[lammps-users] lammps

Dear LAMMPS Users,

I am working on ice and THF hydrate growth on the copper surface using lammps simulation.
My results show that the copper surface can not promote ice or hydrate nucleation and we have no formation on the sub-cooled surface.
but in my experimental section i have a cell which is sealed by copper plate and ice can form on the copper.
How can I explain the reason for the difference between experimental and simulation results?
best regards

hi farzane
I think in order to give similar results to experiments, the potential must be very accurate and appropriate. Check if the potentials are suitable to replicate this type of phenomenon. If not then you may want to look for more appropriate one. Even you may have to look for generating your own potential files.

There is a large amount of published literature on simulations of aqueous films on metal surfaces. There was research from classical MD and MC simulations on such systems even before I was a graduate student >25 years ago. It is good practice to read up on that and learn from the experience of other researchers before running your own.

It also is not as straightforward to compare a (macro scale) experiment to a (nano scale) simulation. Both time and length scales are significantly different and that has a significant impact when studying activated processes.

Even without knowing the details of your simulation setup and assuming you are using only features that are included in the LAMMPS distribution (and not external user packages), there are several things to consider:

  • what is the time scale that you see nucleation on in your experiment and how does this correspond to the duration of your simulation?
  • have you considered that you are (likely) simulating a perfect surface without defects and that you may have some hysteresis effects due to that as you would depend on fluctuations to create some trigger for nucleation? there is published literature on supercooled water which should give you some impression on how far you will be off with your simulation considering the system size?
  • have you considered that (simple) water potentials have a phase diagram that (somewhat) differs from that of real water? most are parameterized to represent bulk water around ambient conditions.
  • you are likely not considering the (dynamic) polarization of the metal due to the interaction with the polar molecules. a simple way to model this would be the “image charge model”, which does not exist in the official LAMMPS distribution since it is notoriously difficult to make it work with LAMMPS’ domain decomposition parallelization, and nobody has yet proposed an implementation that is suitable. using pair style hybrid (which is what is commonly used for this kind of system with LAMMPS) won’t be a good model to represent this property of such a system because you would be forced to use a very simplified pair-wise model to represent the interaction of the metal with the adsorbed molecules.


Thank you for the information.