1: \documentclass[twocolumn,preprintnumbers,amsmath,amssymb]{revtex4}
2: \usepackage{graphicx,epsfig}
3: \usepackage{dcolumn}
4: \usepackage{bm}
5: %\parskip=0.5cm
6: \begin{document}
7:
8:
9: \title {Dynamical Heterogeneity and the interplay between activated and
10: mode coupling dynamics in supercooled
11: liquids
12: }
13: \author{ Sarika Maitra Bhattacharyya$^{\dagger}\footnote
14: {Electronic mail~:sarika@sscu.iisc.ernet.in}$,
15: Biman Bagchi$^{\dagger}$\footnote{Electronic mail~:bbagchi@sscu.iisc.ernet.in} \\
16: and \\
17: Peter G. Wolynes$^{\ddagger}$\footnote
18: {Electronic mail~:pwolynes@chem.ucsd.edu}}
19:
20: \affiliation{$^{\dagger}$ Solid State and Structural Chemistry Unit,
21: Indian Institute of Science, Bangalore 560 012, India.\\
22: $^{\ddagger}$Department of Chemistry and Biochemistry,
23: University of California at San
24: Diego, La Jolla, California 92093-0371}
25:
26:
27:
28: \begin{abstract}
29: We present a theoretical analysis of the dynamic structure factor (DSF)
30: of a liquid at
31: and below the mode coupling critical temperature $T_c$, by developing
32: a self-consistent theoretical treatment which includes the contributions both from
33: continuous diffusion, described using general two coupling parameter ($F_{12}$)
34: mode coupling theory (MCT), and from the
35: activated hopping, described using the random first order transition (RFOT)
36: theory, incorporating the effect of dynamical heterogeneity.
37: The theory is valid over the whole temperature plane and shows
38: correct limiting MCT like behavior above $T_{c}$ and goes over to the RFOT
39: theory near the glass transition temperature, $T_{g}$. Between
40: $T_{c}$ and $T_{g}$, the theory predicts that neither the continuous diffusion,
41: described by pure mode coupling theory,
42: nor the hopping motion alone suffices
43: but both contribute to the dynamics while interacting
44: with each other.
45: We show that the interplay between the two contributions conspires to modify the
46: relaxation behavior of the DSF from what would be predicted by a theory with
47: a complete static Gaussian barrier distribution in a manner that may be
48: described as a facilitation effect.
49: Close to $T_c$, coupling between the short
50: time part of MCT dynamics and hopping reduces the stretching given by the F$_{12}$-MCT
51: theory significantly and
52: accelerates structural relaxation. As the temperature is progressively lowered
53: below $T_c$, the equations yield a crossover from MCT dominated regime to the
54: hopping dominated regime.
55: In the combined theory the dynamical heterogeneity is modified because the
56: low barrier components interact with the MCT dynamics to enhance the relaxation rate
57: below $T_c$ and reduces the stretching that would otherwise arise from an
58: input static
59: barrier height distribution. Many of these results can be explained
60: from an analytical treatment of the combined
61: equation of motion.
62: \end{abstract}
63: \maketitle
64: \section{Introduction}
65:
66: In an earlier article we showed how to connect self-consistently the mode
67: coupling theory (MCT) with the random first order transition theory (RFOT)
68: to describe the dynamics of a liquid above and below the mode coupling
69: transition temperature, $T_{c}$ \cite{sbp}.
70: The resulting dynamics
71: includes both the diffusive dynamics described by MCT and the hopping dynamics
72: described by RFOT theory. Although other earlier attempts to include
73: both hopping and mode coupling dynamics within one theoretical
74: scheme have been made \cite{gotze,das}, the merit of our calculation was
75: the use of
76: hopping dynamics, determined
77: via the
78: RFOT theory, thus acknowledging in accord with experiments
79: that the hopping rate decreases with the configurational entropy. Because of
80: the feedback between the structural
81: relaxation (which includes contribution from both continuous and hopping
82: dynamics) and the viscosity, hopping has a non-linear effect on the total
83: dynamics. Due to the self-consistent nature of the
84: calculation there was hopping induced softening of the growth of the frequency
85: dependent viscosity with decreasing temperature
86: and this in turn helped the relaxation of the MCT
87: contribution to the structural relaxation. Thus, the theory
88: predicts that below $T_{c}$ along with the input hopping dynamics
89: there is an additional hopping
90: induced continuous diffusion which was absent when hopping was frozen.
91: The time scale of relaxation
92: is thus found to be faster than that predicted by hopping motion alone
93: because it now also includes the contribution from the continuous dynamics.
94: This effect is key to showing explicitly that no strict
95: localization transition
96: takes place at $T_{c}$, in accord with long standing arguments
97: \cite{kirk,biroli}.
98:
99: To keep the
100: theory analytically tractable in our earlier work we
101: used a simpler version of the MCT which
102: included only one coupling parameter $\lambda$ and neglected the
103: distribution of barrier heights in the hopping dynamics predicted by RFOT
104: theory \cite{sbp}. As a result the $\alpha$ relaxation was nearly exponential.
105: In this present article in
106: order to address the origin of the stretching of the long time
107: $\alpha$ relaxation
108: dynamics we examine not only a two coupling parameter MCT
109: (Gotze's $F_{12}$ model)
110: having both $\lambda_{1}$ and
111: $\lambda_{2}$ term which in combination can directly result
112: in stretching \cite{gotze} but we also incorporate the static barrier
113: height distribution, from
114: RFOT theory, in the
115: hopping dynamics.
116: The term containing $\lambda_{1}$
117: describes the coupling of the density relaxation
118: to a static field (which may describe a localized
119: defect or a static inhomogeneity in the density of the
120: system) and that containing $\lambda_{2}$ describes the self coupling.
