Multiscale approach to explore the potential energy surface of water clusters (H2O)nn <or= 8.

We propose a multiscale method to explore the energy landscape of water clusters. An asynchronous genetic algorithm is employed to explore the potential energy surface (PES) of OSS2 and TTM2.1-F models. Local minimum structures are collected on the fly, and the ultrafast shape recognition algorithm was used to remove duplicate structures. These structures are then refined at the B3LYP/6-31+G* level. The number of distinct local minima we found (21, 76, 369, 1443, and 3563 isomers for n = 4-8, respectively) reflects the complexity of the PES of water clusters.