ExcitedStatesQM8

The data was used in our first application of machine learning modeling of excited state properties across chemical space.

Electronic spectra from TDDFT and machine learning in chemical space
Raghunathan Ramakrishnan, Mia Hartmann, Enrico Tapavicza, O. Anatole von Lilienfeld
Journal of Chemical Physics, 143 (2015) 084111 (1-8).
https://doi.org/10.1063/1.4928757

The dataset comprises

• Structures (Cartesian coordinates in Angstroms) of 21786 (22k) molecules relaxed at the B3LYP/6-31G(2df,p) level. This method was used to generate structures of the QM9 dataset (see Scientific Data, 1 (2014) 140022).

• Valence electronic excitation energies (in hartree) and oscillator strengths (in atomic unit, length representation) computed at the levels

  • RI-CC2/def2TZVP
  • LR-TDDFT(PBE0/def2SVP)
  • LR-TDDFT(PBE0/def2TZVP)
  • LR-TDDFT(CAMB3LYP/def2TZVP)

  How to download the data?

  git clone git@github.com:raghurama123/ExcitedStatesQM8.git

  cd ExcitedStatesQM8

  cd 22k_electronic_spectra_TDDFT_CC2

The file

• gdb8_22k_elec_spec.txt contains electronic excitation energies (in hartree) and oscillator strengths (in atomic unit, length representation)
• XYZ_B3LYP_631G2dfp.xyz contains equilibrium geometries relaxed at the B3LYP/6-31G(2df,p) level.

Data-mining platform

You can perform data-analysis on this dataset at the MolDis platform https://moldis.tifrh.res.in/db/dbqm8ex

References

ExcitedStatesQM8 dataset

Electronic spectra from TDDFT and machine learning in chemical space
Raghunathan Ramakrishnan, Mia Hartmann, Enrico Tapavicza, O. Anatole von Lilienfeld
Journal of Chemical Physics, 143 (2015) 084111 (1-8).
https://doi.org/10.1063/1.4928757

QM9 dataset

Quantum chemistry structures and properties of 134 kilo molecules
Raghunathan Ramakrishnan, Pavlo O. Dral, Matthias Rupp, O. Anatole von Lilienfeld
Scientific Data 1, 140022 (2014).
https://doi.org/10.1038/sdata.2014.22