Potential theory on Sierpiński carpets. With applications to uniformization (Q2192230)

The goal of this book is to establish a uniformization result for planar Sierpiński carpets. In this purpose it continues a tradition of research in the uniformization of metric spaces, as the author expounds in the seven-page introduction (Chapter 1). The necessary background is prepared in Chapter 2. Let \(\Omega\subset {\mathbb C}\) be an open Jordan region, \((Q_i)_{i\in {\mathbb N}}\) a family of open Jordan regions compactly contained in \(\Omega\) with disjoint closures. If \(S:=\bar{\Omega}\setminus\cup_{i\in {\mathbb N}}Q_i\) has empty interior and is locally connected, it is called a \textit{Sierpiński carpet}, the \(Q_i\)'s its \textit{peripheral disks}. A key notion for the study is that of a \textit{Sobolev function} on \(S\). Its definition applies the concept of an \textit{upper gradient} for a function. \par Let \(g:S\cap\Omega\rightarrow {\mathbb R}\cup\{-\infty,\infty\}\). A nonnegative sequence \((\rho(Q_i))_{i\in {\mathbb N}}\) is called an \textit{upper gradient} for \(g\) if there exists a family \(\Gamma_0\) of paths in \(\Omega\) with \(\text{mod}(\Gamma_0)=0\) such that for all paths \(\gamma\not\in\Gamma_0\) in \(\Omega\) and \(x, y\in\gamma\cap S\) it holds \(g(x), g(y)\neq\pm\infty\) and \[ |g(x)-g(y)|\le\sum_{i: Q_i\cap\gamma\neq\emptyset}\rho(Q_i)\,. \] Here, the \textit{carpet modulus} \(\text{mod}(\Gamma_0)\) is defined as \(\inf\sum_{i\in {\mathbb N}\cup\{0\}}\sigma(Q_i)^2\) (\(Q_0:={\mathbb C}\setminus \bar{\Omega}\)), where \((\sigma(Q_i))_{i\in {\mathbb N}\cup\{0\}}\) is a nonnegative admissible sequence, meaning that \[ \sum_{i: Q_i\cap\gamma\neq\emptyset}\sigma(Q_i)\ge 1 \] for all \(\gamma\in \Gamma_0\) with \(\mathcal{H}^1 (\gamma\cap S)=0\) (\(\mathcal{H}\) denoting the Hausdorff measure). \par Setting \(M_{Q_i}(g)=\sup_{x\in\partial Q_i}g(x)\), \(m_{Q_i}(g)=\inf_{x\in \partial Q_i}g(x)\), \(\text{osc}_{Q_i}(g)=M_{Q_i}(g)-m_{Q_i}(g)\) (\(i\in {\mathbb N}\)), \(g\) is called a \textit{local Sobolev function} if for every open ball \(B\) relatively compact in \(\Omega\), \[ \sum_{i\in I_B} M_{Q_i}(g)^2 \text{diam}(Q_i)^2 <\infty\;\text{and}\; \sum_{i\in I_B}\text{osc}_{Q_i}(g)^2 <\infty \] (\(I_B:=\{i\in {\mathbb N}: B\cap Q_i\neq\emptyset\}\)) and \((\text{osc}_{Q_i} (g))_{i\in {\mathbb N}}\) is an upper gradient for \(g\). If these inequalities hold for the full sums over \(i\in {\mathbb N}\), the Sobolev property is termed \textit{global}. \par Now, a local Sobolev function \(u\) is called \textit{carpet-harmonic} if for every open set \(V\) relatively compact in \(\Omega\) and every Sobolev function \(\zeta\) supported on \(V\) it holds \[ D_V(u)\le D_V(u+\zeta)\,, \] the so-called Dirichlet energy functional being defined by \(D_V(f)=\sum_{i\in I_V}\text{osc}_{Q_i}(f)^2 \in [0,\infty ]\). The author presents several properties of carpet-harmonic functions, as the Caccioppoli inequality, for instance (Section 8 of Chapter 2). \par In the sequel we indicate how this notion of harmonicity is applied to obtain the uniformization result. After fixing four points on \(\partial\Omega\), this boundary is decomposed into closed sides \(\Theta_1,\ldots,\Theta_4\), enumerated counterclockwise, where \(\Theta_1\) and \(\Theta_3\) are opposite. Calling a Sobolev function \(g\) on \(S\), continuous up to \(\partial\Omega\), \textit{admissible for the free boundary problem} if \(g|_{\Theta_1}=0\) and \(g|_{\Theta_3}=1\), the author shows that there is a unique carpet-harmonic function \(u\) that minimizes \(D_\Omega (g)\) over all admissible functions \(g\). This \(u\) is continuous up to \(\partial\Omega\), \(u|_{\Theta_1}=0\), and \(u|_{\Theta_3}=1\). \par By ``integrating'' the ``gradient'' of \(u\) along its level sets, as the author puts it, the \textit{conjugate function} \(v\) of \(u\) is defined (Section 6 of Chapter 3). The function \(v:S \rightarrow [0, D_\Omega (u)]\) is continuous, \(v|_{\Theta_2}=0\), \(v|_{\Theta_4}=D_\Omega (u)\), but it is unclear if this \(v\) is carpet-harmonic. In any case, \(f:=(u,v)\) is the desired uniformization map. It maps \(S\) homeomorphically onto a so-called square Sierpiński carpet, that is, a carpet whose peripheral disks are squares and whose underlying open Jordan region is a rectangle. A condition, however, for this uniformization result is that the peripheral disks \(Q_i\) be uniformly quasiround (there exists \(K_0\ge 1\) such that for each \(Q_i\) there are concentric balls \(B(x,r)\), \(B(x,R)\) with \(B(x,r)\subset Q_i\subset B(x,R)\) and \(\frac{R}{r}\le K_0\)) and uniformly Ahlfors 2-regular (there exists \(K_1>0\) such that for every \(Q_i\) and every \(B(x,r)\) with \(x\in Q_i\) and \(r<\text{diam}(Q_i)\) it holds \(\mathcal{H}^2 (B(x,r)\cap Q_i)\ge K_1 r^2\)). Furthermore, this \(f\) has a property called \textit{carpet-quasiconformality}. Finally, the author presents certain refinements under additional geometric assumptions.