The classical Boltzmann equation of a gas of

diatomic molecules
Cite as: J. Chem. Phys. 75, 376 (1981); https://doi.org/10.1063/1.441792
Published Online: 31 August 1998

C. F. Curtiss

ARTICLES YOU MAY BE INTERESTED IN

The classical Boltzmann equation of a molecular gas

The Journal of Chemical Physics 97, 1416 (1992); https://doi.org/10.1063/1.463267

Non-equilibrium effects of diatomic and polyatomic gases on the shock-vortex interaction

based on the second-order constitutive model of the Boltzmann-Curtiss equation
Physics of Fluids 30, 016109 (2018); https://doi.org/10.1063/1.5009122

Kinetic Theory of Nonspherical Molecules. V

The Journal of Chemical Physics 38, 2352 (1963); https://doi.org/10.1063/1.1733510

J. Chem. Phys. 75, 376 (1981); https://doi.org/10.1063/1.441792 75, 376

© 1981 American Institute of Physics.

The classical Boltzmann equation of a gas of diatomic
moleculesa)
C. F. Curtiss
Theoretical Chemistry Institute and Department of Chemistry, University of Wisconsin-Madison, Madison,
Wisconsin 53706
(Received 16 February 1980; accepted 25 March 1981)

A classical Boltzmann equation and collision integral of a gas of nonvibrating diatomic molecules is
developed. It is shown that the generalization of the usual equation for structureless particles involves
additional integrations over the angular momentum vector of the second molecule and two angles associated
with the phases of the rotational motion of the two molecules. These latter angles are constants of the free
motion of the molecules defined in a manner analogous to that of the usual impact parameter, i.e., they are
determined by the free molecule trajectory as extrapolated to the point of closest approach.

The statistical mechanical theory of the nonequilibri- 8 =lim exp( - T,c (2) )exp[T(,ca + ,cb)], (3)
T-~

um behavior of low density gases may be based on the

Boltzmann equation or a generalization of this equation. where
The development of a generalization of the classical
Boltzmann equation of a gas of point particles to the .£
(2)
=.£a+.£b - "T (acp
"\'
aqla)
a acp
apla) + aqlb) apl b)
a) (4)
description of a low density gas of molecules with struc-
ture is discussed in an earlier paper l and a number of is the Liouville operator in the phase space of two in-
special cases are considered. In a later paper, Hoffman teracting molecules. The effect of this operator is to
and Dahler 2 have discussed a number of properties of transform the phase space coordinates of the pair of
the resulting classical Boltzmann equation and in par- molecules, first backwards to time along the actual tra-
ticular, the properties of the linearized collision in- jectories until the molecules are no longer interacting
tegral describing collisions between molecules of an and then forward the same distance in time along tra-
arbitrary type. In this paper, we consider the applica- jectories which would arise if the molecules did not in-
tion of the earlier results to the special case of non- teract. In the present discussion, we consider the
vibrating diatomic molecules and further reduce the special case in which the molecules are nonvibrating
collision integral to a form similar to that which arises diatomic molecules interacting through an arbitrary
in the spherical case by carrying out, explicitly, one "soft"potential.
of the integrations. This integration leads to an expres- The nonvibrating diatomic molecules are idealized as
sion which involves only information about complete pairs of mass points at a fixed distance from one an-
collisions. The result is a generalization of the colli- other. The configuration coordinates of such a mole-
sion integral describing rigid body collisions discussed cule may be taken to be the position vector r of the cen-
by Hoffman, 3 and is also the classical limit of the col- ter of mass, and the polar angles e, and cp of thE\ inter-
lision operator of the quantum mechanical treatments nuclear axis. The conjugate momenta are then p, Pe,
of Waldmann 4 and Snider. 5 and P",. and the Hamiltonian is
Let X=QI,q2,q3"'" PI,P2,P3"" represent the set of
phase space coordinates of a single molecule. Then the
1 Ip
H= 2m 12 + 2I1 ( Pe+ sin12e p",2)
2
, (5)
generalization of the Boltzmann equation, developed by
Curtiss and Dahler, 1 which describes the time evolution where m is the mass and I is the moment of inertia of
of the distribution function j(x, t) in the one molecule a molecule.
phase space is The Cartesian components of the angular momentum
vector J, associated with the rotational motion of such
a diatomic molecule are
J 1 = - Pe sincp - p", cote coscp ,
where
J2=PeCoSCP -p", cote sincp , (6)
.£ _ L ( aH _a_ _ aH _a_) (2) J 3 =p", ,
i aPi aqi aqi api
is the one molecule Liouville operator, H is the Hamilto- and the magnitude of this vector is
nian of one molecule, and cp is the intermolecular poten- p2 )1/2
tial. The subscripts and superscripts a and b are used J= ( p~+ sin"'2e . (7)
to distinguish the phase space coordinates of two colliding
molecules. The operator ,~ is defined by The angular momentum J is a constant of the "free mol-
ecule" motion, i. e., the trajectory of a single noninter-
acting molecule. This suggests a transformation of
a)This research was carried out under Grant CHE79-05685 variables from the phase space variables, e, cp, Pe' and
from the National Science Foundation. p", to a new set of four variables such that three are

