cond-mat0506447/hs.tex
1: %
2: \documentclass[12pt]{iopart}
3: \usepackage{graphicx}
4: % Uncomment next line if AMS fonts required
5: %\usepackage{iopams}  
6: 
7: \begin{document}
8: 
9: \title{Configurational entropy of hard spheres}
10: 
11: \author{Luca Angelani$^1$ and \ Giuseppe Foffi$^{2}$}
12: 
13: 
14: \address{$^1$ Research center SMC INFM-CNR, c/o Universit\`a
15: di Roma ``La Sapienza'', P.le A. Moro 2, I-00185, Roma, Italy}
16: 
17: \address{$^2$ Institut Romand de Recherche Num\'erique en
18: Physique des Mat\'eriaux IRRMA,
19:  and Institute of Theoretical Physics (ITP), Ecole Polytechnique F\'ed\'erale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland
20: }
21: 
22: 
23: \begin{abstract}
24: We  numerically calculate the configurational entropy $S_{conf}$ of a binary mixture 
25: of hard spheres, by using a perturbed Hamiltonian method trapping the system inside a given state,
26: which requires less assumptions than the previous methods
27: [R.J.~Speedy, Mol. Phys. {\bf 95}, 169 (1998)].
28: We find that  $S_{conf}$ is  a decreasing function of  packing fraction $\varphi$
29: and extrapolates to zero at the Kauzmann packing fraction
30: $\varphi_{_K} \simeq 0.62$, suggesting the possibility of an ideal 
31: glass-transition for hard spheres system.
32: Finally, the Adam-Gibbs relation is found to hold.
33: \end{abstract}
34: 
35: 
36: 
37: 
38: \section{Introduction}
39: 
40: The idea that the glass transition is driven by a decreasing of the number of
41: accessible states upon lowering temperature (or raising density) is quite old
42: \cite{kauzmann,AG,GM}.  In this picture,
43: if crystallization is avoided,
44: an ideal glass transition is expected
45: to happen at the point where the configurational entropy $S_{conf}$ (the
46: logarithm of the number of states) vanishes.  When the liquid enters in the
47: the supercooled region, the dynamics becomes slower and slower and the
48: particles get trapped for an increasingly longer time
49:  inside the ``cages'' made by their
50: neighbors: the dynamics of the system can be successfully described as a
51: ``fast'' motion of the representative point in the $3N$ configuration space
52: inside metastable states, and a ``slow'' motion corresponding to jumps among
53: states. Entering more in the supercooled region the number of accessible
54: metastable states decreases
55: and the extrapolation to zero of $S_{conf}$ defines the ideal glass
56: transition.  In experiments (or numerical simulations) the region close to the
57: ideal glass transition is unreachable, due to the ``apparent'' arrest of the
58: system at the so called glass-transition temperature (or density) when
59: relaxation times become longer than experimental time scale.  The above
60: scenario has been shown to be valid for many interacting systems, based on
61: smooth pair-potential (as Lennard-Jones liquids), for which the Potential
62: Energy Landscape (PEL) approach
63: \cite{deb_nature,stilweb,stil,skt,scala,land_sastry,land_buc,land_sastry2,keyes_inm,fsschool}
64: and the replica method \cite{mezard,coluzzi0} have allowed to give numerical
65: estimations of $S_{conf}$ and of the ideal glass transition.
66: 
67: The overall picture is still not well established for Hard Spheres (HS), for
68: which the existence of a glass transition is still an
69: open question \cite{rt,robles,speedy3,tarzia,donev}.
70: A particularly important role seems to be played by the dimesionality of the
71: system. In particular in d=2 dimension there are numerical \cite{donev,sen_kra}
72: and theoretical \cite{tarzia2,zam_and}
73: evidences of the absence of a thermodynamic glass transition, while the
74: opposite seems to be true in d=3~\cite{speedy3,paza}.  
75: Moreover, the step-wise
76: form of the interparticle potential does not allow a PEL analysis 
77: and different
78: approaches have to be taken in consideration in order to calculate the
79: configurational entropy $S_{conf}$.  In the past, attempts to estimate
80: $S_{conf}$ have been performed based on different evaluations of the entropy
81: in each single state \cite{speedy3,speedy2,aste}.  Recently the replica method
82: has been extended to the HS case for one-component systems \cite{zam_and,paza}.
