Characterization of hydrogen dissociation over aluminium-doped zinc oxide using an efficient massively parallel framework for QM/MM calculations

Abstract

A task-farm parallelization framework has been implemented in the ChemShell computational chemistry environment to provide a facility for parallelizing common chemical calculations, including finite-difference Hessian evaluation, the nudged elastic band method for reaction path optimization, and population-based methods for global optimization. The optimization methods are provided by a parallel interface to the DL-FIND optimization library. As ChemShell can already exploit parallel external programs for energy and gradient evaluations, the new methods result in a two-level approach to parallelization that gives significantly improved performance for massively parallel calculations. For typical systems, speed-up factors of five to eight times have been observed compared with non-task-farmed calculations. The task-farming version of ChemShell has been used to study the heterolytic dissociation of a hydrogen molecule over a polar oxygen-terminated surface of aluminium-doped zinc oxide using an embedded cluster hybrid QM/MM approach. We calculate a 42 kcal mol −1 heat of reaction and a 30 kcal mol −1 activation energy, which is equivalent to a high backward reaction barrier of 72 kcal mol −1 per H 2 molecule, in close agreement with temperature programmed desorption experiments. The dissociation path includes a stable intermediate comprising a hydride ion in an oxygen vacancy and physisorbed atomic hydrogen.