You are on page 1of 13

PHYSICAL REVIEW E 87, 032307 (2013)

Phase transitions and phase equilibria in spherical confinement


Alexander Winkler, Antonia Statt, Peter Virnau, and Kurt Binder
Institut für Physik, Johannes Gutenberg-Universität Mainz, Staudinger Weg 7, D-55099 Mainz, Germany
(Received 7 December 2012; published 20 March 2013)
Phase transitions in finite systems are rounded and shifted and affected by boundary effects due to the surface
of the system. This interplay of finite size and surface effects for fluids confined inside of a sphere of radius R
is studied by a phenomenological theory and Monte Carlo simulations of a model for colloid-polymer mixtures.
For this system the phase separation in a colloid-rich phase and a polymer-rich phase has been previously studied
extensively in the bulk. It is shown that spherical confinement can strongly enhance the miscibility of the mixture.
Depending on the wall potentials at the confining surface, the wetting properties of the wall can be controlled, and
this interplay between preferential adsorption of one species to the confining surface and bulk unmixing leads to
very special shapes of the loops observed for the chemical potential of the colloids as a function of their packing
fraction. We also discuss the extent to which concepts used for phase transitions in macroscopic systems, such
as critical exponents of the order parameter distinguishing the phases, or the Kelvin equation describing the shift
of the chemical potential at phase coexistence with the radius R, are applicable.

DOI: 10.1103/PhysRevE.87.032307 PACS number(s): 64.75.Xc, 82.70.Dd, 47.57.J−, 64.60.an

I. INTRODUCTION formation of (macroscopically thick) wetting or drying layers,


and hence wetting transitions [37–39] are also rounded [22].
The recent interest in exploring nanomaterials and nanoflu-
Note that such a system is still infinite in two spatial
idic devices [1–3] urgently emphasizes the need for a better
dimensions, and hence sharp phase transitions still occur,
understanding of the phase behavior of materials when one
which are predicted to fall in the universality class of the two-
constrains them to exist in nanoscopic dimensions. While some
dimensional Ising model (see, e.g., Ref. [22]). If we consider
aspects of the general problem have received longstanding
attention (e.g., it is well-known that very small ferromagnetic cylindrical pores of diameter D instead, the system is quasi-
particles can exist in monodomain states and behave as one-dimensional (if the cylinder length Lcyl tends to infinity).
superparamagnets [4–6], and near the Curie temperature their This fact implies the absence of “true” phase transitions (in the
phase transition to the standard paramagnetic phase is rounded sense of sharp singularities of the thermodynamic potentials),
[7]), phase separation of fluids in spherical confinement has but for finite Lcyl , one still may find rather sharp transitions
received attention only recently; e.g., progress in the prepara- between monodomain and multidomain structures [26]. For
tion of miniemulsions as “containers” for polymer blends [8,9] spherical confinement, however, no length scale larger than
allows for the production polymer nanoparticles that exhibit the sphere diameter can evolve, of course, and hence all phase
phase separation [10]; both “Janus particles” [11–13] and transitions are rounded off. Nevertheless, away from criticality
core-shell nanoparticles [14] are clearly of interest. interfacial widths may be much smaller than the sphere
While the general aspects of the phase behavior of simple diameter, and thus rather well-defined two-phase coexistence
fluids or of binary mixtures of small molecules in the bulk inside the sphere becomes possible. The resulting morphology
is well understood [15,16], progress has also been made in of the system depends on the interplay between the interfacial
clarifying phase diagrams and critical phenomena in various free energy between the coexisting phases (in the bulk) and
complex fluids, such as binary polymer blends [17,18], poly- their surface excess free energies due to the wall, analogous to
mer solutions [19,20], colloid-polymer mixtures [21,22], etc., the case of wetting phenomena at planar surfaces [37,39].
clarifying properties of such systems in confined geometries In the present study, we aim to contribute to the under-
is still an active field of research; see, e.g., Refs. [22–27]. The standing of this problem by Monte Carlo simulations of a
special case that is best understood are systems that have finite simple model, supported by a brief outline of the appropriate
linear dimensions but no surfaces, e.g., a cubic L × L × L phenomenological theory (Sec. II). The model that is studied
box with periodic boundary conditions; this case is relevant here (Sec. III) is the Asakura-Oosawa (AO) model [40] of
for computer simulations [28–30], and the finite-size scaling colloid-polymer mixtures. In this model, colloids are modeled
theory [29] accounts for the corresponding rounding and as hard spheres of radius Rc , which must not overlap with each
shifting of phase transitions. The finite-size effects on phase other nor the polymers. The latter, however, are modeled as
coexistence for such systems have also been analyzed recently soft spheres of radius Rp , which may overlap with no energy
[31–35]. However, for systems of experimental interest in cost, and provide the entropic attraction between the colloid
reduced geometry always both finite-size effects and surface particles causing the phase separation in a colloid-rich and
effects due to the constraining walls are present. When a bulk a polymer-rich phase. Despite its simplicity, this model is a
fluid is exposed to a wall (i.e., a semi-infinite geometry), rather good description of real colloid-polymer mixtures [21],
the interactions of the fluid with the wall may give rise to a which also are a very suitable experimental model system for
wealth of interesting wetting and layering phenomena [36–39]. the study of interfacial phenomena [41], wetting layers [42],
Confining the fluid in a slit pore (e.g., Refs. [22–25]), the etc. Experimental evidence [43] and simulations [44,45] show
finite distance D between the confining walls precludes the that the critical behavior of this system falls in the universality

1539-3755/2013/87(3)/032307(13) 032307-1 ©2013 American Physical Society


WINKLER, STATT, VIRNAU, AND BINDER PHYSICAL REVIEW E 87, 032307 (2013)

