1: \documentclass[aps,twocolumn,showpacs]{revtex4}
2: %\documentclass[aps,showpacs]{revtex4}
3: %\documentclass[aps,preprint,showpacs]{revtex4}
4: \usepackage{graphicx}
5: \usepackage{amsmath}
6: \usepackage{amssymb}
7:
8: \newcommand{\hsm}{{\rm HSM}}
9: \newcommand{\fom}{{\rm FOM}}
10: \newcommand{\AOV}{{\rm AOV}}
11: \newcommand{\e}{{\rm e}}
12: \newcommand{\cv}{\otimes}
13: \newcommand{\rhocluster}{\rho_{\rm pc}}
14: \newcommand{\comment}[1]{{\bf #1}}
15: \newcommand{\myext}{eps}
16: %\newcommand{\myext}{pdf}
17: \newcommand{\etai}{\eta_{i}}
18: \newcommand{\ec}{\eta_c}
19: \newcommand{\ep}{\eta_p}
20: \newcommand{\etam}{\eta_m}
21: \newcommand{\epr}{\eta_p^r}
22: \newcommand{\rv}{{\bf r}}
23: \newcommand{\ex}{{\bf e}_x}
24: \newcommand{\ey}{{\bf e}_y}
25: \newcommand{\ez}{{\bf e}_z}
26: \newcommand{\eal}{{\bf e}_{\alpha}}
27: \newcommand{\crit}{{\rm crit}}
28: \newcommand{\spin}{{\rm spin}}
29:
30: \begin{document}
31:
32: \date{\today}
33:
34: \author{Jos\'e A.\ Cuesta}
35: \author{Luis Lafuente}
36: \affiliation{
37: Grupo Interdisciplinar de Sistemas Complejos (GISC),
38: Departamento de Matem\'aticas, Universidad Carlos III de Madrid, Avenida de la
39: Universidad 30, E-28911 Legan\'es, Madrid, Spain
40: }
41:
42: \author{Matthias Schmidt}
43: \affiliation{
44: Institut f{\"u}r Theoretische Physik II,
45: Heinrich-Heine-Universit{\"a}t D{\"u}sseldorf,
46: Universit{\"a}tsstra{\ss}e 1, D-40225 D{\"u}sseldorf, Germany
47: }
48:
49: \title{Lattice density functional for colloid-polymer mixtures:\\
50: Multi-occupancy versus Highlander version}
51:
52: \begin{abstract}
53: We consider a binary mixture of colloid and polymer particles with
54: positions on a simple cubic lattice. Colloids exclude both colloids
55: and polymers from nearest neighbor sites. Polymers are treated as
56: effective particles that are mutually non-interacting, but exclude
57: colloids from neighboring sites; this is a discrete version of the
58: (continuum) Asakura-Oosawa-Vrij model. Two alternative density
59: functionals are proposed and compared in detail. The first is based
60: on multi-occupancy in the zero-dimensional limit of the bare model,
61: analogous to the corresponding continuum theory that reproduces the
62: bulk fluid free energy of free volume theory. The second is based on
63: mapping the polymers onto a multicomponent mixture of polymer
64: clusters that are shown to behave as hard cores; the corresponding
65: (Highlander) property of the extended model in strong confinement
66: permits direct treatment with lattice fundamental measure
67: theory. Both theories predict the same topology for the phase
68: diagram with a continuous fluid-fcc freezing transition at low
69: polymer fugacity and, upon crossing a tricritical point, a
70: first-order freezing transition for high polymer fugacities with
71: rapidly broadening density jump.
72: \end{abstract}
73:
74: \pacs{61.20.Gy, 82.70.Dd, 64.75.+g}
75:
76: % 61.20.Gy Theory and models of liquid structure
77: % 82.70.Dd Colloids
78: % 64.75.+g Solubility, segregation, and mixing; phase separation
79:
80: \maketitle
81:
82: \section{Introduction}
83:
84: Mixtures of colloidal particles and non-adsorbing polymers suspended
85: in a common solvent \cite{poon02,tuinier03review} have been
86: theoretically investigated on various levels of description, ranging
87: from simplistic to realistic effective interactions between the
88: constituent particles. The prototype of the former is the
89: Asakura-Oosawa-Vrij (AOV) model \cite{asakura54,vrij76} that describes
90: the colloids as hard spheres and the polymers as ideal (i.e.,
91: non-interacting) effective spheres that interact via hard core
92: repulsion with the colloids. Despite its simplicity this model
93: reproduces the essential trends in bulk phase behavior of
94: colloid-polymer mixtures, involving colloidal gas, liquid, and
95: crystalline phases
96: \cite{lekkerkerker92,meijer94,dijkstra99,bolhuis02phasediag}, and
97: proved to be useful for studying a range of interfacial properties
98: \cite{brader03swetl}, like the structure of colloidal gas-liquid
99: interfaces and wetting of substrates, issues that are experimentally
100: relevant \cite{aarts03swet,aarts04codef,wijting03}.
101:
102: Density functional theory \cite{evans79,evans92} is a primary tool to
103: treat spatially inhomogeneous systems. For the common reference model
104: of additive mixtures of hard spheres, the fundamental measures theory
105: (FMT) \cite{rosenfeld89} is many investigators' current choice. This
106: approach was extended to cover the AOV model \cite{schmidt00cip},
107: thereby triggering much further interest in the study of interfacial
108: properties of this model, see e.g.\ \cite{schmidt04aog} for recent
109: work on novel effects in sedimentation-diffusion equilibrium. In a
110: different direction, FMT was recently generalized to hard core {\em
111: lattice} models \cite{lafuente02,lafuente03diagram,lafuente04prl}
112: elucidating the very foundation of the FMT approach.
113:
114: Despite its usefulness for practical applications, the AOV functional
115: of Ref.\ \cite{schmidt00cip} possesses several deficiencies, that are
116: absent in the hard sphere FMT (see Ref.\ \cite{schmidt02cip} for a
117: detailed discussion): i) When applied to one spatial dimension a
118: spurious phase transition is predicted \cite{schmidt02cip}, that is
119: absent (as befits a model with short-ranged interactions) in the exact
120: solution \cite{brader01oned}. ii) In three dimensions the location of
121: the bulk fluid-fluid critical point lies at lower polymer fugacity as
122: compared to the value obtained with simulations. iii) The excess free
123: energy is a linear functional of the polymer density profile.
124:
125: While we will not present a full resolution of the above issues,
126: dealing with a lattice model allows us to go significant steps beyond
127: the recipe of construction of Ref.\ \cite{schmidt00cip}. Hence we
128: consider a simplified version of the AOV model by constraining the
129: position coordinates to an underlying lattice, for convenience taken
130: to be of simple cubic symmetry. We are inspired by the fact that
131: similar lattice models have been proven useful in soft matter research
132: to make qualitative predictions, like e.g.\ for demixing
133: \cite{widom67,louis92} and three-phase equilibria \cite{duijneveldt93}.
134: Also much vital attention has been recently paid to lattice models with
135: short-ranged attractive interactions to address adsorption in
136: disordered porous media
137: \cite{kierlik01,kierlik02,sarkisov02,detcheverry03,woo03}.
