Monte Carlo Procedures for Generation of Nonintersecting Chains

A new method is described for high-speed computer generation of non-self-intersecting random walks for use as models of coiling-type polymer molecules. Previous methods for the unbiased generation of self-avoiding random walks have proved to be statistically inadequate when certain intramolecular interactions are assumed to exist. This situation arises because energetically important configurations may have a very low probability of occurrence in a normal sample. With this problem in mind, a method of generating samples in a biased fashion has been developed. The procedure allows energetically important configurations to be generated in such a way that the actual amount of bias can be calculated and compensated for, and the samples so obtained can be incorporated with those obtained in a normal, unbiased way. Data are presented to show how the attrition attending the generation of samples depends on bias.