[gmx-users] g_bar ... the right answer for the wrong reason?

[Justin A. Lemkul]

On 6/28/12 10:35 AM, Michael Brunsteiner wrote:
>
> Hi,
> i just performed a free energy TI calculation, to get the
> free energy of solvation of water in water (the chemical potential of water)
> i stuck closely to the templates given in the tutoral by justin lemkul.
> The final result i get with g_bar looks good, and the number is within the error-bars
> of both experiment and literature data - still i keep getting a warning ...
> i first turn off the charges and then LJ ... but for the LJ part (only) g_bar keeps saying:
>
> WARNING: Some of these results violate the Second Law of Thermodynamics:
>           This is can be the result of severe undersampling, or (more likely)
>           there is something wrong with the simulations.
>
> i can't think of anything that's wrong with my simulations, nor do i think
> this is undersampling (21 windows each 2 nano seconds)
> also the difference between my mdp and justin's are minor (as i believe, see below)
> why do i get the correct answer but still this warning keeps coming?
>
> at lambda=0 my LJ interactions are fully turned on, and for the DeltaG(0->0.05)
> i,e, the first row in the table below i get from g_bar:
>
>   lam_A  lam_B      DG   +/-     s_A   +/-     s_B   +/-   stdev   +/-
>   0.000  0.050   -2.90  0.03    2.15  0.03 9647088805553410.00 9570465417587940.00    3.30  0.05
>   0.050  0.100   -3.36  0.03    1.48  0.02    1.47  0.03    1.95  0.02
>   0.100  0.150   -0.49  0.00    0.03  0.00    0.03  0.00    0.22  0.00
>   0.150  0.200    0.12  0.00    0.00  0.00    0.00  0.00    0.05  0.00
> etc...
>
> there appears to be a problem with the phase space overlap, but then
> all other values, and the end-result, look ok ... can it be that there is an issue
> with the gmx implementations of soft core potentials and g_bar?
>

The presence of (essentially) infinite values would indicate buggy behavior. 
There have been many changes to the free energy code in advance of version 4.6, 
so perhaps this issue has been solved (silently) already, but if you want to 
file a bug report on redmine.gromacs.org, please do so.

-Justin

-- 
========================================

Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================