Jeremy Miller was supported by NSF Grant DMS2504473 and a Simons Foundation Travel Support for Mathematicians Grant.
Peter Patzt was supported by NSF Grant DMS2202943 and a Simons Foundation Travel Support for Mathematicians Grant.
Introduction
The goal of this paper is to establish a new structural result for the cohomology of symplectic groups of a Dedekind domain R R R with coefficients in the symplectic Steinberg module. We apply our results in the case R R R is a number ring to establish new lower bounds for the cohomology of S p 2 n ( R ) \mathrm{Sp}_{2n}(R) Sp 2 n ( R ) and its congruence subgroups. Before we state our result, we recall basic definitions and Borel–Serre duality.
Borel–Serre duality
1.1. Let F F F be the field of fractions of R R R . Let V V V be an n n n -dimensional F F F -vector space and let T ( V ) \mathcal{T}(V) T ( V ) denote the geometric realization of the poset of proper nonzero subspaces of V V V ordered by inclusion. Let W W W be a 2 n 2n 2 n -dimensional symplectic vector space and let T ω ( W ) \mathcal{T}^{\omega}(W) T ω ( W ) denote the geometric realization of the poset of nonzero isotropic subspace of W W W ordered by inclusion. The Solomon-Tits theorem implies that both of these spaces are spherical. The Steinberg modules and symplectic Steinberg modules can be defined as the top homology of these simplicial complexes:
St ( V ) : = H ~ n − 2 ( T ( V ) ) and St ω ( W ) : = H ~ n − 1 ( T ω ( W ) ) . \operatorname{St}(V) := \widetilde{H}_{n-2}(\mathcal{T}(V)) \quad\text{and}\quad\operatorname{St}^{\omega}(W) := \widetilde{H}_{n-1}(\mathcal{T}^{\omega}(W)). St ( V ) := H n − 2 ( T ( V )) and St ω ( W ) := H n − 1 ( T ω ( W )) .
One motivation for considering these representations is their appearance in Borel–Serre duality. Assume R R R is a number ring with r 1 r_1 r 1 real and r 2 r_2 r 2 many complex conjugate pairs of complex conjugate embeddings. Borel–Serre proved that S L n ( R ) \mathrm{SL}_n(R) SL n ( R ) is a virtual duality group of dimension
ν n S L : = r 1 ( ( n + 1 ) n − 2 2 ) + r 2 ( n 2 − 1 ) − n + 1 \nu_n^{\mathrm{SL}} := r_1\left(\frac{(n+1)n-2}{2}\right) + r_2(n^2-1)-n+1 ν n SL := r 1 ( 2 ( n + 1 ) n − 2 ) + r 2 ( n 2 − 1 ) − n + 1
with dualizing module St ( F n ) \operatorname{St}(F^n) St ( F n ) and S p 2 n ( R ) \mathrm{Sp}_{2n}(R) Sp 2 n ( R ) is a virtual duality group of dimension
ν 2 n ω : = r 1 ( n ( n + 1 ) ) + r 2 ( n ( 2 n + 1 ) ) − n \nu_{2n}^{\omega} := r_1\bigl(n(n+1)\bigr) + r_2\bigl(n(2n+1)\bigr)-n ν 2 n ω := r 1 ( n ( n + 1 ) ) + r 2 ( n ( 2 n + 1 ) ) − n
with dualizing module St 2 n ω ( F 2 n ) \operatorname{St}_{2n}^{\omega}(F^{2n}) St 2 n ω ( F 2 n ) , with F 2 n F^{2n} F 2 n equipped with the standard symplectic form. In particular,
H ν 2 n ω − i ( S p 2 n ( R ) ; Q ) ≅ H i ( S p 2 n ( R ) ; St ω ( F 2 n ) ⊗ Q ) . H^{\nu_{2n}^{\omega}-i}(\mathrm{Sp}_{2n}(R);\mathbb{Q}) \cong H_i(\mathrm{Sp}_{2n}(R);\operatorname{St}^{\omega}(F^{2n})\otimes\mathbb{Q}). H ν 2 n ω − i ( Sp 2 n ( R ) ; Q ) ≅ H i ( Sp 2 n ( R ) ; St ω ( F 2 n ) ⊗ Q ) .
We call the above groups the codimension-i i i rational cohomology groups since the rational cohomology vanishes above the virtual cohomological dimension.
Since the homology with Steinberg coefficients determines the untwisted rational cohomology when R R R is a number ring, there is no loss of information in considering homology with Steinberg coefficients. Moreover, the homology with Steinberg coefficients has additional algebraic structure.
Comodule algebras and symplectic groups
1.2. Fix a coefficient ring k k k which we will assume is a field in this section. Let
B : = ⨁ [ P ] , i H i ( G L ( P ) ; St ( P ⊗ R F ) ⊗ k ) \mathbf{B} := \bigoplus_{[P],i} H_i(\mathrm{GL}(P);\operatorname{St}(P\otimes_R F)\otimes k) B := [ P ] , i ⨁ H i ( GL ( P ) ; St ( P ⊗ R F ) ⊗ k )
with the sum taken over all isomorphism classes of finitely generated projective R R R -modules and let
A : = ⨁ n , i H i ( S p 2 n ( R ) ; St ω ( F 2 n ) ⊗ k ) . \mathbf{A} := \bigoplus_{n,i} H_i(\mathrm{Sp}_{2n}(R);\operatorname{St}^{\omega}(F^{2n})\otimes k). A := n , i ⨁ H i ( Sp 2 n ( R ) ; St ω ( F 2 n ) ⊗ k ) .
We prove the following.
Theorem 1.1. Let R R R be a Dedekind domain. Then A \mathbf{A} A is a graded-commutative comodule-algebra over the graded-commutative Hopf algebra B \mathbf{B} B .
That B is a graded-commutative Hopf algebra was previously known by work of [1 ] and Brown–Chan–Galatius–Payne [2 ] . It follows from a theorem of Leray that graded-commutative Hopf algebras over fields of characteristic zero are free as graded-commutative algebras. This allows one to easily produce classes in the rational cohomology of S L n ( R ) \mathrm{SL}_n(R) SL n ( R ) for R R R a number ring. We prove a similar freeness result for comodule algebras Corollary 5.6 . We use our structural results to deduce the following numerical consequences.
Corollary 1.2. Let R R R be a number ring with class number c c c . Then
dim Q H 2 n − 2 m ω ( S p 2 n ( R ) ; Q ) ≥ ( c − 1 ) n − m \dim_{\mathbb{Q}} H^{\omega}_{2n-2m}(\mathrm{Sp}_{2n}(R);\mathbb{Q}) \ge(c-1)^{n-m} dim Q H 2 n − 2 m ω ( Sp 2 n ( R ) ; Q ) ≥ ( c − 1 ) n − m for all m ≤ n m \le n m ≤ n .
See Theorem 6.8 and Remark 6.11 for a stronger result along these lines and Example 6.9 for an application. Corollary 1.2 generalizes Brück–Himes [3 Theorem 1.1] who proved this result in the case m = 0 m=0 m = 0 .
Example 1.3. If R R R is the ring of integers in Q [ − 23 ] \mathbb{Q}[\sqrt{-23}] Q [ − 23 ] , then r 1 = 0 r_1=0 r 1 = 0 , r 2 = 1 r_2=1 r 2 = 1 , and c = 3 c=3 c = 3 and hence
dim Q H 2 n − 2 ( S p 2 n ( R ) ; Q ) ≥ 2 n − 1 \dim_{\mathbb{Q}} H^{2n-2}(\mathrm{Sp}_{2n}(R);\mathbb{Q}) \ge2^{n-1} dim Q H 2 n − 2 ( Sp 2 n ( R ) ; Q ) ≥ 2 n − 1
for n ≥ 1 n \ge1 n ≥ 1 .
Remark 1.4. Pal–Payne [15 ] construct a comodule algebra structure on the Steinberg homology of S p 2 n ( Z ) \mathrm{Sp}_{2n}(\mathbb{Z}) Sp 2 n ( Z ) using a symplectic analogue of the Waldhausen S-construction, building on arguments from Brown–Chan–Galatius–Payne [2 ] . They work linear-dually with the corresponding module coalgebra, and identify the action on the coaugmentation class with the weight-zero inclusion in Deligne’s mixed Hodge structure and show that the coproduct is a morphism of mixed Hodge structures. Their results give freeness over the weight zero Hopf algebra and strong lower bounds for Betti numbers and dimensions of weight-graded pieces.
Duoidal comodule algebras and congruence subgroups
Theorem 1.1 is the avatar of algebraic structure on the Steinberg modules themselves. The collection of representations
{ S t ( P ⊗ R F ) } \{\mathrm{St}(P \otimes_R F)\} { St ( P ⊗ R F )}
has the structure of a graded-commutative duoidal bialgebra and we prove that the collection of representations
{ S t ω ( W ) } \{\mathrm{St}^{\omega}(W)\} { St ω ( W )}
has the structure of a graded-commutative duoidal comodule algebra over this duoidal bialgebra in Theorem 4.10.
This duoidal perspective is essential for applications to congruence subgroups. For I I I an ideal in R R R , let S p 2 n ( I ) \mathrm{Sp}_{2n}(I) Sp 2 n ( I ) denote the kernel of mod-I I I reduction map
S p 2 n ( R ) → S p 2 n ( R / I ) . \mathrm{Sp}_{2n}(R) \to\mathrm{Sp}_{2n}(R/I). Sp 2 n ( R ) → Sp 2 n ( R / I ) .
When R = Z R=\mathbb{Z} R = Z , the classifying space B S p 2 n ( j Z ) B\mathrm{Sp}_{2n}(j\mathbb{Z}) B Sp 2 n ( j Z ) has the homotopy type of the moduli stack of dimension-n n n principally polarized abelian variety with level-j j j structure. Theorem 4.10 has the following numerical consequence.
Corollary 1.5. Let p p p be an odd prime and k \mathbb{k} k be a field. The codimension i i i -cohomology satisfies
dim k H ( n + 1 ) 2 − i ( S p 2 n + 2 ( p Z ) ; k ) ≥ p 2 n + 1 ( p − 1 2 ) dim k H n 2 − i ( S p 2 n ( p Z ) ; k ) . \dim_{\mathbb{k}} H^{(n+1)^2-i}(\mathrm{Sp}_{2n+2}(p\mathbb{Z});\mathbb{k}) \ge p^{2n+1}\left(\frac{p-1}{2}\right)\dim_{\mathbb{k}} H^{n^2-i}(\mathrm{Sp}_{2n}(p\mathbb{Z});\mathbb{k}). dim k H ( n + 1 ) 2 − i ( Sp 2 n + 2 ( p Z ) ; k ) ≥ p 2 n + 1 ( 2 p − 1 ) dim k H n 2 − i ( Sp 2 n ( p Z ) ; k ) .
Example 1.6. Since the 0th cohomology group of any group is k \mathsf{k} k , an elementary proof by induction shows that
dim k H n 2 − m 2 ( Sp 2 n ( p Z ) ; k ) ≥ p n 2 − m 2 ( p − 1 2 ) n − m \dim_{\mathsf{k}} H^{n^{2}-m^{2}}(\operatorname{Sp}_{2n}(p\mathbb{Z});\mathsf{k}) \geq p^{n^{2}-m^{2}}\left(\frac{p-1}{2}\right)^{n-m} dim k H n 2 − m 2 ( Sp 2 n ( p Z ) ; k ) ≥ p n 2 − m 2 ( 2 p − 1 ) n − m
for n ≥ m n \geq m n ≥ m .
Capovilla-Searle [6 Corollary C] obtained the same lower bound for m = 0 m=0 m = 0 .
Overview
In § 2 , we recall basic properties of Steinberg modules. In § 3 , we describe the necessary categorical background such as duoidal categories. Then, in § 4 we prove the symplectic Steinberg modules are duoidal comodule algebras and prove that A A A is a comodule algebra. In § 5 , we prove our freeness results. We apply these results to the cohomology of symplectic groups in § 6 and to their congruence subgroups in § 7 .
Acknowledgments
We thank Erik Panzer, Dan Petersen, Andrew Snowden, Robin Sroka, and Jennifer Wilson for helpful conversations. ChatGPT was used for proofreading and reference requests.
Review of Steinberg modules
In this section, we recall basic facts and constructions related to Steinberg modules.
Steinberg modules associated to general linear groups
We begin by recalling the Steinberg modules associated to the general linear groups and their presentation.
Convention 2.1. Let R R R be a Dedekind domain with field of fractions F F F . We will also fix a commutative coefficient ring k \mathsf{k} k . Also homology groups will be taken with k \mathsf{k} k -coefficients, all tensor products without subscript will be over k \mathsf{k} k , etc.
Definition 2.2. Let M M M be a projective R R R -module. Let T ( M ) \mathcal{T}(M) T ( M ) denote the geometric realization of the poset of proper nonzero summands ordered by inclusion.
Lemma 2.3. Let M M M be a projective module over a Dedekind domain R R R with field of fractions F F F . Sending a summand N N N to N ⊗ R F N \otimes_R F N ⊗ R F induces an isomorphism T ( M ) → T ( M ⊗ R F ) \mathcal{T}(M) \to\mathcal{T}(M \otimes_R F) T ( M ) → T ( M ⊗ R F ) .
Proof. The inverse is given by sending a subspace V ⊂ M ⊗ R F V \subset M \otimes_R F V ⊂ M ⊗ R F to V ∩ M V \cap M V ∩ M (compare with Church–Farb–Putman [5 Lemma 2.2(c)] ). □
The Solomon–Tits theorem implies T ( M ) \mathcal{T}(M) T ( M ) is highly connected. Define the rank of M M M to be dim F ( M ⊗ F ) \dim_F(M \otimes F) dim F ( M ⊗ F ) .
Theorem 2.4 (Solomon–Tits). Let M M M be a rank-n n n projective module over a Dedekind domain R R R . Then T ( M ) ≃ S n − 2 \mathcal{T}(M) \simeq S^{n-2} T ( M ) ≃ S n − 2 .
Definition 2.5. Let M M M be a rank-n n n projective module over a Dedekind domain R R R . If n > 0 n > 0 n > 0 , let St ( M ) = H ~ n − 2 ( T ( M ) ) \operatorname{St}(M) = \widetilde{H}_{n-2}(\mathcal{T}(M)) St ( M ) = H n − 2 ( T ( M )) . If n = 0 n = 0 n = 0 , let St ( M ) = k \operatorname{St}(M) = \mathsf{k} St ( M ) = k .
By convention, H ~ − 1 ( ∅ ) = k \widetilde{H}_{-1}(\varnothing) = \mathsf{k} H − 1 ( ∅ ) = k . The Solomon–Tits theorem also provides generators of the Steinberg modules. Let β = ( v ⃗ 1 , … , v ⃗ n ) \beta= (\vec{v}_1,\ldots,\vec{v}_n) β = ( v 1 , … , v n ) be an ordered basis of V = M ⊗ R F V = M \otimes_R F V = M ⊗ R F . Let A ( β ) A(\beta) A ( β ) denote the full subcomplex of T ( V ) \mathcal{T}(V) T ( V ) with vertices subspaces given by the span of proper nonempty subsets of { v ⃗ 1 , … , v ⃗ n } \{\vec{v}_1,\ldots,\vec{v}_n\} { v 1 , … , v n } . A ( β ) A(\beta) A ( β ) is homeomorphic to S n − 2 S^{n-2} S n − 2 and this sphere comes with an orientation induced by the order on the basis. The apartment class [ β ] ∈ St ( V ) [\beta] \in\operatorname{St}(V) [ β ] ∈ St ( V ) is the image of the fundamental class of A ( β ) A(\beta) A ( β ) . The Solomon–Tits theorem states that these are generators of St ( V ) ≅ St ( M ) \operatorname{St}(V) \cong\operatorname{St}(M) St ( V ) ≅ St ( M ) . Lee–Szczarba [10 ] determined the relations between these generators.
Theorem 2.6 (Lee–Szczarba). Let V V V be an n n n -dimensional vector space over a field F F F . Then St ( V ) \operatorname{St}(V) St ( V ) has the following presentation. It is generated by apartment classes [ v ⃗ 1 , … , v ⃗ n ] [\vec{v}_1,\ldots,\vec{v}_n] [ v 1 , … , v n ] associated to ordered bases and the relations are as follows:
Let u 1 , … , u n ∈ F × u_1,\ldots,u_n \in F^{\times} u 1 , … , u n ∈ F × . Then
[ v ⃗ 1 , … , v ⃗ n ] = [ u 1 v ⃗ 1 , … , u n v ⃗ n ] . [\vec{v}_1,\ldots,\vec{v}_n]=[u_1\vec{v}_1,\ldots,u_n\vec{v}_n]. [ v 1 , … , v n ] = [ u 1 v 1 , … , u n v n ] . Let σ ∈ S n \sigma\in S_n σ ∈ S n and let sgn : S n → { ± 1 } \operatorname{sgn}:S_n \to\{ \pm1\} sgn : S n → { ± 1 } denote the sign function. Then
[ v ⃗ 1 , … , v ⃗ n ] = sgn ( σ ) [ v ⃗ σ ( 1 ) , … , v ⃗ σ ( n ) ] . [\vec{v}_1,\ldots,\vec{v}_n]=\operatorname{sgn}(\sigma)[\vec{v}_{\sigma(1)},\ldots,\vec{v}_{\sigma(n)}]. [ v 1 , … , v n ] = sgn ( σ ) [ v σ ( 1 ) , … , v σ ( n ) ] . Let v ⃗ 0 = ∑ i = 1 j u i v ⃗ i \vec{v}_0=\sum_{i=1}^{j}u_i\vec{v}_i v 0 = ∑ i = 1 j u i v i with u i ∈ F × u_i \in F^{\times} u i ∈ F × . Then
∑ i = 0 j ( − 1 ) i [ v ⃗ 0 , … , v ⃗ i − 1 , v ⃗ i + 1 , … , v ⃗ j , … , v ⃗ n ] = 0. \sum_{i=0}^{j}(-1)^i[\vec{v}_0,\ldots,\vec{v}_{i-1},\vec{v}_{i+1},\ldots,\vec{v}_j,\ldots,\vec{v}_n]=0. i = 0 ∑ j ( − 1 ) i [ v 0 , … , v i − 1 , v i + 1 , … , v j , … , v n ] = 0.
Steinberg modules associated to symplectic groups
We now describe the analogous story for symplectic groups.
