Simulated annealing-based algorithms for the studies of the thermoelastic scaling behavior

Abstract

Simulated annealing is a robust and easy-to-implement algorithm for material simulation. However, it consumes a huge amount of computational time, especially on the studies of percolation networks. To reduce the running time, we parallelize the simulated annealing algorithm in our studies of the thermoelastic scaling behavior of percolation networks. The critical properties of the thermoelastic moduli of percolation networks near the threshold p/sub c/ are investigated by constructing a square percolation network. The properties are tested by simulations of a series of two-dimensional (2-D) percolation networks near p/sub c/. The simulations are performed using a novel parallelizing scheme on the simulated annealing algorithm. To further accelerate the computational speed, we also propose a new conjectural method to generate better initial configurations, which speeds up the simulation significantly. Preliminary simulation results show surprisingly that the percolating phenomenon of thermal expansion does exist under certain conditions. The behavior seems to be governed by the elastic properties of a percolation network.