121: As shown by Gotze and co-workers, the
122: $F_{12}$ model which formally describes static inhomogeneity
123: present in the system predicts a stretching of the $\alpha$
124: relaxation dynamics above $T_{c}$ \cite{gotze}.
125: It is not clear precisely how such a static inhomogeneity would in fact
126: be generated
127: above the microscopic $T_{c}$, however, such a scenario is
128: perhaps viable at temperatures below $T_c$ but there
129: the hopping dynamics must also contribute significantly.
130: %**************************************************88
131:
132: Computer simulation studies of atomic displacements in supercooled binary mixture systems
133: strongly suggest
134: the coexistence of continuous diffusion and hopping as mechanisms
135: of mass transport \cite{sarikajcp}.
136: These studies show that hopping events are often followed by enhanced
137: continuous diffusion. These two mechanisms can obviously, therefore,
138: interact cooperatively with
139: each other.
140:
141: The present analysis provides a quantitative description of the non-linear
142: interaction between continuous diffusion and dynamically heterogeneous
143: activated hopping. It is shown that at and just below $T_C$,
144: hopping helps unlock the continuous diffusion which now becomes more effective
145: than the hopping would be by itself.
146: We further find that below
147: $T_c$, the stretching of the relaxation combines the effects of activation
148: inhomogeneity and static inhomogeneity. The barrier height distribution
149: takes
150: care of the dynamic inhomogeneity in the system and becomes the primary
151: source of the stretching of
152: the dynamics much below $T_{c}$, but the MCT effects play a role in
153: the low barrier components to enhance the rate of short time diffusion.
154:
155: The organization of the rest of the paper is as follows. In the next section we
156: describe the theoretical scheme. In section III we present several analytical results
157: that can be derived for the combined theory. Section IV contains numerical results.
158: Section V concludes with a discussion on the results.
159:
160: %*****************************************************************
161:
162:
163: \section{Theoretical Scheme}
164:
165: In our earlier article we showed that activated dynamics
166: or hopping opens up an
167: extra channel for the structural relaxation \cite{sbp}. The continuous dynamics was
168: calculated using the one coupling parameter ($\lambda$)
169: MCT.
170: In describing the activated
171: motions the probability of a single hop was calculated from RFOT theory which
172: connects the height of the free energy barrier to
173: the configurational entropy. For simplicity we had considered a single value of the
174: barrier height for each temperature although RFOT theory shows that this
175: barrier is in fact distributed. The present theory uses the same scheme
176: of calculation with some modifications to understand the relation to
177: previous MCT efforts to cope with nonexponential relaxation.
178: $(1)$ The one coupling parameter ($\lambda$) MCT is extended to incorporate
179: the two coupling parameters
180: ($\lambda_{1}$, $\lambda_{2}$) MCT (the $F_{12}$ model)
181: which Gotze has used to address the effect of static inhomogeneity on the dynamics above $T_{c}$.
182: $(2)$ The single valued barrier height for the activated dynamics is
183: replaced by a distribution of barrier heights in accord with RFOT theory.
184:
185: The previous article used two different mathematical schemes to combine the
186: hopping and the continuous dynamics (described by MCT)\cite{sbp}.
187: In one of the schemes
188: the full intermediate scattering function was written as
189: a product of a hopping
190: and a MCT part using the separation of timescales between the MCT dynamics and
191: the hopping dynamics. In the second scheme the
192: strict parallelism of hopping and continuous motion was more transparent.
193: The structure of the equation in the second scheme is similar
194: to that obtained by Gotze and coworkers \cite{gotze} and Das and Mazenko
195: \cite{das} from more detailed microscopic derivations.
196: Both
197: the schemes give nearly identical results. This further adds credence
198: to the first scheme. In the present paper, therefore,
199: we will work
200: with the first scheme (easier to implement and also in this scheme
201: the continuous diffusion
202: and hopping dynamics can be investigated separately)
203: although the extensions made here can also
204: be incorporated into the second scheme in a similar manner.
205:
206: The total intermediate scattering function can be written as,
207: \begin{equation}
208: \phi(q,t)\simeq \phi_{MCT}(q,t)\phi_{hop}(q,t). \label{fqttot}
209: \end{equation}
210: \noindent
211: Here $\phi_{MCT}(q,t)$ is the MCT part of the intermediate
212: scattering function, which is now self consistently calculated
213: with $\phi(q,t)$, and its equation of motion is given by,
214:
215: \begin{eqnarray}
216: \ddot{\bf\phi}_{MCT}(t)
217: +\gamma \dot{\bf \phi}_{MCT}(t)
218: &+& \Omega_{0}^{2}
219: {\bf\phi}_{MCT}(t)\\ \nonumber
220: &+& \lambda_{1} \Omega_{0}^{2}
221: \int_{0}^{t} \:dt^{\prime} {\bf\phi}(t^{\prime})
222: \dot{\bf\phi}_{MCT}(t-t^{\prime}) \\ \nonumber
223: &+& \lambda_{2} \Omega_{0}^{2}
224: \int_{0}^{t} \:dt^{\prime} {\bf\phi}^{2}(t^{\prime})
225: \dot{\bf\phi}_{MCT}(t-t^{\prime}) = 0 \label{fqtmct}
226: \end{eqnarray}
227: \noindent
228:
229: In the above equation the fourth term on the left hand side describes the
230: coupling of $\phi_{MCT}(q,t)$ with a static field which is meant to describe
231: the defects
232: or the inhomogeneity in the system, according to Gotze \cite{gotze}.
