1: %\documentclass[twocolumn,a4paper,aps,prl]{revtex4}
2: \documentclass[a4paper,aps,10pt]{revtex4}
3: \usepackage{amsmath}
4: \usepackage{graphics}
5: \usepackage{mathrsfs}
6:
7: \begin{document}
8:
9: \title{Liquid-glass transition of a fluid confined in a disordered
10: porous matrix: A mode coupling theory}
11:
12: \author{V. Krakoviack}
13:
14: \affiliation{Laboratoire de Chimie, \'Ecole Normale Sup\'erieure de
15: Lyon, 46, All\'ee d'Italie, 69364 Lyon Cedex 07, France}
16:
17: \date{\today}
18:
19: \begin{abstract}
20: We derive an extension of the mode coupling theory for the
21: liquid-glass transition to a class of models of confined fluids, where
22: the fluid particles evolve in a disordered array of interaction
23: sites. We find that the corresponding equations are similar to those
24: describing the bulk, implying that the methods of investigation which
25: were developed there are directly transferable to this new domain of
26: application. We then compute the dynamical phase diagram of a simple
27: model system and show that new and nontrivial transition scenarios,
28: including reentrant glass transitions and higher-order singularities,
29: can be predicted from the proposed theory.
30: \end{abstract}
31:
32: \maketitle
33:
34: Since the mode coupling theory (MCT) of the liquid-glass transition
35: was proposed in the mid-eighties \cite{leu84pra,bengotsjo84jpc}, it
36: has acquired a central role in this field of research
37: \cite{leshouches,gotsjo92rpp,got99jpcm}. Indeed, on the experimental
38: or numerical side, it is very often to the predictions of the MCT that
39: new data are first confronted \cite{gotsjo92rpp,got99jpcm}, and on the
40: theoretical side, models of increasing complexity are regularly
41: investigated within the MCT framework as a means to unveil potential
42: new phenomena \cite{highorder}. The reason for this strong influence
43: of the MCT lies in its ability to reproduce important phenomenological
44: aspects of the dynamics of supercooled liquids: First, of course, the
45: slowing down of the structural relaxation when density is increased or
46: temperature decreased, but other more specific features as well, like
47: the two-step relaxation scenario. Moreover, it makes a number of
48: precise universal predictions especially suitable for comparisons with
49: experiment or simulation results, and, for simple enough systems for
50: which the MCT is tractable as a first principle theory, it provides
51: detailed predictions concerning nonuniversal aspects of the dynamics
52: as well, allowing thus extensive quantitative tests of the theory
53: \cite{gotsjo92rpp,got99jpcm}.
54:
55: In the past few years, a rapidly growing interest for the dynamics of
56: liquids under confinement has built into the glass transition
57: community \cite{proceedings}, with the aim of clarifying the concept
58: of cooperativity, a key ingredient of many glass transition theories
59: \cite{heterogeneities}. Indeed, confinement has appeared as a means to
60: impose to a glassforming system a new characteristic lengthscale (pore
61: size, film thickness\dots), which should interact with any correlation
62: or cooperativity length developing in it, possibly leading to indirect
63: informations on the nature and evolution with temperature of
64: cooperativity. In the course of these investigations, at least for
65: some systems studied by computer simulation
66: \cite{galrovspo00prl,galpelrov02el,schkolbin04jpcb}, many features of
67: the dynamics of bulk glassforming liquids which had found an
68: interpretation in the framework of the MCT have been uncovered. It
69: then appeared natural to compare the corresponding data with the
70: predictions of the MCT, even if the theory had been designed for bulk
71: fluids, and, because the tests were quite successful, the idea emerged
72: that a mode coupling scenario was at work in confined fluids as well.
73:
74: It seems thus that there is a clear need for an extension of the MCT
75: to confined glassforming liquids. First, if this theory turned out to
76: be similar enough to the theory for the bulk, this would put the
77: studies of simulation data mentioned above on firmer grounds, and more
78: generally this would provide a framework for the analysis of
79: experimental or numerical data on confined fluids. Second, and maybe
80: more importantly, by applying the new theory to various models, as it
81: was done with its bulk counterpart, a thorough exploration of the
82: phenomenology of confined glassforming systems would become possible,
83: potentially allowing to disentangle the different physical effects
84: which interplay in these systems.