138:
139: We consider the case of small particles sizes, namely such that
140: particles only repel other particles from nearest neighbor sites. As
141: the polymers are ideal, each site may be occupied by more than one of
142: those particles. We compare two different DFTs for the binary AOV
143: lattice model, both based on the lattice fundamental measure theory
144: (LFMT) \cite{lafuente02,lafuente03diagram,lafuente04prl}. The first
145: approach allows for multi-occupancy of polymers in the
146: zero-dimensional limit, and hence constitutes the lattice version of
147: the (continuum) theory of Ref.\ \cite{schmidt00cip}. The second
148: approach relies on the introduction of {\em polymer clusters} as
149: quasi-species that feature hard-core interactions only. We hence
150: arrive at an extended model that possesses a ``Highlander''
151: \cite{onlyOne} property: In an appropriately small cavity, there can
152: only be one particle. Exploiting further the fact that LFMT can cope
153: directly with small non-additivities \cite{lafuente02},
154: we arrive at a theory for the AOV
155: model that is exact when applied to one dimension; hence it does
156: correctly predict the absence of phase transitions. We find that
157: although the derivations of the two DFTs, as well as their appearance
158: at first glance differ markedly, the results for bulk phase behavior
159: in three dimensions are very similar, demonstrating the internal
160: consistency of FMT.
161:
162: The paper is organized as follows. In Sec.\ \ref{SECmodel} we define
163: the lattice AOV model. Sec.\ \ref{SECtheory} is devoted to the
164: construction of both density functional theories. In Sec.\
165: \ref{SECresults} results for the bulk phase diagram are presented and
166: we conclude in Sec.\ \ref{SECconclusions}.
167:
168:
169:
170: \section{The model}
171: \label{SECmodel}
172:
173: \begin{figure}
174: \begin{center}
175: \includegraphics[width=0.90\columnwidth]{model.\myext}
176: \caption{(a) Illustration of the lattice AOV model realized in
177: $d=2$ spatial dimensions. Colloidal particles (dark rhombi)
178: exclude their site and their nearest neighbor sites to colloids
179: and polymers. Polymer particles (light rhombi) exclude colloids
180: from their site and their nearest neighbor sites, but can
181: overlap freely with other polymers. Integers indicate the number
182: of (overlapping) polymers at occupied sites. (b) Equivalent
183: non-additive multi-component hard core mixture. Clusters of
184: overlapping polymers in (a) are viewed as quasi-particles (light
185: rhombi) that exclude their site to other polymer clusters, and
186: their site and their nearest neighbor sites to
187: colloids. Integers enumerate the different species of polymer
188: clusters and correspond to the number of overlapping polymers on
189: the lattice sites in (a). Colloids (dark rhombi) behave as in
190: (a).}
191: \label{FIGmodel}
192: \end{center}
193: \end{figure}
194:
195: We consider a binary mixture of particles representing colloids
196: (species $c$) and polymers (species $p$) on the $d$-dimensional simple
197: cubic lattice $\mathbb{Z}^d$. The interaction between colloids is that
198: of site-exclusion and nearest neighbor exclusion, corresponding to
199: the pair interaction potential
200: \begin{equation}
201: V_{cc}(\rv) = \begin{cases} \infty & \mbox{if $|\rv| \leq 1$,} \\
202: 0 & {\rm otherwise,} \end{cases}
203: \end{equation}
204: where $\rv\in\mathbb{Z}^d$
205: is the center-center distance between the particles.
206: Colloids and polymers interact similarly:
207: \begin{equation}
208: V_{cp}(\rv) = \begin{cases} \infty & \mbox{if $|\rv| \leq 1$,} \\
209: 0 & {\rm otherwise,} \end{cases}
210: \end{equation}
211: while polymers are ideal (non-interacting),
212: \begin{equation}
213: V_{pp}(\rv) = 0.
214: \end{equation}
215: In essence this is a discretized AOV model with equal-sized
216: components; see Fig.~\ref{FIGmodel}a for a sketch.
217:
218: \section{Density functional theories}
219: \label{SECtheory}
220:
221: \subsection{Overview of LFMT}
222: \label{sec:DFToverview}
223: As is customary in density functional theory \cite{evans79,evans92},
224: we express the Helmholtz free energy functional as $F=F_{\rm
225: id}+F_{\rm ex}$, where the free energy of the binary ideal lattice gas
226: is
227: \begin{equation}
228: F_{\rm id}[\rho_c,\rho_p] = \sum_{i=c,p}\sum_{\rv\in\mathcal{L}}
229: \rho_i(\rv)[\ln(\rho_i(\rv))-1],
230: \end{equation}
231: with $\rho_i(\rv)$ being the occupancy probability of site $\rv$ by
232: particles of species $i=c,p$, and $\mathcal{L}$ denoting the lattice,
233: here $\mathcal{L}=\mathbb{Z}^d$. For the excess contribution to the
234: total free energy, $F_{\rm ex}$, we will in the following use two
235: different implementations of LFMT. This theory permits to obtain an
236: approximation to $F_{\rm ex}$ for a lattice model from the exact
237: solution obtained in finite (and small) sets of lattice sites,
238: alternatively called nodes
239: \cite{lafuente02,lafuente03diagram,lafuente04prl}. The essential step
240: determining the accuracy of the theory is the choice of {\em maximal
241: cavities} \cite{lafuente04prl}. Those are maximal in the sense that
242: any further cavity taken into consideration can be obtained as an
243: intersection of maximal cavities. The common choice (inherited from
244: the continuum version of the theory \cite{RSLTshort,Tarazona97,
245: Tarazona00}) for these maximal cavities is to set them equal to
246: zero-dimensional (0d) cavities, as detailed below. Once chosen, the
247: cavities uniquely determine the form of the functional under the
248: requirement that it yields the exact result when evaluated at 0d
249: density profiles \cite{lafuente04prl}. For a general binary mixture
250: (with species labeled by $c$ and $p$) the final result possesses the
251: form
252: \begin{equation}
253: F_{\rm ex}[\rho_c,\rho_p]=\sum_{{\cal C}\,\rm cavities}a(\mathcal{C})
254: F_{\rm ex}^{\mathcal{C}}[\rho_c,\rho_p],
255: \end{equation}
256: where the summation runs over the set of maximal cavities and their
257: nonempty intersections; the $a(\mathcal{C})$ are uniquely determined
258: integer coefficients \cite{lafuente04prl}, and $F_{\rm ex}^{\mathcal{C}}$
259: is the {\em exact} excess free energy functional of the system when
260: confined to cavity $\mathcal{C}$.
261:
262:
263: \subsection{Multi-occupancy version of DFT}
264: \label{sec:DFTA}
265:
266: \begin{figure}
267: \begin{center}
268: \includegraphics[width=0.65\columnwidth]{cavities.\myext}
269: \caption{Maximal cavities along the X axis (the ones along the
270: Y axis are just 90$^{\circ}$ rotations of these ones) for
271: the multi-occupancy (a) and the Highlander (b) DFTs. Dark
272: nodes can be occupied either by colloidal or by polymer
273: particles; light nodes can only be occupied by colloidal
274: particles.}
275: \label{FIGcavities}
276: \end{center}
277: \end{figure}
278:
279: The definition of 0d cavities for lattice models with only hard-core
280: interactions is unambiguous: Any set of nodes that can accommodate at
281: most one particle constitutes a 0d cavity. For the present model,
282: however, this is problematic: As polymers are ideal, even a single
283: lattice site can be multiply occupied by polymers, hence no 0d cavity
284: can contain polymers, and the resulting functional would account for
285: them only through the ideal part, entirely ignoring the
286: colloid-polymer interaction. One way out is to relax the definition:
287: Any set of nodes in which any two particles present {\em necessarily
288: overlap} constitutes a 0d cavity \cite{schmidt00cip}. Hence, as
289: opposed to the first definition, the presence of more than one
290: particle is not excluded a priori. (For mixtures with only hard core
291: interactions, both definitions are equivalent.)