233: The fifth term on the left describes
234: the coupling of $\phi_{MCT}(q,t)$ with itself (the self coupling term).
235: Unlike the earlier model the present model contains two order parameters.
236: In the absence of hopping,
237: the MCT transition would now take place not at a single point but at many points
238: on the ${\lambda_{1}}-{\lambda_{2}}$ plane.
239:
240: In eq.\ref{fqttot} the hopping part of the intermediate scattering function
241: is give by $\phi_{hop}(q,t)$. The contribution from a single hopping event
242: to the scattering function was derived in our earlier paper \cite{sbp}.
243: It can be written as,
244: \begin{eqnarray}
245: \phi_{hop}^{s}(q)=\frac{1}{s+K_{hop}(q)} \label{fqthop}
246: .
247: \end{eqnarray}
248: \noindent
249: where,
250: \begin{eqnarray}
251: K_{hop}(q)=\frac{P}{v_{p}}\Bigl[v_{0}-8
252: &&\int_{\pi/\xi}^{\infty} \:dq_{1} q_{1}^{2}
253: e^{-q_{1}^{2}d_{l}^{2}}\\\nonumber
254: &&\times \Bigl\{
255: \Bigl(\frac{-(q-q_{1})\xi \: cos((q-q_{1})\xi)}{(q-q_{1})^{3}}\\\nonumber
256: &&+\frac
257: {sin((q-q_{1})\xi )}
258: {(q-q_{1})^{3}}\Bigr)^{2}\Bigr\}\Bigr] \label{khop}
259: \end{eqnarray}
260: \noindent
261: In the above expression of the hopping kernel, $P$ is the average hopping
262: rate which is a function of the free energy barrier height, $\Delta F$
263: and is given by $P=\frac {1}{\tau_{0}}exp(-\Delta F/k_{B}T)$ \cite{lubwoly}.
264: The
265: free energy barrier is calculated from RFOT theory \cite{lubwoly}.
266: $v_{0}= \frac{4}{3}\pi \xi^{3}$
267: is the region participating in hopping where $\xi$ is calculated from RFOT
268: theory. $v_{p}$ is the volume of a single particle in the system. $d_{L}$
269: is the Lindemann length. In this model kernel a typical hopping event
270: involves an uncorrelated
271: displacement of particles by a Lindemann length. More
272: complex kernels that encode correlations between movements are also possible.
273:
274:
275: Now if we consider a distribution of barrier heights then the contribution
276: from multiple hoppings to the intermediate structure function can be written as,
277:
278: \begin{eqnarray}
279: \phi_{hop}(q,t)
280: &=&\int \phi^{s}_{hop}(t) {\cal P}(\Delta F) d\Delta F \nonumber \\
281: &=&\int e^{-tK_{hop}(\Delta F)} {\cal P}(\Delta F) d\Delta F \label{hopdist}
282: \end{eqnarray}
283: \noindent
284:
285: ${\cal P}(\Delta F)$ is considered to be Gaussian. With a Gaussian
286: distribution of barrier heights the relaxation function is known to
287: fit well to a stretched exponential, where the stretching depends on
288: the width of the Gaussian \cite{xiawoly}.
289:
290: \begin{figure}
291: \epsfig{file=fig1.eps,height=5cm,width=7cm,angle=0}
292: \caption { The $\phi(t)$ calculated from the unified theory, and the results of the idealized MCT are plotted
293: for three temperatures. The solid lines A, B, C correspond to the full $\phi(t)$ for
294: $\epsilon=-0.1$, $\epsilon=0$ and $\epsilon=0.5$, respectively. The
295: dashed lines
296: are the idealized MCT results.
297: In this plot $\lambda=0.568 $
298: which means that the MCT dynamics without hopping above $T_{c}$ is
299: stretched with stretching parameter $\beta_{MCT}=0.8$. The activated
300: dynamics is calculated with barrier height distribution such that
301: $\beta^{static}_{hop}=0.5$.}
302: \end{figure}
303:
304: In figure 1 we plot the results of the unified theory and also that of
305: the idealized MCT ($F_{12}$ model). The plots are given for
306: $\epsilon=-.1$, $\epsilon=0$ and $\epsilon=0.5$ which correspond to
307: temperatures above, at and below $T_{c}$ respectively. The calculations
308: are done at $\lambda=0.568$ which for idealized MCT
309: above $T_{c}$ (for $\epsilon < 0$ )
310: predicts a stretched relaxation with the MCT stretching parameter,
311: $\beta_{MCT}=0.8$.
312: For the activated dynamics we have considered
313: Gaussian distribution of barriers that would predict the stretching parameter
314: for the hopping dynamics,
315: $\beta^{static}_{hop}=0.5$. As expected,
316: hopping does not have much effect above $T_{c}$ but below $T_{c}$ the unified
317: theory continues to show structural relaxation where
318: the idealized MCT would have predicted
319: strict localization transition. The longtime dynamics is stretched at all
320: the temperatures. But as will be discussed later the stretching parameter
321: for the total dynamics is different from $\beta_{MCT}$ and
322: $\beta^{static}_{hop}$. Its value depends on the interaction between the
323: hopping and the MCT dynamics and changes with temperature.
324:
325: \section{The effect of hopping on the MCT dynamics above, at and below $T_{c}$
326: :Analytical results}
327:
328: In the earlier article we showed numerically that hopping has very little
329: effect on the MCT dynamics above $T_{c}$ and has a nonlinear effect on the MCT
330: relaxation timescale at $T_{c}$. In this article we analytically investigate
331: the effect of hopping on the MCT dynamics above, at and below $T_{c}$.
332: For this study we first examine the one parameter MCT dynamics, i.e. we
333: consider $\lambda_{1}=0$. This is a special case of the two parameter model.
