Multiresolution analysis on product of zero-dimensional Abelian groups (Q641675)

After reviewing the theory of multiresolution analyses on zero-dimensional groups previously presented by the author in a paper in Russian [Sb. Math. 201, No. 5, 669--691 (2010; Zbl 1201.42026)], a theory for corresponding multiresolution analyses on products of such groups is developed, at least in a special case. By a zero-dimensional group \({\mathfrak G}\), the author means the union of a decreasing chain of subgroups \({\mathfrak G}_n\), \(n\in\mathbb Z\), such that each \({\mathfrak G}_n\) is a compact abelian group and the quotient \({\mathfrak G}_n\) \(/{\mathfrak G}_{n+1}\) is finite of prime order \(p_n\). Elements \(x\in{\mathfrak G}\) can be expressed in the form \(x=\sum_{n=-\infty}^\infty a_n g_n\) with \(g_n\in{\mathfrak G}_n\) and \(a_n\in \{0,\dots, p_n-1\}\). The \(g_n\) form a basic system. The sum defining \(x\) has at most finitely many negative index terms, that is, \(x=\sum_{n=-N}^\infty a_n g_n\) for some \(N=1,2,\dots\). The key to defining a multiresolution analysis is first to define a dilation \(A:{\mathfrak G}\to{\mathfrak G}\) via \(A(\sum a_n g_n)=\sum a_n g_{n-1}\). This operator is additive and \(A{\mathfrak G}_n ={\mathfrak G}_{n-1}\). If one calls characters in the dual \({\mathfrak G}_{n+1}^\bot\setminus{\mathfrak G}_n^\bot\) Rademacher functions \(r_n\), then for the special case \(p_n=p\) for all \(n\), one has \(r_n =r_0 A^n\). A multiresolution analysis of \(L^2({\mathfrak G})\) is defined in the usual way as nested subspaces \(V_n\subset V_{n+1}\) such that \(\bigcup V_n\) is dense in \(L^2({\mathfrak G})\) and \(\bigcap V_n=\{0\}\), \(f(x)\in V_n\Leftrightarrow f(x)\in V_{n+1}\), \(V_0\) is invariant under shifts by \(H_0=\{\sum_{j<0} a_j g_j\}\), and, finally, there is a \(\varphi\in V_0\) such that \(\{\varphi(x-h): h\in H_0\}\) forms an orthonormal basis for \(V_0\). Then \(\varphi\) satisfies a scaling relation \(\varphi(x)=\sum_{h\in H_0} \varphi (Ax-h)\), possibly having only finitely many nonzero terms. It was shown in the author's previous work that a scaling function \(\varphi\) generates an MRA for \(L^2({\mathfrak G})\) if \(|\widehat{\varphi}(\chi)|\) is the indicator function of \({\mathfrak G}_0^\bot\). The main steps for proving this are also outlined here. If \({\mathfrak G}\) is such that \({\mathfrak G}_n/{\mathfrak G}_{n+1}=p\) and if one sets \({\mathfrak G}_n^d={\mathfrak G}_n\times\cdots\times {\mathfrak G}_n\) (\(d\)-times), then \({\mathfrak G}_n^d/{\mathfrak G}_{n+1}^d\) has cardinality \(p^d\), which is not prime. The multiresolution theory for zero-dimensional groups relies on the prime cardinality of these quotients (to define dilations). The author refines the scheme by identifying intermediate groups \({\mathfrak G}_{nk}\) such that \({\mathfrak G}_{n0}={\mathfrak G}_n^d\), \({\mathfrak G}_{nd}={\mathfrak G}_{n+1}^d\), and \({\mathfrak G}_{n,k-1}/{\mathfrak G}_{nk}=p\), \(k=1,\dots, d\), and uses this refinement to define a suitable dilation operator that leads to a corresponding suitable notion of MRA on \(d\)-fold products of zero-dimensional groups.