1: \documentclass[twocolumn,showpacs,superscriptaddress,aps,prl]{revtex4}
2: %\documentclass[preprint,showpacs,superscriptaddress,aps,prl,endfloats*]{revtex4}
3:
4: \topmargin -0.75in
5: \renewcommand{\paragraph}[1]{\textbf{#1.}---}
6:
7: \usepackage{graphicx} % Include figure files
8: \usepackage{bm} % bold math
9:
10: \newcommand{\figurewidth}{86mm}
11: %\newcommand{\figurewidth}{130mm}
12:
13: \renewcommand{\UrlFont}{\rm} % MEF prefers this
14: \bibpunct[]{[}{]}{,}{n}{}{,} % make multiple entries work properly
15:
16: \begin{document}
17:
18: \title{Universality Class of Criticality in the Restricted Primitive Model
19: Electrolyte}
20:
21: \author{Erik Luijten}
22: \altaffiliation[Now at ]{Dept.\ of Materials Science and Engineering,
23: University of Illinois, Urbana, IL 61801}
24: \email[]{luijten@uiuc.edu}
25:
26: \author{Michael E. Fisher}
27: \thanks{Corresponding author: fax (301) 314-9404.}
28: \affiliation{Institute for Physical Science and Technology, University of
29: Maryland, College Park, Maryland 20742}
30:
31: \author{Athanassios Z. Panagiotopoulos}
32: \affiliation{Department of Chemical Engineering, Princeton University,
33: Princeton, New Jersey 08540}
34:
35: \date{\today}
36:
37: \begin{abstract}
38: The 1:1 equisized hard-sphere electrolyte or \emph{restricted primitive
39: model} has been simulated via grand-canonical fine-discretization Monte
40: Carlo. Newly devised unbiased finite-size extrapolation methods using
41: temperature-density, ($T, \rho$), loci of inflections, $Q \equiv \langle
42: m^2\rangle^2/\langle m^4\rangle$ maxima, canonical and $C_V$ criticality,
43: yield estimates of ($T_c$, $\rho_c$) to $\pm$(0.04, 3)\%. Extrapolated
44: exponents and $Q$-ratio are ($\gamma$, $\nu$, $Q_c$) = [1.24(3), 0.63(3);
45: 0.624(2)] which support Ising ($n \! = \! 1$) behavior with ($1.23_9$,
46: $0.630_3$; $0.623_6$), but exclude classical, XY ($n \! = \! 2$), SAW ($n \!
47: = \! 0$), and $n \!= \! 1$ criticality with potentials
48: $\varphi(r)>\Phi/r^{4.9}$ when $r \to \infty$.
49: \end{abstract}
50:
51: \pacs{02.70.Rr, 05.70.Jk, 64.60.Fr, 64.70.Fx}
52:
53: \maketitle
54:
55: Since the experiments of Singh and Pitzer in 1988~\cite{fisher-review,weing},
56: an outstanding experimental and theoretical question has been: What is the
57: universality class of Coulombic criticality? Early experimental data for
58: electrolytes exhibiting phase separation driven by long-range ionic forces
59: suggested classical or van der Waals (vdW) critical behavior, with exponents
60: $\beta=\frac{1}{2}$, $\gamma=1$, $\nu=\frac{1}{2}$,
61: etc.~\cite{fisher-review,weing}: But the general theoretical consensus has been
62: that asymptotic Ising-type criticality, with $\beta \simeq 0.326$, $\gamma
63: \simeq 1.239$, $\nu \simeq 0.630$, etc., should be
64: expected~\cite{fisher-review,stell94,fisher-lee96}. Naively, one may argue that
65: the exponential Debye screening of the direct ionic forces results in
66: \emph{effective short-range} attractions that can cause separation into two
67: neutral phases: ion-rich and
68: ion-poor~\cite{fisher-review,weing,stell94,fisher-lee96,fisher-levin93}; the
69: order parameter, namely, the ion density or concentration difference, is a
70: scalar; so Ising-type behavior is indicated. Field-theoretic approaches support
71: this picture~\cite{moreira94}.
