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 {Bridging the Gap Between the Mode Coupling and the Random First
10: Order Transition Theories of Structural Relaxation in Liquids}
11:
12: \author{ Sarika Maitra Bhattacharyya$^{\dagger}\footnote
13: {Electronic mail~:sarika@sscu.iisc.ernet.in}$,
14: Biman Bagchi$^{\dagger}$\footnote{Electronic mail~:bbagchi@sscu.iisc.ernet.in} \\
15: and \\
16: Peter G. Wolynes$^{\ddagger}$\footnote
17: {Electronic mail~:pwolynes@chem.ucsd.edu}}
18:
19: \affiliation{$^{\dagger}$ Solid State and Structural Chemistry Unit,
20: Indian Institute of Science, Bangalore 560 012, India.\\
21: $^{\ddagger}$Department of Chemistry and Biochemistry,
22: University of California at San
23: Diego, La Jolla, California 92093-0371}
24:
25:
26: \begin{abstract}
27: A unified treatment of structural
28: relaxation in a deeply supercooled glassy liquid
29: is developed which extends the existing
30: mode coupling theory (MCT)
31: by incorporating the effects of activated
32: events by using the concepts from the
33: random first order transition (RFOT) theory. We show how
34: the decay of the dynamic
35: structure factor is modified by localized activated events
36: (called instantons) which lead to the spatial reorganization of molecules
37: in the region where the instanton pops up.
38: The instanton vertex added to the usual MCT depicts the
39: probability and consequences of such an event
40: which can be derived from the random first
41: order transition theory. The vertex is proportional to $exp(-A/s_{c})$ where $s_{c}$
42: is the configurational entropy.
43: Close to the glass transition temperature, $T_{g}$,
44: since $s_{c}$ is diminishing, the activated process
45: slows beyond the time window and this eventually leads to an arrest of the
46: structural relaxation as expected for glasses. The combined treatment describes
47: the dynamic structure factor in deeply supercooled liquid fairly well, with
48: a hopping dominated decay following the MCT plateau.
49: \end{abstract}
50:
51: \maketitle
52: %\section{Introduction}
53: Inelastic neutron scattering has given detailed structural information on
54: how transport in liquids changes as they are cooled in the glassy state
55: \cite{book, fayer}. Many
56: of the structural details are well described by mode coupling theory (MCT)
57: but
58: that approach does not describe well the dynamics of
59: the deeply supercooled state
60: \cite{rajesh25,leu,beng,gotze,sjogren,sarirev}.
61: The random first order transition (RFOT) theory of glass explains this
62: lacuna of MCT \cite{xiawoly,Lubwoly,lubwoly2}.
63: In this paper we show how RFOT theory can bridge the gap between the
64: onset of glassy dynamics at high temperatures and the low temperature behavior
65: and thus provide a unified view of the
66: structural dynamics in deeply supercooled region.
67: Over the last few decades, experimental and theoretical
68: studies on the phenomenological aspects of
69: the glass transition have been the subject of a vast scientific literature.
70: A glass forming liquid is characterized by its ability to circumvent
71: crystallization on rapid cooling to a temperature well below its freezing
72: temperature $T_{m}$.
73: In contrast to crystal formation what is most remarkable about solidification
74: into an amorphous state is that the structure
75: changes accompanying vitrification
76: are either very small or unobservable.
77: However, on passage through supercooled regime the
78: dynamic and thermodynamic behavior of the liquid
79: exhibits a number of anomalies as it eventually transforms to an
80: amorphous solid, called a glass.
81:
82:
83: The density-density correlation function at finite wave vector probed by
84: neutron scattering, $\phi_{k}(t)$ for a system
85: approaching glass transition first
86: decays via a fast microscopic process which is followed by a plateau in the
87: intermediate time. The dynamics while approaching the plateau is
88: non-exponential, power law in nature and is called $\beta$ relaxation.
89: The time scale of the $\beta$ relaxation is predicted to show a
90: power law divergence. The
91: decay from the plateau known as the initial $\alpha$ relaxation
92: is also given by a power law, known as von Schweidler law. The
93: exponent in the $\beta$ regime and the initial $\alpha$ relaxation
94: are different in value but related to each other. The slow $\alpha$
95: relaxation which appears at long times is described by an
96: stretched exponential or Kohlrausch-Williams-Watts (KWW)
97: function \cite{fayer}.
98:
99: The thermodynamic and kinetic anomalies at $T_{g}$ are known experimentally
100: to be correlated.
101: One experimentally finds a sharp rise in the measured heat capacity of a liquid
102: during heating which follows prior cooling at a constant rate\cite{heatcp},
103: the cycle being well extended on either sides of the glass transition region.
104: The over shoot of the heat capacity is taken to be a thermodynamic
105: signature of a glass
106: to liquid transition. However, what lies at the heart of glass
107: transition is the
108: dramatic slowing down of molecular motion that occurs progressively
109: on cooling after the plateau in the inelastic structure factor appears.
110: This is manifested by a phenomenal increase in shear viscosity $\eta$
111: (also the characteristic structural relaxation time $\tau$) by several
112: orders of magnitude for a relatively modest decrease of temperature
113: (by few tens of degrees). The glass transition seen in the laboratory
114: is in fact best described
115: as a kinetic phenomenon which marks the falling out of the equilibrium
116: due to the inability of molecular rearrangement within the experimental
117: timescale as temperature T is lowered. A conventional definition
118: of the transition temperature $T_{g}$ is the temperature
119: at which the viscosity, $\eta$
120: reaches a value $10^{13} P$.
121:
122: The mode coupling theory makes detailed predictions of the
123: temperature
124: dependence of the time evolution of the density-density correlation
125: function \cite{rajesh25}.
126: The emergence of a plateau in the neutron scattering
127: is one of the important predictions of the ideal mode coupling theory.
128: At a quantitative level both the emergence of a plateau in the dynamic
129: structure factor at high temperature and the exponential slowing of
130: the dynamics
131: at low temperature are also explained by the random first order transition
132: (RFOT) theory \cite{xiawoly,Lubwoly}.
133: In the high temperature limit, RFOT contains the essential elements of MCT as a
134: stability limit.
135: While idealized MCT can also correctly predict the power law dynamics in
136: the $\beta$ and the initial $\alpha$
137: relaxation regime and the relation between the two exponents in these
138: two regimes, it fails to explain the slow $\alpha$ relaxation
139: which appears at long times. MCT strictly applied, predicts
140: a premature arrest of
141: the decay of the density-density correlation function.
142: The dynamical plateau of MCT was contemporaneously obtained
143: by a variety of static
144: approaches based on self-consistent phonon calculations and density functional
145: theory which form the origin of the RFOT theory \cite{stossel1,stossel}.
146: MCT would appear to predict a dynamic arrest of the
147: liquid structural relaxation, arising solely from a nonlinear
148: feedback mechanism on the memory function. This feedback mechanism
149: was first proposed
150: by Geszti \cite{ges}
151: to describe the growth of viscosity in a previtrification region
152: and involves a self-consistent calculation of the dynamical correlation
153: function and transport coefficients. In RFOT theory one sees the dynamic
154: arrest is connected with the emergence of a mean field free energy landscape.
155:
156: According to the MCT, the essential driving force for glass
157: transition is the slowing down of the
158: density fluctuations near the wavenumber at $q\simeq q_{m}$ where the static
159: structure factor is sharply peaked. Thus, prima facie the glass transition does not appear
160: to be caused
161: by a small wavenumber infra red singularity observable in the static
162: structure factor, but rather it is a phenomenon
163: where the intermediate wavenumbers are important, (although a diverging
164: length associated with the four point correlations is expected in the mean
165: field RFOT theory \cite{kw1987,biroli}). There is a softening
166: of the heat mode near $q \simeq q_{m}$.
167: Kirkpatrick and Wolynes \cite{kw1987} showed that this mode coupling
168: transition was equivalent to the transition to an aperiodic crystal
169: predicted by the quasi-static theories. Near the dynamical transition there
170: is a general slowing down of all the dynamical quantities,
171: and this slowing down
172: is most effectively coupled to wavenumbers near $q_{m}$.
173: The glass transition temperature predicted by MCT always exceeds the
174: laboratory $T_{g}$.
175: The explanation for this was pointed out by Kirkpatrick and
176: Wolynes \cite{kw1987}: Since there are exponentially many aperiodic structures
177: any single one of them is unstable to transformation to the other.
178: At $T_{c}$ the system is not frozen because activated events allow
179: interconnection between these structures.
180: Thus $T_{c}$ can be identified as a crossover temperature below which
181: activated motions occur in which groups of particles move in a
182: cooperative way over local barriers in the free energy landscape.
183: Computer simulation studies have amply demonstrated this change of transport
184: mechanism and have explicitly revealed the presence of such
185: hopping motions in deeply supercooled liquids when the usual convective
186: diffusive
187: motions cease to exist \cite{wang,arnab}. Simulations also
188: show that these hopping events relax the local stress in the system
189: \cite{anistress}.
