On the Stokes and Navier-Stokes equations in a perturbed half-space (Q2504050)

The authors study the Stokes- and Navier-Stokes equations on a finitely perturbed half space \(\Omega\). That is, given the halfspace \(H=\{(x_1,\dots,x_n)\in\mathbb R^n,\) \(x_n>0\}\) there is a sphere \(B_R=\{|x|\leq R, x\in \mathbb R^n\}\) such that \(\Omega-B_r=H-B_r\), i.e. \(\Omega\) differs from \(H\) by a finite smooth bulge. The equations studied are \[ u_t-\Delta u+\nabla\pi=0\text{ on }\Omega\times (0,\infty), \tag{1} \] \[ u_t-\Delta u+(u\nabla)u+\nabla\pi=0\tag{2} \] subject to the conditions \[ \text{div\,}u=0;\quad u(x,t)=0\text{ on }\partial\Omega\times (0,\infty); \quad u(x,0)=a(x)\text{ on }\Omega.\tag{3} \] Function spaces are defined as usual: (a) \(J^p(\Omega)\) is the \(L^p(\Omega)\)-closure of \(u\in C^\infty_0(\Omega)^n\), \(\text{div\,}u=0\), while \(G^p(\Omega)=\{\nabla v\in L^p(\Omega)^n, \;v\in L^p_{\text{loc}}(\overline \Omega)\}\). Based on the well known Helmholtz decomposition \(L^p(\Omega)^n=J^p(\Omega)\oplus G^p(\Omega)\) with associated projection \(P\) onto \(J^p(\Omega)\) (Farwig, Sohr) the authors first investigate the holomorphic semigroup \(T(t)\), \(t\geq 0\) generated by the Stokes operator \(P_\Delta=-A\). They prove a number of results on the behaviour of \(T(t)\), among others, the important Thm. 1.2, which expresses \(L^p\)-\(L^q\) estimates. It states that for all \(t>0\), \(f\in J^p(\Omega)\) and \(1\leq p\leq q\leq \infty\) \((p\neq \infty,q\neq 1)\) one has \[ \|T(t)\,f\|_{L^q(\Omega)^n}\leq C_{pq}t^{-\frac n2(\frac1p-\frac1q)}\|f\|_{L^p(\Omega)^n}.\tag{4} \] Similar estimates hold for \(\|\nabla T(t)\|_{L^q(\Omega)^{n^2}}\). The authors now apply these results to the Navier-Stokes equations (2)+(3) so as to get results about the decay of global solutions induced by small initial data. Thm. 1.3 asserts that there is \(\delta=\delta(\Omega,n)>0\) as follows: if \(a\in J^n(\Omega)\) satisfies \(\|a\|_{L^n}\leq \delta\) then eq. (2)+(3) admits a unique strong solution \(u(t)\), \(t\geq 0\) with \(u(0)=a\) and such that \[ \begin{aligned} & \|u(t)\|_{L^r}=o\left(t^{-\frac12+\frac n{2r}}\right),\\ & \|\nabla u(t)\|_{L^r}=o\left(t^{-\frac12+\frac n{2r}}\right),\quad n\leq r<\infty.\end{aligned}\tag{5} \] The proofs, too involved to be discussed here, rely on earlier results of the authors and others (Farwig, Sohr) concerning the solution of the Stokes problem \[ (\lambda-\Delta)u+\nabla\pi=f,\quad\text{div}\,u=0,\quad u=0\text{ on }\partial\Omega,\text{ with solution }R(\lambda)f=u,\quad \Pi(\lambda)=\pi.\tag{6} \] A remarkable feature of the results obtained by the authors is that they are better than the corresponding results in exterior domains.