Utilizing relativistic effective core potentials for accurate calculations of molecular polarizabilities on transition metal compounds.

The requirements necessary to extend an ECP basis set for the calculation of electric and linear optical properties to the transition metals are studied. Previously an augmentation of the SBK basis set for 39 elements with s and p electron only valences (H-Rn, excluding Ga, In, and Tl) [J. Comput. Chem., 2005, 26, 1464-1471] was presented. In this work, electric dipole moments, polarizabilities, and anisotropies of selected metal hydrides, sulfides, and bromides, cisplatin, and the Fe, Ru, and Os metallocene derivatives along with many other systems are calculated and discussed. ECP calculations of molecules containing 3d and 4d metal centers among main group atoms have good agreement, often within 1-2% of the all-electron result at the time-dependent Hartree-Fock (TDHF)/Sadlej level of theory. Molecules with a 5d metal center have a large difference from and are more accurate than the all-electron results due to the inclusion of relativistic effects in the ECPs. The polarizability as calculated with and without ECPs of metallic clusters and surfaces is increasingly different as atomic number increases, again due to a lack of relativistic effects in the all-electron calculations. The augmented ECP calculations are consistent with relativistic all-electron results, while the Sadlej calculations are consistent with other nonrelativistic results. Both relativistic and basis set effects are less noticeable when the metal is in a formally positive state.

Silicon radicals in silicon oxynitride: a theoretical ESR study.

Quantum mechanical calculations are performed on a series of silicon radical defects. These are the upward arrow Si[triple bond]O(3-x)Nx, upward arrow Si[triple bond]N(3-x)Si(x), and upward arrow Si[triple bond]Si(3-x)Ox defects, where x takes on values between 0 and 3. The defects under study constitute a central silicon radical, upward arrow Si, with differing first-nearest-neighbor substitution, as may be found at a Si/SiOxNy interface. These first-nearest neighbor atoms are connected to the silicon radical via three single covalent bonds, denoted as " [triple bond] ". A hybrid defect, upward arrow Si[triple bond]ONSi, is also included. Calculations are performed on gas-phase-like cluster models, as well as more-constrained hybrid quantum and molecular mechanical (QM/MM) models. The isotropic hyperfine coupling constants of these defects are calculated via density functional theory (DFT). Trends in these calculated hyperfines are consistent between the different models utilized. Analysis of the electronic structure and geometries of defects correlate well with trends in the electronegativity of the first-nearest-neighbor atoms. Changes in radical hybridization, induced by changes in the first-nearest-neighbor composition, are the primary factor that affects the calculated hyperfines. Furthermore, comparisons to experimental results are encouraging. Agreement is found between experiments on amorphous to crystalline materials.

An augmented effective core potential basis set for the calculation of molecular polarizabilities.

Calculations of molecular polarizabilities require basis sets capable of accurately describing the responses of the electrons to an external perturbation. Unfortunately, basis sets that yield suitable quantitative results have traditionally been all-electron sets with large numbers of primitives, making their use computationally intractable even for moderately sized systems. We present a systematic augmentation of the effective core potential basis set of Stevens et al. [J Chem Phys 81, 12 (1984), Can J Chem 70, 612 (1992)] for 39 main group elements based on the procedure used to construct diffuse and polarization functions in the well-known Sadlej basis sets [Collec Czech Chem Comm 53, 1995 (1988)]. Representative calculations have been performed and we have shown that results to within 1% of all-electron calculations using the Sadlej basis set can be obtained for <1-35% of the computational cost using this new basis set.