Preamble

Evaluation of CENDL-3.2 and CENDL-TMSR-V1 on Zero Power Benchmark of Molten Salt Reactor Experiment

Jing-shen Zhao,¹,² Jian Guo,³,² Rui Yan,³,² Yang Zou,³,² Gui-feng Zhu,³,² and Xue-chao Zhao³

¹State Key Laboratory of Thorium Energy, Shanghai Institute of Applied Physics, Chinese Academy of Sciences, Shanghai 201800, China
²University of Chinese Academy of Sciences, Beijing 100049, China
³National Key Laboratory of Thorium Energy, Shanghai Institute of Applied Physics, Chinese Academy of Sciences, Shanghai 201800, China

The 2019 edition of the International Reactor Physics Evaluation Project (IRPhEP) Handbook includes the Molten Salt Reactor Experiment (MSRE) benchmark, whose calculation results comprise the effective multiplication factor (k_eff) and control rod worth obtained using several evaluated nuclear data libraries, including ENDF/B-VII.1, ENDF/B-VIII.0, etc. The benchmark refers to the first criticality experiment at zero power, with stationary salt and uniform temperature using ²³⁵U fuel. This benchmark serves as the first comprehensive reference case for Molten Salt Reactors (MSRs) and has been widely used to assess the consistency and accuracy of Monte Carlo codes and nuclear data libraries in molten salt reactor simulations.

Since 2011, the Thorium Molten Salt Reactor (TMSR) nuclear energy system has been under development at the Shanghai Institute of Applied Physics, Chinese Academy of Sciences, with the aim of facilitating thorium utilization. To support this project, the China Nuclear Data Center has developed a nuclear data library optimized for the thorium-uranium fuel cycle, CENDL-TMSR-V1. However, the verification of CENDL-3.2 and CENDL-TMSR-V1 for molten salt reactor simulations remains unexplored.

In this work, a high-fidelity MSRE model was developed with OpenMC in reference to the benchmark model, and comparative analyses were carried out with four evaluated nuclear data libraries: ENDF/B-VII.1, ENDF/B-VIII.0, CENDL-3.2, and CENDL-TMSR-V1. The accuracy of the model developed in this work is proven by a discrepancy of 34 pcm in k_eff and the consistency of neutronic parameters, including neutron spectrum, axial neutron flux and production distributions, rod-shadowing effect, and control rod worth, with ENDF/B-VII.1, which was used in the benchmark model calculations. Based on the verified model, the applicability of CENDL libraries was validated by comparisons of neutronic parameters, including k_eff, control rod worth, radial neutron flux distribution, and depletion calculation results, with those obtained using ENDF/B-VII.1. The k_eff obtained with CENDL-TMSR-V1 is 56 pcm lower than that obtained with ENDF/B-VII.1, while that of CENDL-3.2 is 72 pcm lower. The maximum rod bank worth deviation between CENDL libraries and ENDF/B-VII.1 is approximately 50 pcm.

Keywords: Molten salt reactor experiment, Zero-power experiment, OpenMC, CENDL-TMSR-V1, CENDL-3.2

Introduction

As one of the Generation IV reactor systems, MSRs feature revolutionary technological advantages because of liquid fuel. First, MSRs use molten fluoride salts as fuel, eliminating the fabrication of traditional solid fuel elements and thereby significantly reducing fuel costs. Second, MSRs have a large negative temperature coefficient of reactivity, which inherently suppresses chain reactions as core temperature rises, making the reactor self-regulating. Third, molten salt reactors have online refueling and reprocessing capability that allows prompt removal of fission products via helium purging and chemical separation, not only maintaining low excess reactivity but also providing a new way of dealing with long-lived radionuclides. For example, a small modular molten salt reactor started with plutonium can both transmute plutonium from pressurized water reactor spent fuel and achieve efficient thorium utilization with the ²³³U-²³²Th cycle. In addition, the liquid fuel system achieves higher fuel burn-up compared to solid fuel reactors, improving resource utilization, which serves as a critical technological foundation for sustainable nuclear energy.

MSR engineering practice began with the MSRE at Oak Ridge National Laboratory (ORNL) in the 1960s. Operating from 1965 to 1969 with LiF-BeF₂-ZrF₄-UF₄ fuel salt, the MSRE achieved criticality, accumulated over 13,000 operating hours, and reached thermal power levels of up to 7.3 MWth. It successfully validated the chemical stability of the fuel salt and the viability of technologies such as online refueling. MSRE experimental data remain the "gold standard" for validating molten salt reactor physics models. Its benchmark is not only applicable to liquid-fueled reactors but also provides important references for novel designs such as the Molten Fluoride Fast Reactor (MFFR).