class of the three-dimensional Ising model [15]. Since colloidal symmetric polymer blend, H is a chemical potential difference
particles have radii of about 1 μm, the direct experimental between the species; for the colloid-polymer mixture, H is the
counterpart to our simulation results (Sec. IV) would require difference between the colloid chemical potential μ and its
us to confine such systems in cavities with radii 10 to 100 μm. value μcoex (ηpr ) at two-phase coexistence, H = μ − μcoex (ηpr ).
Presumably, such studies will be difficult, and one might Here, we have noted the fact that in systems lacking any
ask why we do not use models for polymer mixtures (e.g., symmetries between the coexisting phases, such as colloid-
Refs. [46,47]) in view of the fact that first experiments to study polymer mixtures, phase coexistence occurs at a nontrivial
phase separation of polymer blends confined in miniemulsions value of the chemical potential (unlike the Ising magnet where
(of diameters in the 100-nm range) are available [10]. However, symmetry requires phase coexistence at magnetic field H =
since such studies would require huge computational effort 0), which depends on the temperature-like control variable, ηpr .
(unlike the case of colloid-polymer mixtures the internal The latter is given by the product of fugacity exp(μp /kB T )
conformational degrees of freedom of the polymers need to be of the polymers (μp being their chemical potential) and the
accounted for [46,47]) and the existence of another large length polymer volume, ηpr = (4π Rp3 /3) exp(μp /kB T ). Varying this
scale, namely the correlation length at which the mean-field so-called “polymer reservoir packing fraction”, the density
to Ising crossover [17,46] occurs, complicates the quantitative of polymers in the suspension is controlled. In the Ising
understanding of this system. However, we feel that the case of magnet, ηpr would correspond to inverse temperature (kB T )−1 ,
the AO-model in spherical confinement can serve as a generic and in the polymer blend the corresponding variable is the
model for the interplay of finite-size and surface effects of Flory-Huggins parameter [17]. Following a path H = 0,
all fluids and fluid mixtures separating into coexisting phases the quantities defined above exhibit the well-known critical
under spherical confinement. behavior in the bulk [15]
 β  −ν
M ∝ ηpr − ηp,critr
, ξ ∝ ηpr − ηp,crit
r  , (3)
II. PHENOMENOLOGICAL CONSIDERATIONS
 −γ
A. The order parameter distribution near criticality kB T χ = V (M 2  − M2 ) ∝ ηpr − ηp,crit
r  (4)
A basic quantity often used to discuss finite-size effects on and the critical exponents describing these singularities have
systems exhibiting phase transitions from a homogeneous state their well-known values [49] β ≈ 0.325, ν ≈ 0.63, and γ ≈
toward phase coexistence is the distribution function PL (M) 1.24.
of the “order parameter” M used to distinguish the phases. In As has been said above, in the finite system these singular-
an Ising ferromagnet, M would be the magnetization per spin, ities are rounded off, and one has instead (for H = 0 we need
while in a binary polymer blend (A,B) it would be the relative to take |M| rather than M in the finite system [50])
monomer concentration of one of the species. In the colloid-
polymer mixture, we may take M = ηc − ηccrit , ηc being the |M| = L−β/ν M̃(L/ξ ), χ = Lγ /ν χ̃ (L/ξ ), (5)
colloid packing fraction in the considered volume V and ηccrit where M̃ and χ̃ are scaling functions, with M̃(0), χ̃(0) being
its value at the critical point, Nc being the number of colloids constants. Thus, |M| does not vanish in the finite system at
of radius Rc in the system criticality, and χ does not diverge but reaches a maximum of
   finite height (proportional to Lγ /ν ). Considered as a function
ηc = 4π Rc3 3 Nc /V . (1)
of ηpr , both |M| and χ (and any moments M k  of the order
Strictly speaking, the optimum choice for the order param- parameter distribution function) are analytic. Nevertheless, the
eter actually is a linear combination [48] of ηc − ηccrit and location of the critical point can be inferred from suitable
ηp − ηpcrit , ηp being the packing fraction of the polymers, ratios, such as [50] M 4 /M 2 2 . As an example, Fig. 1
defined analogously [ηp = (4π Rp3 /3)Np /V , with Np being shows corresponding data for the choice Rp /Rc = 0.8 and
the number of polymers in the system]. However, this three choices of V : one sees that any such ratio M 4 /M 2 2
complication is not essential for what follows and, hence, shall is a smooth function of ηpr (actually a straight line for ηpr
r r
be ignored here. close enough to ηp,crit ), but ηp,crit is well defined nevertheless
r
In the distribution function PL (M), the index L stands for from the intersection point of these curves at ηp,crit , since
the linear dimension of the considered volume V [we take L = the ratio U = M 4 /M 2 2 = Ũ (L/ξ ) = Ũ (0) when ξ → ∞,
V 1/3 for a cubic box, but L = R = (3V /4π )1/3 for a spherical so at criticality there is no longer L dependence. Figure 1
cavity]. If one disregards any effects due to the surface of this also illustrates that the order parameter distribution PL (M)
volume V , which is appropriate only for the case of periodic of the AO model at criticality converges to that of the
boundary conditions, of course, PL (M) near the critical point three-dimensional Ising model, but this still is a double-peaked
can be written as [28–30] distribution: if we would locate criticality from the value
ηpr where the shape of the distribution changes from double
PL (M) = Lβ/ν P̃ (MLβ/ν , H L(β+γ )/ν , L/ξ ), L → ∞, (2)
peak to single peak (far in the one phase region PL (M) is
where β and γ are the standard critical exponents of the always a simple Gaussian, see, e.g., Ref. [50]), we would
order parameter M and the susceptibility kB T χ , ν is the
r
underestimate ηp,crit systematically. Thus, while for ηpr 
r
critical exponent of the correlation length ξ of order parameter ηp,crit one may simply estimate the packing fractions of the
fluctuations, P̃ is a scaling function, and H the thermodynamic two coexisting phases from the peak positions of PL (M),
r
variable conjugate to the order parameter. For the Ising this gives systematically wrong results near ηp,crit : we find
magnet, H would simply be the magnetic field; for the a distance between the peak positions of order L−β/ν still for

032307-2
PHASE TRANSITIONS AND PHASE EQUILIBRIA IN . . . PHYSICAL REVIEW E 87, 032307 (2013)

R = L). Just like bulk order parameter and susceptibil-


ity follow from the derivatives M = −(∂fb /∂H )T , χ =
−(∂ 2 fb /∂H 2 )T , keeping the temperature (or temperature-like
variable ηpr , respectively) constant, one finds also the local
order parameter M1 and surface excess order parameter Ms
from analogous derivatives:
M1 = −(∂fs /∂H1 )T ,H ; Ms = −(∂fs /∂H )T ,H1 . (8)
Noting that the free-energy density f can also be expressed
in terms of the order parameter distribution PL (M) as

FIG. 1. (Color online) (a) Plot of the moment ratio M 4 /(M 2 )2
of the AO-model with Rp /Rc = 0.8 versus ηpr for different volumes f (H,H1 ,L) = −(kB T /V ) ln dMPL (M) + const, (9)
as indicated, yielding an estimate of the critical point from the unique
r
intersection point at ηp,crit = 0.7670, consistent with values from it is easily seen that the singular parts of f and fs satisfy
literature [44]. Lengths are measured in units of the colloid  radius, similar scaling properties as written in Eq. (7), in particular
Rc = 1. (b) Rescaled order parameter distribution P (M)/ M 2  fs (H,H1 ,L) = ξ −2 f˜s (H L(β+γ )/ν ,H1 L 1 /ν ,L/ξ ), L → ∞.
versus M/M 2 , with M = ηc − ηc,crit , taken at ηp,crit
r
, for the largest
volume included in (a). The result for the three-dimensional Ising (10)
model is included for comparison. From Eqs. (8) and (10), one finds that the critical behavior
of the surface excess order parameter is
ηpr  ηp,crit
r
. Fitting the power law M ∝ (ηpr − η̃p,crit
r
)β to the
r r
data, using η̃p,crit as the value of ηp , where the minimum in Ms (H,H1 ) ∝ ξ β/ν−1 M̃s (H1 ξ 1 /ν ,H ξ (γ +β)/ν ), (11)
PL (M) disappears, one automatically must find the mean field where M̃s is another scaling function. From Eq. (11), one
value β = 1/2, since for finite volume PL (M) is an analytic concludes that Ms diverges as one approaches criticality, and
function of ηpr [51]. its response to H1 is also divergent, implying that the order
Unfortunately, it is not straightforward to obtain any simple parameter in the finite system must deviate very strongly from
results in the case of a finite system with surfaces. Then the its value in the bulk, since Eq. (7) also implies
generalization of Eq. (2) involves an additional “field” H1 ,
C
which describes the preferential attraction of one of the species M(H,H1 ,L) = Mbulk (H ) + Ms (H,H1 ). (12)
to the surface [52,53] L
While the divergence of Ms (0,0) in the finite system again
PL (M) = Lβ/ν P̃ (MLβ/ν ,H L(β+γ )/ν ,H1 L 1 /ν ,L/ξ ), will be rounded off, it is clear that the interplay of finite-size
(6)
L → ∞, effects on bulk and surface excess properties at criticality must
lead to a rounding and shifting of the singularities, which
where the associated critical exponent 1 ≈ 0.47 ± 0.03 [51]. strongly depends on H1 .
While in an Ising ferromagnet H1 is a local magnetic field
acting only on the surface layer of the system, and for small B. Interplay of finite-size and surface effects
magnetic particles such a field normally will not be present, we on phase coexistence
do expect some preferential attraction (which we assume to be
of short range here) of one species of a mixture to the surface, We now focus on the regime where two-phase coexistence
both in the case of polymer blends and of colloid-polymer occurs in the bulk, i.e., ηpr  ηp,crit
r
, and hence one reaches
mixtures. In the latter case, an entropic attraction of colloids easily the limit L  ξ , which means that Eqs. (7)–(10) are
to a wall occurs if the wall is purely repulsive for both colloids useful for the interpretation of the actual simulation data. As
and polymers (see, e.g., Ref. [22]); however, if one adds a soft a first step, we ignore the finite-size effects completely and
repulsive potential acting on the colloids [27], one can change take only the surface effects into account. Denoting the phase
the behavior of a planar wall from complete wetting conditions rich in colloids as “liquid” ( ) and the phase rich in polymers
to partial wetting or even drying. Thus, in an experiment it (and hence poor in colloids) as “vapor” (v), we note that at
is likely that one can control the parameter H1 by suitable phase coexistence the free-energy densities fv and f of the
coating of the wall to some extent, but it is not likely that one two coexisting phases are equal, and using Eq. (7), this implies
realizes the special case H1 = 0 over an extended regime of (we now use L = R here for spherical geometry)
the temperature-like control parameter (ηpr )−1 . 3
To clarify the physical effects of H1 , we recall that the free- fv (H,H1 ,R) = fb,v (H ) + fs,v (H,H1 )
R
energy density f of a system in the presence of a surface (or 3
surfaces, respectively) can be split into bulk (fb ) and surface = f (H,H1 ,R) = fb, (H ) + fs, (H,H1 ).
R
(fs ) contributions, in the limit L → ∞
(13)
C
f (H,H1 ,L) = fbulk (H ) + fs (H,H1 ), L → ∞, (7) Of course, all these bulk and surface free energies depend
L on ηpr , but to simplify our notation this dependence is not
where C is a geometrical factor (C = 2 for a thin film with explicitly written here. A more subtle point is the fact that
two equivalent surfaces, while C = 3 for a sphere of radius for finite R one must also expect a curvature dependence