Definition 2.7. Let W W W be a finite rank free R R R -module. We say that a bi-linear form ω \omega ω on W W W is symplectic if there is an ordered basis ( a ⃗ 1 , b ⃗ 1 , … , a ⃗ n , b ⃗ n ) (\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_n,\vec{b}_n) ( a 1 , b 1 , … , a n , b n ) with
ω ( a ⃗ i , a ⃗ j ) = ω ( b ⃗ i , b ⃗ j ) = 0 for all i and j \omega(\vec{a}_i,\vec{a}_j)=\omega(\vec{b}_i,\vec{b}_j)=0\text{ for all }i\text{ and }j ω ( a i , a j ) = ω ( b i , b j ) = 0 for all i and j and
ω ( a ⃗ i , b ⃗ j ) = − ω ( b ⃗ i , a ⃗ j ) = δ i , j for all i and j . \omega(\vec{a}_i,\vec{b}_j)=-\omega(\vec{b}_i,\vec{a}_j)=\delta_{i,j}\text{ for all }i\text{ and }j. ω ( a i , b j ) = − ω ( b i , a j ) = δ i , j for all i and j . We say ( a ⃗ 1 , b ⃗ 1 , … , a ⃗ n , b ⃗ n ) (\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_n,\vec{b}_n) ( a 1 , b 1 , … , a n , b n ) is a symplectic basis and ( W , ω ) (W,\omega) ( W , ω ) is a symplectic R R R -module. If ( W , ω ) (W,\omega) ( W , ω ) is a symplectic R R R -module, then we say ( x ⃗ 1 , y ⃗ 1 , … , x ⃗ n , y ⃗ n ) (\vec{x}_1,\vec{y}_1,\ldots,\vec{x}_n,\vec{y}_n) ( x 1 , y 1 , … , x n , y n ) is a generalized symplectic basis if
ω ( x ⃗ i , x ⃗ j ) = ω ( y ⃗ i , y ⃗ j ) = 0 for all i and j \omega(\vec{x}_i,\vec{x}_j)=\omega(\vec{y}_i,\vec{y}_j)=0\text{ for all }i\text{ and }j ω ( x i , x j ) = ω ( y i , y j ) = 0 for all i and j and
ω ( x ⃗ i , y ⃗ j ) = 0 for all i ≠ j and ω ( x ⃗ i , y ⃗ i ) ≠ 0. \omega(\vec{x}_i,\vec{y}_j)=0\text{ for all }i\ne j\text{ and }\omega(\vec{x}_i,\vec{y}_i)\ne0. ω ( x i , y j ) = 0 for all i = j and ω ( x i , y i ) = 0. A summand U ⊆ W U\subseteq W U ⊆ W is called isotropic if the form vanishes on U U U and is called symplectic if the form restricts to a symplectic form. Maximal rank isotropic summands are called Lagrangian . We call n n n the genus of W W W . If W W W is a genus n n n symplectic module over a Dedekind domain R R R and J ⊂ W J\subset W J ⊂ W is a rank m m m isotropic summand, then the symplectic form on W W W induces a symplectic form on J ⊥ / J J^\perp/J J ⊥ / J and J ⊥ / J J^\perp/J J ⊥ / J has genus n − m n-m n − m . In particular, J ⊥ / J J^\perp/J J ⊥ / J is a free R R R -module of rank 2 ( n − m ) 2(n-m) 2 ( n − m ) .
Definition 2.8. Let ( W , ω ) (W,\omega) ( W , ω ) be a symplectic module over a Dedekind domain R R R . Let T ω ( W ) \mathcal{T}^{\omega}(W) T ω ( W ) denote the geometric realization of the poset of nonzero isotropic summands.
As before, T ω ( W ) ≅ T ω ( W ⊗ R F ) \mathcal{T}^{\omega}(W)\cong\mathcal{T}^{\omega}(W\otimes_R F) T ω ( W ) ≅ T ω ( W ⊗ R F ) if R R R is a Dedekind domain. The Solomon–Tits theorem implies that T ω ( W ) \mathcal{T}^{\omega}(W) T ω ( W ) is highly connected and provides generators for the top homology.
Theorem 2.9 (Solomon–Tits). Let ( W , ω ) (W,\omega) ( W , ω ) be a genus-n n n symplectic module over a Dedekind domain R R R . Then T ω ( W ) ≃ ⋁ S n − 1 \mathcal{T}^{\omega}(W)\simeq\bigvee S^{n-1} T ω ( W ) ≃ ⋁ S n − 1 .
Let ( W , ω ) (W,\omega) ( W , ω ) be a genus-n n n symplectic module over a field F F F . Given an ordered generalized symplectic basis β = ( a ⃗ 1 , b ⃗ 1 , … , a ⃗ n , b ⃗ n ) \beta=(\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_n,\vec{b}_n) β = ( a 1 , b 1 , … , a n , b n ) of ( W , ω ) (W,\omega) ( W , ω ) , let B ( β ) B(\beta) B ( β ) denote the full subcomplex of T ω ( W ) \mathcal{T}^{\omega}(W) T ω ( W ) on vertices corresponding to nonzero isotropic summands that are equal to the span of a subset of β \beta β . Note, unlike in the case of general linear groups, we do not take all proper nonempty subsets, as not all such subsets span isotropic subspaces. We have that B ( β ) ≅ S n − 1 B(\beta)\cong S^{n-1} B ( β ) ≅ S n − 1 with a natural orientation coming from the order. Let [ β ] [\beta] [ β ] denote the image of fundamental class of B ( β ) B(\beta) B ( β ) in St ω ( W ) \operatorname{St}^{\omega}(W) St ω ( W ) .
Let S 2 n ω ≤ S 2 n S^{\omega}_{2n}\leq S_{2n} S 2 n ω ≤ S 2 n be the subgroup of permutations such that induce permutations on the following set of subsets
{ { 1 , 2 } , { 3 , 4 } , … , { 2 n − 1 , 2 n } } . \{\{1,2\},\{3,4\},\ldots,\{2n-1,2n\}\}. {{ 1 , 2 } , { 3 , 4 } , … , { 2 n − 1 , 2 n }} .
Abstractly, S 2 n ω ≅ S 2 ≀ S n S^{\omega}_{2n}\cong S_2\wr S_n S 2 n ω ≅ S 2 ≀ S n . Let
χ : S 2 n ω → { ± 1 } \chi:S^{\omega}_{2n}\to\{\pm1\} χ : S 2 n ω → { ± 1 }
be the product of the sign representations of each S 2 S_2 S 2 and the sign representation of S n S_n S n .
The following presentation can be extracted from Pal, Corollary 1.4, Lemma 4.3, Proof of Theorem 1.5; see also Tóth [Tó05, Theorem 2].
Theorem 2.10 (Pal, Tóth). Let ( W , ω ) (W,\omega) ( W , ω ) be a genus n n n symplectic vector space over a field F F F . Then St ω ( W ) \operatorname{St}^{\omega}(W) St ω ( W ) has the following presentation. The generators are given by symplectic apartment classes [ a ⃗ 1 , b ⃗ 1 , … , a ⃗ n , b ⃗ n ] [\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_n,\vec{b}_n] [ a 1 , b 1 , … , a n , b n ] associated to ordered generalized symplectic bases and the relations are as follows:
(1) Let u 1 , … , u 2 n ∈ F × u_1,\ldots,u_{2n}\in F^{\times} u 1 , … , u 2 n ∈ F × . Then
[ a ⃗ 1 , b ⃗ 1 , … , a ⃗ n , b ⃗ n ] = [ u 1 a ⃗ 1 , u 2 b ⃗ 1 , … , u 2 n − 1 a ⃗ n , u 2 n b ⃗ n ] . [\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_n,\vec{b}_n]=[u_1\vec{a}_1,u_2\vec{b}_1,\ldots,u_{2n-1}\vec{a}_n,u_{2n}\vec{b}_n]. [ a 1 , b 1 , … , a n , b n ] = [ u 1 a 1 , u 2 b 1 , … , u 2 n − 1 a n , u 2 n b n ] . (2) Let σ ∈ S 2 n ω \sigma\in S^{\omega}_{2n} σ ∈ S 2 n ω . Then
[ v ⃗ 1 , … , v ⃗ 2 n ] = χ ( σ ) [ v ⃗ σ ( 1 ) , … , v ⃗ σ ( 2 n ) ] . [\vec{v}_1,\ldots,\vec{v}_{2n}]=\chi(\sigma)[\vec{v}_{\sigma(1)},\ldots,\vec{v}_{\sigma(2n)}]. [ v 1 , … , v 2 n ] = χ ( σ ) [ v σ ( 1 ) , … , v σ ( 2 n ) ] . (3)
[ a ⃗ 1 , b ⃗ 1 , a ⃗ 2 , b ⃗ 2 , … , a ⃗ n , b ⃗ n ] = [ a ⃗ 1 + b ⃗ 1 , b ⃗ 1 , a ⃗ 2 , b ⃗ 2 , … , a ⃗ n , b ⃗ n ] + [ a ⃗ 1 , a ⃗ 1 + b ⃗ 1 , a ⃗ 2 , b ⃗ 2 , … , a ⃗ n , b ⃗ n ] . [\vec{a}_1,\vec{b}_1,\vec{a}_2,\vec{b}_2,\ldots,\vec{a}_n,\vec{b}_n]=
[\vec{a}_1+\vec{b}_1,\vec{b}_1,\vec{a}_2,\vec{b}_2,\ldots,\vec{a}_n,\vec{b}_n]+[\vec{a}_1,\vec{a}_1+\vec{b}_1,\vec{a}_2,\vec{b}_2,\ldots,\vec{a}_n,\vec{b}_n]. [ a 1 , b 1 , a 2 , b 2 , … , a n , b n ] = [ a 1 + b 1 , b 1 , a 2 , b 2 , … , a n , b n ] + [ a 1 , a 1 + b 1 , a 2 , b 2 , … , a n , b n ] . (4)
[ a ⃗ 1 , b ⃗ 1 , a ⃗ 2 , b ⃗ 2 , … , a ⃗ n , b ⃗ n ] = [ a ⃗ 1 , b ⃗ 1 − b ⃗ 2 , a ⃗ 1 + a ⃗ 2 , b ⃗ 2 , … , a ⃗ n , b ⃗ n ] + [ b ⃗ 1 − b ⃗ 2 , a ⃗ 2 , a ⃗ 1 + a ⃗ 2 , b ⃗ 1 , … , a ⃗ n , b ⃗ n ] . [\vec{a}_1,\vec{b}_1,\vec{a}_2,\vec{b}_2,\ldots,\vec{a}_n,\vec{b}_n]=
[\vec{a}_1,\vec{b}_1-\vec{b}_2,\vec{a}_1+\vec{a}_2,\vec{b}_2,\ldots,\vec{a}_n,\vec{b}_n]+[\vec{b}_1-\vec{b}_2,\vec{a}_2,\vec{a}_1+\vec{a}_2,\vec{b}_1,\ldots,\vec{a}_n,\vec{b}_n]. [ a 1 , b 1 , a 2 , b 2 , … , a n , b n ] = [ a 1 , b 1 − b 2 , a 1 + a 2 , b 2 , … , a n , b n ] + [ b 1 − b 2 , a 2 , a 1 + a 2 , b 1 , … , a n , b n ] .
Algebraic structure
In this section, we describe monoidal structures on categories of representations of general linear groups and symplectic groups.
Monoidal structures
Definition 3.1. For R R R a ring, let VB ( R ) \operatorname{VB}(R) VB ( R ) denote the groupoid of finitely generated projective R R R -modules and isomorphisms. Let VB ω ( R ) \operatorname{VB}^{\omega}(R) VB ω ( R ) denote the groupoid of finite genus symplectic R R R -modules and form-preserving isomorphisms.
Definition 3.2. Let R R R and k k k be rings. A VB ( R ) \operatorname{VB}(R) VB ( R ) -module over k k k is a functor X : VB ( R ) → Mod k X:\operatorname{VB}(R)\to\operatorname{Mod}_k X : VB ( R ) → Mod k . Let Mod VB ( R ) ( k ) \operatorname{Mod}_{\operatorname{VB}(R)}(k) Mod VB ( R ) ( k ) denote the category of VB ( R ) \operatorname{VB}(R) VB ( R ) -modules over k k k . Similarly define VB ( R ) \operatorname{VB}(R) VB ( R ) -chain complexes and the corresponding symplectic versions.
Definition 3.3. Given two V B ( R ) \mathrm{VB}(R) VB ( R ) -modules X \mathbf{X} X and Y \mathbf{Y} Y , let X ⊗ L Y \mathbf{X}\otimes_{\mathrm{L}}\mathbf{Y} X ⊗ L Y be the V B ( R ) \mathrm{VB}(R) VB ( R ) -module with value on a projective R R R -module M M M is given by
( X ⊗ L Y ) ( M ) = ⨁ A ⊕ B = M X ( A ) ⊗ Y ( B ) . (\mathbf{X}\otimes_{\mathrm{L}}\mathbf{Y})(M)=\bigoplus_{A\oplus B=M}\mathbf{X}(A)\otimes\mathbf{Y}(B). ( X ⊗ L Y ) ( M ) = A ⊕ B = M ⨁ X ( A ) ⊗ Y ( B ) . Let X ⊗ P Y \mathbf{X}\otimes_{\mathrm{P}}\mathbf{Y} X ⊗ P Y be the V B ( R ) \mathrm{VB}(R) VB ( R ) -module with value on a projective R R R -module M M M is given by
( X ⊗ P Y ) ( M ) = ⨁ A ⊆ M X ( A ) ⊗ Y ( M / A ) . (\mathbf{X}\otimes_{\mathrm{P}}\mathbf{Y})(M)=\bigoplus_{A\subseteq M}\mathbf{X}(A)\otimes\mathbf{Y}(M/A). ( X ⊗ P Y ) ( M ) = A ⊆ M ⨁ X ( A ) ⊗ Y ( M / A ) . Here A A A runs over all summands of M M M .
The products ⊗ L \otimes_{\mathrm{L}} ⊗ L and ⊗ P \otimes_{\mathrm{P}} ⊗ P endow Mod V B ( R ) ( k ) \operatorname{Mod}_{\mathrm{VB}(R)}(k) Mod VB ( R ) ( k ) with monoidal structures. The unit of ⊗ L \otimes_{\mathrm{L}} ⊗ L and ⊗ P \otimes_{\mathrm{P}} ⊗ P are given by the V B ( R ) \mathrm{VB}(R) VB ( R ) -module
M ↦ { k , if M ≅ 0 , 0 , otherwise. M\mapsto
\begin{cases}
k, & \text{if }M\cong0,\\
0, & \text{otherwise.}
\end{cases} M ↦ { k , 0 , if M ≅ 0 , otherwise. While ⊗ L \otimes_{\mathrm{L}} ⊗ L is symmetric monoidal, ⊗ P \otimes_{\mathrm{P}} ⊗ P is not. These two structures are compatible in the sense that they form a duoidal category. This will be discussed in § 3.2 . Part of this data is an interchange natural transformation.
Definition 3.4. Let X , Y , Z , W \mathbf{X},\mathbf{Y},\mathbf{Z},\mathbf{W} X , Y , Z , W be V B ( R ) \mathrm{VB}(R) VB ( R ) -modules. Let
ζ X , Y , Z , W : ( X ⊗ P Y ) ⊗ L ( Z ⊗ P W ) ⟶ ( X ⊗ L Z ) ⊗ P ( Y ⊗ L W ) \zeta_{\mathbf{X},\mathbf{Y},\mathbf{Z},\mathbf{W}}:(\mathbf{X}\otimes_{\mathrm{P}}\mathbf{Y})\otimes_{\mathrm{L}}(\mathbf{Z}\otimes_{\mathrm{P}}\mathbf{W})\longrightarrow(\mathbf{X}\otimes_{\mathrm{L}}\mathbf{Z})\otimes_{\mathrm{P}}(\mathbf{Y}\otimes_{\mathrm{L}}\mathbf{W}) ζ X , Y , Z , W : ( X ⊗ P Y ) ⊗ L ( Z ⊗ P W ) ⟶ ( X ⊗ L Z ) ⊗ P ( Y ⊗ L W ) be the following map. For M M M a finitely generated projective R R R -module,
( ( X ⊗ P Y ) ⊗ L ( Z ⊗ P W ) ) ( M ) = ⨁ A ⊕ B = M ( X ⊗ P Y ) ( A ) ⊗ ( Z ⊗ P W ) ( B ) = ⨁ A ⊕ B = M A ′ ⊆ A , B ′ ⊆ B X ( A ′ ) ⊗ Y ( A / A ′ ) ⊗ Z ( B ′ ) ⊗ W ( B / B ′ ) . \begin{aligned}
((\mathbf{X}\otimes_{\mathrm{P}}\mathbf{Y})\otimes_{\mathrm{L}}(\mathbf{Z}\otimes_{\mathrm{P}}\mathbf{W}))(M)
&=\bigoplus_{A\oplus B=M}(\mathbf{X}\otimes_{\mathrm{P}}\mathbf{Y})(A)\otimes(\mathbf{Z}\otimes_{\mathrm{P}}\mathbf{W})(B)\\
&=\bigoplus_{\substack{A\oplus B=M\\ A'\subseteq A,\ B'\subseteq B}}\mathbf{X}(A')\otimes\mathbf{Y}(A/A')\otimes\mathbf{Z}(B')\otimes\mathbf{W}(B/B').
\end{aligned} (( X ⊗ P Y ) ⊗ L ( Z ⊗ P W )) ( M ) = A ⊕ B = M ⨁ ( X ⊗ P Y ) ( A ) ⊗ ( Z ⊗ P W ) ( B ) = A ⊕ B = M A ′ ⊆ A , B ′ ⊆ B ⨁ X ( A ′ ) ⊗ Y ( A / A ′ ) ⊗ Z ( B ′ ) ⊗ W ( B / B ′ ) . Similarly,
( ( X ⊗ L Z ) ⊗ P ( Y ⊗ L W ) ) ( M ) = ⨁ M ′ ⊆ M ( X ⊗ L Z ) ( M ′ ) ⊗ ( Y ⊗ L W ) ( M / M ′ ) = ⨁ M ′ ⊆ M C ⊕ D = M ′ , C ′ ⊕ D ′ = M / M ′ X ( C ) ⊗ Z ( D ) ⊗ Y ( C ′ ) ⊗ W ( D ′ ) . \begin{aligned}
((\mathbf{X}\otimes_{\mathrm{L}}\mathbf{Z})\otimes_{\mathrm{P}}(\mathbf{Y}\otimes_{\mathrm{L}}\mathbf{W}))(M)
&=\bigoplus_{M'\subseteq M}(\mathbf{X}\otimes_{\mathrm{L}}\mathbf{Z})(M')\otimes(\mathbf{Y}\otimes_{\mathrm{L}}\mathbf{W})(M/M')\\
&=\bigoplus_{\substack{M'\subseteq M\\ C\oplus D=M',\ C'\oplus D'=M/M'}}\mathbf{X}(C)\otimes\mathbf{Z}(D)\otimes\mathbf{Y}(C')\otimes\mathbf{W}(D').
