Differences of functions and group generators for compact Abelian groups (Q1126392)

Let \(G\) be a compact connected abelian group, \(L^2(G)\) the space of square-summable functions, and \(L_0^2(G)\) the null space of the Lebesgue integral. A \(difference\) is a function in \(L_0^2(G)\) of the form \(g - \delta _x\star g\). Meisters and Schmidt proved that a function \(f \in L_0^2(G)\) can be written as the sum of three differences. They showed that there were some functions for which two differences were not enough. B. E. Johnson showed that, if \(G\) has connected component \(C\) such that \(G/C\) is topologically finitely generated, then \(f \in L_0^2(G)\) can still be written as the sum of a finite number of differences. (The case with \(G/C\) finite was proved in the original paper by Meisters and Schmidt.) The purpose of this paper is to try and relate the number of topological generators for \(G/C\) to the minimum number of differences required to represent the function \(f\). In the event, the authors use a subspace \(H(G)\) of \(L_0^2(G)\), and it is with respect to these functions that the comparison is made. In fact, several subspaces can be used for \(H(G)\), and they are defined as follows: For \(0 < \epsilon < 1\), let \[ A_\epsilon(G) = \{f \in L_0^2(G): \sum_{\gamma \in \hat{G}}|\hat{f}(\gamma)|^\epsilon < \infty\}. \] Then we can take for \(H(G)\) the space \(A_\epsilon(G)\) or \(\cup\{ A_\epsilon(G): 0 < \epsilon < \delta \}\). Now, if \(G\) is connected, any function in \(H(G)\) is a difference, and the main result of this paper is that if \(G/C\) is topologically generated by \(n\) elements, then any \(f\) in \(H(G)\) can be written as the sum of \(n\) differences.