General Full-Frequency Convergence and Efficient Implementation in RMC of the Flight-Based Monte Carlo Method for Neutron Noise

Convergence is a critical challenge in Monte Carlo simulations of neutron noise, and existing solvers have not yet achieved full-frequency convergence. Building on Fauvel’s convergence analysis for simple systems, we generalized the full-frequency convergence proof of the flight-based method beyond simple systems, showing that for nonzero frequencies, the overall eigenvalue k˜ and the within-generation eigenvalue c˜ of the extended noise equation are strictly less than unity.We then developed a noise solver in the general-purpose Monte Carlo code RMC, implementing both flight-based and collision-based transport methods together with norm-based and real/imaginary-part–based roulette methods and a binning-based weight cancellation technique, and introduced targeted optimizations to address the known efficiency bottleneck of the flight-based method.Numerical tests were performed on two benchmarks spanning the full-frequency spectrum: a two-dimensional fuel assembly with a cross-section oscillation source and a one-dimensional rod with a mechanical vibration source, both evaluated with and without weight cancellation. In both cases, the flight-based method with norm-based roulette achieved full-frequency convergence, whereas the collision-based method may lose stable convergence at sufficiently high frequencies, even with increased η and weight cancellation; the derived expression for the extended space eigenvalue k˜ was further validated through power iteration computations in RMC.After optimization, the flight-based method matched or exceeded the efficiency of the collision-based method at low to intermediate frequencies and became clearly superior at very high frequencies, where it also retained robust convergence. These results indicate that the flight-based method with norm-based roulette is the preferred choice when full-frequency coverage, in particular high-frequency behavior, or guaranteed convergence robustness is required, while remaining competitive with the collision-based method at low to intermediate frequencies.