292:
293: For the present model it is straightforward to show that according to
294: the modified (multi-occupancy) definition, any pair of adjacent nodes
295: on the simple cubic lattice constitutes a maximal 0d cavity
296: (Fig.~\ref{FIGcavities}a): Such an
297: arrangement can either hold one colloid or an arbitrary number of
298: (overlapping) polymers. Adding an additional node destroys the 0d
299: character as (at least) two nonoverlapping particles will fit.
300: Implementing the scheme of Ref.~\onlinecite{lafuente04prl} with this
301: choice of basic cavities we obtain
302: \begin{equation}
303: \beta F_{\rm ex}[\rho_c,\rho_p] =\sum_{\rv}\Phi(\rv),
304: \end{equation}
305: with $\beta=1/(k_BT)$, where $T$ is temperature and $k_B$ is the
306: Boltzmann constant, and the (scaled) free energy density is
307: \begin{equation}
308: \begin{split}
309: \Phi(\rv) =&\left[\sum_{\alpha=1}^d \phi^{\rm AOV}_0\Big(n^{(\alpha)}_c(\rv),
310: n^{(\alpha)}_p(\rv) \Big) \right] \\
311: &-(2d-1)\phi^{\rm AOV}_0\Big(\rho_c(\rv),\rho_p(\rv)\Big),
312: \end{split}
313: \label{eq:PhiAOV}
314: \end{equation}
315: depending on weighted densities
316: \begin{equation}
317: n^{(\alpha)}_i(\rv) \equiv\, \rho_i(\rv)+\rho_i(\rv+\eal),
318: \qquad i=c,p,
319: \end{equation}
320: where $\eal$ denotes the unit Cartesian vector along the
321: $\alpha=1,\ldots,d$ axis and
322: \begin{equation}
323: \phi^{\rm AOV}_0(\ec,\ep)=(1-\ec-\ep)\ln(1-\ec)+\ec
324: \label{eq:phi0AOV}
325: \end{equation}
326: is the excess free energy of a 0d cavity for the AOV model
327: \cite{schmidt00cip}; we give a derivation of $\phi^{\rm
328: AOV}_0(\eta_c,\eta_p)$ in the Appendix. Since $\phi^{\rm AOV}_0(\ec,\ep)$
329: depends linearly on $\ep$ as
330: \begin{equation}
331: \phi^{\rm AOV}_0(\ec,\ep)=\phi_0(\ec)+\ep\phi'_0(\ec),
332: \end{equation}
333: where
334: \begin{equation}
335: \phi_0(\eta)\equiv\eta+(1-\eta)\ln(1-\eta),
336: \label{eq:phi0}
337: \end{equation}
338: (\ref{eq:PhiAOV}) admits the rewriting
339: \begin{equation}
340: \begin{split}
341: \Phi =&\left[\sum_{\alpha=1}^d \phi_0(n^{(\alpha)}_c)\right]
342: -(2d-1)\phi_0(\rho_c) \\
343: &-\rho_p\ln\left(\frac{\prod_{\alpha=1}^d(1-n^{(\alpha)}_c)
344: (1-\bar n^{(\alpha)}_c)}{(1-\rho_c)^{2d-1}}\right),
345: \end{split}
346: \end{equation}
347: where, for the sake of notational simplicity, we omit here and in the
348: following the dependence of $\rho_c,\rho_p, n_c^{(\alpha)}$ on $\rv$
349: and have introduced the short-hand notation $\bar n_c^{(\alpha)}\equiv
350: n_c^{(\alpha)}(\rv-\eal)$.
351:
352: The linear dependence of the functional on $\rho_p(\rv)$ yields a
353: particularly simple form of the Euler-Lagrange equation for the polymer
354: density distribution,
355: \begin{equation}
356: \rho_p=z_p\frac{\prod_{\alpha=1}^d(1-n^{(\alpha)}_c)
357: (1-\bar n^{(\alpha)}_c)}{(1-\rho_c)^{2d-1}},
358: \label{eq:rhopAOV}
359: \end{equation}
360: being explicit in $\rho_p$ (i.e., independent on the right hand side
361: of $\rho_p$). Given that we have eliminated $\rho_p(\rv)$ as a
362: functional of $\rho_c(\rv)$, $\rho_c(\rv\pm\eal)$, $\alpha=1,
363: \dots,d$, and $z_p(\rv)$, the Legendre-transformed functional
364: $\beta\Upsilon_{\rm ex}[\rho_c,z_p]=\sum_{\rv}\Phi^{\rm eff}(\rv)$,
365: with
366: \begin{equation}
367: \Phi^{\rm eff}=\rho_p\ln(\rho_p/z_p)-\rho_p+\Phi,
368: \end{equation}
369: is an excess free energy functional for the {\em effective}
370: one-component fluid of colloidal particles interacting through polymer
371: depletion; $\rho_p(\rv)$ is obtained from (\ref{eq:rhopAOV}). The
372: strength of the depletion interaction is controlled by $z_p(\rv)$,
373: which in turn can be determined by an external potential acting on the
374: polymers and by the polymer fugacity. (A more detailed account of the
375: procedure to obtain functionals for effective fluids can be found in
376: Ref.~\onlinecite{Cuesta99}.) With the appropriate substitutions, we
377: obtain
378: \begin{equation}
379: \begin{split}
380: \Phi^{\rm eff} &=\left[\sum_{\alpha=1}^d \phi_0(n^{(\alpha)}_c)\right]
381: -(2d-1)\phi_0(\rho_c) \\
382: &-z_p\frac{\prod_{\alpha=1}^d(1-n^{(\alpha)}_c)
383: (1-\bar n^{(\alpha)}_c)}{(1-\rho_c)^{2d-1}}.
384: \end{split}
385: \label{eq:effAOV}
386: \end{equation}
387: This completes the prescription for the lattice analog of the
388: continuum AOV functional of Ref.\ \cite{schmidt00cip}.
389:
390:
391: \subsection{Highlander version of DFT}
392: \label{sec:DFTB}
393: \subsubsection{Strategy of polymer clusters as quasi-particles}
394: Here we elude the presence of several particles inside a cavity and
395: aim at sticking to the stronger (Highlander) definition: {\em There
396: can be only one} \cite{onlyOne} (here: particle in the cavity). For
397: this purpose, polymer ideality constitutes a major problem, as in the
398: AOV lattice model each site occupied by a polymer can be occupied
399: simultaneously by an arbitrary number of further polymers. To
400: circumvent this problem we will hence map the model onto an extended
401: one: We refer to a piling of $n$ polymers at the same node as a
402: ``polymer cluster'' and treat polymer clusters as quasi-particles of a
403: set of new species labeled by $n$. Thus there is a one-to-one
404: correspondence between each configuration of polymers of the original
405: model and a corresponding (unique) configuration of polymer {\em
406: clusters} of the new model. (Colloids are treated as before.) We
407: regard polymer clusters as having shapes smaller than the original
408: polymer, see Fig.\ \ref{FIGmodel}b. As a consequence, polymer clusters
409: behave as hard core particles with i) site exclusion to other polymer
410: clusters and ii) site exclusion and nearest neighbor exclusion to
411: colloids. Clearly, feature ii) is directly inherited from the
412: colloid-polymer interaction. Feature i) may be unexpected at first
413: glance. Consider that a site occupied by an $n$-polymer cluster cannot
414: be {\em simultaneously} occupied by an $m$-polymer cluster. Although
415: in the original model the site can well be occupied by $m+n$ polymers,
416: in the extended model such a configuration corresponds to one single
417: $(m+n)$-polymer cluster. Hence polymer clusters repel like hard cores
418: do.
