Introduction
Understanding algebraic cycles on moduli spaces of vector bundles over curves is a classical problem in algebraic geometry. While the cohomology of these moduli spaces are comparatively well-understood, describing their Chow groups is substantially more subtle. One approach is to study the Chow groups of the universal moduli spaces, where the variation of the underlying curve imposes additional rational equivalences on algebraic cycles. This leads naturally to Franchetta-type questions: to what extent are cycles on a universal family determined by their cohomology classes on individual fibers. The following conjecture was made by the second author in [31 ] .
Conjecture A ([31 ] ). Let V e c H ( r , d , g ) \mathcal{V}ec^{\mathcal{H}}(r,d,g) V e c H ( r , d , g ) be the universal moduli stack of vector bundles of rank r r r , degree d d d over hyperelliptic curves of genus g g g . For any g ≥ 2 g \ge2 g ≥ 2 , d ∈ Z d \in\mathbb{Z} d ∈ Z and a hyperelliptic curve C C C of genus g g g , the composition of restriction and cycle-class map
A ∗ ( V e c H ( r , d , g ) ) ⟶ H ∗ ( Bun r , d ( C ) ) A^*(\mathcal{V}ec^{\mathcal{H}}(r,d,g)) \longrightarrow H^*(\operatorname{Bun}_{r,d}(C)) A ∗ ( V e c H ( r , d , g )) ⟶ H ∗ ( Bun r , d ( C )) is an isomorphism onto H κ ∗ H^*_{\kappa} H κ ∗ where H κ ∗ H^*_{\kappa} H κ ∗ is the cohomology subring generated by twisted-kappa classes.
In loc. cit. the second author proves Conjecture A for rank 2 bundles on genus 2 curves with arbitrary degree. In this paper, we study this question for rank 2 bundles over smooth hyperelliptic curves of any genus g ≥ 2 g \ge2 g ≥ 2 . Our main result is the following.
Theorem 0.1. Conjecture A is true for r = 2 r = 2 r = 2 , any d ∈ Z d \in\mathbb{Z} d ∈ Z and g ≥ 2 g \ge2 g ≥ 2 .
We note that Conjecture A is false over M g \mathcal{M}_g M g for g ≥ 3 g \ge3 g ≥ 3 because of the Gross-Schoen cycle (see Remark 4.3 ). We also obtain an explicit description of the rational Chow ring of V e c H ( 2 , d , g ) \mathcal{V}ec^{\mathcal{H}}(2,d,g) V e c H ( 2 , d , g ) when d d d is odd in terms of generators and relations among twisted κ \kappa κ classes. In particular, using a monodromy argument, we obtain the following theorem.
Theorem 0.2. Let π : V e c H ( 2 , d , g ) ⟶ H g \pi: \mathcal{V}ec^{\mathcal{H}}(2,d,g) \longrightarrow\mathcal{H}_g π : V e c H ( 2 , d , g ) ⟶ H g be the universal stack. Fix a hyper-elliptic curve C ∈ H g C \in\mathcal{H}_g C ∈ H g . Then we have the isomorphism of graded rings
A ∗ ( V e c H ( 2 , d , g ) ) → ≅ H 2 ∗ ( Bun 2 , d ( C ) , Q ) π 1 ( H g , [ C ] ) A^*(\mathcal{V}ec^{\mathcal{H}}(2,d,g)) \xrightarrow{\cong} H^{2*}(\operatorname{Bun}_{2,d}(C),\mathbb{Q})^{\pi_1(\mathcal{H}_g,[C])} A ∗ ( V e c H ( 2 , d , g )) ≅ H 2 ∗ ( Bun 2 , d ( C ) , Q ) π 1 ( H g , [ C ])
The central result of [24 ] may be re-stated as follows.
Theorem ([24 ] ). The composition of restriction and the cycle class map gives the following isomorphism of graded rings for any hyper-elliptic curve C C C of genus g g g ,
A ∗ ( V e c H ( 1 , d , g ) ) → ≅ H 2 ∗ ( Bun 1 , d ( C ) , Q ) π 1 ( H g , [ C ] ) . A^*(\mathcal{V}ec^{\mathcal{H}}(1,d,g)) \xrightarrow{\cong} H^{2*}(\operatorname{Bun}_{1,d}(C),\mathbb{Q})^{\pi_1(\mathcal{H}_g,[C])}. A ∗ ( V e c H ( 1 , d , g )) ≅ H 2 ∗ ( Bun 1 , d ( C ) , Q ) π 1 ( H g , [ C ]) .
So Theorem 0.2 is a natural extension of Larson’s result and we give a relatively simpler proof of Larson’s result (see Theorem 6.1 ).
A central ingredient in our argument is a relative version of the motivic decomposition for moduli spaces of stable bundles as in [11 ] . We extend this decomposition to families of curves over a smooth base. More precisely, let π : C → B \pi: \mathcal{C} \to B π : C → B be a smooth projective family of curves of genus g ≥ 2 g \ge2 g ≥ 2 , where B B B is smooth. U C / B ( 2 , d ) \mathcal{U}_{\mathcal{C}/B}(2,d) U C / B ( 2 , d ) is the relative moduli space of stable rank 2 bundles of odd degree d d d . Write J ( C / B ) \mathcal{J}(\mathcal{C}/B) J ( C / B ) for the relative Jacobian and C ( n ) \mathcal{C}^{(n)} C ( n ) for the relative symmetric powers. We prove the following result in Section 2 (see Theorem 2.6 ).
Theorem 0.3. Let U B : = U C / B ( 2 , d , g ) U_B := \mathcal{U}_{\mathcal{C}/B}(2,d,g) U B := U C / B ( 2 , d , g ) be the relative moduli space of rank two stable bundles of degree d d d with d d d odd. We have the following isomorphism in M ( B ) \mathrm{M}(B) M ( B ) , the category of relative smooth
Chow motives:
h B ( U B ) ≃ h B ( Jac ( C / B ) ) ⊗ ( h B ( [ C B ( g − 1 ) ] ( g − 1 ) ) ⊕ ⨁ i = 0 g − 2 h B ( [ C B ( i ) / S i ] ) ⊗ ( Q ( i ) ⊕ Q ( 3 g − 3 − 2 i ) ) ) . h_B(U_B) \simeq h_B(\operatorname{Jac}(\mathcal{C}/B)) \otimes\left( h_B\left([\mathcal{C}_B^{(g-1)}](g-1)\right) \oplus\bigoplus_{i=0}^{g-2} h_B\left([\mathcal{C}_B^{(i)}/S_i]\right) \otimes\left(\mathbb{Q}(i) \oplus\mathbb{Q}(3g-3-2i)\right) \right). h B ( U B ) ≃ h B ( Jac ( C / B )) ⊗ ( h B ( [ C B ( g − 1 ) ] ( g − 1 ) ) ⊕ i = 0 ⨁ g − 2 h B ( [ C B ( i ) / S i ] ) ⊗ ( Q ( i ) ⊕ Q ( 3 g − 3 − 2 i ) ) ) .
We apply Theorem 0.3 to the universal curve over the hyperelliptic locus and results from [33 ] to establish a strong Franchetta Property (see Definition 1.3.1) for d d d odd. The identification of the image of the cycle class map on the cohomology of the fibers is taken up in § 8 . We give two proofs of surjectivity. In the first proof, we use representation theory of the symplectic group Sp 2 g \operatorname{Sp}_{2g} Sp 2 g to compute dimensions of A i ( J 2 , g d × H g J 2 , g d ′ ) A^i(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'}) A i ( J 2 , g d × H g J 2 , g d ′ ) and A i ( C 2 , g ( n ) × H g J 2 , g g ) 1 ≤ n ≤ g + 6 A^i(\mathcal{C}_{2,g}^{(n)} \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^g)_{1\leq n\leq g+6} A i ( C 2 , g ( n ) × H g J 2 , g g ) 1 ≤ n ≤ g + 6 and use a monodromy argument to identify the image. This also enables us to get an explicit Hilbert series for U H ( 2 , d , g ) \mathcal{U}^H(2,d,g) U H ( 2 , d , g ) when d d d is odd, giving another numerical proof of surjectivity using the strong Franchetta property. The case d d d is even is taken up in § 9 . We study the moduli stack H g , n w \mathcal{H}_{g,n}^w H g , n w of hyperelliptic curves with n n n distinct marked points and an additional Weierstrass point which can possibly coincide with one of the n n n -markings. We show the following, after a modification of the arguments in [8 ] (see Theorem 7.3 ).
Theorem 0.4. For all g ≥ 2 g \geq2 g ≥ 2 , 0 ≤ n ≤ 2 g + 4 0 \leq n \leq2g+4 0 ≤ n ≤ 2 g + 4 , the pullback map A ∗ ( H g , n ) → A ∗ ( H g , n w ) A^*(\mathcal{H}_{g,n}) \to A^*(\mathcal{H}_{g,n}^w) A ∗ ( H g , n ) → A ∗ ( H g , n w ) is an isomorphism.
This enables us to prove the strong Franchetta property and Conjecture A for r = 2 r=2 r = 2 and all even d d d .
Acknowledgements. SG would like to thank his PhD advisor, Junliang Shen, for his interest in the project, and David Bai and David Fang for helpful discussions. SS would like to thank his PhD advisor, Elham Izadi, for her guidance and support, and Qizheng Yin and Samir Canning for helpful discussions.
Acknowledgement of AI use. The authors acknowledge the use of AI in the preparation of this manuscript. Google Gemini 3.1 Deep Think model via Perplexity was used to assist with the proofs of Theorems 6.6 , 6.12 , and the code in Figure 1. Although the proof of Theorem 6.6 is the authors’ own, the discussions helped simplify the proof of Theorem 6.3 . We also give a proof of the main theorems independent of Theorem 6.12 . The authors also acknowledge the usage of Open AI’s ChatGPT Plus to do an extensive literature search on the representation theory of the symplectic group used in the proofs of § 8 . All the mathematical arguments and proofs appearing in the paper were written and verified by the authors. The authors take sole responsibility for all statements and arguments.
Notation
All schemes in this paper are defined over the field of complex numbers. All Chow rings and groups are taken with rational coefficients. We use the subspace convention for projective bundles.
We denote the locus of hyperelliptic curves of genus g g g with n n n -marked points with an additional marked Weierstrass point as H g , n w ↪ H g , 1 × H g H g , n \mathcal{H}_{g,n}^w \hookrightarrow\mathcal{H}_{g,1}\times_{\mathcal{H}_g}\mathcal{H}_{g,n} H g , n w ↪ H g , 1 × H g H g , n , and put C 2 , g , n w \mathcal{C}_{2,g,n}^w C 2 , g , n w as the universal curve over H g , n w \mathcal{H}_{g,n}^w H g , n w with the Weierstrass section { σ g , i , n : H g , n w → C 2 , g , n w } \{\sigma_{g,i,n}:\mathcal{H}_{g,n}^w\to\mathcal{C}_{2,g,n}^w\} { σ g , i , n : H g , n w → C 2 , g , n w } .
Analogous to [8 ] , we set D i , j : = { ( C , p 1 , … , p n ) ∈ H g , n ∣ p i = p j ‾ } D_{i,j}:=\{(C,p_1,\ldots,p_n)\in\mathcal{H}_{g,n}\mid p_i=\overline{p_j}\} D i , j := {( C , p 1 , … , p n ) ∈ H g , n ∣ p i = p j } which is the locus where p i p_i p i and p j p_j p j are conjugate under the hyper-elliptic involution.
For a fibre product ∏ i ∈ I / S X i \prod_{i\in I/S}X_i ∏ i ∈ I / S X i over a scheme S S S , and a subset J ⊂ I J\subset I J ⊂ I , we denote the projection onto the sub-product ∏ j ∈ J / S X j \prod_{j\in J/S}X_j ∏ j ∈ J / S X j by p J p_J p J .
By Hilbert Series of a graded Q \mathbb{Q} Q -vector space V ∙ V^\bullet V ∙ , we refer to the series hilbser ( V ∙ ) : = ∑ i ≥ 0 dim Q ( V i ) T i \operatorname{hilbser}(V^\bullet):=\sum_{i\geq0}\dim_{\mathbb{Q}}(V^i)T^i hilbser ( V ∙ ) := ∑ i ≥ 0 dim Q ( V i ) T i where T T T is an indeterminate. For a stack X \mathcal{X} X , we refer to the series hilbser ( X ) : = ∑ i ≥ 0 T i dim A dim X − i ( X ) \operatorname{hilbser}(\mathcal{X}):=\sum_{i\geq0}T^i\dim A^{\dim\mathcal{X}-i}(\mathcal{X}) hilbser ( X ) := ∑ i ≥ 0 T i dim A d i m X − i ( X ) .
For a stack X \mathcal{X} X , H a l g ∗ ( X ) H^*_{\mathrm{alg}}(\mathcal{X}) H alg ∗ ( X ) denotes the algebraic cohomology subring of H ∗ ( X ) H^*(\mathcal{X}) H ∗ ( X ) , i.e., the image of the cycle class map A ∗ ( X ) → H ∗ ( X ) A^*(\mathcal{X})\to H^*(\mathcal{X}) A ∗ ( X ) → H ∗ ( X ) .
We denote the universal Jacobian over hyperelliptic curves of genus g g g with degree d d d as J 2 , g d \mathcal{J}^{d}_{2,g} J 2 , g d and its rigidification by J 2 , g d \mathcal{J}^{d}_{2,g} J 2 , g d . The associated G m \mathbb{G}_m G m -rigidification map is denoted by ν 1 , d : J 2 , g d → J 2 , g d \nu_{1,d}:\mathcal{J}^{d}_{2,g}\to\mathcal{J}^{d}_{2,g} ν 1 , d : J 2 , g d → J 2 , g d .
Preliminaries
G m \mathbb{G}_m G m -gerbes
For technical simplicity, we shall use the following notion of a torsion-G m \mathbb{G}_m G m -gerbe.
Definition 1.1.1 ([32 ] ). A torsion-G m \mathbb{G}_m G m -gerbe is a gerbe Y → X \mathcal{Y}\to\mathcal{X} Y → X , banded by G m \mathbb{G}_m G m , such that the associated cohomology (Brauer-Grothendieck) class is torsion. That is, [ Y / X ] ∈ H 2 ( X e ˊ t , G m ) t o r s [\mathcal{Y}/\mathcal{X}]\in H^2(\mathcal{X}_{\mathrm{\acute{e}t}},\mathbb{G}_m)_{\mathrm{tors}} [ Y / X ] ∈ H 2 ( X e ˊ t , G m ) tors .
Note that any G m \mathbb{G}_m G m -gerbe defined over a regular, integral, and noetherian scheme is a torsion-G m \mathbb{G}_m G m -gerbe [15 Corollaire 1.8] . The rigidification map J g d → J g d \mathcal{J}^{d}_{g}\to\mathcal{J}^{d}_{g} J g d → J g d is also a torsion-G m \mathbb{G}_m G m -gerbe by [29 Theorem 6.4] .
Theorem 1.1 ([24 ] ). Let Y → X \mathcal{Y}\to\mathcal{X} Y → X be a torsion-G m \mathbb{G}_m G m -gerbe. There exists c 1 ∈ A 1 ( Y ) c_1\in A^1(\mathcal{Y}) c 1 ∈ A 1 ( Y ) such that
A ∗ ( Y ) ≅ A ∗ ( X ) [ c 1 ] . A^*(\mathcal{Y})\cong A^*(\mathcal{X})[c_1]. A ∗ ( Y ) ≅ A ∗ ( X ) [ c 1 ] .
Proof. Let n ∈ N n\in\mathbb{N} n ∈ N be such that n [ Y / X ] = 0 n[\mathcal{Y}/\mathcal{X}]=0 n [ Y / X ] = 0 . Applying [27 Lemma 3.1.1.8] , there exists a μ n \mu_n μ n -gerbe Y → Y n \mathcal{Y}\to\mathcal{Y}_n Y → Y n such that
Y → Y n ≅ X × B G m ↓ X \begin{CD}
\mathcal{Y} @>>> \mathcal{Y}_n \cong\mathcal{X}\times B\mathbb{G}_m \\
@. @VVV \\
@. \mathcal{X}
\end{CD} Y Y n ≅ X × B G m ↓ ⏐ X
Therefore, we have A ∗ ( Y ) ≅ A ∗ ( Y n ) A^*(\mathcal{Y})\cong A^*(\mathcal{Y}_n) A ∗ ( Y ) ≅ A ∗ ( Y n ) . Applying the classical equivariant-Chow argument from [10 ] , we have
A ∗ ( X × B G m ) ≅ A ∗ ( X ) ⊗ A ∗ ( B G m ) ≅ A ∗ ( X ) [ c 1 ] . A^*(\mathcal{X}\times B\mathbb{G}_m)\cong A^*(\mathcal{X})\otimes A^*(B\mathbb{G}_m)\cong A^*(\mathcal{X})[c_1]. A ∗ ( X × B G m ) ≅ A ∗ ( X ) ⊗ A ∗ ( B G m ) ≅ A ∗ ( X ) [ c 1 ] .
Twisted Sheaves
We follow Lieblich’s notion of twisted sheaves [26 ] . A twisted-sheaf on a smooth DM stack X X X is a coherent sheaf on X \mathcal{X} X where X → X \mathcal{X}\to X X → X is a gerbe banded by a finite abelian group. To emphasize this distinction between twisted sheaves and coherent sheaves, we shall use coherent sheaves and honest coherent sheaves interchangeably.
For a torsion Brauer class α ∈ B r ( X ) \alpha\in Br(X) α ∈ B r ( X ) such that n α = 0 n\alpha=0 n α = 0 , let X \mathcal{X} X be the associated G m \mathbb{G}_m G m -gerbe. We can descend honest coherent sheaves on X \mathcal{X} X using the following theorem.
Theorem 1.2 ([26 ] ). Suppose we have a coherent sheaf F → X \mathcal{F} \to\mathcal{X} F → X , then F \mathcal{F} F descends to X X X as a twisted-sheaf.
Proof. We use the Kummer sequence and the long exact sequence in étale cohomology
1 → μ n → G m → ( − ) n G m → 1. 1 \to\mu_n \to\mathbb{G}_m \xrightarrow{(-)^n} \mathbb{G}_m \to1. 1 → μ n → G m ( − ) n G m → 1.
We have the exact sequence
H 1 ( X , G m ) → H 2 ( X , μ n ) → i H 2 ( X , G m ) → H 2 ( X , G m ) . H^1(X,\mathbb{G}_m) \to H^2(X,\mu_n) \xrightarrow{i} H^2(X,\mathbb{G}_m) \to H^2(X,\mathbb{G}_m). H 1 ( X , G m ) → H 2 ( X , μ n ) i H 2 ( X , G m ) → H 2 ( X , G m ) .
Since n α = 0 n\alpha=0 n α = 0 , we have α ∈ Im ( i ) \alpha\in\operatorname{Im}(i) α ∈ Im ( i ) . That is, there exists β ∈ H 2 ( X , μ n ) \beta\in H^2(X,\mu_n) β ∈ H 2 ( X , μ n ) such that i ( β ) = α i(\beta)=\alpha i ( β ) = α . The class β \beta β corresponds to a μ n \mu_n μ n -gerbe X → X \mathcal{X} \to X X → X .
The restriction of α \alpha α to X \mathcal{X} X is trivial [32 Theorem 1.2] since 0 = β ∣ X 0=\beta|_{\mathcal{X}} 0 = β ∣ X maps to α ∣ X \alpha|_{\mathcal{X}} α ∣ X under i ∣ X i|_{\mathcal{X}} i ∣ X . Hence, we have the Cartesian diagram
X × B G m → α ∣ X X s ↓ ↓ X → X . \begin{CD}
\mathcal{X} \times B\mathbb{G}_m @>{\alpha|_{\mathcal{X}}}>> \mathcal{X} \\
@V{s}VV @VVV \\
\mathcal{X} @>>> X.
\end{CD} X × B G m s ↓ ⏐ X α ∣ X X ↓ ⏐ X .
The pullback ( α ∣ X ∘ s ) ∗ F (\alpha|_{\mathcal{X}} \circ s)^*\mathcal{F} ( α ∣ X ∘ s ) ∗ F is an honest coherent sheaf on X \mathcal{X} X . Hence, we have a twisted sheaf on X X X with the claimed property. □
Definition 1.2.1. We shall refer to these descended twisted-sheaves on X X X as α \alpha α -twisted sheaves.
The Strong Franchetta Property
For a morphism f : F → Z f:\mathcal{F} \to\mathcal{Z} f : F → Z between smooth, finite-type algebraic stacks, we say that f f f satisfies the Franchetta property if:
For every Γ ∈ A ∗ ( F ) \Gamma\in A^*(\mathcal{F}) Γ ∈ A ∗ ( F ) such that Γ ∣ F z \Gamma|_{\mathcal{F}_z} Γ ∣ F z is trivial in H ∗ ( F z ) H^*(\mathcal{F}_z) H ∗ ( F z ) for a very general geometric point z ∈ ∣ Z ∣ z \in|\mathcal{Z}| z ∈ ∣ Z ∣ , the restriction Γ ∣ F z \Gamma|_{\mathcal{F}_z} Γ ∣ F z is trivial in A ∗ ( F z ) A^*(\mathcal{F}_z) A ∗ ( F z ) for all z ∈ ∣ Z ∣ z \in|\mathcal{Z}| z ∈ ∣ Z ∣ .
This property has been studied in [25 , 5 ] . We refer to these sources for more details. For our purposes, we define a variant of the property below.
Definition 1.3.1. We say that a family ϕ : X → S \phi:\mathcal{X} \to S ϕ : X → S satisfies the strong Franchetta property (SFP) if the map A ∗ ( X ) → H ∗ ( X s ) A^*(\mathcal{X}) \to H^*(X_s) A ∗ ( X ) → H ∗ ( X s ) is injective for all fibers X s X_s X s over geometric points s : Spec C → S s:\operatorname{Spec}\mathbb{C} \to S s : Spec C → S .
Theorem 1.3. Let X → S \mathcal{X} \to S X → S be a family satisfying SFP, and let Y → X \mathcal{Y} \to\mathcal{X} Y → X be a torsion-G m \mathbb{G}_m G m -gerbe. Then Y → S \mathcal{Y} \to S Y → S satisfies SFP.
Proof. For a geometric point s ∈ ∣ S ∣ s \in|S| s ∈ ∣ S ∣ , and the fibers Y s \mathcal{Y}_s Y s , X s \mathcal{X}_s X s of Y \mathcal{Y} Y , X \mathcal{X} X over it respectively, we have
Y s → X s \mathcal{Y}_s \to\mathcal{X}_s Y s → X s
is a G m \mathbb{G}_m G m -gerbe. Applying Theorem 1.1 and [19 Lemma 5.13] , we have that there exists c 1 ∈ A 1 ( Y ) c_1 \in A^1(\mathcal{Y}) c 1 ∈ A 1 ( Y ) such that
A ∗ ( Y ) ≅ A ∗ ( X ) [ c 1 ] , H ∗ ( Y s ) ≅ H ∗ ( X s ) [ c 1 ∣ Y s ] . A^*(\mathcal{Y}) \cong A^*(\mathcal{X})[c_1],\quad H^*(\mathcal{Y}_s) \cong H^*(\mathcal{X}_s)[c_1|_{\mathcal{Y}_s}]. A ∗ ( Y ) ≅ A ∗ ( X ) [ c 1 ] , H ∗ ( Y s ) ≅ H ∗ ( X s ) [ c 1 ∣ Y s ] .
::: SFP of X → S \mathcal{X} \to S X → S gives the following diagram:
A ∗ ( X ) [ c 1 ] ⇢ H ∗ ( X s ) [ c 1 ∣ Y s ] ↑ ↑ A ∗ ( X ) ⟶ H ∗ ( X s ) \begin{array}{ccc}
A^{*}(\mathcal{X})[c_{1}] & \dashrightarrow& H^{*}(X_{s})[c_{1}|_{Y_{s}}] \\
\uparrow& & \uparrow\\
A^{*}(\mathcal{X}) & \longrightarrow& H^{*}(X_{s})
\end{array} A ∗ ( X ) [ c 1 ] ↑ A ∗ ( X ) ⇢ ⟶ H ∗ ( X s ) [ c 1 ∣ Y s ] ↑ H ∗ ( X s )
Injectivity of the dashed horizontal arrow is immediate from the diagram. Consequently, we have SFP for Y → S \mathcal{Y} \to S Y → S . □
Theorem 1.4. Let f : X → Y f:\mathcal{X}\to\mathcal{Y} f : X → Y be a morphism of families over S S S , with X , Y \mathcal{X},\mathcal{Y} X , Y smooth such that f ∗ : A ∗ ( Y ) → A ∗ ( X ) f^{*}:A^{*}(\mathcal{Y})\to A^{*}(\mathcal{X}) f ∗ : A ∗ ( Y ) → A ∗ ( X ) is injective. The family Y → S \mathcal{Y}\to S Y → S satisfies SFP if we have the same for X → S \mathcal{X}\to S X → S .
Proof. For fibers Y s , X s Y_{s},X_{s} Y s , X s of Y , X \mathcal{Y},\mathcal{X} Y , X respectively, over a geometric point s ∈ ∣ S ∣ s\in|S| s ∈ ∣ S ∣ , we have the following diagram due to SFP of X → S \mathcal{X}\to S X → S :
A ∗ ( Y ) ⇢ H ∗ ( Y s ) ↓ f ∗ ↓ A ∗ ( X ) ⟶ H ∗ ( X s ) \begin{array}{ccc}
A^{*}(\mathcal{Y}) & \dashrightarrow& H^{*}(Y_{s}) \\
\downarrow f^{*} & & \downarrow\\
A^{*}(\mathcal{X}) & \longrightarrow& H^{*}(X_{s})
\end{array} A ∗ ( Y ) ↓ f ∗ A ∗ ( X ) ⇢ ⟶ H ∗ ( Y s ) ↓ H ∗ ( X s ) Injectivity of the dashed arrow follows from the diagram, which shows SFP for Y → S \mathcal{Y}\to S Y → S . □
The following theorem follows from a straightforward application of the projective bundle formula in Chow and cohomology.
Theorem 1.5. Let Y → X \mathcal{Y}\to\mathcal{X} Y → X be a projective bundle and X → S \mathcal{X}\to S X → S be a family. We have that X → S \mathcal{X}\to S X → S satisfies SFP if and only if Y → S \mathcal{Y}\to S Y → S satisfies SFP.
A motivic decomposition over families
Chow Motives over a Base
Let B B B be a non-singular variety over C \mathbb{C} C . Following [17 ] , we denote the category of relative smooth Chow motives over B B B with Q \mathbb{Q} Q -coefficients as M ( B ) \mathsf{M}(B) M ( B ) . The objects in M ( B ) \mathsf{M}(B) M ( B ) are ( X → B , p , m ) (X\to B,p,m) ( X → B , p , m ) where X → B X\to B X → B is smooth projective, p ∈ Corr B 0 ( X , X ) p\in\operatorname{Corr}^{0}_{B}(X,X) p ∈ Corr B 0 ( X , X ) is a projector and m ∈ Z m\in\mathbb{Z} m ∈ Z records the Tate twists. Further, we have
Hom M ( B ) ( ( X , p , m ) , ( Y , q , n ) ) = q ∘ Corr B n − m ( X , Y ) ∘ p . (1) \operatorname{Hom}_{\mathsf{M}(B)}((X,p,m),(Y,q,n))=q\circ\operatorname{Corr}^{n-m}_{B}(X,Y)\circ p. \tag*{(1)} Hom M ( B ) (( X , p , m ) , ( Y , q , n )) = q ∘ Corr B n − m ( X , Y ) ∘ p . (1)
Further, we denote by A ( B ) A(B) A ( B ) the full pseudo-abelian tensor category generated additively and tensorially by motives of relative curves in M ( B ) \mathsf{M}(B) M ( B ) .
Theorem 2.1. [[17 Corollary C] ] Let h h h be a motive in A ( B ) A(B) A ( B ) , let η \eta η be the generic point of B B B with residue field C ( η ) \mathbb{C}(\eta) C ( η ) , let L / C ( η ) L/\mathbb{C}(\eta) L / C ( η ) be a field extension, and set ξ : = Spec L \xi:=\operatorname{Spec}L ξ := Spec L . Any endomorphism of a motive in A ( B ) A(B) A ( B ) whose pull-back to ξ \xi ξ is nilpotent modulo numerical equivalence is nilpotent modulo rational equivalence over the base scheme B B B .
Corollary 2.1.1. Suppose we have two motives h 1 , h 2 h_{1},h_{2} h 1 , h 2 in A ( B ) A(B) A ( B ) and Γ ∈ Hom M ( B ) ( h 1 , h 2 ) \Gamma\in\operatorname{Hom}_{\mathsf{M}(B)}(h_{1},h_{2}) Γ ∈ Hom M ( B ) ( h 1 , h 2 ) , Ψ ∈ Hom M ( B ) ( h 2 , h 1 ) \Psi\in\operatorname{Hom}_{\mathsf{M}(B)}(h_{2},h_{1}) Ψ ∈ Hom M ( B ) ( h 2 , h 1 ) be two correspondences such that for all b ∈ B b\in B b ∈ B closed points,
Γ b ∈ Hom M ( C ) ( ( h 1 ) b , ( h 2 ) b ) \Gamma_b \in\operatorname{Hom}_{\mathrm{M}(\mathrm{C})}((h_1)_b,(h_2)_b) Γ b ∈ Hom M ( C ) (( h 1 ) b , ( h 2 ) b ) , Ψ b ∈ Hom M ( C ) ( ( h 2 ) b , ( h 1 ) b ) \Psi_b \in\operatorname{Hom}_{\mathrm{M}(\mathrm{C})}((h_2)_b,(h_1)_b) Ψ b ∈ Hom M ( C ) (( h 2 ) b , ( h 1 ) b ) satisfies Γ b ∘ Ψ b = Id ( h 2 ) b \Gamma_b \circ\Psi_b = \operatorname{Id}_{(h_2)_b} Γ b ∘ Ψ b = Id ( h 2 ) b and Ψ b ∘ Γ b = Id ( h 1 ) b \Psi_b \circ\Gamma_b = \operatorname{Id}_{(h_1)_b} Ψ b ∘ Γ b = Id ( h 1 ) b . Then Γ ∈ Hom M ( B ) ( h 1 , h 2 ) \Gamma\in\operatorname{Hom}_{\mathrm{M}(\mathrm{B})}(h_1,h_2) Γ ∈ Hom M ( B ) ( h 1 , h 2 ) is an isomorphism.
Proof. Consider the correspondences E 1 = Id h 1 − Ψ ∘ Γ E_1 = \operatorname{Id}_{h_1} - \Psi\circ\Gamma E 1 = Id h 1 − Ψ ∘ Γ and E 2 = Id h 2 − Γ ∘ Ψ E_2 = \operatorname{Id}_{h_2} - \Gamma\circ\Psi E 2 = Id h 2 − Γ ∘ Ψ . Then we have by condition that E 1 ∣ b = 0 , E 2 ∣ b = 0 E_1|_b = 0, E_2|_b = 0 E 1 ∣ b = 0 , E 2 ∣ b = 0 for all b ∈ B b \in B b ∈ B closed points. Applying [35 Lemma 2.1.2] , it follows that ( E 1 ) η = 0 , ( E 2 ) η = 0 (E_1)_\eta= 0, (E_2)_\eta= 0 ( E 1 ) η = 0 , ( E 2 ) η = 0 where η \eta η is the generic point of B B B and hence by Theorem 2.1 , we see that E 1 ∈ End M ( B ) ( h 1 ) , E 2 ∈ End M ( B ) ( h 2 ) E_1 \in\operatorname{End}_{\mathrm{M}(\mathrm{B})}(h_1), E_2 \in\operatorname{End}_{\mathrm{M}(\mathrm{B})}(h_2) E 1 ∈ End M ( B ) ( h 1 ) , E 2 ∈ End M ( B ) ( h 2 ) are nilpotent endomorphisms. Therefore Ψ ∘ Γ \Psi\circ\Gamma Ψ ∘ Γ and Γ ∘ Ψ \Gamma\circ\Psi Γ ∘ Ψ are isomorphisms. It follows that Γ \Gamma Γ is an isomorphism. □
Motive of Moduli of Stable Bundles over a Base
Let C → B \mathcal{C} \to B C → B be such that for any closed point b ∈ B b \in B b ∈ B , the fiber C b \mathcal{C}_b C b is a smooth, proper curve of genus g ≥ 2 g \ge2 g ≥ 2 . Let U C / B ( r , d , g ) U_{\mathcal{C}/B}(r,d,g) U C / B ( r , d , g ) be the relative coarse moduli space of stable bundles of rank r r r and degree d d d with r , d r,d r , d coprime.
We have U C / B ( r , d , g ) → B U_{\mathcal{C}/B}(r,d,g) \to B U C / B ( r , d , g ) → B is a smooth proper morphism of relative dimension m : = r 2 ( g − 1 ) + 1 m := r^2(g-1)+1 m := r 2 ( g − 1 ) + 1 .
For notational convenience, set U : = U C / B ( r , d , g ) U := U_{\mathcal{C}/B}(r,d,g) U := U C / B ( r , d , g ) . We shall use a relative version of Beauville’s diagonal argument as in [4 ] to show h B ( U ) h_B(U) h B ( U ) is Kimura finite-dimensional. We refer to [22 ] and [17 ] for the necessary definitions. For our case, we shall only need that the relative motive is in A ( B ) A(B) A ( B ) .
Lemma 2.2. Let Δ U / B \Delta_{U/B} Δ U / B be the image of the embedding U ↪ U × B U U \hookrightarrow U \times_B U U ↪ U × B U . We have
[ Δ U / B ] ∈ Q . c m ( − E x t π 13 ! ( π 12 ∗ E ′ , π 23 ∗ E ′ ′ ) ) [\Delta_{U/B}] \in\mathbb{Q}.c_m\left(-\mathcal{E}xt^!_{\pi_{13}}\left(\pi_{12}^*\mathcal{E}',\pi_{23}^*\mathcal{E}''\right)\right) [ Δ U / B ] ∈ Q . c m ( − E x t π 13 ! ( π 12 ∗ E ′ , π 23 ∗ E ′′ ) ) in A m ( U × B U ) A^m(U \times_B U) A m ( U × B U ) where the morphisms π i j \pi_{ij} π ij ’s are given by the projections in the diagram
U × B C × B U → π 23 C × B U π 12 ↓ U × B U π 13 ↑ \begin{CD}
U \times_B \mathcal{C} \times_B U @>{\pi_{23}}>> \mathcal{C} \times_B U \\
@V{\pi_{12}}VV @. \\
U \times_B U @. \\
@A{\pi_{13}}AA @.
\end{CD} U × B C × B U π 12 ↓ ⏐ U × B U π 13 ⏐ ↑ π 23 C × B U
Proof. Let α \alpha α be the Brauer class of the G m \mathbb{G}_m G m -gerbe U → U \mathcal{U} \to U U → U where U \mathcal{U} U is the unrigidified moduli stack of fiber-wise stable, rank r r r bundles with relative degree d d d for the family C → B \mathcal{C} \to B C → B . We note that α ∈ Br ( U ) \alpha\in\operatorname{Br}(U) α ∈ Br ( U ) is torsion by [13 Theorem B] . Let E ′ , E ′ ′ \mathcal{E}',\mathcal{E}'' E ′ , E ′′ be twisted universal sheaves on U × B C , C × B U U \times_B \mathcal{C},\mathcal{C} \times_B U U × B C , C × B U respectively, defined using α \alpha α and the universal sheaves on U × B C , C × B U \mathcal{U} \times_B \mathcal{C},\mathcal{C} \times_B \mathcal{U} U × B C , C × B U respectively by Theorem 1.2 .
