Global invertibility of Sobolev maps (Q831231)

Let \(\Omega\subset{\mathbb R}^n\) an open bounded set, \(p>n-1\). The authors consider the class of functions \({\boldsymbol u}\in W^{1,p}(\Omega,{\mathbb R}^n)\cap W^{1,p}(\partial\Omega,{\mathbb R}^n)\) such that \(\det D\boldsymbol{u}\in L^1(\Omega)\). Here \(\det D\boldsymbol{u}\) denotes the pointwise determinant of \(D\boldsymbol{u}\), and \(\det D\boldsymbol{u}\) its distributional determinant. The condition \[ \int_{\partial\Omega}\phi({\boldsymbol x})(\textrm{adj}D {\boldsymbol u}({\boldsymbol x}){\boldsymbol g}({\boldsymbol u}({\boldsymbol x})))\cdot {\boldsymbol \nu}({\boldsymbol x})d{\mathcal H}^{n-1}({\boldsymbol x})- \int_\Omega [\textrm{adj}D {\boldsymbol u}({\boldsymbol x}){\boldsymbol g}({\boldsymbol u}({\boldsymbol x}))]\cdot D\phi({\boldsymbol x})d{\boldsymbol x}=\] \[ =\int_D\det D\boldsymbol{u}(\boldsymbol{x})\phi(\boldsymbol{x})\operatorname{div}\boldsymbol{g}(\boldsymbol{u}(\boldsymbol{x})) d\boldsymbol{x} \tag{1}\] for all \(\boldsymbol{g}\in C_c^1({\mathbb R}^n,{\mathbb R}^n)\) and \(\phi\in C^\infty(\overline\Omega),\) extends the equality between the pointwise determinant and the distributional determinant of \(D\boldsymbol{u}.\) Notice that if \(\phi\in C_c^\infty(\Omega)\) then the first term in (1) vanishes. If \(\bar{\mathcal A}_p(\Omega)\) denotes the class of functions \({\boldsymbol u}\in W^{1,p}(\Omega,{\mathbb R}^n)\cap W^{1,p}(\partial\Omega,{\mathbb R}^n)\) such that \(det D\boldsymbol{u}\in L^1(\Omega)\) and for which condition (1) holds, the main result of the paper (Theorem 1.1) states the following global invertibility result: \textbf{Theorem}: Let \(p>n-1,\) and let \(\boldsymbol{u},\boldsymbol{u}_0 \in\bar{\mathcal{A}}_p(\Omega)\) be such that \(\boldsymbol{u}\vert_{\partial\Omega}=\boldsymbol{u}_0\vert_{\partial\Omega}.\) Assume that \(\det D\boldsymbol{u} > 0\) a.e., \(\det D\boldsymbol{u}_0 \geq 0\) a.e. and \(\boldsymbol{u}_0\) is injective a.e. Then \(\boldsymbol{u}\) is injective a.e. Using this global invertibility result, the authors prove under suitable assumptions the existence of the minimizers for the typical hyperelasticity functional \[ {\boldsymbol u}\mapsto \int_\Omega W({\boldsymbol x},{\boldsymbol u}({\boldsymbol x}),D{\boldsymbol u}({\boldsymbol x})) d{\boldsymbol x}. \]