An inverse theorem for an inequality of Kneser (Q2423221)

Let \(G\) be a compact connected abelian group, and let \(\mu_{G}\) be its Haar proba\-bility measure. In his paper [Math. Z. 66, 88--110 (1956; Zbl 0073.01702)], \textit{M. Kneser} showed that if \(A\) and \(B\) are non-empty compact subsets of \(G\), then \[ \mu_{G}(A + B) \geq \min\bigl(\mu_{G}(A) + \mu_{G}(B),1\bigr). \] The main result of the paper under review establishes that when the equality in the preceding estimation is almost attained, the sets \(A\) and \(B\) are close to some canonic subsets of \(G\), called Bohr sets. Precisely, a Bohr set in \(G\) is a subset of the form \(\varphi^{-1}(I)\), where \(\varphi\) is a continuous surjective character of \(G\) and \(I\) is a compact arc in \(\mathbb{R}/\mathbb{Z}\). Two Bohr sets \(\varphi^{-1}(I)\) and \(\psi^{-1}(J)\) are said to be parallel if \(\varphi = \psi\). The author proves: Theorem 1.3 Let \(\varepsilon > 0\), and suppose \(\delta > 0\) is sufficiently small depending of \(\varepsilon\). Then, for any compact subsets \(A\) and \(B\) of \(G\) with \[ \mu_{G}(A), \mu_{G}(B), 1 - \mu_{G}(A) - \mu_{G}(B) \geq \varepsilon \] and \[ \mu_{G}(A + B) \leq \mu_{G}(A) + \mu_{G}(B) + \delta, \] there exist parallel Bohr sets \(\varphi^{-1}(I)\) and \(\varphi^{-1}(J)\) such that \[ \mu_{G}(A \triangle \varphi^{-1}(I)), \mu_{G}(B \triangle \varphi^{-1}(J)) \leq \varepsilon, \] where \(A \triangle B\) denotes the symmetric difference of \(A\) and \(B\). Theorem 1.5 Let \(\varepsilon > 0\), and suppose \(\delta > 0\) is sufficiently small depending of \(\varepsilon\). Then, for any measurable subsets \(A\) and \(B\) of \(G\) with \[ \mu_{G}(A), \mu_{G}(B), 1 - \mu_{G}(A) - \mu_{G}(B) \geq \varepsilon \] and \[ \mu_{G}(A +_{\delta} B) \leq \mu_{G}(A) + \mu_{G}(B) + \delta, \] there exist parallel Bohr sets \(\varphi^{-1}(I)\) and \(\varphi^{-1}(J)\) such that \[ \mu_{G}(A \triangle \varphi^{-1}(I)), \mu_{G}(B \triangle \varphi^{-1}(J)) \leq \varepsilon, \] where \(A +_{\delta} B = \{x \in G : \int_{G}1_{A}(y) 1_{B}(x-y) d\mu_{G}(y) \geq \delta\}\).