some questions about the autocas test of benzene

[histhinking]

1. What kind of orbitals and orbital indices are used in active space search?

We know that benzene has a total of 42 electrons forming 21 pairs. The log files of both autocas jobs show that the indice of highest occupied orbital is 20, in line with the manual of ORCA 5.0.4 (p.118).

Initial active space CAS(e, o): (30, 30) Orbital indices: [6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35] Orbital occupations: [2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]

Since the orbital indice in ORCA starts from 0, I assume that here the orbital indice in autocas DMRG calculation also starts from 0. And the orbital used here is RHF orbital, right? Besides, the range of orbital indices in the entanglement diagram (see the attachment) is from 5 to 34, which further shifts 1 unit leftwards in the number line.

2. Why don't autocas with the command line and yml (most of options are taken from the template of online document) give the same results of active space?

(a) Using "python3 -m scine_autocas --xyz_file benzene.xyz --basis_set DEF2-TZVP --plot --interface Molcas"

Starting final calculation with optimized CAS. Final active space CAS(e, o): (6, 6) Final orbital indices: [16, 19, 20, 21, 22, 30] Final orbital occupation: [2, 2, 2, 0, 0, 0]

(b) using the yml file (see the attachment)

Starting final calculation with optimized CAS. Final active space CAS(e, o): (6, 6) Final orbital indices: [16, 19, 20, 21, 22, 29] Final orbital occupation: [2, 2, 2, 0, 0, 0]

The distribution of orbital indices in (b) is in agreement with the known active space of benzene calculated by ORCA.

3. In what circumstance should we use "large_cas"?

When I set "large_cas=true" and "max_orbitals=6", autocas did not find any active space, but said, "No active space method required here".

autocas_benzene_yml.log autocas_benzene_command_line.log entanglement_diagram.pdf benzene.yml.txt

Thanks!

[mmoerchen]

Hi,

1. The default are RHF orbitals, but localization is supported by the OpenMolcas interface. Unfortunately the indexing in the entanglement diagram has a bug in the current version, so that the indices are shifted.

2. From the log files it seems that you used the def2-svp for the "yaml" calculation and def2-tzvp for the "CLI" calculation.

3. "large_cas" should mainly be used if the valence space is too big for the default algorithm. You can check out this article https://doi.org/10.1002/jcc.25869 for more information on that topic.