85:
86: It is the aim of this Letter to provide such an extension of the MCT
87: for a particular class of confined systems, the so-called
88: ``quenched-annealed'' (QA) binary mixtures. In these systems, first
89: introduced by Madden and Glandt \cite{MG1988}, the fluid molecules
90: equilibrate in a matrix of particles frozen in a disordered
91: configuration sampled from a given probability distribution. The
92: models studied in Refs.~\cite{galpelrov02el,kim03el,chajagyet04pre}
93: belong to this class of systems, as does the Lorentz model, which
94: corresponds to a zero fluid density limit. Besides the proposed theory
95: will borrow ideas from the mode coupling approaches to the
96: diffusion-localization transition in this model
97: \cite{gotleuyip81pra,leu83pra,sza04el}.
98:
99: The theory is derived using the projection operator method described
100: in Ref.~\cite{leshouches}. In the present problem, the inner product
101: of two arbitrary dynamical variables $A$ and $B$ is given by
102: $\overline{\langle A B^* \rangle}$, where $^*$ denotes complex
103: conjugation, $\langle \cdots \rangle$ a thermal average \emph{taken
104: for a given realization of the matrix} and $\overline{\cdots}$ a
105: \emph{subsequent} average over the matrix realizations. We thus
106: consider a fluid made of $N_f$ particles of mass $m$, adsorbed in a
107: homogeneous disordered matrix consisting of $N_m$ immobile sites. The
108: system has volume $V$, hence the fluid and matrix densities are
109: respectively $n_f=N_f/V$ and $n_m=N_m/V$. As in the bulk MCT, the
110: dynamical variables of interest are the Fourier components of the
111: microscopic fluid density, $\rho^f_\mathbf{q}(t)=\sum_{j=1}^{N_f} e^{i
112: \mathbf{q} \mathbf{r}_j(t)}$, where $\mathbf{q}$ denotes the
113: wavevector and $\mathbf{r}_j(t)$ is the position of the fluid particle
114: $j$ at time $t$. But, before proceeding with the dynamical theory, one
115: has to take care of certain peculiarities of the statics of QA
116: systems. Indeed, because of the presence of the quenched component,
117: for a given matrix realization, the translational invariance of the
118: system is broken. This implies that, at variance with bulk fluids,
119: time-persistent density fluctuations exist at equilibrium,
120: i.e., $\langle \rho^f_\mathbf{q} \rangle \neq 0$. This fact is well
121: known from the derivation, using the replica method, of the
122: Ornstein-Zernike (OZ) equations describing this type of systems
123: \cite{G1992,RTS1994,MD1994}, where it leads to the splitting of the
124: total and direct correlation functions of the fluid, respectively
125: $h^{ff}(r)$ and $c^{ff}(r)$, into two contributions, connected
126: [$h^{c}(r)$ and $c^{c}(r)$] and blocked or disconnected [$h^{b}(r)$
127: and $c^{b}(r)$]. A similar splitting occurs for the structure factor
128: of the fluid $S^{ff}_q= \overline{\langle \rho^f_\mathbf{q}
129: \rho^{f}_\mathbf{-q} \rangle} / N_f =1+n_f \hat{h}^{ff}_q$, leading to
130: $S^{ff}_q = S^{c}_q + S^{b}_q$ with $S^{c}_q = \overline{\langle(
131: \rho^f_\mathbf{q} - \langle\rho^f_\mathbf{q}\rangle)
132: (\rho^{f}_\mathbf{-q}-\langle\rho^{f}_\mathbf{-q} \rangle)\rangle} /
133: N_f=1+n_f \hat{h}^{c}_q$ and $S^{b}_q =
134: \overline{\langle\rho^f_\mathbf{q}\rangle \langle\rho^{f}_\mathbf{-q}
135: \rangle}/N_f=n_f \hat{h}^{b}_q$, where $\hat{f}_q$ denotes the Fourier
136: transform of $f(r)$. For future reference, we define the matrix-matrix
137: and fluid-matrix structure factors and total correlation functions as
138: well, which are given by $S^{mm}_q = \overline{\langle
139: \rho^m_\mathbf{q} \rho^{m}_\mathbf{-q}\rangle}/ N_m=1+n_m
140: \hat{h}^{mm}_q$ and $S^{fm}_q = \overline{\langle \rho^f_\mathbf{q}
141: \rho^{m}_\mathbf{-q} \rangle} / \sqrt{N_f N_m}= \sqrt{n_f n_m}
142: \hat{h}^{fm}_q$; $\rho^m_\mathbf{q}=\sum_{j=1}^{N_m} e^{i \mathbf{q}
143: \mathbf{s}_j}$, where $\mathbf{s}_j$ is the fixed position of the
144: matrix particle $j$, is the $\mathbf{q}$ Fourier component of the
145: quenched microscopic matrix density.