83: 
84: 
85: 
86: In this paper we  follow an approach, based on the Frenkel-Ladd method
87: \cite{frenkel_book} and recently introduced in the study of
88: Lennard-Jones systems \cite{coluzzi} and attractive colloids
89: \cite{noi,moreno}, to numerically estimate $S_{conf}$ for binary hard
90: spheres.  As in previous studies, the calculation is reduced to that
91: of vibrational entropy $S_{vib}$, using the fact that the total
92: entropy $S$ can be decomposed into the sum of a configurational
93: contribution $S_{conf}$ and a vibrational one $S_{vib}$:
94: \begin{equation}
95: S = S_{conf} + S_{vib} \ .
96: \label{entropy}
97: \end{equation}
98: This expression is consistent with the idea that, at high enough density,
99: there are two well separated time scales: a fast one, related to the motion
100: inside a local state (the rattling in the cage), and a slow one associated to 
101: the exploration of different states.\\
102: The total entropy $S$ is obtained by thermodynamic integration, starting from
103: the ideal gas state. The quantity $S_{vib}$ is calculated using a perturbed
104: Hamiltonian, adding to the original Hamiltonian an harmonic potential around a
105: given reference configuration. Calculating the mean square displacement from
106: the reference configuration and making an integration over the strength of the
107: perturbation, it is possible to estimate the vibrational entropy \cite{noi}.
108: The difference $S-S_{vib}$ provides an estimate of the configurational entropy
109: $S_{conf}$ as a function of packing fraction $\varphi$ (or density $\rho$).
110: 
111: The main findings of the present work are the following.  
112: {\it i}) $S_{conf}$ is a decreasing function of packing fraction $\varphi$, 
113: and a suitable extrapolation to zero provides and estimate of the ideal 
114: phase transition point (Kauzmann packing fraction) 
115: $\varphi_{_K} \simeq 0.62$.
116: {\it ii}) The diffusivity
117: $D$ and configurational entropy $S_{conf}$ are related through the Adam-Gibbs
118: relation, in agreement with previous claims \cite{speedy3}.
119: 
120: 
121: \section{The model}
122: 
123: The studied system is a binary mixture $50-50$ of hard spheres, $A$ and $B$, with diameters
124: ratio $\sigma_{_B}/\sigma_{_A}$=$1.2$.  The collision diameters are
125: $\sigma_{_{AA}}$=$\sigma_{_A}$, $\sigma_{_{BB}}$=$\sigma_{_B}$ and
126: $\sigma_{_{AB}}$=$(\sigma_{_A}+\sigma_{_B})/2$.  The particles ($N$=$256$) are
127: enclosed in a cubic box with periodic boundary conditions.   We use the
128: following units: $\sigma_{_B}$ for length and $m_A=m_B=1$ for mass. Moreover
129: we chose $k_B=1$ and $\hbar=1$.  The density is measured by the packing
130: fraction $\varphi$ that is related to the number density $\rho=N/V$ by
131: $\varphi$=$\rho \pi (\sigma_{_A}^3+\sigma_{_B}^3)/12$. We analyzed systems in
132: the range $\varphi$=$0.425-0.580$. Hard spheres systems depend only trivially
133: on temperature that sets an overall scale for the dynamics, consequently we
134: perform all our simulations at $T=1$. We performed standard event-driven
135: molecular dynamics \cite{rapaport} and we stored several equilibrated
136: configurations at different density.
137: \begin{figure}[t]
138: \vspace{0.8cm}
139: \begin{center}
140: \includegraphics[width=.6\textwidth]{fig1}
141: \end{center}
142: \caption{ Mean square displacements (MSD) of the species $A$ (top) and
143: $B$ (bottom). Both MSDs has been normalized by the value of the
144: diameter squared.  The dashed lines represent the cage size squared,
145: $\delta r^2$.}
146: \label{fig1}
147: \end{figure}
148: 
149: \section{Diffusivity}
150: 
151: The diffusion coefficients $D$ of the two species have been extracted from the
152: long time limit of the mean square displacements (MSD) $\langle r^2(t)\rangle$=$N^{-1}\langle [{\bf
153: r}(t)-{\bf r}(0)]^2 \rangle$ (${\bf r}$ is the $3N$-vector of
154: the coordinates):
155: \begin{equation} 
156: \lim_{t\rightarrow \infty}\frac{\langle r^2(t)\rangle}{t}\simeq 6D \ .
157: \label{einstein} 
158: \end{equation} 
159: To improve the statistical significance of the data, an average over $10$
160: independent runs have been performed. 
