1: \documentclass[aps,preprint,nofootinbib]{revtex4}
2: \usepackage{graphicx}
3: \usepackage{color}
4: \usepackage{epsfig}
5: %\usepackage[dvipdfm,colorlinks]{hyperref}
6: %\hypersetup{c
7: % pdftitle={},
8: % pdfauthor={Ki-Joo Kim},
9: % pdfkeywords={pdf, latex, tex, ps2pdf, dvipdfm, pdflatex},
10: % bookmarksnumbered,
11: % pdfstartview={FitH},
12: % urlcolor=blue,
13: % citecolor=magenta,
14: %}
15:
16: \begin{document}
17: \title{Growing length and time scales in glass forming liquids}
18:
19: \author{Smarajit Karmakar$^{1}$}
20: \email{smarajit@physics.iisc.ernet.in}
21: %\thanks{To whom correspondence should be addressed}
22: \author{Chandan Dasgupta$^{1,2}$}
23: \email{cdgupta@physics.iisc.ernet.in}
24: \author{Srikanth Sastry$^{2}$}
25: \email{sastry@jncasr.ac.in}
26:
27: \affiliation{$^1$ Centre for Condensed Matter Theory, Department of Physics,
28: Indian Institute of Science, Bangalore, 560012, India}
29: \affiliation{$^2$ Jawaharlal Nehru Centre for Advanced Scientific Research, Bangalore
30: 560064, India.}
31:
32: \maketitle
33:
34: {\bf The nature of the glass transition, whereby a variety of liquids
35: transform into amorphous solids at low temperatures, is a subject of
36: intense research despite decades of investigation. Explaining the
37: enormous increase in viscosity and relaxation times of a liquid upon
38: supercooling is essential for an understanding of the glass
39: transition. Although the notion of a growing length scale of
40: ``cooperatively rearranging regions'' (CRR), embodied in the
41: celebrated Adam-Gibbs relation, has been invoked for a long time to
42: explain dynamical slow down, the role of length scales relevant to
43: glassy dynamics is not well established. Recent studies of spatial
44: heterogeneity in dynamics provide fresh impetus in this direction. A
45: powerful approach to extract length scales in critical phenomena is
46: finite size scaling, wherein a system is studied for a range of sizes
47: traversing the length scales of interest. We perform finite size
48: scaling for the first time for a realistic glass former, using
49: computer simulations, to evaluate a length scale associated with
50: dynamical heterogeneity which grows as temperature decreases.
51: However, relaxation times which also grow with decreasing temperature,
52: do not show the same kind of scaling behavior with system size as the
53: dynamical heterogeneity, indicating that relaxation times are not
54: solely determined by the length scale of dynamical heterogeneity. We
55: show that relaxation times are instead determined, for all studied
56: system sizes and temperatures, by configurational entropy, in
57: accordance with the Adam-Gibbs relation, but in disagreement with
58: theoretical expectations based on spin-glass models that
59: configurational entropy is not relevant at temperatures substantially
60: above the critical temperature
61: of mode coupling theory. The temperature dependence of the
62: heterogeneity lengthscale shows significant deviations from
63: theoretical expectations, and the length scale one may extract from
64: the system size dependence of the configurational entropy has much
65: weaker temperature dependence compared to the heterogeneity length
66: scale at all studied temperatures. Our results provide new insights
67: into the dynamics of glass-forming liquids and pose serious challenges
68: to existing theoretical descriptions.}
69:
70: Most approaches to understanding the glass transition and slow dynamics
71: in glass formers
72: \cite{adam-gibbs,jack,debstil,Gotze,inhmct,rfotpra,rfotrev,mezard,kinetic}
73: are based on the intuitive picture that the movement of their
74: constituent particles (atoms, molecules, polymers) requires
75: progressively more {\it cooperative} rearrangement of groups of
76: particles as temperature decreases (or density increases). Structural
77: relaxation becomes slow because the concerted motion of many
78: particles is infrequent. Intuitively, the size of such ``cooperatively
79: rearranging regions'' (CRR) is expected to increase with decreasing
80: temperature. Thus, the above picture
81: naturally involves the notion of a growing length scale, albeit implicitly in
82: most descriptions. The notion of such a length scale, related to the
83: configurational entropy $S_c$ (see methods), forms the basis of
84: rationalizing \cite{adam-gibbs,rfotpra,rfotrev} the
85: celebrated Adam-Gibbs (AG) relation \cite{adam-gibbs}
86: between the relaxation time and $S_c$.
87:
88: More recently, a number of theoretical approaches have explored the
89: relevance of a growing length scale to dynamical slow down
90: \cite{rfotrev,inhmct,kinetic}. A specific motivation for some of these
91: approaches arises from the study of heterogeneous dynamics in glass
92: formers\cite{edigerhet,onuki,harrowell,donati}. In particular,
93: computer simulation studies\cite{onuki,harrowell,donati} focused
94: attention on spatially correlated groups of particles which exhibit
95: enhanced mobility, and whose spatial extent grows upon
96: decreasing temperature. The spatial correlations of local relaxation permits
97: identification of a dynamical (time dependent) length scale, $\xi$,
98: through analysis of a {\it four point} correlation function first
99: introduced by Dasgupta {\it et al} \cite{chandan} (see methods), and the
100: associated dynamical susceptibility $\chi_4$\cite{silvio,sharon}. These
101: quantities have been studied recently {\it via} inhomogeneous
102: mode coupling theory (IMCT) \cite{inhmct} and estimated from simulation and
103: experimental
104: data\cite{inhmct,berthier_fss,berthier_wg,berthier_science,berthier_physreve,zamponi}.
