Seminar by Prof. Volker Deringer on machine-learning-based simulation of amorphous materials

OCAMM presents a special seminar by Prof. Volker Deringer from the University of Oxford, UK on machine-learning-based simulation of amorphous materials. This invited seminar will take place at the Department of Chemistry and Materials Science, Aalto University on 18 March 2024 @ 14:15 in lecture hall A304 (Ke2) at the main building of the School of Chemical Engineering in the Otaniemi campus, Kemistintie 1, 02150 Espoo. The seminar is open to all. Please join us in Otaniemi to delve into the intricate atomic structure and fascinating properties of amorphous materials!

Title

Machine-learning-driven advances in modelling amorphous materials

Abstract

Understanding the connections between the atomic-scale structure of materials and their macroscopic properties is among the most important research challenges in solid-state and materials chemistry. Atomistic simulations based on quantum-mechanical methods have played a key role in this – but they are computationally demanding, and therefore they will inevitably reach their limits when materials with highly complex structures are to be studied. Machine learning (ML) based interatomic potentials are a rapidly emerging approach that helps to overcome this limitation: being “trained” on a suitably chosen set of quantummechanical data, ML potentials achieve comparable accuracy whilst giving access to much larger-scale simulations – with thousands or even millions of atoms.
In this presentation, I will showcase some recent advances in the modelling and understanding of inorganic materials that have been enabled by ML-driven simulations. I will argue that ML potentials are particularly useful for modelling non-crystalline (amorphous) structures that are difficult to characterise experimentally. I will survey recent work ranging from structural transitions in amorphous silicon [1] to multicomponent systems – specifically, chalcogenide phase-change materials used in digital data storage [2]. I will also discuss methodological aspects, including a perspective for using large synthetic datasets to pre-train neural-network potentials which can subsequently be fine-tuned on quantum-mechanical data [3]. The development of new, accurate and efficient atomistic ML models promises a way to more fully understand the structure and properties of amorphous materials on the atomic scale.
[1] V. L. Deringer, N. Bernstein, G. Csányi, C. Ben Mahmoud, M. Ceriotti, M. Wilson, D. A. Drabold, S. R. Elliott, Nature 2021, 589, 59.
[2] Y. Zhou, W. Zhang, E. Ma, V. L. Deringer, Nat. Electron. 2023, 6, 746.
[3] J. L. A. Gardner, K. T. Baker, V. L. Deringer, Mach. Learn.: Sci. Technol. 2024, 5, 015003.