Boundary regularity for the steady generalized Newtonian flow with shear thickening viscosity (Q831429)

The authors consider the Navier-Stokes problem \(-\nabla \cdot \sigma +u\cdot \nabla u=-\nabla p+f\), \(\operatorname{div} u=0\), posed in the half-space \(\Omega = \mathbb{R}_{+}^{3}\), where the deviatoric stress is defined by \(\sigma _{ij}=\left\vert D(u)\right\vert ^{q-2}D_{ij}(u)\), with \(D(u)=\frac{1}{2} (\nabla u+(\nabla u)^{T})\) and \(2<q<3\). Homogeneous Dirichlet boundary conditions \(u=0\) is imposed on \(\partial \Omega \). The authors prove boundary regularity for weak solutions to this problem. For \(f\in L^{3q/(4q-3)}(\Omega )\), they first define the notion of weak solution as a function \(u\in \widehat{W}_{0,\sigma }^{1,q}(\Omega )\) which satisfies a variational formulation, the notion of suitable weak solution if \(u\) further satisfies \( \int_{\Omega }\left\vert D(u)\right\vert ^{q}dx\leq \int_{\Omega }f\cdot udx\), and the notion of strong solution as a couple \((u,p)\in \widehat{W}_{0,\sigma }^{1,q}(\Omega )\times L_{loc}^{q^{\prime }}(\Omega )\) which further satisfies \(\sigma _{ij}\in W^{1,s}(\Omega )\) and \(D_{i}p\in L^{s}(\Omega )\) for some \(s\geq 1\). The main result proves that if \(f\in L^{3q/(4q-3)}(\Omega )\cap W^{1,q^{\prime }}(\Omega )\), every weak solution to the Navier-Stokes problem is strong, \(\nabla u\in L^{s}(\Omega )\), for every \(1\leq s<\frac{5q}{2}\), and \(D_{3}\sigma _{ij},D_{3}p\in L^{s}(\Omega ) \) for every \(1\leq s<\frac{2q^{\prime }}{3}\). This implies that \(u\) is Hölder continuous up to the boundary, that is \(u\in C^{\gamma }(\overline{ \Omega })\), for every \(0\leq \gamma <1-\frac{6}{5q}\). For the proof, the authors follow that they used in their previous paper [J. Differ. Equations 258, No. 11, 3811--3850 (2015; Zbl 1315.35046)], but they introduce a localized approximate system \(-\nabla \cdot \sigma ^{\varepsilon ,N}+\nabla p_{\varepsilon ,N}=-U_{N}\cdot \nabla (U_{N}+w_{\varepsilon ,N})+f\), \( \operatorname{div} w_{\varepsilon ,N}=0\), posed in \(U_{2r}^{+}=B_{2r}^{\prime }(x_{0})\times (0,2r)\), for \(x_{0}\in \{x\in \mathbb{R}^{3}:x_{3}=0\}\) and \( 0<r<\infty \). They also use a linearization of the non-linear term \(u\cdot \nabla u\). The homogeneous Dirichlet boundary condition \(w_{\varepsilon ,N}=0\) is imposed on \(\partial U_{2r}^{+}\). Here \(\sigma ^{\varepsilon ,N}=2\varepsilon D(u_{\varepsilon ,N})+\left\vert D(u_{\varepsilon ,N})\right\vert ^{q}-2D(u_{\varepsilon ,N})\), \(u_{\varepsilon ,N}=U_{N}+w_{\varepsilon ,N}\), and \(U_{N}\) is a mollification of \(u\). The authors prove estimates of the second gradient of \(u_{\varepsilon }\) in the tangential and normal directions and higher \(L^{p}\)-estimates of \(\nabla u\) using weighted Sobolev embeddings. They finally pass to the limits when \( \varepsilon \rightarrow 0\) and \(N\rightarrow \infty \).