najoNanosystems: Physics, Chemistry, MathematicsНаносистемы: физика, химия, математика2220-80542305-7971Университет ИТМО10.17586/2220-8054-2025-16-5-619-627najo-1533Research ArticleCHEMISTRY AND MATERIALS SCIENCEХИМИЯ И НАУКА О МАТЕРИАЛАХBridging accuracy and efficiency: assessing universal ML potentials for niobiumoxygen clustersСовмещая точность и эффективность: оценка универсальных машиннообучаемых межатомных потенциалов для кластеров ниобийкислородhttps://orcid.org/0000-0001-8312-7071ПоповИ. С.PopovI. S.

Ilya S. Popov

Yekaterinburg

popov@ihim.uran.ruhttps://orcid.org/0000-0003-1656-732XВалееваА. А.ValeevaA. A.

Albina A. Valeeva

Yekaterinburg

anibla_v@mail.ruЕняшинА. Н.EnyashinA. N.

Andrey N. Enyashin

Yekaterinburg

enyashin@ihim.uran.ruInstitute of Solid State Chemistry, Ural Branch of Russian Academy of SciencesRussian Federation202505112025165619627Copyright © Popov I.S., Valeeva A.A., Enyashin A.N., 20252025Попов И.С., Валеева А.А., Еняшин А.Н.Popov I.S., Valeeva A.A., Enyashin A.N.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://nanojournal.ifmo.ru/jour/article/view/1533

Machine Learning Interatomic Potentials (MLIPs) promise to combine the accuracy of DFT with the speed of classical force fields. However, their reliability for complex, multi-component systems requires rigorous validation. Here, we perform a targeted evaluation of three leading universal MLIPs using niobium oxide clusters (NbnOm, n ≤ 6, m ≤ 6) as a challenging test case. The Nb–O system is very well suited for this task due to its complex electronic interactions, manifested in existence of the bulk phase with 25% vacancy-ordered lattice and, at nanoscale, by a diverse range of non-stoichiometric clusters. We employ a dataset of global minima structures identified via DFT-based evolutionary search as a strict reference. A comparative analysis is then performed by executing evolutionary searches with the MLIPs. By directly comparing predicted structures, energies, and relative stability, we provide a comprehensive assessment of the accuracy and limitations of current universal potentials for modeling complex nanoscale oxides.

Машиннообучаемые межатомные потенциалы (MLIP) обещают сочетание вычислительной точности теории функционала плотности (DFT) со скоростью типичной для методов молекулярной механики. Однако их надежность для сложных многокомпонентных систем требует тщательной валидации. В данной работе мы проводим оценку трех ведущих универсальных MLIP на примере кластеров оксида ниобия (NbnOm, n ≤ 6, m ≤ 6), как сложнейшего теста. Система Nb-O хорошо подходит для этой задачи благодаря сложным межатомным взаимодействиям, проявление которых в макрокристаллической фазе приводит к стабилизации решётки с 25% упорядоченных вакансий, а на наноуровне – к широкому спектру нестехиометрических кластеров. В качестве референса используется набор данных о глобальных минимумах структур кластеров Nb-O, идентифицированных с помощью эволюционного поиска (USPEX) и DFT расчетов. Путем прямого сравнения структур, энергий и относительной стабильности, предсказываемых эволюционным поиском с использованием MLIP разного уровня, мы даём всестороннюю оценку точности и ограничений современных универсальных потенциалов для моделирования сложных наноразмерных оксидов.

