Small noise limit and convexity for generalized incompressible flows, Schrödinger problems, and optimal transport (Q2296933): Difference between revisions

The authors consider six optimization problems: the quadratic Monge-Kantorovich optimal transport problem, the Schrödinger problem, a Brenier relaxed model for incompressible fluids, a Brödinger problem for incompressible viscid fluids, the multiphase Brenier model and the multiphase Brödinger problem, which all consist in minimizing a kinetic action and/or a relative entropy of some path measures with respect to the reversible Brownian motion. The purpose of the paper is to establish \(\Gamma\)-convergence relations between the corresponding problems, as the viscosity parameter \(\nu\) goes to 0, and to prove the convergence of the associated pressures arising from the incompressibility constraints. The authors first present the six problems and some properties of their solutions. The first main result proves the convergence of the Brödinger problem towards the Brenier model as the diffusivity \(\nu \) tends to 0. The key tools of the proof are a Cameron-Martin formula and a narrow continuity of the optimal action in Brenier's model proposed by the first author in [Calc. Var. Partial Differ. Equ. 58, No. 1, Paper No. 25, 22 p. (2019; Zbl 1484.35029)]. The second main result gives a new proof of the convergence of the Schrödinger problem to the optimal transport one, which is based on an explicit PDE construction of recovery sequences. The third main result proves the convergence of the multiphase Brödinger problem towards the multiphase Brenier model and the convergence of the associated pressures. In the last part of their paper, the authors present new results concerning the time-convexity of the entropy along solutions.