Graphical splittings of Artin kernels (Q2037523)

Let \(\Gamma\) be a graph with the set of vertices \(V(\Gamma )\) and the set of edges \(E(\Gamma )\). As it is well known, the right-angled Artin group (RAAG) associated to \(\Gamma\) is the group with presentation \(A_{\Gamma}=\langle V(\Gamma )\mid [x,y]=1 \text{ for all} \{x, y\}\in E(\Gamma )\rangle\). As RAAGs interpolate between free groups and free abelian groups, they have been studied extensively but the study is not exhausted. They display surprisingly rich geometric behavior and a very diverse collection of subgroups. An Artin kernel is a normal subgroup \(K\) of \(A_{\Gamma}\) such that the quotient \(A_{\Gamma} /K\) is infinite cyclic. In [\textit{M. Bestvina} and \textit{N. Brady}, Invent. Math. 129, No. 3, 445--470 (1997; Zbl 0888.20021)], the Artin kernel of the homomorphism \(A_{\Gamma}\longrightarrow\mathbb{Z}\) obtained by sending every vertex to \(1\in \mathbb{Z}\) is studied. In [\textit{K.-U. Bux} and \textit{C. Gonzalez}, J. Lond. Math. Soc., II. Ser. 60, No. 3, 793--801 (1999; Zbl 1025.20026); \textit{J. Meier} et al., Comment. Math. Helv. 73, No. 1, 22--44 (1998; Zbl 0899.57001); \textit{J. Meier} and \textit{L. VanWyk}, Proc. Lond. Math. Soc. (3) 71, No. 2, 263--280 (1995; Zbl 0835.20037); \textit{S. Papadima} and \textit{A. I. Suciu}, Adv. Math. 220, No. 2, 441--477 (2009; Zbl 1208.57002)] have been studied finiteness properties of Artin kernels considering more general homomorphisms from RAAGs to infinite cyclic groups, induced by mapping the vertices of \(\Gamma\) to more collections of numbers. Unlike right-angled Artin groups themselves, Artin kernels are very far away from being classified by their underlying graph. A characterization of Artin kernels as far their finite generation is given in [\textit{J. Meier} and \textit{L. VanWyk}, Proc. Lond. Math. Soc. (3) 71, No. 2, 263--280 (1995; Zbl 0835.20037)] and in [\textit{R. Bieri} et al., Invent. Math. 90, 451--477 (1987; Zbl 0642.57002)] (see Corollary 2.13 in the present paper). Before we state the main results of the paper under consideration we need some more terminology and quote the known results. Let \(\Gamma _{1},\Gamma _{2},\,\Gamma _{3}\) be pairwise distinct non-empty full subgraphs of a graph \(\Gamma\) such that \(\Gamma = \Gamma _{1}\cup \Gamma _{2}\) and \(\Gamma _{1}\cap \Gamma _{2}= \Gamma _{3}\), then it is said that the triple \((\Gamma _{1},\Gamma _{2},\Gamma _{3})\) is a splitting of \(\Gamma\). If \(\Gamma _{3}\) is connected, then \((\Gamma _{1},\Gamma _{2},\Gamma _{3})\) is a connected splitting of \(\Gamma\). It is easy to check that a splitting of \(\Gamma\) induces a splitting of \(A_{\Gamma }\) as \(A_{\Gamma }=A_{\Gamma _{1}} \ast_{ A_{\Gamma _{3}}} A_{\Gamma _{2}}\). Theorem A. (Theorem 3.2 in the paper) Let \((\Gamma _{1},\Gamma _{2},\Gamma _{3})\) be a splitting of \(\Gamma\), let \(f:A_{\Gamma } \longrightarrow \mathbb{R}\) be discrete, and let \(f_{k}:A_{\Gamma } \longrightarrow \mathbb{R}\) be its restriction for \(k=1,2,3\). Then \(\ker f\) splits as a graph of groups \(\mathcal{G}\) with group vertices isomprphic to \(\ker (f_{k})\) for \(k=1,2\) and group edges isomprphic to \(\ker (f_{3})\). The number of the group vertices and group edges in the decomposition above is \([f(A_{\Gamma}):f(A_{\Gamma _{k}})]\) and \([f(A_{\Gamma}):f(A_{\Gamma _{3}})]\), respectively. In particular, \begin{itemize} \item[(1)] If \(f_{3}\not =0\), then \(\ker f\) splits as a finite graph of groups \(\mathcal{G}\). \item[(2)] If \(f_{3}=0\), then \(\ker f\) surjects onto \(F_{\infty}\). \end{itemize} Interesting particular cases are developed in Corollaries 3.4 and 3.6. The two cases in the theorem above are not disjoint from each other, i.e. the same character \(f\) can display tame behavior with respect to one splitting and wild behavior with respect to another (see Remark 3.3 in the paper). In order to obtain a complete dichotomy, the authors restrict to chordal graphs. A graph \(\Gamma\) is chordal (triangulated graph) if, for every cycle of length at least four, there is a chord, i.e. an edge connecting two non-adjacent vertices of the cycle. Theorem B. (Theorem 3.9 in the paper) Let \(\Gamma\) be a chordal graph, and let \(f:A_{\Gamma } \longrightarrow \mathbb{R}\) be discrete. Then exactly one of the following holds: \begin{itemize} \item[(1)] (wild) \(\ker f\) surjects onto \(F_{\infty}\). \item[(2)] (tame) \(\ker f\) splits as a finite graph of groups \(\hat{\mathcal{G}}\) with finitely generated free abelian edge and vertex groups. \end{itemize} When the graph is not chordal, it is possible to construct characters that do not obey the tame-wild dichotomy. See Proposition 3.11 in the paper. The rest of the paper is devoted to obtain explicit decompositions and rank formulas for tame Artin kernels over graphs with cut vertices. See Proposition 3.14 and Theorem 3.17 in the paper.