Question about Dispersion Correction

Dear gmx-users,
I wan to know how to use the “DispCorr” in mdp file. Should I use the option “DispCorr=EngPres” in Energy minimization?
If I simulate a box filled with water molecules with periodic boundary conditions in NPT and the vdwtype is Cutoff, I should use the DispCorr = EngPres, Am I right? But under what circumstances do we have to give up this? For example, what about a big box with several water molecules?
Could anyone give me some advice? I am a new one in MD! Maybe the question is trivial ! Thank you in advance!

BW!
LM

Use dispersion correction if the force field was parametrized to use it. The nonbonded scheme should be the same for all steps in your preparation and performance of the production simulation.

Thank you for the kind help!@jalemkul And I have two more questions:

  1. The dispersion correction is computed with the assumption that the rdf g®=1 beyond the cutoff(gmx manual 2020.2),but if the box is not filled with the species(for example in energy minimization),the condition g®=1 may not hold on. under these circumstances, there is no need to use the dispersion correction. Am I right?
  2. I want to simulate several molecules in a very small box(eg:101010A) at initial preparation using PBC(NVT), but the usual cutoff in my force field is 12A , I think I have to reduce the rvdw in my simulation. Am I right?

BW!
LM

The box is still filled with something - the dispersion correction algorithm assumes you have bulk (homogeneous) solvent beyond the cutoff, not necessarily that there is only one species.

Don’t change cutoffs. Make your box larger. A cutoff is part of the assumption of the parametrized model. You can’t arbitrarily change it.

@jalemkul
"The box is still filled with something - the dispersion correction algorithm assumes you have bulk (homogeneous) solvent beyond the cutoff, not necessarily that there is only one species."

According to your instant reply, the gmx take the solvent into the consideration automatically. But if I want to simulate a polymer chain without solvent ,(In fact, my box is only one type molecule ) Will it lead to the unphysical effect in my simulation using the non-existing solvent in real?If the solvent doesn’t exist, will it be better turn off the dispersion correction?

Don’t change cutoffs. Make your box larger. A cutoff is part of the assumption of the parametrized model. You can’t arbitrarily change it.

Maybe I didn’t explain my issue clearly.I am so sorry about that! I saw a paper simulated a single polymer chain(polymerized by 4 monomers) in NVT to get its equilibrium conformation in long chain preparation, but box side length a obtained by density is too small to contain it(of course a < rvdw in force field). Under the conditions, can I make the box larger?

BW!
LM

Dispersion correction doesn’t require solvent. Truncation of LJ interactions leads to missing long-range (dispersion) interactions. Some force fields then add dispersion correction to the truncated short-range forces. If your force field is parametrized this way, you need to apply dispersion correction.

If a box vector is shorter than any of the cutoffs, the simulation violates the minimum image convention and will suffer from multiple counting of interactions and therefore spurious forces.

Great Thanks! @jalemkul

Could you recommend any method to check whether my forcefield is parametrized to use dispersion correction? I’m using amber99sb-ildn for protein-ligand complex simulation in dodecahedron box, tip3p water, 12A cutoff and I’m not sure how to set DispCorr parameter in my .mdp files.

Hi,

The best is to check in literature or to cross check with the force field authors

Here setting for amber99sb-ildn. Note the following mdp setting depends on the force field.
constraints = h-bonds ; bonds involving H are constrained
rcoulomb = 1.0 ; short-range electrostatic cutoff (in nm)
rvdw = 1.0 ; short-range van der Waals cutoff (in nm)
vdw-modifier = Potential-shift-Verlet ; Amber specific
DispCorr = EnerPres ; account for cut-off vdW scheme
coulombtype = PME ; Particle Mesh Ewald for long-range electrostatics
fourierspacing = 0.125 ; grid spacing for FFT

Thank you very much for the reply. Could you please explain the choice of fourierspacing and DispCorr values, are those force field specific parameters?
Kind regards,
Barbara

Hi,
I have defined like fourier-spacing/rcoulomb = 0.125. For constant accuracy one should keep fourier-spacing/rcoulomb constant. So a longer cut-off allows for a larger fourier-spacing.
For DispCorr you can cross check with the original literature on AMBER force field
Alessandra

Thank you very much for your help