Exploiting a Shortcoming of Coupled-Cluster Theory: The Extent of non-Hermiticity as a Diagnostic Indicator of Computational Accuracy

The fundamental non-Hermitian nature of the forms of coupled-cluster (CC) theory widely used in quantum chemistry has usually been viewed as a negative, but the present letter shows how this can be used to advantage. Specifically, the non-symmetric nature of the reduced one-particle density matrix (in the molecular orbital basis) is advocated as a diagnostic indicator of computational quality. In the limit of full coupled-cluster theory (which is equivalent to full configuration interaction (FCI)), the electronic wavefunction and correlation energy are exact within a given one-particle basis set and the symmetric character of the exact density matrix is recovered. The extent of the density matrix asymmetry is shown to provide a measure of ``how difficult the problem is'' (like the well-known T$_1$ diagnostic), but its variation with level of theory also gives information about ``how well this particular method works'', irrespective of the difficulty of the problem at hand. The proposed diagnostic is described and applied to a select group of small molecules, and an example of its overall utility for the practicing quantum chemist is illustrated through its application to the beryllium dimer (Be$_2$). Future applications of this idea to excited states, open-shell systems, symmetry-breaking problems and extension of the method to the two-particle density are then proposed.