A memory function analysis of non-exponential relaxation in viscous liquids

TL;DR

This paper introduces a memory function approach with a negative power-law tail to analyze relaxation in glass-forming liquids, revealing a temperature-independent exponent and a fragile-to-strong crossover.

Contribution

It demonstrates that a memory function with a power-law tail effectively models non-exponential relaxation, challenging the traditional stretched exponential fit and identifying a temperature-independent tail exponent.

Findings

Power-law tail exponent is approximately 1.58/1.50 across temperatures.

A fragile-to-strong crossover occurs around T=0.40 in 3D systems.

The short-time decorrelation rate follows an Arrhenius temperature dependence.

Abstract

We analyze data from simulations of 2D and 3D glass-forming liquids using a correlation function defined in terms of a memory function with a negative inverse power-law tail. The self-intermediate function and the autocorrelation functions of pressure and shear stress are analyzed; the obtained fits are very good, at least as good as with a stretched exponential. In contrast to the stretched exponential, the key shape parameter--the exponent of the power-law tail--seems to be the same for all three correlation functions. It decreases from a value around 2 at high temperature to a value close to 1.58 (2D), 1.50 (3D) at low temperatures. The amplitude of the tail increases towards towards a value corresponding to a diverging relaxation time, which is related to anomalous diffusion. On the other hand, careful analysis of the long time behavior in the case of suggests that the memory function is cut-off exponentially, which avoids the divergence of the relaxation time. Repeating the fits with an exponential cut-off included indicates that the power-law exponent is in fact independent of temperature and close to 1.58/1.50 over the whole range. Instead of the divergence, a fragile-to-strong crossover in the dynamics, estimated to occur around $T=0.40$ for the 3D Kob-Andersen system. Another key parameter of the fitting procedure may be interpreted as a short-time rate, the amount of decorrelation that occurs in a fixed, relatively short time interval (compared to the alpha time). This quantity is observed to have a near-Arrhenius temperature dependence, while its wavenumber dependence seems to be diffusive ($q^{-2}$) over a wider range than that of the relaxation itself, a further indication that this "bare relaxation rate" is simpler than the full dynamics.