detci_sym in src/mcscf

[fumihikoaiga]

Dear NWChem developers and Forum members,

 

I am considering performing relatively large CI calculations in a memory‑intensive manner using NWChem’s src/mcscf.

Since src/lucia appears to store CI vectors in files on disk, I would prefer not to use it in this case.

Within src/mcscf, there are two implementations, detci and detci_sym. I understand that the default build uses detci. However, in src/detci/detciP.fh, the following parameters are defined:

parameter(detci_maxorb=14)

parameter(detci_maxelec_tot=18)

parameter(detci_maxelec=10)

 

whereas in src/detci_sym/detciP.fh, they are defined as:

parameter(detci_maxorb=32)

parameter(detci_maxelec_tot=20)

parameter(detci_maxelec=20)

 

Given these larger limits, it seems that larger calculations may be possible when using detci_sym. Therefore, I would like to use detci_sym if possible.

In this regard, I would appreciate your guidance on the following points:

 

(1)   Even when the code is built with detci, does specifying the irreducible representation of the target state in a CI calculation result in diagonalization of the Hamiltonian restricted to that symmetry subspace?

(2)   Why is the default build configured to use detci rather than detci_sym? Are there any known issues or limitations associated with using detci_sym?

(3)   If detci_sym can be used, I would like to build mcscf with detci_sym. In that case, would it be sufficient to modify src/mcscf/GNUmakefile as follows and then rebuild?

 

SUBDIRS = detci_sym

 

Thank you very much for your time and assistance.

 

Best regards,

Fumihiko Aiga