Author: Eiko

Tags: geometry, algebraic geometry, quiver variety, representation theory

Time: 2024-11-10 13:23:04 - 2024-11-11 00:58:34 (UTC)

I might reorganize these information into separate pieces, but now let me put them all together here.

Sources We Need To Look At

  • Prop 2.8 in Nakajima’s book on Hilbert Schemes

  • Prop 6.2 of Le Bruyn

  • Lemma 2.2 of Punctual Hilbert Schemes for Kleininan Singularities as Quiver Varieties

What we expect: Prop 2.8 + Prop 6.2 should give us the claim that for \(\theta\in F\), every closed orbit in the \(\theta\)-semistable locus of \(\mu^{-1}(0)\) has \(j=0\).

Prop 2.8

Lemma 2.2

Variation of GIT Parameters

Let \(\Theta_v = \{\theta:\mathbb{Z}^{Q_0}\to \mathbb{Q}: \theta(v)=0\}\) be the linear span of characters, where characters of \(G\), \(\chi_\theta=\prod_i \det(g_i)^{\theta_i}:G\to \mathbb{C}^\times\) are the integral sub-lattice of \(\Theta_v\).

  • Recall that given parameter \(\theta\in \Theta_v\), a \(\Pi\)-module is \(\theta\)-stable if \(\theta(M)=0\) and \(\theta(M')>0\) for proper subrepresentations. Use \(\ge 0\) for \(\theta\)-semistable.

  • Two \(\theta\)-semistable modules \(M,N\) are semiequivalent if their orbit closure meet in semistable locus, representation theoretically, this means there are semistable filtrations

    \[0=M_0\subset \dots \subset M_{s_1}=M, \quad 0=M_0'\subset \dots\subset M_{s_2}'=M'\]

    whose sum of factors are isomorphic \(\bigoplus M_i/M_{i-1} \cong \bigoplus M'_j/M'_{j-1}\).

    (This hints us that orbit closures are disjoint unions of orbits, these sum of factors are limits of the biggest orbit.)

  • Every semiequivalence class has a unique \(\theta\)-polystable representative, which are the direct sum of \(\theta\)-stable modules.

Consider \(\mathcal{M}_\theta = (\mu^{-1}(0)/\!\!/_\theta G)_{red}\), the GIT quotient with reduced scheme structure, it is the categorical quotient of the locus of \(\chi_\theta\)-semistable points of \(\mu^{-1}(0)\) by \(G\). Also identified as the coarse moduli space of \(\theta\)-semistable representations of \(\Pi\) of dimension \(v\).

It is known that every \(\theta\in \Theta_v\) gives an irreducible and normal \(\mathcal{M}_\theta\) with symplectic singularities.

Remarks

  • The stability only depends on the dimensions of subrepresentations of the representations.

The Lemma On Surjectivity Of Variation

Lemma Let \(\theta\ge \theta'\in \Theta_v\), then \(\pi : \mathcal{M}_\theta \to \mathcal{M}_{\theta'}\) obtained by variation of GIT quotient is a surjective, projective and birational morphism of varieties over \(\mathrm{Sym}^n(\mathbb{C}^2/\Gamma)\).

Prop 6.2

Notations

  • \(\Theta_v\) are the vectors that have \(0\) paring with \(v\),

  • \(\Pi\) is the preprojective algebra of the framed McKay quiver, whose ideal is generated by

    \[ ij=0, \quad ji + \sum aa^* - \sum a^*a = 0 \]

  • \(A\) is the algebra \(\Pi\) quotient by \(j\), i.e. the subscheme where \(j=0\).

  • \(\mathcal{M}_\theta = \mathrm{Rep}(\Pi,v)/\!\!/_\theta G(v)\). It is reduced and irreducible, and is the coarse moduli space of semi-equivalence classes of \(\theta\)-semistable representations of \(\Pi\).

  • \(F=\{\theta\in \Theta_v : \theta(\delta)>0, \theta(\rho_i)>0, 1\le i\le r\}\), is the fundamental domain.

  • The full GIT chamber decomposition of \(\Theta_v\) is described in [2], but we can restrict to those characters lie in \(\overline{F}\) by the action of Namikawa-Weyl group. For each chamber \(C\subset F\), variation of GIT quotient defines a projective symplectic (hence crepant) resolution \(\mathcal{M}_C\to \mathcal{M}_0\) of singularities.

The Proposition

Prop. For \(\theta\in F\), The subscheme \(\mathrm{Rep}(A,v)\subset \mathrm{Rep}(\Pi,v)\) gives an isomorphism of schemes \(\mathrm{Rep}(A,v)/\!\!/_\theta G(v) \cong \mathcal{M}_\theta\), as an isomorphism over \(\mathcal{M}_0 = \mathrm{Sym}^n(\mathbb{C}^2/\Gamma)\).

That (seems to) means, if your \(\theta\) is generic and in fundamental domain, any representation in \(\mathcal{M}_\theta\), or \(\theta\)-semistable representation can be represented by a \(\Pi\)-module that has \(j=0\).

Question: Does this mean any \(\theta\)-semistable representation in \(\mathrm{Rep}(\Pi,v)\) is semi-equivalent to a \(\Pi\)-module with \(j=0\)?

The Proof

  • For \(\theta\in F\) generic, \(C\) be the GIT chamber containing \(\theta\).

  • If \(\theta\in C_+\),