SOVIET PHYSICS JETP VOLUME 4, NUMBEH 5 JUNE, 1957

The Establishment of Thermal Equilibrium between

Quanta and Electrons*

A.S.KOMPANEETS
Institute of Chemical Physics, Academy of Sciences, USSR
(Submitted to JETP editor October 15, 1955)
J. Exptl. Theoret. Phys. (U .S.S.H.) 31, 876-885 (November, 1956)

The role of the Compton effect in the establishment of equilibrium between quanta and
electrons is considered in the nonrelativistic approximation.

Planck's distribution is established only much

l. INTRODUCTION
later, since it exists as the result of induced emis-
JN order that radiation be considered as a closed sion.
The physical conditions under which Wien's
system, it is appropriate to assume that it is
contained in a vessel with perfectly reflecting distribultion can exist can be represented in the
walls. By itself, radiation in such a vessel can- following way. Let the matter be momentarily in a
not achieve thermodynamic equilibrium: the equa- state with very high temperature, so that all the
tions of electrodynamics are strictly linear and atoms are completely ionized, but assume radia-
the exchange of energy between vibrations of dif- tion has not yet been produced. Then the absorp-
ferent frequency, direction of propagation and tion and emission of quanta will occur at the ex-
polarization does not occur. Therefore, we must pense of a "free-free" mechanism. The corre-
consider thalt there is in the container, besides the sponding emission process is nothing else than
radiation, a suitably small "carbon particle" the bremsstrahlung of the electrons. This process
which does not appreciably disturb the field at is the more probable the lower the frequency of the
any given moment, but which is capable of ab- emitted quantum. The same applies to the proba-
sorption and radiation of energy of all frequencies. bility of the reverse process, the "free-free" ab-
Over a sufficiently long time, the absorption and sorption; therefore, for sufficiently low frequen-
emission of quanta by the "particle" leads to the cies, thermal equilibrium will be established by
establishment of thermodynamic equilibrium. means of the absorption and emission of quanta.
What will happen if we put a free electron in For high frequencies, the probability of Compton
place of the material particle in the container? scattering exceeds the absorption probability.
The container is taken to be large enough that we Since the quanta are scattered by the moving elec-
can neglect the quantization of the energy of the trons, their frequency can also be increased in the
electron. The free electron does not absorb and scattering. Initially, the number of quanta is not
emit, but only scatters quanta; therefore, the ~arge: they will all tend toward thermal equilibrium
total number of quanta in the vessel does not mdependently of each other (i.e., induced proces-
change. What sort of equilibrium is established? ses will not constitute any significant part of all
If the occupation numbers n of the individual the scattering processes). The Wien distribution
states are small in comparison with unity, then we
with mean energy 3 kT, is established among such
can neglect induced transitions and assume that quanta.
the probabililty of scattering a quantum in a cer- For this reason, the transition of energy from
tain state doo~s not depend on the number of quanta matter to radiation will be much more rapid than if
occupying this state. In other words, we replace it took place at the expense of bremsstrahlung of
the factor l + n in the transition probability by l. the electrons. The bremsstrahlung quanta which
Then the same distribution will be set up among have frequencies w greater than a certain definite
the quanta as in an ideal gas with a constant frequency w 0 , will quickly and irreversibly in-
number of particles; i.e., we obtain a distribution crease their energy by a Compton mechanism,
of the form n = e·hw/kT (Wien's law). In this tending to go to the maximum energy 3 kT.
case the mean energy of the quantum is 3 kT. In the Sections below, the distribution function
will be found for quanta which undergo the Comp-
* This research was completed in 1950 at the Insti- ton process. In the Appendix we shall consider
tute of Chemical Physics, Academy of Sciences, USSH, what energy the quanta extract from the electrons in
Heport No. 336. a body of finite dimensions.

