On approximation of continuous functions by entire functions on subsets of the real line (Q1959705)

For a closed unbounded set \(E\subset\mathbb R\), let \(BC(E)\) denote the class of bounded continuous functions on \(E\), \(E_{\sigma}\) be the class of entire functions of exponential type at most \(\sigma>0\), and \[ A_{\sigma}(f,E):=\inf_{g\in E_{\sigma}}\|f-g\|_{C(E)},\qquad f\in BC(E), \] with the uniform norm \(\|\cdot\|_{C(E)}\) over \(E\). The set \(\mathbb R\setminus E\) consists of a finite or infinite number of disjoint open intervals \(J_j=(a_j,b_j)\). There exist vertical intervals \(J_j'=(u_j,u_j+v_j)\) in the upper half-plane \(\mathbb H\), \(u_j\in\mathbb R\), \(v_j>0\), and a conformal mapping \(\phi:\mathbb H\to\mathbb H_E:=\mathbb H\setminus(\bigcup_j J_j')\) normalized by \(\phi(\infty)=\infty\), \(\phi(i)=i\), such that \(\phi\) can be extended continuously to \(\overline{\mathbb H}\), and \(\phi(J_j)=J_j'\). For \(x_1,x_2\in E\) and \(x_1<x_2\), set \(\rho_E(x_1,x_2)=\rho_E(x_2,x_1):=\text{diam\,}\phi([x_1,x_2])\). The main result of the paper is the relation \(A_{\sigma}(f,E)=O(\sigma^{-\alpha})\) as \(\sigma\to\infty\) for \(f\in BC(E)\) and \(0<\alpha<1\), which is valid if and only if \(\omega^*_{f,E}(\delta):=\sup|f(x_2)-f(x_1)|=O(\delta^{\alpha})\) as \(\delta\to+0\). The supremum is taken over all \(x_1,x_2\in E\) such that \(\rho_E(x_1,x_2)\leq\delta\). This is an analogue of Bernstein's results for \(E=\mathbb R\), and it follows immediately from the direct Theorem 2 and the inverse Theorem 3. Let a positive non-decreasing function \(\omega(\delta)\), \(\delta>0\), satisfy \(\omega(+0)=0\) and \(\omega(t\delta)\leq2t\omega(\delta)\), \(t>1\). Denote by \(BC^*_{\omega}(E)\) the class of functions \(f\in BC(E)\) for which \(\omega^*_{f,E}(\delta)\leq\omega(\delta)\). Theorem 2: For \(f\in BC^*_{\omega}(E)\) and \(\sigma\geq1\), \[ A_{\sigma}(f,E)\leq C_8\left(\frac{\|f\|_{C(E)}}{\sigma}+\omega\left(\frac{1}{\sigma}\right)\right). \] Theorem 3: Let \(f\in BC(E)\) and \[ A_{\sigma}(f,E)\leq\omega\left(\frac{1}{\sigma}\right),\qquad\sigma\geq1. \] Then, for \(x_1,x_2\in E\), \[ \Big|f(x_2)-f(x_1)\Big|\leq C^{}_9\Omega\Big(\rho_E(x_1,x_2)\Big), \] where \[ \Omega(\delta):=\delta\left(\|f\|_{C(E)}+\int^1_{\delta}\frac{\omega(t)}{t^2}dt\right) \qquad\text{for}\qquad 0<\delta\leq\frac{1}{2}, \] and \(\Omega(\delta):=\Omega(1/2)\) for \(\delta>1/2\).