RE: Scaling solvent-solute intermolecular interactions

From: Michael Bellucci (
Date: Tue May 13 2014 - 15:09:06 CDT

I see...This is a good point, but then this raises another question. Perhaps, I am confused, and maybe you or someone else could provide clarity. Say for example, one wishes to compute the free energy of solvation of a solute A in a solvent B with thermodynamic integration. The mixing Hamiltonian can be written...

H_lambda=(1-lambda)H_1+lambda H_2

where lambda is a scale factor. Let


Then the mixing Hamiltonian can be written as...


which is just a standard Hamiltonian with a simple scale factor (1-lambda) on the solute-solvent intermolecular interaction. If the long-range coulomb electrostatics of solute-solvent interactions cannot be scaled since it is a many-bodied term as you suggest, then how exactly does the thermodynamic integration code in NAMD handle the solute-solvent scale factor? This kind of scaling is similar to what I would like to achieve. Thank you in advance!


From: Axel Kohlmeyer []
Sent: Tuesday, May 13, 2014 2:56 PM
To: Michael Bellucci
Subject: Re: namd-l: Scaling solvent-solute intermolecular interactions

On Tue, May 13, 2014 at 2:34 PM, Michael Bellucci <<>> wrote:

Dear NAMD users,

I would like to know if there is a way to scale down solvent-solute intermolecular interactions while leaving the solute-solute and solvent-solvent intermolecular interactions the same? The closest thing I could find in the NAMD manual was the nonbonded scaling keyword, but this would scale down all intermolecular interactions. I also saw that NAMD is now capable of using tabulated potentials for specific pairwise interactions, but this seems to only be applicable to the VDW intermolecular interaction and not the coulomb interaction. Is there anyway to scale down solvent-solute intermolecular interactions without modifying the source code? Any help or suggestions would be appreciated!

​how do you propose to handle long-range electrostatics for this? ​unlike the short range coulomb term, this is not pairwise additive but rather a manybody term and thus you cannot partition the interactions in the way you want. basically, the amount of scaling would depend on the real space cutoff, while normally it can be chosen rather freely to find the optimal partitioning between the real space and reciprocal space coulomb force calculation.

so you'd be forced to use cutoff-only interactions and then you could just set all charges to zero and use tabulation. or use a different MD code that has more flexibility (and typically correspondingly less performance).



Dr. Axel Kohlmeyer<>
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:24 CST