ON THE ACCURACY OF DENSITY-FUNCTIONAL THEORY FOR ION-MOLECULE CLUSTERS - A CASE-STUDY OF PL(N)(+) CLUSTERS OF THE FIRST AND 2ND ROW HYDRIDES

PL(n)(+) clusters (n = 1, 2 and L = NH3, OH2, FH, PH3, SH2, ClH) in bo th their triplet and singlet states have been characterized by common approximate density functional methods, SVWN, BVWN, BLYP, and B3LYP. T he phosphorus-ligand distances (R), dissociation energies (D-0), tripl et-singlet gaps (Delta(t-s)), and several bond properties, such as the electron density (rho(r(c))), the Laplacian (del(2) rho(r(c))), and t he local energy density H(r(c)) at the bond critical point, were compa red with those obtained by accurate ab initio molecular orbital theory , namely, second-order Moller-Plesset (MP2) and G2 theory. In general, it is observed that the local spin density approximation (SVWN) yield s stronger bonds than ab initio molecular orbital theory. However, add ition of gradient corrections to the exchange functional (BVWN) yields ion-molecule bonds that are too weak. Finally, taking account also of gradient corrections to the correlation functional (BLYP) leads to ve ry close agreement with ab initio results. Among these functionals, Be cke's hybrid functional, B3LYP, best fit the second-order Moller-Pless et and G2 data, reproducing the qualitative trends observed for the ab ove-mentioned properties of phosphorus clusters, except for del(2) rho (r(c)). This fit is particularly good for distances, dissociation ener gies, and electron densities at the bond critical point, and both meth ods show similar deviations of the values of binding energies and trip let-singlet gap with respect to the G2 data. Compared with our most ac curate ab initio molecular orbital data, namely G2, significant overbi nding for the singlets, larger for one-ligand than for two-ligand comp lexes, and significant overestimation of the triplet-singlet gap for o ne-ligand complexes is observed for both methods, namely, B3LYP and MP 2. The deviations at the second-order Moller-Plesset level of theory a re mainly due to the lack of quadratic configuration interaction (QCI) corrections, and this deficiency is also present to some extent in B3 LYP. However, for larger clusters these corrections are smaller, there fore the B3LYP functional is expected to lead to accurate descriptions .