334: The MCT transition takes place at $\lambda_{2}=4$ for $\lambda_{1}=0$.
335: Initially we also take the
336: barrier height distribution to be a delta function corresponding to a single
337: hopping barrier. With these simplifications we use
338: eqs.\ref{fqttot}-\ref{hopdist} for the following analysis of the effect of
339: hopping on the MCT timescale.
340:
341: When $\lambda_{1}=0$,
342: Eq.\ref{fqtmct} can be rewritten as,
343: \begin{eqnarray}
344: \frac{\Omega^{2}_{o} \phi_{MCT}(z)}{1+z\phi_{MCT}(z)}=z+i\gamma
345: + LT \lambda_{2}\Omega^{2}_{o} [\phi^{2}(t)](z)\label{fqtlap}
346: \end{eqnarray}
347: \noindent
348: where LT stands for Laplace transform.
349: Since we are interested in the $\alpha$ relaxation timescale we will take the
350: longtime limit of the above equation. In the longtime limit where
351: $z\rightarrow 0$, eq.\ref{fqtlap} reduces to,
352:
353: \begin{eqnarray}
354: \frac{\Omega^{2}_{o} \phi_{MCT}(z)}{1+z\phi_{MCT}(z)}=i\gamma
355: + LT \lambda_{2}\Omega^{2}_{o} [\phi^{2}(t)](z)\label{fqtlap1}
356: \end{eqnarray}\noindent
357:
358: In the above equation the second term on the right hand side is the self
359: coupling term which makes dominant contribution at high density. But at low
360: liquid density the self coupling term can be neglected and the solution of eq.\ref{fqtlap1}
361: is given by,
362:
363: \begin{equation}
364: \phi_{MCT}(t)=e^{-K_{o}t}
365: \end{equation}
366: \noindent
367: where $K_{o}=\frac{\Omega^{2}_{o}}{\gamma}$ is the inverse
368: timescale of longtime decay in the normal liquid regime.
369:
370: Eq.\ref{fqtlap1} in the time plane can be written as,
371: \begin{eqnarray}
372: {\bf\phi}_{MCT}(t)=
373: \frac{1}{K_{o}}\dot{\bf \phi}_{MCT}(t)
374: +\lambda_{2}
375: \int_{0}^{t} \:dt^{\prime} {\bf\phi}^{2}(t^{\prime})
376: \dot{\bf\phi}_{MCT}(t-t^{\prime}) \label{fqtalpha}
377: \end{eqnarray}
378: \noindent
379: From the earlier analysis \cite{sbp,leu} we know that the solution of
380: eq.\ref{fqtalpha}
381: is exponential in the longtime. So we simplify $\phi_{MCT}(t)$
382: using its longtime form, $\phi_{MCT}(t)=a e^{-K_{MCT}t}$,
383: where $a$ is the prefactor.
384: Presently we also take the hopping part of the intermediate scattering function
385: to be an exponential, $\phi_{hop}(t)=e^{-K_{hop}t}$.
386: Using these expressions for
387: $\phi_{MCT}$ and $\phi_{hop}$ in eq.\ref{fqtalpha} we get an expression
388: for $K_{MCT}$ in terms of $K_{hop}, K_{o},\lambda_{2}$ and $a$.
389: \begin{eqnarray}
390: K_{MCT}=\frac{1}{2}&\Bigl[&-(2K_{hop}-(1-a^{2}\lambda_{2})K_{o})\\\nonumber
391: &+&\sqrt{(2K_{hop}-(1-a^{2}\lambda_{2})K_{o})^{2}+ 8 K_{hop} K_{o}}\Bigr]\label{kmct}
392: \end{eqnarray}
393: \noindent
394:
395: In the absence of hopping the above expression reduces to,
396:
397: \begin{equation}
398: K_{MCT}=(1-a^{2}\lambda_{2})K_{o} \label{kmctred}
399: \end{equation}
400: \noindent
401: Thus we see that the self coupling term leads to an increases in the
402: timescale of relaxation when compared with the bare relaxation timescale $K^{-1}_{o}$.
403: Eq.\ref{kmctred} further predicts that $K_{MCT}$ goes to zero or the
404: relaxation time approaches infinity
405: as $a^{2}\lambda_{2}$ approaches one. Thus the analysis shows that in the
406: absence of hopping strict localization takes place at
407: $a^{2} \lambda_{2}=1$. From previous studies we know at $T_{c}$, $a=1/2$
408: and $\lambda_{2}=4$ \cite{sbp,leu,beng},
409: thus at $T=T_{c}$, $a^{2} \lambda_{2}$ indeed becomes unity.
410:
411: We will now analyze $K_{MCT}$ in the presence of hopping
412: in three different regions, above
413: $T_{c}$, at and around $T_{c}$ and below $T_{c}$.
414:
415: \subsection{Above $T_{c}$}
416:
417: Above the mode coupling transition temperature $T_{c}$, it was shown by
418: explicit calculation that the timescale of
419: hopping dynamics is so much longer than direct relaxation
420: in the system that it can be neglected \cite{sbp}.
421: The expression of $K_{MCT}$ then reduces to,
422: \begin{equation}
423: K_{MCT}\simeq (1-a^{2}\lambda_{2})K_{o}
424: \end{equation}
425: \noindent
426: Thus in accord with our earlier numerical calculation \cite{sbp}
427: above the mode coupling transition temperature the hopping does not have
428: significant effect on the $\alpha$ relaxation timescale.
429:
430:
431: \subsection{At and around $T_{c}$}
432:
433: From the earlier studies we know that
434: at the transition temperature, $a^{2}\lambda_{2}=1$ \cite{sbp,leu}.
