Reflection triangles in Coxeter groups and biautomaticity (Q5314950)

Let \((W,S)\) be a Coxeter system of finite rank (i.e., \(|S|<\infty\)). Let \(t_1,t_2,t_3\) be three reflections in \(W\). The set \(T=\{t_1,t_2,t_3\}\) is called a reflection triangle if the order of \(t_it_j\) is finite for all \(i\neq j\) and if \(T\) is not contained in a rank 2 parabolic subgroup of \(W\). It is known that, given a triangle \(T\), there exists a triangle \(T'\) such that \(\langle T\rangle=\langle T'\rangle\) and \({\mathcal M}(T):=(\langle T\rangle,T')\) is a Coxeter system whose type is determined uniquely by \(T\). Say \(T\) is affine if \({\mathcal M}(T)\) is affine. The main result of the present paper is to prove that when \((W,S)\) contains no affine reflection triangle, there are only finitely many conjugacy classes of reflection triangles. By applying some results of Krammer and Deodhar [see \textit{D. Krammer}, ``The conjugacy problem for Coxeter groups'' Ph.D. thesis, Univ. Utrecht (1994); \textit{V. V. Deodhar}, Commun. Algebra 10, 611-630 (1982; Zbl 0491.20032)], the authors further show that the converse is also valid, i.e., if there are only finitely many conjugacy classes of reflection triangles in \((W,S)\), then \((W,S)\) contains no affine reflection triangle. Let \({\mathcal K}(W,S)\) be the Coxeter cubing associated with \((W,S)\). This is a locally finite, finite-dimensional CAT(0) cube complex on which \(W\) acts properly discontinuously. Then by applying the main result of the paper, the authors prove the following result: the action of the Coxeter group \(W\) on \({\mathcal K}(W,S)\) is cocompact if and only if \((W,S)\) contains no affine reflection triangle. When the equivalent conditions hold, the Coxeter group \(W\) is biautomatic. Note that \textit{B. Brink} and \textit{R. B. Howlett} [in Math. Ann. 296, No. 1, 179-190 (1993; Zbl 0793.20036)], showed that every Coxeter group of finite rank is automatic. However, the question of determining whether or not Coxeter groups satisfy the stronger condition of biautomaticity remains open.