Argonne’s Computational Science Division is researching novel algorithms and techniques for performing particle transport aimed at applications in nuclear energy and high energy physics. Our research focuses on enabling simulation of physical phenomena with higher fidelity than is currently possible through the coupling of particle transport to other systems of equations, such as computational fluid dynamics, structural mechanics, and activation/transmutation. These multiphysics simulations, while enabling predictive simulation, are computationally expensive; thus, a central component of our research is to ensure that simulations can make effective use of leadership-class supercomputing facilities—including traditional CPU architectures, GPUs, and other next-generation architectures.

While various solution methods for particle transport have been formulated, the Monte Carlo method is the most accurate because as it can exactly simulate particle behavior in a system free from approximations. In certain fields, however, there is limited use of the Monte Carlo method because of its computational cost. Improving the efficiency, usability, and parallel scaling of Monte Carlo particle transport methods is one of the primary goals of our research and development activities in this area. In addition, we are exploring new, highly efficient solution methods, like the random ray method, that can provide solutions for time-dependent problems for which the use of the Monte Carlo method would require an unacceptably long time to solution.