\end{aligned} (( X ⊗ L Z ) ⊗ P ( Y ⊗ L W )) ( M ) = M ′ ⊆ M ⨁ ( X ⊗ L Z ) ( M ′ ) ⊗ ( Y ⊗ L W ) ( M / M ′ ) = M ′ ⊆ M C ⊕ D = M ′ , C ′ ⊕ D ′ = M / M ′ ⨁ X ( C ) ⊗ Z ( D ) ⊗ Y ( C ′ ) ⊗ W ( D ′ ) . A collection A , B , A ′ , B ′ A,B,A',B' A , B , A ′ , B ′ in the first indexing set determines a collection C , D , C ′ , D ′ , M ′ C,D,C',D',M' C , D , C ′ , D ′ , M ′ in the second indexing set as follows. Let M ′ = A ′ ⊕ B ′ M'=A'\oplus B' M ′ = A ′ ⊕ B ′ , C = A ′ C=A' C = A ′ , D = B ′ D=B' D = B ′ , C ′ = A / A ′ C'=A/A' C ′ = A / A ′ , D ′ = B / B ′ D'=B/B' D ′ = B / B ′ and identify M / ( A ′ ⊕ B ′ ) M/(A'\oplus B') M / ( A ′ ⊕ B ′ ) with ( A / A ′ ) ⊕ ( B / B ′ ) (A/A')\oplus(B/B') ( A / A ′ ) ⊕ ( B / B ′ ) . The identity maps A ′ → C A'\to C A ′ → C and B ′ → D B'\to D B ′ → D , the inverses of the natural isomorphism C ′ ≅ A / A ′ C'\cong A/A' C ′ ≅ A / A ′ and D ′ ≅ B / B ′ D'\cong B/B' D ′ ≅ B / B ′ , and the symmetry isomorphism Y ( C ′ ) ⊗ Z ( D ) ≅ Z ( D ) ⊗ Y ( C ′ ) \mathbf{Y}(C')\otimes\mathbf{Z}(D)\cong\mathbf{Z}(D)\otimes\mathbf{Y}(C') Y ( C ′ ) ⊗ Z ( D ) ≅ Z ( D ) ⊗ Y ( C ′ ) induce ζ X , Y , Z , W \zeta_{\mathbf{X},\mathbf{Y},\mathbf{Z},\mathbf{W}} ζ X , Y , Z , W .
Note that the symmetry isomorphism Y ( C ′ ) ⊗ Z ( D ) ≅ Z ( D ) ⊗ Y ( C ′ ) \mathbf{Y}(C')\otimes\mathbf{Z}(D)\cong\mathbf{Z}(D)\otimes\mathbf{Y}(C') Y ( C ′ ) ⊗ Z ( D ) ≅ Z ( D ) ⊗ Y ( C ′ ) includes a sign of ( − 1 ) rk ( C ′ ) rk ( D ) (-1)^{\operatorname{rk}(C')\operatorname{rk}(D)} ( − 1 ) rk ( C ′ ) rk ( D ) .
There are symplectic versions of ⊗ L ω \otimes^\omega_{\mathcal L} ⊗ L ω and ⊗ P ω \otimes^\omega_{\mathcal P} ⊗ P ω . Recall that for ( W , ω ) (W,\omega) ( W , ω ) a symplectic module and M ⊂ W M \subset W M ⊂ W , the symplectic complement
M ⊥ = { w ∈ W ∣ ω ( w , m ) = 0 for all m ∈ M } . M^\perp=\{w\in W\mid\omega(w,m)=0\text{ for all }m\in M\}. M ⊥ = { w ∈ W ∣ ω ( w , m ) = 0 for all m ∈ M } .
If M M M is symplectic, then M ⊥ M^\perp M ⊥ is symplectic and W = M ⊕ M ⊥ W=M\oplus M^\perp W = M ⊕ M ⊥ . If R R R is a Dedekind domain, then the genus of W W W is the sum of the genus of M M M and the genus of M ⊥ M^\perp M ⊥ . If M M M is isotropic M ⊂ M ⊥ M\subset M^\perp M ⊂ M ⊥ and M ⊥ / M M^\perp/M M ⊥ / M has a natural symplectic structure. If R R R is a Dedekind domain, then the genus of M ⊥ / M M^\perp/M M ⊥ / M is the genus of W W W minus rank of M M M .
Definition 3.5. Let U and V be V B ω ( R ) \mathrm{VB}^{\omega}(R) VB ω ( R ) -modules and let X be a V B ( R ) \mathrm{VB}(R) VB ( R ) -module. Let U ⊗ L ω V \mathrm{U}\otimes^\omega_{\mathcal L}\mathrm{V} U ⊗ L ω V be the V B ω ( R ) \mathrm{VB}^{\omega}(R) VB ω ( R ) -module with value on a symplectic module W W W given by
( U ⊗ L ω V ) ( W ) = ⨁ A ⊕ B = W U ( A ) ⊗ V ( B ) . (\mathrm{U}\otimes^\omega_{\mathcal L}\mathrm{V})(W)=\bigoplus_{A\oplus B=W}\mathrm{U}(A)\otimes\mathrm{V}(B). ( U ⊗ L ω V ) ( W ) = A ⊕ B = W ⨁ U ( A ) ⊗ V ( B ) . Here A A A and B B B are symplectic submodules of W W W and B = A ⊥ B=A^\perp B = A ⊥ . Mod V B ω ( R ) ( k ) \operatorname{Mod}_{\mathrm{VB}^{\omega}(R)}(\mathbb{k}) Mod VB ω ( R ) ( k ) is monoidal category with ⊗ L ω \otimes^\omega_{\mathcal L} ⊗ L ω .
Let X ⊗ P ω V \mathrm{X}\otimes^\omega_{\mathcal P}\mathrm{V} X ⊗ P ω V be the V B ( R ) ω \mathrm{VB}(R)^\omega VB ( R ) ω -module with value on a symplectic module W W W given by
( X ⊗ P ω V ) ( W ) = ⨁ J ⊂ W X ( J ) ⊗ V ( J ⊥ / J ) . (\mathrm{X}\otimes^\omega_{\mathcal P}\mathrm{V})(W)=\bigoplus_{J\subset W}\mathrm{X}(J)\otimes\mathrm{V}(J^\perp/J). ( X ⊗ P ω V ) ( W ) = J ⊂ W ⨁ X ( J ) ⊗ V ( J ⊥ / J ) .
Here J J J is an isotropic summand. The category and operation ( Mod V B ω ( R ) ( k ) , ⊗ L ω ) (\operatorname{Mod}_{\mathrm{VB}^{\omega}(R)}(\mathbb{k}),\otimes^\omega_{\mathcal L}) ( Mod VB ω ( R ) ( k ) , ⊗ L ω ) has the structure of a monoidal module category over Mod V B ( R ) ( k ) \operatorname{Mod}_{\mathrm{VB}(R)}(\mathbb{k}) Mod VB ( R ) ( k ) (sometimes called a Mod V B ( R ) ( k ) \operatorname{Mod}_{\mathrm{VB}(R)}(\mathbb{k}) Mod VB ( R ) ( k ) -actegory).
We now describe a symplectic analogue of the interchange.
Definition 3.6. Let X and Z be V B ( R ) \mathrm{VB}(R) VB ( R ) -modules and let Y and W be V B ω ( R ) \mathrm{VB}^{\omega}(R) VB ω ( R ) -modules. Let
ξ X , Y , Z , W : ( X ⊗ P ω Y ) ⊗ L ω ( Z ⊗ P ω W ) ⟶ ( X ⊗ L Z ) ⊗ P ω ( Y ⊗ L ω W ) \xi_{\mathrm{X,Y,Z,W}}\colon(\mathrm{X}\otimes^\omega_{\mathcal P}\mathrm{Y})\otimes^\omega_{\mathcal L}(\mathrm{Z}\otimes^\omega_{\mathcal P}\mathrm{W})\longrightarrow(\mathrm{X}\otimes_{\mathcal L}\mathrm{Z})\otimes^\omega_{\mathcal P}(\mathrm{Y}\otimes^\omega_{\mathcal L}\mathrm{W}) ξ X , Y , Z , W : ( X ⊗ P ω Y ) ⊗ L ω ( Z ⊗ P ω W ) ⟶ ( X ⊗ L Z ) ⊗ P ω ( Y ⊗ L ω W ) be the following map. For W W W a finitely generated symplectic R R R -module,
( ( X ⊗ P ω Y ) ⊗ L ω ( Z ⊗ P ω W ) ) ( W ) = \bigl((\mathrm{X}\otimes^\omega_{\mathcal P}\mathrm{Y})\otimes^\omega_{\mathcal L}(\mathrm{Z}\otimes^\omega_{\mathcal P}\mathrm{W})\bigr)(W)= ( ( X ⊗ P ω Y ) ⊗ L ω ( Z ⊗ P ω W ) ) ( W ) = ⨁ U ⊕ V = W ( X ⊗ P ω Y ) ( U ) ⊗ ( Z ⊗ P ω W ) ( V ) = \bigoplus_{U\oplus V=W}(\mathrm{X}\otimes^\omega_{\mathcal P}\mathrm{Y})(U)\otimes(\mathrm{Z}\otimes^\omega_{\mathcal P}\mathrm{W})(V)= U ⊕ V = W ⨁ ( X ⊗ P ω Y ) ( U ) ⊗ ( Z ⊗ P ω W ) ( V ) = ⨁ U ⊕ V = W J ⊂ U , K ⊂ V X ( J ) ⊗ Y ( J ⊥ U / J ) ⊗ Z ( K ) ⊗ W ( K ⊥ V / K ) . \bigoplus_{\substack{U\oplus V=W\\J\subset U,K\subset V}}\mathrm{X}(J)\otimes\mathrm{Y}(J^{\perp_U}/J)\otimes\mathrm{Z}(K)\otimes\mathrm{W}(K^{\perp_V}/K). U ⊕ V = W J ⊂ U , K ⊂ V ⨁ X ( J ) ⊗ Y ( J ⊥ U / J ) ⊗ Z ( K ) ⊗ W ( K ⊥ V / K ) . Here U U U , V V V are symplectic, J J J , K K K are isotropic, and J ⊥ U J^{\perp_U} J ⊥ U and K ⊥ V K^{\perp_V} K ⊥ V are the complements of J J J in U U U and K K K in V V V , respectively. Similarly,
( ( X ⊗ L Z ) ⊗ P ω ( Y ⊗ L ω W ) ) ( W ) = \bigl((\mathrm{X}\otimes_{\mathcal L}\mathrm{Z})\otimes^\omega_{\mathcal P}(\mathrm{Y}\otimes^\omega_{\mathcal L}\mathrm{W})\bigr)(W)= ( ( X ⊗ L Z ) ⊗ P ω ( Y ⊗ L ω W ) ) ( W ) = ⨁ Q ⊂ W ( X ⊗ L Z ) ( Q ) ⊗ ( Y ⊗ L ω W ) ( Q ⊥ / Q ) = \bigoplus_{Q\subset W}(\mathrm{X}\otimes_{\mathcal L}\mathrm{Z})(Q)\otimes(\mathrm{Y}\otimes^\omega_{\mathcal L}\mathrm{W})(Q^\perp/Q)= Q ⊂ W ⨁ ( X ⊗ L Z ) ( Q ) ⊗ ( Y ⊗ L ω W ) ( Q ⊥ / Q ) = ⨁ Q ⊂ W C ⊕ D = Q , C ′ ⊕ D ′ = Q ⊥ / Q X ( C ) ⊗ Z ( D ) ⊗ Y ( C ′ ) ⊗ W ( D ′ ) . \bigoplus_{\substack{Q\subset W\\C\oplus D=Q,\ C'\oplus D'=Q^\perp/Q}}\mathrm{X}(C)\otimes\mathrm{Z}(D)\otimes\mathrm{Y}(C')\otimes\mathrm{W}(D'). Q ⊂ W C ⊕ D = Q , C ′ ⊕ D ′ = Q ⊥ / Q ⨁ X ( C ) ⊗ Z ( D ) ⊗ Y ( C ′ ) ⊗ W ( D ′ ) . Here Q Q Q is isotropic and C ′ C' C ′ , D ′ D' D ′ are symplectic. A collection U U U , V V V , J J J , K K K in the first indexing set determines a collection Q Q Q , C C C , D D D , C ′ C' C ′ , D ′ D' D ′ in the second indexing set as follows. Let Q = J ⊕ K Q=J\oplus K Q = J ⊕ K ,
C = J C=J C = J , D = K D=K D = K , C ′ = J ⊥ U / J C'=J^{\perp_U}/J C ′ = J ⊥ U / J and D ′ = K ⊥ V / K D'=K^{\perp_V}/K D ′ = K ⊥ V / K . Note that J ⊥ U / J ⊕ K ⊥ V / K = Q ⊥ / Q J^{\perp_U}/J\oplus K^{\perp_V}/K=Q^\perp/Q J ⊥ U / J ⊕ K ⊥ V / K = Q ⊥ / Q , where Q ⊥ Q^\perp Q ⊥ is the complement of Q Q Q in W W W . The identity maps J → C J\to C J → C , K → D K\to D K → D , the projection isomorphisms J ⊥ U → C ′ J^{\perp_U}\to C' J ⊥ U → C ′ , K ⊥ V / K → D ′ K^{\perp_V}/K\to D' K ⊥ V / K → D ′ , and the symmetry Y ( C ′ ) ⊗ Z ( D ) → Z ( D ) ⊗ Y ( C ′ ) Y(C')\otimes Z(D)\to Z(D)\otimes Y(C') Y ( C ′ ) ⊗ Z ( D ) → Z ( D ) ⊗ Y ( C ′ ) induce ξ X , Y , Z , W \xi_{X,Y,Z,W} ξ X , Y , Z , W .
Again, the symmetry isomorphism Y ( C ′ ) ⊗ Z ( D ) → Z ( D ) ⊗ Y ( C ′ ) Y(C')\otimes Z(D)\to Z(D)\otimes Y(C') Y ( C ′ ) ⊗ Z ( D ) → Z ( D ) ⊗ Y ( C ′ ) includes a sign of ( − 1 ) genus ( C ′ ) rk ( D ) (-1)^{\operatorname{genus}(C')\operatorname{rk}(D)} ( − 1 ) genus ( C ′ ) rk ( D ) .
Duoidal compatibility
Duoidal categories provide a framework for describing the relationship between P \mathfrak{P} P and L \mathfrak{L} L .
Definition 3.7. A duoidal category is a category C \mathcal{C} C equipped with two monoidal structures ( C , ⋄ , I ) (\mathcal{C},\diamond,I) ( C , ⋄ , I ) and ( C , ⋆ , J ) (\mathcal{C},\star,J) ( C , ⋆ , J ) , such that the functor ⋆ : C × C → C \star:\mathcal{C}\times\mathcal{C}\to\mathcal{C} ⋆ : C × C → C and the object J : 1 → C J:1\to\mathcal{C} J : 1 → C are lax monoidal functors with respect to the monoidal structure ( C , ⋄ , I ) (\mathcal{C},\diamond,I) ( C , ⋄ , I ) .
The data of a duoidal category is the following:
An interchange natural transformation:
ζ X , Y , Z , W : ( X ⋆ Y ) ⋄ ( Z ⋆ W ) ⟶ ( X ⋄ Z ) ⋆ ( Y ⋄ W ) \zeta_{X,Y,Z,W}:(X\star Y)\diamond(Z\star W)\longrightarrow(X\diamond Z)\star(Y\diamond W) ζ X , Y , Z , W : ( X ⋆ Y ) ⋄ ( Z ⋆ W ) ⟶ ( X ⋄ Z ) ⋆ ( Y ⋄ W ) A morphism for the unit of ⋄ \diamond ⋄ :
Δ I : I ⟶ I ⋆ I \Delta_I:I\longrightarrow I\star I Δ I : I ⟶ I ⋆ I A morphism for the unit of ⋆ \star ⋆ :
μ J : J ⋄ J ⟶ J \mu_J:J\diamond J\longrightarrow J μ J : J ⋄ J ⟶ J A morphism relating the two units:
δ : I ⟶ J \delta:I\longrightarrow J δ : I ⟶ J
These morphisms satisfy compatibility axioms. We will often omit units and counits from the notation.
See Kupers–Rudenko–Sierra [9 Section 2.2] for a proof that ( ModVB ( R ) ( k ) , ⊗ P , ⊗ L ) (\operatorname{ModVB}(R)(\Bbbk),\otimes_{\mathfrak{P}},\otimes_{\mathfrak{L}}) ( ModVB ( R ) ( k ) , ⊗ P , ⊗ L ) is a duoidal category in the case R R R is a field. A similar argument works for Dedekind domains.
Definition 3.8. A duoidal bialgebra in a duoidal category ( C , ⋆ , J , ⋄ , I , ζ ) (\mathcal{C},\star,J,\diamond,I,\zeta) ( C , ⋆ , J , ⋄ , I , ζ ) is an object B B B , a monoid structure ( B , μ , ν ) (B,\mu,\nu) ( B , μ , ν ) on B B B with respect to ⋄ \diamond ⋄ , and a comonoid structure ( B , Δ , ϵ ) (B,\Delta,\epsilon) ( B , Δ , ϵ ) with respect to ⋆ \star ⋆ such that the following diagrams commutes:
1.
B ⋄ B → Δ ⋄ Δ ( B ⋆ B ) ⋄ ( B ⋆ B ) μ ↓ ↓ ( μ ⋆ μ ) ∘ ζ B → Δ B ⋆ B \begin{CD}
B\diamond B @>{\Delta\diamond\Delta}>> (B\star B)\diamond(B\star B) \\
@V{\mu}VV @VV{(\mu\star\mu)\circ\zeta}V \\
B @>{\Delta}>> B\star B
\end{CD} B ⋄ B μ ↓ ⏐ B Δ ⋄ Δ Δ ( B ⋆ B ) ⋄ ( B ⋆ B ) ↓ ⏐ ( μ ⋆ μ ) ∘ ζ B ⋆ B 2.
I ← μ I I ⋄ I ν ↓ ↓ ν ⋄ ν B ← μ B ⋄ B \begin{CD}
I @<{\mu_I}<< I\diamond I \\
@V{\nu}VV @VV{\nu\diamond\nu}V \\
B @<{\mu}<< B\diamond B
\end{CD} I ν ↓ ⏐ B μ I μ I ⋄ I ↓ ⏐ ν ⋄ ν B ⋄ B 3.
J → Δ J J ⋆ J ϵ ↑ ↑ ϵ ⋆ ϵ B → Δ B ⋆ B \begin{CD}
J @>{\Delta_J}>> J\star J \\
@A{\epsilon}AA @AA{\epsilon\star\epsilon}A \\
B @>{\Delta}>> B\star B
\end{CD} J ϵ ⏐ ↑ B Δ J Δ J ⋆ J ⏐ ↑ ϵ ⋆ ϵ B ⋆ B
I → ν B δ ↓ ↑ ε J = J (1) \begin{CD}
I @>{\nu}>> B \\
@V{\delta}VV @AA{\varepsilon}A \\
J @= J \tag*{(1)}
\end{CD} I δ ↓ ⏐ J ν B ⏐ ↑ ε J (1)
Definition 3.9. A duoidal module category over a duoidal category ( C , ⋆ c , J c , ⋄ c , I c ) (\mathcal{C},\star^{c},J^{c},\diamond^{c},I^{c}) ( C , ⋆ c , J c , ⋄ c , I c ) is a monoidal category ( D , ⋄ d , I d ) (\mathcal{D},\diamond^{d},I^{d}) ( D , ⋄ d , I d ) , a left ( C , ⋆ c , J c ) (\mathcal{C},\star^{c},J^{c}) ( C , ⋆ c , J c ) -module structure ⋆ d : C × D → D \star^{d}:\mathcal{C}\times\mathcal{D}\to\mathcal{D} ⋆ d : C × D → D such that the functor ⋆ d : C × D → D \star^{d}:\mathcal{C}\times\mathcal{D}\to\mathcal{D} ⋆ d : C × D → D is lax monoidal with respect to the monoidal structure coming from ( C , ⋄ c , I c ) (\mathcal{C},\diamond^{c},I^{c}) ( C , ⋄ c , I c ) and ( D , ⋄ d , I d ) (\mathcal{D},\diamond^{d},I^{d}) ( D , ⋄ d , I d ) .
