Ligand substitution: An assessment of the reliability of ab initio calculations

Density functional methods and conventional ab initio molecular orbital methods have been compared in an investigation of the model ligand substitution reaction Pd(N-O)(CH3)(PH3) + CO → Pd(N-O)(CH3)(CO) + PH3 (N-O = NHCHCOO-). Full geometry optimizations were performed for each stationary point using a range of different basis sets, relativistic effective core potentials (RECPs), and levels of theory in order to ascertain a reliable and computationally inexpensive method for probing such reaction mechanisms. For geometries that compare well to experiment, electron correlation and double-ζ valence basis sets with d-type polarization functions on the main group elements are essential. Geometries determined using small-core RECPs and nonlocal density functional theory compared favorably with those produced at the second-order Møller-Plesset (MP2) level, and good agreement with available experimental data was observed. The density functional reaction energetics, however, were poor, apparently due to an unrealistically high Pd-CO binding energy relative to the Pd-PH3 binding energy. This error was reduced by the use of the Becke three-term hybrid exchange functional. The reaction barrier and reaction energy were predicted to be +15.0 kJ/mol and +5.1 kJ/mol, respectively, at the MP2 level with a medium basis set. Approximate large basis set CCSD(T) energetics were determined using an additivity scheme. This method predicted a stabilization of the 5-coordinate transition structure with respect to the 4-coordinate species, reducing the reaction barrier to +4.9 kJ/mol relative to the separated reactants. The associated reaction energy was +8.9 kJ/mol. © 1996 American Chemical Society.