Accelerated sampling of the free gas resonance elastic scattering kernel

Abstract

In this work, we present the derivation and investigation of a new Doppler broadening rejection sampling approach for the exact treatment of resonance elastic scattering in Monte Carlo neutron transport codes. Implemented in OpenMC, this method correctly accounts for the energy dependence of cross sections when treating the thermal motion of target nuclei in elastic scattering events. The method is verified against both stochastic and deterministic reference results in the literature for 238 U resonance scattering. Upscatter percentages and mean scattered energies calculated with the method are shown to agree well with the reference scattering kernel results. Additionally, pin cell and full core k eff results calculated with this implementation of the exact resonance scattering kernel are shown to be in close agreement with those in the literature. The attractiveness of the method stems from its improvement upon a computationally expensive rejection sampling procedure employed by an earlier stochastic resonance scattering treatment. With no loss in accuracy, the accelerated resonant target velocity sampling algorithm is shown to reduce overall runtime by 3-5% relative to the Doppler broadening rejection correction method for both pin cell and full core benchmark problems. This translates to a 30-40% reduction in runtime overhead.

URANOS v1.0 – the Ultra Rapid Adaptable Neutron-Only Simulation for Environmental Research

“…An algorithm has to be applied which preserves the thermally averaged reaction rate, i.e., taking into account the Doppler effect on the cross-section evaluation. Such has been introduced by Gelbard (1979), whereas this modified version follows the implementation by Romano and Forget (2013) and Walsh et al (2014). By using the effect of thermal motion on the interaction probability…”

Resonance Scattering Treatment with the Windowed Multipole Formalism

“…The relative velocity sampling (RVS) method [38,31] was created to ameliorate the high rejection rates characteristic to the rejection algorithms used to model resonance upscatter. These schemes, in essence, sample probability distributions proportional to f (x)g(x), where f (x) is a distribution and g(x) ∈ [0, 1].…”