Subsystem constraints in variational second order density matrix optimization: curing the dissociative behavior

TL;DR

This paper introduces subsystem constraints in variational second-order density matrix optimization to address dissociation issues, successfully improving results for diatomic molecules with minimal extra computational effort.

Contribution

The paper proposes a new class of N-representability conditions called subsystem constraints that fix dissociation problems in variational density matrix theory.

Findings

Subsystem constraints cure dissociation issues in diatomic molecules.

Application to BeB+ demonstrates improved potential energy surfaces.

Method extends to polyatomic molecules with additional subsystem options.

Abstract

A previous study of diatomic molecules revealed that variational second-order density matrix theory has serious problems in the dissociation limit when the N-representability is imposed at the level of the usual two-index (P, Q, G) or even three-index (T1, T2) conditions [H. van Aggelen et al., Phys. Chem. Chem. Phys. 11, 5558 (2009)]. Heteronuclear molecules tend to dissociate into fractionally charged atoms. In this paper we introduce a general class of N-representability conditions, called subsystem constraints, and show that they cure the dissociation problem at little additional computational cost. As a numerical example the singlet potential energy surface of BeB+ is studied. The extension to polyatomic molecules, where more subsystem choices can be identified, is also discussed.