PHYSICAL REVIEW VOLUME 115, NUMBER 6 SEPTEMBER 15, 1959

Fourth Virial Coefficient for the Square Well Potential

SHIGETOSEiI KATSURA
Department of 3 pplied Science, Tohoku University, Senda~, Japan
(Received September 24, 1958)

The fourth virial coefficient D and its components D1, D&, D3 for molecules interacting with a square-well
potential have been obtained as functions of the temperature by the use of Fourier transformation and the
addition theorem of Bessel functions. Each component and their sum are nearly constant above the Boyle
temperature and tend to negative infinity as temperature tend to zero. The total D is expressed by
D/b'=0 28642+. 1 5397f . 23.55
4f +53 1ft 14.777fe
645f'+. 69 859f' .170.0
where f=e'ter
1, and well width is equal to the hard core diameter. Corrections of the critical data and
some related problems are discussed.

l. INTRODUCTION as follows: The multiple integral in the configuration

HE isotherm of the imperfect gas is expressed by
the virial series
pv/kT=1+Be-'+Ca '+Du '+
where p, n, and T are the pressure, the volume, and the
temperature, respectively. 8, C, and D . are the
second, the third, and the fourth virial coefficients,
respectively. Mayer's theory' of the imperfect gas shows
that the kth virial coe%cient can be obtained from the
intermolecular potential by evaluating the 3k-dimen-
sional integral whose integrand is composed of irre-
ducible cluster graphs.
Virial coeKcients already calculated are as follows:
Hard sphere up to the fourth (Boltzmann, ' Happel, '
Majumdars), square well potential up to the third
(Kihara'), Lennard-Jones potential up to the third
(Kihara'), etc. These are discussed in detail by Kihara'
and by Hirschfelder. 7
The fourth virial coefficient is often used as a criterion
for the validity and the self-consistency of an approxi-
method.
mation used in the distribution-function
The only exact value of D known is that of the hard
sphere, and we do not know how D depends on the
temperature when attractive forces exist. The reason
for this difhculty is that the method of carrying out
the multiple integration for the configuration in Fig.
1(c) has not yet been found. This paper reports a
method of calculation of the integration in Fig. 1(c)
and the values of the fourth virial coefficient and its
components D&, D&, and D3 in the case of the square
well potential. The method of the integration for D3 is
' J. E. Mayer and M. G. Mayer, Strttistical Mechanics (John
Wiley 8z Sons, Inc. , New York, 1940).
'L. Boltzmann, Amsterdam Ber. 477 (1899); Wiss. Abhandl.
von L. Boltzmann III, 658 (1909).
' H. Happel, Ann. Physik 30, 246 (1906).
3
R. Majumdar, Bull. Calcutta Math. Soc. 21, 107 (1929).
T. Kihara, ppon Sugaku-Buturigakukaisi 17, 11 (1943).
T. Kihara, Revs. Modern Phys. 2S, 831 (1953).
7Hirschfelder, Curtiss, and Bird, Jjrlolecular Theory of Gases
and Liquids (John Wiley 8t Sons, Inc. , New York, 1954).
' G. S. Rushbrooke and H. I. Scoins, Phil. Mag. 42, 582 (1951);
Proc. Roy. Soc. (London) A216, 203 (1953).
' K. Hiroike, J. Phys. Soc. Japan 12, 326, 864 (1957).
' A. E. Rodriguez, Proc. Roy. Soc. (London) A239, 373 (1957).
14i
space is transformed to that in t space by the Fourier
transformation. The factors of the integrand in this
space are decomposed by the use of the addition
theorem of the Bessel functions. Then the angular part
of the integrals can be evaluated separately and the
whole integral is expressed as a series, each term of
which consists of a product of three infinite integrals of
one variable. These integrals are calculated in the
Appendix. D& and D2 are calculated in a simple way.
The results of the present paper show that above the
Hoyle temperature the magnitudes of D1, D2, and D3
are approximately constant of comparable order and
approaches the value for hard spheres, respectively. At
D3 contributes mainly to D, and
low temperatures
becomes negative infinity of the order of O(es'"r) as T
tends to zero. This makes us suggest that the ring
approximation for the virial coeKcients might not be
good for both the high- and low-temperature regions.
The present method can be applied also to the higher
virial coefficients, but the applications are postponed
to another occasion.
2. CALCULATION OF D3
We consider a system of molecules whose interaction
is given by the potential(Fig. 2)
= ~,
U(r) r(rr
o. &r & go. (2.1)
=0, go &r
where o. is twice the diameter of the hard sphere, and g
is assumed to be 2, for the convenience of the calcu-
lation, since it is known that g=1.6 2.0 fits well with
D=D, + D, +
D,
1
8
(&) (b) (~)
FIG. 1. The fourth virial coe%cient D is composed of three
parts, each of which consists of diferent multiple integrations.
The integrands of these multiple integrations are expressed in Fig.
1 (a), (b), and (c), and the contributions of these integrals to the
total D are denoted D1, D2, and D3 hereafter.
 SH I GETOSH I KATSURA
For the square well potential given by (2. 1),

v(t) = (1+f)vo(t) Sfvo(2t),

'
FIG. 2. Square well potential. = (2~
'(cost )=
sint I;'
(t)
vo(t) (2. 10)
I

L~)
I I
E to to) I

where vo(t) is the corresponding Fourier transform for

the hard sphere.