419:
420: Despite being nonadditive, the new model belongs to a class of
421: nonadditive lattice models which are amenable to LFMT and whose
422: one-dimensional version was shown to be exact \cite{lafuente02}. The
423: 0d cavities for the new model are nontrivial and consist of two
424: adjacent nodes available for the colloids, one of which (but only one)
425: is also available to a quasi-particle (polymer cluster) of any
426: species (Fig.~\ref{FIGcavities}b).
427: Clearly, the presence of either one colloidal particle or one
428: quasi-particle excludes other particles from the cavity, and such
429: cavities are obviously maximal.
430:
431: Here we will construct the corresponding free energy functional in two
432: steps. First, for simplicity, we consider the quasi-particle model
433: with a single quasi-particle species, $n=1$. Second, we will handle
434: the full (infinite) number of quasi-particle species. Subsequently
435: mapping the result back yields a DFT for the lattice AOV model.
436:
437:
438: \subsubsection{Binary mixture with one quasi-species}
439: \label{sec:binaryMixtureWithOneQuasiSpecies}
440: For the binary mixture of colloids and $1$-polymer clusters, with
441: density fields $\rho_c(\rv)$ and $\rho_1(\rv)$, respectively,
442: according to Refs.~\onlinecite{lafuente02,lafuente04prl}, the excess
443: free energy density $\Phi(\rv)$ is given by
444: \begin{eqnarray}
445: \Phi(\rv) &=& \sum_{i=1}^d\left\{
446: \phi_0\left(n_c^{(\alpha)}(\rv)+\rho_1(\rv)\right) \right.
447: \nonumber \\
448: &&\left. +
449: \phi_0\left(n_c^{(\alpha)}(\rv-\eal)+\rho_1(\rv)\right)-
450: \phi_0\left(n_c^{(\alpha)}(\rv)\right)
451: \right\} \nonumber \\
452: &&- (2d-1)\phi_0\Big(\rho_c(\rv)+\rho_1(\rv)\Big),
453: \label{eq:binary}
454: \end{eqnarray}
455: with $\phi_0(\eta)$ defined in Eq.~(\ref{eq:phi0}). Thus, given a
456: local fugacity field $z_1(\rv)$, the Euler-Lagrange equation for the
457: 1-polymer cluster density is
458: \begin{equation}
459: \rho_1=z_1\frac{\prod_{\alpha=1}^d(1-n_c^{(\alpha)}-\rho_1)
460: (1-\bar n_c^{(\alpha)}-\rho_1)}{(1-\rho_c-\rho_1)^{2d-1}}.
461: \label{eq:ELrhop}
462: \end{equation}
463: where again the dependence on $\rv$ has been omitted. Eq.\
464: (\ref{eq:ELrhop}) is an implicit algebraic equation for $\rho_1(\rv)$,
465: so it determines $\rho_1(\rv)$ as a functional of $\rho_c(\rv)$ and
466: $z_1(\rv)$. This permits to obtain, as in the previous case, a
467: functional $\beta\Upsilon_{\rm ex}[\rho_c,z_1]$ for the effective
468: one-component fluid.
469:
470:
471: \subsubsection{Mixture with an infinite number of quasi-species}
472:
473: Incorporating the infinite number of quasi-species amounts to first
474: correctly modifying the ideal free energy functional to $\beta F_{\rm
475: id}[\rho_c,\{\rho_n\}]=\sum_{\rv}\Phi_{\rm id}(\rv)$, with
476: \begin{equation}
477: \begin{split}
478: \Phi_{\rm id}(\rv)=&\,
479: \rho_c(\rv)\ln(\rho_c(\rv))-\rho_c(\rv) \\
480: &+ \sum_{n=1}^{\infty}\Big[\rho_n(\rv)\ln\Big({\cal V}_n\rho_n(\rv)
481: \Big) -\rho_n(\rv)\Big],
482: \end{split}
483: \label{eq:phiideal}
484: \end{equation}
485: where $\rho_n(\rv)$ denotes the density profile of the quasi-species
486: of $n$-polymer clusters, and ${\cal V}_n$ a thermal ``volume''
487: accounting for the internal partition function of the polymers inside
488: the cluster. (It is not hard to guess that $\mathcal{V}_n=n!$;
489: nevertheless, this result will emerge explicitly from the subsequent
490: analysis.) Secondly, we use again LFMT for the excess part of the free
491: energy. The necessary modification over the binary case (Sec.\
492: \ref{sec:binaryMixtureWithOneQuasiSpecies}) in order to include the
493: infinite number of quasi-species is actually very simple
494: \cite{lafuente02}: we replace $\rho_1(\rv)$ in (\ref{eq:binary}) by
495: the total density of polymer clusters (pc),
496: \begin{equation}
497: \rhocluster(\rv)\equiv\sum_{n=1}^{\infty}\rho_n(\rv).
498: \label{eq:rhoinf}
499: \end{equation}
500:
501: As our aim is to describe the lattice AOV model, we have to ``project''
502: the quasi-species back to polymers. Clearly, the total cluster density
503: and the polymer density are related via
504: \begin{equation}
505: \rho_p(\rv)=\sum_{n=1}^{\infty}n\rho_n(\rv).
506: \label{eq:constraint}
507: \end{equation}
508: For convenience we will again resort to an effective one-component
509: description. Hence we transform to the semi-grand potential
510: \begin{equation}
511: \beta\Upsilon[\rho_c,z_p]=\min_{\{\rho_n\}}\sum_{\rv}\Big(\Phi_{\rm id}(\rv)+
512: \Phi(\rv)-\rho_p(\rv)\ln z_p(\rv)\Big),
513: \label{eq:upsilon}
514: \end{equation}
515: where $\Phi_{\rm id}(\rv)$ is given by (\ref{eq:phiideal}), and
516: $\Phi(\rv)$ by (\ref{eq:binary}) with $\rho_1(\rv)$ replaced by
517: $\rhocluster(\rv)$. With the same notation as in (\ref{eq:ELrhop}),
518: performing the minimization in (\ref{eq:upsilon}) yields
519: \begin{equation}
520: \rho_n=\frac{z_p^n}{\mathcal{V}_n}\frac{\prod_{\alpha=1}^d
521: (1-n_c^{(\alpha)}-\rhocluster)(1-\bar n_c^{(\alpha)}-\rhocluster)}
522: {(1-\rho_c-\rhocluster)^{2d-1}},
523: \label{eq:ELrhon}
524: \end{equation}
525: which using (\ref{eq:rhoinf}) leads to
526: \begin{equation}
527: \rhocluster=\zeta(z_p)\frac{\prod_{\alpha=1}^d(1-n_c^{(\alpha)}-\rhocluster)
528: (1-\bar n_c^{(\alpha)}-\rhocluster)}{(1-\rho_c-\rhocluster)^{2d-1}},
529: \label{eq:ELrhoinf}
530: \end{equation}
531: with $\zeta(z_p)\equiv\sum_{n=1}^{\infty}(z_p^n/\mathcal{V}_n)$.
