P 8 YSI CAL R EVI E%' VOLUME i59, NUMBER 5 JULY 1967

Comyuter "Exyeriments" on Classical Fluids. I. Thermodynamical

Properties of Lennard, -Jones Molecules~
Loup VzzLzrt
Bdfer Gradlaie School of Sciercce, Feehioa UNioersicy, Pm Fork, Peto Fork
(Received 30 January 1967)

The equation of motion of a system of 864 particles interacting through a Lennard-Jones potential has
been integrated for various values of the temperature and density, relative, generally, to a Quid state. The
equilibrium properties have been calculated and are shoran to agree very vreH vrith the corresponding
properties of argon. It is concluded that, to a good approximation, the equilibrium state of argon can be
described thlough a t&o-twdy potential.

I. INTRODUCTION results, summarized in Table I, are suf6cicntly numerous

'HK "exact" machine to allow a comparison on the whole density range of
~
relative to
computations
thc Auld state Rnd on R wide tcIDpcrature IRngc which
clRsslcRl Aulds hRvc ScvcI'Rl RlIDs: It
18 posslblc
essentially exclud. es the extremely high temperatures.
to realize "experiments" in which the intermolecular
The ovcr-all agreement is surprisingly good. It appears
forces are known; approximative theories can thus be
that the many-body forces, if they RIc at all important,
UnRmblguously tested Rnd soIQc guidelines RI'c provldcd
behave so Rs to rcRllzc RD elective lntcI'Rcflon which ls
to build such thcoI'lcs whenever they do not cxlst. Thc
state independent to a good approximation.
comparison of the results of such computations with
The correlation functions are described and, discussed
real experiments is the best way to obtain insight into
ln R scpR1 atc pRpcl. Thc formalism ncccssRI'y to cxpI'css
the interaction between molecules in the high-density
the Quctuations in the microcanonical ensemble was
states. discussed recently. s It can be applied to calculate the
The Monte Carlo method. inifiated by the I.os Alamos derkvatnres of the thermodynamtc functions (e.g. , the
group' is a 6rst example of these "exact" methods. It specific heat and c)P/c)p) in terms of fluctuations
amounts to a d, irect computation of the integrals in-
averaged, over time. The results are not very precise
volved ln thc CRnonlcRl RvcI'ages. It Is cRsy to CRrry out& and will only be presented as an illustration of these
with thc lncoIlvcIllencc) howcvcl ) of pl"ovldlng no theoretical conslderatlons.
information on the time properties of the system.
The results on the time-dependent properties wiH be
Thc dynamics of Rn lsolRtcd systcIQ CRn Rlso bc reported later.
considered and used to calculate time averages and
time-dependent properties. The case of hard. spheres
Rnd hRrd spheres surroUndcd by R Square wcH hRS bccn
'
extensively studied by AMer et al. In the case of a two-
body interaction simulating more closely the interaction %e consider a system of 864 particles, enclosed. in a
between the molccules, it is possible to integrate directly cUbc of side L, with periodic boUDdaIy condltlons
thc cqURtlon of thc motions of RboUt R thoUsRnd interacting through a two-body potential of the
particles, as brilliantly demonstrated by Rahman. ' The Lennard- Jones type
present paper presents some of the results which have
been obtained, using a technique inspired by Rahman's
work, for R system of 864 particles interacting through This potcntlRl ls cut Rt f'1, = 2.50' ln Dlost of oui cxpcll-
a Lennard- Jones potentiaL ments, or, in some of them at r, =3.30. The problem is
In Sec. II we give some tcchnical details on the to intcgrRtc thc cqURtlon of lnotion
method which we use; in particular, we describe a book-
d2r'
keeping device that cuts the computing time by a factor
of the order of 10.
rN =Q f(r;;).
jets
In Sec. III we give and, discuss the results obtained
for the pressure, the internal energy, the high-frequency Kc choose the following units: The lengths are ex-
elastic moduli, and the isotopic separation factor. These pressed. in units of o (a =3.405 A for argon), and the
energies in units of e (a=119.8'K for argon). The
~ Supported
by the U. S. Air Force 0$ce of Scienti6c Research thermodynamic quantities will thus be measured in the
Grant No. 508-66.
)Permanent address: Faculte des Sciences, Laboratoire de
Physique Theorique et Hautes Energies, Batiment 211, 91-0rsay. 4 L, Verlet
(to be published).
' %.%. Wood and F. R. Park. er, J. Chem. Phys. 27, 720 (1957), ' J. L. Lebowitz and J. K. Percus, Phys. Rev. 124, 1673 (1961).
~ S.
$. Alder and X'. E. Wainwright, J. Chem. Phys. 33, 1439 6
$. L. Lebovatz, $. K. Percus, and L. Verlet, Phys. Rev. 153,
(1960). 250 (1967).
3 A. Rahman, Phys. Rev. 136, A405 (1964). ' A. Michels and H. Vhjker, Physica 1S, 627 (1949).
 COM PUTER ''EXPERIMENTS'' ON CLASSICAL FLUI DS
usual "reduced" units. The time unit is chosen so that With this device, the time spent for an integration
m=48&0. ", it turns out to be, for argon, equal to step at the density 0.45 is about 12 sec on the UNIVAC
"
3)(10 sec. This time is of the ord, er of the kinetic 1107 of the Faculte des Sciences, Orsay, where the first
relaxation times of the system in the case considered in calculations were made, and ten times less (with careful
this paper. With this in mind, we have, for the force machine coding of the time-consuming subroutines) on
acting on particle i in the x direction, the CDC6600 of New York University, where the

f, (r;,) = m(x, —x,) (r;; "—

greatest part of the results reported here were obtained.
0.5r;; ) (3) With that machine, a typical "experiment" takes
To integrate (2), we use the very simple algorithm about 1 h. It goes as follows: The positions are initially
taken, in general, at the nodes of a face-centered, -cubic
r, (t+h) = —r;(t —h)+2r;(t)+p f(r,;(t))hs, (4) lattice which has the desired density, and the velocities
are chosen at random with a Gaussian probability law.
Three hundred, steps in time are sufFicient, in general, to
where h is the time increment which we take equal to
reach equilibrium. The computation is then carried on
0.032. This is practically the value chosen by Rahman for 1200 steps in time (this corresponds, for argon, to
(i.e., 10 '4 sec in the case of argon). We have checked 1.2X10 "sec). The main part of the computation con-
that this time increment is adequate and even super- cerns the study of equilibrium quantities (thermo-
Quously small in most cases. For instance, for T= 1.38,
dynamic functions: temperature, pressure, internal
p=0. 55 (i.e., temperature just above critical, density energy, specific heat, etc. ; time-ind. ependent correlation
almost twice critical), we have performed two integra- functions) and of nonequilibrium quantities (velocity
tions up to the time t=4. In one case we have taken autocorrelation function, elastic constants, viscosities,
h=0. 032, in the other k=0.016, with the same initial heat conductivity, etc. ). The necessary technical details
conditions. The di6erence in position at time t is will be given when these results are reported.
typically of the order of 0.001 and the difference in the
thermodynamic quantities, at the same final time, III. THERMODYNAMIC QUANTITIES
amounts to 1/10 000. In this kind of calculation, most
of the time is spent in computing the force. If no special A. Temyerature
devices are introduced, we must, at each step, compute
At each step in time the velocities are calculated
'$(tV —1) terms, most of which turn out to be zero.
—,
simply by the formula
We introduce the following bookkeeping device which
cuts the computing time by a factor of the order of 10: v;(t) = $r;(t+h) $/2h—
r, (t h)—
Every nth step, we compute all the -,'X(X—1) distances,
The temperature is, at time t, -', of the kinetic energy,
and, given a particle i, we make a table of all the
in our units
particles which are within a distance r~ of that particle.
—
Then, for the next n 1 steps in time, we take into T=16 Q s'/E.
account only the particles in the tables. There is no
error as long as r~ is sufFiciently larger than r„so that
no particle outside the table traverses the "skin" of The error entailed by the use of (6) is of the order of
—
depth r~ r„and gets into the range r„of the potential. 1/1000. This error is of no consequence, except that it
The feasibility of such a procedure can be easily gives rise to small irregularities in the total energy
appreciated by giving some orders of magnitude: I.et 8 which shou M be otherwise strictly constant. The
be the root-mean-square velocity in our units, it is temperature, averaged over the time, is affected by a
typically of the order of 0.3; if this is so, no error is made statistical error of the order of 0.004.
as long as
r~ —r„&n8h. (5) B. Pressure
If, for instance —
n=16, n8h=0. 15. By choosing r~ 3.3 The pressure is calculated from the virial theorem
for r„=2.5, thecondition (5) is largely met, and at the
same time the "skin" depth stays reasonably small. We
have checked, by following some systems for several
——
p
pkT
=1
1

6EkT
P' P
~&i
r, ,
9'v,

Br;;
; —p
6kT „,
r
9'v

Br
g(r)dr—
hundred steps in time, that no difference at all was (8)
observed when n was reduced, . Moreover, the conserva-
The second term of (8) is the time average of the virial.
tion of the total energy and of the total velocity, which
The last term is a correction term which takes into
stays of the order of 10 ~, is a guarantee of the sound-
account the effect on the pressure of the tail of the
ness of the whole procedure.
potential which has been neglected, in the dynamics.
The influence on the main term of (8) of the tail of
'Some time-saving tricks have been considered before: See
Ref. 2 and, for the hard-sphere case, A. Rotenberg )New York the potential, which has been neglected in the dynamics
University Report No. NYO-1480-3, 1964 (unpublished)g. can be appreciated, by considering that cutoff tail as a
100 LOUP VERLET
TABLE I. List of thermodynamical results: p is the pa, rticle
density, aud T the temperature; PP/p is the compressibility factor,
00' is equal to (V'I/")/m, E„and G„are the in6nite-frequency bulk
and shear moduli, respectively, K„/G„ their ratio, aud r, the
distance at which the potential is cut. Reduced units are used
throughout.

p T PP/p U' &O' Z G„Z„/G„r,

0,88 1.095 3.48 —5.66 23.3 58.2 32.1 1.81 2.5
0.88 0.94 2, 72 —5.84 21.4 52.4 29.8 1.76 2.5
0.88 0.591 —0.18 —6.53 17.3 42. 5 26.2 1.62 2.5
0.85 2.889 4.36 —4.25 35.1 90.7 44.5 2.04 2.5
0.85 2.202 4.20 —4.76 30.9 77.4 39.3 1.97 2.5
0.85 1.214 3.06 —5.60 22.4 54.7 30.3 1.81 2.5
0.85 1.128 2.78 —5.69 21.5 52.2 29.3 1.78 2.5
0.85 0.880 1.64 —5.94 18.8 45.0 26.4 1.70 2.5
0.85 0.782 0.98 —6.04 17.8 42.1 25.3 1.66 2.5
0.85 0.786 0.99 —6.05 17.8 42.2 25.4 1.67 2.5
0.85 0.760 0.78 —6.07 17.5 41.4 25.0 1.66 2.5
0.85 0.719 0.36 —6.12 17.0 39.9 24.4 1.63 2.5
0.85 0.658 —0.20 —6.19 16.4 38.3 23.8 1.61 2.5
0.85 0.591 —1.20 -6.26 15.5 35.7 22.7 1.57 2.5
0.75 2.849 3.10 —4.07 26.8 30.2
59.3 1.96 2.5
0.75 1.304 1.61 —5.02 ~ ~ ~
36.1
20.9 1.73 2.5
0.75 1.069 0.90 —5.19 15.4 32.0
19.3 1.65 2.5
0.75 1.071 0.89 —5.17 15.4 31.6
19.1 1.66 3.3
0.75 0.881 —0.12 —5.31 14.1 27.g 17.5 1.56 2.5
0.75 0.827 —0.54 —5.38 13.2 26.9 17.2 1.53 2.5
0.65 2.557 2.14 —3.78 35.8 19.3 1.85 3.3
0.65 1.585 1.25 —4.23 ~ ~ ~
25.5 15.0 1.70 2.5
0.65 1.036 —0.11 —4.52 11.5 19.7 12.7 1.55 3.3
0.65 0.900 —0.74 —4.61 10.4 17.8 11.9 1.50 2.5 l. 5
0.55 2.645 1.63 —3.24 23.0 12.7 1.80 2.5
0.5426 3.26 1.86 —3.00 26.1 13.9 1.88 2.5 Fro. t. The compressibility factor Pp/p as a function of P for
0, 5426 1.404 0.57 —3.63 14.2 9.0 1.58 2.5 the isochores p=0. 88, 0.85, 0.75, 0.65, and 0.45 (solid lines), as
0.5426 1.326 0.42 —3.66 ~ ~ ~
13.9 8.9 1.56 3.5 compared with the experimental data for argon (Refs. 10-12)
0.5 1.36 3.40 —3.38 8.5 11.4 7.4 1.54 2.3 (dots). The dashed curves correspond to the gas-liquid and solid-
0.45 4.625 1.68 —2.22 14.9 21.4 11.2 1.91 2.5 Quid coexistence lines.
0.45 2.935 1.3g —2.60 11.0 14.8 8.3 1.79 2.5
0,45 1.744 0.74 —2.90 8.20 11.0 6.7 1.64 2.5
0.45 1.764 0.76 -2.89 8.20 11.0 6.7 1.64 2.5 density (p=0.88, 0.85, 0.75, 0.65, and 0.45), and the
0.45 1.710 0.74 —2.95 8.20 11.0 6.7 1.64 2.5
0,45 1.552 0.75 —2.98 7.7 10.1 6.4 1.58 2.5 comparison of the equation of state determined by vari-
0.4 0.41 —2.72 7.27 4.77 1.53 2.5 ous experimental groups in Amsterdam 7, io Toronto
0.4
0.35
1,462
1,424
1.620
0,38
0.58
—2.73
—2.31
7,22 4.75
5.70 3.72
1.52
1.54
3.3
2.5
and Louvain "
0.35 1.418 0,40 —2.21 4.92 3.31 1.49 2.5 The over-all agreement between "theory" and. experi-
ment is surprisingly good: It appears that the Lennard-
Jones potential is a quite satisfactory interaction as far
weak long-range perturbation. ' The effect on the main as the equilibrium properties of argon are concerned.
term of (8) is quite small: It amounts to 0.006 in the It is to be noted that if, instead. of argon, xenon were
one case studied (p=0. 75, T=1.05). It is smaller than chosen for the comparison, the agreement would not be
the statistical error which is generally of the order of so good. For instance, with, for T= j..35, p=0. 75, the
0.01 when 1200 steps in time are considered. The main value of pp/p from molecular dynamics is 0.86; the
error comes certainly from the correction term which same quantity is equal to 0.86 in argon. For xenon (with
is very large at high density and low temperature where the reduction parameters e/k=225. 3'K, o =4.07 A),
it is of the order of 1 for r, =2.5. the value j..05 is obtained. "
The replacement of g(r) by 1 in the correction term On Fig. 2 are represented three isotherms: One at
leads, for r, = 2.5, to an error which may reach 0.05 at high temperature (T=2.74) where comparison can be
the highest density and lowest temperature considered made with the results obtained a long time ago by Wood
here. Such a replacement is, however, not necessary as and Parker, ' and also with a I ennard-Jones potential
the extrapolated values of g(r)4 may be used. through the Monte Carlo technique; the low-
As a whole, we believe that the over-all error on p/pkT temperature isotherm T = 1; and the isotherm T = 1.35,
is probably of the order of 0.01 around the critical which is also an isotherm for which Monte Carlo com-
density and may reach 0.05 in the high-d. ensity and.
low-temperature region. "$.
" M. H. Levelt, Physica 26, 361 (1960).
The results so far obtained have been gathered in W. Van Witzenburg, University of Toronto dissertation
Table I. Fig. I shows p/pkT for some isochores at high (unpublished); copy available in Dissertation Abstracts 25, 1268
(1964).
' J. Lebowitz
~A. Van Itterbeck, 0. Verbeke, and K. Staes, Physica 29,
(private communication). 742 (1963}.
 OMPU TER gi ExpERI ME& ON CLASSICA FLUIDS 10&