146:
147: It thus follows that, if one is only interested in the relaxing part
148: of the fluid density fluctuations, one has to consider the dynamical
149: variable $\delta\rho^f_\mathbf{q}(t)=\rho^f_\mathbf{q}(t)-\langle
150: \rho^f_\mathbf{q} \rangle$ rather than $\rho^f_\mathbf{q}(t)$ itself.
151: Using the standard method, a generalized Langevin equation for the
152: time evolution of the normalized autocorrelation function of the
153: connected density fluctuations $\phi_q(t)= \overline{\langle
154: \delta\rho^f_\mathbf{q}(t) \delta\rho^f_\mathbf{-q}\rangle}/ (N_f
155: S^{c}_q)$ can then be derived, which is formally identical to the
156: equation for bulk fluids, i.e.,
157: \begin{equation}\label{langcoll}
158: \ddot{\phi}_{q}+\Omega_{q}^2 \phi_{q}+ \Omega_{q}^2 \int_0^t d\tau
159: M_q(t-\tau) \dot{\phi}_{q}(\tau)=0,
160: \end{equation}
161: with $\Omega_{q}^2=q^2 k_B T /(m S^{c}_q)$, where $T$ is the
162: temperature and $k_B$ the Boltzmann constant. The memory function is
163: given by $\Omega_{q}^2 M_q(t)= \overline{\langle
164: R_\mathbf{q}(t)\,R_\mathbf{-q}\rangle}/(N_f m k_B T)$, where
165: $R_\mathbf{q}(t) =\exp[i (1-\mathscr{P}) \mathscr{L} (1-\mathscr{P})
166: t] i (1-\mathscr{P}) \mathscr{L} g^f_\mathbf{q}$ is the projected
167: random force obtained from the longitudinal fluid momentum density
168: fluctuation $g^f_\mathbf{q}(t)$. $\mathscr{L}$ is the Liouville
169: operator of the system and $\mathscr{P}$ is the projector onto the
170: subspace of dynamical variables spanned by $\delta\rho^f_\mathbf{q}$
171: and $g^f_\mathbf{q}$.
172:
173: We now obtain the slow decaying portion of the memory kernel with a
174: mode coupling approach, assuming that the slow decay of the projected
175: random force autocorrelation function is dominated by couplings to two
176: types of quadratic dynamical variables,
177: $B^{(1)}_{\mathbf{q,k}}=\delta\rho^f_\mathbf{k}
178: \delta\rho^f_\mathbf{q-k}$, in analogy with the bulk MCT, and
179: $B^{(2)}_{\mathbf{q,k}}=\delta\rho^f_\mathbf{k} \rho^m_\mathbf{q-k}$,
180: following previous work on the Lorentz gas
181: \cite{gotleuyip81pra,leu83pra,sza04el}. Defining the projector
182: $\mathcal{P}$ on the subspace spanned by the $B_{\mathbf{q,k}}$'s, the
183: mode coupling part of the memory function is then expressed as
184: $\Omega_{q}^2 M^{(\text{MC})}_q(t)=\overline{\langle
185: \mathcal{P}R_\mathbf{q}(t)\,\mathcal{P}R_\mathbf{-q} \rangle}/(N_f m
186: k_B T)$. To complete the calculation of $M^{(\text{MC})}_q$, two
187: steps remain. First, a factorization approximation is required to
188: express four-point density correlations into products of two-point
189: density correlations. Following the usual mode coupling prescription
190: then leads to (we note $\mathscr{Q}=1-\mathscr{P}$)
191: \begin{gather*}
192: \overline{\langle (e^{i\mathscr{QLQ}t} \delta\rho^f_\mathbf{k}
193: \delta\rho^f_\mathbf{q-k} ) \delta\rho^f_\mathbf{-k'}
194: \delta\rho^f_\mathbf{-q+k'} \rangle} \simeq (\delta_{\mathbf{k,k'}}
195: + \delta_{\mathbf{k,q-k'}}) N_f^2 S^{c}_k S^{c}_{|\mathbf{q-k}|}
196: \phi_{k}(t) \phi_{|\mathbf{q-k}|}(t),\\
197: %
198: \overline{ \langle (e^{i\mathscr{QLQ}t} \delta\rho^f_\mathbf{k}
199: \rho^m_\mathbf{q-k})\delta\rho^f_\mathbf{-k'} \rho^m_\mathbf{-q+k'}
200: \rangle} \simeq \delta_{\mathbf{k,k'}} N_f N_m S^{c}_k
201: S^{mm}_{|\mathbf{q-k}|}\phi_{k}(t),\\
202: %
203: \overline{\langle ( e^{i\mathscr{QLQ}t} \delta\rho^f_\mathbf{k}
204: \delta\rho^f_\mathbf{q-k} ) \delta\rho^f_\mathbf{-k'}
205: \rho^m_\mathbf{-q+k'} \rangle} \simeq 0.
206: \end{gather*}
207: A crucial point here is that, since the matrix is quenched,
208: $\rho^m_\mathbf{q}$ shows no thermal fluctuations: $\overline{ \langle
209: \delta\rho^f_\mathbf{q}(t) \rho^m_\mathbf{-q}\rangle}$ is thus
210: identically zero [remember that $\delta\rho^f_\mathbf{q}(t) =
211: \rho^f_\mathbf{q}(t)-\langle \rho^f_\mathbf{q} \rangle$].
212:
213: Second, one needs to calculate $\overline{\langle R_\mathbf{q}
214: B^{(l)}_{\mathbf{-q,-k}} \rangle} = \overline{\langle
215: i\mathscr{L}g^f_\mathbf{q}
216: B^{(l)}_{\mathbf{-q,-k}}\rangle}-\overline{\langle
217: i\mathscr{PL}g^f_\mathbf{q} B^{(l)}_{\mathbf{-q,-k}}\rangle}$. The
218: first term is readily handled by application of the Yvon theorem, just
219: as in the bulk MCT. The second one is far more delicate, since it
220: involves three-point connected static correlations of the QA
221: system. Usually, such terms are estimated using the so-called
222: convolution approximation \cite{JacFee62rmp} which leads to remarkable
223: simplifications in the resulting mode coupling equations. An
224: extension of the convolution approximation to QA systems has thus been
225: developed which gives
226: \begin{align*}
227: \overline{\langle \delta\rho^f_{\mathbf{q}}
228: \delta\rho^{f}_{\mathbf{-k}} \delta\rho^{f}_{\mathbf{-q+k}} \rangle}&=
229: N_f S^{c}_q S^{c}_k S^{c}_{|\mathbf{q-k}|},\\ \overline{\langle
230: \delta\rho^f_{\mathbf{q}} \delta\rho^{f}_{\mathbf{-k}}
231: \rho^{m}_{\mathbf{-q+k}}\rangle}&= \sqrt{N_f N_m} S^{c}_q S^{c}_k
232: S^{fm}_{|\mathbf{q-k}|}.
