Optimising differential data formats for Monte-Carlo radiation transport in the fusion regime

Monte-Carlo based radiation transport codes are widely used to simulate the behaviour of nuclear facilities, to aid their design, operation and inform on safety aspects. The accuracy, and hence reliability, of the output of transport codes is entirely dependent on the underlying nuclear data. The formalism used in the processed data directly impacts on the computational performance for such calculations. For example, low data density results in faster computation possibly made at the expense of accuracy and vice versa. A current standard transport code, MCNP, requires pre-processing and reformatting of raw nuclear data files (ENDF) before their use within the simulation. The incident particles angular distribution is usually processed to an equal probability histogram with 32 channels, or to a set of data points with linear interpolation across the angular and energy range. The former method is fast to sample, yet sacrifices the accuracy of the data representation. The latter method provides a better representation of the original format, but has a tendency to produce larger data files that are correspondingly slower to sample. This study considers the relative accuracies of these processing formalisms and the relationship to computational efficiency via an in-house, simplified Monte-Carlo code. For future code developments within the fusion community, we consider variations upon both methods in an effort to determine the optimal balance between computational efficiency and accurate data representation. We also attempt to quantify the effects of a lesser accuracy versus high computational burden in the simulation of fusion reactor design.