[gmx-users] Thermodynamic Integration

[Berk Hess]

>Thank you David for responding,
>
>On a less serious note, how should one choose the soft-core parameter
>sc_alpha or are the calculations even all that sensitive to this
>parameter?
>
>Thanks,
>David

sc_alpha has a large influence on LJ interactions, as these go as r^-12.
But the influence on Coulomb interactions is very small, as can be seen in 
the plot
on page 74 of the manual. So for the ion calculation the choice of sc_alpha 
is not critical.

The free-energy contribution of the Ewald sum is just the energy with the 
B-state charges
minus the energy with the A-state charges. When exclusions are present, the 
contribution
of the excluded pairs needs to be subtracted. This would require two extra 
Ewald calculations.
Unfortunately this is not implemented yet.

But for the simple case where an ion disappears, you can perform two times 
mdrun -rerun
for each run, one with the A-state and one with the B-state charges. This 
would give you
the PME contribution you are looking for.

One final note: you need to use stochastic dynamics for correct sampling, as 
in the final
state the ion does not interact with the rest of the system and would end up 
moving with
constant velocity (or worse, accelerate due to the Berendsen coupling).
With stochastic dynamics you are sure you get a canonical ensemble.

Berk.

_________________________________________________________________
STOP MORE SPAM with the new MSN 8 and get 2 months FREE* 
http://join.msn.com/?page=features/junkmail