Hypersurfaces of constant curvature in hyperbolic space II (Q973865)

This is the second of a series of papers in which the authors aim at studying complete manifolds of constant curvature and prescribed asymptotic boundary at infinity, like hypersurfaces \(\Sigma\subset {\mathbb H}^{n+1}\), where \[ {\mathbb H}^{n+1}\equiv\{(x^\alpha)\equiv(x^k,u)\in{\mathbb R}^{n+1},\hskip 3pt u>0\}, \] equipped with an hyperbolic metric. [For part I see the authors paper in J. Geom. Anal. 19. No.\,4, 772--795 (2009; Zbl 1184.53066).] In this paper, they focus on graphs over a domain with nonnegative mean curvature. The main result states that there exists a complete hypersurface \(\Sigma\subset {\mathbb H}^{n+1}\) satisfying the following conditions: {\parindent=8mm \begin{itemize}\item[(i)] \(f(\kappa[\Sigma])=\sigma\), \(\partial\Sigma=\Gamma\), where: (a) \(\kappa[\Sigma]=(\kappa_1,\cdots,\kappa_n)\in K\subset{\mathbb R}^n\) denotes the hyperbolic principal curvatures of \(\Sigma\) and \(\sigma\in(0,1)\) is a constant; (b) \(f\) is a smooth symmetric function of \(n\) variables; (c) \(\Gamma\equiv\partial\Omega\times\{0\}\subset{\mathbb R}^{n+1}\) with \(\Omega\) a bounded smooth domain in \({\mathbb R}^n\); (d) \(|\kappa[\Sigma]|\leq C\) on \(\Sigma\). \item[(ii)] The mean curvature \({\mathcal H}_{\partial\Omega}\) is nonnegative and \(\sigma\in(0,1)\) satisfies \(\sigma > \sigma_0\) where \(\sigma_0\) is the unique zero in \((0,1)\) of the function \(\phi(a)\equiv {{8}\over{3}}a+{{22}\over{27}}a^3-{{5}\over{27}}(a^2+3)^{3/2}\). \item[(iii)] The function \(f\) satisfies the following constraints: (e) \((\partial\lambda_i.f)(\lambda)>0\) in \(K\), \(1\leq i\leq n\), where \(K\) is an open symmetric convex cone such that \(K^+\equiv\{\lambda\in{\mathbb R}^n | \forall \lambda_i>0\}\subset K\); (f) \(f(\lambda)>0\), \(\lambda\in K\) ; (g) \(f(\lambda)=0\), \(\lambda\in\partial K\); (h) \(f\) is homogeneous of degree \(1\); (l) \(\mathop{\lim}\limits_{R\to\infty}f(\lambda_1,\dots,\lambda_{n-1},\lambda_n+R)\geq 1+\epsilon_0\), uniformly in \(B_{\delta_0}({\mathbf 1})\), for some \(\epsilon_0>0\) and \(\delta_0>0\), where \(B_{\delta_0}({\mathbf 1})\) is the ball of radius \(\delta_0\) centered at \(\mathbf{1}\equiv(1,\dots,1)\in{\mathbb R}^n\). \item[(iv)] \(\Sigma\) is the graph of a unique admissible solution \(u\in C^\infty(\Omega)\bigcap C^1(\overline{\Omega})\) of the Dirichlet problem in (1). \[ \begin{aligned} \left\{ \begin{aligned} &G(D^2u,Du,u)=\sigma\\ &u(x)>0,\;x\in\Omega\subset{\mathbb R}^n\\ & u(x)=0,\;x\in\partial\Omega \end{aligned} \right\}\;+\;&\text{additional constraints}: \\ &\left\{ \begin{aligned} &u^2\in C^\infty(\Omega)\bigcap C^{1,1}(\overline{\Omega})\\ &\sqrt{1+|Du|^2}\leq {{1}\over{\sigma}},\;x\in\Omega\\ &u|D^2u|\leq C,\;x\in\Omega\\ &\sqrt{1+|Du|^2}={{1}\over{\sigma}},\;x\in\partial\Omega. \end{aligned} \right\}. \end{aligned} \tag{1} \] (For the definition of the nonlinear second order differential operator \(G\), see the paper.) Furthermore, for any \(\sigma\in(0,1)\) and \(\epsilon>0\) sufficiently small, there exists a unique admissible solution \(u^{\epsilon}\in C^\infty(\overline{\Omega})\) such that conditions in (2) are satisfied. \[ \left\{ \begin{aligned} &u^\epsilon(x)-\epsilon=0,\;x\in\partial\Omega\\ &\sqrt{1+|Du^{\epsilon}|^2}\leq {{1}\over{\sigma}},\;x\in\Omega\\ &u^{\epsilon}|D^2u^{\epsilon}|\leq C/\epsilon^2,\;x\in\Omega \end{aligned} \right\}, \tag{2} \] where \(C\) is a constant independent of \(\epsilon\). \end{itemize}} After a detailed introduction the paper splits in five more sections. 2. Vertical graphs and the linearized operator. 3. Height estimates and asymptotic angle condition. 4. The approximating problems and the continuity method. 5. Boundary estimates for second derivatives. 6. Global estimates for second derivatives. Remark. It is interesting to recast the problem considered in this paper in the framework of the geometric theory of PDEs. In fact, equation \(G-\sigma=0\) in (1) can be considered an analytic PDE of the second order \(E_2\subset J{D}^2(W)\) on the analytic fiber bundle \(\pi:W\equiv{\mathbb R}^{n+1}\to{\mathbb R}^n\), \((x^k,u)\mapsto(x^k)\), \(1\leq k\leq n\), that identifies a completely integrable PDE. However, the Dirichlet problem, considered in (1), does not admit regular solutions in \(E_2\). Therefore, in order to find solutions, one must enlarge the point of view and consider the singular PDE \({}^SE_2\equiv E_2\cup S_2\subset J{D}^2(W)\), where \(S_2\subset J{D}^2(W)\) is the analytic submanifold of \(J{D}^2(W)\), \((x^k,u,u_{i},u_{ij})\), identified by the constraint \(u=0\), and endowed with the Cartan distribution of \({\mathbf E}_2(W)|_{S_2}\), where \({\mathbf E}_2(W)\) is the Cartan distribution of \(J{ D}^2(W)\). Thus, even if there are not regular solutions \(V\subset E_2\), solving problem (1), we can find singular solutions \({}^SV\subset{}^SE_2\), with \({}^\epsilon V\equiv{}^SV\cap E_2\neq\emptyset\), the asymptotic part of such solutions just contained in \(E_2\). (Compare with (2).) Then, complete solutions \({}^CV\) are \({}^CV\equiv\overline{{}^\epsilon V}\), i.e., the closure of \({}^\epsilon V\) in \({}^SV\). (For details on this geometric approach, introduced by the reviewer of this paper, to solve nonlinear singular boundary value problems, see \url{arXiv:1006.4483[math.GT]} and references quoted there.)