Analytic wavelets in multiply connected domains with circular boundaries (Q2341991): Difference between revisions

The main achievement of this paper is the construction of bases in Hardy-type spaces of analytic functions in the domain \(\tilde K=\mathbb D\setminus\cup_{k=1}^m D(z_k, r_k)\); here \(\mathbb D\) denotes the unit disk and \(D(z, r)\) the disk of centre \(z\) and radius \(r\). The disks \(D(z_k, r_k)\) are assumed to be contained in \(\mathbb D\) (\(r_k<1-|z|\)) and pairwise disjoint. The spaces \(H^p(\tilde K)\) under consideration are defined in the following way. Let \(\rho\) denote the minimum of the distances between the boundaries of \(\tilde K\) (the unit circle and the circles \(C(z_k,r_k)\)). Let \[ a_0(f)=\sup_{1-\rho<r<1}\left(\frac 1{2\pi} \int_0^{2\pi} |f(re^{i\theta})|^p d\theta\right)^{1/p}\;, \] \[ a_l(f)=\sup_{r_l<r<r_l+\rho}\left(\frac 1{2\pi} \int_0^{2\pi} |f(z_l+re^{i\theta})|^p d\theta\right)^{1/p},\;l=1,\dots,m. \] For \(1\leq p<\infty\) the space \(H^p(\tilde K)\) consists of the analytic functions in \(\tilde K\) such that \[ \|f\|_p=\max_{l=0,\dots,m} a_l(f) <\infty\;. \] This can be viewed as a generalisation of the classical Hardy spaces in the unit disk. For the case \(p=\infty\) the definition is however different: \(f\in H^\infty(\tilde K)\) if \(f\) is analytic in \(\tilde K\) and continuous in \(\overline{\tilde K}\). In analogy with the classical Hardy spaces, it is shown that functions in \(H^p(\tilde K)\) have non-tangential boundary values a.e. and that these boundary values determine the function. The main theorem of the paper gives a representation of functions in \(H^p(\tilde K)\) in terms of a wavelet system. For \(n=2^{j-1}+1\), \(j>0\), and \(k=0,\dots,2^{j-1}-1\), let \[ A_0(z)\equiv 1\;, \] \[ A_n(z)=2^{1-j/2}\sum_{\nu=1}^\infty z^\nu \hat\theta(\frac \nu{2^j}) \sin\left(\frac{2\pi \nu (k+0.5)}{2^j}\right)\;. \] Here \(\hat\theta\) is the even, twicely continuously differentiable, non-negative function defined by \[ \hat\theta(\zeta)=\hat\varphi^2(\zeta/2)-\hat\varphi^2(\zeta), \] where \(\hat\varphi\in\mathcal C^2(\mathbb R)\) is a cut-off function such that, for some \(\epsilon\in (0,1/3)\), \[ \hat\varphi(\zeta)=1\qquad |\zeta|\leq\frac{1-\epsilon}2 \] \[ \hat\varphi(\zeta)=0\qquad |\zeta|\geq\frac{1+\epsilon}2 \] \[ \hat\varphi^2(\zeta+1/2)-1/2\quad \text{is odd for} \;|\zeta|\leq\epsilon/2. \] {Theorem 1.1} (Main Theorem). The system of functions \[ A_0(z), \quad A_n(z), \quad A_n\bigl(\frac {r_l}{z-z_l}\bigr),\qquad n\geq 1, l=1,\dots,m \] forms a basis in \(H^p(\tilde K)\), \(1\leq p\leq \infty\). The coefficients of the expansion \[ f(z)\sim \sum_{n=0}^\infty \alpha_n A_n(z) +\sum_{n=1}^\infty \sum_{l=1}^m \alpha_{n,l} A_{n,l} (z), \] whose partial sums converge in the norm \(\|\cdot\|_p\), can be computed from the boundary values of \(f\) by the formulas \[ \alpha_n=\langle f,A_n\rangle_0\quad,\quad \alpha_{n,l}=\langle f,A_{n,l}\rangle_l\;, \] where \[ \langle f, g\rangle_0=\frac 1{2\pi}\int_0^{2\pi} f(e^{i\theta}) \overline{g(e^{i\theta})}\; d\theta, \] \[ \langle f, g\rangle_l=\frac 1{2\pi}\int_0^{2\pi} f(z_l+r_le^{i\theta}) \overline{g(z_l+r_le^{i\theta})}\; d\theta,\quad l=1,\dots,m. \] The second part of the paper is devoted to study the rate of convergence of the partial sums.