435: Thus eq.\ref{kmct}
436: reduces exactly to,
437:
438: \begin{equation}
439: K_{MCT}=-K_{hop}
440: +\sqrt{K^{2}_{hop}+ 2 K_{hop} K_{o}}
441: \end{equation}
442: \noindent
443: In the above expression if we take some reasonable value for $K_{o}$ and
444: $K_{hop}$, we find that,
445:
446: \begin{equation}
447: K_{MCT}\simeq\sqrt{2 K_{hop} K_{o}} \label{kmcttc}
448: \end{equation}
449: \noindent
450: Thus we find that $K_{MCT}$ has a nonlinear dependence on $K_{hop}$ which is
451: also in accord with our earlier numerical calculations \cite{sbp}. The coupling
452: of the hopping dynamics with the short time part of the MCT dynamics
453: (liquid like dynamics) leads to this nonlinear
454: dependence. Thus we find that coupling between the short time part of the
455: MCT dynamics and hoppings leads
456: to an MCT part of the structural relaxation timescale
457: which is much faster than the hopping
458: timescale. This is a critical effect and is found at and near $T_{c}$.
459: This result corroborates our earlier findings that a
460: single hopping event leads to many continuous
461: diffusion events which are the primary means of structural relaxation
462: in this region.
463: In the next subsection we will find that this scenario changes as we go lower
464: and lower in temperature.
465:
466:
467: \subsection{Below $T_{c}$}
468:
469: Much below the transition temperature, $a^{2}\lambda_{2}>>1$.
470: Eq.\ref{kmct} can be rewritten as,
471: \begin{eqnarray}
472: K_{MCT}=\frac{1}{2}&\Bigl[&((a^{2}\lambda_{2}-1)K_{o}+2K_{hop})\\\nonumber
473: &&\times{\Bigl\{} -1+
474: \sqrt{1+\frac {8 K_{hop} K_{o}}{((a^{2}\lambda_{2}-1)K_{o}+2K_{hop})^{2}}}
475: {\Bigr\}}
476: \Bigr]\label{kmct1}
477: \end{eqnarray}
478: \noindent
479: Since at low temperatures $((a^{2}\lambda_{2}-1)K_{o}+2K_{hop})^{2}
480: >> 8 K_{hop} K_{o}$ thus we can write,
481: \begin{equation}
482: K_{MCT}=\frac {2 K_{hop} K_{o}}{((a^{2}\lambda_{2}-1)K_{o}+2K_{hop})}
483: \end{equation}
484: \noindent
485: Now if we further take into consideration that $(a^{2}\lambda_{2}-1)K_{o} >>
486: 2K_{hop}$ then the above equation reduces to,
487: \begin{equation}
488: K_{MCT}\simeq \frac{2 K_{hop}}{(a^{2}\lambda_{2} -1)}\label{kmctbelow}
489: \end{equation}
490: \noindent
491: Thus much below the transition temperature the timescale of the MCT dynamics
492: and also the total dynamics becomes slaved to the hopping timescale.
493: Analyzing eq.\ref{kmctbelow}, an important observation can be made about
494: the relaxation timescale. It is known that as we lower the temperature both
495: $\lambda_{2}$ and $a$ increases. Thus the denominator in
496: eq.\ref{kmctbelow} will increase as we lower the temperature. From the
497: analysis in the earlier subsection we know that initially to
498: start with, below $T_{c}$ $K_{MCT}>> K_{hop}$, then as we keep lowering
499: the temperature then depending on the temperature dependence of $\lambda_{2}$
500: and $a$ ,$K_{MCT}\simeq K_{hop}$. But as we further lower the temperature then
501: slowly $K_{MCT}<< K_{hop}$. Thus in this regime although there will be hopping
502: induced continuous diffusion but the primary mode of the structural relaxation
503: becomes direct activated hopping itself.
504:
505: \section {Interplay between hopping and MCT dynamics at and below $T=T_{c}$
506: : Numerical results}
507:
508: Continuing from our analysis where both MCT and hopping dynamics are assumed
509: to be exponential,
510: here we will
511: present some numerical results for the general case where both hopping and
512: MCT dynamics can be stretched. For these general cases we will try to
513: understand the effect of hopping on the MCT dynamics.
514:
515: For the calculation we need to solve eq.\ref{fqtmct} numerically.
516: It is well known that due to the self consistent nature of the equation,
517: its numerical solution becomes a nearly Herculean task around
518: the mode coupling
519: transition temperature. In the presence of hopping due to the disparate
520: timescales present in the system and the convolution in eq.\ref{fqtmct}
521: which involves all these timescales, the time of calculation increases
522: many fold. However the scheme proposed by Fuchs {\it et al}
523: \cite{hofac} allows the calculation to be done much faster.
524: Both $\phi(t)$
525: and $\phi_{MCT}(t)$ vary more slowly for longer times than they do
526: at short times.
527: The essential idea involved in the scheme is to separate the slow and the fast
528: variables and treat them differently in the convolution.
529: The short time part of $\phi(t)$ and $\phi_{MCT}(t)$ are calculated exactly
530: with very small stepsize and they are then used as input to carryout
531: the calculation for the long time part of the same. We have exactly
532: followed the scheme presented
533: in reference \cite{hofac} except for a minor modification (one extra term)
534: in eq.29 of
535: reference \cite{hofac} when the integration timestep is an odd number.
536: In our calculations the total timestep N=1000.
537:
538:
539:
540: For this study we have
541: picked three points on the $\lambda_{1}$, $\lambda_{2}$ plane.
