2016-02-10
10:15 at HCI J 574

A truly complex fluid: particles with random interactions

Yitzhak Rabin

Department of Physics and Institute for Nanotechnology and Advanced Materials, Bar−Ilan University, Israel

We use molecular dynamics simulations to study multi-component systems in which all the particles are different (APD). The particles are assumed to interact via Lennard-Jones (LJ) potentials, with identical size parameters but their pair interaction parameters are generated at random from some distribution We analyze both the global and the local properties of these systems at temperatures above and below the freezing transition and find that APD fluids relax into a non-random state characterized by clustering of particles according to the values of their pair interaction parameters (Neighborhood Identity Ordering - NIO). In order to separate NIO from the freezing transition we introduce the Random Bond Lattice model in which N(N-1)/2 bond parameters between neighboring particles are generated at random (quenched realization of bonds). The particles are then placed on a hexagonal lattice and the system is studied using MC simulations with particle exchange. We find that similarly to spin glasses, there exists a temperature above which the statistical properties of the system can be calculated using annealed averaging over all possible realizations of the bond parameters. However, the presence of particle exchange leads to a new phenomenon: the annealed approximation becomes accurate at any temperature, if the size of the system is increased above a certain (temperature-dependent) value. We discuss the ramifications of this result for the statistical properties of NP-hard problems.