Skip to content

Conversation

@RDReynolds
Copy link

No description provided.

Basically copied the MS-CASPT2 scripts with changes to headers and namespace
These are actually copies of an outdated version of RelCASPT2.  I ended up not using them and just updating the
RelCASPT2 generation files with the little bit that was necessary for multistate energies.

The analogous class "MSCASPT2" is I guess only needed for multistate gradients...
…ti-state RelCASPT2.

They seem to have no effect on the final energies...  This removes them.
Sometimes when building the Fock matrix over with closed or virtual indices, for certain terms no valid contraction would be found.
In these cases, oldspin is a null pointer, and we can have segmentation faults when trying to generate code.
Now it will just skip that contribution in such cases.

This bug does not show up when using only general or active indices, which is apparently why it was silent in previous uses of SMITH3.
With this change, I can get the same results for MS-RelCASPT2 using either the full (g, g) Fock matrix or building each block separately.
Emerges when we substitute   T H|0> --> E(0) T |0>
because we only have the full Hamiltonian over active orbitals.
Note that we no longer need the cancellation of diagrams.  I had mis-ordered my operators in yesterday's commit...
# Conflicts:
#	prep/generate_relcaspt2.cc
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant