Accelerating Monte Carlo neutron transport by approximating thermal cross sections with functional forms

Monte Carlo methods suffer from a high computational cost. One bottleneck is looking-up cross sections, requiring a binary search and expensive memory access. In this paper, a simple method to accelerate this process is proposed. It is based on the fact that all cross sections are smooth in the thermal energy range, thus they can be easily fitted with a polynomial or exponential curve. Then, to retrieve a cross section for an energy lower than a certain threshold, a simple curve evaluation can take place. This method, implemented in the Monte Carlo code SCONE, resulted in a speed-up of up to 15% in some thermal reactor models, without impacting the memory usage. The approximation error introduced is generally insignificant, except in some cases where a small bias is observed in the results. Therefore the method is unsuitable for benchmark applications, whereas it could still be useful for other purposes.