Speaker
Description
BiGmax is a Max Planck Network of 10 Max Planck Institutes that addresses the various challenges of big-data-driven materials. It is coordinated by Peter Benner (Max Planck Institute for Dynamics of Complex Technical Systems, Magdeburg) and Matthias Scheffler [1]. The NOMAD Laboratory contributes with projects in heterogeneous catalysis (see poster #10) and together with Claudia Draxl on the FAIRmat project (see Poster #1).
This poster here describes our project on “crystal-structure identification via Bayesian deep learning”. Due to their ability to recognize complex patterns, neural networks can drive a paradigm shift in the analysis of materials science data. Here, we introduce ARISE, a crystal-structure identification method based on Bayesian deep learning [2]. As a major step forward, ARISE is robust to structural noise and can treat more than 100 crystal structures, a number that can be extended on demand. While being trained on ideal structures only, ARISE correctly characterizes strongly perturbed single- and polycrystalline systems, from both synthetic and experimental resources. The probabilistic nature of the Bayesian-deep-learning model allows obtaining principled uncertainty estimates, which are found to be correlated with crystalline order of metallic nanoparticles in electron tomography experiments. Applying unsupervised learning to the internal neural-network representations reveals grain boundaries and (unapparent) structural regions sharing easily interpretable geometrical properties. This work enables the hitherto hindered analysis of noisy atomic structural data from computations or experiments.
References
[1] The webpage of BiGmax is https://www.bigmax.mpg.de
[2] Leitherer, A. Ziletti, and L. M. Ghiringhelli, Robust recognition and exploratory analysis of crystal structures via Bayesian deep learning, Nat Commun 12, 6234 (2021). https://doi.org/10.1038/s41467-021-26511-5
Addresses
(a) Present address: Bayer Pharmaceuticals Berlin, Germany
(b) Present address: The FAIRmat Consortium of the NFDI, Humboldt-Universität zu Berlin, Berlin, Germany
| Abstract Number (department-wise) | SG 06 |
|---|---|
| Department | Scheffler Group |