Explicitly, this means there exist the following structural morphisms:
(1) An interchange natural transformation:
ξ X , Y , Z , W : ( X ⋆ d Y ) ⋄ d ( Z ⋆ d W ) ⟶ ( X ⋄ c Z ) ⋆ d ( Y ⋄ d W ) \xi_{X,Y,Z,W}:(X\star^{d}Y)\diamond^{d}(Z\star^{d}W)\longrightarrow(X\diamond^{c}Z)\star^{d}(Y\diamond^{d}W) ξ X , Y , Z , W : ( X ⋆ d Y ) ⋄ d ( Z ⋆ d W ) ⟶ ( X ⋄ c Z ) ⋆ d ( Y ⋄ d W ) (2) A morphism for the unit of ⋄ d \diamond^{d} ⋄ d :
Δ I : I d ⟶ I c ⋆ d I d \Delta_{I}:I^{d}\longrightarrow I^{c}\star^{d}I^{d} Δ I : I d ⟶ I c ⋆ d I d These structural morphisms must satisfy a set of coherence axioms ensuring compatibility with the associativity and unitality constraints of the monoidal and module structures.
A similar argument to that of Kupers–Rudenko–Sierra [9 ] gives the following.
Proposition 3.10. For R R R a Dedekind domain, ( Mod V B ω ( R ) ( k ) , P , L ) (\operatorname{Mod}_{\mathrm{VB}^{\omega}(R)}(\mathsf{k}),\mathcal{P},\mathcal{L}) ( Mod VB ω ( R ) ( k ) , P , L ) is a duoidal module category over ( Mod V B ( R ) ( k ) , P , L ) (\operatorname{Mod}_{\mathrm{VB}(R)}(\mathsf{k}),\mathcal{P},\mathcal{L}) ( Mod VB ( R ) ( k ) , P , L ) .
Definition 3.11. Let ( D , ⋄ d , I d , ξ ) (\mathcal{D},\diamond^{d},I^{d},\xi) ( D , ⋄ d , I d , ξ ) be a duoidal module category over a duoidal category ( C , ⋄ c , I c , ⋆ c , J c ) (\mathcal{C},\diamond^{c},I^{c},\star^{c},J^{c}) ( C , ⋄ c , I c , ⋆ c , J c ) . Let ( B , μ B , Δ B , ε B , ν B ) (B,\mu_{B},\Delta_{B},\varepsilon_{B},\nu_{B}) ( B , μ B , Δ B , ε B , ν B ) be a duoidal bialgebra in ( C , ⋄ c , I c , ⋆ c , J c ) (\mathcal{C},\diamond^{c},I^{c},\star^{c},J^{c}) ( C , ⋄ c , I c , ⋆ c , J c ) . Let ( A , μ A , ν A ) (A,\mu_{A},\nu_{A}) ( A , μ A , ν A ) be a monoid object in ( D , ⋄ d , I d ) (\mathcal{D},\diamond^{d},I^{d}) ( D , ⋄ d , I d ) with a comodule structure map Δ A : A → B ⋆ d A \Delta_{A}:A\to B\star^{d}A Δ A : A → B ⋆ d A . Then A A A is called a duoidal comodule algebra if the following diagrams commute:
(1)
A ⋄ d A → Δ A ⋄ Δ A ( B ⋆ d A ) ⋄ d ( B ⋆ d A ) μ A ↓ ↓ ( μ B ⋆ d μ A ) ∘ ξ A → Δ A B ⋆ d A \begin{CD}
A\diamond^{d}A @>{\Delta_{A}\diamond\Delta_{A}}>> (B\star^{d}A)\diamond^{d}(B\star^{d}A) \\
@V{\mu_{A}}VV @VV{(\mu_{B}\star^{d}\mu_{A})\circ\xi}V \\
A @>{\Delta_{A}}>> B\star^{d}A
\end{CD} A ⋄ d A μ A ↓ ⏐ A Δ A ⋄ Δ A Δ A ( B ⋆ d A ) ⋄ d ( B ⋆ d A ) ↓ ⏐ ( μ B ⋆ d μ A ) ∘ ξ B ⋆ d A (2)
I d → Δ I d I d ⋆ d I d η A ↓ ↓ ν B ⋆ d ν A A → Δ A B ⋆ d A \begin{CD}
I^{d} @>{\Delta_{I}^{d}}>> I^{d}\star^{d}I^{d} \\
@V{\eta_{A}}VV @VV{\nu_{B}\star^{d}\nu_{A}}V \\
A @>{\Delta_{A}}>> B\star^{d}A
\end{CD} I d η A ↓ ⏐ A Δ I d Δ A I d ⋆ d I d ↓ ⏐ ν B ⋆ d ν A B ⋆ d A
Algebraic Structure of Steinberg modules
In this section, we will prove that symplectic Steinberg modules are a duoidal comodule algebra and their homology is a comodule algebra.
Structure maps
We first recall the product and coproduct structure on general linear group Steinberg modules. Then we extend these constructions to symplectic Steinberg modules.
Definition 4.1. Let St \operatorname{St} St denote the V B ( R ) \mathrm{VB}(R) VB ( R ) -module with value on M M M given by St ( M ) \operatorname{St}(M) St ( M ) . Let St ω \operatorname{St}^{\omega} St ω denote the V B ω ( R ) \mathrm{VB}^{\omega}(R) VB ω ( R ) -module with value on W W W given by St ω ( W ) \operatorname{St}^{\omega}(W) St ω ( W ) .
For VB-modules X , Y , Z \mathbf{X},\mathbf{Y},\mathbf{Z} X , Y , Z , note that a morphism
f : X ⊗ L Y ⟶ Z f:\mathbf{X}\otimes_L\mathbf{Y}\longrightarrow\mathbf{Z} f : X ⊗ L Y ⟶ Z
is determined by the data of GL ( M ) × GL ( N ) \operatorname{GL}(M)\times\operatorname{GL}(N) GL ( M ) × GL ( N ) -equivariant linear maps
f M , N : X ( M ) ⊗ Y ( N ) ⟶ Z ( M ⊕ N ) . f_{M,N}:\mathbf{X}(M)\otimes\mathbf{Y}(N)\longrightarrow\mathbf{Z}(M\oplus N). f M , N : X ( M ) ⊗ Y ( N ) ⟶ Z ( M ⊕ N ) .
Definition 4.2. Let μ : St ⊗ L St ⟶ St \mu:\operatorname{St}\otimes_L\operatorname{St}\longrightarrow\operatorname{St} μ : St ⊗ L St ⟶ St have structure maps
μ M , N : St ( M ) ⊗ St ( N ) ⟶ St ( M ⊕ N ) \mu_{M,N}:\operatorname{St}(M)\otimes\operatorname{St}(N)\longrightarrow\operatorname{St}(M\oplus N) μ M , N : St ( M ) ⊗ St ( N ) ⟶ St ( M ⊕ N ) [ v ⃗ 1 , … , v ⃗ p ] ⊗ [ w ⃗ 1 , … , w ⃗ q ] ⟼ [ v ⃗ 1 , … , v ⃗ p , w ⃗ 1 , … , w ⃗ q ] . [\vec{v}_1,\ldots,\vec{v}_p]\otimes[\vec{w}_1,\ldots,\vec{w}_q]\longmapsto[\vec{v}_1,\ldots,\vec{v}_p,\vec{w}_1,\ldots,\vec{w}_q]. [ v 1 , … , v p ] ⊗ [ w 1 , … , w q ] ⟼ [ v 1 , … , v p , w 1 , … , w q ] . Here v ⃗ 1 , … , v ⃗ p \vec{v}_1,\ldots,\vec{v}_p v 1 , … , v p is a basis of M ⊗ R F M\otimes_R F M ⊗ R F and w ⃗ 1 , … , w ⃗ q \vec{w}_1,\ldots,\vec{w}_q w 1 , … , w q is a basis of N ⊗ R F N\otimes_R F N ⊗ R F .
See [12 ] , [1 ] , and Kupers–Rudenko–Sierra [9 ] for a proof that μ \mu μ is well defined, associative and graded-commutative. Given a k \mathbb{k} k -module T T T , let T [ 0 ] T[0] T [ 0 ] denote the VB( R ) (R) ( R ) -module with value T T T on rank-0 R R R -modules and value 0 0 0 otherwise. Note that k [ 0 ] \mathbb{k}[0] k [ 0 ] is the unit of Mod V B ( R ) ( k ) \operatorname{Mod}_{\mathrm{VB}(R)}(\mathbb{k}) Mod VB ( R ) ( k ) . The unit of St \operatorname{St} St is the map k → St \mathbb{k}\to\operatorname{St} k → St induced by the isomorphism k → St ( 0 ) \mathbb{k}\to\operatorname{St}(0) k → St ( 0 ) .
For VB-modules X , Y , Z \mathbf{X},\mathbf{Y},\mathbf{Z} X , Y , Z , note that a morphism
f : X ⟶ Y ⊗ P Z f:\mathbf{X}\longrightarrow\mathbf{Y}\otimes_P\mathbf{Z} f : X ⟶ Y ⊗ P Z
is the same data as linear maps
f M ′ ⊆ M : X ( M ) ⟶ Y ( M ′ ) ⊗ Z ( M / M ′ ) f_{M'\subseteq M}:\mathbf{X}(M)\longrightarrow\mathbf{Y}(M')\otimes\mathbf{Z}(M/M') f M ′ ⊆ M : X ( M ) ⟶ Y ( M ′ ) ⊗ Z ( M / M ′ )
which are equivariant with respect to the parabolic subgroup that is the stabilizer of M ′ M' M ′ in GL ( M ) \operatorname{GL}(M) GL ( M ) and which are zero for all but finitely many summands M ′ M' M ′ of M M M .
Definition 4.3. Let Δ : St ⟶ St ⊗ P St \Delta:\operatorname{St}\longrightarrow\operatorname{St}\otimes_P\operatorname{St} Δ : St ⟶ St ⊗ P St have structure maps
Δ M ′ ⊆ M : St ( M ) ⟶ St ( M ′ ) ⊗ St ( M / M ′ ) \Delta_{M'\subseteq M}:\operatorname{St}(M)\longrightarrow\operatorname{St}(M')\otimes\operatorname{St}(M/M') Δ M ′ ⊆ M : St ( M ) ⟶ St ( M ′ ) ⊗ St ( M / M ′ ) given by the following formula. Let v ⃗ 1 , … , v ⃗ n \vec{v}_1,\ldots,\vec{v}_n v 1 , … , v n be a basis of M ⊗ R F M\otimes_R F M ⊗ R F . If there is a shuffle σ ∈ S n \sigma\in S_n σ ∈ S n and m ≤ n m\leq n m ≤ n with
M ′ = span ( v ⃗ σ ( 1 ) , … , v ⃗ σ ( m ) ) ∩ M , M'=\operatorname{span}(\vec{v}_{\sigma(1)},\ldots,\vec{v}_{\sigma(m)})\cap M, M ′ = span ( v σ ( 1 ) , … , v σ ( m ) ) ∩ M , then let
Δ M ′ ⊆ M ( [ v ⃗ 1 , … , v ⃗ n ] ) = sgn ( σ ) [ v ⃗ σ ( 1 ) , … , v ⃗ σ ( m ) ] ⊗ [ v ⃗ σ ( m + 1 ) , … , v ⃗ σ ( n ) ] . \Delta_{M'\subseteq M}([\vec{v}_1,\ldots,\vec{v}_n])=\operatorname{sgn}(\sigma)[\vec{v}_{\sigma(1)},\ldots,\vec{v}_{\sigma(m)}]\otimes[\vec{v}_{\sigma(m+1)},\ldots,\vec{v}_{\sigma(n)}]. Δ M ′ ⊆ M ([ v 1 , … , v n ]) = sgn ( σ ) [ v σ ( 1 ) , … , v σ ( m ) ] ⊗ [ v σ ( m + 1 ) , … , v σ ( n ) ] . Here we view v ⃗ σ ( m + 1 ) , … , v ⃗ σ ( n ) \vec{v}_{\sigma(m+1)},\ldots,\vec{v}_{\sigma(n)} v σ ( m + 1 ) , … , v σ ( n ) as vectors in ( M / M ′ ) ⊗ R F (M/M')\otimes_R F ( M / M ′ ) ⊗ R F . Otherwise, let
Δ M ′ ⊆ M ( [ v ⃗ 1 , … , v ⃗ n ] ) = 0. \Delta_{M'\subseteq M}([\vec{v}_1,\ldots,\vec{v}_n])=0. Δ M ′ ⊆ M ([ v 1 , … , v n ]) = 0.
See [1 ] and Kupers–Rudenko–Sierra [9 ] for a proof that Δ \Delta Δ is well defined and coassociative. We now describe symplectic versions.
For V B ω \mathrm{VB}^{\omega} VB ω -modules X , Y , Z \mathbf{X},\mathbf{Y},\mathbf{Z} X , Y , Z , note that a morphism
f : X ⊗ L ω Y ⟶ Z f:\mathbf{X}\otimes_L^\omega\mathbf{Y}\longrightarrow\mathbf{Z} f : X ⊗ L ω Y ⟶ Z
is the same data as Sp ( W ) × Sp ( V ) \operatorname{Sp}(W)\times\operatorname{Sp}(V) Sp ( W ) × Sp ( V ) -equivariant linear maps
f W , V : X ( W ) ⊗ Y ( V ) ⟶ Z ( W ⊕ V ) . f_{W,V}:\mathbf{X}(W)\otimes\mathbf{Y}(V)\longrightarrow\mathbf{Z}(W\oplus V). f W , V : X ( W ) ⊗ Y ( V ) ⟶ Z ( W ⊕ V ) .
Definition 4.4. Let μ : St ω ⊗ ω St ω → St ω \mu:\operatorname{St}^{\omega}\otimes_{\omega}\operatorname{St}^{\omega}\to\operatorname{St}^{\omega} μ : St ω ⊗ ω St ω → St ω have structure maps
μ W , V : St ω ( W ) ⊗ St ω ( V ) → St ω ( W ⊕ V ) \mu_{W,V}:\operatorname{St}^{\omega}(W)\otimes\operatorname{St}^{\omega}(V)\to\operatorname{St}^{\omega}(W\oplus V) μ W , V : St ω ( W ) ⊗ St ω ( V ) → St ω ( W ⊕ V ) [ a ⃗ 1 , b ⃗ 1 , … , a ⃗ p , b ⃗ p ] ⊗ [ c ⃗ 1 , d ⃗ 1 , … , c ⃗ q , d ⃗ q ] ↦ [ a ⃗ 1 , b ⃗ 1 , … , a ⃗ p , b ⃗ p , c ⃗ 1 , d ⃗ 1 , … , c ⃗ q , d ⃗ q ] . [\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_p,\vec{b}_p]\otimes[\vec{c}_1,\vec{d}_1,\ldots,\vec{c}_q,\vec{d}_q]\mapsto[\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_p,\vec{b}_p,\vec{c}_1,\vec{d}_1,\ldots,\vec{c}_q,\vec{d}_q]. [ a 1 , b 1 , … , a p , b p ] ⊗ [ c 1 , d 1 , … , c q , d q ] ↦ [ a 1 , b 1 , … , a p , b p , c 1 , d 1 , … , c q , d q ] . Here a ⃗ 1 , b ⃗ 1 , … , a ⃗ p , b ⃗ p \vec{a}_1,\vec{b}_1,\ldots,\vec{a}_p,\vec{b}_p a 1 , b 1 , … , a p , b p is a generalized symplectic basis of W ⊗ R F W\otimes_R F W ⊗ R F and c ⃗ 1 , d ⃗ 1 , … , c ⃗ q , d ⃗ q \vec{c}_1,\vec{d}_1,\ldots,\vec{c}_q,\vec{d}_q c 1 , d 1 , … , c q , d q is a generalized symplectic basis of V ⊗ R F V\otimes_R F V ⊗ R F .
We now check that this is well defined.
Proposition 4.5. The map μ W , V : St ω ( W ) ⊗ St ω ( V ) → St ω ( W ⊕ V ) \mu_{W,V}:\operatorname{St}^{\omega}(W)\otimes\operatorname{St}^{\omega}(V)\to\operatorname{St}^{\omega}(W\oplus V) μ W , V : St ω ( W ) ⊗ St ω ( V ) → St ω ( W ⊕ V ) is well defined.
Proof. Let a ⃗ 1 , b ⃗ 1 , … , a ⃗ p , b ⃗ p \vec{a}_1,\vec{b}_1,\ldots,\vec{a}_p,\vec{b}_p a 1 , b 1 , … , a p , b p be a symplectic basis of W ⊗ R F W\otimes_R F W ⊗ R F and c ⃗ 1 , d ⃗ 1 , … , c ⃗ q , d ⃗ q \vec{c}_1,\vec{d}_1,\ldots,\vec{c}_q,\vec{d}_q c 1 , d 1 , … , c q , d q be a symplectic basis of V ⊗ R F V\otimes_R F V ⊗ R F . If you apply one of the relations in Theorem 2.10 to [ a ⃗ 1 , b ⃗ 1 , … , a ⃗ p , b ⃗ p ] [\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_p,\vec{b}_p] [ a 1 , b 1 , … , a p , b p ] or [ c ⃗ 1 , d ⃗ 1 , … , c ⃗ q , d ⃗ q ] [\vec{c}_1,\vec{d}_1,\ldots,\vec{c}_q,\vec{d}_q] [ c 1 , d 1 , … , c q , d q ] , then you can undo that by performing the corresponding relation (and possibly also relation (2)) to [ a ⃗ 1 , b ⃗ 1 , … , a ⃗ p , b ⃗ p , c ⃗ 1 , d ⃗ 1 , … , c ⃗ q , d ⃗ q ] [\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_p,\vec{b}_p,\vec{c}_1,\vec{d}_1,\ldots,\vec{c}_q,\vec{d}_q] [ a 1 , b 1 , … , a p , b p , c 1 , d 1 , … , c q , d q ] . □
Proposition 4.6. ( St ω , μ ) (\operatorname{St}^{\omega},\mu) ( St ω , μ ) is a unital graded-commutative monoid object in ( Mod V B ω ( R ) , ⊗ ω ) (\operatorname{Mod}_{\mathrm{VB}^{\omega}}(R),\otimes^{\omega}) ( Mod VB ω ( R ) , ⊗ ω ) .
