Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MM: Fachverband Metall- und Materialphysik
MM 4: Data Driven Material Science: Big Data and Workflows I
MM 4.2: Vortrag
Montag, 18. März 2024, 10:30–10:45, C 243
Universally Accurate or Specifically Inadequate? Stress-Testing General Purpose Machine Learning Interatomic Potentials — •Konstantin Jakob1, Karsten Reuter1, and Johannes T. Margraf1,2 — 1Fritz-Haber-Insttitut der MPG, Berlin — 2Universität Bayreuth
Machine learning interatomic potentials (MLIPs) have revolutionized the field of atomistic materials simulation, both due to their remarkable accuracy - when trained adequately - and their computational efficiency compared to established ab initio methods. Very recently, several general purpose MLIPs have been reported, which are broadly applicable across the periodic table. These represent a fascinating opportunity for materials discovery, provided that they are robust and transferable. In this context, metastability and polymorphism pose significant challenges, as the underlying training data sets cannot cover the full space of such structures and compositions. In order to stress test current general purpose MLIPs, we evaluate models based on the M3GNet and MACE architectures on a unique set of inorganic, crystalline materials generated by atom substitutions. Validating these two models, we shine light on both successes and drawbacks of using general purpose MLIPs and evaluate the opportunities that further research can hold.
Keywords: Density functional theory; MLIPs