sSAPT Scaling term and F-SAPT

Hello,

I saw in Parker et al. that an improvement to the performance of SAPT0 could be had by a scaling function that was fit to the results of the benchmark. I’ve noticed that this energy is also printed out for the total interaction energy when using F-SAPT, but not for the total interaction energies of functional groups in the F-SAPT post-processing analysis.

Is it possible to obtain this empirical correction functional group by functional group? Is it not done in post because, though possible, this is somehow the wrong methodological choice, or unreliable?

Thanks in advance for straightening me out.

Regards,
Louis

There is a keyword option SSAPT0_SCALE that will drop the F-SAPT analysis with sSAPT0 exchange scaling into the directory s-fsapt/ in the current working directory. You can change the output directory with the keyword option FISAPT_FSSAPT_FILEPATH.

I do not know whether it is appropriate to use the sSAPT0 scaling with the F-SAPT decomposition. The scaling factor was chosen to match reference total interaction energies over a particular dataset, so the scaling factor might be inappropriate for functional group analysis or may even vary by functional group. Perhaps the developers could address this?

Thanks. That is the keyword I was looking for. I don’t think that the F-SAPT paper addresses whether this scaling would produce consistent results for fragmentation schemes. At least it does not in my reading of it. Hopefully I can get some additional input from the developers…

@ccavender is exactly right in their assessment. The scaling factor has only been optimized to reproduce total interaction quantities, and we haven’t done a detailed study to validate exchange scaling on a fragment–fragment interaction basis, so it’s not currently a reliable correction to apply to individual fragments.

Quite honestly, while the total F-SAPT interaction energy can and should be taken as SAPT0-quality gospel, the fragment–fragment total interaction energies and the individual fragment interaction terms should instead be taken as semi-quantitative indications of interaction motif rather than absolute truth.