Representations of the Kauffman bracket skein algebra III: closed surfaces and naturality (Q2422570)

This paper is the third entry in the authors' series devoted to analysis of finite dimensional representations of the Kauffman bracket skein algebra \(\mathcal{S}^A(S)\) of a surface \(S\), where \(A\) is a parameter. In previous works it was shown that when \(A\) is an \(N\)-th primitive root of \(-1\) and \(S\) is a closed oriented surface every irreducible finite dimensional representation \(\rho\) corresponds to a unique character \(r_\rho\) called its classical shadow. For the Witten-Reshetikhin-Turaev quantum group based representation the shadow is the trivial character. The main result of the present paper is a proof of the converse -- to every character \(r:\pi(S)\to SL_2(\mathbb{C})\) (modulo trace equivalence) there corresponds an irreducible representation. An alternative construction based on Frobenius algebras was offered by \textit{N. Abdiel} and \textit{C. Frohman} [Algebr. Geom. Topol. 17, No. 6, 3341--3373 (2017; Zbl 1421.57015) and J. Knot Theory Ramifications 25, No. 4, Article ID 1650016, 25 p. (2016; Zbl 1378.57019)]. Frohman, Kania-Bartoszynska and Lê [\textit{C. Frohman} et al., Invent. Math. 215, No. 2, 609--650 (2019; Zbl 1491.57014)] proved that the correspondence is \(1\)-\(1\) on an open Zariski dense subset (it fails for non-generic characters). The proof strategy consists in considering a sufficiently complicated triangulation \(\lambda\) of \(S\) and puncturing \(S\) at its vertices (``drill, baby, drill'', as the authors nickname it) to obtain a representation \(\rho_\lambda\) of the punctured surface \(S_\lambda\), which is much easier. Surprisingly, the representation space of \(\rho_\lambda\) contains a non-trivial invariant subspace \(F_\lambda\), called the total off-diagonal kernel, such that the restriction of \(\rho_\lambda\) to an irreducible component of it gives the sought after irreducible representation \(\rho_r\) of \(\mathcal{S}^A(S)\) (the authors conjecture that generically \(F_\lambda\) is already irreducible). For generic \(r\) the \(\rho_r\) does not depend, up to isomorphism, on \(\lambda\) and other choices, but in general \(\rho_r\) is determined only up to composing with a sign-reversal symmetry. The proof is based on checking invariance under the Pachner moves on triangulations. The authors also obtain surprisingly explicit lower bounds on the dimension of \(F_\lambda\), namely \(\dim\geq N^{3(g-1)}\) for \(g\geq2\), \(\geq N\) for \(g=1\) and \(\geq1\) for \(g=0\), where \(g\) is the genus of \(S\). On an explicit open Zariski dense subset the inequalities turn into equalities, and the authors conjecture that this is always the case.