Regensburg 2022 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MM: Fachverband Metall- und Materialphysik
MM 29: Data Driven Materials Science: Design of Functional Materials
MM 29.2: Vortrag
Donnerstag, 8. September 2022, 10:30–10:45, H45
Accelerating the High-Throughput Search for new Thermal Insulators with Symbolic Regression — •Thomas Purcell1, Matthias Scheffler1, 2, Luca M. Ghiringhelli1, 2, and Christian Carbogno1 — 1The NOMAD Laboratory at Fritz-Haber-Institut der Max-Planck-Gesellschaft — 2FAIRmat at Humboldt Universität zu Berlin, Berlin, Germany
Reliable artificial-intelligence models are key to accelerate the discovery of new functional materials for various applications. Here, we present a general, data-driven framework that combines symbolic regression with sensitivity analysis to create hierarchical workflows. We illustrate the power of this new framework by screening for new thermally insulating materials. We first use the sure-independence screening and sparsifying operator (SISSO) [1] to build an analytical model that describes the thermal conductivity of a material and then extract out the most important input properties using a variance-based sensitivity analysis [2]. Using the information gained from the analysis we screen over a set of 732 materials and find the region of space most likely to contain strong thermal insulators. Finally we confirm these predictions by calculating thermal conductivities using the ab initio Green-Kubo technique [3].
[1] R. Ouyang, et al.. Phys. Rev. Mat. 2, 083802 (2018)
[2] S. Kucherenko, S. Tarantola, and P. Annoni. Comput. Phys. Commun. 183, 937 (2012)
[3] C. Carbogno, R. Ramprasad, and M. Scheffler. Phys. Rev. Lett. 118, 175901 (2017)