E-polynomials of \(\mathrm{SL}(2,\mathbb{C})\)-character varieties of complex curves of genus 3 (Q315843)

Let \(X\) be a smooth complex projective curve of genus \(g\geq1\) and \(G\) a complex reductive group. The moduli space \(\mathcal{M}(G)\) of semisimple representations of \(\pi_1(X)\) into \(G\) is called the \(G\)-character variety of \(X\) and is given explicitly as the GIT quotient \[ \mathcal{M}(G)=\left\{(A_1,B_1,\cdots,A_g,B_g)\in G^{2g}\left|\prod_{j=1}^g[A_j,B_j]=\text{Id}\right.\right\}\Bigg/\!\Bigg/G, \] where \(G\) acts by simultaneous conjugation. More generally, when \(G\) is \(\text{GL}(n,\mathbb{C})\) or \(\text{SL}(n,\mathbb{C})\), one can consider the twisted \(G\)-character varieties \[ \mathcal{M}^d(G)=\left\{(A_1,B_1,\cdots,A_g,B_g)\in G^{2g}\left|\prod_{j=1}^g[A_j,B_j]=e^{2\pi i/n}\text{Id}\right.\right\}\Bigg/\!\Bigg/G. \] The space \(\mathcal{M}^d(G)\) is also known as the Betti moduli space. Recall that the complex cohomology groups with compact supports of any complex quasi-projective variety \(Z\) possess mixed Hodge structures, leading to the definition of the Hodge numbers \(h^{k,p,q}_c(Z)=h^{p,q}(H^k_c(Z))\). The Hodge-Deligne polynomial (or E-polynomial) of \(Z\) is then defined by \(e(Z)(u,v):=\sum_{p,q,k}(-1)^kh_c^{k,p,q}(Z)u^pv^q\). One can recover both topological and geometrical information about \(Z\) from \(e(Z)\). A key fact about the E-polynomials, which assists in computing them, is that, if \(Z\) is expressed as a disjoint union of locally closed subvarieties \(Z_i\), then \(e(Z)=\sum e(Z_i)\). When \(h_c^{k,p,q}(Z)=0\) for \(p\neq q\), \(e(Z)\) is expressible as a polynomial in the single variable \(q:=uv\); this is the case with all the varieties considered in this paper. \textit{T. Hausel} and \textit{F. Rodriguez-Villegas} [Invent. Math. 174, No. 3, 555--624 (2008; Zbl 1213.14020)] began the study of the E-polynomials of \(G\)-character varieties and in particular calculated \(e(\mathcal{M}(G))\) for \(G=\text{GL}(n,\mathbb{C})\). Following this, \textit{M. Mereb} [Math. Ann. 363, No. 3--4, 857--892 (2015; Zbl 1333.14041)] studied the case \(G=\text{SL}(n,\mathbb{C})\) and obtained in particular an explicit formula for \(e(\mathcal{M}^1(\text{SL}(2,\mathbb{C}))\). The methods employed were arithmetic. The authors of the current paper and the reviewer [Rev. Mat. Complut. 26, No. 2, 635--703 (2013; Zbl 1334.14006)] used geometric arguments to obtain formulae for the E-polynomials of character varieties for \(g=1\) and \(g=2\). For the same values of the genus, they also computed the E-polynomials for the twisted character varieties of \(\mathcal{M}^\xi(\text{SL}(2,\mathbb{C}))\) (corresponding to representations for which \(\prod_{j=1}^g[A_j,B_j]=\xi\), where \(\xi\) is an arbitrary element of \(\text{SL}(2,\mathbb{C})\)). The methods were geometric, using stratifications and explicit matrix formulae for elements of each stratum. In the present paper, the authors use these geometric methods to calculate the E-polynomials of \(\mathcal{M}(\text{SL}(2,\mathbb{C}))\) and \(\mathcal{M}^1(\text{SL}(2,\mathbb{C}))\). The stratification employed is similar to that of the previous paper, but the arguments involving Hodge monodromy fibrations have to be extended to a more complicated situation (the earlier paper required only fibrations over one-dimensional bases, whereas here the bases can be two-dimensional). For \(\mathcal{M}(\text{SL}(2,\mathbb{C})\), one also has to deal with reducible representations. Here the method of the earlier paper involves a large number of strata and it turns out to be easier to obtain the E-polynomial of the irreducible locus by subtracting that of the reducible locus from that of the whole. As a byproduct of the analysis, the authors are able to compute the E-polynomial of \(\mathcal{M}^\xi(\text{SL}(2,\mathbb{C}))\) when \(\xi=\begin{pmatrix} \lambda & 0\\ 0 & \lambda^{-1} \end{pmatrix}\). Since completing this paper, the authors have succeeded in computing the E-polynomials of \(\text{SL}(2,\mathbb{C})\)-character varieties for all \(g\geq3\) and all \(\xi\) [the authors, Int. Math. Res. Not. 2016, No. 3, 926--961 (2016; Zbl 1353.14013)]. This does not supersede the current paper as the case \(g=3\) presents special features and also serves as the basis for an induction on \(g\).