Machine learning unifies the modeling of materials and molecules

By A. P. Bart'ok, Sandip De, Carl Poelking, Noam Bernstein, James R. Kermode, Gábor Csányi, Michele Ceriotti. 2017

652 citations recorded by the source corpus.

Statistical learning based on a local representation of atomic structures provides a universal model of chemical stability. Determining the stability of molecules and condensed phases is the cornerstone of atomistic modeling, underpinning our understanding of chemical and materials properties and transformations. We show that a machine-learning model, based on a local description of chemical environments and Bayesian statistical learning, provides a unified framework to predict atomic-scale properties. It captures the quantum mechanical effects governing the complex surface reconstructions of silicon, predicts the stability of different classes of molecules with chemical accuracy, and distinguishes active and inactive protein ligands with more than 99% reliability. The universality and the systematic nature of our framework provide new insight into the potential energy surface of materials and molecules.

Subfields: Computational Theory and Mathematics, Materials Chemistry, Computational Chemistry & Drug Discovery, AI for Physical Sciences

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Theory and Mathematics

In Materials Chemistry

In Computational Chemistry & Drug Discovery

In AI for Physical Sciences

Citations in the corpus

Relevance of Rotationally Equivariant Convolutions for Predicting Molecular Properties (2020, background): “…properties, which depend only on the atomic distance graph, are commonly predicted by kernel methods or Gaussian process regression [14, 15, 16, 17] or graph neural networks [18, 19], where ENNs are usually employed for predicting physical properties, which depend on atomic displacement…”
Graph Networks as a Universal Machine Learning Framework for Molecules and Crystals (2018, background): “Machine learning (ML) 1,2 has emerged as a powerful new tool in materials science, 3–14 driven in part by the advent of large materials data sets from high-throughput electronic structure calculations 15–18 and/or combinatorial experiments.”
Machine-learned potentials for next-generation matter simulations (2021, methodology): “…of ML potentials combined with active learning approaches, in particular for high-energy configurations, enables the reliable simulation of various phases of single-component materials such as carbon 68,109 , silicon 78,87,104 or boron 89 , as well as more complex materials such as GeTe (ref.”
Machine Learning a General-Purpose Interatomic Potential for Silicon (2018, methodology): “The energy differences are much smaller than the target (and assumed) error in the GAP model, and as such this level of detail is not robust: The earlier variant of the potential fitted to a slightly different database [in ways unrelated to the (111) surface] show the 7 × 7 DAS structure as the glob”