161: In Fig \ref{fig1} the mean squared displacements for the slowest cases, 
162: i.e. $\phi >0.56$, are presented for both species.  
163: It is clear that on increasing the density the MSDs develop the typical 
164: two step relaxation pattern. The first part of the MSD is purely ballistic while, 
165: at later stage, it reaches the diffusive regime, described by Eq.~\ref{einstein}.  
166: Between this two regimes a plateau starts to develop. This is the clear indication 
167: of a caging effect. Each particle starts to feel the crowding of its neighbors and 
168: it is trapped in a cage for longer and longer time on increasing the density.
169: The height of the plateau is the typical ``cage diameter squared'', $\delta r^2$. 
170: For both species we find $\delta r^2 =3\cdot 10^{-2} \sigma_\alpha^2$ 
171: for $\alpha=A \mbox{ or } B$ represented by a dashed line in Fig.~\ref{fig1}. 
172: This is a clear evidence that the two species have the same caging effect. 
173: We shall return on the value of $\delta r^2$ later on.  
174: 
175: In Fig. \ref{fig2} the diffusivities $D$ of A and B particles are
176: plotted as a function of the packing fraction $\varphi$. 
177: Dashed lines in the figure
178: are power-law fits $D = C\;(\varphi_c - \varphi)^\gamma$ of the high packing-fraction data 
179: ($\varphi \geq 0.53$), 
180: as predicted by mode-coupling theory.
181: The fitted parameters are: 
182: $\varphi_c$=$0.583$, $\gamma$=$2.27$, $C$=$9.50$ for A
183: particles and $\varphi_c$=$0.583$, $\gamma$=$2.47$, $C$=$11.66$ for B
184: particles.  We note that both diffusivities give rise to the same
185: mode-coupling packing fraction $\varphi_c$, in agreement with the
186: prediction of the theory \cite{Voigtmann} and with previous
187: simulations of the same model \cite{foffietal}.
188: \begin{figure}[t]
189: \vspace{0.8cm}
190: \begin{center}
191: \includegraphics[width=.6\textwidth]{fig2}
192: \end{center}
193: \caption{
194: Diffusivity for A and B particles as a function of packing fraction $\varphi$.
195: The lines are  power-law fits for $\varphi \geq 0.53$,
196: $C\;(\varphi_c - \varphi)^\gamma$, 
197: with:
198: $\varphi_c$=$0.583$, $\gamma$=$2.27$, $C$=$9.50$ for A particles;
199: $\varphi_c$=$0.583$, $\gamma$=$2.47$, $C$=$11.66$ for B particles.
200: }
201: \label{fig2}
202: \end{figure}
203: 
204: 
205: \section{Configurational entropy}
206: 
207: We now turn to the calculation of configurational entropy.  The method we
208: follow to estimate $S_{conf}$ requires the computation of the total entropy
209: $S$ and vibrational entropy $S_{vib}$.  The total entropy $S$ is calculated
210: {\it via} a thermodynamic integration from ideal gas and can be expressed as
211: \begin{equation}
212: S (\rho) = S_{id}(\rho) + S_{ex}(\rho) \ ,
213: \label{S2}
214: \end{equation}
215: where $S_{id}$ is the entropy of the ideal gas and $S_{ex}$ is the 
216: excess entropy with respect to the ideal gas.
217: For a binary mixture, the ideal gas entropy is:
218: \begin{equation}
219: \frac{S_{id}(\rho)}{N} = \frac{5}{2}-\ln{\rho}-3\ln{\lambda}+\ln{2} \ ,
220: \label{Sid}
221: \end{equation}
222: where $\lambda=(2\pi\beta\hbar^2/m)^{\frac{1}{2}}$ is the De Broglie
223: wavelength ($\hbar$ is the Planck's constant and has been set to unitary value),
224: and the term $\ln{2}$ takes into account the mixing contribution.
225: The term $S_{ex}$ can be expressed in the following form
226: \begin{equation}
227: S_{ex}(\rho)  = - \frac{N}{T}
228: \int_{0}^{\rho} \frac{d\rho}{\rho^2}\  P_{ex} \ ,
229: \label{Sex}
230: \end{equation}
231: with $P_{ex}$ the excess pressure. We extracted $P_{ex}$ from the zero density
232: limit up to the densities of interest, performing numerical simulations and
233: fitting the results of the pressure with a high order polynomial in $\rho$.
234: In Fig. \ref{fig3} we show the numerically calculated excess entropy $S_{ex}$
235: (symbols) together with the analytic estimate provided by the
236: Carnahan-Starling (CS) equation of state, extended to hard sphere mixtures
237: \cite{al,cs,nota_cs}.  We note that at high densities the CS equation of state
238: overestimates the entropy of about $7\%$. This discrepancy, however, is
239: not sufficiently significant to affect the resulting $S_{conf}$ values, in
240: particularly closed to the glass transition.
241: 
242: 
243: \begin{figure}[t]
244: \vspace{0.8cm}
245: \begin{center}
246: \includegraphics[width=.6\textwidth]{fig3}
247: \end{center}
248: \caption{
249: Excess entropy $S_{ex}$ for the mixture of hard spheres as obtained from our simulations
250: (symbols) compared with the analytical Carnahan-Starling expression (line) \cite{cs,nota_cs}.
251: }
252: \label{fig3}
253: \end{figure}
254:  
255: The method we use for the calculation of $S_{vib}$ is based on the 
256: investigation of a perturbed system
257: \begin{equation}
258: \beta H' = \beta H + \alpha N ({\bf r} - {\bf r}_0)^2 \ ,
259: \end{equation}
260: where $H$ is the unperturbed hard spheres Hamiltonian, $\alpha$ is the
261: strength of the perturbation, ${\bf r}_0$ specifies the particles coordinates
262: of a reference configuration and $({\bf r} - {\bf r}_0)^2 \equiv
263: N^{-1}\sum_{i=1}^{N} ({\vec r}_i - {\vec r}_{0,i})^2$.  The reference
264: configuration ${\bf r}_0$ is chosen from equilibrium configurations at the
265: considered density (randomly extracted from the stored configurations obtained
266: during molecular dynamics simulations).  With this choice one is sure that the
267: estimated vibrational entropy (see formula below) pertains to the correct
268: state at the studied density.  The vibrational entropy can be obtained from
269: the formula (see Ref. \cite{noi} for details):
270: \begin{equation}
271: \frac{S_{vib}}{N} = \int_{\alpha_0}^{\alpha_{\infty}} d\alpha' \langle ({\bf r} - {\bf r}_0)^2 \rangle_{\alpha'}
272: - \frac{3}{2} \ln\left(\frac{\alpha_{\infty} \lambda^2}{\pi}\right) + \frac{3}{2} \ ,
273: \label{svib2}
274: \end{equation}
275: where $\alpha_{0,\infty}$ are the lower/upper limit of integration, 
276: and $\langle ... \rangle_{\alpha'}$ is the canonical average 
277: for a given $\alpha'$.
278: The choice of $\alpha_0$ deserves few comments.  If the system were confined
279: to move inside a given local free-energy minimum, 
280: for a correct estimation of $S_{vib}$ one would take the
281: lower limit $\alpha_0$=$0$ in the integral in Eq. \ref{svib2}.  As the system,
282: at enough low value of $\alpha$, begins to sample different states (the
283: harmonic force due to the perturbation is no more able to constrain the system
284: inside one state), $\alpha_0$ has to be chosen in such a way that the system
285: has not yet left the state: the underlying idea is that Eq. \ref{svib2} gives
286: a correct estimation of $S_{vib}$ until the system remains trapped in the
287: state.  An appropriate choice in our case seems to be
288: $\alpha_0^{(1)}$=$2^{2.5}$  for all the densities 
289: , as, close to this point, one observes a crossover
290: for all the investigated densities (more pronounced for low density data).  In
291: Fig. \ref{fig4} the quantity $\langle ({\bf r} - {\bf r}_0)^2
292: \rangle_{\alpha}$ is reported as a function of $\alpha$. An arrow indicates
293: the chosen value $\alpha_0$=$\alpha_0^{(1)}$, below which one observes the
294: crossover associated to the exploration of different states.
295: 
296: It is worth noting that different choices of $\alpha_0$ are in principle
297: possible, giving rise to different estimations of the vibrational entropy
298: term.  However, even though a kind of arbitrariness is present in the method,
299: one can argue that a reasonable choice should be for values above
300: $\alpha_0^{(1)}$ (close to the crossover corresponding to the exploration of
301: different states) and below an upper value $\alpha_0^{(2)}$ at which one is
302: sure that the system is still confined in a single state.  The latter value
303: can be estimated requiring that the MSD $\langle ({\bf r} - {\bf r}_0)^2
304: \rangle$ is always close/below the cage diameter squared $\delta r ^2 \simeq
305: 3 \cdot 10^{-2} \sigma_\alpha^2$ (with $\alpha=A \mbox{ or }B$) (this has been
306: estimated from the {\it plateau} of the mean square displacement, see
307: Fig.~\ref{fig1}).
