Incorporating dynamic correlation into the variational determination method of the second‐order reduced density matrix in the doubly occupied configuration interaction space

• Published in: International journal of quantum chemistry Vol. 120; no. 15
• Main Authors: Alcoba, Diego R., Torre, Alicia, Lain, Luis, Oña, Ofelia B., Ríos, Elías, Massaccesi, Gustavo E.
• Format: Journal Article
• Language: English
• Published: Hoboken, USA John Wiley & Sons, Inc 05-08-2020; Wiley Subscription Services, Inc
• Subjects: Chemistry; Configuration interaction; Correlation analysis; Density functional theory; dynamic correlation; Electrons; pair density functional theory; Physical chemistry; Quantum physics; reduced density matrix; static correlation; variational methods; Wave functions
• ISSN: 0020-7608; 1097-461X
• DOI: 10.1002/qua.26256

The variational determination of the second‐order reduced density matrices arising from N‐electron doubly occupied configuration interaction wave functions has recently been proposed as a method to describe systems possessing static correlation in their ground states. In this work, we propose to combine this approach with the on‐top pair density functional theory in order to incorporate dynamic correlations and improve the description of systems possessing both types of electron correlation. The procedure requires ensuring that the second‐order reduced density matrix elements satisfy determined N‐representability variational conditions, as well as other constraints arising from the doubly‐occupied configuration interaction wave function features. An analysis of the results obtained through this method, in studies of strongly correlated many‐electron systems, shows the usefulness of our proposals.
A hybrid method, which combines variational determination of the second‐order reduced density matrix arising from double‐occupied configuration interaction wave functions with density functional theory procedures, allows us to efficiently account for static and dynamic electron correlations, requiring a moderate computational cost.