What atom types should we use in wall-atomtype for surface tenstion calculation?

GROMACS version: 2021.2
GROMACS modification: No
Dear all, I’m trying to calculate the surface tension of an organic solvent. I couldn’t find a thorough explanation on the internet of how to calculate the surface tension of a solvent but from what I found we should use an NVT ensemble with two walls in one direction (z for my case). Then we can use gmx energy to calculate the surface tension (#Surf*SurfTen) by dividing the obtained value by two (because we have two surfaces).

My question is in the “wall-atomtype” of the mdp file what atom type should we use and does it really affect the results or not?

I’ve attached the mdp file that I tried to use for your reference.

Any suggestion is appreciated,

Ehsannvt_st.mdp (2.6 KB)


you don’t necessarily need walls to measure the surface tension. The important thing is that your solvent slab should not interact notably with itself across a periodic boundary along z. You can estimate this effect by repeating the surface tension measurement for a few different box sizes along z: as you increase the box size the value should converge.

If you do want to use walls I’d recommend to add an atomtype that only has a repulsive interaction with other atoms to your topology, and use that. Otherwise individual solvent molecules which detach from your solvent might get stuck at the walls, which could impact your measurement. You can add an atomtype to your topology (for example by copying, say, the Argon or Carbon entry and changing its name) and set the LJ-sigma parameter to a negative value. A negative value means that only the repulsive 1/r^12 term from the potential is used, with the absolute value of the input as the sigma.


1 Like

Thanks for your reply.

Appearantly, the value of LJ parameters doesn’t change much the surface tension results, but for consistency in the results the negative sigma parameter is a good approach.

Why would you want to use a wall at all? This can affect the pressure calculation and I see no advantage.