Proof. The image of the unit map is the empty apartment class and concatenation with the empty apartment class is the identity operation. Thus St ω \operatorname{St}^{\omega} St ω is unital. Associativity follows from the fact that
μ ( μ ( [ a ⃗ 1 , … , a ⃗ 2 n ] , [ b ⃗ 1 , … , b ⃗ 2 m ] ) , [ c ⃗ 1 , … , c ⃗ 2 k ] ) \mu\left(\mu([\vec{a}_1,\ldots,\vec{a}_{2n}],[\vec{b}_1,\ldots,\vec{b}_{2m}]),[\vec{c}_1,\ldots,\vec{c}_{2k}]\right) μ ( μ ([ a 1 , … , a 2 n ] , [ b 1 , … , b 2 m ]) , [ c 1 , … , c 2 k ] ) = [ a ⃗ 1 , … , a ⃗ 2 n , b ⃗ 1 , … , b ⃗ 2 m , c ⃗ 1 , … , c ⃗ 2 k ] =[\vec{a}_1,\ldots,\vec{a}_{2n},\vec{b}_1,\ldots,\vec{b}_{2m},\vec{c}_1,\ldots,\vec{c}_{2k}] = [ a 1 , … , a 2 n , b 1 , … , b 2 m , c 1 , … , c 2 k ] = μ ( [ a ⃗ 1 , … , a ⃗ 2 n ] , μ ( [ b ⃗ 1 , … , b ⃗ 2 m ] , [ c ⃗ 1 , … , c ⃗ 2 k ] ) ) . =\mu\left([\vec{a}_1,\ldots,\vec{a}_{2n}],\mu([\vec{b}_1,\ldots,\vec{b}_{2m}],[\vec{c}_1,\ldots,\vec{c}_{2k}])\right). = μ ( [ a 1 , … , a 2 n ] , μ ([ b 1 , … , b 2 m ] , [ c 1 , … , c 2 k ]) ) . Graded commutativity follows from the fact that
μ ( [ a ⃗ 1 , … , a ⃗ 2 n ] , [ b ⃗ 1 , … , b ⃗ 2 m ] ) = ( − 1 ) n m μ ( [ b ⃗ 1 , … , b ⃗ 2 m ] , [ a ⃗ 1 , … , a ⃗ 2 n ] ) . \mu([\vec{a}_1,\ldots,\vec{a}_{2n}],[\vec{b}_1,\ldots,\vec{b}_{2m}])=(-1)^{nm}\mu([\vec{b}_1,\ldots,\vec{b}_{2m}],[\vec{a}_1,\ldots,\vec{a}_{2n}]). μ ([ a 1 , … , a 2 n ] , [ b 1 , … , b 2 m ]) = ( − 1 ) nm μ ([ b 1 , … , b 2 m ] , [ a 1 , … , a 2 n ]) . □
For V B ω \mathrm{VB}^{\omega} VB ω -modules X , Z \mathbf{X},\mathbf{Z} X , Z , and a V B \mathrm{VB} VB -module Y \mathbf{Y} Y , note that a morphism
f : X → Y ⊗ P ω Z f:\mathbf{X}\to\mathbf{Y}\otimes_{\mathcal{P}}^{\omega}\mathbf{Z} f : X → Y ⊗ P ω Z
is determined by linear maps
f J ⊂ W : X ( W ) → Y ( J ) ⊗ Z ( J ⊥ / J ) f_{J\subset W}:\mathbf{X}(W)\to\mathbf{Y}(J)\otimes\mathbf{Z}(J^{\perp}/J) f J ⊂ W : X ( W ) → Y ( J ) ⊗ Z ( J ⊥ / J )
which are equivariant with respect to the parabolic subgroup that is the stabilizer of J J J in Sp ( W ) \operatorname{Sp}(W) Sp ( W ) and which are zero for all but finitely many isotropic summands J J J of W W W .
Definition 4.7. Let Δ : St ω → St ⊗ P ω St ω \Delta:\operatorname{St}^{\omega}\to\operatorname{St}\otimes_{\mathcal{P}}^{\omega}\operatorname{St}^{\omega} Δ : St ω → St ⊗ P ω St ω have structure maps associated to an isotropic J J J
Δ J ⊂ W : St ω ( W ) → St ( J ) ⊗ St ω ( J ⊥ / J ) \Delta_{J\subset W}:\operatorname{St}^{\omega}(W)\to\operatorname{St}(J)\otimes\operatorname{St}^{\omega}(J^{\perp}/J) Δ J ⊂ W : St ω ( W ) → St ( J ) ⊗ St ω ( J ⊥ / J ) given by the following formula. Let ( v ⃗ 1 , … , v ⃗ 2 n ) (\vec{v}_1,\ldots,\vec{v}_{2n}) ( v 1 , … , v 2 n ) be an ordered symplectic basis of W ⊗ R F W\otimes_R F W ⊗ R F . If there is no σ ∈ S 2 n ω \sigma\in S_{2n}^{\omega} σ ∈ S 2 n ω and m ≤ n m\le n m ≤ n with
J = span ( v ⃗ σ ( 1 ) , v ⃗ σ ( 3 ) , … , v ⃗ σ ( 2 m − 1 ) ) ∩ W , J=\operatorname{span}(\vec{v}_{\sigma(1)},\vec{v}_{\sigma(3)},\ldots,\vec{v}_{\sigma(2m-1)})\cap W, J = span ( v σ ( 1 ) , v σ ( 3 ) , … , v σ ( 2 m − 1 ) ) ∩ W , then let
Δ J ⊂ W ( [ v ⃗ 1 , … , v ⃗ 2 n ] ) = 0. \Delta_{J\subset W}([\vec{v}_1,\ldots,\vec{v}_{2n}])=0. Δ J ⊂ W ([ v 1 , … , v 2 n ]) = 0.
Otherwise, fix such a σ \sigma σ . For i > 2 m i>2m i > 2 m , view v ⃗ σ ( i ) \vec{v}_{\sigma(i)} v σ ( i ) as an element of the quotient space J ⊥ / J J^{\perp}/J J ⊥ / J . Let
Δ J ⊂ W ( [ v ⃗ 1 , … , v ⃗ 2 n ] ) = χ ( σ ) [ v ⃗ σ ( 1 ) , v ⃗ σ ( 3 ) , … , v ⃗ σ ( 2 m − 1 ) ] ⊗ [ v ⃗ σ ( 2 m + 1 ) , v ⃗ σ ( 2 m + 2 ) , … , v ⃗ σ ( 2 n − 1 ) , v ⃗ σ ( 2 n ) ] . \Delta_{J\subset W}\left([\vec{v}_1,\ldots,\vec{v}_{2n}]\right)=\chi(\sigma)[\vec{v}_{\sigma(1)},\vec{v}_{\sigma(3)},\ldots,\vec{v}_{\sigma(2m-1)}]\otimes[\vec{v}_{\sigma(2m+1)},\vec{v}_{\sigma(2m+2)},\ldots,\vec{v}_{\sigma(2n-1)},\vec{v}_{\sigma(2n)}]. Δ J ⊂ W ( [ v 1 , … , v 2 n ] ) = χ ( σ ) [ v σ ( 1 ) , v σ ( 3 ) , … , v σ ( 2 m − 1 ) ] ⊗ [ v σ ( 2 m + 1 ) , v σ ( 2 m + 2 ) , … , v σ ( 2 n − 1 ) , v σ ( 2 n ) ] .
Note that this equation is independent of which σ ∈ S 2 n J \sigma\in S_{2n}^{J} σ ∈ S 2 n J satisfying
J = span ( v ⃗ σ ( 1 ) , v ⃗ σ ( 3 ) , … , v ⃗ σ ( 2 m − 1 ) ) ∩ W J=\operatorname{span}(\vec{v}_{\sigma(1)},\vec{v}_{\sigma(3)},\ldots,\vec{v}_{\sigma(2m-1)})\cap W J = span ( v σ ( 1 ) , v σ ( 3 ) , … , v σ ( 2 m − 1 ) ) ∩ W
we use.
We now check that Δ \Delta Δ is well defined.
Proposition 4.8. The map Δ J ⊂ W : St ω ( W ) → St ( J ) ⊗ St ω ( J ⊥ / J ) \Delta_{J\subset W}:\operatorname{St}^{\omega}(W)\to\operatorname{St}(J)\otimes\operatorname{St}^{\omega}(J^{\perp}/J) Δ J ⊂ W : St ω ( W ) → St ( J ) ⊗ St ω ( J ⊥ / J ) is well defined.
Proof. Let β = ( a ⃗ 1 , b ⃗ 1 , … , a ⃗ n , b ⃗ n ) \beta=(\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_n,\vec{b}_n) β = ( a 1 , b 1 , … , a n , b n ) be an ordered generalized symplectic basis of W W W . It is clear that if we apply relations (1) and (2) from Theorem 2.10 to β \beta β , then the output of Δ J ⊂ W \Delta_{J\subset W} Δ J ⊂ W is changed by apply relations (1) and (2) from Theorem 2.6 and Theorem 2.10 . Now apply relation (3) from Theorem 2.10 to β \beta β . The case J = 0 J=0 J = 0 is easy so assume J ≠ 0 J\ne0 J = 0 . By applying relation (2) from Theorem 2.10 , we may assume we are in of the following cases:
J ⊗ R F = span ( a ⃗ 1 , … , a ⃗ m ) J\otimes_R F=\operatorname{span}(\vec{a}_1,\ldots,\vec{a}_m) J ⊗ R F = span ( a 1 , … , a m ) ,
J ⊗ R F = span ( a ⃗ 1 + b ⃗ 1 , a ⃗ 2 , … , a ⃗ m ) J\otimes_R F=\operatorname{span}(\vec{a}_1+\vec{b}_1,\vec{a}_2,\ldots,\vec{a}_m) J ⊗ R F = span ( a 1 + b 1 , a 2 , … , a m ) ,
J ⊗ R F = span ( a ⃗ 2 , … , a ⃗ m + 1 ) J\otimes_R F=\operatorname{span}(\vec{a}_2,\ldots,\vec{a}_{m+1}) J ⊗ R F = span ( a 2 , … , a m + 1 ) ,
J ⊗ R F = span ( a ⃗ 2 + b ⃗ 2 , a ⃗ 3 , … , a ⃗ m + 1 ) J\otimes_R F=\operatorname{span}(\vec{a}_2+\vec{b}_2,\vec{a}_3,\ldots,\vec{a}_{m+1}) J ⊗ R F = span ( a 2 + b 2 , a 3 , … , a m + 1 ) ,
J ⊗ R F J\otimes_R F J ⊗ R F is not the span of a subset of β ∪ { a ⃗ i + b ⃗ i } \beta\cup\{\vec{a}_i+\vec{b}_i\} β ∪ { a i + b i } .
In case (i),
Δ J ⊂ W ( [ a ⃗ 1 , b ⃗ 1 , a ⃗ 2 , b ⃗ 2 , … , a ⃗ n , b ⃗ n ] ) = [ a ⃗ 1 , … , a ⃗ m ] ⊗ [ a ⃗ m + 1 , b ⃗ m + 1 , … , a ⃗ n , b ⃗ n ] , \Delta_{J\subset W}([\vec{a}_1,\vec{b}_1,\vec{a}_2,\vec{b}_2,\ldots,\vec{a}_n,\vec{b}_n])=[\vec{a}_1,\ldots,\vec{a}_m]\otimes[\vec{a}_{m+1},\vec{b}_{m+1},\ldots,\vec{a}_n,\vec{b}_n], Δ J ⊂ W ([ a 1 , b 1 , a 2 , b 2 , … , a n , b n ]) = [ a 1 , … , a m ] ⊗ [ a m + 1 , b m + 1 , … , a n , b n ] ,
Δ J ⊂ W ( [ a ⃗ 1 + b ⃗ 1 , b ⃗ 1 , a ⃗ 2 , b ⃗ 2 , … , a ⃗ n , b ⃗ n ] ) = 0 , \Delta_{J\subset W}([\vec{a}_1+\vec{b}_1,\vec{b}_1,\vec{a}_2,\vec{b}_2,\ldots,\vec{a}_n,\vec{b}_n])=0, Δ J ⊂ W ([ a 1 + b 1 , b 1 , a 2 , b 2 , … , a n , b n ]) = 0 ,
and
Δ J ⊂ W ( [ a ⃗ 1 , a ⃗ 1 + b ⃗ 1 , a ⃗ 2 , b ⃗ 2 , … , a ⃗ n , b ⃗ n ] ) = [ a ⃗ 1 , … , a ⃗ m ] ⊗ [ a ⃗ m + 1 , b ⃗ m + 1 , … , a ⃗ n , b ⃗ n ] . \Delta_{J\subset W}([\vec{a}_1,\vec{a}_1+\vec{b}_1,\vec{a}_2,\vec{b}_2,\ldots,\vec{a}_n,\vec{b}_n])=[\vec{a}_1,\ldots,\vec{a}_m]\otimes[\vec{a}_{m+1},\vec{b}_{m+1},\ldots,\vec{a}_n,\vec{b}_n]. Δ J ⊂ W ([ a 1 , a 1 + b 1 , a 2 , b 2 , … , a n , b n ]) = [ a 1 , … , a m ] ⊗ [ a m + 1 , b m + 1 , … , a n , b n ] .
This means that applying relation (3) does not change the output in case (i). A similar argument works in case (ii). Now consider case (iii). In case (iii),
Δ J ⊂ W ( [ a ⃗ 1 , b ⃗ 1 , a ⃗ 2 , b ⃗ 2 , … , a ⃗ n , b ⃗ n ] ) = [ a ⃗ 2 , … , a ⃗ m + 1 ] ⊗ [ a ⃗ 1 , b ⃗ 1 , a ⃗ m + 2 , b ⃗ m + 2 , … , a ⃗ n , b ⃗ n ] , \Delta_{J\subset W}([\vec{a}_1,\vec{b}_1,\vec{a}_2,\vec{b}_2,\ldots,\vec{a}_n,\vec{b}_n])=[\vec{a}_2,\ldots,\vec{a}_{m+1}]\otimes[\vec{a}_1,\vec{b}_1,\vec{a}_{m+2},\vec{b}_{m+2},\ldots,\vec{a}_n,\vec{b}_n], Δ J ⊂ W ([ a 1 , b 1 , a 2 , b 2 , … , a n , b n ]) = [ a 2 , … , a m + 1 ] ⊗ [ a 1 , b 1 , a m + 2 , b m + 2 , … , a n , b n ] ,
Δ J ⊂ W ( [ a ⃗ 1 + b ⃗ 1 , b ⃗ 1 , a ⃗ 2 , b ⃗ 2 , … , a ⃗ n , b ⃗ n ] ) = [ a ⃗ 2 , … , a ⃗ m + 1 ] ⊗ [ a ⃗ 1 + b ⃗ 1 , b ⃗ 1 , a ⃗ m + 2 , b ⃗ m + 2 , … , a ⃗ n , b ⃗ n ] , \Delta_{J\subset W}([\vec{a}_1+\vec{b}_1,\vec{b}_1,\vec{a}_2,\vec{b}_2,\ldots,\vec{a}_n,\vec{b}_n])=[\vec{a}_2,\ldots,\vec{a}_{m+1}]\otimes[\vec{a}_1+\vec{b}_1,\vec{b}_1,\vec{a}_{m+2},\vec{b}_{m+2},\ldots,\vec{a}_n,\vec{b}_n], Δ J ⊂ W ([ a 1 + b 1 , b 1 , a 2 , b 2 , … , a n , b n ]) = [ a 2 , … , a m + 1 ] ⊗ [ a 1 + b 1 , b 1 , a m + 2 , b m + 2 , … , a n , b n ] ,
and
Δ J ⊂ W ( [ a ⃗ 1 , a ⃗ 1 + b ⃗ 1 , a ⃗ 2 , b ⃗ 2 , … , a ⃗ n , b ⃗ n ] ) = [ a ⃗ 2 , … , a ⃗ m + 1 ] ⊗ [ a ⃗ 1 , a ⃗ 1 + b ⃗ 1 , a ⃗ m + 2 , b ⃗ m + 2 , … , a ⃗ n , b ⃗ n ] . \Delta_{J\subset W}([\vec{a}_1,\vec{a}_1+\vec{b}_1,\vec{a}_2,\vec{b}_2,\ldots,\vec{a}_n,\vec{b}_n])=[\vec{a}_2,\ldots,\vec{a}_{m+1}]\otimes[\vec{a}_1,\vec{a}_1+\vec{b}_1,\vec{a}_{m+2},\vec{b}_{m+2},\ldots,\vec{a}_n,\vec{b}_n]. Δ J ⊂ W ([ a 1 , a 1 + b 1 , a 2 , b 2 , … , a n , b n ]) = [ a 2 , … , a m + 1 ] ⊗ [ a 1 , a 1 + b 1 , a m + 2 , b m + 2 , … , a n , b n ] .
By relation (3) from Theorem 2.10 , applying relation (3) from Theorem 2.10 does not change the output in case (iii). A similar argument works in case (iv). In case (v), all terms map to 0.
Now consider relation (4) from Theorem 2.10 . The only interesting case to consider is when J ⊗ R F = span ( a ⃗ 2 , … , a ⃗ m + 1 ) J\otimes_R F=\operatorname{span}(\vec{a}_2,\ldots,\vec{a}_{m+1}) J ⊗ R F = span ( a 2 , … , a m + 1 ) as all other cases can be dealt with by arguments similar to uses previously used in this proof. We have
Δ J ⊂ W [ a ⃗ 1 , b ⃗ 1 , a ⃗ 2 , b ⃗ 2 , … , a ⃗ n , b ⃗ n ] = [ a ⃗ 2 , … , a ⃗ m + 1 ] ⊗ [ a ⃗ 1 , b ⃗ 1 , a ⃗ m + 2 , b ⃗ m + 2 , … , a ⃗ n , b ⃗ n ] , Δ J ⊂ W [ a ⃗ 1 , b ⃗ 1 − b ⃗ 2 , a ⃗ 1 + a ⃗ 2 , b ⃗ 2 , a ⃗ 3 , b ⃗ 3 , … , a ⃗ n , b ⃗ n ] = 0 , and Δ J ⊂ W ( [ b ⃗ 1 − b ⃗ 2 , a ⃗ 2 , a ⃗ 1 + a ⃗ 2 , b ⃗ 1 , … , a ⃗ n , b ⃗ n ] ) = [ a ⃗ 2 , … , a ⃗ m + 1 ] ⊗ [ a ⃗ 1 + a ⃗ 2 , b ⃗ 1 , a ⃗ m + 2 , b ⃗ m + 2 , … , a ⃗ n , b ⃗ n ] = [ a ⃗ 2 , … , a ⃗ m + 1 ] ⊗ [ a ⃗ 1 , b ⃗ 1 , a ⃗ m + 2 , b ⃗ m + 2 , … , a ⃗ n , b ⃗ n ] . \begin{aligned}
\Delta_{J\subset W}[\vec{a}_1,\vec{b}_1,\vec{a}_2,\vec{b}_2,\ldots,\vec{a}_n,\vec{b}_n]
&=[\vec{a}_2,\ldots,\vec{a}_{m+1}]\otimes[\vec{a}_1,\vec{b}_1,\vec{a}_{m+2},\vec{b}_{m+2},\ldots,\vec{a}_n,\vec{b}_n],\\
\Delta_{J\subset W}[\vec{a}_1,\vec{b}_1-\vec{b}_2,\vec{a}_1+\vec{a}_2,\vec{b}_2,\vec{a}_3,\vec{b}_3,\ldots,\vec{a}_n,\vec{b}_n]
&=0,\quad\text{and}\\
\Delta_{J\subset W}([\vec{b}_1-\vec{b}_2,\vec{a}_2,\vec{a}_1+\vec{a}_2,\vec{b}_1,\ldots,\vec{a}_n,\vec{b}_n])
&=[\vec{a}_2,\ldots,\vec{a}_{m+1}]\otimes[\vec{a}_1+\vec{a}_2,\vec{b}_1,\vec{a}_{m+2},\vec{b}_{m+2},\ldots,\vec{a}_n,\vec{b}_n]\\
&=[\vec{a}_2,\ldots,\vec{a}_{m+1}]\otimes[\vec{a}_1,\vec{b}_1,\vec{a}_{m+2},\vec{b}_{m+2},\ldots,\vec{a}_n,\vec{b}_n].
