Specific rotations for five notoriously difficult molecules, (S)-methyloxirane, (S)-methythiirane, (S)-2-chloropropionitrile, (1S,4S)-norbornenone, and (1R,5R)-β-pinene, have been computed using coupled cluster (CC) and density functional theory (DFT). The performance of the recently developed LPol basis sets compared to the correlation-consistent sets of Dunning and co-workers has been examined at four wavelengths: 355, 436, 589, and 633 nm. We find that the LPol basis sets are an efficient choice, often outperforming the more commonly used correlation-consistent basis sets of comparable size. The smallest of the four, LPol-ds, performs nearly as well as the rest of the series and often yields results closer to the basis set limit than appreciably larger basis sets. While the performance of the LPol bases is admirable, they still do not alleviate the need for high levels of electron correlation, vibrational corrections, and the inclusion of solvent effects to accurately reproduce experimental rotations. In particular in the case of β-pinene we find that they do not produce agreement between DFT and experiment as was previously suggested.

The effectiveness of the optical rotation prediction (ORP) basis set for computing specific rotations at the coupled cluster (CC) level has been evaluated for a test set of 14 chiral compounds. For this purpose, the ORP basis set has been developed for the second-row atoms present in the investigated systems (that is, for sulfur, phosphorus, and chlorine). The quality of the resulting set was preliminarily evaluated for seven molecules using time-dependent density-functional theory (TD-DFT). Rotations were calculated with the coupled cluster singles and doubles method (CCSD) as well as the second-order approximate coupled cluster singles and doubles method (CC2) with the correlation-consistent aug-cc-pVDZ and aug-cc-pVTZ basis sets and extrapolated to estimate the complete basis-set (CBS) limit for comparison with the ORP basis set. In the compounds examined here, the ORP calculations on molecules containing only first-row atoms compare favorably with results from the larger aug-cc-pVTZ basis set, in some cases lying closer to the estimated CBS limit, while results for molecules containing second-row atoms indicate that larger correlation-consistent basis sets are necessary to obtain reliable estimates of the CBS limit.

The impact of orbital localization on the efficiency and accuracy of the optimized-orbital coupled cluster model is examined for the prediction of chiroptical properties, in particular optical rotation. The specific rotations of several test cases-(P)-[4]triangulane, (S)-1-phenylethanol, and chiral conformers of 1-fluoropentane, heptane, and nonane-were computed using an approach in which localization is enforced throughout the orbital optimization and subsequent linear response computation. This method provides a robust local-correlation scheme for future production-level implementation. Although the cross-over point between the canonical and localized coupled cluster approach lies at larger molecules than for ground-state energies, the scheme presented should still provide reduced scaling sufficient to investigate much larger molecules than are presently accessible.

No abstract

We have investigated 15 excited states of the allyl radical, including the lowest three valence states (two doublets and one quartet) and the n = 3 Ry series, using coupled cluster methods that approximate the correlation effects of connected triple excitations. The quality of the excitation energies is measured on the basis of comparison to existing theoretical and experimental data, as well as on the basis of three diagnostics related to spin contamination and the overall level of excitation of a given state. Basis-set effects are significant for states exhibiting substantial Rydberg character, and the use of molecule-centered diffuse functions appears to provide an accurate description of such states, while avoiding the computational expense of basis sets in which diffuse functions are added to every atom in the molecule. In contrast to earlier observations for linear carbon-chain radicals, coupled cluster methods compare well to both theoretical predictions and experimental band origins, where discrepancies in the latter are sometimes attributable to structural relaxation in the excited state. One of the three lowest (2)B(1) excited states exhibits a twisting of the terminal methylene groups to yield a C(2)-symmetry minimum. The most challenging states for coupled cluster methods are of A(2) symmetry, where both spin contamination and basis-set effects are appreciable.