[gmx-users] affect of water removal on subsequent energy calculations

Justin Lemkul jalemkul at vt.edu
Wed Apr 11 21:33:21 CEST 2018

On 4/11/18 3:09 PM, Alex wrote:
> Mark, Justin:
> This is two against one, even though noone was questioning the 
> additivity of energy in forcefields with constant charges, etc.
> So, let's go back specifically to solvation. Consider a system with 
> two oppositely charged ions (1 and 2) in water of your choosing (group 
> 3). For extreme simplicity, the ions are actually restrained at a 
> distance R from each other, and again for simplicity we're only 
> interested in the Coulomb part of the potential energy. The trajectory 
> contains everything.
> In one instance, we calculate E_tot - E_13 - E_23 - E_33, in another 
> E_12 (as one would easily do for a trajectory that has no water 
> coordinates). Are those two numbers in agreement and, if so, what is 
> it? q1*q2/R or q1*q2/eps/R?

For an additive force field, the two numbers are in agreement, E_tot - 
E_13 - E_23 - E_33 = E_12, by definition. You can easily design a test 
case that will prove this.

The dielectric constant of the medium is not included in the 
calculation, at least not explicitly, but its effects are there in the 
calculation of the forces. GROMACS lets you mess with the relative 
dielectric via epsilon_r in the .mdp file, but if you do that, you break 
the force field. All modern force fields and the water models that go 
along with them assume they are being parametrized relative to vacuum 
permittivity, as explained by Erik in e.g.:


As a result of this convention, water models are parametrized to 
approximate the dielectric constant of water based on their charge 
distribution and hence the forces competing for interactions among 
charged particles to model the screening effect. This is different from 
an implicit or MM/PBSA-type calculation, of course, where setting a 
solvent dielectric is necessary and part of the calculation to compute 
terms in the energy function. In an additive MM force field, the effect 
is baked into the parametrization of the model.




Justin A. Lemkul, Ph.D.
Assistant Professor
Virginia Tech Department of Biochemistry

303 Engel Hall
340 West Campus Dr.
Blacksburg, VA 24061

jalemkul at vt.edu | (540) 231-3129


More information about the gromacs.org_gmx-users mailing list