Abstract:

A new global optimization-based method for deriving parameters of atom-atom potentials for crystals has been developed. The method consists of two parts (force minimization and a novel Monte Carlo method, Vector Monte Carlo [VMC]) to minimize a vector target function consisting of three components. The first component depends on the rank of the minimized experimental structure among all structures found by a global search; the second one is a function that allows minimization of deviations between experimental and minimized experimental structures; and the third one is a penalty function that provides the best fit to the available heats of sublimation. The method enables the limitations of any chosen form of the potential to be evaluated and finds the best possible set of parameters for a given form. The method has been applied to sets of saturated hydrocarbon and ether molecules. For hydrocarbons, the resulting set of potential parameters describes observed crystal structures with high accuracy. Significant improvement was achieved in the case of ethers for which the experimental structures of the selected molecules (except dioxane) were stable upon local energy minimization and were found as global minima of the refined potential. Our results for dioxane are in agreement with those of Mooij et al. (J. Phys. Chem. A 1999, 103, 9883) and show that potential energy calculations with simple "6-exp-1" energy functions are not adequate for predicting the crystal structure of dioxane.