Quilts and \(T\)-systems (Q1895590)

Let \(G\) be a finite group. In the case \(G =\) Monster, \textit{S. P. Norton} [Generalized moonshine, Proc. Symp. Pure Math. 47, 208-209 (1987; Zbl 0642.10025)] proposed to associate to pairs \((g,h) \in G \times G\) holomorphic functions \(F(g,h;z)\) on the complex upper half \(z\)-plane enjoying \(F(g,gh,z) = F(g,h,z + 1)\), \(F(h,g^{-1}, z) = F(g,h, - 1/z)\) and the invariance under conjugation: \(F(g^\gamma, h^\gamma, z) = F(g,h,z)\). Moreover, Norton briefly indicated an idea of constructing from \((g,h) \in G \times G\) a trivalent polyhedron homeomorphic to the Riemann surface associated with the invariance group of the expected \(F(g,h;z)\), in order to suggest a ``possible link between the Monster and the genus 0 property.'' In this paper, the combinatorial process of this construction is clarified and established in detail. An orbit of the \(B_3\) (the 3-string braid group)-action on \(G \times G\) generated by \((g,h) \mapsto (g,gh)\), \((g,h) \mapsto (gh,h)\) is called a \(T\)-system. For each given \(T\)-system, the authors construct a certain diagram (2-complex plus additional data), called a quilt, by associating to each suborbit by the center of \(B_3\) a half-edge of the diagram called a ``seam''. The main result states that the \(T\)-systems correspond bijectively to the (equivalence classes of) quilts. Examples show that the notion of quilts is efficient even when identifying fairly sizeable \(T\)-systems.