[gmx-users] Question about decomposition of Lennard-Jones contributions in free energy calculations with Gromacs

Hannes Loeffler Hannes.Loeffler at stfc.ac.uk
Mon Nov 7 17:36:40 CET 2016

On Mon, 7 Nov 2016 16:16:33 +0100
David van der Spoel <spoel at xray.bmc.uu.se> wrote:

> On 07/11/16 15:55, Joel Jose Montalvo Acosta wrote:
> > Dear Dr. van der Spoel,
> >
> > My name is Joel Montalvo, I'm a PhD student computational chemistry
> > in the University of Strasbourg (France). I'm interested in
> > computing solvation free energy of solutes with apolar character as
> > fullerenes and graphene in organic solvents, using free energy
> > perturbation (FEP) and thermodynamic integration (TI) techniques as
> > implemented in gromacs.
> >
> > I write because I would like to decouple independently the
> > dispersion and repulsion components of the Lennard-Jones term in a
> > FEP/TI run such as it happens with electrostatics and Lennard-Jones
> > interactions. How can I modify the gromacs source code in order to
> > make this decouple independenly?. I image that in the mdp input
> > file, the options couple-lambda0 and couple-lambda1 could accept
> > new values as "disp" for dispersion and "rep" for repulsion. Also,
> > it is possible to have new lambda-vector sections, "disp-lambdas"
> > and "rep-lambdas" for controlling independently both contributions
> > during a FEP/TI run.
> >
> > I find interesting this decomposition because, first, the
> > convergence profiles for both contributions are different, so I
> > could use a different number of lambda for dispersion than for
> > repulsion, similar to what happens with electrostatics and
> > Lennard-Jones contributions. Second, it will be valuable to have
> > the dispersion/repulsion contribution ratio for solvation free
> > energies of different solutes.
> >
> > Thank you in advances for any comment or help
> >
> > Joel Montalvo Acosta
> >  
> This is a regular gromacs question so please continue the discussion
> on the gromacs user list.
> I think you can simply define A and B values for C6 and C12 in your 
> topology file and the do the TI in multiple steps. You will have to
> turn off the C6 first I guess, since the goes to -infinity. However,
> then I don't see how you could ever get converged values for the next
> step, since then you are left with a completely repulsive potential
> (it might work in NVT though).

Sounds like what Joel wants is WCA separation for free energy
calculations.  The answer may still be the same as almost 10 years ago:


> However before you even start you should ask yourself what it means. 
> Will you get any meaningful results? Decomposing in enthalpy and
> entropy is meaningful, but energy components is not.


More information about the gromacs.org_gmx-users mailing list