[gmx-users] computation of energy for energygrps + problem using g_enemat

[Justin Lemkul]

Please make sure to keep the discussion on the mailing list.

On 9/1/15 9:35 AM, ANTHONY C MANSON wrote:
> Thank You Justin:
>
> Can you tell me where in the code the actual energies for these groups are
> calculated.

See gmxlib/nonbonded/nonbonded.c, particularly do_nonbonded(); there are 
probably other locations in the code where bookkeeping and things are done but 
this is what seems to be accumulating all the pairwise interactions (I don't 
know the guts of this part of the code well).

>   I noted that some energy group calculations are done in nb_kernel_allvsall.c
> using a unit dielectric constant.
> This function is called indirectly by do_force_lowlevel(...) for each frame.
> I do not, however, think this is the correct code because the
> energies should be calculated every nstenergy frames.
>

nstenergy is for output being written to the .edr.

> It is clear that these numbers seem high.  I am trying to understand WHY they
> are so high.
> Is it the impact of the cutoff? especially in the Protein-SOL Coul-SR
> calculation.  Can the
> cutoff be reduced say in a rerun type calculation ( I have reduced the cutoff in
> a rerun and the
> numbers scale down dramatically) -or- Are there still factors that cause the
> calculation to give large results? (such as not factoring in the dielectric
> constant of ~80 typically
> ascribed to water calculations)

Cutoffs are of course going to impact the magnitude of the value, as you've 
noticed, because shorter cutoffs simply mean fewer interactions are calculated 
in the short-range list.  There is no need for dielectric values here; these are 
built into the parametrization of the water model (not all of them get the 
dielectric right).  GROMACS allows for scaling of the vacuum dielectric (via 
epsilon_r), but this should not be altered for standard simulations as all force 
fields are parametrized using epsilon_r = 1 (you know, real physics :)

-Justin


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

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

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