190:
191:
192: These properties of the activated events below $T_{c}$ can be represented
193: by the random first order transition theory in a way superficially analogous
194: to the theory of nucleation at random first order transition (although there
195: are some differences).
196: Calculation of fluctuations and transport
197: in deeply supercooled or
198: glassy medium are most easily made where the reference is an
199: ideal glassy state and where the configurational entropy has vanished
200: \cite{xiawoly,kw,ktw}. The barriers for activated dynamics vanish as the
201: temperature of the system approaches the dynamical transition temperature
202: from below much like at a spinodal.
203: The barrier for activated events can be described as
204: a nucleation phenomenon where the free energy change is a sum of
205: the entropy change
206: due to the formation of an entropy droplet and a surface term with a radius
207: dependent surface energy \cite{xiawoly,Lubwoly}.
208: A more explicit view of the similarities and the differences from nucleation
209: appears in the recent work of Lubchenko and Wolynes and the elegant analysis
210: of Biroli and Bouchaud \cite{biroli}.
211: Deep in the supercooled regime, the nucleation free energy barrier is given
212: by the following simple expression
213: \begin{equation}
214: \Delta F^{\star}/k_{B}T = 32 k_{B}/s_{c} \label{barrier}
215: \end{equation}
216: where $s_{c}$ is the configuration entropy of a single moving bead i.e
217: equivalent spherical particle.
218: The coefficient $32$ arises from a specific microscopic calculation and
219: reflects the entropy cost of localizing a particle to vibrate in the cage,
220: explicitly observed in the plateau of the dynamic structure factor.
221: If one approximates
222: $s_{c}$ near the
223: Kauzmann temperature $T_{K}$ by $s_{c}=\Delta c_{P} (T-T_{K})/T_{K}$,
224: one then obtains the empirical Vogel-Fulcher-Tammann (VFT) law, with a relation
225: between the liquid's fragility and the heat capacity jump, $\Delta c_{P}$.
226: This expression agrees well with experimental results.
227: However, Eq.\ref{barrier} is an asymptotic form of the barrier, valid only
228: near
229: $T_{K}$ and as will be discussed later, if we take into account the
230: barrier softening effect near $T_{c}$ then the value of the barrier
231: will be considerably
232: smaller than that given by Eq.\ref{barrier} \cite{Lubwoly}. This effect is
233: found to be stronger as we move away from $T_{K}$.
234: Once the estimation of the barrier height is made the relaxation time
235: $\tau$ is calculated using, $\tau/\tau_{0}=exp(F^{*}/k_{B}T)$.
236: The random first order phase transition theory
237: also provides an explanation of the dominant source of non-exponentiality
238: in relaxation
239: dynamics in terms of
240: distribution of the configuration entropy of the activated droplet.
241: If one assumes a Gaussian distribution of the barrier, then the standard
242: deviation of the barrier height at $T_{g}$ also turns out to be
243: related to the specific
244: heat discontinuity of the supercooled liquid. This correlation between
245: nonexponentiality and $\Delta C_{p}$ is observed
246: in the laboratory \cite{angell}.
247:
248: In this article we use the concepts of mode coupling theory and random
249: first order transition theory to include both the convective diffusional motion
250: (given by MCT) and the hopping motion (described by RFOT) in the dynamics.
251: Thus the present treatment also correlates the kinetic anomalies
252: (like stretching of the relaxation time) with the thermodynamic
253: anomalies (jump in the heat capacity) primarily through the RFOT theory.
254: Here we first discuss the effect of hopping alone on the structural
255: density fluctuations.
256: The rate of hopping and the size of the region involved in hopping are
257: calculated from
258: RFOT theory \cite{Lubwoly}. It is shown that hopping opens up a new channel
259: for the structural relaxation.
260: Next we present the equation of motion of the density-density correlation
261: function where the diffusional and the hopping motions are considered as
262: two parallel channels for the structural relaxation in the liquid. In the same
263: spirit as in the case of the idealized MCT the new
264: equation of motion which now has hopping motion in it,
265: is calculated self-consistently with the longitudinal
266: viscosity. The feedback mechanism which gave rise to the divergence of the
267: viscosity in the idealized MCT \cite{leu}
268: is also present in our theory but due to the
269: presence of hopping, the divergence is shifted to a lower temperature where
270: the RFOT hopping rate now finally vanishes only
271: as the configurational entropy goes to zero.
272: In our proposed equation of motion the diffusional and the hopping motions are
273: considered as parallel channels for the density relaxation.
274: In the first approximate scheme we write the total density correlation
275: in the time plane and in the second approximate scheme
276: it is written in the frequency plane. We show that not only both
277: the approximations (although they appear to be quite different)
278: lead to similar
279: effects on the density-density correlation function, although.
280: the predicted longtime decays are somewhat different.
281:
282:
283: Before discussing our own theoretical scheme we would like to mention that
284: Das and Mazenko \cite{das} in their fluctuating nonlinear
285: hydrodynamic theory and later Gotze and Sjogren \cite{gotze,sjogren}
286: in their extended MCT
287: have elegantly demonstrated that in the relaxation kernel (memory function), in addition to the density contribution,
288: if a coupling to the current modes is also included,
289: then below $T_{c}$,
290: the dynamic structure factor decays from the plateau value in the long time.
291: Gotze and Sjogren \cite{gotze,sjogren} have
292: shown that this modification eventually leads to a parallel decay
293: channel of the dynamic structure factor which eliminates the ergodic to
294: non-ergodic transition at $T_{c}$. However, although the authors call this
295: extra
296: decay channel a hopping mode, there is no microscopic connection made in
297: those theories to the
298: single particle or collective hopping or
299: the system crossing over a free energy barrier in the
300: free energy landscape. In those treatments,
301: the strength and the timescale of this so
302: called
303: hopping mode was determined by fitting to experimental results, and an
304: Arrhenius temperature dependence of the rate of "hopping" was assumed
305: in this fitting.
306:
307:
308:
309: \section{Theoretical Scheme}
310: We propose a unified structural description
311: covering the whole temperature regime, from the
312: high temperature microscopic dynamics described by collisional physics
313: to the low temperature collective many particle activated dynamics.
314: The mode coupling theory analyzes liquid state dynamics by starting
315: from a gas-like mode of transport but including correlated collisions and
316: back flow. MCT does a good job describing
317: structural correlations observed by neutron scattering in the normal
318: liquid regime. MCT can also describe the dynamics approaching
319: the supercooled regime but fails below $T_{C}$,
320: where activated processes also contribute substantially.
321: A deeply supercooled liquid explores a very rugged underlying free energy
322: surface. The dynamics
323: within each well may still be described by the idealized MCT
324: which does not take
325: into account the possibility of hopping between wells.
326: Activated events represent essential singularities in the noise strength
327: and thus are not contained in the perturbative MCT.
328: To include the effect of activated events
329: on dynamical structure factor, we need to calculate two different
330: quantities. First, we need to calculate the effect
331: such an event (which we call "popping up of an instanton") would have on the
332: dynamic structure factor. The impact of such an instanton depends
333: both on its size, that is, how many molecules are involved in the
334: activated event and
335: on how far particles are displaced by an activated transition.
336: In RFOT theory, these quantities are the dynamical correlation length
337: $\xi$ and the Lindemann length $d_{L}$.
338: Second, we also need to know the
339: probability of such an activated event occurring.
340: We can estimate both this probability and the other quantities from
341: the random first order transition theory using thermodynamic and structural
342: information alone.
343:
344: A similar underlying physical picture to that presented here has been
345: put forward by Fuchizaki and Kawasaki \cite{kawasaki} using what they
346: called dynamic density functional theory
347: (DDFT) to go beyond the MCT.
348: In this coarse-grained description one considers the time evolution
349: of the density field on a lattice.
350: The dynamics is supposed to be given
351: by a master equation and the free energy functional is taken from
352: Ramakrishna-Yussouff (RY) theory \cite{ry}. As shown by Singh, Stoessel and
353: Wolynes \cite{stossel}, RY theory can produce multiple minima.
354: Fuchizaki and Kawasaki have shown that
355: transitions between such minima allow
356: the
357: structural relaxation of the coarse grained
358: density field. They show these transitions give rise to a stretched
359: exponential relaxation.
360: In their calculation,
361: the overlap between the time
362: dependent density field and the reference one of the same lattice site,
363: takes a constant
364: value, close to unity, in the liquid
365: but in the supercooled state, the overlap is less than
366: one and decreases via discrete jumps which they have connected to activated
367: hopping transition over free-energy barriers. They have
368: further defined an order parameter which is the overlap of the
369: density fields at different lattice sites. The distribution of the order
370: parameter, in the normal liquid regime is sharply peaked around unity
371: but at higher density the distribution is broad and peaked around a value
372: less than unity. The broad distribution reflects the multi-minima
373: aspects of the free-energy landscape.
374:
375: The scheme presented by FK was implemented numerically and
376: does not give as explicit picture of the microscopic
377: mechanism of hopping as the current treatment does.
378: Moreover, due to the temporal coarse graining the
379: initial decay of the dynamic structure factor, $\phi(t)$
380: and the subsequent plateau are missed in their analysis. In addition,
381: their analysis did not highlight the
382: connection with the
383: configurational entropy of the system.
384: In contrast the present unified theoretical treatment provides a $\phi(t)$
385: which produces a correct behavior over the full
386: time and temperature plane, and also provides an analytical expression of
387: the dynamic structure factor where the hopping alone is considered from a
388: microscopic point of view. In the following section, we
389: find $\phi_{hop}$ or the effect of hopping on the density relaxation.
390: Subsequently we incorporate $\phi_{hop}(t)$ in the total $\phi(t)$
391: in a self-consistent way.
392:
393:
394: \section{Effect of hopping on density relaxation}
395: Here we describe the hopping in terms of the popping up of a single
396: instanton which is an
397: activated event whose probability is obtained from the random first order
398: transition theory.
399: Let us say that an instanton pops up at a position $R$.
400: We will take this region of influence to be spherical on the average.
401: Near $T_{A}$, good arguments suggest the region actually
402: is fractal or string-like \cite{wolyunpub}. But the effect will ultimately
403: be spherically averaged anyway
404: as we shall see below.
405: The particles within the sphere of radius $\xi$ around the position $R$
406: will be displaced by a Lindemann length, $d_{l}$.
407: Now the change in density $\rho(r)$ due to a
408: single instanton popping up can be written as,
409: \begin{eqnarray}
410: \delta \rho^{new}({\bf r},t+\delta t)&=& \delta \rho({\bf r},t )+
411: \theta ((r-R)<\xi)\nonumber\\
412: &&
413: \times \Biggl[
414: \int dt^{\prime}
415: \int_{{\cal D}(R)}
416: d{\bf r}^{\prime} \:\delta \rho({\bf r,}^{\prime}, t^{\prime}) \nonumber\\
417: &&~~~~~
418: \times G ({\bf r},{\bf r}^{\prime},t-t^{\prime})
419: \theta((r^{\prime}-R)<\xi)\nonumber\\
420: &&~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
421: - \delta \rho ({\bf r},t) \Biggr]
422: \end{eqnarray}
423: \noindent
424: Where $\theta(({\bf r} -{\bf R})<\xi)$ provides the effect of the
425: instanton felt at the position ${\bf r}$, provided it
426: is within the radius $\xi$. $G ({\bf r}, {\bf r}^{\prime}t-t^{\prime})$ is the Greens
427: function which determines the effect of instanton
428: in moving particles from
429: position ${\bf r}$ to a new one ${\bf r}^{\prime}$,
430: typically Lindemann length
431: away during time $t$ and
432: $t+\delta t$. ${\cal D}(R)$ determines
433: the region where the effect of the instanton is felt.
434:
435:
436: Since a given instanton pops up at a specific location the transformation
437: is not
438: translationally invariant, but when averaged over the droplet location
439: and orientation we restore translational and rotational invariance,
440:
441: \begin{eqnarray}
442: \overline {\delta {\rho}}^{new}({\bf r},t+\delta t)&=& \delta \rho({\bf r},t) +
443: \frac{1}{V}
444: \int d{\bf R}\:
445: \theta (({\bf r}-{\bf R})<\xi) \nonumber\\
446: &&\times \Biggl[
447: \int dt^{\prime} \int_{{\cal D}(R)}
448: d{\bf r}^{\prime} \:\delta \rho({\bf r}^{\prime}, t^{\prime}) \nonumber\\
449: &&~~~~~~~~~~
450: \times
451: G ({\bf r}, {\bf r}^{\prime}, t-t^{\prime})
452: \theta(({\bf r}^{\prime}-{\bf R})<\xi)\nonumber\\
453: &&~~~~~~~~~~~~~~~~~~~~~~
454: - \delta \rho ({\bf r},t) \Biggr] \nonumber \\
455: &=&
456: \delta \rho({\bf r}, t) +
457: \frac{1}{V}
458: \int dt^{\prime}\int_{{\cal D}(R)}
459: d{\bf r}^{\prime} \:\delta \rho({\bf r}^{\prime}, t^{\prime}) \nonumber\\
460: &&~~~~~~~~~~~~~~~~~
461: \times
462: G ({\bf r}, {\bf r}^{\prime}, t-t^{\prime})
463: \Omega({\bf r}^{\prime}-{\bf r}) \nonumber\\
464: &&~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
465: - \frac {v_{0}}{V} \delta \rho ({\bf r}, t)
466: \end{eqnarray}
467: \noindent
468: Where,
469: \begin{eqnarray}
470: \Omega({\bf r}-{\bf r}^{\prime})&=& \int d{\bf R}
471: \:\theta(({\bf r}-{\bf R})<\xi)
472: \theta(({\bf r}^{\prime}-{\bf R})<\xi) \nonumber\\
473: &=&\frac{1}{12} \pi (4\xi+|{\bf r}-{\bf r^{\prime}}|)
474: (2\xi-|{\bf r}-{\bf r^{\prime}}|)^{2}, \nonumber\\
475: &&~~~~~~~~~~~~~~~~~~~~~~~~~~for
476: ({\bf r}- {\bf r}^{\prime}) \leq 2\xi \nonumber\\
477: &=& 0~~~~~~~~~~~~~~~~~~~~~~~~for
478: ({\bf r}- {\bf r}^{\prime})> 2\xi
479: \end{eqnarray}
480: \noindent
481: is the overlap volume of two spheres of radius $\xi$ centered at $r$ and
482: $r^{\prime}$, respectively,and
483: \begin{equation}
484: v_{0}=\int d{\bf R} \: \theta(({\bf r}-{\bf R})<\xi)
485: =\frac{4}{3} \pi \xi^{3}
486: \end{equation}
487: \noindent
488: is the volume of the region participating in hopping.
489: $\xi$ defines the region of hopping which is calculated
490: from RFOT theory \cite{Lubwoly}.
491:
492: The RFOT $\xi$ is of the order of a molecular length,'a' at $T_{c}$ but
493: diverges as $(T-T_{K})^{-2/3}$ as the Kauzmann temperature is approached. Near
494: $T_{g}$ (where the timescale is of the order of 1 hour) $\xi$ is universally
495: $\sim$ 5 molecule spacings \cite{Lubwoly}.
496:
497:
498: Thus expanding
499: $\overline {\delta {\rho}}^{new}({\bf r},t+\delta t)$ about $t$ we get,
500:
501: \begin{eqnarray}
502: \frac {d \overline {\delta {\rho}}^{new}({\bf r},t)}{dt}\times \delta t
503: &=&\frac{1}{V}\Biggl[ \int dt^{\prime}\int_{{\cal D}(R)}
504: d{\bf r}^{\prime} \:\delta \rho({\bf r}^{\prime}, t^{\prime})\nonumber\\
505: &&~~~~~~\times
506: G ({\bf r}, {\bf r}^{\prime}, t-t^{\prime})
507: \Omega({\bf r}^{\prime}-{\bf r})\nonumber\\
508: &&~~~~~~~~~~~~~~~~~~~~~
509: - v_{0} \delta \rho ({\bf r}, t) \Biggr ]
510: \end{eqnarray}
511: \noindent
512: Now lets say that the rate of a particle hopping is given by $p_{h}$.
513: Thus the probability of any particle hopping in the region of volume $V$
514: during the time interval
515: $\delta t$ is $p_{h}V \times {\delta t}/v_{P}$, where $v_{p}$ is the volume of a
516: particle ( a particle will be a bead according to RFOT theory).
517: If we include this information in the above equation and take a
518: Laplace transform we get,
519:
520: \begin{eqnarray}
521: s\overline {\delta \rho ^{new} ({\bf q},s)}&-&
522: \overline {\delta \rho ^{new} ({\bf q},t=0)}\nonumber\\
523: &=& \frac {p_{h}}{v_{p}}\Biggl[ \int d{\bf q}_{1} G({\bf q}_{1}, s)
524: \Omega ({\bf q}-{\bf q}_{1}) \delta \rho({\bf q},s)\nonumber\\
525: &&~~~~~~~~~~~~~~~~~~~ - v_{0}
526: \delta \rho({\bf q},s)\Biggr]
527: \end{eqnarray}
528:
529: Thus we obtain,
530: \begin{equation}
531: \phi_{hop}(q,s)=\frac{\phi_{hop}(q,0)}{s+(v_{0}-{\widehat {\Omega G}}({q},s)
532: )p_{h}/v_{p}} \label{phihop8}
533: \end{equation}
534: \noindent
535: In calculation of the hopping rate we have considered that hopping is
536: instantaneous. But simulations have shown \cite{anistress} that once it occurs,
537: hopping happens over a period of time. This would give a further frequency
538: dependence to $\Omega$.
539: The fluctuations of entropic driving force leads to a distribution of
540: hopping rates $p_{h}$ \cite{lubwoly2}.
541: To account for this we can consider that there are n-type of
542: instanton with lifetimes ($\tau_{j}$) and with occurrence probability
543: $p_{h}^{i}$. In this case equation \ref{phihop8}
544: can be averaged over different instanton lifetimes and rates.
545: Here, for simplification we replace $p_{h}$ by the average hopping
546: rate $P$ which is given by $P=\frac{1}{n}\Sigma_{i} p_{h}^{i}$. Since
547: hopping rate is small (that is, hops are rare), we can assume that
548: the duration or lifetime of a hop (instanton popping up) is
549: infinitesimally small.
550:
551: In the above equation,
552: \begin{eqnarray}
553: {\widehat {\Omega G}}({q}, s)=\int d{\bf q}_{1} G ({\bf q}_{1}, s)
554: \Omega ({\bf q}- {\bf q}_{1})
555: \end{eqnarray}
556: In the above theory the length scale of the Greens function is given by the
557: Lindemann length $d_{l}$ where as the length scale of $\Omega$ is given by
558: $\xi$. The random first order transition theory gives us the estimate of
559: $\xi$, that is the spatial extent of the instanton and also the rate
560: at which the instantons pop up that is $p_{h}$ or more properly its distribution.
561: The latter should be related
562: to the waiting time distribution of hopping, as discussed by De Michele and
563: Leporini \cite{lepo} .
564:
565: Since $\xi >> d_{l}$, $G(q,s)$ has a much slower decay with $q$ than
566: $\Omega(q-q_{1})$ which implies that the product decays slowly with $q$.
567: The propagator $G(q,s)$ should be just the short time part of
568: the dynamic structure factor.
569:
570: Now we can write,
571: \begin{eqnarray}
572: \phi_{hop}(q,s)&=&
573: \frac{1}{s+(v_{0}-{\widehat {\Omega G}}({ q},s)
574: )P/v_{p}} \nonumber\\
575: &=&\frac{1}{s+K_{hop}(s)}
576: \label{fhop}.
577: \end{eqnarray}
578: \noindent
579:
580: As mentioned earlier, the length
581: scale of the Greens function is determined by the
582: Lindemann length $d_{l}$. If we neglect its frequency dependence then
583: the Greens function can be written as,
584: $G(q) = exp(-q^{2}d_{l}^{2})$.
585: %*******************************************88
586: \subsection{Calculation of the hopping rate from RFOT theory}
587: In this section we briefly discuss the calculation of the
588: hopping rate $P$ from the RFOT theory.
589: As has been discussed before, the hopping rate of an instanton
590: is connected to the free
591: energy barrier, which in turn is connected to the
592: configurational entropy of the system \cite{xiawoly,Lubwoly}
593: and an asymptotic form of that is given by Eq.\ref{barrier}.
594: Note that, in glasses, we need the interface energy between
595: two alternative amorphous packings and not between two thermodynamic
596: phases, like a solid and a delocalized uniform liquid.
597: The alternative phases become more alike at $T_{c}$.
598: The interfacial energy (surface tension) that enters
599: in the calculation of the instanton barrier is smaller at $T_{c}$ than at
600: lower temperature.
601: This effect has been discussed by Lubchenko and Wolynes who termed it
602: as barrier softening. A rigorous estimate of the barrier requires complex
603: replica instanton calculations as recently outlined by Franz \cite{franz}
604: and by Dzero,
605: Schmalian and Wolynes \cite{wolycond}. An alternative method motivated by variational reasoning proposed by
606: Lubchenko and Wolynes is to write a simple interpolation formula for the free
607: energy which allows the system to choose the smallest available barrier,and is
608: given by \cite{Lubwoly},
609:
610: \begin{equation}
611: F(r)= \frac{\Gamma_{K} \Gamma_{A}}{\Gamma_{K}+ \Gamma_{A}} -\frac {4 \pi}{3}
612: r^{3} T s_{c}\:, \label{bsoft}
613: \end{equation}
614: \noindent
615: where $\Gamma_{K}$ and $\Gamma_{A}$ are the surface energy terms at
616: $T_{K}$ and $T_{A}$ respectively.
617: $T_{A}$ gives an estimate of temperature at which a
618: soft crossover to the onset of activated transport takes place or the
619: temperature at which the barrier disappears when approached from below.
620: The expression for the surface energy at $T_{K}$ is,
621: $\Gamma_{K}=4 \pi \sigma_{o}a^{2} (r/a)^{3/2}$, where
622: $\sigma_{o}=\frac{3}{4}(k_{B}T/a^{2}) ln((a/d_{L})^{2}/\pi e)$ is the
623: surface tension coefficient and $a$ is the molecular bead-size which refers
624: to the size of the quasi-spherical parts of any a spherical molecule. The bead
625: count 'b' for each molecule is also be calculated within RFOT theory.
626: The expression for the surface energy at $T_{A}$ is, $\Gamma_{A}=
627: 4 \pi \Delta f a^{3} (r/a)^{2}$, where
628: $\frac{\Delta f}{T_{A}s^{A}_{c}}= 4 \bigl(\frac{t}{t_{K}}\bigr)^{3/2}
629: (1/\sqrt{1+3t/t_{K}})$,
630: where $t=(1-T/T_{A})$ and $\Delta f$, the excess free energy density computed
631: from the density functional.
632: Note that Eq.\ref{bsoft} tends to $\Gamma_{K}$ as $
633: T \rightarrow T_{K}$ and similarly, to $\Gamma_{A}$ as $T \rightarrow T_{A}$.
634: From the above expression of the barrier, the critical
635: barrier free energy, $F^{*}(r)$, can be calculated as
636: the maximum value of $F(r)$.
637: Microscopic calculation giving $T_{A}$ can then be used with
638: Eq.\ref{bsoft} to calculate the value of viscosity,
639: $\eta/\eta_{0}=exp(F^{\star}/k_{B}T)$, (where $\eta_{0}$ is the viscosity
640: at the
641: melting temperature, $T_{m}$ and is obtained from experiments). Rather than
642: carry out completely microscopic calculations, Lubchenko and Wolynes
643: used this approach to fit
644: the experimental values of viscosity, where the values of the transition temperature
645: $T_{A}$ and $s_{fit}$
646: ($s_{c}=s_{fit}(1-T_{K}/T)$) are used as the fitting parameters.
647: $T_{K}$ is obtained from thermodynamics. For example,
648: $T_K$ is known to be 175K for Salol.
649:
650: For our calculation of the hopping rate $P$, we use Eq.\ref{bsoft} with the
651: fitted values of $T_{A}$ and $s_{fit}$ from Lubchenko and Wolynes
652: \cite{Lubwoly}.
653: Thus $P=\frac {1}{\tau_{0}}exp(-\Delta F^{\star}/k_{B}T)$,
654: where $\tau_{0}$ is the timescale of the system at the melting point, $T_{m}$
655: which is taken from experimentally known value \cite{stickel}.
656: The above expression would predict that rate of hopping
657: becomes of the order of hours at $T=T_{g}$ and vanishes at $T=T_{K}$
658: \cite{Lubwoly}.
659:
660: We also need an estimate of the size of the hopping region $\xi$ which has a temperature
661: dependence and is equated to the $r$ value where in Eq.\ref{bsoft}, F(r)=0.
662:
663:
664: \section{The total density-density time correlation function}
665:
666: In the last section we have evaluated the change in the
667: dynamic structure factor due to hopping alone.
668: In this section we incorporate this effect along with mode coupling feedback terms
669: in the full
670: density-density correlation function in a self-consistent manner.
671:
672:
673: As mentioned earlier MCT describes well the dynamics for a moderately
674: supercooled system. It also works well for a deeply supercooled system confined
675: within a single free energy minimum. These features are preserved when the
676: hopping term or instanton vertex is added.
677: In the idealized MCT, the equation of motion for the
678: dynamic structure factor,${\bf\phi}^{id}_{MCT}(q,t)$ is first simplified by
679: neglecting the coupling between
680: different wave vectors and considering the contribution from the static and
681: dynamic quantities at a single
682: wave number, $q=q_{m}$, ($q_{m}$ is the wavenumber where the peak of the
683: structure factor appears). The idealized MCT equation
684: can be written as \cite{leu,beng},
685: \begin{eqnarray}
686: &&\ddot{\bf\phi}^{id}_{MCT}(t)
687: +\gamma \dot{\bf \phi}^{id}_{MCT}(t)
688: + \Omega_{0}^{2}
689: {\bf\phi}^{id}_{MCT}(t)\nonumber\\
690: &&~~+ 4 \lambda \Omega_{0}^{2}
691: \int_{0}^{t} dt^{\prime} {{\bf\phi}^{id}_{MCT}}^{2}(t^{\prime})
692: \dot{\bf\phi}^{id}_{MCT}(t-t^{\prime}) = 0 \label{fqtleu}
693: \end{eqnarray}
694: \noindent
695: where ${\bf\phi}^{id}_{MCT}$ is the normalized density time correlation function.
696: This integro-differential equation
697: has an unusual structure but it can be solved numerically and many of its properties are now
698: analytically understood. The frequency of the free oscillator can be
699: approximated as, $\Omega_{0}^{2}=k_{B}Tq_{m}^{2}/m S(q_{m})$
700: and the damping constant, $\gamma$,
701: which is the short time part of the memory kernel,
702: is taken to be proportional to $\Omega_{0}$.
703: This is because randomizing collisions occur in Lennard-Jones liquids on nearly the same timescale as the
704: vibrations.
705: The fourth term on the left hand side has the form of a memory kernel
706: and its strength is controlled by the dimensionless coupling constant which
707: can be taken to be,
708: $\lambda=(q_{m} A^{2}/8 \pi^{2} \rho) S(q_{m})$, where $A \delta(q-q_{m})
709: =S(q_{m})-1 $ \cite{beng}.
710: It is useful to transform to
711: the Laplace frequency plane, where the Laplace transform is defined as
712: $\phi(s)={\cal L} [\phi(t)]$. Now the equation can be rewritten as,
713:
714:
715: \begin{eqnarray}
716: {\bf \phi}^{id}_{MCT} (s)=\frac{1}{s+ \frac {\Omega_{0}^{2}}
717: {s+\eta^{id}_{l}(s)}}\label{fqtmcts}
718: \end{eqnarray}
719: \noindent
720: where the longitudinal viscosity $\eta^{id}_{l}$ is given by
721: \begin{eqnarray}
722: \eta^{id}_{l}(s)=\gamma + 4 \lambda \Omega_{0}^{2} {\cal L} [
723: {{\bf \phi}^{id}_{MCT}}^{2}(t)]
724: \label {etal}
725: \end{eqnarray}
726: \noindent
727: The above equations are nonlinear in nature and when Eq.\ref{fqtleu}
728: is solved
729: numerically or Eqs.\ref{fqtmcts} and \ref{etal} are solved self consistently
730: for different values of
731: $\lambda$, a dynamical arrest of the liquid structural
732: relaxation was predicted at $T=T_{c}$ \cite{leu,beng,gotze}.
733: However, $T_{c}$ is always higher than $T_{g}$. The origin
734: for this inconsistency is the exclusion of the hopping motion
735: in the perturbative MCT dynamics.
736:
737: Hopping and continuous motion are essentially
738: two distinct channels for the structural
739: relaxation in a supercooled liquid, although they get coupled due to
740: self-consistency, as described below. In the deeply supercooled
741: regime these two styles of motion
742: are
743: usually well separated in their timescales. Thus to include hopping in the
744: dynamics of structural relaxation, as a first approximation, valid in the
745: deeply supercooled regime, the
746: full intermediate scattering function can be written as
747: a product of two functions \cite{bbhop},
748: \begin{equation}
749: \phi(q,t)\simeq \phi_{MCT}(q,t)\phi_{hop}(q,t) \label{fqthop}
750: \end{equation}
751: \noindent
752: where $\phi_{MCT}(q,t)$ is the mode coupling part and $\phi_{hop}(q,t)$ is the
753: contribution from the hopping motion, which is already discussed in the last
754: subsection.
755: %$F_{mct}(k,t=0)=S(k)$ and $F_{hop}(k,t=0)=1$.
756:
757: We now derive the equation for $\phi_{MCT}(q,t)$ consistent
758: with Eq.\ref{fqthop}. To begin with, note that
759: the equation of motion ${\bf\phi}_{MCT}(t)$ should remain the same as
760: ${\bf\phi}^{id}_{MCT}(t)$ (Eq.\ref{fqtleu})
761: but due to the presence of hopping, the memory function will
762: now be modified. The memory function, which is the longitudinal
763: viscosity, is now determined by the full dynamic structure factor which
764: is given by Eq.\ref{fqthop}. Therefore, ${\bf\phi}_{MCT}(t)$ is now solved
765: self consistently with the full
766: dynamic structure factor, ${\bf\phi}(t)$, and the equation of motion is
767: written as,
768: \begin{eqnarray}
769: &&\ddot{\bf\phi}_{MCT}(t)
770: +\gamma \dot{\bf \phi}_{MCT}(t)
771: + \Omega_{0}^{2}
772: {\bf\phi}_{MCT}(t)\nonumber\\
773: &&~~~~+ 4 \lambda \Omega_{0}^{2}
774: \int_{0}^{t} \:dt^{\prime} {\bf\phi}^{2}(t^{\prime})
775: \dot{\bf\phi}_{MCT}(t-t^{\prime}) = 0 \label{fqtmod}
776: \end{eqnarray}
777: \noindent
778: Note that for simplicity we have removed explicit $q$ dependence
779: in the above equation and all the quantities are calculated at $q=q_{m}$.
780: Although, the structure of Eq.\ref{fqtleu} and Eq.\ref{fqtmod} are quite
781: similar, the memory function in the former is calculated using
782: only $\phi^{id}_{MCT}(t)$ whereas in the later it is calculated using the
783: full intermediate
784: scattering function, $\phi(t)$. Thus ${\phi}_{MCT}(t)$ is now dependent
785: on the hopping motion.
786: The equation in the Laplace frequency plane, for the modified
787: ${\bf\phi}_{MCT}(s)$ remains the same as Eq.\ref{fqtmcts}, except that the
788: ideal longitudinal viscosity $\eta^{id}_{l}$ is replaced by the following
789: longitudinal viscosity given by,
790:
791: \begin{eqnarray}
792: \eta_{l}(s)=\gamma + 4 \lambda \Omega_{0}^{2} {\cal L} [\bf \phi^{2}(t)]
793: \label {etalmod}
794: \end{eqnarray}
795: \noindent
796: We see at this order, the longitudinal viscosity is modified
797: which in turn modifies the decay of ${\bf\phi}_{MCT}(t)$.
798:
799: Combining Eq.\ref{fqthop} and Eq.\ref{fqtmod} we can now write a equation
800: of motion for the full structure factor, $\phi(t)$. The equation becomes simple when
801: $\phi_{hop}(t)$ can be approximated as an exponential with time scale $1/K_{hop}$.
802: In the limit of small $K_{hop}$ the equation of motion for $\phi(t)$ can be
803: written as,
804:
805: \begin{eqnarray}
806: \ddot{\bf\phi}(t)
807: &+&\gamma \dot{\bf \phi}(t)
808: + (K_{hop}\gamma+\Omega_{0}^{2})
809: {\bf\phi}(t) \nonumber\\
810: &+& 4 \lambda \Omega_{0}^{2}
811: \int_{0}^{t} \:dt^{\prime} \phi_{hop}(t^{\prime}){\bf\phi}^{2}(t^{\prime})
812: \nonumber\\
813: &&~~~~~~~\times
814: \biggl[\dot{\bf\phi}(t-t^{\prime})+K_{hop}{\bf\phi}(t-t^{\prime})\biggr] = 0
815: \label{schm1}
816: \end{eqnarray}
817: \noindent
818:
819: The effects of hopping motion in the full structural relaxation can be
820: incorporated directly in the frequency plane
821: which is
822: suggested by Eq.\ref{fhop}
823: and Eq.\ref{fqtmcts}. Exploiting the strict parallelism
824: of hopping and convective motion,
825: these two equations can be combined
826: to write an extended equation for the dynamic
827: structure factor in the frequency plane as,
828:
829: \begin{eqnarray}
830: \bf \phi (s)&=&\frac{1}{s+
831: (v_{0}-{\widehat {\Omega G}}({ q_{m}},s)
832: )P/v_{p}+
833: \frac {\Omega_{0}^{2}}{s+\eta_{l}(s)}}\nonumber\\
834: &=&\frac{1}{s+K_{hop}(s) + K_{MCT}(s)} \label{fqt2s}
835: \end{eqnarray}
836: \noindent
837: where
838: \begin{eqnarray}
839: &K_{hop}(s)=(v_{0}-{\widehat {\Omega G}}({q_{m}},s)
840: )P/v_{p} \nonumber\\
841: &and \nonumber\\
842: &K_{MCT}(s)= \frac {\Omega_{0}^{2}}{s+\eta_{l}(s)} \label{mem2s}
843: \end{eqnarray}
844: \noindent
845: and $\eta_{l}(s)$ is given by Eq.\ref{etalmod}.
846:
847: $K_{MCT}(s)$ and $K_{hop}(s)$ are the contributions to the rate
848: of relaxation from MCT and hopping
849: modes, respectively, and they act as two parallel channels for the decay
850: of $\phi(t)$.
851: At high temperature and low density $K_{MCT} >> K_{hop}$, thus the decay of
852: $\phi(t)$ is determined primarily by the MCT term. But as we approach
853: $T_{c}$, the relaxation of $\phi (t)$ slows down which increases the
854: longitudinal viscosity value (given by Eq.\ref{etal}). This in turn
855: causes $K_{MCT}(s)$ to decrease much faster than $K_{hop}(s)$ at low
856: frequency.
857: Now if we neglect the frequency dependence of $K_{hop}$ then the equation of
858: motion for $\phi(t)$ for small $K_{hop}$ value can be written as,
859: \begin{eqnarray}
860: \ddot{\bf\phi}(t)
861: &+&\gamma \dot{\bf \phi}(t)
862: + (K_{hop}\gamma+\Omega_{0}^{2})
863: {\bf\phi}(t) \nonumber\\
864: &+& 4 \lambda \Omega_{0}^{2}
865: \int_{0}^{t} \:dt^{\prime}{\bf\phi}^{2}(t^{\prime})\nonumber\\
866: &&~~~~~\times
867: \biggl[\dot{\bf\phi}(t-t^{\prime})+K_{hop}{\bf\phi}(t-t^{\prime})\biggr] = 0
868: \label{schm2}
869: \end{eqnarray}
870: \noindent
871:
872: Note that as expected,
873: while Eq.\ref{schm2} has similar structure as Eq.\ref{schm1},
874: the memory function is different and the consequence of this will be discussed
875: later.
