Electronic dipole polarizabilities of polychlorinated dibenzofurans and semiempirical PM6 level performance

Abstract

Polychlorinated dibenzofurans (PCDFs) are widespread and persistent environmental contaminants [1]. Electronic dipole polarizabilities (a) of PCDFs were previously computed at the B3LYP level with cc-pVDZ, 6-31G and 6-31G basis sets in order to elucidate the effect of the substituent position on the congener specific toxicity [2,3] and aqueous solubility [4]. Recently, semiempirical PM6 method [5] has been implemented in MOPAC 2007 package [6], giving satisfactory estimates of molecular properties such as heats of formation [5] and electronic a values [7,8]. This work is principally concerned on the validation of the PM6 method in the determination of a values, focusing attention on DF and the 135 PCDF congeners (Fig. S1 of the Supporting Material). Static aij (i, j = x, y, z) components were calculated at the AM1, PM3 and PM6 levels. Additionally, we computed aij values for DF and its octacloro substituted congener at the HF, MP2 and PBE0 levels with aug-cc-pVDZ basis set on the B3LYP/6-31G geometry. Present computations were performed with MOPAC 2007 [6] and PC GAMESS [9,10] programs. Calculated average polarizability, hai 1⁄4 1=3ðaxx þ ayy þ azzÞ, and polarizability anisotropy, Da 1⁄4 f2 1⁄2ðaxx ayyÞ 2 þ ðaxx azzÞ þ ðayy azzÞ þ 6ðaxy þ axzþ ayzÞ g , are given in Tables S1–S3 of the Supporting Material. The results show that PM6 is noticeably superior to both the commonly employed semiempirical AM1 and PM3 methods, reproducing the PBE0/aug-cc-pVDZ (and also MP2/aug-cc-pVDZ) hai values of DF and 1,2,3,4,5,6,7,8-OCDF within 5 a.u. (2–3%) and Da data within 8–11 a.u. (3–8%), geometrical effects (PM6 vs. B3LYP/6-31G ) being almost negligible. Note that the corresponding deviations for hai obtained using the AM1, PM3 and B3LYP/6-31G [3] data are substantially larger, being 36–94 a.u. (25–34%), 41–76 a.u. (27–28%), 24–47 a.u. (16–17%), respectively, while those for Da are 22–25 a.u. (9–20%), 16–43 a.u. (12–18%) and 11–14 a.u. (4– 11%), respectively. However, least-mean squared fitting linear relationships between the semiempirical and B3LYP/6-31G hai and Da data (See Figs. S2 and S3 of the Supporting Material) are