Dynamical susceptibility of glass formers: Contrasting the predictions of theoretical scenarios Cristina Toninelli, 1 Matthieu Wyart, 2 Ludovic Berthier, 3 Giulio Biroli, 4 and Jean-Philippe Bouchaud 2,5 1 ENS 24 rue Lhomond, 75231 Paris Cedex 05, France 2 Service de Physique de l’État Condensé Orme des Merisiers—CEA Saclay, 91191 Gif sur Yvette Cedex, France. 3 Laboratoire des Verres UMR 5587, Université Montpellier II and CNRS, 34095 Montpellier, France 4 Service de Physique Théorique Orme des Merisiers—CEA Saclay, 91191 Gif sur Yvette Cedex, France 5 Science & Finance, Capital Fund Management 6-8 Bd Haussmann, 75009 Paris, France Received 15 December 2004; published 14 April 2005 We compute analytically and numerically the four-point correlation function that characterizes nontrivial cooperative dynamics in glassy systems within several models of glasses: elastoplastic deformations, mode- coupling theory MCT, collectively rearranging regions CRR’s, diffusing defects, and kinetically constrained models KCM’s. Some features of the four-point susceptibility 4 tare expected to be universal: at short times we expect a power-law increase in time as t 4 due to ballistic motion t 2 if the dynamics is Brownian followed by an elastic regime most relevant deep in the glass phasecharacterized by a t or t growth, depending on whether phonons are propagative or diffusive. We find in both the and early regime that 4 t , where is directly related to the mechanism responsible for relaxation. This regime ends when a maximum of 4 is reached at a time t = t * of the order of the relaxation time of the system. This maximum is followed by a fast decay to zero at large times. The height of the maximum also follows a power law 4 t * t * . The value of the exponents and allows one to distinguish between different mechanisms. For example, freely diffusing defects in d =3 lead to =2 and = 1, whereas the CRR scenario rather predicts either = 1 or a logarithmic behavior depending on the nature of the nucleation events and a logarithmic behavior of 4 t * . MCT leads to = b and =1/ , where b and are the standard MCT exponents. We compare our theoretical results with numerical simulations on a Lennard-Jones and a soft-sphere system. Within the limited time scales accessible to numerical simulations, we find that the exponent is rather small, 1, with a value in reasonable agreement with the MCT predictions, but not with the prediction of simple diffusive defect models, KCM’s with noncooperative defects, and CRR’s. Experimental and numerical deter- mination of 4 tfor longer time scales and lower temperatures would yield highly valuable information on the glass formation mechanism. DOI: 10.1103/PhysRevE.71.041505 PACS numbers: 64.70.Pf I. INTRODUCTION The idea that the sharp slowing down of supercooled liq- uids is related to the growth of a cooperative length scale dates back at least to Adam and Gibbs 1. But it is only a few years back that this idea has started being substantiated by convincing experiments 2–6, numerical simulations 7–14, and simple microscopic models 15–25,27. One of the basic problems has been to find an observable that allows one to define and measure objectively such a cooperative length scale. An interesting quantity, proposed a few years ago in the context of mean-field p-spin glasses 28see 29 for an important early insightand measured in simulations, is a four-point density correlator, defined as G 4 r , t= 0,00, tr ,0r , t - 0,00, tr ,0r , t , 1 where r , trepresents the density fluctuations at position r and time t. In practice one has to introduce an overlap func- tion w 28to avoid a singularity due to the evaluation of the density at the same point or consider slightly different corre- lation functions 30. This quantity measures the correlation in space of local-time correlation functions. Intuitively, if at point 0 an event has occurred that leads to a decorrelation of the local density over the time scale t, G 4 r , tmeasures the probability that a similar event has occurred a distance r away within the same time interval t see, e.g., 31. There- fore G 4 r , tis a candidate to measure the heterogeneity and cooperativity of the dynamics. The best theoretical justifica- tion for studying this quantity is to realize that the order parameter for the glass transition is already a two-body object—namely, the density-density correlation function Ct= 0,00, t—which decays to zero in the liquid phase and to a constant value in the frozen phase. The four- point correlation G 4 r , ttherefore plays the same role as the standard two-point correlation function for a one-body order parameter in usual phase transitions. Correspondingly, the associated susceptibility 4 tis defined as the volume inte- gral of G 4 r , tand is equal to the variance of the correlation function 28,32,33. The susceptibility 4 thas been com- puted numerically for different model glass formers and in- deed exhibits a maximum for t = t * , the relaxation time of the system 11–14. The peak value 4 t * is seen to in- crease as the temperature decreases, indicating that the range of G 4 r , t * increases as the system becomes more sluggish. The dynamical correlation length 4 t * extracted from G 4 r , t * in molecular dynamics simulations grows and be- comes of the order of roughly 10 interparticle distances when the time scale is of the order of 10 5 microscopic time scales 0 with 0 0.1 ps for an atomic liquid. In experiments close PHYSICAL REVIEW E 71, 041505 2005 1539-3755/2005/714/04150520/$23.00 ©2005 The American Physical Society 041505-1