The energy of a many-particle system is not convex with respect to particle number for r−k interparticle repulsion potentials if k > log34 ≈ 1.262. With such potentials, some finite electronic systems have ionization potentials that are less than the electron affinity: they have negative band gap (chemical hardness). Although the energy may be a convex function of the number of electrons (for which k = 1), it suggests that finding an analytic proof of convexity will be very difficult. The bound on k is postulated to be tight. An apparent signature of non-convex behavior is that the Dyson orbital corresponding to the lowest-energy mode of electron attachment has a vanishingly small amplitude.
REFERENCES
The restriction to spinless fermions (e.g., electrons with the same spin) is probably not necessary because the large and divergent interparticle repulsion should prevent multiple particles from occupying the central site as the classical limit is approached. However, the author has not proved this. If it were not true, building a Hamiltonian for non-spin-free fermions would require a non-Coulomb on-site potential. (Specifically, an on-site potential that is more tightly localized to the site.)