532: Also, from Eqs.~(\ref{eq:constraint}), (\ref{eq:ELrhon}) and
533: (\ref{eq:ELrhoinf}) we get
534: \begin{equation}
535: \rho_p=z_p\frac{\zeta'(z_p)}{\zeta(z_p)}\rhocluster,
536: \label{eq:rhoprhoinf}
537: \end{equation}
538: establishing a simple proportionality between $\rho_p(\rv)$ and
539: $\rhocluster(\rv)$.
540:
541: The function $\zeta(z_p)$ can be easily obtained realizing that it is
542: {\em independent} of all densities appearing in (\ref{eq:ELrhoinf}),
543: rather it depends solely on $z_p(\rv)$. Thus if we particularize
544: (\ref{eq:ELrhoinf}), e.g.\ for $\rho_c(\rv)=0$, we obtain the simple
545: relationship
546: \begin{equation}
547: \rhocluster=\zeta(z_p)(1-\rhocluster),
548: \end{equation}
549: from which
550: \begin{equation}
551: \rhocluster=\frac{\zeta(z_p)}{1+\zeta(z_p)}.
552: \end{equation}
553: Then, according to Eq.~(\ref{eq:rhoprhoinf}),
554: \begin{equation}
555: \rho_p=z_p\frac{\zeta'(z_p)}{1+\zeta(z_p)}
556: =z_p\Big(\ln(1+\zeta)\Big)',
557: \end{equation}
558: and since in the absence of colloidal particles, the polymers form
559: an ideal gas (hence $\rho_p=z_p$), we get
560: \begin{equation}
561: \Big(\ln(1+\zeta)\Big)'=1.
562: \end{equation}
563: The solution to this equation satisfying $\zeta(0)=0$ is
564: \begin{equation}
565: \zeta(z_p)=e^{z_p}-1=\sum_{n=1}^{\infty}\frac{z^n}{n!},
566: \label{eq:zinfzp}
567: \end{equation}
568: providing the confirmation of the relationship $\mathcal{V}_n=n!$.
569:
570: \subsubsection{Functional for the lattice AOV model}
571:
572: Returning to Eq.~(\ref{eq:ELrhoinf}), we observe that this equation
573: for $\rhocluster(\rv)$ is formally equivalent to Eq.~(\ref{eq:ELrhop})
574: for $\rho_1(\rv)$ with $z_1(\rv)$ replaced by the
575: $\zeta\big(z_p(\rv)\big)$ given by (\ref{eq:zinfzp}). Recalling the
576: origin of Eq.~(\ref{eq:ELrhop}), this means that we can describe the
577: original AOV model in terms of the free energy functional $\beta
578: F[\rho_c,\rhocluster]=\sum_{\rv}\left\{ \Phi_{\rm
579: id}(\rv)+\Phi(\rv)\right\}$, with
580: \begin{eqnarray}
581: \Phi_{\rm id} &=& \rho_c(\ln\rho_c-1)+
582: \rhocluster(\ln \rhocluster-1) \nonumber \\
583: \Phi &=& \sum_{i=1}^d\left [
584: \phi_0(n_c^{(\alpha)}+\rhocluster) +
585: \phi_0(\bar n_c^{(\alpha)}+\rhocluster) \right. \nonumber \\
586: &&-\left.\phi_0(n_c^{(\alpha)})\right ]
587: - (2d-1)\phi_0(\rho_c+\rhocluster),
588: \end{eqnarray}
589: with the fugacity for the total density of polymer clusters,
590: $\rhocluster(\rv)$, given by
591: \begin{equation}
592: \zeta(\rv)=e^{z_p(\rv)}-1,
593: \label{eq:zinfzp2}
594: \end{equation}
595: and the polymer density obtained from $\rhocluster(\rv)$ as
596: \begin{equation}
597: \rho_p(\rv)=\frac{z_p(\rv)}{1-e^{-z_p(\rv)}}\,\rhocluster(\rv).
598: \label{eq:rhoprhoinf2}
599: \end{equation}
600: This completes the prescription of the Highlander functional for the
601: AOV model.
602:
603: As a final note, we can alternatively obtain the effective functional
604: $\beta\Upsilon_{\rm ex}[\rho_c,z_p]=\sum_{\rv}\Phi^{\rm eff}(\rv)$,
605: which, after a few algebraic manipulations, can be written as
606: \begin{equation}
607: \begin{split}
608: \Phi^{\rm eff}=&\,
609: \rho_c\ln\rho_c-(2d-1)(1-\rho_c)
610: \ln(1-\rho_c-\rhocluster) \\
611: &+\sum_{\alpha=1}^d\big[(1-n_c^{\alpha})\ln(1-n_c^{\alpha}-\rhocluster) \\
612: &+(1-\bar n_c^{\alpha})\ln(1-\bar n_c^{\alpha}-\rhocluster) \\
613: &-(1-n_c^{\alpha})\ln(1-n_c^{\alpha}) \big],
614: \end{split}
615: \label{eq:finalPhieff}
616: \end{equation}
617: with $\rhocluster(\rv)$ being the solution of (\ref{eq:ELrhoinf}).
618:
619: \subsection{Relationship between both DFTs}
620:
621: Assuming small polymer fugacity, $z_p(\rv)\ll 1$, for the Highlander
622: DFT, from Eqs.~(\ref{eq:zinfzp2}) and (\ref{eq:rhoprhoinf2}) it
623: follows that $\zeta(\rv)\approx z_p(\rv)$ and $\rhocluster(\rv)\approx
624: \rho_p(\rv)$. The functional is then approximately given by
625: Eq.~(\ref{eq:binary}). On the other hand, small fugacities imply small
626: densities, so we can expand (\ref{eq:binary}) to linear order in
627: $\rho_p(\rv)$. Doing so and taking into account that the sum over
628: $\rv$ allows us to shift the arguments of the functions, we realize
629: that the expansion is precisely functional (\ref{eq:PhiAOV}). So both
630: DFTs coincide at low polymer fugacities.
631:
632:
633: \section{Results}
634: \label{SECresults}
635:
636: As an application we consider the bulk phase diagram for the lattice
637: AOV model and compare the predictions from both DFTs. Guided by its
638: continuum counterpart, we expect the phase behavior of the lattice
639: model to be determined by the interplay of hard core colloid-colloid
640: repulsion and the (short-ranged) colloid-colloid attraction induced by
641: polymer depletion. Thus a clear framework to study the model is the
642: effective one-component description, cf.~Eqs.~(\ref{eq:effAOV}) and
643: (\ref{eq:finalPhieff}). As in other (one-component) models with hard
644: core repulsion and short-range attraction, both condensation
645: (corresponding to demixing from the viewpoint of the binary mixture)
646: and freezing may occur. We can study the former via a convexity
647: analysis of the effective (semi-grand) free energy of a uniform fluid,
648: and the latter by considering spatially inhomogeneous density
649: distributions characteristic for crystalline states. Here the
650: candidate is a face-centered cubic (fcc) structure that constitutes
651: the close packed state for the colloids. In the following we will
652: carry out this program for either DFT.
653:
654:
655: \subsection{Thermodynamics from the multi-occupancy DFT}
656: \label{subsec:thermomulti}
657: As an appropriate thermodynamic potential we consider a (scaled)
658: semi-grand free energy per volume, $Y\equiv
659: \rho_c[\ln(\rho_c)-1]+\beta\Upsilon_{\rm ex}/V$. For a uniform fluid at
660: given polymer fugacity $z_p$, according to (\ref{eq:effAOV}) this is
661: given by
662: \begin{equation}
663: \begin{split}
664: Y=&\,\frac{\eta_c}{2}\ln\left(\frac{\eta_c}{2}\right)+d(1-\eta_c)\ln(1-\eta_c) \\
665: &-(2d-1)\left(1-\frac{\eta_c}{2}\right)
666: \ln\left(1-\frac{\eta_c}{2}\right) \\
667: &-z_p\frac{(1-\eta_c)^{2d}}{\left[1-(\eta_c/2)\right]^{2d-1}},
668: \end{split}
669: \label{eq:YunifA}
670: \end{equation}
671: where $\eta_c=2\rho_c$ is the colloid packing fraction. Convexity
672: implies $Y''(\eta_c)>0$, so the spinodal condition, $Y''(\eta_c)=0$,
673: yields
674: \begin{equation}
675: \eta^{r, \spin}_p(\eta_c)=\frac{2\left(1-\frac{\eta_c}{2}\right)^{2d}
676: [1+(d-1)\eta_c]}{d(2d-1)\eta_c(1-\eta_c)^{2d-1}},
677: \label{eq:condensA}
678: \end{equation}
679: where we have used the polymer reservoir packing fraction, $\eta^{\rm
680: r}_p=2z_p$, as an alternative thermodynamic variable. The minimum of
681: this curve determines the critical point, given by one of the roots of
682: the cubic polynomial
683: \begin{equation}
684: (d-1)(\eta_c^\crit)^3+2(d-1)^2(\eta_c^\crit)^2+(2d+1)\eta_c^\crit-2=0,
685: \end{equation}
686: and the corresponding value $\eta_p^{r,\crit}$ obtained from
687: (\ref{eq:condensA}). For $d=3$ the critical point is at
688: $\eta_c^\crit=\sqrt{3/2}-1\approx 0.225$
689: (see Fig~\ref{FIGbulkPhaseDiagram}). The liquid-gas
690: binodal can be obtained from (\ref{eq:YunifA}) via a double tangent
691: construction, in practice carried out numerically.
692:
693: For crystals we are guided by the close-packed state of the colloids,
694: and distinguish two fcc sublattices of the underlying simple cubic
695: lattice, each formed by the nearest neighbor nodes of the other. The
696: colloid density at nodes of sublattices $a$ and $b$ are denoted by
697: $\rho_a$ and $\rho_b$, respectively. Hence the average colloid packing
698: fraction is
699: \begin{equation}
700: \eta_c=\rho_a+\rho_b.
701: \label{eq:crystalDensityConstraint}
702: \end{equation}
703: The semi-grand potential is given by
704: \begin{equation}
705: \begin{split}
706: Y=&\, \frac{\rho_a}{2}\ln\rho_a+\frac{\rho_b}{2}\ln\rho_b
707: +d(1-\eta_c)\ln(1-\eta_c) \\
708: &-\left(d-{\textstyle\frac{1}{2}}\right)(1-\rho_a)\ln(1-\rho_a) \\
709: &-\left(d-{\textstyle\frac{1}{2}}\right)(1-\rho_b)\ln(1-\rho_b) \\
710: &-\frac{z_p}{2}\left[\frac{(1-\eta_c)^{2d}}{(1-\rho_a)^{2d-1}}
711: +\frac{(1-\eta_c)^{2d}}{(1-\rho_b)^{2d-1}}\right].
712: \end{split}
713: \end{equation}
714: The equilibrium density distribution for given values of $\eta_c$ can
715: be obtained by minimizing $Y$ with respect to $\rho_a$ and $\rho_b$
716: under the constraint (\ref{eq:crystalDensityConstraint}), leading to
717: the condition
718: \begin{equation}
719: \begin{split}
720: \rho_a & (1-\rho_a)^{2d-1}\exp\left(-(2d-1)z_p\left(\frac{1-\eta_c}{1-
721: \rho_a}\right)^{2d}\right) \\
722: &=\rho_b(1-\rho_b)^{2d-1}\exp\left(-(2d-1)z_p\left(\frac{1-\eta_c}{1-
723: \rho_b}\right)^{2d}\right).
724: \end{split}
725: \end{equation}
726: This equation possesses one solution characteristic for fluid states,
727: $\rho_a=\rho_b=\eta_c/2$, which is locally stable as long as it
728: minimizes $Y(\rho_a,\rho_b)$, i.e.\ up to the value of $\eta_c$ at
729: which
730: \begin{equation}
731: \frac{\partial^2}{\partial\rho_a^2}Y(\rho_a,\eta_c-\rho_a)\bigg|_{\rho_a=
732: \eta_c/2}=0,
733: \label{eq:freezingcondition}
734: \end{equation}
735: defining the freezing spinodal, that we can explicitly obtain as
736: \begin{equation}
737: \eta^{\rm r,fr}_p(\eta_c)=\frac{2\left(1-\frac{\eta_c}{2}\right)^d
738: (1-d\eta_c)}{d(2d-1)\eta_c(1-\eta_c)^{2d}}.
739: \label{eq:freezingA}
740: \end{equation}
741:
742: If freezing is second order, Eq.\ (\ref{eq:freezingA}) directly gives
743: the transition line in the phase diagram. In the case of a first-order
744: transition again a numerical double tangent construction is required
745: to obtain the coexistence densities.
746:
747:
748: \subsection{Thermodynamics from the Highlander DFT}
749: In the fluid phase we obtain the semi-grand potential
750: \begin{equation}
751: \begin{split}
752: Y=&\,\frac{\eta_c}{2}\ln\left(\frac{\eta_c}{2}\right)-
753: (2d-1)\left(1-\frac{\eta_c}{2}\right)
754: \ln\left(1-\frac{\eta_c}{2}-\rhocluster\right) \\
755: &+d(1-\eta_c)\ln\left(\frac{(1-\eta_c-\rhocluster)^2}{1-\eta_c}\right),
756: \end{split}
757: \end{equation}
758: with $\rhocluster$ being the solution of
759: \begin{equation}
760: \left(1-\frac{\eta_c}{2}-\rhocluster\right)^{2d-1}\rhocluster=
761: \zeta(z_p)(1-\eta_c-\rhocluster)^{2d}.
762: \label{eq:t-eta0}
763: \end{equation}
764: Again, the fluid-fluid spinodal can be determined by $Y''(\eta_c)=0$;
765: necessary expressions for $\rhocluster'(\eta_c)$ and
766: $\rhocluster''(\eta_c)$ can be obtained implicitly from
767: (\ref{eq:t-eta0}). The spinodal polymer cluster density is
768: \begin{equation}
769: \rhocluster^{\spin}(\eta_c)=
770: \frac{(1-\eta_c)\big[1+(d-1)\eta_c\big]}{1+(d-1)(2d+1)\eta_c},
771: \label{eq:spin1}
772: \end{equation}
773: which inserted into (\ref{eq:t-eta0}) gives the liquid-gas spinodal
774: \begin{equation}
775: \begin{split}
776: \eta_p^{\rm r,spin}(\eta_c)=&\,2\ln\Bigg(1+
777: \frac{2(2d-1)^{2d-1}}{\big[4d(d-1)\big]^{2d}}
778: \big[1+(d-1)\eta_c\big] \\
779: &\times
780: \frac{\big[2d-1-(d-1)\eta_c\big]^{2d-1}}{\eta_c(1-\eta_c)^{2d-1}}
781: \Bigg),
782: \end{split}
783: \end{equation}
784: [notice that $\eta_p^{\rm r}=2\ln(1+\zeta)$].
