Skip to content

OrbNetMethod

The OrbNetMethod object specifies the settings for using OrbNet [1] [2] as the method for a calculation. OrbNet is a family of machine learning methods, suitable for the calculation of energies, structures, and non-covalent interactions in large molecular systems. OrbNet provides the accuracy of dispersion-corrected range-separated hybrid DFT with a roughly 1000-fold reduction in computational cost. OrbNet is parameterized for neutral, closed-shell species containing C, H, B, O, N, F, P, S, Cl, Si, Br, I. Chemical coverage of OrbNet will increase in future releases.

Fields

All of the fields in MethodBase and the following:

model

Specify the OrbNet model major version

  • Type: OrbNetMajorVersionEnum
  • Default: denali (since Sierra v1.4.1)
  • Values:
    • fuji
    • denali (since Sierra v1.4.1)
version

Specify the OrbNet model minor version

  • Type: str
  • Default: 'latest'

Note

When using OrbNet Denali for geometry optimization, the default thresholds are different than for other methods, in order to increase robustness.