876:
877: The second scheme (given by Eq.\ref{fqt2s}) appears to be rather similar to
878: extended mode coupling theory of Gotze and Sjogren\cite{gotze,sjogren}.
879: In the extended MCT the current contribution is incorporated in the memory
880: kernel. The modified memory kernel leads to an extra decay channel of
881: the dynamic structure factor thus successfully
882: predicting the decay of $\phi(t)$
883: below $T_{c}$. The rate of decay for the second channel, $\delta$,
884: is called the rate due to the hopping mode.
885: However, note that there is no direct connection between
886: the expression of $\delta$ in the Gotze-Sjogren \cite{gotze} analysis
887: and the hopping motion. Since the microscopic
888: calculation of $\delta$ can never predict a zero hopping rate they used
889: $\delta$ as a fitting parameter to explain the
890: experimental results.
891:
892:
893:
894: \section{Numerical calculations and results}
895:
896: %\subsubsection {Choice of parameters and methods of solution}
897:
898:
899: For this paper we solved
900: Eqs.\ref{fqtleu}, \ref{fqtmod},\ref{schm1} and \ref{schm2}
901: are solved numerically.
902: Although in principle Eqs.\ref{fqt2s}-\ref{mem2s} can be solved self
903: consistently with Eq.\ref{etalmod} in an
904: iterative way, it requires a very large number of iterations
905: to converge near the glass transition temperature.
906: An alternative way is to numerically solve directly Eq.\ref{schm2}.
907: Both these calculations were done for a model system and for concreteness were applied to a specific
908: molecular system, Salol.
909:
910: \begin{figure}
911: \epsfig{file=fig1.eps,height=7cm,width=8cm,angle=0}
912: %\includegraphics[width=12cm]{fig1.ps}
913: \caption{The idealized MCT result (given by Eq.\ref{fqtleu})
914: and the modified
915: full $\phi(t)$ (given by Eq.\ref{schm1})
916: have been plotted against log(t ps) for different values of $K_{hop}$.
917: All the plots are for $\lambda=1$ that is at $T=T_{c}$, the
918: mode coupling transition temperature. For different values of
919: $K_{hop}$ the curve follows the idealized MCT result till it starts
920: decaying from the plateau. The smaller the $K_{hop}$ value the longer
921: is the plateau and the slower the long time decay.}
922: \end{figure}
923:
924: \begin{figure}
925: \epsfig{file=fig2.eps,height=7cm,width=8cm,angle=0}
926: %\includegraphics[width=12cm]{fig2.ps}
927: \caption{The total $\phi(t)$(given by Eq.\ref{schm1}), the modified
928: $\phi_{MCT}(t)$(given by Eq.\ref{fqtmod}) and the idealized
929: $\phi^{id}_{MCT}(t)$(given by Eq.\ref{fqtleu}) are plotted
930: against time for $\lambda=1$ ( i.e $T=T_{c}$) and for $K_{hop}=.001$.
931: After the initial short time decay,
932: both the total structural relaxation $\phi(t)$ and the MCT part of the
933: structural relaxation $\phi_{MCT}(t)$ decays with time where as the
934: idealized MCT theory predicts no decay of the structural relaxation.
935: This figure shows that the
936: decay of $\phi_{MCT}(t)$ due to hopping induces
937: continuous diffusion in the system. }
938: \end{figure}
939:
940: \begin{figure}
941: \epsfig{file=fig3.eps,height=7cm,width=8cm,angle=0}
942: %\includegraphics[width=12cm,angle=0]{fig3.ps}
943: \caption{$ln(\phi(t))$, $ln(\phi_{MCT}(t))$ and
944: $ln(\phi^{id}_{MCT}(t))$ have been plotted against time where the parameters
945: are the same as in figure 2. The plot shows that in the longtime
946: both $\phi(t)$ and $\phi_{MCT}(t)$ decays exponentially with different
947: timescales. Due to the presence of the explicit hopping term
948: (see Eq.\ref{fqthop}) $\phi(t)$ decays faster than $\phi_{MCT}(t)$.}
949: \end{figure}
950: \subsection{Results for the schematic model}
951: In the numerical calculations of the schematic model system,
952: we took $\Omega_{0}=1$,
953: and $\gamma=\Omega_{0}$ fixed. We varied $\lambda$ between .1-1, knowing that $T_{c}$ is
954: reached for $\lambda=1$ and varied $K_{hop}$ from
955: 0.01 to 0.0001.
956:
957: %\subsubsection{Results}
958: In {\bf figure 1} we plot both the idealized MCT result (given by Eq.\ref{fqtleu})
959: and the modified
960: full $\phi(t)$ (given by Eq.\ref{fqthop})
961: dynamic structure factor against log(t).
962: We see that although the ideal MCT saturates to a plateau value,
963: the extended theory exhibits a
964: hopping induced decay following the MCT plateau value -- the duration
965: of the plateau depends on the rate of hopping.
966:
967: In {\bf figure 2} we plot the total $\phi(t)$ and $\phi_{MCT}(t)$
968: against time.
969: The plot shows that hopping not only leads to the decay
970: of the total $\phi(t)$ (given by Eq.\ref{fqthop}), but
971: it also slows down the growth of the
972: longitudinal viscosity thus facilitating the structural relaxation
973: even through the mode coupling channel.
974:
975: The long time behavior of $ln(\phi(t))$ and $ln(\phi_{MCT}(t))$ are plotted
976: against time in
977: {\bf figure 3}, for $\lambda=1$ and $K_{hop}=.001$.
978: The semilog plot shows straight lines in the long time with different
979: slopes for the MCT part and the total dynamic structure factor, thus
980: indicating that both the functions are exponential
981: in the long time but with somewhat different time constants
982: as evident in figure 2.
983:
984:
985:
986: Next we compare the two different schemes at $T=T_{c}$ (that is, for
987: $\lambda=1$) and for $K_{hop}=.001$. The solution of Eq.\ref{schm1}
988: and Eq.\ref{schm2} are plotted in {\bf figure 4}
989: against log(t).
990: Both the approximate schemes provide similar results but the first scheme
991: (Eq.\ref{schm1}) shows a faster decay implying that hopping has a stronger
992: effect at this level.
993: A comparison between Eq.\ref{schm1} and Eq.\ref{schm2}
994: shows that although the structures of the equations are similar, in
995: the memory function of Eq.\ref{schm1} there is an extra product
996: with $\phi_{hop}$ which
997: eventually leads to a faster decay of the total function as seen in the figure.
998: The advantage of the first approximate scheme is that
999: we can separately calculate the MCT part of the dynamic structure factor and
1000: can then demonstrate explicitly how hopping induces a
1001: decay of $\phi_{MCT}$. This
1002: implies that hopping opens up continuous diffusion channels in the system.
1003:
1004: In the schematic calculation we have independently varied $\lambda$ and $K_{hop}$
1005: while keeping parameters $\Omega_{0}$ and $\gamma$ constant.
1006: In real systems all of these are
1007: functions of density and temperature and needs to varied simultaneously.
1008: Although the temperature dependence of $\Omega_{0}$ and $\gamma$ are not too
1009: strong but $\lambda$ depends more strongly on T and $K_{hop}$ is expected to depend more
1010: strongly than exponential on $T$. This means that
1011: as we increase the timescale of hopping (which is achieved by lowering the
1012: temperature), the $\lambda$ value also increases.
1013: Thus for real systems, unlike as shown in figure 1,
1014: as $1/K_{hop}$ gets stretched, the MCT plateau value also should
1015: increase somewhat. This is consistent with density functional
1016: theory of the aperiodic solid.
1017: In the next subsection we apply our theory to a real system
1018: chosen as Salol and include these effects.
1019: \begin{figure}
1020: \epsfig{file=fig4.eps,height=7cm,width=8cm,angle=0}
1021: %\includegraphics[width=12cm]{fig4.ps}
1022: \caption{To compare the two different
1023: approximate schemes, give by Eq.\ref{schm1} (scheme 1) and
1024: by Eq.\ref{schm2} (scheme 2) we plot $\phi(t)$
1025: against log(t ps) for $\lambda=1$ (that is, at $T=T_{c}$)
1026: and for $K_{hop}=.001$. The initial short time dynamics
1027: is identical for both the schemes but scheme 1 predicts a slightly
1028: faster longtime
1029: decay (see text for detailed discussion).}
1030: \end{figure}
1031:
1032: \subsection{Salol}
1033:
1034: Salol (phenyl salicylate) is a fragile glass forming substance.
1035: It is intra-molecularly hydrogen bonded, a van der Walls system
1036: which is considered as a model substance for the study of glass
1037: transition and molecular mobility in the supercooled liquid.