72:
73: However, the theoretical arguments are by no means rigorous and have not, so
74: far, been tested by precise calculations for appropriate models. To do that is
75: the aim of the researches reported here. We have studied a finely-discretized
76: version~\cite{azp} of the simplest continuum model (considered by Debye and
77: H\"uckel in 1923~\cite{fisher-review,weing}, three years before Ising's work),
78: namely, the \emph{restricted primitive model} (RPM), consisting of $N=N_+ +
79: N_-$ equisized hard spheres of diameter~$a$, precisely half carrying a
80: charge~$+q_0$ and half $-q_0$, in a medium (representing a solvent) of
81: dielectric constant~$D$. At a separation $r \geq a$, like (unlike) ions
82: interact through the potential $\pm q_0^2/Dr$; thus appropriate reduced
83: density, $\rho=N/V$ for volume~$V$, and temperature variables are
84: \begin{equation}
85: \rho^* = \rho a^3 \;, \quad
86: T^* = k_B T D a / q_0^2 \;, \quad
87: t= (T-T_c)/T_c \;.
88: \label{eq:criticalvalues}
89: \end{equation}
90:
91: Except at low densities and high temperatures, when the inverse Debye length
92: $\kappa_D a = (4\pi\rho^*/T^*)^{1/2}$ is small, the RPM is intractable
93: analytically or via series expansions~\cite{fisher-review,stell94,bekiranov99}.
94: However, it has been much studied by Monte Carlo (MC)
95: simulations~\cite{fisher-review,caillol,orkoulas,jcp} which have recently
96: approached the consensus $T_c^* \simeq 0.049$, $\rho_c^* = 0.060$--$0.085$.
97: However, these values have been derived by assuming Ising-type criticality: on
98: that basis Bruce--Wilding extrapolation procedures have been
99: employed~\cite{caillol,orkoulas} (which, even then, neglect potentially
100: important, asymmetric `pressure-mixing' terms~\cite{fisher-orkoulas00}.) It
101: must be stressed that implementing appropriate finite-size extrapolation
102: methods constitutes the heart of the computational task since a grand-canonical
103: (GC) system confined in a simulation `box' of dimensions $L \times L \times L$
104: (with, say, periodic boundary conditions~\cite{ewald}) \emph{cannot} exhibit a
105: sharp critical point; a finite canonical system may become critical but can
106: display \emph{only classical} or vdW behavior~\cite{ofp}.
107:
108: Thus, while previous RPM simulations~\cite{caillol,orkoulas} demonstrate
109: \emph{consistency} with Ising (or $n=1$) behavior, \emph{no other} universality
110: classes are ruled out: see also~\cite{jcp,ofp,camp}. Putative `nearby'
111: candidates are XY or $n=2$ systems (with $\gamma \simeq 1.316$, $\nu \simeq
112: 0.670$), self-avoiding walks (SAWs, $n=0$: with $\gamma \simeq 1.159$, $\nu
113: \simeq 0.588$)~\cite{ofp,guida-zj} and long-range, $1/r^{d+\sigma}$ scalar
114: systems (with $d=3$, $\sigma < 2-\eta$)~\cite{camp,lrfss}. On the other
115: hand, in a preparatory GCMC study \textbf{II}~\cite[(b)]{ofp} of the hard-core
116: square-well (HCSW) fluid---for which Ising criticality has long been
117: anticipated---new, \emph{unbiased}, finite-size extrapolation techniques
118: enabled the $n=2$ and~$0$ classes to be convincingly excluded.
119:
120: \begin{figure}
121: \includegraphics[width=\figurewidth]{phasediag.eps}
122: \caption{Approximate coexistence curve of the RPM in the $(T,\rho)$ plane: open
123: circles and fitted line. The estimated critical point is shown as an
124: uncertainty bar. The dashed curves are loci of $C_V(T)$ maxima at fixed~$\rho$
125: for $L^* \equiv L/a = 8$, $10$, and~$12$. The loci labeled $k=1$, $\vartheta$,
126: and~$Q$ are explained in the text. The inset shows the canonical critical
127: points $T_c^0(L)$, $\rho_c^0(L)$ (squares), and corresponding GC mean densities
128: $\rho_c^\dagger(L)$ (crosses) for $L^*=9$--$12$, the $C_V(L)$ extrema
129: $T_c^-(L), \rho_c^-(L)$, for $L^*=7$--$10$ and~$12$ (solid circles), and the
130: $\sqrt{\rho}$ diameter, $\bar\rho^*_{1/2}(T)$, defined in the text (open
131: squares).}
132: \label{fig:coex}
133: \end{figure}
134:
135: \paragraph{Present approach}%
136: We have now applied the methods of \textbf{II} to the RPM; however, the extreme
137: \emph{asymmetry} of the critical region in the model (see Fig.~\ref{fig:coex})
138: has demanded further developments. By extending finite-size scaling
139: theory~\cite{kim} and previous applications of the Binder parameter or
140: fourth-moment ratio~\cite{lrfss,kim,binder}
141: \begin{equation}
142: Q_L(T; \rho) \equiv \langle m^2\rangle^2/\langle m^4\rangle
143: \quad\text{with}\quad m = \rho - \langle\rho\rangle \;,
144: \label{eq:Q}
145: \end{equation}
146: \cite{note-expectation} to systems lacking symmetry, we have assembled
147: evidence, outlined below, that excludes not only classical criticality in the
148: RPM but also the XY and SAW universality classes and $(d=3)$ long-range
149: Ising criticality with $\sigma \lesssim 1.9$.
150:
151: Our work employs multihistogram reweighting~\cite{histogram} and a ($\zeta \! =
152: \! 5$)-level fine-discretization formulation (with a fine-lattice spacing
153: $a/\zeta$~\cite{azp}). Since $\zeta < \infty$, \emph{non}universal parameters,
154: such as $T_c^*$, will deviate slightly from their continuum limit
155: ($\zeta\to\infty$)~\cite{azp,note-discretization}; but, at this level, there
156: are no serious grounds for contemplating changes in universality class. For the
157: critical parameters we find $T_c^* = 0.05069(2)$ and $\rho_c^*=0.0790(25)$: the
158: confidence limits in parentheses refer, here and below, to the last decimal
159: place quoted. The inset in Fig.~\ref{fig:coex} shows how these values are
160: approached (i) by the canonical values $T_c^0(L)$, $\rho_c^0(L)$ and
161: $\rho_c^\dagger(L)$ ($=\langle \rho\rangle_{T_c^0(L),\,
162: \mu_c^0(L)}$~\cite{note-expectation}) derived from the isothermal density
163: histograms [see \textbf{II}(2.18)--(2.23), Figs. 1,~3], (ii) by $T_c^-(L)$ and
164: $\rho_c^-(L)$, from the isochoric maxima of $C_V(T; \rho; L)$ [see
165: Fig.~\ref{fig:coex} and \textbf{II} Sec.~III, Fig.~7], and (iii) by the
166: $\sqrt{\rho}$ diameter, $\bar\rho_{1/2}(T)$, defined below.
167:
168: \begin{figure}
169: \includegraphics[width=\figurewidth]{gamma_eff.eps}
170: \caption{Effective susceptibility exponent $\gamma_{\text{eff}}^+(T)$ for
171: $\rho=\rho_c$ (solid curves) and $\tilde\gamma^+_{\text{eff}}(T)$ on the theta
172: locus (dashed; see text), for sizes $L^*=7$--$12$ and~$15$. Values for
173: vdW and for $n=0$, $1$, and~$2$ are marked on the $\gamma$ axis.}
174: \label{fig:gamma}
175: \end{figure}
176:
177: \paragraph{Exponents $\bm\gamma$ and $\bm\nu$}%
178: Before justifying the precision of our $(T_c, \rho_c)$ estimates, we consider
179: their implications. The solid curves in Fig.~\ref{fig:gamma} portray the
180: effective susceptibility exponent $\gamma_{\text{eff}}^+(T; L)$ on the critical
181: isochore above $T_c$, as derived from $\chi_{NN} \equiv V\langle m^2\rangle =
182: k_B T \rho^2 K_T$: see \textbf{II}(3.7). Within statistical precision the data
183: are independent of the $(T_c, \rho_c)$ uncertainties.
184:
185: Also presented in Fig.~\ref{fig:gamma} are the modified estimators
186: $\tilde\gamma^+_{\text{eff}}(T)$ [defined as in \textbf{II}(3.7) but with $t$
187: replacing $t'$] evaluated on the `theta locus,' $\rho_\vartheta(T) =
188: \rho_c[\vartheta + (1-\vartheta)(T_c/T)]$. This relation approximates an
189: \emph{effective symmetry locus}~(\textbf{II}) above $T_c$, derived from the
190: behavior of the isothermal inflection loci $\rho_k(T; L)$, on which $\chi^{(k)}
191: \equiv \chi_{NN}(T, \rho; L)/\rho^k$ is maximal [see
192: \textbf{II}(2.26)--(2.32)]. The $k=1$ loci are shown in Fig.~1 for $L^* \equiv
193: L/a = 6,8,10,12$; the selected value $\vartheta \!= \! 0.20$ corresponds
194: roughly to $k\simeq 0.60$ (which may be identified with an optimal value: see
195: \textbf{II} and~\cite{kim}). However, the variation of the $k$ loci when $L$
196: increases is significantly more complicated in the RPM than in the
197: HCSW fluid~[11(c), 23].%\cite[(c)]{jcp} \cite{note-estimates}.
198: % There is no way to get this cited properly in RevTeX
199:
200: Extrapolation of the effective susceptibility exponents in Fig.~\ref{fig:gamma}
201: and those on the $k=0$ locus, etc.~\cite[(c)]{jcp}, to $t=0$ indicates
202: $\gamma=1.24(3)$, upholding Ising-type behavior while both XY and SAW values
203: are implausible.
204:
205: To determine the exponent~$\nu$ we have examined the peak positions, $T_j(L)$,
206: of various properties, $Y_j(T; L)$, on the critical isochore. Finite-size
207: scaling theory~\cite{kim} yields $\Delta T_j(L) \equiv T_j(L)-T_c \sim
208: L^{-1/\nu}$: Figure~\ref{fig:nu} demonstrates the estimation of $1/\nu$
209: (unbiased except for the imposed $T_c$ estimate) from the ratios $\Delta
210: T_j(L_1) / \Delta T_j(L_2)$ for various $j$ (see \cite[(c)]{jcp}), using an
211: established approach [see \textbf{I}(7)--(13), Fig.~1; \textbf{II}(3.1)]. The
212: data indicate $\nu = 0.63(3)$, excluding classical but supportive of Ising
213: ($n=1$) criticality, while $n=2$ and~$0$ seem less probable.
214:
215: \begin{figure}
216: \includegraphics[width=\figurewidth]{nu.eps}
217: \caption{Estimation of the correlation exponent~$\nu$ from the deviations
218: $T_j(L)-T_c$ for various properties $Y_j(T)$ on the critical isochore: see text
219: and \protect\cite[(c)]{jcp}. Values for $n=0,1,2$, i.e., SAW, Ising, and XY,
220: and classical (vdW) criticality are indicated.}
221: \label{fig:nu}
222: \end{figure}
223:
224: \paragraph{Estimation of $\bm{T_c^*}$}%
225: Consider, now, $Q_L(T; \rho)$ in~(\ref{eq:Q}), when $L\to\infty$. In \emph{any
226: single}-phase region of the $(T, \rho)$ plane $Q_L \to\frac{1}{3}$, indicative
227: of Gaussian fluctuations about $\langle \rho\rangle$; conversely, \emph{within}
228: a two-phase region, $\rho_-(T) < \rho < \rho_+(T)$, one finds $Q_L\to 1$
229: \emph{on} the \emph{diameter}, $\bar\rho(T) \equiv \frac{1}{2}(\rho_- +
230: \rho_+)$ for $T<T_c$, while, more generally,
231: \begin{equation}
232: 1 \geq Q_\infty(T; \rho) = 1 - 4y^2/(1+6y^2+y^4) > \textstyle{\frac{1}{2}} \;,
233: \label{eq:qcoex}
234: \end{equation}
235: where $|y|=2|\rho-\bar\rho(T)|/(\rho_- + \rho_+) < 1$. Finally, \emph{at
236: criticality}, $Q_L(T_c; \rho_c)$ approaches a \emph{universal value} $Q_c$
237: which, for cubic boxes with periodic boundary conditions, is $Q_c =
238: 0.4569\cdots$ for classical (vdW)~\cite[(b)]{binder} or $\infty$-range
239: systems~\cite[(c)]{binder} but $Q_c(n \! =\! 1) = 0.6236(2)$ for
240: Ising~\cite[(d),(e)]{binder} and $Q_c(n \! = \! 2) = 0.8045(1)$ for
241: XY~\cite[(f)]{binder} systems, while $Q_c(n \! =\! 0) = 0$~\cite[(b)]{binder}.
