xnn.common.data.hub.ethanol_response#

Ethanol response-property dataset builder (the FieldSchNet reference data).

10,000 ethanol conformations, drawn at random from MD17 and recomputed at the PBE0/def2-TZVP level (ORCA 4.0.1.2) with energies, forces, dipole moments, polarizability tensors and nuclear shielding tensors. Released with the FieldSchNet paper and the training data of the PaiNN spectra experiment (8000 / 1000 / 1000 structures for training, validation and test).

References

M. Gastegger, K. T. Schuett, K.-R. Mueller, “Machine learning of solvent effects on molecular spectra and reactions”, Chem. Sci. 12, 11473 (2021). K. T. Schuett, O. T. Unke, M. Gastegger, “Equivariant message passing for the prediction of tensorial properties and molecular spectra”, ICML 2021. Data: http://www.quantum-machine.org/datasets/ (ethanol_vacuum.tgz).

Notes

  • Upstream units are atomic units throughout (Bohr, Hartree, Hartree/Bohr, e Bohr, Bohr^3); by default they are converted to Angstrom, eV, eV/Angstrom, e Angstrom and Angstrom^3 (units="eV"). units="au" keeps them.

  • The archive holds an ASE database, so reading it needs ase.

  • The published split is a random 8000 / 1000 / 1000 partition; the builder draws it from a seeded permutation (seed), so it is reproducible but not the authors’ own.

Classes

EthanolResponseBuilder()

Builder for the ethanol energies / forces / dipoles / polarizabilities set.
Show Source