Monte Carlo–Informed Surrogate Model for the Analysis of Decay Heat Behavior in Molten Salt Reactors

Noble metal fission products in molten salt reactors (MSRs) deposit on primary loop surfaces, where their decay heat becomes safety-relevant during drain scenarios: Once the salt is discharged, the deposit remains without active cooling and can drive localized overheating. Predicting it requires tracking the full nuclide inventory across core, bulk, and deposition regions, for which repeated Monte Carlo (MC) depletion is prohibitive. This work develops a MC-informed surrogate built on a sparse transition operator spanning neutron-induced reactions, decay, advection, and deposition, with region-wise mass transfer correlations setting the deposition rates. The neutron-induced rates are extracted from a single MC simulation and rescaled with power under a constant-spectrum assumption, so that the ordinary differential equations (ODEs) system is solved without further transport runs, for a substantially larger nuclide set than historical zero-dimensional models. Applied to the Molten Salt Reactor Experiment (MSRE) fuel loop, the predicted mass transfer coefficients reproduce the original MSRE values, and the calculated 132Te distribution is comparable across surfaces for both the 233U and 235U campaigns without per-nuclide tuning. Benchmarked against a self-consistent MC workflow on a simplified generic MSR case, the surrogate reproduces the total decay heat of both the bulk and deposition nodes from sub-second to yearly time scales, to within ±0.3% at fixed power and ~2% under sinusoidal modulation. It does so at a cost low enough for design exploration and for a digital-twin representation of the primary loop. An uncertainty quantification shows that fission yields dominate the nuclear data budget and that the deposited inventory is markedly more uncertain than the bulk, partly due to non–noble-metal precursors in the yield normalization constraint.