A foundation model for atomistic materials chemistry

• DOI: 10.48550/arxiv.2401.00096
• Publication Date: 2024-01-01
• Authors: Ilyes Batatia; Philipp Benner; Yuan Chiang; Alin M. Elena; Dávid Péter Kovács; Janosh Riebesell; Xavier R. Advincula; Mark Asta; William J. Baldwin; Noam Bernstein; Arghya Bhowmik; Samuel M. Blau; Vlad Cărare; James P. Darby; Sandip De; Flaviano Della Pia; Volker L. Deringer; Rokas Elijošius; Zakariya El-Machachi; Edvin Fako; Andrea C. Ferrari; Annalena R. Genre...; Janine George; Rhys E. A. Goodall; Clare P. Grey; Shuang Han; Will Handley; Hendrik H. Heenen; Kersti Hermansson; Christian Holm; Jad Jaafar; Stephan Hofmann; Konstantin S. Jakob; Hyunwook Jung; Venkat Kapil; Aaron D. Kaplan; Nima Karimitari; Namu Kroupa; Jolla Kullgren; Matthew C. Kuner; Domantas Kuryla; Guoda Liepuoniute; Johannes T. Margraf; Ioan-Bogdan Magdău; Angelos Michaelides; J. Harry Moore; Aakash Ashok Naik; Samuel P. Niblett; Sam Walton Norwood; Niamh O’Neill; Christoph Ortner; Kristin A. Persson; Karsten Reuter; Andrew Rosen; Lars L. Schaaf; Christoph Schran; Eric Sivonxay; Tamás K. Stenczel; Viktor Svahn; Christopher Sutton; van der Oord; Eszter Varga-Umbrich; Tejs Vegge; Martin Vondrák; Yangshuai Wang; William C. Witt; Fabian Zills; Gábor Csányi
• Keywords: Transferability; Foundation (evidence)

Machine-learned force fields have transformed the atomistic modelling of materials by enabling simulations ab initio quality on unprecedented time and length scales. However, they are currently limited by: (i) significant computational human effort that must go into development validation potentials for each particular system interest; (ii) a general lack transferability from one chemical to next. Here, using state-of-the-art MACE architecture we introduce single general-purpose ML model, trained public database 150k inorganic crystals, is capable running stable molecular dynamics molecules materials. We demonstrate power MACE-MP-0 model - its qualitative at times quantitative accuracy diverse set problems in physical sciences, including properties solids, liquids, gases, reactions, interfaces even small protein. The can be applied out box as starting or "foundation model" any interest thus step towards democratising revolution lowering barriers entry.