also simulations have been carried out, though generally over
a lesser range (Gozadinos et al 2001).
Thus, a comprehensive description of the structure of the
plasmasheath can now be given, but it must be recognized
that for radio-frequency (rf) generated plasmas a fully selfconsistent treatment including generation is not yet available.
However, we will attempt to relate the rf situation to the dc,
indicating what is generally agreed, and what remains to be
done.

3. Plasmasactive and quiescent

Historically, the first plasmas studied were gas discharges
where the positive column provided examples of steady
state plasmas over a wide range of the similarity variable
pressure plasma sizefrom 0.1 to 103 Pa m in a number
of electropositive gases. In general, ionization occurred by
electron impact in the volume, the volume generation being
balanced by losses to the wall. Such plasmas have number
density distributions that are a function of position with a
central maximum and decreasing monotonically towards the
wall. Such a plasma is described as active.
More recently, efforts have been made to create uniform
plasmas by separating the generation region, and thus
the plasma being studied is generationless. Examples of
devices producing such plasmas are Q-machines, thermally
produced plasmas, and multipolar machines or bucket
sources. Such plasmas are described as quiescent. Typically,
these devices operate at 0.1 Pa m or less, and the presence of
magnetic fields in some cases is over-riding in terms of the
description of the plasma and its boundary regions.

4. Why is the Bohm criterion so universal?

To answer this question we will concentrate on establishing
the fact that in a simplistic fluid model the plasma equations
are singular where the criterion is satisfied, regardless of the
collision model for the ions and regardless of the generation
mechanism for the charged particles.
We take the equations:
d(ni vi )
= G(ne , x)
dx

(generation)

(1)

Topical Review

d(ni vi2 )
ni eE
+ C(ni , vi , x) =
(ion momentum) (2)
dx
M



eV
ne = ne0 exp
(Boltzmann relation) (3)
kTe
and using the plasma aproximation ne = ni = n, find that with
E = dV /dx, and ne0 the electron density where V = 0,
C + 2vi G
dn
= 2
dx
vi (kTe /M)
and

Cvi + G(vi2 + kTe /M)

dvi
=
dx
n(kTe /M vi2 )

Thus, both equations are singular, regardless of the functional

form of G and C the functions describing generation and
collisions, provided that the numerators are everywhere
positive, and that is certainly so if both C and G
0, and
one of them is >0. Equations of this form with C = 0 and for
all geometries were given in Rosa and Allen (1970), and for
generation other than electron impact in Johnson et al (1978).
It follows that there cannot be a collisionally modified
Bohm criterion as has been suggested (Godyak and Sternberg
1990a, b).
In this simplified fluid model due to Woods (1965) we
have neglected the stress tensor.
We add here for completeness, an example of a situation
where the Bohm criterion is not relevant. That is, in plasma
diodes where electrons and ions are generated in equal numbers
at the hot boundaries, the gas being an alkali metal vapour
(Fang et al 1969, Phelps and Allen 1976, Braithwaite and Allen
1981).
We also note that in high current discharges it may be
necessary to take into account the fact that the neutral density
is not constant but recycling of ions from the walls leads to
density depletion in the centre (Allen and Thonemann 1954,
Allen and Coville 1961, Franklin 1963). The Bohm criterion
still applies as noted above even though G is a function of x.

5. Collisionless sheath boundary conditions

The usual conditions that are applied at large distances from the
wall in a collisionless sheath are that the electric field and the
net space charge become vanishingly small. The potential
thus becomes constant at the plasma potential and the electron
and ion densities become equal.
In some circumstances, it turns out to be more convenient
to examine d/dV and treat it as the vanishingly small quantity.
Examples where this has proved particularly useful, relate
to sheaths adjacent to longitudinally uniform plasmas. The
case of a double sheath between two plasmas was treated by
Andrews and Allen (1971). The structure of the double sheath
associated with a hot cathode was described by Prewett and
Allen (1976), who obtained the expression for the equivalent
of the Bohm criterion for this case. This agreed with that given
by Pak and Emeleus (1971) for ion waves in a beam-generated
plasma, in the limit 0.
The expansion of a plasma into a vacuum proved to
be another case where this boundary condition was relevant
(Wickens and Allen 1979).

These examples show how in some situations it is

both convenient and insightful to use the potential as the
independent variable.

6. The structure of collisionless plasma and

collisionless sheath
In an earlier section we demonstrated how universal the Bohm
criterion is.
The essential problem in joining plasma and sheath under
collisionless conditions is how to pass beyond the singularity
in the plasma solution where vi2 = kTe /M. This is coupled
with the fact that the generationless, collisionless sheath
solution is such that all spatial derivatives go to zero at
distances that are many Debye lengths from the wall. The
two solutions are disjoint and it is necessary to introduce
another region to achieve smooth joining and it is this transition
layer that has been the subject of Lam (1965), Franklin and
Ockendon (1970), Riemann (1977, 1991, 1997), and Slemrod
and Sternberg (2001).
The various treatments have subtle differences depending
on the plasma model, but they do have one feature in common,
namely that there is a transition layer and it has a scale length
p
D0 L1p where 1 > p > 0, with D0 the central Debye length,
and thus intermediate in size between plasma and sheath.
For the particular case of the collisionless fluid ion model
we show the structure of the ion and electron spatial density
distributions near the wall obtained by integrating the basic
equations (1)(3) above with G = Zne , where Z is the
ionization rate, C = 0, and also taking into account Poissons
equation d2 V /dx 2 = e(ne ni )/0 .
The central boundary conditions are ne = ne0 , vi = 0,
V = 0, E = 0, and these when fed into equations (1)(3) plus
Poissons equation lead to the requirement in plane geometry
that ni0 /ne0 = 1 + (ne0 /ni0 )2 (D0 Z/cs )2 , while at the wall we
require that the ion directed flux and the electron random flux
should be equal. A convenient parameter is = D0 Z/cs and
we introduce X = xZ/cs , so that the eigenvalue of the problem
is the value of X at the wall Xw LZ/cs .
The method of matched asymptotic approximations gives
this eigenvalue in plane geometry asXw = (/2 1)
[1 + K1 (/Xw )4/5 + K2 (/Xw ) + ] where the expansion
parameter is /Xw D0 /L. K1 is a universal constant
4.4 . . ., and K2 depends on the ion mass because of the wall
boundary condition. However, this method does not give a
solution that is uniformly valid throughout all three regions,
and that was recognized in Franklin and Ockendon (1970).
Nevertheless, we can use the results from it to obtain
estimates of the electric field at different locations in the plasma
since the potential is normalized to kTe /e throughout and thus
the fields in the different regions scale to this quantity divided
by their extent. Furthermore, we can estimate the values of
the field at the point where the plasma and transition layer
merge by estimating the field in mid transition layer to be
4/5
kTe /eDs L1/5 while that at the boundary between transition
layer and sheath it is kTe /eDs . This points to the field at
3/5
the plasma edge being kTe /eDs L2/5 the value advanced
by Riemann (1997).
Displayed in figure 1 are the computed solutions of the
full plasmasheath equations, without any approximation, for
R311

Topical Review

Figure 1. The ion and electron density distributions for a collisionless plasmacollisionless sheath for D0 Z/cs = 0.002 showing
(a) the plasma solution ( ), (b) the full computed solutions (), and (c) the sheath solutions (- - - -) obtained by integrating inwards
from the wall. The asymptotic behaviour is clearly seen and the problem of identifying a simple single plasmasheath boundary manifested.
In order to make the structure visible we show only 7% of the total dimension, i.e. 0.93 < x/L < 1, but we give the full solutions for
0 < x/L < 1 as an inset.

parameters = 0.002 and m/M = 1.36105 corresponding

to argon. Also in the same figure is the plasma solution
corresponding to = 0, and the sheath solutions obtained
by taking the wall values from the integration of the full
equations, and then, with appropriate rescaling, integrating the
sheath equations inwards from the wall. This shows that the
full solution breaks away from the plasma solution near to,
but shortly before, the singular point. That was the Tonks
and Langmuir sheath edge and which, with present insights,
is more appropriately named the plasma edge. After the
transition layer the full solution becomes asymptotic to the
sheath solution and one might then talk of a sheath edge
in this vicinity. This is where the sheath solution and the
transition region solution, echoing the words of Tonks and
Langmuir, are such that the electron and ion densities of the
two solutions are once again within a certain fractional part.
It is important to note that it is within the transition layer that
the Bohm criterion is satisfied, and that this point cannot be
located with precision. This fact is clearly demonstrated in
figure 4.11 of Franklin (1976) where a relatively large value of
(0.0144) was chosen to make it visible on a diagram.
Furthermore, for this floating wall situation, the electron
density at the wall is 1.46% of its central value in hydrogen,
0.23% in argon and 0.095% in mercury; i.e. the electron density
while small, is not vanishingly small.
For such a small value of as given above (0.002), the
analytical intermediate solution is effectively indistinguishable
from the full computed solution, and this is borne out by
the relative values of the eigenvalue obtained analytically
and computationally, viz 0.6030, cf 0.6033 for argon and
= 0.002.
It has been suggested that neglect of volume ionization
gives rise to the problem of simply joining the plasma and
sheath solution, usually described as patching, at some suitably
selected point. But in the intermediate region, while the
relative magnitude of the terms concerned is reflected in the
rescaling in the analytic method, none is neglected, and it is
for this reason that the asymptotic behaviour shown in figure 1
is as it is.
R312

Figure 2. The plasma density N , the electron density Ne , and the

ion density Ni near the plasma edge for a collisional
plasmacollisional sheath for the constant ion mean free path case.
4/5
The spatial coordinate Y (xp x)/D0 L1/5 . Thus, the plasma
solution is zero at x = xp N n/n0 (L/D0 )4/5 . The extent of the
sheath is determined by the wall boundary conditions, scales as
4/5
D0 L1/5 , and the sheath solutions join the plasma smoothly after a
few scale lengths.

We discuss in a later section the magnitude of the errors

introduced when one introduces the expediency of simply
patching.

7. The structure of collisional plasmacollisional

sheath
In order to emphasize the difference between the collisionless
case and the collision-dominated we show in figure 2 the
structure of the region near the wall for the collisional situation
with constant ion mean free path (Benilov and Franklin 2002).
It is clear that the plasma solution which varies
parabolically near N = 0, and the sheath solution, join
smoothly. N the normalized density here is (n/n0 )(L/D )4/5 .
There is no need for an intermediate layer.
The scaling of the collisional sheath is similar to that of
the collisionless transition layer in terms of the parameters

Topical Review

involved, but that has to do with the fact that the density varies
parabolically near the plasma edge in both cases, rather than
any deeper physical reason.
If the ion motion model is of constant collision frequency
then the density varies linearly and the sheath dimension is of
2/3
order D0 L1/3 (Blank 1968).

8. When is a sheath thick?

In the situation of the positive column at low pressure with
a floating wall the structure of the plasmasheath is in the
notation used by Lam (1967)plasma, transition layer, thin
electron sheath.
If the wall is an electrode whose potential is determined
by other external variablesthis could equally well be a
negatively biased probethen the structure has additionally
what Lam called a thick ion sheath. For such a situation, the
ChildLangmuir description is a good approximation of that
region, and this yields an expression for the extent of the thick
ion sheath25/4 /3(eVs /kTe )3/4 Ds or 25Ds for Vs the bias
voltage 100kTe /e. For such a situation, the thick ion sheath
may well be more extensive than the transition layer.
This specific case is the subject of the asymptotic
treatment of Benilov (2000) which contains additional insights,
effectively identifying the scale of the thin electron sheath.
When the sheath is thick the ions strike the wall with
energies such that secondary electron emission may occur and
this adds a further degree of complication. Such a situation
was analysed by Prewett and Allen (1976), though the specific
case they covered was an electron emitting cathode adjacent to
a plasma. This has been followed by further work by Benilov
and Coulombe (2001) in relation to high pressure lamps.

