A survey of factors contributing to accurate theoretical predictions of atomization energies and molecular structures.

High level electronic structure predictions of thermochemical properties and molecular structure are capable of accuracy rivaling the very best experimental measurements as a result of rapid advances in hardware, software, and methodology. Despite the progress, real world limitations require practical approaches designed for handling general chemical systems that rely on composite strategies in which a single, intractable calculation is replaced by a series of smaller calculations. As typically implemented, these approaches produce a final, or "best," estimate that is constructed from one major component, fine-tuned by multiple corrections that are assumed to be additive. Though individually much smaller than the original, unmanageable computational problem, these corrections are nonetheless extremely costly. This study presents a survey of the widely varying magnitude of the most important components contributing to the atomization energies and structures of 106 small molecules. It combines large Gaussian basis sets and coupled cluster theory up to quadruple excitations for all systems. In selected cases, the effects of quintuple excitations and/or full configuration interaction were also considered. The availability of reliable experimental data for most of the molecules permits an expanded statistical analysis of the accuracy of the approach. In cases where reliable experimental information is currently unavailable, the present results are expected to provide some of the most accurate benchmark values available.