Recent advances and applications of deep learning methods in materials science

Abstract

Deep learning (DL) is one of the fastest-growing topics in materials data science, with rapidly emerging applications spanning atomistic, image-based, spectral, and textual data modalities. DL allows analysis of unstructured data and automated identification of features. The recent development of large materials databases has fueled the application of DL methods in atomistic prediction in particular. In contrast, advances in image and spectral data have largely leveraged synthetic data enabled by high-quality forward models as well as by generative unsupervised DL methods. In this article, we present a high-level overview of deep learning methods followed by a detailed discussion of recent developments of deep learning in atomistic simulation, materials imaging, spectral analysis, and natural language processing. For each modality we discuss applications involving both theoretical and experimental data, typical modeling approaches with their strengths and limitations, and relevant publicly available software and datasets. We conclude the review with a discussion of recent cross-cutting work related to uncertainty quantification in this field and a brief perspective on limitations, challenges, and potential growth areas for DL methods in materials science.

Multi-task graph neural networks for simultaneous prediction of global and atomic properties in ferromagnetic systems *

“…where ∥•∥ * may denote the ℓ 1 -norm or ℓ 2 -norm as explained below. The objective function to be minimized in (11) interprets the constraint as a penalization term with the penalization multiplier λ. This is a weak formulation of the constraint added to the original objective function.…”

“…This is a weak formulation of the constraint added to the original objective function. Standard ℓ 1 or ℓ 2 regularizations correspond to choosing c(w) = w and g = 0, and they recast the constraint term in (11) from the strong formulation to the weak formulation using the ℓ 1 -norm and the ℓ 2 -norm, respectively. In MTL, c(w) are the predictions of additional target properties whose target values are stored in g, which allows to recast the global objective function used in equation (11) as the global loss function for MTL defined in equation (9).…”

“…Standard ℓ 1 or ℓ 2 regularizations correspond to choosing c(w) = w and g = 0, and they recast the constraint term in (11) from the strong formulation to the weak formulation using the ℓ 1 -norm and the ℓ 2 -norm, respectively. In MTL, c(w) are the predictions of additional target properties whose target values are stored in g, which allows to recast the global objective function used in equation (11) as the global loss function for MTL defined in equation (9). For a fair comparison and to determine the benefit of using other tasks as a mutual regularizer, we do not use additional ℓ 1 and ℓ 2 regularizers for the STL training in this work.…”

“…Once HydraGNN is trained, it is much faster than DFT. material science is most often small in volume due to expensive experiments and time-consuming numerical simulations, which challenges, but does not preclude, the use of DL models in the field [11]. Recent efforts have proposed transfer learning [12] and injecting the model with pre-existing physical knowledge [13][14][15] to overcome this data constraint.…”

A simple denoising approach to exploit multi-fidelity data for machine learning materials properties

“…With the considerable increase in available data, chemistry and materials science are undergoing a revolution as attested by the multiplication of data-driven studies in recent years. [1][2][3][4][5] Among all these investigations, the prediction of properties from the atomic structure (or sometimes just the chemical composition) is an extremely important topic. 6 Indeed, having an accurate predictor can be very useful for accelerating high-throughput material screening, 7,8 generating new molecules, 3 or classifying reactions.…”