A fission collision separation method for efficient incident flux response expansion coefficient generation

The 3D C5G7 benchmark problem was used to test the accuracy and efficiency of the new response coefficient generation method at both lattice (local) and core (global) levels. For lattice calculations, the response coefficients computed by the new method were compared with those computed by the direct Monte Carlo method. It was found that the truncation errors of the fission neutron separation method at expansion order 3 are much lower than the stochastic uncertainties of the response coefficients and consequently can be ignored. The average truncation errors of the new method ranged from 0% to 0.004% and 0% to 0.08% for the surface-to-surface and fission density response coefficients, respectively. For core level calculations, the COMET whole-core solutions based on the response coefficient libraries generated by the new method and the interpolation method were compared with direct (reference) Monte Carlo calculations using the MCNP code. It was found that the error in the eigenvalues predicted by COMET based on the new method is within 11-26 pcm for three core configurations with and without control rods. These eigenvalue results are about 15 pcm more accurate than those based on the interpolation method, although still within the overall uncertainty. The corresponding assembly and pin fission densities are almost identical for the two methods. These comparisons indicate that the new method is as accurate as or slightly more accurate than the interpolation method, depending on the comparison parameter. The new method was found to be about 5 times faster than the interpolation method for the generation of the response function library.