Using a localized basis

[JuleH21]

Dear autoCAS-Team,

in your autoCAS paper you suggested to use localized orbitals to define the active space. However, in my case, when requesting localized orbitals, Openmolcas seems not to be able to calculate the localized orbitals, the output gives a non-zero return code. Also no molden file of the localized orbitals is generated, which should be done be default by Openmolcas. AutoCAS proceeds anyways, I think that the HF orbitals are taken for the DMRG-CI instead? I attach the autocas-Input and the Output for the HF calculation + orbital localization for a small example, N2. I tried also many other different molecules and basis sets and also ran the calculation on a cluster, but the localization never succeeded. Can you give a hint?

Thank you in advance!

n2.txt
autocas.log

[weymutht]

Hi, thanks for giving autoCAS a try! From your output files, I can see that the localization procedure returns a non-zero exit code, so OpenMolcas has a problem here. However, from the outputs you provided, I'm not sure what exactly the error is. Normally, OpenMolcas writes a dedicated error file (usually ending with ".err") which might help you further.

[JuleH21]

Thank you for your reply!
The error file from Openmolcas does not really help, I attached it here.
In addition to the issue that the localization crashes the question would be if autocas then proceeds with the HF orbitals instead of the localized orbitals? Because apparently the program still continues but the program does not tell that the localization did not succeed. So I am not really sure which basis is used for the further procedure.
autocas_err.txt

[mmoerchen]

Hi, autoCAS probably continues with HF orbitals, because OpenMolcas creates the .scf.h5 file already after the HF-optimization.
However, I can not reproduce the crash in the localization procedure.
Which OpenMolcas version do you use?

[JuleH21]

Hi, we use version V24.02 on a cluster and 23.10 locally, it crashes in both cases