Ensemble learning for predicting formation energy and band gap of actinide materials
Preprint
- 28 April 2026
- preprint
- Published by Elsevier in SSRN Electronic Journal
Abstract
The deployment of advanced nuclear energy systems and closed fuel cycles requires improved understanding of actinide materials characteristics under electrochemical processing conditions, i.e., for spent nuclear fuel (SNF) recycling via molten-salt pyroprocessing. Meanwhile, the relationship between oxygen stoichiometry, dopant incorporation, and electronic properties in uranium and mixed-oxide fuels remains insufficiently characterized. This study addresses these challenges by developing machine learning models to predict both formation energies and band gaps of actinide materials. Using data from the Open Quantum Materials Database, this work represents the first application of ML to characterize actinide band gap evolution. The models were developed using Random Forest and Extreme Gradient Boosting (XGB) algorithms, with XGB become the best model (R2 = 0.97 for formation energy and 0.90 for band gap prediction). Feature importance analysis identified electronegativity and Mendeleev number as key descriptors governing thermodynamic stability and electronic band gap. The trained models were then applied to simulate oxygen reduction and dopant incorporation on uranium, plutonium, and mixed-oxide systems, including incorporation of fission products and noble metals, to evaluate their effects on formation energy and band gap evolution. These results demonstrate the capability of machine learning as a computational framework for the preliminary evaluation of SNF processing, providing critical insights into fuel design and electrode selection.Keywords
This publication has 37 references indexed in Scilit:
- Prediction of thermodynamic stability of actinide compounds by machine learning modelCeramics International, 2023
- Role of nuclear energy in carbon mitigation to achieve United Nations net zero carbon emission: evidence from Fourier bootstrap Toda-YamamotoEnvironmental Science and Pollution Research, 2023
- Thermal conductivity prediction of UO2-BeO composite fuels and related decisive features discovery via convolutional neural networkActa Materialia, 2022
- Policies toward net-zero: Benchmarking the economic competitiveness of nuclear against wind and solar energyApplied Energy, 2022
- Review of computational approaches to predict the thermodynamic stability of inorganic solidsJournal of Materials Science, 2022
- An interpretable hybrid Machine learning prediction of dielectric constant of alkali halide crystalsChemical Physics, 2022
- A machine learning approach to thermal conductivity modeling: A case study on irradiated uranium-molybdenum nuclear fuelsComputational Materials Science, 2019
- Matminer: An open source toolkit for materials data miningComputational Materials Science, 2018
- Electrochemical properties of noble metal anodes for electrolytic reduction of uranium oxideJournal of Radioanalytical and Nuclear Chemistry, 2016
- Materials Design and Discovery with High-Throughput Density Functional Theory: The Open Quantum Materials Database (OQMD)JOM, 2013