9. Patching plasma and sheath

There have been many attempts over the years to modify the
plasma and sheath solutions to enable them to be simply joined.
Basically what is envisaged is that the plasma solution is taken
to be valid up to some specific point and the generationless
sheath solution is joined to it and then integrated to the
wall. An interesting example is to be found in Allis and
Buchsbaum (1967)historically they were post-Bohm, but
without explicit recognition of the transition region, they were
unable to achieve their object of a smooth joining of plasma
and sheath.
It appeared that the problem had been satisfactorily
resolved with the work of Franklin and Ockendon (1970),
and Riemann (1977, 1991). However, Godyak and Sternberg
(1990a, b) intervened asserting that the transition was not
necessary, and that the mathematical ideas behind it introduced
unphysical complexity.
This has caused a considerable degree of confusion
especially since that paper also advanced the concept of a
collisionally modified Bohm criterion.
As we have indicated above the plasma solution is
singular where the Bohm criterion is satisfied regardless of
the collisionality.
What happens as the plasma becomes collisional is that
the density at the plasma edge falls, the corresponding Debye
length increases, and at some point ions undergo at least

Table 1. Normalized wall values of the electron density, ion density,

ion flux, ion speed, potential, field, and physical size, obtained by
computing the full plasmasheath equations and by patching. The
gas is hydrogen and the parameter = ZD0 /cs = 0.002. The
difference in percentage terms is indicated as the error. The effect of
ignoring the transition layer is evident. The wall values for the
patched case correspond to those given in figure 3. w is the
eigenvalue ZLw /cs .
New

Niw

Fluxw

Uw

w

Ew

w

Computed 0.0298 0.1989 0.5102 2.566 3.513 411.5 0.5999

Patched
0.0292 0.1935 0.5000 2.584 3.536 534.5 0.5776
Error %
2.1
2.8
2.0
0.7
0.6
29.9 3.9

one collision in traversing the transition region and they

consequently do not pass through the Bohm speed. Then
the structure of plasmatransition layercollisionless sheath
gives way to that of collisional plasmacollisional sheath as
described for the probe situation by Su (1967) and Blank (1968)
for the active positive column.
This picture has been confirmed by recent extensive
computations carried out by Franklin and Snell (2000, 2001).
In order to be specific we give in table 1 the error in using
patching, for the case of a floating wall and = 0.002, the gas
chosen was hydrogen, and the patch point at the point where
E = kTe /eDs . It is seen that the error is tolerable in some
quantities, but not in others. Again, it must be emphasized that
these are wall values.
Since the wall potential is determined by flux equality, it
shows the least error, the ion speed at the wall Uw is likewise
little in error. However, the fact that the transition layer has
been excised affects the densities and the ion flux because
of ionization occurring in the transition layer and the total
dimension is significantly in error. But the field at the wall is
the quantity that is most seriously in error because the potential
difference has been forced to occur over a significantly shorter
distance.
We give as figures 3(a)(c) the spatial variations in
the volume of the charged particle densities, ion speed and
potential, and finally, the field. The gradient of the field, or
net space charge, as seen in figure 3(c), is zero on the sheath
side of the patch point but kTe /e2D0 on the plasma side. It
is far from continuous, and so the idea that patching can give
smooth solutions is not well-founded.

10. When is the ion motion inertial and collisional?

The simplified ion fluid equation of motion that we have been
using is



vi dvi
eE
Zne
(4)
vi + i vi =
+
ni
M
dx
and the first term on the left-hand side represents inertia and
the second and third friction, the fluid speed vi is continously
renormalized to ensure that momentum is conserved and the
ions can be described by a single speed. We can therefore
talk in terms of inertial motion or collisional, depending on
which terms dominate. This section and the subsequent one
will concentrate on what can be learned from this ion fluid
equation of motion.
R313

Topical Review

(a)

(b)

(c)

Figure 3. Patched solutions showing (a) charged particle densities

(b) ion speed and field, and (c) the field gradient or space charge.
While (a) and (b) are continuous, (c) is discontinuous at the patch
point.

In the plasma ne = ni , and thus we are comparing

U dU/dX with (1 + i )U , where U = vi /cs , i = i /Z and
X = xZ/cs .
From the plasma solution, i.e. (1)(3) with ne = ni we
can derive expressions for both of these quantities and identify
a value of U = U where they are equal. This effectively
marks the transition from collisional motion to inertial, and
we follow the ion motion thereafter asking whether, within the
plasma, once the ions move out of collisional equilibrium with
the field, they ever return. U can readily be shown to be given
by [i /2(1 + i )]1/2 , and the singular point occurs at U = 1.
But, in general, X = U/(1 + i ) + (2 + i )/(1 + i )3/2
1
tan U (1 + i )1/2 so that we can readily determine the number
of collisions made between disequilibrium setting in and the
plasma edge being reached. It is given by i (Xw X ), and
we show this function in figure 4. Clearly if the ions move
out of equilibrium within the plasma, they continue in that
mode since the number of collisions is always less than one,
regardless of the collisionality.
R314

Figure 4. The number of collisions after disequilibrium sets in, and

before the plasma edge over the full range of collisionality for the
plasma solution. Note that it is always less than one.

However, this raises another pertinent question as to

whether the sheath can be sufficiently long and collisional
that the ions once again return to collisional equilibrium with
the field in a thick ion sheath. It is well-known from sheath
solutions, be they collisionless or collisional (see, e.g. Cobine
(1941)) that the ion speed increases monotonically as does the
electric field and the potential. And it is perfectly possible for
ions to be travelling in collisional equilibrium with the field
in the sheath and their speed exceed the ion sound speed for
the plasma from which they derived, though their motion at
this stage is not linked in any way to the ion sound speed.
Experimental data on ion drift shows this to be true for Te
of the order of a few electronvolts (McDaniel 1964, Brown
1966).
In order to gain further enlightenment, we solved the full
plasmasheath equations for situations where the sheath length
is comparable to the plasma dimension, plotting the ratio of the
inertial term to the electric field, i.e. the ratio of the first term
in (4) to the right-hand side, for a wide range of values of
i . What is immediately obvious from figure 5 is that there
is a significant break point in the solutions near the plasma
sheath boundary, and we will examine that further later. But
it is also true that for values of i
100, while there is some
influence on the ion motion near the boundary, the ions do not
move significantly out of equilibrium as the electron density
goes to zero, and soon return to that condition. However,
the rapid change in field in this region means that there is
an influence on the ion motion that is still discernible at quite
high collisionalities.
We then sought to characterize the point of maximum
disequilibrium by determining the electric field at that point.
The results are shown in figure 6 where it is seen that,
over a wide range of collisionality, the field is given by
eED0 /kTe 1. This is reminiscent of the value proposed
by Godyak and Sternberg (1990a, b) for patching of plasma
and sheath. But it is important to note that here we have the
central plasma Debye length, whereas the Debye length at the
1/4
plasma edge Ds tends to infinity as D0 i as i becomes
large. Thus, we are not simply recovering by another route
eEDs /kTe = 1.
However, on the basis of this evidence it is possible to
conclude that in situations where there is a thick ion sheath,
it may be possible to obtain solutions of sufficient accuracy
for some purposes by patching plasma and sheath. This

