Value ranges of univalent self-mappings of the unit disc (Q499253)

Let \(S_T\), \(T>0\), be the class of holomorphic univalent self-mappings \(f\) of the unit disk \(\mathbb D=\{z:|z|<1\}\) normalized by \(f(0)=0\) and \(f'(0)=e^{-T}\). In the paper, the authors determine the value set \[ V_T(z_0)=\{f(z_0):f\in S_T\},\;\;z_0\in\mathbb D\setminus\{0\}. \] Some results on variations of the set \(V_T(z_0)\) are known: there are the classical setting of the Schwarz and Rogosinski's lemma, the description of the set \(\{\log(f(z_0)/z_0)\}\) for bounded univalent functions in \(\mathbb D\) due to Goryainov and Gutlyanskii, and a paper by Roth and Schleissinger for univalent self-mappings \(f\) of \(\mathbb D\), \(f(0)=0\), \(f'(0)>0\). The authors apply Pontryagin's maximum principle to the radial Loewner equation directly to determine \(V_T(z_0)\). This is a different and more straightforward approach to solve the problem. The main result reduced to the case \(z_0\in(0,1)\) is proved in the following theorem. {Theorem 1.} Let \(z_0\in(0,1)\). For \(x_0\in[-1,1]\) and \(T>0\), let \(r=r(T,x_0)\) be the unique solution to the equation \[ \frac{(1-z_0)^2}{1-x_0}\log\frac{1-r}{1-z_0}+\frac{(1+z_0)^2}{1+x_0}\log\frac{1+r}{1+z_0}=\frac{1-2x_0z_0+z_0^2}{1-x_0^2}\log\frac{re^T}{z_0} \] and let \[ \sigma(T,x_0)=\frac{2(1-z_0^2)\sqrt{1-x_0^2}}{1-2x_0z_0+z_0^2}(\text{arctanh}\;z_0-\text{arctanh}\;r(T,x_0)). \] For fixed \(T\geq0\), define the two curves \(C_+(z_0)\) and \(C_-(z_0)\) by \[ C_{\pm}(z_0):=\{w_{\pm}(x_0):=r(T,x_0)e^{\pm i\sigma(T,x_0)}:x_0\in[-1,1]\}. \] Then, if \(\text{arctanh}\;z_0<\pi/2\), then \(V_T(z_0)\) is the closed region whose boundary consists of the two curves \(C_+(z_0)\) and \(C_-(z_0)\), which only intersect at \(x_0\in\{-1,1\}\). For \(\text{arctanh}\;z_0\geq\pi/2\), there are two different cases. First, assume that \(T\) is large enough that the equation \((1-2xz_0+z_0^2)\sigma(T,x)=\pi(1+2xz_0+z_0^2)\) admits a solution \(x\in[-1,1]\). Then the curves \(C_+(z_0)\) and \(C_-(z_0)\) intersect more than twice. There is a \(\chi\in(-1,1)\) such that \(\tilde C_+(z_0)\cup\tilde C_-(z_0)\) is a closed Jordan curve, where \(\tilde C_{\pm}(z_0):=\{w_{\pm}(x_0):x_0\in[\chi,1]\}\), and an \(\mathcal N\in(-1,1)\) such that \(\hat C_+(z_0)\cup\hat C_-(z_0)\) is a closed Jordan curve, where \(\hat C_{\pm}(z_0):=\{w_{\pm}(x_0):x_0\in[-1,\mathcal N]\}\). Then \(V_T(z_0)\) is the closed region whose boundary is \(\tilde C_+(z_0)\cup\tilde C_-(z_0)\cup\hat C_+(z_0)\cup\hat C_-(z_0)\). For smaller \(T\) that do not admit a solution to the above equation, the set \(V_T(z_0)\) can be described exactly as in the case of \(\text{arctanh}\;z_0<\pi/2\). Similarly, the authors describe the value set \(\{f^{-1}(z_0)\}\) for \(f\in S_T\) and \(z_0\in f(\mathbb D)\).