1: %\documentclass[aps,pre,amsmath,floatfix,showpacs,preprint]{revtex4}
2: \documentclass[twocolumn,floatfix,aps,pre,amsmath,showpacs]{revtex4}
3:
4: \usepackage{graphicx}
5:
6: \begin{document}
7: \title{Democratic particle motion for meta-basin transitions in simple glass-formers}
8: \author{ G. A. Appignanesi$^\dag$, J. A. Rodr\'{\i}guez Fris$^\dag$, R. A. Montani$^\dag$
9: and W. Kob$^\ddag$}
10: \affiliation{
11: $^\dag$ Laboratorio de Fisicoqu\'{\i}mica, Departamento de Qu\'{\i}mica, Universidad Nacional del Sur, Av. Alem 1253,
12: 8000 Bah\'{\i}a Blanca, Argentina.\\
13: $^\ddag$ Laboratoire des Collo\"ides, Verres et Nanomat\'eriaux, Universit\'e Montpellier 2,
14: 34095 Montpellier, France.\\
15: }
16: \date{\today}
17:
18: \begin{abstract}
19:
20: We use molecular dynamics computer simulations to investigate the local
21: motion of the particles in a supercooled simple liquid. Using the concept
22: of the distance matrix we find that the $\alpha-$relaxation corresponds
23: to a small number of crossings from one meta-basin to a neighboring
24: one. Each crossing is very rapid and involves the collective motion of
25: $O(40)$ particles that form a relatively compact cluster, whereas
26: string-like motions seem not to be relevant for these transitions. These
27: compact clusters are thus candidates for the cooperatively rearranging
28: regions proposed long times ago by Adam and Gibbs.
29:
30: \end{abstract}
31:
32: \pacs{61.20.Ja; 61.20c.Lc; 64.70.Pf}
33:
34: \maketitle
35:
36: In recent years significant progress has been made in our understanding
37: of the relaxation dynamics of glass-forming liquids at intermediate and
38: low temperatures. Sophisticated experiments and computer simulations
39: have identified many of the salient features of this dynamics, and
40: theoretical approaches have helped to rationalize them, at least to
41: some extend~\cite{ediger_96,gotze_99,debenedetti_01,ngai_02}. Despite
42: this progress, many of the most elementary questions have
43: not been answered so far and among them is the nature of the
44: motion of the particles in the $\alpha-$relaxation regime at
45: low temperatures. Experiments and simulations have demonstrated
46: that this dynamics is quite heterogeneous and therefore can
47: be used to explain the observed stretching of the time correlation
48: functions~\cite{schmidt_91,butler_91,cicerone_95,kob_97,kegel_00,weeks_00,richert_02}.
49: This heterogeneous dynamics has been shown to be related to cooperative
50: motion in which a small number of particles (a few percent)
51: undergo a collective relaxation dynamics in that they move, often
52: in a string-like fashion, by a distance that is comparable to the
53: one between neighboring particles~\cite{donati_98,kegel_00,weeks_00}.
54: Since a qualitatively similar heterogeneous dynamics has also been found
55: in simple lattice models that show a glassy dynamics and for which it
56: is well known that the $\alpha-$relaxation is intimately connected
57: to the dynamical heterogeneities (DH), it has been proposed that
58: these DH are crucial for the relaxation dynamics of all glass-forming
59: systems~\cite{garrahan_02}. However, since in these lattice models all
60: elastic or quasi-elastic effects are completely neglected, it is not at
61: all clear whether or not the DH are indeed the only relevant mechanism
62: for the relaxation.
63:
64: Another approach to describe the relaxation dynamics on the time scale
65: of the $\alpha-$relaxation is by means of the so-called potential energy
66: landscape (PEL)~\cite{angell_91,debenedetti_01b,monthus_96} (or more
67: precisely the free energy landscape) and hence to describe the dynamics
68: of the system by considering its trajectory in configuration space. This
69: PEL is rugged due to the presence of barriers in the free energy and
70: hence at low temperatures the resulting dynamics will be slow. Using the
71: concept of the inherent structures, evidence has been given that (roughly
72: speaking) the motion of the system in the PEL can be decomposed into two
73: types of movements~\cite{doliwa_03,vogel_04}: In the first type the system
74: explores some minima which are locally connected to each other and are not
75: separated by a significant barrier. Therefore such a collection of minima
76: is called ``meta-basin'' (MB) and its exploration corresponds to the
77: $\beta-$relaxation. As a second type of motion, the system overcomes the
78: barriers that surround a MB and enters a new MB, a motion that has been
79: believed to correspond to the $\alpha-$relaxation~\cite{debenedetti_01}
80: or to the elementary events of the $\alpha-$relaxation~\cite{vogel_04}.
