PERFORMANCE OF DENSITY-FUNCTIONAL THEORY ON THE POTENTIAL-ENERGY SURFACE OF THE H+OCS SYSTEM

The H + OCS potential-energy surface (PES) was used to evaluate the pe rformance of density functional theory by comparing the results to ab initio calculations at the QCISD(T)//UMP2 and UMP2 levels using the au g-cc-pVTZ and 6-311+G(2df, 2p) basis sets. The two major reaction path s on this PES involve formation of OH((II)-I-2) + CS((1)Sigma) (reacti on I) and SH((II)-I-2) + CO((1)Sigma) (reaction II). Experimental and QCISD(T)//UMP2/aug-cc-pVTZ activation barriers for (II) and reaction e nthalpies for (I) and (II) were compared to values calculated by sever al density functionals (BLYP, B3LYP, B3PW91, BPW91, BP86, and B3P86) u sing the aug-cc-pVTZ basis set. All DFT/aug-cc-pVTZ predictions, excep t for the B3LYP prediction of the enthalpy of reaction I, were outside the range of experimental uncertainty. B3LYP predictions were in clos est agreement with the experimental values and QCISD(T) predictions. B 3LYP, BPW91, and B3PW91 predictions of the rate-limiting barrier to re action II are within 3.5 kcal/mol of the QCISD(T) prediction, and all DFT values are below that of the QCISD(T). Reaction enthalpies for (I) and (II) were calculated using the BHandHLYP density functional and t he 6-311+G(2df,2p) basis set. These predictions were closer to experim ent and QCISD(T) values than any other DFT calculations, and the predi cted enthalpy for reaction I is within the range of experimental value s. The second portion of the study compared B3LYP and BLYP predictions of the 12 transition states and 6 stable intermediates within this PE S with previously reported QCISD(T)//UMP2/6311+G(2df,2p) predictions. The complexity of this surface allows for the evaluation of barrier he ights for 28 reactions involving hydrogen addition, elimination, isome rization, migration, and radical diatomic elimination. With the except ion of five reactions, all B3LYP barrier heights are within 3.7 kcal/m ol of the QCISD(T) predictions and in several cases are in as good or better agreement than the UMP2 predictions. In addition, all but one o f the B3LYP barriers lie below the QCISD(T) values. The most significa nt differences between the ab initio and DFT predictions were in the s addle points for radical elimination or addition. BLYP/6-311+G(2df, 2p ) failed to find the two transition states associated with SH eliminat ion from the cis- and trans-HSCO species. B3LYP located the saddle poi nt for SH elimination from cis-HSCO, but its prediction of a saddle-po int structure for SH elimination from trans-HSCO has an energy (withou t zero-point corrections) lower than that of the products. These trans ition states were subsequently optimized using the BHandHLYP functiona l and the 6-311+G(2df,2p) and 6-31G* basis sets. The geometries of th ese saddle points were in better agreement with UMP2/6-311+G(2df,2p) p redictions than were the BLYP and B3LYP predictions. The BLYP predicti ons are in overall worse agreement with the QCISD(T) results than are the B3LYP predictions.