242: For long-range, $1/r^{3+\sigma}$ systems, $Q_c(\sigma)$ and also
243: $\gamma(\sigma)$, increase almost linearly from vdW to Ising values in the
244: interval $\frac{3}{2} \leq \sigma \leq (\gamma/\nu)_{n=1} \simeq 1.96_6$ with
245: $Q_c(\sigma \! = \! 1.9) \simeq 0.600$ and $\gamma(\sigma \! = \! 1.9) \simeq
246: 1.20_5$~\cite[(b)]{lrfss}.
247:
248: The result~(\ref{eq:qcoex}) leads us to propose $Q$-loci, $\rho_Q(T; L)$, on
249: which $Q_L(T; \rho)$ is maximal at fixed~$T$. For $T<T_c$ these loci are
250: observed to approach the diameter $\bar\rho(T)$ when $L$ increases. (For $T
251: \lesssim T_c$, but \emph{not} above $T_c$, the $Q$-loci also follow the $k=0$
252: loci quite closely.)
253:
254: \begin{figure}
255: \includegraphics[width=\figurewidth]{q_qmax.eps}
256: \caption{Plots of $Q_L(T; \rho)$ on the $Q$-loci, $\rho_Q(T; L)$, providing
257: estimates for $T_c$ and $Q_c$. Classical, XY and Ising values of $Q$ are
258: shown.}
259: \label{fig:Q}
260: \end{figure}
261:
262: Figure~\ref{fig:Q} displays $Q_L(T; \rho)$ on the $Q$-loci $\rho_Q(T; L)$, for
263: $L^*=7$--$12$. As often seen in plots for \emph{symmetric}
264: systems~\cite{binder}, inflection points and successive intersections,
265: $T_Q(L)$, almost coincide! Scaling yields $Q_L(T_c; \rho_c) \sim
266: L^{-\theta/\nu}$ and $|T_c-T_Q(L)| \sim L^{-\varphi}$ with $\varphi =
267: (1+\theta)/\nu$, where $\theta$ $(=\omega\nu)$ is the leading
268: correction-to-scaling exponent; for classical and Ising criticality one has
269: $(\theta/\nu, \varphi) = (1, 3)$, $\simeq (0.82, 2.41)$~\cite{guida-zj}. With
270: this guidance, the large-scale inset in Fig.~\ref{fig:Q} leads to our estimate
271: $T_c^* \simeq 0.05069(2)$ but also yields $Q_c \simeq 0.624(2)$: this is
272: surprisingly close to the Ising value~\cite{note-q12} and far from the vdW, XY,
273: and SAW values---an unexpected bonus! Likewise, $1/r^{3+\sigma}$ effective
274: potentials with $\sigma \leq 1.9$ are excluded.
275:
276: \begin{figure}
277: \includegraphics[width=\figurewidth]{rho.eps}
278: \caption{Estimation of $\rho_c^*$ from plots of $(k=0)$ and $Q$ locus values at
279: $T_c$ (open and solid squares) vs $A/(L^*+\ell_0)^\psi$ for various values of
280: $\psi$ and optimal shifts~$\ell_0$. The scale parameter~$A$ has been invoked
281: merely for graphical clarity. Note $\psi < 1.6$ requires smaller shifts
282: tending to exclude vdW criticality ($\psi=2$).}
283: \label{fig:rhoc}
284: \end{figure}
285:
286: \paragraph{Estimation of $\bm{\rho_c}$}%
287: Finally, we examine $\rho_0^c(L)$ and $\rho_Q^c(L)$, i.e., the ($k=0$) and $Q$
288: loci intersections with the estimated critical isotherm, $T=T_c$. According to
289: scaling, the deviations, $\Delta \rho_0^c$ and $\Delta \rho_Q^c$, decay as
290: $L^{-\psi}$ with $\psi=(1-\alpha)/\nu$~\cite{kim}, so we may suppose $1.2 <
291: \psi \leq 2$~\cite{guida-zj}. Figure~\ref{fig:rhoc} displays the deviations vs
292: $L^{-\psi}$ for $\psi = 1.2$, $1.4$, $1.7$ and~$2$ with `$\ell_0$ shifts'
293: [\textbf{I}(19), Fig.~2; \textbf{II}(3.1)] chosen to provide linear plots.
294: From these and further plots~\cite[(c)]{jcp} we conclude $\rho_c^*=0.0790(25)$.
