A functional approach to geometry optimization of complex systems

Abstract

A quadratically convergent procedure is presented for the geometry optimization of complex systems, such as biomolecules and molecular complexes. The costly evaluation of the exact Hessian is avoided by expanding the density functional to second order in both nuclear and electronic variables, and then searching for the minimum of the quadratic functional. The dependence of the functional on the choice of nuclear coordinate system is described, and illustrative geometry optimizations using Cartesian and internal coordinates are presented for Taxol™.

Acknowledgements

This work was supported by Q-Chem Inc. through the NIH Small Business Innovation Research programme (grant 1R43GM067335-01), a Research Corporation Cottrel College Science grant (grant CC5459) and an NSF MRI grant (NSF-DMR0116315).

Notes

 For a summary of covariant and contravariant matrix representations in a non-orthogonal basis, see, for example, [42].