Database-driven High-Throughput Calculations and Machine Learning Models for Materials Design

Abstract

This paper reviews past and ongoing efforts in using high-throughput ab-inito calculations in combination with machine learning models for materials design. The primary focus is on bulk materials, i.e., materials with fixed, ordered, crystal structures, although the methods naturally extend into more complicated configurations. Efficient and robust computational methods, computational power, and reliable methods for automated database-driven high-throughput computation are combined to produce high-quality data sets. This data can be used to train machine learning models for predicting the stability of bulk materials and their properties. The underlying computational methods and the tools for automated calculations are discussed in some detail. Various machine learning models and, in particular, descriptors for general use in materials design are also covered.

arXiv comment19 pages, 2 figures

References in corpus (22)

• Restoring the density-gradient expansion for exchange in solids and surfaces
• Generalized gradient approximation for solids and their surfaces
• Van der Waals Density Functional for General Geometries
• A Higher-Accuracy van der Waals Density Functional
• Fast and Accurate Modeling of Molecular Atomization Energies with Machine Learning
• A More Accurate Generalized Gradient Approximation for Solids
• SchNet - a deep learning architecture for molecules and materials
• First-principles prediction of redox potentials in transition-metal compounds with LDA+U
• Long-range correlation energy calculated from coupled atomic response functions
• Low Hole Effective Mass p-type Transparent Conducting Oxides: Identification and Design Principles
• An exchange functional that tests the robustness of the plasmon description of the van der Waals density functional
• Machine Learning Energies of 2 M Elpasolite (ABC${}_2$D${}_6$) Crystals
• Electronic Structure Calculations with the Tran-Blaha Modified Becke-Johnson Density Functional
• Alchemical and structural distribution based representation for improved QML
• Deep Neural Networks for Accurate Predictions of Garnet Stability
• Defect Formation Energies without the Band-Gap Problem: Combining DFT and GW for the Silicon Self-Interstitial
• Density Functionals that Recognize Covalent, Metallic, and Weak Bonds
• Assessing the SCAN functional for itinerant electron ferromagnets
• Universal correction for the Becke-Johnson exchange potential
• High-Throughput Screening of Perovskite Alloys for Piezoelectric Performance and Formability
• Comparison between exact and semilocal exchange potentials: An all-electron study for solids
• Energetics of the AK13 Semi-Local Kohn-Sham Exchange Energy Functional

Cited by in corpus (11)

• OPTIMADE, an API for exchanging materials data
• How to verify the precision of density-functional-theory implementations via reproducible and universal workflows
• Autonomous Reaction Network Exploration in Homogeneous and Heterogeneous Catalysis
• Developments and applications of the OPTIMADE API for materials discovery, design, and data exchange
• Common workflows for computing material properties using different quantum engines
• An Ontology for the Materials Design Domain
• Exhaustive characterization of modified Si vacancies in 4H-SiC
• Absorption $\text{versus}$ Adsorption: High-Throughput Computation of Impurities in 2D Materials
• Discovery of Atomic Clock-Like Spin Defects in Simple Oxides from First Principles
• Na in Diamond: High Spin Defects Revealed by the ADAQ High-Throughput Computational Database
• Making atomistic materials calculations accessible with the AiiDAlab Quantum ESPRESSO app