295:
296: In further support of our $\rho_c$ estimate, we mention first that when the
297: coexistence curve, $\rho_\pm(T)$, is plotted vs $\sqrt{\rho^*}$---as is
298: reasonable since all powers $\rho^{j/2}$ for integral~$j$ appear in virial
299: expansions for the RPM~\cite{bekiranov99}---it becomes markedly more
300: symmetrical [resembling $(\rho, T)$ plots for the HCSW and other simple
301: fluids]. Then, the \emph{corresponding} diameter,
302: $\sqrtsign[\bar\rho^*_{1/2}(T)] = \frac{1}{2}(\sqrt{\rho^*_-} +
303: \sqrt{\rho^*_+})$, is only mildly curved and naive extrapolations to $T_c$
304: yield $\rho_c^* = 0.078(4)$.
305:
306: \paragraph{In conclusion}%
307: By implementing recently tested~\cite{ofp} and newly devised extrapolation
308: techniques for nonsymmetric critical systems, our extensive grand-canonical
309: Monte Carlo simulations for the RPM have provided, \emph{in toto}, convincing
310: evidence to exclude classical, XY ($n=2$), or SAW ($n=0$) critical behavior as
311: well as long-range (effective) Ising interactions decaying more slowly than
312: $1/r^{4.90}$. Rather, the estimates for the exponents $\nu$ and $\gamma$, and
313: for the critical fourth-moment ratio, $Q_c$, point to standard, short-range
314: Ising-type criticality. Studies underway~\cite[(c)]{jcp} should provide
315: further confirmation and additional quantitative results, such as the scale,
316: $R_0$, of the equivalent single-component short-range attractions generated by
317: the RPM near criticality.
318:
319: \begin{acknowledgments}
320: We are indebted to Young C. Kim for extensive assistance in the numerical
321: analysis, for his elucidation of the finite-size scaling properties of
322: asymmetric fluid criticality~\cite{kim}, and his discovery of effective
323: estimators for $\nu$. The support of the National Science Foundation
324: (through Grant No.\ 99-81772, MEF) and the Department of Energy, Office of
325: Basic Energy Sciences (DE-FG02-01ER15121, AZP) is gratefully acknowledged.
326: \end{acknowledgments}
327:
328: \begin{thebibliography}{99}
329:
330: \bibitem{fisher-review} See (a) M. E. Fisher, J. Stat. Phys. \textbf{75}, 1
331: (1994), (b) J. Phys. Condens. Matter \textbf{8}, 9103 (1996), and references
332: therein.
333:
334: \bibitem{weing} H. Weing\"artner and W. Schr\"oer, Adv. Chem. Phys.
335: \textbf{116}, 1 (2001).
336:
337: \bibitem{stell94} See G. Stell, J. Stat. Phys. \textbf{78}, 197 (1994).
338:
339: \bibitem{fisher-lee96} M. E. Fisher and B. P. Lee, Phys. Rev. Lett.
340: \textbf{77}, 3561 (1996).
341:
342: \bibitem{fisher-levin93} M. E. Fisher and Y. Levin, Phys. Rev. Lett.
343: \textbf{71}, 3826 (1993).
344:
345: \bibitem{moreira94} See, e.g., A. G. Moreira, M. M. Telo da Gama and M. E.
346: Fisher, J. Chem. Phys. \textbf{110}, 10058 (1999) and references therein.
347:
348: \bibitem{azp} (a) A. Z. Panagiotopoulos and S. K. Kumar,
349: Phys. Rev. Lett. \textbf{83}, 2981 (1999); (b) A. Z. Panagiotopoulos,
350: J. Chem. Phys. \textbf{112}, 7132 (2000); (c) A. Z. Panagiotopoulos,
351: J. Chem. Phys. [in press].
352:
353: \bibitem{bekiranov99} S. Bekiranov and M. E. Fisher, Phys. Rev. E \textbf{59},
354: 492 (1999).
355:
356: \bibitem{caillol} J. M. Caillol, D. Levesque and J. J. Weis,
357: Phys. Rev. Lett. \textbf{77}, 4039 (1996); J. Chem. Phys. \textbf{107}, 1565
358: (1997).
359:
360: \bibitem{orkoulas} G. Orkoulas and A. Z. Panagiotopoulos,
361: J. Chem. Phys. \textbf{101}, 1452 (1994); \textbf{110}, 1581 (1999);
362: Q. Yan and J. J. de Pablo, J. Chem. Phys. \textbf{111}, 9509 (1999).
363:
364: \bibitem{jcp} E. Luijten, M. E. Fisher and A. Z. Panagiotopoulos, (a) J. Chem.
