Supervised deep learning prediction of the formation enthalpy of the\n full set of configurations in complex phases: the \σ-phase as an\n example

2020/11/21 by Jean‐Claude Crivello, Crivello, Jean-Claude, Nataliya Sokolovska +3
Chemistry · Computer Science · Materials Science · #Computational Drug Discovery Methods #Crystallography and molecular interactions #FOS: Computer and information sciences #FOS: Physical sciences #Machine Learning (cs.LG) #Machine Learning in Materials Science #Materials Science (cond-mat.mtrl-sci)

paper · pdf · doi:10.48550/arxiv.2011.10883

openalex publication_date 2020/11/21 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28

Abstract

Machine learning (ML) methods are becoming integral to scientific inquiry in\nnumerous disciplines, such as material sciences. In this manuscript, we\ndemonstrate how ML can be used to predict several properties in solid-state\nchemistry, in particular the heat of formation of a given complex\ncrystallographic phase (here the \σ-phase, tP30, D8_b). Based on an\nindependent and unprecedented large first principles dataset containing about\n10,000 \σ-compounds with n=14 different elements, we used a supervised\nlearning approach, to predict all the \∼500,000 possible configurations\nwithin a mean absolute error of 23 meV/at (\∼2 kJ.mol^-1) on the heat\nof formation and \∼0.06 Ang. on the tetragonal cell parameters. We showed\nthat neural network regression algorithms provide a significant improvement in\naccuracy of the predicted output compared to traditional regression techniques.\nAdding descriptors having physical nature (atomic radius, number of valence\nelectrons) improves the learning precision. Based on our analysis, the training\ndatabase composed of the only binary-compositions plays a major role in\npredicting the higher degree system configurations. Our result opens a broad\navenue to efficient high-throughput investigations of the combinatorial binary\ncalculation for multicomponent prediction of a complex phase.\n

Citations

• Projector augmented-wave method
• From ultrasoft pseudopotentials to the projector augmented-wave method
• Generalized Gradient Approximation Made Simple
• VESTA 3 for three-dimensional visualization of crystal, volumetric and morphology data
• Special points for Brillouin-zone integrations
• Ab initiomolecular dynamics for liquid metals
• Efficient iterative schemes forab initiototal-energy calculations using a plane-wave basis set
• Self-Consistent Equations Including Exchange and Correlation Effects

Related

• Machine Learning for Improved Density Functional Theory Thermodynamics
• Enhancing crystal structure prediction by combining computational and experimental data via graph networks
• Interpretable machine learning to understand the performance of semi local density functionals for materials thermochemistry
• Formation Energy Prediction of Material Crystal Structures using Deep Learning
• Machine learning approaches for feature engineering of the crystal\n structure: Application to the prediction of the formation energy of cubic\n compounds