Predicting glass properties from structure data and intermolecular forces

Beside energy, entropy is a basic property determining the relative stabili ty of different states of matter. For glass both properties cannot be deter mined by measuring energy exchange with the surroundings. However, given in teraction potentials, both energy and entropy can be determined from molecu lar distributions. The entropy function S = S(E) is related to structure by partial derivative S-2/partial derivativeE(2) = - <(DeltaE)(2)> (-1), wher e <(DeltaE)(2)> are the spatial energy fluctuations, expressible in terms o f low order molecular distributions. The functional dependence on E implies certain constraints kept fixed while the energy E is varied. Such constrai nts are controllable by external forces when the system is in internal equi librium, or can be explicitly introduced in computer experiments, but are o nly assumed to exist in the case of glass, practically preventing it from c hanging with time, although the glass phase is unstable in principle. Integ ration of the differential equation for S(E) is performed with the aid of a modeling of the radial distribution g(r) in form of a parametrized analyti c function, g(r) = g(r; L, D), where L is a lattice characterizing the domi nant local configurations of atoms and D is a 'structural diffusion' parame ter specifying the degree of spatial decay of coherence between local struc tures. The modeling provides a representation of structure by a point in th e low dimensional parameter space {L, D}. This space includes also the orde red state (L) equivalent to (L, 0), for which S = S(L) = 0, as follows from the third law of thermodynamics. Thus integration can be performed along a (virtual) path connecting (L, D) to (L, 0). The method is illustrated by e valuating the entropy of a model metal in the liquid and glass state. (C) 2 001 Elsevier Science B.V. All rights reserved.