A proposed Kernel Density Estimator method for Monte Carlo eigenvalue calculations

A Kernel Density Estimator (KDE) method for Monte Carlo fission source iteration is described. KDE is used to estimate the fission source distribution at the end of each generation and realizations from the estimated source distribution are used as the starting locations for the next generation. The methodology is illustrated by applications to 1D and 3D configurations. The source convergence is measured by the relative source entropy. Significant source convergence improvement is observed for the proposed KDE method compared to the conventional Monte Carlo fission source iteration. Brief mention is made of the application of KDE to represent Monte Carlo tallies. (authors)