Issue with energy differences in charged unit cell

[alexandra_mcisaac]

Hello,

I am doing some benchmarking calculations between VASP and QChem for the attached two molecules, which are reactant and product in a reaction involving a transition metal complex and organic ligands. This is a molecular (non-periodic) system which I am running in VASP using a large vacuum between unit cells (about 19 A between the closest points) and both the reactant and product have a +1 charge.

I followed the directions in the VASP manual and on this forum about running charged calculations (setting IDIPOL=4, NELECT, and using a large box), but the energy difference for the reaction (e.g. E(TMB29_T1)-E(TMB29_C2)) is off by about 120 kcal/mol compared to QChem. I have done plane wave and gaussian basis set extrapolations and the error in both cases is low (~1 kcal/mol) so I don't think this is an issue, and the other (uncharged) molecules I have tested have very low difference in reaction energy between VASP and QChem. I am having similar issues with another, similar system that also has a +1 charge, and am wondering if there is anything else I should be doing to account for the charge, or if the error is just due to inherent issues with charged periodic simulation cells?

Thanks in advance!

[jonathan_lahnsteiner2]

Dear alexandra_mcisaac,

I was checking your calculations. I think the big deviations
in your energy differences arise due to spurious interactions
between the charged molecules due to the periodic boundary
conditions vasp is using.
I would recommend to switch on the LMONO tag.
wiki/index.php/LMONO
This will add monopole corrections to your computation and should improve
the computed energies.

All the best Jonathan