Open metma99 opened 1 month ago
Sure. It's not super high on the list. At the moment, the biggest problem is that GFN2 gives a very small barrier to proton transfer reactions (protonation / deprotonation) vs. DFT or reality. We note that a few times in the paper - the site prediction model and the low influence of the $\Delta E$ protonation as a molecular property on the QupKake model.
So if I were looking for "what parts of the QM properties would I want to improve," the protonation energies is far above any inaccuracy in the solvation corrections. Now if you think fixing CPCM-X / SMD improves the protonation energies .. then let's talk. 😄
(Both probably need improvements - I'm just guessing that fixing solvation probably won't improve the model measurably without better protonation energies.)
Hello: Outstanding work. Congratulations on your publications!
If you have not seen, I would like to direct your attention to the https://github.com/grimme-lab/xtb/releases/tag/v6.7.0
It offers the cpcm-x implicit solvent calculations with nice SMD corrections https://pubs.acs.org/doi/10.1021/acs.jpca.3c04382
In my hands cpcm-x really performs excellent.
Any plans in including this model?
If not, I might give a a shot?
Best regards, Markus