032307-3
WINKLER, STATT, VIRNAU, AND BINDER PHYSICAL REVIEW E 87, 032307 (2013)

of the wall excess free energies fs,v , fs, . Such a curvature


dependence is familiar from the problem that spherical liquid
droplets coexist with surrounding vapor [54]. If we generalize
the standard expression for the vapor to liquid transition [33]
to the present case, we obtain

fs (H,H1 ,R) = fs (H,H1 ,∞)/[1 + 2δ/R + 2( /R)2 ], (14)

where fs (H,H1 ,∞) is nothing but the surface excess free


energy due to a planar wall, δ is the analog of a Tolman [54]
length, and is another characteristic length appearing in the
FIG. 2. Schematic colloid density profiles ρc (r) versus r of
higher correction 1/R 2 . However, to the two leading orders in
coexisting vapor-like and liquid-like states in a spherical cavity of
1/R, for which we wish to study the shift of the location of
radius R, assuming that the wall exhibits a preferential attraction to
the transition, this term does not yet enter. Since in all known
the colloids rather than the polymers. Note that the actual colloid
cases of interest the Tolman length δ is very small, we thus density profiles of colloids will exhibit layering close to the wall of
neglect it here as well. the cavity. For detailed discussion see text.
We then use the expansions (remember that H = μ − μcoex
when we deal with a colloid-polymer mixture) of the transition, it predicts for H = Ht a sharp jump from

fb,v (H ) = fb,v (0) − ρc,v H − 12 χv H 2 , 3


(15) ρc,v (Ht ,R) = ρc,v + χv Ht + ρs,v (0,H1 ) + · · · (19)
R
fb, (H ) = fb, (0) − ρc, H − 12 χ H 2 , (16) to
fs,v (H,H1 ) = fs,v (0,H ) − ρs,v (H1 )H, 3
ρc, (Ht ,R) = ρc, + χ Ht + ρs, (0,H1 ) + · · · . (20)
(17) R
fs, (H,H1 ) = fs, (H,H1 ) − ρs, (H1 )H.
3
For simplicity, only terms in first order in R
will be kept
Here, ρc,v , ρc, denote the colloid densities of the bulk here, so that
coexisting phases. In the Ising magnet, when H denotes a
magnetic field, we would instead have ±msp , the spontaneous ρc, (Ht ,R) − ρc,v (Ht ,R)

magnetization. In analogy to the Ising system, we denote the 3 ρb, (0,M1 ) − ρs,v (0,H1 )
= (ρc, − ρc,v ) 1 +
coefficients χv , χ of the higher-order terms as “susceptibil- R ρc, − ρc,v
ities.” Using now the fact that for bulk phase coexistence
(χ − χv )(fs, − fs,v )
fb,v (0) = fb, (0), Eqs. (15)–(17) and (13) yield, to order 1/R 2 + . (21)
(ρc, − ρc,v )2
3 fs, (0,H1 ) − fs,v (0,H1 ) In typical cases the correction term is negative, and so
Ht ≈
R ρc, − ρc,v the magnitude of the density jump is reduced. But it also is
 rounded over a finite interval H around Ht . To estimate this
3 ρs, (H1 ) − ρs,v (H1 )
× 1− interval, we note that in the finite volume 4π R 3 /3, the system
R ρc, − ρc,v

2
at H = H1 can jump back and forth between both coexisting
χ v − χ 3 fs, (0,H1 ) − fs,v (0,H1 ) 2 phases with densities ρc, (Ht ,R) and ρc,v (Ht ,R), and hence
+ .
2(ρc, − ρc,v ) R ρc, − ρc,v the fluctuation relation of the order parameter susceptibility
(18) yields [cf. Eq. (4)]
kB T χ (H = Ht ) = (4π R 3 /3)[(ρc, (Ht ,R) − ρc,v (Ht ,R)]2 .
Thus, any difference in the surface excess free energy due to
(22)
the wall of the spherical container shifts the chemical potential
μ at which phase coexistence occurs away from its bulk value Thus, the isotherm ρ(H ) as function of H has at H = Ht a
μcoex . This is the analog of the phenomenon of “capillary very steep slope, which scales linearly with the volume of the
condensation,” familiar from slit pores and cylindrical pores sphere. In the interval H around Ht , the variation of H takes
[23–25]. To leading order this shift is proportional to 1/R, the system from ρc,v (Ht ,R) to ρc, (Ht ,R), and hence we have
but in the regime of interest for our simulations the next
ρc, (Ht ,R) − ρc,v (Ht R) = H χ , (23)
order term (proportional to 1/R 2 ) must not be neglected. Note
that the last term on the right-hand side of Eq. (18) does not and using Eq. (22) this yields a rounding that scales inversely
occur for the Ising model, where χv = χ due to spin reversal with the volume,
symmetry. Note that this transition is not a transition between
H /kB T = {(4π R 3 /3)[ρc, (H1 ,R) − ρc,v (H1 ,R)]}−1 . (24)
homogeneous states, like in the bulk, but rather between
different density profiles (Fig. 2). The detailed features of While for systems with periodic boundary conditions,
these density profiles will be discussed at the end of Sec. IV. rounding and shifting of a first order transition both scale
Equation (18) is our first important result of this section, inversely with the system volume, in the case of spherical
and it will be put to a numerical test in Sec. IV. However, our confinement the situation is different, the shift (∝1/R) is in
treatment so far completely neglects the finite-size rounding general much larger than the rounding (∝1/R 3 ).

032307-4
PHASE TRANSITIONS AND PHASE EQUILIBRIA IN . . . PHYSICAL REVIEW E 87, 032307 (2013)

these interfacial contributions, the difference f (ρ) of the free


energy f (ρ) and its value at phase coexistence, f (ρc,v (R))
has a double-minimum shape, but the hump in between the
vapor-like phase [at colloid density ρc,v (R)] and the liquid
phase [at colloid density ρc, (R)] also scales like 1/R, and
hence f (ρc ) smoothly converges toward the double-tangent
construction that results for R → ∞, where f (ρc ) = 0 for
ρc,v  ρc  ρc, . In Fig. 3(b), f (ρc ) for finite R has been
shifted upward for the sake of clarity.
Finally, we address the interesting issue of the crossover
of the results of Eqs. (18) and (24) to the scaling description
(a)
based on Eq. (6). We note that near criticality we may write
ρc, − ρc,v ∝ ξ −β/ν and using the hyperscaling relation 3 =
2β/ν + γ /ν, one readily concludes from Eq. (24) that near
criticality a rounding results:

H ξ (γ +β)/ν ∝ (ξ/R)3 . (25)

Thus, when ξ is of order R(= L), we recover the scaling


dependence written in Eq. (6), as it should be. Recalling the
scaling of surface excess free energies [53], f˜s being a suitable
scaling function,

fs = ξ −2 f˜s [H ξ (γ +β)/ν ,H1 ξ 1 /ν ], (26)


(b)
we recognize from Eq. (18) that to leading order near criticality
FIG. 3. Schematic comparison of phase coexistence in a spherical the shift becomes, with f˜s = f˜s, − f˜s,v ,
cavity, as predicted in the grand-canonical ensemble (a) and the

canonical ensemble (b). In (a) the colloid density ρc is obtained as a ξ
Ht ξ (γ +β)/ν ∝ f˜s (H1 ξ 1 /ν ), (27)
statistical average as function of the chemical potential μ, while in R
(b) the chemical potential μ and the free-energy difference f (ρc )
are plotted as a function of ρc , treating ρc as the fixed thermodynamic where putting ξ of order R(=L), a scaling behavior compatible
variable. Only for R → ∞ is μ = μ(ρc ) simply the inverse function with Eq. (6) indeed is found. In any case, for R  ξ , the fact
of ρc = ρc (μ). Dash-dotted curves refer to finite R. Broken horizontal that the shift scales like ξ/R while the rounding scales like
straight lines in (b) indicate the Maxwell construction (left part) or (ξ/R)3 is maintained. However, a detailed test of Eqs. (25)–
double-tangent construction (right part), respectively. Free-energy (27) will be left to future work.
differences f (ρc ) refer to H = 0(R → ∞) or H = Ht (finite R),
respectively. For further explanations cf. text.
III. COMMENTS ON THE SIMULATED MODEL
Figure 3 gives a qualitative sketch of the situation, where the We here use the standard Asakura-Oosawa model [40],
reader is also reminded of the fact that the observed finite-size where colloids are represented as hard spheres of radius Rc = 1
effects depend upon the statistical ensemble that is chosen. In (being our unit of length), while polymers are soft spheres of
the grand canonical ensemble, with the chemical potential μ radius Rp = 0.8, which may overlap with each other without
as an independently fixed thermodynamic variable controlling energy cost, while the overlap of polymers with colloids is also
the system, the conjugate variable (the colloid density ρc ) is strictly forbidden. As has already been stated in Sec. II A, in
a monotonically increasing function of μ. If we choose the the simulations it is convenient to work in the grand-canonical
conjugate statistical ensemble, in the thermodynamic limit the ensemble with the chemical potentials μp and μ of polymers
existence of a Legendre transformation implies that nothing and colloids as thermodynamic parameters that are varied.
changes, ensemble equivalence implies that all what happens Note that there does not occur any energy parameter in this
is an interchange of ordinate and abscissa, the jump from ρc,v model; the effective interactions between the particles are
to ρc, at μ = μcoex becomes a horizontal plateau for μ in the purely entropic in origin. However, one can use the “polymer
region ρc,v  ρ  ρc, , at a value μcoex . Still, ρ as a function of reservoir packing fraction” ηpr = (4π Rp3 /3) exp(μp /kB T ) as
increasing μ also increases monotonically. This is not true in a parameter playing the role of inverse temperature, when we
finite systems, however: now we obtain a loop in the μ versus compare the phase behavior of this system to other models.
ρ curve (Fig. 3). In the finite system, the curve μ as function One motivation for the choice of this model is the fact that its
of ρ (ρ being given and μ a fluctuating observable) is not the bulk properties have been studied extensively in previous work,
inverse function of the ρ versus μ curve (μ being given and r
the critical point occurs at ηp,crit = 0.767 (see also Fig. 1),
ρ a fluctuating observable). As we shall see later, the detailed and the phase diagram in the bulk is also known with good
shape of the loop in the canonical ensemble reflects interfacial accuracy [44,45]. As a potential between the particles and
contributions to the free energy of the finite system. Due to confining walls, we use a hard-core repulsion as well, both for

032307-5
WINKLER, STATT, VIRNAU, AND BINDER PHYSICAL REVIEW E 87, 032307 (2013)

colloids (Ucw ) and polymers (Upw ), oriented perpendicular to the wall. This special case of “neutral
walls” is also of particular interest since then in Eq. (18) the
Ucw (z) = ∞, z < σcw , Ucw (z) = 0, z > σpw (28) coefficient of the leading 1/R term of the shift of the transition
Upw (z) = ∞, z < σpw , Upw (z) = 0, z > σpw , (29) vanishes, and a much weaker shift (∝1/R 3 ) remains.
Since the methodology yielding the data shown in Fig. 4
where z is the normal distance between the center of mass has been explained elsewhere [27], we immediately turn to the
position of a particle and the wall. We only consider the results obtained for the AO model in spherical confinement,
special case where the ratio of the repulsion ranges σpw /σcw using radii in the range 5  R  10. In Fig. 2, schematic
is the same as the ratio of particle’s radii, σpw /σcw = 0.8. colloid density profiles for spherical confinement are shown.
For this special case, the range σwc has been varied over a Phase coexistence in the cavity is shifted to a value of
wide range, and for one value of ηpr , where the system in the chemical potential μ at which the bulk system is off
the bulk is well separated, namely ηpr = 0.94, the wetting coexistence in the vapor-like phase [Ht < 0, ρc (Ht ) < ρc,v ]. A
properties of planar walls have been studied [27]. At this value liquid-like wetting layer at the wall, the average colloid-density
of ηpr , the coexistence packing fractions of the colloids have ρc,v (Ht ,R) in the cavity typically will exceed ρc,v . This
been estimated as ηc,v = 0.0077 and ηc, = 0.3117 [27], while wetting-like layer is separated by a vapor-like interface (of
the interfacial tension fint between both coexisting phases is thickness 2 ξ ) from the central vapor region. On the other
fint = 0.2029 (note that a factor 1/kB T has been absorbed hand, in the corresponding liquid-like state, the average colloid
in all interfacial and surface excess free energies throughout) density in the pore, ρc, (Ht ,R), still can be less than the bulk
[27]. Using the so-called “ensemble switch method” [27], the coexistence density ρc, : As coexistence occurs at Ht < 0, the
difference fs between the surface excess free energies of liquid density ρ (ms) (Ht ), which would correspond to a bulk
the vapor-like phase and the liquid-like phase at a planar wall, metastable (ms) state, because at Ht < 0 only the vapor-like
where the potentials Eqs. (28) and (29) act, can be found. Since state is stable, is less than ρc, .
this difference is related to the contact angle θ via Young’s
equation [36–39],
IV. SIMULATION RESULTS FOR SPHERICALLY
cos θ = fs /fint , fs = fs,v − fs, , (30) CONFINED COLLOID-POLYMER MIXTURES
the ratio at the right-hand side of this equation is plotted versus The first case that shall be considered is σwc = 0.5 (and,
the range of the wall potential σwc in Fig. 4. One can see consequently, σwp = 0.4), a case which, according to Fig. 4,
that the wetting transition occurs for σwc = 0.59, while the for ηpr = 0.94 is far away from the wetting transition, deep
drying transition occurs for σwc = 0.78. Thus, a relatively inside the region when for a planar wall at phase coexistence
small variation of σwc suffices to change the state of the complete wetting of the colloid-rich phase occurs. From the
(planar) wall from wet to dry. For σwc ≈ 0.6755, we have general theory of wetting phenomena [37–39], it then follows
the special case of neutral walls, fs = 0: Then, the contact that for all smaller values of ηpr (closer to the critical point)
angle θ = π/2 implies that vapor-liquid interfaces have to be the system is also in the complete wetting regime. Although
one might expect that for very large ηpr one encounters the
1.5 wetting transition, where the state of the planar wall changes
sim. results from complete to partial wetting, it is not clear whether or not
fit this regime has been reached in our simulations, which were
1
restricted to ηpr < 1.5 (note that the cluster algorithm [44,45]
that is needed for the implementation of the grand-canonical
0.5
ensemble simulation becomes inefficient for larger ηpr ). As
Δ fs / fint

