1: %\documentstyle[aps,prl,preprint]{revtex}
2: %\documentstyle[preprint,aps,prl]{revtex}
3: \documentstyle[aps,prl,twocolumn]{revtex}
4: \draft
5: \begin{document}
6:
7: \include{psfig}
8: \draft
9: %\twocolumn[\hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse\endcsname
10:
11: \title{Extension of the Fluctuation-Dissipation theorem
12: to the physical aging of a model glass-forming liquid}
13:
14: %\title{
15: %Physical aging of a model glass-forming liquid:
16: %on the applicability of the fluctuation-dissipation theorem
17: %}
18:
19: \author{Francesco Sciortino and Piero Tartaglia}
20:
21: \address{Dipartimento di Fisica, Universit\'a di Roma
22: "La Sapienza" and Istituto Nazionale per la Fisica della Materia,
23: Piazzale Aldo Moro 2, I-00185, Roma, Italy}
24:
25: \date{Draft: \today}
26: \maketitle
27:
28: \begin{abstract}
29: We present evidence in favor of the possibility of treating
30: an out-of-equilibrium supercooled
31: simple liquid as a system in quasi-equilibrium.
32: Two different temperatures, one controlled by the external bath and one
33: internally selected by the system characterize the quasi-equilibrium state.
34: The value of the
35: internal temperature is explicitly calculated within
36: the inherent structure thermodynamic formalism.
37: We find that
38: the internal temperature controls the relation between the response
39: to an external perturbation and the long-time
40: decay of fluctuations in the liquid.
41: %The temperature calculated using the fluctuation-dissipation theorem
42: %agrees very well with the one calculated theoretically.
43: \end{abstract}
44:
45: In the last decade, several efforts have been devoted to the
46: understanding of the glass-transition
47: phenomenon\cite{debenedetti,review,goetze}, one of the open
48: fundamental problems of condensed matter.
49: %\cite{anderson}.
50: In particular, the thermodynamics of supercooled liquids and the
51: relations between dynamics and thermodynamics have received
52: considerable interest. Novel
53: approaches\cite{parisi,franz1,coniglio,teo,speedy} and detailed analyses of
54: computer generated configurations of supercooled liquids
55: \cite{heuer,thomas,barbara,prlentro,srinew}
56: are supporting and formalizing the picture --- rooted in old
57: ideas\cite{goldstein} --- of a glass as a system trapped in one local
58: free energy basin. Correspondingly, the equilibrium dynamics of deep
59: supercooled liquids is interpreted as motion among different
60: basins.
61:
62: In this Letter, we show
63: that the thermodynamic approach also allows us to interpret
64: the behavior of supercooled model liquids under non-equilibrium
65: (aging) situations. The model liquid we study is the Lennard Jones
66: binary (80:20) mixture (BMLJ) which has been extensively investigated
67: in the past\cite{kob,system}. The formalism we adopt is the so-called
68: inherent structure ($IS$) formalism\cite{stillinger}, which focus on
69: the local minima of the potential energy surface and on the
70: corresponding basins of attraction. In this formalism, each point in
71: the 3$N$-dimensional configuration space is unambiguously associated
72: to the basin of attraction of the closest $IS$. The configuration
73: space is partitioned in a sum of basins (usually labeled according to
74: the value $e_{IS}$ of the potential energy in the $IS$) and the
75: partition function is written as a sum of basin partition functions.
76: In the thermodynamic limit, the system free energy $F(T)$ is written
77: as\cite{stillinger,costantV}
78:
79: \begin{equation}
80: F(T) = - T S_{conf}(e_{IS}(T)) + f_{basin}(T, e_{IS}(T))
81: \label{eq:freee}
82: \end{equation}
83:
84: \noindent
85: where $- T S_{conf}(e_{IS}(T)) $ account for the entropic contribution
86: arising from the number of basins of depth $e_{IS}$ and $f_{basin}(T,
87: e_{IS})$ describes the free energy of the system constrained
88: in one characteristic $e_{IS}$ basin. The separation of the free
89: energy in two parts
90: %--- $-T S_{conf}(e_{IS})$ and
91: %$f_{basin}(T,e_{IS})$ ---
92: reflects the separation of time scales which
93: is observed in the supercooled liquid state. Indeed, the intra-basin
94: dynamics is much faster ($ps$-time scale in real liquids) compared to
95: the inter-basin dynamics (whose time scale diverges upon cooling). We
96: stress that the $IS$ expression for the liquid free energy,
97: Eq.~(\ref{eq:freee}) is analogous to the free-energy, derived using
98: the Thouless-Anderson-Palmer (TAP)\cite{tap} approximation in the
99: $p-$spin models, once a basin is identified with one TAP solution and
100: $f_{basin}(T, e_{IS})$ with the TAP free-energy.
101:
102: In the case of glass-forming liquids, the intrabasin free-energy
103: $f_{basin}$ is usually written as $e_{IS}+ f_{vib}(T, e_{IS})$ to
104: separate the value of the potential energy in the minimum ($e_{IS}$)
105: from the vibrational free-energy contribution. At low $T$,
106: $f_{vib}(T, e_{IS})$ can be calculated in the harmonic approximation,
107: expanding the potential energy around the $IS$ configuration, as
108:
109: \begin{equation}
110: f_{vib}(T, e_{IS}) = k_B T \sum_{i=1}^{3N-3}
111: ln(\hbar \omega_i(e_{IS})/k_B T )
112: \label{eq:hfvib}
113: \end{equation}
114:
115: \noindent
116: where $\omega_i$ is the frequency of the $i$-th normal mode\cite{omexp}.
117: If all basins had the same curvature, then $f_{vib}$ would be only
118: function of $T$. In the BMLJ system, basins with different depth have
119: different curvatures and hence $f_{vib}$ is a function of both $T$ and
120: $e_{IS}$. Fig.~\ref{fig:tap} shows the $T$-dependence of
121: $f_{basin}(T,e_{IS})$ in the
122: harmonic approximation, for different $e_{IS}$
123: values. Curves for different $e_{IS}$ values are not parallel, since
124: the basin curvatures depend on the basin depth, as shown in the inset
125: of the same figure.
126:
127: In equilibrium, at each temperature $T_{eq}$, the system populates
128: basins of different depth $e_{IS}(T_{eq})$\cite{nature}. The
129: $e_{IS}(T_{eq})$ value is fixed by the condition of $F$ being a
130: minimum, i.e. by
131: \begin{equation}
132: {{\partial F} \over { \partial e_{IS} }} =
133: -T_{eq} {{ \partial S_{conf}(e_{IS}) }\over {\partial e_{IS} } } +
134: {{ \partial f_{basin}(T_{eq},e_{IS}) }\over { \partial e_{IS} }}
135: %-T_{eq} {{ \partial S_{conf}(e_{IS}) }\over {\partial e_{IS} } } +
136: %1+ {{ \partial f_{vib} }\over { \partial e_{IS} }}
137: =0
138: \label{eq:teq}
139: \end{equation}
140: where the term ($ {{\partial f_{basin} }\over { \partial e_{IS} }}$)
141: describes the free energy change associated to the change in the basin shape
142: with $e_{IS}$ and the term (${{ \partial S_{conf} }\over
143: {\partial e_{IS} } }$) describes the entropic change associated to the
144: different basin degeneracy. As shown in Ref.\cite{nature}, the onset
145: of slow dynamics correlates with a sharp drop in the $T$-dependence of
146: $e_{IS}$. This observation holds for all model systems studied so
147: far. For the BMLJ system, the $e_{IS}$ basins which are populated at
148: the different temperatures (in the $T$-range accessible to equilibrium
149: simulations) are indicated in Fig.~\ref{fig:tap}. As indicated by
150: Eq.~(\ref{eq:teq}), the $T$-dependence of $e_{IS}$ arises from a
151: balance between the change of basin free energy and the change in
152: configurational entropy. Of course, the monotonic relation
153: $e_{IS}(T_{eq})$ can be inverted to give the temperature at which the
154: equilibrium system populates basins of depth $e_{IS}$,
155: $T_{eq}(e_{IS})$\cite{epl}.
156:
157: The analysis of the non-equilibrium dynamics presented in this Letter
158: is based on the assumption that the separation of intrabasin
159: and interbasin time scales--- which characterize equilibrium
160: supercooled liquid states --- retains its validity during aging. In
161: particular, we assume that after a temperature jump from $T_i$ to
162: $T_f$, the vibrational intrabasin dynamics thermalizes quickly to the
163: thermostat value $T_f$. The thermalization of the entire system is
164: instead very slow, requiring a search for the low energy minima. Such
165: a search in configuration space, which is the essence of the aging
166: phenomenon in liquids, is so slow that quasi-equilibration on basins
167: of depth $e_{IS}$ might be faster than the decrease of the $e_{IS}$
168: value.
169:
170: In the case of BMLJ, the out-of-equilibrium dynamics following a
171: $T$-jump have been recently studied\cite{kobbarrat,epl}.
172: Ref.\cite{epl} indeed suggested the possibility that the equilibration
173: in configuration space proceeds via quasi-equilibrium steps. The
174: proposed equilibration process is schematically depitched with arrows
175: in Fig.~\ref{fig:tap}, for the case of a jump from $T_i=0.8$ to
176: $T_f=0.25$. In a time much shorter than the any basin change,
177: $f_{basin}$ assumes the value characteristic of the final temperature
178: (full line arrow in Fig.~\ref{fig:tap}). The fast equilibration of the
179: intra-basin degrees of freedom is then followed by a much slower
180: process (dashed arrow in Fig.~\ref{fig:tap}) during which the system
181: populates deeper and deeper $e_{IS}$ levels. For the BMLJ case,
182: $e_{IS}(t)$ after a $T$-jump
183: %from $T_i=0.8$ to $T_f=0.25$
184: is reported in Fig.~\ref{fig:teff}.
185:
186: If the hypothesis of quasi-equilibrium is correct, then we can ask
187: which is the value of the internal temperature $T_{int}(e_{IS},T_f)$
188: selected by the system when it is populating basins of
189: depth $e_{IS}$\cite{temp}.
190: To calculate $T_{int}(e_{IS},T_{f})$, we again search for solutions of
191: Eq.~(\ref{eq:teq}) but, in contrast to the equilibrium case, we
192: consider the value $e_{IS}$ to be fixed and solve for the unknown
193: temperature, obtaining
194:
195: %\begin{equation}
196: %1+ {{ \partial f_{vib}(T_f,e_{IS}) }\over { \partial e_{IS} }}-
197: %T_{int}(e_{IS},T_{f})
198: %{ {\partial S_{conf} } \over {\partial e_{IS}} }=0
199: %\end{equation}
200: %\noindent
201: %or
202:
203:
204: \begin{equation}
205: T_{int}(e_{IS},T_{f}) = { {1+ {{ \partial } \over { \partial e_{IS} }} f_{vib} (T_f,e_{IS}) }\over{
206: { {\partial } \over {\partial e_{IS}} } S_{conf}(e_{IS}) }}
207: \label{eq:teff}
208: \end{equation}
209:
210:
211: \noindent
212: Note that, differently from Eq.~(\ref{eq:teq}), $f_{vib}$ is now
213: evaluated at the thermostat temperature $T_f$, since the fast
214: intrabasin degrees of freedom are already thermalized to $T_f$.
215: This expression for $T_{int}$ coincides with the expression proposed
216: by Franz and Virasoro\cite{silvio} in the context of $p-$spin systems,
217: once the basin free energy is identified with the TAP free energy. We
218: note that $ {{\partial S_{conf}} \over {\partial e_{IS} } }$ can be
219: evaluated from equilibrium conditions Eq.~(\ref{eq:teq}), and thus,
220: once a model for $f_{vib}$ is chosen, $T_{int}(e_{IS},T_{f})$ can be
221: calculated. If $T_f$ is small (as it usually is), the harmonic
222: approximation for $f_{vib}$ can be confidently used. In this case,
223: from Eq.~(\ref{eq:teff}) and Eq.~(\ref{eq:teq}) we obtain
224:
225: \begin{equation}
226: T_{int}(e_{IS},T_{f}) =
227: {{1+k_B T_{f}
228: \sum_i {{ \partial ln [ \hbar \omega_i / k_B T_f] } \over { \partial
229: e_{_{IS}} }}
230: }\over{ 1+k_B T_{eq}(e_{IS})
231: \sum_i {{ \partial ln [\hbar \omega_i/k_B T_e ] } \over { \partial
232: e_{IS} }}
233: }} T_{eq}(e_{IS})
234: \label{eq:teff2}
235: \end{equation}
236:
237:
238: \noindent
239: Fig.~\ref{fig:teff}-left shows $T_{int}(e_{IS},T_{f})$ for the BMLJ
240: system. We note that, if basin curvatures were independent on the
241: $e_{IS}$ value, then the derivatives in Eq.~(\ref{eq:teff2}) would be
242: zero. Hence, $T_{int}(e_{IS},T_{f})$ would not depend of $T_{f}$ and
243: $T_{int}$ would coincide with $T_{eq}(e_{IS})$. In this limit,
244: $T_{int}$ plays the same role as the fictive temperature introduced in
245: the analysis of experimental data in aging systems\cite{hodge}.
