Reduced time step when water boils
I am trying to model the conductive heating process in water-saturated soil and I found the time step is reduced to a quite small value when water starts boiling. Is that normal to have that when new phase (water vapor in my case) is being generated?
It's quite normal. As you might expect, the properties of the water are undergoing significant changes over a short amount of time when a phase shift occurs. Consequently, the simulation slows down enough to handle this situation properly. If you expect this to occur quite frequently, I would suggest reducing the WNR factor in PARAM.3 to something around 0.98 to 0.99 (by default, WNR=1). This invokes a relaxation method to reduce the amount of phase shifting that occurs between Newton Raphson (NR) iterations within a time step. The result is that the simulation progresses through the phase shift with larger time steps than it would otherwise. However, when no phase shifts are taking place, it will slow the convergence down somewhat.
If you're running with TOUGH3 or iTOUGH2, you can also change MOP2(13) in the MOMOP block, which will automatically decrease WNR when oscillations occur within a time step (which can happen often with phase shifts). It will then increase back up (eventually to 1) after each time step that has enough NR iterations that converge without oscillations. If you decide to go with this option, I would suggest not setting MOP2(13) greater than 2.