Optimization of biasing parameters in particle transport simulations

A technique for the optimization of biasing parameters in Monte-Carlo simulation of particle transport problems is described. This technique is a variant of Spanier's multi-stage procedure. It is based on stochastic evaluation of the derivatives of the second moment using small samples. Application of the technique to tracklength biasing schemes employed in deep penetration problems is discussed. Various leakage estimators are considered to demonstrate the applicability of the technique. Optimum biasing parameters are obtained for various values of scattering probability and are compared with those obtained through Spanier's technique. The advantages of our technique over Spanier's technique are discussed. (author)