The largest eigenvalue of a convex function, duality, and a theorem of Slodkowski (1503.02231v2)

Abstract: First, we provide an exposition of a theorem due to Slodkowski regarding the largest "eigenvalue" of a convex function. In his work on the Dirichlet problem, Slodkowski introduces a generalized second-order derivative which for $C^2$ functions corresponds to the largest eigenvalue of the Hessian. The theorem allows one to extend an a.e lower bound on this largest "eigenvalue" to a bound holding everywhere. Via the Dirichlet duality theory of Harvey and Lawson, this result has been key to recent progress on the fully non-linear, elliptic Dirchlet problem. Second, we give a dual interpretation of this largest eigenvalue using the Legendre-Fenchel transform, and use this dual perspective to provide an alternative proof to an important step in the proof of the theorem.