Since 2011, led by the Shanghai Institute of Applied Physics (SINAP) of the Chinese Academy of Sciences, research on thorium-based molten salt reactors has been conducted, pioneering the closed thorium-uranium fuel cycle. In October 2023, the 2 MWth liquid-fueled thorium molten salt reactor designed and built by SINAP achieved criticality for the first time, marking SINAP as a major force in molten salt reactor research.

Evaluated nuclear data significantly influence the accuracy of reactor physics simulation results. Major international evaluated nuclear data libraries such as ENDF/B-VIII.0, JEFF-3.3, and JENDL-5 have significantly improved the accuracy of actinide cross sections, fission yields, and thermal scattering data through continuous updates. For example, ENDF/B-VIII.0 refined neutron capture cross sections for uranium and plutonium, JEFF-3.3 improved decay data evaluations, and JENDL-5 integrated photonuclear data. These advances provide a robust basis for the design of complex MSR systems.

The China Nuclear Data Center (CNDC) developed the CENDL-3.2 library, whose key nuclides in nuclear applications, such as U, Pu, and Th, have been revised and improved. It has demonstrated excellent performance in shielding calculations for pressurized water reactors (PWRs) and fast critical benchmark calculations. Simulations of the CNP-1000 reactor at the Fuqing Nuclear Power Plant confirmed that CENDL-3.2 satisfies the requirements for PWRs. To address the specific needs of the TMSR nuclear energy system, CNDC developed the specialized library CENDL-TMSR-V1. Validated against 1,245 criticality benchmarks from the International Criticality Safety Benchmark Evaluation Project (ICSBEP) and LR-0 benchmarks, CENDL-TMSR-V1 achieves calculation errors below 0.5% in benchmarks containing ²³³U, Li, and F. However, errors exceed 0.5% in systems containing Th/Pu, indicating the need for further refinement of thorium nuclear data.

International research demonstrates the indispensable role of benchmarks in nuclear data validation. For example, the minimum plutonium loaded for a small modular MSR using plutonium as the initial fuel (SM-MSR-Pu) criticality during 2000 days of operation revealed discrepancies of up to 10% between JEFF-3.1 and JENDL-4.0. In liquid metal fast reactor simulations, ENDF/B-VIII.0 and JEFF-3.3 show deviations in k_eff of more than 1.7σ, mainly due to different assessments of plutonium fission. The MSRE benchmark is emerging as a prime validation resource due to its uniquely detailed documentation of operational parameters.

This study evaluates the performance of CENDL-3.2 and CENDL-TMSR-V1 for the MSRE benchmark. A high-fidelity model of the MSRE first criticality experiment is constructed using OpenMC, and its accuracy is verified through comparisons of neutronic parameters with those of the benchmark model. The HDF5 format nuclear data required by OpenMC to solve continuous-energy problems were produced by converting ACE files produced with the NJOY nuclear data-processing system that converted raw ENDF/B data into linearly-interpolable data. The validation of CENDL-3.2 and CENDL-TMSR-V1 through the MSRE benchmark provides critical insights for optimizing domestic nuclear libraries and advancing MSR design.

The paper is organized as follows: Section II details the core structure and the model developed in this work for the MSRE first criticality experiment, emphasizing the geometric differences between this model and the benchmark model. Section III introduces OpenMC and CENDL libraries and describes the workflow for generating continuous-energy cross sections in HDF5 format for OpenMC calculations. Section IV comparatively analyzes the performance of different libraries in the MSRE benchmark, including k_eff, control rod worth, radial neutron flux distribution, and results from the customized depletion calculation. Finally, Section V summarizes this study.

II. High-Fidelity OpenMC Model of MSRE

The 2019 edition of the IRPhEP includes a model of the MSRE first criticality experiment. The reactor neutronic parameters, such as control rod worth and the reactivity coefficient of ²³⁵U concentration, calculated based on this model show good agreement with experimental values. Based on this model, researchers have used the Monte Carlo codes Shift of SCALE, MCNP, and Serpent with ENDF/B-VII.0, ENDF/B-VII.1, and ENDF/B-VIII.0 to calculate k_eff. These studies compared the results of different Monte Carlo codes and evaluated nuclear data libraries.

In reference [34], a detailed comparison of this model and a CAD model developed by Copenhagen Atomics, which is available as an OpenMC code, was conducted using OpenMC. The CAD model establishes the exact geometry of the lower head instead of using the homogenization treatment in the benchmark model. The k_eff of the CAD model is approximately 1% lower than the benchmark model, demonstrating closer agreement with experiment. The computed k_eff is 2.154% higher than the experimental value, which is a 5σ difference. The origin of this discrepancy has remained an open question. However, the primary objective of this study is to evaluate the performance of CENDL-3.2 and CENDL-TMSR-V1 in the MSRE benchmark. Therefore, we developed a constructive solid geometry (CSG) model referring to the benchmark model and compared the results with those of the benchmark model.