365: Phys. \textbf{114}, 5468 (2001); (b) a preliminary account of the present
366: work was presented at \textsc{STATPHYS 21} in Cancun, Mexico, on 16 July
367: 2001; (c) to be published.
368:
369: \bibitem{fisher-orkoulas00} M. E. Fisher and G. Orkoulas,
370: Phys. Rev. Lett. \textbf{85}, 696 (2000).
371:
372: \bibitem{ewald} We use Ewald summations with conducting boundary conditions
373: ($\varepsilon_\infty \to \infty$)~\protect\cite[(c)]{azp}, a screening
374: parameter $\kappa = 6/L$, chosen so as to ensure a sufficiently fast decay of
375: the real-space charge distribution, and 1152 wave vectors (to accomodate the
376: relatively rapid variation of the charge distribution).
377:
378: \bibitem{ofp} (a) G. Orkoulas, A. Z. Panagiotopoulos and M. E. Fisher, Phys.
379: Rev. E \textbf{61}, 5930 (2000); (b) G. Orkoulas, M. E. Fisher and A. Z.
380: Panagiotopoulos, Phys. Rev. E \textbf{63}, 051507 (2001); to be denoted
381: \textbf{I} and \textbf{II}.
382:
383: \bibitem{camp} P. J. Camp and G. N. Patey, J. Chem. Phys. \textbf{114}, 399
384: (2001).
385:
386: \bibitem{guida-zj} R. Guida and J. Zinn-Justin, J. Phys. A \textbf{31}, 8103
387: (1998).
388:
389: \bibitem{lrfss} E. Luijten, (a) Phys. Rev. E \textbf{60}, 7558 (1999); (b)
390: \textit{Interaction range, Universality and the Upper Critical Dimension}
391: (Delft Univ. Press, Delft, 1997).
392:
393: \bibitem{kim} Y. C. Kim and M. E. Fisher (to be published).
394:
395: \bibitem{binder} (a) K. Binder, Z. Phys. B \textbf{43}, 119 (1981); (b) E.
396: Br\'ezin and J. Zinn-Justin, Nucl. Phys. B \textbf{257}, 867 (1985); (c) E.
397: Luijten and H. W. J. Bl\"ote, Int. J. Mod. Phys. C \textbf{6}, 359 (1995);
398: (d) H. W. J. Bl\"ote, E. Luijten and J. R. Heringa, J. Phys. A \textbf{28},
399: 6289 (1995); (e) H. W. J. Bl\"ote, L. N. Shchur and A. L. Talapov, Int. J.
400: Mod. Phys. C \textbf{10}, 1137 (1999); (f) M. Campostrini \emph{et al.},
401: Phys. Rev. B \textbf{63}, 214503 (2001).
402:
403: \bibitem{note-expectation} All expectation values, $\langle \bm{\cdot}
404: \rangle$, pertain to finite GC systems at fixed~$T$ and chemical
405: potential~$\mu$, chosen, when appropriate, to provide a mean density $\langle
406: \rho\rangle$ which we denote merely by $\rho$ when used as a variable, as in
407: $Q_L(T; \rho)$.
408:
409: \bibitem{histogram} A. M. Ferrenberg and R. H. Swendsen,
410: Phys. Rev. Lett. \textbf{63}, 1195 (1989).
411:
412: \bibitem{note-discretization} Compare J. M. Romero-Enrique \emph{et al.}, Phys.
413: Rev. Lett. \textbf{85}, 4558 (2000), using $\zeta=10$, with Q. L. Yan and J.
414: J. de Pablo, Phys. Rev. Lett. \textbf{86}, 2054 (2001) and see
415: \protect\cite[(c)]{azp}.
416:
417: \bibitem{note-estimates} For this reason the estimates quoted in the \emph{Note
418: added in proof} in \protect\cite[(a)]{jcp} are inaccurate.
419:
420: \bibitem{note-q12} A similar analysis of $Q_L^{(1,2)} \equiv \langle
421: |m|\rangle^2/\langle m^2\rangle$ on the RPM $Q$-loci confirms $T_c^*$ and
422: yields $Q_c^{(1,2)} \simeq 0.807$ close to the corresponding Ising value
423: $0.8070(9)$ (E. Luij\-ten, unpublished).
424:
425: \end{thebibliography}
426:
427: \end{document}
428: