Browsing by Subject "Coupled cluster"
Now showing 1 - 2 of 2
Results Per Page
Sort Options
Item High-accuracy ab initio thermochemistry : application to hydrocarbons(2013-08) Ferguson, Michael Eric; Stanton, John (John F.)This work focuses on an examination of the high-accuracy extrapoloated ab initio thermochemistry (HEAT) protocol of determining molecular atomization energies. The HEAT protocol does not utilize experimental data or empirical scaling effects. The accuracy of the approach is tested via comparison to ATcT data, and all molecules fall within 1 kcal/mol of accepted values. There are several important points to note about this treatment: namely, that we have used atomic natural orbital (ANO) basis sets for the calculation of the zero point energy and that we have made determinations for larger molecules than previously done with HEAT. The molecules in this paper were chosen to provide benchmark numbers for the homodesmotic reaction heirarchy as described by Wheeler et al.[3] The relative accuracy of the approach is considered, as well as a discussion of possible remaining sources of error.Item Non-orthogonal spin-adaptation and application to coupled cluster up to quadruple excitations(2014-08) Matthews, Devin Alexander; Stanton, John (John F.)The theory of non-orthogonal spin-adaptation for closed-shell molecular systems is presented, with an emphasis on application to the coupled cluster family of electronic structure methods. To aid in the derivation of efficient and compact working equations, a new diagrammatic interpretation of the Goldstone diagrams is derived which only requires a small number of the many distinct diagrams and which directly produces equations in a factored form in terms of “spin-summed” tensor elements. This diagrammatic interpretation is applied to coupled cluster methods with quadruple excitations (CCSDTQ), including coupled cluster with a perturbative correction for quadruple excitations (CCSDT(Q)) and to CCSDTQ gradients and properties. The advantages of the non-orthogonal spin-adaption with respect to simplification and factorization of the working equations and to efficient implementation are presented and discussed. Additionally, specific optimizations of the implementation for often-overlooked issues such as tensor transposition, disk access, and removal of redundant and/or unnecessary operations are detailed. The resulting algorithm is implemented for the CCSDTQ and CCSDT(Q) methods and compared to existing codes, where a one to two order-of-magnitude improvement in efficiency is observed. The new implementation is also used for calculations on several larger molecular systems to illustrate the scalability of the method.