Density-Functional Theory (DFT) is highly desirable for modeling biomolecules due to its accuracy. While DFT calculations of the energy are now common for large molecules, the use of DFT geometry optimizers is still confined to relatively small molecules containing no more than thirty atoms. The key limitation is that the number of steps required to reach the optimum geometry increases with the number of atoms in the molecule. Furthermore, conventional optimizers perform particularly poorly for the floppy molecules typically encountered in biochemistry. This proposal is based on a radical change in the algorithm for DFT geometry optimization, potentially enabling a quantum leap in the size of molecules that can be optimized. The new optimizer takes a much more accurate geometry step than conventional optimizers, so that the optimum geometry is reached in a small number of steps (about 10 for Taxol). Each of these 'smart steps' is obtained via the iterative solution of an equation which is ill-conditioned for floppy molecules. In phase I we devised an economical procedure to solve this ill-conditioned equation, and in phase II we propose to complete the development of a geometry optimizer based on this smart step. The proposed algorithm will be implemented in the Q-Chem software package and, if successful, widely distributed through Q-Chem, Inc. and Spartan, Inc. ? ?