DFTмашиннообучаемые потенциалыMLIPэволюционный алгоритмоксид ниобияNbOнаночастицакластерUSPEXDFTMachine learning potentialMLIPEvolutionary algorithmNiobium oxideNbONanoparticleClusterUSPEXThis work was supported by the Russian Science Foundation (project 19-73-20012-Π, https://rscf.ru/en/project/19-73-20012/) and was performed at the Institute of Solid State Chemistry UB RAS.ReferencesWan K., He J., Shi X. Construction of High Accuracy Machine Learning Interatomic Potential for Surface/Interface of Nanomaterials – A Review. Adv. Mater., 2024, 36(22), P. 2305758.Wan K., He J., Shi X. Construction of High Accuracy Machine Learning Interatomic Potential for Surface/Interface of Nanomaterials – A Review. Adv. Mater., 2024, 36(22), P. 2305758.Wang G., Wang C., Zhang X., Li Z., Zhou J., Sun Z. Machine learning interatomic potential: Bridge the gap between small-scale models and realistic device-scale simulations. iScience, 2024, 27(5), P. 109673.Wang G., Wang C., Zhang X., Li Z., Zhou J., Sun Z. Machine learning interatomic potential: Bridge the gap between small-scale models and realistic device-scale simulations. iScience, 2024, 27(5), P. 109673.Mortazavi B., Zhuang X., Rabczuk T., Shapeev A.V. Atomistic modeling of the mechanical properties: the rise of machine learning interatomic potentials. Mater. Horiz., 2023, 10, P. 1956–1968.Mortazavi B., Zhuang X., Rabczuk T., Shapeev A.V. Atomistic modeling of the mechanical properties: the rise of machine learning interatomic potentials. Mater. Horiz., 2023, 10, P. 1956–1968.Barroso-Luque L., Shuaibi M., Fu X., Wood B.M., Dzamba M., Gao M., Rizvi A., Zitnick C.L., Ulissi Z.W. Open Materials 2024 (OMat24) Inorganic Materials Dataset and Models. arXiv, 2024, [cond-mat.mtrl-sci], arXiv:2410.12771.Barroso-Luque L., Shuaibi M., Fu X., Wood B.M., Dzamba M., Gao M., Rizvi A., Zitnick C.L., Ulissi Z.W. Open Materials 2024 (OMat24) Inorganic Materials Dataset and Models. arXiv, 2024, [cond-mat.mtrl-sci], arXiv:2410.12771.Ramlaoui A., Siron M., Djafar I., Musielewicz J., Rossello A., Schmidt V., Duval A. LeMat-Traj: A Scalable and Unified Dataset of Materials Trajectories for Atomistic Modeling. arXiv, 2025, [cs.LG], arXiv:2508.20875.Ramlaoui A., Siron M., Djafar I., Musielewicz J., Rossello A., Schmidt V., Duval A. LeMat-Traj: A Scalable and Unified Dataset of Materials Trajectories for Atomistic Modeling. arXiv, 2025, [cs.LG], arXiv:2508.20875.Jain A., Ong S.P., Hautier G., Chen W., Richards W.D., Dacek S., Cholia S., Gunter D., Skinner D., Ceder G., Persson K.A. Commentary: The Materials Project: A materials genome approach to accelerating materials innovation. APL Materials, 2013, 1, P. 011002.Jain A., Ong S.P., Hautier G., Chen W., Richards W.D., Dacek S., Cholia S., Gunter D., Skinner D., Ceder G., Persson K.A. Commentary: The Materials Project: A materials genome approach to accelerating materials innovation. APL Materials, 2013, 1, P. 011002.Schmidt J., Cerqueira T.F.T., Romero A.H., Loew A., Jager F., Wang H.-C., Botti S., Marques M.A.L. Improving machine-learning models in ¨ materials science through large datasets. Materials Today Physics, 2024, 48, P. 101560.Schmidt J., Cerqueira T.F.T., Romero A.H., Loew A., Jager F., Wang H.-C., Botti S., Marques M.A.L. Improving machine-learning models in ¨ materials science through large datasets. Materials Today Physics, 2024, 48, P. 101560.Saal J.E., Kirklin S., Aykol M., Meredig B., Wolverton C. Materials Design and Discovery with High-Throughput Density Functional Theory: The Open Quantum Materials Database (OQMD). Jom., 2013, 65(11), P. 1501–1509.Saal J.E., Kirklin S., Aykol M., Meredig B., Wolverton C. Materials Design and Discovery with High-Throughput Density Functional Theory: The Open Quantum Materials Database (OQMD). Jom., 2013, 65(11), P. 