GROMACS version: 2021
GROMACS modification: No
I am trying to calculate the surface tension between a ‘surrogate’ solid substrate I obtain by restraining atomic positions and either a fluid or a gas phase. The results I get are a bit strange, in the sense that I obtain a negative estimate for the surface tension at the solid/something interface, in both cases.
By definition, surface tension is computed as the difference between normal and parallel components (w.r.t. the interface) of the pressure tensor; my understanding is that the pressure tensor is in turn computed from the virial, which may be ill-defined when position restrains are present.
When doing NPT one should select a proper
refcoord-scaling, but what about NVT calculation? Is the virial always well-defined? Or maybe never well defined, and so one should never do a NVT simulation to estimate surface tension when position restreaints are present? Is it possible to extract the pos. res. contribution to surface tension alone?