Dynamical optimization for partition theory

How to partition a chemical system into its constituent parts is a classic problem of theoretical chemistry. A formally exact solution has recently been developed, partition theory (PT), based on density functional theory [Cohen, M. H.; Wasserman, A. J. Phys. Chem. A 2007, 111, 2229]. PT presents a constrained optimization problem to which the Car-Parrinello (CP) method of electronic structure theory is well suited. We propose here a generalization of the CP method suitable for PT and thereby make way for its practical numerical implementation. We demonstrate that this CP implementation of PT need not increase the complexity of the computation of the system's electronic structure. The scheme provides an exact DFT formulation of, e.g., atoms in molecules theory that is amenable to numerical implementation.