- Main
Particle Transport in non-Markovian Stochastic Media
- Lioce, Dominic Andrew
- Advisor(s): Vujic, Jasmina L;
- Prinja, Anil K
Abstract
Particle transport in stochastic media is important in the simulation of many applications, including inertial confinement fusion, pebble bed nuclear reactors, atmospheric science, and supernovae explosions. Classical stochastic media transport models often assume Markovian mixing statistics, in which material chord lengths are exponentially distributed and the current material label is sufficient to determine the probability of encountering the next material interface. Many physically relevant random media, however, contain finite inclusions, grains, droplets, or other structured features whose chord length distributions are not necessarily exponential. In these non-Markovian media, the probability of crossing a material interface depends on the distance already traveled through the current material, so accurate transport models must retain additional geometric memory. This dissertation develops age-based particle transport methods for binary stochastic media described by an alternating renewal statistical process. The central idea is to augment the joint stochastic state space with an age variable, defined as the distance traveled since the previous material interface. For transport in purely absorbing media, this augmented state yields a Markovian description in the expanded stochastic state space, allowing for the derivation of a generalized master equation for transport in renewal media. A stochastic simulation algorithm based on this formulation is shown to reproduce benchmark leakage quantities, scalar flux moments, and transmission distributions for several non-exponential chord length distributions. These results demonstrate that replacing renewal statistics with a Markovian approximation can produce significant errors even when mean chord lengths and volume fractions are preserved. The age-based framework is then extended to include particle scattering in 1D planar geometry. Scattering introduces an additional transport memory effect because particles may reverse direction and move back toward previously sampled material. Several Monte Carlo algorithms are developed to approximate the age update after scattering, including the Backscatter-Resample and Age-Decreasing methods, and these methods are compared with the heuristic renewal statistics extensions of the Markovian stochastic media algorithms Chord Length Sampling (CLS) and Local Realization Preserving (LRP). The results show that methods preserving more local geometric memory generally improve accuracy relative to the benchmark in leakage values and scalar flux, with the Age-Decreasing method often producing accuracy comparable to LRP while retaining a simpler memory structure. The methods are also extended to 3D stochastic media and tested against explicit spherical inclusion benchmark calculations. In 3D, material interfaces are sampled along full particle trajectories, and the treatment of local geometric memory after scattering becomes approximate because scalar interface memory cannot fully describe the surrounding 3D geometry. Deterministic and probabilistic memory-update treatments are compared, and the probabilistic rule is adopted for consistency with previous 3D LRP implementations. The 3D results show the same broad method ordering observed in 1D, with the Age-Decreasing method and LRP generally improving accuracy relative to the Backscatter-Resample method and FLRP. The spherical inclusion benchmarks also show that a randomly distributed radius can influence scalar flux and leakage, emphasizing the importance of accurate characterization of real-world stochastic media transport applications. Overall, this work provides an extension of stochastic media transport that preserves prescribed chord length distributions for renewal statistical media more directly than previous heuristic adaptations of Markovian models. The resulting age-based methods provide a computationally efficient framework for modeling ensemble-averaged transport in non-Markovian media.