Congratulations to Jonathan Crozier, doctoral student at North Carolina State University Department of Nuclear Engineering, and recipient of the US Department of Energy (DOE) Innovations in Nuclear Energy Student Paper Award. He is 1 of 14 students from 10 universities recognized for published graduate work on the leading edge of nuclear energy research.

“It is an immense honor for this work to be recognized by the DOE-NEUP, and I’m grateful for the guidance and collaboration with Dr. Ayman Hawari which have allowed this work to flourish (even beyond the results presented in the awarded paper). I’m also grateful for the support of the LEIP group, and early PhD mentors (Dr. N. Colby Fleming and Dr. Cole A. Manring) who set a high standard for engineering creativity and excellence. This award is particularly meaningful as it coincides with finalizing my PhD and enhances my ability to support my family as a graduate student”. 

Crozier’s paper, “Embedding Neural Thermal Scattering (NeTS) Modules in Serpent for Higher Fidelity Advanced Reactor Analysis”, examines the following –

When a neutron born in fission thermalizes to the order of , it’s de-Broglie wavelength and energy approach the order of interatomic spacing and elementary lattice oscillations, respectively. S(a,b,T), or the scattering law, quantify these temperature-dependent crystallographic contributions to total cross section (or reaction rate). In a Monte Carlo analysis, cumulative distribution functions (CDFs) of S(a,b,T) are loaded to memory from “A Compact ENDF” (ACE) files for stochastically selecting thermal scattered neutron trajectories. In this work, novel neural thermal scattering (NeTS) modules for S(a,b,T) CDFs are designed, trained, serialized, and embedded within SERPENT using Python’s limited C-API for on-the-fly deployment of crystalline graphite S(a,b,T) sampling. Torchscript tracing and Numba just-in-time (JIT) compilation streamline neural inference on NVIDIA GPUs with CUDA libraries.

Demonstrations of bare sphere thermalization of fast and thermal sources show excellent agreement between embedded NeTS in SERPENT and MCNP. With an explicit model of the reactor, NeTS can predict on-the-fly changes in TREAT neutron spectra as a function of local temperature, which can serve to improve transient and accident predictions in a multiphysics analysis framework. This framework can be further extended to account on-the-fly for changes in local graphitic microstructure to scattering cross sections and outlines a novel coupling of modern machine learning with state-of-the-art reactor physics methods.