730
 THERMAL EQUILIBRIUM 731

The laws of energy and momentum conservation

2. THE KINETIC EQUATION in the nonrelativistic approximation are written
Let us write down the kinetic equation for the in the form:
distribution function n of the quanta in an un-
bounded medium for the case of a single scatter- (hw I c) n + p = (hw' I c) n' + p', (3)
ing. For generality, we shall at first not neglect
induced processes. The equation has the following hw + p I 2m= hw' + p'
2 2 12m.
form:

(~~)c =-~d-e ~[n(l +n')N(E) (l) Here p 2 /2 m = E, p and p' are the momenta of the
electron before and after the collision, n and n'
~~ t?e d.irecti,ons of propagation of the quantum.
- n' (1 + n) N (E + hw- hw')] dW. Ehrrnnatmg p from these equations, we obtain an
equation which determines w' as a function of
w, p and the angles of scattering. In this equation,
Here N is the distribution function for the free we set w = w '+ 1'1. and limit ourselves to terms
electrons, d 'T is the element of phase space of that are linear in /'1., at least while we determine
the electrons, dQ is the differential probability of the integral which contains /'1.2. After simple
transition from a given state into another, com-
transformations, we get
patible with the laws of conservation of energy
and momentum. The index C on dn/dt denotes h/'1. =_ hcoo (p, n- n') + (hoo)2 (1- nn')
that only Compton processes are considered in the mc 2 [1+ +
(hoo I mc 2 ) (1 nn')- pn I me} • (4 )
equation. Statistical equilibrium among the elec-
trons in the plasma is established very quickly, It is valid to replace the brackets in the denomina-
ind~p~mdently of the radiation; therefore, we can tor of Eq. (4) with unity for the case of interest to
consider the distribution function N (E) to be Max-
us. The first term in the numerator has the order
wellian. Then, if we replace n in Eq. (l) by the
Planck distribution n = ( ehw/ k T - l)- 1 , then the of magnitude (kT/mc 2 ) 3 1 2 and the second term the
order (kT/mc 2 ) 2 • But we shall determine the
right and the left parts vanish, as they must.
We shall consider that the energy of the elec- integral of 1'1. 2 ; therefore, the contribution of the
trons is nonrelativistic; i.e., we assume that the first term in the averaging over the angles does
inequality kT << mc2 holds. Then the energy not vanish. Consequently, upon substitution in
Eq. (2), we must compute the following integral:
transferred in each separate act is small in com-
parison with the energy quantum h w: w' - w 1'1.= I= (hw I mc) 2 ~d-e ~ dW (p, n- n') 2 N (s). (5)
<<w.
Making use of this inequality, we expand the
integrand of Eq. (l) in a power series in 1'1. up to After averaging over all directions of p we get
terms of the second order, inclusively. We set
hw/kT =x:
iJn)
( 7ft [ iJn
c= ax-+n(l
Jh \ \
+n) kT·jd-c ~dWN(s)~
(2)
The first integral is equal to 2m (3kT /2)
= 3 mkT. In the second integral, we replace the
Compton cross section in the nonrelativsitic ap-
proximation by the Thomson cross section, which
is symmetric relative to scattering over the angles
() amd (77- ()),so that J nn'dW = 0. The Thom-
son cross section does not depend on the energy
The second integral on the right side of Eq. (2) is of the quantum; consequently,
much easier to compute than the first ; but it suf-
fices to compute only one of the two integrals. The I= (hw) 2 (kT I mc 2 ) c I l, (7)
other is determined from the condition that the
equation ought to guarantee conservation of the where l is the Compton range, which is determined
total number of quanta in the scattering. by the total cross section ( 811/3 ){e 2/mc 2 ).
732 A. S. KOMPANEETS