308:  
309: 
310: The chosen value in our case is $\alpha_0^{(2)}$=$2^{4.5}$
311: (indicated by an arrow in Fig. \ref{fig4}).
312: We then repeated the same calculation of $S_{vib}$ using Eq.\ref{svib2} with 
313: the lower bound in the integral $\alpha_0=\alpha_0^{(2)}$.
314: In this way we obtain a lower and upper bound for the quantity of interest $S_{vib}$,
315: by using respectively $\alpha_0^{(1)}$ or $\alpha_0^{(2)}$ in the 
316: expression of $S_{vib}$ in Eq.\ref{svib2}.
317: 
318: \begin{figure}[t]
319: \vspace{0.8cm}
320: \begin{center}
321: \includegraphics[width=.6\textwidth]{fig4}
322: \end{center}
323: \caption{
324: The quantity $\langle ( {\bf r} - {\bf r}_0 )^2 \rangle_{\alpha}$
325: plotted vs. $\alpha$ in logarithmic scale for different packing fractions
326: $\varphi$.
327: Vertical lines are the values  $\alpha_0^{(1)}=2^{2.5}$ and 
328: $\alpha_0^{(2)}=2^{4.5}$ used as $\alpha_0$-value in the integral 
329: in Eq.~\ref{svib2} for the calculation of $S_{vib}$.
330: }
331: \label{fig4}
332: \end{figure}
333: 
334: 
335: 
336: Fig. \ref{fig5} shows $S_{conf}$ as a function of $\varphi$. The
337: configurational entropy $S_{conf}$ is calculated using Eq. \ref{entropy},
338: where the two entropies $S$ and $S_{vib}$ are obtained from Eq.s \ref{S2} and
339: \ref{svib2} respectively.  Due to the fact that the correct integral for the
340: estimation of $S_{vib}$ should be done from $\alpha_0$=$0$, but with the
341: system always inside a given state, we have added to the expression in Eq.
342: \ref{svib2} the term $\alpha_0 \; \langle ({\bf r} - {\bf r}_0)^2
343: \rangle_{\alpha_0}$, corresponding to assume a constant value of $\langle
344: ({\bf r} - {\bf r}_0)^2 \rangle$ below $\alpha_0$ and using a zero value for
345: the lower limit of the integral in Eq. \ref{svib2}.  Fig. \ref{fig5} shows the
346: two estimates of $S_{conf}$, corresponding to the two different values of
347: $\alpha_0$: $\alpha_0^{(1)}$=$2^{2.5}$ (open symbols) and
348: $\alpha_0^{(2)}$=$2^{4.5}$ (full symbols).  One observes that the discrepancy
349: between the two estimations decreases by increasing the packing fraction,
350: suggesting that, at high density, the method used to calculate
351: $S_{conf}$ is less affected by the choice of the free parameters entering in
352: its evaluation.  This is probably due to the fact that increasing the density
353: the system tends to be more trapped in a local free-energy minimum.
354: Indeed, it is only at high density that the method is expected to work better,
355: due to the better definition of two time scales corresponding to local-fast 
356: and global-slow dynamics (see Fig. 1). At low density, instead, the two are
357: less separated and this corresponds to a difficulty in the extrapolation for 
358: $\alpha_0 \to 0$ of the quantity reported in Fig. 4. The low density data 
359: show a  more clear crossover on lowering $\alpha$, and then
360: a worst definition of state in this limit. As we are interested in the high 
361: packing fraction extrapolation, this fact do not affect our prediction on the 
362: Kauzmann density value.
363: \begin{figure}[t]
364: \vspace{0.8cm}
365: \begin{center}
366: \includegraphics[width=.6\textwidth]{fig5}
367: \end{center}
368: \caption{ Configurational entropy $S_{conf}$ as function of packing fraction
369:   $\varphi$.  Open symbols are data using $\alpha_0^{(1)}=2^{2.5}$, full
370:   symbols using $\alpha_0^{(2)}=2^{4.5}$ (see text).  Dashed and dot-dashed
371:   lines are from Speedy \cite{speedy3} for binary and monatomic hard-spheres
372:   respectively.  Thin full line is the analytical computation of Parisi and
373:   Zamponi for monatomic hard-spheres \cite{paza}.  Thick lines are polynomial
374:   extrapolations of our data in the high packing fraction region, giving rise
375:   to the same Kauzmann packing fraction estimation for which
376:   $S_{conf}(\varphi_K)$=$0$: $\varphi_{_K} \simeq 0.62$.  }
377: \label{fig5}
378: \end{figure}
379: Also reported in the
380: figure are the curves obtained by Speedy \cite{speedy3} using a 
381: different method 
382:  (assuming a particular form of the vibrational entropy, a Gaussian 
383: distribution of states and involving some free parameters)
384: for the estimation of $S_{conf}$, for monatomic (dot-dashed
385: line) and binary (dashed line) hard-spheres (with the same diameters ratio
386: $1.2$ and composition $50:50$ as in our case).  
387: It is worth noting that our method improves on Speedy's one, as, 
388: even though requiring some accuracy in 
389: the choice of the $\alpha_0$ parameter, has the advantage to be less affected by 
390: the presence of many free parameters and particular assumptions.
391: We note that the data of
392: Speedy for the binary case do agree very well with our data with
393: $\alpha_0$=$\alpha_0^{(2)}$, suggesting the possibility that the choice of
394: $\alpha_0$=$\alpha_0^{(2)}$ is more accurate for the estimation of $S_{vib}$
395: and so of $S_{conf}$.  As a comparison, in Fig.~\ref{fig5} is also reported an
396: analytic estimation of $S_{conf}$ recently provided by Parisi and Zamponi
397: \cite{paza} for monatomic hard-spheres.  From the $\varphi$-dependence of the
398: configurational entropy one can determine the packing fraction at which
399: $S_{conf}$ extrapolates to zero, corresponding to the ideal phase transition
400: point (Kauzmann packing fraction $\varphi_{_K}$) $S_{conf}(\varphi_{_K})$=$0$.
401: Using a polynomial extrapolation \cite{nota_poly} 
402: for the two set of data (corresponding to the
403: different $\alpha_0$ values) we obtain an estimated Kauzmann packing fraction
404: value $\varphi_{_K} \simeq 0.62$ (see Fig.\ref{fig5}).
405: It is worth noting that, even
406: though the two curves 
407: are quite different, the estimated value of $\varphi_{_K}$ is the same, again
408: suggesting the robustness of the method in the high density region and then in
409: the estimation of the Kauzmann packing fraction.
410: 
411: 
412: 
413: 
414: \section{Adam-Gibbs relation}
415: \begin{figure}[t]
416: \vspace{0.8cm}
417: \begin{center}
418: \includegraphics[width=.6\textwidth]{fig6}
419: \end{center}
420: \caption{The
421: Adam-Gibbs relation 
422: $D$=$D_{\infty}\exp[-N\Delta/(TS_{conf})]$ 
423: ($T$=$1$)
424: for the two species A and B:
425: $D_{\infty}$=$3.66$, $\Delta$=$9.8$ for A particles;
426: $D_{\infty}$=$4.59$, $\Delta$=$10.5$ for B particles.
427: The data of $S_{conf}$ are obtained with $\alpha_0$=$\alpha_0^{(2)}$.
428: }
429: \label{fig6}
430: \end{figure}
431: 
432: In this Section we explore the validity of the Adam-Gibbs relation,
433: linking dynamic quantities, like diffusivity, 
434: to $S_{conf}$.
435: In Fig.~\ref{fig6} we report the diffusivities $D$
436: for A and B particles vs. 
437: the quantity $(TS_{conf})^{-1}$,
438: with $S_{conf}$ obtained for the value $\alpha_0$=$\alpha_0^{(2)}$.
439: We find that the Adam-Gibbs relation
440: \begin{equation}
441: D=D_{\infty} e^{-N\Delta/TS_{conf}}
442: \end{equation} 
443: is verified (lines in the figure), with: 
444: $D_{\infty}$=$3.66$, $\Delta$=$9.8$ for A particles;
445: $D_{\infty}$=$4.59$, $\Delta$=$10.8$ for B particles.
446: A similar behavior is obtained using $S_{conf}$ calculated with
447: $\alpha_0$=$\alpha_0^{(1)}$ (not shown in the figure), with the values:
448: $D_{\infty}$=$24.5$, $\Delta$=$9.5$ for A particles;
449: $D_{\infty}$=$37.5$, $\Delta$=$10.5$ for B particles, suggesting that, in this range of
450: diffusivity values,  the AG expression is not able to discriminate between the two different
451: estimations of $S_{conf}$.
452: 
453: 
454: 
455: \section{Conclusions}
456: 
457: In conclusion, we have calculated $S_{conf}$
458: for binary mixture hard spheres,
459: by numerically estimating the total entropy $S$ ({\it via} thermodynamic
460: integration from ideal gas) and the vibrational entropy $S_{vib}$  
461: using a numerical procedure based on Frenkel-Ladd method and recently applied
462: in the analysis of Lennard-Jones systems and attractive colloids:
463: the system is constrained inside a given ``{\it state}''
464: through an harmonic perturbed term in the Hamiltonian.
465: We found, in agreement with analytical and simulation results in the literature,
466: that $S_{conf}$ is a decreasing function of the packing fraction
467: $\varphi$, suggesting the possibility of a vanishing of $S_{conf}$ around
468: the Kauzmann point $\varphi_{_K}$=$0.62$.
469: Moreover, by studying the relationship between $S_{conf}$ and the diffusion
470: constant $D$, the Adam-Gibbs relation is found to reasonably hold for the analyzed system.
471: 
472: 
473: 
474: %\begin{acknowledgments}
475: \ack
476: We thank G Ruocco, F Sciortino and F Zamponi for useful discussions and suggestions.
477: G F acknowledges the support of the Swiss Science Foundation (Grant No. 200021-105382/1).
478: %\end{acknowledgments}
479: 
480: 
481: \section*{References}
482: 
483: \begin{thebibliography}{}
484: 
485: 
486: \bibitem{kauzmann}
487:  Kauzmann A W  1948
488: {\it Chem. Rev.}  {\bf 43}  219 
489: 
490: \bibitem{AG}
491: Adam G  and Gibbs J H  1958
492: {\it J. Chem. Phys.}  {\bf 43}  139 
493: 
494: \bibitem{GM}
495: Gibbs J H  and Di Marzio E A  1958
496: {\it J. Chem. Phys.}  {\bf 28}  373
497: 
498: \bibitem{deb_nature}
499: Debenedetti P G  and Stillinger F H  2001
500: {\it Nature}  {\bf 410}  259 
501: 
502: \bibitem{stilweb}
503: Stillinger F H  and Weber T A  1984
504: {\it Science}  {\bf 225}  983 
505: 
506: \bibitem{stil}
507: Stillinger F H 1995
508: {\it Science}  {\bf 267}  1935 
509: 
510: \bibitem{skt}
511: Sciortino F, Kob W and Tartaglia P 1999
512: {\it Phys. Rev. Lett.}  {\bf 83}  3214 
513: 
514: \bibitem{scala}
515: Scala A, Starr F W, La Nave E, Sciortino F and Stanley H E 2000
516: {\it Nature}  {\bf 406}  166 
517: 
518: \bibitem{land_sastry}
519: Sastry S, Debenedetti P G and Stillinger F H 1998
520: {\it Nature}  {\bf 393}  554 
521: 
522: \bibitem{land_buc}
523: B\"uchner S  and Heuer A 2000
524: {\it Phys. Rev. Lett.}  {\bf 84}  2168 
525: 
526: \bibitem{land_sastry2}
527: Sastry S  2001
528: {\it Nature}  {\bf 409}  164 
529: 
530: \bibitem{keyes_inm}
531: Keyes T  1997
532: {\it J. Phys. Chem.}  {\bf 101}  2921 
533: 
534: \bibitem{fsschool} 
535: Sciortino F  2005
536: {\it J. Stat. Mech.}  P05015 
537: 
538: 
539: \bibitem{mezard}
540: M\'ezard M  and Parisi G  1999
541: {\it J. Chem. Phys.}  {\bf 111}  1076 
542: 
543: \bibitem{coluzzi0}
544: Coluzzi B , Parisi G  and Verrocchio P  2000
545: {\it J. Chem. Phys.}  {\bf 112}  2933 
546: 
547: \bibitem{rt}
548: Rintoul M D  and Torquato S  1996
549: Phys. Rev. Lett. {\bf 77}  4198\\
550: Rintoul M D  and Torquato S  1996
551: {\it J. Chem. Phys.}  {\bf 105}  9258 
552: 
553: \bibitem{robles}
554: Robles M, L\`opez de Haro M, Santos A  and  Bravo Yuste S 1998
555: {\it J. Chem. Phys.}  {\bf 108}  1290
556: 
557: \bibitem{speedy3}
558: Speedy R J  1998
559: {\it Mol. Phys.}  {\bf 95}  169 
560: 
561: \bibitem{tarzia}
562: Tarzia M, de Candia A, Fierro A, Nicodemi M  and Coniglio A  2004
563: {\it Europhys. Lett.}  {\bf 66}  531
564: 
565: \bibitem{donev}
566: Donev A, Stillinger F H and Torquato S
567: 2006
568: {\it Phys. Rev. Lett.}  {\bf 96}  225502
569: 
570: \bibitem{sen_kra}
571: Santen L and Krauth W
572: 2000
573: {\it Nature} {\bf 405} 550
574: 
575: \bibitem{tarzia2}
576: Tarzia M
577: 2007
578: {\it J. Stat. Mech.} P01010
579: 
580: \bibitem{zam_and}
581: Zamponi F 2006 
582: {\it Phil. Mag.} {\bf 87}  485
583: 
584: \bibitem{paza}
585: Parisi G  and Zamponi F  2005
586: {\it J. Chem. Phys.}  {\bf 123}  133501
587: 
588: \bibitem{speedy2}
589: Speedy R J  1999
590: {\it J. Chem. Phys.} {\bf 110} 4559\\
591: Speedy R J  2001
592: {\it J. Chem. Phys.} {\bf 114}  9069
593: 
594: 
595: \bibitem{aste}
596: Aste T and Coniglio A  2003
597: {\it Physica A } {\bf 330}  189 \\
598: Aste T  and Coniglio A  2003
599: {\it J. Phys.: Condens. Matter } {\bf 15}  S803 \\
600: Aste T  and Coniglio A  2004
601: {\it Europhys. Lett.}  {\bf 67}  165
602: 
603: 
604: 
605: \bibitem{frenkel_book}
606: Frenkel D  and Smit B  2001
607: {\it Understanding Molecular Simulation} 
608: (Academic Press, London, 2nd)
609: 
610: \bibitem{coluzzi}
611: Coluzzi B, M\'ezard M, Parisi G, and Verrocchio P 1999
612: {\it J. Chem. Phys.}  {\bf 111}  9039 
613: 
614: \bibitem{noi}
615: Angelani L, Foffi G, Sciortino F and Tartaglia P 2005
616: {\it J.  Phys.: Condens. Matter}  {\bf 17}  L113;
617: Foffi G and Angelani L (in preparation) 
618: 
619: \bibitem{moreno} Moreno A J, Buldyrev S V, La Nave E,
620: Saika-Voivod I, Sciortino F, Tartaglia P  and  Zaccarelli E  2005
621: {\it Phys. Rev. Lett.}  {\bf 95}  157802
622: 
623: \bibitem{rapaport}
624: Rapaport D C  1997
625: {\it The art of computer simulations} 
626: (Cambridge Univ Press, London, 2nd)
627: 
628: \bibitem{Voigtmann}
629: Voigtmann Th 2003
630: {\it Phys. Rev. E}  {\bf 68} 051401
631: 
632: 
633: \bibitem{foffietal}
634: Foffi G, G\"otze W, Sciortino F, Tartaglia P and Voigtmann Th 2003
635: {\it Phys. Rev. Lett.}  {\bf 91}  085701 \\
636: Foffi G, G\"otze W, Sciortino F, Tartaglia P and Voigtmann Th 2004
637: {\it Phys. Rev. E}  {\bf 69}  011505
638: 
639: \bibitem{al}
640: Alder B J  1964
641: {\it J. Chem. Phys.}  {\bf 40}  2724 
642: 
643: \bibitem{cs}
644: Mansoori G A, Carnahan N F, Starling K E, and Leland T W Jr 1971
645: {\it J. Chem. Phys.}  {\bf 54}  1523 
646: 
647: 
648: \bibitem{nota_cs}
649: The analytical CS expression for $S_{ex}$ is given by Eq. $9$ in Ref. \cite{cs}
650: omitting the last term $\ln Z$, as in our case the excess entropy is defined with respect
651: to the ideal gas at the same density instead of at the same pressure as in Ref. \cite{cs}
652: 
653: 
654: \bibitem{nota_poly}
655: We use a polynomial function of the form
656: $A\;(\varphi_{_K} - \varphi) - B(\varphi_{_K} - \varphi)^2$
657: to fit the data with $\varphi>0.5$. Adding more polynomial orders do not 
658: significantly affect the extrapolated Kauzmann packing fraction value.
659: 
660: 
661: \end{thebibliography}
662: 
663: 
664: 
665: 
666: 
667: 
668: 
669: 
670: \end{document}
671: 
672: 
673: