Molecular equilibrium geometries based on coupled-cluster calculations including quadruple excitations

Using analytic gradient techniques and an additivity scheme for the various electron correlation contributions, i.e. core-correlation, contribution due to full treatment of triple excitations and contributions due to quadruple excitations calculated with different basis sets, the accuracy of computed geometrical parameters are analysed in comparison with experiment. For a test set of 12 closed-shell and 5 open-shell molecules, it is found that inclusion of quadruple excitations is essential to reach agreement with experiment. The mean error of 0.002 pm and the standard deviation of 0.040 pm of the present CCSD(T)/cc-pV6Z + core(CCSD(T)/cc-pCVQZ) + T/cc-pVTZ + Q/cc-pVDZ results for the closed-shell systems underline the importance of quadruple excitations, in particular, as corresponding calculations without quadruple excitations exhibit significantly larger error. Quadruples contributions for multiply bonded systems as well as the challenging F2 molecule are as large as 0.1 to 0.3 pm, while for single bonds the effects are typically significantly smaller than 0.1 pm.