Topical Review

Figure 5. Values of 1 R where R is the ratio of the inertial term in the positive ion fluid momentum equation to the electric field for
collisionality i /Z = 1, 10, 30, 100, 1000, and 10 000. R Mv(dv/dx)/eE. In the first three cases, the Bohm criterion is satisfied and the
ions remain in disequilibrium even though the thick ion sheath is comparable in size with the plasma. In the second three, inertia never
dominates and for 10 000 the sheath edge is scarcely detectable, i.e. the ions are everywhere in collisional equilibrium. = 0.002.

11. Determining the return to collisional equilibrium

Figure 6. The value of the quantity eED0 /kTe at the point of

maximum disequilibrium over a wide range of collisionality
1 < i /Z < 1000 showing that this quantity varies by less than a
factor of 2 over the range.

clearly requires the thick ion sheath to be much larger than

the combined extent of the transition layer and the thin
electron sheath when the plasma is collisionless, and very
many Debye lengths long when collisional. Furthermore, it
does not negate the existence of the detailed structure found
at low collisionality (Franklin and Ockendon 1970, Riemann
1991, 1997, Slemrod and Sternberg 2001, Kaganovich 2002a).
And the changes in the physical quantities in these regions
should be properly taken into account, before patching.
Thus, unless one recognizes that the plasma edge and the
sheath edge are at different spatial positions albeit only a few
Debye lengths apart, there will be problems arising from the
different conceptions of what constitutes plasma and what
sheath. It is for this reason that the recognition of a transition
layer as necessary to achieve a smooth variation of all physical
quantities is so important, and why we have pointed up the
conceptual differences between quotations (1) and (2).

In this section we examine in greater detail the situation in

the sheath when it is so extensive that the ions do return to
collisional equilibrium.
This involves solving the ion equation of motion in the
situation where there is no generation because the electron
density has gone to zero and thus the relevant equations are
vi dvi /dx + i vi = eE/M, or vi dvi /dx + vi2 /i = eE/M, and
dE/dx = ni e/0 , ni vi = constant = ns vs . Not surprisingly
it turns out to be easier, and mathematically more elegant, to
treat this problem in time in the constant collision frequency
case.
We give that case first, substituting d/dt for vi d/dx.
The combined differential equations reduce to the relatively
simple form
vi + i vi =

nis vis e2
2
vis
= pis
0 M

(5)

where  denotes differentiation with respect to time. This

2
vis t/i +
can readily be integrated to givevi = vis + pis
2
2
i t
2
1] and integrating this again
(vis pis /i cs /D0 i ) [e
with respect to time allows one to relate vi and E to distance
in the thick sheath. Clearly, the motion settles down to
equilibrium with a time constant 1/i , and does so from above
2
or vis > 2Ds i /D0 , and otherwise from
if vis > cs2 i /D0 pis
2
vis t/i as
below. Indeed the long time behaviour is vi pis

expected from (5) when vi  vi i . Since then the inertial term
is no longer of significance in determining the ion motion.
The constant ion mean free path case has correspondingly
vi dvi /dx+vi2 /i = eE/M, ni vi = nis vis and dE/dx = ni e/0 .
2
vis with a corresponding large
This gives vi + 2vi vi /i = pis
t solution vi = vis i /(i + 2vis t) + pis (vis ti )1/2 i /4t +
for t  i /vis .
R315

Topical Review

The constant collision frequency model allows a smooth

transition as it increases to the case considered by Schottky
(1924) with constant ion mobility. However, the constant
ion mean free path is more accurate when charge exchange
collisions are the dominant collisions that the ions undergo.
Of course there are physical limitations to these
descriptions but they are not significant until the ion energies
reach 1020 keV when the effective ion collision cross-section
decreases with energy, and thus there is the possibility of ion
runaway (Massey and Burhop 1952).

12. The stress equation for collisionless plasma and

for collisionless sheath
If one takes the Poissons equation for a collisionless,
generationless sheath and integrates it, one derives a stress
equation that describes the physical situation.
Thus, we begin with d2 V /dx 2 = e(ne ni )/0 and
then substituting ne = ns exp(eV /kTe ) and ni vi = ns vs ,
with vs2 = kTe /M and vi2 = vs2 eV /M, obtain an
equation which reads2ns kTe = ns kTe exp(eV /kTe ) + ns kTe
(12 eV/Mvs2 )1/2 0 /2(dV /dx)2 .
This is the stress balance equation for a collisionless sheath
(Allen 1974) and holds throughout the sheath from the plasma
to the wall.
However, it does not give credence, or validity, to the
suggestion that the point where 0 (dV /dx)2 = ns kTe has the
significance suggested by Godyak and Sternberg. The three
terms on the right-hand side are such that the first goes rapidly
to zero, while the others grow at a rate that is limited by the lefthand side. Thus, on this basis, the so-called Godyak point is
not significant. Throughout the sheath once the electron term
has gone to zero, 0 (dV /dx)2 and ns kTe may be of similar
magnitude, but where they are equal is of no significance.
Equally, one can compare the magnitudes of the terms
ne kTe , ni Mvi2 , and 0 E 2 /2 in the plasma solution. The ratio
of the first two is simply U 2 and that of the first and last can
be shown to be ( 2 /2Ne )(U/(1 U 2 ))2 . Thus, the electric
stress within the plasma is of significance only within a Deybe
length or two of the boundary, i.e. where we have placed the
plasma edge.

13. Criteria for the development of a sheath

Here we give comments stimulated by the recent work of
Valentini (2000) who sought to characterize space charge
separation and how and why it occurred.
His standpoint then was to ask the question What
conditions must be fulfilled for the electron density to fall more
rapidly than the ion density while still essentially within the
plasma?.
In order to answer the question he posed, it is necessary
once again to separate the equations involved into a hierarchy.
Thus, we maintain an electron distribution given by ne =
ne0 exp(eV /kTe ), but describe the ions by their equations of
motion and of generation, viz vi dvi /dx + vi i + vi Zne /ni =
eE/M, d(ni vi )/dx = ne Z, and then ask under what conditions
is |dne /ne dx| > |dni /ni dx|, or rather, |dne /ne dx| >
|dni /ni dx|. This, when we take into account the fact that vi
R316

is monotonically increasing, i.e. dvi /dx > 0 gives the two

conditions



eE(1 vi2 /cs2 )
eE
Zne
i +
> vi >
M
ni
M(i + 2Zne /ni )
We conclude that a sheath always develops if there are
collisions, and also if Mvi2 > kTe at some point. This is, in
itself, a significant insight. It demonstrates that there are two
regimes namely the collision-dominated, and the collisionless.
The latter is characterized by the Bohm criterion, but the
former is one where the Bohm criterion has no significance.

14. Generalizations of the Bohm criterion

The descriptions given above have all been limited by their
assumption that the ions behave as a simplified fluid,
i.e. have a common speed, and the electrons obey a
Boltzmann relation. These restrictions have been relaxed by
Harrison and Thompson (1959) who showed that for an ion
function fi (vi ) the corresponding expression was
distribution

2
f
(v
)/Mv
i
i
i dvi
0 fi (vi )/kTe dvi .
0
The further generalization to allow the electrons to have
a non-Maxwellian distribution was given by Riemann (1995),
and can be written as


2
(1/mve )fe (ve )/ve dve
0 fi (vi )/Mvi dvi

+ 
=0
0 fi (vi ) dvi
fe (ve ) dve
These expressions place requirements on the behaviour of both
fi and fe near vi = 0 and ve = 0.
Other generalizations can be found in situations where
there are more than one species of charged particle. At
low pressures when one has two negatively charged species
obeying Boltzmann distributions with different temperatures
Te and Tn , Braithwaite and Allen (1988), considering the
plane sheath equations for an electronegative plasma, and
also the probe-sheath equations, showed that the analogous
expression was
Mvi2 =

kTe Tn (ne + nn )
ne T n + n n T e

(6)

Equally, they showed that for a bi-Maxwellian electron

distribution the expression was (6), with Th replacing Te and Tc
replacing Tn , because if the negative particles obey their own
Boltzmann relation they are independent of mass. Franklin
and Snell (1992), considering the active plasma equations,
showed that they were singular where (6) is satisfied. These
expressions can be rewritten in terms of density ratios and
temperature ratios, and the density ratios in particular, may
well be local values rather than quantities that are constant
throughout the plasma. Further subtleties of such plasmas will
be considered in a later section.
The case of two positive ion species was considered by
Tokar (1994), but had been examined earlier (Cooke 1981).
Designating the species 1 and 2, the sheath equations allow
one to obtain (n1 + n2 )/kTe = n1 /M1 v12 + n2 /M2 v22 . The
same expression can be shown to correspond to a singularity
in the plasma equations (Franklin 2000), but again the densities
are local quantities and so the position where this composite
Bohm criterion is satisfied will not in general coincide with

Topical Review

either species satisfying its own Bohm speed. That occurs

for collisionless active plasmas generated by electron impact
(Franklin 2000), but not for collisional (Franklin 2001).
Further expressions for multi-species cases have been given
by Valentini (1988) and Benilov (1996).
Other workers have examined the ion distribution near
the wall at a level of detail where it is possible to
distinguish those ions that have made a collision in the
sheath region and those that have fallen through the full
plasmawall potential. In this connection we mention the
experimental and theoretical work of Riemann et al (1992),
the analytical/computational kinetic treatments of Vasenkov
and Shigal (2002a, b) and of Riemann (2003).
We conclude this section by noting that some workers have
linked the Bohm criterion with limits in the propagation of ion
waves. The simplest observation is that the long wavelength
dispersion is /k = vD cs in a uniform plasma and thus
the zero frequency, i.e. dc limit coincides with vD = cs . More
detailed consideration has been given by Stangeby and Allen
(1970) and Allen (1976), where the concepts of a Mach surface
and sonic flow are introduced and discussed. For rf generated
plasmas, Riemann (1992) and Allen and Skorik (1993) have
considered the propagation of ion waves and their limits.

15. Can the Bohm criterion be satisfied at more than

one point in a plasma?
The fact that this is so has been known for some time, and
could have been deduced in connection with electronegative
plasmas by Boyd and Thompson (1959), but it was not made
explicit until Braithwaite and Allen (1988). They pointed
out that for certain values of density ratios and temperature
ratios there was a choice to be made because the criterion
in terms of local variables was apparently satisfied for three
different sets of parameters. But the significance of the ratios
had already been appreciated in Wickens and Allen (1979) for
two electron temperature
plasmas, Teh and Tec where the ratio

of Teh /Tec = 5 + 24 10 being a critical value appeared for

the first time.
The same ambiguity was noted for an active electronegative plasma in Franklin and Snell (1992). Sato and Miyawaki
(1992) treated the situation by introducing a double layer
between two different plasmas, but their treatment though
imaginative was not fully self-consistent. And it was the
work of Kono (1999) on probes in electronegative plasmas
that really brought to attention the fact that at low pressures,
where Boltzmann relations apply for all negative species, there
can exist separate plasma regions each terminating in its own
Bohm criterion. The colder (heavier) negative species is the
more confined. In the region between the two different criteria being satisfied, there exists a space charge region which
has been called a quasi-plasma. It is essentially a hot negative species-positive ion plasma, but because particles enter
it under conditions determined by the inner plasma, it is not
in equilibrium, and this gives rise to spatial oscillations in the
variables concerned. The wavelength of the oscillations is of
the order of the local Debye length, but their magnitude and
extent depends on the model employed. For instance, the freefall (TonksLangmuir) model imposes monotonicity of potential and then they are almost tantamount to a double layer. This,

as noted above, was anticipated in the model described by Sato

and Miyawaki (1992). A collisionless fluid model gives them
free rein, while it is known that collisions, diffusion and ion
trapping can all contribute to damping the oscillations (Kono
(2003) and references therein).
A treatment of this problem giving the insights that
matched asymptotic approximations can give, albeit for a
particular model of ion motion, was carried out by Benilov
and Franklin (1999).
Simulations have revealed the same phenomenon, but
to date there have been no experiments under the necessary
conditions of low collisionality, significant temperature ratio,
and a significant proportion of the colder (heavier) negative
species to test theory.

16. Non-local plasmas

One of the most significant developments in the understanding
of gas discharges in the late twentieth century has been that of
treating the positive column as a situation where all physical
quantities, not only the charged particle densities, vary within
the spatial dimension of the plasma.
Much of this development was due to Tsendin (1974)
who concentrated on the electron energy distribution and its
variation with position. The revision of basic ideas in this
connection has been considerably assisted by Ingold (1997)
and his attention to detail.
But since our concern here is with the boundary regions
we concentrate on recent results revealing the nature of the
electron energy distributions near the wall. In particular we
refer to the work of Uhrlandt (2002) which shows that, even
with this refinement, the electron energy distribution near the
wall is closely Maxwellian.
Thus, it seems probable that an equivalent Bohm criterion
could if necessary, be called into play. However, methods of
solution for non-local treatments do not require the criterion
to be invoked in order to solve the problem since Poissons
equation is retained throughout.
But they do indicate why in so many real plasmas,
experimentalists relying on the visual aspect, may have judged
the sheath to be more extensive than a simple generationless
collisionless sheath is.
So one question for the future is When is the Bohm
criterion a useful concept?

17. Rf plasmasheaths
The situation in the positive column plasmasheath with the
electric field that determines the basic plasma parameters being
orthogonal to the radius makes it have a degree of simplicity
which does not necessarily apply when the plasma is generated
by rf fields. The case where the plasma is generated between
parallel plate electrodes is usually described as capacitively
coupled, and that we will treat briefly.
The fact that the driving field is alternating introduces a
new element in that there are time variations, which have to
be taken into account. These include the driving frequency, ,
the electron plasma frequency, pe , the ion plasma frequency,
pi , the ion collision frequency, i , and the ion sheath transit
time i . Now i Ds /cs 1/pi , which reduces the number
R317

Topical Review

of parameters. In the most usual situation pe > > pi

though the inequalities may not be strong, but this enables one
to construct a model which separates out the motion of the
charged particles so that the ions are described by their time
averaged behaviour while the electrons respond to the driving
field. This has the effect that the sheath expands and contracts
with the driving field. Furthermore, the rf fields supply the
energy that causes ionization and thus one has to consider the
mechanisms by which the electrons gain their energy.
One such mechanism that is operative at higher pressures
is Ohmic heating deriving from the electrons colliding with the
background gas, but at lower pressures the sheath dynamics
of the electrons is an important heating mechanism, usually
referred to as stochastic heating. Early work (Lieberman
1988) suggested that this heating might be analogous to one
mechanism believed to be of significance in the acceleration
of cosmic rays, however the hard wall model (Godyak 1976)
on which this was based has recently undergone significant
revision (Gozadinos et al 2001a, b), although the need had
been recognized for some time (Skorik et al 1992).
The problem of describing how the electric field penetrates
the plasma has not yet been solved in a totally self-consistent
manner. Early work was done by Riemann (1992) and
Allen and Skorik (1993) and progress is currently being made
(Kaganovich 2002b, Slemrod 2003).
It has been established for some time that the Bohm
criterion is still relevant at the plasma edge, even though it
is time-varying in position (Riemann 1995).
A further factor which arises is that the sheath acts as
a rectifier and thus there can be at one or both electrodes
depending on the driving field arrangements, a significant dc
bias voltage and the sheath is many Debye lengths long, though
the electrons sweep in and out of this sheath, so that it is not a
simple replication of the dc sheath.
This, in turn, leads to ions striking the electrodes with
energies sufficient to give rise to secondary emission of
electrons, indeed one mode of such discharges known as the
gamma mode (Raizer 1997) relies on secondary emission to
generate the plasma.
The principal parameter describing the oscillating sheath,
in what has come to be called the moving step model because
the electron density is taken to go to zero on a scale shorter
than others under consideration, has been given a different
symbol by different workers, but once one recognizes that
they are related, then it is possible to understand and compare
the underlying physics. The quantity H used by Lieberman
(1988), used by Godyak and Sternberg (1990a, b) and used
by Benilov (2003) are simply related by 2 H 2 = = 4 and
= (J0 /Ji )(pi /) that is the ratio of the oscillating rf current
to the dc ion current through the sheath times the ratio of the
ion plasma frequency to the driving frequency, and so it has a
relatively simple physical interpretation. This quantity, which
is proportional to the ratio of the sheath oscillation amplitude to
the Debye length, is usually greater than one, and the treatment
of Benilov is notable because it solves the problem to higher
order using the method of matched approximations and thus is
able to bridge 1. A different approach is that of Skorik
reported in Allen (1995).
Slemrod (2003) takes the problem one stage further
and seeks to give a description of the whole plasmasheath.
R318

He recovers results that link directly to the fluid model for the
dc case as given in Franklin and Ockendon (1970), but it has to
be recognized that the generation rate is treated as a separately
determined constant and independent of position.
Kaganovich (2002b) has sought to give a description of
the penetration of the rf field into the plasma, but introduces
an arbitrary parameter being the ratio of the ion density in the
sheath to that in the bulk plasma.
Thus, the treatments available to date are not fully selfconsistent and it is for this reason that much use has been
made of simulations in this area, however we will not attempt
to review them here, basically because of the present heavy
computational cost in carrying out simulations over a sufficient
range of parameters to give sufficient physical insight to link
them back into more simple treatments.

18. Experimental work

It is intrinsically difficult to make accurate measurements of the
relevant quantities in the plasmasheath region. Introducing
any material physical probe causes a disturbance on the scale
of interest and thus it is necessary to try to devise means that
are not so intrusive.
The first technique that we are aware of was the use of
electron beams to measure the electric field, and this was used
in relation to the hot cathodeplasma sheath by Crawford and
Cannara (1965), and Goldan (1970) in a more conventional
plasmasheath situation. Goldan obtained good agreement
with the full TonksLangmuir (1929) description covering
the region from plasma to within 2.2 mm = 4.4D of the
wall. In the hot cathode case, theory followed experiment,
but good agreement was found by Prewett and Allen (1976),
who developed the relevant theory.
The technique of laser-induced fluorescence (LIF) allows
the possibility of measuring ion velocity distributions in the
plasmasheath boundary region and from the ion mean speed
to deduce the ion density. This has been applied using argon
ions by Goeckner et al (1992) and Severn et al (2003). These
latter have made comparison with the theoretical description
of Riemann (1997) for the case of a single ion species. And
for two positive ion species, comparison can be made with its
extension (Franklin 2003).
It should be noted that good agreement with Riemann
(1997), which described a quiescent plasmasheath for
a single ion species, had been reported earlier using a
combination of more conventional probe techniques by Oksuz
and Hershkowitz (2002).
A recent diagnostic of the plasmasheath region has
arisen in connection with dusty plasmas, or more correctly, in
situations where there is dust in plasmas, when the motion of
dust particles can be used to effectively diagnose the potential
distribution, we mention in this connection the work reported in
Tomme et al (2000). This brings into perspective the relatively
small differences that exist so far as the potential variation with
distance is concerned, and it is approximately parabolic in the
many treatments we have described.
The LIF technique has been used with effect in rf
discharges to give both space and time resolution by Czarnetzki
et al (1999) and a more general discussion of its use in plasma
diagnostics is given in Freegarde and Hancock (1997).

Topical Review

19. Conclusions
The structure of the boundary region between plasma and
sheath is more complicated than Langmuir envisaged when
he introduced the two terms. Interestingly plasma became
a universal term, whereas sheath got translated into other
languages.
The structure varies with pressure in a significant way,
particularly according as the Debye length associated with the
edge of the plasma Ds is greater than or less than the ion mean
free path i . There are also other subtleties associated with the
nature of the plasma itself.
The Bohm criterion has proved useful in understanding
structure, and provides a check for more complicated
treatments than the monoenergetic ion model that he used,
as it often applies in some limit. Though, in spatial terms,
precisely where it is met, when D and i are finite, is not
easy to locate. However, the fact that most practical plasmas
are such that D /L is a small number, and that in the limit of
D /L 0 it survives as a singularity in the plasma equation
for so many different models of plasmas, has given it an
enduring significance, particularly once its generalized forms
are appreciated.
Practising plasma physicists need to ask themselves the
question How much of the detail of the structure is needed in
my particular application?.
This paper is intended to assist in answering that question,
and to indicate where to look for help.

Acknowledgments
I would like to thank those with whom I have interacted over the
years in developing the understanding of the plasma boundary
regions. Some of them are contributors to this volume, some
have predeceased me, and some as research students were
demanding at the time.
I have tried to make the description as comprehensive and
up-to-date as possible, and am grateful for early sight of some
of the work referred to.
I acknowledge several fruitful discussions with Professor
J E Allen.
Note added in proof A recent paper by Sternberg and
Godyak (2003) recovers all of the results given for plane
geometry and the fluid model for a collisionless plasma by
Franklin and Ockendon (1970) and therefore represents a
considerable retraction on their part.

References
Allen J E and Thonemann P C 1954 Proc. Phys. Soc. B 67 768
Allen J E and Coville C 1961 Proc. IVth ICPIG (Munich) p 278
Allen J E 1974 Plasma Physics ed B Keen (London: Institute of
Physics Publishing) p 136
Allen J E 1976 J. Phys. D: Appl. Phys. 9 2331
Allen J E and Skorik M A 1993 J. Plasma Phys. 50 243
Allen J E 1995 Proc. XXIIth ICPIG (Hoboken) p 316
Allis W P and Buchsbaum S J 1967 Plasma Theory in Electrons,
Ions and Waves (Cambridge: MIT MA) p 146
Andrews J G and Allen J E 1971 Proc. R. Soc. Ser. A 320 459
Benilov M S 1996 J. Phys. D: Appl. Phys. 29 364
Benilov M S and Franklin R N 1999 J. Plasma Phys. 62 541
Benilov M S 2000 IEEE Trans. Plasma Sci. 28 2207

Benilov M S and Coulombe S 2001 Phys. Plasmas 8 4227

Benilov M S and Franklin R N 2002 J. Plasma Phys. 67 163
Benilov M S 2003 IEEE Trans. Plasma Sci. 31 678
Blank J L 1968 Phys. Fluids 11 1686
Bohm D 1949 The Characteristics of Electrical Discharges in
Magnetic Fields ed A Guthrie and R K Wakerling (New York:
McGraw-Hill) chapter 3, p 77
Boyd R L F and Thompson J B 1959 Proc. R. Soc. Ser. A 252 102
Braithwaite N S and Allen J E 1981 Int. J. Electron. 51 637
Braithwaite N S and Allen J E 1988 J. Phys. D: Appl. Phys. 21 1733
Brown S C 1966 Basic Data of Plasma Physics
(Cambridge, MA: MIT)
Caruso A and Cavaliere 1962 Nuovo Cimento 26 1389
Cobine J D 1941 Gaseous Conductors (New York: McGraw-Hill)
pp 1239
Cohen I M 1963 Phys. Fluids 6 1492
Cooke M J 1981 DPhil. Thesis Oxford University
Crawford F W and Cannara A B 1965 J. Appl. Phys. 36 3135
Czarnetzki U, Luggenholscher D and Dobele H F 1999 Plasma
Sources Sci. Technol. 8 230
Fang M T C, Fraser D A and Allen J E 1969 Brit. J. Appl. Phys.
2 229
Franklin R N 1963 Proc. Vth ICPIG (Paris) p 157
Franklin R N and Ockendon J R 1970 J. Plasma Phys. 4 371
Franklin R N 1976 Plasma Phenomena in Gas Discharges
(Oxford: Oxford University Press)
Franklin R N and Snell J 1992 J. Phys. D: Appl. Phys. 25 453
Franklin R N and Snell J 2000 Phys. Plasmas 7 3077
Franklin 2000 J. Phys. D: Appl. Phys. 33 3186
Franklin R N and Snell J 2001 Phys. Plasmas 8 643
Franklin R N 2001 J. Phys. D: Appl. Phys. 34 1959
Franklin R N 2003 J. Phys. D: Appl. Phys. 36 1806
Freegarde T G M and Hancock G 1997 J. Phys. IV (France) 7 C4-15
Godyak V 1976 Sov. J. Plasma Phys. 2 78
Godyak V and Sternberg N 1990a IEEE Trans. Plasma Sci. 18 159
Godyak V and Sternberg N 1990b Phys. Rev. A 42 2299
Godyak V and Sternberg N 2002 Phys. Plasmas 9 4427
Goeckner M J, Goree J and Sheridan T E 1992 Phys. Fluids
B 4 1663
Goldan P 1970 Phys. Fluids 13 1055
Gozadinos G, Vender D and Turner M M 2001a J. Comput. Phys.
172 348
Gozadinos G, Turner M M and Vender D 2001b Phys. Rev. Lett. 87
135004
Harrison E R and Thompson W B 1959 Proc. Phys. Soc. 74 2145
Ingold J H 1997 Phys. Rev. E 56 5932
Johnson P C, Cooke M J and Allen J E 1978 J. Phys. D: Appl. Phys.
11 1877
Kaganovich I G 2002a Phys. Plasmas 9 4788
Kaganovich I G 2002b Phys. Rev. Lett. 89 265006
Kono A 1999 J. Phys. D: Appl. Phys. 32 1357
Kono A 2003 J. Phys. D: Appl. Phys. 36 465
Lam S H 1965 Phys. Fluids 8 73
Lam S H 1967 Proc. VIIIth Int. Conf. Phenomena in Ionized Gases
Invited Lectures (Vienna: IAEA) p 545
Langmuir I 1923 Gen. Elec. Rev. XXVI 731
Langmuir I 1928 Proc. Natl Acad. Sci. XIV 627
Langmuir I 1932 J. Franklin Inst. CCXIV 275
Lieberman M A 1988 IEEE Trans. Plasma Sci. 16 638
Massey H S W and Burhop E H S 1952 Electronic and Ionic Impact
Phenomena (Oxford: Clarendon)
McDaniel E W 1964 Collision Phenomena in Ionized Gases
(New York: Wiley)
Oksuz L and Hershkowitz N 2002 Phys. Rev. Lett. 89 145001
Pak T S and Emeleus K G 1971 Proc. Xth ICPIG (Oxford) p 361
Phelps A D R and Allen J E 1976 Proc. R. Soc. Ser. A 348 221
Prewett P D and Allen J E 1976 Proc. R. Soc. Ser. A 348 435
Raizer Y P 1997 Gas Discharge Physics (Berlin: Springer)
Riemann K-U 1977 PhD Thesis Ruhr-Universitat, Bochum
Riemann K-U 1991 J. Phys. D: Appl. Phys. 24 493
Riemann K-U, Ehlemann U and Wiesemann K 1992 J. Phys. D:
Appl. Phys. 25 620

R319

Topical Review

Riemann K-U 1992 Phys. Fluids B 4 2693

Riemann K-U 1995 IEEE Trans. Plasma Sci. 23 321
Riemann K-U 1997 Phys. Plasmas 4 4158
Riemann K-U 2000 J. Technol. Phys. 41 89
Riemann K-U 2003 J. Phys. D: Appl. Phys. 36
Rosa R and Allen J E 1970 J. Plasma Phys. 4 195
Sato K and Miyawaki F 1992 Phys. Fluids B 4 1247
Schmitz H and Riemann K-U 2001 J. Phys. D: Appl. Phys. 34 1193
Schottky W 1924 Phys. Z 25 635
Severn G et al 2003 Phys. Rev. Lett. 90 145001
Skorik M A , Braithwaite N StJ and Allen J E 1992 XIth ESCAMPIG
vol 16C (Europhysics Conference Abstracts) p 1973
Slemrod M 2002 Euro. J. Appl. Math. 13 663
Slemrod M 2003 SIAM J. Appl. Math. submitted
Slemrod M and Sternberg N 2001 J. Nonlinear Sci. 11 193
Stangeby P C and Allen J E 1970 J. Phys. A: Gen. Phys. 3 304
Sternberg N and Godyak V 2003 IEEE Trans. Plasma Sci.

R320

Su C H and Lam S H 1963 Phys. Fluids 6 1479

Su C H 1967 Proc. VIIIth Int. Conf. Phenomena in Ionized Gases
Invited Lectures (Vienna: IAEA) p 575
Tokar M Z 1994 Cont. Plasma Phys. 34 139
Tomme E B, Law D A, Annaratone B M and Allen J E 2000 Phys.
Rev. Lett. 85 2518
Tonks L and Langmuir I 1929 Phys. Rev. 34 876
Tsendin L D 1974 Sov. Phys. JETP 39 805
Uhrlandt D 2002 J. Phys. D: Appl. Phys. 35 2159
Valentini H-B 1988 J. Phys. D: Appl. Phys. 21 311
Valentini H-B 2000 Plasma Sources Sci. Technol. 9 574
Van Dyke M 1964 Perturbation Methods in Fluid Mechanics
(New York: Academic)
Vasenkov A V and Shizgal B D 2002a Phys. Plasmas 9 691
Vasenkov A V and Shizgal B D 2002b Phys. Rev. E 65 046404
Wickens L M and Allen J E 1979 J. Plasma Phys. 22 167
Woods L C 1965 J. Fluid Mech. 23 315