1: \documentclass[12pt,showpacs,showkeys]{revtex4}
2:
3: % packages
4: \usepackage{amssymb}
5: \usepackage{amsmath}
6: \usepackage{graphicx}
7:
8: % new commands
9: \newcommand{\bq}{\begin{eqnarray}}
10: \newcommand{\eq}{\end{eqnarray}}
11: \newcommand{\bqn}{\begin{eqnarray*}}
12: \newcommand{\eqn}{\end{eqnarray*}}
13: \newcommand{\rr}{{\mathbf r}}
14:
15: % double space
16: %\renewcommand{\baselinestretch}{2}
17:
18: \begin{document}
19: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
20:
21: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
22:
23: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
24:
25: \title{Computer simulation study of the closure relations in
26: hard sphere fluids}
27:
28: \author{R. Fantoni\footnote{e-mail: {\rm rfantoni@ts.infn.it}}}
29: %\email{rfantoni@ts.infn.it}
30: \affiliation{Dipartimento di Fisica Teorica dell' Universit\`a
31: and Istituto Nazionale di Fisica della Materia, Strada Costiera 11,
32: 34014 Trieste, Italy}
33: \author{G. Pastore\footnote{e-mail: {\rm pastore@ts.infn.it}}}
34: %\email{pastore@ts.infn.it}
35: \affiliation{Dipartimento di Fisica Teorica dell' Universit\`a
36: and INFM DEMOCRITOS National Simulation Center, Strada Costiera 11,
37: 34014 Trieste, Italy}
38: \date{\today}
39:
40: \begin{abstract}
41: \noindent
42: We study, using Monte
43: Carlo simulations, the cavity and the bridge functions of
44: various hard sphere fluids:
45: one component system,
46: equimolar additive and non additive binary mixtures.
47: In particular, we numerically check the
48: assumption of local dependency
49: of the bridge functions from the indirect correlation functions, on
50: which most of the existing integral equation theories hinge.
51: We find that this condition can be violated
52: either in the region around the first and second neighbors shell,
53: or inside the hard core,
54: for the systems here considered.
55: The violations manifest themselves clearly in the so called
56: Duh-Haymet plots of the bridge functions versus the indirect
57: correlation functions and become amplified as the coupling
58: of the system increases.
59:
60: \end{abstract}
61:
62: \pacs{61.20.Ja,61.20.Gy}
63:
64: \keywords{liquid state theory, Monte Carlo simulation,
65: bridge functions, correlation functions}
66:
67: \maketitle
68: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
69: \newpage
70: \section{Introduction}
71:
72: A central problem in the theory of the static structure of classical
73: liquids is to find a simple and efficient way to obtain the pair
74: correlation functions from the inter-particle forces in pairwise
75: interacting fluids. Exact statistical mechanics
76: \cite{Hansen,Caillol02}
77: allows to write the
78: formal solution of such problem as the coupled set of equations:
79: \begin{equation}
80: 1 + h_{ij}(r) = \exp[-\beta \phi_{ij}(r) + h_{ij}(r) - c_{ij}(r) +
81: B_{ij}(r)] \label{gr}
82: \end{equation}
83: and
84: \begin{equation}
85: h_{ij}(r) = c_{ij}(r) + \sum_{l} \rho_{l} \int d\rr^\prime
86: c_{il}(r^\prime) h_{lj}(|\rr-\rr^\prime |) \label{OZ}
87: \end{equation}
88: where $h_{ij}(r)$ and $c_{ij}(r)$ are the (total) and indirect
89: correlation functions for atomic pairs of species $i$ and $j$,
90: $\rho_{l}$ is the number density of the $l$-th component and $\beta =
91: 1/kT$. The functions $B_{ij}(r)$, named bridge functions after their
92: diagrammatic characterization \cite{Hansen} are {\em functionals} of
93: the total correlation functions, i.e. their value at distance $r$
94: depends on the values of all the correlation functions at all
95: distances.
96:
97: The basic difficulty with equations (\ref{gr}) and (\ref{OZ}) is that
98: we do not have an explicit and computationally efficient relation
99: between
100: $B_{ij}(r)$ and the correlation functions,
101: so we have to resort to approximations.
102: The results of the last three decades of research have shown that it
103: is possible to make progress by approximating the bridge functionals
104: $B_{ij}(r)$ by {\em functions } of the indirect correlation functions
105: $\gamma_{ij}(r) = h_{ij}(r) - c_{ij}(r)$ (approximate closures).
106: Once we have an explicit form for $B_{ij}(\gamma_{ij}(r))$,
107: the resulting integral
108: equations (\ref{gr}) and (\ref{OZ}), although approximate, can
109: provide excellent results for the static structure of
110: liquids. Moreover, besides the original focus on the structural
111: properties, in recent years interest has grown toward using
112: approximate integral equations to obtain thermodynamics and the phase
113: diagrams of liquids and liquid mixtures \cite{Caccamo}.
114:
115: In particular, Kjellander and Sarman
116: \cite{Kjellander89} and Lee \cite{Lee92} have derived
117: an approximate but useful formula for the chemical potential
118: of a fluid requiring only the knowledge of the correlation
119: functions at the thermodynamic state of interest. Their formula is
120: based on two main approximations. The first is the same assumption from
121: which
122: integral equations are derived, i.e. that the bridge
123: functions $B_{ij}(r)$ are local functions of
124: the corresponding indirect correlation functions.
