NEW ALGORITHM FOR THE OPTIMIZATION OF GEOMETRIES IN LOCAL DENSITY FUNCTIONAL THEORY

We present a new algorithm for optimizing geometries, inspired by both the recent unified approach for molecular dynamics and density functional theory of Car and Parrinello, as well as the traditional iterative SCF approach to solving the Kohn-Sham equations with quasi-Newton algorithms to update geometries. Tests on organic and organometallic molecules and transition metal clusters indicate that the proposed "hybrid" algorithm, within a particular form of local density functional theory, the LCGTO-MCP-LSD method, is a marked improvement over either approach used alone. © 1990.