Roadmap on Machine learning in electronic structure

Abstract

In recent years, we have been witnessing a paradigm shift in computational materials science. In fact, traditional methods, mostly developed in the second half of the XXth century, are being complemented, extended, and sometimes even completely replaced by faster, simpler, and often more accurate approaches. The new approaches, that we collectively label by machine learning, have their origins in the fields of informatics and artificial intelligence, but are making rapid inroads in all other branches of science. With this in mind, this Roadmap article, consisting of multiple contributions from experts across the field, discusses the use of machine learning in materials science, and share perspectives on current and future challenges in problems as diverse as the prediction of materials properties, the construction of force-fields, the development of exchange correlation functionals for density-functional theory, the solution of the many-body problem, and more. In spite of the already numerous and exciting success stories, we are just at the beginning of a long path that will reshape materials science for the many challenges of the XXIth century.

Physics-based representations for machine learning properties of chemical reactions

“…Because QML representations are rooted in foundational laws of nature, they are extremely transferable and do not need to be adapted to each specific learning task. Given their transferability, generality, and deep connection to electronic targets, QML representations have been the forefront of machine learning applied to solve chemical problems [2][3][4][5][6][7].…”

Computational vibrational spectroscopy of molecule–surface interactions: what is still difficult and what can be done about it

“…3.2.2 Examples of applications to molecule-surface systems. Applications of the variational approach are necessarily limited to systems where sufficiently good (spectroscopically accurate 87 ) PESs are available. This outright excludes the vast majority of molecule-surface systems actively studied in laboratories.…”

Electronic-Structure Properties from Atom-Centered Predictions of the Electron Density

“…In the past few years, many efforts have been devoted to the integration of supervised learning techniques within state-of-the-art electronic-structure methods, aiming at accelerating the calculation of properties beyond ground-state total energies and forces. − Different quantities have been considered as the target of machine learning (ML) models, including molecular multipoles − and polarizabilities, electronic density of states and bandgaps, − kinetic density functionals, − exchange-correlation potentials, − single-particle wave functions, − and Hamiltonians. ,, For problems where DFT is the method of choice, it is attractive to think that one could access all ground-state properties of a system at once by simply being able to predict its real-space electronic density. However, it is not always obvious whether one can actually benefit from computing derived electronic properties from the predicted densities.…”