Diffuse and polarization functions have been optimized for the LANL2DZ basis set for elements in groups 14-17. The optimized exponents are in most cases similar to those optimized with different effective core potentials, valence basis sets, or computational models. The average of the LANL2DZ results for different models is taken to be the best generalized set of exponents. The extended basis set gives good results (average deviation from experiment 0.11 eV) for atomic electron affinities with the B3LYP model, but is consistently low with the MP2 model. The extended basis set gives similar performance to the all-electron 6-31+G(d) basis set in calculations of vibrational frequencies and bond energies in selected main-group compounds, and is intermediate in speed between the 6-31+G(d) basis set and the unmodified LANL2DZ basis set. IntroductionTo support our experimental program on the energetics of hypervalent compounds, 1 we need to calculate structures, energetics, vibrational frequencies, and other properties for a number of species containing heavy main-group atoms, such as I 5 -, SbF 6 -, and BiCl 4 -. We have therefore become interested in expanding the range of available basis sets for heavy maingroup atoms.Basis sets used in calculations on these systems must meet certain criteria. 2 Computational efficiency generally dictates the use of effective core potentials (ECPs). Use of a relativistic core potential (RECP) can also largely account for the relativistic effects that become significant for heavier atoms. Because many of the systems are anionic, diffuse functions are needed on all of the atoms. 3 Also, the basis set used on the central atom should include at least one set of d-polarization functions for the valence-p orbitals. In addition to modifying the shape of the valence orbitals, d functions allow probing of the controversial nature of bonding in hypervalent systems; that is, whether the three-center, four-electron bonding model 4 (where p orbitals on the central atom are primarily involved in the bonding) or the older expanded octet model 5 (where d orbitals on the central atom contribute substantially to the bonding) is more defensible.Finally, the possible use of different basis sets on different atoms require basis set balance to avoid inaccuracies in the charge distribution in the molecule. "Balance" means that for each atom, the last functions of each symmetry added to the basis set contribute equally to energy lowering. 6-8 For example, adding diffuse functions to double-basis sets has a significant effect on the energy even for atoms. 9 Thus, using a small basis set for heavy atoms and a larger basis set for lighter atoms in the same molecule can lead to inaccurate electron distributions, energetics, and other properties. In practice, balance can be approached by using basis sets of a similar size for each atom.There is no comprehensive basis set for the main-group elements in general use that includes ECPs, diffuse functions,

[reaction: see text] Computational studies of three different reaction types involving hydrocarbons (homolytic C-C bond breaking of alkanes, progressive insertions of triplet methylene into C-H bonds of ethane, and [2+2] cyclizations of methyl-substituted alkenes to form polymethylcyclobutanes) show that the B3LYP model consistently underestimates the reaction energy, even when extremely large basis sets are employed. The error is systematic and cumulative, such that the reaction energies of reactions involving hydrocarbons with more than 4-6 C-C bonds are predicted quite poorly. Energies are underestimated for slightly and highly methyl-substituted cyclic and acyclic hydrocarbons, so the errors do not arise from structural issues such as steric repulsion or ring strain energy. We trace the error associated with the B3LYP approach to its consistent underestimation of the C-C bond energy. Other DFT models show this problem to lesser extents, while the MP2 method avoids it. As a consequence, we discourage the use of the B3LYP model for reaction energy calculations for organic compounds containing more than four carbon atoms. We advocate use of a collection of pure and hybrid DFT models (and ab initio models where possible) to provide computational "error bars".

The xenon-fluoride bond dissociation energy in XeF3- has been measured by using energy-resolved collision-induced dissociation studies of the ion. The measured value, 0.84 +/- 0.06 eV, is higher than that predicted by electrostatic and three-center, four-electron bonding models. The bonding in XeF3- is qualitatively described by using molecular orbital approaches, using either a diradical approach or orbital interaction models. Two low-energy singlet structures are identified for XeF3-, consisting of Y- and T-shaped geometries, and there is a higher energy D3h triplet state. Electronic structure calculations predict the Y geometry to be the lowest energy structure, which can rearrange by pseudorotation through the T geometry. Orbital correlation diagrams indicate that that ion dissociates by first rearranging to the T structure before losing fluoride.

The strengths of the F 2 ClP-Cl -, POCl 3 -Cl -, and PSCl 3 -Clbonds have been determined by measuring thresholds for collision-induced dissociation in a flowing afterglow-tandem mass spectrometer. The results are combined with previously determined values for the PF 4 -, PF 3 Cl -, POF 4 -, and PCl 4systems to determine the effect of adjacent ligands on hypervalent bond strengths. Although the addition of electronegative equatorial ligands strengthens bonding to axial halides, the effect is, in some cases, outweighed by the rearrangement energy of the dissociation products. Computational results indicate that the B3LYP/aug-cc-pVTZ method gives particularly good agreement with experiment among the models used here; however, several less resourceintensive methods give acceptable agreement.