Convexity of trace functionals and Schrödinger operators (Q2491612)

The paper is concerned with the convexity of certain trace functionals with applications in quantum mechanics. They are of the form \[ \Phi(U)=\text{tr\,}(F(H+U-\varepsilon(U)))-\varepsilon(U), \] where \(U\) is a bounded, self-adjoint operator and \( \varepsilon(U) \) is a normalizing real function -- the Fermi level. In applications, \(H+U\) is a Hamiltonian with kinetic energy part \(H\) and potential energy part \(U\). Let \(f\) be a strictly decreasing thermodynamic equilibrium distribution function for the quantum system. Then the Fermi level \(\varepsilon(U)\) is defined such that \(f(H+U-\varepsilon(U))\) is a state (density matrix) and the function \(F\) is defined by setting \[ F(t)=\int_t^\infty f(s)\,ds. \] In simple cases, it is well-known that \(\Phi\) is convex. The paper extends this convexity result to unbounded operators \(H\) in a quite general setting.