Per-Åke Malmqvist

The well-known natural orbitals are defined as eigenfunctions of a one-particle reduced density operator, and can be obtained from a computed density matrix by diagonalization. Similarly, in this article we define the binatural orbitals, which are obtained for a pair of wave functions by a singular value decomposition of a reduced transition density matrix. The pair of states would usually be eigenstates of the electronic Hamiltonian, and the binatural orbitals then serve as a useful tool for the analysis of the transition between these states. More generally, application to any two state functions gives important information as to how the two states differ. Some examples are shown.