[gmx-developers] solvation free energy with couple-intramol set to 'no'

[Vytautas Gapsys]

Hey,

I have encountered an unexpected outcome when calculating solvation free 
energies and using the option: couple-intramol = no

The observation is that the forces, overall electrostatic energy and the 
final free energy depend on the rcoulomb cutoff, in spite of using PME.

I speculate that the following might be happening: for this simulation, 
intramolecular interactions are replaced by explicit pair interactions, 
however, this apparently is done for both states - coupled and 
decoupled. This way, in the coupled state those intramolecular 
electrostatic interactions that are outside of the rcoulomb cutoff are 
counted twice (when using PME): once they are counted explicitly 
ignoring the rcoulomb cutoff, second time they appear in the reciprocal 
part of PME.

This can be verified by extending rcoulomb up to the point where all the 
intramolecular interactions are within the rcoulomb cutoff. When 
extending the cutoff further from that point the overall electrostatic 
energy becomes rcoulomb independent.

Such a dependence on rcoulomb is problematic, as tunepme can change 
rcoulomb value during runtime.

Let me know if I am overlooking something and maybe this behaviour can 
be circumvented.

Best Regards,
Vytas

-- 
Vytautas Gapsys
Project Leader at Max Planck Institute for Biophysical Chemistry
Computational biomolecular dynamics group
Am Fassberg 11
D-37077 Goettingen, Germany
Email: vgapsys[at]gwdg.de
Phone: +49 551 201-2310