[gmx-developers] modifing nonbonded interactions

[Mark Abraham]

Nazish Hoda wrote:
> This is what I am thinking. I know it is an approximation, but I don't know
> how bad would this be.
> Let me know what you people think.
> In PME,
> E_coul = E_coul^SR +  E_coul^LR. (and the same holds for force) 
> 
> My cut off for dielectric permittivity, which I call r_1, is very small, and
> would 
> only affect E_coul^SR (F_coul^SR). So what I think is that just changing
> E_coul^SR (F_coul^SR) 
> might work. 

Sure, subject to the foregoing force discontinuity issues, but what 
purpose might this serve? IIRC the purpose of a distance-dependent 
dielectric is to model heuristically either an effective dielectric 
constant, or one that allows the Coulomb sum to converge faster, or 
both. Since under PME the convergence of the short-range sum is already 
rapid, and common water models like SPC and TIPxP are parameterized for 
the use of a constant dielectric, you'd have many stones to roll up the 
hill to possibly get an inconclusive result.

Mark