Let N ∈ N N \in\mathbb{N} N ∈ N be the order of α \alpha α and G → g U \mathcal{G} \xrightarrow{g} U G g U be the associated μ N \mu_N μ N -gerbe. It follows that E ′ \mathcal{E}' E ′ is q 1 ∗ α q_1^*\alpha q 1 ∗ α -twisted and E ′ ′ \mathcal{E}'' E ′′ is r 2 ∗ α r_2^*\alpha r 2 ∗ α -twisted where q 1 : U × B C → U q_1 : U \times_B \mathcal{C} \to U q 1 : U × B C → U , r 2 : C × B U → U r_2 : \mathcal{C} \times_B U \to U r 2 : C × B U → U are the projections.
We denote the associated honest coherent sheaves on G × B C , C × B G \mathcal{G} \times_B \mathcal{C},\mathcal{C} \times_B \mathcal{G} G × B C , C × B G as E ~ ′ , E ~ ′ ′ \widetilde{\mathcal{E}}',\widetilde{\mathcal{E}}'' E ′ , E ′′ respectively with projections π ~ 12 : G × B C × B G → G × B C \widetilde{\pi}_{12} : \mathcal{G} \times_B \mathcal{C} \times_B \mathcal{G} \to\mathcal{G} \times_B \mathcal{C} π 12 : G × B C × B G → G × B C , π ~ 23 : G × B C × B G → C × B G \widetilde{\pi}_{23} : \mathcal{G} \times_B \mathcal{C} \times_B \mathcal{G} \to\mathcal{C} \times_B \mathcal{G} π 23 : G × B C × B G → C × B G onto (first, second) and (second, third) factors respectively.
The pull-backs π ~ 12 ∗ E ~ ′ , π ~ 23 ∗ E ~ ′ ′ \widetilde{\pi}_{12}^*\widetilde{\mathcal{E}}',\widetilde{\pi}_{23}^*\widetilde{\mathcal{E}}'' π 12 ∗ E ′ , π 23 ∗ E ′′ are honest coherent sheaves on G × B C × B G \mathcal{G} \times_B \mathcal{C} \times_B \mathcal{G} G × B C × B G . Consequently, R π ~ 13 ∗ R H o m ( R π ~ 12 ∗ E ~ ′ , R π ~ 23 ∗ E ~ ′ ′ ) R\widetilde{\pi}_{13*}R\mathcal{H}om(R\widetilde{\pi}_{12}^*\widetilde{\mathcal{E}}',R\widetilde{\pi}_{23}^*\widetilde{\mathcal{E}}'') R π 13 ∗ R H o m ( R π 12 ∗ E ′ , R π 23 ∗ E ′′ ) can be represented on G × B G \mathcal{G} \times_B \mathcal{G} G × B G by a two-term complex K 0 → K 1 K^0 \to K^1 K 0 → K 1 .
For a geometric point ( ξ , η ) ∈ G × B G (\xi,\eta) \in\mathcal{G} \times_B \mathcal{G} ( ξ , η ) ∈ G × B G mapping to ( x , y ) ∈ U × B U (x,y) \in U \times_B U ( x , y ) ∈ U × B U , the fiber of the complex computes R H o m ( E x , E y ) R\mathcal{H}om(E_x,E_y) R H o m ( E x , E y ) and R 0 H o m ( E x , E y ) = { C x = y 0 x ≠ y R^0\mathcal{H}om(E_x,E_y) = \begin{cases} \mathbb{C} & x = y \\ 0 & x \ne y \end{cases} R 0 H o m ( E x , E y ) = { C 0 x = y x = y . Using the Porteous formula for smooth DM stacks, we get
[ G × U G ] ∈ Q . c m ( − R π ~ 13 ∗ R H o m ( R π ~ 12 ∗ E ~ ′ , R π ~ 23 ∗ E ~ ′ ′ ) ) ∈ A ∗ ( G × B G ) . [G \times_U G] \in\mathbb{Q}.c_m\left(-R\widetilde{\pi}_{13*}R\mathcal{H}om\left(R\widetilde{\pi}_{12}^*\widetilde{\mathcal{E}}',R\widetilde{\pi}_{23}^*\widetilde{\mathcal{E}}''\right)\right)\in A^*(G\times_B G). [ G × U G ] ∈ Q . c m ( − R π 13 ∗ R H o m ( R π 12 ∗ E ′ , R π 23 ∗ E ′′ ) ) ∈ A ∗ ( G × B G ) .
Since ( g × B g ) ∗ : A ∗ ( G × B G ) → A ∗ ( U × B U ) (g\times_B g)_*:A^*(G\times_B G)\to A^*(U\times_B U) ( g × B g ) ∗ : A ∗ ( G × B G ) → A ∗ ( U × B U ) is an isomorphism, we have
[ Δ U / B ] ∈ Q . c m ( − E x t π ~ 13 ! ( π 12 ∗ E ′ , π 23 ∗ E ′ ′ ) ) . [\Delta_{U/B}]\in\mathbb{Q}.c_m\left(-\mathcal{E}xt_{\widetilde{\pi}_{13}}^!\left(\pi_{12}^*\mathcal{E}',\pi_{23}^*\mathcal{E}''\right)\right). [ Δ U / B ] ∈ Q . c m ( − E x t π 13 ! ( π 12 ∗ E ′ , π 23 ∗ E ′′ ) ) .
Lemma 2.3. Let M M M , S S S , S ′ S' S ′ , M ′ M' M ′ be smooth proper B B B -varieties. If α ∈ Corr B ( M , S ) \alpha\in\operatorname{Corr}_B(M,S) α ∈ Corr B ( M , S ) , β ∈ Corr B ( S , M ′ ) \beta\in\operatorname{Corr}_B(S,M') β ∈ Corr B ( S , M ′ ) , α ′ ∈ Corr B ( M , S ′ ) \alpha'\in\operatorname{Corr}_B(M,S') α ′ ∈ Corr B ( M , S ′ ) , β ′ ∈ Corr B ( S ′ , M ′ ) \beta'\in\operatorname{Corr}_B(S',M') β ′ ∈ Corr B ( S ′ , M ′ ) , then
( β ′ ∘ α ′ ) ⋅ ( β ∘ α ) = γ ∘ δ (\beta'\circ\alpha')\cdot(\beta\circ\alpha)=\gamma\circ\delta ( β ′ ∘ α ′ ) ⋅ ( β ∘ α ) = γ ∘ δ for some γ ∈ Corr B ( S × B S ′ , M ′ ) \gamma\in\operatorname{Corr}_B(S\times_B S',M') γ ∈ Corr B ( S × B S ′ , M ′ ) , δ ∈ Corr B ( M , S × B S ′ ) \delta\in\operatorname{Corr}_B(M,S\times_B S') δ ∈ Corr B ( M , S × B S ′ ) .
Proof. It follows from a straightforward computation that
( β ′ ∘ α ′ ) ⋅ ( β ∘ α ) = ( P 13 ′ ∗ β ⋅ P 23 ′ ∗ β ′ ) ∘ ( P 12 ∗ α ⋅ P 13 ∗ α ′ ) (\beta'\circ\alpha')\cdot(\beta\circ\alpha)=\left(P_{13}^{\prime*}\beta\cdot P_{23}^{\prime*}\beta'\right)\circ\left(P_{12}^*\alpha\cdot P_{13}^*\alpha'\right) ( β ′ ∘ α ′ ) ⋅ ( β ∘ α ) = ( P 13 ′ ∗ β ⋅ P 23 ′ ∗ β ′ ) ∘ ( P 12 ∗ α ⋅ P 13 ∗ α ′ )
where P 13 ′ P'_{13} P 13 ′ , P 23 ′ P'_{23} P 23 ′ , P 12 P_{12} P 12 , and P 13 P_{13} P 13 are the projection morphisms.
By Grothendieck-Riemann-Roch, we have
ch ( E x t π 13 ! ( π 12 ∗ E ′ , π 23 ∗ E ′ ′ ) ) = ( ch ( E ′ ′ ) ⋅ q 1 ∗ td C ) ∘ ( ch ( E ′ ) ∨ ⋅ p 2 ∗ td C ) (2) \operatorname{ch}\left(\mathcal{E}xt_{\pi_{13}}^!\left(\pi_{12}^*\mathcal{E}',\pi_{23}^*\mathcal{E}''\right)\right)=\left(\operatorname{ch}(\mathcal{E}'')\cdot q_1^*\sqrt{\operatorname{td}_C}\right)\circ\left(\operatorname{ch}(\mathcal{E}')^\vee\cdot p_2^*\sqrt{\operatorname{td}_C}\right) \tag*{(2)} ch ( E x t π 13 ! ( π 12 ∗ E ′ , π 23 ∗ E ′′ ) ) = ( ch ( E ′′ ) ⋅ q 1 ∗ td C ) ∘ ( ch ( E ′ ) ∨ ⋅ p 2 ∗ td C ) (2)
where q 1 : C × B U → C q_1:C\times_B U\to C q 1 : C × B U → C , p 2 : U × B C → C p_2:U\times_B C\to C p 2 : U × B C → C are the projections, and td C \sqrt{\operatorname{td}_C} td C is the formal square root of the unit td C ∈ A ∗ ( C ) \operatorname{td}_C\in A^*(C) td C ∈ A ∗ ( C ) . For notational convenience, we shall denote × B n C \times_B^n C × B n C by C n C^n C n for all n ∈ N n\in\mathbb{N} n ∈ N .
Lemma 2.4. For any k ≥ 1 k\geq1 k ≥ 1 , there exist δ α ∈ Corr B ( U , C k α ) \delta_\alpha\in\operatorname{Corr}_B(U,C^{k_\alpha}) δ α ∈ Corr B ( U , C k α ) , γ α ∈ Corr B ( C k α , U ) \gamma_\alpha\in\operatorname{Corr}_B(C^{k_\alpha},U) γ α ∈ Corr B ( C k α , U ) with k α ≤ k k_\alpha\leq k k α ≤ k such that
c k ( − E x t π 13 ! ( π 12 ∗ E ′ , π 23 ∗ E ′ ′ ) ) = ∑ α γ α ∘ δ α . c_k\left(-\mathcal{E}xt_{\pi_{13}}^!\left(\pi_{12}^*\mathcal{E}',\pi_{23}^*\mathcal{E}''\right)\right)=\sum_\alpha\gamma_\alpha\circ\delta_\alpha. c k ( − E x t π 13 ! ( π 12 ∗ E ′ , π 23 ∗ E ′′ ) ) = α ∑ γ α ∘ δ α . Proof. By (2), we have
ch ( E x t π 13 ! ( π 12 ∗ E ′ , π 23 ∗ E ′ ′ ) ) = γ ∘ δ \operatorname{ch}\left(\mathcal{E}xt_{\pi_{13}}^!\left(\pi_{12}^*\mathcal{E}',\pi_{23}^*\mathcal{E}''\right)\right)=\gamma\circ\delta ch ( E x t π 13 ! ( π 12 ∗ E ′ , π 23 ∗ E ′′ ) ) = γ ∘ δ for γ ∈ Corr B ( C , U ) \gamma\in\operatorname{Corr}_B(C,U) γ ∈ Corr B ( C , U ) , δ ∈ Corr ( U , C ) \delta\in\operatorname{Corr}(U,C) δ ∈ Corr ( U , C ) . We have
ch = r + c 1 + ( 1 2 c 1 2 − c 2 ) + ⋯ = ∑ j = 0 ∞ P j \operatorname{ch}=r+c_1+\left(\frac{1}{2}c_1^2-c_2\right)+\cdots=\sum_{j=0}^{\infty}P_j ch = r + c 1 + ( 2 1 c 1 2 − c 2 ) + ⋯ = j = 0 ∑ ∞ P j where P j P_j P j is a homogeneous polynomial with Q \mathbb{Q} Q coefficients in the Chern classes c 1 , c 2 , … , c j c_1,c_2,\ldots,c_j c 1 , c 2 , … , c j with deg c k = k \deg c_k=k deg c k = k and the coefficient of c j c_j c j in P j P_j P j is non-zero. A straightforward induction and Lemma 2.3 show that
c k ( − E x t π 13 ! ( π 12 ∗ E ′ , π 23 ∗ E ′ ′ ) ) = ∑ α γ α ∘ δ α c_k\left(-\mathcal{E}xt_{\pi_{13}}^!\left(\pi_{12}^*\mathcal{E}',\pi_{23}^*\mathcal{E}''\right)\right)=\sum_\alpha\gamma_\alpha\circ\delta_\alpha c k ( − E x t π 13 ! ( π 12 ∗ E ′ , π 23 ∗ E ′′ ) ) = α ∑ γ α ∘ δ α
where δ α ∈ Corr B ( U , C k α ) \delta_{\alpha} \in\operatorname{Corr}_{B}(U,C^{k_{\alpha}}) δ α ∈ Corr B ( U , C k α ) , γ α ∈ Corr B ( C k α , U ) \gamma_{\alpha} \in\operatorname{Corr}_{B}(C^{k_{\alpha}},U) γ α ∈ Corr B ( C k α , U ) with k α ≤ k k_{\alpha} \leq k k α ≤ k .
We present a relative version of [B̈20, Theorem 0.1] below.
Corollary 2.4.1. We have h B ( U ) ∈ A ( B ) h_{B}(U) \in A(B) h B ( U ) ∈ A ( B ) .
Proof. Let I ⊂ Corr B ( U , U ) I \subset\operatorname{Corr}_{B}(U,U) I ⊂ Corr B ( U , U ) denote the two-sided ideal generated by relative correspondences which factor through some relative power of C C C . More precisely, we have
I = ⟨ β ∘ α ∣ α ∈ Corr B ( U , C k ) , β ∈ Corr B ( C k , U ) ⟩ . I = \langle\beta\circ\alpha\mid\alpha\in\operatorname{Corr}_{B}(U,C^{k}), \beta\in\operatorname{Corr}_{B}(C^{k},U) \rangle. I = ⟨ β ∘ α ∣ α ∈ Corr B ( U , C k ) , β ∈ Corr B ( C k , U )⟩ .
We note that I I I is closed under taking intersection products by Lemma 2.3. Applying Lemma 2.2 and Lemma 2.4, there exist relative correspondences δ i ∈ Corr B ( U , C k i ) \delta_{i} \in\operatorname{Corr}_{B}(U,C^{k_{i}}) δ i ∈ Corr B ( U , C k i ) , γ i ∈ Corr B ( C k i , U ) \gamma_{i} \in\operatorname{Corr}_{B}(C^{k_{i}},U) γ i ∈ Corr B ( C k i , U ) , and k i ∈ N k_{i} \in\mathbb{N} k i ∈ N such that
[ Δ U / B ] = ∑ i γ i ∘ δ i . [\Delta_{U/B}] = \sum_{i} \gamma_{i} \circ\delta_{i}. [ Δ U / B ] = i ∑ γ i ∘ δ i .
Viewing δ : = ⨁ i δ i \delta:= \bigoplus_{i}\delta_{i} δ := ⨁ i δ i as a morphism of motives
⨁ i h B ( C k i ) ( d i − 2 k i ) → δ h B ( U ) , \bigoplus_{i} h_{B}(C^{k_{i}})(d_{i}-2k_{i}) \xrightarrow{\delta} h_{B}(U), i ⨁ h B ( C k i ) ( d i − 2 k i ) δ h B ( U ) ,
we obtain the right inverse γ : = ⨁ i γ i \gamma:= \bigoplus_{i}\gamma_{i} γ := ⨁ i γ i . Equivalently, we have a section of δ \delta δ
h B ( U ) → γ ⨁ i h B ( C k i ) ( d i − 2 k i ) . h_{B}(U) \xrightarrow{\gamma} \bigoplus_{i} h_{B}(C^{k_{i}})(d_{i}-2k_{i}). h B ( U ) γ i ⨁ h B ( C k i ) ( d i − 2 k i ) .
Hence h B ( U ) h_{B}(U) h B ( U ) is a direct summand of ⨁ i h B ( C k i ) ( d i − 2 k i ) \bigoplus_{i} h_{B}(C^{k_{i}})(d_{i}-2k_{i}) ⨁ i h B ( C k i ) ( d i − 2 k i ) . Consequently, we have the claimed containment.
A Relative Motivic Decomposition for Moduli of Bundles
Let C → B C \to B C → B be a family of smooth curves as above. Fix d ∈ Z d \in\mathbb{Z} d ∈ Z odd for the rest of this section. The following is proved in [11 Theorem 4.3]
Theorem 2.5 ([11 ] ). For a smooth projective curve C C C of genus g g g over an arbitrary field k k k of char 0, let U : = U C ( 2 , d ) U := U_{C}(2,d) U := U C ( 2 , d ) be the moduli space of stable bundles of rank 2 and degree d d d . We have the following isomorphism in M ( k ) M(k) M ( k )
h ( U ) ≃ h ( J a c ( C ) ) ⊗ ( h ( S y m g − 1 C ) ( g − 1 ) ⊕ ⨁ i = 0 g − 2 h ( S y m i C ) ⊗ ( Q ( i ) ⊕ Q ( 3 g − 3 − 2 i ) ) ) . h(U) \simeq h(Jac(C)) \otimes\left(h(Sym^{g-1}C)(g-1) \oplus\bigoplus_{i=0}^{g-2} h(Sym^{i}C) \otimes\left(\mathbb{Q}(i) \oplus\mathbb{Q}(3g-3-2i)\right)\right). h ( U ) ≃ h ( J a c ( C )) ⊗ ( h ( S y m g − 1 C ) ( g − 1 ) ⊕ i = 0 ⨁ g − 2 h ( S y m i C ) ⊗ ( Q ( i ) ⊕ Q ( 3 g − 3 − 2 i ) ) ) .
Proof. The proof in [11 ] goes through the existence of a line bundle of odd degree, which we want to avoid. So we indicate a sketch of the above decomposition following their argument. No novelty is claimed as to the proof.
It follows from [9 Theorem 4.11] and the argument in [11 Theorem 4.3] that the following relation holds in K 0 ( M n u m ( k ) ) K_{0}(M_{\mathrm{num}}(k)) K 0 ( M num ( k )) :
[ h ( U ) ] = [ h ( J a c ( C ) ) ] ( [ h ( S y m g − 1 C ) ] L g − 1 + ∑ i = 0 g − 2 [ h ( S y m i C ) ] ( L i + L 3 g − 3 − 2 i ) ) . [h(U)] = [h(Jac(C))]\left([h(Sym^{g-1}C)]\mathbb{L}^{g-1} + \sum_{i=0}^{g-2}[h(Sym^{i}C)]\left(\mathbb{L}^{i}+\mathbb{L}^{3g-3-2i}\right)\right). [ h ( U )] = [ h ( J a c ( C ))] ( [ h ( S y m g − 1 C )] L g − 1 + i = 0 ∑ g − 2 [ h ( S y m i C )] ( L i + L 3 g − 3 − 2 i ) ) .
Moreover we know that h ( U ) , h ( J a c ( C ) ) , h ( C ) , L ∈ A ( k ) h(U), h(Jac(C)), h(C), L \in A(k) h ( U ) , h ( J a c ( C )) , h ( C ) , L ∈ A ( k ) by 2.4.1, [9 Theorem 4.5] (this is where we use char k = 0 \operatorname{char} k = 0 char k = 0 , as both 2.4.1 and [9 Theorem 4.5] require that hypothesis). By the results in [21 ] , we can lift this to an isomorphism of numerical motives and since both sides are of abelian type, we can further lift them to an isomorphism in M ( k ) M(k) M ( k ) .
The objective of this section is to extend the above result for smooth curve fibrations C → B C \to B C → B .
Theorem 2.6. Let U B : = U C / B ( 2 , d , g ) U_B := U_{C/B}(2,d,g) U B := U C / B ( 2 , d , g ) be the relative moduli space of rank two stable bundles of degree d d d . We have the following isomorphism in M ( B ) M(B) M ( B ) :
h B ( U B ) ≃ h B ( J a c ( C / B ) ) ⊗ ( h B ( [ × B g − 1 C / S g − 1 ] ( g − 1 ) ) ⊕ ⨁ i = 0 g − 2 h B ( [ × B i C / S i ] ) ⊗ ( Q ( i ) ⊕ Q ( 3 g − 3 − 2 i ) ) ) h_B(U_B) \simeq h_B(Jac(C/B)) \otimes\left(h_B\left([\times_B^{g-1} C/S_{g-1}](g-1)\right) \oplus\bigoplus_{i=0}^{g-2} h_B\left([\times_B^i C/S_i]\right) \otimes\left(\mathbb{Q}(i) \oplus\mathbb{Q}(3g-3-2i)\right)\right) h B ( U B ) ≃ h B ( J a c ( C / B )) ⊗ ( h B ( [ × B g − 1 C / S g − 1 ] ( g − 1 ) ) ⊕ i = 0 ⨁ g − 2 h B ( [ × B i C / S i ] ) ⊗ ( Q ( i ) ⊕ Q ( 3 g − 3 − 2 i ) ) )
Proof. Let η ∈ B \eta\in B η ∈ B be the generic point with residue field k ( η ) k(\eta) k ( η ) . Applying Theorem 2.5, we have the following isomorphism in M ( k ( η ) ) M(k(\eta)) M ( k ( η )) :
h ( U η ) ≃ h ( J a c ( C η ) ) ⊗ ( h ( S y m g − 1 C η ) ( g − 1 ) ⊕ ⨁ i = 0 g − 2 h ( S y m i C η ) ⊗ ( Q ( i ) ⊕ Q ( 3 g − 3 − 2 i ) ) ) . h(U_\eta) \simeq h(Jac(C_\eta)) \otimes\left(h(Sym^{g-1} C_\eta)(g-1) \oplus\bigoplus_{i=0}^{g-2} h(Sym^i C_\eta) \otimes\left(\mathbb{Q}(i) \oplus\mathbb{Q}(3g-3-2i)\right)\right). h ( U η ) ≃ h ( J a c ( C η )) ⊗ ( h ( S y m g − 1 C η ) ( g − 1 ) ⊕ i = 0 ⨁ g − 2 h ( S y m i C η ) ⊗ ( Q ( i ) ⊕ Q ( 3 g − 3 − 2 i ) ) ) .
Let Γ η \Gamma_\eta Γ η be a correspondence that induces an isomorphism above with inverse given by Ψ η \Psi_\eta Ψ η , let M : = h B ( U ) \mathcal{M} := h_B(U) M := h B ( U ) , and let N \mathcal{N} N be the motive on the right of the Theorem statement.
Using the “spread-out” argument by Voisin [34 ] , there exist Γ ∈ Hom M ( B ) ( M , N ) \Gamma\in\operatorname{Hom}_{M(B)}(\mathcal{M},\mathcal{N}) Γ ∈ Hom M ( B ) ( M , N ) , Ψ ∈ Hom M ( B ) ( N , M ) \Psi\in\operatorname{Hom}_{M(B)}(\mathcal{N},\mathcal{M}) Ψ ∈ Hom M ( B ) ( N , M ) such that Γ \Gamma Γ specializes to Γ η \Gamma_\eta Γ η and Ψ \Psi Ψ specializes to Ψ η \Psi_\eta Ψ η at η \eta η .
By specialization it follows that for every closed point b ∈ B b \in B b ∈ B , we have Γ b ∘ Ψ b = I d \Gamma_b \circ\Psi_b = Id Γ b ∘ Ψ b = I d and Ψ b ∘ Γ b = I d \Psi_b \circ\Gamma_b = Id Ψ b ∘ Γ b = I d .
Applying Corollary 2.1.1, we have that Γ \Gamma Γ is an isomorphism since Γ η , Ψ η \Gamma_\eta, \Psi_\eta Γ η , Ψ η are inverses of each other and M , N ∈ A ( B ) \mathcal{M}, \mathcal{N} \in A(B) M , N ∈ A ( B ) .
Algebraic and tautological cohomology for bundles on a curve
Let C → B C \to B C → B be as in § 4.2 . We recall that we have the following equality from 2.2 and 2.4 and
[ Δ U / B ] = ∑ i γ i ∘ δ i (3) [\Delta_{U/B}] = \sum_i \gamma_i \circ\delta_i \tag*{(3)} [ Δ U / B ] = i ∑ γ i ∘ δ i (3)
for δ i ∈ Corr B ( U , C k i ) \delta_i \in\operatorname{Corr}_B(U,C^{k_i}) δ i ∈ Corr B ( U , C k i ) and γ i ∈ Corr B ( C k i , U ) \gamma_i \in\operatorname{Corr}_B(C^{k_i},U) γ i ∈ Corr B ( C k i , U ) . Further, it follows from the proof of Lemma 2.4 that { γ α } \{\gamma_\alpha\} { γ α } can be expressed as a polynomial in { c l i ( E i ′ ′ ) } 1 ≤ l i ≤ r , 1 ≤ i ≤ k \{c_{l_i}(\mathcal{E}_i^{\prime\prime})\}_{1 \le l_i \le r,\ 1 \le i \le k} { c l i ( E i ′′ ) } 1 ≤ l i ≤ r , 1 ≤ i ≤ k , where E i ′ ′ \mathcal{E}_i^{\prime\prime} E i ′′ is the pull back of the, possibly twisted in the sense of §1.2, Poincaré bundle E ′ ′ \mathcal{E}^{\prime\prime} E ′′ on U × B C U \times_B C U × B C via the projection p i × B 1 U : C k × B U → C × B U p_i \times_B 1_U : C^k \times_B U \to C \times_B U p i × B 1 U : C k × B U → C × B U .
Theorem 3.1. Fix co-prime pairs ( r i , d i ) i = 1 , … , N (r_i,d_i)_{i=1,\ldots,N} ( r i , d i ) i = 1 , … , N and let E i \mathcal{E}_i E i be a possibly twisted Poincaré bundle on U C ( r i , d i ) × B C U_C(r_i,d_i) \times_B C U C ( r i , d i ) × B C . Then the total Chow group A ∗ ( U C ( r 1 , d 1 ) × B ⋯ × B U C ( r N , d N ) ) A_*(U_C(r_1,d_1) \times_B \cdots\times_B U_C(r_N,d_N)) A ∗ ( U C ( r 1 , d 1 ) × B ⋯ × B U C ( r N , d N )) is Q \mathbb{Q} Q -linearly generated by
{ p U ∗ ( ∏ i = 1 m 1 ∏ j = 1 l 1 c d j ( 1 ) ( E 1 ( i ) ) ⋯ ∏ i = 1 m N ∏ j = 1 l N c d j ( N ) ( E N ( i ) ) ρ ∗ ( Γ ) ) } Γ ∈ A ∗ ( C m 1 × B ⋯ × B C m N ) , d i ( j ) ∈ Z ≥ 0 (4) \left\{p_{U*}\left(\prod_{i=1}^{m_1}\prod_{j=1}^{l_1} c_{d_j^{(1)}}(\mathcal{E}_1^{(i)}) \cdots\prod_{i=1}^{m_N}\prod_{j=1}^{l_N} c_{d_j^{(N)}}(\mathcal{E}_N^{(i)}) \rho^*(\Gamma)\right)\right\}_{\Gamma\in A^*(C^{m_1} \times_B \cdots\times_B C^{m_N}),\ d_i^{(j)} \in\mathbb{Z}_{\geq0}} \tag*{(4)} { p U ∗ ( i = 1 ∏ m 1 j = 1 ∏ l 1 c d j ( 1 ) ( E 1 ( i ) ) ⋯ i = 1 ∏ m N j = 1 ∏ l N c d j ( N ) ( E N ( i ) ) ρ ∗ ( Γ ) ) } Γ ∈ A ∗ ( C m 1 × B ⋯ × B C m N ) , d i ( j ) ∈ Z ≥ 0 (4)
where E j → C × B UC ( r j , d j ) \mathcal{E}_j \to C \times_B \operatorname{UC}(r_j,d_j) E j → C × B UC ( r j , d j ) are possibly twisted Poincaré bundles with E j ( i ) : = q i , j ∗ E j \mathcal{E}^{(i)}_j := q_{i,j}^{*}\mathcal{E}_j E j ( i ) := q i , j ∗ E j , m i : = r i 2 ( g − 1 ) + 1 m_i := r_i^2(g-1)+1 m i := r i 2 ( g − 1 ) + 1 are dimensions of UC ( r i , d i ) \operatorname{UC}(r_i,d_i) UC ( r i , d i ) , and q i , j q_{i,j} q i , j , ρ \rho ρ , p U p_U p U are the projections
q i , j : C m 1 × B ⋯ × B C m N × B UC ( r 1 , d 1 ) × B ⋯ × B UC ( r N , d N ) ⟶ C m j × B UC ( r j , d j ) → i t h factor C × B UC ( r j , d j ) , q_{i,j}: C^{m_1} \times_B \cdots\times_B C^{m_N} \times_B \operatorname{UC}(r_1,d_1) \times_B \cdots\times_B \operatorname{UC}(r_N,d_N) \longrightarrow C^{m_j} \times_B \operatorname{UC}(r_j,d_j) \xrightarrow{i^{\mathrm{th}}\ \text{factor}} C \times_B \operatorname{UC}(r_j,d_j), q i , j : C m 1 × B ⋯ × B C m N × B UC ( r 1 , d 1 ) × B ⋯ × B UC ( r N , d N ) ⟶ C m j × B UC ( r j , d j ) i th factor C × B UC ( r j , d j ) ,
ρ : C m 1 × B ⋯ × B C m N × B UC ( r 1 , d 1 ) × B ⋯ × B UC ( r N , d N ) ⟶ C m 1 × B ⋯ × B C m N , \rho: C^{m_1} \times_B \cdots\times_B C^{m_N} \times_B \operatorname{UC}(r_1,d_1) \times_B \cdots\times_B \operatorname{UC}(r_N,d_N) \longrightarrow C^{m_1} \times_B \cdots\times_B C^{m_N}, ρ : C m 1 × B ⋯ × B C m N × B UC ( r 1 , d 1 ) × B ⋯ × B UC ( r N , d N ) ⟶ C m 1 × B ⋯ × B C m N ,
p U : C m 1 × B ⋯ × B C m N × B UC ( r 1 , d 1 ) × B ⋯ × B UC ( r N , d N ) ⟶ UC ( r 1 , d 1 ) × B ⋯ × B UC ( r N , d N ) . p_U: C^{m_1} \times_B \cdots\times_B C^{m_N} \times_B \operatorname{UC}(r_1,d_1) \times_B \cdots\times_B \operatorname{UC}(r_N,d_N) \longrightarrow\operatorname{UC}(r_1,d_1) \times_B \cdots\times_B \operatorname{UC}(r_N,d_N). p U : C m 1 × B ⋯ × B C m N × B UC ( r 1 , d 1 ) × B ⋯ × B UC ( r N , d N ) ⟶ UC ( r 1 , d 1 ) × B ⋯ × B UC ( r N , d N ) .
Proof. For simplicity of notation, we prove the case N = 1 N=1 N = 1 . The general case follows from a similar argument and the fact that [ Δ UC ( r 1 , d 1 ) × B ⋯ × B UC ( r N , d N ) / B ] = ∏ j = 1 N π j ∗ [ Δ UC ( r j , d j ) / B ] [\Delta_{\operatorname{UC}(r_1,d_1)\times_B\cdots\times_B\operatorname{UC}(r_N,d_N)/B}] = \prod_{j=1}^{N}\pi_j^{*}[\Delta_{\operatorname{UC}(r_j,d_j)/B}] [ Δ UC ( r 1 , d 1 ) × B ⋯ × B UC ( r N , d N ) / B ] = ∏ j = 1 N π j ∗ [ Δ UC ( r j , d j ) / B ] where π j : UC ( r 1 , d 1 ) × B ⋯ × B UC ( r N , d N ) → UC ( r j , d j ) \pi_j:\operatorname{UC}(r_1,d_1)\times_B\cdots\times_B\operatorname{UC}(r_N,d_N)\to\operatorname{UC}(r_j,d_j) π j : UC ( r 1 , d 1 ) × B ⋯ × B UC ( r N , d N ) → UC ( r j , d j ) is the j j j th projection. We have
Δ UC ( r , d ) / B = ∑ α γ α ∘ δ α \Delta_{\operatorname{UC}(r,d)/B}=\sum_{\alpha}\gamma_\alpha\circ\delta_\alpha Δ UC ( r , d ) / B = α ∑ γ α ∘ δ α
where δ α ∈ A ∗ ( UC ( r , d ) × B C m ) \delta_\alpha\in A^{*}(\operatorname{UC}(r,d)\times_B C^m) δ α ∈ A ∗ ( UC ( r , d ) × B C m ) , γ α = P α ( ⋯ c d j ( E ( i ) ) ⋯ ) ∈ A ∗ ( C m × B UC ( r , d ) ) \gamma_\alpha=P_\alpha(\cdots c_{d_j}(\mathcal{E}^{(i)})\cdots)\in A^{*}(C^m\times_B\operatorname{UC}(r,d)) γ α = P α ( ⋯ c d j ( E ( i ) ) ⋯ ) ∈ A ∗ ( C m × B UC ( r , d )) is a polynomial in the Chern classes of E ( i ) \mathcal{E}^{(i)} E ( i ) , the pull-back of E \mathcal{E} E on C × B UC ( r , d ) C\times_B\operatorname{UC}(r,d) C × B UC ( r , d ) to C m × B UC ( r , d ) C^m\times_B\operatorname{UC}(r,d) C m × B UC ( r , d ) via the i i i th projection.
Let z ∈ A ∗ ( UC ( r , d ) ) z\in A^{*}(\operatorname{UC}(r,d)) z ∈ A ∗ ( UC ( r , d )) . It follows from (3) that
z = Δ ∗ ( z ) = ∑ α ( γ α ∘ δ α ) ∗ ( z ) = ∑ α ( P α ( ⋯ c d j ( E ( i ) ) ⋯ ) ) ∗ ( ( δ α ) ∗ ( z ) ) \begin{aligned}
z=\Delta_{*}(z) \\
&=\sum_{\alpha}(\gamma_\alpha\circ\delta_\alpha)_{*}(z) \\
&=\sum_{\alpha}\left(P_\alpha(\cdots c_{d_j}(\mathcal{E}^{(i)})\cdots)\right)_{*}\left((\delta_\alpha)_{*}(z)\right)
\end{aligned} z = Δ ∗ ( z ) = α ∑ ( γ α ∘ δ α ) ∗ ( z ) = α ∑ ( P α ( ⋯ c d j ( E ( i ) ) ⋯ ) ) ∗ ( ( δ α ) ∗ ( z ) )
Since ( δ α ) ∗ ( z ) ∈ A ∗ ( C m ) (\delta_\alpha)_{*}(z)\in A^{*}(C^m) ( δ α ) ∗ ( z ) ∈ A ∗ ( C m ) , each term in the above sum is a Q \mathbb{Q} Q -linear combination of expressions as in (4), and this completes the proof. □
Universal curves on universal hyperelliptic Jacobians
Powers of the universal hyperelliptic curve.
Definition 4.1.1. Let C 2 , g n : = × H g n H g , 1 C^n_{2,g}:=\mathop{\times}\limits_{\mathcal{H}_g}^{n}\mathcal{H}_{g,1} C 2 , g n := H g × n H g , 1 be the n n n -th fibered power of the universal hyperelliptic curve over the hyperelliptic locus H g ↪ M g \mathcal{H}_g\hookrightarrow\mathcal{M}_g H g ↪ M g . Additionally, let ψ i : = p i ∗ K ∈ A 1 ( C 2 , g n ) \psi_i:=p_i^{*}K\in A^1(C^n_{2,g}) ψ i := p i ∗ K ∈ A 1 ( C 2 , g n ) be the pullback of the relative canonical bundle K K K of H g , 1 → H g \mathcal{H}_{g,1}\to\mathcal{H}_g H g , 1 → H g , and let Δ S \Delta_S Δ S be the S S S -diagonal for any subset S ⊂ [ n ] S\subset[n] S ⊂ [ n ] .
