Best MD Protocol for Phase Equilibria and Surface Tension (LJ Fluid Case)

45 views

Skip to first unread message

Ehsan Heidaryan

unread,

Mar 21, 2025, 2:59:16 PMMar 21

to MBX-users

Hi everyone,

I have a question regarding the best practices for determining phase densities, saturation pressure, and interfacial properties like surface tension using molecular dynamics (MD).

While I’ve successfully implemented and coded GEMC for VLE, I’m finding MD approaches for these properties to be less straightforward. The literature seems to suggest various protocols—for example, some use two consecutive NVT simulations, others follow NVT with NPT, and some apply anisotropic pressure coupling (e.g., restricting pressure in only two dimensions) after dual NVT steps.

In one case, using a simple Lennard-Jones fluid initialized in an FCC lattice, I found that two NVT runs of 2 million steps worked reasonably well. But when I added an NPT step afterward, the system collapsed into a single phase, filling the entire box—clearly not what I expected.

That’s why I’m reaching out. I’d really appreciate any advice or suggestions, and if anyone is willing to share sample input files (even just as references), that would be incredibly helpful.

Thanks in advance!

Best regards,
Ehsan

Ethan Bull-Vulpe

unread,

Mar 24, 2025, 5:49:17 PMMar 24

to MBX-users

Hi Ehsan,

We have studied the vapor-liquid coexistence with MB-pol in the past, and you can find that work here:
https://pubs.aip.org/aip/jcp/article/154/21/211103/595418
See the "Simulation and analysis details" subsection of the "Methods" section for information on how we set up the two-phase coexistence simulations.

The procedure we recommend is to first equilibrate the liquid (in NPT) and then expand the box along one direction to create the "vapor" area. Then re-equilibrate at the desired T (in NVT) before running production simulations (also in NVT).

Its important that you do the vapor-liquid simulations in NVT, since if you allow the volume to change as in NPT, the box size will collapse back down to the desired pressure. This is probably what you are seeing with the LJ fluid.

You can look at the above paper's data repository:
https://datacommons.princeton.edu/discovery/catalog/doi-10-34770-nfcx-rb66
which contains all the input files we used.