PHYSICAL REVIEW E 69, 031111 共2004兲

Pressure tensor and heat flux vector for inhomogeneous nonequilibrium fluids under the influence
of three-body forces
Junfang Zhang and B. D. Todd*
Centre for Molecular Simulation, Swinburne University of Technology, P.O. Box 218, Hawthorn Victoria 3122, Australia
共Received 11 September 2003; published 31 March 2004兲
We present a derivation of the pressure tensor and heat flux vector for inhomogeneous fluids under the
influence of three-body forces. The derivation is based on the method of planes formalism of Todd, Evans, and
Daivis 关Phys. Rev. E 52, 1627 共1995兲; 51, 4362 共1995兲兴. Our derivation is validated against nonequilibrium
molecular dynamics simulations of a confined fluid acted upon by a two-body Barker-Fisher-Watts force
coupled with the Axilrod-Teller three-body force. Our method of planes calculations agree perfectly with the
equivalent mesoscopic route of integrating the momentum and energy continuity equations directly from the
simulation data. Our calculations reveal that three-body forces have an important consequence for the isotropic
pressure, but have negligible influence on the shear stress 共hence viscosity兲 and heat flux vector 共hence thermal
conductivity兲 for a confined simple fluid.

DOI: 10.1103/PhysRevE.69.031111 PACS number共s兲: 05.60.Cd, 05.20.Jj, 05.10.⫺a, 66.20.⫹d

I. INTRODUCTION that results from an expansion of the difference of ␦ func-

tions specifying atomic positions. For homogeneous fluids
Practical implementations of the theory and computa- this operator is unity and poses no complication: the result is
tional methods of statistical mechanics have never been more the standard Irving-Kirkwood formulas for the pressure ten-
sought after, due mainly to rapid developments in nano- sor and heat flux vector. The operator is not unity for inho-
science and their technological offshoots. These applications mogeneous fluids, and while it is possible to obtain exact
range from the transport of natural gas through zeolites expressions of the position-dependent momentum and energy
through to the operation and dynamics of protein motors. fluxes by this route 关15兴 it is an involved computation.
While many of these applications are relevant to equilibrium In Refs. 关13兴, 关14兴 it was demonstrated how a formulation
situations, there is a growing interest in applying the prin- of the Irving-Kirkwood procedure in reciprocal space could
ciples of nonequilibrium statistical mechanics to molecular greatly simplify the calculation of momentum and energy
fluxes. This formulation, known as the ‘‘method of planes’’
fluids under flow conditions. Simulations of shear-induced
共MOP兲, is valid for systems with planar geometry and flow
flow by homogeneous nonequilibrium molecular dynamics
in one direction. While the original papers were devoted to
共NEMD兲 methods are now well established and routine, with
formally deriving nonequilibrium statistical mechanical ex-
particular technological relevance to lubrication and polymer pressions for the pressure tensor and heat flux vector for
processing 关1–5兴. Over the past decade interest has also pairwise additive potentials, it was later shown 关16兴 how the
grown in the application of inhomogeneous NEMD tech- method may be generalized to compute any other relevant
niques to fluids confined at the nanoscale 关6 –10兴. For non- property exactly, such as the position-dependent density,
equilibrium environments one is interested in not only the temperature and streaming velocity.
standard thermodynamic and structural information relevant The method of planes formalism demonstrated that the
to equilibrium fluids, but also their transport properties, Irving-Kirkwood gauge is a direct and natural consequence
namely the transport of mass, momentum and energy. of solving for the microscopic momentum and energy fluxes
The first simulation of planar Couette flow by inhomoge- via the hydrodynamic continuity equations formulated in re-
neous NEMD methods 共i.e., atomic fluid confined between ciprocal space. No heuristic assumptions were made about
atomistic walls moving at constant and opposite velocities pressure being the ‘‘force across unit area;’’ rather it is a
with respect to each other兲, was performed by Liem et al. natural consequence of the formalism. Since these deriva-
关11兴. Their simulations showed that in the limit of large wall tions were published there have been a number of studies
spacing, inhomogeneous and homogeneous NEMD methods that have used the methodology in practical simulations of
were consistent with each other. A problem still remained in equilibrium and nonequilibrium fluids, e.g., Refs. 关17兴, 关18兴,
the formal application of nonequilibrium statistical mechan- as well as more theoretical papers addressing the vagueness
ics to compute the stress tensor and heat flux vector. For of the definition of the pressure tensor, e.g., Ref. 关19兴. The
homogeneous flows the standard Irving-Kirkwood procedure method has been used for confined alkanes 关20兴, polymers
关12兴 is well suited, but for strongly inhomogeneous flows 关21兴, the computation of elastic constants in thin films 关22兴,
共e.g., fluids confined by structured walls on the nanoscale兲 it and has recently been formulated for flows in cylindrical
cannot be used. This was clearly demonstrated by Todd et al. geometry 关23兴.
关13,14兴 to be a consequence of the differential operator O i j In this paper we extend the method of planes formalism
one step further, by formally deriving expressions for the
pressure tensor and heat flux vector for fluids under the in-
*Email address: [email protected] fluence of three-body forces. We test our theoretical expres-

1063-651X/2004/69共3兲/031111共11兲/$22.50 69 031111-1 ©2004 The American Physical Society

J. ZHANG AND B. D. TODD PHYSICAL REVIEW E 69, 031111 共2004兲

sions by applying them to simulations of a confined atomic

fluid under the influence of gravity driven flow. The fluid
interacts with itself and the walls by a combination of a
two-body Barker-Fisher-Watts potential 关24兴 and a three-
body Axilrod-Teller potential 关25兴. Our MOP calculations for
the pressure tensor and heat flux vector are validated against
equivalent calculations involving direct integration of the
momentum and energy continuity equations 共the so-called
IMC and IEC methods 关13,14兴兲 and are in excellent agree-
ment. Our simulations demonstrate that the inclusion of
three-body forces has a significant influence on the isotropic
pressure, in agreement with previously reported results for
homogeneous NEMD simulations of three-body fluids 关26兴.
However, for simple fluids the transport of momentum and
energy is largely dependent only on the two-body potential.
This agrees with previously reported results that indicated
the three-body potential only contributes about 3% of the
total shear viscosity 关27兴.

II. THEORY FIG. 1. Planar geometry of the flow for a confined fluid under
the influence of an external field. The z axis is normal to the page.
In the derivations that follow we designate the two-body
and three-body force contributions to the total interatomic In k space the spatially average momentum density is 关13兴
force as F(2)i and F(3) (3)
i , respectively. Fi j is defined here to be
the contribution to the total three-body force on atom i due to 1
atom j. If ␾ (2)ij ⫽␾
(2)
(ri ,r j ) is the two-body potential and
J ␣共 k y 兲 ⫽
A 兺i m v ␣ i e ik y ,
y i 共4兲
␾ i jk ⫽ ␾ (ri ,r j ,rk ) is the three-body potential, then
(3) (3)

冉 冊
where A is the area of the x-z surface that has its normal in
⳵ ␾ (2)
i ⬅
F(2) 兺j i j ⫽⫺
F(2) 兺j ij
⳵ ri
, 共1兲
the y direction. m and v ␣ i are the mass and laboratory ve-
locity of particle i, respectively, and ␣ ⫽x, y, z. Similarly,
the k-space momentum continuity equation is

i ⬅
F(3) 兺jk 共 F(3)
i j ⫹Fik 兲
(3)
⳵ J ␣共 k y 兲
⫽ik y 关 P ␣ y 共 k y 兲 ⫹F兵 ␳ 共 y 兲 u ␣ 共 y 兲 u y 共 y 兲 其 兴 , 共5兲
⳵t
⫽⫺ 冋兺 再冉 冊 冉 冊冎册
jk
⳵ ␾ (3)
i jk
⳵ ri j
⫹
⳵ ␾ (3)
i jk
⳵ rik
共 ri j ⬅ri ⫺r j 兲 , where F兵 其 denotes the Fourier transform of the quantity in
brackets and u is the streaming velocity of the fluid. Substi-
共2兲 tuting Eq. 共4兲 into Eq. 共5兲, isolating the pressure tensor, and
finally inverse transforming gives the kinetic (K) and poten-
⳵ ␾ (3)
i jk tial (U) contributions to the pressure tensor as
i j ⬅⫺
F(3) . 共3兲
⳵ ri j
1 p ␣ i p yi
The geometry of our system is planar and is shown sche- P K␣ y 共 y 兲 ⫽
A 兺i m
␦ 共 y i ⫺y 兲 共6兲
matically in Fig. 1. A three-dimensional fluid is confined
between planar parallel walls separated by length L y in the y and
direction. A constant field drives the system away from equi-
librium and is directed in the x direction. Thus, all thermo- 1
dynamic and transport properties are functions of only y. ␣y共 y 兲⫽
PU
2A 兺i F ␣ i sgn共 y i ⫺y 兲 , 共7兲

A. Pressure tensor
where p ␣ i and p yi are peculiar 共i.e., thermal兲 momenta, and
The method of planes derivation of the pressure tensor F ␣ i is the ␣ component of the total force on atom i 共i.e., it
follows closely the original derivation in Ref. 关13兴. Briefly, includes both two- and three-body contributions兲. The deri-
the method involves defining the microscopic expressions for vation up to this point is identical to the original derivation
the mass and momentum densities, then Fourier transforming and the reader should refer to Ref. 关13兴 for further details. It
and integrating over the x, z directions 共as physical proper- illustrates that Eqs. 共6兲 and 共7兲 are actually completely gen-
ties are assumed uniform in x, z). One also takes the Fourier eral for this type of planar symmetry and is valid for n-body
transform of the momentum continuity equation, substitutes forces. In what follows we specifically consider the case of
in the microscopic k-space momentum flux and solves for three-body forces and use symmetry relations to generate a
the k-space pressure tensor. Finally one performs the inverse useful expression for the potential contribution to the three-
Fourier transform to recover the r-space pressure tensor. body pressure. The kinetic contribution remains unchanged

031111-2
PRESSURE TENSOR AND HEAT FLUX VECTOR FOR . . . PHYSICAL REVIEW E 69, 031111 共2004兲

as it implicitly contains the full two-plus three-body force

contributions in the particle momenta.
We now separate out the two- and three-body contribu-
tions to the pressure

1
␣y共 y 兲⫽
PU
2A 兺i F ␣ i sgn共 y i ⫺y 兲
1
⫽
2A 兺i 共 F ␣(2)i ⫹F ␣(3)i 兲 sgn共 y i ⫺y 兲
⫽ P (2)U
␣y 共 y 兲⫹ P␣y 共 y 兲,
(3)U
共8兲

where P (2)U (3)U

␣ y (y) and P ␣ y (y) are contributions from two-
body and three-body forces, respectively. In Ref. 关13兴 it was
shown that

1
␣y 共 y 兲⫽
P (2)U
2A 兺i F ␣(2)i sgn共 y i ⫺y 兲

⫽
1
4A 冋兺 ij
␣ i j sgn共 y i ⫺y 兲
F (2)
FIG. 2. Triangular configuration of atoms and the plane located

册
at y⫽y 0 . Contributions to the pressure tensor are included from
⫹ 兺i j F (2)
␣ ji sgn共 y j ⫺y 兲
atoms 1, 2, and 3 along the vectors r12 and r13 .

⫽
1
4A 再兺 ij
␣ i j 关 sgn共 y i ⫺y 兲 ⫺sgn共 y j ⫺y 兲兴
F (2) 冎 ⫽
1
6A 再兺 ij
␣ i j 关 sgn共 y i ⫺y 兲 ⫺sgn共 y j ⫺y 兲兴
F (3)

⫽
1
兺i j F (2)
␣ i j 关 ⌰ 共 y i ⫺y 兲 ⌰ 共 y⫺y j 兲
⫹ 兺ik F ␣(3)ik 关 sgn共 y i ⫺y 兲 ⫺sgn共 y k ⫺y 兲兴
2A

⫺⌰ 共 y j ⫺y 兲 ⌰ 共 y⫺y i 兲兴 , 共9兲 ⫹ 兺jk F ␣(3)jk 关 sgn共 y j ⫺y 兲 ⫺sgn共 y k ⫺y 兲兴 冎 共11a兲

where ⌰ is the Heaviside step function.

Making similar use of particle exchange symmetry, the
three-body contribution to the pressure tensor can be ex-
pressed as
⫽
1
3A 再兺 ij
␣ i j 关 ⌰ 共 y i ⫺y 兲 ⌰ 共 y⫺y j 兲 ⫺⌰ 共 y j ⫺y 兲 ⌰ 共 y⫺y i 兲兴
F (3)

1 ⫹ 兺ik F ␣(3)ik 关 ⌰ 共 y i ⫺y 兲 ⌰ 共 y⫺y k 兲 ⫺⌰ 共 y k ⫺y 兲 ⌰ 共 y⫺y i 兲兴

P (3)U
␣ y (y)⫽ 兺i F ␣(3)i sgn(y i ⫺y)

冎
2A
兺jk F (3)
⫽
1
6A 冋 兺i F ␣(3)i sgn(y i ⫺y)⫹ 兺j F (3)
␣j sgn(y j ⫺y)
⫹ ␣ jk 关 ⌰ 共 y j ⫺y 兲 ⌰ 共 y⫺y k 兲 ⫺⌰ 共 y k ⫺y 兲 ⌰ 共 y⫺y j 兲兴

共11b兲
.

⫹ 兺k F ␣(3)k sgn(y k ⫺y) 册 . 共10兲 Equation 共11兲 demonstrates that the potential contribution to
the three-body pressure at a plane located at y occurs when
components of the three-body force intersect that plane, in
Substitution of Eq. 共2兲 into Eq. 共10兲 yields complete analogy with the two-body force contributions. For

冋兺
example, consider the situation shown in Fig. 2, in which a
1
␣y 共 y 兲⫽
P (3)U
6A ij
␣ i j sgn共 y i ⫺y 兲 ⫹
F (3) 兺ik F ␣(3)ik sgn共 y i ⫺y 兲 triangular configuration of three particles is shown. Only the
force contributions along the vectors r12 and r13 intersect the
plane at y⫽y 0 and contribute to the three-body pressure at
⫹ 兺ji F (3)
␣ ji sgn共 y j ⫺y 兲 ⫹ 兺 F ␣ jk sgn共 y j ⫺y 兲
jk
(3) this plane.
Finally, in a molecular dynamics simulation the full pres-

册
sure tensor is computed by time averaging over the simula-
⫹ 兺ki F ␣(3)ki sgn共 y k ⫺y 兲 ⫹ 兺
kj
␣ k j sgn共 y k ⫺y 兲
F (3) tion phase space trajectory. In the case of the potential con-
tributions this time averaging is straightforward. In Ref. 关13兴

031111-3
J. ZHANG AND B. D. TODD PHYSICAL REVIEW E 69, 031111 共2004兲

it was shown that the time averaged kinetic component of the We start the derivation at the Fourier transformed energy
pressure tensor can be usefully expressed as density continuity equation 关14兴

1 ⳵
具 P K␣ y 共 y 兲 典 ⫽ lim
t→⬁
兺 兺i p ␣ i共 t i,m 兲 sgn关 p yi共 t i,m 兲兴 .
At 0⬍t i,m ⬍t ⳵t
␳ e 共 k,t 兲 ⫽ik• 关 Jq 共 k,t 兲 ⫹F兵 ␳ eu其 ⫹F兵 P•u其 兴 . 共13兲

共12兲
Here it is noted that particle i crosses the plane at y at a set If pi is defined here as the laboratory momentum of atom i,
of times (t i,m ;i⫽1,...,N;m⫽1,2,...). then the total energy of atom i is

B. Heat flux vector

p2i 1 1
As with the pressure tensor derivation, we follow the
method of planes formalism developed in Ref. 关14兴 for the
e i⫽
2m
⫹
2 兺j ␾ (2)
i j ⫹ 兺 ␾ i jk .
3 jk
(3)
共14兲
heat flux vector. That approach uses the microscopic defini-
tions of the local energy density and the Fourier transform of
the energy continuity equation to obtain a k-space expression Noting that ␳ eu(k,t)⫽ 兺 i e i u(ri ,t)e ik•ri and ␳ e(k,t)
for the heat flux vector, which is again back transformed into ⫽ 兺 i e i e ik•ri 关14兴, we first compute the time derivative of the
r-space. energy density

⳵␳ e 共 k,t 兲
⳵t
⫽ik• 冉兺 i
vi e i e ik•ri
冊 ⫹ 兺i mvi •v̇i e ik•ri
⫹
1
2 兺i j 冉 ṙi •
⳵ ␾ (2)
ij
⳵ ri
⫹ṙ j •
⳵ ␾ (2)
ij
⳵ rj
冊
e ik•ri ⫹
1
3 兺
i jk
冉 ṙi •
⳵ ␾ (3)
i jk
⳵ ri
⫹ṙ j •
⳵ ␾ (3)
i jk
⳵ rj

⫹ṙk •
⳵ ␾ (3)
i jk
⳵ rk
冊
e ik•ri ⫽ik• 冉兺 i
vi e i e ik•ri ⫹ 冊 兺i vi • 共 F(2)
i ⫹Fi 兲 e
(3) ik•r
⫺ 兺 共 vi •F(2)
2 ij
i
1
i j ⫹v j •F ji 兲 e
(2) ik•r i

兺冉 冊 冉兺 冊
1 ⳵ ␾ (3) ⳵ ␾ (3) ⳵ ␾ (3) 1
⫹
3 i jk
ṙi •
i jk
⳵ ri
⫹ṙ j •
i jk
⳵ rj
⫹ṙk •
i jk
⳵ rk
e ik•ri ⫽ik•
i
vi e i e ik•ri ⫹
2 兺i j vi •F(2)
ij 共e
ik•r
⫺e ik•r 兲
i j

⫹ 兺i vi •F(3)
i e
ik•r
⫹ 兺
3 i jk
i
1
冉 ṙi •
⳵ ␾ (3)
i jk
⳵ ri
⫹ṙ j •
⳵ ␾ (3)
i jk
⳵ rj
⫹ṙk •
⳵ ␾ (3)
i jk
⳵ rk
冊
e ik•ri . 共15兲

Consider now the two terms containing the three-body forces. The first term may be symmetrized as

兺i vi •F(3)
i e
ik•r
⫽ 兺 vi •F(3)
3 i i e
i ik•r
1
j
冋
⫹ 兺 v j •F(3)
j e
ik•r
⫹ 兺 vk •F(3)
k
k e
ik•r i j k
册
⫽
1
3 冋兺
i jk
vi • 共 F(3)
i j ⫹Fik 兲 e
(3) ik•ri
⫹ 兺
i jk
v j • 共 F(3)
ji ⫹F jk 兲 e
(3) ik•r
⫹ 兺 vk • 共 F(3)
i jk
ki ⫹Fk j 兲 e
j (3) ik•r k
册 . 共16兲

The second term containing three-body forces in Eq. 共15兲 may be similarly expanded:
1
3 兺
i jk
冉 ṙi •
⳵ ␾ (3)
i jk
⳵ ri
⫹ṙ j •
⳵ ␾ (3)
i jk
⳵ rj
⫹ṙk •
⳵ ␾ (3)
i jk
⳵ rk
e ik•ri ⫽冊1
3 兺
i jk
冋 冉 vi •
⳵ ␾ (3)
i jk
⳵ ri j
⫹
⳵ ␾ (3)
i jk
⳵ rik
冊 冉
⫹v j •
⳵ ␾ (3)
i jk
⳵ r ji
⫹
⳵ ␾ (3)
i jk
⳵ r jk
⫹vk •
⳵ ␾ (3)
i jk
⳵ rki
冊 冉
⫹
⳵ ␾ (3)
⳵ rk j
冊册
i jk
e ik•ri

1
⫽⫺
3 兺
i jk
共 vi •F(3)
i j ⫹vi •Fik ⫺v j •Fi j ⫹v j •F jk ⫺vk •Fik ⫺vk •F jk 兲 e
(3) (3) (3) (3) (3) ik•r
. i

共17兲

Defining S (3) as the sum of Eqs. 共16兲 and 共17兲 gives

1
S (3) ⫽
3 兺
i jk
关 v j •F(3)
ij 共e
ik•r
⫺e ik•r 兲 ⫺v j •F(3)
i
jk 共 e
j ik•r
⫺e ik•r 兲 ⫹vk •F(3)
ik 共 e
ik•r
⫺e ik•r 兲 ⫹vk •F(3)
i
jk 共 e
j ik•r
⫺e ik•r 兲兴 .i k i k 共18兲

031111-4
PRESSURE TENSOR AND HEAT FLUX VECTOR FOR . . . PHYSICAL REVIEW E 69, 031111 共2004兲

We now permute the triplet indices in Eq. 共18兲 such that all J qy 共 y,t 兲 ⫽J Kqy 共 y,t 兲 ⫹J U
qy 共 y,t 兲 . 共23兲
velocities are in terms of the index i to obtain
The kinetic contribution is, as with the pressure tensor, iden-
1 1 tical in form to the original MOP derivation in Ref. 关14兴:
S (3)
⫽
3 兺
i jk
vi • 关 F(3)
ik ⫹Fi j 兴共 e
(3) ik•ri
⫺e ik•rk 兲 ⫹
3 兺
i jk
vi • 关 F(3)
ij
1
1 J Kqy 共 y,t 兲 ⫽ 兺i 关v yi ⫺u 共 y 兲兴 U i ␦ 共 y⫺y i 兲 共24兲
⫹F(3)
ik 兴共 e
ik•ri
⫺e ik•r j 兲 ⫽
3 兺ik vi •F(3)
i 共e
ik•r
⫺e ik•r 兲i k
A

except that here U i is the internal energy of a particle defined

1
兺i j
as
⫹ vi •F(3)
i 共e
ik•ri
⫺e ik•r j 兲 共19兲
3
1 1 1
Substituting Eq. 共19兲 back into Eq. 共15兲 gives
U i ⫽ m 关 vi ⫺u共 y i 兲兴 2 ⫹
2 2 兺j ␾ (2)
i j ⫹ 兺 ␾ i jk .
3 jk
(3)
共25兲

⳵␳ e 共 k,t 兲
⳵t
⫽ik• 冉兺 i
冊
vi e i e ik•ri ⫹
1
2 兺i j vi •F(2)
ij 共e
ik•r
⫺e ik•r 兲
i j
The potential contribution to the heat flux vector is

1
1 qy 共 y,t 兲 ⫽⫺
JU 兺i j 关 vi ⫺u共 y 兲兴 •F(2)
i j 关 sgn共 y⫺y i 兲
⫹
3 兺ik vi •F(3)
i 共e
ik•ri
⫺e ik•rk 兲
4A

1
1 ⫺sgn共 y⫺y j 兲兴 ⫺ 兺i j 关 vi ⫺u共 y 兲兴 •F(3)
ij
⫹
3 兺i j vi •F(3)
i 共e
ik•ri
⫺e ik•r j 兲 . 共20兲
6A

⫻ 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y j 兲兴

Substitution of Eq. 共20兲 into the Fourier transformed energy 1
continuity equation 关i.e., Eq. 共13兲兴 yields ⫺
6A 兺ik 关 vi ⫺u共 y 兲兴 •F(3)
ik

ik•Jq 共 k,t 兲 ⫽ik• 冉兺 i

关 vi ⫺u共 ri ,t 兲兴 e i e ik•ri 冊 ⫻ 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y k 兲兴 ⫺
1
6A 兺
i jk
vi •F(3)
ij

1
⫹
2 兺i j vi •F(2)
ij 共e
ik•r
⫺e ik•r 兲
i j
⫻ 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y k 兲兴 ⫺
1
兺 vi •F(3)
ik
6A i jk
1
⫹
3 兺i j vi •F(3)
i 共e
ik•r
⫺e ik•r 兲
i j
⫻ 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y j 兲兴 ⫹
1
兺jk u共 y 兲 •F(3)
jk
6A
1
⫹
3 兺ik vi •F(3)
i 共e
ik•r
⫺e ik•r 兲 ⫺ik•F兵 P•u其 .
i k ⫻ 关 sgn共 y⫺y j 兲 ⫺sgn共 y⫺y k 兲兴 . 共26兲

共21兲 An alternative, more concise form of Eq. 共26兲, may be de-

rived by substituting the unsymmetrised form of the pressure
Integrating over x and z, dividing by ik y and taking the tensor 关i.e., Eq. 共8兲兴 into Eq. 共22兲. This is demonstrated in the
inverse Fourier transform yields Appendix and the result is quoted here as

1
AJ qy 共 y,t 兲 ⫽ 兺i 共 v yi ⫺u y 兲 e i ␦ 共 y⫺y i 兲 qy 共 y,t 兲 ⫽⫺
JU
2A 兺i 关 vi ⫺u共 y 兲兴 •F(2)
i sgn共 y⫺y i 兲

1 1
⫺
4 兺i j vi •F(2)
i j 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y j 兲兴 ⫺
2A 兺i 关 vi ⫺u共 y 兲兴 •F(3)
i sgn共 y⫺y i 兲

1 1
⫺
6 兺i j vi •F(3)
i 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y j 兲兴 ⫹
6A 兺
i jk
vi •F(3)
i 关 sgn共 y⫺y i 兲 ⫹sgn共 y⫺y j 兲

1 ⫹sgn共 y⫺y k 兲兴 . 共27兲

⫺
6 兺ik vi •F(3)
i 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y k 兲兴
The last three terms in Eq. 共26兲 and the last term in Eq.
⫺A 兵 P•u其 y . 共22兲 共27兲 are not direct analogies of the two-body heat flux, unlike
the case in the pressure tensor three-body expressions, which
To complete the derivation, we substitute the MOP expres- are direct analogies. They are a result of particle velocities
sion for the pressure tensor, Eq. 共11a兲, into Eq. 共22兲 to give coupling to three-body forces, which does not occur in the
the kinetic and potential contributions to the heat flux vector pressure tensor calculation. However, it will be shown in

031111-5
J. ZHANG AND B. D. TODD PHYSICAL REVIEW E 69, 031111 共2004兲

Sec. IV that these additional terms are negligible, if not zero. TABLE I. Parameters for the Barker-Fisher-Watts potential.
As was the case for the kinetic term for the pressure tensor,
the kinetic term for the heat flux vector in Eq. 共24兲 can be Argon
written in a more useful way for computer simulation as 关14兴 ␧/k(K) 142.095
␴ (A) 3.3605
1
兺
r m (A) 3.7612
J Kqy 共 y 兲 ⫽ lim U i sgn关 c yi 共 t i,m 兲兴 , 共28兲
t→⬁ At 0⬍t i,m ⬍t
Barker-Pompe Bobetic-Barker
␧/k(K) 147.70 140.235
where ci ⬅vi ⫺u(y) is the plane peculiar velocity of atom i.
r m (A) 3.7560 3.7630
␴ (A) 3.341 3.3666
III. SIMULATIONS A0 0.2349 0.29214
A. Two- and three-body potentials A1 ⫺4.7735 ⫺4.41458
A2 ⫺10.2194 ⫺7.70182
Our simulations are performed on fluid and solid atoms
A3 ⫺5.2905 ⫺31.9293
that interact via the Barker-Fisher-Watts two-body potential
A4 0.0 ⫺136.026
关24兴 and Axilrod-Teller three-body potential 关25兴. The total
intermolecular potential 共␾兲 is a contribution from two-body A5 0.0 ⫺151.00
interactions ( ␾ (2) ) and three-body dispersion interactions C6 1.0698 1.11976
( ␾ (3) ): C8 0.1642 0.171551
C 10 0.0132 0.013748
␾ 共 r 兲 ⫽ ␾ (2) 共 r 兲 ⫹ ␾ (3) 共 r 兲 . 共29兲 ␣ 12.5 12.5
␦ 0.01 0.01
The two-body interaction of argon is well represented by the
Barker-Fisher-Watts 共BFW兲 potential 关24兴. The BFW poten-
tial is a linear combination of the Barker-Pompe 关28兴 ( ␾ BP) cell is thus periodic in x, y, and z. The total number of
and Bobetic-Barker 关29兴 ( ␾ BB) potentials atoms is N⫽324, which includes 270 liquid atoms and 54
wall atoms (N w ⫽18 atoms per layer兲. In what follows all
␾ (2) 共 r 兲 ⫽0.75␾ BB共 r 兲 ⫹0.25␾ BP共 r 兲 , 共30兲 quantities are expressed in reduced units. The density of the
fluid is 0.44 and the wall density is 0.84. The cell dimensions
where the potentials of Barker-Pompe and Bobetic-Barker
are L x ⫽5.0565, L y ⫽27.5143, L z ⫽5.0565. The thickness of
have the following form:
the walls is ⌬y w ⫽2.5143. We use the same method in Ref.

␾ (2) 共 r 兲 ⫽␧ 冋兺5

i⫽0
A i 共 x⫺1 兲 i exp关 ␣ 共 1⫺x 兲兴 ⫺
2
C
兺 2 j⫹6
j⫽0 ␦ ⫹x
2 j⫹6 册 .
关13兴 to approximate the accessible width of the fluid channel,
which gives an effective pore width of l y ⫽24.0871. A cutoff
potential radius of L x /2⫽2.5282 was used for the two-body
共31兲 force calculation, whereas a value of L x /4⫽1.2641 was used
for the three-body force. These were optimal values, based
Here, x⫽r/r m where r m is the intermolecular separation at
upon the work performed in Ref. 关31兴. In our simulations we
which the potential has a minimum value and the other pa-
used a truncated and shifted version of the BFW and AT
rameters are listed in Table I.
potentials, so that long-range corrections need not be consid-
The triple-dipole Axilrod-Teller 共AT兲 potential 关25兴 is
ered. In this way the potential is zero at and beyond the
v DDD 共 1⫹3 cos ␪ i cos ␪ j cos ␪ k 兲 cutoff value. We justify this as our goal is to verify the MOP
␾ (3) 共 ri ,r j ,rk 兲 ⫽ , 共32兲 expressions derived in Sec. II for the pressure tensor and heat
共 r i j r ik r jk 兲 3 flux vector, rather than to accurately reproduce experimental
results.
where v DDD is the nonadditive coefficient, and the angles
The equations of motion used to simulate wall and fluid
and intermolecular separations refer to a triangular configu-
atoms were developed in Ref. 关13兴 and quoted here as fol-
ration of atoms. The nonadditive coefficient for argon is
lows, modified for three-body forces. We note that wall at-
518.3 a.u. 关30兴. Recent work 关31兴 has demonstrated that the
oms interact via the two plus three body forces in addition to
Axilrod-Teller term can significantly improve the prediction
a harmonic spring force that tethers them together. For the
of liquid phase properties.
wall particles,

B. Geometry and equations of motion pi

ṙi ⫽
The geometry of our simulation cell is shown in Fig. 1. m
An atomic fluid is confined between atomistic walls as de-
i ⫹Fi ⫺ ␣ pi ⫺j␭ L n ,
ṗi ⫽⫺K 共 ri ⫺qi 兲 ⫹F(2) (3)
picted. Our geometry is such that y⫽0 defines the center of i苸L n .
the fluid channel. A field 共e.g., gravity兲 representing a con- 共33兲
stant pressure head drives the fluid and is directed in the x
direction. Each wall is three atomic layers thick, and the K is the spring force constant and was set to 57.15 in all
second wall is just the periodic image of the first. The entire simulations. qi is the equilibrium ‘‘frozen’’ lattice position of

031111-6
PRESSURE TENSOR AND HEAT FLUX VECTOR FOR . . . PHYSICAL REVIEW E 69, 031111 共2004兲

atom i. ri is the laboratory position of atom i and pi here

refers to the laboratory momentum of atom i. As the walls
are not under the influence of a gravitational force the
streaming velocity is zero and the peculiar 共thermal兲 and
laboratory momenta are equivalent. j is the unit vector in the
y direction and the layer multiplier ␭ L n ensures that the cen-
ter of mass of each wall layer stays fixed, where the index
n⫽1, 2, 3, refers to the wall layer. This is important other-
wise the walls separate as the fluid heats up under flow. ␣ is
a thermostat multiplier used to keep the temperature of the
walls fixed 共in our simulations the wall temperature was
fixed at 0.722兲. The layer multiplier and thermostat are com-
puted as
N
j w
␭ Ln⫽ 兺
N w i苸L n
关 ⫺K 共 ri ⫺qi 兲 ⫹F(2)
i ⫹Fi 兴 ,
(3)
共34兲

where
3 Nw FIG. 3. Number density profile for the fluid system.
兺 兺
L n ⫽1 i苸L n
1⫽3N w
IV. RESULTS AND DISCUSSION

and In Figs. 3 and 4 we plot the density and streaming veloc-

ity profiles, respectively. The streaming velocity is seen to be
3N
i ⫹Fi ⫺j␭ L n 兴 •pi 其
兺 i苸Lw 兵 关 ⫺K 共 ri ⫺qi 兲 ⫹F(2) (3)
well represented by a symmetric quadratic function in y, in
␣⫽ 3N . 共35兲 conformity with hydrodynamics 关13兴. Only near the walls
兺 i苸Lw p2i
does the streaming velocity deviate from quadratic behavior.
Here L⫽ 兵 L 1 ,L 2 ,L 3 其 . It is well known that hydrodynamics breaks down at smaller
The fluid atoms obey Newton’s equations of motion channel widths 关32兴.
In Fig. 5 we plot MOP calculations of P y y , the y compo-
pi nent of the pressure in the direction normal to the wall sur-
ṙi ⫽ face, for both the BFW fluid and the BFW fluid with the
m
inclusion of the AT three body forces. For mechanical stabil-
ṗi ⫽F(2)
i ⫹Fi ⫹iF e ,
(3)
共36兲 ity P y y must be constant throughout the channel, and this is
indeed seen to be the case. Also shown is the pressure cal-
where F e is the external driving field and i is the unit vector
in the x direction and we again note that ri and pi refer to the
laboratory position and momentum of atom i, respectively.
In our simulations the field strength used was F e ⫽0.2.
The equations of motion were solved with a fifth order
Gear predictor-corrector scheme with an integration time
step of ␶ ⫽0.001. Our simulations were first run for a total of
106 time steps to reach a nonequilibrium steady state. Once
steady state was achieved, production runs of a total of 107
time steps were run with averages accumulated in blocks of
50 000 time steps. The errors presented in our plots represent
the standard error in the mean.
In our simulations we do not assume any functional form
for the streaming velocity. Rather, we first run a steady-state
simulation of ⬃106 time steps and compute a time-averaged
velocity profile at planes, using the procedure developed in
Ref. 关16兴. These plane velocity values are then used as the
streaming velocity u x (y) in MOP calculations of the pressure
tensor and heat flux vector in all subsequent production runs.
Finally, we note that a total of 200 planes were used in the
MOP calculations, though not all planes data are plotted on FIG. 4. Streaming velocity profile 共circle data points兲 for the
the figures presented in this work for clarity of visualization. fluid. Superimposed 共solid curve兲 is a symmetric quadratic fit, in
For further details of the simulation methodology, readers are conformity with hydrodynamic prediction. Error bars are the size of
referred to Refs. 关13兴, 关14兴, 关16兴. the plotting symbol.

031111-7
J. ZHANG AND B. D. TODD PHYSICAL REVIEW E 69, 031111 共2004兲

FIG. 5. P y y as a function of y for the BFW and BFW⫹AT

fluids. The pressure is computed by the method of planes. Also
shown is the pressure at the walls.

culated on the system walls. This is computed from the total

y component of the force per unit area exerted on the wall
atoms by fluid atoms on one side of the wall. We see perfect
agreement between wall and fluid P y y values, as expected.
Clearly the addition of the three-body force significantly af-
fects the pressure P y y . Neglecting to include three-body
forces overestimates the pressure by almost 11%.
Of greater interest to us is the shear stress (⫺ P xy ). In Fig.
6共a兲 we plot P xy as a function of y for the BFW and BFW
⫹AT fluids. We show the results of our MOP calculations
and compare them with direct integration of the momentum
continuity equation 共the IMC method of Ref. 关13兴, given as
P xy (y)⫽F e 兰 0y dy ⬘ n(y ⬘ ), where n(y) is the number density兲.
Error bars are of the order of the size of the plotting symbols. FIG. 6. 共a兲 P xy as a function of y for the BFW and BFW⫹AT
We find excellent agreement between both methods, demon- fluids. P xy computed by the MOP and IMC methods are shown, as
strating that the MOP calculations are correct. The value of well as P xy computed at the walls. 共b兲 As with 共a兲 but magnified in
P xy calculated at the walls is also included and seen to be the range 5.7⭐y⭐6.1.
consistent with both the MOP and IMC values. Note that the
stress deviates from the linear hydrodynamics prediction Ref. 关14兴 given as J qy (y)⫽⫺ 兰 0y dy ⬘ P xy (y ⬘ ) ␥˙ (y ⬘ ), where ␥˙
close to the walls, as is to be expected for such an inhomo- is the strain rate, ␥˙ ⫽ ⳵ u x (y)/ ⳵ y]. Error bars are of the order
geneous system. From Fig. 6共a兲 it is clear that three-body of plotting symbol sizes. Once again, excellent agreement is
forces have negligible effect on the shear stress. This is seen found between both methods, confirming the validity of the
more clearly in Fig. 6共b兲, in which the region between 5.7 MOP expressions. For a channel this size it is clear that the
⭐y⭐6.1 is magnified. This is consistent with the observa- classical cubic heat flux profile is obeyed. The value of the
tions reported in Ref. 关27兴 that showed that three-body forces heat flux at the walls was also computed by noting that the
only affected the shear viscosity by approximately 3%. The Gaussian thermostat acting on the walls removes heat at a
precise degree of influence is likely to depend on tempera- rate of
ture and density, and to a lesser degree the number of atoms,
so we do not say anything conclusive at this stage. We also N
p2i
note that our potentials are shifted and truncated and include
no long-range corrections.
Q̇ 共 t 兲 ⫽ ␣ 共 t 兲 兺
i⫽1 m
. 共37兲

In Fig. 7共a兲 we plot the heat flux vector as a function of y

for the BFW and BFW⫹AT fluids. Classical hydrodynamics The heat flux at the walls is therefore
predicts a cubic heat flux profile. We again show the results
of our MOP calculations and compare them with direct inte- 1
具 J qy 共 y⫽y wall 兲 典 ⫽ 具 K w ␣ 典 , 共38兲
gration of the energy continuity equation 关the IEC method of A

031111-8
PRESSURE TENSOR AND HEAT FLUX VECTOR FOR . . . PHYSICAL REVIEW E 69, 031111 共2004兲

FIG. 8. Individual terms of the three-body contribution to the

heat flux vector. Term 1 represents the second and third terms of Eq.
共26兲 and is the direct analogy of the two-body contribution to the
heat flux vector. Term 2 represents the last three terms in Eq. 共26兲.

body contribution to the heat flux vector. Term 2 共diamonds兲

represents the additional last three terms in Eq. 共26兲. It is
clear that these last three terms are negligible, if not zero.

V. CONCLUSIONS

In this paper we have derived method of planes expres-

sions for the pressure tensor and heat flux vector for a fluid
under the influence of three-body forces. Our derivations
have been validated against numerical simulations of gravity
driven flow by nonequilibrium molecular dynamics methods.
The MOP calculations are in excellent agreement with inde-
FIG. 7. 共a兲 Heat flux vector J qy (y) as a function of y for the
BFW and BFW⫹AT fluids. J qy (y) computed by the MOP and IEC pendent calculations based upon direct integration of the hy-
methods are shown, as well as J qy (y) computed at the walls. 共b兲 As drodynamic momentum and energy continuity equations.
with 共a兲 but magnified in the range 4.0⭐y⭐7.0. Our results show that the isotropic pressure is sensitive to the
presence of three-body forces, whereas the shear stress and
where K w is the kinetic energy of the wall atoms and the heat flux vector seem to be largely independent of them.
angle brackets indicate a time average 关14兴. The heat flux at Further work is required to study the temperature and density
the walls is in excellent agreement with the MOP and IEC dependence on the relative magnitudes of the two to three-
values at the wall-fluid interface. body force contributions. While such effects are clearly small
We again observe that the presence of three-body forces for noble gas fluids such as argon, they will most likely play
has very little influence in the transportation of energy across a significant and important role for heavier atomic and mo-
the channel. In Fig. 7共b兲 the region between 4.0⭐y⭐7.0 is lecular fluids and liquid metals. It is hoped that our MOP
magnified. Three-body forces contribute a very small but no- expressions will be useful for the study of such liquids in the
ticeable effect on the heat flux, slightly increasing its magni- future.
tude. Again, this effect is likely to be temperature and density
dependent and we refrain from specific conclusions at this
stage. What is import for our purposes is the excellent agree- ACKNOWLEDGMENTS
ment between the MOP and IEC methods, clearly visible in
this figure. J.Z. acknowledges the Australian government for financial
Finally, in Fig. 8 we plot several contributions to the assistance. Computations were performed on the supercom-
three-body component of the heat flux vector. Term 1 puting facilities at the Australian Partnership for Advanced
共circles兲 represents the second and third terms in Eq. 共26兲. Computing 共APAC兲, through a generous grant of computer
These terms are the direct three-body analogy of the two- time.

031111-9
J. ZHANG AND B. D. TODD PHYSICAL REVIEW E 69, 031111 共2004兲

APPENDIX 3
An alternative form of Eq. 共26兲 may be obtained by using
qy 共 y,t 兲 ⫽⫺
J U(3)
6A 兺i vi •F(3)
i sgn共 y⫺y i 兲

the alternative 共unsymmetrised兲 form of the pressure tensor

1
in Eq. 共8兲, i.e., ⫹
6A 兺i vi •F(3)
i sgn共 y⫺y i 兲

1
␣ y 共 y,t 兲 ⫽⫺
P U(3) 兺i F ␣(3)i sgn共 y⫺y i 兲 . 共A1兲
⫹
1
兺ik vi •F(3)
i sgn共 y⫺y k 兲
2A
6A
Substituting Eq. 共A1兲 into Eq. 共22兲 gives 1
⫹
6A 兺i j vi •F(3)
i sgn共 y⫺y j 兲
1
AJ qy 共 y,t 兲 ⫽ 兺i 共 v yi ⫺u y 兲 e i ␦ 共 y⫺y i 兲 ⫺ 4 兺i j vi •F(2)
ij
3
⫹
6A 兺i u•F(3)
i sgn共 y⫺y i 兲
2
⫻ 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y j 兲兴 ⫺
6 兺i vi •F(3)
i
1
⫽⫺
2A 兺i 关 vi ⫺u共 y 兲兴 •F(3)
i sgn共 y⫺y i 兲
1
⫻sgn共 y⫺y i 兲 ⫹
6 兺ik vi •F(3)
i sgn共 y⫺y k 兲
1
⫹
6A 兺
i jk
vi •F(3)
i 关 sgn共 y⫺y i 兲 ⫹sgn共 y⫺y j 兲
1 3
⫹
6 兺i j vi •F(3)
i sgn共 y⫺y j 兲 ⫹ 兺 u•Fi
6 i
(3)
⫹sgn共 y⫺y k 兲兴 . 共A3兲
⫻sgn共 y⫺y i 兲 . 共A2兲 To prove that Eq. 共A3兲 关i.e., three-body contribution to Eq.
共27兲兴 is identical to the three-body contribution of Eq. 共26兲,
The-three body potential contribution can be isolated as let us expand Eq. 共A3兲:

1 1
qy 共 y,t 兲 ⫽⫺
J U(3)
2A 兺i 关 vi ⫺u共 y 兲兴 •F(3)
i sgn共 y⫺y i 兲 ⫹
6A 兺
i jk
vi •F(3)
i 关 sgn共 y⫺y i 兲 ⫹sgn共 y⫺y j 兲 ⫹sgn共 y⫺y k 兲兴

1 1
⫽⫺
2A 兺i vi •F(3)
i sgn共 y⫺y i 兲 ⫹
2A 兺
i
u•F(3)
i sgn共 y⫺y i 兲

1
⫹
6A 兺
i jk
vi •F(3)
i 关 sgn共 y⫺y i 兲 ⫹sgn共 y⫺y j 兲 ⫹sgn共 y⫺y k 兲兴

1 1
⫽⫺
2A 兺i vi •F(3)
i sgn共 y⫺y i 兲 ⫹
6A 兺
u•
i jk
兵 F(3)
i sgn共 y⫺y i 兲 ⫹F j sgn共 y⫺y j 兲 ⫹Fk sgn共 y⫺y k 兲 其
(3) (3)

1
⫹
6A 兺
i jk
vi •F(3)
i 关 sgn共 y⫺y i 兲 ⫹sgn共 y⫺y j 兲 ⫹sgn共 y⫺y k 兲兴

冦 冧
兺i j F(3)
i j 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y j 兲兴 ⫹

3 1
兺i vi •F(3) u• 兺 Fik 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y k 兲兴 ⫹
(3)
⫽⫺ i sgn共 y⫺y i 兲 ⫹
6A 6A ik

兺jk F(3)
jk 关 sgn共 y⫺y j 兲 ⫺sgn共 y⫺y k 兲兴

031111-10
PRESSURE TENSOR AND HEAT FLUX VECTOR FOR . . . PHYSICAL REVIEW E 69, 031111 共2004兲

1 1
⫹
6A 兺
i jk
vi • 共 F(3)
i j ⫹Fik 兲 sgn共 y⫺y i 兲 ⫹
(3)
6A 兺
i jk
vi • 共 F(3)
i j ⫹Fik 兲
(3)

1
⫻sgn共 y⫺y j 兲 ⫹
6A 兺
i jk
vi • 共 F(3)
i j ⫹Fik 兲 sgn共 y⫺y k 兲
(3)

1 1
⫽⫺
6A 兺i j 关 vi ⫺u共 y 兲兴 •F(3)
i j 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y j 兲兴 ⫺
6A 兺
ik
关 vi ⫺u共 y 兲兴 •F(3)
ik 关 sgn共 y⫺y i 兲 ⫺sgn共 y

1 1
⫺y k 兲兴 ⫺
6A 兺
i jk
vi •F(3)
i j 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y k 兲兴 ⫺
6A 兺
i jk
vi •F(3)
ik 关 sgn共 y⫺y i 兲 ⫺sgn共 y⫺y j 兲兴

1
⫹
6A 兺jk u共 y 兲 •F(3)
jk 关 sgn共 y⫺y j 兲 ⫺sgn共 y⫺y k 兲兴 . 共A4兲

Equation 共A4兲 is just the three-body contribution to Eq. 共26兲, as required.

关1兴 M. Kroger and S. Hess, Phys. Rev. Lett. 85, 1128 共2000兲. 关16兴 P. J. Daivis, K. P. Travis, and B. D. Todd, J. Chem. Phys. 104,
关2兴 J. D. Moore, S. T. Cui, P. T. Cummings, and H. D. Cochran, 9651 共1996兲.
AIChE J. 43, 3260 共1997兲. 关17兴 A. Baranyai, Phys. Rev. E 54, 6911 共1996兲.
关3兴 S. Bair, C. McCabe, and P. T. Cummings, Phys. Rev. Lett. 88, 关18兴 T. Ikeshoji, B. Hafskjold, and H. Furuholt, Mol. Simul. 29, 101
058302-1 共2002兲. 共2003兲.
关4兴 P. J. Daivis, M. L. Matin, and B. D. Todd, J. Non-Newtonian 关19兴 B. Hafskjold and H. Furuholt, Phys. Rev. E 66, 011203 共2002兲.
Fluid Mech. 111, 1 共2003兲. 关20兴 S. T. Cui, P. T. Cummings, and H. D. Cochran, J. Chem. Phys.
关5兴 L. I. Kioupis and E. J. Maginn, J. Phys. Chem. B 103, 10 781 114, 7189 共2001兲.
共1999兲. 关21兴 F. Varnik, J. Baschnagel, and K. Binder, J. Chem. Phys. 113,
4444 共2000兲.
关6兴 K. P. Travis and K. E. Gubbins, J. Chem. Phys. 112, 1984
关22兴 K. Van Workum and J. J. de Pablo, Phys. Rev. E 67, 031601
共2000兲.
共2003兲.
关7兴 S. T. Cui, C. McCabe, P. T. Cummings, and H. D. Cochran, J.
关23兴 O. G. Jepps and S. K. Bhatia, Phys. Rev. E 67, 041206 共2003兲.
Chem. Phys. 119, 8941 共2003兲.
关24兴 J. A. Barker, R. A. Fisher, and R. O. Watts, Mol. Phys. 21, 657
关8兴 F. Varnik and K. Binder, J. Chem. Phys. 117, 6336 共2002兲.
共1971兲.
关9兴 J. Delhommelle and D. J. Evans, Mol. Phys. 100, 2857 共2002兲.
关25兴 B. M. Axilrod and E. Teller, J. Chem. Phys. 11, 299 共1943兲.
关10兴 X. J. Fan, N. Phan-Thien, N. T. Yong, and X. Diao, Phys. 关26兴 G. Marcelli, B. D. Todd, and R. J. Sadus, Phys. Rev. E 63,
Fluids 14, 1146 共2002兲. 021204 共2001兲.
关11兴 S. Y. Liem, D. Brown, and J. H. R. Clarke, Phys. Rev. A 45, 关27兴 S. H. Lee and P. T. Cummings, J. Chem. Phys. 99, 3919
3706 共1992兲. 共1993兲; 101, 6206 共1994兲.
关12兴 J. H. Irving and J. G. Kirkwood, J. Chem. Phys. 18, 817 关28兴 J. A. Barker and A. Pompe, Aust. J. Chem. 21, 1683 共1968兲.
共1950兲. 关29兴 M. V. Bobetic and J. A. Barker, Phys. Rev. B 2, 4169 共1970兲.
关13兴 B. D. Todd, D. J. Evans, and P. J. Daivis, Phys. Rev. E 52, 关30兴 P. J. Leondard and J. A. Barker, in Theoretical Chemistry:
1627 共1995兲. Advances and Perspectives, edited by H. Eyring and D. Hend-
关14兴 B. D. Todd, P. J. Daivis, and D. J. Evans, Phys. Rev. E 51, erson 共Academic, London, 1975兲 Vol. 1.
4362 共1995兲. 关31兴 G. Marcelli and R. J. Sadus, J. Chem. Phys. 111, 1533 共1999兲.
关15兴 D. R. J. Monaghan and G. P. Morriss, Phys. Rev. E 56, 476 关32兴 K. P. Travis, B. D. Todd, and D. J. Evans, Phys. Rev. E 55,
共1997兲. 4288 共1997兲.

031111-11