Weird GCMC of Water Result

I have a couple of thoughts. First, what are the pressures in the systems before and after the transition? Do these pressure look like vapor pressures you might expect for the water model in question at these conditions? Second, it could be that this is a vapor-liquid phase change you are observing. If you start from one of the configurations that is in a liquid state and change the chemical potential to -40 kJ/mol, does it stay a liquid or convert to vapor?

"It struck me as odd that there would be a "cliff" chemical potential where the density goes to 0"

This is normal for a sub-critical fluid. Above the critical point, all densities will be stable. Below the critical point, there will be densities intermediate to the liquid and vapor densities that are unstable. If you were to run an NVT simulation at one of these intermediate densities, the system would phase separate into liquid droplets separated by vapor (it may look like vacuum on the length scales of the simulation). In the grand canonical ensemble, there will be a chemical potential (mu.coexist) at which both the vapor and liquid densities are stable. In a GCMC simulation at mu.coexist, the system will spend extended periods of time at vapor densities and then rapidly transition to liquid densities and vice versa. At mu slightly greater than mu.coexist, the liquid phase will be thermodynamically stable but the vapor phase will be metastable--that is there is a free energy barrier to the vapor-to-liquid transition, but once the system makes the transition to the liquid phase it is unlikely to return to vapor phase.

Ryan