A note on quasiconformal maps with Hölder-continuous dilatation (Q476111)

The authors give a new proof to the theorem of \textit{J. Mateu} et al. [J. Math. Pures Appl. (9) 91, No. 4, 402--431 (2009; Zbl 1179.30017)] which concerns conditions on the Beltrami coefficient \(\mu\) that guarantee that the solution of the Beltrami equation \[ \frac{\partial \Phi}{\partial \overline{z}}=\mu(z)\frac{\partial \Phi}{\partial z},\quad z\in \mathbb{C}, \] is bi-Lipschitz. Homeomorphic solutions of the Beltrami equation are the quasiconformal mappings of the plane. A necessary and sufficient condition for a quasiconformal map to be bi-Lipschitz is unknown, but in [\textit{K. Astala} et al., Elliptic partial differential equations and quasiconformal mappings in the plane. Princeton, NJ: Princeton University Press (2009; Zbl 1182.30001)] it is shown that \(\Phi\) is of class \(C^{1+\epsilon}\) provided \(\mu\) is Hölder continuous and compactly supported. Thus \(\Phi\) is bi-Lipschitz. Mateu et al. [loc. cit.] found a class of non-smooth dilatations which determine bi-Lipschitz quasiconformal mappings \(\Phi\). Theorem. Let \(\{\Omega_j\}\), \( 1\leq j\leq N\), be a finite family of disjoint bounded domains of the plane with boundary of class \(C^{1+\epsilon}\), \(0<\epsilon<1\), and let \(\mu=\sum_{j=1}^n\mu_j\chi_{\Omega_j}\), where \(\mu_j\) is an \(\epsilon\)-Hölder continuous functions on \(\Omega_j\) and \(||\mu_j||_{\infty}<1\) for \(j=1,\ldots, N\). Then the principal solution associated to the \(\mu\)-Beltrami equation is a bi-Lipschitz quasiconformal map. The proof of Mateu et al. [loc. cit.] is analytic and based on properties of the Beurling-Ahlfors transform. In this article, the authors give another proof using the more geometric methods of [\textit{S. Rohde}, Ann. Acad. Sci. Fenn., Math. 22, No. 2, 465--474 (1997; Zbl 0903.30018)]. First, the authors simplify the assumptions to \(j=1\), and then they reduce the theorem to the case of the unit disc \(\mathbb{D}\). Then it is enough to consider a homeomorphism \(f\) that satisfies \(\overline{\partial}f=\mu_f\partial f\) almost everywhere in \(\mathbb{C}\) with \(\mu_f\) an \(\epsilon\)-Hölder continuous function in \(\mathbb{D},\) \(\mu_f =0\) outside \(\mathbb{D},\;||\mu_f||_{\infty}<1\) and \(f(z)=z+O(\frac1z)\) near \(\infty\). By the Koebe distortion theorem, the authors get that \(f\) is bi-Lipschitz at least outside \(\frac32 \mathbb{D}\). Then to show that \(f\) is bi-Lipschitz also in \(2\mathbb{D}\), they use the fact from [Zbl 0903.30018] that a \(K\)-quasiconformal homeomorphism \(f\) of \(\mathbb{C}\) is bi-Lipschitz in \(E\subset \mathbb{C}\) if the difference in conformal modulus of \(A\) and \(f(A)\) is bounded above by a constant \(N\) for all annuli \(A\) centered at points of \(E\) with the property that both boundary circles meet \(E\). And to show that this difference in conformal modulus is bounded above they use Lehto's inequality from [Ann. Acad. Sci. Fenn., Ser. A I 275, 28 p. (1960; Zbl 0090.05301)] which says that \[ |M(A)-M(f(A)) |\leq C(K)\int_A\frac{|\mu_f(y)|}{|x-y|^2}\, dy. \] These geometric methods can be applied also to domains without smooth boundary as long as there is some knowledge about the behavior of conformal maps on such domains. The authors show that this theorem extends to domains with positive angle corners if the dilatation takes certain allowable values which depend on the angle at the corner.