There are several models available to simulate water (e.g., http://www1.lsbu.ac.uk/water/water_models.html) in molecular dynamics simulations. Most of the these models are developed/targeted to simulate pure liquid water. However, in a situation like moist N2, most of the collisions of a water molecule happen with nitrogen molecules. So, the first starting point to simulate moist N2 would be finding out which water model would be best suitable to model these intermolecular interactions?

One possible way out (as done in most of the available literature) is to use SPC/E or TIP4P model and then apply the Lorentz–Berthelot mixing rule to get interaction parameters (sigma and epsilon) to model intermolecular interaction between N2 and H2O. The problem with the use of such a mixing rule to simulate a collision between H2O and N2 molecule is this--- water is a polar molecule, and therefore, most of the water-models use a combination of coulombic and LJ interaction to simulate 'water - water' interaction. Point electric charges are placed on three sites (on each atom), and one LJ site (Lennard-Jones potential) is placed at/near the oxygen atom. On the other hand, N2 is a nonpolar and poorly polarizable molecule and is usually modeled with a LJ potential. No charge is placed on any N atoms. During a collision between N2 and H2O molecule, the N atom (of the N2 molecule) can interact with both the hydrogen and oxygen atoms (of the H2O molecule). However, if we apply a mixing rule between the LJ potential of N2 and that placed on O atom of H2O, we will be modeling only the O and N interaction and ignoring the H and N interaction. It does not seem ok to me, especially if rotational energy exchange is our primary concern, as even small force applied on H atom can result in a meaningful torque. It should be noted that the center of mass of H2O molecule would be very close to O atom (precisely, at the angle bisector H-O-H bong at a distance of '0.068d' from the O atom, where 'd' is the bond length of O-H bond) and H atoms are far from the C.O.M.

So, to properly simulate an N2 and H2O collision, we need an interaction model that takes into account interaction between the N and O as well as N and H. Do there exist any such model? Or any other suitable approximation?