Direct limits, multiresolution analyses, and wavelets (Q961503): Difference between revisions

Building on earlier work of the authors [J. Fourier Anal. Appl. 15, No. 5, 616--633 (2009; Zbl 1190.42016)], generalized multiresolution analyses (GMRAs) are studied by expressing a Hilbert space \(H\) as a direct limit of an increasing union of closed subspaces \(V_n\), using universal properties to provide efficient proofs of known results concerning classical wavelets, wavelets on fractals due to \textit{D. E. Dutkay} and \textit{P. E. T. Jorgensen} [Rev. Mat. Iberoam. 22, No. 1, 131--180 (2006; Zbl 1104.42021)], and the supports of measures in realizations of MRAs in \(L^2\)-spaces on solenoids as discussed by \textit{M. Ionescu} and \textit{P. S. Muhly} [in: Groupoid methods in wavelet analysis. Contemporary Mathematics 449, 193--208, American Mathematical Society (AMS) (2008; Zbl 1161.22001)]. The key hypothesis in the present work is that the isometry \(S_m\) associated with a filter is a pure isometry, that is, its Wold decomposition has no unitary summand. In the basic setup, \(\Gamma\) is a countable abelian group with dual \(\widehat{\Gamma}\) and \(\alpha\) is an endomorphism of \(\Gamma\) with finite index \(N\). If \(D\) is a unitary ``dilation'' operator, and \(T: \Gamma\to U(H)\) is a unitary ``translation'' operator, an MRA for \((H,D,T)\) consists of an increasing sequence of closed subspaces \(V_n\) whose union is dense, whose intersection is \(\{0\}\), satisfying \(D(V_n)=V_{n+1}\), and possessing a ``scaling function'' \(\phi\in V_0\) such that \(\{T_\gamma \phi:\, \gamma\in\Gamma\}\) forms an orthonormal basis for \(V_0\). In a GMRA, the existence of \(\phi\) is relaxed to the requirement that \(V_0\) is invariant under \(T\). In the authors' earlier work (loc. cit.), GMRA's were associated with direct limits. Here, \(S\) is an isometry on a Hilbert space \(H\) to which one associates a direct limit \((H_\infty,U_n)\) and a unitary operator \(S_\infty\) on \(H_\infty\) such that \(S_\infty U_n=U_n S=U_{n-1}\) satisfying the properties (i) \(V_n=U_n H\) (\(n\geq 0\)) and \(V_n=S_\infty^{|n|} (V_0)\) (\(n<0\)), (ii) \(V_n\subset V_{n+1}\), (iii) \(\cup V_n\) is dense in \(H_\infty\), and (iv) \(\cap V_n=\{0\}\) precisely when \(S\) is a pure isometry. In this case, \(H_\infty =\oplus W_n\) where \(W_n=V_{n-1}\ominus V_n\). The first main result, (Theorem 3.1) states that if \(B\) is a Borel subset of \(\widehat{\Gamma}\) and \(m:\widehat{\Gamma}\to\mathbb{C}\) is a Borel function such that \[ \sum_{\alpha^\ast(\zeta)=\omega} |m(\zeta)|^2 =N\chi_B(\omega) \] almost everywhere on \(\widehat{\Gamma}\), and if one defines \(S_m: L^2(B)\to L^2(B)\) by \((S_mf)(\omega)=m(\omega)f(\alpha^\ast(\omega))\), then if either \(\widehat{\Gamma}\setminus B\) has positive Haar measure, or \(|m(\omega)|\neq 1\) is a set of positive measure, then \(S_m\) is a pure isometry. The second main general result provides an identification of the direct limit \(H_\infty\) a follows: Suppose that \(\mu:\Gamma\to U(H)\) is a unitary representation and \(S\) is an isometry on \(H\) such that \(S \mu_\gamma=\mu_{\alpha(\gamma)} S\) for \(\gamma\in \Gamma\). Suppose that \(\lambda:\Gamma\to U(K)\) is a unitary representation and \(D\) is a unitary operator on \(K\) such that \(D\lambda_\gamma D^\ast=\lambda_{\alpha(\gamma)}\) for \(\gamma\in \Gamma\). If there is an isometry \(R: H\to K\) such that (a) \(RS=DR\) and (b) \(R\mu_\gamma=\lambda_\gamma R\) for \(\gamma\in \Gamma\) then there is an isomorphism \(R_\infty\) of \(H_\infty\) onto the closure of the union of the spaces \(D^{-n} R(H)\) (\(n\geq 0\)) of \(K\) such that \(R_\infty S_\infty R_\infty^\ast=D\) and \(R_\infty \mu_\infty R_\infty^\ast=\lambda\). The subspaces \(D^{-n}R(H)\) form a GMRA of \(R_\infty(H_\infty)\) relative to \(D\) and \(\lambda\) if and only if \(S\) is a pure isometry.