Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MM: Fachverband Metall- und Materialphysik

MM 6: Topical Session: Data Driven Materials Science - Materials Design II (joint session MM/CPP)

MM 6.4: Vortrag

Montag, 16. März 2020, 12:30–12:45, BAR 205

Data-Efficient Machine Learning for Crystal Structure Prediction — •Simon Wengert¹, Gábor Csányi², Karsten Reuter¹, and Johannes T. Margraf¹ — ¹Chair of Theoretical Chemistry, TU Munich, Germany — ²Department of Engineering, University of Cambridge, UK

The combination of modern machine learning (ML) approaches with high-quality data from quantum mechanical (QM) calculations can yield models with an unrivalled accuracy/cost ratio. However, such methods are ultimately limited by the computational effort required to produce the reference data. In particular, reference calculations for periodic systems with many atoms can become prohibitively expensive. This trade-off is critical for crystal structure prediction. Here, a data-efficient ML approach would be highly desirable, since screening a huge space of polymorphs with small stability differences requires the assessment of a large number of trial structures with high accuracy.

In this contribution, we present tailored hybrid-ML models that allow screening a wide range of crystal candidates while adequately describing the subtle interplay between intermolecular interactions such as H-bonding and many-body dispersion effects. This is achieved by enhancing a physics-based description of long-range interactions ---for which an efficient implementation is available--- with a short-range ML model trained on high-quality first-principles reference data. The presented workflow is broadly applicable to different molecular systems, without the need for a single periodic calculation on the reference level of theory.