co co co
In order that Eq. (2) guarantee the conservation
of the total number of quanta, we must set £~ ~ nx dx
3 = 4 ~ nx dx -3 ~ nx4 dx. (14)
0 0 0
(8)
~ d'" ~ dW' N (s) il While h w is still small in comparison with kT, we
can neglect the second integral on the right in
= (kT I mc 2 ) (c ll) hw (4- hw I kT). . . te 4 Y. This means that the time
(14). Here x = x lnl
for the energy to increase by a factor e is equal to
Equation (8) can he obtained in the following way. '"c = (mc 2 14kT) ll c. (15)
The kinetic equation for the quanta must look like
a sort of conservation law 3, BREMSSTRAHLUNG SPECTRUM
The bremsstrahlung spectrum in the general case
appears rather complicated. We assume that such
where j is the "flow" of quanta in frequency conditions are satisfied in which the Born approxima-
space. Since Eq. (2) is of second order relative tion is applicable for the determination of the
to x and contains the second derivative d2n/dx 2 spectrum. These conditions are satisfied in the
linearly, without a coefficient depending on n, the light elements for sufficiently high temperature
current must have the form of the sum of the first (from several tens of kilovolts and higher). At
derivative an; ax and a certain function depend- the same time, it is taken into account that the
ing on n. But in the state of total equilibrium, inequality kT << mc 2 is satisfied. We shall not
when n = ( e" - 1 )-l, the flow vanishes. In this take into account the scattering of electrons on
case, an/ax= -n( l + n ). It therefore follows electrons, with emission of quanta, since such a
that process makes a significant contribution only in
very light materials and in each case it does not
j=g(x)[anlax+n(I+n)], (10) exceed the bremsstrahlung emission on nuclei,
always remaining smaller than it.
wher~ the function g ( x) must be determined. Sub- In the Born approximation, and in the nonrela-
stituting (10) in Eq. (9), and comparing with (2), tivistic case, the bremsstrahlung cross section is
we find that g = -x2, and the unknown function (Ref. 1, p. 183, Eq. (18)]:
f.. = x ( 4 - x ), which is equivalent to (8). Equa-
dw (16)
tion (8) is in qualitative agreement with Wien'~
law: the energy of the quantum increases so long
as hw < 4kT.
We now introduce the dimensionless time Y by In order to go over to the total number of quanta of
the expression a given frequency that are emitted per second, we
need to multiply the cross section by LN ( i) vd T
t = (mc 2 1kT) (ll c) = y. (ll) ( L is the number of nuclei per unit volume, v is the
In these units, the kinetic equation is written in velocity of the electron), integrate over all states
the following form: of the electron and refer to the number of states of
a quantum of given frequency per unit volume. Thus
on )c ------x
(-ny 1 a ,4 (an
-- x iJx -ax +n+n 2)
. (12) we obtain the equation
2

If n << 1, then this equation transforms into the

1 -
:;.-;;· - 3
v'_rr_a_ za
8m3 kT
[2
hw 3
es -hw/2kT
e
( hw )
Ko \ 2kT (17)
lipear relation
r_an ') __1__ ~_ x4 (an
\iJy 1 c - x 2 ox OX
+ n). (13)

This equation will also be solved. According to

Here Ko is the Macdonald function (Ref. 2, p. 206),
Eq. (13), it is easy to compute the time in which
T~ is the magnitude of the time interval which
the energy of the quantum increases by a factor e
as a result of the Compton effect. Multiplying characterizes the bremsstrahlung. If we take
both sides of Eq. (13) by x 3 and integrating, we into consideration the inverse abs~rption of quanta
and their induced emission, then we obtain the
get
 THERMAL EQUILIBRIUM 733

rate of change of the quanta as a result of brems- not with respect to number ) is in the fashion of
strahlung and free-free absorption: Brownian motion. The limit of the frequency, for
which the Compton process takes up the quanta
(dn I dt)B = (1 I '=V [(I + n)- neh"' I kT]. ( 18) more rapidly than it succeeds in being absorbed,
is determined by the following estimate: Tc/TB rv l.
The derivative vanishes if we put the Planck dis- Substituting TC and TB in it, we get
tribution in place of n
Digressing from the Compton process, and assum- (23)
ign that the total change dn/ dt appears on the left
side of Eq. (18), we can integrate (18). If n (0)
= 0, then we get
The entire process considered (i.e., the estab-
(19)
lishment of equilibrium by the Compton effect) pre-
n = 1
ehw 1 hT -1 (1- exp {- -:...!~-
.. (ehwJkT- 1)}) . sents interest only in the case in which h w << kT
0 '
since only then is there a frequency interval,
w >> w 0 , in which the Compton effect plays an
It then follows that the relaxation time for essential role. For small x 0 , the function K 0 can
thermal equilibrium of the quanta, for a pure brems- be replaced by an expansion in x 0 (Ref. 2, p. 96 ):
strahlung mechanism, is equal to K 0 (xoJ2) = ln(4/yx 0 ), where lny= 0.577. Thus
"B = 't~ 1(ehw 1kT _ 1). (20) x 0 is determined by the equation

