Arboreal Galois representation for a certain type of quadratic polynomials (Q2297520)

The results of this paper are nicely described by the author's introduction: ``Consider the quadratic polynomial \(f(x)=x^2+a\in \mathbb Z[x]\) and consider the Galois groups of polynomials of the form \(f^n(x)\), where \(f^n=f\circ\cdots\circ f\) is the \(n\)-th iterate of \(f\). Let \(K_n\) be the splitting field of \(f^n\) over \(\mathbb Q\) and denote by \(G_n=\mathrm{Gal}(K_n/\mathbb Q)\). \(G_n\) has been called arboreal Galois group because it can be made to act on trees. In fact, let \(T_n\) be the graph whose vertex set is \(\cup_{0\le i\le n}f^{-i}(0)\) and where we draw an edge from \(\alpha\) to \(\beta\) if \(f(\alpha)=\beta\). Then clearly \(G_n\) acts faithfully on \(T_n\), so that \(G_n\) can be considered as a subgroup of \(\Aut(T_n)\), the automorphism group of \(T_n\). We write \(G_\infty\) as the inverse limit of the Galois groups \(G_n\) and write \(T_\infty\) as the increasing union of \(T_1\subset T_2\subset\cdots\). Therefore, the injections \(G_n\hookrightarrow \Aut(T_n)\) naturally extend to the injection \(G_\infty\hookrightarrow \Aut(T_\infty)\). In this article, we suppose that \(-a\) is not a square in \(\mathbb Z\). This is equivalent to saying that all iterates of \(f\) are irreducible over \(\mathbb Q\), and hence ensures that \(T_n\) is a complete binary rooted tree of height \(n\) (so there are \(2n\) leaves at the top). In this case, we can describe \(\Aut(T_n)\) as the \(n\)-fold iterated wreath product of the cyclic group \(C_2\) of two elements, as shown by \textit{R. W. K. Odoni} [Mathematika 35, No. 1, 101--113 (1988; Zbl 0622.12010)]. We also neglect the case that \(a=-2\). This is the only case that \(f(x)=x^2+a\) is so called post-critically finite, which means that the forward orbit of its critical points is finite. In fact, when \(f(x)=x^2-2\), it is well known that \(G_n\cong C_{2^n}\) and hence \([\Aut(T_\infty):G_\infty]=\infty\). Under these assumptions (that is \(-a\) is not a square in \(\mathbb Z\) and \(a\ne -2)\), many evidences indicate that one should expect \([\Aut(T_\infty):G_\infty]=\infty\). For example, \textit{C. Gratton} et al. [Bull. Lond. Math. Soc. 45, No. 6, 1194--1208 (2013; Zbl 1291.37121)] proved that this is the case assuming the ABC conjecture. The study of the case \(a=1\) which is a problem originally posed by J. McKay, began by Odoni [loc. cit.] and \textit{J. E. Cremona} [Mathematika 36, No. 2, 259--261 (1989; Zbl 0699.12018)]. Later, by extending Odoni's theory and techniques for \(x^2+1\) to the general case, and by adding an ingenious pair of new ideas of his own, \textit{M. Stoll} [Arch. Math. 59, No. 3, 239--244 (1992; Zbl 0758.11045)] proved as his main result that if \(a\in\mathbb Z\) has one of the following properties: (a) \(a>0\) and \(a\equiv 1\pmod 4\) or \(a\equiv 2\pmod 4\); (b) \(a<0\), \(a\in 4\mathbb Z\) and \(-a\) is not a square in \(\mathbb Z\), then \(G_n=\Aut(T_n)\) for all \(n\ge 1\). Not much seems to be known for all other cases where \(-a\) is not a square.'' In this paper, we solve another piece of this puzzle. We prove that if \(a<0\) and \(a\equiv -1\pmod 4\), then \([\Aut(T_\infty):G_\infty]\le 2\). More precisely, we have \(G_n=\Aut(T_n)\) for all \(n\ge1\) if \(-a\) is not of the form \(4k^2+1\) and \([\Aut(T_n):G_n]=2\) for all \(n\ge 2\), otherwise. The author's approach follows the methods used by Odoni and Stoll.