125: The second stronger
126: assumption is that the only dependence of the bridge
127: functions from the thermodynamic state is through the indirect
128: correlation functions. Thus, the functional dependence of
129: $B_{ij}(\gamma_{ij})$ is the same for all the states.
130:
131: In this paper we want to investigate via direct numerical
132: computer simulation the two approximations.
133:
134: Up to now, numerical studies of the bridge functions and of the
135: accuracy of the local approximation have been limited to the case of
136: one component systems \cite{Restrepo92,Duh95} or
137: electrolytic solutions \cite{Duh92}.
138: We feel that two-component
139: systems deserve more interest for many reasons: i) there are strong
140: indications that the approximate universality of the bridge functions
141: \cite{Rosenfeld79} is not valid in multicomponent systems, ii) the
142: phase diagrams of multicomponent systems are richer and more
143: interesting than those of pure fluids, and iii) it turns out that
144: modeling the bridge functions for multicomponent systems is much more
145: difficult than for pure systems.
146:
147: We have studied, through Monte Carlo simulation, the
148: bridge functions of a few systems of non-additive hard spheres (NAHS)
149: mixtures, including the limiting cases of additive (AHS) mixtures and
150: one component system.
151: In particular we are interested in a direct check of the local
152: hypothesis for the functional relations between bridge and
153: correlation functions in binary mixtures. To this aim we use the so
154: called Duh-Haymet plots
155: \cite{Duh92}. These are
156: plots of the partial bridge functions $B_{ij}$ as a function of the
157: partial indirect correlation functions $\gamma_{ij}$.
158:
159: The paper is organized as follows. In section \ref{sec:cavity} we
160: summarize the equations we used to evaluate the cavity correlation
161: functions from which the bridge functions can be easily obtained and
162: we provide the relevant technical details of the numerical
163: calculations. In section \ref{sec:res} we present and discuss our
164: numerical results.
165:
166: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
167:
168: \section{Calculation of the cavity and bridge functions}
169: \label{sec:cavity}
170: \subsection{Theory}
171:
172: The binary NAHS system is a fluid made of hard spheres
173: of two species. One specie, here named 1, with
174: diameter $R_{11}$ and number density $\rho_1$ and another
175: specie (2) with diameter $R_{22}$ and number density $\rho_2$,
176: with a pair interaction potential
177: that can be written as follows
178: \bq \label{pp}
179: \phi_{ab}(r)=\left\{\begin{array}{ll}
180: \infty & r<R_{ab}\\
181: 0 & r>R_{ab}
182: \end{array}\right.~,
183: \eq
184: where $R_{12}=(R_{11}+R_{22})/2+\alpha$, with $\alpha$ being the non
185: additivity parameter. We will also study various special cases as the
186: one component system, and the binary mixture of additive
187: hard spheres (AHS) $\alpha=0$.
188: We can rewrite Eq. (\ref{gr}) to obtain the partial bridge functions
189: \bq \label{loc:B}
190: B_{ab}(r)=\ln y_{ab}(r)-\gamma_{ab}(r)~,
191: \eq
192: where $y_{ab}(r)$ are the partial cavity functions
193: \bq
194: y_{ab}(r)= g_{ab}(r)\exp[\beta\phi_{ab}(r)]~,
195: \eq
196: here $g_{ij}(r) = 1 + h_{ij}(r)$ are the partial radial distribution
197: functions. Notice that both the cavity functions and the indirect
198: correlation functions are everywhere continuous then also the bridge
199: is so.
200:
201: In the region outside the hard cores, in a hard sphere (HS) system,
202: the cavity correlation functions coincide with the pair distribution
203: functions $g_{ij}(r)$.
204: In order to determine the relationship between the partial bridge
205: functions and the partial indirect correlation functions within the
206: hard cores, we need to calculate the partial cavity functions.
207: There are two distinct methods for calculating them \cite{Restrepo92}:
208: the one which uses Henderson's
209: equation \cite{Henderson83} and the direct simulation method of Torrie
210: and Patey \cite{Torrie77}. We decided to use the first method which
211: is accurate at small $r$.
212:
213: For a binary mixture the like cavity functions can be obtained from
214: the following canonical average
215: \bq \nonumber
216: y_{aa}(r_{1_a 2_a})&=&\frac{Vz_a}{N_a}
217: \bar{y}_{aa}(r_{1_a 2_a})\\ \label{loc:y11}
218: &=&\frac{Vz_a}{N_a}\left\langle
219: \exp\left\{-\beta\left[\sum_{i_a>2}^{N_a+1}\phi_{aa}
220: (r_{1_a i_a})+\sum_{i_b=1}^{N_b}\phi_{ab}
221: (r_{1_a i_b}) \right]\right\}\right\rangle_{N_1,N_2,V,T}~,
222: \eq
223: where $a,b=1,2$ with $b\neq a$, $r_{i_a
224: j_b}$ is the distance between particle $i$ of specie $a$ and
225: particle $j$ of specie $b$,
226: $z_a=\exp(\beta\mu_a)/\Lambda^3$ is the activity of specie
227: $a$, $\mu_a$ its chemical potential, and $\Lambda$ the de Broglie
228: thermal wavelength, $V$ is the volume, $N_a$ the
229: number of particles of specie $a$, so that the prefactor
230: $Vz_a/N_a = \exp(\beta\mu^{exc}_a)$ where
231: $\mu^{exc}_a$ is the excess chemical potential of specie
232: $a$. The notation $\langle\ldots\rangle_{N_1,N_2,V,T}$ indicates
233: the canonical average at fixed number of particles, volume and
234: temperature.
