Application of the Random Ray Method to Variance Reduction in Radiation Shielding Problems

Deep penetration radiation shielding problems often use Monte Carlo (MC) augmented with weight windows (WWs) to obtain sufficiently low variances in acceptable run times. In this paper, the random ray (RR) method is proposed as an approach to generating WWs. Compared to deterministic alternatives, it allows for straightforward spatial discretization and continuously samples angular phase space. The methods implemented were tested on two shielding problems, examining global variance reduction (GVR) and source-detector variance reduction. In both cases, the figure of merit of the RR-based methods is at least 50 times higher than analog MC, and in the GVR model, it is comparable to that of MAGIC, a popular fully MC approach to WW generation. The methods implemented are also robust to modeling choices like the source of multigroup cross sections (NJOY or MC), scattering anisotropy, and the order of the RR spatial source representation.