81:
82: Although this picture for the motion of the system within the PEL is
83: certainly appealing from a qualitative point of view, it does not
84: give any insight on the nature of the dynamics of the particles on
85: the microscopic level. Furthermore it remains unclear whether the DH
86: mentioned above have anything to do with the barrier-crossing of the
87: system moving in the PEL. The goal of the present work is therefore to
88: clarify this issue and thus to make an advancement in our understanding
89: of the relaxation dynamics of supercooled liquids. To this aim we have
90: done molecular dynamics computer simulations of a simple glass-former
91: in order to identify the presence of the MB and to investigate the
92: nature of the motion of the particles during the transition from one
93: MB to another. The so obtained results can then be compared with the
94: heterogeneous motion of the particles in order to see to what extend
95: the two motions are related to each other.
96:
97: The system considered is a binary mixture of Lennard-Jones (LJ) particles.
98: In previous investigations it has been shown that this system shows
99: many features of glass-forming liquids and can thus serve as a simple
100: model for such liquids~\cite{kob_95}. The interaction between two atoms
101: of type $A$ (80\%) and $B$ (20\%) is given by $V_{\alpha \beta}(r)
102: = 4 \epsilon_{\alpha \beta} \{ (\sigma_{\alpha \beta}/r)^{12} -
103: (\sigma_{\alpha \beta}/r)^{6} \}$, where $\alpha, \beta \in \{A,B\}$. The
104: LJ parameters used are $\epsilon_{AA} = 1.0$, $\sigma_{AA} = 1.0$,
105: $\epsilon_{AB} = 1.5$, $\sigma_{AB} = 0.8$, $\epsilon_{BB} = 0.5$,
106: and $\sigma_{BB} = 0.88$. These interactions have been truncated and
107: shifted at $r_{\rm cutoff}=2.5\sigma_{\alpha\beta}$. In the following
108: we will use $\sigma_{AA}$ and $\epsilon_{AA}$ as units of length and
109: energy, respectively, and measure time in units of $(m \sigma_{AA}^{2}
110: / 48 \epsilon_{AA})^{1/2}$. The equations of motion were solved for the
111: NVE ensemble at a particle density of 1.2, using the velocity form of
112: the Verlet algorithm with a time step of 0.02. All the presented results
113: correspond to the situation in equilibrium.
114:
115: In order to identify the MBs we define the following ``distance matrix''
116: (DM)~\cite{ohmine_95}:
117: \vspace*{-4mm}
118:
119: \begin{equation}
120: \Delta^2(t',t'') = \frac{1}{N}\sum_{i=1}^N |{\bf r}_i(t')-{\bf r}_i(t'')|^2
121: \quad ,
122: \label{eq1}
123: \end{equation}
124:
125: \noindent
126: where ${\bf r}_i(t)$ is the position of particle $i$ at time $t$. Thus
127: $\Delta^2(t',t'')$ gives the system averaged squared displacement
128: of a particle in the time interval that starts at $t'$ and ends at
129: $t''$. Note that the time average of $\Delta^2(t',t'+\theta)$
130: over $t'$ gives the $r-$average of $G_s(r,\theta)$, the self-part of
131: the van Hove correlation function for time displacement $\theta$. The
132: same is true if one averages over a very large system. Since we are
133: interested in {\it individual} MB-MB transitions, one has to avoid that
134: the presence of several independent local rearrangements, that will
135: occur in a large system almost simultaneously, obscures the analysis of
136: the individual event. Therefore we have considered a rather small system
137: of 150 particles, the smallest possible system that does not affect the
138: interactions, i.e. the box size was two times $r_{\rm cutoff}$ for the
139: $A-A$ interaction~\cite{footnote1}.
