Parallel-hat tempering: A Monte Carlo search scheme for the identificationof low-energy structures

A new parallel-hat tempering algorithm has been developed for Monte Carlo s imulations, in which a composite ensemble of noninteracting replicas of the molecule system at different temperatures is simulated, and the Markov pro cess of each replica is driven by a hatlike weight factor with exponentiall y enhanced probability in both low- and high-energy regions. To test the al gorithm, the methodology is applied to a homopolymeric protein chain locate d on a face-centered cubic lattice. We demonstrate that the ability of the current approach to search for low-energy molecule structures is significan tly better than other Monte Carlo techniques found in the literature. (C) 2 001 American Institute of Physics.