Investigation of prolate and near spherical geometries of mid-sized silicon clusters

We have applied a new strategy to search for the ground-state geometry of silicon clusters. It is a knowledge-based procedure that severely restricts the search space. Detailed analysis of the geometry of small clusters indicates the ubiquitous presence of certain structural elements. These elements can be created from smaller units already present in a cluster. Such investigation leads to the proposal of assembly procedures (CAP's). Starting with optimized structures of small clusters, we apply one or more CAP's to create larger clusters. This approach generates the initial population for a genetic algorithm. This procedure is tested on mid-sized silicon clusters because they show an intriguing coexistence of two different shapes: a prolate and a more spherical structures. Our method proved so efficient that for most test cases Si n (7 ≤ n ≤ 10 and 14 ≤ n ≤ 18) a simple local optimization procedure could find almost degenerate clusters of both shapes.