Nuclear Science and Engineering / Volume 155 / Number 2 / February 2007 / Pages 155-167
Technical Paper / Mathematics and Computation, Supercomputing, Reactor Physics and Nuclear and Biological Applications / dx.doi.org/10.13182/NSE07-A2653
A new hybrid Monte Carlo-Deterministic technique is presented for simulating global particle transport problems, in which flux estimates are desired at all physical locations in the system. This technique has two steps: First, an inexpensive deterministic global estimate of the forward flux is obtained; then Monte Carlo is used to estimate the multiplicative correction to the deterministic flux estimate. We call the multiplicative correction to the deterministic flux the correcton flux, and the Monte Carlo particles that estimate this flux correctons. For deep-penetration problems, the correcton flux has significantly less spatial variation than the physical flux. Therefore, the Monte Carlo process automatically distributes correctons much more uniformly across the system than it distributes Monte Carlo particles for the original angular flux. In the "deep" parts of the problem, at locations far from the source, this results in a greatly reduced variance and a greatly increased figure of merit.