The stochastic properties of a molecular-dynamical Lennard-Jones system in equilibrium and nonequilibrium states

The K-entropy and the t(m) time of dynamical memory (the time of forgetting initial conditions during numerical integration) of a classical system of particles whose interactions are governed by the Lennard-Jones potential were calculated by the method of molecular dynamics. The K value was a characteristic of a system of many particles, and the t(m) value proved to increase logarithmically as fluctuations of the total energy of the system decreased; that is, as the accuracy of numerical integration increased. Two different K-entropy values corresponding to the same total energy of the system were found to exist, namely, K-e for the equilibrium and K-n for the nonequilibrium state. The rate of kinetic energy relaxation (t(r)(-1)) was shown to equal K-n, and the K-n value was found to be a more fundamental characteristic than t(r)(-1). The density dependences of K-e (monotonic) and K-n (nonmonotonic) were calculated. The transition from dynamical (Newtonian) correlations to stochastic for the velocity autocorrelation function was considered. The reasons for the finiteness of dynamical memory in physical processes are discussed. The duration of dynamical correlations in real systems is limited by quantum uncertainty and is of the order of picoseconds. (C) 2001 MAIK Nauka/Interperiodica.