experiments. "
We denote the position vector of the
ith molecule by r;, and define the function f(r) by
When we insert (2.8) for the fourth, fifth, and sixth
factors in the integrand of (2.6) and integrate over )oo,
)oo, and )o4, and use (2.9) and (2. 10), then

r&o
o. (r(go. Do= o~ol, " q
9/2

t ()t ()t (p4)v'")

gr.
(2.2) '(2 j I
J J
ger
X exp{i[t (eo )o4)+t (p4
)oo)
Let
r, r& o-lo i=2, 3, 4
+ t" ()oo ]}d)nod)nod)o4d )oo) td t'dt"
(2.3)
f(lrl) =h(l el) =h(p)
As is well known, the fourth virial coe%cient D is v(t) v(t') v(t") v(l t' t" I)
given by
J
D= Di+Do+Do, (2 4) Xv(l t" tl)v(l t t'I)dtdt'dt" (2. 11)
where = o ' 2
a&. . .a6
(1+f)"'( Sf)"'
1 1 XI( a,iao)ao')a4, ao)ao), (2.12)
r r, l)
J f(l
D4 ril) f
(l ro ril) f(l r4
8 V~ where

Xf(lro r4I)f(l r~l)f(l r~ r~l)dr, dr, dr4dr4, (2.5) I(ai, a2, ao, a4, ao, ao)

and V is the volume of the total system. The integrands

vo(ait)vo(aot')vo(aot )vo(a4I t t I)
of D&, D2, and D& are schematically expressed in Figs. J J
1 (a), (b), and (c). The exact definitions of Di and Do
are given by in (3.1) and (4.1), respectively. Xvo(aol t" tl)vo(l t t'I)'" (2 13)
(2.5) can be integrated with respect to the coordinate
of center of gravity and is expressed by The summation in (2.12) means the summation for all
64 cases where all a, take the values 1 and 2; and e~
Do= o~' " h(po)h(po)h(p4)h(l eo e4I) and e2 are the number of u; which take the values 1
and 2, respect. ively (Ni+44o 6). (2.
11) is reduced to
{(1+f)'I(1)1)1) 1,1,1)
Do//)To= -',

-3(1+f) (Sf) [I(211 1)+I(1 211)j

Let the Fourier transform of h(l pl) be v(l tl): +3(1+f)'(Sf)'[I(1,2, 2; 1,1,1)+I(1,1,1; 1,2, 2)
+2I(2, 1, 1; 1,2, 1)+I(2,1,1; 2, 1,1)]
v(t)=v(ltl)=I I h(lel) exp( it
'

e)de
(1+f)'(Sf)'[I(1,1, 1; 2, 2, 2)+3I(1,2, 2; 2, 1, 1)
L2~) J +6I(2, 1,1; 2, 1,2)+6I(1,2, 2; 1,2, 1)

i. .
" +3I(2, 1,1; 1,2, 2)+I(2, 2, 2; 1,1, 1)]
(2) / sintp
+3(1+f)'(Sf)'[I(1,2, 2; 1,2, 2)+2I(1,2, 2; 2, 1,2)
ph(p) dp, (2 7)
i , +I(2, 2, 2; 2, 1, 1)+I(2,1,1; 2, 2, 2) )
/'1'I
3(1+f) (8f)'[ I(2, 2, 2; 1,2, 2)+I(1,2, 2; 2,2, 2)]
h(p)=I(foal)=l I
)
v(ltf) exp(it ).)dt + (Sf)'I(2, 2, 2; 2,2, 2) }. (2. 14)
The polar coordinates of t, t', t" are denoted by
t'2)
(~) ,
' /'"
tv(t)
sinpt

p
dt. (2.8)
t, n,P; t', n', P', t", n", P", respectively, and their solid
angle elements are denoted by dQ, dQ', and dQ",
respectively (dQ=sinndndP). Now we use the addition
 FOURTH VI RIAL COEFFICIENT FOR SQUARE FAVELL POTENTIAL 1419

theorem of the Bessel function to decompose We substitute this into (2. 18):
vo(a 'I),
I

~, (aI -'I) (2. 18) =4m Q Q(2u+1) (2u+1)

u

J;(aI 'I) X P(cosy")P(cosy") dQdQ".

(a 'I)-: After the integration over dQ, only the term for u=m
J;+(at) J,*+(at') does not vanish. Hence
= 2'P (-', ) P (-,'+u) C''(cosy), (2. 15)
(at)'* (')'*
(2.18) = (4n)' g(2u+1) dQ"
where y, y', p" are the angles between t and t', t' and
'g

t", t" and t, respectively, i.e., = (4~)' g(2u+ 1). (2.21)

cosy= cosu cosa'+sinn sinn' cos(P P'), etc.
Since the nonvanishing terms in the right-hand side
C"(cos&) is the Gegenbauer polynomial and is ex- of (2. 18) are those for u=n=w, the nonvanishing terms
pressed in terms of Legendre polynomials by in Pi P Pin (2.17) are those for which I, m, and u
dP~y(cos't) are all even or all odd. Denoting Min(t, m, n) by M,
C:(cosy) = the upper limit in the summation of (2.21) is M, and
d cosp hence we have
u =s (2. 18) = (2.21) = (4w)'X ~ (M+1) (M+2).
(2u+1)P-( os') (2 16)
u =even or odd We use the result of the integration of the angular
u&0
part in (2. 15),
Then
aa) '
I( a, re,
a' 3a4, ae, ae)
I(a, , a, , a; a, a, ,a,) =[2lI'(-')]'(a a a) '*(a
= P' P' P' ' (M+1) (M+2) N , (ag, a5, a, )
X~
1=0
Z
00

~ n=0 m=O
00 00

(-;+I)(-', +m)(-,'+u)
f " J;(a t)
l m n

XA'ni(a~, ae, a4)Nr~(a3, a4, ar), (2.22a)

where
X J;(a2t) J,*(aat) J*,+t(a4t') J;+i(a4t")J;+(a5t")
N.(a, , a, , aI,) = 24.i[(-,'+m) (-', +u) $*'(a,a, ap) '
X JH. (an't) JH. (act)J;+(act')C&l(cosp)C '(cosp') XW;, ~ i l(a;, a;,aj,). (2.22b)
XC'(cosp )t 'dtt 'dt t 'dt dQdQ dQ . (2.17)
II pl ' (alqa'21 ' 'an)
n2 nn
With use of (2. 16), the angular part of (2.17) becomes
00
p
f J,g(a, t) J(a2t) . J(at)t"dt, (2.23)
I Ci'(cosy)C l(cosy')C '(cosy")dQdQ'dQ" JO

which is the generalization of Weber-Schafheitlin s

J J
X (2u+1) (2m+1) (2m+1)P(cosy)
First we consider
(2.19) vanishes for
discontinuous integral, and appears frequently in the
n =even or odd v =even or odd w =even or odd following sections. Pi'g 'P'
in (2.22) means the
summation taken for the points where I, m, e are zero,
all positive even integers, and all positive odd integers.
The necessary W;, H. + '(a, , a, ,ai, ) can be calculated
XP, (cosy') P(cosy")dQdQ'dQ". (2. 18)
~
from W~r;+, '(a, b, b) for a, b=1, 2. These integrals
are calculated in Appendix I. In the numerical calcu-
lation we truncate the triple in6nite summation
' P'
by the Rnite summation in which (t, m, u)
t P (cosy)P(cosy')dQ'. (2. 19) is taken to be (0,0,0), (0,0, 2), (0,2, 2), (0,0,4), (2, 2, 2),
(1,1,1), (1,1,3), (1,3,3), and their permutations. These
use. When u=v, it becomes values are listed in Table II of Appendix I. Then we
have
4m.
P(cosy") . (2 20) D,/b'= [1.2713 2.9763f+17.182f' 8.4200 f'
-',

2u+1 +20.190f' 35.434f'+14.640f'J (2.24)

1420 SH I GETOSH I KATSURA

where b=3mo' is four times the volume of the hard Applying the Fourier transformation to the fourth and
sphere. The 6rst term inside the square bracket is the fifth factors in the integrand, we have
value for the hard sphere and its value cited by Riddell
and Uhlenbeck" is 1.2728. The accuracy of each term Pi 1' t'
in (2.24) can be estimated from Table II in the Appen- h(p )h(p )h(p )v(t)v(t')
dix. [By adding more terms for (t,m, n) to Table II,
(g) J J
we can make the coeKcient of (2.24) as accurate as Xexp(i[t (ps
y4)+t' (p&
ys)7)dp&dg&dy4dtdt'. (4.2)
we desire. ] Integrating over y2 and g4, we have
3. CALCULATION OF Dz
f
so ~' ' '
r

h(p&)h(p4)h(l tise4l) h(ps)des [v(&)7'[~(&')7'

Xh(l p2 ps l)dp2dysdy4. (3.1) X exp[its (t t') ]dtdt', (4.3)

Applying the Fourier transformation to the third and
fourth factors in the integrand, and integrating over y3,
we have the 8-function and (3.1) is transformed to
h(p )de b (t)7' exp(it. 9 )(4. 4)

This is equivalent to the result of Rushbrooke and

h(P )h(P )~(t)V(t') Scoins. '
i2~) &
Inserting (2.9) for y(t) and using (2.8) for the
Xexp[it (ps p4)+it' (ps ps)]dpsdpsdg4dtdt' integration over t,
r" (
Ds= l(So')(2~)' '

p'dp '
l
(1+f)~ (&)

