Unifying machine learning and quantum chemistry with a deep neural network for molecular wavefunctions

Machine learning (ML) methods reach ever deeper into quantum chemistry and materials simulation, delivering predictive models of interatomic potential energy surfaces1,2,3,4,5,6, molecular forces7,8, electron densities9, density functionals10, and molecular response properties such as polarisabilities11, and infrared spectra12. Large data sets of molecular properties calculated from quantum chemistry or measured from experiment are equally being used to construct predictive models to explore the vast chemical compound space13,14,15,16,17 to find new sustainable catalyst materials18, and to design new synthetic pathways19. Recent research has explored the potential role of machine learning in constructing approximate quantum chemical methods20, as well as predicting MP2 and coupled cluster energies from Hartree–Fock orbitals21,22. There have also been approaches that use neural networks as a basis representation of the wavefunction23,24,25. Most existing ML models have in common that they learn from quantum chemistry to describe molecular properties as scalar, vector, or tensor fields26,27.