Coarse grained Molecular Modelling and Information Loss

Tracking atomistic details in molecular simulations generates a lot of data and reveals a lot of statistical and thermodynamic properties of system under study. Due to large system sizes and time scales in a realistic molecular system, current atomistic simulations needs to cover up to 10-12 orders of magnitude computation time or system size compare to what current computers can tackle. A coarser representation of molecular system is usually employed to reduce the spatial and time scales, proportionally computation time. Fine details of the system is removed and
effective interactions or physical characteristics are introduced instead.(See a review by Pep Espanol Statistical Mechanics of Coarse Graining).

An important problem of information loss in this coarse graining procedure is recently investigated by using relative entropy approach, where entropy of reduce system compared to detailed system by a group in California [link].

A similar effort in this direction to model water (see my short blog as well [link])
is developed by Van gunsteren group coarse graining polarizable water [link].

Pair potentials are not that bad: Water & Silica

One of the challenges in modeling water or silica is the figure out correct description of inter-atomic potential. Usually used models for water in-cooperate 3-body interactions and charges, B.Guillot has a review on the subject [doi]. However there are attempts to use only pair potential for water [doi], as well as for amorphous silica [doi].