Understanding the links between structures and properties of materials is an important challenge in chemical research. Atomistic simulations based on density-functional theory (DFT) have played important roles in this – but they are computationally expensive, and therefore they can describe structurally complex materials only in small model systems. Machine learning (ML) based force fields are a rapidly emerging approach that aims to overcome this limitation: being "trained" on DFT data for suitably chosen reference configurations, they achieve close-to-DFT accuracy at a small fraction of the computational cost, and therefore give access to substantially larger system sizes.<br/><br/>In this presentation, I will showcase some recent advances in the modeling and understanding of inorganic materials that have been enabled by ML force fields. I will argue that these methods are particularly useful for studies of structurally complex inorganic solids, for example, of non-crystalline materials – affording the required accuracy and flexibility whilst allowing simulations to reach beyond the nanometer length scale. These ML-driven studies range from the structural characterization of amorphous elements to the computational modeling of materials for practical applications – for example, of disordered battery materials. I will also discuss envisioned synergies between ML force fields, established DFT methods, and experiments, which are hoped to be fully exploited in materials chemistry in the years ahead.

Machine-learning-driven atomistic simulations for inorganic materials chemistry

ACS Fall 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.