Biasing secondary particle interaction physics and production in MCNP6

•Biasing secondary production and interactions of charged particles in the tabular energy regime.•Examining lower weight window bounds for rare events when using Russian roulette.•The new biasing strategy can speedup calculations by a factor of 1million or more. Though MCNP6 will transport elementary charged particles and light ions to low energies (i.e. less than 20MeV), MCNP6 has historically relied on model physics with suggested minimum energies of ∼20 to 200MeV. Use of library data for the low energy regime was developed for MCNP6 1.1.Beta to read and use light ion libraries. Thick target yields of neutron production for alphas on fluoride result in 1 production event per roughly million sampled alphas depending on the energy of the alpha (for other isotopes the yield can be even rarer). Calculation times to achieve statistically significant and converged thick target yields are quite laborious, needing over one hundred processor hours. The MUCEND code possess a biasing technique for improving the sampling of secondary particle production by forcing a nuclear interaction to occur per each alpha transported. We present here a different biasing strategy for secondary particle production from charged particles. During each substep, as the charged particle slows down, we bias both a nuclear collision event to occur at each substep and the production of secondary particles at the collision event, while still continuing to progress the charged particle until reaching a region of zero importance or an energy/time cutoff. This biasing strategy is capable of speeding up calculations by a factor of a million or more as compared to the unbiased calculation. Further presented here are both proof that the biasing strategy is capable of producing the same results as the unbiased calculation and the limitations to consider in order to achieve accurate results of secondary particle production. Though this strategy was developed for MCNP6 the technique can be leveraged in any fusion Monte Carlo code using library data with the condensed random walk algorithm.