Cyclic covers and Ihara's question (Q2336056)

Let \(k\) be a number field, and let \(G_k\) be the absolute Galois group of \(k\). Let \(\ell\) be a rational prime, and let \(S_0\) be a \(G_k\)-stable finite subset of \(\mathbb{P}^1_{\bar k}\) containing \(\{0,1,\infty\}\). Consider a \(\bar k\)-morphism \(X \to \mathbb{P}^1_{\bar k}\) where \(X/\bar k\) is a complete smooth irreducible curve such that the Galois closure of the morphism has degree equal to a power of \(\ell\). This setup admits a natural action of the pro-\(\ell\) fundamental group of \( \mathbb{P}^1 - S_0\), and the authors of the paper are interested in the canonical pro-\(\ell\) outer Galois representation \(\Phi_k : G_k \to \mathrm{Out}( \pi_1^{\mathrm{pro-}\ell}(\mathbb{P}^1 - S_0))\). Anderson and Ihara studied this setup, and introduced a combinatorial object \(\mathbb S\) consisting of finite subsets \(S\) of \(\mathbb P^1_{\bar k}\) obtained by pulling back \(S_0\) through a finite combination of certain linear fractional transformations and \(\ell\)th power mappings. They proved that \textit{ given a monic polynomial \(f(x)\in k[x]\) and \(\lambda\in k\), the \(\ell\)-power torsion of the Jacobian variety \(J/k\) defined by the superelliptic curve \(y^{\ell^n} = \lambda f(x)\) is defined over \(\mathrm{ker}(\Phi_{\mathbb Q})\) if \(S_0=\{0,1,\infty\}\), \(k\) is fixed by \(\mathrm{ker}(\Phi_{\mathbb Q})\), and there are two subsets \(S,S'\) in \(\mathbb S\) such that \(\lambda \in S\) and \(S'\) contains the roots of \(f(x)\).} The authors of the paper under review attempt to generalize the Anderson-Ihara result by avoiding the use of \(\mathbb S\). Let \(H_k\) be the maximal pro-\(\ell\) extension of \(k(\mathbf{\mu_{\ell^\infty}})\) unramified away from \(\ell\) where \(\mu_{\ell^\infty}\) is the subset of \(\bar{\mathbb Q}\) consisting of the \(\ell\)-power roots of unity. If \(S_0=\{0,1,\infty\}\), then \(\mathrm{ker}(\Phi_k) \subset H_k\), and in addition if \(k=\mathbb Q\) and \(\ell\) is an odd regular prime, then Sharifi showed that \(\mathrm{ker}(\Phi_k) = H_k\) by the Delign-Ihara Conjecture, which is now a theorem. The authors of the paper under review prove the following version. Let \(J/k\) be the Jacobian variety defined by a superelliptic curve \(y^{\ell^n} = f(x)\) where \(f(x)\in k[x]\), and let \(S\) be the branch locus of the curve. \textit{ If \(J\) has good reduction away from \(\ell\), and \(S\subset H_k\), then the \(\ell\)-power torsion of \(J\) is defined over \(H_k\).}