Closed obackhouse closed 3 years ago
This looks really great overall! I don't have much ADC experience. Do you have any recommendations for reviewers? If not we can ping the rest of the maintainers here.
I can ask either of the adcc
developers, who I believe are familiar with psi4? Up to you. I will get to your review comments this weekend.
@obackhouse That would be fantastic on the adcc
developers! Someone closer to these codes is always very welcome.
@maxscheurer Thanks for the review! Can you approve and merge?
I have no privileges to merge PRs 😄
Description
This PR adds reference implementations for the ADC(2) method, with a separate implementation for ionization potentials, electron affinities and excitation energies. The implementations are based on an RHF reference, but use spin-orbitals as the equations are nicer there. There is also a helper script with a function to perform the Davidson algorithm for an arbitrary matrix, or function defining a matrix-vector product. The latter is adapted from the EOM-CCSD implementation. These are not particularly efficient, but should be a simple enough showcase of the iterative eigensolver to keep the scaling down.
I have tested these scripts against PySCF for the IP and EA, and against adcc for the EE. The codes will not account for degeneracy in
n_states
nor separate singlets and triplet EEs as adcc does, but the energies themselves match.What are your new additions? Please provide a brief list.
Any questions for the community?