= s (32m')o'J (y(t))'Pdt. (3.2) J;(2t) p ' sintp

+8f tdt
0
(2t)' i I

p
This is the result obtained by Montroll and Mayer. '2
Inserting (2.9) for y(t), we have p2 l" ( J;(t)
+f PdP l (1+f)
Di= s'(327r')O' P (1+f)"'(Sf)"'(aiasasa4) &o

X W;, ;, *(aa,, aa4), (3.3) +Sf

J;(2t)
*

(2t)l
) ' sintp
]
l
p
tdh (4.5)
P means the summation for all 16 cases where ai, as,
a~, and a4 take the values 1 and 2, i.e.,
= 4s32o'~4 (1+f)'W **(1,1,p)
Di
pdp ~
-', (32s4) a'[(1+f)4W~,
;, *;4(1,1,1)11) 0
4(1+f)'Sf2='W;, ;, ;, *4(1,, 1)1)2)
16
+
(1+f)'(Sf)' 'W-:. :.:, 1, , ) (, (1+f)fW-:, :. (1,2,P)
4(1+f) (Sf)'2 'l'W * ~,*'(1,2,2, 2) 2
+
(8f)'2 'W. ;;
a'(2, 2, 2, 2) 7. (3.4) 2 p2
The necessary W;, **,, ,*(a&,a&, as, a4) values are given in +Sf'W:, -*.
,
l'*(2, 2,P) +f PdPL 7', (4 6)
Appendixes II and III. Using these values, we have i

Di/b' = (1/560) [544 4075f+ 35 007f' where the content of the second square bracket is the
99 687f'+139 215f47. (3.5) same as that in the first. W,*, ;, tl(ai, as, p) is given in
The first term, 544/560, is the value for hard spheres
Appendix IV for the cases where a, and 2. From =i
this we have"
and has been given by Rushbrooke and Scoins. '
= (1/4480) [6347+ 27 369f 156f'
Ds 'o'
, .
4. CALCULATION OF

h(p2)h(ps) h
(p4) h( l
D2

ys
pg l )
D&/h'
+594 272f' 1 518 980f4+918 540f'7 (4.
The first term, 6347/4480= 1.41670, is the value for
184
7).
Xh(I es es l)desdesde4 (4 1) hard spheres and agrees with the value of Rushbrooke
and Scoins. 8
"R. J. Riddell and G. E. Uhlenbeck, J. Chem. Phys. 21, 2056
'8 K. Hiroike has informed the author that he has obtained the
(1953).
"E. W. Montroll and I. E. Mayer, J. Chem. Phys. 9, 626 same results as (3.5) and (4.7) for Di and D2 independently
(1941). (private communication).
 FOURTH VIRIAL COEFFICIENT FOR SQUARE WELL POTENTIAL 1421

5. CONCLUSIONS AND DISCUSSIONS Bx2

In the preceding sections the integrals D~, D2, and D3

have been obtained as functions of the temperature
variable f=e'I" 1
in (3 5), (4.7), and (2.26). The suni

the third '

of them gives the total D. We summarize the second,
and the fourth virial coefficients of the,
square well potential (g=2 in Fig. 1) and the fifth
virial coeKcient of the hard sphere.
B/b=1 7 f, (5.1) -20

C/b'= (5 17f+136f'162f')/8~ (5.2)

D/b'=0. 28642+1.5397f 23.554f +53.645 f'
+69.859f' 170 01f' . 14 7
77 f' . (5 3)
E/b4 = 0.1 15+ (5.4) -10
0.2 0.4 0.6 0.-8 1.0
The terms of the lowest order are the hard-sphere I i I l I

X
values.
The value of D for the hard sphere obtained by
-10
Boltzmann' and by Majumdar4 is'4

D 1283 3 7342+81&(17(arctan&2 ~/4)

+
b' 8960 2 32X35~
FIG. 4. Temperature dependence of Bx Cx D1x' D2x'2 D3x'2.
=0.2869, (5.5) D=D1+D2+D3) x=e ''

