Abstract
We show that following the standard mantra of quantum chemistry and diagonalizing the Born-Oppenheimer (BO) Hamiltonian H_ BO(R) is not the optimal means to construct potential energy surfaces. A better approach is to diagonalize a phase-space electronic Hamiltonian, H_ PS(R,P), which is parameterized by both nuclear position R and nuclear momentum P. The foundation of such a non-perturbative phase-space electronic Hamiltonian can be made rigorous using a partial Wigner transform and the method has exactly the same cost as BO for a semiclassical calculation (and only a slight increase in cost for a quantum nuclear calculation). For a three-particle system, with two heavy particles and one light particle, numerical results show that a phase space electronic Hamiltonian produces not only meaningful electronic momenta (which are completely ignored by BO theory) but also far better vibrational energies. As such, for high level results and/or systems with degeneracies and spin degrees of freedom, we anticipate that future electronic structure and quantum chemistry packages will need to take as input not just the positions of the nuclei but also their momenta.