Wild quotient surface singularities whose dual graphs are not star-shaped (Q906644)

Let \(G\) be a finite group acting on a regular local noetherian ring \(A\). Then the invariant ring \(R=A^G\) is a normal local noetherian ring and the ring extension \(R\subset A\) is finite. If the induced \(G\)-action on \(X=\text{Spec}(R)\) is free in codimension one, the ring \(R\) and its spectrum \(Y=\text{Spec}(R)\) are called a quotient singularity. A quotient singularity is called tame if the order of \(G\) is prime to the characteristic exponent of the residue field \(k=A/\mathfrak m_A\) and wild otherwise (the characteristic exponent of a field is \(1\) if the field has characteristic zero, and equal to the characteristic of the field otherwise). Lorenzini initiated a systematic study of wild quotient singularities in dimension two in a series of papers [\textit{D. Lorenzini}, Math. Z. 275, No. 1--2, 211--232 (2013; Zbl 1309.14003); Algebra Number Theory 8, No. 2, 331--367 (2014; Zbl 1332.14029); ``Wild quotients of products of curves'', preprint, \url{http://www.math.uga.edu/-lorenz/paper.html}]. The goal of the paper under review is to further investigate such singularities. Let \(R\) be a local complete normal noetherian ring with separably closed residue field. Then there exists a resolution of singularities \(f\colon X\to \text{Spec}(R)\) [\textit{J. Lipman}, Ann.\ Math. (2) 107, 151--207 (1978; Zbl 0349.14004)]. Let \(E\subset X\) be the exceptional divisor, viewed as a closed reduced subscheme of \(X\), and let \(E_1,\ldots,E_r\) be its irreducible components. The authors always assume that \(E\) has simple normal crossings. Attached to this situation is the dual graph of \(E\); it is the graph whose vertices correspond to the curves \(E_1,\ldots,E_r\). Dual graphs of tame quotient singularities are always star-shaped, i.e., contain at most one node; a node of a graph is a vertex which has at least three neighboring vertices. \textit{E. Brieskorn}'s [Invent. Math. 4, 336--358 (1968; Zbl 0219.14003)] characteristic zero classification of quotient singularities shows that the number of nodes in the dual graph is either zero or one. In the wild case, the dual graphs may become rather complicated. \textit{D. Lorenzini} [``Questions on wild \(\mathbb Z/p\mathbb Z\)-quotient singularities in dimension 2'', preprint, \url{http://www.math.uga.edu/-lorenz/paper.html}] asked whether there are wild quotient singularities whose dual graph has no nodes, or whose dual graph has at least two nodes. In the first part of this paper--sections 2-4--the authors deal with this question. In Theorem 2.1 they show that the local fundamental group \(\pi^{\text{loc}}_1(R)\) of a Hirzebruch-Jung singularity \(R\) of type \(m/b\) is isomorphic to the prime-to-\(p\) part of the cyclic group \(\mathbb Z/m\mathbb Z\), \(p\) being the characteristic exponent of the residue field of \(R\). From this result they deduce in Cor.\ 2.2 that the dual graph attached to a wild quotient singularity in dimension two contains at least one node, answering in the negative the question of Lorenzini. In order to determine the local fundamental group, the authors use \textit{K. Kato}'s theory of log structures [Am. J. Math. 116, No. 5, 1073--1099 (1994; Zbl 0832.14002)]. The second part--sections 5--8 deal with the construction of wild quotient singularities having at least two nodes. Let \(k\) be an algebraically closed field of characteristic \(p\), and let \(q=p^m\). Let \(C\) be the smooth affine curve \(y^q -y=x^{q+1}\) over \(k\). The authors define a particular group \(G\) of automorphisms of \(C\); this group has order \(q^3\) and is a Sylow-\(p\)-subgroup of the automorphism group of \(C\) when \(q\neq2\); moreover, \(G\) is a special \(p\)-group, i.e., its Frattini subgroup, its commutator subgroup and its center coincide. In Theorem 6.1 the authors show that the dual graph for the minimal resolution of singularities for the wild quotient singularity \(G/(C\times C)\) contains at least two nodes. In the remaining sections the authors study in more detail the special \(p\)-group \(G\). They determine, e.g., its higher ramification groups, and give results on the \(l\)-adic cohomology group \(H^1(C,\mathbb Q_l)\) where \(l\neq p\) is a prime and \(\mathbb Q_l\) is the field of \(l\)-adic numbers.