an example for the data that can be obtained, Fig. 5 shows


0 for ηpr = 1.40 the effective free energy F (Nc ) as function of
the number Nc of colloidal particles in a sphere of radius
-0.5 R = 10. The chemical potential μ of the colloids was adjusted
according to the equal weight rule [55,56], so there occurs
-1 phase coexistence between the state containing Nc = 177
colloids and the state containing Nc = 2909 colloids. Indeed,
-1.5 in the vapor-like state (Nc = 177) all colloidal particles are
0.5 0.6 0.7 0.8 very close to the confining surface, but clearly their total
σw number is far too small to enable the formation of a precursor
of a wetting layer: the latter is only seen deep inside of the
FIG. 4. (Color online) Normalized surface free-energy difference two-phase coexistence regime, such as Nc = 877. The state at
fs /fint of the AO model at ηpr = 0.94 plotted vs. the range σwc of the top of the barrier is characterized by a kind of core-shell
the colloid-wall potential. When fs /fint < −1, complete “drying” structure, where the polymer-rich phase forms the core and
occurs (i.e., polymers form a wetting-like layer at the wall in the the colloid-rich phase forms the shell. While we expect that on
colloid-rich phase). For −1 < fs /fint < +1, the wall is partially average the structure exhibits spherical symmetry, this is not
wet. Therefore, the cases fs /fint = ±1, where the (first-order) so for individual microstates of the system, of course, since
wetting or drying transition takes place, are highlighted by horizontal entropy favors that the center of mass of the polymer-rich core
straight lines. coincide with the center of the sphere, in average only, but not

032307-6
PHASE TRANSITIONS AND PHASE EQUILIBRIA IN . . . PHYSICAL REVIEW E 87, 032307 (2013)

Nc=177 Nc=877

Nc=2130 Nc=2711
(a)

0.5
Nc=2909
βμ

-0.5

-1
c)
(c)
c)
0 1000 2000 3000
(b)
Nc

FIG. 5. (Color online) (a) Effective free energy βF (Nc ) of a colloid-polymer mixture confined into a sphere of radius R = 10 with repulsive
wall potential ranges σwc = 0.5, σwp = 0.4 at ηpr = 1.40 plotted vs. the number of colloids Nc . The dots show the choices of Nc for the snapshots
in part (b). Note that the zero βF (Nc ) = 0 was arbitrarily chosen. (b) Snapshots showing typical particle configuration in the sphere (one half of
the sphere is cut off, so one looks on a cross section plane containing the center of the sphere). Colloids are shown as yellow spheres; polymers
are black. States refer to different numbers of colloids Nc , as indicated. (c) Chemical potential difference βμ = ∂(βF (Nc ))/∂Nc plotted vs.
Nc , for the data shown in (a).

in individual configurations. In addition, the interface between are separated by a barrier. Therefore, if one only applies
colloid-rich and polymer-rich phases is rather rough, and a one-dimensional order parameter (such as the number of
there occur significant shape fluctuation of the polymer-rich particles), the transition point can be obscured because the
“droplet” in the center of the system. This is also true for simulation has to push the configuration over the barrier. This
Nc = 2711, where the polymer-rich phase takes only a very may explain the differences between the two measurements of
small fraction of the total volume, and due to entropic reasons, the chemical potential, as seen for the cylinder-slab transition
the probability that the droplet is in the center is rather small. in Fig. 6(a).
We emphasize that this description of phase coexistence For spherical confinement, no evidence for the presence of
inside of a spherical cavity with a wall that favors one of any entropic barriers was found, and hence we feel that our
the phases is very different from what one sees when one data should be fully reliable.
studies cubic boxes without walls but rather choosing periodic Figure 7 now summarizes some basic findings of the
boundary conditions (Fig. 6). While one finds spherical present study. While in fully grand-canonical representation
droplets of the minority phases close to the coexistence (ηpr versus μ) the capillary-condensation-type shift of phase
packing fractions of the colloids (and polymers), the peri- coexistence caused by the wall is rather regular and not very
odic boundary conditions also stabilize structures, such as large [Fig. 7(a)], the distortion of the phase diagram in the
cylindrical droplets or slab-configurations, that one would (ηpr , ηc ) representation is rather dramatic [Fig. 7(b)]. Note that
never find for the physically more realistic case of spherical the data included here are simply the minima recorded for the
confinement. Since for the transition from the cylindrical potential βF (Nc ) versus Nc at the various choices of R and
morphologies to the slab one has to overcome rather high ηpr , for the chemical potential μ at which phase coexistence
entropic barriers in configuration space, studies with periodic according to the “equal weight rule” occurs, as mentioned
boundary conditions are plagued with hysteresis effects, and above [Fig. 5(a)]. We caution the reader that in finite systems
the accurate estimation where these morphological transitions the finite-size effects do depend on the type of statistical
occur is difficult [31–35]. This problem is illustrated by ensemble that is chosen, as discussed in Sec. II. This means,
the deviations between the estimates for μ found directly that in Fig. 7(a) only the transition line for R → ∞ is well
from the Widom [57] particle insertion method and from defined, while for finite values of R the transition is rounded.
the derivative of F (Nc ). For the gas-droplet transition, two However, on the scales of Fig. 7(a), the rounding, in fact, is
peaks in the energy emerge at the transition point [58], which much less than the size of the symbols [recall that the shift

032307-7
WINKLER, STATT, VIRNAU, AND BINDER PHYSICAL REVIEW E 87, 032307 (2013)

(a)

(b)
(c)

FIG. 6. (Color online) (a) Effective free energy βF (Nc ) of a colloid-polymer mixture in a cubic L × L × L box with L = 26 and periodic
boundary conditions. The dots show the choices of Nc for the snapshots in panel (b). All data refer to ηpr = 1.30. (b) Snapshots showing typical
particle configurations in the cubic box (colloids are shown in yellow; polymers are dark). States refer to the choices Nc = 121 (colloidal
droplet of spherical shape), Nc = 426 (cylindrical droplet of colloids), Nc = 777 ( slab configuration), Nc = 1060 (cylindrical polymer-rich
droplet) and Nc = 1365 (spherical polymer-rich droplet). (c) Chemical potential μc as a function of Nc , taken as a derivative of the data from
part (a), red curve, and determined directly by the Widom particles insertion method. The maximum (minimum) of this curve corresponds to
the droplet (bubble) evaporation/condensation transition [31–35], while the steps signify the transitions in droplet (bubble) morphology.

μcoex (R) − μcoex (∞) ∝ R −1 , while the rounding scales prediction, it is tempting to try such an analysis also with
as R −3 ]. In the fully canonical representation of the phase the data shown in Fig. 7(b) for finite R. This is shown in
diagram, ηp versus ηc [Fig. 7(c)], size effects are somewhat Fig. 8: while with three adjustable parameters reasonably good
less pronounced than in Fig. 7(b). fits can be obtained, one must not attribute much physical
Since we have seen that in the bulk the difference ηc, − ηc,v significance to the effective critical exponents βeff (R) and
r
should follow a power law [Eq. (3)] with the critical exponent effective critical points ηp,c (R): in a finite volume, no true
β = 0.325, and the actual data for the bulk (taken from [44,45]) critical behavior anywhere can occur, since the growth of
that are included in Fig. 7(b) are nicely compatible with this critical correlations is limited by the finite sphere diameter.