140:
141: \begin{figure}[tb]
142: \includegraphics[width=0.9\linewidth]{fig1.eps}
143: \caption{
144: Distance matrix $\Delta^2(t',t'')$ of the system for $T=0.50$.
145: The gray level correspond to values of $\Delta^2(t',t'')$ that are
146: given to the right of the figure.}
147: \vspace*{-3mm}
148:
149: \label{fig1}
150: \end{figure}
151:
152: \begin{figure}[tb]
153: \vspace*{4mm}
154:
155: \includegraphics[width=1.0\linewidth]{fig2.eps}
156: \caption{
157: Solid line (right scale): Averaged squared displacement $\delta^2(t,\theta)$ for
158: the trajectory given in Fig.~\ref{fig1}. The value of $\theta$ is 160.
159: Vertical bars (left scale): The function $m(t,\theta)$ which gives the fraction of
160: mobile particles, i.e. particles that moved more than the threshold value
161: $r_{\rm th}= 0.3$ in the time interval $[t,t+\theta]$, using $\theta=40$.}
162: \label{fig2}
163: \end{figure}
164:
165: Figure~\ref{fig1} shows a typical graph of the DM at $T=0.50$ as
166: a function of the two time arguments $t'$ and $t''$, with darker areas
167: corresponding to configurations that have a smaller distance. At this
168: temperature the dynamics is already slow with an $\alpha-$relaxation time
169: $\tau$ of the order of $4\cdot 10^3$~\cite{kob_95}. ($\tau$ can, e.g. be defined
170: by requiring that the self intermediate scattering function $F_s(q,t)$ has
171: decayed to 10\% of its initial value.) From this figure we see immediately
172: that the dynamics of the system is quite heterogeneous {\it in time} in
173: that it stays for a significant time relatively close to one region in
174: configuration space, dark square-like regions, before it finds a pathway
175: to a new region. Thus this is clear evidence that the system explores the
176: present MB before it moves on to a neighboring one. At this temperature
177: the typical sojourn time within one MB is around 300-800 time units,
178: which is around 10\% of $\tau$. Thus this sojourn time corresponds to the
179: time scale of $t^*$, the time that previous investigations have shown to
180: be relevant for the dynamical heterogeneities in the system and which is
181: defined as the time at which one observes the maximum in the non-gaussian
182: parameter $\alpha_2(t)$~\cite{kob_97,appignanesi_04}, and which at this
183: temperature is $400$ time units~\cite{kob_95}. Hence we can conclude
184: that the MB-MB transitions are relevant for the DH whereas it takes about
185: 5-10 such transitions in order to make an $\alpha-$relaxation. We also
186: point out that from Fig.~\ref{fig1} it becomes evident that the time
187: for a MB-MB transition is quite short, on the order of 100~time units,
188: which thus corresponds to about 20\% of $t^*$. Thus the transition is
189: significantly faster than the $\alpha-$relaxation times $\tau$.
190:
191: \begin{figure}[tb]
192: \includegraphics[width=0.9\linewidth]{fig3.eps}
193: \caption{
194: The distribution function $\widehat{G}_s(r,t,t+\theta)$ for different
195: values of $t$ (curves with symbols). The value of $\theta$ is 40. The
196: bold curve is $G_s(r,\theta)$, the self part of the van Hove function. a) Values
197: of $t$ in which the system is inside a MB. b) Values of $t$ for which the system
198: is about to leave a MB.}
199: \label{fig3}
200: \end{figure}
201:
202: In Fig.~\ref{fig2} we show for the same run and time interval
203: $\delta^2(t,\theta)$, the (particle) averaged squared displacement (ASD)
204: of the particles within a time interval $\theta$ (solid curve). This
205: function is defined as
206: \vspace*{-4mm}
207:
208: \begin{eqnarray}
209: \delta^2(t,\theta) & = & \Delta^2(t-\theta/2,t+\theta/2) \\
210: & = &
211: \frac{1}{N}\sum_{i=1}^N |{\bf r}_i(t-\theta/2)-{\bf r}_i(t+\theta/2)|^2\quad .
