Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
INIS-mf—8683
FJ.F. van Odenhoven
KINETIC THEORYOF TRANSPORT PROCESSES
IN PARTIALLY IONIZED GASES
KINETIC THEORY OF TRANSPORT PROCESSES
IN PARTIALLY IONIZED GASES
PROEFSCHRIFT
TER VERKRIJGING VAN DE GRAAD VAN DOCTOR IN DE
TECHNISCHE WETENSCHAPPEN AAN DE TECHNISCHE
HOGESCHOOL EINDHOVEN, OP GEZAG VAN DE RECTOR
MAGNIFICUS, PROF.DR. S.T.M. ACKERMANS, VOOR
EEN COMMISSIE AANGEWEZEN DOOR HET COLLEGE VAN
DECANEN IN HET OPENBAAR TE VERDEDIGEN OP
VRIJDAG 18 FEBRUARI 1983 TE 16.00 UUR
DOOR
FERDINAND JOAN FRANCISCUS VAN ODENHOVEN
GEBOREN TE EINDHOVEN
DIT PROEFSCHRIFT IS GOEDGEKEURD
DOOR DE PROMOTOREN
PROF.DR.IR. P.P.J.M. SCHRAM
EN
PROF.DR. M.P.H. WEENINK
Wees niet bang voor het
langzaam voorwaarts gaan,
wees slechts bevreesd
voor het blijven staan
(Chinees gezegde)
Contents
page
I Introduction I
References 5
II Basic equations 6
References 15
III Very weakly ionized gases 16
III-l The electron distribution function 17
III-2 The electron macroscopic equations 23
III-3 Form relaxation of the electron distribution 30
III-4 The inclusion of Coulomb collisions 34
References 37
IV Weakly ionized gases 38
IV-1 Heavy particle results 39
IV-2 Perturbation solution of the electron
distribution function 46
IV-3 The macroscopic electron equations 51
IV-4 The first order isotropic correction 57
IV-5 Electron transport coefficients 61
IV-6 Modifications for a seeded plasma 67
References 69
V Strongly ionized gases 70
V-l Heavy particle results 72
V-2 The electron kinetic equation 77
V-3 The electron macroscopic equations 82
V-4 The nonisotropic part of the
electron distribution 86
References 93
VI Numerical results 94
VI-1 The isotropic correction 95
VI-2 Electron transport coefficients 108
VI-3 Electrical conductivity of
cesium seeded argonplasma 118
References 120
VII Summary and conclusions 121
Appendices
A Expansion of electron-heavy particle collision integrals
A-l Electron-atom collisions 124
A-2 Electron-ion collisions 128
A-3 Moments of the electron-heavy particle
collision integral 130
B Some H-theorems and properties of collision integrals
B-l The zeroth order electron-atom
collision operator 133
B-2 The zeroth order electron-ion
collision integral 134
B-3 H-theorems for the ion distribution function 135
C Harmonic tensors 137
D The Landau collision integral for identical particles
D-l The Landau collision integral 140
D-2 The linearized Landau collision operator
for like particles 143
D-3 Matrix elements for the operators obtained from
the Landau collision integral 147
E Renormalization of the ion multiple collision term 150
References to the appendices 151
Samenvatting .152
Nawoord 154
Korte levensloop 154
-1-
I INTRODUCTION
One can state that the modern kinetic theory of non-equili-
brium processes in dilute gases came to maturity with the works
of Chapman and Enskog^The book by Chapman and Cowling2 has
never ceased to be an Indispensable textbook on this matter.
Since then there have been written many new textbooks , and
much has been added to the theory, especially to the kinetic
theory of plasmas. More complete historical summaries can be
found in the references^.
The method of multiple scales is one of the important tools
used in this thesis. First introduced by Sandri e.a.4 it has
developed into a valuable mathematical device5. It has also
proved to be very succesful in deriving kinetic equations from
the BBGKY-hierarchy6.
The purpose of the present work is the description of transport
processes and the calculation of transport coefficients of
partially ionized gases. The calculations are restricted to
elastic collision processes. This is certainly justified if the
kinetic energy of the electrons is much smaller than the
excitation energy of the first atomic energy level. There are
of course, always inelastic collisions involving high energy
electrons, but their influence on the values of the transport
coefficients is small, because these result from integrations
over the entire velocity space.
In chapter II the basic equations and the multiple time scale
formalism are expounded. The electrons are of special interest,
since they contribute significantly to all transport processes.
Because of their small mass the electrons often have a tempera-
ture different than the one of the heavy particles. If there
are only very few electrons the isotropic part of the electron
distribution function can deviate significantly from an equili-
brium Maxwelllan as a consequence of fields, gradients and
temperature differences which may be present. There are two
limiting cases in which the situation is relatively simple.
-2-
In the fully ionized or Spltzer limit the isotropic part of the
electron distribution function is a Maxwellian and the non-
isotropic part has been computed numerically by Spitzer and
HSrm^. Within the framework of the Landau kinetic equation this
solution is exact.
In the Lorentz limit (very small degree of ionization but
finite electron-atom mass ratio), on the other hand, the
isotropic part is found to be a so-called Davydov distribution
function8. If the neutrals are sufficiently cold, the
Druyvesteyn distribution is a special case of this distribution
for the hard spheres interaction model.
One can distinguish four domains for the electron density with
different orderings in terms of the small parameter e which is
the square root of the electron-atom mass ratio:
hG = (m /m )6 Si
(i-D
Two of these domains contain the already mentioned cases of
very low respectively high degree of ionization. The
definition of the different regions in terms of the ratio of
the electron-electron to electron-atom collision frequency,
which is proportional to the electron-atom density ratio, is
now as follows:
Very WeaklyIonized Gas
Vee „ 2<< e
Vea
NonlinearRegion
" 6 e =#(e 2)ea
Weakly IonizedGas
ea
Strongly IonizedGas
V
ea
Adjacent to the region of the very weakly ionized gases lies a
region where the equation for the electron distribution
function in zeroth order of e is non-linear and the form of the
distribution function varies with the electron density between
a Davydov and a Maxwell distribution.
In chapter III the first two regions are considered. An order-
ing different from the work of van de Water10 is assumed. Some
results additional to his are obtained.
-3-
The strongly Ionized domain is defined as the region where all
collision frequencies of the electrons are of the same order of
magnitude. This region is investigated in chapter V.
It contains as a special case the fully ionized limit, as far
as the electron equations are concerned.
The equation determining the nonisotropic part of the electron
distribution function is written in the form of a differential
equation, which permits easier calculations. In the fully
ionized limit the integro-differential equation solved at first
by Spitzer and Harm is shown to reduce to a simple second order
differential equation.
Between this region and the nonlinear one a fourth region of
interest is situated. Here the electron mutual collision
frequency is smaller than the electron-atom collision frequency
by a factor e. Plasmas in this region are referred to as weakly
ionized. The interesting feature of this region is the appear-
ance of an isotropic correction to the Maxwellian distribution
function which is found in zeroth order of e.
The necessity of an isotropic correction had already been
indicated by van de Water10. His work was, however, restricted
to a Lorentz like plasma with Maxwell interaction between
electrons and atoms. The equation for this isotropic correction
is solved analytically in chapter IV. This correction leads to
contributions to the transport coefficients which are nonlinear
in the fields and gradients. In this way one gets for instance
a correction to the electrical conductivity which depends
quadratically on the electric field. There also appear new
transport processes partly also nonlinearly depending on fields
and gradients. The Onsager symmetry relations do not hold for
these contributions to the transport cofficients. Other contri-
butions are due to the influence of the Coulomb collisions on
the electron-atom collisions, i.e. multiple collisions. These
are linear and obey Onsagers' theorem.
Much work in the field of transport coefficients in partially
fts was motivated by the possibility of direct energy
-4-
conversion by means of an MHD-generator'1. Therefore some
attention is also paid in this thesis to new transport
processes and higher order corrections to transport coeffi-
cients in alkali seeded noble gas plasmas. This attention is
rewarding, because for these plasmas a better comparison with
experiments appears to be possible.
All results of the calculations and the comparisons with
experiments are collected in chapter VI.
The method used in this thesis consists of an expansion of the
unknown quantities into an asymptotic series in the small
parameter e supplemented by the method of multiple time scales.
The general form of the solution f of the relevant kinetic
equation in each order is found in terms of an expansion into
harmonic tensors (see appendix C):
f(c) = f(0)(c) +
+e2(f(2)(c) + f(2)(c).c + f(2)(c):<cc>) +
+ (1-2)
where c is the peculiar velocity, <cc> is the harmonic tensor
of second rank and f denotes an isotropic correction of
order n. Nonisotropic parts give rise to expressions for the
transport coefficients, isotropic parts appear in the contribu-
tions of the nonisotropic part.s in higher order. The expansion
generally used in the litterature is a two-term expansion of
the form:
f(c) = f(0)(c) + f(1)(c).c , (1-3)
which is sufficient for the calculation of transport coeffi-
cients in lowest order. The method applied in this thesis gives
results up to second order in e and describes both fast and
slow transport phenomena by means of the multiple time scales
formalism.
-5-
References
1. S .Chapman, Phil.Trans.R.Soc,216(1916)279,217 (1917) 118,
Proc.Roy.Soc.,A98(1916)1.
Q.Enskog,Inaugural dissertation,Uppsala 1917.
2. S.Chapman and T.G.Cowling:"The mathematical theory of non-
uniform gases",Camgridge University Press 1970.
3. J.O.Hirschfelder,C.F.Curtiss and R.B.Bird:"Molecular theory
of gases and liquids",.J.Wiley 1954.
L.Waldmann:"Transporterscheinungen in Casen von mittleren
Druck",in:iïandbuch der Physik, Springer 1958.
CCercignani:"Mathematical methods in kinetic theory",
Plenum press 1969.
J.H.Ferziger and H.G.Kaper:"Mathematical theory of trans-
port processes in gases", North Holland Publ. Comp. 1972.
4. G.Sandri,Ann.Physics,24(1963)332,380.
E.A.Frieman.J.Math.Phys. ,4^1963)410.
J.E.McCune,T.F.Morse and G.Sandri:Rar.Gas Dynam.J_(1963)115.
5. A.H.Nayfeh:"Perturbation methods", J.Wiley 1973.
6. P.P.J.M.Schram,:"Kinetic equations for plasmas",
Ph.D.thesis Utrecht 1964.
7. L.Spitzer and R.Harm, Phys.Rev. ,89_( 1953)977.
8. B.Davydov,Phys.Zeits.der Sowjetunion,8Q935)59 .
9. M.J.Druyvesteyn,Physica,1£( 1930)61, j_( 1934)1003.
10. W.van de Water,Physica,85C(1977)377.
11. M.Mitchner and C.H.Kruger:"Partially ionized gases",.J.Wiley
1973.
-6-
In order to describe a partially ionized gas one needs at
least three kinetic equations. Henceforth a plasma is
considered which consists of one-atomic neutral particles, ions
and of course electrons. Ionizing collisions assure the
presence of charged particles, but will just as the other
inelastic collisions be neglected when determining the distri-
bution functions for calculations of transport coefficients. If
the plasma is close to equilibrium one may use Sana's equation
to calculate the electron density from the electron
temperature. When the departure from equilibrium is larger, for
example because of radiation losses, it is assumed that the
electron density has been determined by other means. Thus the
collision terras in the Boltzraann equations of the three compo-
nents consist of a sum over all possible elastic collisions
that may.occur:
3f ,— s + vVf + — F «V f = ) J (f ,f ). (2-1)3t - s m -s v s L. st s' t x
s t=e,i,a
The left-hand side of this equation gives the total time
derivative of the distribution function of particles s under
the influence of a force F , for example external forces or as™ s
a result of a self-consistent electric field- The right-hand
side of equation (2-1) describes the variation of f caused bys
all possible elastic collisions.
Macroscopic quantities appear as so-called moments of the
distribution function f . Important quantities are:s
the density n , the hydrodynamic velocity in the laboratorys
frame w , the temperature T , the pressure P and the thermal-s s =s
heat flux <j . These are defined as follows:
n (r,t) = It (r,v,t)d3v, n w (r,t) = /vf (r,v,t)d3v,s - s - - s-s - - s - -
Pg(r,t) = /mgcscsfg(r,v,t)d3v, SB<r,t) = A n ^ c ^ f g(r,v, t)d
3v,
- 7 -
f n k T ( r , t ) = Am c^f ( r ,v , t )d J v , (2-2)i S S ™ S S S " ~
where the peculiar velocity c = v - w .—s — —s
If equation (2-1) is multiplied by appropriate functions of
velocity and integrations over the entire velocity space are
performed one obtains so-called moment equations. Choosing
these functions as: 1, m v, and ^m v2 the moment equations ares - s
the conservation equations for the particle number density,
momentum and energy respectively:3n
j^-+ V-(ngWg) = 0, (2-3a)
3w
s s 3t -s -s -s s-s s~ t*s S t S C (2-3b)
= Am v2{ [J (f f)^v,S t*s St S C (2-3c)
In the energy equation the following notation was introduced:
C = I kT + \m w2. (2-4)
The conservation equation for the particle number density is
called the equation of continuity. Equations (2-3b) and (2-3c)
are also frequently called equation of motion and of energy
respectively. In the right-hand side of these equations the
term corresponding to t=s disappears because it represents
collisions between identical particles for which the above
functions of velocity are collisional invariants1"2. Physically
this means that there is no net exchange of momentum and of
energy between like particles. One could have simplified
equation (2-3c) further with the aid of equation (2-3b) and
have arrived at the following form of the energy equation:
, 3T4 n k{-r-£ + w «7T } + 7«q + P : Vw » ftii c 2 I J d3v, (2-5)2 s l3t -s s ' a s - s - s J s s L st 't*s
- 8 -
a result that can also be obtained directly from equation (2-1)
with the velocity function \m c2. Another quantity of impor-s s
tance is the mass-velocity or plasma-velocity defined as:
£ m n w
s s s
It is possible to define diffusion velocities U with respect
to this plasma-velocity:
U := w - w . (2-7)-s -s -in
In a weakly ionized gas (WIG), however, the density numbers of
the charged particles are small. It follows that the mass
velocity almost equals the hydrodynamic velocity of the neutral
component. For later use diffusion velocities u are defined:
u := w - w . (2-8)-s -s -a
Now return to equation (2-1) and consider the right-hand side
of this equation. It consists of a sum of collision integrals
describing the variation in time of the distribution function
f due to elastic encounters only. One can distinguish two
different types of interaction: one based on a short-range
intermolecular potential and one of the Coulomb type, which
varies as l/r, r being the distance between two interacting
particles. The first of these applies to all collisions between
charged particles and neutral particles and between neutral
particles mutually, and will be described by the well known
Boltzraann collision integral:
J s t ( f s , f t ) = 2/d3M386<l2«1.*>{f8Cr ^ )ft(v+S+ ^ ) +t s t s
-fs(v)ft(v+g)}. (2-9)
Here g = v - v is equal to the relative velocity just before a
collision. The validity of the Boltzmann collision integral is
based on the stnallness of the number of particles in a sphere
-9-
with radius equal to the characteristic range of the potential,
i.e. the Boltzraann parameter. The notation in (2-9) is such
that it shows the integrations to be performed explicitly.
Indicating post-collision variables with a prime, the veloci-
ties just after a collision read:
ra Jl m S.vf = v — ~̂ ~"™" , v l = v + g + — — — , (2—10)— — m +tn —t — a m +nr
s t s t
where I = g1- g denotes the difference in relative velocities
just before and after a collision. The factor I(g,JL) is the
differential cross section and is defined as:
rl . (2-iD
where b is the impact parameter and x *s t n e scattering angle.
It contains the geometry of the collision. The 6-Dirac function
with argument A2-l-2g• Ji assures energy conservation.
Collisions between charged particles are more difficult to
treat because of the 1/r potential. The Landau collision
integral^ will be used, which can be obtained from the
3oltzraann collision integral in the impulse approximation,
based on the assumption that collisions change the velocity
only slightly. But one can also derive the Landau integral
directly from the well known BBGKY-hierarchy. The Landau
collision integral reads:
8!££ i iJ (f ,f ) = C V •ƒ[—- )•{- 7 - - V }f (v)f (v )d3v .st s' t st v n 3 ' lm v m v ' s - t -t t
Ê S C t
(2-12)
For reasons of simplicity only the velocity dependence of the
distribution functions in equations (2-9) and (2-12) has been
indicated. The constants C are given by:s c
q2q2lnA
To s
where q and q are the charges of the collision partners andS L
-10-
inA is the so-called Coulomb logarithm. Herein A is the inverse
of the plasmaparameter e , and is proportional to the number of
electrons in a sphere with radius equal to the Debye lenght r :
A = 7 - ner3. (2-14)
P
In a plasma one distinguishes three characteristic lenghts: the
Debye length A^, which is a measure of the distance over which
the potential of a charged particle is shielded by the surroun-
ding charged particles, the mean interpartlcle distance r and
the Landau lenght r , which is the distance of closest approachLi
between two like charged particle? with thermal velocities.
These lenghts are defined as:
rL = 7Tx> ro=*-1/3' ' D = £ T A C2~15)
One can verify that the plasma parameter is proportinal to the
ratio of the Landau- to Debye lenght, but also that the plasma
parameter connects all three characteristic lenghts in (2-15).
The condition for these lenghts to be well separated is that
the plasma parameter should be very small. The plasma is then
called ideal.
The Landau collision integral results after making two cutt-
off's: in the derivation of this expression there appears an
integral over the interaction distance diverging at zero and
infinity. The approximation made is that one introduces the
lenghts r and r as integration boundaries. This leads to the
factor lnA. This factor has to be much greater than unity.
Speaking in more physical terras one could say that the Landau
lenght is so small that there are relatively very few short
range collisions. Because of the effect of screening the upper
boundary can be replaced by the Debye lenght: collisions with
larger impact parameter contribute little to the collision
integral.
Next the electron Boltzraann equation will be considered in more
-11-
detail. To solve this complicated equation an expansion into a
small parameter e will be used, e being the square root of the
electron-atom mass ratio:
e = <me/maA (2-16)
This choice seems obvious and the next step is that all
dimensionless numbers, obtainable from the dimensionless
electron Boltzraann equation, are expressed as powers of e. The
equations will, however, not be made dimensionless. All terras
will be multiplied by the appropriate power of e. The distribu-
tion functions will be expanded into e and in the end e is put
equal to unity, so that e merely plays a bookkeeping role. For
a weakly ionized gas the electron Boltzraann equation reads as
follows:
3f eE-r-̂ - + evVf - e — «V f - u) (vxb) «V f = eJ + J + eJ .,9t - e m v e ce - - v e ee ea eie
(2-17)wherein b is a unit vector in the direction of a constant
external magnetic field B. The electron cyclotron frequency:
ui = — , has been taken of the order of the electron-atomce m
ecollision frequency:
w T =0(1). (2-18)ce ea
Here T is the mean collision time between two successiveea
collisions of an electron with a neutral atom:
v— = v = n v _ C r ' = r—£ . (2-19)T ea a Te^ea X v 'ea ea
Thermal velocities are defined as v =(kT /ra ) and Q is theAS S o S C
elastic collision cross section for momentum transfer of
particles s with particles t defined as follows:.. . IT
Q s t (g) = J2ita(g,x)(l-cosx)sinxdx, (2-20)o
where g is the relative speed of the colliding particles.
-12-
In expressions like (2-19) some characteristic value for g will
be substituted e.g. v . Furthermore the mean free path A hasle ea
been introduced.
The electric field has been scaled in such a way that the
energy gain of an electron in this field between two successive
collisions with neutral atoms will be compensated on the
average by the energy loss as a result of these collisions.
Then the following order relation holds:
e Te
Concerning the inhomogeneitits the Knudsen number ex defined as
the ratio of \ to some macroscopic length scale L reads:
xp0 (2-22)
where the ordering is in accordance with equation (2-17). The
order of magnitude estimation of the collision terms on the
right-hand side of equation (2-17) depends on the degree of
ionization and the kind of interaction. Because of the long
range of the Coulomb potential the Coulomb collision cross
section for momentum transfer is about 104 times larger than
the electron-neutral cross section. Coulomb collision cross
sections are defined on the basis of a 90° deflection. This is
necessary because of the weak interaction. Scattering is the
result of many grazing encounters.
A weakly ionized gas is defined such that the ratio of the
electron-electron to electron-atom collision frequencies equals
e:
J v n v_ Q ( 1 )
ee H ee e T ^_ H = _ ^ _
ea ea n v„ Qa Texea
The same holds for the electron-ion collision integral.
A strongly ionized gas will be defined as a plasma in which the
collision frequencies satisfy the conditions: v < v = v ..^ J ea ~ ee ei
-13-
Next the heavy particle Boltzraann equations have to be
considered. For a weakly ionized gas one obtains:
3f—«-.3t
- E2V • Vf =a ae
h Jaa
£2J , , (2-24)ai
3f eS-r-1 + E2v7f. + E2(— + ID ,VXb)'V f, = e \ j . + E J, + E2J . . .3t - i m c i - - v i ie ia ii
(2-25)
Some extra assumptions have been made in these equations. The
time variable has been scaled with T , so that in these equa-ea ^
tions the choice v = v = v /E has been made. The heavyea aa ia
particle electron collision terms receive an additional factor
e2 because of the fact that momentum transfer in these colli-
sions is rather inefficient. From these assumptions it followsthat Q *- EQ - Q. , which is reasonable provided that
xea xaa 4ia ' r
charge transfer is not taken into account.
At the same time it is assumed that the temperatures of the
different components are of the same order of magnitude, so
that v = v ~ EV . In the chapters to follow solutions ofTi 13 Te
kinetic equations will be found by means of a perturbation
expansion:f= (E.Y.O = fo
0)(E»Y,t) + ef ^ r . v . t ) +.... (2-26)s s s
It is known that such an expansion nay often lead to secular
behaviour, i.e. it contains terras f ,. and f such that thes,n+i s,n
ratio f ,/f goes to infinity with increasing time, sos,n+l s,n
that the expansion fails. One possibility to avoid these
secularities is to make use of the multiple time scale forma-
lism1*"7. For that purpose it is observed that there are
different time scales to be distinguished: tg is called the
fastest time scale which is connected with the mean free time
between two successive collisions of an electron with an atom;tn = T . Then successive time scales are defined in the0 ea
following manner: t1 = tQ/e, t2 = tQ/e2 etc. The t2 time scale
will appear to be the timescale on which energy relaxation
-14-
between electrons and atoms takes place. In the multiple time
scale formalism new time variables T are defined as follows:n
rn := ent, (2-27)
so that the time derivative transforms as:
IF * l?0+ 'h^^h^ <2~28>
Thus the formalism consists of a transformation from one time
variable to a certain number of time variables T which aren
treated as independent. In this way extra freedom is created,
that will be used to eliminate the secularities which may
occur. This is the essence of the multiple time scales forma-
lism. The expansion (2-26) then transforms as:f (r,v,t) + f (r,v,x ,T ,..) + f (r,v,x ,T ,..) +... (2-29)
The procedure is then as follows: the collision integrals are
also expanded in powers of £ and the expansion (2-29) is
substituted into the Boltzmann equation. Terms of equal power
of e are collected and equated to zero. The resulting equations
are then solved for the functions f . The conservation equa-
tions will be treated in a similar manner, and will serve to
find solutions to the kinetic equations. Substituting the
resulting solutions into the general expressions (2-2) trans-
port coefficients are obtained, mostly as integrals over the
electron-atom cross scetions. For realistic cross sections
numerical integration schemes have to be resorted to.
-15-
References
1. S.Chapman and T.G.Cowling:"The mathematical theory of nc
uniform gases", Cambridge University Press, 197
2. J.H.Ferziger and H.G.Kaper:"Mathematical theory of
transport processes in gases",North Holland
Publishing Company, 1972.
3. L.D.Landau,Phys.Zelts.der Sowjetunion,10(1936)154.
4. G.Sandri,Ann.Phys.24/1963)332,380.
5. E.A.Frieman:J.Math.Phys._4(1963)410.
6. J.E.McCune,G.Sandri and E.A.Frleman,
in Rar.Gas Dynara.j_ (1963)102.
7. G.Sandri,in:"Nonlinear partial differential equations",
ed.W.F.Ames, 1967.
-16-
III VERY WEAKLY IONIZED GASES
In the first chapter several categories of plasmas were
distinguished on the basis of the degree of ionization. In this
chapter the case of a very weakly ionized gas is considered.
Here the degree of ionization is so low that the effect of
Coulomb collisions is relatively small or even negligible. The
latter case has been considered by van de Water1. In the
following two sections a similar type of analysis is given for
a different ordering of some parameters. Inhomogeneities are
now assumed to be of the order e, whereas the influence of the
background neutrals is reduced as compared to his work.. The
ordering is then identical to the one used by Bernstein2.
Ir this chapter only the electron component is considered. The
distribution function of the neutral atoms is assumed to be a
local Maxwellian, of which the macroscopic quantities satisfy
the Euler equations.
In the third section the form-relaxation of the zeroth order
electron distribution function in a homogeneous plasma is
described for an arbitrary electron-atom cross section. This
differs from van de Water's work, in which also an inhomo-
geneous plasma is investigated but then restricted to a Maxwell
interaction between electrons and atoms.
In the last section collisions between charged particles are
included. The ratio of electron-electron to electron-atom
collision frequency is assumed to be of the order e2. The
influence of the electron-electron collisions on the electron
distribution function is nevertheless large. The form of the
zeroth order electron distribution function Is shown to be
governed by a non-linear integro-differential equation. The
asymptotic form of this equation describes the competition
between a Davydov and a Maxwell distribution function.
-17-
III-l The electron distribution function
In the Boltzmann equation for the electron distribution
function in a very weakly ionized gas only electron-atom
collisions are to be considered. Only the term J is thusea
retained in the right-hand side of equation (2-17). The heavy
atoms possess a local Maxwellian:
fa(r,Y,t)3/2
exp{-Y-wa(r,t)
3
[> (3-D
where the macroscopic quantities obey the Euler equations:
3n
3t+ V-(nawa) = O,
3w( —3 i /
f- (n T"3/2) - 0.dt ^ a a
The Mach number is assumed to be of the order unity:
M : =wI-a IV T a
= Oil).
(3-2)
(3-3)
(3-4)
(3-5)
From equation (3-3) the instationary inertial term is estimated
by means of the pressure terra:
V2
= 0{-T^)' (3-6)9t
VP,n na a
If the electron and atom temperatures are of the same order and
a velocity transformation is applied according to:
v -»• c = v - w (r, t),- - - -a -
the electron Boltzmann equation takes the following form:
(3-7)
3f«
3t
eE 3wec«Vf + e2w «Vf - { e - = - + e 3 f - r ^ - + ( w «7)w ) + e 2 ( c«7 )w— e —a e m dt —a —a — —;
+ eto w xb}«7 f - to c«(bxV f ) = J (f ),ce-a -' c e ce- - c e ea e
a
(3-S)
-•18-
where the ordering indicated earlier appears explicitly.
The solution of this equation is sought in the form of an
expansion of f in the small parameter e. At the same time the
multiple time scale formalism is applied; cf. chapter II.
The expansion of the electron-atom collision integral can be
found in appendix A. In zeroth order the following equation is
obtained from (3-8):
3f(0)
(3-9)ce- v-
It is possible to derive an H-theorem from this equation. In
velocity space a spherical co-ordinate
system with c directed along the unit
vector b is introduced. See fig. 3-1.
Equation (3-9) then reads:
3f(0) (0)
3fwce W (3-10)
fig. 3-1.
Multiplication of this equation by
(l+ln(f )) and an integration over
the entire velocity space results in:
3H(0)
W%f<°>ea
)d3c < 0
(3-11)where the inequality is proved in appendix B. Thus it is seen
that the zeroth order electron distribution function relaxes
towards an isotropic function when T Q + », since that is the
general solution of the equation J (f) = 0.
The first order part of equation (3-8) reads:
3TI " O
eE
J (3-12)
In a formal procedure one may separate the distribution
-19-
functions in an asymptotic part on the tQ time scale and a
remaining transient part:
f(O) = f(o) (0) (0) = llm f(o)#
e e,as e,t • e.as T + „ e
Then equation (3-12) is integrated with respect to TQ:
3f(0)
eE' T 3f(0)
-cVf +!i.Vf(°> +jC0)(f(D ,}+ /°{- fe.t _ c.Vf(0)e,as m c e,as ea e,as 3T 1 - e,t
c ^ < t ?Ve!t ea^t
where E' = E + w xB. (3-15)
If it is assumed that the integral in this equation remains
finite when Tg+ °», the first part in the right-hand side would
increase without bounds with T Q except if it is demanded that:
af(0) eE,
- e' a s+ c.7f
(0) - 5 . .v f<°> = j(°>(f(1> ) + u c-(bxV f(1) ).3T - e,as m c e,as ea v e,as' ce- y- c e.as'
This equation can be solved easily if f is expanded in
Insertion of this expansion in equation (3-16) then gives with
the aid of appendix A and definition (4-61) for M, .:=(n)
I {( l - na) bx) f ( 1> ( c ) } . ^ = J M . ,• f ( 1 ) ( c ) - ^ 1 ^ =, T^ v(c) ce~ n-e,asv Jn - L =(n) n-e,asv 'n -
n«l (n) n
g/' 3 8 » 0. (3-18b)
The latter equation is the isotropic part of equation (3-16).
-20-
Frora the right-hand side of (3-18a) it appears that n=l gives
the only contribution apart from an isotropic function f satis-
fying the homogeneous equation:
Jea ) ( f ) + uce£'(*xVcf) = °* ( 3" 1 9 )
Thus the solution for f reads:e,as
eE'9f(0)
fCD ( c ) = fd) ( c ) + T ( C ) C .M-1 . ( ^ _ e ' a s - Vf(0> ).e,as - e,as (1) - =(1) vm 3c e,asv
(3-20)
In second order equation (3-8) yields:
»f(2) 8 f
( 1 > 3f ( 0 ) ,,, . E. ,,,
For reasons of simplicity this equation will be dealt with in
the limit i •*• «>. The isotropic part can easily be separated
from the rest by means of the othogonality property of the
harmonic tensors (see appendix C) :
(0) ~(1)^e.as + £e,as + c^.^l) _ ±_ . a ( 3f(l) } + (0)« 2 ox1 3 -e,as „ 2 3c1- -ejas-* -a e,as
3f(0)
- | -.^V.w = J^)(f(0> ). (3-22)3 oc -a ea e,as
This equation may be integrated over T^, if w is assumed to be
stationary on this time scale. This is in accordance with the
Chapman-Enskog theory of the heavy particle gas. Then the
following equation results from elimination of the secular
terms:
3f(1>
= 0. (3-23)
With the results in appendix A and expression (3-20) for f-e,as
equation (3-22) is written as follows:
-21-
,f(0)
if;"
m ca
3
where <A' = V -*— . This equation has been derived earliere
by Bernstein2 and 0len3. An isotropic correction Is not
mentioned by these authors. The non-isotropic part of (3-21)
when TQ-»- °° reads:
, n ... eE'«c ... eE' 3f ( 1 )
c-Vf(1) +<cc>:7f(1) - ^ ^ f ^ ) - <cc>: - ^ ^ e ' a s
- e,as — -e,as ra c e,as — m c oc
3f(0)
-<cc>:Vw ~—-e'as = ai c«(bxV f ( 2 ) ) + J ( 0 ) (f ( 2 ) ). (3-25)-a c 3c ce- - c e,as e,as e,as v '
(2)Insertion of an expansion like (3-17) for fv ' leads to the
e,asfollowing solution of equation (3-25):
3f(0)
~(2)where the isotropic part f is as yet undetermined.
e j as
The third order part of the electron Boltzmann equation (3-8)
is:
„(0) „(1) „(2) „(3)! \ ^ ' < 2 ) < l ) <0>
dw- T^a«V f(0)- (c-Vw )-V f(1)- Ü) c(bxV f
dt c e - -a c e ce- - c e
where IF = IF + v 7 -The isotropic part of this equation can again be separated from
-22-
the rest. When Tg+ • this isotropic part yields the following
equation for the first order isotropic correction:
3f(0)e,as e.as
3~(2)e.as
eg1
3m c 2
e
3T,4
3» —3c
f
-e,as-a
?
e,as
c e3 3c
-e,asJea
fe.as
Insertion of expression (3-26a) for f
then gives:
(0) ~(1) ~(2)
(2)f clS
3f
TT,e,as e.as
3T1 -a e.as
(3-28)
and using appendix A
ofc e,as_
— — —— ' V*w —3 3c -a
^ f- e,as ^ 1)
kT
m cf }.e,asJ
(3-29)
Equation (3-29) may be Integrated over T^. Elimination of
secular behaviour then leads to the following equations:
3 f
3f
(2)e,as
1
(0)e, as
(3-30a)
3ff(1)
e,as
-a e,as
3c -xj1 .<Af
(1)=(1) - e,as
3fC_ ;
3 3c
m ca
( 1 )
—a
TÜ)
kT
m ce
fe,as
(3-30b)The latter equation for f is almost equal to equation
(0) e > a s
(3-24b) for f , which is homogeneous. Equation (3-3Ob) has a
source terra containing the zeroth order.distribution function.
These equations are different from the corresponding equations
of van de Water1, due to the different ordering.
The inhomogeneity of equation (3-3Ob) obstructs the absorbtion
of the first order isotropic correction into the zeroth order
distribution function, which was an assumption made by
Bernstein2. The equation for f is of second order in the
variable c. In the following section two conditions will be
given which determine the two constants of integration.
-23-
IIT.-2 The electron macroscopic equations
The macroscopic equations for the electrons can be
obtained from equation (3-8) through multiplication by the
appropriate functions of velocity and subsequent integration
over the entire velocity space. The following equations are
then obtained:
3n— e 0, (3-31)
3u dwm n f-r— + e(u 'V)u + e2(w «V)u } +e2m n (u »V)w + e3m n -r—e e'3t -e -e -a -eJ e e -e -a e edt
eV«P + een E + m n Ü) (u + EW )xb = Jm cJ (f ,f )d3c, (3-32)=e e- e e ce -e -a - e- ea e a
dln f-r-e + eu »V£ + e2w «Vê 1 + eV«(q + P «u ) + een u «E +p^-at- — o p - a e ' -»e = e - e P - P -
dw-a
+ era n u *(w xb) + e3m n u —r— + e2fp + m n u u ) : 7 w +e e-e
= A m c2J (f ,f )d3c.e ea e a
(3-33)
Note the transformation that has been made according to (3-7).
Therefore t is now defined slightly different from (2-4) as:
£ e = -| kTe + ^meu|. (3-34)
The macroscopic quantities are also expanded in powers of e and
the multiple time scale formalism (MTS) is applied. From the
above balance equations the following equations are obtained in
zeroth order of e:
(0) aï(0)3n
(0)3u
In first order one obtains:
3n on /rt\ /n\^ e f ><0>_u
m c
0,
(3-35)
(3-36)
(3-37)
-24-
„ ( 0 ) { ^ +!£e +( U<°).7)4O )
+ Io c e41>xb} + 7.P<
0)+ enf >E'
„ . 3u ( 0 )
'(1)
_ƒ_£!_ f(Dd3 (3_38)JT (c) e
+ en^0)E'«u^0) = 0. (3-39)
When T Q + «> the zeroth order electron distribution function
relaxes towards an isotropic function of velocity as was shown
in the previous section. This means that in this limit the
diffusion velocity u and the heat flux £ vanish.
Equations (3-37) and (3-39) take the following form when T.-»- =>:
9n(°> 3T<°>_e,as = _e,as = Q> (3_40)
en(0) (E'+ u(1> xB) + VP(0) + f^L-,f ( 1 ) d3c, (3-41)e,as - -e,as - r e a s 'iAc) e a s
where P <0 ) = n(°> kT
e,as e,as e,as
Thus it is seen that many terms in these equations vanish when
TQ+ a». The expression for f found in (3-20) may be
substituted into equation (3-41) which then yields an identity.
The second order equations are given in the limit T Q+ « in
order to reduce the complexity of the equations:3n 3ne a s _ e , a s +
a (1)
(0) (1)
-e,as e.as^1 vae,as 6e,as -e
d T ( ) ( )3 (0) (^e,as ^e.as (1) (0) j+ (1) (0) (1)z^eas kar 3T -eas eas^1 vaeas 6eas - e ^
-25-
+ en<°> E - ^ + P( 0 ) :Vw - - ̂ J l + ~*" f)f(0)e,as- -e,as =e,as -a m -'T, x(c) ̂ ra c 3c •* e,as
a (L) e
(3-44)
From equations (3-18a), (3-20) and (3-24a) it is inferred that:
3u ( 1 ) 3n(1> 3T ( 1 )
_^e,as = _e,as _ _e,as = Q< ( 3_ 4 5 )
It will be assumed now that the following first order quanti-
ties are zero:
n ( 1 ) = T ( 1 ) = 0 , (3-46)e,as e,as
which are the additional conditions needed for a unique solu-
tion of equation (3-29). Such conditions can in fact be chosen
without loss of generality on the basis of the arbitrariness of
the expansions of the initial conditions in powers of e. Since
moreover f is isotropic the second order equations nowe,as r n
°. (3-47)
= 0, (3-48)
reduce to:
8ne,as (0
(0) (2)en u xB +
e,as-e,as -
3 (0) rdTe,asIne,asKidT2
e,as-e,as -
) ( u ( i , +
m c
T ( 1 ) ( c ) f
• + P( o ) ve,as
w )
e,as
(0)e,as
- a
.( (1) + (0) (1)e,as-e,as'
(3-49)
As all quantities occurring here are functionals of f , seee g as
equation (3-20), (3-26) and (3-29), it appears that these equa-
tions do not contain any variations with T,, SO that the t.time scale has no physical meaning in this particular situa-
(2)tion. Insertion of expression (3-26) for fv ' into (3-48)
G y 3 S
leads to:
/c2<A'f(1') d3c = 0. (3-50)
This equation can be further evaluated to give:
-26-
Vp ( 1 ) + en ( 1 ) E' = O, (3-51)*e,as e,as- '
which ts satisfied through the requirements (3-46).
With the aid of equation (3-47) the energy equation can be
written in the following form:
- p(0) D ln{n(0) (0) 3 / 2 } + (1) (0) E, + (0) +
e,as D T 2 L e,as e,as ' -e,as e,as- re,as
k T . ,_.(1+ -Ji.|-)f<0> d 3 c . (3-52)v tn c 3c' e,as
where: £_ = f- + u(1) -V = f- + (w + u(1) ) -V.DT dT -e,as 9t2 -a -e,as
At this point it is suitable to introduce transport
coefficients. The first order electron diffusion velocity can
be calculated with the aid of expression (3-20):
n(0)u(D =-Iö(1>.E'+V.(n<°>D
(1)), (3-53)
e,as-e,as e = k e,as= ;'
3f(0)
where g ( 1 ) = - 3 ^ ; CT ( 1 )(c)^/ )-^e' a Sd3 c, (3-54)
S(1)-^öT-^2\i)fc>B?i)O3c- (3~55)
e,as
are the conductivity and diffusion tensors respectively.
If the solution of equation (3-24) for the zeroth order
electron distribution function is known, the transport
coefficients can be calculated. In a simple theory the
following approximation is often made:
f(0) = n(°> f (c), (3-56)e,as e,as 0
where f (c) depends on c only, so that the space and time
dependencies occur through n solely. With this assumption aG I 3S
diffusion equation may be obtained from equation (3-47):
^ 8 0 > s , . 0 i ( 3_ 5 7 )
where the neutral component has been assumed to be homogeneous
-27-
in space. The assumption in (3-56) also implies a uniform
electron temperature. Refinements can be obtained by making an
expansion of f in the spatial derivatives of n : see1 e,as e,as
e.g. reference 4.
These equations are used for the determination of electron-atom
cross sections from diffusion experiments5.
The thermal heat flux is also calculated with the aid of
expression (3-20):
3( 1> = - S
(1).E'- X(1).Vln(T(0) ) + V(n D ( 1 ) ) , (3-58)ae,as aq - = v
e,as' v e,as=q " v '
W i t h :
we,as e,as
5kT(0)
(1) = _ ^ £=
^ J T c V l f C ) d 3 c . (3-61)6 J (1) =(1) e,as
(2)It appears from expression (3-26a) for f that corrections
to the transport coefficients are given by the same expressions
if f<0) is replaced by f(1) .e,as e,as
From equation (3-50) one may infer then that in the special
case of Maxwell interaction between electrons and atoms the(2)
second order diffusion velocity u vanishes. The second™G j aS
order thermal heat flux reduces in this special case to:
3(2) = M-! 7J(2) j(2) . _ V 0 i ; 4~(l) d3c>ae,as =(1) q ' q 6 ' eas v '
The first order fluxes reduce to the following expressions in
case of Maxwell interaction between electrons and atoms:
kT C 0 )
-28-
showing that there are no cross effects In this case.
In third order of e the moment equations, when considered
asymptotically on the to-time scale, read:
3n(0) a (2)
e,as + _^,as + 7.(n<°>g_u<2>s) - 0, (3-64)
_
e
, (1) (2)(0) { 3 , a s + ^ e , a s ( 1 ) + (1) (1)e,as l3T2 3TJ v üe,as -aJ -e .as K-e,as }-a
+ V«P + en E + m n u (u + w )xb ==e,as e,as- e e,as ce -e,as -a -
m m c kT , , . . .= _£ /{.-ÊZ _ «CA ( _L_) _ k T 7 ( J L _ ) }f <D d 3 c ( 3 _ 6 5 )
ma T ( l ) 2 ~ C T ( D l c t ( l ) 6 l "
3£(0) 3Ï ( 2 )
(0) ,^e ,as + ffe.as (2) ^(0) } (2) (0) (2)ne,as^3T 3 ^ ye,as y°e,asJ ^3e,as Êe,as - e , a s ;
+ en ( 0 ) u ( 2 ) .K' + . n ( 0 ) u u ( 2 ) . ( . xb) =e,as-e,as - e e,as ce-e,as x-a -
m 3kT me 2
kT c i ^ -L- ) )^ 1 ) d3c. (3-66)a ^ T ( D
e'as
Again an Ansatz Is made, namely:
which can be justified In the same manner as in (3-46).
Equations (3-64) and (3-66) may then be written as follows:
3n(0)
eas -eas a e a s m 'x,.. m c
From (3-23) and (3-26a) It can be deduced that:
-29-
jf'as = 0. (3-70)
Equation (3-65) may therefore be written as follows:
,n. Du^ Dw ...(0) r -e,as + -a, (2) =
e e.as^Di^ Dt J =e,as
= j-!*! {1- flT(1) i-fc1* |-(J—))}f(1) d3c, (3-71)3 V ( 1 ) 2mec^
3c 9C T(l) "e'as
The survey of the moment equations has now been carried out up
to third order. The equations of this chapter are useful in the
process of solving the kinetic equations.
In the following section the equation for the zeroth order
electron distribution function will be solved in a special
case.
-30-
III-3 Form relaxation of the electron distribution function.
In this section the equation for the zeroth order
electron distribution function is examined for the case of a
homogeneous plasma without a magnetic field.
Equation (3-24) then takes the following form:
af(0) *
_ e> a s = A i_(-£Lfi + __s. i_)f(o) , (3_72)3T, jacW,, l^ucac^e.as'1 ( 3 7 2 )
2 m cz (1) e2* m
ale E T(l) ( c ))a ( l ) )
in which: Ta = Ta + < , (3-73)e
is a function of c. The relevant macroscopic aquations read:
"37e 'a S = 0, (3-74)
(0) p *3 (0) , 3 T e , a s rae,mec , _ k T a 8 , . ( 0 ) , 3-^e.as^, = -—/r~(1+i-c-^c-)fe,asd c- <
2 a (1) e
Equation (3-72) may be solved by means of the method of
separation of variables. Insertion of
f ( 0 ) = n ( 0 ) f(c)h(T2) <e,as e,as 2
into equation (3-72) results in the following eigenvalue
problem for the function f:
2 h 7 ^ ^ £ °' (3"77)m c' (1) eaand a simple equation for the function h:
~ + Xh = 0. (3-78)a T2
If X=0, equation (3-77) can be directly integrated. The
solution y0, the eigenfunction for \=0, then reads:
c m c 'dc 'y0 - A exp{- ƒ — ; }. (3-79)
o kTa(c ')
-31 -
This is the asymptotic solution of (3-72) when T 2+ °°, and is
known as the Davydov distribution function^. It will be demon-
strated now that all other eigenvalues are positive.
Define:
f(c) = yo(c)<j>(c). (3-80)
Substitution into (3-77) and subsequent multiplication by <j> and
integration then leads to:
(3-81)
ƒ yo(c)<)>2(c)c2dc
c2kTwhere yo(c) and p(c) = ^—r- (3-82)
aT(ir
are positive functions, so that all eigenvalues except X=0 are
positive indeed. Expression (3-81) also gives a device for the
calculation of the eigenvalues and eigenfunctions by means of a
variational principle. From equation (3-77) one can deduce that
all eigenvalues are orthogonal with weighting function c2yQ:
Jyoc2(j) (j> dc = 0, n+ra. (3-83)
o
The variational principle then reads as follows:
= min R(<f>) = R(<t>n); |yo(c)<|>n(c)<j>m(c)c2dc = 0, m=0,1,... ,n-l.
(3-84)
where: R(<|)) = — . (3-84a)
Jyo(c)<t)2(c)c2dc
o
A Rayleigh-Ritz method may be used to approximate the first N
eigenvalues and eigenfunctions. In the special case of Maxwell
interaction between electrons and atoms the eigenvalue equation
can be solved directly. Then the collision time T. . is a* (l)
constant, so T does not depend on c either. The eigenvalueSi
equation after a transformation of variables reads:
-32-
m w2
where w = j- . Equation (3-85) is the differential equation2kT
a
of Laguerre. The eigenvalues and eigenfunctions are thus equal
to:
y = 4 n )(w), A = | 2 - , n=0,l,2 (3-86)n * n T(l)
The Davydov distribution function is now a Maxwellian with*
temperature equal to T .3
In the case of a hard spheres interaction model one has:
T ( 1 ) ( C ) - | , (3-87)
where I is a constant mean free path. A straightforward calcu-
lation shows that the Davydov distribution is now equal to:
2 . m m (AeE)2
y0 = C exp(-ac2)(l + j~) , a = 5--=- , A = -2 (3-88)
a 3m2kTe a
where the constant C is fixed by:
00
/yo(c)c2dc = 1. (3-89)
0
In the cold gas limit T •»• 0, the Druyvesteyn distribution7 is
recovered:3m3
y0 = C exp(-Yc't), y = . (3-90)
4m (AeE)2
a
If the eigenvalues and eigenfunctions are known, the initial
value problem may be solved, i.e. equation (3-72) supplemented
by the condition:
The formal solution reads:
-33-
with: a => Jfo(c)<J> (c)c2dc, (3-93)
o
if the eigenfunctions are orthonormal:
J*2(c)yo(c)c2dc = 1, n=0,l,2,... (3-94)
o
and form a complete set.
The same problem has been investigated by Braglia et al8, who
calculated the temporal behaviour of the distribution function
for various cross sections.
-34-
III-4 The inclusion of Coulomb collisions.
In the foregoing sections the Coulomb collisions have
been neglected entirely. If, however, the electron density is
such that the ratio of the electron-electron to electron-atom
collision frequency is of the order ra /m , i.e.:e a
n 0 'a ea
the e-e and e-i collision terras appear in the second order
equation of section 1. When T Q + <=°, only the isotropic part
changes, and the equation for the zeroth order electron distri-
bution function now becomes a nonlinear integro-differential
equation:
»jlO,l+ (0) c*jlO,ls 1_ , 3 . 1 , (0) ,3T -a e,as 3 3c -a 3c - (1)=(1) - e.as"1
2
, me 3 r c3 ., a 3 .-(0) •> ,r(0) ,(0)
m c 3c lx n. m c 3c e,as ee e,as e,asa (l) e
In third order of e the results of section 1 change as follows.
To the nonisotropic part of the electron distribution function
terms proportional to c are added coming from the Coulomb
collisions and the equation for the isotropic correction in
first order becomes of the same type as equation (3-96). In
order to study the nature of equation (3-96) this equation will
be considered in the special case of a homogeneous plasma
without a magnetic field. With appendix D-l one obtains:
or £.\J « /rw / ̂ \ "̂-*- oi.ni r J
e,as ee 3 t A U ) (c)\n ' fii e,as e,as •>3TT ~ -> 3c^ e,as^ ' L e,as*- m c ~ 5c '
2 m c z e
J ( v c ) v f = (v) (
m ca
-35-
The asymptotic solution of this equation may be considered as
the result of the competition between a Maxwell and a Davydov
distribution function. Omitting the time derivative and
integrating once one obtains the following equation for the
asymptotic solution f :A
2C « 31n(f (v)) 31n(f (c))\-s—j(v3~có)vf. (v) T r [dv +
m l3 J A lv 3v c 8c 'e c
kT 31n(f ) m c3 kT 31n(f )n . f l + — - — r — — ) ] + — fl+ — T — — ) = 0 , (3 -39 )
A*- m e 9c Ji m T , , . 1 1 m c 3c ' ' v '
e a (1 ) ewhere the constant of integration has vanished by consideration
of the limit c
following form:
of the limit c + °°. The equation for f can be written in theA
IT W
*
T.JL-ËI) + v (i+ ÈL) = o, (3-99)T. dw ee dw
Am w 3 / 2 m e 2 2C n
where B(w) = 2/2 — - r-rr , w = s p _ v = e e A
m aT ( 1 )(w) 2kTA ee ^kT m 3/2
V T = > fA = nA ( y T 7 ) exp(y(w)). (3-99a)
i e e A A ZTTKIA
The following normalizations should then be imposed on a
solution of equation (3-99):
/exp(y)w%dw = — , /exp(y)w3/2dw = ̂ ~- , (3-100)o o
in order to determine the integration constant and the
temperature T . If w » l , then the solution of (3-99) may beA
approximated by the solution of the following equation:
A
The solution of this first order differential equation is:
-36-
w B(w') + vy(w) = - ƒ{ 5E ££ }dw' + C, (3-102)
o B(w')T /T. + va A ee
where the Integration constant C and the temperature T^ areA
fixed by conditions (3-100).
The problem has been Investigated earlier by Lo Surdo9, who
obtained solutions for simple electron-atom cross sections by
means of an iterative numerical procedure. It seems that,
because equation (3-99) is of a simpler form than his equation,
the results of this section might lead to simpler numerical
techniques to obtain a solution.
-37-
References
1. W. van de Water, Physlca 850(1977)377.
2. I.B.Bernstein, in: Advances in plasma physics vol.3 (1969)
3. A.0ien, J.Plasma physics, 26^(1981)517.
4. L.G.H.Huxley and R.W.Crompton, "The diffusion and drift of
electrons in gases", J.Wiley (1974).
5. H.B.Milloy et al, Austr.J.Phys. 30(1977)61.
6. B.Davydov, Phys.Zeits.der Sowjetunion £(1935)59.
7. M.J.Druyvesteyn, Physica 10(1930)61,1(1934)1003.
8. Braglia et al, II nuovo cimento 62B(1981)139.
9. C.Lo Surdo, II nuovo cimento 52B(1967)429.
-38-
IV WEAKLY IONIZED GASES
In chapter II a weakly ionized gas (WIG) was defined as a
plasma in which the ratio of electron-electron to electron-atom
collision frequencies is of the order e (cf. equation (2-23)).
This means that the degree of ionizatton is very low. Since the
Coulomb collisions become more important at lower temperatures
the degree of ionization should be assumed to decrease with
temperature in order to satisfy the ordering mentioned above.
In this chapter the procedure is as follows. Firstly the heavy
particles are considered, because they can be treated as almost
independent from the electrons, i.e. as a binary mixture.
Because the degree of ionization is low the usual Chapman-
Enskog equations are only slightly modified. Then the electron
Boltzmann equation which gives more interesting results will be
dealt with. The isotropic correction to the zeroth order
Maxwellian electron distribution function is not adequately
dealt with in other theories, with the exception of van de
Water's paper^. It also appears in references 3 and 4, but does
not receive the attention it deserves. The expansion of the
electron distribution function in powers of e leads to some
results which are not found with the usual harmonic tensor
expansion-*. .
The isotropic correction results from the competition between
the mutual electron collisions which try to establish a
Maxwellian and the disturbing effect of electric fields,
temperature differences between electrons and heavy particles
and temperature- and pressure gradients.
The domain of the degree of ionization in a WIG can be roughly
devided into two regions. At lower degrees of ionization the
isotropic correction is important whereas the corrections due
to multiple collisions dominate at higher degree of
ionization. Expressions for the electron transport coefficients
will be derived and finally the modifications in case of a
seeded plasma are given.
-39-
IV-1 Heavy particle results
The heavy particle Boltzniann equations valid in a WIG
were already given in chapter II, equations (2-24) and (2-25).
The distribution functions are expanded in powers of e and the
multiple time scales formalism (MTS) is applied. Up to second
order the results are:
( 0 ) 3f ( 1 )
l- + ! l a =J ( f V l + J (f (1 ) ,f (0 )), (4-2)3TX 3 T 0 aa a ' a aa a a '
f(0) 3 f(D af(2)
^ = 0, (4-4)
af(0) 3 f(D
< <l) «i 2 ) (0) «5 (0)I72 +Tt\ + T F ' +Y.Vf[O) + ( +̂<,clvxb).7vf5°> =
(0) (1) (1) (0) (0) (0)Jia(fi 'fa } + Jia(fi 'fa ) + Jii(fi 'fi )# < 4" 6 )
By means of an H-theorem obtainable from equation (4-1) it
follows that f ' relaxes to a Maxwellian when t n+ "•• This
a u
limit will be indicated by a subscript "as" so that:
(0) . n(0). n((0)
_
a,as a,as
In order to proceed the moment equations are needed. The
balance equations for the neutral particles read:
-40-
3n-^ + e2V-(nawa) = 0, (4-8)
3wm n (^a + e2(w -v)wJ + e2V-P = e2jm vJ (f f )d3v, (4-9)3 3 OC ~3 3 ~3 3 3x 3 1
(4-10)
in which the interaction terms between the heavy particles and
the electrons are omitted because these are of the order e1*.
The macroscopic variables are also expanded in powers of e and
the MTS formalism is exploited. Up to second order the results
from these equations are:
3n(°> 3g(0> 3w (°>
< < <
3n ( 1 ) 3n ( 2 )
< TF; i%i°> 0. (4-13)
3w(0) 3w<^ 3w<2> „.
; a ; 4 0 ) ^
From equations (4-7) and (4-12) and the definition (2-4) of
chapter II it is concluded that -r^ f£ - 0. Then equation (4-2)
becomes in the limit T.+ °°, indicated by a subscript "as":
J (f<0) ,f<1> ) + J (f ( 1 ) ,f(°> ) - 0. (4-16)aa a,as a,as aa a,as a,as
This equation possesses the following general solution2:
-41-
a.as " vul T *2mï T U 3 V ^ a , a s ' V H i / J
where o.(r,TltT2,...) are at this point arbitrary functions of
position and time. The Chapman-Enskog choice:
n(D = W ( D = T ( D = Qa,as -a,as a,as
makes these functions zero, so that the first order correction
to f vanishes:a,as
f ( 1 ) = 0. (4-19)a,as '
Next equation (4-5) will be considered in the limit T Q+ °°:
T F = Jia(fi as'a as'1 ( 4" 2 0 )
This equation also possesses an Il-theorem implying that £.
relaxes to a Maxwell distribution function, when T Q+ ", with a
temperature and a hydrodynamic velocity equal to the neutral
ones:
- „(0) "
T ^
aA
A subscript "A" denotes the limit Tj+ «. The ion balance
equations read:3nj^ + E2V.(n1?i) = 0, (4-22)
3w.
e/m±yJia(f1,fa)d3v, (4-23)
(4-24)
After expansion in powers of e and using the MTS formalism the
results up to second order of e are:
-42-
3n<°>
a (0) (1)
(4-27)
(0) (1) (2)3n. 3n 3n+ ' + J
i l ci-i
(4-30)
When T.* 00 f^0) is a Maxwellian with w[°^= w( ^ and T ^ = T^K1 1 -IA -aA iA aA
Furthermore the first order corrections n.. and T., areiA iA
assumed to have vanished. Equations (4-29) - (4-31) then read:JO)^ * ,.c.2>.S') • 0. (
-43-
where p£^= n^'^aA^ and -'= - + -aA>x-* (*"35)
if n /m = 0(1).With the definition of the total derivative:
fc/fc/sS»-. '«')equations (4-32) and (4-34) can be written as follows:
< + n(0)^(0) = 0>
dw(0) '(O)-aA (0)_ (0) _ " v '"̂
miniA "dT2 piA eniA -
(4-39)
And for the neutrals the Euler equations are obtained:
.(0)
. (C)
C°>
When equations (4-39) and (4-42) are compared with each other
it appears that there is no net energy exchange on the t2-time
scale between ions and neutrals in first order:
which is compatible with the choice T^, = 0.1A
-44-
Now equations (4-3) and (4-6) can be treated. When TQ* "° these
equations read as follows:
3£(0)
aA
which are the Chapman-Enskog equations for f . and f.. . The
left-hand sides of these equations can be brought into a more
familiar form through a transformation in velocity space:(0)
v •>• v - w . crom the laboratory frame to a frame moving with— — aA
the zeroth order hydrodynamic velocity of the neutrals.
With the aid of the macroscopic equations (4-37) - (4-42) the
source terras of equations (4-44) and (4-45) become:
(4-46)
aA aA
"i -nu ,rvv eE'
aA aA aA
(4-47)
where c = v - w . , which is the peculiar velocity defined in
chapter III. The equations (4-44) and (4-45) are consistent
with the traditional Chapman-Enskog procedure, see e.g.
Chmieleski and Ferziger3.
If one considers the heavy particle results of reference 3 in
the case n « n the equations (4-44) and (4-45) are recovered
with source terras (4-46) and (4-47) respectively. The solution
of these equations is standard2. If resonant charge transfer
instead of elastic scattering is the main mechanism for the
-45-
ion-neutral interaction the zeroth order ion distribution
function will in general not be a Maxwellian. When a constant
cross section for the charge exchange process is assumed the
deviations from a Maxwellian are not very large, even in the
absence of ion-ion collisions5.
-46-
IV-2 The electron Boltzmann equation
The electron Boltzmann equation for a WIG has already
been given in chapter II: equation (2-17). Contrary to the
heavy particle equations a transformation in velocity space
from the variable v to the variable c = v - w will be started- - - -a
with. Equation (2-17) then reads:
3f-^e + e(c + ewa)-7fe- oi^c-CbxV^) +
eE 3w- { e — + eu w xb + er^a + e2f(c + ew )«v)w 1-V f =1 m ce-a - 3t kV- -a J~a' c e
e
where b is a unit vector in the direction of B and w = — is
ce m
the electron cyclotron frequency. The hydrodynaraic velocity of
the neutrals has been taken of the order of the thermal veloci-
ty, i.e. the Mach number is of the order unity. The electron-
heavy particle collision integrals are expanded in powers of e,
the velocity variables are assumed of thermal order. The
results are presented in appendix A. In the expansion of J
the first order term vanishes because of the transformation in
velocity space mentioned above. After substitution of the
expansion for e and exploiting the MTS formalism the results
from equation (4-48) up to second order are:
e , e . _ „,(0) r -
D c«(bx7 f̂ ;̂ce- - c e
-47-
(0) (1) (2)
"• + ii- +ii- +c.w<i>+»«
o>.vf<o>-. (w(i)xb).vf(°
3 T 2 3TJ dTQ - e -a e cev-a - c e
a (0) . (1)_ ( f ! + u w(0)xb).? f(D_ & +!£a +c.Vw(0)}.vf(0)+
''ia c e - a -' c e <*Ti ^ T o - - a - ' c e
3 ( 0 )
1
(4-51)
From equation (4-49) an H-theoretn can be derived, see also
chapter I I I , :
,„«»e < 0, (4-52)
° H< 0 ) = ft ̂0)ln(f ̂0))d3c, (4-53)
so that again the zeroth order distribution function relaxes to
an isotropic function when T--> °°. In that limit one obtains
from (4.50):
3f(0> , „ eE' ( 0 )
e a s ( 0 ) ^- e» a s + c.(7f(0) - - ^ T -e ' a s l - u (bx7T, - ̂ e,as m c 3c ; ce^- fc
J<°>(f(1) ) + J (f<0) ,f ( 0 ) ), (4-54)ea e,as ee e,as e,as ' v '
where J . (f ) vanished because of the isotropy of f .ex e,as e,as
Isotropic and nonisotropic parts of equation (4-54) can be
readily separated, see appendix C, so that the following
equations are obtained:
3f(0)
ILe.as = (0) (0)3TJ Jeeue,as'£e,as;'
U c ( b x V f ( 1 ) ) + J ( 0 ) ( f ( 1 ) ) - c ^ f ( 0 ) (4-56)ce- - c e,as ea e,as e,as v '
eE' 3where: cA' = V -r- , as in chapter III. Equation (4-55)
-48-
also permits an H-theorem which states that f relaxes to ae,as
local Maxwell distribution function when T,+ °°:
(0), me'eA (
A
m c'e
2kT(0)•}•
eA eAA solution of equation (4-56) can easily be obtained if fis developed into harmonic tensors (see appendix C):
(4-57)
(0)e.as
f (c) f(1) (c)«c + f(1) (c):<cc>e,as -e,as - =e,as —
f(1) (C)KCS ,n~e'aS n "
(4-58)
where f is a tensor of rank n and • denotes an n-fold dotn-e,as nproduct.Insertion of this expansion into equation (4-56) gives:
n=l l(l)'- no> bxl f(1) (c)^.<c^ = lM
ce- ;n-e,as Jn - => b fce- ;n-e,as
n=i/1N. f
(1) (c)-<cn>(1) n-e,as n -
where : 2irn c/o(c,x) (l-P (cosx) )sinxdx,a Q n
(4-59)
(4-60)
(see appendix A).
If b is directed along the z-axis M, . in index notation reads:
M — o — no) T (c) £ b » (4—61a}(n)ij ij ce (n)v ikj k'
M,nü)ceT(n)(c)eikjbk
(4-61b)
Only the first two terms in the expansion of f are non-zero
so that the solution of (4-59) is:
e.as
where f
i 44(1) 2? (4-62)
6j asis a yet undetermined isotropic contribution. It is
in fact the homogeneous part of equation (4-56). In much the
same way equation (4-51) will be treated. When TQ+ °° the
isotropic part of this equation reads:
-49-
df<°>df 3f 2 , „ 3f ^ e E a , „e,as e,as c^ (1) _ c e.as (0)_ - 3 ( 3 (1)
di 2 9Tj 3 -e,as 3 3c -a 3ra c^ 3c*- -e,as-
= -!S- [JEL ( l +fa_ 3 }f(0) + (0) ?(D +o LT,1X
V- m c 9c ' e , a s J ee e ,as e ,asm c^ (1) e ' ' 'a
+ J ( f ( 1 ) ,f ( 0 ) ) . (4-63)ee e ,as e ,as
This is the Chapman-Enskog-like equation determining the first
order isotropic correction. The non-isotropic parts give the(2)
following solution for f in the same way as in the case of6 y 3S
the first order part:
, _ ( 0 ) (0 ) ( 0 )/ T \ i / n \ i C . n . u , of
f(2) - V l ,M: ; , . { -^ f ( 1 ) - e l 1>a s- i>a s —e>a
-e.as (1)=(1) l - e,as ^ 3c
2C ^ Q )
- e i i'aB f ( 1 ) + .J (f ( 1 ) )}, (4-64a)^ -e.as 1 ee -e.as >'
m e 3 '
3f ( 0 )
f<2> - x„,M:ï,.l-^f(1) -lJ-e'aS7w<°> }, (4-64b)
=e,as (2)=(2) l - -e,as c 3c -a,asJ
where the constant C is defined in appendix A.ei
Again there appears a yet undetermined isotropic contribution.
In equation (4-64a) the following linearized electron-electron
collision term was introduced:
J ( f ( 1 ) ) : = J ( f ( 0 ) , f ( 1 ) ) + J ( f ( 1 ) , f ( 0 ) ) =ee v e , a s ee e , a s e , a s ee e , a s e , a s
This expansion is justified because the collision operator is
rotationally invariant in velocity space.
In expression (4-64a) for the correction f the contribution-e,as
of the first order isotropic correction appears. The last two
terms between braces express the influence of the Coulomb
-50-
collisions on the electron-atom interaction and are referred to
as the effect of multiple collisions. The first term of these
may lead to divergent expressions because of the factor c~3. It
becomes even worse in higher order terms. In appendix E It is
shown that one can replace m c3 by [m c3 + 2C .n^ T/i\(c)] *n
6 6 cl 1)3S ^ i j
the denomerator of that specific terra. This is actually an
improvement because it results from renormalization of that
term.
The foregoing procedure can be continued up to arbitrary order,
but it will not be done here. The higher order equations can in
principle be solved, but the increasing complexity impedes the
actual calculations to be done. When T.+ °°, it has been deraon-(0)
strated that f . is a Maxwelllan and a solution of (4-63) can
be constructed. Before doing so the electron balance equations
will be dealt with first.
- 5 1 -
IV—3 The macroscopic electron equations
The moment equations for the electron component of a WIG
can be obtained from equation (4-48) by the normal procedures.
With the definitions of the diffusion velocities u (see-s
chapter II) and the definition c = v - w one can see that:— —a
— Set (r,v,t)d3c = w - w = u . (4-66)
e
The electron balance equations for a WIG provided with the
appropriate powers of e then read as follows:3n-^e + eV-(neue) + e
27-(ngwa) = 0, (4-67)
3um n f-̂ -e+ e(u «V)u + e2(w «7)u } + eV«P + een E +e el3t -e -e -a -e' =e e-
3w+ emenehiïa + e ( - e ' V ) - a + e 2 ^ a " V ) -a^ + meneuce (-e+ e - a ) x -
= /n?e^^Jea(fe'fa) + e J ei ( f e ' f i} ^ 3 c ' ( 4 " 6 8 )
3 5 en (-T- + eu •?£ + e2w «V£ ) + eV«(g + P «u ) + em n u *(w xb)
ev.gt -e e -a e' -*e =e -e e e-e - a -
3w+ een u »E + em n u •f-rr- + e2w «Vw 1 + e2fp + m n u u 1: Vw =
e-e - e e-e '•St - a -a.' l=e e e-e~eJ -a
= Amec2(jea(£e,fa) + e^^f^fpjdSc. (4-68)
Where now, slightly different from equation (2-4):
le = ̂ e + ̂ meUl ' (4'7O)
All macroscopic quantities are expanded in powers of e and
again the MTS formalism is used. In zeroth order the results
from equations (4-67)-(4-69) are:
3n(°> 8*<°>
(4-72)
-52-
And in first order of e:
_ « + ._« + ,.<„<%<<»> . 0, (4-73,
. (0) , (1) , < = )
& 1 ^E * . „ » (u „ > b • . „< l )
U u(
e - e e ce -e -a - e e ce-e
s L f(D d3 c _ 2C .„(O)jJ.el 3c3
e„(0)E.u(0) + u . „(0)»<0).(»(0),b) - 0. (4-75)e - -e ce e e -e -a -
Here and in the sequel the results are used that were obtained
in preceding sections, e.g. —- w' '=0. When T + <*>, equations
(A-73)-(4-75) can be further simplified, because then f ^ isCO) (0) e'aS
isotropic, which implies: u = q = 0 etc. In this limit-e,as -*e,as
the first order equations become:
8n<°> 3T(0)
TF^ S = < * a S " 0, (W6).(1)
7p̂ 0^ + en(0) E + « n
W M (u(1> + w
(0) ) xb + h^H^c =0.re,as e,as- e e,as cev-e,as -a,as - JT n.(c)
(1) ( W 7 )
Substituting the expression for f as given in equation6|aS
(4-62) obviously renders equation (4-77) into an identity.
Further observation shows that equation (4-77) 'closes' when
T. .(c) does not depend on c, i.e. the case of Maxwell Inter-
action. The electron-atom interaction potential Is then assumed
to vary as r"1*.
Equation (4-77) in case of Maxwell interaction reads:
s^l)~e,as *
(4-78)
o> ( ) b
e e,as ce-e,as -e,as -
- 5 3 -
which is the generalized law of Ohm in first order.
In second order the balance equations will be considered
asymptotically when TQ* ", then they read as follows:
8n(0) 3 n(De.as e.as { (0) ( (1) + (0) ̂
3 T 2 3T 1 *• e,asv-e,as -a,as ' '
^ T ( 0 ) 3T3 , r e , a s L e,as ̂ (1) ̂ ,(0) , ̂ (0)3 , r e , a s L e,T ne,ask{HT2
+ TT,
+ V . ( 3( 1 )
+ P ( 0 ) . u ( 1 ) ) + P ( 0 > :Vw ( 0 ) =-*e,as =e,as -e,as =e,as -a,as
m tn c2 kT^0) . . . .= _ _£ƒ_«_ (i+ _ £ i S £ 3 ) f ( ° ) d 3 c . (4-81)
m JTfl.v m e 8c' e ,as
The derivative T — was defined in equation (4-36) of section 1.
Now the following Ansaz is made:
(1) = T ( D = 0 (4-82)
e,as e,as ' v '
which will be verified in the next section.
The equations (4-79)-(4-81) then assume the following form when
Tj+ » (subscript A):
(0)
lï3 C ^ u ' " . <4-84>
.w
-aA
-54 -
Next the local entropy density in zeroth order is introduced:
Then it is possible to rewrite equation (4-85) as an entropy
balance equation:
3s , q . /, x ,„N ,-... i . *X + g „ »XeA (jeA (1) (0) (0) } = -eA -m ^eA -q
3T, „(0) -eA -aA ' eA ' m(0)TeA xeA
- ̂ V 1 m c 2^1 f e f(0) ,3 .,
Ö i TT feA d C' (4"
ÏTÖJ OÖ) TTeA eA
where the thermodynamic forces:
kT (0 )
K » - - | - t 5 f +-T i Vln(p2 ) )} (*-88«)e
X := -Vln(T(^) (4-88b)
have been introduced. The first terra in the right-hand side of
equation (4-87) gives the entropy production which is positive
definite. This may be proved by means of Schwartz' inequality
with the aid of expression (4-62) for f . . It also gives the
relations between the fluxes i . = m n . u . , q . and the-eA e eA -eA -*eA
forces as defined in (4-88a,b). These relations, which also
obey the Onsager reciprocity relations, read as follows:i(l) = . n(0)u(l) . m n(0) ( 2 0(l). x. + D ( D . X j-eA e eA -eA e eA *• Te = -m =T -q;
s ü ) - ' r l ^ 1 ) ^ + i(1)-V (4"89b)
in which the transport coefficients are tensors because of the
magnetic field. The subscripts "T" and "D" stand for thermal
diffusion and Dufour effect respectively. The expressions for
these coefficients are:
D ( 1 ) = -ifljy /T(1)(e)c2f2)H"1
1 jd'c. (4-90.)
- -f-JfeA =CO d ' (4~99b)"eA """eA
- 5 5 -
/1 \ kT * t n c c 2 / rw i
*( 1} • - r* ^(D^>c2(r!(ó)" 4) f e ^ a )ttcT ,
eA
The divergence term In equation (4-87) contains the entropy
flux, consisting of a thermal and a convective part. The last
term in this equation represents the entropy exchange between
the electrons and the neutrals.
Finally the third order equations in the limit T,* °° will be
given. Again an Ansatz is made:
which will be verified later on. In third order of c there
results from equations (4-67)-(4-69) when Tj+ °°:
3 (0)
A («»>+ ̂ ?>)xb = - A i f(3)d3C - 2Ce eA ce -eA -aA - rd> e^ e
+ —— ra tS°,K .u^ + —(vl 2 A'X1 + B«X ), (4-93)3 v ^ j - e iA ei-lA m^ Te = -m = -q •"
eA TeA
e/ aA .
^ 2 ?8 v f pi e i r
(?>e r e__r., aA a \v(l).3 , i n i \
The tensors A and 3 in equation (4-93) are defined as follows:
V 2 5 (0) -1 3 (A"95)
(1) 2fcT(0) 2 eA-(l)
-56-
kT ( 0 )
, , . 1 aA 9 r u 3 1 ïwhere: a(c) = - — ?rlc 'aT ~Z— J-
T(l) 2c1* 3 c 3c T(l)
In equation (4-94) the entropy exchange with the ions appears.
Because of the conditions (4-82) the first order part of the
entropy vanishes:
0. (4-96)
Then one may add equations (4-87) and (4-94) to obtain the
total entropy balance equation up to third order.
The entropy production term in equation (4-94) can be evaluated
using the expressions (4-62) and (4-64a) in the formulas for
the fluxes. It appears that those parts corresponding to the
multiple collision terms in (4-64a) give positive definite
contributions to the entropy production. This could have been
anticipated because these contributions depend linearly on the
forces defined in (4-88).
Another important conclusion that can be drawn is the follow-
ing: if T--. is independent of c great simplifications occur in
the momentum and energy equations, see e.g. equation (4-78).
In equation (4-94) the second term on the right-hand side which
contains the isotropic correction, vanishes because of the
conditions (4-82). Further it is observed also that in the case
of a constant Tf-i\ t n e cross effects are absent in first order.
-57-
IV-4 The first order isotropic contribution
In section IV-2 the equation for the first order
correction, equation (4-63) has been derived. When Tg+ «• f
is a Maxwell distribution function- The equation for the
isotropic correction f . then reads:(0) e A
2
£ 2 ? VeA
+ -S_J_f_£l f a -l If (0) 1 (4-97)a ( U eA
where:
e eA
The left-hand side of equation (4-97) contains the linearized
collision integral defined in (4-65) which is from now on to be
understood as follows:
Jee<f> mJM<''f™> + Jee«™-V- (4~99)
i.e. asymptotically on the t. time scale. The moment equations
(4-83) and (4-85) will be used to eliminate the time derivative
in the first terra in the right-hand side of equation (4-97).
The Coulomb collision integral can be written as a divergence,
see appendix D. When the following integral operators are
defined:
c °°
I (f) = *jj Jvp+2f(v)dv, J (f) = *£ /vp1"2f(v)dv, (4-100)
P CP 0 C C
it is possible to integrate equation (4-97) once. The
integration constant vanishes, as can readily be verified. The
result is then:
m lL o eA ' ,, (0)*- *K eA ' lv eA '}i eA
m e
-58-
e eA eA e eA
p(0) m <0)lT(0) LJix feA Ci « T l T(0) ^ e A
eA a eA eA eA
V.lJf2T(1)(c)M-1
1).{x;-^) + X q(-^L - f)f<°>}j. (4-101)
eA eA
It is then advantageous to make a change of variables from
(c,r,x2) to (w,r,T2) where:
m e 2 kT^?* , .w := -^TTT, , — ^ | - - f- , V > V + wX | - . (4-102)
2 k T ( 0 ) ' mec 3c 3w ' -q 3w
eA
The functional notations are not altered after this
transformation. Further the function g is introduced which is
related to £ , according to:
g(w) := (1+ — ) f ^ ' . (4-103)
Finally an integral equation for the function g is obtained:
iw 3/2°°
F(w)g(w) - j / x 3 / 2 g(x)dx - -*~- /g(x>d x = Kw>> (4-104)0 W
in which the source terra is defined by:
VTe
3m2 . T ^ !
J$) (0)eA neA
)ee eA
? ,(0) " e 7 ' ^ , Ame7ieA (0) +7^=(0) 7ieA (0) + 7 ^ = J0 T (x)
eA K1e m"7F ° U ;eA ma"7F ° U ; (4-105)
-59-
where: T (w) = —- , T = ( n ^ v Jl Q ) - 1 , (4-106)Q(w)/w a
w , w2 w 2 ( w - j )Gi<w> = TTT̂ T ' G 2 ( w ^ = w Q ( w ) . G3(w> = 7T7TX • Gu(w> = — F 7 ^ — »
G5(w) = eW J { 2 - ^ So - - ^ M-l }e-Xdx,
w 7,2v^ x ( x - f ) , , _xe J {—— Rn ~/_\ ïT7iTv l e <*x»
5. - x
The function F(w) is defined as follows:
F(w) := irV*erf(wS/4 - w^/2 , (4-108)
of which some properties are:
F(w) =Ö"(w3/2), w*0 ; |^(F(w)e"W) = ^ w V W . (4-109)
It can be verified that exp(-w) is a solution of the homo-
geneous part of the integral equation (4-104). The integral
operator is symmetric and real, thus exp(-w) is also a solution
of the homogeneous adjoint equation. Then it is required that:
GO
/e~Wb(w)dw = 0. (4-110)
o
This equation turns out to be the energy equation of the
electrongas. By means of a special operation on equation
(4-104) it is possible to obtain the following simple ordinary
differential equation for the isotroplc correction:
where the new source term J(w) is connected with b(w) in
equation (4-104) via the relation:
(4-112)
-60-
When J is put equal to zero in (4-112) a second order homo-
geneous differential equation is obtained for b which has two
solutions: b=constant and b=w^/2. This means that the second
part of expression (4-105) does not contribute to the final
solution. From equation (4-111) it follows that the general
solution for f . reads:eA
00 Ot>
f ^ = e~W ƒ ƒ eXJ(x)dxdw' + Cie~W + C2we~
W. (4-113)w w'
The constants C, and C2 are fixed through the requirements in
(4-82) leading to:
ƒ rt V'2<lw = ƒ fUAV
/2dw = 0. (4-114)J eA ' eAo o
Thus it was legitimate to make the Ansatz (4-82).
Again in the special case of Maxwell interaction between
electrons and atoms (T,,.= constant, i.e. Q(w) proportional to
w ) the source term b reduces to:
60 e ^ 0> [ {^ -« eA ))SÜJ11 eepeA
Only those parts relevant for the solution are given here. This
expression vanishes for a homogeneous electron temperature.
-61-
IV-5 Electron transport coefficients
With the results of the foregoing sections it is now a
matter of straightforward substitution to obtain the electron
fluxes. This section will be restricted to the case without a
magnetic field. This means that the tensor MT.1. becomes equal
to the unit tensor. As a consequence the tensors G5 and G_6, SQ,
RQ become also proportional to the unit tensor so that e.g.
G5 = G5Ï. The electron fluxes in first order then read:
a<l> := Am c2cf (1>d3c - &°y\} = Xn{v-2 RQX + L0X },aeA e - eA 2 eA -eA 0 Te 0-m 0-q'
(4-116b)
^T^ 4Tn(?)(kT(J))2 » w ( w 5 VeA , eA v eA ̂ T f v 2 .
D 0 " ~ ' X0 ' L0 = J Q 7 ^ — dw'in 3/7 m 3/7 o Wl ; ., ,,_.
e e (4-117)
Observe that 3 . is the thermal heat flux, defined in terms of
the peculiar velocity of the electrons, see also expressions
(4-89) and (4-90). In second order the electron fluxes become:i(P - /m cf(Pd3c = m n^Djgv-^-S , + — S )x +-eA e- eA e eA "L Tev ei «- ee;-m
/IT
' ^R" R ^ + ^ °l " ^ f ö )eA
eE 5k 3k T
where: B~ = (Ö")'{knvTe -ra + ( k 3 i " ~~F^~ 1 ± — ("7?) ~1 ^kT . m SL2 T .
eA a eA
5i eA -m , 6 i reA -q , 3i-m -q , , „o . . , „ „ .+ TTTV + (ft) ^ + ; — * + k
4 lx q ' (4-120)
eA Te peA Te
-62-
In the expressions (4-118),(4-119) the following coefficients
were introduced:
Sei = Iei(1)' Rei = Iei(w-!>' Lei = ^ « " " f ^ -
Ree = W1'*"!*' Lee
:= ƒ f<w>™fdw i (f ,g) = /f(„)/{g(w)}dw. (4-121)o Q(w){w2Q(w)+e} e e o
where £ is a linear integro-differential operator, see appendix
D. This operator also plays a role in the so-called Spitzer
problem, see chapter V. Also appearing in (4-118),(4-119) are:
3m T ( 0 )
£i = ( V i-off»»!+ *icf*ffi+ {k"^]" k"x<a eA eA
+ (k, - k,,)X «X - kc.(V«X - X2) - k,.V»X }x +6i 3i -q -p 5i -p -p 6x -qJ-q
eV(V«E)
4 i q 51l Si—W- + k517 ( VV +
eA
e(k - k ) V(E»X )
CO)eA
(4-122)
where: X := -V l n ( p ^ ) . (4-l22a)
A mean free pathlenght I has been defined by:
I '.= v TJ , v|e = kT^)/me. (4-123)
The parameter fj is of the order e and is proportional to the
e-e to e-a collision frequency ratio:
n(0)c
sLsse e e e 2m v3 e e Snem
e Te o e
-63-
The coefficients k. . appearing in the expressions above are
defined as follows:
00 0O CO
- w r r r x ,k := /e"W{ ƒ ƒ eXJi(x)dxdw' + C + C21wJH.(w)dw (4-125)J o w w'
w w _ dGwith: J.(w) = 4-[-=r-r f e x -j- dxl.
ï dwLF(w) J dx J
o(4-126)
The functions H. are defined by:
H 1 ( W ) = Q T W V H2<W> = f X ' H3(w> = ^
H^(W) = — K — 'ÜÜM , Ht(w) =ü_(wH, ). (4-127)dw b dw ^
Note that in general the coefficients k . depend on the cross
section and on the temperature as well; this as a result of the
definition of the variable w.
The complexity of the second order results makes It desirable
to restrict the calculations to a number of special cases. In
table (4-1) five different situations are specified.
1
2
3
4
5
E = 0
X = --P
X = --q
i "
a eA
Vlnl
Vln(
0
0
T eA ) }
* eS
+ eE
= 0
= 0
- k T ( 0 ) ( k X + XK eA U T - q -
- k T ( 0 ) r k x + xeA A-q
\ v — p / e
table
(4-1)
In each of these situations the expressions for the electron
transport coefficients are much simplified. In general these
are defined as follows:
-64-
,(2) _ % (2) (0)r (2)„ .n(2) •> (2)-eA e - e eA l -p T -qJ -ex '
Hl (2} (2\ (2\ (2\q , — — O h. + A_. X + A X + q ('t—J./OD)aeA q - T) -p -q aex
(2) (2)In these equations the second order fluxes I and q are
^ -ex ^ex
different from the other terms because they are not
proportional to E, X or X . They give no corrections to the
first order fluxes, but are new effects. Their general form is:
in which the vectors Y are defined as follows:
i kT(0) -^ kT(0) - -J kT(0) - -peA eA eA
eA
Y 7 = VXZ, Ya = V(V«X ) , YQ = 7(X «X ) . ( 4 - 1 3 0 )— / p — o —q * — y —p — q '
The ccifficlents a. and b can be expressed in terms of k :
a l ' k l i ' b l " km«
2 "" if i» 2 ~ 4 4 '
a 3 = " a 4 = a 5 = k 5 1 ' b 3 = " b 4 " b 5 = k 5 4 '
a 6 " k 6 1 ' b 6 = k 6 4 '
a 7 "~ 11 mm 51 ' 7 14 ~ 54'5 5
8 ~ 2 .11 3 1 51 ' 8 ^ 1 4 34 54»
a9 = k31 k61' b9 = k34 ~ k64" (4-131)
This section is concluded with some expressions for the
transport coefficients in some special cases mentioned in table
(4-1). The electrical conductivity in case 3, i.e. when no
-65-
teraperature gradient is present, reads:
eA eA
. /? T<°>(ö)7-I * k52<% ~ *?l + k22 4^(^ÖT "OeA eA
JUwhere:
aee p°ei"(4-132)
0 ^ • (4
3/7r e
The thermal heat conductivity in case 4, where there is no
first order electrical current, is:
k5H7 '5p + ^ 5 4 + k 1 4 ) x ] + - £ _ - K j - g - - i )
Ei L .
eA
7(Lee" Vee> " «L.i" V.i> ̂ (4"133)where:
K2 Y26 K T T 2 5 + 5Y24 K34
K D = Y, c - k Y-, c + ^Y^, ~ k,, ,3 36 1*35 3k 6 4 '
K4 = k26 " kTk25 ~ ^k24»
] f 4i T 1̂ 2 T l ~i T Li'
Y2i = 6i ~ r 5i 3i ~ i^li ~ T 11'
Y3i = k6i " kTk5i* (4-133a)
The thermal diffusion coefficient for the electrons up to
second order in case 4 reads:
-66-
m fü T ( 0 )
00 21 + k 2 3 ) ^ 4 ë ( - ^ "O - » a l + f Ree+
3 eA
+ • r T ^ l l c 5 X *X + K R V ' X + K 7 x 2 " k s i V # X + ( k m + k i i > x 2 l 1»120 ->-q - p 6 -q ~ q 51 - p 51 11 p ' '
(4 -134)
where:KS = Y23 " ^Y21 + k 3 1 " k S l 2 k
Tk l l »
K6 = Y33 " ^Y31 ~ k61»K? = Y13 " Kil ~ K l + k T k U ' (4 -134a)
Finally, again In case 4, the coefficient for the Dufour effect
for the electrons reads into second order reads:
A^>+ X̂ 2> = An{R0 + k25 ̂ 4 f e -1) - BR , + ̂ R +D D 0' ° 25 m 4B
V (0) J ei /- eeeA
All the foregoing results for case 4 can be transformed to
those of case 5 by simply replacing k_, by k •
Further It is observed that only those parts in expression
(4-134) and (4-135) that originate from the isotroplc
correction do not obey the Onsager symmetry relation.
Numerical examples are worked out in chapter VI.
-67-
IV-6 Modifications for a seeded plasma
Alkali seeded noble gases are of practical Importance for
MHD-generators, which operate at low temperatures. It is then
nevertheless possible to obtain a sufficient degree of ioniza-
tlon because the seed is easily Ionized. The partial seeding
pressure will be rather low, but the elastic cross section for
momentum exchange is rather high as compared with noble gas
atoms. Therefore the case where the electron-atom collision
frequencies of the noble gas and of the alkali seed atoms are
of the same order of magnitude will be considered. In the
electron Boltzmann equation a term J , (f ,f, ) is added where b
eb e b
denotes the seed. In the expressions for f one has to
replace x. . by T. . defined as:
a M bT + TT(D CD
( 4" 1 3 6 ;
Inserting the expressions for T.'. one obtains (see (4-10Ó)):
-aT
Where an electron-atom cross section for the seeded plasma has
been introduced. When the seed has the same temperature as the
neutral gas it follows that T /T = n./n i.e. proportional toD 3.
the relative seed concentration.
The collision terms in second order of c are also influenced by
the mass difference between the gas and the seed atoms. It is
not difficult to see that for an isotropic function f:
a s,ra e s,T
where:
-68-
Td) b T d ) Td) b a T d )
If the heavy particles are in thermal equilibrium these
collision times are equal. The first terra in the right-hand
side of equation (4—63) should be replaced by expression
(4-138) with f = f . In short the modifications to be made6 y ELS
for a seeded plasma are: replace evrywhere T.. . by expression
(4-136) except in the energy equation: (4-31),(4-85),(4-87) and
(4-94) where (4-139) is to be used. This is also necessary in
expression (4-97) where the energy equation has been used. The
last term in equation (4-97) should then read:
( 4 " U 0 )
m e a
where
*T = Ta
c 3
now:
(0) _aA
[ a 1TeA°
se maT(l)
)f ( 0 )] .
T T ( 0 )
s,m eA „,? - =(D Kl
1 Te
X'-m
(4-141)
The general formulas which are obtained up to now will be used
in chapter VI to calculate the transport coefficients for
actual situations with realistic cross sections. These calcula-
tions will be compared with the results of mixture rules and
with experimental results.
-69-
References
1. W. van de Water, Physica 85C(1977)377.
2. S.Chapman and T.G.Cowling, "The mathematical theory of
non-uniform gases", Cambridge University Press, 1970.
3. R.M.Chmieleski and J.H.Ferziger, Phys.Fluids 10(1967)364.
4. V.G.Molinari.F.Pizzio and G.Spiga, II nuovo cirnento
53^(1979)95.
5. I.P.Shkarofsky,T.W.Johnston and M.P.Bachynski, "The
particle kinetics of plasmas", Addison Wesley, 1966.
6. P.M.Banks and G.J.Lewak, Phys.fluids 11(1968)804.
-70-
V STRONGLY IONIZED GASES
A strongly ionized gas was defined in chapter IT as a plasma
in which all elastic collision frequencies of the electrons are
of the same order of magnitude. Then the parameter e is absent
in the right-hand side of the electron kinetic equation and the
solution of this equation should be valid for arbitrary degree
of ionization. Unfortunately this cannot be fully exploited in
practice for the following reasons.
Firstly the isotropic part of the electron distribution
function shows strong deviations from a Maxwellian as demon-
strated in the preceding chapters, whereas In the present
chapter it shows up in second order. In the second place the
polynomial expansion mostly used to approximate the solution
for the non-isotropic part converges very badly for low degrees
of ionization, especially in the case of argon because of the
Rarasauer minimum*. That is why the restriction has been made
that all collision frequencies of the electrons shall be of the
same order of magnitude, except for the fully ionized limit,
which can be taken without any severe problems. The fully
ionized case Is thus a special case of the results of this
chapter, as far as the electrons are concerned.
The equation for the non-isotropic part of the electron distri-
bution function for a fully Ionized plasma has been solved
numerically by Spitzer and Harm2. Sonine polynomial approxima-
tions were used by Landshof3 and Kaneko4 among others. With the
inclusion of a neutral species the problem has been attacked by
many authors5"9. In this chapter this problem is reconsidered
and it is shown that the equations can be written in the form
of a self-adjoint differential equation, which permits easier
calculations. The connection with the weakly ionized case as
treated in chapter IV is also demonstrated.
When the electron collision frequencies in the Boltzmann equa-
tion are of the same order of magnitude the following order of
magnitude relation for the densities holds:
-71-
7 (5-1)n Ql ;
The electron-electron collision cross section Q is much
m e
larger than Q : the electron-atom cross section. Therefore
the assumption will be made that the degree of ionization is of
the order e:•jp =<?(£), (5-2)
a
which, of course, implies a limitation for the validity of the
heavy particle equations. With the assumption (5-2) the heavy
particle Boltzraann equations read:
3f— a + e 2v7f = e3J + J + eJ . , (5-3)3t - a ae aa ai3f-r- + E2vVf. + e2^ (E + vxB)»Vf. = e3J, + eJ. , + eJ. . (5-4)dt — l m , — — i ie ii i.a
The right-hand side of equation (5-4) contains extra factors e
because the fastest time scale corresponds to the e-a collision
time. It has been assumed that the electron-atom and atom-atom
collision frequencies are of the same order of magnitude. Heavy
particle-electron collision integrals receive an extra factor
e 2 because of the inefficient momentum transfer process (cf,
equations (2-24) and (2-25) of chapter II).
The electron kinetic equation now reads:
3f eE 3w— e + ec«Vf + e2w «Vf - (e— + eu w xb + z-f? +e2(c«V)w +3t e -a e *• ra ce-a - 3t - - a
e
+ e 3 ( w »V)w ) . V £ - u> c ( b x V f ) =• J + J + J , , (5-5)-a -aJ c e ce- - c e ea ee ei*
where the transformation to the hydrodynaraical velocity of the
neutral gas has been made as in chapter IV:
c := v - w . (5-7)_ _ ~3
In the following sections a similar procedure as in chapter IV
will be followed. Firstly the heavy particles are dealt with;
after that the electrons.
-72-
V-l Heavy particle results
The heavy particle equations are only slightly altered
when compared with the weakly ionized gas in chapter IV sect.l.
Equations (4-1) and (4-4) remain unchanged so that f is a
Maxwellian as in (4-7). The equation of continuity for the
atoms is also identical to (4-8). The factor e2 in the right-
hand sides of (4-9) and (4-10) is now replaced by e. The
results in zeroth order from the balance equations are again:
3n(°> 3w<°> 3£<°>3. a a -. . — _ ,.
•57 = -rr = r- = 0 . (5-7)3 T 0 3T Q 9 T 0
The results from the continuity equations are the same as in
the case of a WIG. (Jp to second order they read:
3n<°> an*" 3n(°> 3n(1> 3n<2>
< - A ~~ °' < + < + < + V'(ns as0>> = 0, (5-8)where s=a or s=i. The macroscopic equations for the ions are
all the same as in the WIG, see equations (4-25)-(4-39)•
The momentum- and energy equations for the atoms now yield up
to second order the following results:
au(0) (1)
3»<0>
) ,(1) ,(2) ,(0) ,<l)
(5-12)
-73-
In order to proceed the kinetic equations have to be considered
simultaneously. From the equations (5-3) and (5-4) the follow-
ing equations are obtained in first order of e:
3f(0) a,(U£ • + " « .j ( f V ' j t J (f(1)f(0)) + J f(f<
0\f<0)>,3T. 3xn aa a a aa a a ai a ' 1 '
(5-13)
It is shown in appendix B that from these equations an(0)
H-theorera can be derived Implying that f. relaxes to a
Maxwellian, with a hydrodynamical velocity and a temperature
equal to those of the atoms, when T,+ °°:
f(0) <0>, "I , , il a l ,
Contrary to the case of a WIG the conclusion that f doesa,as
not depend on Tj cannot be drawn. From equation (5-13) it
appears that only when T,* °° the same equatTon for the firstorder contribution f . as in the case of the WIG is obtained:aA
J C f ^ f ^ + J (f^f^) = 0. (5-16)aa aA aA aa aA aA '
With the Chapman-Enskog choice:
it can be concluded that the first order correction is absent
if T,+ »:
^ = 0. (5-18)
The second order equations derived from (5-3) and (5-4) read:
3f(0) af(l) af(2)
f'. 3f(2)i _j
O
-74-
eE
mi
p(0)
_ (0) (1) (i) (0) (0)" Jia(fi 'fa )+Jia(fi 'fa )+Jii(fl
When Tj+ o» these equations reduce to the Chapraan-Enskog
equations for the corrections f.. and f . :
f{0)>.(5-20)
^ a A v.vf
3(0)aA
ff(0) f ( 2 ) ) + J
aa UaA • aA ; J(2) f(0)aA ' aA ; J
3f(0)
r(0) f(l)j
(5-21)
(0) (1)Jii(fiA 'fiA
(1) (0)Jii(fiA >fiA
(5-22)
In order to evaluate the right-hand sides of these equations
the balance equations have to be considered in the limit T,*
As far as the ions are concerned they are given by equations
(4-37),(4-38) and (4-39),which will be repeated here:
(5-23)
dw
The macroscopic atom equations read:
(5-25)
(5-26)
,(0)d2aA (0) (0)iA (5-27)
(0) 2
a I - -aA(5-28)
-75-
Addition of (5-25) and (5-28) gives with the aid of (5-23) and
(5-26):
From (5-25) and (5-28) it can then be concluded that there is
no energy exchange between Ions and neutrals on the t
scale in this order:
where c = v - w , a slightly different definition as used in
(5-6). This result is the same as obtained in case of a WIG(see
equation (4-43)).
The energy equations thus reduce to the Euler adiabatic
equations of state. Momentum transfer, however, does take place
on the t2-timescale. Addition of the equations (5-24) and
(5-27) gives:
5' = 0, (5-31)
where ph = m/»+ .f™. ph - p ^ + pg', Th = T^\ (5-32)
The left-hand sides of the equations (5-21) and (5-22) can now
be evaluated in terms of the macroscopic quantities. After a
transformation from the variable v to the new velocity variable
c = v - w one finally obtains:
aa aA ' aA ai aA ' iA
£ . { ( ^ - |)Vln(Th) + ^ l ] ^ . (5-33)naA
- 76 -
( f f ) + J Cf ( 1 ) f ( 0 ) )i i l t l A * iA ; J i a U t A ' aA ;
+ ^ ) 5 l a } ] f <J\ (5-34)h h n,
iA
where n, = nV. + n*. and the diffusion driving forces are
defined by:
n.. (ID — in )n.a n m n , n ,
-ia - a i »• r^ J n p ^ph
(5-35)
These equations can be seen as a special case of the ones
obtained by Chmieleski and Ferziger8. This is due to the
restriction made in relation (5-2). Their equations for the
heavy particles are coupled, whereas here equation (5-34) can
be solved independently for f.A . Substitution of the solution(2)
into (5-33) then gives an equation for f . The solutions canaA
be obtained by means of a traditional Sónine polynomial
expansion10.
The ion- and atom Chapman-Enskog equations (5-33) and (5-34)
are thus seen to be only weakly coupled due to the choice of
the specific domain of degree of ionization. The coupling
becomes stronger when the ion density increases. See also the
corresponding equations (4-46), (4-47) for the case of a WIG.
-77-
V-2 The electron kinetic equation
The kinetic equation for the electron distribution
function, equation (5-5) will now be treated along the familiar
lines. The zeroth order equation reads:
(5-36)
It is easily shown that from this equation an H-theorera can be
derived implying that the zeroth order electron distribution
function relaxes to a local Maxwellian when T_+ °°:
e,asexp{-
T
e,as2kT(0)
"I- (5-37)
e.as
The left-hand side of the first order equation is the same as
in equation (4-50), whereas the right-hand side now becomes:
( 0 ) ( f a )J ( 0 ) ( f a ) ) + J (f(e a e e e e
J (fe e e
+ Jei { e ; + Jei U e >£i >m
When T Q+ °° the first order equation reads:
(5-38)
eE.
<f ) + J ( f ) + J ! ( f ,fi ) ,ee e,as ei e.as' ei v e,as' i,as"
(5-39)
where J (f) is the linearized collision operator (see (4-65)).
In the next section it is shown that n and T do note,as e,as
depend on tp hence from (5-37):
TF,"e,as _
E 0. (5-40)
Then equation (5-39) becomes:
ce - c e,as - ea e,as e,as ei
-78-
p >e,as
„(0) C3e,as
where appeadix A and expression (5-37) have been used to
evaluate the right-hand side; u. Is a diffusion velocity,
see (2-8). The isotropic part of this equation simply reads:
J (f ( 1 ? ) = 0. (5-42)ee e,as
The function f is assumed to be expanded as in equatione f as(4-58). The general solution of equation (5-42) is:
HI TOT)e,as
see chapter IV section 4. In (5-43) A and B are as yet
arbitrary functions of space and time. The choice:
n ( 1 ) - T ( 1 ) = 0 , (5-44)e,as e,as '
makes them zero. Then the conclusion is that there is no first
order isotropic correction in a strongly ionized gas:
From equation (5-41) it appears that f is proportional to c
only. The magnetic field makes it necessary to separate the
components of f in the following way:
n(0)
-fe!L=-7̂ f Hf^ + f^+f^},, (5-46)
where: A = (b'A)b, A. = A - A.., A = b*A. (5-47)
The vector A stands for one of the vectors between braces in
eE'(5-41): Vln(p(0) ), Vln(T(0) ) , —^rrr- . »i°^ . The summationV ie,as" e.as'' ,_(0) -i.as
-79-
in (5-46) is over these different possibilities. The general
form of the equation determining f then reads:1 61 as
2C ,ii(0) B(0)
ai 6. f, +ce it i
ra c"e
1_ ie e .cis , , vJfj = ^QJ—*— b(c),ne,as (5-48)
where i = ll,l,t; the subscript k has been omitted and 5 Is
the Kronecker delta. If A = 71n(p ) then b(c) » 1, and so
on, see equation (5-41). Equation (5-48) will be dealt with
further in section 4. The operator J was defined in (4-65).1 ee
Without the term (T (c))-1f the equation is Identical to the
equation that has been solved numerically by Spitzer and Harm1.
In second order the electron Boltzmann equation has the same
left-hand side as equation (4-51) of the WIG. The right-hand
side now reads:
eaJ(2V°\f(0)) + J (ea e a ee
ee
( 0 )) (0))j ( fJei ( e j ( fJei U e ' (5-49)
In the limit T Q * <*>, the isotroplc part of the second order
equation reads:
df
f"*... 3kTvf ywe.as -a,as -e,as
e.as
(ea e,as e,as
f(0) p (1) . (1) + j ?(2)eas u ee -eas eas - ee easee -e,as e,as - ee e,as
\\fW -c.f5°> )ei -e,as - i,as (5-50)fi^hf ,fS ),° ei e,as i,as
where Pfl Is the operator which when operating on some function
gives the isotropic part of that function. In general (see also
appendix C):
1c
- 8 0 -
Equation (5-50) is an equation for the isotropic correction~f 2 }
f and is of the same type as equation (4—97) for the first
order isotropic correction in the case of a WIG.
When Tg* » the nonisotropic part of the second order equation
reads:3f( 1 ) eE' m e 3w( 0 )
e,as
+ J!f_ f(°> <cc>:Vw<0) - u, c (bx 7 f(2) ) =kT(0) e,as — -a,as ce- - c e,as
e,as
- J(0)(f(2) ) + J (f(2) ) + J«\fW ) + J™<fW .ff(1) )ea e,as ee e,as ei e,as ei e,as i,as
(2 )Next f is expanded into irreducible harmonic tensors. Then
e,as v
equation (5-52) can be separated into two equations: one for
-e^as atld one fGr =e2=the aid of appendix A.
The equatit
and reads:
f and one for f • The collision terms are evaluated with-e.as =e.a«=
The equation for f is of the same type as the one for f^ -e.as 1V -e.as
<2>(2)
o bxfce- -e,as
e,as
•s»< 0 )
e,as
(2)(2)The equation for f takes the following form:
=e,as
2u,c, bfce- =e,as
' ee,as
-81-
-e,as m c -e,as ._(0) e,as -a,ase,as
C n(0) ( 0 )
ei i,as-i,as{6 4 8 )f(l) _f (1) (1) _5
m lc° cH3cJ-e,as * ee -e,as - -e,as -
When T,+ <» these equations simplify further since then uj =0.~1 ) 3.S
Equations (5-50),(5-53) and (5-54) can in principle be solvedIf the first order contributions f\ and f are known. The
-i,as -e,as
equation for the latter will be discussed In section 4.
Contrary to the case of the WIG it was demonstrated in this
section that the equations for the isotropic and rconisotropic
parts of the electron distribution function are found in the
same order. It also appeared that there is no need for a first
order isotropic correction as In the case of a WIG.
-82-
V-3 The electron macroscopic equations
The moment equations in a strongly ionized gas will be
treated with the aid of the corresponding equations (4-67) to
(4-69) of the WIG. The only alteration to be made is to drop
the factor e in front of the e-i collision term in equations
(4-68) and (4-69). The zeroth order equations now read:
e e 1 9 T 0 ce-e -> ' l i 3 • ' e - e
In f i r s t order of e:
^ • w; + v.<.f'4»', - o,
(0) (1) (0)/ _ » Oil OU /r\\ / / - \ \ / i \ oU
n . (u w ) b . n ( 1 > . u ( 0 ) xb =e e c e - e -a - e e ce-e -
( 0 >
(0) £(1) m c
c f d c _ 2C ( 0 ) ; £e - e e i i J c J e
(5-58)
(5-59)
ral
terms in these equations to vanish. The equations then read as
(0)When T.+ °°, f has become isotrcplc, which causes several
6 « clS
-83-
follows:
a (0) .„(0)on oT
IT6'38 = IT6'38 = °' (5-60)
„ (0) . (0) (0) (1)Vp + en E + m n io u *b =e,as e,as- e e,as ce-e,as -
?r n ( 0 )
_ƒ(_!_ + ei i,as)m (l) d 3 c + _ 8 _ m n(0) y (0)T(l) m e 3 e~ e' a S 3/27 e e' a S ei-l'as
where: vfil = " [ ^ 0 ^ ^In contrast to the case of a WIG this equation does not 'close'
when T,.. Is Independent of c, I.e. the case of Maxwell inter-
action.
In order to reduce the size of the formulas the second order
equations are only given in the limit TQ->- <»:
9n ( 0 )
if'** + 7-(^1S4:L) = 0, (5-62)
xb +rf-2— + ei i.aSN- e.as r. . ei-e,as_
(5-63)
3 (0) {^e,as (1) (0) , + 7 . ( (1) + (0) (1) } +Te.as Idi, ae,as e,asJ ^e.as Fe,as-e,as'1
e,as-e,as - e,as -a,as ei i,as-i,as •* 'c3 e,as
or ™ n ( 0 ) t ^ 0 5 a_ /Leimenl,asr fl +
KIi,as _ ) f ( 0 ) d 3 c +
m1 J l c m c2 3c e.as
e
(0) _ 2
a ' a S . . e —+ k-T ; el-f-1—) If4 ; d3c. (5-64)
See appendix A for the evaluation of the moments of the
collision integrals. When use is made of the fact that f
a Maxwellian further simplifications can be obtained.
-84-
If the local entropy density is defined as:
it is possible to write the energy equation (5-64) in the
following way:
9 Q ( 0 ) .(I) , (1)
2 T ' ' ' Te,as e,as
(0) o (0)in T . ra c ... 8m v . ._. T.er a,as n -il f e „(O) 3 e e l (0)
7 ? ^ 5 r ?e,as e,as i e,as
- 2C ̂ i0) OJ°> •ƒ V0 d3c, (5-66)
ei i,as-i,as ' cJ e,as
where i , X' and X are defined, just as in chapter of the~6) 3S *"* ni "~T
WIG, as:
1^> :=mn<°> u^> , X' = - *• (E' + ^ L , P ^ ),-e,as e e,as-e,as -m m v— e e,asJ
X = - Vln(T(0) ). (5-67)
-q e,as
Equation (5-66) is the entropy balance equation. The first term
on the right-hand side is the entropy production, wich will be
shown to be positive definite. When Tj+ <*> it is clear from
(5-64) that in the case without a magnetic field the correction
f . had the general form:— GA
£ '" T^T- [ A ("^ X + B<">x ' (Te ee eA
where w is now the new independent velocity variable:
(5-69)m c 2
T ,eA
The functions A and B are solutions of the following equations
(see also appendix C and next section):
-85-
= w 3 / 2, (5-7Oa)
JCB = w3 / 2(w - 5 /2) , (5-7Ob)
where «6 is a symmetric, negative-definite integral operator.
With the definition of the fluxes 1 , and q . and the infor--eA aeA
nation just given the relations between the fluxes and the
forces read:1 ( D = m n
(0)|D ( 1 )v- 2X + D(1)X 1-eA e eA ' Te-m T -q
(5-71)
-q'
V 2
where: D U ; = -D JAJÜAdw, D^X) = -TlJvZAdv, D =3 . .ee
- (0 )v'To
(5-72)
' g j " ~ > --J ~gj » "g
o o 3v il
i< 1 5 - i TAZRHW i ( 1 ) - ! 7 R 2 R ^ i - "V'eA "Te\ n = -A JA*Bdw, A = -A JBotBdw, A = —————-—U eo 6o 6 3v /2
ee
One can clearly see that D and A are positive-definite
and that the Onsager reciprocity relations hold:m n . 1/ — v_, Art • \j~~to)
The entropy production rate is equal to:
( 0 ) . ( 1 )1 r e e A ( 1 ) ? r ( 0 / ( 1 ) TJ i ( 1 ) P I
Trv\ i ^ x™ + lm«n^A " T + JXm*X„ + ^ ^ i (5-74)(0) 2
m e e A T V2 ~m ~q q
eA Te Tewhich is proportional to:
» X 2 « X «X2 ) / 2 ^ 2 Z (5-75)
This expression is positive-definite if:
{JAZBdw} - /A^Adw/B/Bdw < 0, (5-76)O 0 0
which can be proved with the aid of the Schwartz Inequality.
In the following section the solution of the equations
(5-7Oa,b) will be decussed In detail.
-86-
V-4 The nonisotropic part of the electron distribution
In section 2 it was shown that the first order
contribution to the electron distribution function is
proportional to c only. It is also clear from the equations
that the inclusion of a magnetic field does not introduce extra
difficulties. In case of a zero magnetic field equation (5-48)
for f = f reads (see appendix D):
wjw °°j(-|x5/2 - ̂ x3/2)f(x)dx +(|w5/2 - |w3/2)/f(x)dx + 2w3/2f(w)0 W
|^(2wF(w)|£) - v(w)f(w) = w3/2b(w), (5-77)
/* , w 3 / 2 ,where: v(w) = -r— [v . + Jexp(w). (5-7S)
ee " T ( 1 ) ( W ) V T
The function F(w) was defined in (4-108), see also appendix D.
The problem is now reduced to solving euation (5-77), or:
2f = (J5 - v)f = w3/2b(w), (5-79)
where X is the part of X coming from the e-e collision term.
The other collisions are present in the function v(w). One can
easily verify that <£ and X are symmetric operators. When «C is
differentiated once a pure differential-equation of the Sturra-
Liouville type is obtained:
" W | f =-|j(ge"W), (5-80a)
(5-80b)
In the same way an operator J> can be obtained from
£ "W f | (5-81a)
v(w) )e"2w |f j +
+ e~W -|j(v(w)e"W)g(w), (5-81b)
Inspection of this operator shows that the last term of it
vanishes when the plasma is fully ionized, since then v(w)<* e .
-87-
For that case, which Is referred to as the Spitzer problem, the
operator 3) Is defined as followed:
J£] - [4F(w) + -^ eW]e-2wh(w), (5-82a)
J^h] =£&, h - Jf , (5-82b)
where t, = n.Z^/n Is the Ionic charge number,i i e
The accepted method to solve (5-77) is through an expansion
into a finite number of orthogonal Sonine-(Laguerre-)
polynomials11, which gives a set of linear algebraic equations
for the unknown coefficients in the expansion. This method
essentially is the Galerkin method, which can, of course also
be applied to the equations in differential form. There are,
however, some difficulties. It appears that the operator 3> is
not symmetric in all cases. By means of partial integrations
the following relation is obtained, valid for functions that
are bounded at w=0:
00 3f2 3fj -
J f ^ d w = [v(w){fl(w)_ - f2(w)— \]^Q + Jf/fjdw. (5-83)f2(w)0 0
From equations (5-70a,b) it appears that two source terms are
relevant: b=l and b=w-5/2. This makes the use of Sonine
polynomials of order 3/2 obvious, if one examines some
properties of these functions:
s(0) . ! s(l) = 5/2 - „ S( n ) = V r(n+5/2)(-w)k
b3/2 L' b3/2 D / Z W' b3/2 k^ o (n-k)!k!r(k+5/2) '
^ 3 ; 2 ^ 6nm. (5-84)
o
With the expressions for uC given in (5-77) it is possible to
calculate the coefficients:oo
X = /wpe~w £(wqe~w)dw. (5-85)q o
This is done in appendix D. The approach is some what different
from that in the litterature: the calculations presented in
-88-
appendix D are valid for arbitrary values of p and q, whereas
in the other calculations p and q are restricted to integer
values2/7/11. In terms of the operator «& the problem stated in
(5-79) reads:
jög = e~WJ^(w3/2b(w)). (5-86)
By means of partial integrations one can show that if gj and g2
are solutions of (5-86) with corresponding source terras bj and
b 2 and if f^ and f2 are the related solutions of (5-77) the
following identity holds:
00 CO
ƒfj2f2dw = /g1»g2dw, (5-87)o o
which directly gives the transport coefficients in (5-72) in
terms of the solutions of (5-86).
The matrix elements for the operator 3> are defined as:CO
6 = fwpA(wq)dw, (5-88)pq J
o
which are also given in appendix D. The calculation of these
coefficients is easier than for X ; they are also valid forpq
non-integer p and q. There still is one little problem: the
matrix is not symmetric for every set of functions, according
to equation (5-83). If p and q are natural numbers there is no
problem and 3> is symmetric. If p and q are non-negative
integers there is only one pair (p,q) for which the symmetry
relation does not hold:
*01 - A + ?10 = " W + k J(w-Dw2e~WQ(w)dw], (5-89)J01 " 4
" o
where Q(w) is related to the e-a collision cross section and is
defined in (4-106). In (5-89) it is also assumed that
lim W 3 / 2 T , 7 > ( W ) = 0. The parameter B is defined as:
6 = ™ee'2 = -^~ — = -^~ ̂ —^- {—J£-) , (5-90)
where r is the Landau length.
-89-
In the Galerkin method the solution of, say, equation (5-86) is
approximated by a linear combination of a finite number of so-
called co-ordinate functions <j> (w):
N
8M(w) = I a • (w) (5-91)n=0 n n
All these functions d> satisfy the boundary conditions and then
constants a are then fixed by the requirements:00
o ' w
If the functions to be chosen are <J> = w , the equations (5-92)
take the form:
N °°
£ \ n a n = /e"W|^[w3/2b(w)]<t,k(w)dw, (5-93)n=0 o
so that the matrix of the equations is not completely
symmetric. It can, however, be made symmetric if the first
equation (k=0) is replaced by:
00
/{o?fN - w3/2b(w)}e~Wdw = 0, (5-94)o
where X and f are related to J> and gN according to (5-81)—w
respectively. The function e is the solution of the homo-
geneous equation f=0. Therefore:CO CO COt —W "^ t —W f —W 3 t —W *\Je ££ dw = - Jv(w)e fN
dw = -Jv(w)e ~%~ie 8„Jd w- (5-95)o o o
Integrating by parts and using: e —[v(w)e J = ó61
one obtains:N f^ N
E a n / * n * l d w = ( 4 + ^00^a0 + Io " "* n=0 o n=0/e~W ̂ fNdw = ̂ Ia 0 + f an /«/id» = (^ + \QUQ + f . «
n=0
(5-96)
which shows that indeed the matrix is symmetric now. The system
of equations is not inconsistent, because the relation that
should be valid if (5-96) and the k=0-equation of (5-93) both
-90-
hold reads:
£N(°>
This relation, however, follows directly from the integro-
differentia! equation (5-77) for the exact solution. Thus it
may be expected that (5-97) is approximately satisfied with an
accuracy increasing with N.
Next an example is given: the calculation of the electrical
conductivity in the Spitzer limit with the aid of the operators *
jt> . In terms of this operator the problem then reads:
»ap = —(F(w)e"W - -J), p(w) = — h(w), (5-98)
where the right-hand side results from integration of equation
(5-86) with b=l; see also (4-109). The constant of integration
is chosen such that if w + « the right-hand side of (5-98)
becomes zero. On the basis of the general relation for the
diffusion velocity u . the first order electrical conductivity
is equal to:
, , 6itm v3 e2/2if
0 ( D = K e T e ° = 3j Kco (5-99)
2 ee2lnA e
where a . is the Lorentz conductivity of a fully ionized
plasma, i.e. taking only electron-ion collisions into account.
The constant K is related to the solution of (5-98) as follows:
O3 CO ^
K = -i- - — Jp*Spdw = i + J w V w /p(w')dw'dw, (5-100)
fy o o o
which again shows that the conductivity is always positive as
the operator Sb is negative-definite. The exact value of K has
been calculated numerically by Spitzer and Ha'rra1 and is equal
to 1.975 if c=l.
An approximation with polynomials can be made as follows:
NP(w) " P™ " I « »n« (5-101)
N n=0 n
If N is not too large there is no need for Sonine polynomials.
-91-
For N=l a system of two equations for aQ and
leading to:
K = I
results,
(5-l02)C 64<;2 + 244G/T + 288
To obtain the same result with the operator X, which has been
done by Landshof2 and Kaneko^, one has to solve three equations
for three unknowns. The numerical values of K for higher N
fully agree with their results. Substantial improvements,
however, can be obtained if non-integer powers of w are
admitted as co-ordinate functions. If p is approximated by:
p = pN
NI
n=0a wn
n/2 (5-103)
the result for N=l is even better than the fourth approximation
of Landshof. If N=2 the result cannot be distinghuished from
the exact Spitzer and Harm result, see table (5-1). If N=l the
result for K with approximation (5-103) becomes:
135ÏÏ- 32TT
256(
K = 17TT)
(15 4 2)ir - (2
(5-104)
N
1
2
3
Landshof2
1.9320
1.9498
1.9616
App.(5-101)
1
1
1
.9498
9616
9657
App.(5-103)
1
1
1
.9620
9751
9757
table (5-1):
values of K
for 5-1.
Near the origin the solution of (5-98) can be represented by a
Taylor series in powers of w 1 / 2 which could be an explanation
for the good results obtained with approximation (5-103).
-92-
This section is concluded with an examination of the limit of
very small degree of ionlzation. Equation (5-79) can be written
as follows (see also (4-106)):
^ w2Q(w)eWf(w) = - w3/2b(w) + (JS- £peW)f(w). (5-105)
If the degree of ionization is small 6 is a small parameter.
The solution of (5-105) may then be sought in the form of an
expansion in the parameter $. One then finds:00
f(w) = I f (w)Bn, (5-106)n=0 n
where: f. = - *£ M^£l , f = - j C L ^ e » - 4 - £}f . n > l ./w Q(w)/w " W2Q(w) /IT " l
(5-107)
It is readily verified that the first two terms of (5-106) are
equal to the first order contribution plus the multiple
collision parts of the second order contribution of the
function f . in case of a WIG. Thus the connection with the
weakly ionized gas theory has been verified.
-93-
References
1. L.Spltzer and R.Harm, Phys.Rev. 89(1953)977.
2. R.Landshof, Phys.Rev. ̂ É/ 1^ 9) 9 0 4» 82(1951)442.
3. S.Kaneko, J.Phys.Soc.Japan 15(1960)1685, 17(1962)390.
4. R.S.Devoto, Phys.of Fluids £(1966)1230, 22.(1967)354,2105.
5. W.L.Nigham, Phys.of Fluids 12(1969)162.
6. CH.Kruger.M.Mitchner and U.Daybelge, AIM J. £(1968)1712.
7. C.H.Kruger and M.Mitchner, Phys.of Fluids 10(1967)1953.
8. R.M.Chmieleski and J.H.Ferziger, Phys.of Fluids
10(1967)364,2520.
9. L.C.Johnson, Phys.of Fluids 10(1967)1080.
10. J.H.Ferziger and H.G.Kaper: "The mathematical theory of
transport processes in gases",
North Holland Publ. Comp. 1972.
11. M.Mitchner and C.H.Kruger: "Partially ionized gases",
J.Wiley, 1973.
-94-
VI NUMERICAL RESULTS
In this chapter the results of chapters IV and V are
applied to several practical situations. The shape of the
isotroptc correction is computed numerically for different
electron-atom cross sections. These are the hard spheres inter-
action model and the cross sections for neon and argon accord-
ing to experimental data obtained from litterature. The values
of the 36 basic coefficients k. ., which appear in the
expressions for the electron transport coefficients are given
for these cross sections. For other cross sections than the
constant hard spheres cross section these coefficients are
functions of the electron temperature.
Transport coefficients are calculated in several special cases
and are compared with results obtained by means of mixture
rules and with experimental results. When comparison with
experiment is made one has to bear in mind that not all
processes and effects have been taken into account such as
Inelastic collisions and impurities. On the other hand experi-
mental data suffer from rather large inaccuracies. These are
due to Several causes such as the lack of thermal equilibrium
and the presence of impurities.
Results obtained with the equations of the strongly ionized gas
(SIG) are also given and are included in some of the figures.
The better convergence with other functions than polynomials,
as shown already in chapter V for a fully ionized plasma, is
also observed in plasmas of a much lower degree of ionization.
In all calculations mentioned above it appeared that the cross
section of argon presents some difficulties, following from the
fact that it possesses a so-called Ramsauer minimum In the
energy range considered.
-95-
VI-1 The isotropic correction
In chapter IV the general solution for the first order
isotropic correction in a weakly ionized gas was given in
equation (4-113). There are six different functions J so that
there are in fact six isotropic corrections. See expressions
(4-107) and (4-126). In the numerical procedures the following
integration Is actually performed:
w G,(w')(6-1)
The solution of the homogeneous equation is then added after
the constants Cj and C. have been fixed by the requirements
(4-114). The isotropic corrections are given in figures (6-1)
to (6-3) for the cross sections of the hard spheres model
(hereafter denoted by HSM), and of neon and argon. The cross
sections for neon and argon were taken from references 1 and 2
respectively. The different isotropic corrections are numbered
according to the indices of the function G. ; see (4-107).
The reference cross section QQ has been chosen 10~20 mZ^ SQ
that the diraensionless functions Q(w) and hence the isotropic
functions are uniquely determined.
Characteristic for all isotropic correction functions is the
rather large peak near w=0 and the occurrence of two positive
zeros. The resemblance of the functions for neon and for the
HSM possibly implies that the HSM is not a bad approximation
for neon. The functions for argon have the same shape except
for the last two, and the magnitudes are larger than for the
other cross sections. This must be a consequence of the
Ramsauer minimum, which is absent in the neon cross section.
The coefficients k. . are also computed for these different
types of cross sections. See equations (4-125)-(4-127) for the
definition of these coefficients. They are the basic coeffi-
cients for the contributions of the isotropic correction to all
transport coefficients. Except for the HSM these coefficients
-96-
are functions of the electron temperature. In table (6-1) these
coefficients are give for three different cross sections. The
calculations were performed with a possible error of about one
percent, which is good enough when compared to the accuracy
with which the cross section data have been determined. The
constants for argon are significantly larger in absolute value
than for the other cross sections. Again this is due to the
Rarasauer minimum. This may invalidate the ordering and hence
severely restrict the applicability of the results.
The coefficients in expressions (4-132) to (4-135) for the
electron transport coefficients are algebraic functions of the
coefficients k .. Table (6-2) gives the values for the HSM,
while the results for neon are plotted as functions of the
electron temperature in figure (6-4). The temperature scale is
thereby chosen such that an atmospheric plasma in thermal
equilibrium in this temperature range is weakly ionized.
The effect of the isotropic correction can be demonstrated by
adding the zeroth order Maxwellian. This has been done In
figures (6-5) and (6-6) for an atmospheric argon plasma. The
other cross sections give similar results, see figures (6-1) to
(6-3). Figure (6-5) shows the influence of an electric field on
the isotropic electron distribution function and figure (6-6)
shows a similar effect due to a temperature difference between
electrons and heavy particles. From the source term (4-105) for
the equation of the isotropic correction it appears that
isotropic correction for a homogeneous plasma increases with
the square of the electric field and is proportional to the
temperature difference. The direction of the effect is the same
if the electrons have a higher temperature than the heavy
particles, as can be seen from figures (6-5) and (6-6).
When gradients are present the isotropic corrections numbered 3
to 6 are needed. For the special case of Maxwell interaction
between electrons and atoms there is an isotropic correction
only if a temperature gradient is present. See equation (4-115)
which gives the source term in that case. Therefore this model
-97-
seems to be less suited for a description of the electron-atom
interaction than the hard spheres model.
j-
0.40
-0.87
-0.87
-6-41
-0.10
0.93
-1.26
2.35
2.35
15.2
0.30
-2.50
0.81
-1.75
-1.75
-12.8
-0.21
1.85
-3.43
8.02
8.02
63.9
0.93
-8.49
3.96
-7.63
-7.63
-50.9
-0.96
8.11
-9.29
21.9
21.9
176
2.52
-23.1
Table (6-la): k. . constants for hard spheres model (HSM)
.078
-0.30
-.046
-0.57
-.0026
.046
-2.74
7.78
1.14
12-4
.075
-1.19
0
-0
-0
-1
-.
0
.14
.77
.12
.63
0058
.12
-1
6
0
.25
.15
.95
12.0
.
-0
046
.95
6.
-20
-2.
-32
-0.
3.
99
.2
98
.6
19
10
-3.
17
2.
34
0.
-2.
59
.9
77
.7
13
75
Table (6-1b): k. . constants for neon at T = 5000K.ij e
-4.
21
-4.
14
-3.
7.
59
.9
30
.2
16
16
-101
113
-54.4
148
-7.16
23.2
-45
56
-25
69
-5.
13
.5
.0
.6
.9
05
.7
16
-56
12
-39
7.
-16
.1
.2
.5
.9
16
.6
208
-226
110
-301
14.3
-44.3
102
-134
57.0
-158
11.1
-30.5
Table (6-lc):k. , constants for argon, T = 5000K, data Milloy2
-93-
3.32
11.2
-0.79
1.00
-2.30
4.39
-78.1
112
-52.0
133
1.01
11.9
-35.4
64.9
-27.3
67.5
-2.15
11.0
7.48
-43.7
10.9
-26.1
4.39
-10.9
160
-215
103
-264
-4.69
-18.7
87.6
-147
63.5
-159
2.50
-21.1
Table (6-ld): k . constants for argon, T = 5000K,
data Frost and Phelps3.
k 1 2 =
k52 =k 2 2 =
K ~
K2 "k 24 =
S "
-1.26
0.30
2.35
149
31.1
8.02
14.1
k 1 4
K3k 5 1 +
K5
k 5 1k n
= -3.43
= -13.6
= 0.93
= -2.61
= 0.38
= -0.10
= 0.40
k 23 =
<? =
k 2 1 =
k 2 5 =
^25 =
Y3 5 "
-1.75
5.02
-0.87
-7.63
-62.5
-7.93
7.63
Table (6-2): Some coefficients for the HSM appearing in
equations (4-132)-(4-135).
-99-
5
-1 -
-20 1 2 3 4 5 6 7 8
-40 1 2 3 4 5 6 7 8
Fig.(6-la,b,c) Isotropic correction functions for the HSM.
-100-
0 1 2 3 4 5 6 7 8
0 1 2 . 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
Fig.(6-ld,e,f) Isotropic correction functions for the HSM.
-101-
1
n
i
2
30
//
JJ1
1 1 1
I I I
2 3 4W
1 a
NEONTe=5000K .
| |
5 6 7
3
M-l
0 1 2 3 4 5 6 7
0 1 2 3 4 5 6 7
Fig.(6-2a,b,c) Isotroplc correction functions for neon1,
at T = 5000K.e
-102-
1 2 3 4 5 6 7
. 1
* O
- . 10 1 2 3 4 5 6 7
3Is
-1
-2
1 1 1
J1 1 f
1
• ^ - ^
1
1 ' f
NEON '
V 5000K
0 1 2 3 4 5 6 7UI
Fig.(6-2d,e,f) Isotropic correction functions for neon1,
a t T = 5000K.e
-103-
1 2 3 4 5
Fig.(6-3a,b,c) Isotropic correction functions for argon,
at T = 5000K.e
-104-
0 1 2 3 4 5 6 7
Fig.(6-3d,e,f) Isotropic correction functions for argon,
at T = 5000K.e
-105-
m
• y
•
•
/
1 1 !
11 -
4000 5000 6000 7000TEMP. (KELVIN)
10
8
6
4
Z
-2
^ —
k
^ — — ;
. k u •
SI
1 , 1 . "
4000 5000 6000 7000TEHP. (KELVIN)
4000 5000 6000 7000TEMP. (KELVIN)
60
50
40
30
20
10
—
'- - y
- &s1 1 . . _
• i
i f•
• t
i
s
4000 5000 6000 7000TEHP. (KEIVIN)
4000 SOOO 6000 7000TEHP. (KELVIN)
-1-44000 5000 6000 7000
TEMP. (KELVIN)
Fig.(6-4) Several coefficients appearing in the transport
coefficients for neon1; see equations (4-132)-(4-135)
-106-
0 1 2 3 4 S 6
Fig.(6-5a,b) Effect of an electric field on the isotropic part
of the electron distribution function( ) in
the case of argon2:
p = latm., n = 1.3 1018nT3,T = T = 5000K.
M: zeroth order Maxwellian without isotropic correction.
-107-
Fig.(6-6) Effect of a temperature difference on the isotropic
part of the electron distribution function in case of
argon2: p = 1 atm., n = 3.4 1017nT3,
T = 4500K, T = 1000K.e ' a
M: zeroth order Maxwellian without isotropic correction.
-108-
VI-2 Electron transport coefficients
In this section the electron transport coefficients in
weakly ionized gases (WIG) are calculated from the expressions
(4-132) to (4-138). Some results of the strongly ionized gas
(SIG) of chapter V are also given. The first order parts
contain the coefficients S0,R0 and LQ which are functions of
the electron temperature except for the case of the HSM. If the
electron-atom interaction potential is assumed to vary with
some power of the interaction distance, the collision cross
section is proportional to a power of the relative velocity. If
this model is adopted one has:
T ( 1 ) ( » ) = V 'n , (6-2)
so that:
-, . -(nri-l)/2 T - 1 . (6-3)Q(w) = q W v ' , q = —pr , T = V '
m T (2v ) m / 2 n /2 v„ Qnm Te a Te O
The coefficients mentioned above are then easily calculated
giving the following results:
The hard spheres (HSM) corresponds to m=-l with q-^1.
The following coefficients can be calculated exactly for the
interaction model (6-2):
bee qra m/2,m/2 ' ee qm l m/2,ra/2+l
Lee " qm ^m/2+l,m/2+l 5Xm/2+l,ra/2+ 4 -1' ( 6 5 )
where the coefficients X are defined in (D-41) of appendix
D3. A tedious but straightforward calculation gives for the
HSM:
S = - [-Ü- + TT - T-lnC1-̂ ) 1*̂ » -0.2276e e 30^2 1 5 4
-109-
Re e
= " \r^l= -%~ r-lnd+^J/ÏT = 0.6436e e 60/2 iU ö
[ gln(l+/2)]v^ -1.8775- (6-6)e e 120/2 b
If more realistic cross sections, which are available in the
form of tables, are used these coefficients have to be calcula-
ted numerically. It turns out that the coefficients in (4-121)
for the multiple collisions are very sensitive to the precise
shape of the Rarasauer minimum of argon. This is demonstrated by
using two different cross sections, one from Milloy2 and one
obtained earlier by Frost and Phelps3. Figure (6-7) shows a
sketch of these cross sections and in figure (6-8) a plot is
given of the function:
SF(w) = ƒ -̂ Z{—~ }dx, (6-7)o Q(x)/x Q(x)/x
which is related to one of the coefficients in (4-121), namely
S = S17(«>). One can then see that the main contribution to the
integral S comes from the Ramsauer minimum. It is clear thatee
the sharper minimum in the cross section data of Milloy et al.
results in a much larger value of See
The electrical conductivity will now be calculated as a
function of the parameter 3» defined as:n , » rT 2
. - pr e inA L , , „.3 - veeT/2 = — - ^ J J - , (6-8)
a u
where r is the Landau length, see chapter II. In the presence
of an electric field and a temperature difference between the
electrons and heavy particles the electrical conductivity in a
uniform WIG, where 0 is of the order e, up to second order
reads:
a(1) + a(2) =
• (6-9)e a
-110-
From this expression one can infer that when g is either very
small or very large, singularities occur originating from the
fact that the ordering has a restricted region of validity.
Comparison is now made with three other calculations of the
electrical conductivity. Firstly the addition mixture rule
introduced by Lin et.al.1*, which is defined as follows:
where a = OQSQ is a result of electron-atorn collisions only
and a of electron-ion collisions only:ei
64/27e2v3
oel := - 2 J 2 . , (6-11)Z1e
2lnA
and Yp *s t n e well-known Spitzer factor: y„ = 0.582, so that
Y„a . is the electrical conductivity of a fully ionizedE ei
plasma5. Mixture rules proposed by Frostfa, use the lowest order
expressions for the transport coefficients in a WIG, but add to
the electron-atom collision frequency a modified electron-ion
collision frequency in order to obtain simple formulae for the
transport coefficients which might be reasonable approximations
for arbitrary degrees of ionization, from the weakly ionized
gas up to the fully ionized plasma. Care has been taken that
the expressions give the correct answer in the fully ionized
limit. In case of the electrical conductivity the Frost mixture
rule reads:
P t w*5 ^e dw
o* = oj £_222 . (6-12)o {w3/3Q(w)+0.952s}
The third way of calculating the electrical conductivity is
based on the equations of the SIG, see section 4 of chapter V.
For the case of a HSM the convergence is good, especially when
powers of half an odd integer are admitted as co-ordinate
functions. In table (6-3) two sets of co-ordinate functions are
compared with each other. One consists of the classical Sonine-
-111-
or Laguerre polynomials and the other one is a set of orthogo-
nal functions constructed by means of Gram Schmidt's method
from the following functions:
3/21, w, w
The function
(6-13)
is not permitted because it results in
infinitly large matrix elements 6 . In table (6-3) values of
-(A,J5A) appearing in equation (5-72) are tabulated for neon at
an electron temperature of 5000K, with an increasing number of
co-ordinate functions up to eight.
Number of
functions
2
3
4
5
6
7
8
Polynomials
1.3393
1.3842
1.4128
1.4264
1.4325
1.4354
1.4370
Functions
in (6-13)
1.3393
1.4069
1.4370
1.4401
1.4402
1.4402
1.4402
Table (6-3), values of -(A,J5A) for the HSM with 0=1.
When 3 * 0 all calculations of the electrical conductivity
except (6-9) converge to the first order part of expression
(6-9), because none of them takes any deviation from a
Maxwellian electron distribution into account.
Figure (6-9) gives the results of the calculations for the
HSM; i.e. when Q(w) = 1. To obtain clear pictures the conducti-
vity is normalized to QQSQ for low values of 3 and to the
Spltzer conductivity y„o , for the higher values. The relation
between these normalizations is the following:
qosoVel
3» (6-14)
-112-
One can see in figure (6-9a) that the electric field suppresses
the electrical conductivity below the common limit of the other
calculations. Figure (6-9b) shows the strongly ionized domain
where the addition rule gives much higher and the Frost mixture
rule gives lower values for the electrical conductivity than
the SIG calculations. Figure (6-10) shows similar results for
neon.
Calculations of the thermal heat conductivity are given in the
next two figures. Figure (6-11) shows the results for the HSM
and (6-12) for neon. As can be seen from these figures, the
Frost mixture rule gives rather good results in the SIG domain.
When gradients are weak the expression for the thermal heat
conductivity up to second order reads:
where <k = k26 - kTk25 - %k2k.
For large 3 only the Frost mixture rule and the SIG results are
shown in figures (6-11) and (6-12), normalized to the Spitzer
value. For small values of g the normalization is done with
respect to (L - k R0)XQ, i.e the first order contribution.
When T /T = 0.9 and T = 5000K, the thermal heat conductivitya e e
at lower values is higher than the results of the Frost mixture
rule and The SIG in the case of neon, see fig (6-12). At higher
electron temperatures the effect changes sign because < does,
see fig (6-4).
The cross section of argon leads to many difficulties, because
of the large values of the occurring coefficients. If the
fields and gradients are small enough, reliable results may be
obtained for low degrees of ionization.
-113-
Fig.(6-7) Plot of the data for the electron-argon cross section
for momentum transfer as obtained by Milloy2( )
and by Frost and Phelps3(- - - - ) .
ÜJ
v
Fig.(6-8) Plot of the function S (w), see (6-7) for the cross
section data of Milloy(M) and of Frost and Phelps(FP)
at T = 5000K.e
-114-
1.1
to 10
Fig.(6-9a) Electrical conductivity normalized to the zeroth
order value a S for the HSM.o o
1
• 9
• 8
.7
_2L"6
V».4
.3
.2
.1
0
HSM.....I
10° 10 JO*(3
10'
Fig.(6-9b) Electrical conductivity normalized to the value of
the fully ionized plasma for the HSM.
B = 2 v T , AR: Addition mixture rule, FR: Frost mixture rule,ee
-115-
1.1
1
9 •
•6 I-
•5
•10"
/ NEON• Te*S00OK
" e E = 0.01
WI6\
-..AR
\
-J • x
10 10
o-t 10°
Fig.(6-10a) Electrical conductivity normalized to the zeroth
order value for neon at T = 5000K.e
10 10
Fig.(6-10b) Electrical conductivity normalized to the value of
the fully ionized plasma for neon at T » 5000K.\ e
f$ = 2 v T , AR: Addition mixture rule, FR: Frost mixture rule.
-116-
10 10
Fig.(ó-lla) Thermal heat conductivity normalized to the zeroth
order value for the HSM.
10
Fig.(6-llb) Thermal heat conductivity normalized to the value
of the fully ionized plasma for the HSM.
8 = 2 v T , FR: Frost mixture rule.ee '
-117-
1-2
Fig.(6-12a) Thermal heat conductivity normalized to the zeroth
order value for neon at T = 5000K.e
10
Fig.(6-12b) Thermal heat conductivity normalized to the value
of the fully ionized plasma for neon at T = 5000K.
6 = 2 v T , FR: Frost mixture rule,ee
-118-
VI-3 Electrical conductivity of cesium seeded argon plasma
When for a seeded plasma calculations are compared with
experimental results there is the advantage that for these low
temperature plasmas the experimental conditions are well
defined. On the other hand, however, the low temperature and
the relatively high degree of ionization tend to give high
values of (3 outside the scope of the present theory of the UIG.
It therefore appears that the experimental values are not in
the region where the isotropic correction influences the
transport coefficients significantly. Calculations have been
performed for a cesium seeded argon plasma of atmospheric
pressure. Two different cesium-cross sections have been used
together with the argon cross section of Milloy2: one has been
obtained by Postma7 and the other one by Stefanov8. Figure
(6-13) shows these rather different cross section data.
In figure (6-14) the experimental points are from Harris9,
which contain a possible error of 30%. The results with
Stefanov1s cress section are In better agreement with the
experimental points simply because he used Frost's mixture rule
and the data of Harris to fit a curve for the cross scetion of
cesium. Postma, however, used electron drift-velocity measure-
ments and numerical integrations of the electron Boltzraann
equation to obtain his curve. The results of Postma are not too
far away from the experimental points. The fact that the
experimental results are larger than the theoretical ones for
higher cesium-pressures has also been observed by Kruger et
al.10. It might thus appear that curve fitting of cross
sections by means of experimental data is rather inaccurate, if
possible experimental errors are high. The difference between
the Frost mixture rule and the present work for Stefanov's
cross section is due to the fact that the minimum in his data
lies at the same energy value as the Rarasauer minimum of argon
and thus reinforces the Influence of the latter.
-119-
f * 7 •iitcirr
Fig.(6-13) electron-cesium momentum-
transfer cross section
data as obtained by
Postma7(P) and by Stefanov8(S).
1300 1400 1S00 ISOO 1700 1600TEHP. (KELVIN)
10
0.1 torr Cs
Stttanov
1300 14D0 1S00 I tOO 1700 1100TEn*. (KELVIN)
1300 1400 1500 1600 1700 1800TEflP. (KELVIN)
10
£ 10
10'
10
1 torr Cs
Stefanov
1300 1400 isoo isoo 1700 leoo
Fig.(6-14) Electrical conductivity of cesium seeded argon
plasma as a function of the electron temperature.
e-Ar cross section data of Milloy2
e-Cs cross section data of Postma7 and of Stefanov0.
argon pressure is 1 atm.
-120-
References
1. A.G.Robertson, J.Phys.B 5^(1972)648.
2. H.B.Milloy et al, Austr.J.Phys. 30(1977)61.
3. L.S.Frost and A.V.Phelps, Phys.Rev. 136(1964)A1538.
4. S.C.Lin,E.L.Resler and A.Kantrowitz,
J.Appl.Phys. 26(1955)95.
5. L.Spitzer and R.H'arm, Phys.Rev. 89(1953)997.
6. L.S.Frost, J.Appl.Phys. 32(1961)2029.
7. A.J.Postnia, Physica 43(1969)465.
8. B.Stefanov, Phys.Rev. A22_( 1980)427.
9. L.P.Harris, J.Appl.Phys. 34(1963)2958.
10. CH.Kruger.M.Mitchner and U.Daybelge, AIAA J. £(1968)1712,
-121-
VII SUMMARY AND CONCLUSIONS
The work presented in this thesis shows that a perturba-
tion expansion in the framework of the multiple time scale
formalism is well suited to attack the complicated set of
kinetic equations describing transport phenomena in a partially
ionized gas.
The equations are limited to elastic collision processes only.
The aim of the present work is to describe transport phenomena
at thermal energies and to calculate transport coefficients. \s
partially ionized gases are in general low temperature plasmas
of thermal energies well below the first excitation level the
restriction above is not unrealistic.
In chapter II the basic considerations are given. For the
description of the Coulomb collisions the Landau collision
integral Is used, which assumes a static screening of the
charged particles in the collision process. The other
collisions will be described by the Boltzraann collision
integral, valid for short range interaction potentials and
assuming only binary interactions of the collision partners.
Diverse parameters in the problem are related to the principal
small parameter: the square root of the electron-atom mass
ratio. Among these are the electric field and the Knudsen
number but also the degree of ionization. The latter is used to
make a division of partially ionized gases into four categories
from very weakly ionized to strongly ionized.
In chapter III the necessity of a first order isotropic correc-
tion in a very weakly ionized gas, when only electron-atom
collisions are taken into account, is proved for a situation
with the same ordering of parameters as that of Bernstein. He,
however, incorrectly assumed that a possible correction could
be absorbed into the zeroth order electron distribution
function. When the plasma is homogeneous the equation for the
zeroth order electron distribution function describes the
relaxation towards the Davydov distribution on the timescale
-122-
for energy relaxation between electron» and atoms. This equa-
tion can be completely solved by means of separation of
variables and subsequent solution of an eigenvalue problem.
The case of a very weakly ionized gas, with Coulomb collisions
included, is complicated because of the nonlinearity of the
equation for the zeroth order electron distribution function.
This equation describes the competition between the tendencies
to establish a Davydov or a Maxwell distribution function. All
corrections to this function are functionals of it so that
solving that equation is essential for that region. The
equation has been brought into a rather simple form, which may
lead to possible analytic and-or numerical solutions. An analy-
tic solution for the tail of the electron distribution function
has been obtained.
Also for a weakly ionized gas, as studied in chapter IV, an
isotropic correction has to be introduced. The integro-
differential equation for it is solved analytically. It appears
that for a given electron-atom cross section there are in fact
six different isotropic correction functions. Six moments of
each of these functions are needed to evaluate the corrections
on the transport coefficients. Thus there are 36 coefficients,
which, apart from the hard spheres interaction model, are
functions of the electron temperature. New transport phenomena
are found which depend nonlinearly on the gradients and forces
or involve second order derivatives.
The second order corrections to the transport coefficients can
be devided into two groups: one of them consists of nonlinear
contributions from the isotropic correction, the other
encompasses the effects of multiple collisions. The latter give
linear relations between fluxes and forces and thus obey
Onsagers' relations.
Chapter V deals with the strongly ionized domain. The equation
for the first order non-isotropic part of the electron distri-
bution function has been cast in the form of a fourth order
differential equation.
-123-
The calculation of the coefficients needed for the Galerkin
method happens to be easier in this formulation. Th- Spitzer
equation for the non-isotropic part of the electron distribu-
tion function reduces to a differential equation of even second
order, which provides a more convenient basis for the calcula-
tion of the transport coefficients than the equation of Spitzer
and Harm. When powers of half an odd integer are admitted as
co-ordinate functions the convergence of the Galerkin method
becomes much better.
The results of the numerical calculations in realistic cases
are collected in chapter VI. The domain in which the results of
the weakly ionized gas can be fruitfully applied strongly
depends on the energy dependence of the electron-atom cross
section for momentum transfer. Especially in the ca^e of argon
the results are so poor that the domain almost vanishes. This
is due to the well-known Ramsauer minimum.
Of the mixture rules that are tested, the addition rule gives
too high values for the transport coefficients while the Frost
mixture rule appears to be relatively reliable. The electrical
conductivity is also calculated for a cesium seeded argon
plasma and compared with experimental results. This is done for
two rather different experimentally obtained sets of data for
the cesium cross section, of which those obtained by Postma are
more reliable than those obtained by Stefanov.
At lower cesium pressures the agreement between theory and
experiment is satisfactory, while at higher pressures the
theoretical values are lower than the experimental ones. It
should be remembered that the experimental conditions are at
the border of the range of validity of the present theory for
the weakly ionized gas.
-124-
APPENDIX A
Expansion of the electron-heavy particle collision Integrals
A-l Electron-atom collisions
The Boltzraann collision term describing elasic collisions
between particles of species s and t in a dilute gas is usually
given in the following form:
SC» f ) = / d 3 g / b d b d < f > g { f ( v ' ) f ( v ' ) - f ( v ) f . ( v ) } , ( A - l )
t S C- u S C C
where b denotes the impact parameter and g is the relative
velocity: g = v - v. Primes indicate post collision variables,
which are defined by the relations:
ml
Conservation of momentum is guaranteed by these expressions.
Conservation of energy in the centre of mass reference frame
requires: g'=g- This can also be expressed by the relation:
I2 + 2g.* = 0. (A-3)
It is now possible to show the integrations to be performed
explicitly if the differential cross section o(g,x) *s defined
by:
J4H
bdbdij) = b — dxdifi = - a(g, x)sinxdXdli>> (A-4)
where x is tne scattering angle in the centre of mass system.
Figure A-l shows an encounter in that system. With the help of
relation (A-3) the integration over x a"d <f> is written as an
integration over the complete Jl-space. Expression (A-l) then
reads as follows:
J (f ,f ) = 2/d3g/d3JlI(g,Jl)6(A2 + 2g.*)xst s t - -
Xt f s<v - !TTm-)ft<2 + 8 + TlZr> - f Js t s t
-125-
<A-6)where: I(g,*) = o(g,X) = " ^ ||^| •
The 6-Dirac function takes care of relation (A-3).
The collision Integral J describing the collisions betweenea
electrons and atoms will be expanded in powers of the small
parameter e:
e = (m /m ) .v e aJ(A-7)
If the electron temperature is of the order of the atom
temperature the velocity variables in the thermal range scale
with e. A Taylor expansion then gives the following result:
ea ea
èa
ea
ea(A-8)
(A-8a)
/d3v v f (v Wdfla(v,x)2(n2 - I)-Vf(u) +3. a a. a V G
' j 3 v Z f <*><J"dnVJva<v>x)){f (u) - f (v)}, (A-8b)
^d3v av av af a(v a):/dn{Vv7v(va(v,X))(f e(u) - fg(v)) +
-126-
4V (vo(v,x))(I-nn)«V f (u) + 4va(v,x)(I-nn)(I-nn): V V f (u)} +
2m—-/d 3v f (v )Vm a a ~a va
- nn)-vva(v,x)f (u),— — e —
(A-8c)
In these expressions u Is the electron velocity after a
collision with an atom of infinite mass, so that u=v. The unit
bisector of the angle between u and v is n, see figure A-2.
'Jhat is left from the
integration over A-space
is an integration over
dfl = sinxdxdiji.
The introduction of the
unit vector n permits
the following notation:
u = (2nn-I)«v. (A-9)
A transformation in velocity space to a reference frame moving
with the hydrodynaraical velocity of the atoms makes the first
order term (A-8b) equal to zero. The expression can be simpli-
fied further if it is assumed that f is a Maxwellian. In that
case the third order contribution vanishes also:
Jea
J(0>ea
ea ea
/diJv(v,X)[f (u) - f (v)],
kT
(A-10)
(A-lOa)
7 f (u)v e -
n va(v,x)3.
(A-10b)where: v(v,x)
The collision operator permutes with the rotation operator in
velocity space. Therefore the spherical harmonics or equi-
valeutly the harmonic tensors of appendix C are eigenfunctions.
If f is expanded into harmonic tensors:
-127-
(A-ll)n=0 " c
and if this expansion is substituted into (A-lOa) the n-th term
reads:
nfe(v)«/dfiv(v,x)[<un> - <vn>]. (A-12)
Consider now the right-hand factor of this n-fold dot product.
After having performed the integration over ds} = sinxdxd<|> where
X is the angle between u and v, the result still is a harmonic
tensor of rank n, therefore:
gn ('
where g is as yet undetermined. By taking the n-fold inner
product of this expression with <v > one can show (see app. C)
tha t g Cx»*!") = P (cosx) so t h a t :n n
J ( f (v )*<v>) = f (v) »<v >/dS2v(v,x) JP (COSY) -1 }. (A-14)ea n— n — n— n— l n '
In general with expansion (A-ll) for some function f :
(0) v i riJ (f ) = - T T~ 1 (v) f (v)«<v> (A-15)ea e _. (1) n—e n —
with: T 1
( y ) = /dJMv,x)[l - Pn(cosx)].
Note that the term n=0 is absent: J is zero for an arbitrarye a (0)
isotropic function. In appendix B it is proved that J is
symmetric and negative-definite. Also an H-theorera will be
derived.(2)
The expression for J becomes also very simple if f isisotropic:
If f is a Maxwellian too, this expression becomes:e
Which shows that this term vanishes in thermal equilibrium.
-128-
A-2 Electron-ion collisions
The interaction between charged particles will be
described by the Landau collision integral; see also chapter II
and appendix D. The general form reads as follows:
2 1 —/f » f
o ) = C OV • / — . { — 7 - — V , }f (v)f o
a ' ?>' a6 v J ? lm v mn v' ' a - gg a p
— 7 - — V , }f ( v ) f o ( v ' ) d 3 v f
(A-17)where: g = v 1 - v and C i s a cons tan t defined by:
- - - ap
q l
SLÊ (A-13)
o a
The case in which the particles have equal masses will be
treated in appendix D. The electron-ion collision integral can
easily be expanded in powers of e by means of Taylor
expansions; the results up to second order are:
(0) l , ,v *•= v e;'ei m v *•= v e
e
j _ei ra
e
- _£ijVf (V>)d3vt.v .[(V V ) . v f 1 (A-20)ra - I - v L v= v eJ
C n 2 v CU > = _ e i 1 7 r _ ^ f > + e i_ j . , , f ( v . ) d 3 v » : 7 . [ ( v v V ) - V f 1e i m v ^ 3 eJ 2ra - - i - v •• v v = v e J
(A-21)where the tensor V is defined as follows:
v 2 l - vvV := = 7 7 v. (A-22)= 2 v v
Some properties of V, which are easily verified are:
V«v - 0, (A-23a)
2vV .v = - -^-, (A-23b)v " V3
v-7 V = -V, (A-23c)
-129-
v ( V V V) = -2v V , (A-23d)- v v= v=
A V = 7 •? V = — c <w>, (A-23e)
v= v v= vD — '
= f-(v2 |-) + v37 «(V-V ). (A-23f)
With these properties It can be shown that:
co C n.
<O> ei i
(f ) I n(n+l) ei i f ( v ) ' ^ 1 ^ , (A-24)
ei e L, a n-e n -n=l m v*
e
and if f is isotropic:C 2v
m
i_I2|_[I|l). (A.26)e
The second term on the right-hand side of (A-26) clearly
vanishes if f is also isotropic. If the electron and ion
distribution functions are Maxwellians with hydrodynamical
velocity equal to w and with different temperatures one finds
that the first contribution to J is given by (A-26) and
reads:
e e
See also equation (A-16a) which is of the same type.
-130-
A-3 Moments of the electron-heavy particle collision Integral
All collision integrals considered are elastic and thus
preserve the number of particles:
/J (f ,f )d3v = 0, s,t = i,e,a. (A-28)st s t
In general the moment with some function <j> of v reads:
/Jst(fs,ft)<t.(v)d3v. (A-29)
If J is the Boltzmann collision integral of (A-5) and thes c
following transformations:m %
I + -I, g * g + *, Y. + Y. " ~-' (A-30)o
are applied in this order one obtains for the moment (A-29):
/d3V(j)(v)J (f ,f ) = - 2/d32d3vd3g6(£2 + 2g-A)Kg, £)*- st s t - -
m i" [ • (v - r ^ 1 ) ~ • (v)J f ( v ) f . ( v + g ) , (A-31)
o
which indeed is zero if $=1. If $=m v equation (A-31) becomes:
2m m2/d3vm vJ = 2_£/,i3va3gd3u6(Jl2 + 2g-l)I(g,l)f (v)f (v + g)0
(A-32)Firstly the integration over £-space is performed in spherical
co-ordinates with g as polar axis:
fdH A6(A2 + 2*.g)I(g,A) = - -— T-Y , (A-33)n T ( r 8 J
where T-1. is defined as in (A-15). Thus:
( A" 3 4 )
If s=e,t=a this term can be expanded in powers of e. It is then
convenient to perform an integration over v'= v + g instead of
one over g. The# result of the expansion for this case is:
- 1 3 1 -
m yf (v) m m v kT/d3vm vJ = - J 6 e , \ d3v + e 2 - i r f e " _ a v A (—£___)+
e— ea J T . 1 . ( V ) m T , . . ( V ) 2 — v k i , . , ( v ) J
- kT V [ ^ -T- l l f (v)d3v +&(t1*). (A-35)a V t ( l ) W e ~
Next iji(v) = 5jm v2 i s inserted into (A-31):
mm m X.2
Am v2J d3v = - 2 - VS SC Ïs st m o v m - -' i
o tx f (v)f (v + g). (A-36)
S ™ C "™ *•*
With the aid of (A-33) and s similar integral:
2
t (1)
the moment integral with ̂ m v2 can be written as follows:
smm m g 2 f (v)f (v + g)
Am v2J d3v = -=-=vfd3vd3g(—=—+ g - v } - ^ — " . . — . (A-38)s st mQ m o -T n tT (1 )(g)
If again s=e and t=a the following expansion in powers of c is
obtained by means of Taylor expansions of the integrand:
ra 3kT - ra v 2
Am v 2 J d 3 v = e 2 —ƒ f r—r 1- kT v-r—f ) If ( v ) d 3 v + Oiz*)s ea m J L T ^ . - C V ) a 3 V V T . 1 ^ ; J e -
(A-39)
Finally the results are given of the expansions of the moments
of the electron-ion collision integrals:
vfm vJ .d 3 v = - 2C , n . f — f (v )d 3 v +
e- a i e i xJ 3 e -
3<v2> 2m v- e2C . / v ' f . ( v ' ) d 3 v ' « f f ( v ) d 3 v - e2—^C .n,J—f ( v ) d 3 v +
ei - i - v 5 e - m1 ei i Jv 3 e -
15<v3>- e2 /y 'v 'f i(v l)d3v< : / —fg(v)d3v + #(e3) . (A-40)
vAmev
2J fd3v = -e2Cel/v
1f1(v')d3v t ' f—i (v)d3v +
-132-
2C m <v2>_ e 2 — Ê l ^ / i f (V)d3v -e26C . / v 1 v ' f , ( v l ) d 3 v 1 : f ~ — f (v)d3vm. J v e - e l 1 - J c e -1 v 3
Some caution must be taken with integrals like:
<V2>ƒ——f(v)d3v. (A-42)v5
At first glance this integral would be zero If f is Isotropic.
But this is only true if f(0)=0. The result of the integration
must be an isotropic tensor of rank two (see also app.C):
<V2>ƒ f(v)d3v = -j/VvVv(i)f(v)d
3v = A(v)I. (A-43)v5
Contraction on both sides gives:
3A = j/Av(^)f(v)d3v = - YL/5(v)f(v)d3v = - *ff!;(0). (A-44)
so that: ƒ——f(v)d;iv = - ^v I. (A-45)
A generalization of (A-43) is:
JV^[i)f(v)d3v = An(v)nI, (A-46)
where I is an isotropic tensor of rank n. For odd n there isn-
no such tensor so that these integrals are zero. This follows
also from the fact that the Integrand is an odd function of v.If n is even A can be determined by a contraction over all
n
indices and a subsequent calculation of the Integral as
follows:
|A"/2(|)f(v)d3v = - 41r/(AjJ/2"1ó(v)]f(v)d3v. (A-47)
If E can be differentiated a sufficient number of times this
integral can be calculated.
-133-
APPENDIX B
Some H-theorems and properties of collision integrals
B-l The zeroth order electron-atom collision operator
The zeroth order electron atom collision operator J has
ea
been derived in appendix A. In this section two properties and
an H-theorem will be derived. With tne aid of the 6-Dirac
function the collision integral .7 can be written as follows:J(0)(f) = n /d3£6(Jl2 + 2A.y)I(v,je){f(y - 1) - f(y)}. (B-l)
By multiplication with some function g and integration over the
entire velocity space the following inner product is obtained:
f),g) = n /d3vd3£6U2 + 2y««)I(v,JO{f(y-£) - f(y)}g(y).
(B-2)If the velocity transformations v •*• v - X. and I •*• -I are
performed in this order, (B-2) is found to be equal to:
(0)*• e a a — — '
x{f(v-A) - f(v)}jg(v) - g(y-A)} = (J^^g)^], (B-3)
so that J is a symmetric operator. From this expression oneGel / r\\
finds immediately that J is negative-definite:
(j^(f),f) < 0. (B-4)
Finally an H-theorera can easily be proved as follows. Suppose
that the following equation holds for f:
| | = JgaJ(f)- (B-5)
The quantity H is now defined as:
H = /fln(f)d3v. (B-6)
If equation (B-5) is multiplied by: 1+ ln(f),and integrated
over v-space, tha following inequality can be obtained:
-134-
3H tta f ( V >
= ^Jd3v43lS(l9 i 2 v « j ) I ( v A ) l n ( 3 ) { f ( v - A ) - f ( v ) J < O,•— = ^ • J d v 4 l S ( l i 2 ( )
(3-7)
because f is assumed to be positive everywhere. At the same
time H is bounded from below, so that when t •* °° the integral
in (B-7) is equal to zero and f has become an isotropic
function, i.e. depending only on |v|.
B-2 The zeroth order electron-ion collision integral
The above derivations suggest that the properties of J(0) e a
also hold for J , . They will be proven below. One has:
^ jpV^V v ve e (B-8)
which is symmetric because V is symmetric. If g=f the
integrand in the right-hand side of (B-8) takes the form:a2v2 -' (a«v)2
a-V«a = — — ? 0, (B-9)
v3
which shows that J , is a non-positive operator. Finally an H-
theorem is derived. Again H is the quantity defined in (B-6).
If in (B-8) g = 1+ ln(f) and f obeys:
it is easily demonstrated that:
-!•£ = - 1 el/i-V f«V«V fd3v < 0. (B-ll)3t m Jf v = v3t m Jf v
e
-135-
B-3 H-theorems for the ion distribution function
In this section an H-theorem will be derived for the
zeroth order ion distribution function in two different cases.
The first is that of a weakly ionized gas, the starting point
is then equation (4-20). A function $ is defined by:
f.3/2 v-w
(0) , 2
i.as i i.as ,„,,T(u; 2kTtU;Y1 I i ~ -a.asl 1 , x c5 exp{~ ~TI(ö) 1 " *i(ï)fi
a,as a,as (B-12)
With the results from the moment equations it can be shown that
-g7±M = 0. (B-13)
Equation (4-20) is now multiplied by <j>.(v) and integrated:
3*i (0)f%Y~ fiM d 3 v = 2/d3vd3£d3g6(je2 + 2g«l)I(g, SL)ta as(v+g)f 1M(v)x
m I
*• i - m i - ' i - 'o
where m = m + tn • Application of the following transforma-0 3. X
m Itions in the given order: I * -I, v + v — , g + g + I is
oequivalent to interchanging direct and inverse collisions. Withthis transformation equation (B-14) can be written as follows:
3Hi 3 3 3 (0)3Tl v g g _ g, 1 M v a > a s _ g x
m I 2
i — i *~ IDo
where: H = /f <f>2d3v. (B-16)
The conclusion is then that <t> =1 when Tj+ «, for H is non-
negative and decreases with time, i.e.:
m |v-w<0)|2
fiA} = 4 A ) ^ T I 7 Ö T ) exp(" i," (!) }- (B"17)aA
-136-
The second case to be considered is the relaxation of the
zeroth order ion-distribution function in a strongly ionized
gas. This time two equations are needed: equations (5-13) and
(5-14) of chapter V. When T Q + » they read:
a.as = (0) (1) + (1) (0) , + j (f(0) (0) ,3Tj aa a,as a,as aa a,as a,as ai a,as i,as
(B-1S)
8f ( 0 )
dti,as _ (0) r(0) . . . ,.(0) -(0) . f Q.3TJ ii i,as i,as ia I,as a,as
Again f' ' is a Maxwellian. If equation (B-18) is multiplieda y as
by (l+ln(f )) and integrated over the entire velocity spacea y as
the following equation is obtained:
->>™ - /(l+ln(f<°> ))J (f<°> fi0) )d3v1 • > *•>
where: H ( 0 ) = If. ln(f(0) )d3v. (B-21)a,as a,as a,as
The terms containing J vanish because (l+ln(f )) consistsaa a yas
of mere collision invariants. Equation (B-19) is multiplied by(l+ln(f. )) and integrated too:
x f as
»M. • "\"
Then the following inequality can be proved1:
H4,l<V«°- (B-24>(0)
from which the conclusion can be drawn that when T,+ «, f,1 ' i,as
relaxes to a Maxwellian with a temperature and a hydrodynaraic
velocity equal to those of the atoms as in the case of a WIG.
-137-
APPENDIX C
Harmonic tensors
The harmonic tensors that are used throughout this thesis are
completely equivalent to the familiar spherical harmonics, as
has been demonstrated by Johnston2. The harmonic tensor of rank
n is defined as follows:
n 2n+l
^ ^ := (2n-l)! ! VvM" (C-l)
It is an irreducible tensor, i.e. it is symmetric and a
contraction over any two indices makes the tensor equal to
zero, because v~1 is a solution of the equation of Laplace. The
harmonic tensor <y > can be seen as the irreducible part of rhe
tensor v := vvv...v (n vectors). The first few harmonic
tensors written in index notation are:
<v°> = 1, <vl>£ = v±, <v2>i:j = v l V j - f 6±y
<V3>_ _ = v v v - — (v ó.,+ v.6 + v 6 . ) • (C-2)
Any tensor can be made irreducible in a unique way, see Grad3.
One can also prove a kind of orthogonality relation, see
Wilhelm and Winkler4, which reads as follows:
2n,< h(v)>, (C-3)
where h is an arbitrary tensor of rank n and < h> is then- J n-
irreducible part of h. The following expansion is very
n-2 I n(n-l)(n-2)(n-3) i+r-r? n-41
1 + 8(2n-l)(2n-3) v të2Y ] " • • •(C-A)
where the square brackets denote the symmetric part, obtainable
by adding all the permutations and deviding the result by n!.
The inner product of v and <v > will again be an irreducible
tensor but now of rank n-1, and will thus be proportional to
<v >. This tensor also has an expansion as in (C-4).
-138-
The inner product v«<v > is thus equal to some factor times
<v >, which will appear in the right-hand side of (C-4) after
having performed the inner product with v. The tensor v«[ly ]
possesses 2(n-l)! permutations equal to v , therefore:
n 2 n-1 n(n-l) 2(n-l)! 2 n-1 ^ nv2 n-1
v.<v> = V2v - 5 ^ — — V2v + ...=—< v >
(C-5)
This result can easily be generalized to:
k n n(n-l) (n-k+1) 2k n-k- &• (2n-l)(2n-3) (2n-2k+l) v <v- >• n>k' (C"6)
With the aid of the definition (C-l) it can be shown that:
n 2n+l, n n+1 , _vVv<v > = (v<v > - <v >), (C-7)
v2
From which immediately follows with (C-5):(C-8)
If again h(v) is an irreducible tensor the following relation
holds:
v ^ V h = vn+1« h. (C-9)- - n n - - n n -With the expansion (C-4) for <v > this equation becomes:
n , n+1 , n(n+l)v2 r_ n-1 i , , ,.,v<v >• h = <v >• h + •,; ' n llv • h. (C-10)- - n n- - n n- 2(2n+l) l=- Jn n-
The fact that h is irreducible reduces this result to:n-
v < vn > . h = < v
n + 1>. h + T 2 ^ < vn ~ 1 > . 1 h. (C-ll)
- - n n- - n n- 2n+l - n-1 n- '
This relation has been employed to derive the following useful
result:
S ) = <ï >n v ̂ + ^ >n-l "2ÏT 17^v
V
-139-
The relation between the harmonic tensors and Legendre
polynomials can be Inferred from the following formula:
. , n n+1 _n .
P (Cose) = lzil_ï !_ (I), (C-13)n n! 3 n V
J 'z
where v 2 = v 2 + v 2 + v2 and v = vcos8.x y z z
Comparing (C-l) and (C-14) one obtains:
n
Let u be a vector in the direction of the z-axls; the following
result is then readily obtained:
gfr Pn(cos6) .Finally a projection operator P is defined by:
n W / V ^4irv n! S2
whih then permits the following notation for the orthogonality
property (C-3):
-140-
APPENDIX D
The Landau collision integral for identical particles
D-l The Landau collision integral
The formulation of a kinetic equation for a plasma is more
difficult than for an ordinary gas, because of the long range
of the Coulomb potential. When the number of particles in a
Debye sphere is large enough, one may use a cut-off potential.
See also chapter II. The Landau-" collision integral reads:
J „(f ,fj = C J .fG-f- Ï - - ? . ] f (v)f,(v')dV, (D-l)a8 a S aS v J= ^m v m v' ' a 8
a 3
q2q2lnA g2l - ggwhere: C „ = -2-2 , c = , g = v ' - v. (D-2)
af5 Sire2» g 3 " "o a
The following properties of G will be often used:
G«v = G»v' , G = V V g. (D-3)
If the distribution functions in (D-l) are isotropic the Landau
collision integral reduces to:
C °° 3f m 3f
a6 m v ^ ̂ v v v —^v 8 3v m v' a 3v' 'a o 3
The integration over ft , is done in spherical co-ordinates with
v along the polar axis (vv'= vv'cos6). Then:
2-rr ir j_
v~ ~r v ^ ~ivv cua a "
O O
3
, = ƒ / {v 2 + v ' 2 -2vvlcos8}2sinGd9d<)) =
Two different cases have to be distinguished:
if V < v :
if V>v : |i(^i^)
- 1 4 1 -
Straightforward differenti.itton of these expressions yields:
V 7 /gdtf . = 4IT{- iv2> + V}, if v'<v,v v ö v' ' c - ='
vD
= §J,I , if v'>v. (0-7)
Insertion of these results into (D-4) then leads to:
2C ra 9f
\z = "mf 2 I JThe functionals I and J are defined as:
P P
I ( f) _ * I ƒ xpf2f(x)dx, J = ̂ ƒ xpf2f(x)dx. (D-9)
P v p 0 p v p v
In the case of identical particles the expression (D-8) can be
written in an elegant way. If f3=oc (D-8) becomes:
2C . . 31n(f )
Two of the integrals are evaluated as follows:
In(f ) = |4itx2f dx - /4wx2f dx = n - 4ir/f x2dx, (D-lla)0 v v
00 00 3n kT »
I 2( f ) = ±2L fx»f dx - li Jx4f dx = _ 2 _ « _ *i jx«»f dx. (a v2 0 a v 2 v a m v 2 v 2 v a
aThe expression in (D-10) between braces is then equal to:
kT 91n(f ) « 31n(f )( 2 a ) f { 2 I « 3 3 }) 4Trf{x2 + I _«_(v3-X3)}f (x)dx.
^ V 3V v 3V 9V a (The second terra in (D-12) can be written in a more symmetric
form by the observation that:
3 « 3 3f » 81n(f )
-/x2fa(x)dx = f fa(v) + ƒ |-_«dx = i /(x3-v3)fa(x)^r2^x,
(D-13)
so that the collision integral when operating on isotropic
functions can be written as follows:
-142-
2C kT 31n(f )
= ——2 "5~[f (v) ln (1+ —~ 5—~) +
a a
, °° , 31n(f ) , 31n(ftiff r j 1 * 1 ré ( i-
+ — J(xJ - \
One can now see that this collision integral is zero if f is a
Maxwellian, i.e. if:
31n(f ) m m 3/2 m v 2
a ^ 2 ^ Ha a a
The collision Integral in (D-14) is still nonlinear. In the
remainder of this appendix the linearized Landau collision
integral will be investigated.
-143-
D-2 The linearized Landau collision operator for like particles
The linearized Landau collision integral for collisions
between identical particles is defined as follows:
Ja« ( f ) = Jaa ( f' faM ) + Jaa ( faM' f )' ( D" 1 6 )
where f w is the Maxwell distribution, see (D-15). If f isan
isotropic too, expression (D-10) may be used. The linearized
collision operator then reads:2C kT
|_[i (f|[i (f ) ( 1 +2 9v ° aM m v 3v
m ? aa
This operator appears in the equation for the isotropic
correction in chapter IV section 4. A generalization of (D-17)
obviously reads:
J <aa n- 'n -
C) - - 2 » ?
m va
v v'
xff(v) f(v')«<vtn> - fM(vf) f(v)«<vn>}d3v'. (D-
1 M n— n — M n— n — '
With the properties of the harmonic tensors, see appendix C,
the first part of the integrand is found to be:
Concerning the angle integration in (D-18) the following
integrals have to be calculated:
v v v v . (D-20)
The integral in the right-hand side will still be a harmonic
-144-
tensor, built up by the vector y, so that the following Ansatz
is made:
/g<v'n>dfiv, = H (v,v')<vn>. (D-21)
If on both sides of this equation the n-folded inner product
with <y > is formed the following expression for H is
obtained:
H (v,v') = (— )n/gP (cos9)dii ,, (D-22)
where 8 is the angle between v and v'. The result for n=0 was
already obtained in (D-6). After elementary integrations the
result for n=l reads as follows:
Hj = — /gcos9 dU , = -TT —(v'2- 5v 2), if v'< v,V3
= *1 C2-^'2) , if v>> v. (D-23)
For the evaluation of in the case n=l the following expressions
are useful:
Si = ( V (Hx(v,v')vj.v = ii[v'2(5 -2lll)v +^-l], if V< v,
v v U . v 2 l. if
2o - I7.I . lf vl> v« <
where QQ has already been derived in (D-7). With these
definitions the linearized operator in (D-18) for n=l reads:
C » in 3f(v')=-fVv-/f-2l'{^f(v') +i-^p—jf^Cv) +
a o a
3f(v)
i- f (v)f(v')}]v'2dv' (D-26)
oM
-145-
After some tedious but straightforward calculations this
expression is brought into the following form:
C ? m 2 2m
a a
j m 2 •
+ ' " M C T ^ ~ 3vkT J * o t f J ~ l v " ' m v 'aM 3va a a
kT kT „ ,_i_ I T / c \ I ( Ct ! J.
m v^ m v°a a
where the symbol ,J is introduced, see also equation (4-65)1 aa
Next a change of variables is made:
m v 2 kT
a a
so that for example:
111 3 /2
2kT 32f
m va
I (f) = 4irv'2(—2.) xT*"* /xC f r t 'L ; / Z f (x)dx. (D-29)
The operator ,J can then be written as follows:1 aa
f 3/2"W^ (D"30)n C kT
w h e r e : v = -2L«2_ , v | - - ^ , (D-31)2mavTa a
-146-
and the linear integro-differential operator is equal to:
2w3/2f(w) +|-(2wF(w)||]. (D-32)
One can show that JC becomes a pure differential operator by
differentiating once. After some manipulations it can be cast
into the following form:
-w 3 <„,i 32 r_ „, . -2w 32g l 3 r._, . -2w 3e ^f 3 " [2wF(w)e f J [4F(w)e ^= el 3J - 3w-
where: f = |^{e"Wg}. (D-33)
The function F(w) appearing in these expressions was defined in
(D-29) already. An important property is the following:
w , _F(w)e~W = hjx e Xdx. (D-34)
0
F satisfies the following differential problem:
| £ - F = W^, F(0) = 0, (D-35)
from which the following power series valid near w=0 is
obtained:
/ % V k+3/2 ak 1 ,„ _ , .F(w) = 2, a w , a =• , a 0 = -r . (D-36)
k=0 k K + i k + j J
An as3onptotic expansion for w + °° reads:
^/ N i/ /• x L % i v (i)r(k+^) ,n „ s
F(x) ~ j/iexp(x) - W I '|i ̂ y. (D-37)4/iT k=0 x
The special interest in the function F stems from its frequent
occurrence In the operators derived from the Landau collision
Integral.
-147-
D-3 Matrix elements for the operators obtained from the Landau
collision integral
In the chapters V and VI the Galerkin method is used to
calculate the non-isotropic parts of the electron distribution
function in a strongly ionized gas. Several integrals that are
used in this method will be evaluated in this appendix. First
the operator is treated, see (D-32). The matrix element (p,q)
for this operator is defined by:
00
X = fwPe"W/(wqe"W)dw. (D-38)pq J
Q
Straightforward substitution of X into this expression leads to
the following integrals to be calculated:
CO
T = rwme~WY(n,w)dw, m > - l , (D-39)on
o
where Y(m,n) denotes the incomplete gamma function:
w 1
Y(m,n) = Je~Xx dx, n > 0. (D-40)
The matrix elements X turn out to be:pq
ffff5/2r*- f(p+q+3/2)(p+q+5/2) }.
The function I(k) is defined by:CO
I(k) := JwkF(w)e"2wdw = h\ 3 / 2 > k > -5 /2 . (D-42)
It is of special interest because it appears also in the matrix
elements of the differential operators. The following
recurrence relations facilitate the computation of the
coefficients T :mn
-148-
T = mT . + r ( " * n ) T , + T . = r ( m ) r ( n ) . (D-43)mn m - l , n „m+n * m - l , n n- l , ra v ' v '
From these the following expressions are directly obtained:
T 0 , n = F ( n > 2 n • T mnrH = * r 2 ( - « ) , T ^ = 2 n 1 ( n + 2 ) r ( n ) ,
T = (1 -2~n~V(n+l), T, = ^ r ( n ) - 2I(n-l),ni $n
(D-44)%,n
and so on. For I(k) one easily obtains:
= kl(k-l) + r ( ^ 5 ^ } • k >
K O ) = i(|) % , I(%) = Y^- , I(-l) = iir^{ln(l+2%) - 2"^}. (D-45)
The matrix elements for the operator £ are then simply (see
also (5-79)):
\ = /wPe"Wi'(wqe~W)dw = X - /v(w)wpfqe"2wdw =pq 0 pq J
o
' = Xpq " **V(lrf-q) " Jp /wlH'q+2q(w)dW (D-46)
(see (4-106) where Q and (3 were introduced).
The matrix elements for the operator Jb are defined by:
00
6 = JwP.Z>(wq)dw. (D-47)P q o
With the definition (D-33) these are calculated as:
6pq - 2pq{((p+l)(q+l)-4)l(p+q-3) + H ^ " ^ ] , (D-48)
which clearly is a simpler expression than (D-41). The
analogous expression with the operator 3) reads:
00
pq ' i pq
-149-
+ JL- J(w-p)(w-q)wfH"qQ(w)e~Wdw, p,q > 0 . (D-49)
O
In equations (5-83) and (5-88) of chapter V it was already
demonstrated that these coefficients are symmetric for integer
values of p and q with one exception:
\ \ °°3 0 1 = -Si- + % = i-(? + I /(w-l)w
2e"WQ(w)dwJ. (D-50)4 4 P o
Finally the matrix elements for the Spitzer operator A are
given by:
CO
S r P4S Q
pq ^
{2pq+4(p+q)}l(p+q-l) + ^ ^ L + -^rdri-qfl). (!
This section is closed with the remark that in the expressions
for the matrix elements given above p and q need not be
restricted to integer values, but can take arbitrary values as
long as the integrals converge. This is an extension of the
method described in the litterature . The recurrence relations
for T and I(k) permit easy calculations of all coefficients.mn
-150-
APPENDIX E
Renormallzation of the ion multiple collision term
In section 2 of chapter IV the corrections to the electron
distribution function were obtained up to second order of e.
The ion multiple collision term in the second order
contribution is (cf. equation (4-64a)):
/o\ 2C .n} T,,x(c) .,.(2) _ _ ei lras (1)^ (1)
-e.as c 3 -e.ase
It is the solution of the following equation:
J(0)(f(2) ) = - J(?>(f(1) ), (E-2)ea e,as ei e,as
if an arbitrary isotropic function satisfying the homogeneous
equation is momentarily not taken into account. In higher order
the following equations will appear:
AV O - - 40)<C» • -»2-Only solutions proportional to c are relevant, cf. (E-l), so
that the solution in order n reads:
f ( n ) = _ 2 n i ? a s C e i T ( f ( c ) > {
e,as m c3 -e.ase
All of these contributions have a singularity at c=0, namely:
which relation is valid if f is a Maxwellian and T,,.(c)
goes to some constant value if c + 0. Thus infinitely large
contributions to the transport coefficients are obtained as a
result of a nonuniformity of the expansion in powers of e. This
divergence is removed by summation:
-151-
f(2)norm = y f(n) , _ 2Ceini,asT(l)(c) f(l)
-e.as ^ -e.as 3 + (0) -e.as'e ei i,as (l)v
( g_ 6 )
This expression gives convergent contributions to the transport
coefficients and will be used instead of (E-l).
References to the appendices
1. S.Chapman and T.G.Cowling:"The mathematical theory of non-
uniform gases", Cambridge University Press, 1970.
2. T.W.Johnston, J.Math.Phys. 7^1966)1453.
3. H.Grad, Phys.Fluids 4(1961)696.
4. J.Wilhelm and R.Winkler, Beitr.Plasmaphysik 8(1968)167.
5. L.D.Landau, Phys.Zeits.der Sowjetunion 10(1936)154.
6. M.Mitchner and C.H.Kruger:"Partially ionized gases",
J.Wiley, 1973.
Samenvatting
Het werk dat in dit proefschrift wordt gepresenteerd toont aan
dat een perturbatie ontwikkeling in het kader van het meertijd-
schalen formalisme zeer geschikt is om het gecompliceerde
stelsel vergelijkingen aan te pakken welke de transport
verschijnselen in een gedeeltelijk geïoniseerd gas
beschrijven.
De vergelijkingen beschrijven alleen elastische botsingsproces-
sen. Het doel van het huidige werk is de beschrijving van
transportverschijnselen bij energieën van thermisch niveau en
de berekening van transportcoëfficiënten. Aangezien gedeel-
telijk geïoniseerde gassen in het algemeen plasma's van lage
temperatuur zijn met thermische energieën veel lager dan het
eerste excitatieniveau is de bovengenoemde beperking niet
onrealistisch.
De basisvergelijkingen worden in hoofdstuk II gegeven. Voor de
beschrijving van de Coulorab-botsingen wordt de Landau botsings-
integraal toegepast, terwijl de andere botsingen door de
Boltzmann botsingsintegraal worden beschreven.
Diverse parameters die voorkomen, zoals het electrisch veld en
het Knudsen-getal, worden gerelateerd aan de belangrijkste
kleine parameter in het probleem: de wortel uit de electron-
atoom raassaverhouding. Ook de ionisatiegraad wordt aldus
ingeschaald en dit geeft aanleiding tot de indeling van het
gedeeltelijk geïoniseerde gas in vier gebieden van zeer zwak
tot sterk geïoniseerd.
In hoofdstuk III wordt het zeer zwak geïoniseerde gas
behandeld. Wanneer de Coulomb botsingen volledig verwaarloosd
worden beschrijft de vergelijking voor de nulde orde electronen
verdelingsfunctie de relaxatie naar een Davydov-verdeling. Dit
proces vindt plaats op de tijdschaal voor energierelaxatie
tussen electronen en atomen. De noodzaak van een isotrope
korrektie op deze verdeling wordt ook aangetoond.
Wanneer Coulomb botsingen worden meegenomen in het zeer zwak
-153-
geloniseerde gas beschrijft bovengenoemde vergelijking de
competitie tussen de tendensen naar een Davydov- en een Maxwell
verdeling. Deze vergelijking is nu niet lineair ten gevolge van
de electron-electron botsingsintegraal.
Het zwak geïoniseerde gas wordt in hoofdstuk IV behandeld. Ook
hier is een isotrope korrektie op de nulde orde electronen
verdelingsfuctie noodzakelijk. De integro-differentiaal-
vergelijking voor deze functie wordt analytisch opgelost. Het
blijkt dat er voor een gegeven electron-atoora botsingsdoorsnede
in feite zes verschillende isotrope korrektie functies zijn.
Ook verschijnen er nieuwe transport verschijnselen welke op
niet lineaire wijze afhangen van de gradiënten en krachten. De
symmetrierelaties van Onsager zijn hiervoor niet meer geldig.
In hoofstuk V wordt het sterk geïoniseerde gebied behandeld. De
integro-differentiaal-vergelijking voor het niet-isotrope deel
van de electronen verdelinges functie is in de vorm van een
vierde orde gewone DV geschreven. In de limiet van een volledig
geïoniseerd gas gaat deze zelfs over in een tweede orde DV. Dit
betekent een nuttige aanvulling op de theorie van Spitzer en
geeft eenvoudigere berekeningen voor de transportcoëfficiënten.
Numerieke berekeningen in realistische gevallen zijn samengevat
in hoofdstuk VI. De toepasbaarheid van de resultaten in het
zwak geïoniseerde gebied hangen sterk af van de gebruikte
electron-atoom botsingsdoorsnede. Voor argon blijkt het
Rarasauer minimum zware beperkingen aan de toepasbaarheid van de
theorie in te houden. Berekeningen van de transportcoëfficiën-
ten worden ook vergeleken met zogenaamde mengregels. De meng-
regel voorgesteld door Frost blijkt, vooral gezien de onnauw-
keurigheid waarmee de botsingsdoorsnedes bekend zijn, redelijk
betrouwbaar voor de berekening van transportcoëfficiënten.
Er wordt ook aandacht geschonken aan zogenaamde seeded
plasma's. Voor een cesium-seeded argon plasma is het electrisch
geleidingsverraogen berekend. Daarbij worden twee sterk van
elkaar verschillende reeksen metingen van de electron-cesium
botsingsdoorsnede vergeleken.
-154-
Nawoord
Voor de prettige samenwerking en het kritisch volgen van mijn
verrichtingen dank ik Piet Schram.
Voor de nuttige opmerkingen tijdens de besprekingen van het
manuscript wil ik ook Ties Weenink danken.
De (ex-) leden van de werkeenheid gasdynamica wil ik bedanken
voor de werkbesprekingen o.l.v. Rini van Dongen waaraan ik
mocht deelnemen.
Verder bedank ik alle leden van de vakgroep transportfysica
voor de prettige tijd die ik met hen heb beleefd in W&S.
Korte levensloop
Geboren te Eindhoven op 2 november 1952.
Middelbare schoolopleiding Atheneum-B gevolgd aan het
st.Bernardinus college te Heerlen van 1965 tot 1971.
Studie electrotechniek aan de Technische Hogeschool Eindhoven
van 1971 tot 1978.
Van 1978 tot 1982 wetenschappelijk assistent in de werkeenheid
kinetische theorie van de vakgroep transportfysica van de
afdeling natuurkunde aan de TH Eindhoven.
Stellingen behorende bij het proefschrift van
F.J.F, van Odenhoven
Eindhoven, 18 februari 1983.
Bij de zogenaamde hydraulische sprong stroomt het water van de lage naar de
hoge zijde. Dit volgt uit de energiebalans en het feit dat de entropie moet
toenemen. Deze conclusie wordt ook door Landau en Lifshitz bereikt, maar op
grond van een foutieve berekening. Zij laten namelijk de bijdrage tot de
energieflux van de potentiële energie in het gravitatieveld ten onrechte weg.
1) Landau and Lifshitz: A course of theoretical physics, vol.VI,
Fluid Mechanics, Pergamon Press, 1966, p.398.
II
De eerste orde correctie van het over een gyratieperiode gemiddelde magne-
tische moment van een geladen deeltje in een inhomogeen magnetisch veld is
in het kader van de adiabatische theorie gelijk aan nul. Dit resultaat
volgt niet uit de berekeningen van Northrop .
1) T.G.Northrop: "The adiabatic motion of charged particles",
Interscience Publishers, 1963.
Ill
In een zeer zwak geïoniseerd gas is het noodzakelijk een isotrope correctie
op de nulde orde verdelingsfunctie van de electronen toe te laten. Een bewe-
ring van Bernstein van tegengestelde strekking is derhalve onjuist.
1) I.B.Bernstein in: "Advances in Plasma Physics", vol.3, 1969, p.127.
2) Dit proefschrift, hoofdstuk III.
IV
In een instabiele schuiflaag voldoet de gradiënt-lengte van het snelheids-
profiel beter als karakteristieke lengte dan de impulsverliesdikte.
In de behandeling door de Groot et al. , van een electron-foton gas is ten
onrechte de dynamische afscherming geheel buiten beschouwing gelaten.
1) S.R. de Groot et al.: "Relativistic kinetic theory", North Holland
Pub1i shing Company,1980.
VI
De uitdrukking van Rostoker voor het tensoriële geleidingsvermogen van een
plasma is onjuist. Dit blijkt uit het feit dat 2ijn uitdrukking niet isotroop
wordt in de limiet: k -> 0. In een correcte behandeling moet met het inwendige
magnetische veld rekening gehouden worden.
I) N.Rostoker, Nuclear Fusion J_(1961)101.
Vil
De nulde orde verdelingsfunctie van de lichte deeltjessoort in een Lorentz-
gas relaxeert naar een willekeurige isotrope functie. De veronderstelling
van Chapman en Cowling dat. dit een Maxwellverdeling is, volgt niet uit de
Chapman-Enskog theorie
1) Chapman and Cowling: "The mathematical theory of non-uniform gases",
Cambridge University Press, 1970, p.188.
VIII
De in turbulentie-theorieën vaak gemaakte veronderstelling dat het ensemble
van dynamische systemen uniform is , blijkt soms in strijd te zijn met de
dynamica van die systemen .
1) R.C.Davidson: "Methods in nonlinear plasma theory", Academic Press, 1972.
2) I.E.Alber, Proc.R.Soc.Lond. A363(1978)525.
IX
Oplossingen van de electronentemperatuurvergelijking duiden er op dat
macroscopische "runaway" van electronen in een gedeeltelijk geïoniseerd
gas slechts in uitzonderlijke omstandigheden te verwachten valt.
1) Dit proefschrift, hoofdstuk V.
I'te Met betrekking tot de evenwichtige opbouw van onderzoek- en onderwijs-
>'- programma's is het wenselijk om bij de afsluiting van onderzoekcontracten
• ;. met bedrijven een «xtra percentage in rekening te brengen voor gelieerd'•'-', onderzoek van fundamentele aard.