\end{aligned} Δ J ⊂ W [ a 1 , b 1 , a 2 , b 2 , … , a n , b n ] Δ J ⊂ W [ a 1 , b 1 − b 2 , a 1 + a 2 , b 2 , a 3 , b 3 , … , a n , b n ] Δ J ⊂ W ([ b 1 − b 2 , a 2 , a 1 + a 2 , b 1 , … , a n , b n ]) = [ a 2 , … , a m + 1 ] ⊗ [ a 1 , b 1 , a m + 2 , b m + 2 , … , a n , b n ] , = 0 , and = [ a 2 , … , a m + 1 ] ⊗ [ a 1 + a 2 , b 1 , a m + 2 , b m + 2 , … , a n , b n ] = [ a 2 , … , a m + 1 ] ⊗ [ a 1 , b 1 , a m + 2 , b m + 2 , … , a n , b n ] .
Here we use that the symplectic apartment in the output of Δ J ⊂ W \Delta_{J\subset W} Δ J ⊂ W lives in the symplectic Steinberg module of the quotient space J ⊥ / J J^\perp/J J ⊥ / J . This checks that applying relation (4) from Theorem 2.10 does not change the output. □
Proposition 4.9. ( S t ω , Δ ) (\mathrm{St}^{\omega},\Delta) ( St ω , Δ ) is counital coassociative comodule over ( S t , Δ ) (\mathrm{St},\Delta) ( St , Δ ) .
Proof. Using the fact that the unit maps to the empty apartment class, it is not hard to see that ( S t ω , Δ ) (\mathrm{St}^{\omega},\Delta) ( St ω , Δ ) is counital over ( S t , Δ ) (\mathrm{St},\Delta) ( St , Δ ) .
Let W W W be a symplectic R R R -module and let v 1 , v 2 , … , v 2 n − 1 , v 2 n v_1,v_2,\ldots,v_{2n-1},v_{2n} v 1 , v 2 , … , v 2 n − 1 , v 2 n be a symplectic basis of W ⊗ R F W\otimes_R F W ⊗ R F . To check that Δ \Delta Δ is coassociative, We must check that
( Id ⊗ Δ ) ( Δ ( [ v ⃗ 1 , v ⃗ 2 , … , v ⃗ 2 n − 1 , v ⃗ 2 n ] ) ) = ( Δ ⊗ Id ) ( Δ ( [ v ⃗ 1 , v ⃗ 2 , … , v ⃗ 2 n − 1 , v ⃗ 2 n ] ) ) (\operatorname{Id}\otimes\Delta)\left(\Delta([\vec{v}_1,\vec{v}_2,\ldots,\vec{v}_{2n-1},\vec{v}_{2n}])\right)=(\Delta\otimes\operatorname{Id})\left(\Delta([\vec{v}_1,\vec{v}_2,\ldots,\vec{v}_{2n-1},\vec{v}_{2n}])\right) ( Id ⊗ Δ ) ( Δ ([ v 1 , v 2 , … , v 2 n − 1 , v 2 n ]) ) = ( Δ ⊗ Id ) ( Δ ([ v 1 , v 2 , … , v 2 n − 1 , v 2 n ]) )
∈ ⨁ M ⊆ J ⊂ W S t ( M ) ⊗ S t ( J / M ) ⊗ S t ω ( J ⊥ / J ) . \in\bigoplus_{M\subseteq J\subset W}\mathrm{St}(M)\otimes\mathrm{St}(J/M)\otimes\mathrm{St}^{\omega}(J^\perp/J). ∈ M ⊆ J ⊂ W ⨁ St ( M ) ⊗ St ( J / M ) ⊗ St ω ( J ⊥ / J ) .
Here J J J is an isotropic summand of W W W and M M M is a summand of J J J or equivalently we view M M M as an isotropic summand of W W W , J / M J/M J / M as an isotropic summand of M ⊥ / M M^\perp/M M ⊥ / M , and J ⊥ / J J^\perp/J J ⊥ / J as ( J / M ) ⊥ / ( J / M ) (J/M)^\perp/(J/M) ( J / M ) ⊥ / ( J / M ) .
We will check that they are equal by checking they are equal on the summand associated to M ⊂ J M\subset J M ⊂ J . If there is no σ ∈ S 2 n ω \sigma\in S^\omega_{2n} σ ∈ S 2 n ω and k ≤ m ≤ n k\le m\le n k ≤ m ≤ n with
J = span ( v ⃗ σ ( 1 ) , v ⃗ σ ( 3 ) , … , v ⃗ σ ( 2 m − 1 ) ) ∩ W J=\operatorname{span}(\vec{v}_{\sigma(1)},\vec{v}_{\sigma(3)},\ldots,\vec{v}_{\sigma(2m-1)})\cap W J = span ( v σ ( 1 ) , v σ ( 3 ) , … , v σ ( 2 m − 1 ) ) ∩ W
and
M = span ( v ⃗ σ ( 1 ) , v ⃗ σ ( 3 ) , … , v ⃗ σ ( 2 k − 1 ) ) ∩ W , M=\operatorname{span}(\vec{v}_{\sigma(1)},\vec{v}_{\sigma(3)},\ldots,\vec{v}_{\sigma(2k-1)})\cap W, M = span ( v σ ( 1 ) , v σ ( 3 ) , … , v σ ( 2 k − 1 ) ) ∩ W ,
then both terms are zero. Otherwise, both terms are equal to
χ ( σ ) [ v ⃗ σ ( 1 ) , v ⃗ σ ( 3 ) … , v ⃗ σ ( 2 k − 1 ) ] ⊗ [ v ⃗ σ ( 2 k + 1 ) , v ⃗ σ ( 2 k + 3 ) … , v ⃗ σ ( 2 m − 1 ) ] ⊗ [ v ⃗ σ ( 2 m + 1 ) , v ⃗ σ ( 2 m + 2 ) , … , v ⃗ σ ( 2 n ) ] . \chi(\sigma)[\vec{v}_{\sigma(1)},\vec{v}_{\sigma(3)}\ldots,\vec{v}_{\sigma(2k-1)}]\otimes[\vec{v}_{\sigma(2k+1)},\vec{v}_{\sigma(2k+3)}\ldots,\vec{v}_{\sigma(2m-1)}]\otimes[\vec{v}_{\sigma(2m+1)},\vec{v}_{\sigma(2m+2)},\ldots,\vec{v}_{\sigma(2n)}]. χ ( σ ) [ v σ ( 1 ) , v σ ( 3 ) … , v σ ( 2 k − 1 ) ] ⊗ [ v σ ( 2 k + 1 ) , v σ ( 2 k + 3 ) … , v σ ( 2 m − 1 ) ] ⊗ [ v σ ( 2 m + 1 ) , v σ ( 2 m + 2 ) , … , v σ ( 2 n ) ] .
Thus ( S t ω , Δ ) (\mathrm{St}^{\omega},\Delta) ( St ω , Δ ) is a coassociative comodule over ( S t , Δ ) (\mathrm{St},\Delta) ( St , Δ ) . □
Duoidal comodule algebra structure
Kupers–Rudenko–Sierra [9 Section 2.3] checked that ( S t , Δ , μ ) (\mathrm{St},\Delta,\mu) ( St , Δ , μ ) is a graded-commutative duoidal bialgebra in the duoidal category ( Mod V B ( R ) ( k ) , ⊗ P , ⊗ L ) (\operatorname{Mod}_{\mathrm{VB}(R)}(k),\otimes_P,\otimes_L) ( Mod VB ( R ) ( k ) , ⊗ P , ⊗ L ) for R R R a field. A similar argument works for R R R a Dedekind domain.
Theorem 4.10. For R R R a Dedekind domain, ( S t ω , Δ , μ ) (\mathrm{St}^{\omega},\Delta,\mu) ( St ω , Δ , μ ) is a graded-commutative duoidal comodule algebra over ( S t , Δ , μ ) (\mathrm{St},\Delta,\mu) ( St , Δ , μ ) .
Proof. The only interesting condition of Definition 3.11 is Condition 1. To check that, we must show that the following diagram commutes:
St ω ⊗ Q St ω → Δ ⊗ Q Δ ( St ⊗ P St ω ) ⊗ Q ( St ⊗ P St ω ) μ ↓ ↓ ( μ ⊗ P μ ) ∘ ξ St ω → Δ St ⊗ P St ω \begin{CD}
\operatorname{St}^{\omega}\otimes_{\mathbb{Q}}\operatorname{St}^{\omega} @>{\Delta\otimes_{\mathbb{Q}}\Delta}>> (\operatorname{St}\otimes_{\mathbb{P}}\operatorname{St}^{\omega})\otimes_{\mathbb{Q}}(\operatorname{St}\otimes_{\mathbb{P}}\operatorname{St}^{\omega}) \\
@V{\mu}VV @VV{(\mu\otimes_{\mathbb{P}}\mu)\circ\xi}V \\
\operatorname{St}^{\omega} @>{\Delta}>> \operatorname{St}\otimes_{\mathbb{P}}\operatorname{St}^{\omega}
\end{CD} St ω ⊗ Q St ω μ ↓ ⏐ St ω Δ ⊗ Q Δ Δ ( St ⊗ P St ω ) ⊗ Q ( St ⊗ P St ω ) ↓ ⏐ ( μ ⊗ P μ ) ∘ ξ St ⊗ P St ω
Concretely, we must check that the following diagram commutes for every symplectic module W W W :
⨁ U ⊕ V = W St ω ( U ) ⊗ St ω ( V ) → Δ ⊗ Δ ⨁ U ⊕ V = W , J ⊂ U , K ⊂ V St ( J ) ⊗ St ω ( J ⊥ / J ) ⊗ St ( K ) ⊗ St ω ( K ⊥ / K ) μ ↓ ↓ ( μ ⊗ μ ) ∘ ξ St ω ( W ) → Δ ⨁ Q ⊂ W St ( Q ) ⊗ St ω ( Q ⊥ / Q ) . \begin{CD}
\bigoplus_{U\oplus V=W}\operatorname{St}^{\omega}(U)\otimes\operatorname{St}^{\omega}(V) @>{\Delta\otimes\Delta}>> \bigoplus_{\substack{U\oplus V=W,\\J\subset U,K\subset V}}\operatorname{St}(J)\otimes\operatorname{St}^{\omega}(J^{\perp}/J)\otimes\operatorname{St}(K)\otimes\operatorname{St}^{\omega}(K^{\perp}/K) \\
@V{\mu}VV @VV{(\mu\otimes\mu)\circ\xi}V \\
\operatorname{St}^{\omega}(W) @>{\Delta}>> \bigoplus_{Q\subset W}\operatorname{St}(Q)\otimes\operatorname{St}^{\omega}(Q^{\perp}/Q).
\end{CD} U ⊕ V = W ⨁ St ω ( U ) ⊗ St ω ( V ) μ ↓ ⏐ St ω ( W ) Δ ⊗ Δ Δ U ⊕ V = W , J ⊂ U , K ⊂ V ⨁ St ( J ) ⊗ St ω ( J ⊥ / J ) ⊗ St ( K ) ⊗ St ω ( K ⊥ / K ) ↓ ⏐ ( μ ⊗ μ ) ∘ ξ Q ⊂ W ⨁ St ( Q ) ⊗ St ω ( Q ⊥ / Q ) .
Let [ a ⃗ 1 , b ⃗ 1 , … , a ⃗ k , b ⃗ k ] ⊗ [ c ⃗ 1 , d ⃗ 1 , … , c ⃗ j , d ⃗ j ] [\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_k,\vec{b}_k]\otimes[\vec{c}_1,\vec{d}_1,\ldots,\vec{c}_j,\vec{d}_j] [ a 1 , b 1 , … , a k , b k ] ⊗ [ c 1 , d 1 , … , c j , d j ] be an element of the top left. Traversing the diagram counter clockwise yields
Δ ( μ ( [ a ⃗ 1 , b ⃗ 1 , … , a ⃗ k , b ⃗ k ] ⊗ [ c ⃗ 1 , d ⃗ 1 , … , c ⃗ j , d ⃗ j ] ) ) = Δ ( [ a ⃗ 1 , b ⃗ 1 , … , a ⃗ k , b ⃗ k , c ⃗ 1 , d ⃗ 1 , … , c ⃗ j , d ⃗ j ] ) \Delta\left(\mu\left([\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_k,\vec{b}_k]\otimes[\vec{c}_1,\vec{d}_1,\ldots,\vec{c}_j,\vec{d}_j]\right)\right)=\Delta\left([\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_k,\vec{b}_k,\vec{c}_1,\vec{d}_1,\ldots,\vec{c}_j,\vec{d}_j]\right) Δ ( μ ( [ a 1 , b 1 , … , a k , b k ] ⊗ [ c 1 , d 1 , … , c j , d j ] ) ) = Δ ( [ a 1 , b 1 , … , a k , b k , c 1 , d 1 , … , c j , d j ] )
This is the signed sum of apartments associated to isotropic subsets of
{ a ⃗ 1 , b ⃗ 1 , … , a ⃗ k , b ⃗ k , c ⃗ 1 , d ⃗ 1 , … , c ⃗ j , d ⃗ j } \{\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_k,\vec{b}_k,\vec{c}_1,\vec{d}_1,\ldots,\vec{c}_j,\vec{d}_j\} { a 1 , b 1 , … , a k , b k , c 1 , d 1 , … , c j , d j }
tensored with the symplectic apartment associated with the complementary symplectic basis. Traversing the diagram clockwise yields
( μ ⊗ Q μ ) ( ξ ( Δ ( [ a ⃗ 1 , b ⃗ 1 , … , a ⃗ k , b ⃗ k ] ) ⊗ Δ ( [ c ⃗ 1 , d ⃗ 1 , … , c ⃗ j , d ⃗ j ] ) ) ) . (\mu\otimes_{\mathbb{Q}}\mu)\left(\xi\left(\Delta([\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_k,\vec{b}_k])\otimes\Delta([\vec{c}_1,\vec{d}_1,\ldots,\vec{c}_j,\vec{d}_j])\right)\right). ( μ ⊗ Q μ ) ( ξ ( Δ ([ a 1 , b 1 , … , a k , b k ]) ⊗ Δ ([ c 1 , d 1 , … , c j , d j ]) ) ) .
Note that Δ ( [ a ⃗ 1 , b ⃗ 1 , … , a ⃗ k , b ⃗ k ] ) \Delta([\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_k,\vec{b}_k]) Δ ([ a 1 , b 1 , … , a k , b k ]) is the signed sum of apartments associated to isotropic subsets of
{ a ⃗ 1 , b ⃗ 1 , … , a ⃗ k , b ⃗ k } \{\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_k,\vec{b}_k\} { a 1 , b 1 , … , a k , b k }
tensored with the symplectic apartment associated with the complementary symplectic basis and similarly Δ ( [ c ⃗ 1 , d ⃗ 1 , … , c ⃗ j , d ⃗ j ] ) \Delta([\vec{c}_1,\vec{d}_1,\ldots,\vec{c}_j,\vec{d}_j]) Δ ([ c 1 , d 1 , … , c j , d j ]) and has the analogous description. Applying the function ( μ ⊗ μ ) ∘ ξ (\mu\otimes\mu)\circ\xi ( μ ⊗ μ ) ∘ ξ concatenates the symplectic apartments and the isotropic apartments. Since isotropic subsets of
{ a ⃗ 1 , b ⃗ 1 , … , a ⃗ k , b ⃗ k , c ⃗ 1 , d ⃗ 1 , … , c ⃗ j , d ⃗ j } \{\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_k,\vec{b}_k,\vec{c}_1,\vec{d}_1,\ldots,\vec{c}_j,\vec{d}_j\} { a 1 , b 1 , … , a k , b k , c 1 , d 1 , … , c j , d j }
are the same as sets obtained by taking unions of isotropic subsets of
{ a ⃗ 1 , b ⃗ 1 , … , a ⃗ k , b ⃗ k } \{\vec{a}_1,\vec{b}_1,\ldots,\vec{a}_k,\vec{b}_k\} { a 1 , b 1 , … , a k , b k }
and
{ c ⃗ 1 , d ⃗ 1 , … , c ⃗ j , d ⃗ j } , \{\vec{c}_1,\vec{d}_1,\ldots,\vec{c}_j,\vec{d}_j\}, { c 1 , d 1 , … , c j , d j } ,
traversing the diagram clockwise and counterclockwise agree. □
Classical comodule algebra structure
In this section, and whenever talking about B and A , we will assume that k \mathbb{k} k is a field. This is done so that the Künneth map is an isomorphism.
Definition 4.11. Let R R R be a Dedekind domain. For each isomorphism class of finitely generated projective modules [ P ] [P] [ P ] , pick a representative. Let
B [ P ] , i = H i ( GL ( P ) ; St ( P ) ⊗ k ) , (2) \mathbf{B}_{[P],i}=H_i(\operatorname{GL}(P);\operatorname{St}(P)\otimes\mathbb{k}), \tag*{(2)} B [ P ] , i = H i ( GL ( P ) ; St ( P ) ⊗ k ) , (2)
B n , i = ⨁ rk ( [ P ] ) = n B [ P ] , i , B_{n,i}=\bigoplus_{\operatorname{rk}([P])=n} B_{[P],i}, B n , i = rk ([ P ]) = n ⨁ B [ P ] , i ,
and
B = ⨁ n , i B n , i . B=\bigoplus_{n,i}B_{n,i}. B = n , i ⨁ B n , i .
Let
A n , i = H i ( Sp 2 n ( R ) ; St 2 n ω ( R ) ⊗ k ) and A = ⨁ n , i A n , i . A_{n,i}=H_i(\operatorname{Sp}_{2n}(R);\operatorname{St}^{\omega}_{2n}(R)\otimes\mathbb{k})
\quad\text{and}\quad
A=\bigoplus_{n,i}A_{n,i}. A n , i = H i ( Sp 2 n ( R ) ; St 2 n ω ( R ) ⊗ k ) and A = n , i ⨁ A n , i .
Algebra structures on B and A
Let P P P and Q Q Q be finitely generated projective R R R -modules. The product
μ : St ( P ) ⊗ St ( Q ) ⟶ St ( P ⊕ Q ) \mu:\operatorname{St}(P)\otimes\operatorname{St}(Q)\longrightarrow\operatorname{St}(P\oplus Q) μ : St ( P ) ⊗ St ( Q ) ⟶ St ( P ⊕ Q )
induces a map
μ : H i ( GL ( P ) ; St ( P ) ) ⊗ H j ( GL ( Q ) ; St ( Q ) ) ⟶ H i + j ( GL ( P ⊕ Q ) ; St ( P ⊕ Q ) ) . \mu:H_i(\operatorname{GL}(P);\operatorname{St}(P))\otimes H_j(\operatorname{GL}(Q);\operatorname{St}(Q))\longrightarrow H_{i+j}(\operatorname{GL}(P\oplus Q);\operatorname{St}(P\oplus Q)). μ : H i ( GL ( P ) ; St ( P )) ⊗ H j ( GL ( Q ) ; St ( Q )) ⟶ H i + j ( GL ( P ⊕ Q ) ; St ( P ⊕ Q )) .
By picking isomorphisms between P P P , Q Q Q , and P ⊕ Q P\oplus Q P ⊕ Q with our chosen representatives of [ P ] [P] [ P ] , [ Q ] [Q] [ Q ] , and [ P ⊕ Q ] [P\oplus Q] [ P ⊕ Q ] , we obtain a product
B [ P ] , i ⊗ B [ Q ] , j ⟶ B [ P ⊕ Q ] , i + j . B_{[P],i}\otimes B_{[Q],j}\longrightarrow B_{[P\oplus Q],i+j}. B [ P ] , i ⊗ B [ Q ] , j ⟶ B [ P ⊕ Q ] , i + j .
Note that this product does not depend on these choices of isomorphisms since conjugation by an element g g g of a group G G G induces the trivial automorphism on the group homology of G G G . In this way, μ \mu μ induces a graded ring structure on B \mathbf{B} B . Similarly, A \mathbf{A} A has a graded ring structure induced by μ : St ω ⊗ St ω → St ω \mu:\operatorname{St}^{\omega}\otimes\operatorname{St}^{\omega}\to\operatorname{St}^{\omega} μ : St ω ⊗ St ω → St ω .
These products are graded-commutative in these sense that if α ∈ B n , i \alpha\in B_{n,i} α ∈ B n , i and β ∈ B m , j \beta\in B_{m,j} β ∈ B m , j , then
α β = ( − 1 ) n m + i j β α \alpha\beta=(-1)^{nm+ij}\beta\alpha α β = ( − 1 ) nm + ij β α
an similarly for A \mathbf{A} A .
Coalgebra structure on B and comodule structure on A
The coproducts induce coalgebra and comodule structures on B \mathbf{B} B and A \mathbf{A} A respectively. Let M M M be a finitely generated projective R R R -module. The coproduct Δ : St → St ⊗ Z St \Delta:\operatorname{St}\to\operatorname{St}\otimes_{\mathbb{Z}}\operatorname{St} Δ : St → St ⊗ Z St induces a map
Δ : H i ( GL ( M ) ; St ( M ) ) ⟶ H i ( GL ( M ) ; ⨁ M ′ ⊆ M St ( M ′ ) ⊗ St ( M / M ′ ) ) . \Delta:H_i(\operatorname{GL}(M);\operatorname{St}(M))\longrightarrow H_i\left(\operatorname{GL}(M);\bigoplus_{M'\subseteq M}\operatorname{St}(M')\otimes\operatorname{St}(M/M')\right). Δ : H i ( GL ( M ) ; St ( M )) ⟶ H i ( GL ( M ) ; M ′ ⊆ M ⨁ St ( M ′ ) ⊗ St ( M / M ′ ) ) .
We will simplify the codomain using Shapiro’s lemma. Note that the orbit of GL ( M ) \operatorname{GL}(M) GL ( M ) acting on the set of summands of M M M is the set of isomorphism classes of R R R -modules of rank less than or equal to the rank of M M M that are isomorphic to M M M if they have the same rank as M M M . Thus, Shapiro’s lemma provides a isomorphism
H i ( GL ( M ) ; ⨁ M ′ ⊆ M St ( M ′ ) ⊗ St ( M / M ′ ) ) ⟶ ⨁ [ M ′ ] , rk ( M ′ ) ≤ rk ( M ) [ M ′ ] = [ M ] if rk ( M ′ ) = rk ( M ) H i ( Stab M ′ ( GL ( M ) ) ; St ( M ′ ) ⊗ St ( M / M ′ ) ) . H_i\left(\operatorname{GL}(M);\bigoplus_{M'\subseteq M}\operatorname{St}(M')\otimes\operatorname{St}(M/M')\right)
\longrightarrow
\bigoplus_{\substack{[M'],\,\operatorname{rk}(M')\leq\operatorname{rk}(M)\\ [M']=[M]\ \text{if }\operatorname{rk}(M')=\operatorname{rk}(M)}}
H_i\left(\operatorname{Stab}_{M'}(\operatorname{GL}(M));\operatorname{St}(M')\otimes\operatorname{St}(M/M')\right). H i ( GL ( M ) ; M ′ ⊆ M ⨁ St ( M ′ ) ⊗ St ( M / M ′ ) ) ⟶ [ M ′ ] , rk ( M ′ ) ≤ rk ( M ) [ M ′ ] = [ M ] if rk ( M ′ ) = rk ( M ) ⨁ H i ( Stab M ′ ( GL ( M )) ; St ( M ′ ) ⊗ St ( M / M ′ ) ) .
The stabilizer of M ′ M' M ′ in GL ( M ) \operatorname{GL}(M) GL ( M ) , is a parabolic subgroup with a natural split surjective map to the corresponding Levi subgroup:
Stab GL ( M ) ( M ′ ) ⟶ GL ( M ′ ) × GL ( M / M ′ ) . \operatorname{Stab}_{\operatorname{GL}(M)}(M')\longrightarrow\operatorname{GL}(M')\times\operatorname{GL}(M/M'). Stab GL ( M ) ( M ′ ) ⟶ GL ( M ′ ) × GL ( M / M ′ ) .
The action of the parabolic on St ( M ′ ) ⊗ St ( M / M ′ ) \operatorname{St}(M') \otimes\operatorname{St}(M/M') St ( M ′ ) ⊗ St ( M / M ′ ) factors through the Levi so we obtain a map
H i ( GL ( M ) ; St ( M ) ) ⟶ ⨁ [ M ′ ] , rk ( M ′ ) ≤ rk ( M ) H i ( GL ( M ′ ) × GL ( M / M ′ ) ; St ( M ′ ) ⊗ St ( M / M ′ ) ) . H_i(\operatorname{GL}(M);\operatorname{St}(M)) \longrightarrow\bigoplus_{\substack{[M'],\ \operatorname{rk}(M')\leq\operatorname{rk}(M)}} H_i(\operatorname{GL}(M') \times\operatorname{GL}(M/M');\operatorname{St}(M') \otimes\operatorname{St}(M/M')). H i ( GL ( M ) ; St ( M )) ⟶ [ M ′ ] , rk ( M ′ ) ≤ rk ( M ) ⨁ H i ( GL ( M ′ ) × GL ( M / M ′ ) ; St ( M ′ ) ⊗ St ( M / M ′ )) .
After applying the inverse of the Künneth map, we obtain a coalgebra structure:
Δ : B n , i ⟶ ⨁ a + b = n k + j = i B a , j ⊗ B b , k . \Delta:\mathbf{B}_{n,i}\longrightarrow\bigoplus_{\substack{a+b=n\\ k+j=i}}\mathbf{B}_{a,j}\otimes\mathbf{B}_{b,k}. Δ : B n , i ⟶ a + b = n k + j = i ⨁ B a , j ⊗ B b , k .
A similar discussion provides a comodule map
Δ : A n , i ⟶ ⨁ a + b = n k + j = i B a , j ⊗ A b , k . \Delta:\mathbf{A}_{n,i}\longrightarrow\bigoplus_{\substack{a+b=n\\ k+j=i}}\mathbf{B}_{a,j}\otimes\mathbf{A}_{b,k}. Δ : A n , i ⟶ a + b = n k + j = i ⨁ B a , j ⊗ A b , k .
Here we use that if J J J is an isotropic summand of a symplectic R R R -module W W W , then the Levi subgroup corresponding to parabolic subgroup Stab Sp ( W ) ( J ) \operatorname{Stab}_{\operatorname{Sp}(W)}(J) Stab Sp ( W ) ( J ) is
GL ( J ) × Sp ( J ⊥ / J ) . \operatorname{GL}(J)\times\operatorname{Sp}(J^{\perp}/J). GL ( J ) × Sp ( J ⊥ / J ) .
Again, the action of the parabolic on St ( J ) ⊗ St ω ( J ⊥ / J ) \operatorname{St}(J)\otimes\operatorname{St}^{\omega}(J^{\perp}/J) St ( J ) ⊗ St ω ( J ⊥ / J ) factors through the Levi.
4.3.3. Hopf algebra structure on B \mathbf{B} B and comodule algebra structure on A \mathbf{A} A . Since St \operatorname{St} St is a duoidal bialgebra, the product and coproduct on B \mathbf{B} B are compatible in the sense that they endow B \mathbf{B} B with the structure of a bialgebra. Since the product on St \operatorname{St} St is graded-commutative, so is the product on B \mathbf{B} B . Since B 0 , 0 \mathbf{B}_{0,0} B 0 , 0 is the 0th homology of the trivial group with coefficients in k = St ( 0 ) \mathbb{k}=\operatorname{St}(0) k = St ( 0 ) , B 0 , 0 = k \mathbf{B}_{0,0}=\mathbb{k} B 0 , 0 = k . Connected bialgebras are Hopf algebras [11 Proposition 8.2] .
Proposition 4.12. Let k \mathbb{k} k be a field and R R R a Dedekind domain. Then B \mathbf{B} B is a connected graded-commutative Hopf algebra.
Recall that a comodule algebra over a bialgebra B B B is a monoid object in the category of B B B -comodules. The duoidal St ω \operatorname{St}^{\omega} St ω -comodule algebra structure on St ω \operatorname{St}^{\omega} St ω gives the following which is a restatement of Theorem 1.1 .
Theorem 4.13. Let k \mathbb{k} k be a field and R R R a Dedekind domain. Then A \mathbf{A} A is a connected graded-commutative comodule algebra over the Hopf algebra B \mathbf{B} B .
Freeness of comodule algebras
In this section, we prove a structure theorem for module coalgebras, which will give us a lower bound for the comodule algebra A \mathbf{A} A in the following section. Throughout this section, let k \mathbb{k} k be a field, B B B be a graded connected bialgebra over k \mathbb{k} k , and C C C a graded connected B B B -module coalgebra.
We denote by W W W the image
W : = im ( B ≅ B ⊗ C 0 ↪ B ⊗ C → C ) . W:=\operatorname{im}(B\cong B\otimes C_0\hookrightarrow B\otimes C\to C). W := im ( B ≅ B ⊗ C 0 ↪ B ⊗ C → C ) .
Note that W W W is a graded B B B -submodule of C C C , because
b ′ ⋅ ( b ⋅ 1 C ) = ( b ′ ⋅ b ) ⋅ 1 C . b'\cdot(b\cdot1_C)=(b'\cdot b)\cdot1_C. b ′ ⋅ ( b ⋅ 1 C ) = ( b ′ ⋅ b ) ⋅ 1 C .
W W W is also a graded subcoalgebra of C C C because
Δ C ( b ⋅ 1 C ) = Δ B ( b ) ⋅ Δ C ( 1 C ) = Δ B ( b ) ⋅ ( 1 C ⊗ 1 C ) ∈ W ⊗ W . \Delta_C(b\cdot1_C)=\Delta_B(b)\cdot\Delta_C(1_C)=\Delta_B(b)\cdot(1_C\otimes1_C)\in W\otimes W. Δ C ( b ⋅ 1 C ) = Δ B ( b ) ⋅ Δ C ( 1 C ) = Δ B ( b ) ⋅ ( 1 C ⊗ 1 C ) ∈ W ⊗ W .
Further, denote by Q Q Q the graded quotient vector space
Let q : C → Q q : C \to Q q : C → Q denote the canonical quotient map. It follows that Q Q Q is a graded B B B -module. Furthermore, Q Q Q inherits a C C C -comodule structure, as
Δ C ( W ) ⊆ W ⊗ W ⊆ C ⊗ W \Delta_C(W) \subseteq W \otimes W \subseteq C \otimes W Δ C ( W ) ⊆ W ⊗ W ⊆ C ⊗ W
and thus
Δ Q : Q ⟶ C ⊗ Q c + W ⟼ ( id ⊗ q ) Δ C ( c ) \begin{aligned}
\Delta_Q &: Q \longrightarrow C \otimes Q \\
c + W &\longmapsto(\operatorname{id} \otimes q)\Delta_C(c)
\end{aligned} Δ Q c + W : Q ⟶ C ⊗ Q ⟼ ( id ⊗ q ) Δ C ( c )
is well defined. Thus Q Q Q is a B B B -module C C C -comodule. The C C C -comodule primitives of Q Q Q are defined as
P ( Q ) : = { c + W ∈ Q ∣ Δ Q ( c + W ) = 1 C ⊗ ( c + W ) } . P(Q) := \{c + W \in Q \mid\Delta_Q(c + W) = 1_C \otimes(c + W)\}. P ( Q ) := { c + W ∈ Q ∣ Δ Q ( c + W ) = 1 C ⊗ ( c + W )} .
Lemma 5.1. W W W is a subspace of
{ c ∈ C ∣ Δ C ( c ) − 1 C ⊗ c ∈ C ⊗ W } \{c \in C \mid\Delta_C(c) - 1_C \otimes c \in C \otimes W\} { c ∈ C ∣ Δ C ( c ) − 1 C ⊗ c ∈ C ⊗ W } and
P ( Q ) = { c ∈ C ∣ Δ C ( c ) − 1 C ⊗ c ∈ C ⊗ W } / W . P(Q) = \{c \in C \mid\Delta_C(c) - 1_C \otimes c \in C \otimes W\}/W. P ( Q ) = { c ∈ C ∣ Δ C ( c ) − 1 C ⊗ c ∈ C ⊗ W } / W .
Proof. Since Δ C ( w ) ∈ W ⊗ W \Delta_C(w) \in W \otimes W Δ C ( w ) ∈ W ⊗ W for all w w w , so is Δ C ( w ) − 1 C ⊗ w \Delta_C(w) - 1_C \otimes w Δ C ( w ) − 1 C ⊗ w . It remains to prove that Δ C ( c ) − 1 C ⊗ c ∈ C ⊗ W \Delta_C(c) - 1_C \otimes c \in C \otimes W Δ C ( c ) − 1 C ⊗ c ∈ C ⊗ W if and only if Δ Q ( c + W ) = 1 C ⊗ ( c + W ) \Delta_Q(c + W) = 1_C \otimes(c + W) Δ Q ( c + W ) = 1 C ⊗ ( c + W ) . This can be seen because
Δ Q ( c + W ) − 1 C ⊗ ( c + W ) = ( id ⊗ q ) ( Δ C ( c ) − 1 C ⊗ c ) \Delta_Q(c + W) - 1_C \otimes(c + W) = (\operatorname{id} \otimes q)(\Delta_C(c) - 1_C \otimes c) Δ Q ( c + W ) − 1 C ⊗ ( c + W ) = ( id ⊗ q ) ( Δ C ( c ) − 1 C ⊗ c )
and ( id ⊗ q ) ( x ) = 0 (\operatorname{id} \otimes q)(x) = 0 ( id ⊗ q ) ( x ) = 0 if and only if x ∈ C ⊗ W x \in C \otimes W x ∈ C ⊗ W . □
We will now prove our main structure theorem.
Theorem 5.2. There is a graded injection from W ⊗ ( P ( Q ) ⊕ C 0 ) W \otimes(P(Q) \oplus C_0) W ⊗ ( P ( Q ) ⊕ C 0 ) into C C C .
The main part of the proof is in the following lemma.
Lemma 5.3. Let s : W → B s: W \to B s : W → B be a graded section of the surjection B → W B \to W B → W . Then
Φ : W ⊗ P ( Q ) → s ⊗ incl B ⊗ Q → μ Q Q \Phi: W \otimes P(Q) \xrightarrow{s \otimes\operatorname{incl}} B \otimes Q \xrightarrow{\mu_Q} Q Φ : W ⊗ P ( Q ) s ⊗ incl B ⊗ Q μ Q Q is a graded injection.
Proof. It is easy to observe that Φ \Phi Φ is graded as all maps involved are graded. Assume that x ∈ W ⊗ P ( Q ) x \in W \otimes P(Q) x ∈ W ⊗ P ( Q ) is nonzero. We need to show that Φ ( x ) ≠ 0 \Phi(x) \ne0 Φ ( x ) = 0 . Let us assume that
x = x 0 + ⋯ + x n x = x_0 + \cdots+ x_n x = x 0 + ⋯ + x n
with x i ∈ W i ⊗ P ( Q ) x_i \in W_i \otimes P(Q) x i ∈ W i ⊗ P ( Q ) and $x_n \ne 0. We calculate that for any w i ⊗ ( c + W ) ∈ W i ⊗ P ( Q ) w_i \otimes(c + W) \in W_i \otimes P(Q) w i ⊗ ( c + W ) ∈ W i ⊗ P ( Q ) we get
Δ Q ( Φ ( w i ⊗ ( c + W ) ) ) = Δ Q ( s ( w i ) ⋅ ( c + W ) ) = Δ B ( s ( w i ) ) ⋅ Δ Q ( c + W ) = Δ B ( s ( w i ) ) ⋅ ( 1 C ⊗ ( c + W ) ) = ( s ( w i ) ⊗ 1 B + v ) ⋅ ( 1 C ⊗ ( c + W ) ) = w i ⊗ ( c + W ) + u , \begin{aligned}
\Delta_Q\bigl(\Phi(w_i \otimes(c + W))\bigr)
&=\Delta_Q\bigl(s(w_i)\cdot(c+W)\bigr)\\
&=\Delta_B(s(w_i))\cdot\Delta_Q(c+W)\\
&=\Delta_B(s(w_i))\cdot(1_C\otimes(c+W))\\
&=(s(w_i)\otimes1_B+v)\cdot(1_C\otimes(c+W))\\
&=w_i\otimes(c+W)+u,
\end{aligned} Δ Q ( Φ ( w i ⊗ ( c + W )) ) = Δ Q ( s ( w i ) ⋅ ( c + W ) ) = Δ B ( s ( w i )) ⋅ Δ Q ( c + W ) = Δ B ( s ( w i )) ⋅ ( 1 C ⊗ ( c + W )) = ( s ( w i ) ⊗ 1 B + v ) ⋅ ( 1 C ⊗ ( c + W )) = w i ⊗ ( c + W ) + u ,
where v ∈ B < i ⊗ B v\in B_{<i}\otimes B v ∈ B < i ⊗ B with B < i = B 0 ⊕ ⋯ ⊕ B i − 1 B_{<i}=B_0\oplus\cdots\oplus B_{i-1} B < i = B 0 ⊕ ⋯ ⊕ B i − 1 and u ∈ C < i ⊗ Q u\in C_{<i}\otimes Q u ∈ C < i ⊗ Q with C < i = C 0 ⊕ ⋯ ⊕ C i − 1 C_{<i}=C_0\oplus\cdots\oplus C_{i-1} C < i = C 0 ⊕ ⋯ ⊕ C i − 1 . By linearity, we conclude that
Δ Q ( Φ ( x ) ) = x n + u ′ , \Delta_Q(\Phi(x))=x_n+u', Δ Q ( Φ ( x )) = x n + u ′ ,
where u ′ ∈ C < n ⊗ Q u'\in C_{<n}\otimes Q u ′ ∈ C < n ⊗ Q . Since x n + C < n ⊗ Q x_n+C_{<n}\otimes Q x n + C < n ⊗ Q is nonzero, Φ ( x ) ≠ 0 \Phi(x)\ne0 Φ ( x ) = 0 . □
Proof of Theorem 5.2. Let r : Q → C r:Q\to C r : Q → C be a graded section of q : C → Q q:C\to Q q : C → Q . Then W ⊕ r ( Q ) = C W\oplus r(Q)=C W ⊕ r ( Q ) = C . Thus
Ψ : W ⊗ ( P ( Q ) ⊕ C 0 ) ⟶ C w ⊗ ( c + W ) ⟼ r ( s ( w ) ⋅ c + W ) w ⊗ 1 C ⟼ w \begin{aligned}
\Psi: W\otimes(P(Q)\oplus C_0)&\longrightarrow C\\
w\otimes(c+W)&\longmapsto r(s(w)\cdot c+W)\\
w\otimes1_C&\longmapsto w
\end{aligned} Ψ : W ⊗ ( P ( Q ) ⊕ C 0 ) w ⊗ ( c + W ) w ⊗ 1 C ⟶ C ⟼ r ( s ( w ) ⋅ c + W ) ⟼ w
is a monomorphism. □
For more applicability, we finish this section by identifying a particular subspace of P ( Q ) P(Q) P ( Q ) .
Lemma 5.4. ( P ( C ) + W ) / W (P(C)+W)/W ( P ( C ) + W ) / W is a subspace of P ( Q ) P(Q) P ( Q ) , where P ( C ) P(C) P ( C ) are the primitives of C C C .
Proof. This follows immediately because Δ C ( c ) − 1 C ⊗ c = c ⊗ 1 C ∈ C ⊗ W \Delta_C(c)-1_C\otimes c=c\otimes1_C\in C\otimes W Δ C ( c ) − 1 C ⊗ c = c ⊗ 1 C ∈ C ⊗ W for every c ∈ P ( C ) c\in P(C) c ∈ P ( C ) . □
The last lemma has the following direct consequence.
Corollary 5.5. W ⊗ [ C 0 ⊕ P ( C ) / ( P ( C ) ∩ W ) ] W\otimes[C_0\oplus P(C)/(P(C)\cap W)] W ⊗ [ C 0 ⊕ P ( C ) / ( P ( C ) ∩ W )] injects into C C C .
In our applications, we actually want to use the dual statement for B B B -comodule algebras A A A . To this end, the dual of W W W is
V : = im ( A → Δ A B ⊗ A → id ⊗ ϵ A B ⊗ A 0 ≅ B ) V:=\operatorname{im}\left(A\xrightarrow{\Delta_A}B\otimes A\xrightarrow{\operatorname{id}\otimes\epsilon_A}B\otimes A_0\cong B\right) V := im ( A Δ A B ⊗ A id ⊗ ϵ A B ⊗ A 0 ≅ B )
and the dual of P ( C ) = Q ( A ) = A + / A + 2 P(C)=Q(A)=A_+/A_+^2 P ( C ) = Q ( A ) = A + / A + 2 . We denote composition
id ⊗ ϵ A ∘ Δ A \operatorname{id}\otimes\epsilon_A\circ\Delta_A id ⊗ ϵ A ∘ Δ A
by δ : A → B \delta:A\to B δ : A → B and call this the structure homomorphism of the B B B -comodule algebra structure of A A A . This leads to the following statement for a graded connected bialgebra B B B and a graded connected B B B -comodule algebra A A A both of finite type.
Corollary 5.6. A A A surjects onto V ⊗ ( A 0 ⊕ ker ( δ : Q ( A ) → Q ( V ) ) ) V\otimes(A_0\oplus\ker(\delta:Q(A)\to Q(V))) V ⊗ ( A 0 ⊕ ker ( δ : Q ( A ) → Q ( V ))) .
Consequences for cohomology of symplectic groups
The goal of this section is to establish lower bounds for the high-dimensional cohomology of symplectic groups of number rings.
A source of indecomposables
Borel–Serre proved symplectic groups and general linear groups are virtual duality groups.
Theorem 6.1 (Borel–Serre). Let R R R be a number ring with field of fractions F F F . Let r 1 r_1 r 1 be the number real embeddings of F F F and r 2 r_2 r 2 pairs of complex embeddings of F F F . Let M M M be a projective R R R -module of rank n n n and let W W W be a symplectic module of genus n n n . The virtual cohomological dimension of GL ( M ) \operatorname{GL}(M) GL ( M ) is
ν n GL : = r 1 ( n + 1 2 ) + r 2 n 2 − n \nu_n^{\operatorname{GL}} := r_1 \binom{n+1}{2} + r_2 n^2 - n ν n GL := r 1 ( 2 n + 1 ) + r 2 n 2 − n and the virtual cohomological dimension of Sp ( W ) \operatorname{Sp}(W) Sp ( W ) is
ν 2 n ω : = r 1 ( n ( n + 1 ) ) + r 2 ( n ( 2 n + 1 ) ) − n . \nu_{2n}^{\omega} := r_1(n(n+1)) + r_2(n(2n+1)) - n. ν 2 n ω := r 1 ( n ( n + 1 )) + r 2 ( n ( 2 n + 1 )) − n . The virtual dualizing module of Sp ( W ) \operatorname{Sp}(W) Sp ( W ) is St ω ( W ) \operatorname{St}^{\omega}(W) St ω ( W ) and St ( M ) \operatorname{St}(M) St ( M ) is the dualizing module of SL ( M ) \operatorname{SL}(M) SL ( M ) . A tensor product of St ( M ) \operatorname{St}(M) St ( M ) with a one-dimensional representation is the dualizing module of GL ( M ) \operatorname{GL}(M) GL ( M ) ; see Putman–Studenmund [16 Theorem C] .
Recall
B : = ⨁ [ P ] , i H i ( GL ( P ) ; St ( P ⊗ R F ) ) \mathbf{B} := \bigoplus_{[P],i} H_i(\operatorname{GL}(P);\operatorname{St}(P\otimes_R F)) B := [ P ] , i ⨁ H i ( GL ( P ) ; St ( P ⊗ R F ))
and
A : = ⨁ n , i H i ( Sp 2 n ( R ) ; St ω ( F 2 n ) ) . \mathbf{A} := \bigoplus_{n,i} H_i(\operatorname{Sp}_{2n}(R);\operatorname{St}^{\omega}(F^{2n})). A := n , i ⨁ H i ( Sp 2 n ( R ) ; St ω ( F 2 n )) .
Let δ : A → B \delta:\mathbf{A}\to\mathbf{B} δ : A → B denote the structure homomorphism and let V = im ( δ ) ⊆ B \mathbf{V}=\operatorname{im}(\delta)\subseteq\mathbf{B} V = im ( δ ) ⊆ B . In particular, δ \delta δ is given by the composition
H i ( Sp 2 n ( R ) ; St ω ( F 2 n ) ) ⟶ H i ( Sp 2 n ( R ) ; ⨁ L ∈ L ( R 2 n ) St ( L ) ) ≅ ⨁ [ L ] , L ∈ L ( R 2 n ) H i ( Stab Sp 2 n ( R ) ( L ) ; St ( L ⊗ R F ) ) H_i(\operatorname{Sp}_{2n}(R);\operatorname{St}^{\omega}(F^{2n}))
\longrightarrow H_i\left(\operatorname{Sp}_{2n}(R);\bigoplus_{L\in\mathcal{L}(R^{2n})}\operatorname{St}(L)\right)
\cong\bigoplus_{[L],L\in\mathcal{L}(R^{2n})}H_i(\operatorname{Stab}_{\operatorname{Sp}_{2n}(R)}(L);\operatorname{St}(L\otimes_R F)) H i ( Sp 2 n ( R ) ; St ω ( F 2 n )) ⟶ H i Sp 2 n ( R ) ; L ∈ L ( R 2 n ) ⨁ St ( L ) ≅ [ L ] , L ∈ L ( R 2 n ) ⨁ H i ( Stab Sp 2 n ( R ) ( L ) ; St ( L ⊗ R F ))
⟶ ⨁ [ L ] , rk ( L ) = n H i ( GL ( L ) ; St ( L ⊗ R F ) ) , \longrightarrow\bigoplus_{[L],\operatorname{rk}(L)=n}H_i(\operatorname{GL}(L);\operatorname{St}(L\otimes_R F)), ⟶ [ L ] , rk ( L ) = n ⨁ H i ( GL ( L ) ; St ( L ⊗ R F )) ,
where L ( W ) \mathcal{L}(W) L ( W ) denotes the set of Lagrangians of a symplectic R R R -module W W W , the first map is induced by the coproduct ⨁ L ∈ L ( R 2 n ) Δ L ⊂ R 2 n \bigoplus_{L\in\mathcal{L}(R^{2n})}\Delta_{L\subset R^{2n}} ⨁ L ∈ L ( R 2 n ) Δ L ⊂ R 2 n , and the second map by the projection Stab Sp 2 n ( R ) ( L ) → GL ( L ) \operatorname{Stab}_{\operatorname{Sp}_{2n}(R)}(L)\to\operatorname{GL}(L) Stab Sp 2 n ( R ) ( L ) → GL ( L ) .
We will prove that the kernel
ker ( δ : Q ( A ) → Q ( V ) ) \ker(\delta:\mathsf{Q}(\mathbf{A})\to\mathsf{Q}(\mathbf{V})) ker ( δ : Q ( A ) → Q ( V ))
is large for degree reasons.
Lemma 6.2. Let k k k be a field of characteristic zero. Let R R R be a number ring with field of fractions F F F . Let r 1 r_1 r 1 be the number of real embeddings of F F F and r 2 r_2 r 2 be the number of pairs of complex embeddings of F F F . Let
S = { ( n , i ) ∣ i = ν 2 n ω − j with j ≤ 2 ( r 1 + 2 r 2 ) ( n − 1 ) − 1 } ∪ { ( 1 , i ) ∣ i > r 1 + r 2 − 1 } ⊂ N 2 . S=\{(n,i)\mid i=\nu_{2n}^{\omega}-j\text{ with }j\le2(r_1+2r_2)(n-1)-1\}\cup\{(1,i)\mid i>r_1+r_2-1\}\subset\mathbb{N}^2. S = {( n , i ) ∣ i = ν 2 n ω − j with j ≤ 2 ( r 1 + 2 r 2 ) ( n − 1 ) − 1 } ∪ {( 1 , i ) ∣ i > r 1 + r 2 − 1 } ⊂ N 2 . There is a natural injection
⨁ ( n , i ) ∈ S A n , i ↪ ker ( δ : Q ( A ) → Q ( V ) ) . \bigoplus_{(n,i)\in S} A_{n,i}\hookrightarrow\ker(\delta:\mathsf{Q}(\mathbf{A})\to\mathsf{Q}(\mathbf{V})). ( n , i ) ∈ S ⨁ A n , i ↪ ker ( δ : Q ( A ) → Q ( V )) .
Proof. Suppose α ≠ 0 ∈ A n , i \alpha\ne0 \in A_{n,i} α = 0 ∈ A n , i is decomposable. Then n > 1 n > 1 n > 1 and α \alpha α is a sum of products of elements β \beta β and γ \gamma γ of the form β ≠ 0 ∈ A a , b \beta\ne0 \in A_{a,b} β = 0 ∈ A a , b and γ ≠ 0 ∈ A c , d \gamma\ne0 \in A_{c,d} γ = 0 ∈ A c , d with n = a + c n = a + c n = a + c , i = b + d i = b + d i = b + d , and a , c > 0 a,c > 0 a , c > 0 . By Theorem 6.1 , b ≤ ν 2 a ω b \le\nu^\omega_{2a} b ≤ ν 2 a ω and d ≤ ν 2 c ω d \le\nu^\omega_{2c} d ≤ ν 2 c ω . Thus,
i ≤ ν 2 a ω + ν 2 c ω ≤ ν 2 ω + ν 2 n − 2 ω = ν 2 n ω − 2 ( r 1 + 2 r 2 ) ( n − 1 ) . i \le\nu^\omega_{2a}+\nu^\omega_{2c} \le\nu^\omega_2+\nu^\omega_{2n-2}=\nu^\omega_{2n}-2(r_1+2r_2)(n-1). i ≤ ν 2 a ω + ν 2 c ω ≤ ν 2 ω + ν 2 n − 2 ω = ν 2 n ω − 2 ( r 1 + 2 r 2 ) ( n − 1 ) .
The second inequality follows from convexity of the function n ↦ ν 2 n ω n \mapsto\nu^\omega_{2n} n ↦ ν 2 n ω . Hence
⨁ ( n , i ) ∈ S A n , i ↪ Q ( A ) . \bigoplus_{(n,i)\in S} A_{n,i} \hookrightarrow Q(A). ( n , i ) ∈ S ⨁ A n , i ↪ Q ( A ) .
Theorem 6.1 implies B n , i = 0 B_{n,i}=0 B n , i = 0 for i > ν n i>\nu_n i > ν n . In particular, B n , i = 0 B_{n,i}=0 B n , i = 0 if i > ν 2 n − 2 ω − 2 ( r 1 + 2 r 2 ) ( n − 1 ) ≥ ν n i>\nu^\omega_{2n-2}-2(r_1+2r_2)(n-1)\ge\nu_n i > ν 2 n − 2 ω − 2 ( r 1 + 2 r 2 ) ( n − 1 ) ≥ ν n and if n = 1 n=1 n = 1 and i > ν 1 = r 1 + r 2 − 1 i>\nu_1=r_1+r_2-1 i > ν 1 = r 1 + r 2 − 1 . Thus B n , i = 0 B_{n,i}=0 B n , i = 0 if ( n , i ) ∈ S (n,i)\in S ( n , i ) ∈ S . As Q ( V ) Q(V) Q ( V ) is a quotient of B B B , it also vanishes in this range. The claim follows. □
The best known homological stability range for symplectic groups is due to Sierra–Wahl [18 ] .
Theorem 6.3 (Sierra–Wahl). Let R R R be a Dedekind domain and i ≤ 2 n − 8 3 i \le\frac{2n-8}{3} i ≤ 3 2 n − 8 . Then H i ( Sp 2 n ( R ) ) ≅ H i ( Sp 2 n + 2 ( R ) ) H^i(\operatorname{Sp}_{2n}(R)) \cong H^i(\operatorname{Sp}_{2n+2}(R)) H i ( Sp 2 n ( R )) ≅ H i ( Sp 2 n + 2 ( R )) .
Borel [4 ] computed the stable cohomology.
Theorem 6.4 (Borel). Let R R R be a number ring with field of fractions F F F . Let k k k be a field of characteristic zero. Let r 1 r_1 r 1 be the number of real embeddings of F F F and r 2 r_2 r 2 the number of pairs of complex embeddings of F F F . Then the stable cohomology H ∗ ( Sp ∞ ( R ) ) H^*(\operatorname{Sp}_{\infty}(R)) H ∗ ( Sp ∞ ( R )) is the graded algebra
k [ { x i , j ∣ i ≥ 1 , 1 ≤ j ≤ r 1 } ] ⊗ Λ k [ { y a , b ∣ a ≥ 1 , 1 ≤ b ≤ r 2 } ] k[\{x_{i,j}\mid i\ge1,1\le j\le r_1\}]\otimes\Lambda_k[\{y_{a,b}\mid a\ge1,1\le b\le r_2\}] k [{ x i , j ∣ i ≥ 1 , 1 ≤ j ≤ r 1 }] ⊗ Λ k [{ y a , b ∣ a ≥ 1 , 1 ≤ b ≤ r 2 }] with ∣ x i , j ∣ = 4 i − 2 |x_{i,j}|=4i-2 ∣ x i , j ∣ = 4 i − 2 and ∣ y a , b ∣ = 4 a − 1 |y_{a,b}|=4a-1 ∣ y a , b ∣ = 4 a − 1 .
Lower bounds for the image of the structure homomorphism in the vcd
Given a symplectic module W W W , recall that L ( W ) \mathcal{L}(W) L ( W ) denotes the set of Lagrangians of W W W . The structure homomorphism δ : A → B \delta:A\to B δ : A → B is induced by a map
δ ω : St ω ( W ) ⟶ ⨁ L ∈ L ( W ) St ( L ) \delta^\omega:\operatorname{St}^\omega(W)\longrightarrow\bigoplus_{L\in\mathcal{L}(W)}\operatorname{St}(L) δ ω : St ω ( W ) ⟶ L ∈ L ( W ) ⨁ St ( L )
defined as the sum of the maps Δ L ⊂ W \Delta_{L\subset W} Δ L ⊂ W .
For M M M a rank d d d -module, let
δ d − 1 : St ( M ) ⟶ ⨁ N ⊆ M rk ( N ) = d − 1 St ( N ) ⊗ St ( M / N ) \delta^{d-1}:\operatorname{St}(M)\longrightarrow\bigoplus_{\substack{N\subseteq M\\ \operatorname{rk}(N)=d-1}}\operatorname{St}(N)\otimes\operatorname{St}(M/N) δ d − 1 : St ( M ) ⟶ N ⊆ M rk ( N ) = d − 1 ⨁ St ( N ) ⊗ St ( M / N )
be the sum of the maps Δ N ⊆ M \Delta_{N\subseteq M} Δ N ⊆ M with rk ( N ) = d − 1 \operatorname{rk}(N)=d-1 rk ( N ) = d − 1 . Let
D d − 1 : ⨁ J d ⊂ J d + 1 ⊂ ⋯ ⊂ J n rk ( J i ) = i J n ∈ L ( W ) St ( J d ) ⊗ St ( J d + 1 / J d ) ⊗ ⋯ ⊗ St ( J n / J n − 1 ) ⟶ ⨁ J d − 1 ⊂ J d ⊂ ⋯ ⊂ J n rk ( J i ) = i J n ∈ L ( W ) St ( J d − 1 ) ⊗ St ( J d / J d − 1 ) ⊗ ⋯ ⊗ St ( J n / J n − 1 ) D^{d-1}:\bigoplus_{\substack{J_d\subset J_{d+1}\subset\cdots\subset J_n\\ \operatorname{rk}(J_i)=i\\ J_n\in\mathcal{L}(W)}}\operatorname{St}(J_d)\otimes\operatorname{St}(J_{d+1}/J_d)\otimes\cdots\otimes\operatorname{St}(J_n/J_{n-1})
\longrightarrow
\bigoplus_{\substack{J_{d-1}\subset J_d\subset\cdots\subset J_n\\ \operatorname{rk}(J_i)=i\\ J_n\in\mathcal{L}(W)}}\operatorname{St}(J_{d-1})\otimes\operatorname{St}(J_d/J_{d-1})\otimes\cdots\otimes\operatorname{St}(J_n/J_{n-1}) D d − 1 : J d ⊂ J d + 1 ⊂ ⋯ ⊂ J n rk ( J i ) = i J n ∈ L ( W ) ⨁ St ( J d ) ⊗ St ( J d + 1 / J d ) ⊗ ⋯ ⊗ St ( J n / J n − 1 ) ