Force fields (FF) are parametrized based on available properties (experimental or using QM calculations). How FFs can model properly/accuratly interfacial phenomena in multi-structural molecules during the phase transition?
This depends on what properties (e.g., charge, dipole moment, size, van der Waals interactions) of the molecules there are that determine the phase. If these are reproduced in the model everything should be fine - but in that way the phase transitions are a sensitive test of your model.
The idea behind this is that if we can describe the molecule by a few essential molecular properties that are independent of the environment we should be able to produce a transferable force field that is valid under many different conditions. This might require that you include, e.g., polarisabilties to get a good enough description of the molecules.
Having studied "unusual" dynamic behaviour in multi-structural molecules, internal activation MD effects over the interfacial layer is analytically approved to have a determine role in phase transitions.
See
http://www.nature.com/articles/srep25572
Which FF(s) may re-produce this property of these molecules at the interface in multi-phase flows during the phase changes?