非表示:
キーワード:
-
要旨:
Several variants of the quasi-Newton method for nonlinear unconstrained optimization are studied in MNDO calculations for a representative set of 58 hydrocarbon test molecules. The BFGS algorithm with a simple inaccurate line search is recommended for semiempirical geometry optimizations. Compared with standard current MNDO programs such as MOPAC, this approach is found to reduce the computation time for optimizing the test molecules by more than a factor of 2.