The materials of the model in this work are consistent with those of the benchmark model, which were obtained from the available CAD model. The active zone of the MSRE core consists of densely arranged graphite stringers, as shown in Fig. 2(a), with fuel salt flowing through the channels formed by grooves on the sides of the graphite stringers. The central area of the graphite array contains three control rods and a sample basket used for reactivity control and material irradiation testing, respectively.

The core configuration during the first MSRE criticality experiment was: control rod 2 and control rod 3 fully withdrawn (129.54 cm), control rod 1 inserted to 3% of its integral worth (118.364 cm), with a nominal ²³⁵U load of 65.25 kg and a fuel salt flow rate of 4.54 m³/min. The benchmark model simplified the fuel salt as a static medium with uniform temperature distribution. The fuel salt composition at the first criticality experiment is presented in Table 1. The detailed geometric parameters are the hot dimensions in reference [35].

Figure 1 illustrates the geometry of the MSRE model in both the axial and radial directions, including detailed representations of the control rod and sample basket. Although the primary dimensions are consistent with those of the benchmark model, differences exist in certain details, such as the shape of the graphite stringer top and the horizontal graphite lattice.

In the MSRE core, the graphite stringer top adopts a pyramidal shape to prevent the accumulation of residual fuel salt during draining, whereas the benchmark model simplifies this structure to a conical shape. The present model employs the pyramidal configuration, as depicted in Fig. 2(c). The horizontal graphite lattice, illustrated in Fig. 3, has a thickness of 2.54 cm and features support rod holes with a diameter of 2.642 cm arranged in a cruciform pattern. The benchmark model omits the gaps between the lattice holes and the support rods, whereas the present model takes these gaps into account, thereby providing a more accurate representation of real conditions.

The OpenMC calculations employed 100,000 particles per generation, with 700 active and 50 inactive generations. As summarized in Table 2, the k_eff discrepancy caused by the configurations of this model and those of the benchmark model is only 4 pcm, which lies well within one standard deviation, indicating that these modeling details have a negligible impact on reactivity.

Table 3 compares the k_eff between the model in this work, the benchmark model, and other MSRE zero power benchmark Monte Carlo models. With the same ENDF/B-VII.1 library as the benchmark model, the present model yields a k_eff that is 34 pcm higher than the IRPhEP reference. Figure 4 presents comparisons of the core neutron energy spectrum, axial neutron production distribution, and axial neutron flux distribution between the present model and the benchmark model. The close agreement in the core energy spectrum, as well as axial neutron production and flux distributions, demonstrates the fidelity of the present model. Owing to geometric differences relative to the benchmark model, the axial neutron production distribution of the model in this paper exhibits some deviations at the horizontal graphite lattice and the graphite stringer top.

III. Development of HDF5 Data Library

A. OpenMC

OpenMC is a community-developed Monte Carlo neutron and photon transport code. It can perform fixed source, k-eigenvalue, and subcritical multiplication calculations on models built using either constructive solid geometry or CAD representation, enabling high-fidelity modeling of nuclear reactors and other systems. A flexible and efficient tally system enables a wide variety of physical quantities to be tallied and analyzed. Development of OpenMC was spearheaded by the Computational Reactor Physics Group (CRPG) at the Massachusetts Institute of Technology (MIT). Owing to its open-source nature, OpenMC has been a popular choice for coupling schemes in reactor design and analysis. Through a coupling scheme between OpenMC and another open source CFD code, OpenFOAM, researchers simulated and analyzed three proposed geometries of the core and fertile blanket to optimize the conceptual geometric design of the MSFR.

With the effort of the community, a depletion solver has been implemented in OpenMC (available in version 0.11 or later) and verified. The depletion solver enables an in-memory transport-depletion coupling. In this coupling scheme, the nuclear reaction rates in depletable zones, which determine the rate at which nuclides are transmuted, are scored in the transport calculation and then used to solve the Bateman equations in the burnup calculation. While the code is relatively young, it is already being used in a number of advanced R&D projects including the Consortium for Advanced Simulation of LWRs and the ANL Center for Exascale Simulation of Advanced Reactors. In this study, OpenMC version 0.15.1 was used, not only for calculations but also for the generation of continuous energy cross section libraries in HDF5 format from the CENDL libraries, because it provides a convenient Python API.

B. Introduction to CENDL Libraries

CENDL-3.2 is a Chinese Evaluated Nuclear Data Library officially released on 12 June 2020. It adopts the internationally standardized ENDF-6 format and covers 272 nuclides with neutron incident energies ranging from 10⁻⁵ eV to 20 MeV. Compared with CENDL-3.1, the data of key nuclides in nuclear applications like U, Pu, Th, Fe, etc. have been revised and improved. Moreover, model-dependent covariance data for main reaction cross sections are added for 70 fission product nuclides including ¹³⁵I and ¹³⁷Cs. Validated against 1261 critical benchmark experiments in ENDITS-1.0, CENDL-3.2 shows good agreement between calculated results and experimental data for various nuclear systems.

Compared to the JEFF-3.3 and ENDF/B-VIII.0 libraries, CENDL-3.2 performs better in the calculation of ²³³U assemblies. However, for the pusl11 series of Pu devices, CENDL-3.2 has poor performance in criticality calculations and needs further improvement. To satisfy the nuclear data requirements for the design and fuel cycle analysis of the TMSR nuclear energy system, the China Nuclear Data Center has developed a nuclear library for the thorium-uranium cycle, namely CENDL-TMSR-V1. Most of the data in CENDL-TMSR-V1 are selected from international evaluated nuclear data libraries, including CENDL-3.1, ENDF/B-VII.0, ENDF/B-VII.1, JENDL-4.0, JEFF-3.1, and IAEA/ADS-2.0, with the primary criterion of suitability for TMSR physics design. For the key nuclides in the Th-U cycle, ²³²Th and ²³³U, the nuclear data were improved on the basis of CENDL-3.1. The radiative capture cross section and resonance parameters of ²³²Th were re-evaluated. For ²³³U, the resonance parameters, fission cross section, elastic and inelastic scattering cross sections, and elastic angular distribution were improved. Through comparative selection, CENDL-TMSR-V1 was ultimately established, comprising 403 nuclides, including light nuclides, structural materials, fission products, and fission nuclides.

C. Workflow of HDF5 Data Library

To obtain the continuous-energy data of CENDL libraries in HDF5 format required for OpenMC calculations, the following steps are performed. The processing workflow is illustrated in Fig. 5. OpenMC provides the openmc.data module in its Python API, and the IncidentNeutron.from_njoy() method can be used to create OpenMC data instances. For a target nuclide, the IncidentNeutron.from_njoy() method converts raw ENDF/B data into linearly-interpolable ACE format data by running NJOY. Then, the export_to_hdf5() function was used to export incident neutron data to an HDF5 file. The cross section data libraries were finalized after all HDF5 files were collected.

In generating linearly-interpolable data with NJOY, the same seven temperature points used for the continuous-energy neutron cross section library of ENDF/B-VII.1 were adopted, specifically: 293.6 K, 600 K, 900 K, 1200 K, 2500 K, 0.1 K, and 250 K. Owing to the absence of thermal scattering cross sections in the CENDL libraries, the thermal scattering data required by the MSRE benchmark model—specifically those for graphite and hydrogen in water—from ENDF/B-VII.1 were used when generating the HDF5 CENDL libraries. Even though this approach partially compromised the independence of the library assessment, the influence of neutron cross section data on the benchmark model has been evaluated comprehensively.

IV. Results

A. k_eff of First Criticality

Results obtained with different nuclear data libraries are presented in Table 4. The OpenMC simulations used 100,000 particles per generation, with 700 active generations and 50 inactive generations. The same settings were applied in Sections IV B and IV C, yielding a standard deviation of approximately 12 pcm in k_eff. The CENDL-3.2 library produces k_eff 72 pcm lower than that of ENDF/B-VII.1, whereas CENDL-TMSR-V1 shows a 56 pcm reduction relative to ENDF/B-VII.1. Both CENDL libraries demonstrate close agreement with ENDF/B-VII.1 in k_eff calculations.

B. Control Rod Worth

The differential and integral worth of control rod 1 were calculated at the initial critical uranium concentration, with control rods 2 and 3 fully withdrawn. The results were compared against benchmark model predictions and experimental results. Experimental differential worth curves were derived by normalizing the measured values of control rod 1 at various concentrations to the initial critical uranium concentration, while integral worth curves were generated by integrating the experimental differential worth data. Differential worth calculations require movement of the control rod. Figure 6 shows the differential worth of control rod 1 at displacement distances of 5.08 cm and 10.16 cm, respectively.

The differential and integral worth curves calculated with the present model agree well with the experimental data. For 5.08 cm displacements, CENDL-TMSR-V1 yielded the lowest average error (1.085 pcm) between calculated and experimental differential worth. For 10.16 cm displacements, CENDL-3.2 produced the lowest average error (0.8 pcm), while CENDL-TMSR-V1 yielded an average error of 1.258 pcm. In the comparison of integral worth, the benchmark model exhibited a maximum error of 92 pcm at the rod position of 48.3 cm, whereas CENDL-TMSR-V1 yielded a maximum error of 66 pcm at the 114.3 cm position, both within the acceptable range of agreement.

