Stochastic Homogenization of the Hamilton-Jacobi Equation on Manifolds

Abstract: This article establishes a stochastic homogenization result for the first order Hamilton-Jacobi equation on a Riemannian manifold $M$, in the context of a stationary ergodic random environment. The setting involves a finitely generated abelian group $ \mathtt{G}$ of rank $b$ acting on $M$ by isometries in a free, totally discontinuous, and co-compact manner, and a family of Hamiltonians $H: T^*M \times \Omega \to \mathbb{R}$, parametrized over a probability space $(\Omega, \mathbb{P})$, which are stationary with respect to a $\mathbb{P}$-ergodic action of $\mathtt{G}$ on $\Omega$. Under standard assumptions, including strict convexity and coercivity in the momentum variable, we prove that as the scaling parameter $\varepsilon$ goes to $0$, the viscosity solutions to the rescaled equation converge almost surely and locally uniformly to the solution to a deterministic homogenized Hamilton-Jacobi equation posed on $\mathbb{R}^b$, which corresponds to the asymptotic cone of $\mathtt{G}$. In particular, this approach sheds light on the relation between the limit problem, the limit space, and the complexity of the acting group. The classical periodic case corresponds to a randomness set $\Omega$ that reduces to a singleton; other interesting examples of this setting are also described. We remark that the effective Hamiltonian $\overline{H}$ is obtained as the convex conjugate of an effective Lagrangian $\overline{L}$, which generalizes Mather's $\beta$-function to the stochastic setting; this represents a first step towards the development of a stationary-ergodic version of Aubry-Mather theory. As a geometric application, we introduce a notion of stable-like norm for stationary ergodic families of Riemannian metrics on $M$, which generalizes the classical Federer-Gromov's stable norm for closed manifolds.