233: \end{align*}
234:
235: Eventually, assuming that the contributions to the memory kernel not
236: included in $M^{(\text{MC})}_q$ can be replaced by a white noise term
237: $\Gamma_q \delta(t)$, we obtain the mode coupling equations for the
238: collective dynamics of a QA mixture, Eq.~\eqref{langcoll} with
239: $M_q(t)=\Gamma_q \delta(t) + M^{(\text{MC})}_q(t)$ and
240: \begin{subequations}\label{kerncoll}
241: \begin{gather}
242: M^{(\text{MC})}_q(t)=\frac{1}{V} \sum_{\mathbf{k}}
243: V^{(2)}_{\mathbf{q},\mathbf{k}} \phi_{k}(t) \phi_{|\mathbf{q-k}|}(t) +
244: V^{(1)}_{\mathbf{q},\mathbf{k}} \phi_{k}(t),\\
245: %
246: V^{(2)}_{\mathbf{q},\mathbf{k}} = \frac{1}{2} n_f S^{c}_q
247: \left[\frac{\mathbf{q}\cdot\mathbf{k}}{q^2} \hat{c}^{c}_k +
248: \frac{\mathbf{q}\cdot(\mathbf{q-k})}{q^2}
249: \hat{c}^{c}_{|\mathbf{q-k}|}\right]^2 S^{c}_k
250: S^{c}_{|\mathbf{q-k}|},\\
251: %
252: V^{(1)}_{\mathbf{q},\mathbf{k}} = n_m S^{c}_q
253: \left[\frac{\mathbf{q}\cdot(\mathbf{q-k})}{q^2}+n_f
254: \frac{\mathbf{q}\cdot\mathbf{k}}{q^2} \hat{c}^{c}_k \right]^2
255: \frac{(\hat{h}^{fm}_{|\mathbf{q-k}|})^2}{S^{mm}_{|\mathbf{q-k}|}}
256: S^{c}_k,
257: \end{gather}
258: \end{subequations}
259: where the replica OZ equations were used to introduce the relevant
260: direct correlation functions \cite{G1992,RTS1994}.
261:
262: The same procedure can be applied to the dynamics of a tagged particle
263: and the corresponding equations will be reported in a forthcoming
264: paper.
265:
266: Equations \eqref{langcoll} and \eqref{kerncoll} form the proposed MCT
267: for QA mixtures. Like in the bulk, they involve static quantities
268: only, and, more crucially, they retain the mathematical structure of
269: the typical mode coupling equations which have been extensively
270: studied in Ref.~\cite{leshouches}. Thus, all the known properties of
271: the solutions of MCT equations, in particular their universal
272: behaviors, apply \textit{a priori} unchanged to QA mixtures. This
273: means that, in principle, the analysis performed in
274: Ref.~\cite{galpelrov02el} is as legitimate as all the analogous ones
275: done on bulk systems.
276:
277: As one would expect, the present theory integrates the previously
278: known mode coupling theories as limiting cases: In the limit of
279: vanishing matrix density, the bulk MCT \cite{bengotsjo84jpc} is
280: recovered, while in the limit of vanishing fluid density, the MCT
281: equations for the Lorentz gas \cite{sza04el} are obtained. Since both
282: limits show ergodicity-breaking transitions (ideal glass transitions
283: in the first case, diffusion-localization transitions in the second),
284: the present theory, which ``interpolates'' between them, is bound to
285: display such phenomena.
286:
287: To illustrate this point and as a first demonstration of the
288: potentialities of the theory, we have computed the dynamical phase
289: diagram of a simple QA system (the models of
290: Refs.~\cite{galpelrov02el,kim03el} would be quite complex for a
291: preliminary study). This is the one studied in
292: Ref.~\cite{chajagyet04pre}, which consists of a fluid of hard spheres
293: confined in a matrix of hard spheres frozen in an equilibrium
294: configuration. Both the fluid and matrix particles have diameter
295: $\sigma$, and the system is characterized by two volume fractions
296: $\phi_f=\pi n_f\sigma^3/6$ and $\phi_m=\pi n_m\sigma^3/6$. The
297: Percus-Yevick approximation \cite{G1992,MLW96JCP} is used to compute
298: the required structural quantities. Since we are confronted to
299: basically the same equations as in bulk systems, the numerical
300: procedures to track ergodicity-breaking transitions signalled by the
301: appearance of a nonzero infinite time limit to $\phi_{q}(t)$ do not
302: differ from those used there. We have applied the method which is
303: described in Ref.~\cite{frafucgotmaysin97pre}.
