[gmx-users] Re: Potential energy positive. Can be?

Mark Abraham Mark.Abraham at anu.edu.au
Thu Aug 16 17:16:36 CEST 2007

Eudes Fileti wrote:
> Dear Mark, thank you for your reply.
> For equilibration of the system, firstly I generated a ordinated lattice 
> of 10x10x10 molecules.
> After, I carried out successive minimization runs (with STEEP and 
> L-BFGS) up to the system
> to reach the convergence for these methods and in the sequence I 
> performed simulations
> totalizing 300ps, in ensemble NVT. (Could you suggest something more 
> appropriate?)

Check out 
http://wiki.gromacs.org/index.php/Steps_to_Perform_a_Simulation, in 
particular, how did you choose your volume? (and thus density and 
pressure?) Is this reasonable?

> I believe that if I start from a configuration with molecules randomly 
> distributed I would get better
> results than a ordinated configuration.

As soon as you're generating random velocities, it won't matter at all 
whether the starting configuration was ordered or not. After 10ps or 
less you won't be able to see any ordering.

> I have created the potential for the toluene from of the potential of 
> the benzene and
> and it seemed the sufficiently reasonable, compared to others potentials 
> from literature.

"Looking reasonable" is not a sufficient criterion if you want to 
publish these results. You're simulating a complex balance of different 
interactions and our intuition is poorly-trained to see what is good or 
not in this context. The only measure of quality is the ability to 
reproduce (experimental) observables. Hence my point about using an 
existing forcefield in exactly the same way, and not making something

The fact that you are getting positive energies suggests that at least 
on of the above issues is a significant problem for your system.

 > And about this, would you a topology or a equilibrated box for the 

I don't understand you.


More information about the gromacs.org_gmx-users mailing list