Boundary \(C^{2, \alpha}\) estimates for Monge-Ampère type equations (Q2354378): Difference between revisions

Similarly to the optimal transportation, the authors consider the potential function \(u\) satisfying the Dirichlet problem of the Monge-Ampére type PDE \(\det \left( D^{2}u-A\left( \cdot ,Du\right) \right) =f\) \(\;\)in \(\Omega \) and \(u|_{\partial \Omega }=u_{0}\). The matrix-valued function \(A\) is defined by a smooth real-valued function \(c\left( x,y\right) \), \(x,y\in\mathbb{R}^{n}\) (an analogue of the cost function): \(A\left( x,Du\right) =D_{x}^{2}c\left( x,Tx\right) \), \(x\in \Omega \), where \(T\left( x\right) =D_{x}^{-1}c\left( x,Du\left( x\right) \right) \). The solution \(u\) is understood in the viscosity sense. The authors make the following assumptions (similar to the standard assumptions for the cost function \(c\) in optimal transportation): {\parindent=6mm \begin{itemize} \item[(H1)] Given \(x,p\in\mathbb{R}^{n}\), there is a unique \(y\in\mathbb{R}^{n}\) such that \(D_{x}c\left( x,y\right) =p\) and also, given \(y,q\in\mathbb{R}^{n}\), there is a unique \(x\in\mathbb{R}^{n}\) such that \(D_{y}c\left( x,y\right) =q\). \item [(H2)] \(D_{p_{k}p_{l}}^{2}A_{ij}\left( x,p\right) \xi _{i}\xi _{j}\eta _{k}\eta _{l}\geq c_{0}\left| \xi \right| ^{2}\left| \eta \right| ^{2}\) (\(c_{0}>0\)) for all \(\left( x,p\right) \in \Omega \times\mathbb{R}^{n}\) and two vectors \(\xi ,\eta \in\mathbb{R}^{n}\) such that \(\xi \perp \eta \). \end{itemize}} The domain \(\Omega \subseteq\mathbb{R}^{n}\) is assumed to be uniformly \(c\)-convex with \(C^{3}\)-smooth boundary. The main result (Theorem 1.1) states: If \(u_{0}\in C^{3}\left( \partial \Omega \right) ,\) \(f\in C^{a}\left( \overline{\Omega }\right) \) for some \( \alpha \in \left( 0,1\right) \) and \(0<\lambda \leq f\leq \Lambda \), then \( \left\| u\right\| _{C^{2,\alpha }\left( \overline{\Omega }\right) }\leq C\left( n,\alpha ,c,f,u_{0},\partial \Omega \right) \).