304:
305: \begin{figure}
306: \includegraphics{figtrans.ps}
307: \caption{\label{fig1} Dynamical phase diagram of a hard sphere fluid
308: confined in a matrix of identical hard spheres frozen in an equilibrium
309: configuration. $\phi_f$ and $\phi_m$ denote respectively the fluid and
310: matrix compacities. Point E is the common endpoint of the
311: type A and type B transition lines.}
312: \end{figure}
313:
314: The corresponding dynamical phase diagram is reported in
315: Fig.~\ref{fig1}. It consists of two transition lines. On the one
316: hand, starting from the bulk fluid ($\phi_m=0$) transition point and
317: increasing $\phi_m$, one follows a line of discontinuous or type B
318: transitions, where $f_q=\lim_{t\to\infty} \phi_{q}(t)$ jumps
319: discontinuously from zero to a nonzero value when moving from the
320: ergodic liquid phase to the nonergodic glassy phase. Along this line,
321: as $\phi_m$ is increased, the amplitude of the jump decreases to zero
322: and the exponent parameter $\lambda$, which determines many properties
323: of the solutions of the mode coupling equations
324: \cite{leshouches,gotsjo92rpp,got99jpcm}, increases from its bulk value
325: (of about $.73$) to one, its largest allowed value. On the other hand,
326: moving away from the diffusion-localization ($\phi_f=0$) transition
327: point by increasing $\phi_f$, one follows a line of continuous or type
328: A transitions, from which $f_q$ grows continuously from zero when the
329: system enters in the glassy domain. Here, as $\phi_f$ is increased,
330: $\lambda$ grows continuously from zero to one, the allowed interval
331: for type A transitions.
332:
333: Both lines connect smoothly at a common endpoint E, where $\lambda=1$
334: for both. Point E corresponds to a degenerate A$_3$ singularity, a
335: generic type of topologically stable singularities already known from
336: the so-called $F_{12}$ model \cite{f12}. To our knowledge, this widely
337: studied one equation toy model had never found any physical
338: realization until now. This result has important physical
339: implications, since, in the vicinity of such a higher-order
340: singularity, the dynamics are known to display very specific features,
341: most significantly logarithmic decay laws and subdiffusive behaviors
342: \cite{leshouches,highorder,gotspe02pre}.
343:
344: Beside this specific bifurcation scenario, and formally not related to
345: it, another remarkable prediction of the present theory lies in the
346: shape of the phase diagram. Indeed, starting from the zero fluid
347: density limit and increasing $\phi_f$, the matrix density at which the
348: system becomes frozen first increases, reaches a maximum and then
349: decreases until the bulk limit is reached. The last behavior can be
350: easily understood from simple free volume arguments: Because of the
351: volume excluded by the matrix particles, the larger the matrix density
352: is, the smaller the fluid density has to be for structural arrest to
353: occur. The first regime however is quite unexpected and might be due
354: to a partial relaxation of the dynamical correlations which lead to
355: the localization of a single particle moving in the porous medium by
356: the introduction of collective density fluctuations at a finite but
357: small fluid density. Overall, this nonmonotonic behavior of $\phi_m$
358: at the dynamical transition leads to reentrant type A transitions,
359: i.e., for a given matrix density, ergodicity can be broken both by an
360: increase or a decrease of the fluid density.
361:
362: All these predictions (bifurcation scenario, logarithmic decay laws,
363: shape of the transition lines, evolutions of $f_q$ and $\lambda$ along
364: these lines) can be tested by computer simulations to judge of the
365: validity of the present theory. Unfortunately, the focus of the work
366: of Ref.~\cite{chajagyet04pre} was not on a putative mode coupling
367: scenario and thus only indirect and not so convincing comparisons in
368: favor of the theory can be made. For instance, these authors found
369: that at the lowest investigated fluid density ($\phi_f=0.05$), the
370: dependence of the diffusion coefficient on $\phi_m$ was different from
371: the one found at higher fluid densities. This might be a signature of
372: the nonmonotonicity of the transition line in this low fluid density
373: regime. Another of their observations, made in Ref.~\cite{kim03el} as
374: well and not so surprising, is that the inclusion of matrix particles
375: slows down the dynamics more efficiently than the inclusion of the
376: same amount of fluid particles. Here, this is reflected in the fact
377: that the total compacity $\phi_\text{tot}=\phi_f+\phi_m$ at the
378: transition is a decreasing function of $\phi_m$ in the top part of the
379: phase diagram of Fig.~\ref{fig1}. Thus, at a transition point, the
380: corresponding value of $\phi_\text{tot}$ being held fixed, an increase
381: of $\phi_f$ at the expense of $\phi_m$ leads to an ergodic system,
382: while the reverse change drives the system deeper into the nonergodic
383: domain. This looks encouraging, but clearly more simulation work is
384: needed.
385:
386: In summary, we have developed an extension of the MCT to the QA
387: mixture model of confined fluids. The corresponding equations turn out
388: to be similar to those of the MCT for the bulk, so that all the
389: applications of the theory which have been conceivable for the bulk,
390: like tests of its universal predictions or quantitative comparisons
391: with computer simulations for simple models, are transposable to the
392: present class of systems. The calculation of the dynamical phase
393: diagram of a simple system shows that new and complex bifurcation
394: scenarios can be predicted and that a rich phenomenology could be
395: unveiled by a systematic study of models of increasing
396: complexity. Such a work is under way.
397:
398: This of course does not exhaust the question of a general
399: mode coupling description of confined glassforming fluids. Indeed, the
400: QA mixture has the simplifying feature that it corresponds to a
401: statistically homogeneous confinement, while many studies have been
402: done for slit, cylindrical or spherical geometries of the confining
403: medium. The present development should nevertheless represent a
404: valuable means to improve our general understanding of the slow
405: dynamics of confined glassforming liquids.
406:
407: \acknowledgments It is a pleasure to thank W. G{\"o}tze for useful
408: comments and G. Tarjus for fruitful discussions and, together with
409: M.L. Rosinberg and E. Kierlik, for an earlier collaboration on the
410: theory of QA mixtures which made this work possible.
411:
412: \begin{thebibliography}{99}
413: \bibitem{leu84pra} E. Leutheusser, Phys. Rev. A \textbf{29}, 2765
414: (1984).
415: \bibitem{bengotsjo84jpc} U. Bengtzelius, W. G{\"o}tze, and
416: A.~Sj{\"o}lander, J. Phys. C \textbf{17}, 5915 (1984).
417: \bibitem{leshouches} W. G{\"o}tze, in \textit{Liquids, freezing and
418: glass transition}, edited by J.-P.~Hansen, D. Levesque, and
419: J. Zinn-Justin (North Holland, Amsterdam, 1991), pp.~287-503.
420: \bibitem{gotsjo92rpp} W. G{\"o}tze and L. Sj{\"o}gren,
421: Rep. Prog. Phys. \textbf{55}, 241 (1992).
422: \bibitem{got99jpcm} W. G{\"o}tze, J. Phys.: Condens. Matter
423: \textbf{11}, A1 (1999).
424: \bibitem{highorder} A recent example is the study of short-ranged
425: attractive colloids; see F. Sciortino, P. Tartaglia, and
426: E. Zaccarelli, Phys. Rev. Lett. \textbf{91}, 268301 (2003), and
427: references therein.
428: \bibitem{proceedings} See for instance \textit{Proceedings of the
429: International Workshop on Dynamics in Confinement}, J. Phys. (Paris)
430: IV \textbf{10}, Pr7-203 (2000), and \textit{Proceedings of Second
431: International Workshop on Dynamics in Confinement}, Eur. Phys. J. E
432: \textbf{12}, 3-204 (2003).