235:
236: So to calculate $\bar{y}_{aa}(r)$ we need to introduce in
237: the system of $N_a+N_b$ particles labeled
238: $1_b,\ldots,N_b,2_a,\ldots,(N+1)_a$
239: a test particle $1_a$ placed a distance $r$ from particle
240: $2_a$ and calculate, at each Monte Carlo step, the interaction of this
241: particle with all the particles of the system except particle $2_a$.
242:
243: We immediately realize that when $r=0$ we must have
244: \bq \label{loc:ycontact}
245: \bar{y}_{aa}(0)=1~,
246: \eq
247: since the configurations where particle $2_a$ overlaps with other
248: particles of the system are forbidden.
249: Moreover, by taking into account that $y_{ab}(r) = g_{ab}(r)$ for
250: $r > R_{ab}$ and from the asymptotic value of the partial pair
251: distribution functions follows that
252: \bq \label{loc:y11asymptotic}
253: \lim_{r\to\infty}\bar{y}_{aa}(r)=e^{-\beta\mu^{exc}_a}~.
254: \eq
255:
256: The unlike cavity functions can be obtained from the following
257: canonical average
258: \bq \nonumber
259: y_{12}(r_{1_1 1_2})&=&\frac{Vz_1}{N_1}
260: \bar{y}_{12}(r_{1_1 1_2})\\ \label{loc:y12}
261: &=&\frac{Vz_1}{N_1}\left\langle
262: \exp\left\{-\beta\left[\sum_{i_2>1}^{N_2}\phi_{12}
263: (r_{1_1 i_2})+\sum_{i_1>1}^{N_1+1}\phi_{11}
264: (r_{1_1 i_1}) \right]\right\}\right\rangle_{N_1,N_2,V,T}~,
265: \eq
266:
267: So to calculate $\bar{y}_{12}(r)$ we need to introduce in the system
268: of $N_1+N_2$ particles labeled $1_2,\ldots,N_2,2_1,\ldots,(N+1)_1$
269: a test particle $1_1$ placed a distance $r$ from particle
270: $1_2$ and calculate, at each Monte Carlo step, the interaction of this
271: particle with all the particles of the system except particle $1_2$.
272:
273: Now there is no simple argument to guess the contact value of
274: $\bar{y}_{12}$. All we can say is that we must have
275: $\bar{y}_{12}(0)\le 1$. At large $r$ we still have
276: \bq \label{loc:y12asymptotic}
277: \lim_{r\to\infty}\bar{y}_{12}(r)=e^{-\beta\mu^{exc}_1}~.
278: \eq
279:
280: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
281:
282: \subsection{Numerical implementation}
283:
284: Monte Carlo simulations were performed with a standard NVT
285: Metropolis algorithm \cite{AT} using $N=4000$ particles.
286: Linked lists \cite{AT} have been used to reduce the computational
287: cost. To calculate the partial pair distribution functions we
288: generally used $5.2\times 10^8$ Monte Carlo steps, where one step
289: corresponds to the attempt to move a single randomly chosen particle,
290: and incremented
291: the histograms once every $20\times 4000$ steps. To calculate the
292: partial cavity functions we used $1.6\times 10^9$ Monte Carlo steps
293: and incremented the histograms once every $2\times 4000$ steps. The
294: acceptance ratio was adjusted to values between 10\% and 40\%.
295:
296: The Monte Carlo simulation returned the $g_{ab}(r)$ over a range not
297: less than $8.125R_{11} $ for the densest system. In all the studied
298: cases, the pair distribution functions attained their asymptotic value
299: well inside the maximum distance they were evaluated. Thus, it has
300: been possible to obtain accurate Fourier transforms of the total
301: correlation functions [$\hat h_{ab}(k)$] (it was necessary to cure the
302: cusps at contact in the partial pair distribution functions by
303: adding to them $H(R_{ab}-r)g_{ab}(R_{ab})$, $H$ being the Heaviside
304: step function, before taking the Fourier transform and removing
305: its analytical Fourier transform
306: afterwards). To obtain the partial indirect correlation functions we
307: first calculated the partial direct correlation functions
308: [$\hat c_{ab}(k)$] using the Fourier transform of the Ornstein-Zernike
309: equation (\ref{OZ}) and then we got the Fourier transform of the
310: indirect correlation
311: functions $\hat \gamma_{ij}(k)=\hat h_{ij}(k)-\hat c_{ij}(k)$
312: which is the transform of a continuous function in real space and
313: then is safe to transform back numerically to obtain $\gamma_{ab}(r)$.
314:
315:
316: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
317:
318: \section{Numerical Results}
319: \label{sec:res}
320:
321: We carried on simulations on the following systems: (A), one component
322: HS; (B), equimolar binary mixture of AHS; (C), equimolar binary
323: mixture of NAHS with equal like diameters and negative non additivity;
324: (D), equimolar binary mixture of NAHS with equal like diameters and
325: positive non additivity; and (E), equimolar binary mixture of NAHS
326: with different like diameters. In all these cases we have drawn the
327: corresponding Duh-Haymet plots, i.e. we plot, for each distance, the pairs
328: $(B_{ij}(r),\gamma_{ij}(r) )$.
