National Center for Computational Sciences

Neutron transport inside a reactor core is a big topic in the energy modeling world, in both the fission and fusion fields. In general, a nuclear reactor will create a lot of neutrons from the initial reaction. Researchers hope to continue that process and to cause a chain reaction as the neutrons escape from the “epicenter” (i.e., neutron transport), to maximize the energy released within the system.

Denish Kaushik of Argonne National Laboratory and his collaborators are using 25 million hours on the Jaguar petaflops computer at ORNL to develop and test scalable algorithms, performance modeling and prediction, and parallel programming models, to model how neutrons speed around inside the reactor core. They are testing new codes to both simulate reactor cores and to reduce the current uncertainties in design and operation parameters.

The simulation of processes in reactor cores is challenging because of large length scales, a complicated distribution of materials, and the intricacies of the physical data. Calculating and simulating these processes requires simulation over several orders of magnitude and energy, and the resolution of strong local fluctuations. The work on Jaguar will demonstrate the capabilities of a new method of progressively eliminating existing approximations in reactor modeling. The UNIC code will provide reactor analysts the ability to remove as many approximations as possible. As the algorithms are refined, they will be used to solve coupled physics problems in such a way that thermal, hydraulic, and structural feedbacks are accurately represented in realistic reactor transient simulations. This will lead to a significant reduction in cost and better assessments of the safety of fast reactors—nuclear reactors where the fission is sustained by fast neutrons.