A variation of Gaussian-3 (G3) theory is presented in which the geometries and zero-point energies are obtained from B3LYP density functional theory [B3LYP/6-31G(d)] instead of geometries from second-order perturbation theory [MP2(FU)/6-31G(d)] and zero-point energies from Hartree–Fock theory [HF/6-31G(d)]. This variation, referred to as G3//B3LYP, is assessed on 299 energies (enthalpies of formation, ionization potentials, electron affinities, proton affinities) from the G2/97 test set [J. Chem. Phys. 109, 42 (1998)]. The G3//B3LYP average absolute deviation from experiment for the 299 energies is 0.99 kcal/mol compared to 1.01 kcal/mol for G3 theory. Generally, the results from the two methods are similar, with some exceptions. G3//B3LYP theory gives significantly improved results for several cases for which MP2 theory is deficient for optimized geometries, such as CN and O2+. However, G3//B3LYP does poorly for ionization potentials that involve a Jahn–Teller distortion in the cation (CH4+, BF3+, BCl3+) because of the B3LYP/6-31G(d) geometries. The G3(MP2) method is also modified to use B3LYP/6-31G(d) geometries and zero-point energies. This variation, referred to as G3(MP2)//B3LYP, has an average absolute deviation of 1.25 kcal/mol compared to 1.30 kcal/mol for G3(MP2) theory. Thus, use of density functional geometries and zero-point energies in G3 and G3(MP2) theories is a useful alternative to MP2 geometries and HF zero-point energies.

No abstract

No abstract

Quantum scattering calculations on the SN2 reaction Cl − +CH 3 Br→ClCH 3 +Br − Some of the factors affecting the accuracy of following reaction paths and calculating projected frequencies perpendicular to the reaction path have been examined. The S N 2 reaction of Cl Ϫ with CH 3 Cl computed at the HF/6-31G* level of theory has been used as a test case. The symmetric C-H stretching mode couples strongly to the reaction path, and the projected frequency of this mode is very sensitive to the numerical accuracy of the path following and frequency projection methods. The transition state geometry must be converged very tightly so that the path steps in the correct direction. For second order implicit algorithms, improved accuracy can be obtained by computing the tangent used for path following and frequency projection from the displacement along the path rather than from the gradient. An even greater increase in accuracy can be achieved by employing the Hessian, used to compute the frequencies, to take a Newton-Raphson step to improve the convergence of the reaction path following. Taken together, these techniques yield a one to three order of magnitude decrease in the errors in the projected frequencies along the reaction path.