Density Functional Calculations of E2 and SN2 Reactions: Effects of the Choice of Method, Algorithm, and Numerical Accuracy.

Herein we provide a detailed account on how the potential energy surfaces of the E2 and SN2 reactions of X(-) + CH3CH2X (X = F, Cl) depend on various methodological and technical choices in density functional calculations. We cover a choice of density functionals (OLYP, various M06-types, and the new SSB-D), basis sets (up to quintuple- and quadruple-ζ for Gaussian- and Slater-type orbitals, respectively, plus polarization and diffuse functions), and other aspects of the computations (among others: nonrelativistic versus zeroth-order regular approximation relativistic; numerical integration accuracy; all-electron versus frozen core; self-consistent field (SCF) versus post-SCF). The program codes ADF and NWChem are used for calculations with Slater- and Gaussian-type basis sets, respectively. The fluoride systems (X = F) appear to not only depend extremely sensitively on the basis set size (especially the presence of diffuse functions) but also on other technical settings, especially in the case of hybrid meta-generalized gradient approximation functionals. This work complements a recent contribution (Y. Zhao, D. G. Truhlar, J. Chem. Theory Comput. 2010, 6, 1104) and provides recommendations for density functionals, basis sets, and technical settings.