329:
330: When we are outside the hard core the partial bridge functions
331: (\ref{loc:B}) reduces to
332: \bq
333: B_{ab}(r)=\ln g_{ab}(r)-\gamma_{ab}(r)~.
334: \eq
335: and we can obtain the bridge functions directly from the pair
336: correlation functions solving the OZ equation
337: (\ref{OZ}) to get the partial indirect correlation functions
338: $\gamma_{ab}$.
339:
340: To realize the Duh-Haymet plots when we are within the hard core
341: regions, we first calculated the cavity functions $\bar{y}_{ab}$ as
342: explained in section \ref{sec:cavity} and then the bridge functions
343: (up to an additive constant, the excess chemical potential
344: $\beta\mu_a^{exc}~$) from their definition (\ref{loc:B}).
345: Estimating the excess chemical potential from the long range behavior
346: of the cavity functions [see equations (\ref{loc:y11asymptotic}) and
347: (\ref{loc:y12asymptotic})] we where able to find the full bridge
348: functions.
349: Since the cavity functions in proximity of $R_{ab}$ becomes very
350: small, they are subject to statistical errors. In order to obtain
351: smooth Duh-Haymet plots we needed to smooth the cavity functions
352: obtained from the simulation. We did this by constructing the cubic
353: smoothing spline which has as small a second derivative as possible.
354:
355:
356: \subsection{One component HS}
357: \label{subsec:ochs}
358:
359: We carried out two simulations at $\rho_1\simeq 0.650$
360: (with a packing fraction of $\eta=\pi\rho_1R_{11}^3/6=0.340$) and
361: $\rho_1\simeq 0.925$ ($\eta= 0.484$), the former corresponding to an
362: intermediate density case and the latter to a liquid close to the
363: freezing point. In our simulations we use $R_{11}$ as unit of length.
364:
365: Inside the hard core, the bridge and the indirect correlation
366: functions are monotonic and, for the cases here considered, there are
367: no non-localities in the Duh-Haymet plots inside the core. Thus, to
368: search for non-localities it is enough to analyze results in the
369: external region. The resulting curves in the $(B,\gamma)$ plane
370: corresponding to points outside the hard core region are shown in
371: Fig. \ref{boc_ochs}. On the left the intermediate density case and
372: on the right the high density one. We see that, as the density
373: increases, the non-locality becomes more accentuated. Of course, the
374: quality of a local approximation does depend on the choice of the
375: correlation functions used as independent variable:
376: plotting the bridge function as function of the direct correlation
377: function we observed the opposite behavior.
378:
379: In order to compare the computer simulation results with the local
380: approximate $B(\gamma)$ relations used in the integral equations,
381: we have plotted the curves corresponding to different closures:
382: the hyper-netted chain (HNC) \cite{Hansen}:
383: \bq
384: B(\gamma) = 0~,
385: \eq
386: the Percus Yevick (PY) \cite{Hansen};
387: \bq
388: B(\gamma) = log(1+\gamma) - \gamma~,
389: \eq
390: the Martynov Sarkisov (MS) \cite{Martynov83} and its generalization
391: by Ballone, Pastore, Galli, and Gazzillo (BPGG) \cite{Ballone86}:
392: \bq
393: B(\gamma) = (1+\alpha \gamma)^{1/\alpha} - \gamma -1~,
394: \eq
395: (MS corresponds to $\alpha=2$, in the BPGG generalization $\alpha$
396: could be used as state dependent parameter to enforce thermodynamic
397: consistence, here a fixed value of $15/8$ has been used as suggested
398: in \cite{Ballone86}), and the modified Verlet (MV) \cite{Verlet80}:
399: \bq
400: B(\gamma) = \frac{-\gamma^2}{2 \left[ 1 + 0.8 \gamma \right] }~.
401: \eq
402: We can see that the best closures (MS, BPGG and MV), although not passing
403: through the simulation curve, tend to follow its slope and curvature.
404: When looking at Fig. \ref{boc_ochs} one should also bear in mind that the
405: values of the bridge function outside the hard core are not the most
406: relevant for the quality of the structural and thermodynamic results of the
407: closures.
408:
409: \subsection{Equimolar binary mixture of AHS}
410: \label{subsec:ahs}
411:
412: We carried out a simulation at $\rho_1=\rho_2\simeq 0.589$
413: [$\eta=\pi(\rho_1R_{11}^3+\rho_2R_{22}^3)/6= 0.375$] and
414: $\rho_1=0.5$. We chose $R_{11}=1$, $R_{12}=0.8$, and $R_{22}=0.6$.
415:
416: The results outside the hard core region are shown in the insets of
417: the plots of Fig. \ref{bvg_ahs-ed}. There are non-localities in a
418: neighborhood of the origin which corresponds to the large $r$ region.
419: These are more evident in the high density case.
420:
421: The most interesting feature shown in the figure is the difference
422: between the curves at the two different densities. If the hypothesis
423: of closures defined by a
424: unique function $B(\gamma)$ would be exact data for different densities
425: should collapse into a unique curve in these plots.
426: The data shown in Figs. \ref{boc_ochs} and \ref{bvg_ahs-ed} indicate
427: clearly that this not strictly true. However, at low and intermediate
428: densities the quantitative effect of the changing functional form is not
429: dramatic. And even at the highest liquid densities, the success of
430: closures such as MV, MS or BPGG can be probably explained in term of a
431: higher sensitivity of the theory to localized (near the contact)
432: features of the bridge functions more than to the behavior over the
433: whole range of distances.
434:
435: Inside the hard core region the Duh-Haymet plots do not have
436: non-localities. In Fig. \ref{bic_ahs-ed} we show the results for the
437: cavity functions $\bar{y}_{ab}$ for the system at the highest
438: density. The plot for the unlike functions is more noisy
439: than the plots for the like functions because $\bar{y}_{12}$ being
440: smaller than $\bar{y}_{aa}$ for $a=1,2$ is more subject to
441: statistical error.
442:
443: In Fig. \ref{bvg_ahs-ed} we show the full Duh-Haymet plots for the
444: system at the highest density, from the simulation (dots) and from
445: integral equation theories (lines).
446: The plots show how the MV approximation is the best one for
447: this system. The unlike bridge function starts at $r=0$ close to the
448: MV approximation, stays close to this approximation as $r$
449: increases and at some point have a smooth change in behavior and
450: get closer to the PY curve.
451:
452: Fig. \ref{br_ahs-ed} shows the full bridge functions as a function of
453: $r$ for the system at the highest density. It is worth of notice the
454: almost flat region of the unlike bridge near the origin.
455:
456: \subsection{Equimolar binary mixture of NAHS: $R_{11}=R_{22}$, $\alpha
457: < 0$}
458: \label{subsec:nahs-een}
459:
460: We carried out a simulation at $\rho_1=\rho_2\simeq 0.573$
461: ($\eta=0.6$). We chose $R_{11}=R_{22}=1$ and $R_{12}=0.649$
462: ($\alpha=-0.351$). These radii values would be suitable for
463: a reference system to model correlation in molten NaCl
464: \cite{Ballone84}.
465:
466: The results outside the hard core region are shown in the insets of
467: the plots of Fig. \ref{bvg_nahs-een}. There are non-localities in the
468: neighborhood of the origin corresponding to the large $r$ region.
469:
470: In Fig. \ref{bic_nahs-een} we show the results for the cavity
471: functions $\bar{y}_{ab}$.
472:
473: In Fig. \ref{bvg_nahs-een} we show the full Duh-Haymet plots from the
474: Monte Carlo simulation (dots) and from the most common integral
475: equation theories (lines). The approximation which seems to be closer
476: to the like bridge function is MV: only at big $r$ the bridge
477: functions is well approximated by PY, MS, BPGG, and MV. The unlike
478: bridge function starts at $r=0$ close to the PY approximation but as
479: $r$ increases it has a sudden change in behavior which displaces it
480: away from all the approximations. Inside the hard core region the
481: Duh-Haymet plots for the unlike
482: functions exhibit significant non-localities in correspondence with
483: the non monotonic behavior of the unlike cavity function (black dots
484: in Fig. \ref{bic_nahs-een}).
485:
486: Fig. \ref{br_nahs-een} shows the full bridge functions as a function
487: of $r$. The unlike bridge function shows oscillations in a
488: neighborhood of the origin.
489:
490: \subsection{Equimolar binary mixture of NAHS: $R_{11}=R_{22}$, $\alpha
491: > 0$}
492: \label{subsec:nahs-eep}
493:
494: We carried out a simulation at $\rho_1=\rho_2\simeq 0.200$
495: ($\eta=0.209$). We chose $R_{11}=R_{22}=1$ and $R_{12}=1.2$
496: ($\alpha=+0.2$). Notice that this system undergoes phase separation
497: when $\rho=2\rho_1 > 0.42$.
498:
499: The results outside the hard core region are shown in the insets of
500: the plots of Fig. \ref{bvg_nahs-eep}. There are non-localities in a
501: neighborhood of the origin corresponding to large distances.
502:
503: Also for this system, inside the hard core region the Duh-Haymet plots
504: for the unlike functions have non-localities in a neighborhood of
505: $r=0$. These are smaller in extent than the ones found for system
506: C. In Fig. \ref{bic_nahs-eep} we show the results for the cavity
507: functions $\bar{y}_{ab}$.
508:
509: In Fig. \ref{bvg_nahs-eep} we show the full Duh-Haymet plots from
510: the simulation (dots) and from the most common integral equations
511: (lines). The approximations which seem to be closer to the
512: like bridge function is MV and BPGG even if there is always a gap
513: between the approximations and the simulation. The unlike bridge
514: function starts at $r=0$ far away from all the approximations but as
515: $r$ increases it has a smooth change in behavior approaching the BPGG
516: curve.
517:
518: Fig. \ref{br_nahs-eep} shows the full bridge functions as a function
519: of $r$. Again, the unlike bridge function have an almost flat behavior
520: in a neighborhood of the origin.
521:
522: \subsection{Equimolar binary mixture of NAHS: $R_{11}\ne R_{22}$}
523: \label{subsec:nahs-edn}
524:
525: We carried out a simulation at $\rho_1=\rho_2\simeq 0.589$
526: ($\eta=0.375$). We chose $R_{11}=1$ and $R_{12}=R_{22}=0.6$
527: ($\alpha=-0.2$).
528:
529: The results outside the hard core region are shown in the insets of
530: the plots of Fig. \ref{bvg_nahs-edn}. There are non-localities
531: in a neighborhood of the origin which corresponds to the big $r$
532: region.
533:
534: Inside the hard core region the Duh-Haymet plots have no
535: non-localities. In Fig. \ref{bic_nahs-edn} we show the results for the
536: cavity functions $\bar{y}_{ab}$.
537:
538: In Fig. \ref{bvg_nahs-edn} we show the full Duh-Haymet plots from
539: the simulation (dots) and from the most common integral equations
540: (lines). The approximation which is closer to the 11 bridge
541: function is the MV. The one that is closer to the 22 bridge function
542: is the BPGG. The 12 bridge function starts at $r=0$ far away from all
543: the 5 approximations and as $r$ increases has a sudden change in
544: behavior and starts following the BPGG approximation.
545:
546: Fig. \ref{br_nahs-edn} shows the full bridge functions as a function
547: of $r$. The unlike bridge function shows again a qualitatively
548: different behavior near the origin.
549:
550: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
551:
552: \section{Conclusions}
553:
554: From our analysis follows that the non-localities in the function
555: relationship between the bridge functions and the indirect correlation
556: functions may appear either outside of the hard core regions or
557: inside of it. While the non-localities outside the hard core appear
558: both in the like and in the unlike functions, the ones inside the hard
559: core appear only in the unlike functions (see
560: Fig. \ref{bvg_nahs-een} and Fig. \ref{bvg_nahs-eep}),
561: for the systems that we have studied. Their appearance can be
562: directly related to the peculiar behavior of the unlike cavity
563: correlation function inside the hard core.
564:
565: As is shown by a comparison of the plots of
566: Fig. \ref{boc_ochs} and from Fig. \ref{bvg_ahs-ed} the non-localities
567: become more accentuated as we increase the coupling (the density) of
568: the system.
569: Nonetheless Fig. \ref{bvg_nahs-eep} shows that the
570: non-localities may appear even in a weakly coupled system (in this
571: case symmetric NAHS with positive non additivity).
572: Among the systems studied the one which presents the worst
573: non-localities is the equimolar symmetric NAHS with negative non
574: additivity (see Fig. \ref{bvg_nahs-een}) .
575: For this system the Duh-Haymet plot for the unlike bridge
576: function is non-local both in the hard core region (in a neighborhood
577: of $r=0$) and outside of it (at large $r$).
578:
579: We can conclude that the two hypothesis of a local function
580: approximation for the bridge functionals of the indirect
581: correlation functions and the stronger hypothesis of unique functional
582: form independent on the state, are not strictly supported by the
583: numerical data. For the one component system, this finding is
584: consistent with the observed density dependence of the bridge function
585: reported in \cite{Malijevsky87}.
586: We observe clear violations of both the assumptions increasing with
587: the density.
588: This negative statement should be somewhat mitigated by realizing that
589: the violations of the locality, in the systems studied, are limited to
590: the small and large distances regions. The latter, corresponding to
591: the region of the fast vanishing of the bridge functions affect very
592: little the thermodynamic and structural properties of the systems. The
593: former are presumably more important for the level of thermodynamic
594: consistence of the theory but have small effect on quality of the
595: structural results. The well known success of closures like MS, BPGG
596: and MV supports such point of view.
597:
598: From comparison with the simulation data in the cases we have studied,
599: we conclude that the best
600: approximations of the true hard sphere bridge functions are provided by the
601: MV and BPGG even if, especially in the
602: unlike bridge functions, there are a wide variety of characteristic behaviors
603: which are not captured by any of the most popular integral equation
604: approximations.
605: In this respect, we feel that a final comment on the local functional
606: approximation in the case of multicomponent systems is in
607: order. Indeed, density functional theory allows to say that the bridge
608: function $B_{ij}$ should be a functional of all the pair correlation
609: functions, not only the $(i,j)$ one. Thus, we could have a function
610: approximation $B_{ij}(\gamma_{11}(r),\gamma_{12}(r),\gamma_{22}(r))$
611: which would be local in space but not with respect to the
612: components. At the best of our knowledge, up to now no attempt has
613: been done to explore this additional freedom to improve the modeling
614: of the bridge functions in multicomponent systems.
615:
616:
617: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
618:
619: %\begin{acknowledgments}
620:
621: %\end{acknowledgments}
622: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
623:
624: \bibliography{loc2}
625:
626: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
627:
628: \newpage
629: \centerline{\bf LIST OF FIGURES}
630: \begin{itemize}
631: %1
632: \item[Fig. \ref{boc_ochs}] The first two graphs are Duh-Haymet plots
633: (dots), outside the hard core region, for the one component HS
634: system (the lines show the behavior of integral equation closures).
635: On the left $\rho=0.650$ on the right $\rho=0.925$.
636:
637: %2
638: \item[Fig. \ref{bvg_ahs-ed}] Full Duh-Haymet plots obtained by the
639: inversion of the Monte Carlo simulation data (dots) compared with some
640: of the most common integral equation theories (lines) for the
641: equimolar binary mixture of AHS at two different densities (in the
642: second and third plot only results at the highest density are shown).