The total kinetic equation, which takes into ac- Ax;;-2 In (2.35 I X 0 ) ~ 1. (24)
count the Compton and bremsstrahlung processes,
is now written This equation has meaning only if A is a small
number.
(21) We now compare the transfer of energy from the
electrons to the quanta by pure bremsstrahlung and
= kT
mc2
_c_ _1_ _i_
l x 2 iJx
x (anox 1, n + n2)
4
by the Compton mechanism*.
The total energy of the quanta emitted by the
electrons per unit volume is

Going over again to the dimensionless time y, we

get
on _i_ x4 (-iJn + n + n2) (22)
-oy = _x_!___
2 ox ox
(Ref. 2, p. 424). All quanta whose frequencies
4T
+ -~
"o
[(1 + n)- ne-']. are greater than w 0 transfer energy to the electrons
B by the Compton mechanism; on the average the
energy of the quanta approaches 3 kT. Therefore,
The braking time TB decreases with decrease in the energy transferred per unit time is
the frequency, for example, as w 2 ; therefore, the 00

quanta for sufficiently low frequencies will always (dE I dt)c--:-- 3kT ~ "il w2 dw 1-r; 2 c3. (26)
undergo transition to statistical equilihirurn in w,
processes of emission and absorption. The quanta
of high frequencies will he taken up by the Comp- The principal contribution to the integral is pro-
ton process and thus increase their frequency, vided by the low frequencies. Therefore, we can re-
approaching the Wien distribution. It should not place the function K 0 under the integral by its
he thought that frequency of the quantum increases
monotonically; Eq. (12) is of second order, of the
diffusion type. The approximation of quanta to * This comparison was pointed out to us by L. D.
incomplete equilibrium ( with respect to energy hut Landau,
734 A. S. KOMPANEETS

approximate value; to avoid divergence at the

Such a solution is valid, strictly speaking, only
upper limit, we do not integrate to infinity, but to when the exponent is large in comparison with
w = kT/h. We get unity. But if we choose C = -1, then we get an
i<T interpolation formula for the distribution function
( dE) = 3 kT __B_ ~ ~-In 4kT
(27)
which possesses the necessary properties and fo;
dt c crt2 c" h J w yhw
Wo X>> x 0 :

n = -! ( 1- exp {- 2 ~A V ·!In 2 5 }) • (32)

If we do not make use of the approximate expres- For large x/.x 0 , the number of quanta is much less
sion for K 0 , then there will still stand a numerical than the equilibrium number. Adding the factor
component in front of the logarithm in Eq. (27). ( 1 - nx) under the integral (27), a factor which
That is, in place of ln 2 ( 4/ yx 0 ), the integral gives takes into account induced emission and absorption
ln 2 ( 4/ y x)- 0.27. But we can hardly consider of quanta, we get a converging expression which
this to be accurate, since the frequency w 0 is can be integrated from w = 0 :
found by an estimation method. 00

(.~£
The ratio of the energy transferred by the Comp- '3kT B \ 1 (33)
ton mechanism to the energy transferred to the
) -
dt c - ' 1t2 c 3h .) -;r- w2 dw
0
quanta upon emission is equal to

td£' (dE\ 3 2 l;kT

1----) · --; - - I n - (28) X
exp-
{ 2kT VA
hw
• /I n~.
V 2.35 kT}
\ dt /C · \ dt )B - 4 hyw 0 '

The function under the integral vanishes for suffi-

which can reach a few score. ciently small w. The effective limit of integration
If we assume that quanta with frequency less lies at about w"' w 0 ·
than w 0 do not undergo Compton scattering at all,
we commit a certain error. There is another, and 4. GENERAL FORMULAS FOR THE AVERAGE
somewhat more accurate, method of computing FREQUENCY IN THE COMPTON PROCESS
dE/dt. For sufficiently low frequencies, the
state of the quanta is stationary, and for x-+ 0, it If the photon gas is still far from statistical
tends toward 1/x, i.e., toward the limiting form of eq_uilibriurn with the electrons, then the number of
Planck's formula. Therefore, n satisfies the ordi- photons with frequency w > w 0 is small in com-
nary differential equation which is obtained from parison with unity. In this phase of the process,
(22) if we discard an;ay
in it. Moreover, we should we can write down the following linearized kinetic
also discard n in comparison with n2, and conse- equation for quanta with frequency w > w 0 :
quently regard .x as a small quantity in the second
(34)
component of the right side. This gives
d 3 tdn ,\ 4A 235 *
-dxx ~dx·+w)+x-(1-nx)ln~=O. (29) Let us first consider the corresponding homo-
We now setn,=z '/ z,z = x + 1/J, tjJ << x. Then we geneous equation. For this we introduce a new
get the equation unknown function by the formula

__ cj_ x 31ll'' - 4A w' In 2 ·X35 = 0 (30) n = e-:c/ 2 cp (x) / x. (35)

dX 1 X 1

for tjJ '. cp satisfies the equation (we are considering the
This can be integrated approximately (by the homogeneous equation)
WKB method) if we consider the coefficient for tjJ'
(36)
to be a large number. With this accuracy,
a a2 ( x 2) _
Ill' - C ex p (l--x-
I -
2 VA vr I n--:;;
2.35} . (31)
~=x---xm+ --:-+2-- Xcp=A(x)cp.
(}y OX2 T 2 X
A

* It is further assumed that In ( 2.35/x )>> I.

 THERMAL EQUILIBRIUM 735

A
Here A ( x) is an operator on the right-hand side.
Some simplifications of the calculations arise be-
z = In x, x= e-z12f ~'- (z) e-(ll."-'l.lY. (40)
cause of the fact that 1 (x) is a Hermitian oper-
ator*. The function f ( z) satisfies an ordinary differ-
Let us assume that the initial photon distribution ential equatio/: of the same form as the Schrodinger
is given in the form of some function n 0 ( x ); then equation for the vibration levels of the diatomic
the distribution at any later moment is written molecules, if the potential function of the nuclei is
symbolically as taken in the form of a Morse potential:

(37)

Equation (41) is integrated with the help of the

degenerate hypergeometric functions, which we de-
note by the variable x:
We assume that a quantum with frequency w 1 is
emitted at the initial moment. Then n 0 (x)=
(42)
8\( x- x 1 )/xi. We also compute (in dimension-
less form) the mean frequency which this quantum (Ref. 3, Sec. 16 ). Here W . ( x) is the Whittaker
will have over some time interval as a result of 2 ·'fl

the quantum process: function, which is a combination of two ordinary

degenerate hypergeometric functions:
(38)

Here M2 . ( x) is determined by the well-kn(JNn

. ·'fl
senes

(44)

Here we have made use of the fact that a function

of a Hermitian operator is also a Hermitian opera-
tor. In Eq. (38), there enters the function x(x 1 ),
which clearly satisfies, relative to x 1 , the same The function W falls off at infinity as e-x, while
equation (36) as Cf!(X) relative to x. In contrast M does not possess this property. Therefore, X
to Cf!, the function X is subject to a different ini- is also expressed by W.
We now find the function for the discrete spec-
tial condition:
trum, analogous to the vibrational states of the
(39) diatomic molecule (functions of the continuous
X (y = 0) = xie-xt!2.
spectrum correspond to a dissociating molecule).
The discrete spectrum is possible for negative 11 2
In what follows, the index l will be omitted from
x 1' because the "current frequency" x enters no [see Eq. (41)] or for purely imaginary fl· The cor-
further. responding eigenfunctions should be quadratically
Thus, in order to find the mean frequency of the integrable. According to Eqs. (42)-(44), the func-
·quantum, which is emitted with an initial frequency tions W, for small x and purely imaginarY. I+• con-
x, we must solve Eq. (36) with the initial condition sist of two components of the form x-Yz+lfll. How-
(39). For this purpose, we must first determine ever, we cannot integrate the square of the function
A
the spectrum of the Hermitian operator A (x ). We for this upper sign. Consequently, the eigenvalues
set are numbers for which W consists only of functions
with an integrable square. Here the numbers are
A(
* We have introduced the operator x) and found its I fll = 3/2 and IP.l = 1/2, since the first component
spectrum by making use of the results of I. M. Gel'fand. of Eq. (43) vanishes [ [' (- 2) = oo, [' ( -3) = oo ].
736 A. S. KOMPANEETS

The series for the functions M2 , Y. ( x) and

M2 , 3/2 ( x) are broken off. Elementary functions
are obtained which we at once write with normal- ioo
ized coefficients X (' r (a) r (a+4) r (-a -i11- -
3 / 2 ) r (-a+ i11-- J d
3 2)

~ f(-i!l--3/ 2Jf{ifL-a/2) 0
-ico
(45)
X. 1 12 (x) = "}12 (1 - xj2) e-x/2, x_ 312 = 2-I/2xe-x/2,

where the complex integral is taken according to

The functions for the continuous spectrum are
normalized by transition to their normalized asymp- Barnes' formula 3 • Substituting this in Eq. (47),
totic expansion in a way entirely analogous to that making USe of the equation r(u)r( 1-u)=TTCSCTTU,
used in the problem of the hydrogen atom. The ex- and the fact that W2 . (X) is an even function of
·'If
pansions are referred, not to large but to small x, Jl• we get the expressiOn for x in the form of a
because for large x the function W decays ex- complex integral
ponentially. It is seen from Eq. (41) that for ioo
z--> -oo, i.e., for small x, the function {Jl is pro-
portional to cos ( Jl z + ,\). The normalized factor
x= ~2 e-xf2e-9Y 14 ~ es,YW2,-s (x) tg 1tSX sds (48 )
-ioo
of the cosine is ..j 2/ rr • Expressing W in terms of
cos(JlZ + {3)"" cos (Jllnx +.\),we get the normal- 1 1\
ized functions of the continuous spectrum
- 2 (x - 2 -) e-2Y + 3.
(46) It is therefore evident that, as a result of the
Compton process, the mean frequency of any
quantum tends toward 3 kT /h, independently of its
It is now easy to write down the general formula initial frequency. Application of Eq. (48) will be
for :X ( x on the right-hand side gives the initial given in the Appendix.
frequency of the quantum, previously called x 1 ) In conclusion, I wish ~o express my sincere
gratitude to Ia. B. Zel'dovich who set up the
x= ! ex/ 2 [e-9Y/ 4 ~ e-P.,YCp.x_p.dfL (47) present problem and displayed a constant interest
in it, and also to L. D. Landau and I. M. Gel'fand
0
who made a series ofimportant suggestions. Much
+ e-2YCI/2X.I/2 + Cs12X.s/2] ' of the results of Sec. 3 were obtained with the
participation of the late S. P. D'iakov.
where t;;Jl' t;;y, and t;;312 are the expansi,n coeffi-
cients of the function x 2 e-x/ 2 [see Eq. (38)] in APPENDIX
the orthogonal set of functions xJl; 'Y. and ,3/2 THE COMPTON PROCESS IN A BODY OF FINITE
are found in elementary fashion. For the determina- DIMENSIONS
tion of t;; we make use of the integral representa- We can compute the average frequency which is
tion of ~he function W2 ,iJl ( x) (Ref. 3, p. 345 ): possessed by quanta in a body of finite dimensions
x2e-x12 as a result of the Compton process. Since the
w2 • .... (x) =
ju --.-
21tt Thomson cross section does not depend on the
ioo energy, the diffusion coefficient of the quantum is
d
X~ f(a)f(-a-itJ.- 3 / 2)f (-a+itJ.- 3 /2)x 6
r (- itJ.- 8/2) r (itJ.- 3/2) cr. a constant. Here the probability of emission of a
-ioo quantum at the instant of time y depends exponen-
tially on the time only if the initial distribution
corresponds to one of the eigenfunctions of the
With the aid of Eq. (46), we find diffusion problem. It is natural to take a distribu-
tion which corresponds to the ground state eigen-
value since only it of all the functions is constant
in sign. In dimensionless units, the probability of
emission of a quantum from the system is given by
 THERMAL EQUILIBRIUM 737

dw (y) = ~e-f3Ydy. (I)

::=-2,.~tg"V~+9/4 r(--2Vi1+9/4) (V)
r (- 3!2- V ~ + 9/4)
The mean frequency of the emitted quantum in this x H (xo,- Vr + 9/4)
case is less than 3 kT /h. It can be determined if 00

we multiply (I) by x and integrate over y: + 2} (-)k(k + 2)! ~

k=2 (2k + 1) ! ~ + 9/4- (2k + 1)2/4
= \-
x = J xdw (y) =
i~ex/2
--:;2
ico
\
J tg 7tS W 2,-s (x)
(II)
X H (' 0 -
'
2k +
2
1'
) .
-ico
The quantities H ( 0, - ( 2 k + l )/ 2) are decom-
posed into the converging numerical series
H(O, -5/2) = 2.27, H(O, -7/2) = 2.31; there-
The energy transferred to the quanta from the maining expressions of this type are multiplied
electrons under these conditions is equal to the ~y small coefficients. We putH(x 0 ,-..Jf3+9/4)
integral (II), taken over the bremsstrahlung spec- m the form of the difference between two integrals
trum. We introduce the notation
(VI)
co

H (x 0 , s) = ~ M2.-s (x) Ko ( ~) d:a.

(III)
H (xo,- V~ +9 I 4) = ~ M 2 ,-('i•+uJ (x) K 0 ( ~) d:a
0

We then have x,

(x) dx - ~ M2.-('l•+n) (x) Ko ( ~) d:a .

x= = ('J .Xe-x/2
- Ko 2 x
(IV)
0

ico
The first integral [we call it F ( u)] is computed by
= i~ ~ tg TCS {r (_!s ~2~ s) H (x 0 s)
1 expansion of M in a power series, The numerical
-ioo
values of F ( u) are the following:

u = 0.01 0.03 0.1 0.2

F (u)= 10080 1137 107 28.8
x,
u = 0.3 0.4 0.5 0.75 1.0
00 F(u)=l3.7 8.2 5.7 3.3 2.2
+3~ e-·*Ko ( ~):1:.
x,
In the second integral in (VI), we must use the
The complex integral of the first component is com-
expressions M and K 0 for small x. This gives
pleted by the integral over an infinite semicircle
in the left half-plane. Then the entire integral is
expressed in the form of the sum of the residues of
the expression under the integral from poles located
to the left of the imaginary axis. These poles are
r'J xu-I
(Xy) dx = X~u -
In 4 ( 4 -
In yu u1 ) ·.
0
located at the points s = -l/2, -3/2, -5/2 and,
furthermore, at the point s = ..j {3 + 9/4. The resi- As u tends to zero, we obtain the results for an
dues at the first two points are reduced with the unbounded medium that we already know.
last two integrals in Eq. (IV). The residues at
the points s = -5/2, -7/2,-9/2 remain finite l W. Heitler, Quantum theory of radiation.
even for x 0 = 0. Therefore, we can replace x 0
2 G. N. Watson, Theory of Bessel functions.
in them by 0. All these residues give only a
3 E. T. Whittaker and G. N. Watson, Modern Analysis,
small contribution to the integral. The principal 4th ed.
contribution is made by the residue for
s=-../9/4+{3=-(3/2+u).(Ifthe bodyhas Translated by R. T. Beyer
dimensions that are not teo small, then u is a small 182
number.) For~, we get the expression