b'Scalable DomainMultiphysics simulations support the advancement of clean energy solutions.Decomposition AlgorithmsT he main physics involved in nuclear reactors are neutronics, thermal-for Large-scale Monte Carlohydraulics, and fuel thermo-mechanics. Multiphysics simulations are only as accurate as the models used for each of these physics. Though Monte Neutron Transport SimulationsCarlo neutral particle transport methods are widely regarded as the gold-standard on Unstructured Mesh for neutronics simulations, they were long thought impractical for most types of nuclear reactor analysis outside of shielding calculations.However, with the current drive towards higher-fidelity nuclear reactor analyses, as well as the growing interest in small modular reactors, the method has become competitive with traditional deterministic transport algorithms for the same level of accuracy. Before such analysis can be practical, several algorithmic challenges must be TOTAL APPROVED AMOUNT:addressed, particularly with regards to the memory requirements of the method. The $125,000 over 1 year most common approach to these issues is to perform a domain decomposition to share PROJECT NUMBER:the memory burden among numerous nodes of a computing cluster. To date, most 20A1054-019 Monte Carlo codes do not offer domain decomposition as existing implementations lack practicality, performance, or scalability. This limits their use for large-scale simulations, PRINCIPAL INVESTIGATOR:for example of nuclear reactor cores in neutron transport simulations or full tokamak Guillaume Giudicelli models in shielding calculations. While most Monte Carlo codes use simplified CO-INVESTIGATOR: geometric models known as Constructive Solid Geometry, support for computer-aided Logan Harbour, INL design models and unstructured mesh in Monte Carlo codes is increasingly common and required to model complex geometries of advanced reactors or fusion devices. The memory requirements of an unstructured mesh exacerbate the memory issues of the Monte Carlo method and, therefore, the need for domain decomposition.This project first involved the development of a Monte Carlo neutron transport application, MaCaw, for the MOOSE framework, based on physics routines from OpenMC and ray-tracing routines developed in MOOSE for the method of characteristics transport. The application supports both shared memory and distributed memory parallelism. The latter is achieved using domain decomposition. In this situation, the tallies are also domain decomposed, splitting the memory cost of large simulations among all processes. This new application is being tested on a light water reactor full core benchmark. In its final phase, this project addressed the issue of communication of particles between domains and the issue of load balancing to improve scalability.TALENT PIPELINE:Robert Crowder, student at University of Tennessee KnoxvilleFuel pin partitioners available in the MOOSE ecosystem for Monte Carlo particle transport simulations. From left to right: equal volume, default (based on element numbering), 50 minimization of domain surfaces, regular axial grid.'