Benchmarking of emergent radiation transport codes for fusion neutronics applications

Abstract

The accurate and efficient mapping of the radiation environment in a nuclear fusion reactor requires the most advanced radiation transport tools. The Monte Carlo method has long been deployed to deal with the complexity of fusion relevant geometries, with MCNP the adopted industry standard code among the European and wider international community. However, reliance on a single code has driven explorations into alternatives to establish their capabilities and maturity for fusion analyses. These alternatives are being actively developed and increasingly adopted in some types of applied analysis by the user community. In this paper, we explore both experimental and computational benchmarks in order to examine the capabilities of Serpent, OpenMC as well as the framework for allowing CAD based particle transport, DAGMC, over a broad range of fusion relevant nuclear responses and geometries. This spans from more simple parametric models adopted in reactor scoping studies to the current ITER reference model which has been successfully translated to Monte Carlo codes other than MCNP using an open source utility, csg2csg. An assessment for both CSG and CAD based workflows has been conducted as well as a hybrid approach combining the two. The FNG HCPB, Cu and a subset of the FNS experiments were also converted to Serpent and OpenMC input files for comparison of calculation to available experimental data. Good agreement was observed across all codes for the determined tritium production rates and activation foil measurements. Potentially more efficient workflows for complex tokamak models are detailed. For a heterogeneous model of JET octant 1, an optimised CAD based model in Serpent is over 50% faster than the MCNP CSG equivalent model. In the case of the generated OpenMC model of ITER, an order of magnitude reduction in simulation time, including a model loading time of the order of minutes, is reported. Such validation and benchmarking activities should in the future be integrated as part of a testing suite for which a basic framework has been demonstrated here. To conclude, the current limitations and required development are outlined as well as identifying where each code may specialise for a particular application. Based on this future work relevant to both the developers and user community is briefly discussed.