105:
106: The method of finite size scaling, used extensively in numerical studies of
107: critical phenomena \cite{fss}, is uniquely suited for investigations of the
108: presence of a dominant length scale. This method involves a study of the
109: dependence of the properties of a finite system on its size.
110: We study a binary mixture of particles interacting via the
111: Lennard-Jones potential\cite{kob}, originally proposed as a model for
112: $Ni_{80} P_{20}$, and widely studied as a model glass former. We perform
113: constant temperature molecular dynamics simulations at a constant
114: volume (see methods and \cite{sastryPRL} for details), for seven
115: temperatures, and up to a dozen different system sizes for each temperature.
116: For each case, we calculate the dynamic susceptibility $\chi_4(t)$
117: as the second moment of the distribution of
118: a correlation function $Q(t)$ which measures the overlap of
119: the configuration of particles at a given time with the configuration
120: after a time $t$ (see
121: methods). From previous work, it is now well-established that
122: $\chi_4(t)$ has non-monotonic time dependence, and peaks at a time
123: $\tau_4$ that is proportional to the structural relaxation time
124: $\tau$. Such behavior is shown in the inset of Figure 1(a). In Figure
125: 1(a), we show the peak values $\chi_4^p \equiv
126: \chi_4(\tau_4)$ {\it vs} system size (number of particles)
127: $N$ for a range of temperatures. At each temperature, $\chi_4^p$
128: is an increasing function of $N$, saturating at large $N$.
129: The saturation occurs at a larger value of $N$ at lower temperatures.
130: This is precisely the finite-size scaling behaviour expected for a quantity whose
131: growth with decreasing temperature is governed by a dominant
132: correlation length that increases with decreasing temperature.
133:
134: We have estimated the correlation length $\xi$ from finite size
135: scaling of $\chi_4^p(T,N)$, which also involves estimating the value
136: of $\chi_4^p$ as $N \rightarrow \infty$. As the latter estimation is a
137: potential source of error in estimating $\xi$, we employ the Binder
138: cumulant of the distribution of $Q(\tau_4)$ to estimate $\xi$. The
139: Binder cumulant \cite{binder}, defined (see methods) in terms of the
140: fourth and second moments of the distribution, vanishes for a Gaussian
141: distribution, whereas it acquires negative values which are larger for
142: distributions that deviate from a Gaussian more strongly. The Binder
143: cumulant has been used extensively in finite size scaling analysis in
144: the context of critical phenomena, owing to its very useful property
145: that in systems with a dominant correlation length $\xi$, it is a
146: scaling function only of $L/\xi$ (or equivalently, of $N/\xi^3$),
147: where $L$ is the linear dimension of the system. The distributions
148: themselves are shown in the inset of Figure 1(b), for two different
149: system sizes for temperature $T = 0.47$. We see that the distribution
150: is unimodal for the large system size of $N = 1600$ whereas it is
151: strongly bimodal for the small system size of $N = 150$. The same
152: trend is observed as temperature is decreased for a fixed size of the
153: system. The data collapse of the Binder cumulant, from which we
154: extract the correlation length $\xi(T)$, is shown in Figure 1(b). The
155: collapse observed is excellent, confirming that the growth of
156: $\chi_4^p$ with decreasing $T$ is governed by a growing dynamical
157: correlation length. The values of $\xi$ obtained from this scaling
158: analysis are consistent with less accurate estimates obtained from a
159: similar analysis of the $N$-dependence of $\chi_4^p(T,N)$, as well as
160: from the wave vector dependence of the four point dynamic structure
161: factor $S_4(q,\tau_4)$ (see, {\it e. g.} \cite{inhmct}). Estimated $\chi_4^p$ as $N \rightarrow
162: \infty$ compare very well with the $q
163: \rightarrow 0$ limit of $S_4(q,\tau_4)$, up to a
164: proportionality constant, implying that concerns about the dependence
165: of $\chi_4^p$ values on the details of the simulations do not apply in
166: our case \cite{berthierjcp}.
167:
168: The value of $\xi$ (up to a temperature-independent multiplicative
169: factor that can not be determined from scaling, but has been fixed to
170: match results obtained from $S_4(q,\tau_4)$) increases from $2.11$ to
171: $6.23$ as $T$ decreases from $T = 0.70$ to $T = 0.45$. We find that
172: both $\xi$ and the asymptotic, $N \to \infty$ value of $\chi_4^p$
173: deviate from power law behavior as the critical temperature $T_c$ of
174: mode coupling theory ($T_c \simeq 0.435$ in our system) is approached
175: (consistently with previous observations). However, the power-law
176: relationship between $\chi_4^p$ and $\xi$ predicted in IMCT is
177: approximately satisfied by our data. Since the range of the measured
178: values of $\xi$ is small, it is not possible to obtain accurate
179: estimates of the exponents of these power laws.