785: The critical point is at the minimum value of $\eta_p^{\rm r, spin}$
786: versus $\eta_c$, which we can obtain analytically as
787: \begin{equation}
788: \eta_c^\crit=\frac{1-d(d+1)+d\sqrt{2-6d+5d^2}}{(d-1)^2(2d+1)}.
789: \label{eq:criticalPointHighlander}
790: \end{equation}
791:
792:
793: In order to treat crystalline states we use the same density
794: parameterization as in the treatment with the multi-occupancy version,
795: with $\rho_a$ and $\rho_b$ being the densities of two sublattices, and
796: $\eta_c=\rho_a+\rho_b$. The semi-grand potential is obtained as
797: \begin{equation}
798: \begin{split}
799: Y =&\,\frac{\rho_a}{2}\ln\rho_a+\frac{\rho_b}{2}\ln\rho_b \\
800: &-(d-{\textstyle \frac{1}{2}})(1-\rho_a)\ln(1-\rho_a-\rhocluster^{(a)}) \\
801: &-(d-{\textstyle \frac{1}{2}})(1-\rho_b)\ln(1-\rho_b-\rhocluster^{(b)}) \\
802: &+d(1-\eta_c)\ln\left(\frac{(1-\eta_c-\rhocluster^{(a)})
803: (1-\eta_c-\rhocluster^{(b)})}{1-\eta_c}\right),
804: \end{split}
805: \end{equation}
806: and $\rhocluster^{(\alpha)}$, $\alpha=a,b$, fulfill
807: \begin{equation}
808: \rhocluster^{(\alpha)}(1-\rho_{\alpha}-\rhocluster^{(\alpha)})^{2d-1}=
809: \zeta(z_p)(1-\eta_c-\rhocluster^{(\alpha)})^{2d}.
810: \label{eq:t-eta}
811: \end{equation}
812: Minimizing $Y(\rho_a,\eta_c-\rho_a)$ with respect to $\rho_a$ leads to
813: \begin{equation}
814: \rho_a\big(1-\rho_a-\rhocluster^{(a)})^{2d-1}=\rho_b
815: \big(1-\rho_b-\rhocluster^{(b)})^{2d-1}.
816: \label{eq:minimum}
817: \end{equation}
818: We obtain $\partial \rhocluster^{(\alpha)}/\partial\rho_{\alpha}$
819: through implicit differentiation in (\ref{eq:t-eta}).
820:
821: As in section \ref{subsec:thermomulti}, the freezing spinodal is
822: obtained through Eq.~(\ref{eq:freezingcondition}), yielding
823: \begin{equation}
824: \rhocluster^{(a)}\Big|_{\rho_a=\rho_b=\eta_c/2}
825: =\frac{(1-\eta_c)(1-d\eta_c)}{1+d(2d-3)\eta_c},
826: \end{equation}
827: which combined with (\ref{eq:t-eta}) gives the analytic result
828: \begin{equation}
829: \begin{split}
830: \eta_p^{\rm r,fr}(\eta_c) &=2\ln\Bigg[1+
831: \frac{2(2d-1)^{2d-1}}{\big[4d(d-1)\big]^{2d}} \\
832: &\times \frac{(2d-1-d\eta_c)^{2d-1}}{\eta_c(1-\eta_c)^{2d-1}}
833: \big(1-d\eta_c\big)\Bigg].
834: \end{split}
835: \label{eq:freezingspinodal}
836: \end{equation}
837:
838:
839: \subsection{Phase behavior in one dimension}
840: It is clear from a general argument applicable to one-dimensional
841: systems with short-ranged forces that the lattice AOV model does not
842: exhibit a phase transition for $d=1$. In this case the Highlander
843: functional (\ref{eq:binary}), with $\rho_p(\rv)$ replaced by
844: $\rhocluster(\rv)$, is exact \cite{lafuente02}. (Note that the mapping
845: of the polymers to hard-core polymer clusters, to which we apply the
846: theory of \cite{lafuente02}, is an exact transformation.) Hence an
847: immediate consequence is that the Highlander version correctly
848: predicts the absence of phase transitions. We can recover this result
849: explicitly: the position of the liquid-gas critical point, obtained
850: via (\ref{eq:criticalPointHighlander}), is $\eta_c^\crit=1/2$,
851: $\eta_p^{\rm r}=\infty$, consistent with the absence of the
852: transition. The freezing spinodal, (\ref{eq:freezingspinodal}),
853: reduces to $\eta_p^{\rm r} =\infty$, again reflecting the absence of a
854: phase transitions. The multi-occupancy version, however, incorrectly
855: predicts both condensation and freezing, see Eqs.~(\ref{eq:condensA})
856: and (\ref{eq:freezingA}). The one fluid phase is below (in $\eta_p^r$)
857: the liquid-gas critical point located at $\eta_c^\crit=2/3$,
858: $\eta_p^{r,\crit}=4$.
859:
860:
861: \subsection{Phase behavior in three dimensions}
862:
863: \begin{figure}
864: \begin{center}
865: \includegraphics[width=\columnwidth]{pdMultiOccupancy.ps}
866: \includegraphics[width=\columnwidth]{pdHighlander.ps}
867: \caption{Bulk phase diagram of the lattice AOV model as a function
868: of colloid packing fraction, $\eta_c$, and polymer reservoir
869: packing fraction, $\eta_p^{\rm r}$. (a) Result from the
870: multi-occupancy DFT. For small $\eta_p^{\rm r}$, below the
871: tricritical point (triangle), there is a continuous freezing
872: transition into an fcc crystal; its location for $\eta_p^{\rm
873: r}=0$ is at a smaller value of $\eta_c$ as compared to the
874: result by Gaunt \cite{gaunt67} (arrow). Above the tricritical
875: point freezing is discontinuous; coexistence is along horizontal
876: tielines (dotted lines) between the fluid and crystal branches
877: of the binodal (solid lines). Also shown is the freezing
878: spinodal (long-dashed line) where the fluid phase loses its
879: metastability upon increasing $\eta_c$. The gas-liquid binodal
880: (short-dashed line) ending in a lower critical point (filled
881: dot) is metastable with respect to freezing. (b) The same as
882: (a), but obtained from the Highlander DFT. For comparison the
883: tricritical point (open triangle) and the metastable gas-liquid
884: critical point (open dot) obtained from the multi-occupancy DFT
885: are shown.}
886: \label{FIGbulkPhaseDiagram}
887: \end{center}
888: \end{figure}
889:
890: We display the result for the full phase diagram from either DFT in
891: Fig.~\ref{FIGbulkPhaseDiagram}, plotted as a function of colloid
892: packing fraction, $\eta_c$, and polymer reservoir packing fraction,
893: $\eta_p^r$. For $\eta_p^r=0$ the system is a pure (colloid) hard core
894: lattice gas with nearest-neighbor exclusion. Both DFTs reduce to the
895: (same) LFMT that predicts a continuous freezing transition at a
896: (colloid) packing fraction of $\eta_c=1/3$. This is considerably lower
897: than the value obtained from Pad\'e approximants \cite{gaunt67},
898: $\eta_c^{\crit}\approx 0.43$.
899:
900: Increasing $\eta_p^r$ leads to a shift of the transition to smaller
901: values of $\eta_c$, which can be attributed to polymers substituting
902: colloids on crystal lattice positions and hence decreasing the colloid
903: density at the transition. At a threshold value of $\eta_p^r$ the
904: transition becomes first order and a density gap opens up. The
905: location of this tricritical point differs somewhat in both
906: treatments, being located at a larger values of $\eta_p^r$ in the
907: Highlander DFT. Upon increasing $\eta_p^r$ further, the coexistence density
908: gap increases strongly. The liquid-gas transition is found to be
909: metastable with respect to freezing. (The liquid-gas binodal obtained
910: from the Highlander version is again located at higher values of
911: $\eta_p^r$ as compared to the multi-occupancy version.)
912:
913: The occurrence of a tricritical point for the freezing transition is a
914: peculiarity of the lattice model absent in the continuum version,
915: where freezing is first order for all $\eta_p^r$
916: \cite{dijkstra99}. The phase behavior of the continuum AOV model
917: depends sensitively on the polymer-to-colloid size ratio,
918: $q=\sigma_p/\sigma_c$, where $\sigma_p$ and $\sigma_c$ is the polymer
919: and colloid diameter, respectively. The phase diagram obtained for
920: small values of $q\sim 0.1$ indeed roughly resembles the topology of
921: the phase diagram that we find for the lattice AOV model, provided
922: $\eta_p^r$ is sufficiently high. For small values of $\eta_p^r$ the
923: broad coexistence region smoothly approaches the hard sphere
924: crystal-fluid coexistence densities in the continuum model (without an
925: intervening tricritical point).
926:
927: Given the fact that we do find the correct (hence also experimentally
928: observed) fcc crystalline structure, we hence conclude that the
929: current model is very suitable for the study of inhomogeneous
930: situations at high $\eta_p^r$, where a dilute colloidal gas and a
931: dense crystal are the relevant bulk states.
932:
933:
934:
935: \section{Conclusions}
936: \label{SECconclusions}
937:
938: We have carried out a detailed comparison between two density
939: functional theories for a lattice AOV model consisting of a binary
940: mixture of colloidal particles and polymers both with position
941: coordinates on a simple cubic lattice. The (pair) interactions are
942: such that colloids exclude their site and nearest neighbors to both
943: colloids and polymers, and polymers do the same with colloids, but do
944: not interact with other polymers. Relying on the LFMT concept we have
945: obtained two density functionals for arbitrary space dimension $d$ by
946: starting with the zero-dimensional Statistical Mechanics of the model.
947: The multi-occupancy version is an analog of the DFT for the continuum
948: AOV model \cite{schmidt00cip}. The Highlander version exploits the
949: possibility of mapping the lattice model to a cluster model that
950: features hard core interaction only. Both DFTs are exact for $d=0$ by
951: construction. The Highlander version is also exact in $d=1$, where the
952: multi-occupancy version incorrectly gives a phase transition.
953:
954: The predictions for the phase diagram for $d=3$ from both approaches
955: are very similar, and we have argued that the topology resembles that
956: of real colloid-polymer mixtures for small polymer-to-colloid size
957: ratios, where the liquid-gas transition is metastable with respect to
958: freezing into an fcc colloid crystal and the coexistence density gap
959: is very wide for large values of polymer fugacity. We hence conclude
960: that both the model and DFT approaches are well suited to study
961: properties of inhomongeneous situations in colloid-polymer mixtures
962: driven by dilute colloidal fluid and dense colloidal crystal phases.
963:
964:
965: \acknowledgments
966:
967: This work is supported by projects no.~BFM2003-0180 of the Direcci\'on
968: General de Investigaci\'on (Ministerio de Ciencia y Tecnolog\'{\i}a,
969: Spain) and the SFB-TR6 ``Colloidal dispersions in external fields'' of
970: the German Science Foundation (Deutsche Forschungsgemeinschaft).
971:
972: \bigskip
973:
974: \section*{APPENDIX}
975:
976: For notational simplicity, we use ${\bf 1}\equiv\rv$ and ${\bf
977: 2}\equiv\rv+\eal$. If $z_i({\bf x})$ denotes the local fugacity at
978: node ${\bf x}(={\bf 1},{\bf 2})$ of species $i(=c,p)$, then the grand
979: partition function of a maximal 0d cavity will be
980: \begin{equation}
981: \Xi_0=e^{z_p({\bf 1})+z_p({\bf 2})}+z_c({\bf 1})+z_c({\bf 2}).
982: \label{eq:Xi0}
983: \end{equation}
984: Since $\rho_i({\bf x})=z_i({\bf x})\delta\ln\Xi_0/\delta z_i({\bf x})$,
985: from the equation above
986: \begin{eqnarray}
987: \rho_c({\bf x}) &=& \frac{z_c({\bf x})}{\Xi_0},
988: \label{eq:rhocx} \\
989: \rho_p({\bf x}) &=& \frac{z_p({\bf x})e^{z_p({\bf 1})+z_p({\bf 2})}}{\Xi_0}.
990: \label{eq:rhopx}
991: \end{eqnarray}
992: From (\ref{eq:rhocx}),
993: \begin{equation}
994: \eta_c\equiv\rho_c({\bf 1})+\rho_c({\bf 2})=\frac{z_c({\bf 1})+
995: z_c({\bf 2})}{\Xi_0},
996: \end{equation}
997: so eliminating $z_c({\bf 1})+z_c({\bf 2})$ in (\ref{eq:Xi0}) we
998: obtain
999: \begin{equation}
1000: e^{z_p({\bf 1})+z_p({\bf 2})}=\Xi_0(1-\eta_c).
1001: \label{eq:expzp}
1002: \end{equation}
1003: With this Eq.~(\ref{eq:rhopx}) becomes
1004: \begin{equation}
1005: \rho_p({\bf x})=z_p({\bf x})(1-\eta_c),
1006: \label{eq:newrhopx}
1007: \end{equation}
1008: what allows us to obtain from (\ref{eq:expzp})
1009: \begin{equation}
1010: \ln\Xi_0=\frac{\eta_p}{1-\eta_c}-\ln(1-\eta_c)
1011: \label{eq:lnXi0}
1012: \end{equation}
1013: (with the obvious notation $\eta_p\equiv\rho_p({\bf 1})+\rho_p({\bf 2})$).
1014:
1015: Now, as $\phi^{\rm AOV}_0=\beta F_{\rm ex}[\rho_c,\rho_p]$ in this cavity,
1016: \begin{equation}
1017: \phi^{\rm AOV}_0=-\ln\Xi_0+\sum_{i,{\bf x}}
1018: \rho_i({\bf x})\ln\Big(z_i({\bf x})/\rho_i({\bf x})\Big),
1019: \end{equation}
1020: and using (\ref{eq:rhocx}), (\ref{eq:newrhopx}) and (\ref{eq:lnXi0})
1021: we finally obtain (\ref{eq:phi0AOV}).
1022:
1023: \vfill
1024:
1025:
1026: \bibliographystyle{apsrev}
1027: \bibliography{FMF}
1028:
1029:
1030: \end{document}
1031: