openmm / spice-dataset

A collection of QM data for training potential functions
MIT License
155 stars 9 forks source link

Some calculations may converge to unstable wavefunction #110

Open tamaswells opened 2 months ago

tamaswells commented 2 months ago

We checked those confomations with large errors, and found that some of them may be due to unstable wavefunction solution. Using this molecule as an example image We did a normal calculation with Psi4, the final total energy is -718.5454888702812468 Hartree. (in the directory abnormal/sp_psi4) The we do another calcultion with Gaussian with the aim of searching for stable wavefunction, the result shows that the wavefunction may be unstable: The wavefunction has an RHF -> UHF instability., then the Gaussian starts to optimize the wavefunction. Finally, it has a stable wavefunction solution. (in the directory abnormal/from_gaussian) After formatting wavefunction file chk, the MOKIT was used to convert the Gaussian wavefunction to Psi4 wavefunction. We did another calculation using these Psi4 wavefunction which has basis set definition and orbital coefficients. The final energy is -718.5651033410908894 Hartree. (in the directory abnormal/from_gaussian/psi4) This solution should be a proper one. test_orbital_transfer.zip