1038: To perform the microscopic calculations, we
1039: need to map the system
1040: into a Lennard-Jones system and for that we
1041: need the molecular diameter and the well depth.
1042: Although the shape of the molecule is not
1043: spherical we can calculate a hard sphere diameter $\sigma_{hs}=7.22\AA$
1044: \cite{bondi}. Note that in the RFOT theory the elementary
1045: particles in the system
1046: are not the molecules but beads and each molecule is made up of certain
1047: number of beads. We thus need to estimate the bead size {\it a}.
1048: Use of the RFOT theory (with softening) for the fit of the
1049: viscosity leads to
1050: the number of beads in a Salol molecule to be 6.29 \cite{Lubwoly}.
1051: Equating the volume of the molecule with the volume of 6.29 beads,
1052: we get the hard sphere bead size
1053: $a_{hs}=3.095\AA$ which
1054: for the present calculation we equate to the Lennard-Jones bead size {\it a}.
1055: Next we need to estimate the temperature
1056: scaling which will be equivalent to the Lennard-Jones well depth $\epsilon$.
1057: To obtain the temperature scaling we use the fact that for
1058: reduced Lennard-Jones system the mode coupling transition is known to
1059: take place at
1060: reduced density, $\rho^{*}=0.95$ and reduced temperature, $T^{*}=0.57$.
1061: We are not aware of any reported density value of the Salol system,
1062: but a reduced
1063: temperature $T^{*}_{c}=0.57$ seems to be a reasonable value when compared with
1064: simulation results \cite{arnab}.
1065: For a Salol system it is known that the mode coupling
1066: transition temperature, $T_{c}=256 K$
1067: \cite{stickel25}. Thus the temperature is to be scaled by,
1068: $\epsilon/k_{B}=449.122$.
1069:
1070: \begin{table}
1071: \caption [Table n:] { In this table we present the value of the parameters
1072: calculated. $\Omega_{0}$ is the frequency of the free oscillator, $\gamma$
1073: is the short time part of the memory function, acting as a damping constant.
1074: $\lambda$ provides an estimate of the strength of coupling and is a
1075: dimensionless quantity. $\xi$ is the radius of the region of hopping,
1076: $\frac{(v_{0}-{\widehat {\Omega G}}({q_{m}}))}{v_{p}}$
1077: gives the effect of a single hopping on the density and
1078: $P$ is the hopping rate per particle. Although all the parameters vary with
1079: temperature, the temperature dependence of
1080: We find that $\Omega_{0}$ and $\gamma$ have a weak temperature dependence
1081: which implies that the short time dynamics remain unchanged. Although,
1082: $\lambda$ also shows a weak temperature dependence but a small variation in
1083: $\lambda$ leads to a substantial change in the dynamics.
1084: The hopping rate varies strongly with temperature.}
1085:
1086:
1087: \begin{center}
1088: \begin{tabular}{|c|c|c|c|c|c|c|} \hline
1089: & & & & & & \\
1090: T (K) & $\Omega_{0}^{2}(ps^{-2})$ &
1091: $\gamma(ps^{-1})$ &$\lambda$ & $\xi/a$ &
1092: $\frac{(v_{0}-{\widehat {\Omega G}}({q_{m}}))}{v_{p}}$ & P ($ps^{-1}$) \\
1093: & & & & & & \\ \hline
1094: 270 & 1.353 & 1.163 & 0.94 & 1.89 & 1.79 & 3.09 $\times 10^{-4}$ \\
1095: 256 & 1.268 & 1.126 & 1.01 & 2.32 &3.47 & 3.2 $\times 10^{-7}$ \\
1096:
1097: 247 & 1.216 & 1.102 & 1.05 & 2.62 & 5.12 & 2.89 $\times 10^{-8}$ \\
1098:
1099:
1100: &&&&&&\\\hline
1101: \end{tabular}
1102: \end{center}
1103: \end{table}
1104:
1105:
1106: Now that we know the molecular and the thermodynamic
1107: parameters of the Lennard-Jones system, we can calculate
1108: the various parameter values in the integral equation
1109: that we need for a microscopic calculation,
1110: namely, for the MCT part, $\Omega_{0}^{2}=k_{B}Tq_{m}^{2}/m S(q_{m})$,
1111: $\gamma$ and $\lambda=(q_{m} A^{2}/8 \pi^{2} \rho) S(q_{m})$,
1112: where $A \delta(q-q_{m})=(S(q_{m})-1) $ \cite{beng}.
1113: Instead of microscopically calculating the short time part of the memory
1114: kernel, $\gamma$, we again take it to be proportional
1115: to $\Omega_{0}$ with the
1116: proportionality factor assumed to be unity (as our schematic model study).
1117: Gotze
1118: and Sjogren \cite{gotze} suggest that this proportionality factor even
1119: when varied between 1 and 100 does not effect the long time behavior.
1120: In the calculation of $\Omega_{0}$ and $\lambda$,
1121: $S(q_{m})$ is calculated for the above mentioned
1122: Lennard-Jones system.
1123:
1124: Next for the physical quantities in the hopping part of the calculation,
1125: we need to calculate
1126: $\xi$ and $P$. These are temperature dependent quantities. For this
1127: calculation we use Eq.\ref{bsoft} where $T_{A}=333$ and $s_{fit}=2.65$
1128: \cite{Lubwoly}. The values for all the calculated parameters are given in Table 1.
1129:
1130: \begin{figure}
1131: \epsfig{file=fig5.eps,height=7cm,width=8cm,angle=0}
1132: %\includegraphics[width=12cm]{fig5.ps}
1133: \caption{The calculated structural relaxation for Salol
1134: at three state points which also includes the mode coupling transition
1135: temperature $T_{c}=256K$.
1136: Eq.\ref{schm1} is solved with the parameters given in table 1 to obtain
1137: $\phi(t)$. The solid lines A, B, and C correspond to $\phi(t)$ at
1138: temperatures 270K, 256K and 247K respectively. The dot-dashed lines are
1139: the plots at the same temperatures but without hopping (that is, the
1140: idealized MCT result).}
1141: \end{figure}
1142:
1143: The plot for the Salol system, obtained by using the parameters reported
1144: in Table 1 is given in figure 5. The plot appears to be
1145: quite similar to figure 1, at and below the mode coupling
1146: transition temperature $T_{c}=256K$. Due
1147: to the presence of hopping the
1148: structural relaxation decays from the plateau value.
1149: From Table 1 we find that $\Omega_{0}$ and
1150: $\gamma$ have a weak temperature dependence which is reflected in figure 5
1151: in the near invariance of the short time dynamics at all
1152: the three temperatures. On the other hand, although
1153: $\lambda$ shows a weak temperature dependence but a small variation in
1154: $\lambda$ leads to a substantial change in the dynamics. Above
1155: $\lambda=1$ the increase in its magnitude leads to an increase of
1156: the plateau value.
1157: The hopping rate varies strongly with temperature and has a
1158: stronger effect on the dynamics above $\lambda=1$ where at lower
1159: temperature the plateau
1160: is stretched to longer times. Below $\lambda=1$ (or $T>T_{c}$), where
1161: $\phi(t)$ can decay completely via the diffusive channel (MCT part),
1162: the presence of hopping leads to a slightly
1163: faster decay of the structural relaxation.
1164:
1165: \section{Concluding Remarks}
1166:
1167:
1168: Experiments show that the structural relaxation in a supercooled
1169: liquid exhibits rich dynamics over many time and length scales.
1170: Often the physical origin of these different
1171: dynamics are quite disparate. To explain the dynamics of the
1172: liquid over the whole temperature regime under a single
1173: theoretical framework has been
1174: a challenge. The random first order transition theory which
1175: at high temperature also contains the essential elements
1176: described in the perturbative mode
1177: coupling theory\cite{kw1987} has been successful in
1178: explaining the liquid dynamics
1179: around the laboratory glass transition temperature $T_{g}$.
1180: The mode coupling theory,
1181: on the other hand has been able to explain the high temperature dynamics extremely well.
1182: The theory is also known to be accurate for short times in
1183: the supercooled liquid below
1184: $T_{c}$. Below $T_{c}$ on long time scales
1185: the idealized MCT theory fails because of the
1186: exclusion of the activated hopping motion. As RFOT theory makes clear
1187: that hopping
1188: motions correspond to instantons which gives rise to essential
1189: singularities at T$_{K}$.
1190: In this work we have proposed a unified structural description of the
1191: liquid, covering
1192: the whole temperature regime, from above $T_{c}$ to $T_{K}$.
1193:
1194:
1195: Towards this goal we first treated the effect of hopping on the density
1196: fluctuation. In the theory the effect of a single particle hopping
1197: is connected to the neighboring density fluctuations through a propagator.
1198: The typical rate of a particle hopping and the spatial extent of a hopping event are
1199: obtained from RFOT theory. After incorporating all the informations it is
1200: found that hopping acts as a channel for the decay of the structural
1201: relaxation (Eq.\ref{fhop}). Distribution of hopping times are essential
1202: for the non-exponentiality of the $\alpha$ relaxation \cite{lubwoly2}
1203: and will be incorporated in future work.
1204:
1205: The total structural relaxation after incorporating the hopping motion was
1206: calculated using two different approximate schemes.
1207: As in idealized MCT, the density relaxation
1208: is calculated self consistently with the longitudinal viscosity.
1209: In the first scheme (given by Eq.\ref{fqtmod}) the structural
1210: correlations in the time plane are taken as a simple product
1211: of the MCT part and the hopping part.
1212: It is found that due to the inclusion of
1213: the hopping motion the arrest of the
1214: structural relaxation at $T_{c}$, as predicted by the idealized MCT,
1215: disappears.
1216: The theory with thus modified total structural relaxation predicts
1217: a hopping dominated decay following the MCT plateau.
1218: In the first scheme the part of relaxation of
1219: $\phi(t)$ due to the diffusional motion (MCT part)
1220: is calculated separately (Eq.\ref{fqtmod}), but
1221: self-consistently with the full dynamic structure
1222: factor (which now includes the hopping motion).
1223: The theory
1224: predicts that hopping not only relaxes the total density correlation but
1225: {\it also invokes a decay of the MCT part}. This implies that hopping
1226: opens up continuous diffusion channels in the system. This is consistent
1227: with the co-existence of both kinds of motions observed in computer
1228: simulation\cite{arnab}.
1229:
1230: It is interesting to consider the predictions of the present generalized
1231: theory regarding the temperature dependence of viscosity. The
1232: decay of the MCT part of the dynamic structure factor already predicts a
1233: slower rise of the viscosity with the temperature than the ideal
1234: MCT. The interesting point to note is that the hopping rate itself
1235: also decreases with the lowering of temperature. Thus, the dynamic
1236: structure factor now not only has a slow exponential decay
1237: in the long time, but the plateau also gets stretched to longer times.
1238: Both the slow exponential
1239: relaxation at long times and the stretching of the plateau
1240: at 'intermediate times' contribute to the
1241: increase of the viscosity. However, the resulting temperature dependence
1242: can be quite different from the prediction of the ideal MCT. In particular,
1243: viscosity is dominated by the power law part at low supercooling
1244: but at large
1245: supercooling even the duration of the plateau and the power law
1246: decay are determined mostly by the hopping rate.
1247:
1248:
1249: {\bf ACKNOWLEDGEMENT}~
1250: This work was supported in parts from NSF (USA) and DST (India).
1251:
1252: \begin{references}
1253:
1254: \bibitem{book} Articles in {\it Dynamics of Disordered Materials} ed
1255: D. Richter, A. J. Dianoux, W. Petry and J. Teixeira, Berlin:Springer, (1989).
1256:
1257: \bibitem{fayer} G. Hinze, D. D. Brace, S. D. Gottke and M. D. Fayer, J. Chem.
1258: Phys. {\bf 113}, 3723(2000); Phys. Rev. Lett. {\bf 84}, 2437 (2000).
1259:
1260:
1261: \bibitem{rajesh25} W. Gotze, in {\it Liquids, Freezing and Glass Transition, Les Houches session}, edited by J. P. Hansena, D. Levesque and J.
1262: Zinn-Justin (North Holland, Amsterdam, 1991) p 287; W. Gotze and L. Sjogren,
1263: Rep. Prog. Phys. {\bf 55}, 241 (1992).
1264:
1265: \bibitem{leu} E. Leutheusser, Phys. Rev. A {\bf 29}, 2765 (1984).
1266:
1267: \bibitem{beng} U. Bengtzelius, W. Gotze, and A. Sjolander, J. Phy. C {\bf 17},
1268: 5915 (1984).
1269:
1270: \bibitem{gotze} W. Gotze and L. Sjogren, Z. Phys. B- Cond. Mat., {\bf 65},
1271: 415 (1987);W. Gotze and L. Sjogren, J. Phys. C:Solid State Phys. {\bf 21}, 3407 (1988).
1272:
1273:
1274: \bibitem{sjogren} L. Sjogren, Z. Phys. B - Cond. Mat., {\bf 79}, 5 (1990).
1275:
1276: \bibitem{sarirev} B. Bagchi and S. Bhattacharyya, Adv. Chem. Phys. {\bf 116},
1277: 67, (2001).
1278:
1279: \bibitem{xiawoly} X. Xia and P. G. Wolynes, Phys. Rev. Lett. {\bf 86}, 5526
1280: (2001);Proc. Natl. Acad. Sci. U. S. A. {\bf 97}, 2990 (2000).
1281:
1282:
1283: \bibitem{Lubwoly}V. Lubchenko and P. G. Wolynes, J. Chem. Phys. {\bf 119},
1284: 9088 (2003);
1285:
1286: \bibitem{lubwoly2}V. Lubchenko and P. G. Wolynes, J. Chem. Phys. {\bf 121},
1287: 2852 (2004).
1288:
1289: \bibitem{heatcp} C. T. Moynihan, A. J. Easteal, J. Wilder and J. Tucker,
1290: J. Phys. Chem. {\bf 78}, 2673 (1974).
1291:
1292: \bibitem{stossel1} J. P. Stoessel and P. G. Wolynes,
1293: J. Chem. Phys. {\bf 80}, 4502 (1984).
1294:
1295: \bibitem{stossel} Y. Singh, J. P. Stoessel and P. G. Wolynes,
1296: Phys. rev. Lett. {\bf 54}, 1059 (1985).
1297:
1298:
1299:
1300: \bibitem{ges}T. Geszti, J. Phys. C; Solid State Phys. {\bf 16}, 5805 (1983).
1301:
1302: \bibitem{kw1987} T.R. Kirkpatrick and P.G. Wolynes, Phys. Rev. {\bf A 35},
1303: 3072 (1987).
1304:
1305: \bibitem{biroli} J. P. Bouchaud and G. Biroli, Cond. Mat. 0406317 (2004).
1306:
1307: \bibitem{wang}G.Wahnstr${\ddot o}$m, Phys. Rev. A {\bf 44}, 3752 (1991).
1308:
1309: \bibitem{arnab} A. Mukherjee, S. Bhattacharyya and B. Bagchi, J. Chem. Phys.
1310: {\bf 116}, 4577 (2002); S. Bhattacharyya, A. Mukherjee, and B. Bagchi,
1311: J. Chem. Phys.{\bf 117}, 2741(2002).
1312:
1313:
1314: \bibitem {anistress}S. Bhattacharyya and B. Bagchi, Phys. Rev. Lett. {\bf 89},
1315: 025504-1(2002).
1316:
1317: \bibitem{kw}T. R. Kirkpatrick and P. G. Wolynes, Phys. Rev. B {\bf 36},
1318: 8552 (1987).
1319:
1320:
1321: \bibitem{ktw}T. R. Kirkpatrick, D. Thirumalai, and P. G. Wolynes, Phys. Rev. A {\bf 40}, 1045 (1989).
1322:
1323:
1324:
1325:
1326: \bibitem{angell}R. B$\ddot o$hmer, K. L. Ngai, C. A. Angell and D. J. Plazek
1327: {\bf 99}, 4201 (1993).
1328:
1329: \bibitem{das}S. P. Das and G. F. Mazenko, Phys. Rev. A, {\bf 34}, 2265 (1986).
1330:
1331:
1332:
1333: \bibitem{kawasaki} K. Fuchizaki and K. Kawasaki, J. Phys.:Condens. Matter
1334: {\bf 14}, 12203 (2002).
1335:
1336:
1337: \bibitem{ry} T. V. Ramakrishnan and M. Yussouff Phys. Rev. B {\bf19}
1338: 2775 (1979).
1339:
1340:
1341: \bibitem{wolyunpub} J. Stevenson, J. Schmalian and P. G. Wolynes
1342: (unpublished).
1343:
1344: \bibitem{lepo} C. DeMichele and D. Leporini, Phys. Rev. E, {\bf 63},
1345: 036701 (2001).
1346:
1347: \bibitem{franz} S. Franz, cond=mat/0412383.
1348:
1349: \bibitem{wolycond}M. Dzero, J. Schmalian and P. G. Wolynes, cond-mat/050211.
1350:
1351:
1352: \bibitem{stickel}F. Stickel, E. W. Fischer and R. Richert, J. Chem. Phys.
1353: {\bf 102}, 6251 (1995).
1354:
1355:
1356: \bibitem{bbhop} B. Bagchi, J. Chem. Phys. {\bf 101}, 9946 (1994).
1357:
1358:
1359:
1360:
1361:
1362:
1363: \bibitem{bondi}A. Bondi, J. Chem Phys., {\bf 68}, 441 (1964) ;D. Ben-Amotz and
1364: D. R. Herschbach, J. Phys. Chem. {\bf 94}, 1038 (1990).
1365:
1366: \bibitem{stickel25} G. Li, W. M. Du, A. Sakai, and H. Z. Cummins, Phys. Rev. A
1367: {\bf 46}, 3343 (1992).
1368:
1369:
1370: \end{references}
1371:
1372: \end{document}
1373:
1374: