Changing Orca SCF settings for Part1 amd Part2 computations

[ankur56]

Hello,

I am using CENSO to screen metal complexes with different spin multiplicities. In many cases, the SCF convergence is painfully slow, potentially because of the default Orca setting of using the TRAH method quite early during the SCF procedure. Therefore, I would like to add some keywords in the CENSO-generated Orca input files to make the SCF convergence faster. However, I couldn't find any keyword in .censorc that would be useful for my needs. Is there any way I could modify Part-1 and Part-2 generated input files through CENSO?

[fabothch]

Hi,

since CENSO v.1.2.0 a modification of input files for all ORCA calculation is possible:

• A new file censo_ORCA_editable.dat is introduced to the directory ~/.censo_assets/ and will be created upon the next start of CENSO. It contains ORCA keywords which will be added to every ORCA-input generated by CENSO, therefore it can only contain global information independent of specific calculation types. There exists no check for misspelling so typos will lead to crashes. The intention is to enable the user to specify additional input like for example "NOTRAH". This feature is considered to be an expert-user-option. The file looks like this:

  ! smallprint printgap noloewdin
  ! NOSOSCF
  %MaxCore 8000
  %output
     print[P_BondOrder_M] 1
     print[P_Mayer] 1
     print[P_basis] 2
  end

  You can add the NOTRAH keyword if this is better suiting your use case.

All the best,

Fabian

[ankur56]

Hi Fabian,

This is precisely what I wanted. Thank you for your help. There is another small related issue I was wondering if it could be resolved.

I would like to save forces ( and potentially perform wavefunction stability analysis) for all the Part-1 geometries. I know that I could add the engrad keyword to the censo_ORCA_editable.dat file to generate forces for Part-1 structures. However, that would conflict with the engrad keyword added by CENSO in Part 2. Therefore, I was wondering if we could have separate Orca input files for Part-1 and Part-2.

[fabothch]

Glad that this user-editable option to modify the ORCA inputs fixes your convergence issues.

Concerning your second request: this is not implemented and I believe an additional option makes the already complicated setup more user-unfriendly. You could however just run part1 with the keyword engrad and then stop CENSO, remove the keyword from censo_ORCA_editable.dat and restart into part2.

[ankur56]

Hi Fabian,

Thank you for your response. I have thought about using your method to run part1 and part2 computations separately. However, in such a workflow, I won't be able to run part2 computations for some other molecule while I am running part1 computations. Ideally, I would like to have a separate censo_ORCA_editable.dat file for different projects (and also for different parts within the same project). So, censo_ORCA_editable.dat should probably be a local file rather than a global file, primarily because SCF settings are usually system and job dependent.

I personally find CENSO to be quite user-friendly. So, I don't think adding more expert user options would make it complicated. If you ever release a future version, please consider adding such features.