Re: VDW energy: positive or negative?

From: Kenno Vanommeslaeghe (
Date: Wed Mar 05 2014 - 17:21:12 CST

Since we're using CHARMM parameter files, your statement that "the choice
of positive values vs. negative values is just pedagogical" is not
entirely correct. As senior CHARMM developer Lennart Nilsson pointed out:

> The negative sign in the CHARMM parameter file is just a convention
> indicating that the LJ parameters are given as epsilon and rmin/2. In the
> parameter file there are three values for the LJ parameters, and if the
> second one is negative they have the following meanings: ignored,
> -epsilon, rmin/2.
> If the second value is positive (a form which is not used today):
> polarizability, effective number of outer shell electrons, rmin/2. See
> appendix I in J Comp Chem 4:187.

Of course, this doesn't invalidate anything else you wrote, though I
didn't think Panagiota actually believed the negative sign implied an
upside-down L-J potential, which would obviously be catastrophically
unstable. Surely, he or she realized that if that were the case, we
wouldn't be putting all-negative values in our parameter files. We're not
stupid. Stupid is as stupid does. (Hmm, I wonder if
is already taken.)

On 03/04/2014 09:04 PM, Jason Swails wrote:
> On Tue, Mar 4, 2014 at 7:24 PM, PANAGIOTA KYRIAKOU <
> <>> wrote:
> However when I checked the parameter files I use ( CHARMM 36), the
> epsilon values are negative, so the statement above is not true.
> The previous posts have given you useful information, I'll just point out
> the error in this statement. The sign of the epsilon parameters here is
> entirely irrelevant (for instance, in the Amber force fields the epsilons
> for each atom types are positive). The absolute value of the epsilon
> parameter is the depth of the attractive well (but don't forget the LJ
> potential has a repulsive 1/r^12 term that dominates at very low distances
> and an attractive 1/r^6 term that attracts at longer distances).
> The epsilons from each atom type are combined to give a pairwise
> interaction. The canonical combining rules (Lorentz-Berthelot are used by
> CHARMM, I believe) state that the combined epsilon parameters between two
> atom types is defined as sqrt(epsilon_i * epsilon_j). So as you can see,
> you get exactly the same combined LJ parameters whether all epsilons are
> positive or negative. The choice of positive values vs. negative values
> is just pedagogical.
> HTH,
> Jason

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:12 CST