So let’s just do a little back-of-the-envelope estimate. The NaCl conventional unit cell is 0.56x0.56x0.56nm^3 and has 8 atoms. So to reach a micrometer, you would need replicate this 2000x2000x2000 time, which are 64 billion atoms. And that is the NaCl crystal only.
But there are other things to consider:
- how large a water layer do you need around this? so you may be looking at systems 1000 times larger (which is only a 1:10 ratio in each direction)
- LAMMPS has limitations as to how many atoms per MPI rank you can simulate and this would be beyond that
- you will need to do long-range electrostatics which parallelized only O(N log(N)) once the 3d-Fourier transform becomes a relevant term, i.e. when you grow the number of CPUs you use
- what time scale are the processes on that you want to study. there are same as the size limitations, you also have time limitations. while you may be able to run a tiny system for miliseconds (your time step is of the order of a femtosecond), this will be near impossible for a huge system
- what kind of potential can actually model the process accurately? you need something that is good to represent the solid, the interface and the interaction with water at the interface and the solvated ions. not an easy feat.
- what impact does structuring of the surface have? how much do you know whether structural defects matter? or whether dissolving ions from an ideal surface is an activated process?
Overall, I don’t think this is something that is as easily doable with classical MD as you think. Have you done any research of the published literature about what kind of research people have done to study dissolution of salts in water? or any kind of solid into a solvent? For all I know, this is a “hard” problem and most certainly not accessible by a brute force “simulate and a see what happens” approach. From my limited experience, I would expect a method like kinetic monte carlo be more suitable, but then again, how would one parameterize that?