Conference Agenda

Research reactors such as the Jules Horowitz Reactor (JHR), under construction at CEA-Cadarache or the High Flux Isotope Reactor (HFIR), in operation at Oak Ridge National Laboratory, play an essential role in radioisotope production, fuel and material qualification, and neutron-based research. Despite their relevance they have received considerably less attention to our knowledge, than power reactors in comprehensive sensitivity and uncertainty (S/U) studies. This work helps to fill this gap by presenting a systematic S/U analysis of nuclear data (ND) for these reactors, focused on the effective neutron multiplication factor and the main reactivity safety coefficients.

The analysis has been performed with the help of SUMMON (Sensitivity and Uncertainty Methodology for Monte Carlo codes), a deterministic tool developed at CIEMAT that propagates ND uncertainties to reactor responses through the sandwich rule. Sensitivity profiles are obtained from MCNP6.2 (via KSEN cards processed into the SDF format introduced by SCALE), and covariance matrices are taken from the most recent major evaluations: JEFF-3.3, JEFF-4, ENDF/B-VIII.0, JENDL-4.0u and JENDL-5. A 281-group energy structure (Santamarina–Hfaiedh) is used to properly resolve the thermal region of these thermal systems.

For the JHR, calculations rely on a MCNP model provided by CEA in a start-of-cycle configuration with mixed fresh and irradiated U₃Si–Al fuel and a beryllium reflector. Some of the results are compared with the previous CEA study performed within the SANDA project [Truchet, 2022], which used a covariance library that is not publicly available. For HFIR, the ORNL homogenized MCNP model [Ilas et al., 2015] is used at Beginning- and End-of-Cycle to evaluate k_eff, the void reactivity worth in several configurations, and temperature coefficients perturbing the fuel region, the flux trap, and the system isothermally; the void results are additionally compared with the HFIRCE-4 historical experimental measurements [Cheverton and Sims, 1971].

Across both reactors, the total ND-driven uncertainty in k_eff is found to be systematically of the order of 1% (~800–1000 pcm), well above the 300 pcm target accuracy requirement traditionally adopted in design studies. The uncertainty is dominated by ²³⁵U ν̄ and χ, but a distinctive feature of these thermal research reactors — and a notable difference with respect to fast research reactors — is the strong contribution of moderator and structural materials (¹H, ²⁷Al, ¹⁶O). Temperature coefficients are also analysed: the Doppler coefficient is comparatively well constrained in JHR (ND uncertainty of 1.0–1.3% across fuel-temperature variations from 200 to 700 K), while in HFIR its very small magnitude — due to the high enrichment and low ²³⁸U content — precludes a reliable extraction of the ND uncertainty. Coolant- and moderator-driven temperature coefficients exhibit larger ND uncertainties, reaching ~5% in JHR and 1–3% across the HFIR scenarios, with the cross-correlation between ²³⁵U (n,f) and (n,γ) emerging as a leading contributor.

The systematic intercomparison of the five covariance libraries reveals significant inconsistencies, especially in ²³⁵U ν̄ and χ, in the cross-correlations between (n,f) and (n,γ) of ²³⁵U, and in (n,γ)–(n,p) of ²⁷Al, where some evaluations do not even provide the data. These discrepancies translate into a non-negligible spread of the propagated uncertainty depending on the library used, which is a relevant concern for benchmarking and design margins in MTRs.

The contribution presents the full methodology, the results for both reactors with their main uncertainty contributors, the consistency checks across libraries, and the resulting recommendations regarding the cross sections: ²³⁵U ν̄ and χ covariances and ²⁷Al cross-reaction data emerge as priority targets for future ND improvement efforts in support of research-reactor analysis.
References:

R. D. Cheverton and T. M. Sims. HFIR Core Nuclear Design. Oak Ridge, TN, 1971. Tech. rep., ORNL-4621. Oak Ridge National Laboratory
G. Ilas, D. Chandler, B. J. Ade, et al. Modeling and Simulations for the High Flux Isotope Reactor Cycle 400. Oak Ridge, TN, USA, 2015. ORNL/TM-2015/36. Oak Ridge National Laboratory
G. Truchet. JHR Reactivity Sensitivities to Nuclear Cross-Sections. 2022-09-30. Deliverable (Report), D5.3, Work Package WP5. SANDA (Solving Challenges in Nuclear Data for the Safety of European Nuclear facilities).

Advanced computational methodologies are becoming increasingly important for the design, optimization, and operation of modern research reactors. Multipurpose reactor systems with complex geometry and diverse operational objectives require high-fidelity neutronics and multiphysics analysis tools capable of efficiently evaluating reactor performance, neutron utilization, thermal-hydraulic behavior, and control system dynamics. In particular, compact plate-type research reactors present unique challenges for neutron transport, burnup analysis, and coupled multiphysics simulations due to their heterogeneous geometries.

This work, which supports the POLARIS (Physics-Oriented Learning and AI Reactor Integration System) project begins with the application of the RAPID-AI code system to the Brazilian Multipurpose Reactor (RMB) to evaluate the capability of fission matrix methodologies to accurately represent uranium silicide plate-type fuel assemblies. The analyses investigate neutron flux behavior, neutron transport characteristics, and future burnup modeling capabilities while establishing a computational framework for efficient high-fidelity reactor analysis.

In parallel, thermal-hydraulic and conjugate heat transfer (CHT) analyses are being performed using OpenFOAM to investigate coolant behavior, fuel temperature distributions, and hot-channel thermal-hydraulic performance in representative plate-type fuel geometries. Ongoing efforts additionally focus on the development of reduced-order multiphysics methodologies suitable for efficient coupling between neutronics and thermal-hydraulic simulations for digital twin development, autonomous reactor control, and real-time reactor analysis.

The Molten Salt Reactor Experiment (MSRE), operated at Oak Ridge National Laboratory from 1965 to 1969, remains the only liquid-fueled molten salt reactor to have operated at reactor scale, achieving criticality and sustained operation with online fuel management. This work presents a hyper-fidelity computational model of the entire MSRE reactor, developed without geometric simplifications and designed to capture tightly coupled neutronic and thermal–hydraulic phenomena with high spatial resolution.

The numerical model is constructed from open-source CAD geometry. All major internal components are explicitly resolved, including the reactor vessel, fuel inlet volute, downcomer (annulus), lower plenum with anti-swirl vanes, grid support structures, graphite lattice blocks, graphite stringers, control rods, the full array of fuel passages formed by the moderator assembly, and the upper plenum. The graphite core is modelled as an explicit heterogeneous structure, reproducing the formation of 1140 fuel channels.

The steady-state solution is obtained through a fully coupled iterative multiphysics strategy. Spatial power distributions are computed using the Serpent2 Monte Carlo neutron transport and a reference thermal power of 7.4 MW. Serpent2 is directly interfaced with the OpenFOAM-based mesh used in the thermal–hydraulic solver, allowing consistent transfer of temperature and density fields for cross-section feedback. Thermal–hydraulic simulations are performed using the foamForNuclear multiphysics framework, which solves the compressible Navier–Stokes equations with heat transfer and internal volumetric heat generation in both fluid and solid domains. Turbulence is modelled using the k–ω Shear Stress Transport closure, enabling resolution of complex recirculation zones and swirl-driven structures within the vessel.

The coupling procedure follows an iterative fixed-point approach. An initial Monte Carlo calculation provides the spatially resolved power field, which is passed to the thermal–hydraulic solver to compute velocity, temperature, and density distributions. The temperature and density fields are then returned to Serpent2 to update temperature-dependent cross sections and recompute the power distribution. The process is repeated until convergence is achieved, defined as a relative change in temperature distributions below 1%, accounting for feedback-induced power variations.

The results demonstrate strong agreement with available experimental measurements and historical reports. The predicted velocity distribution in the volute aligns well with experimental data, reproducing the angular variation of flow entering the downcomer. Swirl generated in the volute persists through the annular region and is subsequently attenuated by the lower plenum anti-swirl vanes. While the simulations exhibit a slightly non-uniform flow profile in the annulus compared to the more uniform experimental observations, the differences remain modest and do not significantly influence core inlet conditions. The model also captures the formation of a weak off-axis recirculation zone driven by asymmetries in the inlet distribution. Predicted volumetric flow rates in the fuel channels show good agreement with experimental measurements. The simulations reproduce the observed trend in channel-dependent flow behaviour, though the directional asymmetry is less pronounced than reported experimentally. Maximum fuel power occurs slightly off-centre, influenced by the presence of control rod thimbles and the central sample basket. Within the graphite moderator, peak power occurs at the core centre. Axial temperature profiles in the hottest channel closely match historical nuclear analyses based on simplified multi-region representations of the reactor core. The highest temperatures occur below the top of the core and off-axis, consistent with the combined effects of power distribution and flow patterns.

Overall, the hyper-fidelity heterogeneous MSRE model reproduces key hydrodynamic, neutronic, and thermal characteristics of the reactor with strong agreement to available data. The study demonstrates the feasibility of fully resolved, reactor-scale multiphysics simulation for complex liquid-fueled systems. Beyond reconstructing historical behaviour, the framework establishes a validated computational approach applicable to modern small modular reactor designs, enabling high-resolution assessment of coupled feedback mechanisms and supporting future digital twin development for advanced reactor technologies.