376 J. Chem. Phys. 75(1), 1 July 1981 0021·9606/81/130376·03$01.00 © 1981 American Institute of Physics
 C. F. Curtiss: Boltzmann equation of a diatomic gas 377

constant along a free trajectory and the fourth changes
linearly with time. For this purpose, we introduce a
rotation R described by three Euler angles, el, (3, and
y, and an associated matrix with components Rw The
angles el and (3 are determined by the condition that
i. e., the rotation is such that in the rotated frame the
angular momentum is along the z axis. Since the inter-
nuclear axis is orthogonal to the angular momentum,
i. e., the component of the angular momentum along the
internuclear axis is zero, one may fix the third Euler
angle I' by the condition that in the rotated frame the in-
ternuclear axis is in the x direction, i. e., by the condi-
tion that
~j =Rlj ,
where ~ is the unit vector along the internuclear axis.
The "new" coordinates are taken to be el, (3, 1', and
J or symbolically, Rand J. It may be shown that the
Jacobian of the transformation is
a(9,cp,PB'P.) =-Jsin(3 •
a(J, el, (3, y)
It also follows directly from the classical equations of
motion that J, el, and (3 are constants of the free mole-
cule motion and for the free molecule motion,
y=J/I •
In terms of these variables, the free mOlecule Liouville
operator [Eq. (2) becomes
tion coordinates of both distribution functions and in ad-
dition, one may approximate rb by r. in the distribution
functionfb. With these two approximations,
sfafb=sf(p., J., t)f(Pb, J b, t)
= f(p~, J~, t) f(~, J~, t) , (16)
(8)
where for convenience we do not indicate explicitly the
dependence of the distribution functions on the position
coordinate r. which is a parameter of the integral. The
symbol J. is used to indicate the three coordinates J.,
el., and (3., i. e., the magnitude of the angular momen-
tum and the polar angles of the vector. The primes on
the linear and angular momenta indicate the effect of
the operator, 8, and are therefore the precollision val-
ues of these free molecule constants of motion.
(9)
The product of the distribution functions appearing on
the right-hand side of Eq. (16), would without the ap-
proximation just introduced depend on the position co-
ordinates ra and rb both explicitly and through the depen-
dence of the primed variables on the unprimed values.
Because of the translational invariance of the intermo-
(10)
lecular potential, however, this latter, implicit, de-
pendence is not on the variables independently, but only
on the relative pOSitions of the two molecules
(17)
The effect of the approximation is to remove the explicit
(11) dependence on r b • From Eq. (4), one finds that
(18)

J:,={~ • (12) and, thus, for the same reason this operator does not
I By involve the two coordinates independently, but only the
The generalized Boltzmann equation may now be written relative coordinate r. USing as independent variables
in terms. of the new variables. ra and r, one finds from Eq. (12) that

Let .e +.e
• b
1 a .-+g.-+-
=-p
rna
B
ar a
B 1(J -a+ J -
Br I • BYa
B)
b aYb '
(19)
f = f(r, p,J, el, (3, y, t) = j(x, t) (13)
where
represent the distribution function in terms of the new
variables. Then, making use of Eq. (4), the general-
ized Boltzmann equation [Eq. (1)], becomes
is the relative velocity of the two molecules. From
(:t +J:,a)fa= f (J:,a+J:,b-J:,(2»)(~f.fb)JbdJbdRbdrbdPb' these arguments it follows that
(14) (J:,a+J:,b-J:,(2»)Sf.fb =[g.~+.!.(J ~
where the subscripts on the distribution functions and Br I • BY a
coordinates are used to distinguish the two sets of mo-
lecular coordinates, and +J ~)-J:,(2)]f'f' (21)
b aYb a b'
(15) where
indicates an integration over the three Euler angles. f~ = fa(P~, J~, t) (22)
To simplify the expression for the collision integral and it is to be understood that the operator J:, (2) acts
on the right-hand side of the Boltzmann equation, we only on the implicit dependence of the product of the dis-
will assume that the distribution function depends pri- tribution functions which arises through the primed vari-
marily on the free molecule constants of motion. More ables. The primed variables, i. e., the precolUsion
specifically, we will assume that f does not depend on momenta, however, remain constant along a real tra-
y, i. e., on the phase of the rotational motion, and de- jectory and are therefore constants of the motion of the
pends only weakly on the coordinate r so that one may two molecule system. Thus,6 with this interpretation
neglect this dependence over distances of the order of of the action of the operator,
molecular dimensions. With the latter apprOXimation,
one may neglect the effect of the operator s on: the posi- (23)

J. Chem. Phys., Vol. 75, No.1, 1 July 1981

378 C. F. Curtiss: Boltzmann equation of a diatomic gas
Finally, since the left-hand side of Eq. (14) is assumed
to be independent of I' a' one may average the right-hand
side over this variable. Thus, the Boltzmann equation,
[Eq. (14)] becomes
m Pa ._a)
( ~at + ~ ar a fa --~f[('~
21T g ar + ~~ I a'Yb fa f 'J
I aYa + ~~)
One may now carry out one of the integrations on the
right-hand side of the Boltzmann equation by considering
integrals over trajectories in the two molecule phase
space. For this purpose, let b be an arbitrary vector
normal to g (so that b· g = 0) and let
1 .
Yb=lh+ TJbt
The values, b, ifJ a, and ifJ b, thus represent the values at
the point of closest approach of the two molecules, if
they continued along uniterruped free molecule trajec-
tories and t is the time measured from this point. The
angular momentum associated with the relative transla-
tion motion of the two molecules, the "orbital" angular
momentum, is
J r =lJ.rxg
=lJ.bxg,
where IJ. = ml2 is the reduced mass of the pair of collid-
ing molecules.
The two vectors g and J r may now be used to define
a rotation Rr described by three Euler angles ar' (Jr'
and Yr and the associated rotation matrix with ele-
ments, R~J)' We define this rotation so that g is in the
z direction
where g is the magnitude of g, and since J r is orthogonal
to g, J r may be taken to be in the x direction
J rJ =JR~J) ,
where J r is again the magnitude of the vector Jr'
Five of the variables of integration of the collision in-
tegral may now be changed from the set r, I' a' and Yb
to a new set of five: t, J" ifJa, Ij!b' and Yr' The Jacobian
of the transformation is J rl IJ. 2g and in terms of these
variables the operator appearing in the collision inte-
gral is simply a/at. Thus, the collision integral on the
right-hand side of the Boltzmann equation becomes
where
dR~O) =sinf3 bda bdf3 bdifJb ,
i. e., the rotation R~O) is that determined by the Euler
angles a b, f3 b, and Ij!b' and again the time derivative acts
only on the primed variables. The integration .over the
time is now readily carried out and one obtains the dif-
J. Chern. Phys., Vol. 75, No.1, 1 July 1981
ference of the products f~ f~ evaluated as t - + "" and as
t - - "". From the definition of the primed variables,
however, the latter product is simply fafb' Thus, the
generalized Boltzmann equation [Eq. (24)] becomes
I
b (:t +; Pa' a:Jfa = 21T~ 2 f (f~f~ - fafb) (JbJrlg)
(24) (33)
where the primes now indicate the precollision values of
the free mOlecule constants of motion determined by the
postcollision values P. and J a upon which the integral
depends, and the full set of integration variables.
From Eq. (28) it follows that the magnitude of the im-
(25) pact parameter b is JrllJ.g. It also follows from the
definition that the first two Euler angles of R~O) are the
(26) polar angles of the angular momentum vector J b •
Therefore, one may rewrite the collision integral on
the right-hand side of the Boltzmann equation in the al-
(27)
ternate form
1
21T f( fafb - fafb gbdbdYrdPbJbdJbdJbdlj!adlj!b
I ' ) A
(34)
where
dJ b = sin {3bdabd{Jb (35)
indicates an integration over the angles of the angular
momentum vector J b • Thus, this generalization of the
expression which arises in the case of point particles.
involves integrations over the angular momentum vector
(28) J b , and the phase of the rotational motion of molecule b
as described by the angle ifJb and averaging over the
phase of the rotational motion of molecule a as de-
scribed by Ij!a. This latter averaging arises from the
assumed lack of dependence of the distribution functions
on this angle, although, of course, the details of a colli-
sion do depend on ifJ. and Ij!b' The angles Ij!a and ifJb are
defined as constants of the free motion of the molecules
determined by the phase of the rotational motion as de-
scribed by the extrapolated value at the free molecule
(29)
distance of closest approach.
In the limiting case in which the molecules are rigid
bodies the collision integral developed above reduces to
(30) that obtained earlier, 1 except for the additional averag-
ing over ifJ a which is not introduced in the earlier devel-
opment. In a linearized form, this collision integral
leads to expressions for the generalized or "kinetic
theory" cross sections which agree with the classical
limit of those derived from the Waldmann-Snider colli-
sion operator. These results will be developed in later
papers.
IC. F. Curtiss and J. S. Dahler, J. Chern. Phys. 38, 2352
(1963).
(31) 2D • K. Hoffman and J. S. Dahler, J. Stat. Phys. I, 521 (1969).
3D. K. Hoffman, J. Chern. Phys. 50, 4823 (1969).
41.. Waldmann, Z. Naturforsch. Teil A 12, 660 (1957); 13,
609 (1958).
(32) ~. F. Snider, J. Chern. Phys. 32, 10.51 (1960).
Sorhe use of this invariance condition in the development of gen-
eralizations of the Boltzmann equation was apparently first
used by H. S. Green in his discussion of density corrections
to the Boltzmann equation [H. S. Green, Molecular Theory of
Fluids (rnterscience, New York, 1952), Chap. VIllI.