0.288. E/b'
and that obtained by HappeP is is the
Rosenbluth~ "
value for the hard sphere obtained by Rosenbluth and
using the Monte Carlo method.
The temperature dependence of D~, D2, D3, and D is
shown in Fig. 3. As the unit of temperature we choose
the Boyle temperature T&, at which the second virial
-0.9715 coefficient vanishes. Above the Boyle temperature D&,
D2, and D3 are approximately constant and approach
the values for the hard sphere, and the first and the
third are negative while the second is positive. Their
absolute values are of the same order of magnitude and
the contributions to D from these three cancel for the
1.4167 most part. Below the Boyle temperature each of them
becomes negative and approaches negative infinity as
D2 T tends to zero, being of the order of e "~' e T "I'
2
I
3
I
and |, "I~, respectively. Such a feature of the temper-
ature dependence of the fourth virial coeKcient and its
components will also occur in the case of the Lennard-
Jones potential.
~-0.1589 In general, the eth virial coefFicient tends to ~00
D3 &n(n 1) e/2kT 2n(n 1) Hence g& g& D & D
and D3x" for x = e ""r (f=x ' 1) are sho
wn in Fig.
4. Figure 4 shows the contribution of each component
0. 5 of D to the total D, especially at low temperatures.
The D3 contribution is the main part at low tempera-
tures. Such a contribution of D3 to D makes us suggest
that the summation of the ring approximation" of
irreducible cluster integrals, which is equivalent to the
solution of Green's linearized integral equation, is not
FIG. 3. Temperature dependence of D1, D2, D3, and D. The unit
of temperature is the Hoyle temperature. a good approximation over the whole temperature
region. For example, it can be proved that the isotherm
"Boltzmann and Majumdar used the result by Van Laar
I Archives du Musee Joyler ser. 2, 6, p. 237 (i900) =Amsterdam '5M. N. Rosenbluth and A. W. Rosenbluth, J. Chem. Phys.
Alrad. Versl. Jan (1899), p. 350j. 22, 88i (1954).
1422 SH I GETOSH I KATSURA

FIG. 5. Normalized
second, third, and fourth
cluster integrals for the
square well potential.
b4* was calculated using
the value of D in the
present result. This fig-
ure was calculated by
T. Kihara.

obtained from the ring approximation for the square tive comparison, however, wil] depend on the precise
form of the potential.
well
obtained by Rodriguez. "
potential has a singularity similar to the one
The origin of such a singu- The discussions of the distribution-function' "
larity in this isotherm may be attributed to the neglect method often use the comparison of the true D and
of other irreducible graphs. As a criterion of the validity approximate D. One should, correctly, use the compari-
of the ring approximation and similar ones, one should son over the whole temperature range, rather than only
show that the isotherm behaves well at low tempera- the value for hard spheres.
tures. The method of Sec. 2 of the present paper can be
In the case of the lattice gas' the alternative signs applied to D~, giving the result
of the virial coefficients appear. In the present results,
D2 = ' (64m. 4) 0'
8, C, and D are negative at low temperatures. It is an ,
at o ~ ~
a5
interesting question whether the higher virial coeffi- n1+m2 =5
cients E, F, . . . at low temperatures are all negative
at ay=i or 2

as 8, C, and D, or not, since the negative signs of the

virial coefficient below the critical temperature is
X (C18203Q4) *Gs Q (2+2'I)
n=o
assumed in Mayer's theory of condensation.
Kihara-Kaneko's proposition" that the contribution X W;, ; '-+2
, , *(+1,+2 Q5) W,'-, '-,
, '+~
,
, '((Lg ll4 85) ~ (5.6)
of irreducible cluster integrals P~'s for the system of The value of D~/b' for hard spheres from the sum of
g= ~ is a good approximation to the calculation of the the 6rst three terms in Po"
is 1.41618, which agrees
higher cluster integrals b~'s for g=2, is valid at high well with the exact value of (4.7), showing the good
temperatures, but this approximation is not sufficient convergence of this series.
at low temperatures. Figure 5 shows Kihara's normal-
ized cluster integrals. " Moreover, this method can also be applied to virial
coefficients higher than the fourth. In particular, for
The critical data obtained from the second and the some types of irreducible graphs, a certain simple rule
third virial coefficients for the square well potential' has been found which enables one to write down the
(g=2) (po, &O, To) and their corrected values under con- integrand in t space corresponding to (2.13). The
sideration of the fourth virial coefficient (p~, v~, T~) are details will be discussed at another time.
shown in Table I. The critical temperature decreases
9%, the critical volume increases 36%, while p~vj/kT~ ACKNOWLEDGMENTS
increases 9% from povg/kTO. Such a detailed quantita-
'6 A. E. Rodriguez, Proc. Roy. Soc. (London) A196, 73 (1949).
The author wishes to express his sincere gratitude to
' S. Katsura, Progr. Theoret. Phys. (Kyoto) 20, 192 (1958). Professor Y. Nomura and Professor T. Kihara for
"T.Kihara and S. Kaneko, J. Phys. Soc. Japan 12, 994 (1957). valuable discussions.

TABLE I. Correction of the critical data of zeroth approximation by Kihara' due to the fourth virial coe%cient D.
The third line indicates the comparison of the various temperature variables, e/kT, T/T~, f, and g.

Pb/kTa pv/kT
Zeroth approx. (from 8 and C), Kihara 0.06656 1.9184 0.3333 0.3485 0.3831 0.41688 0.8403
First approx. (from 8, C, and D), present results 0.04865 2. 6086 0.3629 0.3819 0.3497 0.4654 0.8261
Hoyle temp. 0.13353 1. 0, 14285 0.9351
& See ref'erence 5.
 F 0 IS R T H VI RIAL C 0EFFI CI EN T F0R SQ UA RE KV ELL P0TFN TIA L 1423

APPENDIX I to the first factor in the integrand of (A. 1). Here is I

lV ;+i,.tg, + . '(a, b, b) the straight line parallel to the imaginary axis crossing
at s= 3. We insert (A. 2) into (A. 1) and shift the
p GO
integration path to I.', which is the straight line parallel
1;+ (at) I;+ (bt) J;+(bt)t ddt. (A. 1)
"0 to the imaginary axis crossing at s= so= ', -,'l e,
We use the integral representation of the Bessel function and change the order of the integrations J'ds and
J dt ( l m ~n & so &-,' ,'l). The integration
I'( s) (at ) ~r'+"
', ',
~ 2
1
J;pi(at) = ds, (A. 2) over dt is the Weber-Schafheitlin s integral, and it
i. I'(-,'+l+1) & 2 )
I I

&
2vri gives

1 1 p
ay-"+'1 r pay"
(A. 1) = b'I ! [
dsl
2 Qn- &2b) 27ri~ z, (2bl

I(-s)l'I ( -s+-
2 (
1

23
- -2j
( -s+1- -l q i~l p s+ 3+m+n+l
l q
l~l
2 )! &
~

(A.3)
m+n 1+2) |' m+n i+5) ( m+n 1+2)
Ii'I s+
l~l -+ ) & 2 )
Since a/2b&1 for a, b=1 or 2, the integration can be the integrand, and only s=0 remains for m=e, and no
obtained from the residue of the poles in the right plane pole remains for n&m The p. oles of I'( s+2) cancel
of L those of the third factor of the denominator and
The necessary integral can be confined to the case s=~i, 3~, -+
3~2m+-', nremain. The poles of I'( s+1)
where m and e are both even or odd and e(m with cancel those of the fourth factor of the denominator and
l=0, and to the case where l and m are both even or s=1, 2. . .-', m n remain for m&n+2, and no pole
',

odd and l(m with n=0. remains for m=e.

First we consider the former case: The poles of I'( s) The sum of the residues can be expressed by the
cancel those of the second factor in the denominator of generalized hypergeometric function:

t
m+n+3q
1(l)1
)
I

2
(A. 1) = al
4(2~) l m+n) mn)
r ( 1+
(m+n+5q
I(-:)Il 1+

fm+n+3 m n m n 3 m+n r ( ) r (-', ) (a/2b)

-',

l, o, l; (a/2b)' I+
2 2 2 2
;
j ( m n+1 ) (m n+1$
r(3)rl irl
j 2 ) I

(m+n+4 m n+1 m n
2 +n+1
m
x4Fai ; ', -'3; (a/2b)'
,
I

2 2 2 2

I ( 1)I'( :)I
]m+n+5 y )a p '
l

2 j (2bj
m+ny )m +n+3y pm ny
r(-,')r Ir il'
) &2)
m+n+5 m n+2 m n 1 m+n+2
x4pal
I
; 2, ';
-'-, (a/2b)' . (A.4)
2 2 2 2 ) I
 SH I GEYOSH I KAYSURA

TAHrE II. The values oh IC, (ab, c)= 4

(2~)&Wt, t+~ t+k(ab, c)=4(2
pr)& JPJt(at)Jt+(bt) Jt+~(ct)t ~dt a, .b, c=1 or 2.

(a, b, c) (2) (122) (222) {121) (212) (112) (221)

EQQ 5/12 2v2/9 185/288 2&Epp (111) Epp(211) (122)

Epp E'pp(211) Epp(122)
E'22 25/1344 0 2649/57344 2&Epp (111) 2'/84 241/43008 E(121} E22 {212)
7/96 0 117/1024 2v2/12 31/768 0
+2Q
&P2
E4Q
E2p(111)
11/960
Epp(211) = 0
0
X Q(122)
549/40960
2' p(111)
2&@

2&E
Q(111)
p(111)
Epp(112) =0
242/96
Epp(122)
31/7680
Epp(121)
0
Epp (122)
z2Q(212)
Epp(122)
EQ4 E4p(111) E4p(211) =0 E4p(122) 2', (111)
o E4p(112) = 0 Epp(122) Epp(121) Epp(212)

Egg 53/480 0 1503/5120 2&Eu (111) 242/30 709/7680 Egg(121) EII (212)
Epp 35/6912 0 10251/131072 2&Epp (111) 2@2/216 3883/442368 Epp (121) Z(212)
Egp 17/960 0 1257/40960 2VCI3(111) 0 333/20480 2v2/30 77/15360
En Erp(111) E&p(211) =0 Egp (111) 2&Egp (111) Egp (112) Eip(221) E,p(121) =0 Egp(212)

In the above 4Fs(nr, np, np, n4, pq, ps, pp, x), when the over dt (Weber-Schafheitlin's integral), we have
number of n; which are equal to zero is larger than that
of P, which are equal to zero, &Fp1. When there exists =
1/r (zero or negative integer) in the constant factor,
this term is put to be equal to zero. 1
The integral in the latter case (v=0, t and m are X (A. S)
both even or odd) can be obtained in the similar way. ', ) s (s
(s+3) (s+ 2) (s+ ') (s
', ) ,

The results are summarized in Table II.

According as a~b or u~b, the integral is given by
the sum of the residues of the poles in the right or left
W, ~~"(a,b, c) =
'
E
(a,b, c). half-plane, respectively.
4(2pr) '
a' 4 4 pbbs 8 hbq'
APPENDIX II 2pr 27 35 (aI (a)
567
(A. 9)
Wb, ;, 1, ~4(a, a, b, b) s 4 (ay' 4 ~a~'
8 ~ay
I
I+35 Eb)
I I I I, a~b,
567 Eb) 27 l b)
J;(at) J;(at) J;(bt) J;(bt)t-'dl. (A.5)
eJ p

We use the integral representation of the product of

Wb, 1, ;, ; (2, 2&1, 1) =-1 263
2m 2835
the Bessel functions to the 6rst and the second factors
in the integrand; then
W~, , 4 p (1,1,1, 1) =- 136
1
(A. 10)
~" - ( )"(/2)'+' (m+4)
2vr 2835
J;(bt) J;(bt) t-4dh
"p m=p m! {r(m+-;)) Wp 1 1 p (2, 2&2, 2) =-1 1088
2m. 2835
Iiao ~ ( ) mr (m+ 2) (at)p+sm
Z=o APPENDIX III
m!r(m+-,s)r(m+4)
X J;(bh) J;(bt)t 'Ch. (A. 6) W. ,
.. .-'(1,2,p) = J*,(t) J;(2t) J;(pt)t
lCh,

The summation for m is replaced by the integral (0~ p~2). (A. 11)
representation with the residue of r( s), We express the Bessel function J(t) by the product
1 1 ~ r ( s)r (s+2) a'+" of the exponential function and the polynomial:

!."
4$-
2~i +~~, i (2)p (1+-iq
:. '(12t)=
r(s+-,')r(s+4) 1
lid-. , *.
F '
e"I
16 Epr)
I l

(2p)'* & 4 t)
I

X Jp(bt) J;(bh) t'dt, (A. 7) il iq s)

e"') ( 1+
t'
+e "]f 1 lye "'] 1
t) 2t)
where I.
is the straight line parallel to the imaginary

axis crossing at s= 3. Carrying out the integration X fe"& e "&)t ldh, (A. 12
)
 FOURTH VI RIAL COEFFICIENT FOR SQUARE WELL POTENTIAL 1425

where I' means the principal value. Let it=s; then values of the above integrals are given by the integrals
over a semicircle of in6nitely small radius, i.e. , J~/2 (8
WI I., I*'(1,2, p)
and J'/2 "d0, respectively. When we expand the
integrand in Laurent series beginning from s ', the

8 (m. ) (2p)'* &;E dss ""+"'I( 1 +

2s 2s2)

1
1+
I' 1
+I' ds s-'e&'+&'I FIG. 6. Integration
2s 2s') paths in (A. 14).

+I'
/"
dss 'e~ '+&'I ( 1
1
1)
2s 2s'i
{a) {b)

+I'
3
'+'~'I ( 1+
1)-
dss 'e& + 2s')
I (A 13)
coeKcients of the negative even powers of s do not
2s
appear and the integrals of the negative odd powers of
s vanish since
In each of the above four integrals the pole of the
integrand is only s=0. The third integral for p~1 and
the first and the second integrals vanish when they are 2n lgs
integrated over the left half-circle and imaginary axis e 2nqd + 0
indented the origin LFig. 6(a)]. The third integral for
p&1 and the fourth integral vanish when they are
integrated over the right half-circle and the imaginary Therefore the terms contributing to the integrals are
axis indented the origin Fig. 6(b)]. Hence the principal
I only the terms of order s '. We pick up these terms; then

1 (2l i ds /1 (3+p)' 3 (3+p)' (3+p)')

*(1,2, p) =
-*
t'
W-:, :, +
3! 2! )
I I
I
16 (m) (2p)'* &cqs E2 4! 2

t ds ( 1 (1+p)' 1 (1+p)' (1+p)') !. ds ( 1 ( 1+p)' 1 ( 1+p)' ( 1+p)')

~cps ( 2 4! 2 3! 2! ) 4c, fop p~1$ E 2 4! 2 3! 2t )
C2 for p ~1

dst'1( 3+p)' 3+p)' 3+ )'p)

t
+-3( +
(
(A 14)
~c2s E2 4t 2 31 2! ) I

where C and C2 are the semicircles around the origin more suitable. The results are
with in6nitely small radius, /23
2
and J'/2 J'
respectively. W;;, ,*'(a,a, p)
Hence we have

W*, , (1,2,p)
Vp
I
-+ I,
(4p 1 (pt')
(3 a 12 (a) ) I I
0~ p&2a. (A. 17)
2(2m)&

1 1 1
30p'+72p 27), 1~p~ 2, =0) 2a~ p.
(p4 (A. 15)
96 +~ Qp
APPENDIX IV

W*,, ,*, ;.;.'(a, 1,1,1)

,
= J.;(at) J;.(t) J;(t)J;(t)t 'dt. (A. 18)

W.; ,*. ;, '(1, 1,p) and W;, I *, '(2, 2,p) can be integrated by A method similar to that of Appendix III can be
this method. The method of Appendix II, however, is applied and gives
 SH I GETOSH I KATSURA

(3+a)'1 (3+a)' ( 3) (3+a)'

(A. 18) =
1 1
-+ 1 I+ 3 1q I+ (3+a)'
t 3+-3)I+ (3+a)'!t'
6! (
I

8! ( 7! ( a)
I

4~ ga 9! a a)
I

a) 5!
(1+a)' 1 (1+a)' ( 1) (1+a)' ( 1) (1+a)' ( 1) (1+a)'
-+ 1+- I+ -1+- I+ -1--I+
7! ( 6! (
I I
9!, a 8! a~ & aJ a&
I
5!
(u
-+, -1+-1) (a 1)' ( 1)I+ (a 1)' t' 1)I+ (a 1)'
1)' 1 (a 1)'!
I+, -1-- I1--
6! ( a)
I

8! (
I
9! a aP 7! ( gJ 5!

(g 3)' 1 (g 3)' (
3) (a 3)'
3) (a 3)' ( 1) (a 3)'
9! g
-+ 1
8! & a)
I+ (3
I

7! E a)
-I
6! 4 a~
3 I+
5!
I I+, (A. 19)

The double sign before the third L ) should be taken as + for a~1 and ss for g~1.
1 1 625i V2 92051
;

W.~~;, (2, 1, 1, 1) = IV.;, ;, 1, ;4 (1,2, 2, 2) = 8W;, ;, ;, ;4(-,', 1, 1, 1) = (A. 20)
v2 m 2'X2835 ~ 2835X2

PIC YSI CAL REVI EW VOLLJM E 115, NUM B ER 6 SEPTEMBER 15, 1959

Aggregation of E Centers in Potassium Iodide

D. H. GooDE AND P. A. ScIIRQEDER
I'hysics Department, University of Canterbury, Christchurch, %ex Zealand
(Received April 23, 1959)

As optical aggregation of F centers proceeds in KI the E, M, and E bands develop, but later lose their
identity in a very broad band, which is absent in KCl and KBr. The shape and position of this band are
temperature independent, and it is removed from the peak of the true colloid band. A new band on the
short-wave]ength side of the F band was observed in crystals containing high concentrations of F centers.

A SERIES F of measurements on the optical aggre-

gation of centers in KCl, KBr, and KI indicates
that the final product of aggregation at room ternpera-
ture in KI is markedly different from those in KCl
and KBr.
Crystals of these materials were grown by the
Kyropoulos technique in an inert nitrogen atmosphere
from Analar reagents, and were suSciently free of heavy
ion impurities to possess no absorption bands in the
range 230020 000 A. Optical samples cleaved from these by the temperature of measurement,
crystals were additively colored with metallic potassium,
quenched from 450'C to 20'C, and irradiated with
light in the F band at i6'C. At various stages of this
illumination the crystal was cooled to liquid air temper-
with the R, M, and 1V bands superimposed upon it.
As the F-band irradiation continued the jV, iV, and l7
bands increased to a maximum and then decreased,
until finally only the broad band remained. KBr and
KC1 crystals do not exhibit this broad band when
treated in a similar way, but it is possible that this
band is similar to the R' band previously formed in
KCl by Scott and Bupp' by thermal treatment. The
shape and position of the broad band is scarcely affected
and its peak
TABLE I. Positions and half-widths of bands in ev at 77'K.
Band R1 R2

ature and the absorption spectra measured. KC1

The results of measurements on a crystal of KI are Energy 2.29 1.89 1.70 1.53 1.28
Half-width 0.23 0.11 0.08 0.08 0.12
shown in Fig. 1. The essential points in these curves KBr
were reproduced with several other crystals and F Energy 2, 05 1.69 1.54 1.37 1.13
center concentrations. Freshly colored crystals exhibited Half-width 0. 14 0.95 0.95 0.145
KI
an F band and small R, 3f, and A" bands superimposed Energy 1.87 1.49 1.36 1.21 1.04~0.02
on the general background of the pure crystal. When Half-width 0.21 0.12 0.08 0.08 0.09&0.02
Ivey's predicted energy 1.73 1.49 1.37 1.24
the crystals were irradiated a very broad band appeared
*Now at Division of Applied Chemistry, National Research
Council Laboratories, Ottawa, Canad@, ',A. B. Scott and
, , L.t P. Buppt Phys. Rev, 79, 341 (1950). ,