212: \label{eq2}
213: \end{eqnarray}
214:
215: \noindent
216: Thus $\delta^2(t,\theta)$ is $\Delta^2(t',t'')$ measured along the
217: diagonal $t''=t'+\theta$ and hence the average of this quantity over
218: different start times $t$ gives the usual mean-squared displacement
219: for time lag $\theta$. A comparison of this ASD with Fig.~\ref{fig1}
220: shows that $\delta^2(t,\theta)$ is showing pronounced peaks exactly then
221: when the system leaves a MB. Thus we see that changing the MB is indeed
222: associated with a rapid motion as measured in the ASD.
223:
224: In order to understand the nature of the motion of the particles when the
225: system leaves a MB, we have calculated $\widehat{G}_s(r,t,t+\theta)$,
226: the distribution of the displacement $r$ of a particle for a
227: given time difference $\theta=40$. (Note that the average of
228: $\widehat{G}_s(r,t,t+\theta)$ over $t$ gives $G_s(r,\theta)$.) This
229: distribution is shown in Fig.~\ref{fig3}a for starting values $t$
230: that correspond to times at which the ASD shows a plateau, i.e. when
231: the system explores a MB. Also included in the graph is the self
232: part of the van Hove function, $G_s(r,\theta)$, and we see that
233: the distributions $\widehat{G}_s(r,t,t+\theta)$ are narrower and
234: more peaked than $G_s(r,\theta)$ thus showing that in a MB the system
235: moves more slowly than on average. In Fig.~\ref{fig3}b we show the same
236: distributions but now for times in which the system is about to leave a
237: MB (compare the values of $t$ with Fig.~\ref{fig1}). For these values
238: of $t$ the distributions are displaced to the right with respect to
239: $G_s(r,\theta)$, showing that in this time regime the motion of the
240: system is faster than on average. Most noteworthy is the observation
241: that this shift is relatively uniform, i.e. a substantial part of the
242: particles moves quicker than on average. Thus we can conclude that the
243: rapid increase of the ASD is {\it not} due to the presence of a {\it
244: few} fast moving particles, but instead to a ``democratic'' movement
245: of {\it many} particles, in contrast to the results for cooperative
246: motion on the time scale of $t^*$ which has been documented in earlier
247: work~\cite{kegel_00,weeks_00,donati_98}. Thus this movement is very
248: different in nature from the ``string-like'' motion found in the context
249: of the dynamical heterogeneities~\cite{donati_98,appignanesi_04}.
250:
251: To demonstrate that the number of particles that participate at
252: this democratic motion is indeed substantial and strongly correlated
253: with a strong increase in the ASD, we have defined as ``mobile'' all
254: those particles that in the time interval $\theta=40$ have moved more
255: than $r_{\rm th}=0.3$, and denote the fraction of such particles by
256: $m(t,\theta)$~\cite{footnote2}. In Fig.~\ref{fig2} we have included the
257: fraction of mobile particles as a function of time (vertical bars) and
258: a comparison of this data with the ASD in the same graph shows that the
259: fraction of mobile particles is indeed large whenever the ASD increases
260: rapidly. This fraction is on the order of 30\% of the particles and
261: thus significantly larger than one would expect from $G_s(r,\theta)$
262: if one integrates this distribution from $r_{\rm th}$ to infinity and
263: which gives 0.09.
264:
265: In order to give an idea on the nature of the motion of the mobile
266: particles during a MB-MB transition, we show in Fig.~\ref{fig4} a typical
267: configuration of mobile particles before such a transition event and attach to each
268: particle an arrow which points to the location of the particle after the
269: event, i.e. a time $\theta=40$ later. From this graph we recognize that
270: the MB-MB transitions correspond to a movement in which the particles
271: form a relatively compact cluster. Thus this is in contrast to the
272: type of motion found in the context of the DH in which the particles
273: form string-like objects~\cite{donati_98,appignanesi_04b}. These
274: compact regions have, at the temperature considered, around 30-60
275: particles and can be considered as potential candidates for the
276: cooperatively rearranging regions (CRR) proposed long time ago by
277: Adam and Gibbs~\cite{adam_65} and which are also at the heart of the
278: approach of Goldstein for the relaxation dynamics of glass-forming
279: systems~\cite{goldstein_69}.
280:
281: Finally we mention that we have found that upon a decrease of
282: temperature the sojourn time of the system within one MB increases
283: rapidly. This is in agreement with previous results in which the
284: concept of inherent structures was used~\cite{doliwa_03,vogel_04},
285: although here we have used a significantly simpler method to identify
286: the MB and which can notably also be used in real experiments such as
287: colloidal systems~\cite{kegel_00,weeks_00}. (We emphasize, however,
288: that we have obtained qualitatively the same results by considering
289: the inherent structures, although this approach is computationally much
290: more involved.) On the other hand an increase of the temperature makes
291: that the structure of the MB is basically washed out and the ASD does
292: no longer show the pronounced peaks.
293:
294: \begin{figure}[tb]
295: \includegraphics[width=0.80\linewidth]{fig4.eps}
296: \caption{
297: Configuration snapshot of mobile particles occurring in the MB-MB
298: transition $t=680 \to t=720$. The spheres (light and dark for the $A$
299: and $B$ particles, respectively) give the location of the particles
300: before the rearrangement and the arrows point to their position after
301: the transition. The black spheres give the location of the corner of
302: the simulation box.}
303: \label{fig4}
304: \end{figure}
305:
306: The present result indicate that the $\alpha-$relaxation is {\it not}
307: directly related to the presence of strings that are formed by a small
308: number of particles but instead is due to a cooperative rearrangement of a
309: substantial fraction of the particles. This cooperative rearrangement is
310: responsible for the transition between adjacent MBs and involves, at the
311: temperature considered, on the order of 40 particles. During such
312: a MB-MB transition the majority of the particles {\it outside} this
313: CRR does not contribute significantly to the relaxation. Therefore one
314: needs on the order of 5-10 such transition events in order to complete an
315: $\alpha-$relaxation on the scale of the local neighborhood. Since these
316: transitions take only a few percent of the $\alpha-$relaxation time, one
317: thus can envision, at least at low temperatures, the $\alpha-$relaxation
318: as a sequence of rapid, localized, cooperative relaxation events in which
319: a substantial number of particles participate. This result is thus in
320: surprisingly good agreement with the picture put forward long time ago
321: by Adam and Gibbs and Goldstein.
322:
323: Finally we mention that in view of the proposed connection between
324: the geometrical properties (distribution of barrier heights,
325: size of meta-basins,...) and the fragility of glass-forming
326: systems~\cite{debenedetti_01,angell_94}, it will be of great interest to do
327: the analysis presented here for a strong glass former. This will allow
328: to obtain a better understanding of the difference between strong and
329: fragile glass-formers~\cite{angell_85}.
330:
331:
332:
333:
334: GAA is research fellow of CONICET and RAM is research fellow of
335: CIC. Financial support from Fundaci\'on Antorchas, CONICET, UNS, SeCyT,
336: DYGLAGEMEM, and ECOS is gratefully acknowledged.
337:
338: \vspace*{-7mm}
339:
340: \begin{thebibliography}{10}
341:
342: \bibitem{ediger_96}
343: M. D. Ediger, C. A. Angell, and S. R. Nagel,
344: J. Phys. Chem. B {\bf 100}, 13200 (1996).
345:
346: \bibitem{gotze_99}
347: W. G\"otze,
348: J. Phys.: Condens. Matter {\bf 10}, A1 (1999).
349:
350: \bibitem{debenedetti_01}
351: P. G. Debenedetti and F. H. Stillinger,
352: Nature {\bf 410}, 259 (2001).
353:
354: \bibitem{ngai_02}
355: K. L. Ngai (Ed.): \textit{Proceedings of the Forth International
356: Discussion Meeting on Relaxations in Complex Systems}, Non-Cryst. Solids
357: {\bf 307-310} (2002).
358:
359: \bibitem{schmidt_91}
360: K. Schmidt-Rohr and H. W. Spiess,
361: Phys. Rev. Lett. {\bf 66}, 3020 (1991).
362:
363: \bibitem{butler_91}
364: S. Butler and P. Harrowell,
365: J. Chem. Phys. {\bf 95}, 4454 (1991).
366:
367: \bibitem{cicerone_95}
368: M. T. Cicerone, F. R. Blackburn, and M. D. Ediger,
369: J. Chem. Phys. {\bf 102}, 471 (1995).
370:
371: \bibitem{kob_97}
372: W. Kob, C. Donati, S. J. Plimpton, P. H. Poole, and S. C. Glotzer,
373: Phys. Rev. Lett. {\bf 79}, 2827 (1997).
374:
375: \bibitem{kegel_00}
376: W. K. Kegel and A. van Blaaderen,
377: Science {\bf 287}, 290 (2000).
378:
379: \bibitem{weeks_00}
380: E. R. Weeks, J. C. Crocker, A. C. Levitt, A. Schofield, and D. A. Weitz,
381: Science {\bf 287}, 627 (2000).
382:
383: \bibitem{richert_02}
384: R. Richert,
385: J. Phys.: Condens. Matter {\bf 14}, R703 (2002).
386:
387: \bibitem{donati_98}
388: C. Donati, J. F. Douglas, W. Kob, S. J. Plimpton, P. H. Poole, and S. C. Glotzer,
389: Phys. Rev. Lett. {\bf 80}, 2338 (1998).
390:
391: \bibitem{garrahan_02}
392: J. P. Garrahan and D. Chandler,
393: Phys. Rev. Lett. {\bf 89}, 035704 (2002).
394:
395: \bibitem{angell_91}
396: C. A. Angell,
397: J. Non-Cryst. Solids {\bf 131}, 13 (1991).
398:
399: \bibitem{debenedetti_01b}
400: P. G. Debenedetti, T. M. Truskett, and C. P. Lewis,
401: Adv. Chem. Eng. {\bf 28}, 21 (2001);
402: F. Sciortino, J. Stat. Mech. (in press) (2005).
403:
404: \bibitem{monthus_96}
405: C. Monthus and J.-P. Bouchaud,
406: J. Phys. A: Math. Gen. {\bf 29}, 3847 (1996).
407:
408: \bibitem{doliwa_03}
409: B. Doliwa and A. Heuer,
410: Phys. Rev. E {\bf 67}, 031506 (2003).
411:
412: \bibitem{vogel_04}
413: M. Vogel, B. Doliwa, A. Heuer, and S. C. Glotzer,
414: J. Chem. Phys. {\bf 120}, 4404 (2004).
415:
416: \bibitem{kob_95}
417: W. Kob and H. C. Andersen,
418: Phys. Rev. E {\bf 51}, 4626 (1995).
419:
420: \bibitem{ohmine_95}
421: I. Ohmine,
422: J. Phys. Chem. {\bf 99}, 6765 (1995).
423:
424: \bibitem{footnote1}
425: In order to check for the presence of finite size effects we have
426: also repeated the presented analysis for a system with 1000 particles
427: and considered only a subbox in this extended system. The results
428: obtained were very similar to the ones presented here and therefore
429: finite size effects can be excluded. More details will be presented
430: elsewhere~\cite{appignanesi_05}.
431:
432: \bibitem{appignanesi_05}
433: G. A. Appignanesi, J. A. Rodr\'{\i}guez Fris, R. A. Montani, and W. Kob,
434: to be published.
435:
436: \bibitem{appignanesi_04}
437: G. A. Appignanesi and R. A. Montani,
438: J. Non-Cryst. Solids {\bf 337}, 109 (2004).
439:
440: \bibitem{footnote2}
441: We mention that if this threshold is changed by a reasonable amount,
442: the number of mobile particles changes but the overall conclusions remain
443: nevertheless valid.
444:
445: \bibitem{appignanesi_04b}
446: G. A. Appignanesi, M. A. Frechero, L. M. Alarcon, J. A. R. Fris, and R. A. Montani,
447: Physica A {\bf 339}, 469 (2004).
448:
449: \bibitem{adam_65}
450: G. Adam and J. H. Gibbs,
451: J. Chem. Phys. {\bf 43}, 139 (1965).
452:
453: \bibitem{goldstein_69}
454: M. Goldstein,
455: J. Chem. Phys. {\bf 51}, 3728 (1969).
456:
457: \bibitem{angell_94}
458: C. A. Angell, P. H. Poole, and J. Shao,
459: Nuovo Cimento D {\bf 16}, 993 (1994).
460:
461: \bibitem{angell_85}
462: C. A. Angell, p. 1 in: K. L. Ngai and G. B. Wright (Eds.) {\it Relaxation
463: in Complex Systems} (US Dept. Commerce, Springfield, 1985).
464:
465: \end{thebibliography}
466:
467: \end{document}
468: