Smooth curves having a large automorphism \(p\)-group in characteristic \(p>0\) (Q1005890)

Let \(k\) be an algebraically closed field of characteristic \(p>0\) and \(C\) a connected nonsingular projective curve over \(k\) with genus \(g\geq2\). This paper is the first of a set of three (together with \textit{M. Rocher}'s [J. Algebra 321, No. 2, 704--740 (2009; Zbl 1185.14025)] and [``Large \(p\)-group actions with \(\frac{|G|}{g^2} \geq \frac{4}{(p^2-1)^2}\)'' (2008; \url{arXiv:0801.3834})]), whose main object is to study \(G\)-actions on the curve \(C\), when \(G\) is a \(p\)-group such that \(|G| > \frac{2\,p}{p-1} \, g\). This paper continues the work initiated in [\textit{C. Lehr} and \textit{M. Matignon}, Compos. Math. 141, No. 5, 1213--1237 (2005; Zbl 1083.14028)]. One of the authors' aims is to display some universal families and to discuss the corresponding deformation space. In positive characteristic \(p>0\), the automorphism group \(\mathrm{Aut}_k(C)\) is known to be finite but, because of the appearance of wild ramification, the Hurwitz linear bound is replaced by a biquadratic one [\textit{H. Stichtenoth}, Arch. Math. 24, 527--544 (1973; Zbl 0282.14006), Arch. Math. 24, 615--631 (1973; Zbl 0282.14007)]. To rigidify the situation as has been done in characteristic zero [\textit{R. S. Kulkarni}, Contemp. Math. 201, 63--79 (1997; Zbl 0863.30050); \textit{T. Breuer}, Characters and automorphism groups of compact Riemann surfaces. London Mathematical Society Lecture Note Series. 280 (2000; Zbl 0952.30001)], an idea is to consider large automorphism \(p\)-groups. In this spirit, the authors focus on what they call \textit{big actions}, that is to say pairs \((C,G)\) with \(G\) a \(p\)-subgroup of \(\mathrm{Aut}_k(C)\) such that \(|G|>\frac{2p}{p-1}\, g.\) Under this condition, \(g_{C/G}=0\) and only one point \(\infty\in C\) is ramified in \(C/G\). Call \(G_i\) the \(i\)-th lower ramification group of \(G\) at \(\infty\). Then, \(G=G_{-1}=G_0=G_1\supsetneq G_2 \neq \{e\}.\) Moreover, the quotient curve \(C/G_2\) is isomorphic to the projective line. It follows that the quotient group \(G/G_2\) acts as a group of translations of the affine line \(C/G_2-\{\infty\}=\mathrm{Spec} \, k[X]\), through \(X \rightarrow X+y\), where \(y\) runs over a subgroup \(V\) of \(k\). This induces the exact sequence \[ 0 \longrightarrow G_2 \longrightarrow G=G_1 \overset{\pi}{\longrightarrow} V \simeq (\mathbb{Z}/\,p\, \mathbb{Z})^v \longrightarrow 0, \] where \(\pi(g)=g(X)-X\) for all \(g \in G\). Then, the purpose of this paper is twofold: on the one hand, to give necessary conditions on \(G_2\) for \((C,G)\) to be a big action and, on the other hand, to display realizations of big actions with \(G_2\) abelian of large exponent. The main results of the first part are gathered in the following: {Theorem:} Let \((C,G)\) be a big action with \(g \geq 2\). 1. Let \(H\) be a normal subgroup of \(G\) such that \(H\subsetneq G_2\). Then \((C/H, G/H)\) is a big action with second ramification group \((G/H)_2=G_2/H\) (Lemma 2.4.2). 2. The group \(G_2\) is equal to \(D(G)\), the commutator subgroup of \(G\) (Thm. 2.7). In particular, \(G\) cannot be abelian. 3. The group \(G_2\) cannot be cyclic unless \(G_2\) has order \(p\) (Thm. 5.1). 4. If \(\frac{|G|}{g^2} \geq \frac{4}{(p^2-1)^2}\), then \(G_2\) is an elementary abelian \(p\)-group with order dividing \(p^3\) (Prop. 4.1). These results highlight the major role played by \(G_2\) in the study of big actions. Indeed, a key idea in studying big actions is to use the first statement to go back to the situation of big actions with a \(p\)-cyclic \(G_2\), a case that has been fully characterized in [\textit{C. Lehr} and \textit{M. Matignon}, loc. cit.]. The second companion paper gives a generalization of this case and study big actions with a \(p\)-elementary abelian \(G_2\). These results led to the classification of the big actions satisfying \(\frac{|G|}{g^2} \geq \frac{4}{(p^2-1)^2}\) (third companion paper). The second part of the paper is devoted to examples of big actions with \(G_2\) abelian, knowing that the authors do not know yet examples of big actions with a nonabelian \(G_2\). Following [\textit{K. Lauter}, J. Number Theory 74, No. 1, 56--72 (1999; Zbl 1044.11054)], the authors consider the maximal abelian extension \(K_S^m\) of \(K:=\mathbb{F}_q(X)\) (where \(q=p^e\)) that is unramified outside \(X=\infty\), completely split over the set \(S\) of the finite rational places and whose conductor is smaller than \(m \, \infty\), with \(m \in \mathbb{N}\). It follows from the uniqueness of \(K_S^m\) that the group of translations \(\{X \rightarrow X+y, \, y \in \mathbb{F}_q \}\) extends to a \(p\)-group of \(\mathbb{F}_q\)-automorphisms of \(K_S^m\), say \(G(m)\), with the exact sequence \[ 0 \longrightarrow \mathrm{Gal}(K_S^m/K) \longrightarrow G(m) \longrightarrow \, \mathbb{F}_q \longrightarrow 0. \] For a well-chosen conductor \(m\), this provides examples of big actions whose \(G_2=G_S(m)\) is abelian of arbitrary large exponent. This also relates the problem of big actions to the search of algebraic curves with many rational points. Indeed, if \(N_m\) denotes the number of \(\mathbb{F}_q\)-rational points of the nonsingular projective curve \(C_m\) with function field \(K^m\), then \(\frac{|G_m|}{g_{C_m}} \sim \frac{N_m}{g_{C_m}}\), when \(e\) grows large.