A note on singular perturbation problems via Aubry-Mather theory (Q2372749)

The authors consider the singular perturbation problem \[ -\varepsilon\Delta v_{\varepsilon}+H(x,\nabla v_{\varepsilon})-\varepsilon c(x)=0~~\text{in }D,\quad v_{\varepsilon}(x)=g(x)\text{ on } \partial D,\tag{1} \] where \(D\) is a bounded set in \(\mathbb R^n\) with Lipschitz boundary, \(H:(x,p)\in D \times\mathbb R^n \to \mathbb R\) is continuous, Lipschitz continuous in \(x,\) convex and coercive in \(p,\) \(H(x,0)\leq 0,\) \(c:\overline{D} \to \mathbb R\) is continuous and nonnegative and \(g\) is a given function. The critical value of \(H\) is assumed to be zero. Then, the limit problem \[ H(x,\nabla v)=0 \quad \text{in } D,\quad v(x) =g(x) \text{ on } \partial D\tag{2} \] has, in general, many viscosity solutions. The paper deals with the behavior of \(v_{\varepsilon},\) as \(\varepsilon \to 0\). Following \textit{A. Fathi} and \textit{A. Siconolfi} [Calc. Var. Partial Differ. Equ. 22, 185--228 (2005; Zbl 1065.35092)], let, for \(x\in \overline{D},\) \(\sigma (x,\overset{.}{~})\) be the support function of the set \(Z(x)=\{p\in \mathbb R^n| H(x,p)\leq 0\}.\) For \(x,y\in\overline{D}\), consider the nonsymmetric distance \(S(x,y)=\text{inf}\{ \int_0^1 \sigma(\Phi(s),\overset{.}{\Phi}(s))\,ds, \Phi\in \mathcal C_{x,y}\},\) where \(\mathcal C_{x,y}\) is the set of \(W^{1,\infty}\) curves , with \(\Phi(0)=x \) and \(\Phi(1)=y.\) Denote by \(\mathcal A\) the associated Aubry set. It is assumed that \(g\) is a real function, defined and continuous on \(\partial D\cup \mathcal A,\) such that \(-S(y,x)\leq g(x)-g(y)\leq S(y,x)\) for \(x,y\in\partial D\cup\mathcal A\). The main result is as follows: under some additional assumptions on \(\mathcal A\) and if \(| \nabla v_{\varepsilon}| _{\infty} \) is bounded by a constant, independent of \(\varepsilon\), then \(v_{\varepsilon}\), solution to (1), converges uniformly to the maximal viscosity solution \(G(x)=\min_{y\in\partial D}\{g(y)+S(y,x)\}\) to (2), as \(\varepsilon \to 0.\) Several examples of Hamiltonians, for which the main result holds, are discussed.