433: \bibitem{heterogeneities} H. Sillescu, J. Non-Cryst. Solids
434: \textbf{243}, 81 (1999); M.D. Ediger,
435: Annu. Rev. Phys. Chem. \textbf{51}, 99 (2000); R. Richert, J. Phys.:
436: Condens. Matter \textbf{14}, R703 (2002).
437: \bibitem{galrovspo00prl} P. Gallo, M. Rovere, and E. Spohr,
438: Phys. Rev. Lett. \textbf{85}, 4317 (2000);
439: J. Chem. Phys. \textbf{113}, 11324 (2000).
440: \bibitem{galpelrov02el} P. Gallo, R. Pellarin, and M. Rovere,
441: Europhys. Lett. \textbf{57}, 212 (2002); Phys. Rev. E \textbf{67},
442: 041202 (2003); \textit{ibid.} \textbf{68}, 061209 (2003).
443: \bibitem{schkolbin04jpcb} P. Scheidler, W. Kob, and K. Binder,
444: J. Phys. Chem. B \textbf{108}, 6673 (2004).
445: \bibitem{MG1988} W.G. Madden and E.D. Glandt,
446: J. Stat. Phys. \textbf{51}, 537 (1988); W.G. Madden,
447: J. Chem. Phys. \textbf{96}, 5422 (1992).
448: \bibitem{kim03el} K. Kim, Europhys. Lett. \textbf{61}, 790 (2003).
449: \bibitem{chajagyet04pre} R. Chang, K. Jagannathan, and A. Yethiraj,
450: Phys. Rev. E \textbf{69}, 051101 (2004).
451: \bibitem{gotleuyip81pra} W. G{\"o}tze, E. Leutheusser, and S. Yip,
452: Phys. Rev. A \textbf{23}, 2634 (1981).
453: \bibitem{leu83pra} E. Leutheusser, Phys. Rev. A \textbf{28}, 2510
454: (1983).
455: \bibitem{sza04el} G. Szamel, Europhys. Lett. \textbf{65}, 498 (2004).
456: \bibitem{G1992} J.A. Given and G. Stell, J. Chem. Phys. \textbf{97},
457: 4573 (1992); E. Lomba, J.A. Given, G. Stell, J.J. Weis, and
458: D. Levesque, Phys. Rev. E \textbf{48}, 233 (1993); J.A. Given and
459: G. Stell, Physica A \textbf{209}, 495 (1994).
460: \bibitem{RTS1994} M.L. Rosinberg, G. Tarjus, and G. Stell,
461: J. Chem. Phys. \textbf{100}, 5172 (1994).
462: \bibitem{MD1994} G. I. Menon and C. Dasgupta,
463: Phys. Rev. Lett. \textbf{73}, 1023 (1994).
464: \bibitem{JacFee62rmp} H.W. Jackson and E. Feenberg,
465: Rev. Mod. Phys. \textbf{34}, 686 (1962).
466: \bibitem{MLW96JCP} A. Meroni, D. Levesque, and J.J. Weis,
467: J. Chem. Phys. \textbf{105}, 1101 (1996).
468: \bibitem{frafucgotmaysin97pre} T. Franosch, M. Fuchs, W. G{\"o}tze,
469: M.R. Mayr, and A.P. Singh, Phys. Rev. E \textbf{55}, 7153 (1997).
470: \bibitem{f12} W. G{\"o}tze, Z. Phys. B \textbf{56}, 139 (1984).
471: \bibitem{gotspe02pre} W. G\"otze and M. Sperl, Phys. Rev. E
472: \textbf{66}, 011405 (2002); M. Sperl, \textit{ibid.} \textbf{68},
473: 031405 (2003); W. G\"otze and M. Sperl, J. Phys.: Condens. Matter
474: \textbf{16}, S4807 (2004).
475: \end{thebibliography}
476:
477:
478: \end{document}
479:
480: