Z.

Physik 228, 99--119 (1969)

Validity Conditions
for Local Thermodynamic Equilibrium
H. W. DRAWIN
Association Euratom-CEA, Dtpartement de la Physique
du Plasma et de la Fusion Contrtlte, Centre d'Etudes Nucltaires,
Fontenay-aux-Roses (France)

Received April 21, 1969

Validity criteria for partial and complete local thermodynamic equilibrium (L.T.E.)
are re-examinated on the basis of generalized cross section formulas which are in
agreement with experiments and with quantum mechanical results. The validity
conditions for L.T.E. obtained deviate both in the absolute values and in the tem-
perature dependence from those given by other authors. The influence of resonance
absorption on the establishment of L.T.E. is considered and quantitatively taken
into account. The influence of atom-atom collisions on the population densities of
highly excited levels is also discussed. A general formula is derived permitting an
estimate wether the excited levels are populated due to electronic or due to atomic
collisions. The condition obtained is quite severe for plasmas in which the electron
temperature is different from the gas temperature. For exarnple: In a helium non-
L.T.E. plasma of 1 ~ ionization degree (for Te=2X 104~ all levels of principal
quantum numbers n > 12 will essentially be populated due to atomic collisions at a
gas temperature of Ta= 1.6 X 104 ~ The influence of diffusion on the L.T.E. con-
ditions is discussed for the various kinds of particle interactions. It is found that
deviations from L.T.E. due to particle diffusion may be more severe than is generally
assumed.

I. Introduction
The concept of (complete) local t h e r m o d y n a m i c e q u i l i b r i u m (com-
plete L.T.E.) plays a n i m p o r t a n t role in the field of p l a s m a spectroscopy.
W h e n complete L.T.E. prevails the p l a s m a state at a spatial p o i n t r at
time t can be calculated if, for instance, the two parameters total pres-
sure p a n d local t e m p e r a t u r e T are k n o w n . More generally, all laws a n d
t h e r m o d y n a m i c relations for complete t h e r m o d y n a m i c e q u i l i b r i u m also
hold in the case of complete local t h e r m o d y n a m i c equilibrium, the only
exception f r o m this rule makes P l a n c k ' s r a d i a t i o n law. But even in those
cases, in which complete L.T.E. does n o t further exist, partial L.T.E.
m a y still be realized, thus, p e r m i t t i n g nevertheless useful applications
of general t h e r m o d y n a m i c formulas u n d e r restricted conditions.
8 Z. Physik, Bd. 228
100 H, W. Drawin:

Deviations f r o m L.T.E. have been discussed by m a n y authors, see

e.g. Refs. a-15
The problem of eventual departures f r o m L.T.E. is reconsidered here.
Using recent cross section data formulas for the validity of partial and
complete L.T.E. are obtained. Our results slightly deviate f r o m those
given by Griem. Additionally, a formula is derived f r o m which the in-
fluence of the heavy particle collisions on the populations of the highly
excited states can be estimated. This is of special importance in plasmas
of low electron and high neutral particle densities. F o r strongly i n h o m o -
geneous plasmas the validity condition is obtained f r o m the " z e r o - o r d e r
m o m e n t " of the Boltzmann collision equation.
All formulas for the cross sections and the rate coefficients applied
in this paper m a y be f o u n d in a separate publication ~6 to be referred as I.

II. Partial L.T.E. in Time-Independent

and Homogeneous Optically Thin Plasmas
We consider the most unfavorable case, namely an optically thin
plasma.
D e n o t i n g by nl the population densities calculated on the basis of
the collisional-radiative model, and by n~saha/ the corresponding Saha-
density as given by

nenz, 1 2gz,~ (2rcmekT)a/2 ( ~(0,,Z_~,Z_~,,Z_~,ZA~,)

tt(Saha) h3 exp (1)
z-l,i gz-~,~ kT
one m a y define a coefficient bi by ni =bin~ saha). A level is said to be in
partial L.T.E. when b, is equal to unity within 1 0 ~ or better. W h e n
this condition is fulfilled one m a y be sure that spectroscopically measured

1 Giovanelli, R. G. : Austral. J. Sci. Research 1, 275 (1948).

2 Uns61d, A. : Z. Physik 171, 44 (1963).
3 Pagel, B. E. J. : Proc. Roy. Soc. (London), Ser. A 306, 91 (1968).
4 Griem, H. R.: Phys. Rev. 131, 1170 (1963).
5 Griem, H. R.: Plasma spectroscopy. New York: McGraw-Hill 1964.
6 Biberman, L. M., Ulyanov, K. N. : Opt. Spectr. (USSR) (English Transl.) 16, 216
(1964).
7 Vorobev, V. S.: High Temperature 4, 468 (1966).
8 Lagarkov, A. N.: High Temperature 4, 301 (1966).
9 Thomas, R. N.: Astrophys. J. 123, 260 (1957); 131, 429 (1960).
10 Jefferies, J. T., Thomas, R. iN. : Astrophys. J. 129, 401 (1959).
11 Jefferies, J. T. : Astrophys. J. 132, 775 (1960).
12 Hummer, D. G.: Monthly Notices Roy. Astron. Soc. 125, 461 (1963).
13 Hummer, D. G. : J. Quant. Spectr. Radiative Transfer 8, 193 (1969).
14 Sampson, D. H.: Astrophys. J. 144, 96 (1966).
15 Deutsch, C.: Phys. Letters 28 A, 525, 752 (1969).
16 Drawin, H. W. : Z. Physik 225, 483 (1969).
 Validity Conditions for Local Thermodynamic Equilibrium 101

population densities ni permit a determination of the electron density

and the electron temperature by means of the Saha-Eggert equation (1).
In order to define a validity condition for the fulfillement of b i = 1
(within 10 ~ or better) one may climb down the level system beginning
at the ionization limit. Due to the large two-body ionization and three-
body collisional recombination rates the highly excited states will be in
a collision-dominated equilibrium with the free electrons. Thus, the
condition bi = 1 is fulfilled.
In climbing further downward the direct collision coupling of the
excited states with the continuum of the free electrons will become looser,
however, the collision excitation and de-excitation rates between the
closely spaced levels are still sufficiently large so that a Saha-population
is accomplished via these many closely spaced levels. The collision-
dominated coupling between the excited states ceases when one arrives
at a l e v e l j = i + 1 for which the spontaneous decay rate nj ~ Ajh is of the
h<j
same order as or larger than the collisional decay rate nj ( a (e) lYe>j-,i=j- 2.
It depends now on the level i = j - 1 wether the enhanced radiative decay
of level j can yet be compensated by a sufficient number of excitation
processes i ~j, or not. It is evident that the population densities nj and
n i can be stabilized by collisions when the following relations are fulfilled

n e nj <0"(e) lYe>j~i >>nj ~ Ajh

h<j
(2)
/'/e nj <0"(e) lYe>i~j >>ni Z A~h, i<j.
h<i
We use for <a(e)IYe)i"-*j and ( a (e)lYe)j....~i the expressions given in Ref. 16
[Eqs. (I,10) (I,11)] containing functions ~(fl, x) and ~2(fl, x). Some
numerical values have been listed in Table 1. ~2 is nearly independent
from fl, if x > 1. In the following we pose ~1(2, x ) - gJl(x) and ~2(1, x ) -=
~P2(x). For practical purposes we additionally introduce the functions
9 l(fi, x)=[~l(fl, x)exp(+x)]-x; ~2(fl, x)=[~2(fl, x)exp(+x)] -~
of which the numerical values have been listed in Table 2. Introducing
these relations into (2) yields for the electron density the conditions

Ajh ~q~l(Uij) cm-3 for atoms (3a)

ne>l.15• 3 gj u~j h<~ x
- gi u~/2 fij (~2(u~j) cm -3 for ions

Z Aih x ~ ' [ ~ i ( u / j ) ] - I cm -3 for atoms (3b)

ne>_l.15xl0SZ 3 u~j h<*
- u~/2 fij ([~2(uij)] -1 cm -3 for ions
8*
 102 H. W. Drawin:

Table 1. Numerical values of the functions gtl (fl, x) and gt2(]l, x)

x ~1(1, x) tPl (2, x) % ( 1 , x) x gtl(1, x) gtl (2, x) % ( 1 , x)

1 9 10 -5 1.11 +1 1.18, -b 1 1.12 q- 1 8 1.43, -- 5 3.87, -- 5 1.14, --4

1 9 10 - 4 8.85 +0 9.54, -b0 8.85 +0 10 1.48, - - 6 4.22, -- 6 1.51, --5
1 9 10 - 3 6.53 -kO 7.22, -b 0 6.53 -4-0 12 1.61, - - 7 4.77 -- 7 2.01, --6
1 9 10 - 2 4.16 -k-O 4.82, +0 4.16 -4-0 14 1.81, - - 8 5.54 -- 8 2.69, -- 7
1 9 10 - 1 1.76 -4- 0 2.26, -t-0 1.79 +0 16 2.10, - - 9 6.57 --9 3.61, --8
0.4 5.90 - - 1 8.59, -- 1 7.29 -- 1 18 2.47, - - 1 0 7.91 --10 4.86, --9
1.0 1.55 --1 2.58, -- 1 2.21 -- 1 20 2.96, --11 9.64 --11 6.53, --10
1.5 6.37 --2 1.14, -- 1 1.10 --1 22 3.58;--12 1.18 --11 8.81, --11
2.0 2.87 -- 2 5.47, -- 2 6.13 --2 24 4.39, - - 1 3 1.47 - - 12 1.19, - - 11
3.0 6.78, -- 3 1.42, -- 2 2.07 -- 2 26 5.42, - - 1 4 1.84 _ 13 1.60, -- 12
4.0 1.79, -- 3 4.01, - - 3 6.91, --3 28 6.73, - - 1 5 2.31 - - 14 2.16, - - 13
6.0 1.48, -- 4 3.71, - - 4 8.73, -- 4 30 8.43, - - 1 6 2.92 - - 15 2.92, - - 14

Table 2. Numerical values of the functions ~l (fl= 2; x) and ~z (P= 1 ; x)

x ~1(2, x) ~2(1, x) x ~1(2, x) ~2(1, x)

--r --~z(x) =--~l(x) -~2(x)

1 - 10 . 5 8.44 --2 8.96 --2 12.0 1.31 1 3.06, 0

1 9 10 - 4 1.05 -- 1 1.13 --1 14.0 1.53 1 3.09, 0
1 9 10 - 3 1.38 -- 1 1.53 -- 1 16.0 1.75 1 3.12, 0
1 " 10 - z 2.05 -- 1 2.38 --1 18.0 1.96 1 3.14, 0
1 9 10 - 1 4.00 -- 1 5.06 -- 1 20.0 2.18 1 3.15, 0
0.4 7.80 -- 1 1.02 0 22.0 2.40 1 3.17, 0
1.0 1.43 0 1.66 0 24.0 2.61 1 3.18, 0
2.0 2.47 0 2.22 0 26.0 2.83 1 3.19, 0
3.0 3.52 0 2.48 0 28.0 3.05 1 3.20, 0
4.0 4.57 0 2.65 0 30.0 3.27 1 3.21, 0
5.0 5.62 0 2.76 0 35.0 3.81, 2 3.22, 0
6.0 6.69 0 2.84 0 40.0 4.36, 1 3.23, 0
7.0 7.75 0 2.90 0 45.0 4.90, 1 3.24, 0
8.0 8.82 0 2.94 0 50.0 5.45, 1 3.25, 0
9.0 9.90 0 2.98 0 60.0 6.54, 1 3.26, 0
10.0 1.10 1 3.01 0 100.0 4.14, 2 3.28, 0

Read: 8.44, - - 2 = 8 . 4 4 x 10 -2, etc.

when partial L.T.E. shall be fulfilled within 10~ or better down to

l e v e l i. Aih is i n s e c - 1 , a n d u i j e q u a l s t h e e n e r g y d i f f e r e n c e ( i n u n i t s kTe)
b e t w e e n t w o a d j a c e n t l e v e l s , h e n c e u l j = ( E j - E ~ ) / k T e, w i t h j = i + 1.
I n t h i s f o r m , t h e v a l i d i t y c o n d i t i o n f o r p a r t i a l L . T . E . is v e r y g e n e r a l
and may be applied to any levels i and j in any atomic or ionic level
system.
 Validity Conditions for Local Thermodynamic Equilibrium 103

A simpler equation is obtained by making the following approxima-

tions valid for sufficiently excited hydrogen and hydrogen-like levels j:
j-1 Z4
~..Ajh~l.6OxlOl~ fL j+l_--_-0.5j; uj, j+l~--2.@.
h<j

j now means the principal quantum number. One obtains in stead of

Eqs. (3 a) and (3 b) the condition

n~_>7.4 x 1018 Z7 ( k T~ ~1/2 x ~ E7'1(2u'/J3)]-i cm-3 for atoms

-- ~\Z2E~/ ([~2(2Ul/ja)-I - l c m -3 forions. (4)

This formula differs from Griem's formula by a factor 1/~Ul(x) for

neutral atoms and by a factor 1/~2(x ) for ions. These factors represent
the quantum mechanical correction of Griem's classical result. The
practical consequence is that formula (4) gives decreasing values for ne
when Te is increased, at least for small and medium quantum numbers.
Compared to Griem's formula 4' s, the Eq. (4) leads to about four times
larger electron densities at low and medium quantum numbers, and to
about ten times smaller values for n~ when the quantum numbers j are
large ( j > 8).

Ill. Complete L.T.E. in Time-Independent

and Homogeneous Optically Thin Plasmas
The method applied for obtaining the validity criteria for partial
L.T.E. in neutral and ionic level systems may also be used for obtaining
a condition for the validity of complete L.T.E. For that complete L.T.E.
be established one has to require that the condition (3 a) is fulfilled for
any two adjacent levels with the ground state included. The most severe
condition, however, does not originate from the energy gap between
the ground and the first excited state in all cases, it is mostly determined
by the largest possible energy gap which can ever be realized within
between two adjacent levels. In hydrogen and helium, for instance, this
energy gap just corresponds to that between the ground and the first
excited state. However, there are many level systems which do not follow
this rule, e.g. 0 II showing two excited levels relatively close to the ground
state (excitation energies El, 2 =3.3 eV, and El, 3 =5.0 eV). Then follows
a large energy gap of approximately AEmax~10 eV before one finds an-
other excited level. It is obvious that the condition for complete L.T.E.
should be applied to the energy gap with AEm,x'~10 eV rather than to
the gaps E1,2 =3.3 eV, and El, 3 =5.0 eV, etc.
104 H. W. Drawin:
///////////////////////////.
--k.
j
- - I
h

A Emox

3 L]---Emin,gmin
/
2

GROUND STATE/, 1
Fig. 1. For the definition of the largest energy gap within a level system

Modifying the condition (3a) in this sense we obtain the following

general condition for the validity of complete L.T.E.:
n~> 1.15 x 108Z 3
gma~ Emax--Emin 2hmax, f ~ ~I(A I/max) cm-3 for atoms
grain k T e u3/2 fra, . . . . . ~bz(A Umax) cm -3 for ions (5)
with
A Umax=-~(Emax-Emln)/k T e .
The index "rain" indicates the level at the low-energy side of the largest
possible energy gap (Fig. 1) and the index " m a x " means the correspond-
ing level on the high-energy side. The ratio ~, Araax/fmi . . . . . may be
' < max
scaled with respect to the energy gap Em~x-Emi. using the relation

,<maxZAmax, i ,..A2,
i n ( Emax__ Emin.~ 2 [ Emax__ Emin ) 2
fL2 ~ Ef ] =1.13• - ~ tn sec -1 (6)
fmin, max = ~
with A2n,l = 4 . 7 x 1 0 s sec -1, flnz=0.42. Thus, one finally obtains the
following general condition for the validity of complete L.T.E. in an
optically thin plasma

ne__>6.5 x 1016
gmx Umax-Em,~ 3 (k [~I(A Umax)cm-3 for atoms (7)
x
gmin \ Ef \--E-~ ] l ~ z (A Umax)cm- 3 for ions

The functions ~1 and ~2 have the same meaning as in Eq. (3a) with
the argument 2ul/j 3 replaced by Aum,x. The numerical values for ne
 Validity Conditions for Local Thermodynamic Equilibrium 105

calculated from relation (7) differ from those of other authors by not
more than a factor three to four, however, the temperature dependence
is different.
This difference originates from the different cross sections used in
deriving the relations (3). The coefficients ~l(x), ~2(x) contain the
quantum mechanical correction of the classical results.
The influence of strong resonance absorption on the validity criteria
for complete L.T.E. is discussed in the next section.

IV. Influence of Resonance Absorption on the Establishment

of Complete L.T.E.
One generally argues that radiative absorption in the first members
of the resonance lines favours the establishment of complete L.T.E., and
that due to the large optical thickness of the resonance lines complete
L.T.E. prevails down to rather small electron densities. This may partly
be true for the inner regions of stellar plasmas, however, for laboratory
plasmas this argument does not hold. In general, only the first two or
three members of the resonance series are completely optically opaque.
All other transitions are, in general, optically thin.
In order to show quantitatively the influence of resonance absorption
on the establishment of complete L.T.E. we have calculated the popula-
tion densities of the hydrogen and of the helium levels from a coupled
system of rate equations for the stationary state assuming different de-
grees of reabsorption. The method consisted in reducing successively the
transition probabilities for the radiative transitions. Although not con-
sistent, this method shows nevertheless how serious may be the departure
from complete L.T.E. in laboratory plasmas of small geometrical dimen-
sions. (In a rigorous treatment one has to solve simultaneously the
coupled system of rate equations a n d the corresponding radiative transfer
equations for the geometrical problem given.)
Fig. 2 shows a typical result for the ground state density of neutral
hydrogen. The coefficient b 1 denotes the calculated number density nl
divided by the number density n(ffaha) calculated from Eq. (1), hence
b 1 = n l / n ( 1 s"ha) .

Curve 1 is for an optically thin plasma (all photons escape from the
plasma). By reducing successively the radiative transition probabilities
one obtains the curves 2 to 8. There is only a small difference between
the bt-values for an optically thin plasma and a plasma with the resonance
lines completely absorbed. Only when the transitions of the higher spec-
tral series are optically thick complete L.T.E. may exist at corresponding-
ly lower electron densities. However, in laboratory plasmas one never
106 H. W. Drawin:

111
10 7
HYDROGEN. Te= 1.2x10 4 ~
STEADY STATE
10 e ( ~ OPTICALLYTHIN
(~)-(~) OPTICALLYTHICK

10 5

10 4

10 3

10 2

1o I

io o
1o 1o 1o 12 1o TM i0 le ne ~
1018em-3
Fig. 2. hi-values for the ground state of hydrogen without and with radiative ab-
sorption. The definition of b I is bl=nl/n~saha). Curve1. All transitions optically thin.
Curve2. Lyman --c~ optically opaque (A12=0). Curve3. Lyman series optically
opaque ( A I j = 0 ) . Curve4. Lyman series, Lyman continuum, and /-/=-line optically
opaque (Al=A1j=A23=O). CurveS. Lyman series, Lyman continuum, Balmer
series and Balmer continuum opaque (AI=AIj=A2=Azj=O). Curve6. Lyman
series, Lyman continuum, Balmer lines, Balmer continuum, Paschen lines, and Paschen
continuum optically opaque (AI=AIj=A2=A2j=A3=A3j=O). Curve7. Lyman
series, Lyman continuum, Balmer lines, Balmer continuum, Paschen lines, Paschen
continuum, Pfund lines, and Pfund continuum optically opaque (AI=Alj=Az=
Azj=A3=A3j----A4=A4j=O). Curve8. All free-free and free-bound transitions
optically opaque

obtains substantial reabsorption in the Balmer-, Paschen-, ... etc. series.

(A similar result has been obtained for He I, He II, and for Li. The
results will be published in a separate paper.) It follows from this, that
even in the case of extremely strong absorption in the resonance lines
the right-hand side of the relations (5) and (10) may be reduced by not
more than a factor 10. This quantitative result is in general agreement
with the assumption made by Griem s
Thus, one obtains as condition for ne

n~6.5 • 1015 )3 x
gmax (Emax--Emin (kTe] ~/2 f ~(Aumax) cm-3f~176 (8)
" ~mi, \ ~ \ E~ ] ~r

when substantial resonance absorption exists.

 Validity Conditions for Local Thermodynamic Equilibrium 107

As examples we consider the level systems of H and O I I :

a) Hydrogen at a Temperature of kTe=l eV. One has Em,x-Emi,=
10.1 eV, gmax/gmln=4, Aumax=(Em,~-Emi,)/kTe=lO.1, q ~ ( 1 0 . 1 ) = l l .
Thus, the right-hand side of (8) becomes 4.2• ~6 cm -3, This is in
agreement with the curve 3 in Fig. 2.
b) Level System of 0 11 at a Temperature of Te=2.12 eV. The two
levels in question have the configurations 2s 2p 4 4p and 2s 2 2p 3 2pO,
respectively. One gets Em~-Emln = 9.9 eV, gma• = 12, groin= 10, A U.... = 6.4,
~2 (A urn,x)= 2.86. With these values follows for the right-hand side of (8)
the value 3 . 6 x l 0 ~ S c m -3 confirming the measurements of Berg x6"
who could not find deviations from L.T.E. for electron densities
n~>l.1 x 1016 cm -3.

V. Influence of Resonance Absorption on the Establishment

of Partial L.T.E.
Absorption of resonance radiation h vlj influences the establishment
of partial L.T.E. This will be the case when the levels j and j - 1 con-
sidered partly decay down to the ground state. The corresponding tran-
sition probabilities appearing in the sum ~ Ajh of (2) should, thus, be
multiplied by a reduction factor Adj. As a rule, all A~ ~ should be multi-
plied by the corresponding coefficient Aij, however only the A l j coef-
ficients associated with the ground state will have an influence, the
number densities n~>l of the excited states are generally too small to
give values A~j< 1.
It can easily be shown that strong resonance absorption does not
appreciably change the conditions (3) and (4) for partial L.T.E. When
strong resonance absorption exists, the right-hand sides of the relations (3)
and (4) should be multiplied by a coefficient not smaller than 0.5. Thus,
one has the numerical values 0.6 • 108 and 3.7 • 108, respectively, in
stead of 1.15 • 108 and 7.4 • 10 is.

VI. Influence of Heavy Particle Collisions on the Equilibrium Population

of Highly Excited Levels
When heavy particle collisions are important the excited states will
show population densities which are determined by the gas temperature Ta
rather than by the electron temperature Te, provided that T e ~=T~. The
heavy particles may be either ions or atoms.
We consider plasmas in which the number density of the neutral
atoms in the ground state is larger than that of the ions. Thus, neutral
particle collisions will be much more important than collisions between
16a Berg, H. F.: Z. Physik 207, 404 (1967).
108 H . W . Drawin:

atoms and ions, since the inelastic cross sections for atom-atom and
atom-ion encounters are not very different from each other in the tem-
perature region of interest.
The relative influence of the electronic and neutral particle collisions
on the population densities of the excited levels may be estimated by
considering the de-excitation rates from a level j to an adjacent level
i=]-1 due to electronic and neutral particle encounters.
Heavy particle collisions will be more efficient than those by elec-
trons if
no (q(a) Va)j...,,.(j- 1) >~ 1 (9)
ne (if(e) Ve) j....:,(J- 1)
where na denotes the number density of the ground state atoms.
Inserting the equations for (o-(~ve}j_~(j_ 1) [see Eq. (I,11)], and for
(o-(")v~}2_~(i_~) [see Eq. (I,18)] into relation (9) yields

2
l + - -
4 (Ta)l/2(me]l/2 me ~l(Uj, j-1) wj, j-1
(_ 2me n~ (10)
~ \~I mn u.#,s-1 1+ 2 ne >>l"
\ maWj, j-1

For sufficiently excited levels holds (with Z = 1)

_ 2ul.

uj, j-,--k--L-jj=- F ,
I +Uj, j-1
#'l(uJ's-l)~ 1

where j is identified with the principal quantum number in the hydro-

genic approximation. With these quantities the relation (10) is fulfilled
for all levels]>]'. This means that the population densities nj belonging
to levels of principal quantum numbers j>j' are determined by the gas
temperature rather than by the electron temperature. As long as Ta = Te,
this is not important. However, when the gas temperature is very different
from Te spectroscopic measurements will give the gas temperature rather
than the electron temperature when spectral lines originating from highly
excited levels are evaluated.
For a numerical example, we consider a helium plasma. We pose:

na/ne= 102, T, = 1.6 x 104 ~ Te =2.0 x 10~ ~

 Validity Conditions for Local T h e r m o d y n a m i c E q u i l i b r i u m 109

and obtain from (10) the relation.

4.6 x 10-5j3 q~l (17.5 1 .1 + .8 . 2.x 10-2j

. 3 >1
\ j3 ] l + l . 2 5 x 1 0 - 1 0 j 6

which is fulfilled for all j > j ' = 12.

The physical reason for this effet originates from the quite different
energy behaviour of the electronic and atomic cross sections. It would
surely be very interesting to check the relation (10) for well-defined
experimental conditions.
One can also show that for highly excited levels three-body collisional
recombination with an atom as the third body may eventually be much
more important than the usual three-body collisional recombination
process with two electrons involved, provided that n , ~ n , . The cor-
responding formulas have already been given in a preceding paper 17.

VII. Validity of Partial and Complete L.T.E.

in Homogeneous Transient Plasmas
When the relaxation times are to be calculated one has to consider
all collisional and radiative processes within a level system. In the
collisional-radiative model, the temporal change of the number density
nz-~, j of level j of a ( Z - 1 ) - f o l d charged particle writes

,~,, - - t c o l l i s i o n -
~ / radiation
Z., " i
i< j
neC}})+ (1-Aij)Aji
p
+ ~ nk [n~ F(k}) + A kj] + nz n~ [Rj + tl e Q(e)]
k>j
j--1 p
-
i<j
2 [ or} )+Ajd-nj Z
k>j
(11)

nj

where the nz_ 1,, and nz denote the number densities of particles of two
successively following ionization stages (Z = 1 for hydrogen). Aj and Aij
are reduction coefficients due to radiative absorption of free-bound and
bound-bound radiation, respectively. On the right-hand side of Eq. (11),
the index Z - 1 has been dropped for the number densities nz_l,i,
HZ_ I,j, etc.
17 D r a w i n , H. W . : Z. Physik 211, 404 (1968).
110 H.W. Drawin:

One sees that the return to a steady state after a sudden change
of ne or Te takes place in a time of order Zz-1,j, where
j-1
1 ~neS~e)+ nzne ( 1 - - A j ) R j + ~ ,_rn~F(Y.)+Aji
"Cz-a.j nz-~,j i<j
(12)
+ ne j k ~ (1--Ajk)Akj .
k>j nz-l,j
This equation together with the condition for the steady state

( 9a n ~z - l ' ~ .t~oni~io _v,-n with j = l , 2 , . , . p = i * (13)

O~ / radiation

yields the relaxation time ~z-~, j for the level j .

A simple analytical formula for %_ ~,j is obtained by retaining only
the largest term on the right-hand side of Eq. (12) and multiplying this
term by a coefficient which accounts for the suppression of the other
terms.
1. Ground Level (j = 1)
In the optically thin case is A~ j = 1. It remains for complete ionization
p

TZ-1, 1 k>l

In most cases holds

p
l~O((7(e)ve)l~2<s(e)<~(cT(e)lge)l"*2; L (G(e)l)e)l~k"~(cT(e)De>l~2,
/~>1

hence, with the approximation S~e) =0.3 n e(cr(e)ve)l~2 one obtains

1
~z- 1 1~ (14)
1.3 ne <(r(e) Ve>l~ 2
where <tT(e)Ve>l~2is the rate coefficient for collisional excitation of the
first excited level of the resonance series. According to Eq. (3b) one
obtains two different analytical expressions for the relaxation time. Put-
ting the relations for the rate coefficient (a(~)v~)~_,2 into Eq. (14) yields

0.89• E ~ - E 1 ( k T e ] I/2
"~z-~, ~(A~,2=l)---
(15)
~[~1 (u12)]- 1 see for atoms
x ([~e2 (u12)]-i sec for ions

with R12 = IE2 -- E l]/kze. Eq. (15) holds for complete ionization.
 Validity Conditions for Local Thermodynamic Equilibrium 111

In the optically thick ease (A1, AI,j< 1 for any degree of reabsorption)
one may write
Zz_I,I(At,2<I)~ [1.3ne(ff(e) Ve)l__,2+ nz-a'2
HZ--1, 1
(1-A1,2)A21 (16)

where the ratio nz_l,2/nz_l,1 has to be calculated from the coupled

system of rate equations (13). However, when the system is close to
the state of complete L.T.E., this ratio may be replaced by the Boltzmann
ratio and one obtains
"~Z- 1, 1(A12 < 1)-~ [1.3 ne(a (e) re) 1_~2 + ( 1 --A12) A12
I_
(lVa)
gz-l,2 exp kT~
gz-l, 1
and with Eq. (14) follows
1 1
Zz- 1, 1(Alz < 1) ~ "Cz-1, a(A,2 =1)
(I7b)
+ gz_l,a (e_Aa2)A21exp ( Ez-E,)
gz-1, 1 kTe "
When the resonance line is completely opaque one has AI 2 - 0 .
The relations (14) to (17) have to be used for an estimate of the
relaxation time necessary for establishing complete L.T.E. with the ad-
ditional condition that the electron density has simultaneously to fulfill
the relations (7) or (8), respectively. These latter ones are just the con-
ditions for that radiative processes are much less important than colli-
sional ones. Thus, the second term of Eqs. (17 a), (17 b) is small compared
to the first one when n~ fulfills the conditions (7) or (8). However, as
soon as these relations are not fulfilled the second term in Eqs. (16), (17)
plays an important role. The relaxation times for partial ionization are
obtained by multiplying the values of Eqs. (15) and (17) by nz/(nz + nz- 1,1).

2. Excited States ( j > 1)

One only needs to retain the first term in the first sum of Eq. (12)
plus the radiative terms, hence
TZ- 1, j ~" [ne (if(e) Oe)j~,(j - 1) + 2 AJi] - 1. (18)
i<j
At low electron densities the relaxation is essentially determined by
the spontaneous transition probabilities, at least for the lower-lying levels.
However, when the electron density fulfills additionally the condition (4)
the radiative terms of Eq. (18) may be neglected and we obtain for the
relaxation time necessary to fulfill partial L.T.E. for the level of principal
112 H.W. Drawin:

quantum number j in the hydrogenic approximation the formula:

Z z _ l , j ~ 4"6x 107 Z 3 ( k T ~ ] ~/2 x ~ [~gl(uj'j-~)]-a sec f~ at~

(19)
nej 4 \ Ef] [[~2(uj, j_O] -1 sec for ions

with Uj,j_ 1 ~ - 2 Z 2 E f / k T e j 3 and Z = 1 for neutral atoms, Z > 1 for ions,

ne in cm -3. The Eq. (19) is practically independent from radiative ab-
sorption.

VIII. Complete L.T.E. in Time-Independent Non-Uniform Plasmas

In non-uniform plasmas the density and temperature gradients will
cause diffusion fluxes of various kinds. When the diffusion velocities
are high and the spatial extensions of the plasma are small relaxation
phenomena may lead to a spatial relaxation displacement of the particle
densities and, thus, of the degree of ionization compared to the static
equilibrium state (Griem s, Drawin etal. 18, Drawin 19, Bergstedt 2~
Tichmann21). Most sensitive are the ground state particles, since their
cross sections are much smaller than those for the highly excited states.
In order to obtain a validity condition for L.T.E. we start with the
"zero-order m o m e n t " of the Boltzmann collision equation for ground
state particles of chemical species Z - 1

/ radiation

where <w> denotes the average diffusion velocity of particles in a co-

ordinate frame fixed in space. The collision term on the right-hand side
provides the coupling with the other particles and with the radiation
field. A similar equation holds for the electrons. For time-independent
plasmas, the first term on the left-hand side is equal to zero.
The remaining term is composed of a particle flux due to a density
gradient and a particle flux due to a temperature gradient. However,
(Wz-1,1> does not equal the mean thermal velocity, it has in principle
to be calculated from a closed system of equations in which the higher
moments of the Boltzmann collision equation interfer. The collisional-
radiative term on the fight-hand side contains all types of collisions,
elastic and inelastic ones.
18 Drawin, H. W., Fumelli, M., Weste, G. : Z. Naturforsch. 20a, 184 (1965).
19 Drawin, H. W.: Z. Physik 186, 99 (1965).
20 Bergstedt, K.: Report IPP 3/64, Institut for Plasmaphysik, Garching, 1967. Z.
Naturforsch. 23a, 299 (1969).
21 Tichmann, K. : Report IPP 3/77, Institut fiir Plasmaphysik, Garching, 1968.
 Validity Conditions for Local Thermodynamic Equilibrium 113

The diffusion terms in Eq. (20) may be approximated by

V-(nz-1, 1 (Wz-1, 1)) = V. (Dz-a, 1 Vnz-1, 1)---Oz-1, 1 V2nz-1, 1

D z _ l l t? ( O n z _ l 1) ~ (21)
-- --
r ' Or r . ~r ' ~ /'/Z-l, 1

where Dz-1,1 is the effective diffusion coefficient of particles of type

( Z - l , 1), and l is a "characteristic length". The latter one may, for
instance, be the diameter d of the plasma or a fraction of d. The choice
of the numerical value of d depends on the inhomogeneity of the plasma
considered.
In an inhomogeneous plasma, the diffusion term does influence the
usual condition for complete L.T.E., if the term due to electronic col-
lisions and radiative processes, @nz_l,1/~t)electronlcconisions is much
radiation
smaller than the expression (21). Retaining in the coUisional radiative
term only the most efficient electronic excitation process (excitation of
the resonance line) one gets (see Eq. (14)) for complete ionization

( "0 n z - l , 1 1 electronic collisions ~ 1.3 ne nz- 1, 1 (O'(e) l)e} 1 - ' 2 - (22)

t / radiation

It should be mentioned that the recombination rate can only then

be neglected when the temperatures are sufficiently high so that
n~ nz_ 1,1 Sz- 1 > nz, 1 n, ~z. A further assumption shall be that the ambi-
polar diffusion is sufficiently reduced by a strong magnetic field.
Thus, we obtain from the condition 1.3n~nz_l,l@r(Ove)l~2>
r/z-a, 1 D z - I , 1/12 the following condition for l
l_>( Dz-l' 1 )1/2 /) ,1/2
--
1.3 n~(~5-v~) 1_.2 = (Zz- 1, 1 ~Z-- 1, 1]
(23)

which must be fulfilled for that diffusion processes, and hence spatial
relaxation are negligible. The value for the electron density appearing
in the relation (23) has additionally to fulfill the above given conditions
(7) or (8) for complete L.T.E. Zz-l,1 is given by Eqs. (15) or (17), re-
spectively. The right-hand side of Eq. (23) has the meaning of a mean
diffusion length ~(diffusion)
zLZ-- 1, 1 which the ground state particles traverse before
they come into equilibrium with the electrons.
The diffusion coefficient D z_ 1,1 is a function of the cross sections
for heavy particle encounters and of the relative velocity between the
collision partners. When no neutrals are present, D z_ 1,1 may be approxi-
mated by the usual ambipolar diffusion coefficient for charged particles
(elastic Coulomb encounters). In the presence of neutral atoms, however,
collisions between neutral particles have also to be considered.
 114 H.W. Drawin:

The first approximation to the coefficient of diffusion of a binary

mixture consisting of the two chemical species (a) and (b) is (see e.g.
Hirschfelder et al. 22)
D(.,b)_ 3(m.+mb) kT 3 (nkT]l/2 1
16nm.mb ~,n(i,
a, b
1) ~--~ \ m. b / n(a~(r)) (24)
where ~ b 1) denotes the collision integral for binary mixtures, in reduced
units, m.,b is the reduced mass, and <a~(T)) the average cross section
at temperature T.

1. Collisions between Neutral Particles

When the degree of ionization is low diffusion is mainly determined
by collisions between neutral particles. For mixtures consisting of two
chemical species (a) and (b) the diffusion coefficient for the rigid sphere
model is (see e.g. Hirschfelder et aL22):
T 3/2 [Maq-Mb~ 1/2 cm 2
D("'b)=2"63xlO-3-p-~.b \ 2 M . M b ] see (25)

with the neutral particle pressure p in Atm, the temperature T in ~

M in relative molecular weight (e.g. M = 1 for hydrogen), the effective
molecular diameter d,b in Angstrom units (d,b=(1/2)(d,+db)). When
only one chemical species is present one has Ma=Mb, da=db, and
Eq. (25) reduces to the so-called self-diffusion coefficient, D (a,"). Thus,
at low ionization degrees x ( x < l % ) , one may identify Dz_l, 1 with
D(,, b) or D ("' a). In connection with the validity condition for complete
L.T.E., however, this case is of minor importance, since complete L.T.E.
demands high electron densities. Thus, collisions with ions will be much
more important.

2. Collisions between Neutral Atoms and Singly Charged Ions

At moderately high ionization degrees resonance charge transfer be-
tween atoms and ions (of the same chemical species) may play an im-
portant role. Identifying (a(T)) in Eq. (24) with the cross section for
resonance charge transfer given by

9. z - , . 1 t~ z , / . . . . o ~ \k-~-z] (26)

one obtains the following estimate for the diffusion coefficient of the
neutral atoms ( Z - 1 , 1)

z - 1,1 = 1-ff~a~z,a ~--~z] \ m------~/ \ E~ ] (27)

22 Hirschfelder, J. O., Curtiss, C.F., Bird, R.B.: Molecular theory of gases and
liquids. New York: John Wiley & Sons 1954, 4th ed. 1967.
 Validity Conditions for Local Thermodynamic Equilibrium 115

where nz, 1 is the number density of the charged particles (in cm -3) and
mz the mass (in g) of the collisions partners.
For ions, the diffusion coefficient writes

n(eh.-tr.) ~ 3 V2
167"ca2nz-1 me 1/2 ( ~k'TZ-111/2~TZ-I~I,2 (28)
i (~Z-1)
~'~-1,1 = , ~z---]l I t E'r I

nz-1,1 is now the number density of the ground state atoms, further
mz = m z - 1 , Tz- 1 = Tz .

3. Collisions between Ions and Ions

In fully ionized plasmas, the collisions are due to the long range
Coulomb interactions. Using the " 9 0 degree" Coulomb scattering cross
section one obtains from Eq. (24) for singly ionized particles the diffusion
coefficient
D(COulomb),~ 3 32 ( _ ~ z ) 1/2 (kTz) 512 Pmax
1, 1 = 16 8n nze~lnA~ ; A s - Pmin (29)

where nz is the number density of the ions and m z the ions mass.

4. Effective Diffusion Lengths 2 (aiff"si~

Using the analytical expressions for "cz_ 1,1 and for the diffusion
coefficients one finally obtains the following formulas for the mean
diffusion lengths 2 Z--1,1
(dirf"si~ 9
diffusion of atoms in the ground state until complete ionization

•(diffusion)
a, 1 (m~11'4 [E2--E 1~1i2
atoms
[l+~ I'm.kT.~'/~l' / ~ n-T C C I t f ~ E~ !
t~l .i (30)

" t E, i t--~, / \-7--~=1

diffusion of ions in the ground state until complete ionization

) (diffusion)~ 0~i TzS/4M z 114n~ l(ln As)- 1/2

ions =

Mz In A, n e \-~-~/ \8"~a, \~a] ] (31)

[.E___22--E1.)li2(kT,~'i'[~x2(~)]-'i2cm
9 \ f,2E~ \---~-ll
9 Z. Physik, Bd. 228
116 H.W. Drawin:

with n, and n e ill cm -a, T in ~ M z = M a in relative atomic weight

units. The quantity n r 2 , o = ( a ~ ( T ) ) denotes the (classical) gas kinetic
cross section for the special type of atomic collisions considered, na is
the number density of neutral atoms, T e the electron temperature, Tz the
ion temperature, T a the temperature of the neutral atoms. The quantity
(8kT,/Ttmo) 1/* appearing in Eq. (30) equals the square root of the mean
thermal velocity of the neutral atoms (in units (cm/sec)l/2). The numerical
value of lnA~ lies generally between 10 and 20. When Eq. (15) is inserted
into Eq. (23) one obtains for c~ and c~ithe following values: ~a = 1.7 x 1011,
~ = 6 . 2 x 107. This corresponds to the optically thin case. In the case
of substantial resonance absorption, ~ and ~ become complicated
functions of A 12, he, and T e. A useful approximation is in this case
c~a x 3.5~ 101~ c~i~ 1.8 • 107.
Diffusion lengths for incomplete ionization are obtained by multiply-
ing the values of Eqs. (30) (31) by [An~n] 1/2, where An/n denotes the
percentage ionization of the particles considered.
In Table 3 we give some numerical values for the diffusion lengths
of the ground state particles of H, He, and Ar atoms at an electron
density of n e = 1 0 1 6 cm -3 as calculated from Eq. (30) for the optically
thin case. The atomic number density n~ has been put equal to zero.

Table 3. Diffusion length ~(diffusion)

"~atoms in [cm] f o r ground state atoms when the electron
density is ne= l • 1016cm -3, after Eq. (30) with % = 1 . 7 • 1011 (optically thin in all
transitions, n a has been put equal to zero)

Gas 12,000 ~ 16,000 ~ 24,000 ~ 48,000 ~

H 7.9 • 10 ~ 1.9 x 10 -1 3.4 • 10 -2 --

He 1.9 • 10 a 1.4 x I0 +1 9.8 >( 10 0 7.8 x 10 -2
Ar 4.5 • 10~ 9.2 • 10 -2 2.1 x 10 -2 --

At low temperatures recombination may considerably contribute to

the collisional-radiative rates, thus, compensating partly the flux of the
particles to be ionized. This will lead to smaller diffusion lengths com-
pared to the values given in Table 3. For plasmas submitted to strong
magnetic fields the numerical values of the Table 3 will probably give
the correct order of magnitude.
It should be emphasized, however, that the diffusion length of the
higher excited particles is much smaller than that of the ground state
atoms. The diffusion length 2~.diffusion)of excited neutral atoms in a state
of principal quantum number j may be estimated from Eq. (23) when
the quantity Zz- 1,1 is replaced by Zz- 1, j, and D z_ 1,1 by D z_ 1, j.
 Validity Conditions for Local Thermodynamic Equilibrium 117

Thus, in the hydrogenic approximation for atoms in level j :

,~(diffusion)~ 8.3 x 1011 1 ( me ] 1/4
J = [ 1 ' nanr~,a [ reek Ta ~1/2]'/z nej---~\ - - ~ /
[ -t- neTZa~ \ m - - ~ ] J (32)

{I~Ze~l/4[kZa]l/4(SkZall/4 [~.tl ~ 2E H ~1 -1,2

-ff 1 cm

It follows from this relation that partial L.T.E. can easily be established
for levels of high principal quantum number j. However, the diffusion
length of lower lying excited states may come into the same order of
magnitude than the geometrical dimensions of the plasma considered.
It follows further that the spectroscopic temperature determination
based on a measurement of the relative intensities originating from highly
excited levels of the same parent (ion) atom is practically not influenced
by diffusion.
However, the measurement of relative intensities of spectral lines
originating from two successively following ionization stages may eventually
lead to temperatures which are not consistent with the assumption of com-
plete L.T.E. required for this kind of measurement.
The Eqs. (30) and (32) are based on the cross sections for symmetrical
or resonance charge transfer of the type A + A + ~ A + +A. The cor-
responding cross sections are large at low impact energies, see Eq. (26).
In gas mixtures one has also to consider the asymmetric charge transfer
collisions A+B+--*A++B. These cross sections are extremely small
(see, e.g. HastedZS). When small traces are added to a main gas, e.g.
a small percentage of hydrogen in argon, the diffusion of the ground
state particles of the tracer gas can not be calculated from Eq. (30).
Due to the smallness of the cross sections for asymmetric charge transfer
the diffusion of the tracer gas is governed by the elastic atom-atom
collisions rather than by the charge transfer collisions. With Eq. (25),
this leads to the following diffusion length for the ground state atoms
of the tracer gas (b) in the main gas (a):

t. . . . r = n e112..l/2{/_b[,-p~xX,112
tta ~\OakZa]/] ~ fl2El~

with 6 = 1.6 x 10 al when the resonance lines of the tracer gaz are optically
thin, and 6 ~ 3 x 101~ when the tracer gaz shows substantial resonance
23 Hasted, J. B.: Physics of atomic collisions. London: Butterworths 1964.
9*
118 H.W. Drawin:

absorption, and where n, in cm -3 is the number density of the main

gas, (a~(T,)> cm 2 the elastic cross section for atom-atom collisions
(tracer gas ~ main gas), and ( E 2 - E 1 ) the excitation energy of the
resonance level of the tracer gas. This equation leads to relatively large
diffusion lengths.

IX. Establishment of Maxwell Distributions

Local thermodynamic equilibrium requires the establishment of a
Maxwellian distribution function for all types of particles, and one has
especially to require that T, = 7", = T+ = T.
In an electronic excitation or ionization process i ~ j electrons lose
the energy amount E = E i j = ] E i - E j [ . The number of such events per
unit volume and unit time is given by ne ni(a(e)ve)i-~;, where ni is the
number density of the lower level. When the cross section a (~)----tr!~)j
and nl are large, the distribution function rapidly loses electrons in the
high-energy tail of the distribution function beyond energies E = E i j .
The energy distribution function is filled up by two processes: firstly,
by super-elastic collisions j ~ i ; secondly, by momentum transfer col-
lisions between electrons and electrons and electrons and heavy particles.
The number of super-elastic collisionsj ~ i is given by nenj (a (~)v~)j_, i,
and the number of electrons transported into the high-energy tail due
to momentum transfer collisions is equal to n~nx (a(~~)Ve)E~E~j, where the
number density of the collision partners of the electrons has been de-
noted by nx. The cross section for the corresponding momentum transfer
collisions is a~(~). The subscript E>E~j indicates that the velocity average
over a(xe) has to be taken for electron energies E>E~j and electron
energies E~E~j, where E may be slightly smaller but not very much
smaller than Eij.
The electrons will show a Maxwellian distribution of temperature Te
(over the entire energy range 0 < E < o o ) when the following condition
is fulfilled for all inelastic transitions i ~ j

n~<o"(e)l)e>i~ j -- n j <o"(e)~)e>j~i ~ 1. (34)

1 + n~(a~e) Ve>E>Ei j
In the case of complete L.T.E. the nominator is equal to zero for all
electron densities, in agreement with the principle of detailed balancing.
As soon as the ratio ny/ni deviates from the Boltzmann ratio the actual
velocity distribution function can principally deviate from a Maxwell
distribution. Wether this deviation is important or not depends on the
frequency of the momentum transfer collisions contained in the de-
nominator of condition (34).
 Validity Conditions for Local Thermodynamic Equilibrium 119

The electron distribution function will be in equilibrium with that

of the heavy particles if the number of m o m e n t u m transfer collisions
between electrons and heavy particles is much larger than the number
of inelastic collisions i ~ j , hence the condition

n e ni<a (e) re>i__,j .~n e nx<a(xe) re>. (35)

In plasmas of low ionization degrees, nx denotes the number density
of the neutral particles and a(~e) is the collision cross section for m o m e n t u m
transfer "electron ~ a t o m " . When the ionization degree is high, nx
equals the ion density and a(9 the "electron ~ i o n " m o m e n t u m trans-
fer collision cross section.
Dr. H. W. Drawin
Association Euratom-CEA
D6partement de la Physique du Plasma
et Fusion Contr616e B.P. No. 6
Centre d'Etudes Nucl6aires
F-92-Fontenay-aux-Roses
France