Search papers, labs, and topics across Lattice.
This paper rigorously compares the Born-Oppenheimer, Born-Huang, and exact factorization approaches to potential energy surfaces using two analytically transparent benchmark models. It establishes closed-form expressions for the molecular spectra and demonstrates the ground-state energy ordering across different mass ratios and coupling strengths, while revealing that this ordering does not extend to excited states. The authors also identify the diagonal Born-Huang correction with the mass-weighted quantum metric, providing a geometric error budget that highlights the limitations of these models in accurately capturing electronic-state changes.
The ground-state energy ordering derived from different potential energy surface constructions reveals critical insights into the limitations of traditional quantum mechanical approximations.
The phrase "potential energy surface" refers to several distinct objects. The Born-Oppenheimer construction gives a clamped-nucleus electronic eigenvalue, the single-surface Born-Huang construction adds the diagonal correction, and exact factorization gives a state-dependent exact scalar potential. These constructions answer different questions and should not be regarded as competing definitions of one universal surface. We compare them in two analytically transparent benchmark models. For Fern谩ndez's bilinearly coupled oscillators, the exact molecular spectrum, the Born-Oppenheimer and Born-Huang spectra, and the ground-state exact-factorization surface are obtained in closed form. We prove the ground-state ordering of these energies for every admissible mass ratio and coupling and show why it does not extend uniformly to excited states. The diagonal Born-Huang correction is identified with the mass-weighted quantum metric, and Fern谩ndez's sixth-order result is recast as a leading geometric error budget involving the metric and a gap-weighted spectral moment of the same derivative couplings. In a linear vibronic-coupling model the metric localizes at an avoided crossing while the total Fubini-Study length remains $蟺/2$, separating the localization of electronic-state change from its total magnitude. Exact factorization is smooth for the nodeless ground state but becomes increasingly ill-conditioned, while remaining finite, when an excited-state nuclear marginal becomes small. These models separate approximation error, geometric correction, and conditioning in a form that can be checked directly.
