The synergy of geometric tolerance factor and machine learning in discovering stable materials

Abstract

Assessing stability remains a fundamental prerequisite for deploying materials across a wide range of applications, including batteries, catalysts, and photovoltaics. However, first-principles stability checks such as phonon dispersion and energy above hull calculations typically require days to weeks of computing time per composition, creating a critical bottleneck for truly high-throughput discovery. In this Perspective, we highlight the underutilized potential of geometric tolerance factors (T_f) as lightweight yet informative indicators for rapid stability assessment. First, we review the T_f developed for representative materials systems, including perovskites, spinels, and garnets, and analyze recent cases where such indicators have been integrated into AI-driven materials discovery. Then, we identify key open challenges in designing T_f that are both accurate and generalizable, as well as in effectively incorporating them into AI frameworks. The potential solutions, including active learning for multi-composition structure, electron density profile-based learning for ionic radii estimation, and diffusion model for thermodynamic and kinetic stability, are proposed to address these challenges. The synergy between T_f-based heuristics and advanced AI models has the potential to triage vast compositional spaces before committing to expensive first-principles stability validation, thereby enabling broader innovations in materials design and deployment.

Keywords

Materials informatics / tolerance factor / machine learning / materials discovery / stability assessment

References

[1] Merchant A,Schoenholz SS,Cheon G.Scaling deep learning for materials discovery.Nature2023;624:80-5 PMCID:PMC10700131

[2] Zeni C,Zügner D.A generative model for inorganic materials design.Nature2025;639:624-32 PMCID:PMC11922738

[3] Xu P,Lu W,Zhao W.Multi-objective optimization in machine learning assisted materials design and discovery.J Mater Inf2025;5:26

[4] Westermayr J,Barrett R.High-throughput property-driven generative design of functional organic molecules.Nat Comput Sci2023;3:139-48

[5] Biyela S,Gero KI.Generative AI and science communication in the physical sciences.Nat Rev Phys2024;6:162-5

[6] Njirjak M,Babić M.Reshaping the discovery of self-assembling peptides with generative AI guided by hybrid deep learning.Nat Mach Intell2024;6:1487-500

[7] Li Z,Park J,Berry JJ.Stabilizing perovskite structures by tuning tolerance factor: formation of formamidinium and cesium lead iodide solid-state alloys.Chem Mater2016;28:284-92

[8] Liu X,Lu ZH.Stabilization of photoactive phases for perovskite photovoltaics.Nat Rev Chem2023;7:462-79

[9] Song Z.Tolerance factor and phase stability of the normal spinel structure.Cryst Growth Des2020;20:2014-8

[10] Song Z,Liu Q.Tolerance factor and phase stability of the garnet structure.Acta Crystallogr C Struct Chem2019;75:1353-8

[11] Wang Z,Han Y.Harnessing artificial intelligence to holistic design and identification for solid electrolytes.Nano Energy2021;89:106337

[12] Zhu T,Wessler GC.I₂–II–IV–VI₄ (I = Cu, Ag; II = Sr, Ba; IV = Ge, Sn; VI = S, Se): chalcogenides for thin-film photovoltaics.Chem Mater2017;29:7868-79

[13] Wang Z,Wang Q,Li J.Unsupervised discovery of thin-film photovoltaic materials from unlabeled data.npj Comput Mater2021;7:596

[14] Curtarolo S,Nardelli MB,Sanvito S.The high-throughput highway to computational materials design.Nat Mater2013;12:191-201

[15] Griesemer SD,Wolverton C.Accelerating the prediction of stable materials with machine learning.Nat Comput Sci2023;3:934-45

[16] Wu Z,Wang X.Leveraging language model for advanced multiproperty molecular optimization via prompt engineering.Nat Mach Intell2024;6:1359-69

[17] Jiang X,Tian S,Lookman T.Applications of natural language processing and large language models in materials discovery.npj Comput Mater2025;11:1554

[18] Bartel CJ,Goldsmith BR.New tolerance factor to predict the stability of perovskite oxides and halides.Sci Adv2019;5:eaav0693 PMCID:PMC6368436

[19] Bassen G,Bunstine W,Siegler MA.Tolerance factor approach for the design of quaternary materials as applied to the A₂Ln₄Cu_2nQ_7+n homologous series.J Am Chem Soc2024;146:25190-9

[20] Molokeev MS.Tolerance factor for huntite-family compounds.Phys Solid State2020;62:2058-62

[21] Mouta R,Paschoal CW.Tolerance factor for pyrochlores and related structures.Acta Crystallogr B Struct Sci Cryst Eng Mater2013;69:439-45

[22] Smith, M.; Li, Z.; Landry, L.; Merz, K. M. Jr.; Li, P. Consequences of overfitting the van der Waals radii of ions. J. Chem. Theory. Comput. 2023, 19, 2064-74.

[23] Marchenko EI,Eremin NN,Goodilin EA.Crystal chemical insights on lead iodide perovskites doping from revised effective radii of metal ions.ACS Materials Lett2021;3:1377-84

[24] Turnley JW,Agrawal R.Rethinking tolerance factor analysis for chalcogenide perovskites.Mater Horiz2024;11:4802-8

[25] Mondal D.Structural distortions in hybrid perovskites revisited.Chem Mater2024;36:4254-61

[26] Antoniuk ER,Wang G,Cai W.Predicting the synthesizability of crystalline inorganic materials from the data of known material compositions.npj Comput Mater2023;9:1114

[27] Vasylenko A,Gusev VV,Dyer MS.Element selection for functional materials discovery by integrated machine learning of elemental contributions to properties.npj Comput Mater2023;9:1072

[28] Jha D,Paul A.ElemNet: deep learning the chemistry of materials from only elemental composition.Sci Rep2018;8:17593 PMCID:PMC6279928

[29] Hargreaves CJ,Daniels LM.A database of experimentally measured lithium solid electrolyte conductivities evaluated with machine learning.npj Comput Mater2023;9:951

[30] Kang Y.ChatMOF: an artificial intelligence system for predicting and generating metal-organic frameworks using large language models.Nat Commun2024;15:4705 PMCID:PMC11148193

[31] Zheng Y,Ju J.Large language models for scientific discovery in molecular property prediction.Nat Mach Intell2025;7:437-47

[32] Butler KT,Cartwright H,Walsh A.Machine learning for molecular and materials science.Nature2018;559:547-55

[33] Dass KT, Hossain MK, Marasamy L. Highly efficient emerging Ag₂BaTiSe₄ solar cells using a new class of alkaline earth metal-based chalcogenide buffers alternative to CdS.Sci Rep2024;14:1473 PMCID:PMC10794422

[34] Talirz L,Passaro E.Materials Cloud, a platform for open computational science.Sci Data2020;7:299 PMCID:PMC7479138

[35] Curtarolo S,Hart GL.AFLOW: an automatic framework for high-throughput materials discovery.Comput Mater Sci2012;58:218-26

[36] Sbailò L,Ghiringhelli LM.The NOMAD Artificial-Intelligence Toolkit: turning materials-science data into knowledge and understanding.npj Comput Mater2022;8:935

[37] Szymanski NJ,Daoutidis P.Topological descriptors for the electron density of inorganic solids.ACS Materials Lett2025;7:2158-64

[38] Hylton-Farrington CM.Dynamic local symmetry fluctuations of electron density in halide perovskites.Chem Mater2024;36:9442-59

[39] Feng C,Jiang B.Efficient sampling for machine learning electron density and its response in real space.J Chem Theory Comput2025;21:691-702

[40] Xie T.Crystal graph convolutional neural networks for an accurate and interpretable prediction of material properties.Phys Rev Lett2018;120:145301