PM3(tm) parameterization using genetic algorithms

摘要

PM3(tm) has great potential in studying transition metals because of its speed and applicability to large complexes. However, its parameterization is not yet available for all 30 d-block metals. In this research, genetic algorithms (GAs) were evaluated for the development of PM3(tm) parameters for technetium (Tc). initial Tc parameters were obtained by interpolation of parameters for the metals flanking it in the periodic table-molybdenum and ruthenium. Prototypical Tc compounds were chosen from the Cambridge Structural Database. The sensitivities of the 21 PM3(tm) parameters were tested using different methods and their impact on molecular geometry assessed. The fitness criterion was based on the root mean square (rns) of the distance matrix between calculated and crystal structures. The GA-optimized parameters improved the calculated structural accuracy by more than 50% versus interpolated parameters. In addition, structural prediction (bond lengths within 0.04 Angstrom, bond angles within 2 degrees, dihedral angles within 4 degrees) with the GA-developed parameters for Tc is competitive with those already available in PMS(tm) and with that expected from high-level ab initio calculations, but in a fraction of the time. (C) 2000 John Wiley & Sons, inc. [References: 36]