542: In all the cases
543: to understand the temperature effect the values
544: of $\lambda_{1}$, $\lambda_{2}$ were varied according to the expressions given
545: bellow,
546:
547: \begin{eqnarray}
548: \lambda_{1}= \frac {(2 \lambda -1)}{\lambda^{2}}+ \epsilon \frac {\lambda}
549: {(1+(1-\lambda)^{2})} \label{l1}
550: \end{eqnarray}
551:
552: \begin{eqnarray}
553: \lambda_{2}= \frac {1}{\lambda^{2}}+ \epsilon \frac {\lambda (1-\lambda)}
554: {(1+(1-\lambda)^{2})}\label {l2}
555: \end{eqnarray}
556: \noindent
557: where the value of $\lambda$ determines the value of
558: $\lambda_{1}$ and $\lambda_{2}$ at $T=T_{c}$. $\epsilon$ is a measure of
559: the distance from the MCT transition temperature. As in our earlier model
560: calculation, \cite{sbp} the values of $\Omega^{\star}_{o}$ and $\gamma^{\star}$
561: are kept unity and the scaling time as 1ps. In the numerical calculations,
562: to understand the temperature effect although we have changed the
563: values of $\lambda_{1}$ and $\lambda_{2}$ but we have kept all the other
564: parameters constant including the hopping barriers.
565: For the calculation of the hopping part we have used eq.\ref{hopdist}
566: with a Gaussian distribution of barriers.
567: According to the RFOT theory, the mean barrier height and thus the hopping
568: timescale should change
569: with temperature \cite{xiawoly,lubwoly}.
570: In its simplest form without taking barrier softening into consideration the
571: mean barrier height can be written as, $\Delta F/k_{B}T=32 k_{B}/s_{c}=
572: \frac {32 k_{B}}{\Delta c_{p}} \frac {T_{K}}{(T-T_{K})}$, where $s_{c}$ is the
573: configurational entropy, $\Delta c_{p}$ is the jump in the specific heat and
574: $T_{K}$ is the Kauzmann temperature \cite{xiawoly,lubwoly}. However in
575: this present calculation to clearly assign any change in dynamics due to
576: the change in $\lambda_{1}$ and $\lambda_{2}$ value we have kept the
577: mean of the distribution
578: fixed at about 8.8 $k_{B}T$.
579: As mentioned earlier the stretching in the hopping dynamics
580: is determined by the width of the Gaussian distribution of the barriers. The
581: broader the distribution the more stretched is the dynamics. We have varied
582: the width of the distribution such that the $\beta^{static}_{hop}$ varies
583: from 0.2 to 0.8. Along with the $\beta^{static}_{hop}$ value the timescale of hopping
584: also changes, which has been taken into account in our calculation.
585:
586: For all the cases the $\beta_{total}$ values are plotted against $\beta^{static}_{hop}$
587: for different $\epsilon$ values, where $\beta_{total}$ is the stretching
588: parameter for $\phi(t)$ and $\beta^{static}_{hop}$ is the same for $\phi_{hop}(t)$.
589:
590: \subsection {Case 1: $\lambda=0.5$}
591:
592: In the first case we consider an example when $\lambda=0.5$.
593: This value of $\lambda$
594: implies that at $T=T_{c}$, $\lambda_{1}=0$ and $\lambda_{2}=4$ and the
595: MCT dynamics just above $T_{c}$ is exponential. For this case we vary
596: $\epsilon$ from $0-1$.
597: In figure 2, the $\beta_{total}$ values are plotted against $\beta^{static}_{hop}$
598: for the three $\epsilon$ values. As expected for $\epsilon=0$ the
599: $\beta_{total} \simeq 1$. This is because although without hopping structural
600: relaxation is arrested but once hopping is present the relaxation is
601: dominated by the MCT dynamics. As we increase the epsilon value we find that
602: the effect of hopping is stronger and the dynamics gets stretched. But we also
603: notice that for very low $\beta^{static}_{hop}$ value the dynamics is less stretched
604: than the static barrier distribution would indicate.
605: This is because lower $\beta^{static}_{hop}$ value implies a broader
606: barrier height distribution which means we populate both lower and
607: higher barrier heights.
608: The small barrier hoppings actually
609: couple to the liquid like part of the MCT dynamics (whose timescale
610: is given by $K_{o}^{-1}$) and the relaxation is
611: faster and dominated by
612: this liquid like MCT dynamics which has a much shorter timescale.
613: As we increase the $\epsilon$ value MCT gets more slaved to hopping and thus
614: the time scale of relaxation increases and the liquid like MCT dynamics becomes
615: less important. If we further lower the temperature (or increase the $\epsilon$
616: value) we would find that even for small $\beta^{static}_{hop}$ values the total
617: dynamics is determined primarily by hopping.
618:
619: \begin{figure}
620: \epsfig{file=fig2.eps,height=6cm,width=7cm,angle=0}
621: \caption{The stretching parameter for the total structural relaxation,
622: $\phi(t)$, $\beta_{total}$ is plotted against the same for $\phi_{hop}(t)$,
623: $\beta^{static}_{hop}$ for three different $\epsilon$ values. The solid circle is for
624: $\epsilon=0$, the solid square is for $\epsilon=0.5$ and the solid diamond
625: is for $\epsilon=1$. In this plot the $\lambda=0.5 $
626: which means that the MCT dynamics without hopping above $T_{c}$ is
627: exponential.}
628: \end{figure}
629: The interplay of hopping with MCT nonlinearities can be thought of as
630: a quantitative formulation of "facilitation effects" \cite{xiawoly}.
631: As Xia and Wolynes pointed out hopping events interact if they occur
632: near each other. This is accounted for by the MCT nonlinearity.
633: In the Xia-Wolynes treatment the corresponding effect led to the
634: cutoff of the relaxation time distribution, on the slow side,
635: owing to the renewal of a mosaic cell's environment through hops.
636: This resulted in an increased $\beta$ from that obtained from
637: the static Gaussian model,
638: as occurs here too \cite{xiawoly}.
639:
640:
641:
642:
643: \begin{figure}
644: \epsfig{file=fig3.eps,height=6cm,width=7cm,angle=0}
645: \caption{The stretching parameter for the total structural relaxation,
646: $\phi(t)$, $\beta_{total}$ is plotted against the same for $\phi_{hop}(t)$,
647: $\beta^{static}_{hop}$ for three different $\epsilon$ values. The solid circle is for
648: $\epsilon=0$, the solid square is for $\epsilon=0.5$ and the solid diamond
649: is for $\epsilon=1$. In this plot the $\lambda=0.568 $
650: which means that the MCT dynamics without hopping above $T_{c}$ is stretched
651: with stretching parameter, $\beta_{MCT}=0.8$. }
652: \end{figure}
653:
654: \subsection {Case 2: $\lambda=0.568$}
655:
656: In this case we consider that $\lambda=0.568$. This value of $\lambda$
657: implies that at $T=T_{c}$, $\lambda_{1}=.4215$ and $\lambda_{2}=3.09958$
658: and the MCT dynamics, (without hopping) just above $T_{c}$, would already
659: be stretched
660: with $\beta_{MCT}=0.8$. For this case we vary
661: $\epsilon$ from $0-1$.
662: In figure 3, the $\beta_{total}$ values are plotted against $\beta^{static}_{hop}$
663: for the three $\epsilon$ values. The results are similar to those obtained
664: for the first case. For $\epsilon=0$ the
665: $\beta_{total} \simeq 0.8 $ which implies that the total dynamics is
666: still determined primarily by the MCT dynamics. We also find that for
667: high $\beta^{static}_{hop}$ values, as $\epsilon$ increases,
668: the dynamics gets more
669: and more dominated by the hopping dynamics. Nevertheless,
670: for small $\beta^{static}_{hop}$
671: the scenario is a little different. For smaller $\epsilon$ values
672: the low barrier hoppings couple to the liquid like part of the MCT dynamics
673: and the dynamics is less stretched and also much faster. However,
674: as we increase
675: $\epsilon$, MCT dynamics gets more and more slaved to hopping dynamics and the
676: MCT relaxation timescale becomes proportional to the hopping relaxation
677: timescale
678: (similar to that shown in Eq.\ref{kmctbelow}). Thus the effect of the coupling
679: of low barrier hoppings with the liquid like part of the MCT dynamics, on the
680: total MCT dynamics, reduces. If we further lower the temperature then we would find
681: that even for $\beta^{static}_{hop}=0.2$ the total dynamics follows the hopping dynamics.
682:
683: The results in case 1 and case 2 look quite similar but more
684: detailed observation reveals that the value of $\beta^{static}_{hop}$,
685: where $\beta_{total}$
686: begins to increase, is smaller for case 2 (where MCT dynamics itself is
687: more stretched) than it is in case 1. As discussed earlier, the reason
688: $\beta_{total}$ increases for small $\beta^{static}_{hop}$ is that the
689: small barrier
690: hopping gets coupled to the liquid like part of the MCT dynamics
691: allowing the
692: total structure to relax. Now in case 2, the MCT dynamics is
693: itself stretched thus
694: the effect of the small barrier hopping on the MCT dynamics will be much less
695: effective when compared to case 1. This trend becomes clearer when we
696: study the next case where the MCT dynamics is much more stretched.
697:
698:
699: \subsection {Case 3: $\lambda=0.75$}
700:
701: In this case we consider that $\lambda=0.75$. This value of $\lambda$
702: implies that at $T=T_{c}$, $\lambda_{1}=.889$ and $\lambda_{2}=1.778$
703: and the MCT dynamics (without hopping) just above $T_{c}$ would already
704: be stretched
705: with $\beta_{MCT}=0.5$. For this case we vary
706: $\epsilon$ from $0-1$.
707: In figure 4, the $\beta_{total}$ values are plotted against $\beta^{static}_{hop}$
708: for the three $\epsilon$ values. The results are similar to that obtained
709: for case 1 and case 2 for $\epsilon=0$. But for higher $\epsilon$ values unlike
710: in case 1 and 2,
711: $\beta_{total}$ continuously decreases with $\beta^{static}_{hop}$. This is because
712: as discussed before, since MCT dynamics is already stretched, the structural
713: relaxation due to low barrier hopping is less effective.
714: Also note that
715: the $\beta_{total}$ decreases with $\beta^{static}_{hop}$ value but it is
716: neither equal to $\beta^{static}_{hop}$, nor equal to the pre-transition
717: $\beta_{MCT}$ value. At these temperatures although MCT dynamics is slaved to hopping
718: but both the channels of relaxation are almost equally effective. If we
719: compare the $\beta_{total}$ values for $\epsilon=$ 0.5 and 1, we will find that
720: in most of the cases $\beta_{total}$ for $\epsilon=1$ has a lower value.
721: This is because at lower temperatures MCT dynamics becomes a less effective
722: relaxation channel.
723: At further lower temperatures (higher $\epsilon$ values) $\beta_{total}$ will
724: follow $\beta^{static}_{hop}$ more closely.
725:
726:
727: \begin{figure}
728: \epsfig{file=fig4.eps,height=6cm,width=7cm,angle=0}
729: \caption{The stretching parameter for the total structural relaxation,
730: $\phi(t)$, $\beta_{total}$ is plotted against the same for $\phi_{hop}(t)$,
731: $\beta^{static}_{hop}$ for three different $\epsilon$ values. The solid circle is for
732: $\epsilon=0$, the solid square is for $\epsilon=0.5$ and the solid diamond
733: is for $\epsilon=1$. In this plot the $\lambda=0.75 $
734: which means that the MCT dynamics without hopping above $T_{c}$ is stretched
735: with stretching parameter, $\beta_{MCT}=0.5$.}
736: \end{figure}
737: \section{Concluding Remarks}
738:
739: MCT and the RFOT theory provide a unified
740: theory of relaxation over the whole temperature plane \cite{sbp}. Even
741: without dynamical heterogeneity the theory
742: successfully predicts the decay of the structural relaxation below mode
743: coupling transition temperature, $T_{c}$,
744: confirming there is no strict localization transition at $T_{c}$.
745: Without dynamical heterogeneity of the instantons the coupled theory lead
746: to an exponential $\alpha$ relaxation, but in the laboratory
747: the $\alpha$ relaxation is generally stretched.
748: In the present article we examined both a two parameter
749: MCT model \cite{gotze} and more realistically one that also included
750: a barrier height distribution that gives rise to stretching in the
751: hopping dynamics by itself \cite{xiawoly}.
752:
753: The study has been carried out for different stretching parameters of the MCT
754: dynamics, that is by changing $\lambda$ values in the $F_{12}$ model
755: and also for different stretching parameters of the hopping dynamics obtained
756: by changing the width of the distribution of barrier heights.
757:
758:
759: To summarize, the main conclusions of the present work is that the
760: continuous dynamics, described here within the $F_{12}-$MCT formalism
761: , and the
762: activated hopping dynamics, described here using RFOT theory, interact
763: in a non-linear fashion to give rise to dynamical features which are
764: distinct from both. MCT by itself of course cannot describe dynamics
765: below its
766: critical temperature. We find that hopping facilitates the continuous dynamics
767: channel and in the process the effects of hopping on the relaxation
768: decreases. Thus, one finds the stretching parameter arising
769: solely from distribution of hopping barrier energies in RFOT is increased
770: by the mode coupling terms.
771: It is also found that when MCT dynamics is less stretched then the
772: effect of hopping
773: on the MCT dynamics is less pronounced and the hopping dominated regime
774: moves to a lower temperature. On the other hand,
775: for more stretched MCT dynamics, due to
776: the larger overlap of MCT and hopping timescales, hopping begins to dominate
777: at a higher temperature.
778:
779: We have already mentioned the need for using a more complex hopping kernel
780: than that given by Eqs.3-5. In particular, one needs to include the effects of
781: mode coupling softening on the barrier height distribution.
782: This is a feed-back effect of unleashing the mode coupling
783: relaxation channels due to hopping, on the barrier height
784: distribution itself. This non-linear feed-back
785: is expected to shift the distribution to lower barrier heights and in turn
786: accelerate mode coupling relaxation which can further enhance hopping. The whole
787: system of equations needs to be solved self-consistently. To achieve this,
788: we need to understand more quantitatively the effects of softening
789: on the barrier height distribution.
790: %*******************************************
791:
792: {\bf ACKNOWLEDGEMENT}
793:
794: This work was supported in parts from NSF (USA) and DST (India).
795:
796:
797:
798:
799: \begin{references}
800:
801: \bibitem{sbp} S. M. Bhattacharyya, B. Bagchi, P. G. Wolynes, Phys. Rev. E
802: {\bf 72} 031509 (2005).
803:
804: \bibitem{gotze} W. Gotze and L. Sjogren, Z. Phys. B- Cond. Mat., {\bf 65},
805: 415 (1987);W. Gotze and L. Sjogren, J. Phys. C:Solid State Phys. {\bf 21}, 3407 (1988).
806:
807:
808: \bibitem{das}S. P. Das and G. F. Mazenko, Phys. Rev. A, {\bf 34}, 2265 (1986).
809:
810:
811: \bibitem{kirk} T.R. Kirkpatrick and P.G. Wolynes, Phys. Rev. {\bf E 35},
812: 3072 (1987).
813:
814: \bibitem{biroli} J. P. Bouchaud and G. Biroli, J. Chem. Phys. {\bf 121},
815: 7347 (2004).
816:
817:
818: \bibitem {sarikajcp}S. Bhattacharyya, A. Mukherjee, and B. Bagchi,
819: J. Chem. Phys.{\bf 117}, 2741(2002); S. Bhattacharyya and B. Bagchi, Phys. Rev. Lett. {\bf 89},
820: 025504-1(2002); A. Mukherjee, S. Bhattacharyya and B. Bagchi, J. Chem. Phys.
821: {\bf 116}, 4577 (2002);
822:
823:
824: \bibitem{xiawoly} X. Xia and P. G. Wolynes, Phys. Rev. Lett. {\bf 86}, 5526
825: (2001);Proc. Natl. Acad. Sci. U. S. A. {\bf 97}, 2990 (2000).
826:
827:
828: \bibitem{lubwoly}V. Lubchenko and P. G. Wolynes, J. Chem. Phys. {\bf 119},
829: 9088 (2003);
830:
831: \bibitem{leu} E. Leutheusser, Phys. Rev. A {\bf 29}, 2765 (1984).
832:
833: \bibitem{beng} U. Bengtzelius, W. Gotze, and A. Sjolander, J. Phy. C {\bf 17},
834: 5915 (1984).
835:
836: \bibitem{hofac}M. Fuchs, W. Gotze, I. Hofacker and A.Latz, J. Phys: Condensed Matter {\bf 3}, 5047 (1991).
837:
838: \end{references}
839: \end{document}
840:
841: