Correlations between experimental and theoretical adiabatic ionization energies for organic compounds and rate constants for atmospheric reactions with hydroxyl radicals

Abstract

Adiabatic ionization energy (AIE) calculations were performed at the AM1, PM3, PM6, PDDG, HF/QZVP, and B3LYP/QZVP levels of theory on 722 atmospherically relevant organic compounds with available experimental k OH . From the starting set of molecules, a final suite of 114 mono-and polyfunctionalized compounds provided converged neutral and cationic geometries without imaginary frequencies for all six levels of theory. NIST evaluated AIEs were available for 54 compounds, providing mean absolute AIE prediction errors of 0.31 (AM1), 0.28 (PM3), 0.50 (PM6), 0.36 (PDDG), 1.22 (HF/QZVP), and 0.20 eV (B3LYP/QZVP). Modest correlations were found between the experimental (r=-0.68, SE=0.81) and computationally estimated (r=-0.77 [AM1], -0.75 [PM3], -0.83 [PM6], -0.79 [PDDG], -0.83 [HF/QZVP], and -0.82 [B3LYP/QZVP]; SE=0.75 [AM1], 0.78 [PM3], 0.66 [PM6], 0.73 [PDDG], 0.67 [HF/QZVP], and 0.68 [B3LYP/QZVP]) AIEs and the corresponding experimental log k OH . Univariate AIE versus k OH correlations are of lower predictive ability than state-of-the-art multivariate techniques, and are limited by the inability to calculate reliable AIEs for large numbers of atmospherically relevant compounds due either to convergence failures at various levels of theory or the presence of imaginary frequencies for converged cationic geometries.The hydroxyl radical ( · OH) plays a fundamental role in the abiotic cycling of organic, organometallic, and inorganic compounds in the troposphere [1][2][3][4]. In addition to experimental approaches, theoretical methods for predicting the rate constants for reaction of organic compounds with the hydroxyl radical (k OH ) have been the focus of a number of studies. In general, these predictive investigations have taken one of three general tactics [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24]: (1) rigorous and time-intensive medium-through high-level computational studies (e.g., density functional theory, composite methods) on all possible mechanistic pathways for the reaction of a particular compound with the hydroxyl radical (i.e., addition to unsaturated groups, hydrogen abstractions, etc.); (2) regression based quantitative structure-activity relationship (QSAR) models employing a range of two-and threedimensional molecular descriptors; and (3) univariate correlations with physicochemical properties (e.g., ionization energies [IEs], bond dissociation enthalpies [BDEs]).In the absence of experimental data, high-level theoretical studies typically offer the most accurate approach. However, they are not suitable for screening large numbers of potential atmospheric contaminants due to the computational costs and substantial user-interaction/expert judgement required on an individual compound caseby-case basis to ensure all relevant mechanistic pathways have been rigorously investigated. Multivariate QSAR models are amenable to rapid screening techniques, and many offer reasonable accuracy. Because of the often arbitrary nature of the variables chosen for multivariate QSARs (whi...