180:
181: Next we consider the dependence of the relaxation time $\tau$ on $T$
182: and $N$. For each case, we calculate the relaxation time from the
183: decay of $<Q(t)>$. The results for $\tau$ are displayed in Figure 2,
184: which shows that $\tau$ increases as the temperature decreases, as
185: expected. However, the observed {\it increase} in $\tau$ with
186: decreasing $N$ for small values of $N$ at fixed $T$ is {\it not}
187: consistent with standard dynamical scaling for a system with a
188: dominant correlation length ({\it e.g.} near a critical point):
189: dynamical finite-size scaling would predict a decrease in $\tau$ as
190: the linear dimension $L$ of the system is decreased below the
191: correlation length $\xi$. Similar finite size effects on relaxation
192: times have been observed in previous simulations of realistic glass
193: formers ({\it e. g.} \cite{yamamoto}) but have not been analyzed in
194: detail. The observed $N$-dependence of $\tau$ is opposite to that
195: found in finite-size scaling studies of mean-field spin-glass models
196: ~\cite{binder_fss,rcr} whose analytically tractable behaviour forms
197: the starting point of most recent theories (including IMCT) of the
198: dynamics of glassy liquids \cite{brumer}.
199:
200:
201: The inset of Figure 3 shows the large-$N$ value of $\tau$ plotted as a
202: function of (a) the correlation length $\xi$ on a double-log scale,
203: and (b) ${\xi \over k_B T}$ on a semi-log scale. The
204: power-law relation between these two quantities predicted in IMCT
205: \cite{inhmct} is found above $T = 0.5$; deviation from a power law is
206: found at lower temperatures. The semi-log plot indicates that an
207: exponential form $\tau \sim \exp\left( k (\xi/k_B T)^\zeta\right)$,
208: with $\zeta = 0.7$, describes the data well in the entire
209: range. Though such a dependence is expected according to the random
210: first order theory (RFOT)\cite{rfotrev}, the exponent value we observe
211: cannot be easily rationalized within that framework. We comment
212: further on the significance of the exponent value later. Figure 3
213: shows relaxation times $\tau(T,N)$ for different $N$ values scaled to
214: the asymptotic $N \rightarrow \infty$ value $\tau(T)$, plotted against
215: values of $\chi_4^p(T,N)$ scaled to the asymptotic $N \rightarrow
216: \infty$ value $\chi_4^p(T)$. If the system size dependence of
217: $\tau$ and $\chi_4^p$ are governed by the same length scale,
218: one must expect a universal dependence of the scaled
219: relaxation times on the scaled $\chi_4^p$ values. From the data shown
220: in Figure 3, it is clear that there is no systematic relation between
221: the scaled $\tau$ and $\chi_4^p$ that describes their variation both with $T$ and
222: with $N$. These results indicate that the growth of $\tau$ with
223: decreasing $T$ is {\it not} governed solely by the correlation length
224: $\xi$ that describes the growth of $\chi_4$.
225:
226: Motivated by the AG relation \cite{adam-gibbs}, $\tau \propto
227: \exp\left({A \over T S_c}\right)$, where $A$ is a constant, we
228: next consider the dependence of $\tau$ on the configurational entropy
229: $S_c$ whose evaluation is described elsewhere \cite{sastryPRL}. As
230: shown in Figure 4 where $\log(\tau)$ is plotted {\it vs.}
231: $\frac{1}{TS_c}$ for all temperatures and system sizes studied, we
232: find a remarkable agreement with the AG relation, not only {\it vs.}
233: $T$ but also for all system sizes. To our knowledge, such a
234: demonstration of the validity of the AG relation for finite or
235: confined systems has not been made earlier. Thus, the $N$-dependence of
236: $\tau$, which can not be understood from dynamical finite-size
237: scaling, can be explained in terms of the $N$-dependence of $S_c$,
238: suggesting that the growth of $\tau$ with decreasing temperature is
239: more intimately related to the change of $S_c$, than to the increase
240: of the correlation length $\xi$ and susceptibility $\chi_4$,
241: associated with dynamical heterogeneity. As $S_c$ at a given
242: temperature varies with system size $N$, it is tempting to inquire if
243: the $N$-dependence of $S_c$ is associated with a length scale. We
244: extract such a length scale from data collapse of $S_c(T,N)$, scaled
245: to its value as $N
246: \rightarrow \infty$, shown in the top inset of Figure 4. We obtain
247: reasonable data collapse, but the extracted length scales turn out to
248: have substantially weaker $T$-dependence compared to $\xi$. The
249: comparison of the length scale obtained from the configurational
250: entropy and that obtained from dynamical heterogeneity is shown in the
251: bottom inset of Figure 4.
252:
253: A central role for the configurational entropy, along with an analysis
254: of a length scale relevant to structural relaxation, are the content
255: of the random first order theory, developed by Wolynes and co-workers
256: \cite{rfotrev}. According to RFOT, the length scale of dynamical
257: heterogeneity is the ``mosaic length'' that represents the critical
258: size for entropy driven nucleation of a new structure in a
259: liquid. Mean-field arguments based on known properties of
260: infinite-range models suggest that the RFOT mechanism is operative for
261: temperatures lower than $T_{MCT}$. In this regime, the dynamics of the
262: system is activated, with the relaxation time expected to vary as $\tau =
263: \tau_0 \exp\left[B \left(\Delta F \over k_B T\right)^\psi\right]$, where
264: $\Delta F$ is the free energy barrier to structural rearrangements,
265: and $\psi$ is an unknown exponent. The free energy barrier in turn
266: depends on the mosaic length as ${\Delta F \over k_B T} \sim
267: \xi^\theta$, where $\theta$ describes the dependence of the surface
268: energy on the size of a region undergoing structural change. The
269: observed validity of the AG relation, and the dependence of the
270: relaxation time on the length scale $\xi$, $\tau \sim \exp\left( k
271: (\xi/k_B T)^\zeta\right)$, with $\zeta = 0.7$, can be rationalized
272: within RFOT if the exponent $\theta$ is assumed to be close to 2.3,
273: and the exponent $\psi$ is close to 0.3. However, this interpretation
274: has the drawback that the exponent $\theta$ does not satisfy the
275: physical bound, $\theta \leq 2$, in three dimensions, and there is no
276: evident explanation for the value of $\psi$. We note that similar
277: conclusions were reached in a recent analysis \cite{zamponi} of
278: experimental data near the laboratory glass transition, on a large
279: class of glass-forming materials. Thus we find puzzling values for
280: the exponents relevant to the applicability of RFOT, which are in
281: need of explanation, and data in \cite{zamponi} indicate that such
282: a result may apply for a wide range of temperatures, all the way to
283: the experimental glass transition.
