Patton L. Fast

Multi-coefficient Gaussian-3 method for calculating potential energy surfaces

Abstract We propose a multi-coefficient modification (MCG3) of the Gaussian-3 (G3) electronic structure method that is suitable for calculating continuous potential energy surfaces. We tested it for atomization energies and found that it improves the accuracy by 8% as compared to G3 and reduces the cost of single-point energy calculations by 50%. The method should be useful for calculating bond energies, potential energy surfaces, and thermochemical data of molecules.

Infinite basis limits in electronic structure theory

We have developed a database of 29 molecules for which we have estimated the complete-one-electron-basis-set limit of the zero-point-exclusive atomization energy for five levels of electronic structure theory: Hartree–Fock (HF) theory, Mo/ller–Plesset second- and fourth-order perturbation theory, coupled cluster theory based on single and double excitations (CCSD), and CCSD plus a quasiperturbative treatment of triple excitations [CCSD(T)], all at a single set of standard geometries. Convergence checks indicate that the estimates are within a few tenths of a kcal/mol of the n=infinity limit of the cc-pVnZ basis set sequence. This data is then used to obtain optimized power-law exponents for extrapolating to the basis-set-limit from correlation-consistent polarized valence double and triple zeta (cc-pVDZ and cc-pVTZ) basis sets. This allows one to get thermochemical accuracy comparable to polarized quadruple or quintuple zeta (cc-pVQZ or cc-pV5Z) basis sets with a cost very comparable to polarized triple z...

The Gaussian-2 method with proper dissociation, improved accuracy, and less cost

The Gaussian-2 method (G2) is modified by deleting the empirical high-level correction and instead using empirical coefficients to extrapolate to full configuration interaction and an infinite basis set. The resulting method, called multicoefficient Gaussian-2 (MCG2) is less expensive than G2 but a factor of 1.7 more accurate for molecules composed of H and first-row atoms.

VARIATIONAL REACTION PATH ALGORITHM

In this paper we propose a new algorithm for calculating a reaction path and a set of local vibrational frequencies along a reaction path for dynamics calculations. The new method yields reasonable vibrational frequencies even when using a large step size. The algorithm is tested by carrying out variational transition state theory calculations including multidimensional semiclassical tunneling contributions, for the reaction OH+H2→H2O+H, and the results are very promising.

Thermochemical analysis of core correlation and scalar relativistic effects on molecular atomization energies

Core correlation and scalar relativistic contributions to the atomization energy of 120 first- and second-row molecules have been determined using coupled cluster and averaged coupled-pair functional methods and the MTsmall core correlation basis set. These results are used to parametrize an improved version of a previously proposed bond order scheme for estimating contributions to atomization energies. The resulting model, which requires negligible computational effort, reproduces the computed core correlation contributions with 88%–94% average accuracy (depending on the type of molecule), and the scalar relativistic contribution with 82%–89% accuracy. This permits high-accuracy thermochemical calculations at greatly reduced computational cost.

Multilevel geometry optimization

Geometry optimization has been carried out for three test molecules using six multilevel electronic structure methods, in particular Gaussian-2, Gaussian-3, multicoefficient G2, multicoefficient G3, and two multicoefficient correlation methods based on correlation-consistent basis sets. In the Gaussian-2 and Gaussian-3 methods, various levels are added and subtracted with unit coefficients, whereas the multicoefficient Gaussian-x methods involve noninteger parameters as coefficients. The multilevel optimizations drop the average error in the geometry (averaged over the 18 cases) by a factor of about two when compared to the single most expensive component of a given multilevel calculation, and in all 18 cases the accuracy of the atomization energy for the three test molecules improves; with an average improvement of 16.7 kcal/mol.

The calculation of kinetic isotope effects based on a single reaction path

In this paper we propose a new method for calculating kinetic isotope effects without calculating a separate reaction path for each isotopically substituted species. The new method yields reasonable kinetic isotope effects from calculations using the same reaction path for all isotopic variations. The method is tested by carrying out variational transition state theory calculations, including multidimensional tunneling contributions, for the reactions OH+H2→H2O+H, CH3+H2→CH4+H, and H+H2→H2+H and nine deuterium-substituted isotopologs of these reactions. The results are very encouraging.

Improved coefficients for the scaling all correlation and multi-coefficient correlation methods

We have re-optimized the coefficients for ten scaling all correlation (SAC) methods, five empirical infinite basis (EIB) methods, and 18 multi-coefficient (MC) correlation methods, including the special cases of multi-coefficient SAC and multi-coefficient Gaussian-2 and Gaussian-3. The new parameterization is based on a training set of 82 atomization energies except for multi-coefficient Gaussian-2, which is restricted to H and the first period (nuclear charge ( 9) and is based on a training set of 52 atomization energies. Each method may be employed with or without including core-correlation effects, which are based on a new set of parameters optimized on a 125-molecule training set. The mean unsigned error in the atomization energies of the 82-molecule set is reduced on average by 20% when the new parameters used here are adopted.