246: We note in passing that an important by-product of the present
247: approach is a free-energy expression for an out-of-equilibrium liquid,
248: that depends only on $e_{IS}$ and $T_{f}$. Such expression offers a
249: detailed interpretation of the free energy expression for glassy
250: systems proposed in recent years\cite{teo}.
251:
252: %\begin{equation}
253: %F_{ne}(e_{IS},T_{vib}) = - T_{int}(e_{IS},T_{f}) S_{conf}(e_{IS}) + e_{IS}+ f_{vib}(T_f, e_{IS})
254: %\label{eq:freeene}
255: %\end{equation}
256: %Such expression may contribute to clarify the issue of
257: %phase separation in amorphous systems, a topic of current debate\cite{water}.
258:
259: To test the predictions of Eq.~(\ref{eq:teff}-\ref{eq:teff2}) ---
260: i.e. the assumption of quasi-equilibrium in the aging system --- we
261: study the response of the liquid, described by an Hamiltonian $H_0$,
262: to an external perturbation switched on at $t=t_w$. If the
263: perturbation adds the term $H_P=-V_o B({\bf r^N}) \theta(t-t_w) $ to
264: the Hamiltonian (where $\theta(t)$ is the Heaviside step function),
265: linear response theory predicts that the time evolution of any
266: variable $A({\bf r^N})$ conjugated to $B$ is given by\cite{hansen}
267:
268: \begin{equation}
269: \langle A(\tau) \rangle = - {{V_o}\over{k_B T}} [
270: \langle A(\tau)B(0) \rangle_0 - \langle A(0)B(0) \rangle]_0]
271: \end{equation}
272: \noindent
273: where $\tau \equiv t-t_w$, $\langle \cdot \cdot \cdot \rangle$ is the
274: ensemble average over the perturbed system ($H_0+H_P$) and $\langle
275: \cdot \cdot \cdot \rangle_0$ is the ensemble average over the
276: unperturbed system ($H_0$). We chose $ B \equiv (\rho^{\alpha}_{\bf
277: k}+\rho^{\alpha*}_{\bf k})$, where $\rho^{\alpha}_{\bf k} \equiv
278: \sum_i^{N_{\alpha}} e^{i {\bf k \cdot r}_i^{\alpha} }/\sqrt(N) $ is
279: the Fourier transform component of the density of $\alpha$ particles
280: at wavevector ${\bf k}$, and study the response of $A \equiv
281: \rho^{\alpha}_{\bf k}$. In this case, $\langle A(t) B(0) \rangle_0$
282: coincides with the dynamical structure factor $S_{\bf
283: k}^{\alpha\alpha}(t) \equiv \langle \rho^{\alpha}_{\bf k}(t)
284: \rho^{\alpha*}_{\bf k}(0) \rangle_0 $. Thus, with the present choice of
285: $A$ and $B$, linear response theory predicts \cite{hansen}
286: \begin{equation}
287: \langle \rho^{\alpha}_{\bf k}(\tau)\rangle = - {{V_o}\over{k_B T}} [ S_{\bf k}^{\alpha\alpha}(\tau)-S_{\bf k}^{\alpha\alpha}(0)]
288: \label{eq:fd}
289: \end{equation}
290: \noindent
291: Eq.~(\ref{eq:fd}) --- also referred as fluctuation-dissipation
292: theorem --- is particularly relevant for our purposes, since it
293: predicts that the response of the system is proportional to $T^{-1}$
294: and thus offers an {\it independent} way to confirm the validity of
295: the quasi-equilibrium hypothesis. Indeed, the quasi-equilibrium
296: hypothesis predicts that for short times (i.e. in the time region
297: where the correlation function assumes values between the value in
298: zero and the plateau value) the relation between correlation and
299: response, Eq.~(\ref{eq:fd}), should be controlled by $V_o/k_BT_f$, since
300: the intrabasin dynamics is probed. Similarly, for long times (i.e. in
301: the time region where the correlation function assumes values smaller
302: than the plateau value) the relation should be controlled by
303: $V_o/k_BT_{int}$, since the inter-basin dynamics is now probed. Thus,
304: we predict that switching on the perturbing field when the system is
305: populating basins of depth $e_{IS}$, the calculated $T_{int}$ should
306: coincide with the temperature at which the system responds to the
307: external perturbation for times longer than the vibrational dynamics.
308:
309: Fig.~\ref{fig:cr}-left shows $ S_{\bf k}^{\alpha\alpha}(\tau)$ and
310: $\rho_{\bf k}^{\alpha}(\tau)$ for two different $t_w$, for the BMLJ
311: case, with $T_i=0.8$ and $T_f=0.25$.
312: The reported data are averaged over 300 different
313: quench-realizations and over more than 60 different
314: independent perturbations $H_{P}$
315: (but with the same $k$ modulus) for each
316: quench-realization\cite{infosim}. Both $ S_{\bf
317: k}^{\alpha\alpha}(\tau)$ and $\rho_{\bf k}^{\alpha}(\tau)$ show the
318: two-step relaxation characteristic of the supercooled state, which has
319: been associated to the separation of intra-basin and inter-basin
320: motion. Fig.~\ref{fig:cr}-right shows the corresponding response
321: vs. correlation plots, Eq.~(\ref{eq:fd}). At short time (intra-basin
322: motion) $\rho_{\bf k}(t)$ vs. $ S_{\bf k}(t)$ is linear with the
323: expected $T_f^{-1}$ slope, properly describing the equilibrium
324: condition of the vibrational dynamics with the external reservoir. At
325: larger times, the intra basin motion sets-in and indeed the slope of
326: $\rho_{\bf k}^{\alpha}(\tau)$ vs. $ S_{\bf k}^{\alpha\alpha}(\tau)$
327: becomes controlled by $T_{int}^{-1}$. As shown in Fig.~\ref{fig:cr},
328: the slope of the response vs. correlation plots are extremely well
329: predicted by Eq.~(\ref{eq:teff2}).
330:
331: The good agreement between the internal temperature selected by the
332: system, as measured by the amplitude of the response of the aging
333: system to the external perturbation, and the temperature predicted
334: theoretically using the $IS$ formalism supports the main hypothesis on
335: which our analysis is based, i.e. the validity of the
336: quasi-equilibrium condition. The quality of the agreement is also a
337: positive test on the validity of the $IS$ formalism and of the
338: proposed free energy expression. This Letter suggests that an aging
339: liquid, notwithstanding its out-of-equilibrium condition, can still be
340: described as a system in quasi-equilibrium, at the expenses of
341: introducing an internal temperature which is a function both of the
342: thermostat temperature and of the ($t$-dependent) state of the system
343: --- expressed by its $e_{IS}$ value.
344: Eq.~(\ref{eq:teff}) more precisely defines the concept of fictive
345: temperature, usually defined by the experimentalist as the temperature
346: at which the ergodicity of the system was broken\cite{hodge}.
347: Eq.~(\ref{eq:teff}) clarifies that $T_{int}$ is function also of $T_f$.
348: We stress that the detailed analysis presented in this Letter for a
349: structural glass is conceptually identical to the analysis performed
350: in recent years to describe the out of equilibrium dynamics of
351: disordered spin models\cite{violation}, even though we prefer to
352: present our results in terms of extended validity (as opposed to
353: violation) of the fluctuation dissipation theorem. Tests of the
354: mean-field theory on finite-size disordered $p$-spin system also
355: support the validity of the thermodynamic approach\cite{crisantinew}.
356: As noted for the case of disordered spin-systems\cite{silvio}, the
357: measurement of the internal temperature in aging
358: experiments\cite{grigera} may provide information, once properly
359: interpreted, on $\partial S_{conf}/ \partial e_{IS}$ i.e. on
360: structural properties of the system which control the equilibrium
361: dynamics as well. Finally, we call attention on the fact that the
362: time window accessed by the numerical experiments is very different
363: from the experimental one. Measurements of the internal
364: temperature in aging experiments are very important to assess the
365: range of validity of the presented approach.
366:
367: FS gratefully acknowledges very stimulating discussions with S. Franz. We also
368: thank A. Crisanti, S. Ciuchi, W. Kob, G. Parisi, G. Ruocco and F. Starr.
369:
370: \begin{references}
371: \bibitem{debenedetti} P.~G. Debenedetti, { Metastable Liquids}
372: (Princeton Univ. Press, Princeton, 1997).
373:
374: \bibitem{review} For recent reviews see articles in {\em Science} {\bf 267},
375: (1995) and M.D. Ediger, C.A. Angell and
376: S.R. Nagel, { J. Phys. Chem} {\bf 100}, 13200 (1996).
377:
378: \bibitem{goetze} W. G\"otze, { J. Phys.: Condens. Matter} {\bf 11}, A1 (1999).
379:
380: %\bibitem{anderson} anderson
381:
382: \bibitem{parisi} M. M\'ezard and G. Parisi,
383: { J. Phys. Cond. Matter} {\bf 11}, A157 (1999)
384: M. Cardenas, S. Franz and G. Parisi, { J. Chem. Phys.} {\bf 110}, 1726 (1999).
385:
386: %\bibitem{thirumalai} T. R. Kirkpatrick and D. Thirumalai,
387: %Phys. Rev. Lett. {\bf 58}, 2091 (1987); A. Crisanti and
388: %H.-J. Sommers, Z. für Physik B {\bf 87}, 341 (1992).
389:
390: \bibitem{franz1} S. Franz and G. Parisi, Phys. Rev. Lett. {\bf 79}, 2486 (1997).
391: \bibitem{coniglio} A.Coniglio, A. de Candia, A. Fierro and M. Nicodemi
392: { J. Phys.: Condens. Matter} {\bf 11}, A167 (1999).
393:
394: \bibitem{teo} Th. M. Nieuwenhuizen,
395: { Phys. Rev. Lett.} {\bf 80}, 5580 (1998); {\it ibid.} {\bf 79},
396: 1317 (1997).
397:
398: \bibitem{speedy} R. Speedy,
399: { J. Phys.: Conds. Matter } {\bf 10}, 4185 (1998); R. Speedy,
400: J. Phys. Chem. B {\bf 103}, 4060 (1999).
401:
402: \bibitem{heuer} A. Heuer, { Phys. Rev. Lett.} {\bf 78}, 4051 (1997);
403: S. Buechner and A. Heuer,
404: { Phys. Rev. E.} {\bf 60}, 6507 (1999).
405:
406: \bibitem{thomas}
407: T. Schr\"oder, S. Sastry, J. Dyre and S. Glotzer, J. Chem. Phys. {112},9834 (2000);
408:
409: \bibitem{barbara} B. Coluzzi, Ph.D Thesis, University of Roma La Sapienza (1999). B. Coluzzi, P. Verrocchio and G. Parisi, Phys. Rev. Lett.
410: {\bf 84}, 306 (2000)
411:
412:
413: \bibitem{prlentro} F. Sciortino, W. Kob and P. Tartaglia,
414: Phys. Rev. Lett. {\bf 83}, 3214 (1999).
415:
416: \bibitem{srinew} S. Sastry, Phys. Rev. Letts. (in press) (2000).
417:
418:
419: %\bibitem{andreacris} A. Crisanti and F. Ritort, Europhysics Lett. (in press)
420:
421: %\bibitem{angelani}
422: %L. Angelani, G. Parisi, G. Ruocco, G. Viliani.
423: %{\em Phys. Rev. Lett.} {\bf 81}, 4648 (1998).
424:
425:
426:
427: \bibitem{goldstein} M. Goldstein, { J. Chem. Phys.} {\bf 51}, 3728 (1969).
428:
429: \bibitem{kob} W. Kob and H.C. Andersen, { Phys. Rev. Lett.} {\bf 73},
430: 1376, (1994); { Phys. Rev. E} {\bf 51}, 4626 (1995);
431: { Phys. Rev.} E {\bf 52}, 4134 (1995).
432: M. Nauroth and W. Kob, { ibid.} {\bf55}, 675 (1997);
433: T. Gleim, W. Kob, and K. Binder, {Phys. Rev. Lett.} {\bf 81},
434: 4404 (1998).
435:
436: \bibitem{system}
437: The system is composed by $N=1000$ particles of a Lennard Jones (LJ)
438: binary (80:20) mixture with interaction parameters chosen so that the
439: liquid does not crystallize and does not demix.
440: The LJ parameters ($\epsilon,\sigma$) of the majority component are
441: the units of energy and length used in this Letter.
442: This model has been
443: extensively studied in the past\protect\cite{kob} since it does not crystallize. It has been shown that
444: the atomic dynamics is well described by mode coupling
445: theory\protect\cite{goetze}, with a
446: critical temperature $T_c$ equal to $0.435$\protect\cite{kob}.
447:
448: %
449: %\bibitem{speedy} R. Speedy,
450: % { J. Phys.: Conds. Matter } {\bf 10}, 4185 (1998);
451: % {\it ibid.} {\bf 9}, 8591 (1997);{\it ibid.} {\bf 8}, 10907 (1996);
452: %J. Phys. Chem. B {\bf 103} 4060 (1999).
453:
454: \bibitem{stillinger}
455: F.H. Stillinger and T.A. Weber, { Phys. Rev. A} {\bf 25}, 978 (1982);
456: { Science} {\bf 225}, 983 (1984).
457: F. H. Stillinger, { Science}, {\bf 267}, 1935 (1995).
458:
459: \bibitem{costantV} The adopted $IS$ formalism is
460: for the NVT ensemble. Generalization to NPT does not pose particular
461: problems. The description we present in this Letter
462: refers to a constant $V$. Thus, it is based on
463: one internal parameter (in the language of Davies and Jones
464: [R.O. Davies and G.O. Jones, { Adv. in Physics} {\bf 2}, 370 (1953).]),
465: which we identify with $e_{IS}$.
466:
467: \bibitem{tap} D.J. Thouless, P.W. Anderson and R.G. Palmer,
468: { Phil. Mag.}, {\bf 35} 593 (1977).
469: A. Crisanti, H. Horner and H.J. Sommers, { Z. Phys.} B {\bf 92}, 257 (1993).
470:
471: %\bibitem{pablo} F.H. Stillinger and P.G. Debenedetti,
472: % { J. Phys. Chem B} .{\bf 103}, 4052 (1999).
473:
474: \bibitem{omexp} $\omega_i$ can be calculated by diagonalizing the Hessian matrix
475: evaluated at the $IS$ configuration (a local minimum of the potential energy).
476:
477: \bibitem{nature}S. Sastry, P. G. Debenedetti, and F. H. Stillinger,
478: {\it Nature} {\bf 393}, 554 (1998).
479: See also H. Jonsson and
480: H.C. Andersen, {Phys. Rev. Lett.} {\bf 60}, 2295 (1988).
481:
482: \bibitem{epl} W. Kob, F. Sciortino, and P. Tartaglia,
483: {Europhys. Lett.} {\bf 49}, 590 (2000).
484:
485: \bibitem{kobbarrat} J.L. Barrat and W. Kob,
486: {\em Europhys. Lett.} {\bf 49}, 590 (2000);
487: W. Kob and J.L. Barrat,{ Europhys. J.} B (in press).
488:
489:
490: \bibitem{temp} L. Cugliandolo, K. Kurchan, L. Peliti, Phys. Rev. E,
491: {\bf 55}, 3898 (1997).
492:
493: \bibitem{silvio} A detailed interpretation of this relation is given in
494: S. Franz, M. A. Virasoro, { J. Phys.} A {\bf 33}, 891 (2000).
495:
496: \bibitem{hodge} For a review see for example I, Hodge, { J. Non-Cryst. Solids}
497: {\bf 131}, 435 (1991); {\bf 169}, 211 (1994).
498:
499: \bibitem{hansen} J.P. Hansen and I.R. McDonald, {\em Theory of Simple
500: Liquids} (Academic Press, London, 1986), 2nd Edition.
501:
502: \bibitem{infosim} We have equilibrated 300 independent configurations
503: at $T=0.8$ in the $NVT$ ensemble (Nos\'e-Hoover thermostat). Each of
504: the configuration has been quenched to $T_f=0.25$ by changing at $t=0$
505: the thermostat temperature to $T_f$. The thermostat constant is chosen
506: in such a way that, within 1000 MD steps, the average kinetic energy
507: thermalizes to $T_f$.
508: Configurations after $t_w=1024$ and $t_w=16386$ MD steps have been
509: saved. Each of the saved configurations has been used as starting
510: configuration for 60 independent simulations (each of them lasting $10^5$
511: MD steps) with
512: Hamiltonian $H+H_P$ ($V_o=0.2)$, each of them with a different ${\bf k}$
513: vector. $|\bf{k}| $ has been constrained to the value 6.7, the
514: location of the first minimum of $S_{\bf k}^{\alpha\alpha}$.
515: The total number of MD step requested in the reported calculations is thus
516: the total number of configurations (300) times the
517: number of different perturbations (60), times the number of MD step of each of them ($10^5$), times number of studied $t_w$ (2). The calculation has
518: requested
519: a cluster of 12 alpha processors running full time for about 6 months.
520:
521:
522: \bibitem{violation} L. F. Cugliandolo and J. Kurchan,
523: {Phys. Rev. Lett.} {\bf 71}, 173 (1993).
524:
525: \bibitem{crisantinew} A. Crisanti and F. Ritort, cond-mat/9911226.
526: see also A. Crisanti and F. Ritort, Europhysics Lett. (in press)
527:
528: \bibitem{grigera} T. S. Grigera and N. E. Israeloff,
529: {Phys. Rev. Lett.} {\bf 83}, 5038 (1999).
530:
531: %\bibitem{andreacris} A. Crisanti and F. Ritort, Europhysics Lett. (in press)
532: %\bibitem{prlentro} F. Sciortino, W. Kob and P. Tartaglia,
533: %Phys. Rev. Lett. {\bf 83} 3214 (1999).
534: %\bibitem{angelani}
535: %L. Angelani, G. Parisi, G. Ruocco, G. Viliani.
536: %{\em Phys. Rev. Lett.} {\bf 81}, 4648 (1998).
537: %\bibitem{barbara} B. Coluzzi, P. Verrochio and G. Parisi, Phys. Rev. Lett.
538: %{\bf 84}, 306 (2000).
539:
540: \end{references}
541:
542: \begin{figure}
543: \centerline{\psfig{figure=fig1.eps,width=8.0cm,angle=0}}
544: \caption{
545: Basin free energy for the BMLJ system in harmonic approximation.
546: Each of the dashed line shows
547: $f_{basin}(e_{IS},T)$ for one specific $e_{IS}$ value
548: (the corresponding $e_{IS}$ value coincides with
549: the $T=0$ value of $f_{basin}$). In all curves,
550: the $e_{IS}$-independent contribution $(3N-3)k_BTln(k_BT)$ (see
551: Eq.\protect\ref{eq:hfvib})
552: has not been included for clarity reasons. Filled
553: circles are the equilibrium values $f_{basin}(e_{IS}(T_{eq}),T_{eq})$ .
554: The arrows indicate the path followed by the system after a
555: quench from $T_i=0.8$ to $T_i=0.25$
556: The inset shows the
557: $e_{IS}$-dependence of $\sum_{i=1}^{3N-3} ln(\hbar \omega_i)$. The value
558: of $\hbar$ is such that the dimension of $\hbar \omega$ (as well as the dimension of $k_BT$)
559: are in units of the LJ potential depth\protect\cite{system}.
560: }
561: \label{fig:tap}
562: \end{figure}
563:
564:
565:
566: \begin{figure}
567: \centerline{\psfig{figure=fig2.eps,width=8.0cm,angle=0}}
568: \caption{Left: Solutions of Eq.~(\ref{eq:teff}) for several $T_f$
569: values for the studied BMLJ system.
570: Right: $e_{IS}$ as a function of time, following a $T$-jump from
571: $T_i=0.8$ to $T_f=0.25$. The arrows show graphically the procedure which connects the $e_{IS}(t)$ value to the $T_{int}$ value, once $T_{f}$ is known.
572: Note that if the curvature of the basins were independent on $e_{IS}$, then
573: curves for different $T_f$ would all coincide with $T_{eq}(e_{IS})$
574: (filled symbols).
575: }
576: \label{fig:teff}
577: \end{figure}
578:
579:
580: \begin{figure}
581: \centerline{\psfig{figure=fig3.eps,width=8.0cm,angle=0}}
582: \caption{Left: time dependence of the
583: responce ($\rho^{\alpha}_{k}$, open symbols)
584: and correlation function ($S^{\alpha\alpha}_{k}$, filled symbols)
585: for $t_w=1024$ (circles) and $t_w=16384$ (squares).
586: Right: parametric plot (in $\tau$) of
587: $\rho^{\alpha}_{k}$ vs $S^{\alpha\alpha}_{k}$ for the two studied $t_w$.
588: Dashed lines have slope $V_o/k_BT_f$, thick lines have slope
589: $Vo/k_BT_{int}$. The two $T_{int}$ values can be read from Fig.\protect\ref{fig:teff}. The modulus of $k$ is $6.7$\protect\cite{infosim}.
590: }
591: \label{fig:cr}
592: \end{figure}
593:
594:
595: \end{document}
596:
597: \setcounter{figure}{0}
598: \eject
599:
600: \begin{figure}
601: \centerline{\psfig{figure=fig1.eps,width=16.0cm,angle=0}}
602: \caption{F. Sciortino and P. Tartaglia }
603: \end{figure}
604:
605: \eject
606:
607: \begin{figure}
608: \centerline{\psfig{figure=fig2.eps,width=16.0cm,angle=0}}
609: \caption{F. Sciortino and P. Tartaglia }
610: \end{figure}
611:
612: \eject
613:
614: \begin{figure}
615: \centerline{\psfig{figure=fig3.eps,width=16.0cm,angle=0}}
616: \caption{F. Sciortino and P. Tartaglia }
617: \end{figure}
618:
619: \end{document}
620:
621:
622:
623:
624:
625:
626:
627: