Last modified date: <%+ tp.file.last_modified_date() %>
- Tags: - #AG #symplectic #todo/too-long - Refs: - https://www.mathi.uni-heidelberg.de/~geodyn/teaching/Theses/MA_2020_Lukas%20D.%20Sauer%20(Albers).pdf#page=3 - Brice Loustau.The complex symplectic geometry of the deformation space of complex projective structures: https://arxiv.org/abs/1406.1821 - Equipped with TQFTs: - https://arxiv.org/pdf/2201.08699.pdf#page=1 - William M. Goldman, The modular group action on real SL(2)-characters of a one-holed torus, Geom. Topol. 7 (2003), 443–486. MR 2026539 - William M Goldman, An exposition of results of fricke, arXiv preprint math/0402103 (2004). - Links: - nonabelian Hodge correspondence - group cohomology - Higgs bundle - Fricke-Vogt theorem - semistable
character variety
Motivations
Relation to stable vector bundles, Teichmüller space. # Definitions
- Setup: $G\in \mathsf{Alg} {\mathsf{Grp}}_{/ {{\mathbf{C}}}} $ a reductive algebraic group and \(\pi\) a finitely generated group.
-
Defining the representation scheme:
- Define \begin{align*}{\mathfrak{X}}(\pi, G) \coloneqq\mathop{\mathrm{Hom}}_{\mathsf{Grp}}(\pi, G)\end{align*} as a set, the representation scheme.
- This admits the stricture of a scheme, and in fact an affine algebraic variety, induced by the structure of \(G\): since it is finitely generated, take a presentation \(\mathsf{Free}_m \to \mathsf{Free}_n \to G\to 0\), so \(G \cong \left\langle{\gamma_1,\cdots, \gamma_n \mathrel{\Big|}R_1,\cdots R_m}\right\rangle\), and embed \({\mathfrak{X}}(\pi, G) \hookrightarrow G{ {}^{ \scriptscriptstyle\times^{n} } }\) by \(\rho \mapsto \rho(\gamma_1) \times \cdots \times \rho(\gamma_n)\).
-
There is an action \(G \curvearrowright{\mathfrak{X}}(\pi, G)\) where \(g\cdot \rho({-}) \coloneqq g\rho({-})g^{-1}\), which is the action of an affine algebraic group on an algebraic set. “Taking orbits” via a stack-theoretic quotient yields the character stack
\begin{align*}\widetilde{{\mathcal{M}}}(\pi, G) \coloneqq{\mathfrak{X}}(\pi, G)/G\end{align*}
- As a stack, this is the functor \({\mathsf{Aff}}{\mathsf{Sch}}\to {\mathsf{Grpd}}\) where $\operatorname{Spec}R \mapsto {}_{R\otimes{\mathbf{C}}[\pi_1 \Sigma_g]}{\mathsf{Mod}} $ which are locally free and rank \(n\) over \(R\).
-
Comparing the character stack to the GIT (categorical) quotient:
- Note: this quotient is the usual definition of character variety.
- Let \({\mathcal{O}}({\mathfrak{X}}(\pi, G))\) be the regular functions on the representation scheme and let \({\mathcal{O}}({\mathfrak{X}}(\pi, G))^G\) be its \(G{\hbox{-}}\)invariant functions under the conjugation action above.
- When it exists, define GIT character stack \begin{align*}{\mathcal{M}}(\pi, G) \coloneqq{\mathfrak{X}}(\pi, G) { \mathbin{/\mkern-6mu/}}G \coloneqq\operatorname{Spec}{\mathcal{O}}\qty{{\mathfrak{X}}(\pi, G)}^G,\end{align*}
- A sufficient condition for existence: \(G\) being reductive.
- General fact: if \(X\) is finite type then \(X/G\) is again finite type?
- Finitely generated because \({\mathfrak{X}}(\pi, G)\) is affine and \(G\) is reductive.
- Often called a variety in the literature, but is generally not irreducible.
- There is a map \(\widehat{{\mathcal{M}}}(\pi, G) \to {{\mathcal{M}}}(\pi, G)\) which is a homeomorphism over the locus of irreps, and otherwise has fibers of reducible reps with equivalent semisimplification.
-
The term character variety is an eponym:
- For \(g\in \pi\), there is an associated character \(\chi_g: {\mathfrak{X}}(\pi, G) \to {\mathbf{C}}\) where \(\rho \mapsto \operatorname{tr}^2 \rho(g)\).
-
Define the character scheme as
\begin{align*}\chi(\pi, G) \coloneqq\operatorname{Spec}{\mathbf{C}} { \left[ \scriptstyle {\left\{{\chi_g {~\mathrel{\Big\vert}~}g\in G}\right\}} \right] } .\end{align*}
- For \(G ={\operatorname{SL}}_n, {\mathsf{Sp}}_{2n}, {\operatorname{SO}}_{2n+1}\), and \({\operatorname{PSL}}_2({\mathbf{C}})\), 1 \begin{align*}{\mathcal{M}}(\pi, G) \cong \chi(\pi, G).\end{align*}
- Why this is true for \({\operatorname{PSL}}_2({\mathbf{C}})\): the invariant regular functions \({\mathcal{O}}({\mathfrak{X}}(\pi, G))^G\) is finitely generated since \({\mathfrak{X}}(\pi, G)\) is affine and \(G\) is reductive, and is in fact generated by finitely many characters of the form \(\chi_g(\rho) \coloneqq{\mathrm{tr}}^2\rho(g)\). Thus \({\mathcal{M}}(\pi, G)\) is a moduli space of \({\mathbf{C}}{\hbox{-}}\)valued characters of \(\pi\) factoring through \(G\).
The tangent spaces
-
For \(\pi\) a surface group, the \(G{\hbox{-}}\)invariant locus \(\chi_G^{\mathrm{irr}}(\pi)\) of irreducible reps with closed \(G{\hbox{-}}\)orbits is a complex manifold with manifold tangent spaces \({\mathbf{T}}_\rho(\chi_G^{\mathrm{irr}}(\pi)) \cong H^1_{\mathsf{Grp}}(\pi, {\mathfrak{g}}_{{ \operatorname{Ad} }\rho})\).
- Given a nondegenerate bilinear form \(B:{\mathfrak{g}}{ {}^{ \scriptstyle\otimes_{{\mathbf{C}}}^{2} } }\), Goldman constructs a symplectic form \(\omega = \omega(B)\) making \((\chi_G^{\mathrm{irr}}, \omega)\) a symplectic manifold.
-
Can associate a larger scheme \(\widehat{\mathop{\mathrm{Hom}}}(\pi, G)\) with a larger structure sheaf whose closed points recover \(\mathop{\mathrm{Hom}}_{\mathsf{Grp}}(\pi, G)\) as a scheme.
- The Zariski tangent spaces are isomorphic to 1-cocycles in group cohomology: \({\mathbf{T}}_\rho \widehat{\mathop{\mathrm{Hom}}}_{\mathsf{Grp}}(\pi, G) \cong Z^1_{\mathsf{Grp}}(\pi, {\mathfrak{g}}_{{ \operatorname{Ad} }\rho})\).
Yields a complex orbifold:
Examples
-
For \(\Sigma_g\) a Riemann surface, there is an embedding of the Fricke space of marked hyperbolic structures \(F(\Sigma_g) \hookrightarrow{\mathfrak{X}}(\pi_1\Sigma_g, {\operatorname{PSL}}_2({\mathbf{R}}))\) (note that \(F(\Sigma_g) \cong {\mathcal{T}}(\Sigma_g)\), the Teichmüller space of complex structures on \(\Sigma_g\))
- This is induced by the fact that a marked hyperbolic surface has a marking \(\phi: \Sigma_g \to {\mathbb{H}}/\Gamma\) inducing \(\pi_1\Sigma_g \to \Gamma\hookrightarrow\mathop{\mathrm{Isom}}^+({\mathbb{H}}) { \, \xrightarrow{\sim}\, }{\operatorname{PSL}}_2({\mathbf{R}})\).
- The image is the connected component of faithful and discrete reps.
-
Of character stacks:
- \({\mathcal{M}}(\pi_1 \Sigma_g, \operatorname{GL}_n({\mathbf{C}})) = {\mathcal{M}}^{\operatorname{Betti}}_{g, n}\) is a Betti moduli space appearing the in nonabelian Hodge correspondence. Concretely, \begin{align*}{\mathcal{M}}_{g, n}^{\operatorname{Betti}}= \left\{{ \left\{{A_i, B_i}\right\}_{1\leq i \leq g} {~\mathrel{\Big\vert}~}\prod_{1\leq i \leq g}[A_i, B_i] = \operatorname{id}}\right\}/\operatorname{GL}_n({\mathbf{C}}),\end{align*} the moduli stack of representations of \(\pi_1 \Sigma_G\).
- \({\mathcal{M}}^{{\mathrm{ss}}}(\pi_1 \Sigma_g, {\operatorname{U}}_n)\) parameterizes semisimple reps, and polystable holomorphic bundles of degree zero on \(\Sigma_g\).
- \({\mathcal{M}}(\mathsf{Free}_2, {\operatorname{SO}}_2) \cong S^1\times S^1\), a torus
Notes
- For \(\pi = \pi_1\Sigma_g\) a surface group, the irreducible locus \({\mathcal{X}}^{\mathrm{irr}}(\pi, G)\) is a complex orbifold (Sik09) and Goldman computes \(\dim_{\mathbf{C}}{\mathcal{X}}^{\mathrm{irr}}(\pi, G) = (2g-2)\dim_{\mathbf{C}}G + 2\dim_{\mathbf{C}}Z(G)\) where \(Z(G)\) is the center of \(G\).
- For \(\pi\) a surface group, the smooth locus consists of irreducible reps.
As a subset of \(G^n\)
Universal representation algebra
The evaluation map
Good representations
Tangent spaces are group cocycles
See group cohomology
Proof
Algebraic proof
Proof of lemma:
Using this to prove the theorem:
Smooth locus
For surface groups
As a complex manifold
The symplectic pairing
See general pairings in group cohomology and
One can use the Killing form.
Showing closedness
Uses symplectic reduction and a moment map:
Including the boundary for 3-manifolds
See 3-manifold.
See antiholomorphic involution
Generalization to Kahlers
See Kahler manifolds and hard Lefschetz.
Main result: Lagrangian submanifold theorem
Ergodicity
Examples
A trace pairing:
Stable/irreducible locus
Relation to moduli
See moduli of curves and Fuchsian group:
See also mapping class group.