Convergence Testing for Mcnp5 Monte Carlo Eigenvalue Calculations

Determining convergence of Monte Carlo criticality problems is complicated by the statistical noise inherent in the random walks of the neutrons in each generation. The latest version of MCNP5 incorporates an important new tool for assessing convergence: the Shannon entropy of the fission source distribution, Hsrc. Shannon entropy is a well-known concept from information theory and provides a single number for each iteration to help characterize convergence trends for the fission source distribution. MCNP5 computes Hsrc for each iteration, and these values may be plotted to examine convergence trends. Convergence testing should include both keff and Hsrc, since the fission distribution will converge more slowly than keff, especially when the dominance ratio is close to 1.0.