A deep neural network for molecular wave functions in quasi-atomic minimal basis representation.

The emergence of machine learning methods in quantum chemistry provides new methods to revisit an old problem: Can the predictive accuracy of electronic structure calculations be decoupled from their numerical bottlenecks? Previous attempts to answer this question have, among other methods, given rise to semi-empirical quantum chemistry in minimal basis representation. We present an adaptation of the recently proposed SchNet for Orbitals (SchNOrb) deep convolutional neural network model [K. T. Schütt et al., Nat. Commun. 10, 5024 (2019)] for electronic wave functions in an optimized quasi-atomic minimal basis representation. For five organic molecules ranging from 5 to 13 heavy atoms, the model accurately predicts molecular orbital energies and wave functions and provides access to derived properties for chemical bonding analysis. Particularly for larger molecules, the model outperforms the original atomic-orbital-based SchNOrb method in terms of accuracy and scaling. We conclude by discussing the future potential of this approach in quantum chemical workflows.

[1]  K. Ho,et al.  Transferability of the slater-koster tight-binding scheme from an environment-dependent minimal-basis perspective , 2005 .

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

[3]  K-R Müller,et al.  SchNet - A deep learning architecture for molecules and materials. , 2017, The Journal of chemical physics.

[4]  H. Eschrig,et al.  An optimized LCAO version for band structure calculations application to copper , 1978 .

[5]  Frank Neese,et al.  The ORCA program system , 2012 .

[6]  J. Smutná,et al.  Systematic development of ab initio tight-binding models for hexagonal metals , 2020 .

[7]  Anders S. Christensen,et al.  Operators in quantum machine learning: Response properties in chemical space. , 2018, The Journal of chemical physics.

[8]  Abhinav Vishnu,et al.  Deep learning for computational chemistry , 2017, J. Comput. Chem..

[9]  Thomas F. Miller,et al.  A Universal Density Matrix Functional from Molecular Orbital-Based Machine Learning: Transferability across Organic Molecules , 2019, The Journal of chemical physics.

[10]  Eamonn F. Healy,et al.  Development and use of quantum mechanical molecular models. 76. AM1: a new general purpose quantum mechanical molecular model , 1985 .

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

[12]  Pekka Koskinen,et al.  Density-functional tight-binding for beginners , 2009, 0910.5861.

[13]  O. A. von Lilienfeld,et al.  Alchemical Predictions for Computational Catalysis: Potential and Limitations. , 2017, The journal of physical chemistry letters.

[14]  Hua Guo,et al.  High-Dimensional Atomistic Neural Network Potentials for Molecule-Surface Interactions: HCl Scattering from Au(111). , 2017, The journal of physical chemistry letters.

[15]  Bin Jiang,et al.  Embedded Atom Neural Network Potentials: Efficient and Accurate Machine Learning with a Physically Inspired Representation. , 2019, The journal of physical chemistry letters.

[16]  J. Behler.  Atom-centered symmetry functions for constructing high-dimensional neural network potentials. , 2011, The Journal of chemical physics.

[17]  J. Behler,et al.  Representing molecule-surface interactions with symmetry-adapted neural networks. , 2007, The Journal of chemical physics.

[18]  Thomas F. Miller,et al.  Transferability in Machine Learning for Electronic Structure via the Molecular Orbital Basis. , 2018, Journal of chemical theory and computation.

[19]  Alexandre Tkatchenko,et al.  Quantum-chemical insights from deep tensor neural networks , 2016, Nature Communications.

[20]  Pavlo O. Dral,et al.  Quantum Chemistry in the Age of Machine Learning. , 2020, The journal of physical chemistry letters.

[21]  N. Marzari,et al.  Maximally-localized Wannier Functions: Theory and Applications , 2011, 1112.5411.

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

[23]  T. Carrington,et al.  An improved neural network method for solving the Schrödinger equation1 , 2009 .

[24]  C. Wang,et al.  Molecule intrinsic minimal basis sets. II. Bonding analyses for Si4H6 and Si2 to Si10. , 2004, The Journal of chemical physics.

[25]  Andrew A Peterson,et al.  Acceleration of saddle-point searches with machine learning. , 2016, The Journal of chemical physics.

[26]  Burke,et al.  Generalized Gradient Approximation Made Simple. , 1996, Physical review letters.

[27]  Sándor Suhai,et al.  Self-consistent-charge density-functional tight-binding method for simulations of complex materials properties , 1998 .

[28]  Daniel W. Davies,et al.  Machine learning for molecular and materials science , 2018, Nature.

[29]  Andrea Grisafi,et al.  Symmetry-Adapted Machine Learning for Tensorial Properties of Atomistic Systems. , 2017, Physical review letters.

[30]  John R. Kitchin,et al.  Machine learning in catalysis , 2018, Nature Catalysis.

[31]  Geoffrey J. Gordon,et al.  A Density Functional Tight Binding Layer for Deep Learning of Chemical Hamiltonians. , 2018, Journal of chemical theory and computation.

[32]  K-R Müller,et al.  SchNetPack: A Deep Learning Toolbox For Atomistic Systems. , 2018, Journal of chemical theory and computation.

[33]  Li Li,et al.  Bypassing the Kohn-Sham equations with machine learning , 2016, Nature Communications.

[34]  Naftali Tishby,et al.  Machine learning and the physical sciences , 2019, Reviews of Modern Physics.

[35]  C Z Wang,et al.  Molecule intrinsic minimal basis sets. I. Exact resolution of ab initio optimized molecular orbitals in terms of deformed atomic minimal-basis orbitals. , 2004, The Journal of chemical physics.

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

[37]  Michihiko Sugawara,et al.  Numerical solution of the Schrödinger equation by neural network and genetic algorithm , 2001 .

[38]  Kristof T. Schütt,et al.  Unifying machine learning and quantum chemistry with a deep neural network for molecular wavefunctions , 2019, Nature Communications.

[39]  Marco Buongiorno Nardelli,et al.  Accurate tight-binding Hamiltonian matrices from ab initio calculations: Minimal basis sets , 2015, 1509.02558.

[40]  R. Bowen,et al.  Machine-learned approximations to Density Functional Theory Hamiltonians , 2016, Scientific Reports.

[41]  F. Weigend,et al.  Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for H to Rn: Design and assessment of accuracy. , 2005, Physical chemistry chemical physics : PCCP.

[42]  Walter Thiel,et al.  Machine Learning of Parameters for Accurate Semiempirical Quantum Chemical Calculations , 2015, Journal of chemical theory and computation.