[gmx-users] Potential energy contributions and total pot ener (2)

[Mark Abraham]

pascal.baillod at epfl.ch wrote:
> Dear community,
> 
> Thanks, Mark! But I can't find any g_energy or log file terms specifically
> related to solvent bonded energy (bonds and angles). The log file bonded terms
> are for the total system, I guess, but they are very close to the ones I get
> when I request their "protein" counterpart from g_energy. My guess would be
> that, in my 103-residue protein (1142 at) system solvated by 14'076 water
> molecules (42'228 at), there should be a large difference between protein bonded
> and solvent bonded terms.

Did you constrain your solvent bonded degrees of freedom to be constant?

> Perhaps water bonded terms are not counted at all, as the water molecule
> geometry dose not vary with the SPC model, during the simulation? 
> 
> By the way, is the assumption of my previous mail correct? The one on the total
> potential energy formulation as a function of g_energy potential energy
> decomposition terms:
> 
> Etot = a + 0.5*b + c (+ d), with
> 
> a) protein-protein (bonded and non-bonded)
> b) protein-solvent (interaction, non-bonded)
> c) solvent-solvent (interaction, non-bonded)
> (d) solvent bonded terms)
> 
> (Of course this only holds when using a non-PME method (i.e. generalised
> reaction field) for long range electrostatics!)

Like I said last time, all the non-total and non-kinetic energies must 
sum to give the potential energy. There's no reason to go introducing 
factors of one-half. There's no purpose in GROMACS reporting a 
double-counted protein-solvent interaction.

Mark