You are on page 1of 7

Oscillatory surface tension due to finite-size

effects
Cite as: J. Chem. Phys. 123, 114702 (2005); https://doi.org/10.1063/1.2018640
Submitted: 18 March 2005 . Accepted: 13 July 2005 . Published Online: 19 September 2005

Pedro Orea, Jorge López-Lemus, and José Alejandre

ARTICLES YOU MAY BE INTERESTED IN

Computer simulations of liquid/vapor interface in Lennard-Jones fluids: Some questions and


answers
The Journal of Chemical Physics 111, 8510 (1999); https://doi.org/10.1063/1.480192

Molecular dynamics simulation of the orthobaric densities and surface tension of water
The Journal of Chemical Physics 102, 4574 (1995); https://doi.org/10.1063/1.469505

The surface tension of TIP4P/2005 water model using the Ewald sums for the dispersion
interactions
The Journal of Chemical Physics 132, 014701 (2010); https://doi.org/10.1063/1.3279128

J. Chem. Phys. 123, 114702 (2005); https://doi.org/10.1063/1.2018640 123, 114702

© 2005 American Institute of Physics.


THE JOURNAL OF CHEMICAL PHYSICS 123, 114702 共2005兲

Oscillatory surface tension due to finite-size effects


Pedro Orea
Ingeniería Molecular, Instituto Mexicano del Petróleo, Eje Central Lázaro Cárdenas 152, México DF
07730, México
Jorge López-Lemus
Facultad de Ciencias, Universidad Autónoma del Estado de México, Avenida Instituto Literario 100, Toluca
CP 50000, México
José Alejandrea兲
Departamento de Química, Universidad Autónoma Metropolitana-Iztapalapa, Avenida San Rafael Atlixco
186, Col. Vicentina, México DF 09340, México
共Received 18 March 2005; accepted 13 July 2005; published online 19 September 2005兲

The simulation results of surface tension at the liquid-vapor interface are presented for fluids
interacting with Lennard Jones and square-well potentials. From the simulation of liquids we have
reported 关M. González-Melchor et al., J. Chem. Phys. 122, 4503 共2005兲兴 that the components of
pressure tensor in parallelepiped boxes are not the same when periodic boundary conditions and
small transversal areas are used. This fact creates an artificial oscillatory stress anisotropy in the
system with even negative values. By doing direct simulations of interfaces we show in this work
that surface tension has also an oscillatory decay at small surface areas; this behavior is opposite to
the monotonic decay reported previously for the Lennard Jones fluid. It is shown that for small
surface areas, the surface tension of the square-well potential artificially takes negative values and
even increases with temperature. The calculated surface tension using a direct simulation of
interfaces might have two contributions: one from finite-size effects of interfacial areas due to box
geometry and another from the interface. Thus, it is difficult to evaluate the true surface tension of
an interface when small surface areas are used. Care has to be taken to use the direct simulation
method of interfaces to evaluate the predicted surface tension as a function of interfacial area from
capillary-wave theory. The oscillations of surface tension decay faster at temperatures close to the
critical point. It is also discussed that a surface area does not show any important effect on
coexisting densities, making this method reliable to calculate bulk coexisting properties using small
systems. © 2005 American Institute of Physics. 关DOI: 10.1063/1.2018640兴

I. INTRODUCTION simulation methods. By using the Gibbs ensemble Monte


Carlo simulations Smit3 found that the phase diagram is very
González-Melchor et al. have found1 in simulations of sensitive to long-range interactions. The critical temperature
liquids and ionic fluids that the finite-size effect and periodic increases by around 20% when the Lennard Jones potential
boundary conditions introduce an artificial oscillatory stress is spherically truncated and shifted 共LJSTS兲 at a distance of
anisotropy in bulk phases when the simulations are per- 2.5␴ compared with the full interaction. The phase equilib-
formed on parallelepiped boxes. This box shape is regularly rium properties can also be obtained by direct simulation of
used to directly simulate the liquid-vapor and liquid-liquid
interfaces, and, for the same potential model, the results are
interfaces. The stress anisotropy is important when the trans-
the same as those obtained using methods without
versal area of the parallelepiped is small and when the liquid
interfaces.4,5 Most simulations to obtain surface tension have
density is close to that of phase equilibrium conditions.
used periodic systems having two liquid-vapor interfaces
González-Melchor et al. used the same definition of surface
where a liquid slab is surrounded by vapor; however, other
tension to calculate the stress anisotropy and found positive
and negative values. At small transversal areas this unphysi- setups have also been used,6 for instance, confining the in-
cal stress anisotropy has the same order of magnitude with terfaces between flat surfaces. The surface tension has also
that calculated in a liquid-vapor interface. An oscillatory be- been obtained using bulk simulations combined with
havior of surface tension with surface area was also obtained histogram-reweighting Monte Carlo 共MC兲 and finite-size
by direct molecular-dynamics 共MD兲 simulations at the scaling techniques.7–9 In direct simulations of interfaces the
liquid-vapor interface of ionic fluids.2 system is not homogeneous and the long-range corrections
The liquid-vapor equilibrium of Lennard Jones 共LJ兲 flu- are not in general easy to include; however, the problem is
ids have been widely studied in the last three decades using eliminated by using a lattice sum method to calculate the
long-range interactions.5
a兲
Author to whom correspondence should be addressed. Electronic mail: During the last two decades the effect that the number
jra@xanum.uam.mx of particles,10 the range of interaction,4,5 and the amount of

0021-9606/2005/123共11兲/114702/6/$22.50 123, 114702-1 © 2005 American Institute of Physics


114702-2 Orea, López-Lemus, and Alejandre J. Chem. Phys. 123, 114702 共2005兲

liquid in the slab11 have on the surface tension ␥ has been When two particles are apart a distance r the LJSTS
analyzed. Trockymchuk and Alejandre4 showed that when potential USTS共r兲 is defined as
the surface tension of a truncated LJ potential is obtained by
MC or MD techniques, the forces at the discontinuity of the
potential have to be included. Ignoring that contribution
USTS共r兲 = 再 ULJ共r兲 − ULJ共Rc兲, r 艋 Rc
0, r ⬎ Rc ,
冎 共1兲

gives a nonconstant normal pressure profile. Recently, Duque


where Rc is the distance where the potential is truncated and
and Vega analyzed12 how the long-range corrections can be
ULJ共r兲 is the Lennard-Jones function given by
added consistently in a simulation to calculate interfacial
properties. By using lattice sum expressions for the disper-
sion interactions of the LJ potential López-Lemus and
Alejandre5 showed that phase equilibria and interfacial prop-
ULJ共r兲 = 4␧ 冋冉 冊 冉 冊 册

r
12


r
6
, 共2兲

erties can be obtained for the full interaction, avoiding the where ␴ is the diameter of a LJ particle and ⑀ the well depth.
use of long-range corrections. On the other hand, the surface The full LJ potential is obtained by using a lattice sum
tension of square-well 共SW兲 fluids of different interaction method 共LSM兲 in the dispersion interactions.5,19,20 The at-
ranges has previously been reported using methods with and tractive term in Eq. 共2兲 for one-component systems can be
without interfaces. The agreement between results obtained written as
using both methods was excellent.13–15 N−1 N

Some simulations have been performed to analyze the
effect that surface area has on surface tension. Direct MD
U6 = 兺 兺
i=1 j⬎i r
ij
6 =U
real
+ Urecip + Uself , 共3兲
ij
simulations at the liquid-vapor interface of LJ fluids and
molten salts were performed by Chen16 and Aguado et al.,17 where N is the number of molecules and ␭ij = −␭i␭ j with
respectively. Monte Carlo simulations of bulk systems using ␭ j = 2冑⑀␴3. The real term is

冉 冊
histogram reweighting and finite-size techniques have also N−1 N
been applied8,9 to calculate the surface tension. All of these a−2 −a2
studies found that surface tension decayed monotonically
Ureal = ␯6 兺 兺
L i=1 j⬎i
兺 ␭ij a−6 + a−4 + 2
e , 共4兲
with surface area as predicted by the capillary-wave theory18
共CWT兲. where L is the lattice vector of the periodic array of image
The main goal of this work is to show that the surface cells, a = 兩ri − r j − RL兩␯, where ri is the position of molecule i
tension of LJ and SW fluids at the liquid-vapor interface and RL is the lattice translation vector. The quantity ␯ is an
oscillates with surface area. The origin of the oscillatory be- adjustment parameter which indicates how fast the real term
havior discussed in this work comes from the use of a par- is damped.
allelepiped shape of the simulation box with a small surface The reciprocal contribution is
area and periodic boundary conditions. Thus, the calculation N N
␲3/2 ␲3/2␯3
of the total surface tension contains an artificial contribution
due to box geometry and has to be taken into account in the
U recip
=− 兺
24V h⫽0
S共h兲S共− h兲B共h兲 + 兺 兺 ␭ij ,
6V i=1 j=1
computer simulation of interfaces. The true surface tension is
共5兲
not properly computed when direct simulations of interfaces
are performed with small surface areas, and it is difficult to where

冋 冉 冊 册
evaluate the results predicted by the capillary-wave theory
using this simulation method. We will show that Chen acci- 1 1 −b2
B共h兲 = h3 ␲1/2 erfc共b兲 + 3 − e , 共6兲
dentally performed16 simulations of LJ fluids at the liquid- 2b b
vapor interface on boxes with dimensions where it is not
possible to observe the oscillation on surface tension. The with b = h / 2␯. In Eq. 共5兲, V is the volume of the unit cell, h
simulations were performed using squared interfacial areas, is the reciprocal lattice vector with magnitude h = 兩h兩, and
and the results are analyzed in terms of the box length where S共h兲 is the structure factor defined as
it is easier to look at the wavelength of the oscillations. N
The remainder of the paper is organized as follows: In S共h兲 = 兺 ␭ jeih·r j . 共7兲
Sec. II is given the potential models and the definition of j=1
calculated properties. The details about MD and MC simula-
The second term in Eq. 共5兲 arises from the reciprocal space
tions are presented in Sec. III. The results of the simulations
when h = 0.
are presented and discussed in Sec. IV. Finally, conclusions
Finally, the self-correction term comes from the exclu-
from our findings are given in Sec. V.
sion of i = j when L = 0 and it is written as
N
␯6
U self
= − 兺 ␭ii . 共8兲
II. POTENTIAL MODELS AND PROPERTIES 12 i=1

In this work we used the LJSTS and the full LJ and SW Expressions for forces and components of the pressure
potentials to simulate the liquid-vapor interface at different tensor are found in Refs. 5 and 20
temperatures and surface areas. The SW potential is defined as
114702-3 Oscillatory surface tension due to finite size effects J. Chem. Phys. 123, 114702 共2005兲

冦 冧
⬁ if r ⬍ ␴ the system. The LSM was used to evaluate the dispersion
u共r兲 = − ⑀ if ␴ 艋 r ⬍ ␭␴ 共9兲 forces of the full LJ model. The truncation distance was 2.5␴
for the LJSTS model and for the short-range forces of the
0 if r 艌 ␭␴ , full LJ. In the LSM, the reciprocal lattice vectors in the long-
where ␴ is the diameter of a hard-sphere particle, ⑀ is the est direction were chosen according to kzmax = 共Lz / Lx兲kmax x ,
well depth, and ␭ is the range of the attractive interaction. with a typical value of kmax x = 6. The error in the reduced
Hereafter, the ␴ and ⑀ parameters have the same value in all energy of dispersion interactions for the full LJ model was
the interaction potentials. less than 10−3. A multiple time-step scheme was used to re-
In liquid-vapor interfaces the vapor pressure by hydro- duce the computational cost of the simulation.22 The interac-
static equilibrium is related with the normal component Pzz tions in the real part were calculated using a reduced time
which can be calculated using the mechanical definition step ⌬t* = ⌬t共⑀ / m␴2兲1/2 = 0.005 共where m is the mass of each
N−1 N particle兲; for the reciprocal part ⌬t* = 0.05 was used. The ca-
1
Pzz = ␳kBT + 兺
V i=1

j⬎i
zijFzij , 共10兲
nonical molecular-dynamics simulations were performed us-
ing the Nosé-Hoover chain thermostat,23 one thermostat per
atom. The thermostat parameter was one hundred times the
where zij = zi − z j, zi being the z coordinate of atom i. The smaller time step. In all cases, the reduced temperatures T*
force component between atoms i and j in the z direction is = kBT / ⑀ varied from the triple to the critical points. The sys-
du共r兲 zij tem was equilibrated for at least 5 ⫻ 105 time steps, and after
Fzij = − . 共11兲 this period the coexisting densities, pressure, and surface ten-
dr r
sion were calculated every 20 configurations in 10 blocks of
The derivative of the potential for the LJSTS and full LJ 105. The error 共=␴ / N1/2 b 兲 in the simulation was calculated
potentials were calculated using Eqs. 共1兲 and 共2兲; in the latter using these Nb blocks, where ␴ is a standard deviation.
case Eq. 共3兲 was used to calculate the dispersion interactions. The range of interaction for the SW potential was
The du共r兲 / dr of the square-well potential was calculated in ␭ = 1.5. The simulations of the SW model were performed
every discontinuity by using a standard MC in the canonical ensemble with a neigh-
du共r兲 bor list to speed up the code.24 The setup of the system was
= − kBT␦共r − ␴兲, 共12兲 similar to that of the LJSTS and full LJ potentials. In this
dr case Lz* = 40 for all the MC simulations and the total reduced
where kB is the Boltzmann’s constant, T is the absolute tem- density was kept constant to ␳* = ␳␴3 = 0.35. The simulations
perature and ␦ is a delta function. We have shown that the were performed in cycles; in every cycle it was attempted to
derivative of discontinuous potentials can be obtained during move the particles within the neighbor list. The system was
MC simulations13,21 if the ␦ function is calculated numeri- equilibrated for 106 cycles and after this period the proper-
cally. ties were calculated every 20 cycles in 10 blocks of 106.
The surface tension of a planar interface was calculated Equation 共12兲 was used to calculate the components of the
from the average components of the pressure tensor by using pressure tensor. The approximated ␦ function was
the mechanical definition calculated13 using ⌬␴ / ␴ = 0.005, 0.010, 0.015, and 0.020.
The results obtained using these ⌬␴s were used to extrapo-
Lz
␥= 兵具Pzz典 − 0.5关具Pxx典 + 具Pyy典兴其, 共13兲 late the pressure components and surface tension when
2 ⌬␴ → 0.
where Lz is the box length in the z direction and the factor The reduced coexisting densities were calculated via the
1 / 2 is because there are two interfaces in the system. density profile ␳共z兲 = 具N共z兲典 / ⌬V, where N共z兲 is the average
The liquid and vapor densities were calculated at the end number of molecules in every slab and ⌬V is its volume.
of simulations from the average density profiles.
IV. RESULTS
III. SIMULATION DETAILS The main result of this work is the surface tension as a
function of surface area for LJSTS, full LJ and SW poten-
The interfacial properties at the liquid-vapor interface of tials. The effect of the surface area on coexisting densities is
the LJSTS and full LJ potentials were calculated using MD also discussed for the LJSTS model. The three models have
simulations, while the MC method was used to simulate the different triple and critical points. The reduced critical tem-
SW model. The simulation cell was parallelepiped in shape perature is around 1.3 for the full LJ, 1.1 for the LJSTS
with sides Lz ⬎ Lx = Ly and surface area A = Lx ⫻ Ly. The dis- truncated at 2.5␴, and 1.22 for the SW model with interac-
tance is made dimensionless by dividing it by ␴. In simula- tion ␭ = 1.5.
tions of LJSTS and full LJ potentials Lz* = 50. The initial
setup of the simulation consisted of a liquid slab surrounded
A. LJSTS potential
by two vacuum regions. The particles in the liquid region
were allocated using a fcc crystalline array. Simulations at Figure 1 shows the surface tension as a function of L*x
different surface areas were performed, and the amount of for the LJSTS model at three temperatures. The results ob-
liquid in the simulation cell, around 20␴ in the z direction, tained by Chen16 are also shown at T* = 0.72. Chen per-
was kept constant by increasing the number of molecules in formed his calculations to show that surface tension decayed
114702-4 Orea, López-Lemus, and Alejandre J. Chem. Phys. 123, 114702 共2005兲

FIG. 2. Coexisting densities as a function of Lx* at T* = 0.72 共open circles兲,


0.8 共open squares兲, and 0.9 共open triangles兲. 共a兲 Liquid and 共b兲 Vapor. The
FIG. 1. Surface tension of the LJSTS fluid as a function of Lx*. The reduced
errors in the results from this work are less than symbols.
temperature is shown in the figure. The open circles are results from this
work and 共 *兲 joined by a dashed line are results of Chen 共Ref. 16兲. The
continuous lines are guides to the eye. The filled circles are results from from this work tend to the same value. Sides et al. used their
Sides et al. 共Ref. 25兲. The errors in the results from this work are “less than”
symbols. results to calculate the width of the interface and test the
predictions of the capillary-wave theory for surface tension.
They found that when finite size effects are negligible, the
monotonically with surface area according to CWT predic- width of the interface ⌬ follows a function,

冉 冊
tions. The dashed line is a fit of the simulation data of Chen
to a function according to a Gaussian model of the CWT, k BT Lx
⌬2 = ⌬20 + ln , 共15兲
2␲␥ ␰
␥共L*x 兲 = ␥共L*x = ⬁兲 + a/L*2
x , 共14兲
where ␰ is the correlation length.
where ␥共L*x = ⬁兲
is the surface tension of an infinite system. The value of ⌬ was obtained by fitting the calculated
The results from this work do not follow the same trend; density profile to an error function and to a hyperbolic tan-
instead, an oscillatory function was found. Chen accidentally gent function. The surface tension calculated using ⌬ fitted
performed simulations using integer values of L*x , and it was to an error function was in good agreement with that ob-
a casualty that the surface tension decayed monotonically. If tained through the pressure tensor components. From Fig. 1
he would have chosen noninteger values of L*x , for instance, we can see that it is not possible to validate, using Eq. 共14兲
5.5 and 6.5, his surface tensions would have increased. His with small surface areas, the predictions of the CWT about
simulations are correct and his results are in good agreement how surface tension decays with surface area using direct
with those from this work at the same value of L*x . In a recent interface simulations.
work we have found1 that when bulk simulations are per- The coexisting densities are essentially independent of
formed on parallelepiped boxes, the system has an artificial surface area. Figure 2 shows the liquid and vapor densities as
stress anisotropy when the transversal area is small. The an- functions of L*x for the LJSTS model at three temperatures. It
isotropy was calculated for simple liquids and ionic fluids is seen that for systems having L*x ⬎ 6 is enough to obtain the
using the same definition that is used to calculate the surface correct densities for the potential model. This means that
tension of an interface, and as a result it was observed that it small systems can be used to obtain coexistence properties
oscillates around zero. For small surface areas it has the using direct simulation of interfaces. The same conclusion
same order of magnitude compared with the surface tension can be drawn from the interface simulations of Chen.
in a real liquid-vapor interface. The stress anisotropy in liq-
uids oscillates due to finite size in transversal areas and pe-
B. Full LJ potential
riodic boundary conditions, the so called periodic error. The
oscillations of surface tension shown in Fig. 1 are also due to The effect of long-range forces on the oscillations of
the fact that simulations are performed in parallelepiped cells surface tension is analyzed by using the full LJ model. Fig-
and the true surface tension of the interface cannot be iso- ure 3 shows the surface tension as a function of L*x for this
lated when small surface areas are used. It is seen in Fig. 1 potential at different temperatures. As expected, the ampli-
that oscillations decay faster with surface area and tempera- tude due to larger correlations is higher in this model than in
ture. In Fig. 1 are also shown results for larger systems the LJSTS model, as can be seen at T* = 0.72. This is ex-
共L*x ⬎ 12 and Lz* ⬎ 127兲 obtained by Sides et al.25 The results pected because at this temperature the full LJ is closer to the
114702-5 Oscillatory surface tension due to finite size effects J. Chem. Phys. 123, 114702 共2005兲

FIG. 5. Surface tension of the square-well fluid as a function temperature.


FIG. 3. Surface tension of the full LJ fluid as a function of Lx*. The reduced The range of interaction is ␭ = 1.5. The values of Lx* used to calculate the
temperatures are shown in the figure. The errors in the results from this work surface area are shown in the figure. The critical point 共Ref. 15兲 is shown
are less than symbols. with 共 *兲. The errors in the results from this work are less than symbols.

triple point. As the temperature increased, the oscillation de- high temperature. It is interesting to note that although this
cays faster and the maxima and minima are slightly shifted potential is very short ranged, the oscillation at T* = 0.9 per-
to larger values of L*x . It was not our interest to obtain the sists up to L*x = 10. The surface tension for systems using
surface tension of the infinite system but to show that surface L*x = 12 are shown with star symbols.13 The results from this
tension in this model also oscillates with surface area. We work decay at the same value. At L*x = 5 the surface tension is
would like to emphasize that T* = 0.72 is close to the triple negative, certainly this is not a real surface tension but the
point. result of using a small surface area in a parallelepiped cell
with periodic boundary conditions.
C. SW potential The effect of temperature on surface tension at a con-
As an example of a discontinuous potential, the SW stant surface area is shown in Fig. 5 for the SW model. The
model was chosen to estimate the surface tension varying the values of ␥* decreases monotonically except in systems with
interfacial area and temperature. The results are shown in the smallest area, A = 5 ⫻ 5␴2, where the surface tension
Fig. 4 for values of L*x changing in steps of 0.5. The same shows an anomalous increase. This fact is due to the use of
trend was found compared with that of LJSTS and full LJ periodic boundary conditions in a cuboid cell; a similar ef-
models. The amplitude on the oscillating function decreases fect was found in bulk simulations using the same geometry
with temperature. Actually, the oscillation is damped faster at in the simulation box. It is interesting to see that at T* = 0.9
the surface tension changes from −0.2 to 0.8 when L*x varies
from 5 to 5.5, respectively. The surface tension as a function
of interfacial area and temperature has a triangle shape with
one vertex in the critical point. Its value up and down with L*x
at a given T* and the result for the largest system are almost
in the middle of the side opposite to that vertex. Most of the
results, as a function of surface area, tend to the critical
point, shown with a star.15

V. CONCLUSIONS

We have used the direct simulation of interfaces to show


that periodic boundary conditions and small surface areas
have a strong effect on surface tension. An oscillatory func-
tion of ␥* as a function of L*x was found in the three potential
models used in this work; i.e., the range of interaction is not
responsible for such oscillations. The wavelength of the os-
cillating function is around ␴ for all the potential models.
The oscillation comes from the use of periodic boundary
FIG. 4. Surface tension of the SW fluid as a function Lx* for ␭ = 1.5. The
conditions in a noncubic cell with small surface areas. The
reduced temperatures are shown in the figure. The 共 *兲 are results taken from
Orea et al. 共Ref. 13兲. The errors in the results from this work are less than geometry of the simulation box introduces an artificial con-
symbols. tribution to the components of the pressure tensor, and it
114702-6 Orea, López-Lemus, and Alejandre J. Chem. Phys. 123, 114702 共2005兲

6
seems to be difficult to evaluate properly the true surface G. A. Chapela, G. Saville, S. M. Thompson, and J. S. Rowlinson, J.
Chem. Soc., Faraday Trans. 2 73, 1133 共1977兲.
tension of the interface using small systems. This has impor- 7
J. J. Potoff and A. Z. Panagiotopoulos, J. Chem. Phys. 112, 6411 共2000兲.
tant consequences in the calculation of interfacial properties 8
J. R. Errington, Phys. Rev. E 67, 012102 共2003兲.
of molecular fluids where direct simulation of an interface 9
J. K. Singh and D. A. Kofke, Mol. Simul. 30, 343 共2004兲.
10
and noncubic cells are generally used. Preliminary results C. D. Holcomb, P. Clancy, and J. A. Sollweg, Mol. Phys. 78, 437 共1993兲.
11
show that the softness of the potential decreases the ampli- J. Weng, S. Park, J. R. Lukes, and C. Tien, J. Chem. Phys. 113, 5917
共2000兲.
tude of the oscillations on surface tension. Work in this di- 12
D. Duque and L. F. Vega, J. Chem. Phys. 121, 8611 共2004兲.
rection is being performed. The finite-size effect of the sur- 13
P. Orea, Y. Duda, and J. Alejandre, J. Chem. Phys. 118, 5635 共2003兲.
face area in coexisting densities is small, and the direct 14
P. Orea, Y. Duda, V. C. Weiss, W. Schroer, and J. Alejandre, J. Chem.
simulation of interfaces can be used to calculate them using Phys. 120, 11754 共2005兲.
15
small systems. J. K. Singh, D. A. Kofke, and J. R. Errington J. Chem. Phys. 119, 3405
共2003兲.
16
L. Chen, J. Chem. Phys. 103, 10214 共1995兲.
ACKNOWLEDGMENTS 17
A. Aguado, W. Scott, and P. A. Madden, J. Chem. Phys. 115, 8612
All the authors are grateful to CONACyT for financial 共2001兲.
18
M. P. Gelfand and M. E. Fisher, Physica A 166, 1 共1990兲.
support. J.L.L. also thanks PROMEP-México for Grant No. 19
J. López-Lemus and J. Alejandre, Mol. Phys. 101, 743 共2003兲.
FE03/2005. P.O. thanks IMR-Mexico for Grant No. 20
N. Karasawa and W. A. Goddard III, J. Phys. Chem. 93, 7320 共1989兲.
21
D.00343. M. González-Melchor, A. Trokhymchuk, and J. Alejandre, J. Chem.
Phys. 115, 3862 共2001兲.
22
1
M. González-Melchor, F. Bresme, P. Orea, J. López-Lemus, and J. Ale- M. E. Tuckerman, B. J. Berne, and G. J. Martyna, J. Chem. Phys. 94,
jandre, J. Chem. Phys. 122, 4503 共2005兲. 6811 共1991兲.
2 23
M. González-Melchor, F. Bresme, and J. Alejandre, J. Chem. Phys. 122, M. E. Tuckerman, Y. Liu, G. Ciccoti, and G. J. Martyna, J. Chem. Phys.
104710 共2005兲. 115, 1678 共2001兲.
24
3
B. Smit, J. Chem. Phys. 96, 8639 共1992兲. D. Frenkel and B. Smit, Understanding Molecular Simulation 共Aca-
4
A. Trockymchuk and J. Alejandre, J. Chem. Phys. 111, 8510 共1999兲. demic, New York, 1996兲.
5
J. López-Lemus and J. Alejandre, Mol. Phys. 100, 2983 共2002兲. 25
S. W. Sides, G. S. Grest, and M. Lacasse, Phys. Rev. E 60, 6708 共1999兲.

You might also like