STRUCTURE AND DYNAMICS OF MOLECULAR CLUSTERS - DIAGNOSTIC-CRITERIA IN MONTE-CARLO COMPUTATIONS

In the course of applying Monte Carlo (MC) simulations to the interpretation of experiments on clusters of polyatomic molecules, various schemes to monitor the transitions from one type of structure (phase) to another were examined. Special attention was paid to melting-like phenomena. Initially the diagnoses followed those in the prior literature on atomic clusters and included such gauges as caloric curves and the Lindemann index. Several problems of interpretation were encountered, however, which made it expedient to introduce some new techniques. First, in order to derive results distorted as little as possible by artifacts, it was necessary to implement a method to null out the stochastic translational and rotational displacements of clusters as a whole characteristic of the MC technique. Second, although it was corroborated that the Lindemann index is a more robust indicator of change than is the configurational energy, its meaning in simulations of clusters is different for rigid polyatomic molecules from that for monatomic molecules. Therefore, a more fundamental gauge of melting was devised, namely one characterizing self-diffusion. An analysis based on a simple random-walk model provided an index of diffusion that was independent of the arbitrary MC step size and rejection rate, and was substantially more sensitive to melting than the Lindemann index. Autocorrelation analyses of configurational energies were also found to reveal structural transitions. Illustrative results are presented for the molecular systems of benzene and tellurium hexafluoride.