FIG. 7. (Color online) Phase diagrams for spherical confinement, comparing three radii R (R = 5, 7.5, and 10, respectively) to the
corresponding bulk behavior (R = ∞, as estimated from cubic boxes with periodic boundary conditions). The packing fraction ηc is defined
as usual via ηc = Nc (Rc /R)3 with Nc being the number of colloids in the sphere with radius R. Case (a) shows the phase boundary in the plane
of variables ηpr , βμ, case (b) shows the phase diagrams in the (ηpr , ηc ) plane, and case (c) in the (ηp , ηc ) plane. The full black circle marks the
critical point. All data are for the case σwc = 0.5 σwp = 0.4. One can see from (c) that spherical confinement strongly enhances miscibility for
ηc < 0.2.

032307-8
PHASE TRANSITIONS AND PHASE EQUILIBRIA IN . . . PHYSICAL REVIEW E 87, 032307 (2013)

0.4
R=10 ηpr = 0.5
R=7.5 ηpr = 0.8
R=5 1.5
ηpr = 1.1
0.3

η(r)
Δ

0.2

0.5
0.1

0 0
0 2 4 6 8 10
0.8 1 1.2 1.4
ηp
r r
FIG. 9. (Color online) Radial density profiles of the colloids
FIG. 8. (Color online) Order parameter ηc, − ηc,v plotted vs.
ηc (r) plotted vs. r for the case R = 10, ηwc = 0.5, ηwp = 0.4 and
polymer reservoir packing fraction ηpr , for three choices of R as indi-
an average colloid packing fraction ηc = 0.214, for three choices of
cated. The full curves through the data are phenomenological fits of
ηpr , as indicated.
the form ηc, − ηc,v = const[ηpr − ηp,c
r
(R)]βeff with {ηp,c
r
(R),βeff } =
{0.876(5), 0.63(2); 0.830(1), 0.597(8); 0.816(1); 0.544(6)} for R =
5,7.5 and 10, respectively.
corresponding loops, μ versus ηc , as they are found in
the canonical ensemble, for the choice ηpr = 0.94 already
It is interesting, nevertheless, to observe that due to the wall highlighted in Fig. 4 as a situation of complete wetting. The
effects the effective critical exponent βeff (R) approaches the qualitative features of the phenomenological theory sketched
mean-field value βMF = 1/2 from above. [Recall that when in Sec. II A are indeed recovered. We indeed see that the regime
the two minima of F (Nc ) merge the exponent βeff (R) must of rounding is much smaller than the shift.
take the mean-field value βMF = 1/2, since F (Nc ) is analytic In order to test the theory more quantitatively, a variation of
in both Nc and ηpr and hence a Landau theory-type expansion the range of the potential σwc of the wall has been performed.
applies.] Also, the convergence of the corresponding colloid As an example, Fig. 11 shows the particularly interesting case
packing fraction at these pseudo- “critical points” (which for of “neutral walls,” i.e., σwc = 0.6755 (cf. Fig. 4). Indeed,
the shown values of R is close to about ηc ≈ 0.21) to the critical one can recognize from the snapshots that in bulk vapor
packing fraction in the bulk (ηccrit = 0.134) [44] is obviously (Nc = 144) and bulk liquid (N = 2036), the minority particles
extremely slow. are randomly distributed and there is no strong attraction to
As discussed in Sec. II B, the strong size effects on the the surface of the sphere. At the top of the (almost symmetric)
order parameter ηc, − ηc,v in the finite-size sphere are a free energy [Fig. 11(a)] the configuration corresponds to the
consequence of the fact that (for the wall potential studied presence of one essentially planar interface through the system.
here so far) the colloids get enriched near the wall, and At other packing fractions, one has either a colloid-rich (Nc =
this enrichment gradually increases as ηpr increases. In order 376) or polymer-rich (Nc = 1753) wall-attached droplet. So
to provide an actual counterpart to the schematic sketch in the free-energy barrier in Fig. 11(a) arises from moving one
Fig. 2, Fig. 9 shows radial density profiles of the colloids, interface that cuts the sphere at a circle of radius Rint through
as observed in our simulations. In the regime where colloids the system, which reaches its maximal area (π R 2 ) when
and polymers are well miscible also in the bulk (ηpr = 0.5), Rint = R. The loops of the isotherms [Fig. 11(c)] are already
we see that ηc is independent of r, apart from the immediate qualitatively similar to those isotherms that one finds in the
neighborhood of the wall, where two layering peaks are visible. bulk [Fig. 6(c)], and they develop in the center a linear part,
At ηpr = 0.8, the system would be demixed in the bulk but with a slope that rapidly decreases as R → ∞.
is still miscible in the cavity: due to the vicinity of bulk At this point we comment on the shape of the free-energy
criticality, the packing fraction profile is not constant even barrier F (x) as as function of the volume fraction x of
far away from the wall, however. For ηpr = 1.1, however, one of the phases near its maximum. For periodic boundary
the system is locally unmixed also in the confined system, conditions, this maximum Fmax is given by two planar
a core-shell-type structure with an interface centered near interfaces (of area L2 each in a L × L × L simulation volume),
r = 6 has been formed. This transition from almost uniform Fmax = 2L2 fint , and the free-energy barrier does not depend
density to core-shell structure is completely gradual, no on x near x = 1/2, as changing the volume fraction changes
sharp transition point for this change can be identified, of the relative sizes of the two domains, but does not change the
course. total interfacial area, as long as we have the slab configuration
In order to make contact with the phenomenological theory with two planar, noninteracting interfaces. However, for a
of Sec. II, Fig. 10 shows plots of ηc  versus μ at various sphere with neutral walls, for a volume fraction of the minority
radii, as obtained in the grand-canonical ensemble, and the phase x < 1/2, we have a reduction of the interfacial area. If

032307-9
WINKLER, STATT, VIRNAU, AND BINDER PHYSICAL REVIEW E 87, 032307 (2013)

FIG. 10. (Color online) (a) “Adsorption isotherm” ηc  plotted as a function of the colloid chemical potential β μ for the case ηpr = 0.94
and complete wetting conditions (σwc = 0.5, σwp = 0.4). Four values of R are shown, as indicated. Also the case of R → ∞ (as estimated
from the system with periodic boundary conditions) is included. Insert shows a log-log plot of the maximum slope versus R. The straight
line indicates the theoretical slope 3. (b) Loops of the chemical potential βμ plotted as function of ηc , for the same case as part (a). The
horizontal dashed lines indicate the estimates of the coexistence chemical potentials of the colloids μc (R). These estimates within numerical
error coincide with the positions of the inflection points in (a). (c) The difference β[μc (∞) − μc (R)] as a function of 1/R. The straight line
illustrates that the relation Ht ≈ 1/R [Eq. (18)] holds for R  10.

the vectors from the center of the sphere to the contact line (in occurs for x = 1/2, with Fmax = π R 2 fint ). For small α, one
a cross-section containing the sphere center) make an angle can show that
2α with respect to each other, one finds that the interface is


not flat but exhibits a radius of curvature Rc = R cos α/ sin α 3α 2
(for α → 0 the interface is flat, of course, and this situation F (α) ≈ π R fint
2
1− , (31)
4

(a)

(c) (b)

FIG. 11. (Color online) (a) Effective energy free energy βF (Nc ) vs. Nc of a colloid polymer mixture confined in a sphere of radius R = 10
with σwc = 0.6755, ηpr = 0.94. The dots show the choices of Nc for which snapshots are shown in (b). (b) Snapshots showing typical particle
configurations inside the sphere (one half of the sphere being cut off, as in Fig. 5). Colloids are shown in yellow, polymers in black. States
refer to different numbers of colloids Nc , as indicated. (c) Chemical potential βμ obtained as derivative of the free energy ∂F (Nc )/∂Nc , for
neutral walls at ηpr = 0.94. Four different choices R are included, as indicated. From F (Nc ), the chemical potentials μRc at coexistence were
also estimated but not included in the figure, for the sake of clarity.

032307-10
PHASE TRANSITIONS AND PHASE EQUILIBRIA IN . . . PHYSICAL REVIEW E 87, 032307 (2013)

6.4 0.4
Ens. switch
± γlv
0.3
6.3 sph. conf.

0.2
6.2

fsv- fsl
0.1
βμc

6.1
σwc = 0.74 0
σwc = 0.6755
6 σwc = 0.625 -0.1
σwc = 0.6
σwc = 0.55
5.9 σwc = 0.5 -0.2

0 0.05 0.1 0.15 0.2 0. 5 0. 6 0. 7 0.8


σwc
1/R
FIG. 13. (Color online) Surface free-energy difference fs,v − fs,
FIG. 12. (Color online) Chemical potential μc (R) at phase coex-
at ηpr = 0.94 plotted as function of σwc . Full curve are the data from
istence plotted vs. 1/R at ηpr = 0.94 for 6 choices of σwc , as indicated.
Fig. 4, while symbols with error bars show the data extracted from the
Straight lines are fits to the form μc (R) = μc (∞) + a/R + b/R 2 ,
coefficient of the fits shown in Fig. 12. The horizontal straight lines
with a,b being fitting parameters. This form is suggested from
at ±fint ≡ ±0.2029 indicated where complete wetting or complete
Eq. (18), since Ht ≡ μc (R) − μc (∞).
drying occurs, respectively.
while the volume fraction scales as
metastable continuations of fs beyond the wetting (or drying,
x = 1/2 − 3α/16. (32) respectively) transition could also be observed. The present
Thus, we find that F varies proportional to (x − 1/2)2 near method using the extrapolation from small spheres avoids this
its maximum, for the case of neutral walls. The barrier is problem of metastability, but the first order wetting and drying
symmetric around x = 1/2, of course. Since the radius of transitions in this finite geometry are somewhat rounded, of
curvature of F (x) in its maximum x = 1/2 diverges like R 2 , course.
for R → ∞ a flat barrier (of relative order F /R 3 ∝ 1/R →
0) results, as expected.
V. CONCLUSIONS
For complete wetting, on the other hand, the situation is
completely different: the system has core shell symmetry, so In this paper, systems undergoing a vapor-liquid-like
an interface occurs at a radius r < R, with a barrier F (α) = transition were studied for conditions of confinement within a
4π r 2 fint . Since r = Rx 1/3 , with x the volume fraction of the (mesoscopic) sphere of radius R. Phenomenological theoreti-
nonwetting phase, we see that (in the limit of R → ∞) we cal concepts were outlined and numerical tests by a simple
simply have F / Fmax = x 2/3 , 0 < x < 1: the maximum of model were provided, namely the Asakura-Oosawa model
the barrier moves toward x = 1 for R → ∞. For finite R, of colloid-polymer mixtures, with purely repulsive walls of
we expect a smooth but distinctly asymmetric barrier shape, variable range σwc between the location of the sphere surface
consistent with observation [Fig. 5(a)]. and the colloid particles. While for small values of this
We now present a test of the variation of μc (R) as a function parameter σwc , one finds complete wetting of the wall by the
of R and σwc (Fig. 12). Indeed, the data are compatible colloid-rich phase, for larger values partial wetting is found,
with the variation μc (R) = μc (∞) + a/R + b/R 2 , that one or even complete wetting of the wall by the colloid-rich phase
expects from Eq. (18). Recalling that for ηpr = 0.94 we know (from the point of view of the colloids, this is a “drying”
both ρc, − ρc, = 0.57442 [44] and the difference of surface behavior of the wall).
excess free energies fs, − fs,v [see Eq. (30)], a quantitative We find that one must distinguish essentially between two
test of the leading term in Eq. (18) actually becomes possible. qualitatively different scenarios. For cases of complete wetting
This is shown in Fig. 13. We emphasize that in Fig. 13 there or complete drying, the phase separation that develops in
is no tunable parameter whatsoever, so the present theory is the sphere exhibits (on average) a spherical core-shell-type
quantitatively reliable, at least to leading order in 1/R, as symmetry. For partial wetting, in particular, close to the
expected. Note that in the complete wetting regime the wall condition of neutral walls, a Janus-type morphology of phase
excess free energy of the vapor phase indeed is [36–39] separation in the sphere develops.
In the case where the core-shell-structure develops, the
fsv = fs + fint , complete wetting, (33)
gradual transition from the homogeneously mixed phase in
because then the surface is coated with a layer of the liquid-like the spherical cavity to the core-shell structure already shows
phase and, hence, a contact of the vapor-like phase with up in radial density profiles (as discussed schematically in
the wall is avoided. Since there is a high barrier against Fig. 2, ignoring possible layering effects at the walls, while
nucleation of such a liquid layer at a wall, when one applies numerically layering can be rather pronounced, Fig. 9): in
the ensemble switch method, such as in Ref. [27], the the one-phase region (ηpr distinctly less than the critical value

032307-11
WINKLER, STATT, VIRNAU, AND BINDER PHYSICAL REVIEW E 87, 032307 (2013)

r
ηp,crit ) the surface enrichment of the component preferred by When we examine (almost) neutral walls, the radial density
the wall decays on the scale of the correlation length ξ (
R) profile is not very useful: the (almost) planar interface
to a constant value prevailing in the center of the sphere, separating the colloid-rich and polymer-rich phases (in the
but there is no well-developed wetting-like layer. Only deep in state representing the maximum of the effective free energy)
the two-phase region near the wall such a layer can be found can be oriented in any direction. This maximum occurs then
(Figs. 2 and 9). For ηpr close to ηp,crit
r
, however, ξ in the bulk approximately for a 50:50 ratio of the two coexisting phases,
would be comparable to R, and then radial density profiles in as expected simply from geometry. For perfect neutrality of
the sphere do not exhibit any constant parts, rather the densities the wall, there would be no shift of order 1/R, and shift
have a nontrivial radial dependence throughout the sphere. and rounding would both be of order 1/R 3 . If one examines
We have paid attention to characterize the transition in in the regime of partial wetting the coefficient of the term
terms of the “adsorption isotherm” ηc  versus μ in the grand- of order 1/R in the shift of the chemical potential, one
canonical ensemble, or in terms of the effective free energy confirms the phenomenological theory [Eq. (18)] that relates
function F (ηc ) at phase coexistence in the canonical ensemble this coefficient to the difference in surface excess free energies
(emphasizing the fact that for finite R no equivalence between of the coexisting vapor-like and liquid-like phases, which also
ensembles can be expected). The “adsorption isotherm” always enters Young’s equation for the contact angle. Determining
is a monotonously increasing function, in the one-phase region this difference independently by an “ensemble switch method”
its maximal slope χ scales like ξ γ /ν away from bulk criticality (Fig. 4), the theory can be tested successfully with no need for
but like R γ /ν near bulk criticality. In the two-phase coexistence adjusting any parameters (Fig. 13).
region, an even steeper variation proportional to R 3 is pointed We hope that our treatment will stimulate also experimental
out, but this steep variation occurs only over a very small work to examine some of our predictions. As mentioned
region of the chemical potential (of width H ∝ 1/R 3 ). in the introduction, various other related systems should
This behavior is the rounded counterpart of the δ-function exhibit related phenomena, such as polymer blends confined
singularity that occurs in the thermodynamic limit due to the in miniemulsions, where indeed Janus-type phase separation
jump from ρc,v to ρc, in the isotherm. However, the shift has already been detected. However, many aspects of our
of the transition relative to its location in the bulk is much description apply also to various types of phase transitions
larger (namely proportional to the inverse radius 1/R) than in small particles, e.g., liquid-to-solid transitions in clusters,
the rounding. The effective free-energy density exhibits two crystallization of polymer globules, etc. The possible non-
minima, the locations of which one can take to define an spherical self-adjusting shape of these particles then presents
effective order parameter of the system. Due to the combined an interesting complication, however.
action of the shift of the corresponding chemical potential and
the wetting layer at the sphere surface, this order parameter
ACKNOWLEDGMENTS
is much smaller than its bulk counterpart; i.e., confinement
enhances the region of the parameter space where no demixing We thank Richard Vink for providing bulk phase diagrams.
occurs. We have also shown that fitting critical power laws We are grateful to the Deutsche Forschungsgemeinschaft for
to this order parameter does not give physically meaningful supporting this research under Grants No. TR6/A5, No. SPP
results. In between these minima, the effective free energy 1296 VI 237/4-3, and No. SFB 625/A17. We also thank the
exhibits an (asymmetrically) located maximum, of order 1/R, John von Neumann Institute for Computing (NIC Jülich) for a
which can be attributed to the core-shell interface. generous grant of computing time.

[1] E. I. Wolf, Nanophysics and Nanotechnology (Wiley-VCH, [11] P. G. de Gennes, Rev. Adv. Phys. 64, 645 (1992).
Weinheim, 2004). [12] T. Nisisako, T. Torii, T. Takahashi, and Y. Takizawa, Advanced
[2] I. M. Squires and S. R. Quake, Rev. Mod. Phys. 77, 977 (2005). Materials 18, 1152 (2006).
[3] K. Landfester, Annual Rev. Mat. Res. 36, 231 (2006). [13] A. Walther, K. Motusck, and A. H. E. Müller, Angew. Chem.
[4] L. Néel, Ann. Geophys. 5, 99 (1966). Int. Ed. 47, 711 (2008).
[5] L. S. Jacobs and C. P. Bean, in Magnetism, edited by G. T. Rado [14] K. Landfester, C. Boeffel, M. Lambla, and H. W. Spiess,
and H. Suhl, Vol. 3 (Academic Press, New York, 1963). Marcomolecules 29, 5972 (1996).
[6] S. Bedanta and W. Kleemann, J. Phys. D: Appl. Phys. 42, 013001 [15] H. E. Stanley, An Introduction to Phase Transitions and
(2009). Critical Phenomena (Oxford University Press, Oxford,
[7] K. Binder, H. Rauch, and V. Wildpaner, J. Phys. Chem. Solids 1971).
31, 391 (1970). [16] J. S. Rowlinson and F. L. Swinton, Liquids and Liquid Mixtures
[8] K. Landfester, R. Montenegro, U. Scherf, R. Güntner, (Butterworths, London, 1982).
U. Asawapirom, S. Patil, T. Kietzke, and D. Neher, Adv. Mater. [17] K. Binder, Adv. Polym. Sci. 112, 181 (1994).
14, 651 (2002). [18] Polymer Thermodynamics, Liquid Polymer-Containing Mix-
[9] T. Kietzke, D. Neher, K. Landfester, R. Montenegro, R. Günther, tures, edited by S. Enders and B. A. Wolf (Springer, Berlin
and U. Scherf, Nat. Mater. 2, 408 (2003). 2011).
[10] T. Kietzke, D. Neher, M. Kumke, O. Ghazy, U. Ziener, and [19] K. Binder, M. Müller, P. Virnau, and L. G. Mac Dowell, Adv.
K. Landfester, Small 6, 1041 (2007). Polym. Sci 173, 1 (2005).

032307-12
PHASE TRANSITIONS AND PHASE EQUILIBRIA IN . . . PHYSICAL REVIEW E 87, 032307 (2013)

[20] K. Binder, B. Mognetti, W. Paul, P. Virnau, and L. Yelash, Adv. [37] S. Dietrich, in Phase Transitions and Critical Phenomena,
Poly. Sci. 238, 329 (2011). Vol. XII, edited by C. Domb and J. L. Lebowitz (Academic
[21] W. Poon, J. Phys.: Condens. Matter 14, R859 (2002). Press, New York, 1988), p. 1.
[22] K. Binder, J. Horbach, R. Vink, and A. De Virgiliis, Soft Matter [38] Liquids at Interfaces, edited by J. Chavolin, J.-F. Joanny, and
4, 1555 (2008). J. Zinn-Justin (North-Holland, Amsterdam, 1990).
[23] L. D. Gelb, K. E. Gubbins, R. Radhakrishnan, and M. Sliwinska- [39] D. Bonn and D. Ross, Rep. Progr. Phys. 64, 1085 (2001).
Bartkowiak, Rep. Progr. Phys. 63, 1573 (1999). [40] S. Asakura and F. Oosawa, J. Polym. Sci. 33, 183 (1958).
[24] M. Schoen and S. Klapp, Reviews in Computational Chemistry, [41] D. G. A. L. Aarts, M. Schmidt, and H. N. W. Lekkerkerker,
Vol. 24 (Wiley-VCH, Hoboken, 2007). Science 304, 847 (2004).
[25] I. Bravchenko and A. Oleinikova, Interfacial and Confined Water [42] Y. Hennequin, D. G. A. L. Aarts, J. O. Indekeu, H. N. W.
(Elsevier, Amsterdam, 2008). Lekkerkerker, and D. Bonn, Phys. Rev. Lett. 100, 178305
[26] A. Winkler, D. Wilms, P. Virnau, and K. Binder, J. Chem. Phys. (2008).
133, 164702 (2010). [43] B.-H. Chen, B. Payandeh, and M. Robert, Phys. Rev. E 62, 2369
[27] A. Statt, A. Winkler, P. Virnau, and K. Binder, J. Phys.: Condens. (2000).
Matter 24, 464122 (2012). [44] R. L. C. Vink and J. Horbach, J. Chem. Phys. 121, 3253 (2004).
[28] K. Binder and D. W. Herrmann, Monte Carlo Simulation in [45] R. L. C. Vink, J. Horbach, and K. Binder, Phys. Rev. E 71,
Statistical Physics: An Introduction, 5th ed. (Springer, Berlin, 011401 (2005).
2010). [46] H.-P. Deutsch and K. Binder, Journal de Physique II 3, 1049
[29] Edited by V. Privman, Finite Size Scaling and Numerical (1993).
Simulaiton of Statistical Systems (World Scientific, Singapore, [47] M. Müller and K. Binder, Macromolecules 31, 8323 (1998).
1990). [48] J. Zausch, P. Virnau, K. Binder, J. Horbach, and R. L. C. Vink,
[30] D. P. Landau and K. Binder, A Guide to Monte Carlo Simulation J. Chem. Phys. 130, 064906 (2009).
in Statistical Physics, 3rd ed. (Cambridge University Press, [49] J. C. Le Guillou and J. Zinn-Justin, Phys. Rev. B 21, 3976
Cambridge, 2009). (1980).
[31] L. G. Mac Dowell, P. Virnau, M. Müller, and K. Binder, J. Chem. [50] K. Binder, Rep. Progr. Phys. 60, 487 (1997).
Phys. 120, 5293 (2004). [51] K. K. Mon and K. Binder, J. Chem. Phys. 96, 6989 (1992).
[32] M. Schrader, P. Virnau, and K. Binder, Phys. Rev. E 79, 061104 [52] D. P. Landau and K. Binder, Phys. Rev. B 41, 4633 (1990).
(2009). [53] K. Binder, in Phase Transitions and Critical Phenomena, Vol. 8,
[33] B. J. Block, S. K. Das, M. Oettel, P. Virnau, and K. Binder, edited by C. Domb and J. L. Lebowitz (Academic Press, London,
J. Chem. Phys. 133, 154702 (2010). 1983), p. 1.
[34] A. Troester, M. Oettel, B. J. Block, P. Virnau, and K. Binder, [54] R. C. Tolman, J. Chem. Phys. 17, 333 (1949).
J. Chem. Phys. 136, 064709 (2012). [55] K. Binder and D. P. Landau, Phys. Rev. B 30, 1477 (1984).
[35] K. Binder, B. J. Block, and P. Virnau, Amer. J. Phys. 80, 1099 [56] C. Borgs and R. Kotecky, J. Stat. Phys. 61, 79 (1990).
(2012). [57] B. Widom, J. Chem. Phys. 39, 2808 (1963).
[36] Edited by C. A. Croxton, Fluid Interfacial Phenomena (Wiley, [58] M. Schrader, Diplomarbeit, Johannes Gutenberg Universiät
New York, 1988). Mainz, 2009 (unpublished).

032307-13

You might also like