Lemma 4.1. Let P \mathcal{P} P be the set of partitions of [ n ] [n] [ n ] . For any P ∈ P P\in\mathcal{P} P ∈ P , let Δ P : = ∏ I ∈ P Δ I \Delta_P:=\prod_{I\in P}\Delta_I Δ P := ∏ I ∈ P Δ I be the corresponding diagonal and ∣ P ∣ |P| ∣ P ∣ be the number of parts of the partition. For n ≤ 2 g + 6 n\leq2g+6 n ≤ 2 g + 6 , the total Chow group A ∗ ( C 2 , g n ) A_{*}(C^n_{2,g}) A ∗ ( C 2 , g n ) is linearly generated by { Δ P , ψ i Δ P } P ∈ P , 1 ≤ i ≤ n \{\Delta_P,\psi_i\Delta_P\}_{P\in\mathcal{P},1\leq i\leq n} { Δ P , ψ i Δ P } P ∈ P , 1 ≤ i ≤ n .
Proof. Let the strict diagonal of Δ P \Delta_P Δ P be denoted by
Δ P 0 : = { ( C , x 1 , ⋯ , x n ) ∣ x i = x j ⟺ ∃ S ∈ P such that i , j ∈ S } . \Delta_P^{0}:=\{(C,x_1,\cdots,x_n)\mid x_i=x_j\Longleftrightarrow\exists S\in P\text{ such that }i,j\in S\}. Δ P 0 := {( C , x 1 , ⋯ , x n ) ∣ x i = x j ⟺ ∃ S ∈ P such that i , j ∈ S } .
We can stratify C 2 , g n C^n_{2,g} C 2 , g n as
C 2 , g n = ⨆ P ∈ P Δ P 0 . C^n_{2,g}=\bigsqcup_{P\in\mathcal{P}}\Delta_P^{0}. C 2 , g n = P ∈ P ⨆ Δ P 0 .
The strict-diagonal Δ P 0 \Delta^{0}_{P} Δ P 0 is naturally isomorphic to H g , ∣ P ∣ \mathcal{H}_{g,|P|} H g , ∣ P ∣ . By [8 Corollary 1.5] , we have that A ∗ ( H g , m ) A^{*}(\mathcal{H}_{g,m}) A ∗ ( H g , m ) is linearly generated by [ H g , m ] [\mathcal{H}_{g,m}] [ H g , m ] , and ψ \psi ψ -classes for m ≤ 2 g + 6 m \leq2g+6 m ≤ 2 g + 6 . The lemma now follows from a straightforward excision argument. □ \square □
Definition 4.1.2. For n ≥ 2 g + 2 n \geq2g+2 n ≥ 2 g + 2 and S = { a 1 , ⋯ , a 2 g + 2 } ⊂ [ n ] S=\{a_{1},\cdots,a_{2g+2}\}\subset[n] S = { a 1 , ⋯ , a 2 g + 2 } ⊂ [ n ] , let
α S t a v : = ∑ I S b i 1 , i 2 ⋯ b i 2 g + 1 , i 2 g + 2 \alpha^{tav}_{S}:=\sum_{I_{S}}b_{i_{1},i_{2}}\cdots b_{i_{2g+1},i_{2g+2}} α S t a v := I S ∑ b i 1 , i 2 ⋯ b i 2 g + 1 , i 2 g + 2 be the cycle defined by Tavakol [33 (2)] , where b i j : = Δ i j − ψ i + ψ j 2 g − 2 b_{ij}:=\Delta_{ij}-\frac{\psi_{i}+\psi_{j}}{2g-2} b ij := Δ ij − 2 g − 2 ψ i + ψ j and I S I_{S} I S is the set of partitions of S S S into g + 1 g+1 g + 1 -subsets with two elements each.
Theorem 4.2. The families C 2 , g n → H g \mathcal{C}^{n}_{2,g}\to\mathcal{H}_{g} C 2 , g n → H g satisfy SFP for 1 ≤ n ≤ 2 g + 6 1\leq n\leq2g+6 1 ≤ n ≤ 2 g + 6 . In particular, we have
A ∗ ( C 2 , g n ) = Q [ Δ a b , ψ i ] 1 ≤ i , a , b ≤ n , a , b d i s t i n c t / ⟨ Δ a b ( ψ a − ψ b ) , ψ a ( ( 2 g − 2 ) Δ a b − ψ b ) , ( 2 g − 2 ) Δ a b 2 + ψ a ψ b , Δ a b − Δ b a , ψ a 2 , Δ a b Δ b c − ψ a Δ b c + ψ b Δ a c + ψ c Δ a b 2 g − 2 + ψ a ψ b + ψ b ψ c + ψ a ψ c ( 2 g − 2 ) 2 , α S t a v ⟩ 1 ≤ a , b , c ≤ n , a , b , c d i s t i n c t , S ⊂ [ n ] , ∣ S ∣ = 2 g + 2 . A^{*}(\mathcal{C}^{n}_{2,g})=\mathbb{Q}[\Delta_{ab},\psi_{i}]_{1\leq i,a,b\leq n,\ a,b\ \mathrm{distinct}}\Big/\left\langle\Delta_{ab}(\psi_{a}-\psi_{b}),\ \psi_{a}((2g-2)\Delta_{ab}-\psi_{b}),\ (2g-2)\Delta_{ab}^{2}+\psi_{a}\psi_{b},\ \Delta_{ab}-\Delta_{ba},\ \psi_{a}^{2},
\Delta_{ab}\Delta_{bc}-\frac{\psi_{a}\Delta_{bc}+\psi_{b}\Delta_{ac}+\psi_{c}\Delta_{ab}}{2g-2}+\frac{\psi_{a}\psi_{b}+\psi_{b}\psi_{c}+\psi_{a}\psi_{c}}{(2g-2)^{2}},\ \alpha^{tav}_{S}\right\rangle_{\substack{1\leq a,b,c\leq n,\ a,b,c\ \mathrm{distinct},\\ S\subset[n],\ |S|=2g+2}}. A ∗ ( C 2 , g n ) = Q [ Δ ab , ψ i ] 1 ≤ i , a , b ≤ n , a , b distinct / ⟨ Δ ab ( ψ a − ψ b ) , ψ a (( 2 g − 2 ) Δ ab − ψ b ) , ( 2 g − 2 ) Δ ab 2 + ψ a ψ b , Δ ab − Δ ba , ψ a 2 , Δ ab Δ b c − 2 g − 2 ψ a Δ b c + ψ b Δ a c + ψ c Δ ab + ( 2 g − 2 ) 2 ψ a ψ b + ψ b ψ c + ψ a ψ c , α S t a v ⟩ 1 ≤ a , b , c ≤ n , a , b , c distinct , S ⊂ [ n ] , ∣ S ∣ = 2 g + 2 .
Proof. We begin with verifying the claimed relations. It is easy to see that
Δ a b c = Δ a b Δ b c = Δ a b Δ b c . \Delta_{abc}=\Delta_{ab}\Delta_{bc}=\Delta_{ab}\Delta_{bc}. Δ ab c = Δ ab Δ b c = Δ ab Δ b c .
The claimed relations are proven in [33 Theorem 3.10 - Corollary 3.3] . For S = { a 1 , a 2 , ⋯ , a t } ⊂ [ n ] S=\{a_{1},a_{2},\cdots,a_{t}\}\subset[n] S = { a 1 , a 2 , ⋯ , a t } ⊂ [ n ] , it is easy to see that Δ S = ∏ 2 ≤ i ≤ t Δ a 1 a i \Delta_{S}=\prod_{2\leq i\leq t}\Delta_{a_{1}a_{i}} Δ S = ∏ 2 ≤ i ≤ t Δ a 1 a i . By Lemma 4.1, we have the claimed generators for A ∗ ( C 2 , g n ) A^{*}(\mathcal{C}^{n}_{2,g}) A ∗ ( C 2 , g n ) .
Let s : Spec C → H g s:\operatorname{Spec}\mathbb{C}\to\mathcal{H}_{g} s : Spec C → H g be a geometric point corresponding to a hyperelliptic curve C C C . Using [33 Proposition 3.16 - Corollary 3.18] , we have that the restriction A ∗ ( C 2 , g n ) → H ∗ ( C n ) A^{*}(\mathcal{C}^{n}_{2,g})\to H^{*}(\mathcal{C}^{n}) A ∗ ( C 2 , g n ) → H ∗ ( C n ) is injective. Hence, we have SFP for C 2 , g n → H g \mathcal{C}^{n}_{2,g}\to\mathcal{H}_{g} C 2 , g n → H g . □ \square □
Corollary 4.2.1. The total Chow group A ∗ ( C 2 , g n ) A_{*}(\mathcal{C}^{n}_{2,g}) A ∗ ( C 2 , g n ) is Q \mathbb{Q} Q -linearly generated by
{ ∏ i ∈ S ψ i ∏ 1 ≤ j ≤ k Δ a 2 j − 1 , a 2 j } k ≥ 0 , S ∪ { a 1 , ⋯ , a 2 k } ⊂ [ n ] , S ∩ { a 1 , ⋯ , a 2 k } = ∅ . \left\{\prod_{i\in S}\psi_{i}\prod_{1\leq j\leq k}\Delta_{a_{2j-1},a_{2j}}\right\}_{k\geq0,\ S\cup\{a_{1},\cdots,a_{2k}\}\subset[n],\ S\cap\{a_{1},\cdots,a_{2k}\}=\varnothing}. ⎩ ⎨ ⎧ i ∈ S ∏ ψ i 1 ≤ j ≤ k ∏ Δ a 2 j − 1 , a 2 j ⎭ ⎬ ⎫ k ≥ 0 , S ∪ { a 1 , ⋯ , a 2 k } ⊂ [ n ] , S ∩ { a 1 , ⋯ , a 2 k } = ∅ .
Universal hyperelliptic Jacobians and SFP
We study SFP for several families that arise as factors in the motivic decompositions introduced in §2, and further studied in §5.
Theorem 4.4. For 1 ≤ n ≤ g + 6 1\leq n\leq g+6 1 ≤ n ≤ g + 6 and d , d ′ ∈ Z d,d'\in\mathbb{Z} d , d ′ ∈ Z , the stacks C 2 , g n × H g J 2 , g d \mathcal{C}^{n}_{2,g}\times_{\mathcal{H}_{g}}\mathcal{J}^{d}_{2,g} C 2 , g n × H g J 2 , g d , C 2 , g n × H g J 2 , g d ′ \mathcal{C}^{n}_{2,g}\times_{\mathcal{H}_{g}}\mathcal{J}^{d'}_{2,g} C 2 , g n × H g J 2 , g d ′ , J 2 , g d × H g J 2 , g d ′ \mathcal{J}^{d}_{2,g}\times_{\mathcal{H}_{g}}\mathcal{J}^{d'}_{2,g} J 2 , g d × H g J 2 , g d ′ , and J 2 , g d × H g J 2 , g d \mathcal{J}^{d}_{2,g}\times_{\mathcal{H}_{g}}\mathcal{J}^{d}_{2,g} J 2 , g d × H g J 2 , g d satisfy SFP over H g \mathcal{H}_{g} H g .
Proof. We show that C 2 , g n × H g J 2 , g d \mathcal{C}^{n}_{2,g}\times_{\mathcal{H}_{g}}\mathcal{J}^{d}_{2,g} C 2 , g n × H g J 2 , g d satisfies SFP. The claim would then follow for C 2 , g n × H g J 2 , g d \mathcal{C}^{n}_{2,g}\times_{\mathcal{H}_{g}}\mathcal{J}^{d}_{2,g} C 2 , g n × H g J 2 , g d by Theorem 1.3 since J 2 , g d → J 2 , g d \mathcal{J}^{d}_{2,g}\to\mathcal{J}^{d}_{2,g} J 2 , g d → J 2 , g d is a torsion-G m \mathbb{G}_{m} G m -gerbe [29 Proposition 6.6] .
We can find integers a 1 , ⋯ , a g ∈ Z ∖ 0 a_{1},\cdots,a_{g}\in\mathbb{Z}\setminus0 a 1 , ⋯ , a g ∈ Z ∖ 0 such that ∑ a i = d \sum a_{i}=d ∑ a i = d . Let ⊗ a : C 2 , g g → J 2 , g d \otimes_{a}:\mathcal{C}^{g}_{2,g}\to\mathcal{J}^{d}_{2,g} ⊗ a : C 2 , g g → J 2 , g d be the map induced by the chosen tuple ( a 1 , ⋯ , a g ) (a_{1},\cdots,a_{g}) ( a 1 , ⋯ , a g ) . The induced map 1 n × ⊗ a : C 2 , g n + g → C 2 , g n × H g J 2 , g d 1^n \times\otimes_a : C_{2,g}^{n+g} \to C_{2,g}^n \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^d 1 n × ⊗ a : C 2 , g n + g → C 2 , g n × H g J 2 , g d is generically finite and surjective. Consequently, we have ( 1 n × ⊗ a ) ∗ : A ∗ ( C 2 , g n × H g J 2 , g d ) → A ∗ ( C 2 , g n + g ) (1^n \times\otimes_a)^* : A^*(C_{2,g}^n \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^d) \to A^*(C_{2,g}^{n+g}) ( 1 n × ⊗ a ) ∗ : A ∗ ( C 2 , g n × H g J 2 , g d ) → A ∗ ( C 2 , g n + g ) is injective, and we have SFP for C 2 , g n × H g J 2 , g d → H g C_{2,g}^n \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^d \to\mathcal{H}_g C 2 , g n × H g J 2 , g d → H g by Theorem 4.2 and Theorem 1.4 .
For J 2 , g d × H g J 2 , g d ′ \mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'} J 2 , g d × H g J 2 , g d ′ , we fix coordinate-wise non-zero tuples a = ( a 1 , … , a g ) , b = ( b 1 , … , b g ) ∈ Z g a=(a_1,\ldots,a_g), b=(b_1,\ldots,b_g)\in\mathbb{Z}^g a = ( a 1 , … , a g ) , b = ( b 1 , … , b g ) ∈ Z g such that d = ∑ a i d=\sum a_i d = ∑ a i , d ′ = ∑ b j d'=\sum b_j d ′ = ∑ b j . The induced map defined by ⊗ a × ⊗ b : C 2 , g 2 g → J 2 , g d × H g J 2 , g d ′ \otimes_a \times\otimes_b : C_{2,g}^{2g} \to\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'} ⊗ a × ⊗ b : C 2 , g 2 g → J 2 , g d × H g J 2 , g d ′ is generically finite and surjective. Repeating a similar set of arguments as for C 2 , g n × H g J 2 , g d C_{2,g}^n \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^d C 2 , g n × H g J 2 , g d above, we have SFP for J 2 , g d × H g J 2 , g d ′ → H g \mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'} \to\mathcal{H}_g J 2 , g d × H g J 2 , g d ′ → H g . A straightforward two-step application of Theorem 1.3 gives SFP for J 2 , g d × H g J 2 , g d ′ \mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'} J 2 , g d × H g J 2 , g d ′ .
Corollary 4.4.1. For 1 ≤ n ≤ g + 6 1 \le n \le g+6 1 ≤ n ≤ g + 6 and d ∈ Z d \in\mathbb{Z} d ∈ Z , the stacks [ C 2 , g n / S n ] × H g J 2 , g d [C_{2,g}^n/S_n] \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^d [ C 2 , g n / S n ] × H g J 2 , g d satisfy SFP over H g \mathcal{H}_g H g .
Proof. The pullback along the symmetrising map × H g n C 2 , g × H g J 2 , g d → [ × H g n C 2 , g / S n ] × H g J 2 , g d \times_{\mathcal{H}_g}^n C_{2,g} \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^d \to[\times_{\mathcal{H}_g}^n C_{2,g}/S_n] \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^d × H g n C 2 , g × H g J 2 , g d → [ × H g n C 2 , g / S n ] × H g J 2 , g d induces an injection on Chow groups. The claim follows from Theorem 1.4 .
Remark 4.5. A similar argument as above shows that J 2 , g d → H g \mathcal{J}_{2,g}^d \to\mathcal{H}_g J 2 , g d → H g also satisfies SFP.
Definition 4.2.1. For any d ∈ Z d \in\mathbb{Z} d ∈ Z , let Ξ d \Xi_d Ξ d be the Poincaré bundle on C 2 , g × H g J 2 , g d C_{2,g} \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^d C 2 , g × H g J 2 , g d , and let p C , p J p_C,p_J p C , p J be projections onto C 2 , g , J 2 , g d C_{2,g},\mathcal{J}_{2,g}^d C 2 , g , J 2 , g d , respectively. Additionally, let p C , p J p_C,p_J p C , p J be projections of C 2 , g × H g J 2 , g C_{2,g} \times_{\mathcal{H}_g} \mathcal{J}_{2,g} C 2 , g × H g J 2 , g onto C 2 , g , J 2 , g C_{2,g},\mathcal{J}_{2,g} C 2 , g , J 2 , g , respectively.
Lemma 4.6. The following cycle is in the image of the pullback ( 1 × ν 1 , d ) ∗ A 1 ( C 2 , g × H g J 2 , g d ) (1 \times\nu_{1,d})^* A^1(C_{2,g} \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^d) ( 1 × ν 1 , d ) ∗ A 1 ( C 2 , g × H g J 2 , g d ) :
Ξ d − p J ∗ p J ∗ ( Ξ d ⋅ p C ∗ K ) 2 g − 2 . \Xi_d-\frac{p_J^*p_{J*}(\Xi_d\cdot p_C^*K)}{2g-2}. Ξ d − 2 g − 2 p J ∗ p J ∗ ( Ξ d ⋅ p C ∗ K ) .
We shall denote this cycle by ξ d \xi_d ξ d . Moreover, we have ξ d ⋅ p C ∗ K = 0 \xi_d\cdot p_C^*K=0 ξ d ⋅ p C ∗ K = 0 .
Proof. We start by showing that Ξ d − p J ∗ p J ∗ ( Ξ d ⋅ p C ∗ K ) 2 g − 2 ∈ Im ( 1 × ν 1 , d ) ∗ \Xi_d-\frac{p_J^*p_{J*}(\Xi_d\cdot p_C^*K)}{2g-2}\in\operatorname{Im}(1\times\nu_{1,d})^* Ξ d − 2 g − 2 p J ∗ p J ∗ ( Ξ d ⋅ p C ∗ K ) ∈ Im ( 1 × ν 1 , d ) ∗ . For a family L → C → π S S \mathcal{L}\to\mathcal{C}\xrightarrow{\pi_S}S L → C π S S defining a morphism S → J 2 , g d S\to\mathcal{J}_{2,g}^d S → J 2 , g d and L S ∈ Pic ( S ) \mathcal{L}_S\in\operatorname{Pic}(S) L S ∈ Pic ( S ) :
L + π S ∗ L S − π S ∗ π S ∗ ( ( L + π S ∗ L S ) ⋅ K π S ) 2 g − 2 = L − π S ∗ π S ∗ ( L ⋅ K π S ) 2 g − 2 . \mathcal{L}+\pi_S^*\mathcal{L}_S-\frac{\pi_S^*\pi_{S*}((\mathcal{L}+\pi_S^*\mathcal{L}_S)\cdot K_{\pi_S})}{2g-2}=\mathcal{L}-\frac{\pi_S^*\pi_{S*}(\mathcal{L}\cdot K_{\pi_S})}{2g-2}. L + π S ∗ L S − 2 g − 2 π S ∗ π S ∗ (( L + π S ∗ L S ) ⋅ K π S ) = L − 2 g − 2 π S ∗ π S ∗ ( L ⋅ K π S ) .
We have the claimed containment in Im ( 1 × ν 1 , d ) ∗ \operatorname{Im}(1\times\nu_{1,d})^* Im ( 1 × ν 1 , d ) ∗ by [30 Theorem 2.10] .
Next, we show that ξ d ⋅ p C ∗ K = 0 \xi_d\cdot p_C^*K=0 ξ d ⋅ p C ∗ K = 0 . We verify the relation over the cover H g , 1 w → H g \mathcal{H}_{g,1}^w\to\mathcal{H}_g H g , 1 w → H g . With slight abuse of notation, we shall identify all previously defined cycles with their respective pullbacks, and denote the projection C 2 , g , 1 w × H g J 2 , g d → H g , 1 w × H g J 2 , g d C_{2,g,1}^w\times_{\mathcal{H}_g}\mathcal{J}_{2,g}^d\to\mathcal{H}_{g,1}^w\times_{\mathcal{H}_g}\mathcal{J}_{2,g}^d C 2 , g , 1 w × H g J 2 , g d → H g , 1 w × H g J 2 , g d by p J p_J p J .
Let i σ : σ g , 1 , 1 × H g J 2 , g d ↪ C 2 , g , 1 w × H g J 2 , g d i_\sigma:\sigma_{g,1,1}\times_{\mathcal{H}_g}\mathcal{J}_{2,g}^d\hookrightarrow C_{2,g,1}^w\times_{\mathcal{H}_g}\mathcal{J}_{2,g}^d i σ : σ g , 1 , 1 × H g J 2 , g d ↪ C 2 , g , 1 w × H g J 2 , g d be the inclusion map. It is enough to check that
ξ d ∣ σ g , 1 , 1 × H g J 2 , g d = 0. \left.\xi_d\right|_{\sigma_{g,1,1}\times_{\mathcal{H}_g}\mathcal{J}_{2,g}^d}=0. ξ d ∣ σ g , 1 , 1 × H g J 2 , g d = 0.
Before we verify the triviality of the claimed restriction, we state the following relation on C 2 , g , 1 C_{2,g,1} C 2 , g , 1
p C ∗ K = ( 2 g − 2 ) σ g , 1 , 1 p_C^*K=(2g-2)\sigma_{g,1,1} p C ∗ K = ( 2 g − 2 ) σ g , 1 , 1
since A 1 ( H g , 1 w ) = 0 A^1(\mathcal{H}_{g,1}^w)=0 A 1 ( H g , 1 w ) = 0 . Consequently, we have
i σ ∗ Ξ d = p J ∗ i σ ∗ i σ ∗ Ξ d = p J ∗ ( σ g , 1 , 1 ⋅ Ξ d ) = i σ ∗ p J ∗ p J ∗ ( σ g , 1 , 1 ⋅ Ξ d ) = i σ ∗ p J ∗ p J ∗ ( Ξ d ⋅ p C ∗ K ) 2 g − 2 . i_\sigma^*\Xi_d=p_{J*}i_\sigma^*i_\sigma^*\Xi_d=p_{J*}(\sigma_{g,1,1}\cdot\Xi_d)=i_\sigma^*p_J^*p_{J*}(\sigma_{g,1,1}\cdot\Xi_d)=i_\sigma^*\frac{p_J^*p_{J*}(\Xi_d\cdot p_C^*K)}{2g-2}. i σ ∗ Ξ d = p J ∗ i σ ∗ i σ ∗ Ξ d = p J ∗ ( σ g , 1 , 1 ⋅ Ξ d ) = i σ ∗ p J ∗ p J ∗ ( σ g , 1 , 1 ⋅ Ξ d ) = i σ ∗ 2 g − 2 p J ∗ p J ∗ ( Ξ d ⋅ p C ∗ K ) .
Universal moduli of rank 2 bundles
Following §2.3, we fix d d d odd for this section.
Slope-stable bundles.
Theorem 5.1. The family U H ( 2 , d , g ) → H g \mathcal{U}^{\mathcal{H}}(2,d,g) \to\mathcal{H}_g U H ( 2 , d , g ) → H g satisfies SFP.
Proof. The stack H g \mathcal{H}_g H g is smooth, Deligne-Mumford, and has a global presentation as the group quotient of a smooth scheme. Applying Theorem 2.6 and a standard Edidin-Graham approximation argument (see [10 Definition-Proposition 1, and Propositions 14-15] ), we get the following motivic decomposition in M ( H g ) \mathrm{M}(\mathcal{H}_g) M ( H g ) :
h H g ( U H ( 2 , d , g ) ) ≃ h H g ( J 2 , g d ) ⊗ ( h H g ( C 2 , g ( g − 1 ) ) ( g − 1 ) ⊕ ⨁ i = 0 g − 2 h H g ( C 2 , g ( i ) ) ⊗ ( Q ( i ) ⊕ Q ( 3 g − 3 − 2 i ) ) ) \mathfrak{h}_{\mathcal{H}_g}(\mathcal{U}^{\mathcal{H}}(2,d,g)) \simeq\mathfrak{h}_{\mathcal{H}_g}(\mathcal{J}^{d}_{2,g}) \otimes\left(\mathfrak{h}_{\mathcal{H}_g}(\mathcal{C}^{(g-1)}_{2,g})(g-1) \oplus\bigoplus_{i=0}^{g-2}\mathfrak{h}_{\mathcal{H}_g}(\mathcal{C}^{(i)}_{2,g}) \otimes\left(\mathbb{Q}(i) \oplus\mathbb{Q}(3g-3-2i)\right)\right) h H g ( U H ( 2 , d , g )) ≃ h H g ( J 2 , g d ) ⊗ ( h H g ( C 2 , g ( g − 1 ) ) ( g − 1 ) ⊕ i = 0 ⨁ g − 2 h H g ( C 2 , g ( i ) ) ⊗ ( Q ( i ) ⊕ Q ( 3 g − 3 − 2 i ) ) )
where C 2 , g ( n ) : = [ C 2 , g n / S n ] \mathcal{C}^{(n)}_{2,g}:=[C^n_{2,g}/S_n] C 2 , g ( n ) := [ C 2 , g n / S n ] for all n ≥ 0 n\geq0 n ≥ 0 . Consequently, we obtain a corresponding decomposition of Chow groups. The claim now follows from Corollary 4.4.1. □
Corollary 5.1.1. The family U H ( 2 , d , g ) → H g \mathcal{U}^{\mathcal{H}}(2,d,g) \to\mathcal{H}_g U H ( 2 , d , g ) → H g satisfies SFP.
Proof. The rigidification map U H ( 2 , d , g ) → U H ( 2 , d , g ) \mathcal{U}^{\mathcal{H}}(2,d,g) \to\mathcal{U}^{\mathcal{H}}(2,d,g) U H ( 2 , d , g ) → U H ( 2 , d , g ) is a torsion-G m \mathbb{G}_m G m -gerbe [13 Theorem B] . The corollary follows from Theorem 5.1 and Theorem 1.3. □
Filtration stacks and the excision.
We show that excision sequences on A ∗ ( V e c H ( 2 , d , g ) ) A^*(\mathop{\mathrm{Vec}}\nolimits^{\mathcal{H}}(2,d,g)) A ∗ ( Vec H ( 2 , d , g )) given by the Harder-Narasimhan (HN) stratification are left exact.
Definition 5.2.1 ([14 , 31 ] ). We shall represent universal filtration stacks over H g \mathcal{H}_g H g of type ( 1 , 2 ) (1,2) ( 1 , 2 ) and degree ( n , d ) (n,d) ( n , d ) as F i l t ( 1 , 2 ) ( n , d ) ( g ) \mathop{\mathrm{Filt}}\nolimits^{(n,d)}_{(1,2)}(g) Filt ( 1 , 2 ) ( n , d ) ( g ) .
Theorem 5.2. The HN-stratification of V e c H ( 2 , d , g ) \mathop{\mathrm{Vec}}\nolimits^{\mathcal{H}}(2,d,g) Vec H ( 2 , d , g ) induces the following:
V e c H ( 2 , d , g ) = U H ( 2 , d , g ) ⊔ ( ∐ n ≥ ⌊ d / 2 ⌋ + 1 F i l t ( 1 , 2 ) ( n , d ) ( g ) ) \mathop{\mathrm{Vec}}\nolimits^{\mathcal{H}}(2,d,g)=\mathcal{U}^{\mathcal{H}}(2,d,g)\sqcup\left(\coprod_{n\geq\lfloor d/2\rfloor+1}\mathop{\mathrm{Filt}}\nolimits^{(n,d)}_{(1,2)}(g)\right) Vec H ( 2 , d , g ) = U H ( 2 , d , g ) ⊔ n ≥ ⌊ d /2 ⌋ + 1 ∐ Filt ( 1 , 2 ) ( n , d ) ( g ) such that the pushforward along the closed immersion
A ∗ ( F i l t ( 1 , 2 ) ( n 0 , d ) ( g ) ) ⟶ A ∗ ( V e c H ( 2 , d , g ) ∖ ∐ n ≥ n 0 + 1 F i l t ( 1 , 2 ) ( n , d ) ( g ) ) A_*\left(\mathop{\mathrm{Filt}}\nolimits^{(n_0,d)}_{(1,2)}(g)\right)\longrightarrow A_*\left(\mathop{\mathrm{Vec}}\nolimits^{\mathcal{H}}(2,d,g)\setminus\coprod_{n\geq n_0+1}\mathop{\mathrm{Filt}}\nolimits^{(n,d)}_{(1,2)}(g)\right) A ∗ ( Filt ( 1 , 2 ) ( n 0 , d ) ( g ) ) ⟶ A ∗ ( Vec H ( 2 , d , g ) ∖ n ≥ n 0 + 1 ∐ Filt ( 1 , 2 ) ( n , d ) ( g ) ) is injective for all n 0 ≥ ⌊ d / 2 ⌋ + 1 n_0\geq\lfloor d/2\rfloor+1 n 0 ≥ ⌊ d /2 ⌋ + 1 .
Proof. The HN-stratification in terms of HN-polygons has been studied extensively. For more details, see [19 §5] . The map
F i l t ( 1 , 2 ) ( n 0 , d ) ( g ) ⟶ V e c H ( 2 , d , g ) ∖ ∐ n ≥ n 0 + 1 F i l t ( 1 , 2 ) ( n , d ) ( g ) (5) \mathop{\mathrm{Filt}}\nolimits^{(n_0,d)}_{(1,2)}(g)\longrightarrow\mathop{\mathrm{Vec}}\nolimits^{\mathcal{H}}(2,d,g)\setminus\coprod_{n\geq n_0+1}\mathop{\mathrm{Filt}}\nolimits^{(n,d)}_{(1,2)}(g) \tag*{(5)} Filt ( 1 , 2 ) ( n 0 , d ) ( g ) ⟶ Vec H ( 2 , d , g ) ∖ n ≥ n 0 + 1 ∐ Filt ( 1 , 2 ) ( n , d ) ( g ) (5)
is a closed immersion by [19 Remark 5.2] .
Applying Theorem 4.4, we have SFP for the families F i l t ( 1 , 2 ) ( n , d ) ( g ) Filt_{(1,2)}^{(n,d)}(g) F i l t ( 1 , 2 ) ( n , d ) ( g ) for all n ≥ ⌊ d / 2 ⌋ + 1 n \ge\lfloor d/2 \rfloor+ 1 n ≥ ⌊ d /2 ⌋ + 1 since the forgetful morphism F i l t ( 1 , 2 ) ( n , d ) ( g ) → F 2 , g n × H g F 2 , g d − n Filt_{(1,2)}^{(n,d)}(g) \to\mathcal{F}_{2,g}^{n} \times_{\mathcal{H}_{g}} \mathcal{F}_{2,g}^{d-n} F i l t ( 1 , 2 ) ( n , d ) ( g ) → F 2 , g n × H g F 2 , g d − n is a vector bundle stack of relative dimension g − 1 + d − 2 n g - 1 + d - 2n g − 1 + d − 2 n [14 Corollary 3.2] . Therefore, we have the following diagram of excision sequences for any hyperelliptic curve C C C of genus g g g :
A ∗ ( V e c H ( 2 , d , g ) ∖ ⋃ n ≥ n 0 F i l t ( 1 , 2 ) ( n , d ) ( g ) ) → H ∗ ( B u n 2 , d ≤ ( n 0 , d ) ) i g , n 0 ∗ ↑ ↑ i C , n 0 ∗ A ∗ ( F i l t ( 1 , 2 ) ( n 0 , d ) ( g ) ) → H ∗ ( F i l t ( 1 , 2 ) ( n , d ) ( C ) ) \begin{CD}
A^{*}\left(\mathop{\mathrm{Vec}}\nolimits^{\mathcal{H}}(2,d,g) \setminus\bigcup_{n\ge n_{0}} Filt_{(1,2)}^{(n,d)}(g)\right) @>>> H^{*}\left(Bun_{2,d}^{\le(n_{0},d)}\right) \\
@A{i_{g,n_{0}*}}AA @AA{i_{C,n_{0}*}}A \\
A^{*}\left(Filt_{(1,2)}^{(n_{0},d)}(g)\right) @>>> H^{*}\left(Filt_{(1,2)}^{(n,d)}(C)\right)
\end{CD} A ∗ ( Vec H ( 2 , d , g ) ∖ n ≥ n 0 ⋃ F i l t ( 1 , 2 ) ( n , d ) ( g ) ) i g , n 0 ∗ ⏐ ↑ A ∗ ( F i l t ( 1 , 2 ) ( n 0 , d ) ( g ) ) H ∗ ( B u n 2 , d ≤ ( n 0 , d ) ) ⏐ ↑ i C , n 0 ∗ H ∗ ( F i l t ( 1 , 2 ) ( n , d ) ( C ) )
We have injectivity of i C , n 0 ∗ i_{C,n_{0}*} i C , n 0 ∗ by [19 Corollary 5.14] . Consequently, we get injectivity of i g , n 0 ∗ i_{g,n_{0}*} i g , n 0 ∗ . □
We have the following corollary as an immediate consequence of the Theorem.
Corollary 5.2.1. We have the following decomposition of the total Chow group of V e c H ( 2 , d , g ) \mathop{\mathrm{Vec}}\nolimits^{\mathcal{H}}(2,d,g) Vec H ( 2 , d , g ) :
A ∗ ( V e c H ( 2 , d , g ) ) = A ∗ ( U H ( 2 , d , g ) ) ⊕ ( ⨁ n ≥ ⌊ d / 2 ⌋ + 1 A ∗ ( F i l t ( 1 , 2 ) ( n , d ) ( g ) ) ) . A_{*}\left(\mathop{\mathrm{Vec}}\nolimits^{\mathcal{H}}(2,d,g)\right) = A_{*}\left(\mathcal{U}^{\mathcal{H}}(2,d,g)\right) \oplus\left(\bigoplus_{n\ge\lfloor d/2\rfloor+1} A_{*}\left(Filt_{(1,2)}^{(n,d)}(g)\right)\right). A ∗ ( Vec H ( 2 , d , g ) ) = A ∗ ( U H ( 2 , d , g ) ) ⊕ n ≥ ⌊ d /2 ⌋ + 1 ⨁ A ∗ ( F i l t ( 1 , 2 ) ( n , d ) ( g ) ) .
Theorem 5.3. The family V e c H ( 2 , d , g ) → H g \mathop{\mathrm{Vec}}\nolimits^{\mathcal{H}}(2,d,g) \to\mathcal{H}_{g} Vec H ( 2 , d , g ) → H g satisfies SFP.
Proof. We have SFP for U H ( 2 , d , g ) \mathcal{U}^{\mathcal{H}}(2,d,g) U H ( 2 , d , g ) by Corollary 5.1.1. For F i l t ( 1 , 2 ) ( n , d ) ( g ) Filt_{(1,2)}^{(n,d)}(g) F i l t ( 1 , 2 ) ( n , d ) ( g ) , SFP was shown in Theorem 5.2. The claim now follows from Corollary 5.2.1. □
Theorem 5.4. Given d ∈ Z d \in\mathbb{Z} d ∈ Z odd, the tautological rings R ∗ ( V e c H ( 2 , d , g ) ) R^{*}(\mathop{\mathrm{Vec}}\nolimits^{\mathcal{H}}(2,d,g)) R ∗ ( Vec H ( 2 , d , g )) and R ∗ ( U H ( 2 , d , g ) ) R^{*}(\mathcal{U}^{\mathcal{H}}(2,d,g)) R ∗ ( U H ( 2 , d , g )) are generated by the twisted-kappa classes { κ − 1 , 2 , 0 , κ 0 , 1 , 0 , κ − 1 , 0 , 1 , κ 0 , 0 , 1 , κ − 1 , 1 , 1 , κ − 1 , 0 , 2 } \{\kappa_{-1,2,0}, \kappa_{0,1,0}, \kappa_{-1,0,1}, \kappa_{0,0,1}, \kappa_{-1,1,1}, \kappa_{-1,0,2}\} { κ − 1 , 2 , 0 , κ 0 , 1 , 0 , κ − 1 , 0 , 1 , κ 0 , 0 , 1 , κ − 1 , 1 , 1 , κ − 1 , 0 , 2 } for all g ≥ 2 g \ge2 g ≥ 2 . Additionally, we have the following relations in A ∗ ( V e c H ( 2 , d , g ) ) A^{*}(\mathop{\mathrm{Vec}}\nolimits^{\mathcal{H}}(2,d,g)) A ∗ ( Vec H ( 2 , d , g )) :
( 2 g − 2 ) n − 1 κ − 1 , 0 , n = κ 0 , 0 , 1 n − 2 ( n ( n − 1 ) ( g − 1 ) κ − 1 , 0 , 2 − ( n 2 − 2 n ) κ − 1 , 0 , 1 κ 0 , 0 , 1 ) ∀ n ≥ 2 , (2g - 2)^{n-1}\kappa_{-1,0,n} = \kappa_{0,0,1}^{n-2}\left(n(n - 1)(g - 1)\kappa_{-1,0,2} - (n^{2} - 2n)\kappa_{-1,0,1}\kappa_{0,0,1}\right)\quad\forall n \ge2, ( 2 g − 2 ) n − 1 κ − 1 , 0 , n = κ 0 , 0 , 1 n − 2 ( n ( n − 1 ) ( g − 1 ) κ − 1 , 0 , 2 − ( n 2 − 2 n ) κ − 1 , 0 , 1 κ 0 , 0 , 1 ) ∀ n ≥ 2 ,
( 2 g − 2 ) n + m − 2 κ − 1 , m , n = ( 2 g − 2 ) m − 2 κ 0 , 0 , 1 n κ − 1 , m , 0 + ( 2 g − 2 ) n − 2 κ 0 , 1 , 0 m κ − 1 , 0 , n + n m κ 0 , 1 , 0 m − 1 κ 0 , 0 , 1 n − 1 ( κ − 1 , 1 , 1 − 1 2 g − 2 ( κ − 1 , 0 , 1 κ 0 , 1 , 0 + d κ 0 , 0 , 1 ) ) ∀ m , n ≥ 1 , (2g - 2)^{n+m-2}\kappa_{-1,m,n} = (2g - 2)^{m-2}\kappa_{0,0,1}^{n}\kappa_{-1,m,0} + (2g - 2)^{n-2}\kappa_{0,1,0}^{m}\kappa_{-1,0,n} + nm\kappa_{0,1,0}^{m-1}\kappa_{0,0,1}^{n-1}\left(\kappa_{-1,1,1} - \frac{1}{2g - 2}\left(\kappa_{-1,0,1}\kappa_{0,1,0} + d\kappa_{0,0,1}\right)\right)\quad\forall m,n \ge1, ( 2 g − 2 ) n + m − 2 κ − 1 , m , n = ( 2 g − 2 ) m − 2 κ 0 , 0 , 1 n κ − 1 , m , 0 + ( 2 g − 2 ) n − 2 κ 0 , 1 , 0 m κ − 1 , 0 , n + nm κ 0 , 1 , 0 m − 1 κ 0 , 0 , 1 n − 1 ( κ − 1 , 1 , 1 − 2 g − 2 1 ( κ − 1 , 0 , 1 κ 0 , 1 , 0 + d κ 0 , 0 , 1 ) ) ∀ m , n ≥ 1 ,
( 2 g − 2 ) m + n − 1 κ 0 , m , n = κ 0 , 1 , 0 m κ 0 , 0 , 1 n ∀ m , n ≥ 0 , (2g - 2)^{m+n-1}\kappa_{0,m,n} = \kappa_{0,1,0}^{m}\kappa_{0,0,1}^{n}\quad\forall m,n \ge0, ( 2 g − 2 ) m + n − 1 κ 0 , m , n = κ 0 , 1 , 0 m κ 0 , 0 , 1 n ∀ m , n ≥ 0 ,
κ i , m , n = 0 ∀ i ≥ 1. \kappa_{i,m,n} = 0\quad\forall i \ge1. κ i , m , n = 0 ∀ i ≥ 1.
Proof. By excision on U H ( 2 , d , g ) ⊂ V e c H ( 2 , d , g ) \mathcal{U}^{\mathcal{H}}(2,d,g) \subset\mathop{\mathrm{Vec}}\nolimits^{\mathcal{H}}(2,d,g) U H ( 2 , d , g ) ⊂ Vec H ( 2 , d , g ) and Theorem 5.3, it is enough to verify the claimed relations in H ∗ ( B u n 2 , d C ) H^{*}(Bun_{2,d}C) H ∗ ( B u n 2 , d C ) for a fixed hyperelliptic curve C C C of genus g g g .
Following the notation in [31 Theorem 3.6.1] , let E d , C → C × B u n 2 , d C \mathcal{E}_{d,C} \to C \times Bun_{2,d}C E d , C → C × B u n 2 , d C be the universal bundle and for i ∈ { 1 , 2 } i \in\{1,2\} i ∈ { 1 , 2 } , let
c i ( E d , C ) : = 1 ⊗ a i + ∑ j = 1 2 g γ j ⊗ b j i + [ p t ] ⊗ f i c_{i}(\mathcal{E}_{d,C}) := 1 \otimes a_{i} + \sum_{j=1}^{2g} \gamma_{j} \otimes b_{j}^{i} + [pt] \otimes f_{i} c i ( E d , C ) := 1 ⊗ a i + j = 1 ∑ 2 g γ j ⊗ b j i + [ pt ] ⊗ f i
be the Künneth decomposition of the Chern classes of E d , C \mathcal{E}_{d,C} E d , C where { γ j } 1 ≤ j ≤ 2 g \{\gamma_j\}_{1\leq j\leq2g} { γ j } 1 ≤ j ≤ 2 g is a symplectic basis of H 1 ( C ) H^1(C) H 1 ( C ) . We have
a 1 = κ 0 , 1 , 0 2 g − 2 , a 2 = κ 0 , 0 , 1 2 g − 2 , f 1 = d , f 2 = κ − 1 , 0 , 1 . a_1=\frac{\kappa_{0,1,0}}{2g-2},\quad a_2=\frac{\kappa_{0,0,1}}{2g-2},\quad f_1=d,\ f_2=\kappa_{-1,0,1}. a 1 = 2 g − 2 κ 0 , 1 , 0 , a 2 = 2 g − 2 κ 0 , 0 , 1 , f 1 = d , f 2 = κ − 1 , 0 , 1 .
It immediately follows that κ i , m , n = 0 \kappa_{i,m,n}=0 κ i , m , n = 0 for all i ≥ 1 i\geq1 i ≥ 1 since [ p t ] 2 = 0 ∈ H ∗ ( C ) [\mathrm{pt}]^2=0\in H^*(C) [ pt ] 2 = 0 ∈ H ∗ ( C ) . On the other hand, we have the following relations in H ∗ ( Bun 2 , d C ) H^*(\operatorname{Bun}_{2,d}C) H ∗ ( Bun 2 , d C ) :
κ 0 , m , n = ( 2 g − 2 ) a 1 m a 2 n = ( 2 g − 2 ) ⋅ κ 0 , 1 , 0 m ( 2 g − 2 ) m ⋅ κ 0 , 0 , 1 n ( 2 g − 2 ) n = κ 0 , 1 , 0 m κ 0 , 0 , 1 n ( 2 g − 2 ) m + n − 1 ∀ m , n ≥ 0 , \kappa_{0,m,n}=(2g-2)a_1^m a_2^n=(2g-2)\cdot\frac{\kappa_{0,1,0}^m}{(2g-2)^m}\cdot\frac{\kappa_{0,0,1}^n}{(2g-2)^n}=\frac{\kappa_{0,1,0}^m\kappa_{0,0,1}^n}{(2g-2)^{m+n-1}}\quad\forall m,n\geq0, κ 0 , m , n = ( 2 g − 2 ) a 1 m a 2 n = ( 2 g − 2 ) ⋅ ( 2 g − 2 ) m κ 0 , 1 , 0 m ⋅ ( 2 g − 2 ) n κ 0 , 0 , 1 n = ( 2 g − 2 ) m + n − 1 κ 0 , 1 , 0 m κ 0 , 0 , 1 n ∀ m , n ≥ 0 ,
κ − 1 , 0 , n = n a 2 n − 1 f 2 − 2 ( n 2 ) a 2 n − 2 ( ∑ j = 1 g b j 2 b j + g 2 ) = n κ 0 , 0 , 1 n − 1 κ − 1 , 0 , 1 ( 2 g − 2 ) n − 1 + n ( n − 1 ) κ 0 , 0 , 1 n − 2 2 ( 2 g − 2 ) n − 2 ( κ − 1 , 0 , 2 − κ 0 , 0 , 1 κ − 1 , 0 , 1 g − 1 ) ∀ n ≥ 2 , \kappa_{-1,0,n}=na_2^{n-1}f_2-2\binom{n}{2}a_2^{n-2}\left(\sum_{j=1}^{g}b_j^2b_{j+g}^2\right)=\frac{n\kappa_{0,0,1}^{n-1}\kappa_{-1,0,1}}{(2g-2)^{n-1}}+\frac{n(n-1)\kappa_{0,0,1}^{n-2}}{2(2g-2)^{n-2}}\left(\kappa_{-1,0,2}-\frac{\kappa_{0,0,1}\kappa_{-1,0,1}}{g-1}\right)\quad\forall n\geq2, κ − 1 , 0 , n = n a 2 n − 1 f 2 − 2 ( 2 n ) a 2 n − 2 ( ∑ j = 1 g b j 2 b j + g 2 ) = ( 2 g − 2 ) n − 1 n κ 0 , 0 , 1 n − 1 κ − 1 , 0 , 1 + 2 ( 2 g − 2 ) n − 2 n ( n − 1 ) κ 0 , 0 , 1 n − 2 ( κ − 1 , 0 , 2 − g − 1 κ 0 , 0 , 1 κ − 1 , 0 , 1 ) ∀ n ≥ 2 ,
κ − 1 , m , n = a 1 m κ − 1 , 0 , n + a 2 n κ − 1 , m , 0 − m n a 1 m − 1 a 2 n − 1 ( ∑ j = 1 g b j 1 b j + g 2 − b j + g 1 b j 2 ) \kappa_{-1,m,n}=a_1^m\kappa_{-1,0,n}+a_2^n\kappa_{-1,m,0}-mn a_1^{m-1}a_2^{n-1}\left(\sum_{j=1}^{g}b_j^1b_{j+g}^2-b_{j+g}^1b_j^2\right) κ − 1 , m , n = a 1 m κ − 1 , 0 , n + a 2 n κ − 1 , m , 0 − mn a 1 m − 1 a 2 n − 1 ( ∑ j = 1 g b j 1 b j + g 2 − b j + g 1 b j 2 )
= κ 0 , 1 , 0 m κ − 1 , 0 , n ( 2 g − 2 ) m + κ 0 , 0 , 1 n κ − 1 , m , 0 ( 2 g − 2 ) n + n m κ 0 , 1 , 0 m − 1 κ 0 , 0 , 1 n − 1 ( 2 g − 2 ) m + n − 2 ( κ − 1 , 1 , 1 − κ 0 , 1 , 0 κ − 1 , 0 , 1 + κ 0 , 0 , 1 κ − 1 , 1 , 0 2 g − 2 ) =\frac{\kappa_{0,1,0}^m\kappa_{-1,0,n}}{(2g-2)^m}+\frac{\kappa_{0,0,1}^n\kappa_{-1,m,0}}{(2g-2)^n}+\frac{nm\kappa_{0,1,0}^{m-1}\kappa_{0,0,1}^{n-1}}{(2g-2)^{m+n-2}}\left(\kappa_{-1,1,1}-\frac{\kappa_{0,1,0}\kappa_{-1,0,1}+\kappa_{0,0,1}\kappa_{-1,1,0}}{2g-2}\right) = ( 2 g − 2 ) m κ 0 , 1 , 0 m κ − 1 , 0 , n + ( 2 g − 2 ) n κ 0 , 0 , 1 n κ − 1 , m , 0 + ( 2 g − 2 ) m + n − 2 nm κ 0 , 1 , 0 m − 1 κ 0 , 0 , 1 n − 1 ( κ − 1 , 1 , 1 − 2 g − 2 κ 0 , 1 , 0 κ − 1 , 0 , 1 + κ 0 , 0 , 1 κ − 1 , 1 , 0 )
∀ m , n ≥ 1. \forall m,n\geq1. ∀ m , n ≥ 1.
Applying [24 Corollary 1.2] , we have the claimed generators for R ∗ ( V e c H ( 2 , d , g ) ) R^*(\mathop{\mathrm{Vec}}\nolimits_{\mathcal H}(2,d,g)) R ∗ ( Vec H ( 2 , d , g )) .
We end this section with a remark which we shall refer to repeatedly in the rest of the paper.
Remark 5.5. Fix C C C , a hyperelliptic curve in H g \mathcal H_g H g . X → H g \mathcal X\to\mathcal H_g X → H g is smooth where X \mathcal X X is one of the stacks considered in Theorems 4.4 , 5.1 and 5.3 . Global cycles in all of these cases give monodromy invariant cohomology after restricting to a fiber over a geometric point.
We elaborate more on this point. Let X ⟶ π B \mathcal X\overset{\pi}{\longrightarrow}B X ⟶ π B be a smooth morphism between smooth algebraic stacks and B B B is Deligne-Mumford. Then the cycle class map gives a group homomorphism cl : A r ( X ) → H 2 r ( X , Q ) ≃ Hom D ( B ) ( Q B , R π ∗ Q X [ 2 r ] ) \operatorname{cl}:A^r(\mathcal X)\to H^{2r}(\mathcal X,\mathbb Q)\simeq\operatorname{Hom}_{D(B)}(\mathbb Q_B,R\pi_*\mathbb Q_{\mathcal X}[2r]) cl : A r ( X ) → H 2 r ( X , Q ) ≃ Hom D ( B ) ( Q B , R π ∗ Q X [ 2 r ]) . Taking 0th cohomology gives us a morphism A r ( X ) → H 0 ( B , R 2 r π ∗ Q X ) A^r(\mathcal X)\to H^0(B,R^{2r}\pi_*\mathbb Q_{\mathcal X}) A r ( X ) → H 0 ( B , R 2 r π ∗ Q X ) . Suppose additionally, that R 2 r π ∗ Q X R^{2r}\pi_*\mathbb Q_{\mathcal X} R 2 r π ∗ Q X is a local system with stalk H 2 r ( X b , Q ) H^{2r}(\mathcal X_b,\mathbb Q) H 2 r ( X b , Q ) . Then we have the isomorphism H 0 ( B , R 2 r π ∗ Q X ) ≃ H 2 r ( X b , Q ) π 1 ( B , b ) H^0(B,R^{2r}\pi_*\mathbb Q_{\mathcal X})\simeq H^{2r}(\mathcal X_b,\mathbb Q)^{\pi_1(B,b)} H 0 ( B , R 2 r π ∗ Q X ) ≃ H 2 r ( X b , Q ) π 1 ( B , b ) , so we finally get a homomorphism
A r ( X ) → H 2 r ( X b , Q ) π 1 ( B , b ) . A^r(\mathcal X)\to H^{2r}(\mathcal X_b,\mathbb Q)^{\pi_1(B,b)}. A r ( X ) → H 2 r ( X b , Q ) π 1 ( B , b ) .
In particular if X ⟶ π B \mathcal X\overset{\pi}{\longrightarrow}B X ⟶ π B is topologically locally trivial upto homotopy type, the cycle class map and restriction to a fiber gives a homomorphism A r ( X ) → H 2 r ( X b , Q ) π 1 ( B , b ) A^r(\mathcal X)\to H^{2r}(\mathcal X_b,\mathbb Q)^{\pi_1(B,b)} A r ( X ) → H 2 r ( X b , Q ) π 1 ( B , b ) .
For smooth proper morphisms this follows from the Leray spectral sequence and Ehresmann’s theorem, and for torsion G m \mathbb G_m G m -gerbes over a smooth proper family this can be shown using Leray-Hirsch. Finally for V e c H ( 2 , d , g ) → H g \mathop{\mathrm{Vec}}\nolimits_{\mathcal H}(2,d,g)\to\mathcal H_g Vec H ( 2 , d , g ) → H g , this follows from the works of Atiyah-Bott in [1 ] . So, in view of SFP, the cycle class map and restriction to a geometric fiber gives an embedding
A ∗ ( X ) ↪ H 2 ∗ ( X [ C ] , Q ) π 1 ( H g , [ C ] ) . A^*(\mathcal X)\hookrightarrow H^{2*}(\mathcal X_{[C]},\mathbb Q)^{\pi_1(\mathcal H_g,[C])}. A ∗ ( X ) ↪ H 2 ∗ ( X [ C ] , Q ) π 1 ( H g , [ C ]) .
We know that the standard monodromy representation on H ∗ ( X [ C ] , Q ) H^*(\mathcal X_{[C]},\mathbb Q) H ∗ ( X [ C ] , Q ) in each of these cases is induced by an algebraic representation of Sp ( V ) \operatorname{Sp}(V) Sp ( V ) where V = H 1 ( C , Q ) V=H^1(C,\mathbb Q) V = H 1 ( C , Q ) with the standard action.
Moreover, from the results of [2 ] (see also [23 Theorem 8.1] ), we know that the image of π 1 ( H g , [ C ] ) \pi_1(\mathcal{H}_g,[C]) π 1 ( H g , [ C ]) under the standard monodromy representation is Zariski dense in Sp ( V ) \operatorname{Sp}(V) Sp ( V ) . This gives us
H 2 ∗ ( X [ C ] , Q ) π 1 ( H g , [ C ] ) = H 2 ∗ ( X [ C ] , Q ) Sp ( V ) . H^{2*}(\mathcal{X}_{[C]},\mathbb{Q})^{\pi_1(\mathcal{H}_g,[C])}=H^{2*}(\mathcal{X}_{[C]},\mathbb{Q})^{\operatorname{Sp}(V)}. H 2 ∗ ( X [ C ] , Q ) π 1 ( H g , [ C ]) = H 2 ∗ ( X [ C ] , Q ) Sp ( V ) .
Proofs of the main theorems
Product of universal Jacobians
The following results are analogues of [30 Theorem 4.4] . We include them below for completeness and to address minor technical difficulties that arise when we generalize the result to higher genus.
We also indicate here how this gives an alternative proof of [24 Theorem 1.1] . We use the fact for any d ∈ Z d \in\mathbb{Z} d ∈ Z , there exists Θ d ∈ A 1 ( J 2 , g d ) \Theta_d \in A^1(\mathcal{J}^{d}_{2,g}) Θ d ∈ A 1 ( J 2 , g d ) such that the restriction of Θ d \Theta_d Θ d to J C d ≃ J C g − 1 J^d_C \simeq J^{g-1}_C J C d ≃ J C g − 1 is Θ C \Theta_C Θ C in H 2 ( J C g − 1 , Q ) H^2(J^{g-1}_C,\mathbb{Q}) H 2 ( J C g − 1 , Q ) where the isomorphism J C d ≃ J C g − 1 J^d_C \simeq J^{g-1}_C J C d ≃ J C g − 1 is given by twisting with an arbitrary divisor of degree d − g + 1 d-g+1 d − g + 1 . This follows from [29 Corollary 6.3(ii)] . Moreover, we can choose Θ d \Theta_d Θ d so that
Θ d : = d κ 0 , 1 − ( g − 1 ) κ − 1 , 2 2 ( g − 1 ) ∈ A 1 ( J 2 , g d ) . (6) \Theta_d:=\frac{d\kappa_{0,1}-(g-1)\kappa_{-1,2}}{2(g-1)}\in A^1(\mathcal{J}^{d}_{2,g}). \tag*{(6)} Θ d := 2 ( g − 1 ) d κ 0 , 1 − ( g − 1 ) κ − 1 , 2 ∈ A 1 ( J 2 , g d ) . (6)
Theorem 6.1. [24 Theorem 1.1] We have the following isomorphism of graded rings
A ∗ ( J 2 , g d ) ≃ H 2 ∗ ( P i c d ( C ) , Q ) π 1 ( H g , [ C ] ) ≃ Q [ κ 0 , 1 , κ − 1 , 2 ] / ⟨ ( d κ 0 , 1 − ( g − 1 ) κ − 1 , 2 ) g + 1 ⟩ A^*(\mathcal{J}^{d}_{2,g})\simeq H^{2*}(\mathop{\rm Pic}\nolimits^d(C),\mathbb{Q})^{\pi_1(\mathcal{H}_g,[C])}\simeq\mathbb{Q}[\kappa_{0,1},\kappa_{-1,2}]/\langle(d\kappa_{0,1}-(g-1)\kappa_{-1,2})^{g+1}\rangle A ∗ ( J 2 , g d ) ≃ H 2 ∗ ( Pic d ( C ) , Q ) π 1 ( H g , [ C ]) ≃ Q [ κ 0 , 1 , κ − 1 , 2 ] / ⟨( d κ 0 , 1 − ( g − 1 ) κ − 1 , 2 ) g + 1 ⟩
Proof. We know J 2 , g d → J 2 , g d \mathcal{J}^{d}_{2,g}\to J^{d}_{2,g} J 2 , g d → J 2 , g d is a torsion G m \mathbb{G}_m G m -gerbe so by SFP and Remark 5.5, we already get an embedding of graded rings A ∗ ( J 2 , g d ) ↪ H 2 ∗ ( P i c d ( C ) , Q ) π 1 ( H g , [ C ] ) = H 2 ∗ ( P i c d ( C ) , Q ) Sp ( V ) A^*(\mathcal{J}^{d}_{2,g})\hookrightarrow H^{2*}(\mathop{\rm Pic}\nolimits^d(C),\mathbb{Q})^{\pi_1(\mathcal{H}_g,[C])}=H^{2*}(\mathop{\rm Pic}\nolimits^d(C),\mathbb{Q})^{\operatorname{Sp}(V)} A ∗ ( J 2 , g d ) ↪ H 2 ∗ ( Pic d ( C ) , Q ) π 1 ( H g , [ C ]) = H 2 ∗ ( Pic d ( C ) , Q ) Sp ( V ) where V = H 1 ( C , Q ) V=H^1(C,\mathbb{Q}) V = H 1 ( C , Q ) . So we only need to show surjectivity and compute the ring of invariants. From the works of [19 ] , we know that there is a graded ring isomorphism
H ∗ ( P i c d ( C ) , Q ) ≃ ⋀ ∗ V ⊗ Q [ u ] , deg u = 2 H^*(\mathop{\rm Pic}\nolimits^d(C),\mathbb{Q})\simeq\bigwedge^* V\otimes\mathbb{Q}[u],\quad\deg u=2 H ∗ ( Pic d ( C ) , Q ) ≃ ⋀ ∗ V ⊗ Q [ u ] , deg u = 2
which is also Sp ( V ) \operatorname{Sp}(V) Sp ( V ) -equivariant (the action of Sp ( V ) \operatorname{Sp}(V) Sp ( V ) on Q [ u ] \mathbb{Q}[u] Q [ u ] is trivial). So we have
H ∗ ( P i c d ( C ) , Q ) Sp ( V ) = Q [ θ ] / ( θ g + 1 ) ⊗ Q [ u ] H^*(\mathop{\rm Pic}\nolimits^d(C),\mathbb{Q})^{\operatorname{Sp}(V)}=\mathbb{Q}[\theta]/(\theta^{g+1})\otimes\mathbb{Q}[u] H ∗ ( Pic d ( C ) , Q ) Sp ( V ) = Q [ θ ] / ( θ g + 1 ) ⊗ Q [ u ]
where θ ∈ ⋀ 2 V \theta\in\bigwedge^2 V θ ∈ ⋀ 2 V is dual to the symplectic form. In fact, from [19 Theorem 4.7] and the calculations in loc. cit., we have the formulas θ = d κ 0 , 1 − ( g − 1 ) κ − 1 , 2 2 ( g − 1 ) \theta=\frac{d\kappa_{0,1}-(g-1)\kappa_{-1,2}}{2(g-1)} θ = 2 ( g − 1 ) d κ 0 , 1 − ( g − 1 ) κ − 1 , 2 and u = κ 0 , 1 2 ( g − 1 ) u=\frac{\kappa_{0,1}}{2(g-1)} u = 2 ( g − 1 ) κ 0 , 1 . This completes the proof of surjectivity since the κ \kappa κ -classes are global cycles on J 2 , g d \mathcal{J}^{d}_{2,g} J 2 , g d . □
Lemma 6.2. For all g ≥ 2 g\geq2 g ≥ 2 , the ring A ∗ ( J 2 , g 0 × H g J 2 , g 0 ) A^*(\mathcal{J}^{0}_{2,g}\times_{\mathcal{H}_g}\mathcal{J}^{0}_{2,g}) A ∗ ( J 2 , g 0 × H g J 2 , g 0 ) is generated by π 1 ∗ Θ 0 , π 2 ∗ Θ 0 , μ ∗ Θ 0 \pi_1^*\Theta_0,\pi_2^*\Theta_0,\mu^*\Theta_0 π 1 ∗ Θ 0 , π 2 ∗ Θ 0 , μ ∗ Θ 0 as a Q \mathbb{Q} Q -algebra where μ : J 2 , g 0 × H g J 2 , g 0 → J 2 , g 0 \mu:\mathcal{J}^{0}_{2,g}\times_{\mathcal{H}_g}\mathcal{J}^{0}_{2,g}\to\mathcal{J}^{0}_{2,g} μ : J 2 , g 0 × H g J 2 , g 0 → J 2 , g 0 is the twisting map and π i \pi_i π i is the projection onto the i i i -th factor for i = 1 , 2 i=1,2 i = 1 , 2 .
Proof. Following the Q \mathbb{Q} Q -linear given by Theorem 3.1 for tuples { ( 1 , 0 ) , ( 1 , 0 ) } \{(1,0),(1,0)\} {( 1 , 0 ) , ( 1 , 0 )} , along with the notation in Lemma 4.6, A ∗ ( J 2 , g 0 × H g J 2 , g 0 ) A^*(\mathcal{J}^{0}_{2,g}\times_{\mathcal{H}_g}\mathcal{J}^{0}_{2,g}) A ∗ ( J 2 , g 0 × H g J 2 , g 0 ) is Q \mathbb{Q} Q -linearly generated by the following cycles:
{ p J × J ∗ ( ∏ 1 ≤ i ≤ 2 g c 1 ( ξ 0 ) a i ρ ∗ ( Γ ) ) } Γ ∈ A ∗ ( C 2 , g 2 g ) , a i ≥ 0 (7) \left\{p_{\mathcal{J}\times\mathcal{J}*}\left(\prod_{1\leq i\leq2g}c_1(\xi_0)^{a_i}\rho^*(\Gamma)\right)\right\}_{\Gamma\in A^*(\mathcal{C}^{2g}_{2,g}),a_i\geq0} \tag*{(7)} { p J × J ∗ ( 1 ≤ i ≤ 2 g ∏ c 1 ( ξ 0 ) a i ρ ∗ ( Γ ) ) } Γ ∈ A ∗ ( C 2 , g 2 g ) , a i ≥ 0 (7) where p J × J , ρ p_{\mathcal{J}\times\mathcal{J}},\rho p J × J , ρ are projections of J 2 , g 0 × H g J 2 , g 0 × H g C 2 , g 2 g \mathcal{J}^{0}_{2,g}\times_{\mathcal{H}_g}\mathcal{J}^{0}_{2,g}\times_{\mathcal{H}_g}\mathcal{C}^{2g}_{2,g} J 2 , g 0 × H g J 2 , g 0 × H g C 2 , g 2 g onto J 2 , g 0 × H g J 2 , g 0 , C 2 , g 2 g \mathcal{J}^{0}_{2,g}\times_{\mathcal{H}_g}\mathcal{J}^{0}_{2,g},\mathcal{C}^{2g}_{2,g} J 2 , g 0 × H g J 2 , g 0 , C 2 , g 2 g respectively.
We analyze the restrictions of these generators, in (6), over [ C ] ∈ H g [C] \in\mathcal{H}_g [ C ] ∈ H g since it is enough to show the claim for the image of the restriction map A ∗ ( J 2 , g 0 × H g J 2 , g 0 ) → H ∗ ( J C 0 × J C 0 ) A^*(\mathcal{J}_{2,g}^0 \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^0) \to H^*(J_C^0 \times J_C^0) A ∗ ( J 2 , g 0 × H g J 2 , g 0 ) → H ∗ ( J C 0 × J C 0 ) for a hyperelliptic curve C C C of genus g g g by Theorem 4.4 .
Fixing a symplectic basis of H 1 ( C , Q ) H^1(C,\mathbb{Q}) H 1 ( C , Q ) and the Künneth formula, an immediate consequence is that A ∗ ( J 2 , g 0 × H g J 2 , g 0 ) A^*(\mathcal{J}_{2,g}^0 \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^0) A ∗ ( J 2 , g 0 × H g J 2 , g 0 ) is a subalgebra of
⋀ 1 ≤ i ≤ 2 g , 1 ≤ j ≤ 2 [ b j i ] . \bigwedge_{\substack{1\leq i\leq2g,\\1\leq j\leq2}} [b_j^i]. 1 ≤ i ≤ 2 g , 1 ≤ j ≤ 2 ⋀ [ b j i ] .
Using the image of the restriction A ∗ ( C 2 , g 2 g ) → H ∗ ( C 2 g ) A^*(C_{2,g}^{2g}) \to H^*(C^{2g}) A ∗ ( C 2 , g 2 g ) → H ∗ ( C 2 g ) , given by Theorem 4.2 , A ∗ ( J 2 , g 0 × H g J 2 , g 0 ) A^*(\mathcal{J}_{2,g}^0 \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^0) A ∗ ( J 2 , g 0 × H g J 2 , g 0 ) is Q \mathbb{Q} Q -linearly generated by
{ ( ∑ 1 ≤ i ≤ g b i 1 b i + g 1 ) a ( ∑ 1 ≤ j ≤ g b j 2 b j + g 2 ) b ( ∑ 1 ≤ k ≤ g b k 1 b k + g 2 + b k 2 b k + g 1 ) c } a , b , c ≥ 0 . \left\{\left(\sum_{1\leq i\leq g}b_i^1b_{i+g}^1\right)^a\left(\sum_{1\leq j\leq g}b_j^2b_{j+g}^2\right)^b\left(\sum_{1\leq k\leq g}b_k^1b_{k+g}^2+b_k^2b_{k+g}^1\right)^c\right\}_{a,b,c\geq0}. ⎩ ⎨ ⎧ ( 1 ≤ i ≤ g ∑ b i 1 b i + g 1 ) a ( 1 ≤ j ≤ g ∑ b j 2 b j + g 2 ) b 1 ≤ k ≤ g ∑ b k 1 b k + g 2 + b k 2 b k + g 1 c ⎭ ⎬ ⎫ a , b , c ≥ 0 .
The claim follows consequently since π 1 ∗ Θ 0 , π 2 ∗ Θ 0 , μ ∗ Θ 0 \pi_1^*\Theta_0,\pi_2^*\Theta_0,\mu^*\Theta_0 π 1 ∗ Θ 0 , π 2 ∗ Θ 0 , μ ∗ Θ 0 can be identified with
∑ 1 ≤ i ≤ g b i 1 b i + g 1 , ∑ 1 ≤ j ≤ g b j 2 b j + g 2 , ∑ 1 ≤ k ≤ g ( b k 1 + b k 2 ) ( b k + g 1 + b k + g 2 ) \sum_{1\leq i\leq g}b_i^1b_{i+g}^1,\qquad
\sum_{1\leq j\leq g}b_j^2b_{j+g}^2,\qquad
\sum_{1\leq k\leq g}(b_k^1+b_k^2)(b_{k+g}^1+b_{k+g}^2) 1 ≤ i ≤ g ∑ b i 1 b i + g 1 , 1 ≤ j ≤ g ∑ b j 2 b j + g 2 , 1 ≤ k ≤ g ∑ ( b k 1 + b k 2 ) ( b k + g 1 + b k + g 2 )
respectively. □
Theorem 6.3. Recall the definition of Θ d \Theta_d Θ d in Equation (6) .
x : = π 1 ∗ Θ d , y : = π 2 ∗ Θ d ′ , z : = μ d , d ′ ∗ Θ d + d ′ − x − y ∈ A ∗ ( J 2 , g d × H g J 2 , g d ′ ) x := \pi_1^*\Theta_d,\quad y := \pi_2^*\Theta_{d'},\quad z := \mu_{d,d'}^*\Theta_{d+d'}-x-y \in A^*(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'}) x := π 1 ∗ Θ d , y := π 2 ∗ Θ d ′ , z := μ d , d ′ ∗ Θ d + d ′ − x − y ∈ A ∗ ( J 2 , g d × H g J 2 , g d ′ ) with π 1 , π 2 \pi_1,\pi_2 π 1 , π 2 projections onto the first and second factors respectively, and μ d , d ′ : J 2 , g d × H g J 2 , g d ′ → J 2 , g d + d ′ \mu_{d,d'}:\mathcal{J}_{2,g}^d\times_{\mathcal{H}_g}\mathcal{J}_{2,g}^{d'}\to\mathcal{J}_{2,g}^{d+d'} μ d , d ′ : J 2 , g d × H g J 2 , g d ′ → J 2 , g d + d ′ the twisting map. The ring A ∗ ( J 2 , g d × H g J 2 , g d ′ ) A^*(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'}) A ∗ ( J 2 , g d × H g J 2 , g d ′ ) is isomorphic to
Q [ x , y , z ] / I \mathbb{Q}[x,y,z]/I Q [ x , y , z ] / I where the ideal I I I is generated by
{ ∑ m + n + p = g + 1 , p + 2 n = t ( g + 1 m , n , p ) x m y n z p } 0 ≤ t ≤ 2 g + 2 . \left\{\sum_{\substack{m+n+p=g+1,\\p+2n=t}}\binom{g+1}{m,n,p}x^m y^n z^p\right\}_{0\leq t\leq2g+2}. ⎩ ⎨ ⎧ m + n + p = g + 1 , p + 2 n = t ∑ ( m , n , p g + 1 ) x m y n z p ⎭ ⎬ ⎫ 0 ≤ t ≤ 2 g + 2 .
Proof. A straightforward adaptation of [6 Theorem A.1] shows that we have an isomorphism
h ( J 2 , g d × H g J 2 , g d ′ ) ≃ h ( J 2 , g 0 × H g J 2 , g 0 ) \mathfrak{h}(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'})\simeq\mathfrak{h}(\mathcal{J}_{2,g}^0 \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^0) h ( J 2 , g d × H g J 2 , g d ′ ) ≃ h ( J 2 , g 0 × H g J 2 , g 0 )
preserving the cup product for all d , d ′ ∈ Z d,d'\in\mathbb{Z} d , d ′ ∈ Z . Therefore, it is enough to show the claim for all d , d ′ = 0 d,d'=0 d , d ′ = 0 . Consequently, we have that A ∗ ( J 2 , g d × H g J 2 , g d ′ ) A^*(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'}) A ∗ ( J 2 , g d × H g J 2 , g d ′ ) is generated by x , y , z x,y,z x , y , z as a Q \mathbb{Q} Q -algebra by Lemma 6.2.
Following the notation established in Lemma 6.2, set u N : = ∑ 1 ≤ l ≤ g ( b l 1 + N b l 2 ) ( b l + g 1 + N b l + g 2 ) u_N:=\sum_{1\leq l\leq g}(b_l^1+Nb_l^2)(b_{l+g}^1+Nb_{l+g}^2) u N := ∑ 1 ≤ l ≤ g ( b l 1 + N b l 2 ) ( b l + g 1 + N b l + g 2 ) for all N ∈ Z N\in\mathbb{Z} N ∈ Z . It is clear that
u N = x + N z + N 2 y , u N g + 1 = 0 ∀ N ∈ Z . u_N=x+Nz+N^2y,\qquad u_N^{g+1}=0\quad\forall N\in\mathbb{Z}. u N = x + N z + N 2 y , u N g + 1 = 0 ∀ N ∈ Z .
The claimed generators are coefficients of N t N^t N t in the expansion of ( x + N z + N 2 y ) g + 1 (x+Nz+N^2y)^{g+1} ( x + N z + N 2 y ) g + 1 for all $0\leq t\leq 2g+2. T h e r e f o r e , t h e g e n e r a t o r s a r e c o n t a i n e d i n . Therefore, the generators are contained in . T h er e f or e , t h e g e n er a t or s a r eco n t ain e d in I.
We show that I I I is the ideal of relations for A ∗ ( J 2 , g d × H g J 2 , g d ′ ) A^*(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'}) A ∗ ( J 2 , g d × H g J 2 , g d ′ ) . It is clear that A ∗ ( J 2 , g d × H g J 2 , g d ′ ) A^*(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'}) A ∗ ( J 2 , g d × H g J 2 , g d ′ ) , Q [ x , y , z ] / I \mathbb{Q}[x,y,z]/I Q [ x , y , z ] / I are graded rings with top degree 2 g 2g 2 g such that x , y , z x,y,z x , y , z are algebraically independent up to degree g g g in both.
We start by showing Poincaré duality for A ∗ ( J 2 , g d × H g J 2 , g d ′ ) A^*(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'}) A ∗ ( J 2 , g d × H g J 2 , g d ′ ) . Set
R : = ⋀ 1 ≤ i ≤ 2 g , 1 ≤ j ≤ 2 [ b i j ] . R := \bigwedge_{\substack{1\leq i\leq2g,\\1\leq j\leq2}} [b_i^j]. R := 1 ≤ i ≤ 2 g , 1 ≤ j ≤ 2 ⋀ [ b i j ] .
As explained in the proof of Lemma 6.2, we can identify x , y , z x,y,z x , y , z with
∑ 1 ≤ i ≤ g b i 1 b i + g 1 , ∑ 1 ≤ j ≤ g b j 2 b j + g 2 , ∑ 1 ≤ k ≤ g b k 1 b k + g 2 + b k 2 b k + g 1 \sum_{1\leq i\leq g} b_i^1b_{i+g}^1,\qquad\sum_{1\leq j\leq g} b_j^2b_{j+g}^2,\qquad\sum_{1\leq k\leq g} b_k^1b_{k+g}^2+b_k^2b_{k+g}^1 1 ≤ i ≤ g ∑ b i 1 b i + g 1 , 1 ≤ j ≤ g ∑ b j 2 b j + g 2 , 1 ≤ k ≤ g ∑ b k 1 b k + g 2 + b k 2 b k + g 1
respectively. Let V V V be a 2 g 2g 2 g -dimensional symplectic vector space with symplectic basis e 1 , f 1 , … , e g , f g e_1,f_1,\ldots,e_g,f_g e 1 , f 1 , … , e g , f g . Consider R ~ : = ⋀ ∙ ( V ⊕ V ) \widetilde{R}:=\bigwedge^\bullet(V\oplus V) R := ⋀ ∙ ( V ⊕ V ) . Sp ( V ) \operatorname{Sp}(V) Sp ( V ) acts diagonally on V ⊕ V V\oplus V V ⊕ V and hence on R ~ \widetilde{R} R . Then we have a canonical isomorphism R ≃ R ~ R\simeq\widetilde{R} R ≃ R given by b i j ↦ e i ( j ) b_i^j\mapsto e_i^{(j)} b i j ↦ e i ( j ) , 1 ≤ i ≤ g 1\leq i\leq g 1 ≤ i ≤ g , b i j = f i ( j ) b_i^j=f_i^{(j)} b i j = f i ( j ) , g + 1 ≤ i ≤ 2 g g+1\leq i\leq2g g + 1 ≤ i ≤ 2 g . From the standard facts about the representation theory of Sp ( V ) \operatorname{Sp}(V) Sp ( V ) , we know R ~ Sp ( V ) = Q [ x , y , z ] \widetilde{R}^{\operatorname{Sp}(V)}=\mathbb{Q}[x,y,z] R Sp ( V ) = Q [ x , y , z ] (this follows from the results in [36 Chapter VI] or [20 Theorem 2] ). Using Theorem 4.4 and Lemma 6.2, we have
A ∗ ( J 2 , g d × H g J 2 , g d ′ ) ≃ R ~ Sp ( V ) . A^*(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'})\simeq\widetilde{R}^{\operatorname{Sp}(V)}. A ∗ ( J 2 , g d × H g J 2 , g d ′ ) ≃ R Sp ( V ) .
Using that Sp ( V ) \operatorname{Sp}(V) Sp ( V ) is reductive and the non-degenerate pairing on R ~ \widetilde{R} R is Sp ( V ) \operatorname{Sp}(V) Sp ( V ) -invariant so the pairing on R ~ Sp ( V ) \widetilde{R}^{\operatorname{Sp}(V)} R Sp ( V ) is also non-degenerate. This follows from the existence of the Reynold’s operator R ~ → R ~ Sp ( V ) \widetilde{R}\to\widetilde{R}^{\operatorname{Sp}(V)} R → R Sp ( V ) , as shown in [28 ] (see Definition 1.5 and the subsequent discussion). This proves Poincaré duality on A ∗ ( J 2 , g d × H g J 2 , g d ′ ) A^*(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'}) A ∗ ( J 2 , g d × H g J 2 , g d ′ ) .
Since x , y , z x,y,z x , y , z are algebraically independent upto degree g g g , we have that
dim Q A k ( J 2 , g d × H g J 2 , g d ′ ) = ( k + 2 2 ) , 0 ≤ k ≤ g . \dim_{\mathbb{Q}} A^k(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'})=\binom{k+2}{2},\qquad0\leq k\leq g. dim Q A k ( J 2 , g d × H g J 2 , g d ′ ) = ( 2 k + 2 ) , 0 ≤ k ≤ g .
Poincaré duality then gives us
dim Q A k ( J 2 , g d × H g J 2 , g d ′ ) = A 2 g − k ( J 2 , g d × H g J 2 , g d ′ ) , 0 ≤ k ≤ g . \dim_{\mathbb{Q}} A^k(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'})=A^{2g-k}(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'}),\qquad0\leq k\leq g. dim Q A k ( J 2 , g d × H g J 2 , g d ′ ) = A 2 g − k ( J 2 , g d × H g J 2 , g d ′ ) , 0 ≤ k ≤ g .
So we get
dim Q A k ( J 2 , g d × H g J 2 , g d ′ ) = ( min { k , 2 g − k } + 2 2 ) . \dim_{\mathbb{Q}} A^k(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'})=\binom{\min\{k,2g-k\}+2}{2}. dim Q A k ( J 2 , g d × H g J 2 , g d ′ ) = ( 2 min { k , 2 g − k } + 2 ) .
In particular, from a straightforward computation, we get
dim Q A ∗ ( J 2 , g d × H g J 2 , g d ′ ) = ( g + 1 ) ( g + 2 ) ( 2 g + 3 ) 6 . \dim_{\mathbb{Q}} A^*(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'})=\frac{(g+1)(g+2)(2g+3)}{6}. dim Q A ∗ ( J 2 , g d × H g J 2 , g d ′ ) = 6 ( g + 1 ) ( g + 2 ) ( 2 g + 3 ) .
We know that there is a surjection Q [ x , y , z ] / I ↠ A ∗ ( J 2 , g d × H g J 2 , g d ′ ) \mathbb{Q}[x,y,z]/I\twoheadrightarrow A^*(\mathcal{J}_{2,g}^d \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'}) Q [ x , y , z ] / I ↠ A ∗ ( J 2 , g d × H g J 2 , g d ′ ) . So we only need to prove a dimension bound for isomorphism. We have the following relations in Q [ x , y , z ] / I \mathbb{Q}[x,y,z]/I Q [ x , y , z ] / I :
0 = ( g + 1 c ) x g + 1 − c z c + ∑ j = 1 ⌊ c / 2 ⌋ ( g + 1 ) ! ( g + 1 − c + j ) ! ( c − 2 j ) ! j ! x g + 1 − c + j y j z c − 2 j (8) 0=\binom{g+1}{c}x^{g+1-c}z^c+\sum_{j=1}^{\lfloor c/2\rfloor}\frac{(g+1)!}{(g+1-c+j)!(c-2j)!j!}x^{g+1-c+j}y^jz^{c-2j} \tag*{(8)} 0 = ( c g + 1 ) x g + 1 − c z c + j = 1 ∑ ⌊ c /2 ⌋ ( g + 1 − c + j )! ( c − 2 j )! j ! ( g + 1 )! x g + 1 − c + j y j z c − 2 j (8)
0 = ( g + 1 c ) y g + 1 − c z c + ∑ j = 1 ⌊ c / 2 ⌋ ( g + 1 ) ! j ! ( c − 2 j ) ! ( g + 1 − c + j ) ! x j y g + 1 − c + j z c − 2 j . 0=\binom{g+1}{c}y^{g+1-c}z^c+\sum_{j=1}^{\lfloor c/2\rfloor}\frac{(g+1)!}{j!(c-2j)!(g+1-c+j)!}x^jy^{g+1-c+j}z^{c-2j}. 0 = ( c g + 1 ) y g + 1 − c z c + j = 1 ∑ ⌊ c /2 ⌋ j ! ( c − 2 j )! ( g + 1 − c + j )! ( g + 1 )! x j y g + 1 − c + j z c − 2 j .
So Q [ x , y , z ] / I \mathbb{Q}[x,y,z]/I Q [ x , y , z ] / I is spanned by { x a y b z c ∣ 0 ≤ c ≤ g , 0 ≤ a , b ≤ g − c } \{x^ay^bz^c\mid0\le c\le g,\ 0\le a,b\le g-c\} { x a y b z c ∣ 0 ≤ c ≤ g , 0 ≤ a , b ≤ g − c } .
This gives us the upper bound
dim Q Q [ x , y , z ] / I ≤ ∑ c = 0 g ( g − c + 1 ) 2 = ( g + 1 ) ( g + 2 ) ( 2 g + 3 ) 6 . \dim_{\mathbb{Q}}\mathbb{Q}[x,y,z]/I\le\sum_{c=0}^{g}(g-c+1)^2
=\frac{(g+1)(g+2)(2g+3)}{6}. dim Q Q [ x , y , z ] / I ≤ c = 0 ∑ g ( g − c + 1 ) 2 = 6 ( g + 1 ) ( g + 2 ) ( 2 g + 3 ) .
It follows that Q [ x , y , z ] / I ↠ A ∗ ( J 2 , g d × H g J 2 , g d ′ ) \mathbb{Q}[x,y,z]/I\twoheadrightarrow A^*(\mathcal{J}_{2,g}^d\times_{\mathcal{H}_g}\mathcal{J}_{2,g}^{d'}) Q [ x , y , z ] / I ↠ A ∗ ( J 2 , g d × H g J 2 , g d ′ ) is an isomorphism. In particular Q [ x , y , z ] / I \mathbb{Q}[x,y,z]/I Q [ x , y , z ] / I also satisfies Poincaré duality. □
We rephrase the above theorem for later use in the following corollary.
Corollary 6.3.1. The subalgebra of ⋀ 1 ≤ i ≤ 2 g , 1 ≤ j ≤ 2 [ b j i ] \bigwedge_{\substack{1\le i\le2g,\\1\le j\le2}}[b_j^i] ⋀ 1 ≤ i ≤ 2 g , 1 ≤ j ≤ 2 [ b j i ] generated by the elements
x : = ∑ 1 ≤ i ≤ g b i 1 b i + g 1 , y : = ∑ 1 ≤ j ≤ g b j 2 b j + g 2 , z : = ∑ 1 ≤ k ≤ g b k 1 b k + g 2 + b k 2 b k + g 1 x:=\sum_{1\le i\le g}b_i^1b_{i+g}^1,\qquad y:=\sum_{1\le j\le g}b_j^2b_{j+g}^2,\qquad z:=\sum_{1\le k\le g}b_k^1b_{k+g}^2+b_k^2b_{k+g}^1 x := 1 ≤ i ≤ g ∑ b i 1 b i + g 1 , y := 1 ≤ j ≤ g ∑ b j 2 b j + g 2 , z := 1 ≤ k ≤ g ∑ b k 1 b k + g 2 + b k 2 b k + g 1 is given by Q [ x , y , z ] / I \mathbb{Q}[x,y,z]/I Q [ x , y , z ] / I where the ideal I I I is generated by
{ ∑ m + n + p = g + 1 p + 2 n = t ( g + 1 m , n , p ) x m y n z p } 0 ≤ t ≤ 2 g + 2 . \left\{\sum_{\substack{m+n+p=g+1\\p+2n=t}}\binom{g+1}{m,n,p}x^my^nz^p\right\}_{0\le t\le2g+2}. ⎩ ⎨ ⎧ m + n + p = g + 1 p + 2 n = t ∑ ( m , n , p g + 1 ) x m y n z p ⎭ ⎬ ⎫ 0 ≤ t ≤ 2 g + 2 .
Corollary 6.3.2. For all g ≥ 2 g\ge2 g ≥ 2 and d d d odd, the ring R ∗ ( Vec H ( 2 , d , g ) ) R^*(\operatorname{Vec}_{\mathcal{H}}(2,d,g)) R ∗ ( Vec H ( 2 , d , g )) is isomorphic to:
Q [ κ 0 , 1 , 0 , κ 0 , 0 , 1 , κ − 1 , 0 , 1 ] ⊗ Q Q [ x , y , z ] / I \mathbb{Q}[\kappa_{0,1,0},\kappa_{0,0,1},\kappa_{-1,0,1}]\otimes_{\mathbb{Q}}\mathbb{Q}[x,y,z]/I Q [ κ 0 , 1 , 0 , κ 0 , 0 , 1 , κ − 1 , 0 , 1 ] ⊗ Q Q [ x , y , z ] / I where I I I is the ideal generated by
{ ∑ m + n + p = g + 1 p + 2 n = t ( g + 1 m , n , p ) x m y n z p } 0 ≤ t ≤ 2 g + 2 \left\{\sum_{\substack{m+n+p=g+1\\p+2n=t}}\binom{g+1}{m,n,p}x^my^nz^p\right\}_{0\le t\le2g+2} ⎩ ⎨ ⎧ m + n + p = g + 1 p + 2 n = t ∑ ( m , n , p g + 1 ) x m y n z p ⎭ ⎬ ⎫ 0 ≤ t ≤ 2 g + 2 and
x : = d κ 0 , 1 , 0 − ( g − 1 ) κ − 1 , 2 , 0 2 ( g − 1 ) , y : = κ − 1 , 0 , 1 κ 0 , 0 , 1 − ( g − 1 ) κ − 1 , 0 , 2 2 ( g − 1 ) , z : = d κ 0 , 0 , 1 + κ 0 , 1 , 0 κ − 1 , 0 , 1 − ( 2 g − 2 ) κ − 1 , 1 , 1 2 ( g − 1 ) . x:=\frac{d\kappa_{0,1,0}-(g-1)\kappa_{-1,2,0}}{2(g-1)},\qquad
y:=\frac{\kappa_{-1,0,1}\kappa_{0,0,1}-(g-1)\kappa_{-1,0,2}}{2(g-1)},\qquad
z:=\frac{d\kappa_{0,0,1}+\kappa_{0,1,0}\kappa_{-1,0,1}-(2g-2)\kappa_{-1,1,1}}{2(g-1)}. x := 2 ( g − 1 ) d κ 0 , 1 , 0 − ( g − 1 ) κ − 1 , 2 , 0 , y := 2 ( g − 1 ) κ − 1 , 0 , 1 κ 0 , 0 , 1 − ( g − 1 ) κ − 1 , 0 , 2 , z := 2 ( g − 1 ) d κ 0 , 0 , 1 + κ 0 , 1 , 0 κ − 1 , 0 , 1 − ( 2 g − 2 ) κ − 1 , 1 , 1 .
Proof. In continuation of Theorem 5.4, recall that E d , C \mathcal{E}_{d,C} E d , C is the Poincaré bundle on C × Bun 2 , d C C\times\operatorname{Bun}_{2,d}C C × Bun 2 , d C and
c i ( E d , C ) : = 1 ⊗ a i + ∑ j = 1 2 g γ j ⊗ b j i + [ p t ] ⊗ f i c_i(\mathcal{E}_{d,C}):=1\otimes a_i+\sum_{j=1}^{2g}\gamma_j\otimes b_j^i+[pt]\otimes f_i c i ( E d , C ) := 1 ⊗ a i + j = 1 ∑ 2 g γ j ⊗ b j i + [ pt ] ⊗ f i
for i = 1 , 2 i=1,2 i = 1 , 2 where a 1 = κ 0 , 1 , 0 2 g − 2 a_1=\frac{\kappa_{0,1,0}}{2g-2} a 1 = 2 g − 2 κ 0 , 1 , 0 , a 2 = κ 0 , 0 , 1 2 g − 2 a_2=\frac{\kappa_{0,0,1}}{2g-2} a 2 = 2 g − 2 κ 0 , 0 , 1 , f 1 = d f_1=d f 1 = d , f 2 = κ − 1 , 0 , 1 f_2=\kappa_{-1,0,1} f 2 = κ − 1 , 0 , 1 . Consequently, we have
κ − 1 , 2 , 0 = d κ 0 , 1 , 0 g − 1 − 2 ( ∑ 1 ≤ i ≤ g b i 1 b i + g 1 ) . \kappa_{-1,2,0}=\frac{d\kappa_{0,1,0}}{g-1}-2\left(\sum_{1\le i\le g} b_i^1b_{i+g}^1\right). κ − 1 , 2 , 0 = g − 1 d κ 0 , 1 , 0 − 2 ( 1 ≤ i ≤ g ∑ b i 1 b i + g 1 ) .
κ − 1 , 1 , 1 = κ 0 , 1 , 0 κ − 1 , 0 , 1 2 g − 2 + d κ 0 , 0 , 1 2 g − 2 − ( ∑ 1 ≤ j ≤ g b j 1 b j 2 + b j 2 b j + g 1 ) . \kappa_{-1,1,1}=\frac{\kappa_{0,1,0}\kappa_{-1,0,1}}{2g-2}+\frac{d\kappa_{0,0,1}}{2g-2}-\left(\sum_{1\le j\le g}b_j^1b_j^2+ b_j^2b_{j+g}^1\right). κ − 1 , 1 , 1 = 2 g − 2 κ 0 , 1 , 0 κ − 1 , 0 , 1 + 2 g − 2 d κ 0 , 0 , 1 − ( 1 ≤ j ≤ g ∑ b j 1 b j 2 + b j 2 b j + g 1 ) .
κ − 1 , 0 , 2 = κ − 1 , 0 , 1 κ 0 , 0 , 1 g − 1 − 2 ( ∑ 1 ≤ k ≤ g b k 2 b k + g 2 ) . \kappa_{-1,0,2}=\frac{\kappa_{-1,0,1}\kappa_{0,0,1}}{g-1}-2\left(\sum_{1\le k\le g}b_k^2b_{k+g}^2\right). κ − 1 , 0 , 2 = g − 1 κ − 1 , 0 , 1 κ 0 , 0 , 1 − 2 1 ≤ k ≤ g ∑ b k 2 b k + g 2 .
Thus, we get the claimed expressions for x , y , z x,y,z x , y , z based on definitions in Corollary 6.3.1. □
Theorem 6.4. Conjecture A holds for all g ≥ 2 g\ge2 g ≥ 2 and d d d odd.
Proof. By Remark 5.5 and SFP for V e c H ( 2 , d , g ) → H g \mathcal{V}ec^{\mathcal{H}}(2,d,g)\to\mathcal{H}_g V e c H ( 2 , d , g ) → H g , we have the embedding
A ∗ ( V e c H ( 2 , d , g ) ) ↪ H 2 ∗ ( Bun 2 , d ( C ) , Q ) Sp ( V ) . A^*(\mathcal{V}ec^{\mathcal{H}}(2,d,g))\hookrightarrow H^{2*}(\operatorname{Bun}_{2,d}(C),\mathbb{Q})^{\operatorname{Sp}(V)}. A ∗ ( V e c H ( 2 , d , g )) ↪ H 2 ∗ ( Bun 2 , d ( C ) , Q ) Sp ( V ) .
so we only need to show that the above map is surjective. By the results of [1 ] (see also [19 Theorem 4.7] ), we have
H 2 ∗ ( Bun 2 , d ( C ) , Q ) = Q [ a 1 , a 2 , f 2 ] ⊗ ( ⋀ ∙ ( V ⊕ V ) ) H^{2*}(\operatorname{Bun}_{2,d}(C),\mathbb{Q})=\mathbb{Q}[a_1,a_2,f_2]\otimes\left(\bigwedge\nolimits^\bullet(V\oplus V)\right) H 2 ∗ ( Bun 2 , d ( C ) , Q ) = Q [ a 1 , a 2 , f 2 ] ⊗ ( ⋀ ∙ ( V ⊕ V ) )
as Sp ( V ) \operatorname{Sp}(V) Sp ( V ) representations where Sp ( V ) \operatorname{Sp}(V) Sp ( V ) acts trivially on Q [ a 1 , a 2 , f 2 ] \mathbb{Q}[a_1,a_2,f_2] Q [ a 1 , a 2 , f 2 ] . By a similar argument as in Theorem 6.3, we have
H 2 ∗ ( Bun 2 , d ( C ) , Q ) Sp ( V ) = Q [ a 1 , a 2 , f 2 ] ⊗ Q [ x , y , z ] / I . H^{2*}(\operatorname{Bun}_{2,d}(C),\mathbb{Q})^{\operatorname{Sp}(V)}=\mathbb{Q}[a_1,a_2,f_2]\otimes\mathbb{Q}[x,y,z]/I. H 2 ∗ ( Bun 2 , d ( C ) , Q ) Sp ( V ) = Q [ a 1 , a 2 , f 2 ] ⊗ Q [ x , y , z ] / I .
The result now follows from the computations in Theorem 5.4 and Corollary 6.3.2. In fact this shows
A ∗ ( V e c H ( 2 , d , g ) ) = R ∗ ( V e c H ( 2 , d , g ) ) ≃ H k ∗ ( Bun 2 , d ( C ) , Q ) = H 2 ∗ ( Bun 2 , d ( C ) , Q ) Sp ( V ) . A^*(\mathcal{V}ec^{\mathcal{H}}(2,d,g))=R^*(\mathcal{V}ec^{\mathcal{H}}(2,d,g))\simeq H_k^*(\operatorname{Bun}_{2,d}(C),\mathbb{Q})=H^{2*}(\operatorname{Bun}_{2,d}(C),\mathbb{Q})^{\operatorname{Sp}(V)}. A ∗ ( V e c H ( 2 , d , g )) = R ∗ ( V e c H ( 2 , d , g )) ≃ H k ∗ ( Bun 2 , d ( C ) , Q ) = H 2 ∗ ( Bun 2 , d ( C ) , Q ) Sp ( V ) .
We note that this also gives a proof Theorem 0.2. □
The rest of this section is devoted to computing Hilbert series of the various stacks we consider.
Slope-unstable locus
Theorem 6.5. For d d d odd, the Hilbert series of A ∗ ( ∐ n ≥ ⌊ d / 2 ⌋ + 1 Filt ( 1 , 2 ) ( n , d ) ( g ) ) A^*\left(\coprod_{n\ge\lfloor d/2\rfloor+1}\operatorname{Filt}_{(1,2)}^{(n,d)}(g)\right) A ∗ ( ∐ n ≥ ⌊ d /2 ⌋ + 1 Filt ( 1 , 2 ) ( n , d ) ( g ) ) is given by
( g + 2 2 ) T g + ∑ 1 ≤ k ≤ g ( g − k + 2 2 ) ( T g − k + T g + k ) ( 1 − T ) 2 ( 1 − T 2 ) . \frac{\binom{g+2}{2}T^g+\sum_{1\le k\le g}\left(\frac{g-k+2}{2}\right)(T^{g-k}+T^{g+k})}{(1-T)^2(1-T^2)}. ( 1 − T ) 2 ( 1 − T 2 ) ( 2 g + 2 ) T g + ∑ 1 ≤ k ≤ g ( 2 g − k + 2 ) ( T g − k + T g + k ) . Moreover, this is the slope-unstable locus in V e c H ( 2 , d , g ) \mathcal{V}ec^{\mathcal{H}}(2,d,g) V e c H ( 2 , d , g ) and has codimension g g g .
Proof. As shown in Theorem 5.2, the slope-unstable locus in V e c H ( 2 , d , g ) \mathcal{V}ec^{\mathcal{H}}(2,d,g) V e c H ( 2 , d , g ) is given by ∐ n ≥ ⌊ d / 2 ⌋ + 1 Filt ( 1 , 2 ) ( n , d ) ( g ) \coprod_{n\ge\lfloor d/2\rfloor+1}\operatorname{Filt}_{(1,2)}^{(n,d)}(g) ∐ n ≥ ⌊ d /2 ⌋ + 1 Filt ( 1 , 2 ) ( n , d ) ( g ) and the HN-stratification induces injective pushforwards on Chow groups. Therefore, it is enough to compute the Hilbert series of A ∗ ( J 2 , g n × H g J 2 , g d − n ) A^*(\mathcal{J}_{2,g}^{n}\times_{\mathcal{H}_g}\mathcal{J}_{2,g}^{d-n}) A ∗ ( J 2 , g n × H g J 2 , g d − n ) for all n ≥ ⌊ d / 2 ⌋ + 1 n\ge\lfloor d/2\rfloor+1 n ≥ ⌊ d /2 ⌋ + 1 by [14 Corollary 3.2] . Applying Poincaré duality and algebraic independence in degree at most g g g in Theorem 6.3 , the Hilbert series of A ∗ ( J 2 , g n × H g J 2 , g d − n ) A^*(\mathcal{J}_{2,g}^n \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d-n}) A ∗ ( J 2 , g n × H g J 2 , g d − n ) is given by
( g + 2 2 ) T g + ∑ 1 ≤ k ≤ g ( g − k + 2 2 ) ( T g − k + T g + k ) . \binom{g+2}{2}T^g+\sum_{1\leq k\leq g}\binom{g-k+2}{2}\left(T^{g-k}+T^{g+k}\right). ( 2 g + 2 ) T g + 1 ≤ k ≤ g ∑ ( 2 g − k + 2 ) ( T g − k + T g + k ) .
Since J 2 , g n × H g J 2 , g d − n \mathcal{J}_{2,g}^n \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d-n} J 2 , g n × H g J 2 , g d − n can be viewed as a tower of torsion-G m \mathbb{G}_m G m -gerbes over J 2 , g n × H g J 2 , g d − n \mathcal{J}_{2,g}^n \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d-n} J 2 , g n × H g J 2 , g d − n , the Hilbert series of A ∗ ( J 2 , g n × H g J 2 , g d − n ) A^*(\mathcal{J}_{2,g}^n \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d-n}) A ∗ ( J 2 , g n × H g J 2 , g d − n ) is given by
( g + 2 2 ) T g + ∑ 1 ≤ k ≤ g ( g − k + 2 2 ) ( T g − k + T g + k ) ( 1 − T ) 2 . \frac{\binom{g+2}{2}T^g+\sum_{1\leq k\leq g}\binom{g-k+2}{2}\left(T^{g-k}+T^{g+k}\right)}{(1-T)^2}. ( 1 − T ) 2 ( 2 g + 2 ) T g + ∑ 1 ≤ k ≤ g ( 2 g − k + 2 ) ( T g − k + T g + k ) .
The claim now follows from the codimension of the filtration strata Filt ( 1 , 2 ) ( n , d ) ( g ) \operatorname{Filt}_{(1,2)}^{(n,d)}(g) Filt ( 1 , 2 ) ( n , d ) ( g ) in V e c H ( 2 , d , g ) \mathcal{V}\!ec^{\mathcal{H}}(2,d,g) V e c H ( 2 , d , g ) being g − 1 + 2 n − d g-1+2n-d g − 1 + 2 n − d for all n ≥ ⌊ d / 2 ⌋ + 1 n\geq\lfloor d/2\rfloor+1 n ≥ ⌊ d /2 ⌋ + 1 . □ \square □
κ \kappa κ -cohomology
We compute the Hilbert series of H κ ∗ ( Bun 2 , d C ) H^*_\kappa(\operatorname{Bun}_{2,d}C) H κ ∗ ( Bun 2 , d C ) where C C C is a curve of genus g g g .
Theorem 6.6. For the graded exterior algebra
⋀ 1 ≤ j ≤ 2 g 1 ≤ i ≤ 2 [ b j i ] \bigwedge_{\substack{1\leq j\leq2g\\1\leq i\leq2}} [b_j^i] 1 ≤ j ≤ 2 g 1 ≤ i ≤ 2 ⋀ [ b j i ] with deg ( b j i ) = 2 i − 1 \deg(b_j^i)=2i-1 deg ( b j i ) = 2 i − 1 for all 1 ≤ i ≤ 2 1\leq i\leq2 1 ≤ i ≤ 2 , 1 ≤ j ≤ 2 g 1\leq j\leq2g 1 ≤ j ≤ 2 g , the Hilbert series of the subalgebra S S S generated by
{ x : = ∑ 1 ≤ i ≤ g b i 1 b i + g 1 , y : = ∑ 1 ≤ j ≤ g b j 2 b j + g 2 , z : = ∑ 1 ≤ k ≤ g b k 1 b k + g 2 + b k 2 b k + g 1 } \left\{x:=\sum_{1\leq i\leq g}b_i^1b_{i+g}^1,\ y:=\sum_{1\leq j\leq g}b_j^2b_{j+g}^2,\ z:=\sum_{1\leq k\leq g}b_k^1b_{k+g}^2+b_k^2b_{k+g}^1\right\} ⎩ ⎨ ⎧ x := 1 ≤ i ≤ g ∑ b i 1 b i + g 1 , y := 1 ≤ j ≤ g ∑ b j 2 b j + g 2 , z := 1 ≤ k ≤ g ∑ b k 1 b k + g 2 + b k 2 b k + g 1 ⎭ ⎬ ⎫ is given by
( 1 − T g + 1 ) ( 1 − T g + 2 ) ( 1 − T 2 g + 3 ) ( 1 − T ) ( 1 − T 2 ) ( 1 − T 3 ) \frac{(1-T^{g+1})(1-T^{g+2})(1-T^{2g+3})}{(1-T)(1-T^2)(1-T^3)} ( 1 − T ) ( 1 − T 2 ) ( 1 − T 3 ) ( 1 − T g + 1 ) ( 1 − T g + 2 ) ( 1 − T 2 g + 3 ) where we half the degree of each monomial in x , y , z x,y,z x , y , z .
Proof. From the proof of Theorem 6.3 , we already know that S = Q [ x , y , z ] ⊂ ⋀ 1 ≤ j ≤ 2 g 1 ≤ i ≤ 2 [ b j i ] S=\mathbb{Q}[x,y,z]\subset\bigwedge_{\substack{1\leq j\leq2g\\1\leq i\leq2}}[b_j^i] S = Q [ x , y , z ] ⊂ ⋀ 1 ≤ j ≤ 2 g 1 ≤ i ≤ 2 [ b j i ] has a basis generated by { x a y b z c : 0 ≤ c ≤ g , 0 ≤ a , b ≤ g − c } \{x^ay^bz^c:0\leq c\leq g,\ 0\leq a,b\leq g-c\} { x a y b z c : 0 ≤ c ≤ g , 0 ≤ a , b ≤ g − c } because the ring is the same and only the grading is different. So we have a homogeneous basis of the ring. Thus we can explicitly compute
hilbser ( S ) = ∑ c = 0 g ∑ a , b = 0 g − c T a + 3 b + 2 c = ( 1 − T g + 1 ) ( 1 − T g + 2 ) ( 1 − T 2 g + 3 ) ( 1 − T ) ( 1 − T 2 ) ( 1 − T 3 ) \begin{aligned}
\operatorname{hilbser}(S)=\sum_{c=0}^{g}\sum_{a,b=0}^{g-c}T^{a+3b+2c} \\
&=\frac{(1-T^{g+1})(1-T^{g+2})(1-T^{2g+3})}{(1-T)(1-T^2)(1-T^3)}
\end{aligned} hilbser ( S ) = c = 0 ∑ g a , b = 0 ∑ g − c T a + 3 b + 2 c = ( 1 − T ) ( 1 − T 2 ) ( 1 − T 3 ) ( 1 − T g + 1 ) ( 1 − T g + 2 ) ( 1 − T 2 g + 3 )
where the second equality is a nice exercise in combinatorics. Another way to see it would be to check that both satisfy the recursion Q g ( T ) = T 2 Q g − 1 ( T ) + ( 1 − T g + 1 ) ( 1 − T 3 g + 3 ) ( 1 − T ) ( 1 − T 3 ) Q_g(T)=T^2Q_{g-1}(T)+\frac{(1-T^{g+1})(1-T^{3g+3})}{(1-T)(1-T^3)} Q g ( T ) = T 2 Q g − 1 ( T ) + ( 1 − T ) ( 1 − T 3 ) ( 1 − T g + 1 ) ( 1 − T 3 g + 3 ) with Q 0 ( T ) = 1 Q_0(T)=1 Q 0 ( T ) = 1 . □ \square □
Theorem 6.7. The Hilbert series of H κ ∗ ( Bun 2 , d ( C ) , Q ) H^*_\kappa(\operatorname{Bun}_{2,d}(C),\mathbb{Q}) H κ ∗ ( Bun 2 , d ( C ) , Q ) , under half of the grading induced by H ∗ ( Bun 2 , d ( C ) , Q ) H^*(\operatorname{Bun}_{2,d}(C),\mathbb{Q}) H ∗ ( Bun 2 , d ( C ) , Q ) , is given by
( 1 − T g + 1 ) ( 1 − T g + 2 ) ( 1 − T 2 g + 3 ) ( 1 − T ) 3 ( 1 − T 2 ) 2 ( 1 − T 3 ) . \frac{(1-T^{g+1})(1-T^{g+2})(1-T^{2g+3})}{(1-T)^3(1-T^2)^2(1-T^3)}. ( 1 − T ) 3 ( 1 − T 2 ) 2 ( 1 − T 3 ) ( 1 − T g + 1 ) ( 1 − T g + 2 ) ( 1 − T 2 g + 3 ) .
Proof. The generators of the ring H κ ∗ ( Bun 2 , d C ) H^*_\kappa(\operatorname{Bun}_{2,d}C) H κ ∗ ( Bun 2 , d C ) were found in Theorem 5.4 . Following Corollary 6.3.1 , we have
H κ ∗ ( Bun 2 , d C ) ≃ Q [ κ 0 , 1 , 0 , κ 0 , 0 , 1 , κ − 1 , 0 , 1 ] ⊗ Q S . H^*_\kappa(\operatorname{Bun}_{2,d}C)\simeq\mathbb{Q}[\kappa_{0,1,0},\kappa_{0,0,1},\kappa_{-1,0,1}]\otimes_{\mathbb{Q}}S. H κ ∗ ( Bun 2 , d C ) ≃ Q [ κ 0 , 1 , 0 , κ 0 , 0 , 1 , κ − 1 , 0 , 1 ] ⊗ Q S .
The degrees of κ 0 , 1 , 0 , κ 0 , 0 , 1 , κ − 1 , 0 , 1 \kappa_{0,1,0},\kappa_{0,0,1},\kappa_{-1,0,1} κ 0 , 1 , 0 , κ 0 , 0 , 1 , κ − 1 , 0 , 1 are 1 , 2 , 1 1,2,1 1 , 2 , 1 respectively. Therefore, the Hilbert series of H κ ∗ ( Bun 2 , d C ) H^*_\kappa(\operatorname{Bun}_{2,d}C) H κ ∗ ( Bun 2 , d C ) is given by
hilb s e r ( H κ ∗ ( Bun 2 , d C ) ) = hilb s e r ( S ) ( 1 − T ) 2 ( 1 − T 2 ) = ( 1 − T g + 1 ) ( 1 − T g + 2 ) ( 1 − T 2 g + 3 ) ( 1 − T ) 3 ( 1 − T 2 ) 2 ( 1 − T 3 ) . \operatorname{hilb}_{\mathrm{ser}}(H^*_\kappa(\operatorname{Bun}_{2,d}C))=\frac{\operatorname{hilb}_{\mathrm{ser}}(S)}{(1-T)^2(1-T^2)}=\frac{(1-T^{g+1})(1-T^{g+2})(1-T^{2g+3})}{(1-T)^3(1-T^2)^2(1-T^3)}. hilb ser ( H κ ∗ ( Bun 2 , d C )) = ( 1 − T ) 2 ( 1 − T 2 ) hilb ser ( S ) = ( 1 − T ) 3 ( 1 − T 2 ) 2 ( 1 − T 3 ) ( 1 − T g + 1 ) ( 1 − T g + 2 ) ( 1 − T 2 g + 3 ) .
:::
Slope-stable bundles
We compute the Hilbert series of U H ( 2 , d , g ) \mathcal{U}^{H}(2,d,g) U H ( 2 , d , g ) for d d d odd.
Definition 6.4.1. Let P g , n ( T ) P_{g,n}(T) P g , n ( T ) be the Hilbert series of C 2 , g ( n ) × H g J 2 , g d C_{2,g}^{(n)}\times_{\mathcal{H}_g}J_{2,g}^{d} C 2 , g ( n ) × H g J 2 , g d for all n ≥ 0 n\geq0 n ≥ 0 where C 2 , g ( n ) : = [ C 2 , g n / S n ] C_{2,g}^{(n)}:=[C_{2,g}^{n}/S_n] C 2 , g ( n ) := [ C 2 , g n / S n ] .
Lemma 6.8. The Hilbert series of U H ( 2 , d , g ) \mathcal{U}^{H}(2,d,g) U H ( 2 , d , g ) is given by
P g , g − 1 ( T ) T g − 1 + ∑ i = 0 g − 2 P g , i ( T ) ( T i + T 3 g − 3 − 2 i ) 1 − T . \frac{P_{g,g-1}(T)T^{g-1}+\sum_{i=0}^{g-2}P_{g,i}(T)(T^i+T^{3g-3-2i})}{1-T}. 1 − T P g , g − 1 ( T ) T g − 1 + ∑ i = 0 g − 2 P g , i ( T ) ( T i + T 3 g − 3 − 2 i ) .
Proof. The map U H ( 2 , d , g ) → U H ( 2 , d , g ) \mathcal{U}^{H}(2,d,g)\to\mathcal{U}^{H}(2,d,g) U H ( 2 , d , g ) → U H ( 2 , d , g ) is a torsion-G m \mathbb{G}_m G m -gerbe by [13 Theorem B] . Applying Theorem 1.1 , we have
hilb s e r ( U H ( 2 , d , g ) ) = hilb s e r ( U H ( 2 , d , g ) ) 1 − T . \operatorname{hilb}_{\mathrm{ser}}(\mathcal{U}^{H}(2,d,g))=\frac{\operatorname{hilb}_{\mathrm{ser}}(\mathcal{U}^{H}(2,d,g))}{1-T}. hilb ser ( U H ( 2 , d , g )) = 1 − T hilb ser ( U H ( 2 , d , g )) .
Plugging the motivic decomposition of U H ( 2 , d , g ) \mathcal{U}^{H}(2,d,g) U H ( 2 , d , g ) in Theorem 2.6 into this expression gives the claimed equality.
We now compute P g , n ( T ) P_{g,n}(T) P g , n ( T ) for 0 ≤ n < g 0\leq n<g 0 ≤ n < g .
Notation 1. The Hilbert series P g , n ( T ) P_{g,n}(T) P g , n ( T ) is independent of the value of d d d by an adaptation of [6 Proposition 0.1] . We shall refer to P g , n ( T ) P_{g,n}(T) P g , n ( T ) as the Hilbert series of C 2 , g ( n ) × H g J 2 , g d C_{2,g}^{(n)}\times_{\mathcal{H}_g}J_{2,g}^{d} C 2 , g ( n ) × H g J 2 , g d from here on.
Theorem 6.9. For 0 ≤ n ≤ g + 6 0\leq n\leq g+6 0 ≤ n ≤ g + 6 , the ring A ∗ ( C 2 , g ( n ) × H g J 2 , g g ) A^*(C_{2,g}^{(n)}\times_{\mathcal{H}_g}J_{2,g}^{g}) A ∗ ( C 2 , g ( n ) × H g J 2 , g g ) is of degree n + g n+g n + g and satisfies Poincaré duality.
Proof. First of all let V = H 1 ( C , Q ) V=H^1(C,\mathbb{Q}) V = H 1 ( C , Q ) which is a symplectic vector space. Then
H ∗ ( C n × J C , Q ) = ( ⨂ i = 1 n ( Q ⊕ V ⊕ Q η i ) ) ⊗ ( ⋀ V ) H^*(C^n\times J_C,\mathbb{Q})=\left(\bigotimes_{i=1}^{n}(\mathbb{Q}\oplus V\oplus\mathbb{Q}\eta_i)\right)\otimes\left(\bigwedge V\right) H ∗ ( C n × J C , Q ) = ( i = 1 ⨂ n ( Q ⊕ V ⊕ Q η i ) ) ⊗ ( ⋀ V )
as Sp ( V ) \operatorname{Sp}(V) Sp ( V ) -modules where Sp ( V ) \operatorname{Sp}(V) Sp ( V ) acts trivially on Q \mathbb{Q} Q and Q η i \mathbb{Q}\eta_i Q η i . It follows that we have a surjective Sp ( V ) \operatorname{Sp}(V) Sp ( V ) -equivariant ring homomorphism
⋀ ∙ ( V 1 ⊕ ⋯ ⊕ V n ⊕ V J C ) ↠ H ∗ ( C n × J C , Q ) \bigwedge^\bullet(V_1 \oplus\cdots\oplus V_n \oplus V_{J_C}) \twoheadrightarrow H^*(C^n \times J_C,\mathbb{Q}) ⋀ ∙ ( V 1 ⊕ ⋯ ⊕ V n ⊕ V J C ) ↠ H ∗ ( C n × J C , Q )
and since Sp ( V ) \operatorname{Sp}(V) Sp ( V ) is reductive, this induces a surjection onto their invariant spaces
[ ⋀ ∙ ( V 1 ⊕ ⋯ ⊕ V n ⊕ V J C ) ] Sp ( V ) ↠ H ∗ ( C n × J C , Q ) Sp ( V ) . \left[\bigwedge^\bullet(V_1 \oplus\cdots\oplus V_n \oplus V_{J_C})\right]^{\operatorname{Sp}(V)} \twoheadrightarrow H^*(C^n \times J_C,\mathbb{Q})^{\operatorname{Sp}(V)}. [ ⋀ ∙ ( V 1 ⊕ ⋯ ⊕ V n ⊕ V J C ) ] Sp ( V ) ↠ H ∗ ( C n × J C , Q ) Sp ( V ) .
It follows from the standard representation theory of Sp ( V ) \operatorname{Sp}(V) Sp ( V ) (see as before [36 ] or [20 ] ) that H ∗ ( C n × J C , Q ) Sp ( V ) H^*(C^n \times J_C,\mathbb{Q})^{\operatorname{Sp}(V)} H ∗ ( C n × J C , Q ) Sp ( V ) is generated as a ring by { ψ i , Δ i j , ξ i , g , p J ∗ Θ g } i ≠ j \{\psi_i,\Delta_{ij},\xi_{i,g},p_J^*\Theta_g\}_{i\ne j} { ψ i , Δ ij , ξ i , g , p J ∗ Θ g } i = j . So we have established using SFP (see Theorem 4.4 and Remark 5.5) that as graded rings
A ∗ ( C 2 , g n × H g J 2 , g g ) ≃ H 2 ∗ ( C n × J C , Q ) Sp ( V ) . A^*(C_{2,g}^n \mathbin{\times_{\mathcal{H}_g}} \mathcal{J}_{2,g}^g) \simeq H^{2*}(C^n \times J_C,\mathbb{Q})^{\operatorname{Sp}(V)}. A ∗ ( C 2 , g n × H g J 2 , g g ) ≃ H 2 ∗ ( C n × J C , Q ) Sp ( V ) .
Further taking S n S_n S n -invariants, we get
A ∗ ( C 2 , g ( n ) × H g J 2 , g g ) ≃ H 2 ∗ ( C ( n ) × J C , Q ) Sp ( V ) . A^*(C_{2,g}^{(n)} \mathbin{\times_{\mathcal{H}_g}} \mathcal{J}_{2,g}^g) \simeq H^{2*}(C^{(n)} \times J_C,\mathbb{Q})^{\operatorname{Sp}(V)}. A ∗ ( C 2 , g ( n ) × H g J 2 , g g ) ≃ H 2 ∗ ( C ( n ) × J C , Q ) Sp ( V ) .
But the Poincaré pairing on H ∗ ( C ( n ) × J C , Q ) H^*(C^{(n)} \times J_C,\mathbb{Q}) H ∗ ( C ( n ) × J C , Q ) is non-degenerate and Sp ( V ) \operatorname{Sp}(V) Sp ( V ) -invariant. Since Sp ( V ) \operatorname{Sp}(V) Sp ( V ) is reductive, it follows that the pairing on H ∗ ( C ( n ) × J C , Q ) Sp ( V ) H^*(C^{(n)} \times J_C,\mathbb{Q})^{\operatorname{Sp}(V)} H ∗ ( C ( n ) × J C , Q ) Sp ( V ) is also non-degenerate and hence Poincaré duality is satisfied. This completes the proof. □
Theorem 6.10. Let ξ i , g \xi_{i,g} ξ i , g be the pullback of ξ g \xi_g ξ g over the projection onto the i i i -th factor C 2 , g n × H g J 2 , g g → C 2 , g × H g J 2 , g g C_{2,g}^n \mathbin{\times_{\mathcal{H}_g}} \mathcal{J}_{2,g}^g \to C_{2,g} \mathbin{\times_{\mathcal{H}_g}} \mathcal{J}_{2,g}^g C 2 , g n × H g J 2 , g g → C 2 , g × H g J 2 , g g . For 1 ≤ n ≤ g + 6 1 \le n \le g+6 1 ≤ n ≤ g + 6 , the ring A ∗ ( C 2 , g n × H g J 2 , g g ) A^*(C_{2,g}^n \mathbin{\times_{\mathcal{H}_g}} \mathcal{J}_{2,g}^g) A ∗ ( C 2 , g n × H g J 2 , g g ) is multiplicatively generated by { ξ i , g , ψ i , p J ∗ Θ d , Δ i j } 1 ≤ i ≠ j ≤ n \{\xi_{i,g},\psi_i,p_J^*\Theta_d,\Delta_{ij}\}_{1\le i\ne j\le n} { ξ i , g , ψ i , p J ∗ Θ d , Δ ij } 1 ≤ i = j ≤ n .
Proof. From the proof of Theorem 6.9 , we get
A ∗ ( C 2 , g n × H g J 2 , g g ) ≃ H 2 ∗ ( C n × J C , Q ) Sp ( V ) . A^*(C_{2,g}^n \mathbin{\times_{\mathcal{H}_g}} \mathcal{J}_{2,g}^g) \simeq H^{2*}(C^n \times J_C,\mathbb{Q})^{\operatorname{Sp}(V)}. A ∗ ( C 2 , g n × H g J 2 , g g ) ≃ H 2 ∗ ( C n × J C , Q ) Sp ( V ) .
The result now follows from observing that the restrictions of { ξ i , g , ψ i , p J ∗ Θ d , Δ i j } 1 ≤ i ≠ j ≤ n \{\xi_{i,g},\psi_i,p_J^*\Theta_d,\Delta_{ij}\}_{1\le i\ne j\le n} { ξ i , g , ψ i , p J ∗ Θ d , Δ ij } 1 ≤ i = j ≤ n give a set of multiplicative generators for H 2 ∗ ( C n × J C , Q ) Sp ( V ) H^{2*}(C^n \times J_C,\mathbb{Q})^{\operatorname{Sp}(V)} H 2 ∗ ( C n × J C , Q ) Sp ( V ) . □
Corollary 6.10.1. Let p J : C 2 , g × H g J 2 , g d → J 2 , g d p_J:C_{2,g}\mathbin{\times_{\mathcal{H}_g}}\mathcal{J}_{2,g}^d\to\mathcal{J}_{2,g}^d p J : C 2 , g × H g J 2 , g d → J 2 , g d be the projection map. We have
A ∗ ( C 2 , g × H g J 2 , g d ) ≃ Q [ a , b , c , d ] / I A^*(C_{2,g}\mathbin{\times_{\mathcal{H}_g}}\mathcal{J}_{2,g}^d)\simeq\mathbb{Q}[a,b,c,d]/I A ∗ ( C 2 , g × H g J 2 , g d ) ≃ Q [ a , b , c , d ] / I where a : = p C ∗ K a:=p_C^*\mathcal{K} a := p C ∗ K , b : = Ξ d b:=\Xi_d b := Ξ d , c : = p J ∗ p J ∗ ( Ξ d 2 ) c:=p_J^*p_{J*}(\Xi_d^2) c := p J ∗ p J ∗ ( Ξ d 2 ) , d : = p J ∗ p J ∗ ( Ξ d p C ∗ K ) d:=p_J^*p_{J*}(\Xi_d p_C^*\mathcal{K}) d := p J ∗ p J ∗ ( Ξ d p C ∗ K ) and the ideal I I I is generated by p C ∗ K 2 p_C^*\mathcal{K}^2 p C ∗ K 2 , ξ d p C ∗ K \xi_d p_C^*\mathcal{K} ξ d p C ∗ K , ξ d 2 + p C ∗ K p J ∗ Θ d g − 1 \xi_d^2+\frac{p_C^*\mathcal{K}p_J^*\Theta_d}{g-1} ξ d 2 + g − 1 p C ∗ K p J ∗ Θ d , p J ∗ Θ d g + 1 p_J^*\Theta_d^{g+1} p J ∗ Θ d g + 1 , ( ξ d − d 2 g − 2 p C ∗ K ) p J ∗ Θ d (\xi_d-\frac{d}{2g-2}p_C^*\mathcal{K})p_J^*\Theta_d ( ξ d − 2 g − 2 d p C ∗ K ) p J ∗ Θ d .
Proof. Applying Theorem 1.1 , and [24 Theorem 1.1] or Theorem 6.1 , we have that A ∗ ( C 2 , g × H g J 2 , g d ) A^*(C_{2,g}\mathbin{\times_{\mathcal{H}_g}}\mathcal{J}_{2,g}^d) A ∗ ( C 2 , g × H g J 2 , g d ) is multiplicatively generated by p C ∗ K , ξ d , p J ∗ κ − 1 , 2 , p J ∗ κ 0 , 1 p_C^*\mathcal{K},\xi_d,p_J^*\kappa_{-1,2},p_J^*\kappa_{0,1} p C ∗ K , ξ d , p J ∗ κ − 1 , 2 , p J ∗ κ 0 , 1 and the claimed ideal of relations. □
Theorem 6.11. For 0 ≤ n ≤ g + 6 0\le n\le g+6 0 ≤ n ≤ g + 6 and 0 ≤ i ≤ g 0\le i\le g 0 ≤ i ≤ g , we have
dim A i ( C 2 , g ( n ) × H g J 2 , g g ) = # { ( a , b , c ) ∈ Z ≥ 0 3 ∣ a + b + c ≤ i , 2 a + b + c ≤ n } . \dim A^i(C_{2,g}^{(n)}\mathbin{\times_{\mathcal{H}_g}}\mathcal{J}_{2,g}^g)=\#\{(a,b,c)\in\mathbb{Z}_{\ge0}^3\mid a+b+c\le i,\ 2a+b+c\le n\}. dim A i ( C 2 , g ( n ) × H g J 2 , g g ) = # {( a , b , c ) ∈ Z ≥ 0 3 ∣ a + b + c ≤ i , 2 a + b + c ≤ n } .
Proof. From the proof of Theorem 6.9 , it’s enough to analyze H ∗ ( C ( n ) × J C , Q ) Sp ( V ) H^*(C^{(n)} \times J_C,\mathbb{Q})^{\operatorname{Sp}(V)} H ∗ ( C ( n ) × J C , Q ) Sp ( V ) . We know H ∗ ( C ( n ) , Q ) = ⨁ r + b ≤ n ⋀ r V H^*(C^{(n)},\mathbb{Q}) = \bigoplus_{r+b\le n} \bigwedge^r V H ∗ ( C ( n ) , Q ) = ⨁ r + b ≤ n ⋀ r V as graded Sp ( V ) \operatorname{Sp}(V) Sp ( V ) -representations with ⋀ r V \bigwedge^r V ⋀ r V in degree r + 2 b r+2b r + 2 b . So we get
dim Q A i ( C 2 , g ( n ) × H g J 2 , g g ) = ∑ r + b ≤ n r + s + 2 b = 2 i dim Q ( ⋀ r V ⊗ ⋀ s V ) Sp ( V ) . \dim_{\mathbb{Q}} A^i(C^{(n)}_{2,g} \times_{\mathcal{H}_g} \mathcal{J}^g_{2,g}) =
\sum_{\substack{r+b\le n\\r+s+2b=2i}}
\dim_{\mathbb{Q}}\left(\bigwedge^r V \otimes\bigwedge^s V\right)^{\operatorname{Sp}(V)}. dim Q A i ( C 2 , g ( n ) × H g J 2 , g g ) = r + b ≤ n r + s + 2 b = 2 i ∑ dim Q ( ⋀ r V ⊗ ⋀ s V ) Sp ( V ) .
From [18 Theorem 5.5.15, Corollary 5.5.16] , we get
⋀ r V = ⨁ c + 2 a = r 0 ≤ c ≤ g 0 ≤ a ≤ g − c P c \bigwedge^r V =
\bigoplus_{\substack{c+2a=r\\0\le c\le g\\0\le a\le g-c}} P^c ⋀ r V = c + 2 a = r 0 ≤ c ≤ g 0 ≤ a ≤ g − c ⨁ P c
where P c P^c P c is the irreducible primitive representation of highest weight ω c \omega_c ω c and dim P c = ( 2 g c ) − ( 2 g c − 2 ) \dim P^c = \binom{2g}{c} - \binom{2g}{c-2} dim P c = ( c 2 g ) − ( c − 2 2 g ) . Moreover these P c P^c P c ’s are self dual and pairwise non-isomorphic. So when we write
⋀ r V ⊗ ⋀ s V = ⨁ c + 2 a = r 0 ≤ c ≤ g 0 ≤ a ≤ g − c ⨁ d + 2 b = s 0 ≤ d ≤ g 0 ≤ b ≤ g − d P c ⊗ P d , \bigwedge^r V \otimes\bigwedge^s V =
\bigoplus_{\substack{c+2a=r\\0\le c\le g\\0\le a\le g-c}}
\bigoplus_{\substack{d+2b=s\\0\le d\le g\\0\le b\le g-d}}
P^c \otimes P^d, ⋀ r V ⊗ ⋀ s V = c + 2 a = r 0 ≤ c ≤ g 0 ≤ a ≤ g − c ⨁ d + 2 b = s 0 ≤ d ≤ g 0 ≤ b ≤ g − d ⨁ P c ⊗ P d ,
we get
dim Q ( ⋀ r V ⊗ ⋀ s V ) Sp ( V ) = # { c ∣ 0 ≤ c ≤ min { r , s , 2 g − r , 2 g − s } , c ≡ r ≡ s ( m o d 2 ) } . \dim_{\mathbb{Q}}\left(\bigwedge^r V \otimes\bigwedge^s V\right)^{\operatorname{Sp}(V)}
= \#\{c \mid0 \le c \le\min\{r,s,2g-r,2g-s\},\ c \equiv r \equiv s \pmod{2}\}. dim Q ( ⋀ r V ⊗ ⋀ s V ) Sp ( V ) = # { c ∣ 0 ≤ c ≤ min { r , s , 2 g − r , 2 g − s } , c ≡ r ≡ s ( mod 2 )} .
Using the above result we get after a straight-forward but tedious computation that
dim Q A i ( C 2 , g ( n ) × H g J 2 , g g ) = ∑ r + b ≤ n r + s + 2 b = 2 i ∑ c + 2 a = r c + 2 e = s 0 ≤ c ≤ g 0 ≤ a , e ≤ g − c 1 = ∑ a , b , c , e ≥ 0 2 a + b + c ≤ n a + b + c + e = i c ≤ g , a , e ≤ g − c 1 = ∑ a , b , c ≥ 0 2 a + b + c ≤ n a + b + c ≤ i 1 , \begin{aligned}
\dim_{\mathbb{Q}} A^i(C^{(n)}_{2,g} \times_{\mathcal{H}_g} \mathcal{J}^g_{2,g})
&= \sum_{\substack{r+b\le n\\r+s+2b=2i}}
\sum_{\substack{c+2a=r\\c+2e=s\\0\le c\le g\\0\le a,e\le g-c}} 1 \\
&= \sum_{\substack{a,b,c,e\ge0\\2a+b+c\le n\\a+b+c+e=i\\c\le g,\ a,e\le g-c}} 1 \\
&= \sum_{\substack{a,b,c\ge0\\2a+b+c\le n\\a+b+c\le i}} 1,
\end{aligned} dim Q A i ( C 2 , g ( n ) × H g J 2 , g g ) = r + b ≤ n r + s + 2 b = 2 i ∑ c + 2 a = r c + 2 e = s 0 ≤ c ≤ g 0 ≤ a , e ≤ g − c ∑ 1 = a , b , c , e ≥ 0 2 a + b + c ≤ n a + b + c + e = i c ≤ g , a , e ≤ g − c ∑ 1 = a , b , c ≥ 0 2 a + b + c ≤ n a + b + c ≤ i ∑ 1 ,
where the last equality follows from i ≤ g i \le g i ≤ g , so the sum remains the same. This completes the proof. □
We have the following corollaries.
Corollary 6.11.1. The ring A ∗ ( C 2 , g × H g J 2 , g d ) A^*(C_{2,g} \times_{\mathcal{H}_g} \mathcal{J}^d_{2,g}) A ∗ ( C 2 , g × H g J 2 , g d ) is multiplicatively generated by p C ∗ K p^*_{C}K p C ∗ K , ξ d \xi_d ξ d , p J ∗ Θ d p^*_{\mathcal{J}}\Theta_d p J ∗ Θ d . In particular, we have
A ∗ ( C 2 , g × H g J 2 , g d ) ≃ Q [ p C ∗ K , ξ d , p J ∗ Θ d ] / ⟨ p C ∗ K 2 , ξ d p C ∗ K , ξ d 2 + p C ∗ K . p J ∗ Θ d g − 1 , p J ∗ Θ d g + 1 , ( ξ d − d 2 g − 2 p C ∗ K ) p J ∗ Θ d g ⟩ . A^*(C_{2,g} \times_{\mathcal{H}_g} \mathcal{J}^d_{2,g}) \simeq\mathbb{Q}[p^*_C K,\xi_d,p^*_{\mathcal{J}}\Theta_d]/\langle p^*_C K^2,\xi_d p^*_C K,\xi_d^2+\frac{p^*_C K.p^*_{\mathcal{J}}\Theta_d}{g-1},p^*_{\mathcal{J}}\Theta_d^{g+1},(\xi_d-\frac{d}{2g-2}p^*_C K)p^*_{\mathcal{J}}\Theta_d^g\rangle. A ∗ ( C 2 , g × H g J 2 , g d ) ≃ Q [ p C ∗ K , ξ d , p J ∗ Θ d ] / ⟨ p C ∗ K 2 , ξ d p C ∗ K , ξ d 2 + g − 1 p C ∗ K . p J ∗ Θ d , p J ∗ Θ d g + 1 , ( ξ d − 2 g − 2 d p C ∗ K ) p J ∗ Θ d g ⟩ .
Corollary 6.11.2. Let p J : C 2 , g × H g J 2 , g d → J 2 , g d p_{\mathcal{J}}: C_{2,g} \times_{\mathcal{H}_g} \mathcal{J}^d_{2,g} \to\mathcal{J}^d_{2,g} p J : C 2 , g × H g J 2 , g d → J 2 , g d be the projection map. We have
A ∗ ( C 2 , g × H g J 2 , g d ) ≃ Q [ a , b , c , d ] / I A^*(C_{2,g} \times_{\mathcal{H}_g} \mathcal{J}^d_{2,g}) \simeq\mathbb{Q}[a,b,c,d]/I A ∗ ( C 2 , g × H g J 2 , g d ) ≃ Q [ a , b , c , d ] / I where a : = p C ∗ K a := p^*_C K a := p C ∗ K , b : = Ξ d b := \Xi_d b := Ξ d , c : = p J ∗ p J ∗ ( Ξ d 2 ) c := p^*_{\mathcal{J}}p_{\mathcal{J}*}(\Xi_d^2) c := p J ∗ p J ∗ ( Ξ d 2 ) , d : = p J ∗ p J ∗ ( Ξ d p C ∗ K ) d := p^*_{\mathcal{J}}p_{\mathcal{J}*}(\Xi_d p^*_C K) d := p J ∗ p J ∗ ( Ξ d p C ∗ K ) and the ideal
I is generated by p C ∗ K 2 , ξ d p C ∗ K , ξ d 2 + p C ∗ K . p J ∗ Θ d g − 1 , p J ∗ Θ d g + 1 , ( ξ d − d 2 g − 2 p C ∗ K ) p J ∗ Θ d g . I\text{ is generated by }p^*_C K^2,\xi_d p^*_C K,\xi_d^2+\frac{p^*_C K.p^*_{\mathcal{J}}\Theta_d}{g-1},p^*_{\mathcal{J}}\Theta_d^{g+1},(\xi_d-\frac{d}{2g-2}p^*_C K)p^*_{\mathcal{J}}\Theta_d^g. I is generated by p C ∗ K 2 , ξ d p C ∗ K , ξ d 2 + g − 1 p C ∗ K . p J ∗ Θ d , p J ∗ Θ d g + 1 , ( ξ d − 2 g − 2 d p C ∗ K ) p J ∗ Θ d g .
Theorem 6.12. Let h U ( T ) h_U(T) h U ( T ) be the Hilbert series of U H ( 2 , d , g ) \mathcal{U}^{H}(2,d,g) U H ( 2 , d , g ) . We have the following:
( 1 − T ) 2 ( 1 − T 2 ) 2 ( 1 − T 3 ) h U ( T ) = 1 − T g − 3 T g + 1 − 3 T g + 2 − T g + 3 + ( 2 g + 3 ) T 2 g + 1 + ( 2 g + 3 ) T 2 g + 2 (1-T)^2(1-T^2)^2(1-T^3)h_U(T)=1-T^g-3T^{g+1}-3T^{g+2}-T^{g+3}+(2g+3)T^{2g+1}+(2g+3)T^{2g+2} ( 1 − T ) 2 ( 1 − T 2 ) 2 ( 1 − T 3 ) h U ( T ) = 1 − T g − 3 T g + 1 − 3 T g + 2 − T g + 3 + ( 2 g + 3 ) T 2 g + 1 + ( 2 g + 3 ) T 2 g + 2 ( 1 − T ) 2 ( 1 − T 2 ) 2 ( 1 − T 3 ) h U ( T ) = − ( 2 g + 3 ) T 2 g + 4 − ( 2 g + 3 ) T 2 g + 5 + T 3 g + 3 + 3 T 3 g + 4 + 3 T 3 g + 5 \phantom{(1-T)^2(1-T^2)^2(1-T^3)h_U(T)={}}-(2g+3)T^{2g+4}-(2g+3)T^{2g+5}+T^{3g+3}+3T^{3g+4}+3T^{3g+5} ( 1 − T ) 2 ( 1 − T 2 ) 2 ( 1 − T 3 ) h U ( T ) = − ( 2 g + 3 ) T 2 g + 4 − ( 2 g + 3 ) T 2 g + 5 + T 3 g + 3 + 3 T 3 g + 4 + 3 T 3 g + 5 ( 1 − T ) 2 ( 1 − T 2 ) 2 ( 1 − T 3 ) h U ( T ) = + T 3 g + 6 − T 4 g + 6 . (9) \phantom{(1-T)^2(1-T^2)^2(1-T^3)h_U(T)={}}+T^{3g+6}-T^{4g+6}. \tag*{(9)} ( 1 − T ) 2 ( 1 − T 2 ) 2 ( 1 − T 3 ) h U ( T ) = + T 3 g + 6 − T 4 g + 6 . (9)
Proof. Let c i , n = # { ( a , b , c ) ∈ Z ≥ 0 3 ∣ a + b + c ≤ i , 2 a + b + c ≤ n } c_{i,n}=\#\{(a,b,c)\in\mathbb{Z}_{\geq0}^{3}\mid a+b+c\leq i,2a+b+c\leq n\} c i , n = # {( a , b , c ) ∈ Z ≥ 0 3 ∣ a + b + c ≤ i , 2 a + b + c ≤ n } . We define the formal power series F ( T , y ) = ∑ i , n ≥ 0 c i , n T i y n F(T,y)=\sum_{i,n\geq0}c_{i,n}T^iy^n F ( T , y ) = ∑ i , n ≥ 0 c i , n T i y n . By summing over the independent variables a , b , c a,b,c a , b , c and encoding the cumulative inequalities via ( 1 − T ) − 1 ( 1 − y ) − 1 (1-T)^{-1}(1-y)^{-1} ( 1 − T ) − 1 ( 1 − y ) − 1 , we obtain the exact rational function:
F ( T , y ) = 1 ( 1 − T ) ( 1 − y ) ( 1 − T y 2 ) ( 1 − T y ) 2 . F(T,y)=\frac{1}{(1-T)(1-y)(1-Ty^2)(1-Ty)^2}. F ( T , y ) = ( 1 − T ) ( 1 − y ) ( 1 − T y 2 ) ( 1 − T y ) 2 1 .
For a fixed n n n , the power series
R n ( T ) : = ∑ i = 0 ∞ c i , n T i R_n(T):=\sum_{i=0}^{\infty}c_{i,n}T^i R n ( T ) := i = 0 ∑ ∞ c i , n T i
is extracted via Cauchy’s integral formula over a small contour:
R n ( T ) = 1 2 π i ∮ γ F ( T , y ) y n + 1 d y with γ : = { y ∈ C : ∣ y ∣ = ϵ } . (10) R_n(T)=\frac{1}{2\pi i}\oint_{\gamma}\frac{F(T,y)}{y^{n+1}}\,dy\quad\text{with }\gamma:=\{y\in\mathbb{C}:|y|=\epsilon\}. \tag*{(10)} R n ( T ) = 2 π i 1 ∮ γ y n + 1 F ( T , y ) d y with γ := { y ∈ C : ∣ y ∣ = ϵ } . (10)
We now construct the palindromic polynomial, by Theorem 6.9 , P g , n ( T ) P_{g,n}(T) P g , n ( T ) directly from the rational function R n ( T ) R_n(T) R n ( T ) . It is easy to see that the sequence of coefficients stabilizes: c i , n = c n , n c_{i,n}=c_{n,n} c i , n = c n , n for all i ≥ n i\geq n i ≥ n . Consequently, the infinite series R n ( T ) R_n(T) R n ( T ) admits the exact rational representation since n < g n<g n < g :
R n ( T ) = ∑ i = 0 g c i , n T i + c n , n T g + 1 1 − T . R_n(T)=\sum_{i=0}^{g}c_{i,n}T^i+c_{n,n}\frac{T^{g+1}}{1-T}. R n ( T ) = i = 0 ∑ g c i , n T i + c n , n 1 − T T g + 1 .
The target polynomial P g , n ( T ) P_{g,n}(T) P g , n ( T ) is palindromic of degree n + g n+g n + g . Because c i , n c_{i,n} c i , n is constant for n ≤ i ≤ g n\leq i\leq g n ≤ i ≤ g , the truncation and palindromic reflection can be captured exactly by adding the reflected rational tail. Evaluating T n + g R n ( T − 1 ) T^{n+g}R_n(T^{-1}) T n + g R n ( T − 1 ) , we get:
P g , n ( T ) = R n ( T ) + T n + g R n ( T − 1 ) . P_{g,n}(T)=R_n(T)+T^{n+g}R_n(T^{-1}). P g , n ( T ) = R n ( T ) + T n + g R n ( T − 1 ) .
We now apply this exact rational identity to h U ( T ) h_U(T) h U ( T ) . Let the weight function be W n ( T ) : = T n + T 3 g − 3 − 2 n W_n(T):=T^n+T^{3g-3-2n} W n ( T ) := T n + T 3 g − 3 − 2 n for 0 ≤ n ≤ g − 2 0\leq n\leq g-2 0 ≤ n ≤ g − 2 , and W g − 1 ( T ) : = T g − 1 W_{g-1}(T):=T^{g-1} W g − 1 ( T ) := T g − 1 . We can express h U ( T ) h_U(T) h U ( T ) entirely in terms of the rational operators:
h U ( T ) = ∑ n = 0 g − 1 W n ( T ) P g , n ( T ) = ∑ n = 0 g − 1 W n ( T ) ( R n ( T ) + T n + g R n ( T − 1 ) ) . h_U(T)=\sum_{n=0}^{g-1}W_n(T)P_{g,n}(T)=\sum_{n=0}^{g-1}W_n(T)\left(R_n(T)+T^{n+g}R_n(T^{-1})\right). h U ( T ) = n = 0 ∑ g − 1 W n ( T ) P g , n ( T ) = n = 0 ∑ g − 1 W n ( T ) ( R n ( T ) + T n + g R n ( T − 1 ) ) .
By interchanging the sum with the contour integral extraction R n ( T ) R_n(T) R n ( T ) in (9), we get the following expression for h U ( T ) h_U(T) h U ( T ) :
h U ( T ) = 1 2 π i ∮ γ [ F ( T , y ) ∑ n = 0 g − 1 W n ( T ) y n + 1 + F ( T − 1 , y ) ∑ n = 0 g − 1 W n ( T ) T n + g y n + 1 ] d y . h_U(T)=\frac{1}{2\pi i}\oint_{\gamma}\left[F(T,y)\sum_{n=0}^{g-1}\frac{W_n(T)}{y^{n+1}}+F(T^{-1},y)\sum_{n=0}^{g-1}\frac{W_n(T)T^{n+g}}{y^{n+1}}\right]\,dy. h U ( T ) = 2 π i 1 ∮ γ [ F ( T , y ) n = 0 ∑ g − 1 y n + 1 W n ( T ) + F ( T − 1 , y ) n = 0 ∑ g − 1 y n + 1 W n ( T ) T n + g ] d y .
By the Global Residue Theorem, evaluating this integral reduces to summing the residues at the exterior poles of F ( T , y ) F(T,y) F ( T , y ) and F ( T − 1 , y ) F(T^{-1},y) F ( T − 1 , y ) , which are y = 1 y=1 y = 1 , y = T ± 1 y=T^{\pm1} y = T ± 1 , and y = ± T ± 1 / 2 y=\pm T^{\pm1/2} y = ± T ± 1/2 based on an arbitrary choice of the two squareroots of T T T . The integrand is ω 1 ( T , y ) + ω 2 ( T , y ) \omega_1(T,y)+\omega_2(T,y) ω 1 ( T , y ) + ω 2 ( T , y ) where:
ω 1 ( T , y ) = 1 ( 1 − T ) ( 1 − y ) ( 1 − T y 2 ) ( 1 − T y ) 2 y ∑ n = 0 g − 1 W n ( T ) y n , \omega_1(T,y)=\frac{1}{(1-T)(1-y)(1-Ty^2)(1-Ty)^2y}\sum_{n=0}^{g-1}\frac{W_n(T)}{y^n}, ω 1 ( T , y ) = ( 1 − T ) ( 1 − y ) ( 1 − T y 2 ) ( 1 − T y ) 2 y 1 n = 0 ∑ g − 1 y n W n ( T ) ,
ω 2 ( T , y ) = − T 4 ( 1 − T ) ( 1 − y ) ( T − y 2 ) ( T − y ) 2 y ∑ n = 0 g − 1 W n ( T ) T n + g y n . \omega_2(T,y)=\frac{-T^4}{(1-T)(1-y)(T-y^2)(T-y)^2y}\sum_{n=0}^{g-1}\frac{W_n(T)T^{n+g}}{y^n}. ω 2 ( T , y ) = ( 1 − T ) ( 1 − y ) ( T − y 2 ) ( T − y ) 2 y − T 4 n = 0 ∑ g − 1 y n W n ( T ) T n + g .
We calculate the explicit residues at the isolated poles of ω \omega ω :
The simple pole at y = 1 y=1 y = 1 : For y = 1 y=1 y = 1 , both ω 1 \omega_1 ω 1 and ω 2 \omega_2 ω 2 have simple poles. We get:
Res y = 1 ω = − 1 ( 1 − T ) 4 ∑ n = 0 g − 1 W n ( T ) − T 4 ( 1 − T ) 4 ∑ n = 0 g − 1 W n ( T ) T n + g . \operatorname{Res}_{y=1}\omega=-\frac{1}{(1-T)^4}\sum_{n=0}^{g-1}W_n(T)-\frac{T^4}{(1-T)^4}\sum_{n=0}^{g-1}W_n(T)T^{n+g}. Res y = 1 ω = − ( 1 − T ) 4 1 n = 0 ∑ g − 1 W n ( T ) − ( 1 − T ) 4 T 4 n = 0 ∑ g − 1 W n ( T ) T n + g .
The double pole at y = 1 / T y=1/T y = 1/ T : This is a pole of order 2 2 2 for ω 1 \omega_1 ω 1 only; ω 2 \omega_2 ω 2 is analytic at y = 1 / T y=1/T y = 1/ T , so its residue is 0 0 0 . We get:
Res y = 1 / T ω = − 3 T ( 1 − T ) 4 ∑ n = 0 g − 1 W n ( T ) T n + 1 − 1 ( 1 − T ) 3 ∑ n = 0 g − 1 W n ( T ) ( n + 1 ) T n + 2 . \operatorname{Res}_{y=1/T}\omega=-\frac{3T}{(1-T)^4}\sum_{n=0}^{g-1}W_n(T)T^{n+1}-\frac{1}{(1-T)^3}\sum_{n=0}^{g-1}W_n(T)(n+1)T^{n+2}. Res y = 1/ T ω = − ( 1 − T ) 4 3 T n = 0 ∑ g − 1 W n ( T ) T n + 1 − ( 1 − T ) 3 1 n = 0 ∑ g − 1 W n ( T ) ( n + 1 ) T n + 2 .
The double pole at y = T y=T y = T : This is a pole of order 2 2 2 for ω 2 \omega_2 ω 2 only; ω 1 \omega_1 ω 1 is analytic at y = T y=T y = T , so its residue is 0 0 0 . We get:
Res y = T ω = − 3 T 3 ( 1 − T ) 4 ∑ n = 0 g − 1 W n ( T ) T g − 1 + T 3 ( 1 − T ) 3 ∑ n = 0 g − 1 W n ( T ) ( n + 1 ) T g − 2 . \operatorname{Res}_{y=T}\omega=-\frac{3T^3}{(1-T)^4}\sum_{n=0}^{g-1}W_n(T)T^{g-1}+\frac{T^3}{(1-T)^3}\sum_{n=0}^{g-1}W_n(T)(n+1)T^{g-2}. Res y = T ω = − ( 1 − T ) 4 3 T 3 n = 0 ∑ g − 1 W n ( T ) T g − 1 + ( 1 − T ) 3 T 3 n = 0 ∑ g − 1 W n ( T ) ( n + 1 ) T g − 2 .
The simple poles at y = ± 1 / T y=\pm1/\sqrt{T} y = ± 1/ T : These are simple poles of ω 1 \omega_1 ω 1 from the factor ( 1 − T y 2 ) (1-Ty^2) ( 1 − T y 2 ) . The roots are y 0 = ± 1 / T y_0=\pm1/\sqrt{T} y 0 = ± 1/ T . We get:
Res y = 1 / T ω = − 1 2 ( 1 − T ) ( 1 − 1 T ) ( 1 − T ) 2 ∑ n = 0 g − 1 W n ( T ) T n / 2 , \operatorname{Res}_{y=1/\sqrt{T}}\omega=-\frac{1}{2(1-T)\left(1-\frac{1}{\sqrt{T}}\right)(1-\sqrt{T})^2}\sum_{n=0}^{g-1}W_n(T)T^{n/2}, Res y = 1/ T ω = − 2 ( 1 − T ) ( 1 − T 1 ) ( 1 − T ) 2 1 n = 0 ∑ g − 1 W n ( T ) T n /2 ,
Res y = − 1 / T ω = − 1 2 ( 1 − T ) ( 1 + 1 T ) ( 1 + T ) 2 ∑ n = 0 g − 1 W n ( T ) ( − 1 ) n T n / 2 . \operatorname{Res}_{y=-1/\sqrt{T}}\omega=-\frac{1}{2(1-T)\left(1+\frac{1}{\sqrt{T}}\right)(1+\sqrt{T})^2}\sum_{n=0}^{g-1}W_n(T)(-1)^nT^{n/2}. Res y = − 1/ T ω = − 2 ( 1 − T ) ( 1 + T 1 ) ( 1 + T ) 2 1 n = 0 ∑ g − 1 W n ( T ) ( − 1 ) n T n /2 .
The simple poles at y = ± T y=\pm\sqrt{T} y = ± T : These are simple poles of ω 2 \omega_2 ω 2 from the factor ( T − y 2 ) (T-y^2) ( T − y 2 ) . The roots are y 0 = ± T y_0=\pm\sqrt{T} y 0 = ± T . We get:
Res y = T ω = T 3 2 ( 1 − T ) ( 1 − T ) ( T − T ) 2 ∑ n = 0 g − 1 W n ( T ) T n + g T − n / 2 , \operatorname{Res}_{y=\sqrt{T}}\omega=\frac{T^3}{2(1-T)(1-\sqrt{T})(T-\sqrt{T})^2}\sum_{n=0}^{g-1}W_n(T)T^{n+g}T^{-n/2}, Res y = T ω = 2 ( 1 − T ) ( 1 − T ) ( T − T ) 2 T 3 n = 0 ∑ g − 1 W n ( T ) T n + g T − n /2 ,
Res y = − T ω = T 3 2 ( 1 − T ) ( 1 + T ) ( T + T ) 2 ∑ n = 0 g − 1 W n ( T ) ( − 1 ) n T n + g T − n / 2 . \operatorname{Res}_{y=-\sqrt{T}}\omega=\frac{T^3}{2(1-T)(1+\sqrt{T})(T+\sqrt{T})^2}\sum_{n=0}^{g-1}W_n(T)(-1)^nT^{n+g}T^{-n/2}. Res y = − T ω = 2 ( 1 − T ) ( 1 + T ) ( T + T ) 2 T 3 n = 0 ∑ g − 1 W n ( T ) ( − 1 ) n T n + g T − n /2 .
Set
X : = ∑ n = 0 g − 1 W n = 1 − T g 1 − T + T g + 1 − T 3 g − 1 1 − T 2 X := \sum_{n=0}^{g-1} W_n = \frac{1-T^g}{1-T}+\frac{T^{g+1}-T^{3g-1}}{1-T^2} X := n = 0 ∑ g − 1 W n = 1 − T 1 − T g + 1 − T 2 T g + 1 − T 3 g − 1
Y : = ∑ n = 0 g − 1 W n T n = 1 − T 2 g 1 − T 2 + T 2 g − 1 − T 3 g − 2 1 − T Y := \sum_{n=0}^{g-1} W_nT^n = \frac{1-T^{2g}}{1-T^2}+\frac{T^{2g-1}-T^{3g-2}}{1-T} Y := n = 0 ∑ g − 1 W n T n = 1 − T 2 1 − T 2 g + 1 − T T 2 g − 1 − T 3 g − 2
Z : = ∑ n = 0 g − 1 W n T n / 2 = 1 − T 3 g − 3 / 2 1 − T 3 / 2 Z := \sum_{n=0}^{g-1} W_nT^{n/2} = \frac{1-T^{3g-3/2}}{1-T^{3/2}} Z := n = 0 ∑ g − 1 W n T n /2 = 1 − T 3/2 1 − T 3 g − 3/2
Z ′ : = ∑ n = 0 g − 1 W n ( − 1 ) n T n / 2 = 1 + T 3 g − 3 / 2 1 + T 3 / 2 Z' := \sum_{n=0}^{g-1} W_n(-1)^nT^{n/2} = \frac{1+T^{3g-3/2}}{1+T^{3/2}} Z ′ := n = 0 ∑ g − 1 W n ( − 1 ) n T n /2 = 1 + T 3/2 1 + T 3 g − 3/2
V : = ∑ n = 0 g − 1 ( n + 1 ) W n T n = g T 2 g + 2 + 1 − ( g + 1 ) T 2 g ( 1 − T 2 ) 2 + T 2 g − 1 ( g − 1 ) + T g − g T ( 1 − T ) 2 V := \sum_{n=0}^{g-1}(n+1)W_nT^n = \frac{gT^{2g+2}+1-(g+1)T^{2g}}{(1-T^2)^2}+T^{2g-1}\frac{(g-1)+T^g-gT}{(1-T)^2} V := n = 0 ∑ g − 1 ( n + 1 ) W n T n = ( 1 − T 2 ) 2 g T 2 g + 2 + 1 − ( g + 1 ) T 2 g + T 2 g − 1 ( 1 − T ) 2 ( g − 1 ) + T g − g T
V ′ : = ∑ n = 0 g − 1 ( n + 1 ) W n = g T g + 1 + 1 − ( g + 1 ) T g ( 1 − T ) 2 + ( g − 1 ) T g + 1 + T 3 g + 1 − g T g + 3 ( 1 − T 2 ) 2 . V' := \sum_{n=0}^{g-1}(n+1)W_n = \frac{gT^{g+1}+1-(g+1)T^g}{(1-T)^2}+\frac{(g-1)T^{g+1}+T^{3g+1}-gT^{g+3}}{(1-T^2)^2}. V ′ := n = 0 ∑ g − 1 ( n + 1 ) W n = ( 1 − T ) 2 g T g + 1 + 1 − ( g + 1 ) T g + ( 1 − T 2 ) 2 ( g − 1 ) T g + 1 + T 3 g + 1 − g T g + 3 .
Adding the seven listed residue values gives us − h U ( T ) -h_U(T) − h U ( T ) . More precisely, we have
h U ( T ) = X 1 + 3 T g + 2 ( 1 − T ) 4 + Y T g + 4 + 3 T 2 ( 1 − T ) 4 + V T 2 ( 1 − T ) 3 − V ′ T g + 1 ( 1 − T ) 3 − Z T g + 2 + T 2 ( 1 − T ) ( 1 − T ) 3 + Z ′ T − T g + 2 2 ( 1 − T ) ( 1 + T ) 3 . \begin{aligned}
h_U(T)={}&X\frac{1+3T^{g+2}}{(1-T)^4}+Y\frac{T^{g+4}+3T^2}{(1-T)^4}+V\frac{T^2}{(1-T)^3}\\
&-V'\frac{T^{g+1}}{(1-T)^3}-Z\frac{T^{g+2}+\sqrt{T}}{2(1-T)(1-\sqrt{T})^3}+Z'\frac{\sqrt{T}-T^{g+2}}{2(1-T)(1+\sqrt{T})^3}.
\end{aligned} h U ( T ) = X ( 1 − T ) 4 1 + 3 T g + 2 + Y ( 1 − T ) 4 T g + 4 + 3 T 2 + V ( 1 − T ) 3 T 2 − V ′ ( 1 − T ) 3 T g + 1 − Z 2 ( 1 − T ) ( 1 − T ) 3 T g + 2 + T + Z ′ 2 ( 1 − T ) ( 1 + T ) 3 T − T g + 2 .
Consequently, we have the claimed expression for ( 1 − T ) 2 ( 1 − T 2 ) 2 ( 1 − T 3 ) h U ( T ) (1-T)^2(1-T^2)^2(1-T^3)h_U(T) ( 1 − T ) 2 ( 1 − T 2 ) 2 ( 1 − T 3 ) h U ( T ) . The associated computations are executed via SymPy and the code is attached in fig:1 . □
Final computations and a numerical proof of Conjecture A
We give a numerical proof of surjectivity and hence of Conjecture A.
Alternative proof of Conjecture A. By Theorems 6.5 and 6.12, the sum of codimension-twisted Hilbert series over HN-strata is given by:
hilb s e r ( U H ( 2 , d , g ) ) 1 − T + T g hilb s e r ( ⨆ n ≥ ⌊ d / 2 ⌋ + 1 Filt ( 1 , 2 ) ( n , d ) ( g ) ) = 1 − T g − 3 T g + 1 − 3 T g + 2 − T g + 3 + ( 2 g + 3 ) T 2 g + 1 + ( 2 g + 3 ) T 2 g + 2 − ( 2 g + 3 ) T 2 g + 4 − ( 2 g + 3 ) T 2 g + 5 + T 3 g + 3 ( 1 − T ) 3 ( 1 − T 2 ) 2 ( 1 − T 3 ) + 3 T 3 g + 4 + 3 T 3 g + 5 + T 3 g + 6 − T 4 g + 6 ( 1 − T ) 3 ( 1 − T 2 ) 2 ( 1 − T 3 ) + T g ( g + 2 2 ) T g + ∑ 1 ≤ k ≤ g ( g − k + 2 2 ) ( T g − k + T g + k ) ( 1 − T ) 2 ( 1 − T 2 ) . \begin{aligned}
\frac{\operatorname{hilb}_{\mathrm{ser}}(\mathcal{U}^{H}(2,d,g))}{1-T}+T^g\operatorname{hilb}_{\mathrm{ser}}\left(\bigsqcup_{n\ge\lfloor d/2\rfloor+1}\operatorname{Filt}_{(1,2)}^{(n,d)}(g)\right) \\
&=\frac{1-T^g-3T^{g+1}-3T^{g+2}-T^{g+3}+(2g+3)T^{2g+1}+(2g+3)T^{2g+2}-(2g+3)T^{2g+4}-(2g+3)T^{2g+5}+T^{3g+3}}{(1-T)^3(1-T^2)^2(1-T^3)} \\
&+\frac{3T^{3g+4}+3T^{3g+5}+T^{3g+6}-T^{4g+6}}{(1-T)^3(1-T^2)^2(1-T^3)}+T^g\frac{\binom{g+2}{2}T^g+\sum_{1\le k\le g}\binom{g-k+2}{2}(T^{g-k}+T^{g+k})}{(1-T)^2(1-T^2)}.
\end{aligned} 1 − T hilb ser ( U H ( 2 , d , g )) + T g hilb ser n ≥ ⌊ d /2 ⌋ + 1 ⨆ Filt ( 1 , 2 ) ( n , d ) ( g ) = ( 1 − T ) 3 ( 1 − T 2 ) 2 ( 1 − T 3 ) 1 − T g − 3 T g + 1 − 3 T g + 2 − T g + 3 + ( 2 g + 3 ) T 2 g + 1 + ( 2 g + 3 ) T 2 g + 2 − ( 2 g + 3 ) T 2 g + 4 − ( 2 g + 3 ) T 2 g + 5 + T 3 g + 3 + ( 1 − T ) 3 ( 1 − T 2 ) 2 ( 1 − T 3 ) 3 T 3 g + 4 + 3 T 3 g + 5 + T 3 g + 6 − T 4 g + 6 + T g ( 1 − T ) 2 ( 1 − T 2 ) ( 2 g + 2 ) T g + ∑ 1 ≤ k ≤ g ( 2 g − k + 2 ) ( T g − k + T g + k ) .
Multiplying by the denominator in Theorem 6.7, ( 1 − T ) 3 ( 1 − T 2 ) 2 ( 1 − T 3 ) (1-T)^3(1-T^2)^2(1-T^3) ( 1 − T ) 3 ( 1 − T 2 ) 2 ( 1 − T 3 ) , we get:
− T g + 1 − T g + 2 + T 3 g + 4 + T 3 g + 5 − T 4 g + 6 + 1 = ( 1 − T g + 1 ) ( 1 − T g + 2 ) ( 1 − T 2 g + 3 ) -T^{g+1}-T^{g+2}+T^{3g+4}+T^{3g+5}-T^{4g+6}+1=(1-T^{g+1})(1-T^{g+2})(1-T^{2g+3}) − T g + 1 − T g + 2 + T 3 g + 4 + T 3 g + 5 − T 4 g + 6 + 1 = ( 1 − T g + 1 ) ( 1 − T g + 2 ) ( 1 − T 2 g + 3 )
which matches the numerator in Theorem 6.7 . The computations are attached in Figure 2. That is, we have
( 1 − T ) 3 ( 1 − T 2 ) 2 ( 1 − T 3 ) ( hilb ser ( V e c H ( 2 , d , g ) ) 1 − T + T g hilb ser ( ⨆ n ≥ ⌊ d / 2 ⌋ + 1 Filt ( 1 , 2 ) ( n , d ) ( g ) ) − hilb ser ( H k ∗ ( Bun 2 , d C ) ) ) = 0. (1-T)^3(1-T^2)^2(1-T^3)\left(\frac{\operatorname{hilb}_{\operatorname{ser}}(\mathcal{V}ec^{\mathcal{H}}(2,d,g))}{1-T}+T^g\operatorname{hilb}_{\operatorname{ser}}\left(\bigsqcup_{n\geq\lfloor d/2\rfloor+1}\operatorname{Filt}^{(n,d)}_{(1,2)}(g)\right)-\operatorname{hilb}_{\operatorname{ser}}(H^*_k(\operatorname{Bun}_{2,d}C))\right)=0. ( 1 − T ) 3 ( 1 − T 2 ) 2 ( 1 − T 3 ) 1 − T hilb ser ( V e c H ( 2 , d , g )) + T g hilb ser n ≥ ⌊ d /2 ⌋ + 1 ⨆ Filt ( 1 , 2 ) ( n , d ) ( g ) − hilb ser ( H k ∗ ( Bun 2 , d C )) = 0.
The conjecture follows from a straightforward adaptation of [31 Theorem 4.1] . □
Even degree bundles
It remains to prove the conjecture for V e c H ( 2 , d , g ) \mathcal{V}ec^{\mathcal{H}}(2,d,g) V e c H ( 2 , d , g ) for d d d even. Throughout this section, we shall use the fact that the pullback of Chow rings along finite surjective maps of smooth DM-stacks is injective without explicit mention.
Lemma 7.1. Let X X X be a smooth DM stack and R ⊂ A ∗ ( X ) R \subset A^*(X) R ⊂ A ∗ ( X ) be a sub-ring. Suppose F 0 ⊂ X F_0 \subset X F 0 ⊂ X is a smooth locally closed substack admitting a filtration F 0 ⊃ F 1 ⊃ ⋯ ⊃ F r F_0 \supset F_1 \supset\cdots\supset F_r F 0 ⊃ F 1 ⊃ ⋯ ⊃ F r by smooth closed substacks of F 0 F_0 F 0 . Set W i = F i ∖ F i + 1 W_i=F_i\setminus F_{i+1} W i = F i ∖ F i + 1 for 0 ≤ i < r 0 \leq i < r 0 ≤ i < r and W r = F r W_r=F_r W r = F r . Assume the following:
(1) For every 0 ≤ i ≤ r 0 \leq i \leq r 0 ≤ i ≤ r , the restriction map R → A ∗ ( W i ) R \to A^*(W_i) R → A ∗ ( W i ) is surjective.
(2) [ F i + 1 ] ∈ A ∗ ( F i ) [F_{i+1}] \in A^*(F_i) [ F i + 1 ] ∈ A ∗ ( F i ) is the restriction of some a i ∈ R a_i \in R a i ∈ R .
Then the restriction map R → A ∗ ( F 0 ) R \to A^*(F_0) R → A ∗ ( F 0 ) is surjective.
Proof. We proceed by descending induction on i i i . For i = r i=r i = r , this is (1) since W r = F r W_r=F_r W r = F r . Assume R → A ∗ ( F i + 1 ) R \to A^*(F_{i+1}) R → A ∗ ( F i + 1 ) is surjective. By the excision exact sequence
A ∗ ( F i + 1 ) → A ∗ ( F i ) → A ∗ ( W i ) → 0 , A^*(F_{i+1}) \to A^*(F_i) \to A^*(W_i) \to0, A ∗ ( F i + 1 ) → A ∗ ( F i ) → A ∗ ( W i ) → 0 ,
and (1), given z ∈ A ∗ ( F i ) z \in A^*(F_i) z ∈ A ∗ ( F i ) we can find z ~ ∈ R \widetilde{z} \in R z ∈ R such that z − z ~ ∣ F i = j ∗ z ′ z-\widetilde{z}|_{F_i}=j_*z' z − z ∣ F i = j ∗ z ′ for some z ′ ∈ A ∗ ( F i + 1 ) z' \in A^*(F_{i+1}) z ′ ∈ A ∗ ( F i + 1 ) where j : F i + 1 ↪ F i j:F_{i+1}\hookrightarrow F_i j : F i + 1 ↪ F i is the inclusion. But by induction hypothesis, z ′ = z ~ ′ ∣ F i + 1 z'=\widetilde{z}'|_{F_{i+1}} z ′ = z ′ ∣ F i + 1 for some z ~ ′ ∈ R \widetilde{z}' \in R z ′ ∈ R so we have j ∗ ( z ′ ) = j ∗ ( z ~ ′ ∣ F i + 1 ) = ( z ~ ′ ∣ F i ) ⋅ j ∗ ( [ F i + 1 ] ) = ( z ~ ′ ⋅ a i ) ∣ F i j_*(z')=j_*(\widetilde{z}'|_{F_{i+1}})=(\widetilde{z}'|_{F_i})\cdot j_*([F_{i+1}])=(\widetilde{z}'\cdot a_i)|_{F_i} j ∗ ( z ′ ) = j ∗ ( z ′ ∣ F i + 1 ) = ( z ′ ∣ F i ) ⋅ j ∗ ([ F i + 1 ]) = ( z ′ ⋅ a i ) ∣ F i . This completes the inductive step and hence the proof. □
Lemma 7.2. Let
Z g , m : = { ( C , p 1 , … , p m ) ∈ H g , m ∣ p 1 is a Weierstrass point } . Z_{g,m}:=\{(C,p_1,\ldots,p_m)\in\mathcal{H}_{g,m}\mid p_1\text{ is a Weierstrass point}\}. Z g , m := {( C , p 1 , … , p m ) ∈ H g , m ∣ p 1 is a Weierstrass point } . We have the closed immersion j m : Z g , m ↪ H g , m j_m:Z_{g,m}\hookrightarrow\mathcal{H}_{g,m} j m : Z g , m ↪ H g , m . Then
j m ∗ : A ∗ ( H g , m ) → A ∗ ( Z g , m ) j_m^*:A^*(\mathcal{H}_{g,m})\to A^*(Z_{g,m}) j m ∗ : A ∗ ( H g , m ) → A ∗ ( Z g , m ) is surjective for m ≤ 2 g + 5 m\leq2g+5 m ≤ 2 g + 5 .
Proof. We adopt the strategy in [8 ] . Let H g , m o \mathcal{H}^o_{g,m} H g , m o (respectively Z g , m o Z^o_{g,m} Z g , m o ) denote the open locus where no two marked points are conjugate under the hyper-elliptic involution. Set W i s m : = { ( C , p 1 , … , p m ) ∈ Z g , m o ∣ p 1 , … , p i are Weierstrass , p i + 1 is not Weierstrass } W_i^{sm}:=\{(C,p_1,\ldots,p_m)\in Z^o_{g,m}\mid p_1,\ldots,p_i\text{ are Weierstrass},\ p_{i+1}\text{ is not Weierstrass}\} W i s m := {( C , p 1 , … , p m ) ∈ Z g , m o ∣ p 1 , … , p i are Weierstrass , p i + 1 is not Weierstrass } .1 In the notation of [8 ] , we have the Cartesian diagram
W i s m → ⊂ o p e n W i ↓ ↓ H g , m o → ⊂ o p e n I g , m o \begin{CD}
W_i^{sm} @>{\subset\mathrm{open}}>> W_i \\
@VVV @VVV \\
\mathcal{H}_{g,m}^{o} @>{\subset\mathrm{open}}>> \mathcal{I}_{g,m}^{o}
\end{CD} W i s m ↓ ⏐ H g , m o ⊂ open ⊂ open W i ↓ ⏐ I g , m o
From [8 Lemma 4.6] , we know A ∗ ( I g , m o ) → A ∗ ( W i ) A^{*}(\mathcal{I}_{g,m}^{o}) \to A^{*}(W_i) A ∗ ( I g , m o ) → A ∗ ( W i ) is surjective for $1 \leq i < m \leq ga n d h e n c e b y e x c i s i o n , w e k n o w and hence by excision, we know an d h e n ce b y e x c i s i o n , w e k n o w A^{}(\mathcal{H}_{g,m}) \to A^{ }(W_i^{sm})i s s u r j e c t i v e f o r $ 1 ≤ i < m ≤ g is surjective for \$1 \leq i < m \leq g i ss u r j ec t i v e f or $1 ≤ i < m ≤ g . From [8 Lemma 3.8] , we know that A ∗ ( W m s m ) = Q A^{*}(W_m^{sm}) = \mathbb{Q} A ∗ ( W m s m ) = Q and hence A ∗ ( H g , m ) → A ∗ ( W m s m ) A^{*}(\mathcal{H}_{g,m}) \to A^{*}(W_m^{sm}) A ∗ ( H g , m ) → A ∗ ( W m s m ) is trivially surjective. Moreover, if we set
F i s m : = { ( C , p 1 , … , p m ) ∈ Z g , m o ∣ p 1 , … , p i are Weierstrass points. } , F_i^{sm}:=\left\{(C,p_1,\ldots,p_m)\in\mathcal{Z}_{g,m}^{o}\mid p_1,\ldots,p_i\text{ are Weierstrass points.}\right\}, F i s m := { ( C , p 1 , … , p m ) ∈ Z g , m o ∣ p 1 , … , p i are Weierstrass points. } ,
then F i + 1 s m ∖ F i s m = W i s m F_{i+1}^{sm}\setminus F_i^{sm}=W_i^{sm} F i + 1 s m ∖ F i s m = W i s m . From [8 Lemma 2.9] , we see that [ F i s m ] [F_i^{sm}] [ F i s m ] is the restriction of a product of divisors. So we can apply 7.1 to conclude that the restriction map A ∗ ( H g , m ) → A ∗ ( Z g , m o ) A^{*}(\mathcal{H}_{g,m})\to A^{*}(\mathcal{Z}_{g,m}^{o}) A ∗ ( H g , m ) → A ∗ ( Z g , m o ) is surjective for 1 ≤ m ≤ g 1\leq m\leq g 1 ≤ m ≤ g . We now prove A ∗ ( H g , m ) → A ∗ ( Z g , m ) A^{*}(\mathcal{H}_{g,m})\to A^{*}(\mathcal{Z}_{g,m}) A ∗ ( H g , m ) → A ∗ ( Z g , m ) is surjective for 1 ≤ m ≤ g 1\leq m\leq g 1 ≤ m ≤ g . For m = 1 m=1 m = 1 , Z g , 1 = H g w \mathcal{Z}_{g,1}=\mathcal{H}_{g}^{w} Z g , 1 = H g w and we know from [7 Lemma 4.1] that A ∗ ( H g w ) = Q A^{*}(\mathcal{H}_{g}^{w})=\mathbb{Q} A ∗ ( H g w ) = Q so A ∗ ( H g , 1 ) → A ∗ ( Z g , 1 ) A^{*}(\mathcal{H}_{g,1})\to A^{*}(\mathcal{Z}_{g,1}) A ∗ ( H g , 1 ) → A ∗ ( Z g , 1 ) is trivially surjective2 . We also note that this implies Z g , 1 \mathcal{Z}_{g,1} Z g , 1 is irreducible and considering the forgetful map Z g , m → Z g , 1 \mathcal{Z}_{g,m}\to\mathcal{Z}_{g,1} Z g , m → Z g , 1 , we can deduce that each Z g , m \mathcal{Z}_{g,m} Z g , m is irreducible. We proceed by induction on m m m . Suppose we have proved A ∗ ( H g , m − 1 ) → A ∗ ( Z g , m − 1 ) A^{*}(\mathcal{H}_{g,m-1})\to A^{*}(\mathcal{Z}_{g,m-1}) A ∗ ( H g , m − 1 ) → A ∗ ( Z g , m − 1 ) is surjective. Set E i , j : = Z g , m ∩ D i , j E_{i,j}:=\mathcal{Z}_{g,m}\cap D_{i,j} E i , j := Z g , m ∩ D i , j for 2 ≤ i < j ≤ m 2\leq i<j\leq m 2 ≤ i < j ≤ m . Forgetting p j p_j p j identifies E i , j E_{i,j} E i , j with an open sub-stack of Z g , m − 1 \mathcal{Z}_{g,m-1} Z g , m − 1 and in particular, we note that E i , j E_{i,j} E i , j is also irreducible. Thus from the commutative diagram
H g , m → Forget j H g , m − 1 ↑ ↑ E i , j → V i , j → ⊂ o p e n Z g , m − 1 \begin{CD}
\mathcal{H}_{g,m} @>{\operatorname{Forget}_{j}}>> \mathcal{H}_{g,m-1} \\
@AAA @AAA \\
E_{i,j} @>>> V_{i,j} @>{\subset\mathrm{open}}>> \mathcal{Z}_{g,m-1}
\end{CD} H g , m ⏐ ↑ E i , j Forget j H g , m − 1 ⏐ ↑ V i , j ⊂ open Z g , m − 1
and the induction hypothesis, we get the restriction map A ∗ ( H g , m ) → A ∗ ( E i , j ) A^{*}(\mathcal{H}_{g,m})\to A^{*}(E_{i,j}) A ∗ ( H g , m ) → A ∗ ( E i , j ) is surjective for 2 ≤ i < j ≤ m 2\leq i<j\leq m 2 ≤ i < j ≤ m . Now since E i , j E_{i,j} E i , j is irreducible, restriction of D i , j ∈ Q > 0 ⋅ [ E i , j ] D_{i,j}\in\mathbb{Q}_{>0}\cdot[E_{i,j}] D i , j ∈ Q > 0 ⋅ [ E i , j ] and in particular, we have [ E i , j ] ∈ A ∗ ( Z g , m ) [E_{i,j}]\in A^{*}(\mathcal{Z}_{g,m}) [ E i , j ] ∈ A ∗ ( Z g , m ) is the restriction of a class in A ∗ ( H g , m ) A^{*}(\mathcal{H}_{g,m}) A ∗ ( H g , m ) . A similar argument as in 7.1 then shows us the restriction A ∗ ( H g , m ) → A ∗ ( Z g , m ) A^{*}(\mathcal{H}_{g,m})\to A_{*}(\mathcal{Z}_{g,m}) A ∗ ( H g , m ) → A ∗ ( Z g , m ) is surjective for 1 ≤ m ≤ g 1\leq m\leq g 1 ≤ m ≤ g in view of the excision exact sequence
⨁ 2 ≤ i < j ≤ m A ∗ ( E i , j ) → A ∗ ( Z g , m ) → A ∗ ( Z g , m o ) → 0. \bigoplus_{2\leq i<j\leq m}A_{*}(E_{i,j})\to A_{*}(\mathcal{Z}_{g,m})\to A^{*}(\mathcal{Z}_{g,m}^{o})\to0. 2 ≤ i < j ≤ m ⨁ A ∗ ( E i , j ) → A ∗ ( Z g , m ) → A ∗ ( Z g , m o ) → 0.
We now tackle the case g + 1 ≤ m ≤ 2 g + 5 g+1\leq m\leq2g+5 g + 1 ≤ m ≤ 2 g + 5 . We use the well-known fact that a smooth hyperelliptic curve C C C admits an embedding into P 1 × P 1 \mathbb{P}^{1}\times\mathbb{P}^{1} P 1 × P 1 as C = { F ( x , y ) = 0 } C=\{F(x,y)=0\} C = { F ( x , y ) = 0 } , where F F F has bi-degree ( g + 1 , 2 ) (g+1,2) ( g + 1 , 2 ) and the hyper-elliptic map is the projection onto the first factor. From [8 ] , we know H g , m o \mathcal{H}_{g,m}^{o} H g , m o is an open substack of P ( F ) \mathbb{P}(\mathcal{F}) P ( F ) where F \mathcal{F} F is a bundle of rank 3 g + 7 − m 3g+7-m 3 g + 7 − m (see [8 Equation 4.11] ) on B g , m o \mathcal{B}_{g,m}^{o} B g , m o where B g , m o \mathcal{B}_{g,m}^{o} B g , m o is the open locus in the configuration stack B g , m \mathcal{B}_{g,m} B g , m consisting of points ( q 1 , … , q g , q g + 2 , … , q m ) ∈ P 1 × P 1 (q_1,\ldots,q_g,q_{g+2},\ldots,q_m)\in\mathbb{P}^{1}\times\mathbb{P}^{1} ( q 1 , … , q g , q g + 2 , … , q m ) ∈ P 1 × P 1 such that the m − 1 m-1 m − 1 points impose linearly independent conditions on equations of bi-degree ( g + 1 , 2 ) (g+1,2) ( g + 1 , 2 ) . Let B g , m o , w \mathcal{B}_{g,m}^{o,w} B g , m o , w denote the open locus consisting of points ( q 1 = ( x 1 , y 1 ) , … , q g = ( x g , y g ) , q g + 2 = ( x g + 2 , y g + 2 ) , … , q m = ( x m , y m ) ) (q_1=(x_1,y_1),\ldots,q_g=(x_g,y_g),q_{g+2}=(x_{g+2},y_{g+2}),\ldots,q_m=(x_m,y_m)) ( q 1 = ( x 1 , y 1 ) , … , q g = ( x g , y g ) , q g + 2 = ( x g + 2 , y g + 2 ) , … , q m = ( x m , y m )) such that the m − 1 m-1 m − 1 points and the condition ∂ y F ( q 1 ) = 0 \partial_yF(q_1)=0 ∂ y F ( q 1 ) = 0 impose linearly independent conditions on equations of bi-degree ( g + 1 , 2 ) (g+1,2) ( g + 1 , 2 ) . Clearly, B g , m o , w ⊂ B g , m o \mathcal{B}_{g,m}^{o,w}\subset\mathcal{B}_{g,m}^{o} B g , m o , w ⊂ B g , m o . For b ∈ B g , m o , w b\in\mathcal{B}_{g,m}^{o,w} b ∈ B g , m o , w as above consider the one dimensional vector space L b = Hom ( T y 1 P 1 , O ( g + 1 , 2 ) ∣ q 1 ) L_b=\operatorname{Hom}(T_{y_1}\mathbb{P}^{1},\mathcal{O}(g+1,2)|_{q_1}) L b = Hom ( T y 1 P 1 , O ( g + 1 , 2 ) ∣ q 1 ) . As b b b varies, this gives us a line bundle L \mathcal{L} L on B g , m o , w \mathcal{B}_{g,m}^{o,w} B g , m o , w . Moreover we have F → d y L \mathcal{F} \xrightarrow{d_y} \mathcal{L} F d y L , which is fiber-wise given by F ↦ ∂ y F ( q 1 ) F \mapsto\partial_yF(q_1) F ↦ ∂ y F ( q 1 ) . By definition of B g , m o , w \mathcal{B}^{o,w}_{g,m} B g , m o , w the kernel of this morphism is a vector bundle F w \mathcal{F}^{w} F w on B g , m o , w \mathcal{B}^{o,w}_{g,m} B g , m o , w of rank rk F − 1 = 3 g + 6 − m \operatorname{rk}\mathcal{F}-1=3g+6-m rk F − 1 = 3 g + 6 − m . We claim that Z g , m o → P ( F ) → B g , m o \mathcal{Z}^{o}_{g,m}\to\mathbb{P}(\mathcal{F})\to\mathcal{B}^{o}_{g,m} Z g , m o → P ( F ) → B g , m o factors through B g , m o , w \mathcal{B}^{o,w}_{g,m} B g , m o , w . To see this, consider the divisor Γ : = 2 p 1 + ∑ j = 2 j ≠ g + 1 m p j \Gamma:=2p_1+\sum_{\substack{j=2\\j\ne g+1}}^{m}p_j Γ := 2 p 1 + ∑ j = 2 j = g + 1 m p j .
By [8 Lemma 4.2] , the evaluation map
H 0 ( P 1 × P 1 , O ( g + 1 , 2 ) ) ⟶ H 0 ( Γ , O Γ ( g + 1 , 2 ) ) H^0(\mathbb{P}^1\times\mathbb{P}^1,\mathcal{O}(g+1,2))\longrightarrow H^0(\Gamma,\mathcal{O}_{\Gamma}(g+1,2)) H 0 ( P 1 × P 1 , O ( g + 1 , 2 )) ⟶ H 0 ( Γ , O Γ ( g + 1 , 2 ))
is surjective since m ≤ 2 g + 5 m\le2g+5 m ≤ 2 g + 5 . In other words, the m − 1 m-1 m − 1 points and the vanishing of the derivative impose linearly independent conditions on equations of bi-degree ( g + 1 , 2 ) (g+1,2) ( g + 1 , 2 ) . So we get the image of Z g , m o \mathcal{Z}^{o}_{g,m} Z g , m o lands in B g , m o , w \mathcal{B}^{o,w}_{g,m} B g , m o , w .
Now consider the Cartesian square
H g , m o × P ( F ) P ( F w ) → o p e n P ( F w ) o p e n ↓ ↓ o p e n H g , m o → o p e n P ( F ) \begin{CD}
\mathcal{H}^{o}_{g,m}\mathop{\times}_{\mathbb{P}(\mathcal{F})}\mathbb{P}(\mathcal{F}^{w}) @>{\mathrm{open}}>> \mathbb{P}(\mathcal{F}^{w})\\
@V{\mathrm{open}}VV @VV{\mathrm{open}}V\\
\mathcal{H}^{o}_{g,m} @>>{\mathrm{open}}> \mathbb{P}(\mathcal{F})
\end{CD} H g , m o × P ( F ) P ( F w ) open ↓ ⏐ H g , m o open open P ( F w ) ↓ ⏐ open P ( F )
We know the top horizontal arrow is an open immersion and it parametrizes precisely the points { ( C , p 1 , p 2 , … , p m ) ∈ H g , m o ∣ p 1 is Weierstrass. } \{(C,p_1,p_2,\ldots,p_m)\in\mathcal{H}^{o}_{g,m}\mid p_1\text{ is Weierstrass.}\} {( C , p 1 , p 2 , … , p m ) ∈ H g , m o ∣ p 1 is Weierstrass. } . So we have H g , m o × P ( F ) P ( F w ) = Z g , m o \mathcal{H}^{o}_{g,m}\mathop{\times}_{\mathbb{P}(\mathcal{F})}\mathbb{P}(\mathcal{F}^{w})=\mathcal{Z}^{o}_{g,m} H g , m o × P ( F ) P ( F w ) = Z g , m o . Now the restriction map A ∗ ( P ( F ) ) → A ∗ ( P ( F w ) ) A^*(\mathbb{P}(\mathcal{F}))\to A^*(\mathbb{P}(\mathcal{F}^{w})) A ∗ ( P ( F )) → A ∗ ( P ( F w )) is surjective and hence so is the restriction map A ∗ ( H g , m ) → A ∗ ( H g , m o ) → A ∗ ( Z g , m o ) A^*(\mathcal{H}_{g,m})\to A^*(\mathcal{H}^{o}_{g,m})\to A^*(\mathcal{Z}^{o}_{g,m}) A ∗ ( H g , m ) → A ∗ ( H g , m o ) → A ∗ ( Z g , m o ) . To show that A ∗ ( H g , m ) → A ∗ ( Z g , m ) A^*(\mathcal{H}_{g,m})\to A^*(\mathcal{Z}_{g,m}) A ∗ ( H g , m ) → A ∗ ( Z g , m ) is surjective, we use the same argument as in 1 ≤ m ≤ g 1\le m\le g 1 ≤ m ≤ g case.
Corollary 7.2.1. We have A ∗ ( Z g , m ) = Q A^*(\mathcal{Z}_{g,m})=\mathbb{Q} A ∗ ( Z g , m ) = Q for 1 ≤ m ≤ 2 g + 5 1\le m\le2g+5 1 ≤ m ≤ 2 g + 5 .
Proof. We know from [8 ] , that for 1 ≤ m ≤ 2 g + 5 1\le m\le2g+5 1 ≤ m ≤ 2 g + 5 , A ∗ ( H g , m ) = Q [ ψ 1 , … , ψ m ] / ( ψ 1 , … , ψ m ) 2 A^*(\mathcal{H}_{g,m})=\mathbb{Q}[\psi_1,\ldots,\psi_m]/(\psi_1,\ldots,\psi_m)^2 A ∗ ( H g , m ) = Q [ ψ 1 , … , ψ m ] / ( ψ 1 , … , ψ m ) 2 . Let C g , m Z → π Z g , m \mathcal{C}^{\mathcal{Z}}_{g,m}\xrightarrow{\pi}\mathcal{Z}_{g,m} C g , m Z π Z g , m denote the universal curve and s 1 , … , s m s_1,\ldots,s_m s 1 , … , s m are the sections. Since s 1 s_1 s 1 is a Weierstrass section, we have ω π = O ( ( 2 g − 2 ) s 1 ) ⊗ π ∗ M \omega_\pi=\mathcal{O}((2g-2)s_1)\otimes\pi^*M ω π = O (( 2 g − 2 ) s 1 ) ⊗ π ∗ M for some line bundle M M M on Z g , 1 = H g w \mathcal{Z}_{g,1}=\mathcal{H}^{w}_g Z g , 1 = H g w . For i > 1 i>1 i > 1 , s 1 , s i s_1,s_i s 1 , s i are disjoint hence ψ i = c 1 ( s i ∗ ω π ) = s i ∗ π ∗ c 1 ( M ) = 0 \psi_i=c_1(s_i^*\omega_\pi)=s_i^*\pi^*c_1(M)=0 ψ i = c 1 ( s i ∗ ω π ) = s i ∗ π ∗ c 1 ( M ) = 0 since A 1 ( H g w ) = 0 A^1(\mathcal{H}^{w}_g)=0 A 1 ( H g w ) = 0 . For i = 1 i=1 i = 1 , we have ψ 1 \psi_1 ψ 1 is pulled back from a divisor in H g w \mathcal{H}^{w}_g H g w and hence 0 0 0 since A 1 ( H g w ) = 0 A^1(\mathcal{H}^{w}_g)=0 A 1 ( H g w ) = 0 . This completes the proof.
Theorem 7.3. For all g ≥ 2 g\ge2 g ≥ 2 , 0 ≤ n ≤ 2 g + 4 0\le n\le2g+4 0 ≤ n ≤ 2 g + 4 , the pullback map A ∗ ( H g , n ) → A ∗ ( H g , n w ) A^*(\mathcal{H}_{g,n})\to A^*(\mathcal{H}^{w}_{g,n}) A ∗ ( H g , n ) → A ∗ ( H g , n w ) is an isomorphism.
Proof. Let q n : H g , n w → H g , n q_n:\mathcal{H}^{w}_{g,n}\to\mathcal{H}_{g,n} q n : H g , n w → H g , n denote the finite etale map of degree 2 g + 2 2g+2 2 g + 2 . It is clear that q n ∗ : A ∗ ( H g , n ) → A ∗ ( H g , n w ) q_n^*:A^*(\mathcal{H}_{g,n})\to A^*(\mathcal{H}^{w}_{g,n}) q n ∗ : A ∗ ( H g , n ) → A ∗ ( H g , n w ) is injective. So we only need to establish surjectivity. For 1 ≤ i ≤ n 1\le i\le n 1 ≤ i ≤ n , set D i = { w = p i } ⊂ H g , n w D_i=\{w=p_i\}\subset\mathcal{H}^{w}_{g,n} D i = { w = p i } ⊂ H g , n w . Note that D i ≃ Z g , n D_i\simeq\mathcal{Z}_{g,n} D i ≃ Z g , n and hence A ∗ ( D i ) = Q A^*(D_i)=\mathbb{Q} A ∗ ( D i ) = Q . Now if we set U : = H g , n w ∖ ⋃ i = 1 n D i U:=\mathcal{H}^{w}_{g,n}\setminus\bigcup_{i=1}^{n}D_i U := H g , n w ∖ ⋃ i = 1 n D i , then U ≃ Z g , n + 1 U\simeq\mathcal{Z}_{g,n+1} U ≃ Z g , n + 1 . So we have for 1 ≤ n ≤ 2 g + 4 1\le n\le2g+4 1 ≤ n ≤ 2 g + 4 , A ∗ ( U ) ≃ Q A^*(U)\simeq\mathbb{Q} A ∗ ( U ) ≃ Q by Corollary 7.2.1. It follows from a straightforward argument involving the excision sequence that
A ∗ ( H g , n w ) = Span Q { 1 , [ D 1 ] , … , [ D n ] } . A^*(\mathcal{H}^{w}_{g,n})=\operatorname{Span}_{\mathbb{Q}}\{1,[D_1],\ldots,[D_n]\}. A ∗ ( H g , n w ) = Span Q { 1 , [ D 1 ] , … , [ D n ]} .
But if C n , g w → π H n , g w \mathcal{C}^{w}_{n,g}\xrightarrow{\pi}\mathcal{H}^{w}_{n,g} C n , g w π H n , g w is the universal curve, we have ω π = O ( ( 2 g − 2 ) w ) \omega_\pi=\mathcal{O}((2g-2)w) ω π = O (( 2 g − 2 ) w ) up to tensoring by a line bundle pulled back from H g w \mathcal{H}^{w}_g H g w . It follows that q n ∗ ψ i = c 1 ( s i ∗ ω π ) = ( 2 g − 2 ) s i ∗ c 1 ( O ( w ) ) = ( 2 g − 2 ) c 1 ( O H g , n w ( D i ) ) = ( 2 g − 2 ) [ D i ] q_n^*\psi_i=c_1(s_i^*\omega_\pi)=(2g-2)s_i^*c_1(\mathcal{O}(w))=(2g-2)c_1(\mathcal{O}_{\mathcal{H}^{w}_{g,n}}(D_i))=(2g-2)[D_i] q n ∗ ψ i = c 1 ( s i ∗ ω π ) = ( 2 g − 2 ) s i ∗ c 1 ( O ( w )) = ( 2 g − 2 ) c 1 ( O H g , n w ( D i )) = ( 2 g − 2 ) [ D i ] . This shows q n ∗ q_n^* q n ∗ is surjective.
Corollary 7.3.1. The pullback map A ∗ ( C 2 , g ) → A ∗ ( × H g w C 2 , g ) A^*(C_{2,g}) \to A^*(\times\mathcal{H}^w_g C_{2,g}) A ∗ ( C 2 , g ) → A ∗ ( × H g w C 2 , g ) is an isomorphism for all g ≥ 2 g \ge2 g ≥ 2 , 0 ≤ n ≤ 2 g + 4 0 \le n \le2g+4 0 ≤ n ≤ 2 g + 4 .
Proof. Using the stratification in Lemma 4.1 and applying Theorem 7.3, we have surjectivity of the pullback map.
Lemma 7.4. Let X , X ′ , Y , Y ′ \mathcal{X}, \mathcal{X}', \mathcal{Y}, \mathcal{Y}' X , X ′ , Y , Y ′ be smooth DM-stacks of finite type such that we have the following Cartesian diagram:
X ′ → f ′ Y ′ g ′ ↓ ↓ g X → f Y \begin{CD}
\mathcal{X}' @>{f'}>> \mathcal{Y}' \\
@V{g'}VV @VV{g}V \\
\mathcal{X} @>{f}>> \mathcal{Y}
\end{CD} X ′ g ′ ↓ ⏐ X f ′ f Y ′ ↓ ⏐ g Y Additionally, assume that f , f ′ f, f' f , f ′ are proper and generically finite, g , g ′ g,g' g , g ′ are finite and surjective, and that the pullback g ′ ∗ : A ∗ ( X ) → A ∗ ( X ′ ) g'^*: A^*(\mathcal{X}) \to A^*(\mathcal{X}') g ′ ∗ : A ∗ ( X ) → A ∗ ( X ′ ) is an isomorphism. Then the pullback g ∗ : A ∗ ( Y ) → A ∗ ( Y ′ ) g^*: A^*(\mathcal{Y}) \to A^*(\mathcal{Y}') g ∗ : A ∗ ( Y ) → A ∗ ( Y ′ ) is also an isomorphism.
Proof. We use the formula g ∗ f ∗ = g ′ ∗ f ∗ ′ g^*f_* = g'^*f'_* g ∗ f ∗ = g ′ ∗ f ∗ ′ as morphisms A ∗ ( X ) → A ∗ ( Y ′ ) A_*(\mathcal{X}) \to A_*(\mathcal{Y}') A ∗ ( X ) → A ∗ ( Y ′ ) . Surjectivity of g ∗ g^* g ∗ follows from the surjectivity of g ′ ∗ , f ∗ ′ . g'^*, f'_*. g ′ ∗ , f ∗ ′ .
Definition 7.0.1. Let V e c w H ( 2 , d , g ) \mathcal{V}\!ec^H_w(2,d,g) V e c w H ( 2 , d , g ) be the universal moduli stack of rank 2 2 2 , degree d d d vector bundles on hyperelliptic curves of genus g g g with a marked Weierstrass point.
Theorem 7.5. The pullback map A ∗ ( V e c H ( 2 , d , g ) ) → A ∗ ( V e c w H ( 2 , d , g ) ) A^*(\mathcal{V}\!ec^H(2,d,g)) \to A^*(\mathcal{V}\!ec^H_w(2,d,g)) A ∗ ( V e c H ( 2 , d , g )) → A ∗ ( V e c w H ( 2 , d , g )) is an isomorphism for all g ≥ 2 g \ge2 g ≥ 2 and d d d odd.
Proof. Following Theorem 4.4 and applying Corollary 7.3.1 and Lemma 7.4 that the pullbacks
A ∗ ( C 2 , g ( n ) × H g J 2 , g d ′ ) → A ∗ ( C 2 , g ( n ) × H g J 2 , g d ′ × H g H g w ) , A ∗ ( J 2 , g d ′ × H g J 2 , g d ′ ′ ) → A ∗ ( J 2 , g d ′ × H g J 2 , g d ′ ′ × H g H g w ) A^*(C_{2,g}^{(n)} \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'}) \to A^*(C_{2,g}^{(n)} \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d'} \times_{\mathcal{H}_g} \mathcal{H}^w_g), \quad A^*(\mathcal{J}_{2,g}^{d'} \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d''}) \to A^*(\mathcal{J}_{2,g}^{d'} \times_{\mathcal{H}_g} \mathcal{J}_{2,g}^{d''} \times_{\mathcal{H}_g} \mathcal{H}^w_g) A ∗ ( C 2 , g ( n ) × H g J 2 , g d ′ ) → A ∗ ( C 2 , g ( n ) × H g J 2 , g d ′ × H g H g w ) , A ∗ ( J 2 , g d ′ × H g J 2 , g d ′′ ) → A ∗ ( J 2 , g d ′ × H g J 2 , g d ′′ × H g H g w )
are isomorphisms for all 0 ≤ n ≤ g + 4 0 \le n \le g+4 0 ≤ n ≤ g + 4 and d ′ , d ′ ′ ∈ Z d',d'' \in\mathbb{Z} d ′ , d ′′ ∈ Z . The claim now follows from Corollary 5.2.1 and Theorem 5.1.
Theorem 7.6. Conjecture A holds for all g ≥ 2 g \ge2 g ≥ 2 and d d d even.
Proof. For e ∈ Z e \in\mathbb{Z} e ∈ Z , let P e , g : = P V e c w H ( 2 , e , g ) ( E e , g w ∣ w × H g w V e c w H ( 2 , e , g ) ) \mathbb{P}_{e,g} := \mathbb{P}_{\mathcal{V}\!ec^H_w(2,e,g)}(\mathcal{E}^w_{e,g}|_{w \times_{\mathcal{H}^w_g} \mathcal{V}\!ec^H_w(2,e,g)}) P e , g := P V e c w H ( 2 , e , g ) ( E e , g w ∣ w × H g w V e c w H ( 2 , e , g ) ) where w : H g w → C 2 , g w w : \mathcal{H}^w_g \to C^w_{2,g} w : H g w → C 2 , g w is the Weierstrass section and E e , g w → C 2 , g w × H g w V e c w H ( 2 , e , g ) \mathcal{E}^w_{e,g} \to C^w_{2,g} \times_{\mathcal{H}^w_g} \mathcal{V}\!ec^H_w(2,e,g) E e , g w → C 2 , g w × H g w V e c w H ( 2 , e , g ) is the Poincaré bundle.
By [31 Theorem 4.2] , we have that P e , g → V e c w H ( 2 , e , g ) \mathbb{P}_{e,g} \to\mathcal{V}\!ec^H_w(2,e,g) P e , g → V e c w H ( 2 , e , g ) is a P 1 \mathbb{P}^1 P 1 -bundle and P e , g ≅ P e − 1 , g \mathbb{P}_{e,g} \cong\mathbb{P}_{e-1,g} P e , g ≅ P e − 1 , g for all e ∈ Z e \in\mathbb{Z} e ∈ Z . Applying the projective bundle formula, we get that
hilbser ( V e c w H ( 2 , d , g ) ) ( 1 + T ) = hilbser ( P d , g ) = hilbser ( P d − 1 , g ) = hilbser ( V e c w H ( 2 , d − 1 , g ) ) ( 1 + T ) \operatorname{hilbser}(\mathcal{V}\!ec^H_w(2,d,g))(1+T) = \operatorname{hilbser}(\mathbb{P}_{d,g}) = \operatorname{hilbser}(\mathbb{P}_{d-1,g}) = \operatorname{hilbser}(\mathcal{V}\!ec^H_w(2,d-1,g))(1+T) hilbser ( V e c w H ( 2 , d , g )) ( 1 + T ) = hilbser ( P d , g ) = hilbser ( P d − 1 , g ) = hilbser ( V e c w H ( 2 , d − 1 , g )) ( 1 + T )
for all g ≥ 2 g \ge2 g ≥ 2 and d d d even. Applying Theorem 7.5, we get that hilbser ( V e c w H ( 2 , d , g ) ) = hilbser ( V e c H ( 2 , d − 1 , g ) ) \operatorname{hilbser}(\mathcal{V}\!ec^H_w(2,d,g)) = \operatorname{hilbser}(\mathcal{V}\!ec^H(2,d-1,g)) hilbser ( V e c w H ( 2 , d , g )) = hilbser ( V e c H ( 2 , d − 1 , g )) for all g ≥ 2 g \ge2 g ≥ 2 and d d d even. The claim follows from Theorem 6.4 and this also gives a proof of Theorem 0.2 in the even case.
Remark 7.7. Note that the proof of Theorem 7.6 does not depend on the explicit computations of the Hilbert series for the odd case.
Remark 7.8. We also note here that our proof strategy combined with the motivic decomposition of the moduli space of stable bundles of rank 3 and degree d d d with gcd ( d , 3 ) = 1 \gcd(d,3)=1 g cd( d , 3 ) = 1 over a smooth curve as in [12 ] , gives us analogues of Conjecture A A A for U H ( 3 , d , g ) \mathcal{U}^{H}(3,d,g) U H ( 3 , d , g ) and U H ( 3 , d , g ) \mathcal{U}^{H}(3,d,g) U H ( 3 , d , g ) for g ≤ 8 g \le8 g ≤ 8 , and V e c H ( 3 , d , g ) \mathcal{V}ec^{H}(3,d,g) V e c H ( 3 , d , g ) for g ≤ 6 g \le6 g ≤ 6 .
Appendix A. Scripts for supporting computations
import sympy as sp
T , g = sp . symbols ( ’T g ’ , positive = True )
X = (1 - T ** g ) /(1 - T ) + ( T **( g +1) -T **(3* g -1) ) /(1 - T **2)
Y = (1 - T **(2* g ) ) /(1 - T **2) + ( T **(2* g -1) - T **(3* g -2) ) /(1 - T )
Z = (1 - T **(3* g - sp . Rational (3 ,2) ) ) /(1 - T ** sp . Rational (3 ,2) )
Zp = (1+ T **(3* g - sp . Rational (3 ,2) ) ) /(1+ T ** sp . Rational (3 ,2) )
W = ( g * T **( g +1) +1 -( g +1) * T ** g ) /(1 - T ) **2 \
+ (( g -1) * T **( g +1) + T **(3* g +1) -g * T **( g +3) ) /(1 - T **2) **2
V = ( g * T **(2* g +2) +1 -( g +1) * T **(2* g ) ) /(1 - T **2) **2 \
+ T **(2* g -1) *(( g -1) + T ** g - g * T ) /(1 - T ) **2
E = X *(1+3* T **( g +2) ) /(1 - T ) **4 \
+ Y *( T **( g +4) +3* T **2) /(1 - T ) **4 \
+ V * T **2/(1 - T ) **3 \
- W * T **( g +1) /(1 - T ) **3 \
- Z *( T **( g +2) + T ** sp . Rational (1 ,2) ) /(2*(1 - T ) *(1 - T ** sp . Rational (1 ,2) ) **3) \
+ Zp *( T ** sp . Rational (1 ,2) -T **( g +2) ) /(2*(1 - T ) *(1+ T ** sp . Rational (1 ,2) ) **3)
F = sp . simplify ( E *(1 - T ) **2*(1 - T **2) **2*(1 - T **3) )
F_simplified = sp . simplify ( sp . together ( F ) )
sp . factor ( F_simplified )
fig:1 . Computations for Hilbert Series of U H ( 2 , d , g ) \mathcal{U}^{H}(2,d,g) U H ( 2 , d , g )
ka = E /(1 - T ) + T ** g *(
( g +2) *( g +1) /2* T ** g
+ sp . summation (( g - k +2) *( g - k +1) /2*( T **( g - k ) + T **( g + k ) ) , (k , 1 , g ) )
) / ((1 - T ) **2*(1 - T **2) )
ka_simplified = sp . simplify ( sp . together ( ka *(1 - T ) **3*(1 - T **2) **2*(1 - T **3) ) )
diff = sp . simplify ( ka_simplified - (1 - T **( g +1) ) *(1 - T **( g +2) ) *(1 - T **(2* g +3) ) )
diff
fig:2 . Checking equality of Hilbert Series
In this section, we indicate how our strategy can be used to prove some partial results towards Conjecture A or Theorem 0.2 for higher rank stable bundles. Let r , d r,d r , d be co-prime integers with r ≥ 1 r \ge1 r ≥ 1 . We start with del Bano’s formula in K ^ 0 ( M ( k ) ) \widehat{K}_0(M(k)) K 0 ( M ( k )) for an arbitrary field k k k of char 0. From [9 Theorem 4.11] , we can write
[ U ( r , d . g ) ] = F r , d , g ( L , [ Jac ( C ) ] , [ Sym 1 C ] , … , [ Sym g − 1 C ] ) ∏ m = 1 r ( 1 − L m ) N m ∈ K ^ 0 ( M ( k ) ) [U(r,d.g)] = \frac{F_{r,d,g}\left(\mathbb{L},[\operatorname{Jac}(C)],[\operatorname{Sym}^{1}C],\ldots,[\operatorname{Sym}^{g-1}C]\right)}{\prod_{m=1}^{r}(1-\mathbb{L}^{m})^{N_m}} \in\widehat{K}_0(M(k)) [ U ( r , d . g )] = ∏ m = 1 r ( 1 − L m ) N m F r , d , g ( L , [ Jac ( C )] , [ Sym 1 C ] , … , [ Sym g − 1 C ] ) ∈ K 0 ( M ( k ))
where F r , d , g F_{r,d,g} F r , d , g is a polynomial with integer coefficients. Further from the expression in loc. cit., if we assign L \mathbb{L} L has deg 0, Jac ( C ) \operatorname{Jac}(C) Jac ( C ) has degree g g g and Sym i C \operatorname{Sym}^{i}C Sym i C has degree i i i , then we can choose an expression where F r , d , g F_{r,d,g} F r , d , g has degree ≤ r g \le rg ≤ r g . Clearing denominators and re-arranging terms with negative coefficients, we get
[ U ( r , d , g ) ] ( 1 + R r , d , g ( L ) ) + Q r , d , g ( L , [ Jac ( C ) ] , [ Sym 1 C ] , … , [ Sym g − 1 C ] ) = P r , d , g ( L , [ Jac ( C ) ] , [ Sym 1 C ] , … , [ Sym g − 1 C ] ) + [ U ( r , d , g ) ] ( E r , d , g ( L ) ) in K ^ 0 ( M ( k ) ) with E r , d , g ( 0 ) = 0 (11) \begin{aligned}
[U(r,d,g)](1+R_{r,d,g}(\mathbb{L}))+Q_{r,d,g}(\mathbb{L},[\operatorname{Jac}(C)],[\operatorname{Sym}^{1}C],\ldots,[\operatorname{Sym}^{g-1}C]) \\
&=P_{r,d,g}(\mathbb{L},[\operatorname{Jac}(C)],[\operatorname{Sym}^{1}C],\ldots,[\operatorname{Sym}^{g-1}C])+[U(r,d,g)](E_{r,d,g}(\mathbb{L})) \text{ in } \widehat{K}_0(M(k)) \text{ with } E_{r,d,g}(0)=0 \tag*{(11)}
\end{aligned} [ U ( r , d , g )] ( 1 + R r , d , g ( L )) + Q r , d , g ( L , [ Jac ( C )] , [ Sym 1 C ] , … , [ Sym g − 1 C ]) = P r , d , g ( L , [ Jac ( C )] , [ Sym 1 C ] , … , [ Sym g − 1 C ]) + [ U ( r , d , g )] ( E r , d , g ( L )) in K 0 ( M ( k )) with E r , d , g ( 0 ) = 0 (11)
where P , Q , R P,Q,R P , Q , R are polynomials with non-negative integer coefficients. Further each P , Q P,Q P , Q has deg ≤ r g \le rg ≤ r g . But each side is in the image of the functor M ( k ) → K ^ 0 ( M ( k ) ) M(k)\to\widehat{K}_0(M(k)) M ( k ) → K 0 ( M ( k )) , so thanks to [12 Proposition 2.2] , we get an isomorphism in M ( k ) M(k) M ( k ) of the form
h ( U ( r , d , g ) ) ⊕ X r , d , g ≃ Y r , d , g ⊕ h ( U ( r , d , g ) ) ⊗ E r , d , g h(U(r,d,g))\oplus X_{r,d,g}\simeq Y_{r,d,g}\oplus h(U(r,d,g))\otimes\mathcal{E}_{r,d,g} h ( U ( r , d , g )) ⊕ X r , d , g ≃ Y r , d , g ⊕ h ( U ( r , d , g )) ⊗ E r , d , g
where Y r , d , g Y_{r,d,g} Y r , d , g is a direct sum of motives built out of Jac ( C ) k ⊗ ∏ j = 1 b ( Sym i j C ) k j \operatorname{Jac}(C)^k\otimes\prod_{j=1}^{b}(\operatorname{Sym}^{i_j}C)^{k_j} Jac ( C ) k ⊗ ∏ j = 1 b ( Sym i j C ) k j with k g + ∑ j = 1 b i j k j ≤ r g kg+\sum_{j=1}^{b}i_jk_j\le rg k g + ∑ j = 1 b i j k j ≤ r g and E r , d , g \mathcal{E}_{r,d,g} E r , d , g is built out of direct sums of powers of L \mathbb{L} L with strictly positive Tate twists. So we get
h ( U ( r , d , g ) ) ↪ Y r , d , g ⊕ h ( U ( r , d , g ) ) ⊗ E r , d , g is a direct summand. h(U(r,d,g))\hookrightarrow Y_{r,d,g}\oplus h(U(r,d,g))\otimes\mathcal{E}_{r,d,g}\text{ is a direct summand.} h ( U ( r , d , g )) ↪ Y r , d , g ⊕ h ( U ( r , d , g )) ⊗ E r , d , g is a direct summand.
Now iterating this, we get
h ( U ( r , d , g ) ) ↪ ( ⨁ j = 0 N − 1 Y r , d , g ⊗ E r , d , g ⊗ j ) ⊕ h ( U ( r , d , g ) ) ⊗ E r , d , g ⊗ N is a direct summand. h(U(r,d,g))\hookrightarrow\left(\bigoplus_{j=0}^{N-1}Y_{r,d,g}\otimes\mathcal{E}_{r,d,g}^{\otimes j}\right)\oplus h(U(r,d,g))\otimes\mathcal{E}_{r,d,g}^{\otimes N}\text{ is a direct summand.} h ( U ( r , d , g )) ↪ ( j = 0 ⨁ N − 1 Y r , d , g ⊗ E r , d , g ⊗ j ) ⊕ h ( U ( r , d , g )) ⊗ E r , d , g ⊗ N is a direct summand.
But Hom M ( k ) ( h ( U ( r , d , g ) ) , h ( U ( r , d , g ) ) ⊗ E r , d , g ⊗ N ) = 0 \operatorname{Hom}_{M(k)}\left(h(U(r,d,g)),h(U(r,d,g))\otimes\mathcal{E}_{r,d,g}^{\otimes N}\right)=0 Hom M ( k ) ( h ( U ( r , d , g )) , h ( U ( r , d , g )) ⊗ E r , d , g ⊗ N ) = 0 for N ≫ 0 N\gg0 N ≫ 0 because of dimension reasons.
⟹ h ( U ( r , d , g ) ) ↪ ( ⨁ j = 0 N − 1 Y r , d , g ⊗ E r , d , g ⊗ j ) is a direct summand. \Longrightarrow\ h(U(r,d,g))\hookrightarrow\left(\bigoplus_{j=0}^{N-1}Y_{r,d,g}\otimes\mathcal{E}_{r,d,g}^{\otimes j}\right)\text{ is a direct summand.} ⟹ h ( U ( r , d , g )) ↪ ( j = 0 ⨁ N − 1 Y r , d , g ⊗ E r , d , g ⊗ j ) is a direct summand.
We indicate how the same thing holds true for smooth curve fibrations. Let Y r , d , g B Y^B_{r,d,g} Y r , d , g B denote the relative versions of the motives we get for a single smooth curve.
h B ( U C / B ( r , d , g ) ) ⇄ Γ Ψ ⨁ j = 0 N − 1 Y r , d , g B ⊗ E r , d , g ⊗ j h_B(U_{C/B}(r,d,g))\ \underset{\Psi}{\overset{\Gamma}{\rightleftarrows}}\ \bigoplus_{j=0}^{N-1}Y^B_{r,d,g}\otimes\mathcal{E}_{r,d,g}^{\otimes j} h B ( U C / B ( r , d , g )) Ψ ⇄ Γ j = 0 ⨁ N − 1 Y r , d , g B ⊗ E r , d , g ⊗ j
Using the result at the generic point and spreading out we get relative correspondences Γ \Gamma Γ , Ψ \Psi Ψ as above such that Ψ η ∘ Γ η = I d \Psi_{\eta} \circ\Gamma_{\eta} = Id Ψ η ∘ Γ η = I d . So we have I d − Ψ ∘ Γ Id - \Psi\circ\Gamma I d − Ψ ∘ Γ is nilpotent thanks to Guletskii’s Theorem 2.1 since everything is in A ( B ) A(B) A ( B ) . So Γ \Gamma Γ is a split injection or in other words the relative motive is a direct summand as well.
By the Edidin-Graham approximation argument used in Theorem 5.1 , we get that
h H g ( U H ( r , d , g ) ) ↪ ⨁ α h H g ( ( J 2 , g d ) k α × H g ( C 2 , g ( i 1 , α ) ) k 1 , α × H g ⋯ × H g ( C 2 , g ( i b , α ) ) k b , α ) ( t α ) h_{\mathcal{H}_g}\left(\mathcal{U}^{\mathcal{H}}(r,d,g)\right) \hookrightarrow\bigoplus_{\alpha} h_{\mathcal{H}_g}\left(\left(\mathcal{J}_{2,g}^{d}\right)^{k_{\alpha}} \times_{\mathcal{H}_g} \left(\mathcal{C}_{2,g}^{(i_{1,\alpha})}\right)^{k_{1,\alpha}} \times_{\mathcal{H}_g} \cdots\times_{\mathcal{H}_g} \left(\mathcal{C}_{2,g}^{(i_{b,\alpha})}\right)^{k_{b,\alpha}}\right)(t_{\alpha}) h H g ( U H ( r , d , g ) ) ↪ α ⨁ h H g ( ( J 2 , g d ) k α × H g ( C 2 , g ( i 1 , α ) ) k 1 , α × H g ⋯ × H g ( C 2 , g ( i b , α ) ) k b , α ) ( t α )
is a direct summand with k α g α + ∑ j = 1 b i j , α k j , α ≤ r g k_{\alpha}g_{\alpha}+\sum_{j=1}^{b}i_{j,\alpha}k_{j,\alpha}\le rg k α g α + ∑ j = 1 b i j , α k j , α ≤ r g . If r g ≤ 2 g + 6 rg\le2g+6 r g ≤ 2 g + 6 , it follows from Theorem 4.2 and the argument of Theorem 4.4 that U H ( r , d , g ) → H g \mathcal{U}^{\mathcal{H}}(r,d,g)\to\mathcal{H}_g U H ( r , d , g ) → H g satisfies SFP whenever r g ≤ 2 g + 6 rg\le2g+6 r g ≤ 2 g + 6 . One can then use our monodromy argument to get explicit isomorphisms with the monodromy invariant cohomology ring for such moduli stacks.
Department of Mathematics, Yale University, New Haven, CT 06511, USA
Email address: soumik.ghosh@yale.edu
Department of Mathematics, University of California, San Diego, La Jolla, CA 92093, USA
Email address: shsaha@ucsd.edu