1501–1509.Riebesell J., Goodall R.E.A., Benner P., Chiang Y., Deng B., Ceder G., Asta M., Lee A.A., Jain A., Persson K.A. A framework to evaluate machine learning crystal stability predictions. Nat. Mach. Intell., 2025, 7, P. 836–847.Riebesell J., Goodall R.E.A., Benner P., Chiang Y., Deng B., Ceder G., Asta M., Lee A.A., Jain A., Persson K.A. A framework to evaluate machine learning crystal stability predictions. Nat. Mach. Intell., 2025, 7, P. 836–847.Okatz A.M., Keesom P.H. Specific heat and magnetization of the superconducting monoxides: NbO and TiO. Phys. Rev. B, 1975, 12, P. 4917.Okatz A.M., Keesom P.H. Specific heat and magnetization of the superconducting monoxides: NbO and TiO. Phys. Rev. B, 1975, 12, P. 4917.Bowman A.L., Wallace T.C., Yarnell J.L., Wenzel R.G. The crystal structure of niobium monoxide. Acta Crystallogr., 1966, 21, P. 843.Bowman A.L., Wallace T.C., Yarnell J.L., Wenzel R.G. The crystal structure of niobium monoxide. Acta Crystallogr., 1966, 21, P. 843.Kurmaev E.Z., Moewes A., Bureev O.G., Nekrasov I.A., Cherkashenko V.M., Korotin M.A., Ederer D.L. Electronic structure of niobium oxides. J. Alloys Compd., 2002, 347, P. 213–218.Kurmaev E.Z., Moewes A., Bureev O.G., Nekrasov I.A., Cherkashenko V.M., Korotin M.A., Ederer D.L. Electronic structure of niobium oxides. J. Alloys Compd., 2002, 347, P. 213–218.Hu Z., Qian G., Li S., Yang L., Chen X., Weng M., Tan W., Pan F. Discovery of space aromaticity in transition–metal monoxide crystal Nb3O3 enabled by octahedral Nb6 structural units. Sci. Bull., 2019, 65, P. 367–372.Hu Z., Qian G., Li S., Yang L., Chen X., Weng M., Tan W., Pan F. Discovery of space aromaticity in transition–metal monoxide crystal Nb3O3 enabled by octahedral Nb6 structural units. Sci. Bull., 2019, 65, P. 367–372.Miura A., Takei T., Kumada N., Wada S., Magome E., Moriyoshi C., Kuroiwa Y. Bonding Preference of Carbon, Nitrogen, and Oxygen in Niobium-Based Rock-Salt Structures. Inorg. Chem., 2013, 52, P. 9699.Miura A., Takei T., Kumada N., Wada S., Magome E., Moriyoshi C., Kuroiwa Y. Bonding Preference of Carbon, Nitrogen, and Oxygen in Niobium-Based Rock-Salt Structures. Inorg. Chem., 2013, 52, P. 9699.Music D., Schmidt P., Mraz S. Adsorption of film-forming species on NbO and NbO ´ 2 surfaces. J. Vac. Sci. Technol. A., 2017, 35, P. 061512.Music D., Schmidt P., Mraz S. Adsorption of film-forming species on NbO and NbO ´ 2 surfaces. J. Vac. Sci. Technol. A., 2017, 35, P. 061512.Fielicke A., Meijer G., Helden G. Infrared Spectroscopy of Niobium Oxide Cluster Cations in a Molecular Beam: Identifying the Cluster Structures. J. Am. Chem. Soc., 2003, 125, P. 3659–3667.Fielicke A., Meijer G., Helden G. Infrared Spectroscopy of Niobium Oxide Cluster Cations in a Molecular Beam: Identifying the Cluster Structures. J. Am. Chem. Soc., 2003, 125, P. 3659–3667.Deng H.T., Kerns K.P., Castleman A.W. Formation, Structures, and Reactivities of Niobium Oxide Cluster Ions. J. Phys. Chem., 1996, 100, P. 13386–13392.Deng H.T., Kerns K.P., Castleman A.W. Formation, Structures, and Reactivities of Niobium Oxide Cluster Ions. J. Phys. Chem., 1996, 100, P. 13386–13392.Popov I.S., Valeeva A.A., Enyashin A.N. Identifying stable Nb–O clusters using evolutionary algorithm and DFT: A foundation for machine learning potentials. Chem. Phys., 2025, 590, P. 112533.Popov I.S., Valeeva A.A., Enyashin A.N. Identifying stable Nb–O clusters using evolutionary algorithm and DFT: A foundation for machine learning potentials. Chem. Phys., 2025, 590, P. 112533.Lyakhov A.O., Oganov A.R., Stokes H., Zhu Q. New developments in evolutionary structure prediction algorithm USPEX. Comp. Phys. Comm., 2013, 184, P. 1172–1182.Lyakhov A.O., Oganov A.R., Stokes H., Zhu Q. New developments in evolutionary structure prediction algorithm USPEX. Comp. Phys. Comm., 2013, 184, P. 1172–1182.Lepeshkin S.V., Baturin V.S., Uspenskii Y.A., Oganov A.R. Method for simultaneous prediction of atomic structure of nanoclusters in a wide area of compositions. J. Phys. Chem. Lett., 2019, 10, P. 102–106.Lepeshkin S.V., Baturin V.S., Uspenskii Y.A., Oganov A.R. Method for simultaneous prediction of atomic structure of nanoclusters in a wide area of compositions. J. Phys. Chem. Lett., 2019, 10, P. 102–106.Heydariyan S., Nouri M.R., Alaei M., Allahyari Z., Niehaus T.A. New candidates for the global minimum of medium-sized silicon clusters: A hybrid DFTB/DFT genetic algorithm applied to Sin, n = 8–80. J. Chem. Phys., 2018, 149, P. 074313.Heydariyan S., Nouri M.R., Alaei M., Allahyari Z., Niehaus T.A. New candidates for the global minimum of medium-sized silicon clusters: A hybrid DFTB/DFT genetic algorithm applied to Sin, n = 8–80. J. Chem. Phys., 2018, 149, P. 074313.Olvera-Neria O., Garc´ıa-Cruz R., Gonzalez-Torres J., Garc´ıa-Cruz L.M., Castillo-Sanchez J.L., Poulain E. Strongly Bound Frenkel Excitons on ´ TiO2 Nanoparticles: An Evolutionary and DFT Approach. Int. J. Photoenergy, 2024, 2024, P. 4014216.Olvera-Neria O., Garc´ıa-Cruz R., Gonzalez-Torres J., Garc´ıa-Cruz L.M., Castillo-Sanchez J.L., Poulain E. Strongly Bound Frenkel Excitons on ´ TiO2 Nanoparticles: An Evolutionary and DFT Approach. Int. J. Photoenergy, 2024, 2024, P. 4014216.Yu X., Oganov A.R., Zhu Q., Qi F., Qian G. The stability and unexpected chemistry of oxide clusters. Phys. Chem. Chem. Phys., 2018, 20, P. 30437.Yu X., Oganov A.R., Zhu Q., Qi F., Qian G. The stability and unexpected chemistry of oxide clusters. Phys. Chem. Chem. Phys., 2018, 20, P. 30437.Mahdavifar Z. Prediction of unexpected BnPn structures: promising materials for non-linear optical devices and photocatalytic activities. Nanoscale Adv., 2021, 3, P. 2846.Mahdavifar Z. Prediction of unexpected BnPn structures: promising materials for non-linear optical devices and photocatalytic activities. Nanoscale Adv., 2021, 3, P. 2846.Sandu M.P., Kovtunov M.A., Baturin V.S., Oganov A.R., Kurzina I.A. Influence of the Pd:Bi ratio on Pd–Bi/Al2O3 catalysts: structure, surface and activity in glucose oxidation. Phys. Chem. Chem. Phys., 2021, 23, P. 14889.Sandu M.P., Kovtunov M.A., Baturin V.S., Oganov A.R., Kurzina I.A. Influence of the Pd:Bi ratio on Pd–Bi/Al2O3 catalysts: structure, surface and activity in glucose oxidation. Phys. Chem. Chem. Phys., 2021, 23, P. 14889.Zhou T., Ma L., Chen H. Electronic structure and stability of Al6CMn (M = Li, Na, K; n = 2, 4, 6) clusters. Comput. Theor. Chem., 2020, 1178, P. 112780.Zhou T., Ma L., Chen H. Electronic structure and stability of Al6CMn (M = Li, Na, K; n = 2, 4, 6) clusters. Comput. Theor. Chem., 2020, 1178, P. 112780.Steshin I.S., Panteleev S.V., Petukhov I.V., Ignatov S.K. Parametrization of Gaussian approximation potential for the global optimization of magnesium clusters MgN (N≤100). Phys. Chem. Chem. Phys., 2025, 27, P. 18960–18977.Steshin I.S., Panteleev S.V., Petukhov I.V., Ignatov S.K. Parametrization of Gaussian approximation potential for the global optimization of magnesium clusters MgN (N≤100). Phys. Chem. Chem. Phys., 2025, 27, P. 18960–18977.Kresse G., Furthmuller J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. ¨ Phys. Rev. B, 1996, 54, P. 11169.Kresse G., Furthmuller J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. ¨ Phys. Rev. B, 1996, 54, P. 11169.Fu X., Wood B.M., Barroso-Luque L., Levine D.S., Gao M., Dzamba M., Zitnick C.L. Learning Smooth and Expressive Interatomic Potentials for Physical Property Prediction. arXiv, 2025, [physics.comp-ph], arXiv:2502.12147.Fu X., Wood B.M., Barroso-Luque L., Levine D.S., Gao M., Dzamba M., Zitnick C.L. Learning Smooth and Expressive Interatomic Potentials for Physical Property Prediction. arXiv, 2025, [physics.comp-ph], arXiv:2502.12147.Rhodes B., Vandenhaute S., Simkus V., Gin J., Godwin J., Duignan T., Neumann M. Orb-v3: atomistic simulation at scale. ˇ arXiv, 2025, [condmat.mtrl-sci], arXiv:2504.06231.Rhodes B., Vandenhaute S., Simkus V., Gin J., Godwin J., Duignan T., Neumann M. Orb-v3: atomistic simulation at scale. ˇ arXiv, 2025, [condmat.mtrl-sci], arXiv:2504.06231.Kim J., Kim J., Kim J., Lee J., Park Y., Kang Y., Han S. Data-efficient multi-fidelity training for high-fidelity machine learning interatomic potentials. arXiv, 2025, [cond-mat.mtrl-sci], arXiv:2504.06231.Kim J., Kim J., Kim J., Lee J., Park Y., Kang Y., Han S. Data-efficient multi-fidelity training for high-fidelity machine learning interatomic potentials. arXiv, 2025, [cond-mat.mtrl-sci], arXiv:2504.06231.Wood B.M., Dzamba M., Fu X., Gao M., Shuaibi M., Barroso-Luque L., Abdelmaqsoud K., Gharakhanyan V., Kitchin J.R., Levine D.S., Michel K., Sriram A., Cohen T., Das A., Rizvi A., Sahoo S.J., Ulissi Z.W., Zitnick C.L. UMA: A Family of Universal Models for Atoms. arXiv, 2025, [cs.LG], arXiv:2506.23971.Wood B.M., Dzamba M., Fu X., Gao M., Shuaibi M., Barroso-Luque L., Abdelmaqsoud K., Gharakhanyan V., Kitchin J.R., Levine D.S., Michel K., Sriram A., Cohen T., Das A., Rizvi A., Sahoo S.J., Ulissi Z.W., Zitnick C.L. UMA: A Family of Universal Models for Atoms. arXiv, 2025, [cs.LG], arXiv:2506.23971.Larsen A.H., Mortensen J.J., Blomqvist J., Castelli I.E., Christensen R., Dułak M., Friis J., Groves M.N., Hammer B., Hargus C., Hermes E.D., Jennings P.C., Jensen P.B., Kermode J., Kitchin J.R., Kolsbjerg E.L., Kubal J., Kaasbjerg K., Lysgaard S., Maronsson J.B., Maxson T., Olsen T., Pastewka L., Peterson A., Rostgaard C., Schiøtz J., Schutt O., Strange M., Thygesen K.S., Vegge T., Vilhelmsen L., Walter M., Zeng Z., Jacobsen ¨ K.W. The atomic simulation environment – a Python library for working with atoms. J. Phys.: Condens. Matter, 2017, 29, P. 273002.Larsen A.H., Mortensen J.J., Blomqvist J., Castelli I.E., Christensen R., Dułak M., Friis J., Groves M.N., Hammer B., Hargus C., Hermes E.D., Jennings P.C., Jensen P.B., Kermode J., Kitchin J.R., Kolsbjerg E.L., Kubal J., Kaasbjerg K., Lysgaard S., Maronsson J.B., Maxson T., Olsen T., Pastewka L., Peterson A., Rostgaard C., Schiøtz J., Schutt O., Strange M., Thygesen K.S., Vegge T., Vilhelmsen L., Walter M., Zeng Z., Jacobsen ¨ K.W. The atomic simulation environment – a Python library for working with atoms. J. Phys.: Condens. Matter, 2017, 29, P. 273002.Thompson A.P., Aktulga H.M., Berger R., Bolintineanu D.S., Brown W.M., Crozier P.S., Veld P.J., Kohlmeyer A., Moore S.G., Nguyen T.D., Shan R., Stevens M.J., Tranchida J., Trott C., Plimpton S.J. LAMMPS – a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales. Comp. Phys. Comm., 2022, 271, P. 10817.Thompson A.P., Aktulga H.M., Berger R., Bolintineanu D.S., Brown W.M., Crozier P.S., Veld P.J., Kohlmeyer A., Moore S.G., Nguyen T.D., Shan R., Stevens M.J., Tranchida J., Trott C., Plimpton S.J. LAMMPS – a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales. Comp. Phys. Comm., 2022, 271, P. 10817.Stukowski A. Visualization and analysis of atomistic simulation data with OVITO – the Open Visualization Tool. Modelling Simul. Mater. Sci. Eng., 2010, 18, P. 015012.Stukowski A. Visualization and analysis of atomistic simulation data with OVITO – the Open Visualization Tool. Modelling Simul. Mater. Sci. Eng., 2010, 18, P. 015012.Hunter J.D. Matplotlib: A 2D Graphics Environment. Computing in Science & Engineering, 2007, 9(3), P. 90–95.Hunter J.D. Matplotlib: A 2D Graphics Environment. Computing in Science & Engineering, 2007, 9(3), P. 90–95.

The authors declare that there are no conflicts of interest present.