De novo exploration and self-guided learning of potential-energy surfaces

Interatomic potential models based on machine learning (ML) are rapidly developing as tools for material simulations. However, because of their flexibility, they require large fitting databases that are normally created with substantial manual selection and tuning of reference configurations. Here, we show that ML potentials can be built in a largely automated fashion, exploring and fitting potential-energy surfaces from the beginning (de novo) within one and the same protocol. The key enabling step is the use of a configuration-averaged kernel metric that allows one to select the few most relevant and diverse structures at each step. The resulting potentials are accurate and robust for the wide range of configurations that occur during structure searching, despite only requiring a relatively small number of single-point DFT calculations on small unit cells. We apply the method to materials with diverse chemical nature and coordination environments, marking an important step toward the more routine application of ML potentials in physics, chemistry, and materials science.

[1]  K. Müller,et al.  Towards exact molecular dynamics simulations with machine-learned force fields , 2018, Nature Communications.

[2]  Georg Kresse,et al.  On-the-fly machine learning force field generation: Application to melting points , 2019, Physical Review B.

[3]  Nongnuch Artrith,et al.  An implementation of artificial neural-network potentials for atomistic materials simulations: Performance for TiO2 , 2016 .

[4]  H. Hillebrecht,et al.  Boron: elementary challenge for experimenters and theoreticians. , 2009, Angewandte Chemie.

[5]  Noam Bernstein,et al.  Exploration, Sampling, And Reconstruction of Free Energy Surfaces with Gaussian Process Regression. , 2016, Journal of chemical theory and computation.

[6]  Noam Bernstein,et al.  A universal preconditioner for simulating condensed phase materials. , 2016, The Journal of chemical physics.

[7]  Muratahan Aykol,et al.  Thermodynamic limit for synthesis of metastable inorganic materials , 2018, Science Advances.

[8]  Yanming Ma,et al.  Ionic high-pressure form of elemental boron , 2009, Nature.

[9]  J. Behler,et al.  Pressure‐induced phase transitions in silicon studied by neural network‐based metadynamics simulations , 2008 .

[10]  J. Behler First Principles Neural Network Potentials for Reactive Simulations of Large Molecular and Condensed Systems. , 2017, Angewandte Chemie.

[11]  Kresse,et al.  Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. , 1996, Physical review. B, Condensed matter.

[12]  Beatriz Cordero,et al.  Covalent radii revisited. , 2008, Dalton transactions.

[13]  M. Dion,et al.  van der Waals density functional for general geometries. , 2004, Physical review letters.

[14]  S. Goedecker,et al.  Two-Dimensional Hexagonal Sheet of TiO2 , 2017 .

[15]  Sebastian E. Ahnert,et al.  Revealing and exploiting hierarchical material structure through complex atomic networks , 2017, npj Computational Materials.

[16]  M I McMahon,et al.  Crystal structures of dense lithium: a metal-semiconductor-metal transition. , 2011, Physical review letters.

[17]  R. Kondor,et al.  Gaussian approximation potentials: the accuracy of quantum mechanics, without the electrons. , 2009, Physical review letters.

[18]  Jonathan Vandermause,et al.  On-the-Fly Bayesian Active Learning of Interpretable Force-Fields for Atomistic Rare Events , 2019 .

[19]  Gábor Csányi,et al.  Gaussian approximation potentials: A brief tutorial introduction , 2015, 1502.01366.

[20]  Alexander V. Shapeev,et al.  Moment Tensor Potentials: A Class of Systematically Improvable Interatomic Potentials , 2015, Multiscale Model. Simul..

[21]  Chris J Pickard,et al.  High-pressure phases of silane. , 2006, Physical review letters.

[22]  Artem R. Oganov,et al.  Structure prediction drives materials discovery , 2019, Nature Reviews Materials.

[23]  Yanming Ma,et al.  Ionic high-pressure form of elemental boron , 2009, Nature.

[24]  Stefano de Gironcoli,et al.  Reproducibility in density functional theory calculations of solids , 2016, Science.

[25]  J. Behler,et al.  Metadynamics simulations of the high-pressure phases of silicon employing a high-dimensional neural network potential. , 2008, Physical review letters.

[26]  Germany,et al.  Neural network interatomic potential for the phase change material GeTe , 2012, 1201.2026.

[27]  Jörg Behler,et al.  Automatic selection of atomic fingerprints and reference configurations for machine-learning potentials. , 2018, The Journal of chemical physics.

[28]  Fujio Izumi,et al.  VESTA 3 for three-dimensional visualization of crystal, volumetric and morphology data , 2011 .

[29]  Georg Kresse,et al.  Phase Transitions of Hybrid Perovskites Simulated by Machine-Learning Force Fields Trained on the Fly with Bayesian Inference. , 2019, Physical review letters.

[30]  Gabor Csanyi,et al.  Achieving DFT accuracy with a machine-learning interatomic potential: thermomechanics and defects in bcc ferromagnetic iron , 2017, 1706.10229.

[31]  Nongnuch Artrith,et al.  Understanding the composition and activity of electrocatalytic nanoalloys in aqueous solvents: a combination of DFT and accurate neural network potentials. , 2014, Nano letters.

[32]  G. Kresse,et al.  From ultrasoft pseudopotentials to the projector augmented-wave method , 1999 .

[33]  Blöchl,et al.  Projector augmented-wave method. , 1994, Physical review. B, Condensed matter.

[34]  O. Isayev,et al.  ible neural network potential with DFT accuracy at force fi eld computational cost † , 2017 .

[35]  D. Bowler,et al.  Van der Waals density functionals applied to solids , 2011, 1102.1358.

[36]  Samad Hajinazar,et al.  Stratified construction of neural network based interatomic models for multicomponent materials , 2016, 1609.08455.

[37]  Xiaojun Wu,et al.  Two-dimensional boron monolayer sheets. , 2012, ACS nano.

[38]  L. Hector,et al.  Geometries of edge and mixed dislocations in bcc Fe from first-principles calculations , 2018, Physical Review Materials.

[39]  Jonathan Vandermause,et al.  On-the-fly active learning of interpretable Bayesian force fields for atomistic rare events , 2020, npj Computational Materials.

[40]  E Weinan,et al.  Deep Potential Molecular Dynamics: a scalable model with the accuracy of quantum mechanics , 2017, Physical review letters.

[41]  Klaus-Robert Müller,et al.  Machine learning of accurate energy-conserving molecular force fields , 2016, Science Advances.

[42]  Gábor Csányi,et al.  Comparing molecules and solids across structural and alchemical space. , 2015, Physical chemistry chemical physics : PCCP.

[43]  Volker L. Deringer,et al.  Extracting Crystal Chemistry from Amorphous Carbon Structures , 2017, Chemphyschem : a European journal of chemical physics and physical chemistry.

[44]  Michele Parrinello,et al.  Generalized neural-network representation of high-dimensional potential-energy surfaces. , 2007, Physical review letters.

[45]  Alexander V. Shapeev,et al.  Active learning of linearly parametrized interatomic potentials , 2016, 1611.09346.

[46]  Noam Bernstein,et al.  Machine Learning a General-Purpose Interatomic Potential for Silicon , 2018, Physical Review X.

[47]  Thomas D. Kuhne,et al.  Ab initio quality neural-network potential for sodium , 2010, 1002.2879.

[48]  Noam Bernstein,et al.  Realistic Atomistic Structure of Amorphous Silicon from Machine-Learning-Driven Molecular Dynamics. , 2018, The journal of physical chemistry letters.

[49]  T. Strobel,et al.  Synthesis of an open-framework allotrope of silicon. , 2015, Nature materials.

[50]  William A. Curtin,et al.  Screw dislocation structure and mobility in body centered cubic Fe predicted by a Gaussian Approximation Potential , 2018, npj Computational Materials.

[51]  Volker L. Deringer,et al.  Similarity Between Amorphous and Crystalline Phases: The Case of TiO2. , 2018, The journal of physical chemistry letters.

[52]  Gus L. W. Hart,et al.  Accelerating high-throughput searches for new alloys with active learning of interatomic potentials , 2018, Computational Materials Science.

[53]  Bjørk Hammer,et al.  Neural-network-enhanced evolutionary algorithm applied to supported metal nanoparticles , 2018 .

[54]  G. Scuseria,et al.  Generalized gradient approximation for solids and their surfaces , 2007, 0707.2088.

[55]  Petros Drineas,et al.  CUR matrix decompositions for improved data analysis , 2009, Proceedings of the National Academy of Sciences.

[56]  Chris J Pickard,et al.  Highly compressed ammonia forms an ionic crystal. , 2008, Nature materials.

[57]  E Weinan,et al.  Active Learning of Uniformly Accurate Inter-atomic Potentials for Materials Simulation , 2018, Physical Review Materials.

[58]  Rampi Ramprasad,et al.  A universal strategy for the creation of machine learning-based atomistic force fields , 2017, npj Computational Materials.

[59]  Noam Bernstein,et al.  Machine learning unifies the modeling of materials and molecules , 2017, Science Advances.

[60]  J S Smith,et al.  ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost , 2016, Chemical science.

[61]  A. J. Morris,et al.  Investigating Sodium Storage Mechanisms in Tin Anodes: A Combined Pair Distribution Function Analysis, Density Functional Theory, and Solid-State NMR Approach. , 2017, Journal of the American Chemical Society.

[62]  Gábor Csányi,et al.  Accuracy and transferability of Gaussian approximation potential models for tungsten , 2014 .

[63]  Noam Bernstein,et al.  Modeling the Phase-Change Memory Material, Ge2Sb2Te5, with a Machine-Learned Interatomic Potential. , 2018, The journal of physical chemistry. B.

[64]  S. Hajinazar,et al.  Multitribe evolutionary search for stable Cu-Pd-Ag nanoparticles using neural network models. , 2019, Physical chemistry chemical physics : PCCP.

[65]  Volker L. Deringer,et al.  Growth Mechanism and Origin of High sp^{3} Content in Tetrahedral Amorphous Carbon. , 2018, Physical review letters.

[66]  Alexander V. Shapeev,et al.  Accelerating crystal structure prediction by machine-learning interatomic potentials with active learning , 2018, Physical Review B.

[67]  Volker L. Deringer,et al.  Data-driven learning and prediction of inorganic crystal structures. , 2018, Faraday discussions.

[68]  Artem R. Oganov,et al.  Synthesis of borophenes: Anisotropic, two-dimensional boron polymorphs , 2015, Science.

[69]  R. Needs,et al.  Hypothetical low-energy chiral framework structure of group 14 elements , 2010 .

[70]  Michele Ceriotti,et al.  Mapping uncharted territory in ice from zeolite networks to ice structures , 2018, Nature Communications.

[71]  Volker L. Deringer,et al.  Data-Driven Learning of Total and Local Energies in Elemental Boron. , 2017, Physical review letters.

[72]  Yanchao Wang,et al.  Accelerating CALYPSO structure prediction by data-driven learning of a potential energy surface. , 2018, Faraday discussions.

[73]  J. Soler,et al.  Efficient implementation of a van der Waals density functional: application to double-wall carbon nanotubes. , 2008, Physical review letters.

[74]  Zhenwei Li,et al.  Molecular dynamics with on-the-fly machine learning of quantum-mechanical forces. , 2015, Physical review letters.

[75]  Volker L. Deringer,et al.  Machine learning based interatomic potential for amorphous carbon , 2016, 1611.03277.

[76]  Nongnuch Artrith,et al.  High-dimensional neural network potentials for metal surfaces: A prototype study for copper , 2012 .

[77]  Michael O'Keeffe,et al.  Identification of and symmetry computation for crystal nets. , 2003, Acta crystallographica. Section A, Foundations of crystallography.

[78]  Atsuto Seko,et al.  Conceptual and practical bases for the high accuracy of machine learning interatomic potential , 2017, 1708.02741.

[79]  R. Kondor,et al.  On representing chemical environments , 2012, 1209.3140.

[80]  Chris J Pickard,et al.  Ab initio random structure searching , 2011, Journal of physics. Condensed matter : an Institute of Physics journal.

[81]  Tomi Laurila,et al.  Reactivity of Amorphous Carbon Surfaces: Rationalizing the Role of Structural Motifs in Functionalization Using Machine Learning , 2018, Chemistry of materials : a publication of the American Chemical Society.

[82]  Y. Vohra,et al.  Omega phase in materials , 1982 .

[83]  G. Scuseria,et al.  Restoring the density-gradient expansion for exchange in solids and surfaces. , 2007, Physical review letters.

[84]  Yu Xie,et al.  Global minimization of gold clusters by combining neural network potentials and the basin-hopping method. , 2015, Nanoscale.