u»
nit cell of the
g cc lattice the x
s from 0 " I.
' For a solid, Pk is of th
f g It pscillates a o
~
m ljtude pf
the prder o If we admit that
metas table ~tates of the solid are short lve
pare to the total t'time of th e computation,
]..2X 0—" secfpr argon ~ an 6 d by trial and error
the melting temp erature. For P= p85 ' asolid con g ura"
tjpn ls sti observed fpr &= 0 695 ' It cprrespo
till o
. s onds ~ «a
f the in't' m erature &0= y 3ps. For
'
y =$.366, melting does ooccur. Kmerim en«lb
p 704. For P
l.5 a solid f«T 0 '
W th &0 f 67, me» Kxper™
melting take P 82 t that density
'
agreem t ent is seen tP be g
wjth exPerimen
I—
C 't cal Co&s

se of comPu a nal errors " the precise dete

mination of the critica ts is d cu ]t We "
elieve, hpwevver th att a careful examin tipnof thedata
~

.5— ' ~

warrants thee fpllowing

o ssion and con clusions.

cri
'
n pf the
a]. temperature y ields resu s
h res towards the
. ' Ood is,
'

I t wjth the Mo nte Carlo results of Leves ue

.
I .2 0000 «n6gura-
. ,&„, tjpns ) Those results~ &o
~

F 2. For the iso T =2.74,

', T'=I;,
y pp is
0f ~' The low-densit& P "„oftire'curves
and in able II.
represented a
as a&unction
(b) The P& I I e q.uation lead " " p45 and p= =0.5
0
(Ref. 13) The
, ~s.
(Refs. I
&,
~

"
h he help 0& the Pg and P Y II equations

&» ~he pots are the exi'e

ts«M Carlo calculations
ions to va ues o f pp/p which ar ~e~, tely too lo~. The trend
tpp l p w values for
densjty is a y
uantities &t irtterrrt&tb
t =0.4 as may b
~ ~

h ne js at the
putatipns havee been made. 13 T
That isotherm wlill be~ used
' ' puutatipnaj. error .
tp exam stion pf the critica ]. constants.
e q uation are corrected
co by 0.
F rlo~valueso&& he~P ressure r cpmputa i " P p 4 and this seemss to be the mpsst that can be
describe met astable stat ' The Periodic bounndary
' x ected " f /owing
e"Pe —' '
t nts are obtaine d:
f inhomogeneitie 'es of a size T', =$.32, O' 032, and PeP f"= ' ' d of those
~

II e uatjp» ~ ~36 P =036

f a two-p ase h hindered Th
p of the isochores be situ
has been o in

n. K uihbnum
d
„, . '"'', ,
b l

urations corresponnd.in
con fi gur
""o from
t
II

M. D.)
eso
PV II equa
1, 5-
p
r

n
'otherm T= . 0

th b
' '
d, rom p atia con
distributions. e PY II M. C. M. D. 5 Vir. Pa e

0 0..506 0.505 0.52 7

in "occurs. Our criterion elting 0.377 0.374 0.40

epos a est v
',
'
reci p rocaal lattice, i.e., k= u, t , 045 0434 025 0 .
o. 0.02 0.38
0.55 0.692 0.45 0.44~
"D. Levesque an L.. Verlet, , Physica
and
'
sica to0 be publishe d )..
s1ca 0.65 1 256 1.22 0.86%0.05 0. . . 0.
'4 P. Nozieres rivate corn
priv
 LOUP VERLET
TsaLE III. Values of Ppjp, of the inverse compressibility should thcQ mRnif est, themselves in thc form of Rn
(Ref. 4), and of the interaction part of the integral as given by
the molecular dynamics (M. D.), the PV II, and the PY equation interaction which is clearly state dependent at high
for T=1.46, p=0.4. density; and this seems to be contradicted by the
excellence of the 6t obtained in the present paper.
M. D.
PP/~ 0.41+0.01 0.40 0.51 D. Internal Energies
p&p/sn 0.54 0.53 0.49
U' —2.72+0.01 —2.'l2 —2.71 The part of the internal energy due to the interaction
U' is calculated as the time average of the sum of the
intcI'pRrticlc intel Rction cncI'gics. A tail col I cctioQ is
(c) The five virial series's gives values of PP/p which included, as for the pressure and the energy; the over-all
are clearly too high by 0.015 at p=0. 3, by 0.05 at error appears to be of the order of 0.02. The agreement
p=0.35, and by 0.06 at p=0. 4. The difference then between our results and the (not too extensive) experi-
dllllllllsllcs Rnd cllallgcs sigil R't lllgll density. T}lc Pad6 mental data" is satisfactory, although the calculated
approximant shows the same behavior but a little less internal energies are a little larger in magnitude than
accentuated. Ke conclude that one or several virial the experimental one. For instance, for p=0. 45, at
cocScients after the fifth one must be negative, and —
T=2.935, U„i,'= 2.60, U, '= 2.49; at the same
, —
that some higher ones must be positive again. —
density, for 7= 1.764, U„»,'= 2.89, U, ~'= 2.84. At —
The critical point deduced, by Barker et ul. '5 from higher densities, the same trend is noticeable, for
their series is situated at T, =1.29, p, =0.26, and p=0. 5426, T= 1.404, U„i,'= —3.63, U, p'= —3.55. It
p, p./p, =0.35. The critical density is quite low, a appears, roughly speaking, that the Lennard-Jones
consequence of the too low pressures at intermediate potential is slightly too deep.
dcnsltlcs.
(d) Table IV summarizes the situation: There a E. High-Frequency Elastic Moduli
comparison with experiment oIl argon is made, which
shows that there exists a discrepancy of at least 5% on Zwanzig and Mountain" have shown that in simple
the critical temperature. This raises the following Auids the infinite-frequency bulk modulus E„and the
problem: It is known nowadays"'7 that the low-density infinite-frequency shear modulus G„can be expressed,
data are not well fitted by a Lennard-Jones potential; in terms of the radial distribution function; and that,
a potcntiRl with R deeper bowl Rnd R wcRkcI' tRil such furthermore, in the case of the Lennard-Jones potential,
as the recently determined Kihara potential, "
seems the necessary integrals can be expressed in terms of the
much more appropriate. It would seem natural to guess pressure and of the internal energy, so that G„and E„
that using such a potential will improve the situation are simply given by the following expressions:
in the critical region. The reverse is true: I.evcsque —
" G„=3p 2p 7 (24/5)—
p U',(9)
and Vieillard-Baron, using the PY II equation, find
R cl'ltlcal telllpcl'R'tlllc of 1.42 (with tllc salllc lllllts Rs E-= (5/3)G. +2(p I») — (&0)
before). The blame for this discrepancy may be put on
many-body forces. Even at density around the critical
Results for E„, G„, and, the ratio E„/G„are given in
Table I. When the temperature is low, the internal™
region, a relatively large amount of those forces would
energy effects predominate and E„/G„ is near (5/3),
be needed. It appears that these many-body forces
the value obtained from Cauchy condition for solids.
Tmr. E IV. Critical constants obtained from the PY II equation Ke may compare the "exact results" given in Table I
with those obtained by Zwanzig Rnd Mountain who
(Ref. 13), from the PYII equation corrected as explained in
Sec. III.A, from the 5-term virial series (Ref. 15) and from experi-
ment (Ref. 10).
used (9) and (10), and the experimental data of Levelt, "
supposing that the interaction between argon atoms is
exactly the Lennard-Jones potentiaL The good agree-
ment between those results and ours shows again-
PV II 1.36a0.04 0.31+0.03 0.36~0.03
Rlthough this proof is not independent of thc picccding
PV II
Maximum correction 1.32 0.31 0.30 —
ones that the Lennard-Jones potential is appropriate
5-term virial 1.29 0.26 0.352 to describe argon in the temperature and. density do-
Experiment 1,26 0, 297 0.316
main considered m. the present paper.

"J.A. Barker, P. j. Leonard, and A. Pompe, $. Chem. Phys. F. Isotopic Separation Factor
44, 4266 {1966).
. "E.
A. Guggenheim and M. C. McGlasham, Proc. Roy. Soc. The separation faction in a liquid-gas mixture of
(London) 22SA, 456 (1960); R. J. Munn and J. Alder, J. Chem. isotopes of argon can be expressed" in terms of the
Phys. 43, 3998 (1965); J. Rowlinson, Discussions Faraday Soc.
40, 19 (1965).
"J.
A. Barker, %. Fock, and F. Smith, Phys. Fluids 7, 429 '9 R. Zwanzig and R. D. Mountain, J. Chem. Phys. 43, 4464
(1964). (1965).
~s D, I.eVezqne any
J. Vieliiard-iiaren (tO be pnb~izimg). "G. Casanova, A. Levi, and N. Terzi, Physica I, 937 (1964).
 COMPUTER ''EXPERIMENTS'' ON CLASSICAL FLU IDS
information (pressure-versus-density curve for the
coexistence curve, as shown on Fig. I), to locate the
temperature of the coexistence line on the isochores. It
may be seen easily, however, that this introduces very
little arbitrariness. A good IIIt of the experiment is ob-
taQlcd) as scen from Flg. 3.

IV. CONCLUSIONS
We have shown, using Rahman's work as a starting
point, how it is possible to integrate the equations of
motion of about a thousand particles in a relatively
easy way. The erst application of this tool is the
thermodynamic study of a Quid composed of atoms
interacting through a Lennard-Jones potentiaL The
striking result of this study is the over-all agreement
between the results thus obtained. and the thermo-
dynamics of real argon. It is likely that those results
can still be slightly improved and that a two-body
potential can 6t the experimental data with a large
dcglcc of success.

ACKNOWLEDGMENTS

f
The author is very grateful for the hospitality of
l. 5 Yeshiva University where most of this work was done.
Fro. 3. The solid hnes show the quantity C=Q&'/2spra
It is a particular pleasure to thank Professor Joel L.
= (n'Vl/96sp for the isochores p=0.85, p=0.75, and p=0.65, as Lebowitz; numerous long discussions have been both
a function of P. The crosses represent the intercepts of those inspiring and stimulating in the series of studies whose
isochores with the experimental liquid-gas coexistence line. It
should be compared with the isotopic separation factor in argon
results we have begun to describe. We are grateful also
(Ref. 20) (triangles). to D. Levesque, J. Vieillard-Baron, and D. Schi8 for
help and discussions. Last, but not least, we must
average value of the Laplacian of the potential (V' V(r) ). acknowledge the profuse computing opportunities and
The quantity Qs'= (I/M)(PV(r)), which is also im- the warm welcome in the Courant Institute of New
portant in the time development of the system, has York University, where the largest part of the computa-
been calculated, and the values we obtained are shown tions were performed, and where the author bene6tted
in Table I. The quantity C=Qss/2sp is plotted for from interesting and helpful discussions with Professor
several isochores in Fig. 3. We have used experimental J. Percus and Dr. M. Kalos.