Force Matching and Iterative Boltzmann Inversion to Improve DFTB Parameters for Liquid Water
Manuel Dömer (Laboratory of Computational Chemistry and Biochemistry, Lausanne)

The parameterization of accurate and transferable repulsive potentials is a key ingredient for the density functional tight-binding method (DFTB). In the conventional parameterization scheme the balanced description of differ- ent chemical environments involves significant human effort and chemical intuition. The present work aims at an in situ parameterization scheme with limited transferability but maximal accuracy for the chemical and physical environment under investigation. In a comparative study we tested the perfor- mance and practical features of two alternative schemes: Force matching and the iterative Boltzmann inversion method. We applied the two schemes to liq- uid water at ambient conditions, a pathological case for conventional DFTB.
Our newly determined parameters significantly improved the structural prop- erties of liquid water.