Analysis of Population Control Techniques for Time-Dependent and Eigenvalue Monte Carlo Neutron Transport Calculations

TL;DR

This paper reviews and analyzes various population control techniques for Monte Carlo neutron transport calculations, introducing a new perspective and evaluating their performance in terms of uncertainty, runtime, and accuracy.

Contribution

It provides a comprehensive review, theoretical analysis, and parallel algorithms for PCTs, and compares their performance in time-dependent and eigenvalue simulations.

Findings

Splitting-Roulette and Combing are the most performant techniques.

DD closely follows in performance.

Theoretical analysis quantifies uncertainty introduced by each PCT.

Abstract

An extensive study of population control techniques (PCTs) for time-dependent and eigenvalue Monte Carlo (MC) neutron transport calculations is presented. We define PCT as a technique that takes a censused population and returns a controlled, unbiased population. A new perspective based on an abstraction of particle census and population control is explored, paving the way to improved understanding and application of the concepts. Five distinct PCTs identified from the literature are reviewed: Simple Sampling (SS), Splitting-Roulette (SR), Combing (CO), modified Combing (COX), and Duplicate-Discard (DD). A theoretical analysis of how much uncertainty is introduced to a population by each PCT is presented. Parallel algorithms for the PCTs applicable for both time-dependent and eigenvalue MC simulations are proposed. The relative performances of the PCTs based on runtime and tally mean error or standard deviation are assessed by solving time-dependent and eigenvalue test problems. It is found that SR and CO are equally the most performant techniques, closely followed by DD.