journal article·2026

Thermodynamic assessment of machine learning models for solid-state synthesis prediction

Jane Schlesinger, Simon Hjaltason, Nathan J. Szymanski, Christopher J. Bartel

Publisher page

Abstract

Machine learning models have recently emerged to predict whether hypothetical solid-state materials can be synthesized. These models aim to circumvent direct first-principles modeling of solid-state phase transformations, instead learning from large databases of successfully synthesized materials. Here, we assess the alignment of several recently introduced synthesis prediction models with material and reaction thermodynamics, quantified by the energy with respect to the convex hull and a metric accounting for thermodynamic selectivity of enumerated synthesis reactions. A dataset of successful synthesis recipes was used to determine the likely bounds on both quantities beyond which materials can be deemed unlikely to be synthesized. With these bounds as context, thermodynamic quantities were computed using the CHGNet foundation potential for thousands of new hypothetical materials generated using the Chemeleon generative model. Four recently published machine learning models for synthesizability prediction were applied to this same dataset, and the resultant predictions were considered against computed thermodynamics. We find these models generally overpredict the likelihood of synthesis, but some model scores do trend with thermodynamic heuristics, assigning lower scores to materials that are less stable or do not have an available synthesis recipe that is calculated to be thermodynamically selective. In total, this work identifies existing gaps in machine learning models for materials synthesis and introduces a new approach to assess their quality in the absence of extensive negative examples (failed syntheses).
Source: openalex_inverted_index · Quality: candidate unverified · Source record ↗

Research context

OpenAlex topics: Machine Learning in Materials Science · Catalysis and Oxidation Reactions · X-ray Diffraction in Crystallography

SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.

Affiliations

Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, MN 55455, USA

Funding

Division of Information and Intelligent Systems

References (55)

Show source-provided references
1. Jain · Nat. Rev. Mater. · 2016 · 1 · 1500410.1038/natrevmats.2015.4 ↗
2. Oganov · Nat. Rev. Mater. · 2019 · 4 · 33110.1038/s41578-019-0101-8 ↗
3. Zunger · Nat. Rev. Chem. · 2018 · 2 · 012110.1038/s41570-018-0121 ↗
4. Horton · Nat. Mater. · 2025 · 24 · 152210.1038/s41563-025-02272-0 ↗
5. Park · Matter · 2024 · 7 · 235510.1016/j.matt.2024.05.017 ↗
6. Jain · APL Mater. · 2013 · 1 · 01100210.1063/1.4812323 ↗
7. Kirklin · npj Comput. Mater. · 2015 · 1 · 1501010.1038/npjcompumats.2015.10 ↗
8. Scheidgen · J. Open Source Softw. · 2023 · 8 · 538810.21105/joss.05388 ↗
9. Calderon · Comput. Mater. Sci. · 2015 · 108 · 23310.1016/j.commatsci.2015.07.019 ↗
10. Bartel · npj Comput. Mater. · 2020 · 6 · 9710.1038/s41524-020-00362-y ↗
11. Deng · Nat. Mach. Intell. · 2023 · 5 · 103110.1038/s42256-023-00716-3 ↗
12. B. M.Wood , M.Dzamba , X.Fu , M.Gao , M.Shuaibi , L.Barroso-Luque , K.Abdelmaqsoud , V.Gharakhanyan , J. R.Kitchin , D. S.Levine , K.Michel , A.Sriram , T.Cohen , A.Das , A.Rizvi , S. J.Sahoo , Z. W.Ulissi and C. L.Zitnick , UMA: A Family of Universal Models for Atoms , arXiv , 2026 , preprint, arXiv:2506.23971 10.48550/arXiv.2506.2397110.48550/arXiv.2506.23971 ↗
13. M.Neumann , J.Gin , B.Rhodes , S.Bennett , Z.Li , H.Choubisa , A.Hussey and J.Godwin , Orb: A Fast, Scalable Neural Network Potential , arXiv , 2024 , preprint, arXiv:2410.22570 10.48550/arXiv.2410.2257010.48550/arXiv.2410.22570 ↗
14. H.Yang , C.Hu , Y.Zhou , X.Liu , Y.Shi , J.Li , G.Li , Z.Chen , S.Chen , C.Zeni , M.Horton , R.Pinsler , A.Fowler , D.Zügner , T.Xie , J.Smith , L.Sun , Q.Wang , L.Kong , C.Liu , H.Hao and Z.Lu , MatterSim: A Deep Learning Atomistic Model Across Elements, Temperatures and Pressures , arXiv , 2024 , preprint, arXiv:2405.04967 10.48550/arXiv.2405.0496710.48550/arXiv.2405.04967 ↗
15. Bartel · J. Mater. Sci. · 2022 · 57 · 1047510.1007/s10853-022-06915-4 ↗
16. Zagorac · J. Appl. Crystallogr. · 2019 · 52 · 91810.1107/s160057671900997x ↗
17. Sun · Sci. Adv. · 2016 · 2 · e160022510.1126/sciadv.1600225 ↗
18. Cavignac · J. Phys. Mater. · 2026 · 9 · 02501410.1088/2515-7639/ae6620 ↗
19. Kovnir · Chem. Mater. · 2021 · 33 · 483510.1021/acs.chemmater.1c01484 ↗
20. Aykol · J. Am. Chem. Soc. · 2021 · 143 · 924410.1021/jacs.1c04888 ↗
21. Neilson · J. Mater. Res. · 2023 · 38 · 288510.1557/s43578-023-01037-2 ↗
22. Aykol · Nat. Commun. · 2019 · 10 · 201810.1038/s41467-019-10030-5 ↗
23. Szymanski · Nat. Commun. · 2023 · 14 · 695610.1038/s41467-023-42329-9 ↗
24. Szymanski · Sci. Adv. · 2024 · 10 · eadp330910.1126/sciadv.adp3309 ↗
25. McDermott · Nat. Commun. · 2021 · 12 · 309710.1038/s41467-021-23339-x ↗
26. McDermott · ACS Cent. Sci. · 2023 · 9 · 195710.1021/acscentsci.3c01051 ↗
27. Kononova · Sci. Data · 2019 · 6 · 20310.1038/s41597-019-0224-1 ↗
28. Chung · Digital Discovery · 2025 · 4 · 243910.1039/D5DD00065C ↗
29. Jang · J. Am. Chem. Soc. · 2020 · 142 · 1883610.1021/jacs.0c07384 ↗
30. Jang · Matter · 2024 · 7 · 229410.1016/j.matt.2024.05.002 ↗
31. Antoniuk · npj Comput. Mater. · 2023 · 9 · 15510.1038/s41524-023-01114-4 ↗
32. Amariamir · Digital Discovery · 2025 · 4 · 143710.1039/D4DD00394B ↗
33. Gleaves · Digital Discovery · 2023 · 2 · 37710.1039/D2DD00098A ↗
34. Animitsa · J. Solid State Chem. · 2016 · 238 · 15610.1016/j.jssc.2016.03.023 ↗
35. Liu · J. Electrochem. Soc. · 2007 · 154 · J18510.1149/1.2735912 ↗
36. Park · Nat. Commun. · 2025 · 16 · 437910.1038/s41467-025-59636-y ↗
37. Szymanski · Mater. Horiz. · 2025 · 12 · 800010.1039/D5MH00010F ↗
38. Riebesell · Nat. Mach. Intell. · 2025 · 7 · 83610.1038/s42256-025-01055-1 ↗
39. Chamorro · Acc. Chem. Res. · 2018 · 51 · 291810.1021/acs.accounts.8b00382 ↗
40. Neilson · J. Mater. Res. · 2023 · 38 · 288510.1557/s43578-023-01037-2 ↗
41. Bartel · Nat. Commun. · 2018 · 9 · 416810.1038/s41467-018-06682-4 ↗
42. Wang · Sci. Data · 2022 · 9 · 23110.1038/s41597-022-01317-2 ↗
43. Wang · Nat. Synth. · 2024 · 3 · 52710.1038/s44160-023-00479-0 ↗
44. Szymanski · Nature · 2023 · 624 · 8610.1038/s41586-023-06734-w ↗
45. Jain · Phys. Rev. B · 2011 · 84 · 04511510.1103/PhysRevB.84.045115 ↗
46. M.Chase , NIST-JANAF Thermochemical Tables , American Institute of Physics , 4th edn, 1998 , vol. 1
47. Virtanen · Nat. Methods · 2020 · 17 · 26110.1038/s41592-019-0686-2 ↗
48. Hjorth Larsen · J. Phys.: Condens. Matter · 2017 · 29 · 27300210.1088/1361-648X/aa680e ↗
49. Deng · npj Comput. Mater. · 2025 · 11 · 910.1038/s41524-024-01500-6 ↗
50. Ong · Comput. Mater. Sci. · 2013 · 68 · 31410.1016/j.commatsci.2012.10.028 ↗
51. Goodall · Nat. Commun. · 2020 · 11 · 628010.1038/s41467-020-19964-7 ↗
52. Zhou · Proc. Natl. Acad. Sci. U. S. A. · 2018 · 115 · E641110.1073/pnas.1801181115 ↗
53. Xie · Phys. Rev. Lett. · 2018 · 120 · 14530110.1103/PhysRevLett.120.145301 ↗
54. Choudhary · npj Comput. Mater. · 2021 · 7 · 18510.1038/s41524-021-00650-1 ↗
55. Schütt · J. Chem. Theory Comput. · 2019 · 15 · 44810.1021/acs.jctc.8b00908 ↗