643: $R_{11}=1$, $R_{12}=0.8$, and $R_{22}=0.6$. The insets shows the
644: portion of the bridge function outside the hard cores.
645:
646: %3
647: \item[Fig. \ref{bic_ahs-ed}] Cavity functions inside the hard core for
648: the equimolar binary mixture of AHS (at the same conditions as in
649: Fig. \ref{bvg_ahs-ed} at the highest density). The plot shows
650: the behavior of the
651: functions defined in (\ref{loc:y11}) and (\ref{loc:y12}) (notice the
652: logarithmic scale on the ordinates), the triangles denote the 22
653: function, the open circles the 11 function, and the closed circle the
654: 12 function.
655:
656: %4
657: \item[Fig. \ref{br_ahs-ed}] Bridge functions $B_{ab}(r)$
658: for the equimolar binary mixture of AHS (at the same conditions as in
659: Fig. \ref{bvg_ahs-ed} at the highest density). The insets shows
660: magnifications of the regions just outside of the hard cores.
661:
662: %5
663: \item[Fig. \ref{bvg_nahs-een}] Full
664: Duh-Haymet plots obtained by the inversion of the Monte Carlo
665: simulation data (dots) and by some of the most common integral
666: equation theories (lines) for the equimolar binary mixture of NAHS
667: with equal like diameters and negative non additivity $\alpha=-0.351$,
668: at $\rho_1=0.589$. $R_{11}=R_{22}=1$ and $R_{12}=0.649$. The insets
669: shows the portion of the bridge function outside the hard cores.
670:
671: %6
672: \item[Fig. \ref{bic_nahs-een}] Cavity functions for the
673: equimolar binary mixture of NAHS with equal like diameters and
674: negative non additivity (at the same conditions as in
675: Fig. \ref{bvg_nahs-een}). The graph shows the behavior of the
676: functions defined in (\ref{loc:y11}) and (\ref{loc:y12}) (notice the
677: logarithmic scale on the ordinates), the open circle denotes the like
678: functions and the closed circle the unlike one.
679:
680: %7
681: \item[Fig. \ref{br_nahs-een}] Bridge functions $B_{ab}(r)$
682: for the equimolar binary mixture of NAHS with equal like diameters and
683: negative non additivity (at the same conditions as in
684: Fig. \ref{bvg_nahs-een}). The insets shows magnifications of the regions
685: just outside of the hard cores.
686:
687: %8
688: \item[Fig. \ref{bvg_nahs-eep}] Full
689: Duh-Haymet plots obtained by the inversion of the Monte Carlo
690: simulation data (dots) and by some of the most common integral
691: equation theories (lines) for the equimolar binary mixture of NAHS
692: with equal like diameters and positive non additivity $\alpha=+0.2$,
693: at $\rho_1=0.200$. $R_{11}=R_{22}=1$ and $R_{12}=1.2$. The insets
694: shows the portion of the bridge function outside the hard cores.
695:
696: %9
697: \item[Fig. \ref{bic_nahs-eep}] Cavity functions for the
698: equimolar binary mixture of NAHS with equal like diameters and
699: positive non additivity (at the same conditions as in
700: Fig. \ref{bvg_nahs-eep}). The graph shows the behavior of the
701: functions defined in (\ref{loc:y11}) and (\ref{loc:y12}) (notice the
702: logarithmic scale on the ordinates), the open circle denotes the like
703: functions and the closed circle the unlike one.
704:
705: %10
706: \item[Fig. \ref{br_nahs-eep}] Bridge functions $B_{ab}(r)$
707: for the equimolar binary mixture of NAHS with equal like diameters and
708: positive non additivity (at the same conditions as in
709: Fig. \ref{bvg_nahs-eep}). The insets shows magnifications of the
710: regions just outside of the hard cores.
711:
712: %11
713: \item[Fig. \ref{bvg_nahs-edn}] Full
714: Duh-Haymet plots obtained by the inversion of the Monte Carlo
715: simulation data (dots) and by some of the most common integral
716: equation theories (lines) for the equimolar binary mixture of NAHS
717: with different like diameters $R_{11}=1$ and $R_{12}=R_{22}=0.6$, at
718: $\rho_1=0.589$. The insets shows the portion of the bridge function
719: outside the hard cores.
720:
721: %12
722: \item[Fig. \ref{bic_nahs-edn}] Cavity functions for the
723: equimolar binary mixture of NAHS with different like diameters (at the
724: same conditions as in Fig. \ref{bvg_nahs-edn}). The graph shows
725: the behavior of the functions defined in (\ref{loc:y11}) and
726: (\ref{loc:y12}) (notice the logarithmic scale on the ordinates), the
727: triangles denote the 22 function, the open circles the 11 function,
728: and the closed circle the 12 function.
729:
730: %13
731: \item[Fig. \ref{br_nahs-edn}] Bridge functions $B_{ab}(r)$
732: for the equimolar binary mixture of NAHS with different like diameters
733: (at the same conditions as in Fig. \ref{bvg_nahs-edn}). The insets
734: shows magnifications of the regions just outside of the hard cores.
735:
736: \end{itemize}
737: \newpage
738:
739: % fig 1
740: \begin{figure}[hbtp]
741: \begin{center}
742: \includegraphics[width=8cm]{boc_ochs_0.3403.eps}
743: \includegraphics[width=8cm]{boc_ochs_0.4843.eps}
744: \end{center}
745: \caption[]{R. Fantoni and G. Pastore
746: \label{boc_ochs}
747: }
748: \end{figure}
749: %
750:
751: % fig 2
752: \begin{figure}[hbtp]
753: \begin{center}
754: \includegraphics[width=8cm]{bvg11_ahs-ed.eps}
755: \includegraphics[width=8cm]{bvg12_ahs-ed.eps}\\
756: \includegraphics[width=8cm]{bvg22_ahs-ed.eps}
757: \end{center}
758: \caption[]{R. Fantoni and G. Pastore
759: \label{bvg_ahs-ed}
760: }
761: \end{figure}
762:
763: % fig 3
764: \begin{figure}[hbtp]
765: \begin{center}
766: \includegraphics[width=8cm]{cav_ahs-ed_0.3753.eps}
767: \end{center}
768: \caption[]{R. Fantoni and G. Pastore
769: \label{bic_ahs-ed}
770: }
771: \end{figure}
772: %
773:
774: % fig 4
775: \begin{figure}[hbtp]
776: \begin{center}
777: \includegraphics[width=8cm]{br11_ahs-ed.eps}
778: \includegraphics[width=8cm]{br12_ahs-ed.eps}
779: \includegraphics[width=8cm]{br22_ahs-ed.eps}\\
780: \end{center}
781: \caption[]{R. Fantoni and G. Pastore
782: \label{br_ahs-ed}
783: }
784: \end{figure}
785: %
786:
787: % fig 5
788: \begin{figure}[hbtp]
789: \begin{center}
790: \includegraphics[width=8cm]{bvg11_nahs-een.eps}
791: \includegraphics[width=8cm]{bvg12_nahs-een.eps}
792: \end{center}
793: \caption[]{R. Fantoni and G. Pastore
794: \label{bvg_nahs-een}
795: }
796: \end{figure}
797: %
798:
799: % fig 6
800: \begin{figure}[hbtp]
801: \begin{center}
802: \includegraphics[width=8cm]{cav_nahs-een_0.6.eps}
803: \end{center}
804: \caption[]{R. Fantoni and G. Pastore
805: \label{bic_nahs-een}
806: }
807: \end{figure}
808: %
809:
810: % fig 7
811: \begin{figure}[hbtp]
812: \begin{center}
813: \includegraphics[width=8cm]{br11_nahs-een.eps}
814: \includegraphics[width=8cm]{br12_nahs-een.eps}
815: \end{center}
816: \caption[]{R. Fantoni and G. Pastore
817: \label{br_nahs-een}
818: }
819: \end{figure}
820: %
821:
822: % fig 8
823: \begin{figure}[hbtp]
824: \begin{center}
825: \includegraphics[width=8cm]{bvg11_nahs-eep.eps}
826: \includegraphics[width=8cm]{bvg12_nahs-eep.eps}
827: \end{center}
828: \caption[]{R. Fantoni and G. Pastore
829: \label{bvg_nahs-eep}
830: }
831: \end{figure}
832: %
833:
834: % fig 9
835: \begin{figure}[hbtp]
836: \begin{center}
837: \includegraphics[width=8cm]{cav_nahs-eep_0.2094.eps}
838: \end{center}
839: \caption[]{R. Fantoni and G. Pastore
840: \label{bic_nahs-eep}
841: }
842: \end{figure}
843: %
844:
845: % fig 10
846: \begin{figure}[hbtp]
847: \begin{center}
848: \includegraphics[width=8cm]{br11_nahs-eep.eps}
849: \includegraphics[width=8cm]{br12_nahs-eep.eps}
850: \end{center}
851: \caption[]{R. Fantoni and G. Pastore
852: \label{br_nahs-eep}
853: }
854: \end{figure}
855: %
856:
857: % fig 11
858: \begin{figure}[hbtp]
859: \begin{center}
860: \includegraphics[width=8cm]{bvg11_nahs-edn.eps}
861: \includegraphics[width=8cm]{bvg12_nahs-edn.eps}\\
862: \includegraphics[width=8cm]{bvg22_nahs-edn.eps}
863: \end{center}
864: \caption[]{R. Fantoni and G. Pastore
865: \label{bvg_nahs-edn}
866: }
867: \end{figure}
868: %
869:
870: % fig 12
871: \begin{figure}[hbtp]
872: \begin{center}
873: \includegraphics[width=8cm]{cav_nahs-edn_0.3753.eps}
874: \end{center}
875: \caption[]{R. Fantoni and G. Pastore
876: \label{bic_nahs-edn}
877: }
878: \end{figure}
879: %
880:
881: % fig 13
882: \begin{figure}[hbtp]
883: \begin{center}
884: \includegraphics[width=8cm]{br11_nahs-edn.eps}
885: \includegraphics[width=8cm]{br12_nahs-edn.eps}\\
886: \includegraphics[width=8cm]{br22_nahs-edn.eps}
887: \end{center}
888: \caption[]{R. Fantoni and G. Pastore
889: \label{br_nahs-edn}
890: }
891: \end{figure}
892: %
893:
894: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
895: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
896: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
897:
898: \end{document}
899: