Non-intrusive Monitoring of Sealed Microreactor Cores Using Physics-Informed Muon Scattering Tomography With Momentum Measurements

TL;DR

This paper presents a physics-informed muon scattering tomography method for non-intrusive detection of missing fuel in sealed microreactor cores, demonstrating high sensitivity and robustness under realistic cosmic-ray conditions.

Contribution

The authors introduce $d$TRec, a novel framework combining muon trajectory reconstruction with Bayesian modeling, significantly improving defect detectability over existing methods.

Findings

Missing fuel flakes detected with 3 million muons at 50 mm voxel resolution.

Incorporating muon momentum enhances detection sensitivity by up to 150%.

The method outperforms PoCA by over 300% in detectability at the same muon count.

Abstract

Next-generation microreactors enable remote deployment and semi-autonomous operation, but compact, sealed, heterogeneous cores limit conventional safeguard approaches that rely on access and bulk accountancy. Limited inspection access and complex internal geometry reduce sensitivity to localized anomalies such as missing fuel. Here we demonstrate missing-fuel detection in microreactor scale geometries using muon scattering tomography under realistic cosmic-ray conditions. We introduce $\mu$TRec, a physics-informed framework that reconstructs event-level curved muon trajectories by combining a Gaussian multiple Coulomb scattering model with Bayesian updating, then maps scattering density through voxel wise M-values for core integrity verification. We evaluate a representative hexagonal core containing 61 fuel flakes with embedded control drums and shutdown rods, using both idealized 5 GeV muons and zenith-angle-dependent 0-60 GeV cosmic-ray spectra. A single missing fuel flake is detected with $3\times 10^{6}$ muons at 50 mm voxel resolution. Incorporating per-muon momentum further increases detectability by up to 149.85% for laser-driven sources and 105.11% for cosmic-ray sources relative to momentum-agnostic reconstruction. The approach remains robust under practical detector limits, with only an 8.88% reduction in detectability for 10 mm spatial resolution and 10% energy resolution. Compared with PoCA, $\mu$TRec delivers 326.13% to 392.14% higher detectability at equal muon counts, enabling faster defect identification.