[gmx-users] Very large fluctuations in dg/dl

[Lars Schaefer]

Stéphane Téletchéa wrote:

> Patrick Fuchs a écrit :
>
>> Hi Gromacs users,
>> I have a few questions related to solvation free energy calculation via
>> thermodynamic integration.
>
>
>> #ifdef FLEX_SPC
>> #include "flexspc.itp"
>> #else
>> #include "spc.itp"
>> #endif
>
>
>> coulombtype            = generalized-reaction-field
>> rcoulomb               = 1.4
>> rvdw                   = 1.4
>> epsilon_rf             = 54.0
>
>
> I cannot state for other parameters, but i'm surprised for the 
> (relatively?) low value for epsilon_rf used here.
>
> For instance in Geerke and van Gunsteren (ChempPhysChem 2006, 7, 
> 674-678), the value used for reaction-field (not generalized) is 61 
> for SPC water (referring to Heinz, van Gunsteren & Hunenberger, J Chem 
> Phys 2001, 115, p1125-1136).
>
> Is this value a special case for 'generalized-reaction-field' ?
>
> Thanks in advance for clarifications.
>
> Stéphane
>

Stéphane ,
as stated earlier in this list by Berk:

For epsilon_rf=54 the prefactor for the RF term is 0.486,
for epsilon_rf=80 it is 0.491. This difference is negligible,
especially compared to the effect that the choice of cut-off
distance will have.
If you want to avoid the discussion, use an epsilon of infinity
(prefactor 0.5), this also gives the least cut-off effects,
as the potential will be exactly zero at the cut-off.
Berk.