In the control rod shadowing effect experiment, control rod 1 served as the control rod in two compensation scenarios: first, control rod 2 served as a compensating rod with control rod 3 fully withdrawn; second, control rods 2 and 3 served as a compensating rod bank at identical positions. In this study, the shadowing effect was evaluated for three uranium loadings (67.94 kg, 69.94 kg, 71.71 kg) through the rod position search method described in reference [32]. During the rod position search, the k_eff error was constrained to within 20 pcm. The calculated results shown in Fig. 7 are in good agreement with experimental data.

Calculations using different libraries were performed to evaluate the control rod bank worths in the current model under three ²³⁵U loadings. Two configurations were considered: (1) control rods 1 and 2 grouped with control rod 3 fully withdrawn; (2) control rods 1, 2, and 3 grouped. The results are presented in Fig. 8. The results of the present model with ENDF/B-VII.1 show good agreement with those of the benchmark model. However, at the ²³⁵U loading of 71.71 kg, both the integral worth of rod bank 1-2 in this study and the IRPhEP results fall below the experimental range. Comparative analysis indicates that CENDL-TMSR-V1 yields lower bank worth than CENDL-3.2, while ENDF/B-VIII.0 produces lower values compared to ENDF/B-VII.1. The discrepancy between the results of CENDL libraries and those of ENDF/B-VII.1 remains within 60 pcm.

C. Neutron Flux Distribution

The radial neutron flux distribution and its standard deviations calculated using ENDF/B-VII.1 in the current model are presented in Fig. 9(a) and (b). Figures 9(c)–(g) present the relative errors compared to ENDF/B-VII.1 of the neutron flux distributions obtained from different libraries and their corresponding standard deviations. CENDL-TMSR-V1 yields a maximum relative error of 4.4% in radial neutron flux compared to ENDF/B-VII.1, whereas CENDL-3.2 exhibits a maximum error of up to 4.6% relative to the same library.

D. Customized MSRE Depletion Case

To verify the applicability of CENDL libraries in MSRE burnup calculations, a customized depletion problem based on the model in this work was calculated. It is a transport-coupled depletion problem, meaning that transmutation reaction rates are obtained by solving the transport equation. The time steps of the depletion calculations are 1, 1, 1, 1, 1, 2, 2, 2, 4, 5, 30, 30, 30, 190, and 200 days, with the core thermal power of 8 MW over the burnup duration of 500 days. OpenMC supports multiple time integration methods for determining material compositions over time. The simplest "predictor" method was used through a depletion integrator, openmc.deplete.PredictorIntegrator, to implement the first-order predictor algorithm.

According to Ref. [7], the volumes of the only two materials containing fissionable nuclides—the fuel salt and the mixture (90.8% fuel salt and 9.2% INOR-8) in the downcomer—were set to be 1.61 m³ and 0.35 m³, respectively. It is notable that all fuel channels in the model in this work were filled with a single material, fuel salt, which means that OpenMC depleted it using a single set of transmutation reaction rates and produced a single new composition for the next time step. With the energy deposition mode "fission-q", the reaction rates were normalized using the product of fission reaction rates and fission Q values from the depletion chain. The depletion chain file used in this study is "ENDF/B-VII.1 Chain (Thermal Spectrum)", which is available on the OpenMC website. It is a complete depletion chain containing all isotopes with cross sections, neutron-induced fission yield data, and decay data as contained in ENDF/B-VII.1. Capture branching ratios are taken to be identical to the default branching ratios used in the Serpent Monte Carlo code, corresponding to a typical PWR spectrum.

To optimize computational efficiency, reduced particle counts were employed: 30,000 particles per generation, 300 active generations, and 50 inactive generations, resulting in a standard error of approximately 30 pcm. The k_eff trend (Fig. 10) shows a maximum discrepancy of approximately 158 pcm between the CENDL-TMSR-V1 and ENDF/B-VII.1 results during depletion. It can be observed that CENDL libraries predicted lower k_eff than ENDF/B libraries during the depletion calculations. Figure 11 shows comparisons of nuclide abundances and their relative errors with respect to ENDF/B-VII.1 across different libraries. Variations in nuclide abundances and library-induced errors were analyzed for ⁸⁵Kr, ¹³⁵Xe, ¹⁴⁹Sm, ²³⁵U, ²³⁹Pu, and ²⁴¹Am. Both CENDL-3.2 and CENDL-TMSR-V1 maintained the relative error within 5% of ENDF/B-VII.1 for all nuclides except ²⁴¹Am. The ²⁴¹Am variation curve shows an initial relative error of 49.84% in CENDL-TMSR-V1, with the error decreasing as burnup progresses. The larger discrepancies are likely attributable to the lower quantities of ²⁴¹Am during the early burnup time steps.

V. Conclusions

In this study, a new MSRE model was developed using OpenMC based on previous research, and the applicability of the CENDL-3.2 and CENDL-TMSR-V1 nuclear data libraries for molten salt reactor benchmark calculations was evaluated. This work extends the IRPhEP evaluation results for MSRE benchmarks by incorporating the Chinese CENDL libraries. The results indicate that CENDL-3.2 and CENDL-TMSR-V1 are suitable for analyses involving reactivity, control rod worth, and depletion. Their accuracy is generally comparable to that of ENDF/B-VII.1, with some results surpassing those of the benchmark model using ENDF/B-VII.1.

The present MSRE model incorporates more realistic geometric features, such as graphite stringer tops and gaps between support rods and the horizontal graphite lattice. Compared to the benchmark model using ENDF/B-VII.1, the new model exhibits a 34 pcm error in k_eff. The neutron energy spectra, axial flux distributions, and neutron production are consistent with IRPhEP results. The integral control rod worth deviates by less than 40 pcm from the experimental measurements, while the control rod bank worth is lower than that of the benchmark model, with errors remaining within 100 pcm. Calculated control rod shadowing effect was in agreement with both experimental data and IRPhEP results. The results calculated by the model in this work using ENDF/B-VII.1, which was employed in the benchmark model, are consistent with those of the benchmark model, confirming the accuracy of the model.

Detailed evaluations of CENDL-3.2 and CENDL-TMSR-V1 were performed using the new MSRE model. For k_eff, CENDL-3.2 and CENDL-TMSR-V1 yield values 72 pcm and 56 pcm lower than ENDF/B-VII.1, respectively. For control rod bank worth, CENDL-3.2 produces higher values than ENDF/B-VII.1, while CENDL-TMSR-V1 produces lower values. Both libraries show control rod bank worth deviations within 60 pcm relative to ENDF/B-VII.1. Moreover, radial neutron flux deviations for both CENDL-3.2 and CENDL-TMSR-V1 remain within 5% relative to ENDF/B-VII.1. A customized MSRE depletion problem showed a maximum k_eff error of approximately -160 pcm for the CENDL libraries relative to ENDF/B-VII.1. This indicates that the lifetime of MSRs designed using the CENDL libraries may be underestimated. Relative errors for ⁸⁵Kr, ¹³⁵Xe, ¹⁴⁹Sm, ²³⁵U, and ²³⁹Pu productions remained within 5%. For ²⁴¹Am, initial errors exceed 45% but decrease with burnup.

This work fills the gap in the validation of CENDL libraries against the MSRE benchmark, demonstrating their applicability for MSRs to some degree. Future work will focus on the analysis of uncertainties in the MSRE benchmark caused by uncertainties of key nuclides and reactions in the CENDL libraries. It is recommended that CENDL libraries provide thermal scattering data such as that for graphite, hydrogen in water, and even Be in BeF₂ and Li in LiF.

Acknowledgments

This study was supported by the Chinese Academy of Sciences Talent Introduction Youth Program (No. SINAP-YCJH-202303) and the Natural Science Foundation of Shanghai (Grant No. 24ZR1478400). We gratefully acknowledge the development team of OpenMC and the China Nuclear Data Center for their efforts in the development of CENDL libraries.

References

[1] J. Serp, M. Allibert, O. Bene, et al., The molten salt reactor (MSR) in generation IV: Overview and perspectives. Prog. Nucl. Energy 77, 308-319 (2014). doi: 10.1016/j.pnucene.2014.02.014

[2] K. F. Ma, C. G. Yu, X. Z. Cai, et al., Transmutation of ¹²⁹I in a single-fluid double-zone thorium molten salt reactor. Nucl. Sci. Tech. 31, 10 (2020). doi: 10.1007/s41365-019-0720-1

[3] X. C. Zhao, Y. Zou, R. Yan, et al., Analysis of burnup performance and temperature coefficient for a small modular molten-salt reactor started with plutonium. Nucl. Sci. Tech. 34, 17 (2023). doi: 10.1007/s41365-022-01155-2

[4] X. C. Zhao, R. Yan, G. F. Zhu, et al., Plutonium utilization in a small modular molten-salt reactor based on a batch fuel reprocessing scheme. Nucl. Sci. Tech. 35, 68 (2024). doi: 10.1007/s41365-024-01428-y

[5] L. E. McNeese, Molten-salt reactor program. Semiannual progress report for period ending February 29, 1976. Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States) (1976)

[6] C. H. Gabbard, Reactor power measurement and heat transfer performance in the molten salt reactor experiment. Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States) (1970). doi: 10.2172/4157686

[7] E. L. Compere, S. S. Kirslis, E. G. Bohlmann, et al., Fission product behavior in the molten salt reactor experiment. Oak Ridge National Lab., Tenn. (USA) (1975). doi: 10.2172/4077644

[8] L. Chen, G. F. Zhu, Z. Wang, et al., Evaluation of ⁹⁹Mo production based on molten salt reactor off-gas extraction. Nucl. Tech. 47, 131-139 (2024). doi: 10.11889/j.0253-3219.2024.hjs.47.080604

[9] X. Y. Jiang, H. J. Lu, Y. S. Chen, et al., Numerical and experimental investigation of a new conceptual fluoride salt freeze valve for thorium-based molten salt reactor. Nucl. Sci. Tech. 31, 16 (2020). doi: 10.1007/s41365-020-0729-5

[10] S. H. Yu, Y. F. Liu, P. Yang, et al., Neutronics analysis for MSR cell with different fuel salt channel geometries. Nucl. Sci. Tech. 32, 9 (2021). doi: 10.1007/s41365-020-00844-0

[11] G. F. Zhu, R. Yan, H. H. Peng, et al., Application of Monte Carlo method to calculate the effective delayed neutron fraction in molten salt reactor. Nucl. Sci. Tech. 30, 34 (2019). doi: 10.1007/s41365-019-0557-7

[12] X. D. Zuo, M. S. Cheng, Y. Q. Dai, et al., Flow field effect of delayed neutron precursors in liquid-fueled molten salt reactors. Nucl. Sci. Tech. 33, 96 (2022). doi: 10.1007/s41365-022-01141-8

[13] Q. Yang, S. H. Yu, Y. Zou, et al., Quantitative similarity analysis of the fluoride salt-cooled pebble bed reactor with HTR-10 and MSRE. Nucl. Tech. 39, 6 (2016). doi: 10.11889/j.0253-3219.2016.hjs.39.010601

[14] C. Q. Chen, X. B. Xia, Z. H. Zhang, et al., Radiological environmental impact analysis of a 2-MW thorium molten salt reactor during an accident. Nucl. Sci. Tech. 30, 78 (2019). doi: 10.1007/s41365-019-0605-3

[15] L. Y. He, S. P. Xia, X. M. Zhou, et al., ThU cycle performance analysis based on molten chloride salt and molten fluoride salt fast reactors. Nucl. Sci. Tech. 31, 83 (2020). doi: 10.1007/s41365-020-00790-x

[16] X. X. Li, D. Y. Cui, C. Y. Zou, et al., Assembly-level analysis on temperature coefficient of reactivity in a graphite-moderated fuel salt reactor fueled with low-enriched uranium. Nucl. Sci. Tech. 34, 70 (2023). doi: 10.1007/s41365-023-01216-0

[17] L. He, J. J. Shen, X. Z. Cai, A thermal-hydraulic model for the graphite-moderated channel-type molten salt reactor. Nucl. Sci. Tech. 36, 59 (2025). doi: 10.1007/s41365-024-01627-7

[18] D. A. Brown, M. B. Chadwick, R. Capote, et al., ENDF/B-VIII.0: the 8th major release of the nuclear reaction data library with CIELO-project cross sections, new standards and thermal scattering data. Nucl. Data Sheets 148, 1-142 (2018). doi: 10.1016/j.nds.2018.02.001

[19] A. J. M. Plompen, O. Cabellos, C. De Saint Jean, et al., The joint evaluated fission and fusion nuclear data library, JEFF-3.3. Eur. Phys. J. A 56, 181 (2020). doi: 10.1140/epja/s10050-020-00114-4

[20] O. Iwamoto, N. Iwamoto, S. Kunieda, et al., Japanese evaluated nuclear data library version 5: JENDL-5. J. Nucl. Sci. Technol. 60, 1-60 (2023). doi: 10.1080/00223131.2022.2141903

[21] Y. F. Liu, R. Yan, Y. Zou, et al., Neutron flux distribution and conversion ratio of critical experiment device for molten salt reactor research. Ann. Nucl. Energy 133, 707-717 (2019). doi: 10.1016/j.anucene.2019.07.018

[22] S. Y. Li, D. Q. Wang, Y. Y. Jiang, et al., Application of evaluated nuclear data in design of heat pipe cooled space reactor. At. Energy Sci. Tech. 52, 1186-1193 (2018). doi: 10.7538/yzk.2018.52.07.1186

[23] Z. G. Ge, R. R. Xu, H. C. Wu, et al., CENDL-3.2: the new version of Chinese general purpose evaluated nuclear data library. EPJ Web Conf. 239, 09001 (2020). doi: 10.1051/epjconf/202023909001

[24] J. J. Hu, B. Zhang, Z. W. Zong, et al., Verification of CENDL-3.2 nuclear data on Venus-3 shielding benchmark by ARES transport code. Sci. Technol. Nucl. Ins. 2021, 13 (2021). doi: 10.1155/2021/6633366

[25] J. Wu, S. C. Liu, Y. X. Chen, Analysis of fast critical benchmark with Be for multi-group cross section library based on CENDL-3.2. Nucl. Tech. 45, 75-80 (2022). doi: 10.11889/j.0253-3219.2022.hjs.45.060601

[26] T. J. Zu, Y. H. Huang, Q. C. Teng, et al., Application of CENDL-3.2 and ENDF/B-VIII.0 on the reactor physics simulation of PWR. Ann. Nucl. Energy 158, 108238 (2021). doi: 10.1016/j.anucene.2021.108238

[27] P. Liu, G. C. Cheng, X. F. Wu, et al., The special nuclear data library CENDL-TMSR for thorium-uranium cycle (in Chinese). Sci. Sin-Phys Mech Astron. 50, 77-91 (2020). doi: 10.1360/SSPMA-2019-0230

[28] Y. F. Liu, J. F. Hu, R. Yan, et al., Validation of CENDL-TMSR-V1 against LR-0 benchmark. At. Energy Sci. Tech. 55, 2070-2078 (2021). doi: 10.7538/yzk.2020.youxian.0860

[29] X. H. Wang, X. Z. Cai, J. L. Han, Benchmark test of CENDL-TMSR-V1 nuclear data library for thorium-uranium fuel cycle. At. Energy Sci. Tech. 53, 1466-1474 (2019). doi: 10.7538/yzk.2018.youxian.0749

[30] C. Wulandari, A. Waris, S. Permana, et al., Evaluating the JEFF 3.1, ENDF/B-VII.0, JENDL 3.3, and JENDL 4.0 nuclear data libraries for a small 100 MWe molten salt reactor with plutonium fuel. Nucl. Sci. Tech. 33, 165 (2022). doi: 10.1007/s41365-022-01141-8

[31] W. Zwermann, I. Hill, Assessment of nuclear data libraries for fast reactor systems with MCNP and NDaST. Ann. Nucl. Energy 178, 109354 (2022). doi: 10.1016/j.anucene.2022.109354

[32] M. Fratoni, D. Shen, G. Ilas, et al., Molten salt reactor experiment benchmark evaluation. DOE-UCB–8542 (2020). doi: 10.2172/1617123

[33] K. T. Clarno, J. E. Barlow, T. Sawyer, et al., Evaluation of SCALE, serpent, and MCNP for molten salt reactor applications using the MSRE benchmark. Ann. Nucl. Energy 194, 110092 (2023). doi: 10.1016/j.anucene.2023.110092

[34] S. Yilmaz, P. K. Romano, L. Chierici, et al., CAD and constructive solid geometry modeling of the molten salt reactor experiment with OpenMC. Front. Nucl. Eng. 3, 1385478 (2024). doi: 10.3389/fnuen.2024.1385478

[35] D. Shen, G. Ilas, J. J. Powers, et al., Reactor physics benchmark of the first criticality in the molten salt reactor experiment. Nucl. Sci. Eng. 195, 825-837 (2021). doi: 10.1080/00295639.2021.1880850

[36] R. Robertson, MSRE design & operations report part 1: Description of reactor design. ORNL-TM-728 (1965). doi: 10.2172/4654707

[37] R. B. Briggs, Molten-salt reactor program semiannual progress report for period ending July 31, 1964. ORNL-3708 (1964). doi: 10.2172/4676587

[38] B. Deng, Y. Cui, J.-G. Chen, et al., Core and blanket thermal-hydraulic analysis of a molten salt fast reactor based on coupling of OpenMC and OpenFOAM. Nucl. Sci. Tech. 31, 85 (2020). doi: 10.1007/s41365-020-00803-9

[39] P. K. Romano, C. J. Josey, A. E. Johnson, et al., Depletion capabilities in the OpenMC Monte Carlo particle transport code. Ann. Nucl. Energy 152, 107989 (2021). doi: 10.1016/j.anucene.2020.107989

[40] P. K. Romano, N. E. Horelik, B. R. Herman, et al., OpenMC: A state-of-the-art Monte Carlo code for research and development. SNA + MC 2013 - Joint International Conference on Supercomputing in Nuclear Applications + Monte Carlo, 06016, Paris, France (2014). doi: 10.1051/snamc/201406016

[41] B. Zhang, X. B. Ma, K. Hu, et al., Performance of the CENDL-3.2 and other major neutron data libraries for criticality calculations. Nucl. Sci. Tech. 33, 8 (2022). doi: 10.1007/s41365-021-00960-0

[42] J. Conlin, D. Parsons, S. Gardiner, et al., Continuous energy neutron cross section data tables based upon ENDF/B-VII.1. LA-UR–13-20137 (2013). doi: 10.2172/1063914

[43] B. E. Prince, S. J. Ball, J. R. Engel, et al., Zero-power physics experiments on the molten-salt reactor experiment. ORNL–4233 (1968). doi: 10.2172/4558029