Convergence problems with EOM-CC3

[mdavis]

Hello,

I am trying to use Psi4’s EOM code to generate some EOM-CC3 reference transition energies. It is my understanding that only one root can be sought at a time for CC3 (please correct me if I am wrong), with the eigensolution being the nth in a given irrep, where n is specified in roots_per_irrep.

This is all very well for the first couple of roots, but with higher roots the computation seems to get stuck with a significant residual norm. This behavior is consistent across basis sets an at least a couple of molecules. Here is a MWE for eom-cc3/cc-pvdz water:

memory 4 GB

molecule {
0 1
O
H 1 1.1
H 1 1.1 2 104.0
}

set basis cc-pvdz
set roots_per_irrep [4,0,0,0]
set cceom {
    cc3_follow_root true #false does not seem to change behavior
}
energy('eom-cc3')

I am running a relatively recent build (1.4a2.dev138, whatever that means) although that shouldn’t matter as the EOM code looks like it hasn’t been touched in some time.

Is there any way to resolve this issue? Thanks for any insight you can offer!

[mdavis]

Solved right after I asked. Increase vecs_cc3 to e.g. 15 or 20 and convergence is slow but does occur.

[jmisiewicz]

For future reference, we have a feature for you to mark solutions now, the checkbox. I went ahead and did it for you.