284:
285: RFOT focuses on behavior near the glass transition, and in the
286: limiting case of the spin glass models where theoretical perditions
287: are available, configurational entropy plays no role in the behavior
288: of the system above the mode coupling temperature. However, there have
289: indeed been attempts to extend the RFOT analysis to temperatures above
290: the mode coupling temperature\cite{stevenson,silvio_kac} and to
291: estimate a mosaic length scale at such temperatures, and we thus
292: compare our results with predictions arising from these analyses.
293: Stevenson {\it et al}\cite{stevenson} have considered the change in
294: morphology of rearranging regions above the mode coupling temperature,
295: and correspondingly the dependence of relaxation times on
296: configurational entropy. The predicted dependence of relaxation times
297: on configurational entropy differs from the Adam-Gibbs form, whereas
298: our results strikingly confirm the Adam-Gibbs form. Franz and
299: Montanari \cite{silvio_kac} have estimated a mosaic length scale in
300: addition to a heterogeneity lenght scale, and have discussed the
301: crossover in the dominant lengthscale near the mode coupling
302: temperature. The significantly smaller mosaic length scale they
303: estimate appears consistent with the smallness of the length scale we
304: obtain from the finite size dependence of the configurational entropy.
305: Nevertheless, this analysis does not contain explicit predictions
306: regarding the relevance of the configurational entropy at temperatures
307: higher than the mode coupling temperature.
308:
309: Our observation that the configurational entropy predicts the
310: relaxation times in accordance with the AG relation for all the
311: temperatures and system sizes we study poses serious challenges to
312: current theoretical descriptions based on the analogy with the
313: behaviour of mean-field models. Although the relevance of the
314: configurational entropy at high temperatures has been observed in
315: earlier simulation studies and analyses based on the inherent
316: structure approach \cite{sastryPRL,sastrynature1,saika}, we emphasize
317: that a theoretical analysis that satisfactorily explains such
318: dependence is not at hand at present, and our results concerning
319: the robustness of the Adam-Gibbs relation in finite systems highlights
320: further the challenge to existing theoretical descriptions. Equally
321: importantly, our results reveal that the length scale associated with
322: dynamical heterogeneity does not play the central role attributed to
323: it in recent analyses, and highlights the necessity to understand the
324: role of other relevant length scales, along the lines of the analysis
325: in \cite{silvio_kac}.
326:
327:
328: %%%% METHODS SECTION
329: %%%% 426 words now, but if one removes all the tex stuff, may be within 300 words.
330: %%%% which is the limit.
331:
332: \noindent{\bf Methods :}
333:
334: \noindent{\it Simulation details:} The system we study is a 80:20
335: (A:B) binary mixture of
336: particles interacting via the Lennard-Jones potential:
337: \begin{equation}
338: V_{\alpha\beta}(r)=4\epsilon_{\alpha\beta}\left[\left(\frac{\sigma_{\alpha\beta}}{r}
339: \right)^{12}-
340: \left(\frac{\sigma_{\alpha\beta}}{r}\right)^{6}\right],
341: \end{equation}
342: where $\alpha,\beta \in \{A,B\}$ and,
343: $\epsilon_{AB}/\epsilon_{AA}=1.5$, $\epsilon_{BB}/\epsilon_{AA}=0.5$,
344: $\sigma_{AB}/\sigma_{AA}=0.80$, $\sigma_{BB}/\sigma_{AA}=0.88$, masses
345: $m_A = m_B$. The interaction potential is cutoff at
346: 2.50$\sigma_{\alpha\beta}$. Length, energy and time are reported in
347: units of $\sigma_{AA}$, $\epsilon_{AA}$ and
348: $\sqrt{\sigma_{AA}^2/\epsilon_{AA}}$, and other {\it reduced
349: units} are derived from these. All simulations are done for number
350: density $\rho = 1.20$. We have used a cubic simulation box with
351: periodic boundary conditions. Simulations are done in the canonical
352: ensemble (NVT) using a modified leap-frog integration scheme. We
353: simulate for seven temperatures in the range $T \in \{0.450,
354: 1.00\}$. The mode coupling temperature for this system has been
355: estimated \cite{kob} to be $T_{MCT} \simeq 0.435$. We equilibrate the system for
356: $\sim 10^7 - 10^8$ MD steps depending on system size and production
357: runs are at least 5 times longer than the equilibration runs. We use
358: integration time steps $dt$ from $0.001$ to $0.006$ for the
359: temperature range $0.800$ to $0.450$. The studied system sizes vary
360: from $N = 50$ to $N = 1600$.
361:
362: \noindent{\it Dynamics:} Dynamics is studied via a two point correlation function, the overlap $Q(t)$,
363: \begin{eqnarray}
364: Q(t)= \int d\vec{r} \rho(\vec{r},t_0)\rho(\vec{r},t+t_0)
365: \sim \sum_{i=1}^{N}w(|\vec{r}_i(t_0)-\vec{r}_i(t_0+t)|)
366: \end{eqnarray}
367: where $\rho(\vec{r},t_0)$ {\it etc} are space-time dependent particle
368: densities, $w(r) = 1$, if $r \le a$ and zero otherwise, and averaging over the
369: initial time $t_0$ is implied. The use of
370: the window function [$a = 0.30$]
371: treats particle positions separated due to small
372: amplitude vibrational motion as the same. The second part of the
373: definition is an approximation which uses only the self-term, which we
374: have verified to be reliable (see \cite{sharon} for
375: details). The structural relaxation time $\tau$ is measured by a
376: stretched exponential fit of the long-time decay of $Q(t)$.
377: %Averages over initial times $t_0$ are performed for all
378: %time-dependent quantities.
379:
380: The fluctuations in $Q(t)$ yields the dynamical susceptibility:
381: \begin{equation}
382: \chi_4(t)=\frac{1}{N}[\langle Q^2(t)\rangle -\langle Q(t)\rangle^2].
383: \end{equation}
384: Previous work \cite{sharon} has shown that $\chi_4(t)$ reaches a maximum for times
385: $\tau_4$ which are proportional to the structural relaxation time $\tau$. We
386: report the values of $\chi_4^p \equiv \chi_4(t=\tau_4)$.
387:
388: The Binder cumulant, which we use for finite size scaling, is defined as
389: \begin{eqnarray}
390: B(N,T) &=& \frac{<[Q({\tau_4}) - <Q({\tau_4})>]^4>}{3<[Q({\tau_4}) -
391: <Q({\tau_4})>]^2>^2} -1.
392: \label{BC_def}
393: \end{eqnarray}
394: $B(N,T) = 0 $, if the distribution $P(Q({\tau_4}))$ is Gaussian, and is a
395: scaling function of $\xi/L$ only (where $L$ is the linear size of the
396: system, and $\xi$ is the correlation length), without any prefactor.
397:
398: \noindent{\it Configurational Entropy :} $S_c$, the
399: configurational entropy per particle, is calculated as the
400: measure of the number of distinct local energy minima, by subtracting
401: from the total entropy of the system the `vibrational' component:
402: \begin{equation}
403: S_c(T) = S_{total}(T) - S_{vib}(T).
404: \end{equation}
405: Details of the calculation procedure are as given in \cite{sastryPRL}.
406:
407:
408:
409: %% Bibliography.
410:
411: \begin{thebibliography}{100}
412:
413: % Adam-Gibbs 1, 2
414:
415:
416: \bibitem{adam-gibbs} Adam, G. and Gibbs, J. H On the Temperature Dependence
417: of Cooperative Relaxation Properties in Glass-Forming Liquids. {\it
418: J. Chem. Phys.} {\bf 43}, 139--146 (1965).
419:
420: \bibitem{jack} Dudowitz, J., Freed, K. F. and Douglas, J. F.
421: Entropy theory of polymer glass formation revisited. I. General Formulation.
422: {\it J. Chem. Phys.} {\bf 124} 064901-064915 (2006).
423:
424: % Energy landscape 3
425:
426: \bibitem{debstil} Debenedetti, P. G. and Stillinger, F. H. Supercooled liquids and the glass transition. {\it Nature} {\bf 410} 259 -- 267 (2001).
427:
428:
429: % mode coupling 4,5
430:
431: \bibitem{Gotze} G\"otze, W. and Sj\"ogren, L., Relaxation Processes in Supercooled Liquids. {\it Rep. Prog. Phys.} {\bf 55}, 241--376 (1992).
432:
433: \bibitem{inhmct} Biroli, G., Bouchaud, J.-P., Miyazaki, K. and Reichman, D. R. Inhomogeneous Mode-Coupling Theory and Growing Dynamic Length in Supercooled Liquids. {\it Phys. Rev. Lett.}, {\bf 97} 195701-1 -- 195701-4 (2006).
434:
435: % RFOT 6,7,8
436:
437: \bibitem{rfotpra} Kirkpatrick, T. R., Thirumalai, D. and Wolynes, P. G. Scaling concepts for the dynamics of viscous liquids near an ideal glassy state. {\it Phys. Rev. A} {\bf 40}, 1045 -- 1054 (1989).
438:
439: \bibitem{rfotrev} Lubchenko, V. and Wolynes, P. G. Theory of Structural Glasses and Supercooled Liquids, {\it Annu. Rev. Phys. Chem.} {\bf 58} 235 -- 266 (2007).
440:
441: %\bibitem{rfotbb} Bouchaud, J.-P. and Biroli, G. On the Adam-Gibbs-Kirkpatrick-Thirumalai-Wolynes scenario for the viscosity increase in glasses. {\it J. Chem. Phys.} {\bf 121} 7347 -- 7354 (2004).
442:
443: \bibitem{mezard} Mezard, M. and Parisi, G. Thermodynamics of glasses: a first principles computation. {\it J. Chem. Phys.} {\bf 111} 1076 --1095 (1999).
444:
445: % Kinetic models 9
446:
447: \bibitem{kinetic} Ritort, F. and Sollich, P. Glassy dynamics of
448: kinetically constrained models. {\it Adv. Phys.}, {\bf 52} 219-342,
449: (2003).
450:
451:
452: %% CRR Experimental using Adam-Gibbs 11,12
453:
454: %\bibitem{yamamuro} Yamamuro, O., Tsukushi, I., Lindqvist, A.,
455: %Takahara, S., Ishikawa, M. and Matsuo, T. Calorimetric Study of Glassy
456: %and Liquid Toluene and Ethylbenzene: Thermodynamic Approach to Spatial
457: %Heterogeneity in Glass-Forming Molecular Liquids. {\it
458: %J. Phys. Chem. B} {\bf 102} 1605--1609 (1998).
459:
460: %heterogeneous dynamics, expt 10
461:
462: \bibitem{edigerhet} Ediger, M. D. Spatially heterogeneous dynamics in supercooled liquids. {\it Annu. Rev. Phys. Chem.} {\bf 51} 99 - 128 (2000).
463:
464: % heterogeneous dynamics simulation 11, 12, 13
465:
466: \bibitem{onuki} Yamamoto, R. and Onuki, A. Kinetic Heterogeneities in a Highly Supercooled Liquid, {\it J. Phys. Soc. Jpn.}{\bf 66}, 2545-2548 (1997).
467:
468: \bibitem{harrowell} Hurley, M. M. and Harrowell, P. Kinetic structure of a two-dimensional liquid. {\it Phys. Rev. E} {\bf 52} 1694 - 1698 (1995).
469:
470: \bibitem{donati} Donati, C., Douglas, J. F.,Plimpton, S. J.,Poole, P. H. and Glotzer, S. C., String-like Cooperative Motion in a Supercooled Liquid. {Phys. Rev. Lett.} {\bf 80}, 2338-2342 (1998).
471:
472: % four point correlation functions 14,15,16
473:
474: \bibitem{chandan}
475: Dasgupta, C., Indrani, A. V., Ramaswamy, S. and Phani, M. K., Is there
476: a growing correlation length near the glass transition?, {\it
477: Europhys. Lett.} {\bf15}, 307-- 312 (1991).
478:
479: \bibitem{silvio} Franz, S. and Parisi, G. On nonlinear susceptibility in supercooled liquids. {\it J. Phys. Condens. Matter} {\bf 12} 6335 -- 6342 (2000).
480:
481: \bibitem{sharon} Donati, C., Franz, F., Parisi, G. and Glotzer, S. C. Theory of Non-linear Susceptibility and Correlation Length in Glasses and Liquids.
482: {\it J. Non-Cryst. Solids} {\bf 307-310} 215 -- 224 (2002).
483:
484: %% finite size scaling 17
485:
486: \bibitem{berthier_fss} Berthier, L. Finite-Size Scaling Analysis of the Glass Transition. {\it Phys. Rev. Lett.} {\bf 91} 055701-1 -- 055701-4 (2003).
487:
488:
489: %% kinetic models + BMLJ analysis 18
490: \bibitem{berthier_wg} Whitelam, S., Berthier, L. and Garrahan, J. P.,
491: Dynamic Criticality in Glass-Forming Liquids. {\it Phys. Rev. Lett.} {\bf 92} 185705-1 -- 185705-4 (2005).
492:
493: %% experimental using dyn. het. 19,20,21
494:
495: \bibitem{berthier_science} Berthier, L. {\it et al}, Direct Experimental
496: Evidence of a Growing Length Scale Accompanying the Glass
497: Transition. {\it Science}, {\bf 310}, 1797 -- 1800 (2005).
498:
499: \bibitem{berthier_physreve}
500: Dalle-Ferrier, C. {\it et al}, Spatial correlations in the dynamics of
501: glassforming liquids: Experimental determination of their temperature
502: dependence. {\it Phys. Rev. E} {\bf 76}, 041510-1 -- 041510-15 (2007).
503:
504: \bibitem{zamponi} Capaccioli, S., Ruocco, G. and Zamponi, F., Dynamically correlated regions and configurational entropy in supercooled liquids, http://arXiv.org/abs/0710.1249
505:
506: %% Finite size scaling 22
507:
508: \bibitem{fss} Privman, V. (ed), Finite size scaling and numerical simulations in
509: statistical systems, World Scientific, Singapore (1990).
510:
511: % Kob-Andersen BMLJ 23
512:
513: \bibitem{kob} Kob, W. and Andersen, H. C. Testing mode-coupling theory
514: for a supercooled binary Lennard-Jones mixture: The van Hove
515: correlation function. {\it Phys. Rev. E} {\bf 51}, 4626--4641 (1995).
516:
517: % Adam-Gibbs simulation evidence and methods 24
518:
519: \bibitem{sastryPRL} Sastry, S. Liquid Limits: The Glass Transition and
520: Liquid-Gas Spinodal Boundaries of Metastable Liquids. {\it
521: Phys. Rev. Lett.} {\bf 85} 590 -- 593 (2000).
522:
523:
524: % Binder cumulant 25
525:
526: \bibitem{binder} Binder, K. Finite Size Scaling Analysis of Ising Model Block Distribution Functions, {\it Z. Phys. B - Condensed Matter} {\bf 43} 119 -- 140 (1980).
527:
528: % analysis of ensemble effects etc
529:
530: \bibitem{berthierjcp} Berthier, L., Biroli, G., Bouchaud, J.-P., Kob, W., Miyazaki, K. and Reichman, D. R. Spontaneous and induced dynamic fluctuations in glass formers. I. General results and the dependence on ensemble and dynamics. {\it J. Chem. Phys.} {\bf 126}, 184503-1 -- 184503-21 (2007).
531:
532: % finite size effects in realistic glass formers
533:
534: \bibitem{yamamoto} Kim, K. and Yamamoto, R. Apparent finite-size effects in the dynamics of supercooled liquids, {\it Phys. Rev. E} {\bf 61} R41 -- R44 (2000).
535:
536: %% finite-size scaling for mean-field Potts glass and ROM model
537:
538: \bibitem{binder_fss} Brangian, C., Kob, W. and Binder, K., Finite-size scaling at the dynamical transition of the mean-field 10-state Potts glass.
539: Europhys. Lett. {\bf 53}, 756 -- 761 (2001).
540:
541: \bibitem{rcr} Rao, F., Crisanti, A. and Ritort, F. Frequency-domain
542: study of relaxation in a spin glass model for the structural glass
543: transition, {\it Europhys. Lett.} {\bf 62} 869 -- 875 (2003).
544:
545: \bibitem{brumer} However, some of these results are
546: controversial. Results contrary to those published in \cite{rcr} have been found
547: in Brumer, E. Y., Ph. D. Thesis, Quantum and Classical Statistical
548: Mechanics of Rough Energy Landscapes, Harvard University (2003).
549:
550:
551:
552:
553:
554:
555:
556: % Instanton result of high theta value 24,25
557:
558: %\bibitem{silviojstat} Franz, S. First steps of a nucleation theory of disordered systems. {\it J. Stat. Mech.} P04001 (2005).
559:
560: %\bibitem{dzero} Dzero, M., Schmalian, J. and Wolynes, P. G. Activated events in glasses: The structure of entropic droplets. {\it Phys. Rev. B} {\bf 72} 100201(R) -- 100205 (R)? (2005).
561:
562: % CRR as dynamical heterogeneities. 27
563:
564: \bibitem{stevenson} Stevenson, J. D., Schmalian, J. and Wolynes, P. G. The
565: shapes of cooperatively rearranging regions in glass-forming liquids.
566: {\it Nature Physics} {\bf 2}, 268 - 274 (2006).
567:
568: % mosaic scale above T_mct 28
569:
570: \bibitem{silvio_kac} Franz, S. and Montanari, A., Analytic determination of dynamical
571: and mosaic length scales in a Kac glass model,
572: {\it J. Phys. A: Math. Theor.}, {\bf 40}, F251-F257 (2007).
573:
574: % inherent structure paper 29,30
575:
576:
577: %\bibitem{scala} Scala, A., Starr, F. W., La Nave, E., Sciortino, F. and
578: %Stanley, H. E. Configurational Entropy and Diffusivity of Supercooled Water.
579: %{\it Nature} {\bf 406} 166--169 (2000).
580:
581: \bibitem{sastrynature1} Sastry, S., Debenedetti, P. G. and Stillinger,
582: F. H. Signatures of Distinct Dynamical Regimes in the Energy Landscape
583: of a Glass Forming Liquid. {\it Nature} {\bf 393}, 554--557 (1998).
584:
585:
586: \bibitem{saika} Saika-Voivod, I., Poole, P. H. and Sciortino, F. Fragile to Strong transition and polyamorphism in the energy landscape of liquid silica. {\it Nature} {\bf 412} 514 -- 517 (2001).
587:
588:
589: % one possible addition is the odagaki work on CRR.
590:
591: \end{thebibliography}
592:
593:
594:
595: \eject
596:
597: %%%% FIGURE CAPTIONS
598:
599: % Figure 1: panel 1. Inset - chi_4 vs t. Main panel - chi_4 vs N.
600: % Figure 1: panel 2. Inset - P(Q) for two N values. Main panel - data collapse of
601: % Binder cumulant.
602:
603: % Figure 2: tau vs T, N
604:
605: % Figure 3: tau(T,N)/tau_0 vs Chi_4(T,N)/chi_4(0) in main panel. tau vs xi, xi/kT in inset.
606:
607: % Figure 4: tau vs 1/TSc, and Sc scaling, Xi vs T in insets
608:
609:
610: \centerline{\LARGE \bf Figure Captions}
611:
612: {\bf Figure 1: System size dependence of dynamic susceptibility
613: $\chi_4^p$, and
614: finite size scaling of the Binder cumulant $B(N,T)$}. {\bf a} {\bf Inset:}
615: $\chi_4(t)$, shown here for $N = 1000$ and selected temperatures,
616: exhibits non-monotonic time dependence, and the time $\tau_4$ at which it has
617: the maximum value has been observed to be proportional to
618: the structural relaxation time $\tau$. {\bf Main panel:} Peak height of the
619: four-point dynamic susceptibility, $\chi_4^p(T,N) \equiv
620: \chi_4(t=\tau_{4}, T, N)$, has been shown as a function of system
621: size $N$ for different temperatures. For each temperature, $\chi_4^p(T,N)$
622: increases with system size, and saturates for large system
623: sizes. $\chi_4^p(T,N)$ also increases as the temperature is lowered. {\bf
624: b} {\bf Inset:} The distribution $P[Q(\tau_4)-<Q(\tau_4)>]$
625: of $Q(\tau_4)-<Q(\tau_4)>$ is shown for two
626: system sizes for T = 0.470. While the distribution for the large size
627: is nearly Gaussian, the small system exhibits a strongly bimodal
628: distribution. Such bimodality is also observed to emerge as the
629: temperature is decreased at fixed system size. {\bf Main Panel:}
630: Binder cumulant B(N,T) (see Methods ) has been plotted as function of
631: $N/\xi^3$ for different temperatures in the range $T \in [0.45,
632: 0.80]$. The correlation length $\xi$ is an unknown,
633: temperature dependent, scaling parameter determined by requiring data
634: collapse for values at different $T$. By construction $B(N,T)$ will
635: approach zero for large system sizes at high temperatures. It changes
636: to negative values as the temperature or the system size is decreased
637: such that $P[Q(\tau_4)-<Q(\tau_4)>]$ becomes bimodal. The correlation length $\xi(T)$ is
638: the only unknown to be determined in order to obtain data collapse for
639: $B(N,T)$ and the quality of the data collapse confirms the reliability
640: of this procedure.
641:
642: {\bf Figure 2: Relaxation times as a function of temperature and system size}.
643: Relaxation time $\tau(T,N)$ for the largest system size increases
644: roughly by three decades from the highest to the lowest temperature
645: shown. For each temperature, $\tau(T,N)$ increases as $N$ is decreased
646: for small values of $N$,
647: displaying a trend that is opposite to that observed near
648: second order critical points. For the smallest temperature, $\tau(T,N)$
649: increases by about a decade from the largest to the smallest system size.
650:
651: {\bf Figure 3: Relationship between the relaxation time $\tau(T,N)$,
652: correlation length $\xi(T)$ and the dynamic susceptibility
653: $\chi_4^p(T,N)$}. {\bf Inset:} (a) $\tau(T,N \to
654: \infty)$ is shown against $\xi(T)$, in a log-log plot (bottom curve). This plot shows that a
655: power-law dependence holds over a temperature range above $T = 0.5$,
656: but breaks down at lower temperatures. (b) $\tau(T,N \to
657: \infty)$ is shown against $\xi(T)/k_B T$, in a semi-log plot (top curve). This plot
658: shows that an exponential dependence $\tau \sim \exp\left( k (\xi/k_B
659: T)^\zeta\right)$, with $\zeta = 0.7$, describes the data well in the
660: entire temperature range. However the observed exponent value $\zeta =
661: 0.7$ is difficult to explain with existing theories.
662: {\bf Main Panel:} Relaxation times $\tau(T,N)/\tau(T,N \to
663: \infty)$ shown against $\chi_4^p(T,N)/\chi_4^p(T,N \to \infty)$
664: in a semi-log plot. While at fixed $N$ both $\tau$ and $\chi_4^p$
665: increase upon decreasing $T$, at fixed $T$, they show opposite trends,
666: with $\tau$ increasing for decreasing $N$ and $\chi_4^p$ increasing
667: for increasing $N$. If $\tau$ and $\chi_4^p$ are determined by the same
668: length scale $\xi$ and further, if their finite size behavior is
669: governed by $N/\xi^3$, the plotted data are expected to lie on a
670: universal curve, which is seen not to be the case.
671:
672: {\bf Figure 4: The dependence of relaxation times $\tau$ on the
673: configurational entropy $Sc$}. {\bf Main Panel:} The relaxation times
674: are shown against configurational entropy in an ``Adam-Gibbs'' plot
675: [$\log(\tau)$ {\it vs} $1/(TS_c)$], for all the temperatures and
676: system sizes studied. The impressive data collapse of all the data
677: onto a master curve indicates that the configurational entropy is
678: crucial for determining the relaxation times. The overall behavior is
679: well described by the AG relation, which requires $\log(\tau)\propto
680: 1/(TS_c)$. Although other powers of $1/(TS_c)$ may also describe the
681: data well, the improvement in fit quality is marginal, and hence we
682: treat the data presented as validating the AG relation. {\bf Top
683: Inset:} The configurational entropy $S_c(T,N)$ scaled to its $N
684: \rightarrow \infty$ value, has been plotted as function of $N/\xi_s^3$
685: for different temperatures in the range $T \in [0.45, 0.80]$, to
686: extract a temperature dependent length scale $\xi_s$ that leads to
687: data collapse. {\bf Bottom Inset:} The length scale obtained from the
688: data collapse of the configurational entropy (green diamonds) compared
689: with the length scale obtained from finite size scaling of the Binder
690: cumulant (red triangles). It is apparent that the length scale from
691: configurational entropy shows very weak temperature dependence, in
692: contrast with the dynamical heterogeneity length scale.
693:
694:
695:
696: \eject
697:
698:
699: %\centerline{\LARGE \bf Figures}
700:
701: \begin{figure}[h]
702: \begin{center}
703: \epsfig{file=./Fig1a.ps,scale=0.5,angle=-90,clip=} \epsfig{file=./Fig1b.ps,scale=0.5,angle=-90,clip=}
704: \caption{\bf Karmakar et al}
705: \end{center}
706: \end{figure}
707:
708: \pagebreak
709:
710: \begin{figure}[h]
711: \begin{center}
712: \epsfig{file=./Fig2.ps,scale=0.5,angle=-90,clip=}
713: \caption{\bf Karmakar et al}
714: \end{center}
715: \end{figure}
716:
717: \pagebreak
718:
719: \begin{figure}[h]
720: \begin{center}
721: \epsfig{file=./Fig3.ps,scale=0.5,angle=-90,clip=}
722: \caption{\bf Karmakar et al}
723: \end{center}
724: \end{figure}
725:
726:
727:
728:
729: \pagebreak
730:
731:
732: \begin{figure}[h]
733: \begin{center}
734: \epsfig{file=./Fig4.ps,scale=0.5,angle=-90,clip=}
735: \caption{\bf Karmakar et al}
736: \end{center}
737: \end{figure}
738:
739:
740: \end{document}
741:
742: