Sei sulla pagina 1di 12

Crystal nucleation in binary hard sphere mixtures: A Monte Carlo simulation study

S. Punnathanam and P. A. Monson

Citation: The Journal of Chemical Physics 125, 024508 (2006); doi: 10.1063/1.2208998
View online: http://dx.doi.org/10.1063/1.2208998
View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/125/2?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in


Free energy barriers for homogeneous crystal nucleation in a eutectic system of binary hard spheres
J. Chem. Phys. 138, 174503 (2013); 10.1063/1.4802777

Doubled heterogeneous crystal nucleation in sediments of hard sphere binary-mass mixtures


J. Chem. Phys. 135, 134115 (2011); 10.1063/1.3646212

Crystallization and dynamical arrest of attractive hard spheres


J. Chem. Phys. 130, 064504 (2009); 10.1063/1.3074310

Confined crystallization of cylindrical diblock copolymers studied by dynamic Monte Carlo simulations
J. Chem. Phys. 124, 244901 (2006); 10.1063/1.2203070

Monte Carlo simulations for the phase behavior of symmetric nonadditive hard sphere mixtures
J. Chem. Phys. 118, 7907 (2003); 10.1063/1.1563595

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Mon, 11 Aug 2014 05:40:41
THE JOURNAL OF CHEMICAL PHYSICS 125, 024508 2006

Crystal nucleation in binary hard sphere mixtures: A Monte Carlo


simulation study
S. Punnathanam and P. A. Monsona
Department of Chemical Engineering, University of Massachusetts, Amherst, Massachusetts 01003
Received 11 January 2006; accepted 8 May 2006; published online 14 July 2006

We present calculations of the nucleation barrier during crystallization in binary hard sphere
mixtures under moderate degrees of supercooling using Monte Carlo simulations in the
isothermal-isobaric semigrand ensemble in conjunction with an umbrella sampling technique. We
study both additive and negatively nonadditive binary hard sphere systems. The solid-fluid phase
diagrams of such systems show a rich variety of behavior, ranging from simple spindle shapes to the
appearance of azeotropes and eutectics to the appearance of substitutionally ordered solid phase
compounds. We investigate the effect of these types of phase behavior upon the nucleation barrier
and the structure of the critical nucleus. We find that the underlying phase diagram has a significant
effect on the mechanism of crystal nucleation. Our calculations indicate that fractionation of the
species upon crystallization increases the difficulty of crystallization of fluid mixtures and in the
absence of fractionation azeotropic conditions the nucleation barrier is comparable to pure fluids.
We also calculate the barrier to nucleation of a substitutionally ordered compound solid. In such
systems, which also show solid-solid phase separation, we find that the phase that nucleates is the
one whose equilibrium composition is closer to the composition of the fluid phase.
2006 American Institute of Physics. DOI: 10.1063/1.2208998

I. INTRODUCTION tem as a function of the nucleus size shows a maximum. The


nucleus for which this maximum occurs is called the critical
It is well known that fluid mixtures are generally harder nucleus and the difference between this maximum of the free
to crystallize compared to pure fluids.15 In addition, the energy and that of the metastable fluid is the work of forma-
solid-fluid phase behavior of mixtures shows a rich variety tion of the critical nucleus G*. Only those crystal nuclei
compared to that of pure fluids. The phase diagrams of even larger than the critical nucleus grow further in size to form
binary mixtures613 vary from simple spindle shaped struc- the new solid phase. Nucleation, hence, is an activated pro-
ture to complex ones showing azeotropes, eutectic points, cess and its rate depends on G*,
peritectic points, compound solid formation, etc. Although
there have been some studies3,4,1417 of crystal nucleation in j = exp G*/kBT, 1
fluid mixtures using molecular simulations, the influence of
where j is the rate of formation of the critical nuclei per unit
these phenomena on the dynamics of crystal formation is
volume per unit time, kB is Boltzmann constant, T is the
largely unknown and it is of fundamental interest to under-
absolute temperature, and is the kinetic factor which gives
stand the mechanism of crystal formation in such systems.
the rate of growth of the critical nucleus. Thus the rate of
The purpose of this paper is to study the mechanism of crys-
nucleation depends upon the thermodynamics properties of
tal formation using molecular simulations in model systems
the critical nucleus. Although crystal nucleation has been
that exhibit the above phenomena. widely studied, our understanding of its mechanism remains
The primary mechanism of crystal formation from a su- incomplete.19 This is mainly because direct experimental ob-
percooled fluid is via nucleation.18,19 A qualitative under- servation of the critical nucleus is very difficult19 since i it
standing of nucleation is provided by the classical nucleation is microscopic in size consisting of few hundreds of par-
theory CNT developed in the works of Volmer,20 Becker,21 ticles, ii its formation is very rare typically
Zeldovich,22 Frenkel,23 and others. According to the theory, 101 106 nuclei cm3 s, and iii its lifetime is very short it
the process of formation of a nucleus requires work to form either rapidly grows to form a solid phase or melts back into
the interface between the crystal and the fluid. This is offset the fluid. However, recent experiments on colloidal
by the reduction in the free energy upon formation of the systems24,25 show considerable promise. Colloidal systems
solid phase. The former is proportional to the surface area of behave similar to atomistic systems and can form
the nucleus and is positive while the latter is proportional to crystals.19,26 However, due to their much larger size, the dy-
the volume of the nucleus and is negative. For small sizes the namics of colloidal crystallization is much slower giving an
surface term dominates while for larger sizes the volume opportunity to observe the critical nucleus directly using
term dominates. As a result the total free energy of the sys- techniques such as confocal microscopy or laser light scat-
tering and obtain information about its composition and
a
Electronic mail: monson@ecs.umass.edu structure.19,25,27

0021-9606/2006/1252/024508/11/$23.00 125, 024508-1 2006 American Institute of Physics


This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Mon, 11 Aug 2014 05:40:41
024508-2 S. Punnathanam and P. A. Monson J. Chem. Phys. 125, 024508 2006

CNT models the critical nucleus as a spherical solid to search out more efficiently the regions of phase space
cluster suspended in the metastable fluid and its thermody- associated with crystal nuclei without the attendant difficul-
namic properties are approximated by the corresponding val- ties of molecular dynamics. They applied their technique to
ues for a bulk solid phase. The value for the interfacial ten- study crystal nucleation in simple atomic systems consisting
sion between the nucleus and the surrounding metastable of Lennard-Jones46 or hard sphere particles47 and showed
fluid is approximated by the planar surface tension between that it was possible to obtain the free energy barrier to nucle-
the two phases at coexistence. For a binary system consisting ation as well as size and structure of the critical nucleus for
of species A and B, G* can be obtained by solving the supercoolings as low as 20%.
following sets of equations:28 In this work, we apply the methods of Frenkel and co-
workers to study crystal nucleation to binary hard sphere
AlP = AsP + P* PvAs , 2 mixtures. Hard sphere systems were studied because they are
the most simple models that exhibit solid-fluid phase
BlP = BsP + P* PvBs , 3 transition48 and their thermodynamic properties are well
known. In the case of hard sphere systems, crystallization is
2 achieved by isothermally compressing the system above its
r* = , 4 freezing pressure. However, we will be using the term super-
P P
*
cooling to denote compression above the freezing pressure
since the dimensionless hard sphere pressure may be inter-
4 *2
G* = r , 5 preted as the reciprocal of a dimensionless temperature and
3 thus compression of a hard sphere system may be interpreted
where i is the chemical potential, vi is the partial molar as an isobaric cooling process.49 The phase diagrams for bi-
volume of component i, P* is the pressure inside the nucleus, nary hard sphere systems have been computed
P is the pressure of the fluid, and r* is the radius of the previously713,50 and they are a good model for studying the
spherical nucleus. The superscripts l and s denote the fluid underlying physics involved in solid-fluid phase equilibria of
and solid phases, respectively. The main criticisms of CNT binary mixtures. Based on a survey of the data for phase
are the use of the properties of the macroscopic phases to diagrams for binary organic mixtures Matsuoka6 were able to
model the critical nucleus and the use of the planar surface classify them into six different types. Binary hard sphere
tension to obtain the surface energy of the nucleus. These systems exhibit five of those six types of phase diagrams,
assumptions get increasingly tenuous at higher degrees of illustrating the importance of packing and molecular size dif-
supercooling when the critical nucleus is microscopic in size ferences upon solid phase stability and crystal structure. The
leading to a breakdown in the predictions of CNT. However, only phase diagrams that binary hard sphere mixtures do not
CNT remains an important framework and a significant exhibit are those that show congruent melting of compound
amount of research,29 as well as our general understanding of solids. However, as we shall see later, congruent melting can
the mechanism, is still grounded in the description provided be obtained if one introduces nonadditivity in the hard sphere
by CNT and captured in Eq. 1. diameter for the interaction between the unlike components.
Due to the microscopic nature of the critical nucleus, This idea comes from studies of solid-fluid phase diagrams
molecular simulations are an attractive tool for studying of symmetric nonadditive Lennard-Jones 12-6 mixtures by
crystal nucleation. As described in a recent review by Auer Vlot et al.51 who showed that such systems can exhibit con-
and Frenkel,30 molecular simulation studies of crystal nucle- gruent melting.
ation using direct molecular dynamics MD simulations The remainder of the paper is organized as follows. In
have many difficulties because nucleation is a rare event and Sec. II we describe the methods used for simulation of crys-
under experimental conditions they occur at time scales that tal nucleation in binary hard sphere systems. The results
are more than 20 orders of magnitude larger than those in- from our simulation study for crystal nucleation in binary
volved with molecular motion. Hence crystallization simula- mixtures are presented in Sec. III. Finally, a summary of our
tions were performed at very large supercoolings3,3144 where results and conclusions is presented in Sec. IV.
the nucleation rates are higher. But as explained by Auer and
Frenkel,30 the mechanism may be different than in experi- II. MODELS AND SIMULATION DETAILS
ments due to the simultaneous formation of multiple inter- A. Models
acting crystal nuclei. To overcome this, they have suggested
an indirect approach4,27,4547 which takes advantage of the The model systems used in our study are binary hard
form of the nucleation rate given by Eq. 1. Instead of di- sphere mixtures comprised of two species denoted as A and
rectly calculating the nucleation rate j, one instead calculates B. The interaction potential uijr between two particles be-
G* and separately and substitutes it back into Eq. 1 to longing to species i and j is given by
obtain j. In this paper our focus is on the calculation of G*
for binary mixtures. The basic idea is to use umbrella sam-
pling combined with replica switch methods to sample the
uijr = if r ij
0 if r ij ,
6

fluctuations that give rise to crystal nuclei and to calculate where r is the intermolecular distance and ij is the distance
the free energy of formation of these nuclei as a function of of closest approach between two particles belonging to spe-
their size. These biased sampling methods make it possible cies i and j. According to this notation, AA is the hard
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Mon, 11 Aug 2014 05:40:41
024508-3 Crystal nucleation in mixtures J. Chem. Phys. 125, 024508 2006

sphere diameter of species A and BB is the hard sphere nuclei containing n particles. Nn is related to Gn, the
diameter of species B. In this paper, we study both additive work of formation of a nucleus containing n particles by27
and nonadditive hard sphere systems. By convention, the
larger species is denoted by A and the smaller species by B.
The ratio of the diameters is given by = BB / AA. The hard
Nn
N
= exp
Gn
k BT
, 9

sphere interaction distance between species A and B is de- where kB is Boltzmanns constant, T is the absolute tempera-
noted by ture, and N is the total number of particles in the system. We
AB = 21 AA + BB1 + , 7 calculate Prn via Monte Carlo simulations. Once Gn is
obtained using Eq. 9 for a range of values of n, we locate
where = 0 for the case of additive hard spheres and 0 the maximum in the plot of Gn versus n. The location of
for nonadditive hard spheres. the maximum determines the critical nucleus and the height
of the maximum gives G*. Although the method described
B. Calculation of phase diagram there pertains to simulation of crystal nucleation in single
Before one proceeds with study of crystal nucleation, it component systems, with suitable modifications we can ex-
is essential to calculate the solid-fluid phase diagram of the tend it to study crystal nucleation in binary systems.
system. The phase diagrams for additive hard sphere systems
studied here have been previously calculated by various 1. Identification of crystal nuclei
authors.1113,50 We have recalculated some of these phase As explained in Ref. 27, the simulation relies on the
diagrams and also the phase diagram for a nonadditive sym- ability to identify a crystalline nucleus inside a metastable
metric hard sphere mixture with = 0.1. In the case of ad- fluid. The method is based on the local bond analysis of
ditive hard sphere systems, we use the equation of state of Steinhardt et al.56 We define for every particle i, a
Mansoori et al.52 for fluid mixtures. The equation of state for 2l + 1-dimensional complex vector qli, with components
fluid mixtures of nonadditive hard spheres and substitution-
Nbi
ally disordered solid mixtures is calculated by isothermal- 1
isobaric semigrand ensemble see Sec. II C 2. These simu- qlmi =
Nbi

j=1
Y lmrij, 10
lations give us the composition of the mixture as a function
of the pressure and the chemical potential difference. Com- where Nb is the number of neighboring particles and Y lm are
bining these calculations with the equation of state for pure the spherical harmonics for the normalized direction vector
hard spheres53,54 gives us the chemical potentials of both rij, between particles i and j. Neighbors are defined as those
species at various compositions and pressures. For the case particles which lie within a given cutoff radius rc. For the
of substitutionally ordered solids, we also need their Gibbs calculations in this paper, we use a value of rc = 1.3AA.
free energies. These are calculated using the method of Fren- Similar to calculations in Ref. 27, we use a value of l = 6. In
kel and Ladd.55 a crystalline environment, these vectors are aligned with
The calculation of the solid-fluid equilibria between fluid each other but in a fluidlike environment they are randomly
mixtures and substitutionally disordered solid solutions is oriented. To obtain a measure of the alignment, we calculate
done by equating the pressures and the chemical potentials of the correlation between the vectors q6 of neighboring par-
each species between the two phases. For the case of the ticles i and j,
equilibrium between the fluid mixture and the substitution- 6
ally ordered solid compound, the equilibrium conditions are
modified by the stoichiometric constraints.8 In general for a
q6i . q6j = q6miq6m
m=6
*
j, 11

compound solid AmBn, where the two species are present in


where
the ratio m : n, we have
q6mi
m + nGAmBn = mAl + nBl , 8 q6mi = . 12
q6mi
where GAmBn is the Gibbs free energy of the compound solid
The asterisk in Eq. 11 denotes the complex conjugate.
and Al and Bl are the chemical potentials of species A and B
Equation 11 is a dot product between two normalized vec-
in the fluid, respectively. At equilibrium the two phases sat-
tors and gives the cosine of the angle between them. A value
isfy Eq. 8 at the same pressure.
of 0.7 is chosen to determine whether the two neighboring
particles are aligned or not. We identify those particles as
C. Simulation of nucleation
solid if the number of aligned neighbors are greater than 5
This paper focuses on the calculation of G*. We closely and as fluid otherwise. Once all the solid particles are iden-
follow the method developed by Frenkel and tified, we then determine the particles to belong to the same
co-workers4,27,4547 for calculating G*. For complete details cluster as they are separated by a distance less than rc.
of the method, the reader should consult Ref. 27. The value We can also analyze the structure of the crystal nuclei
of G* can be obtained from the probability of formation of formed during simulations using these order parameters. The
the critical nucleus. Let Prn be the probability of observing vectors qli depend on the orientation of the coordinate sys-
at least one nucleus containing n particles. When the value of tem used. From these vectors one can obtain local second
Prn is very low, then Nn = Prn, where Nn is the number of order invariants defined as
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Mon, 11 Aug 2014 05:40:41
024508-4 S. Punnathanam and P. A. Monson J. Chem. Phys. 125, 024508 2006

qli = 4
l

qlmi2
2l + 1 m=l
1/2

. 13
mentation, the reader is once again referred to Auer and
Frenkel.27 The biasing potential is a harmonic function
given by
In order to determine the structure of the crystal nucleus, we rN = 21 knrN n0, 15
first calculate the distribution of qli for the particles of the
nucleus. Comparison of the distribution with that of bulk where nrN is the size of the largest cluster and k and n0 are
solid phases will help in identifying the structure of the the parameters of the biasing potential. Application of this
nucleus. biasing potential gives us cluster size distributions centered
around n0 with a width that depends on k. We use a value of
2. Isothermal-isobaric semigrand ensemble k = 0.15 for all the calculations shown in this paper. The free
energy profile is obtained by running simulations with differ-
During the simulations, the crystal nucleus that is
ent values of n0, such that one obtains cluster size distribu-
formed has in general a concentration that is different from
tions for a set of overlapping windows. During simulations,
that of the fluid. Hence, the simulations were performed in
stacking rearrangements in the cluster are very slow. In order
the isothermal-isobaric semigrand ensemble. In this en-
to overcome the barriers towards stacking arrangements, all
semble we fix the temperature T, pressure P, total number of
the windows are simulated in parallel using the parallel tem-
particles N, and the chemical potential difference be-
pering technique of Geyer and Thompson.61 Again our
tween the two species. This allowed us to maintain the fluid
implementation follows very closely with that given in
at the same degree of supercooling even after the crystal
Ref. 27.
nucleus is formed. The simulation box is cubic in shape.
During simulations one of particles is kept on the surface of
4. Augmented semigrand ensemble
the simulation box and its movement is restricted to that one
face of the cube. This particle is the shell particle and it The above method works well when is close to unity.
defines the volume of the system.57,58 The partition function However, when becomes smaller, it becomes harder to
of this semigrand ensemble for a binary mixture is given as change the diameter of a particle and the efficiency of par-


ticle identity changes becomes very poor. As a result it will
1
Y=
B3NN! identities
NA
dV dr2 drNeU+PV , 14 take much longer for the composition of the mixture to reach
the equilibrium value. One way to overcome this problem is
to change the diameter of the particles gradually in several
where U is the potential energy of the system of N particles stages. In the literature, many methods6272 have been pro-
at positions r1 rN, NA is the number of particles of species posed to implement staged insertion and deletion of particles
A, = 1 / kBT, V is the volume of the system, for dense systems. Our method extends the concept of the
= eB3 / A3 , = A B, i is the chemical potential of augmented ensembles developed by Kaminsky67 for the
species i, and i is its de Broglie wavelength. The summa- grand canonical ensemble to the semigrand ensemble. Ac-
tion is done for every particle where its identity changes cordingly, the partition function for the augmented semi-
from one component to another.59 During simulations, we grand ensemble containing m 1 intermediate stages is given
make particle displacements and volume changes. In addi- as
tion, we also switch the identity of each particle between the


m1
two species, i.e., we change the diameter of the particle. The 1
acceptance or rejection of these moves is governed by the Y=
B3NN! i=0
N
i/midentities A dV dr2 drNeU+PV .
chemical potential difference. Typically during simulations,
samples are made after every Monte Carlo move. In the case 16
of the cluster size distribution, however, the calculations in- The system now consists of NA particles of species A,
volved during the sampling are computationally expensive N NA 1 particles of species B, and one particle in the in-
and the samples are strongly correlated. Hence cluster size termediate particle i. There are m 1 stages of the interme-
distributions are calculated after a fixed number of Monte diate particle. The intermediate stage is arbitrary but for
Carlo moves called a trajectory. For all the calculations, the smooth interpolation it is chosen to be a hard sphere particle
averages were collected over 100 000 trajectories following with a diameter whose value lies between that of species A
an equilibration period of 50 000 trajectories. Repeated and B. Each intermediate stage has a particular value of the
simulations indicate that the uncertainties in the values of the hard sphere diameter associated with it. The value of i in-
calculated free energies of formation of the nuclei are kBT. creases from 0 to m and the hard sphere diameter increases
accordingly. When the value of i equals 0, it corresponds to
3. Umbrella sampling with parallel tempering species B and i = m corresponds to species A. We anticipate
In our simulations, we are interested in the properties of that the intermediate particle diameters that give the best
the critical cluster. At the degrees of supercooling that we results are those where the reversible work of increasing the
study, the free energies of formation of these clusters are particle diameter is the same for each stage. As a first ap-
higher than 15kBT. Hence their occurrence in a normal unbi- proximation, we choose particle diameters such that the in-
ased Monte Carlo simulation is very rare. In order to simu- crease in volume of the particle between stages are equal. In
late these large clusters, we resort to umbrella sampling the augmented ensemble, the identity changes are carried out
technique of Torrie and Valleau.60 For details of its imple- only for the intermediate particle. During the identity change
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Mon, 11 Aug 2014 05:40:41
024508-5 Crystal nucleation in mixtures J. Chem. Phys. 125, 024508 2006

TABLE I. Simulation states and results for crystal nucleation in additive binary hard sphere systems. The first
line in the table corresponds to a pure component hard sphere system. P* = P3AA / kBT. = A B.

P* A xA n* G*

1.00 17.000 21.59 1.0 93 18


0.97 18.000 1.44 21.86 0.50 100 19
0.92 21.000 4.21 24.72 0.50 127 22
0.92 22.875 3.32 25.68 0.23 104 18
0.92 22.650 2.29 24.64 0.10 98 18
0.80 21.800 11.00 25.87 0.75 95 25

moves, the value of i is increased or decreased by 1. When and at a given composition forms a substitutionally disor-
no intermediate particles are present, i.e., when i = 0 or i = m, dered solid solution.1113,50 A pure fluid with diameter
one particle is selected at random and the value of i is set to freezes at P3 / kBT 11.7. At equilibrium, the solid phase is
0 if it belongs to species B or to m if it belongs to species A. richer in species A compared to the fluid. However, for the
During simulations only contributions to the partition func- current size difference of = 0.97, the extent of fractionation
tion when i equals 0 or m are considered. This yields correct is very small.
averages for the semigrand ensemble given by Eq. 14. To compare the mechanism of crystal nucleation in a
mixture with that of the pure component, we carried out
III. RESULTS AND DISCUSSION simulations at pressures P* = PAA3
/ kBT 1.5P*f , where P*f is
We carried out Monte Carlo simulations to calculate the the freezing pressure of the fluid. This corresponds to
properties of the critical nucleus in binary hard sphere fluids. roughly 33% supercooling. The total number of particles in
The calculations were performed for both additive and non- the system was 2916. Cluster size distributions were sampled
additive binary hard spheres. For the case of additive hard after every Monte Carlo trajectory which consisted of 20
spheres, systems with = 0.97, 0.92, and 0.80 were studied. displacements per particle, 10 volume moves, and 20 identity
The nonadditive hard sphere system studied had values of changes per particle. In Fig. 2 we show a comparison of the
= 1.0 and = 0.10. The choice of these systems was influ- work of formation of the crystal nucleus for a fluid mixture
enced by the desire to study the effect of fluid composition, with mole fraction xA = 0.5 with that of the pure component.
fractionation, solid-solid immiscibility, and compound solid Table I gives the numerical values from our calculations. The
formation. The states studied are listed in Tables I and II. curves show a maximum indicating that the fluid is meta-
stable and further growth of the crystal nucleus should lead
A. Additive hard sphere system to a decrease in the free energy associated with a phase tran-
sition. We observe that the free energies of formation of the
1. Nucleation in fluid mixtures
nuclei are comparable for both the mixture and the pure
The first system that we studied was a binary hard sphere fluid. Figure 3 shows the composition of the crystal nucleus
mixture with = 0.97. The phase diagram for the as a function of its size. We can clearly see that the crystal
system,1113,50 recalculated by us, is shown in Fig. 1. Since nucleus has a higher concentration of the larger species A
the size difference between the two components is very than the fluid indicating fractionation. Moreover we see that
small, the two components are completely miscible with the fractionation occurs for all sizes, both precritical and
each other in both fluid and solid phases. The mixture is postcritical, implying that fractionation of the components
close to ideal in the fluid phase and shows only mild nonide- precedes the occurrence of crystalline order. This fraction-
ality in the solid phase. The phase diagram is spindle shaped
and there is very little fractionation upon crystallization.
Such phase diagrams are formed by binary mixtures when
the two components are quite similar to each other. In this
case, the solid phase has a face-centered-cubic fcc structure

TABLE II. Simulation states and results for crystal nucleation for symmetric
nonadditive binary hard sphere systems with = 0.10. The first line in the
table corresponds to a pure component hard sphere system. P* = P3AA / kBT.
= A B.

P* B xA n* G*

17.0 21.59 0.0 93 18


18.5 9.75 22.86 0.1 164 31
21.6 9.30 25.54 0.2 170 35 FIG. 1. Phase diagram for a binary hard sphere system with size ratio
24.8 7.70 27.78 0.3 88 24 = 0.97. The solid phase denoted by the letter S is a substitutionally disor-
19.9 3.50 22.33 0.4 68 22 dered fcc crystal and the fluid phase is denoted by the letter L. The symbols
18.5 0.00 19.59 0.5 61 22 represent the coexistence points determined from simulations and the lines
are a guide to the eye.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Mon, 11 Aug 2014 05:40:41
024508-6 S. Punnathanam and P. A. Monson J. Chem. Phys. 125, 024508 2006

FIG. 2. Comparison of the free energy barrier to crystal nucleation between


a pure hard sphere fluid solid line and a binary mixture dashed line with
= 0.97 and xA = 0.5, simulated at conditions shown in Table I.

ation should lead to an increase in the value of G* for the


mixture. However, the extent of fractionation in this mixture
is very small and the differences in G* between the mixture FIG. 4. Color Computer graphics visualization showing a plane cut
and the pure fluid are smaller than the uncertainty in our through a configuration from a Monte Carlo simulation of a crystallizing
estimation of G*. We have also analyzed the structure of binary hard sphere mixture with size ratio = 0.97 and simulated at condi-
the nuclei formed during simulations. In their simulations, tions shown in Table I. Inside the nucleus, particles belonging to species A
are shown in white and those belonging to species B are shown in light gray
Auer and Frenkel27 found that for the pure hard sphere fluid, green. In the bulk, species A are shown in medium gray red and species
the critical nucleus is made of up randomly stacked layers of B are shown in dark gray blue. The hexagonal ordering of a 111 crystal
close packed hexagonal crystal planes. This structure is plane is evident in the cluster.
called random hexagonal close packed structure RHCP.
Figure 4 is visualization of a cross section of the critical is richer in species A, when xA 0.23 the solid is richer in
nucleus for the fluid mixture which also exhibits the RHCP species B and at the azeotropic composition the solid has an
structure. In the following sections, we show calculations of identical composition to that of the fluid. Consequently, we
G* for binary mixtures with smaller values of . These expect little fractionation to occur upon crystallization for
systems show a much larger degree of fractionation upon fluid with compositions close to the azeotropic composition.
crystallization and thus provide good examples for studying Studying crystal nucleation in this system should reveal the
its influence on G*. influence of fractionation upon G*.
We have calculated the nucleation barrier at points with
2. Nucleation in an azeotropic mixture compositions that lie on either side of the azeotrope and at
the azeotropic composition. The calculations were carried
The next system that we considered had a value of out at P* 1.5P*f . Again the simulations were performed on
= 0.92. The phase diagram for this system is shown in Fig. 5. 2916 particles and each Monte Carlo trajectory consisted of
Similar to the previous case the two components are com- 20 displacements per particle, 10 volume moves, and 20
pletely miscible with each other in both phases. However, the identity changes per particle. Figures 6 and 7 show the free
larger size difference leads to an increase in the nonideality, energy change upon crystal nucleation and the composition
especially in the solid phase, and the system now shows an of the crystal nuclei as a function of the size of the nuclei,
azeotrope at xA 0.23. Hence, when xA 0.23 then the solid

FIG. 5. Phase diagram for a binary hard sphere system with size ratio
FIG. 3. Comparison of the composition of the crystal nucleus solid line = 0.92. The solid phase denoted by the letter S is a substitutionally disor-
inside the metastable binary hard sphere fluid having = 0.97 with that to dered fcc crystal and the fluid phase is denoted by the letter L. There is an
the fluid dashed line and the corresponding stable solid dotted line. The azeotrope formation at xA 0.23. The symbols represent the coexistence
simulations are done at conditions given in Table I. points determined from simulations and the lines are a guide to the eye.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Mon, 11 Aug 2014 05:40:41
024508-7 Crystal nucleation in mixtures J. Chem. Phys. 125, 024508 2006

FIG. 6. Comparison of the free energy barrier to crystal nucleation between FIG. 8. Phase diagram for a binary hard sphere system with size ratio
a pure hard sphere fluid and a binary mixture with = 0.92 and simulated at = 0.80. The two components A and B are essentially immiscible in each
conditions shown in Table I. The compositions xA of the bulk fluid are 1.0 other. The solid phases are denoted by the letters A and B which correspond
solid line, 0.5 dashed line, 0.23 dot-dashed line, and 0.1 dotted line. to fcc crystals of pure A and B components, respectively. The fluid phase is
denoted by the letter L. The eutectic point occurs at xA 0.23. The open
respectively. The numerical values from these calculations squares denote state points at which we were successful in studying nucle-
ation and the filled squares denote state point at which we were
are listed in Table I. Based on these calculations we can unsuccessful.
make a number of observations. G* is highest when xA
= 0.50 where correspondingly the degree of fractionation in
the cluster is also the highest. Near the azeotropic composi- in the solid phase. The phase diagram,50 shown in Fig. 8, has
tion of xA = 0.23, the crystal nuclei formed during nucleation a eutectic point at xA 0.23. For pressures P* between 11.7
have a composition similar to that of the fluid phase. In other and 28.0, the fluid mixture coexists with pure A or pure B
words, there is an absence of fractionation. This lack of frac- crystal. At higher pressures, the region of the phase diagram
tionation has a strong effect on the nucleation barrier. As we corresponds to a two phase region of pure A and pure B
can see from Fig. 6, the nucleation barrier at the azeotropic crystals in equilibrium with each other. This type of phase
composition is comparable to that of the pure component. behavior was found to be the most common in the survey by
For the composition of xA = 0.10, the fractionation is now Matsuoka.6 The state points at which nucleation simulations
reversed and the crystal nuclei are richer in component B were carried out are also shown in Fig. 8. These points cor-
which is the smaller sphere. This behavior corresponds to respond to approximately 33% supercooling, i.e., P* P*f .
that of the phase diagram. However, the extent of fraction- We were able to simulate the critical nucleus only at xA
ation involved is very small and hence the nucleation barrier = 0.75. This state point lies in the two phase region made up
is comparable to that of the pure component. These observa- of pure A crystal and the fluid mixture. For this mixture, the
tions lead to the conclusion that crystal nucleation barriers in chemical potential difference between the two species is very
mixtures are strongly influenced by the extent of fraction- large and hence, during simulations we used two intermedi-
ation that is required prior to the formation of a crystalline ate stages to make identity changes between the species.
nucleus. Analysis of the structure of critical nucleus for each Thus during our simulations, on an average only one-third of
of three cases show that they are comprised of hexagonal the configurations contained no intermediate particles. Hence
layers particles randomly stacked one above another and thus for these simulations, each trajectory was now defined as
exhibiting a RHCP structure. consisting of six displacement moves per particle, three vol-
ume moves, and six identity changes per particle where only
3. Nucleation in a system with a eutectic point those moves were considered where the configurations had
When = 0.80, the size difference between the two com- no intermediate particles. The total number of particles in our
ponents is so large that they become essentially immiscible simulations was 4000. The results are shown in Fig. 9 and

FIG. 7. Comparison of the composition of the crystal nucleus solid lines inside the metastable binary hard sphere fluid having = 0.92 with that to the fluid
dashed lines and the corresponding stable solid dotted lines. The simulations are done at conditions given in Table I. The figures correspond to bulk fluid
compositions xA of a 0.5, b 0.23, and c 0.1.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Mon, 11 Aug 2014 05:40:41
024508-8 S. Punnathanam and P. A. Monson J. Chem. Phys. 125, 024508 2006

FIG. 9. Comparison of the free energy barrier to crystal nucleation between FIG. 11. Phase diagram for a symmetric binary hard sphere system with
a pure hard sphere fluid solid line and a binary mixture dashed line with negative nonadditivity of = 0.10. Two types of solid phases occur. The
= 0.80 and xA = 0.75, simulated at conditions shown in Table I. substitutionally disordered fcc solid solution is denoted by the letter S and
the substitutionally ordered compound solid with CsCl structure is denoted
by the letter R. The fluid phase is denoted by the letter L. The squares
Table I. Again we see that the nucleation barrier is higher in indicate the state points at which the nucleation simulations were performed.
the case of the mixture than the pure component. Figure 10 is
a visualization of the critical nucleus. The core of the nucleus
is a pure RHCP crystal of particle A which corresponds to the congruent melting/freezing. Congruent melting/freezing is
phase diagram. We were unable to calculate the nucleation important because the compound solid can melt into a fluid
barriers for other state points shown in Fig. 8. The main having the same composition. Thus congruent melting/
difficulty stemmed from the inability to grow crystal nuclei freezing enables us to study the dynamics of compound solid
from the fluid. Our repeated attempts to simulate nucleation formation from a supercooled fluid in the absence of other
by varying the trajectory lengths and parameters in biasing phase transitions. Such phase behavior is the second most
potential were unsuccessful. The reasons for the lack of suc- common among binary organic mixtures.6 The compound
cess using this method are not clear and further investiga- solids formed by additive spheres do not show congruent
tions will be needed to study crystal nucleation at these state melting/freezing and upon melting they undergo a phase
points. transition to form a fluid with different compositions and a
solid with different structures and compositions. Occurrence
of congruent melting/freezing for symmetric nonadditive
B. Nonadditive hard sphere system mixtures of binary Lennard-Jones particles was first shown
by Vlot et al.51 They showed that binary mixtures can form
Crystal nucleation simulations were also performed on
substitutionally ordered solids that can exhibit congruent
binary mixtures of nonadditive hard spheres. Nonadditive
melting/freezing. Such behavior is typical of chiral com-
hard spheres are interesting because unlike additive hard
pounds and study of these systems could yield insights into
spheres, the compound solids formed by these systems show
the mechanism of solidification in chiral mixtures. We ex-
tended their ideas to compute the phase diagram of the more
simpler symmetric binary hard sphere mixture with negative
nonadditivity. In this model, AA = BB and = 0.1. The
computed phase diagram for this system is shown in Fig. 11.
As we can see, the phase diagram shows a double eutectic.
The system forms a substitutionally ordered solid at a com-
position of xA = 0.5. The compound solid has a CsCl structure
which is a body centered cubic structure with one type of
species at the eight corners of the cube and the other species
at the center of the cube. Moreover, the compound solid
shows congruent melting/freezing.
We performed crystal nucleation simulations at five dif-
ferent compositions of xA = 0.1, 0.2, 0.3, 0.4, and 0.5. The
calculations were performed at the same degree of supercool-
ing, i.e., at pressures P* = 1.5Pz*. These compositions were
chosen to study the influence of the eutectic point and the
effect of formation of the compound solid on nucleation. The
results from these calculations are shown in Table II. Figure
12 gives a comparison of the free energy of formation of
FIG. 10. Color Computer graphics visualization showing a plane cut crystal nuclei inside a metastable fluid with composition xA
through a configuration from a Monte Carlo simulation of a crystallizing
= 0.5 with that of the pure fluid. We immediately see a dif-
binary hard sphere mixture with size ratio = 0.80 and simulated at condi-
tions shown in Table I. The colors used are the same as those in Fig. 4. The ference in the shapes of the two curves. This is an indication
hexagonal ordering of a 111 crystal plane is evident in the cluster. that the structure of the nuclei formed when xA = 0.5 is dif-
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Mon, 11 Aug 2014 05:40:41
024508-9 Crystal nucleation in mixtures J. Chem. Phys. 125, 024508 2006

FIG. 12. Comparison of the free energy barrier to crystal nucleation be- FIG. 14. Comparison of free energy barrier to crystal nucleation inside a
tween a pure hard sphere fluid solid line and a symmetric nonadditive symmetric nonadditive = 0.10 binary hard sphere mixture at various
= 0.10 binary hard sphere mixture dashed line with xA = 0.5. compositions for conditions shown in Table II. These compositions xA are
0.1 solid line, 0.2 dotted line, 0.3 dashed line, and 0.4 dot-dashed line.

ferent from that inside the pure fluid. Visualization of the


crystal nuclei formed when xA = 0.5 shows that they have the mation inside the metastable fluid at various other composi-
CsCl structure see Fig. 13. We also observe from Fig. 12 tions and Fig. 15 shows the composition of the nuclei formed
that although there is no fractionation involved during the compared with that of the bulk fluid and the solid. These
nucleus formation, G* is higher for the mixture than the calculations show the effect of solid-solid phase separation
pure fluid. This result is unlike that of nucleation in an azeo- and the presence of the eutectic point on the mechanism of
tropic fluid where the absence of fractionation leads to an crystal nucleation. When xA = 0.1 and 0.2, the compositions
insignificant change in G* compared to the pure fluid. This of the nuclei are closer to the equilibrium composition of the
is because there is a compositional ordering of the two spe- substitutionally disordered fcc solid than the substitutionally
cies involved prior to formation of the crystalline solid. An- ordered compound solid. The curves for G have shapes
other interesting observation for the xA = 0.5 case is that the similar to the pure component case indicating that the crystal
crystal nucleus has complete substitutional order right up to nuclei formed have a RHCP structure with substitutional dis-
its interface with the fluid. Or in other words, when ones order. We also see that the nuclei are richer in species B
moves across the fluid-solid interface, not only do the par- which is consistent with the phase diagram. Visualizations of
ticles exhibit translational order for a crystalline environment the crystal nuclei confirm this observation. When xA = 0.4, the
but the two species also exhibit substitutional order with al- composition is closer to that of the compound solid than the
most no intermediate behavior between the fluid and the fcc solid phase. The G curve is now similar to that for xA
compound solid. = 0.5 indicating that the crystal nuclei have the CsCl struc-
Figure 14 shows the free energy of crystal nucleus for- ture. The direction of the fractionation is now reversed
Fig. 15 and the nuclei are richer in species A compared to
the fluid. Thus we can say that the structure of the nuclei is
strongly determined by the phase that is closer in composi-
tion to the fluid.
When xA = 0.3, the composition is very close to the eu-
tectic composition and there is a competing influence of both
types of solid phases. This provides a unique challenge in
simulating crystal nucleation since either one of the two solid
phases can crystallize or we can get a nucleus with a hetero-
geneous structure. The calculated G curve shown in Fig. 14
is different in shape from both the previous two cases and
looks like an intermediate between the two. The curve for the
composition dependence of the nuclei as a function of its
size Fig. 15 shows a crossover in the direction of fraction-
ation. The nuclei are richer in species B when n 20 and
vice versa when n 20. If it is consistent with the phase
diagram, this should indicate that the nuclei should have the
RHCP structure with substitutional disorder for extremely
small nuclei and the CsCl structure with substitutional order
FIG. 13. Color Computer graphics visualization showing a plane cut for large nuclei. Visualizations of the larger nuclei show that
through a configuration from a Monte Carlo simulation of a crystallizing they have the CsCl structure. For smaller nuclei, the number
symmetric non-negatively additive binary hard sphere mixture with
of particles was too few to arrive at a definite conclusion.
= 0.10 and simulated at composition xA = 0.5 and at P* = 18.5. The colors
used are the same as those in Fig. 4. The compound CsCl structure is clearly Our study of nucleation at the eutectic composition remains
evident. For clarity, the particles are shown with a diameter AB. inconclusive and further investigations are needed to under-
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Mon, 11 Aug 2014 05:40:41
024508-10 S. Punnathanam and P. A. Monson J. Chem. Phys. 125, 024508 2006

FIG. 15. Comparison of the composition of the nucleus


solid lines with that of the bulk fluid dashed lines for
the symmetric nonadditive = 0.10 binary hard
sphere mixture. The figures correspond to fluid compo-
sitions xA of a 0.10, b 0.20, c 0.30, and d 0.40.
The simulations are performed at conditions shown in
Table II.

stand the mechanism of crystal nucleation under these con- zation at azeotropic composition, the nucleation barrier is
ditions. The present result was obtained only when we ran comparable to the pure component fluids. The role of frac-
our simulations with equilibration periods that were three tionation in increasing the difficulty of crystal nucleation has
times longer as compared to the other cases. For shorter been observed experimentally from crystal nucleation in col-
equilibration periods, we obtained nuclei with a heteroge- loidal systems.7476 Molecular dynamics simulations of crys-
neous structure, i.e., one that contained a mixture of regions tallization in binary hard sphere mixtures described in Ref. 3
with RHCP structure and the CsCl structure. A similar result also show this behavior.
was obtained in the work of Leyssale et al.73 The current We have also studied crystal nucleation of substitution-
method does not distinguish the structure of the crystal nu- ally ordered compound solids in a symmetric nonadditive
clei formed during simulations and only requires that the hard sphere mixture with = 0.10. The phase diagram for
nuclei show crystalline order. This feature is very advanta- this system shows two types of solid phases. A substitution-
geous in that fluid is free to crystallize into any structure and ally disordered solid solution with fcc structure is formed
the nucleation barrier is independent of our choice of order when one of the components is present in much larger quan-
parameters. However, this creates problems when studying tities but a substitutionally ordered solid with CsCl structure
nucleation at eutectic compositions where two different is formed when the two species are present in equal quanti-
structures have similar free energies and we get nuclei with ties. During nucleation, the two crystal structures compete
heterogeneous structure. In order to determine more conclu- with each other in order to determine the structure of the
sively the structure of the critical nucleus, we will have to nuclei. The structure of the nuclei is determined by the extent
choose the order parameters that will force the system to of fractionation required to form the corresponding solid. We
crystallize into a specific structure. The critical nucleus, by found that the phase that nucleates is the one whose equilib-
definition, will be the one with the lowest free energy barrier. rium composition is closer to the composition of the fluid
phase.
IV. SUMMARY AND CONCLUSIONS For future studies, we would like to investigate the role
of proximity to the eutectic composition on crystal nucle-
In this paper, we have presented our calculations of the
ation. As mentioned earlier, our attempts at studying crystal
crystal nucleation barriers for wide variety of binary hard
nucleation for such systems both additive and nonadditive
sphere mixtures both additive and nonadditive. Unlike pure
hard spheres mixtures were only partially successful. It
fluids, the phase behavior for binary mixtures shows a vari-
would be also interesting to apply this method to study for-
ety ranging from simple spindle shaped phase diagrams to
mation of compound solids for additive binary hard sphere
those showing more complex behavior such as occurrence of
mixtures,8 such as AB, AB2, AB13, etc., and the influence of
azeotropes, eutectic points, solid-solid immiscibility, and
fractionation. Although in our simulations, we have not seen
compound solid formation. Our investigations show a sig-
nucleation of metastable phases, experiments9,10,77 on colloi-
nificant effect of the underlying phase diagram on the dy-
dal systems have shown that it is possible to form metastable
namics of crystal nucleation. Our calculations suggest that in
solid phases if the fluid composition is closer to it. This
general, the nucleation barrier for mixtures is higher than
would allow us to investigate the validity of the Ostwald rule
those for pure components under the same degree of super-
of stages.78
cooling. The composition of the solid phase is in general
different from that of the fluid. Hence crystal nucleation in
ACKNOWLEDGMENT
mixtures is preceded by fractionation of the components
which in turn increases the free energy barrier for phase tran- This work was supported by a grant from the U.S. De-
sition. When fractionation is absent or small, e.g., crystalli- partment of Energy Contract No. DE-FG02-90ER14150.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Mon, 11 Aug 2014 05:40:41
024508-11 Crystal nucleation in mixtures J. Chem. Phys. 125, 024508 2006

1
P. Pusey, Liquids, Freezing and The Glass Transition North-Holland, 39
R. D. Mountain and P. K. Basu, J. Chem. Phys. 78, 7318 1983.
40
Amsterdam, 1991, p. 763. R. D. Mountain and P. K. Basu, J. Chem. Phys. 80, 2730 1984.
2
R. J. Speedy, Mol. Phys. 95, 169 1998. 41
J. D. Honeycutt and H. C. Anderson, Chem. Phys. Lett. 108, 535 1984.
3
T. Gruhn and P. A. Monson, Phys. Rev. E 64, 061703 2001. 42
S. Nose and F. Yonezawa, J. Chem. Phys. 84, 1803 1986.
4
S. Auer and D. Frenkel, Nature London 413, 711 2001. 43
J. Yang, H. Gould, and W. Klein, Phys. Rev. Lett. 60, 2665 1988.
5
D. W. Oxtoby, Nature London 413, 694 2001. 44
W. C. Swope and H. C. Anderson, Phys. Rev. B 41, 7042 1990.
6
M. Matsuoka, Bunri Gijutsu Separation Process Engineering 7, 245 45
J. S. van Duijneveldt and D. Frenkel, J. Chem. Phys. 96, 4655 1992.
1977. 46
P. R. ten Wolde, M. J. Ruiz-Montero, and D. Frenkel, Phys. Rev. Lett.
7
X. Cottin and P. A. Monson, J. Chem. Phys. 99, 1993 1993. 75, 2714 1995.
8
X. Cottin and P. A. Monson, J. Chem. Phys. 102, 3354 1995. 47
S. Auer and D. Frenkel, Nature London 409, 1020 2001.
9
M. D. Eldridge, P. A. Madden, and D. Frenkel, Nature London 365, 35 48
W. G. Hoover and F. H. Ree, J. Chem. Phys. 49, 3609 1968.
1993. 49
10 P. A. Monson and D. A. Kofke, Adv. Chem. Phys. 115, 113 2000.
M. D. Eldridge, P. A. Madden, P. N. Pusey, and P. Bartlett, Mol. Phys. 50
X. Cottin, Ph.D. thesis, University of Massachusetts, 1995.
84, 395 1995. 51
11 M. J. Vlot, J. C. van Miltenburg, and H. A. J. Oonk, J. Chem. Phys. 107,
W. G. T. Kranendonk and D. Frenkel, J. Phys.: Condens. Matter 1, 7735
1989. 10102 1997.
52
12
W. G. T. Kranendonk and D. Frenkel, Mol. Phys. 72, 679 1991. G. A. Mansoori, N. F. Carnahan, K. E. Starling, and T. W. Leland, J.
13
D. Kofke, Mol. Simul. 7, 285 1991. Chem. Phys. 54, 1523 1971.
53
14
J. Anwar and P. K. Boateng, J. Am. Chem. Soc. 120, 9600 1998. N. F. Carnahan and K. E. Starling, J. Chem. Phys. 51, 635 1969.
54
15
J. D. Shore, D. Perchak, and Y. Shnidman, J. Chem. Phys. 113, 6278 K. R. Hall, J. Chem. Phys. 57, 2252 1972.
55
2000. D. Frenkel and A. J. C. Ladd, J. Chem. Phys. 81, 3188 1984.
16 56
M. Mucha and P. Jungwirth, J. Phys. Chem. B 107, 8271 2003. P. J. Steinhardt, D. R. Nelson, and M. Ronchetti, Phys. Rev. B 28, 784
17
D. Zahn, Phys. Rev. Lett. 92, 040801 2004. 1983.
57
18
D. Frenkel and J. P. McTague, Annu. Rev. Phys. Chem. 31, 491 1980. D. S. Corti, Phys. Rev. E 64, 016128 2001.
58
19
J. Maddox, Nature London 378, 231 1995. D. S. Corti, Mol. Phys. 100, 1887 2002.
20
M. Volmer and A. Weber, Z. Phys. Chem. 119, 277 1926. 59
D. Frenkel and B. Smit, Understanding Molecular Simulation: From Al-
21
R. Becker and W. Dring, Ann. Phys. 24, 719 1935. gorithms to Applications, 2nd ed. Academic, San Diego, CA, 2002.
22
Y. B. Zeldovich, Zh. Eksp. Teor. Fiz. 12, 525 1942. 60
G. M. Torrie and J. P. Valleau, Chem. Phys. Lett. 28, 578 1974.
23
J. Frenkel, Kinetic Theory of Liquids Dover, New York, 1955, Chap. 7. 61
C. J. Geyer and E. A. Thompson, Autit.: J. Pract. Theory 90, 909 1995.
24 62
J. L. Harland, S. I. Henderson, S. M. Underwood, and W. van Megen, I. Nezbeda and J. Kolafa, Mol. Simul. 5, 391 1991.
Phys. Rev. Lett. 75, 3572 1995. 63
A. P. Lyubartsev, A. A. Martsinovski, S. V. Shevkunov, and P. N.
25
U. Gasser, E. R. Weeks, A. Schofield, P. N. Pusey, and D. A. Weitz, Vorontsov-Velyaminov, J. Chem. Phys. 96, 1776 1992.
Science 292, 258 2001. 64
Z. H. Liu and B. J. Berne, J. Chem. Phys. 99, 6071 1993.
26
P. N. Pusey and W. van Megen, Nature London 320, 340 1986. 65
P. Attard, J. Chem. Phys. 98, 2225 1993.
27
S. Auer and D. Frenkel, J. Chem. Phys. 120, 3015 2004. 66
K. Ding and J. P. Valleau, J. Chem. Phys. 98, 3306 1993.
28
P. G. Debenedetti, Metastable Liquids: Concepts and Principles Prince- 67
R. D. Kaminsky, J. Chem. Phys. 101, 1994 1994.
ton University Press, Princeton, NJ, 1996. 68
29 N. B. Wilding and M. Mller, J. Chem. Phys. 101, 4324 1994.
A. Laaksonen, V. Talanquer, and D. W. Oxtoby, Annu. Rev. Phys. Chem. 69
F. A. Escobedo and J. J. dePablo, J. Chem. Phys. 103, 2703 1995.
46, 489 1995. 70
30 F. A. Escobedo and J. J. dePablo, J. Chem. Phys. 105, 4391 1996.
S. Auer and D. Frenkel, Annu. Rev. Phys. Chem. 55, 333 2004. 71
31
B. J. Alder and T. E. Wainwright, J. Chem. Phys. 33, 1439 1960. D. A. Kofke and P. T. Cummings, Mol. Phys. 92, 973 1997.
72
32
M. J. Mandell, J. P. McTague, and A. Rahman, J. Chem. Phys. 64, 3699 J. W. Schroer and P. A. Monson, J. Chem. Phys. 118, 2815 2003.
73
1976. J. Leyssale, J. Delhommelle, and C. Millot, J. Chem. Phys. 122, 104510
33
M. J. Mandell, J. P. McTague, and A. Rahman, J. Chem. Phys. 66, 3070 2005.
74
1977. S. Martin, G. Bryant, and W. van Megen, Phys. Rev. Lett. 90, 255702
34
M. Tanemura, Y. Hiwatari, H. Matsuda, T. Ogawa, N. Ogita, and A. 2003.
75
Ueda, Prog. Theor. Phys. 58, 1079 1977. S. Martin, G. Bryant, and W. van Megen, Phys. Rev. E 67, 061405
35
C. S. Hsu and A. Rahman, J. Chem. Phys. 70, 5234 1979. 2003.
36 76
C. S. Hsu and A. Rahman, J. Chem. Phys. 71, 4974 1979. S. Martin, G. Bryant, and W. van Megen, Phys. Rev. E 71, 021404
37
F. F. Abraham, J. Chem. Phys. 72, 359 1980. 2005.
77
38
J. N. Cape, J. L. Finney, and L. V. Woodcock, J. Chem. Phys. 75, 2366 A. B. Schofield, Phys. Rev. E 64, 051403 2001.
1981. 78
W. Ostwald, Z. Phys. Chem. 22, 289 1897.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.6.218.72 On: Mon, 11 Aug 2014 05:40:41

Potrebbero piacerti anche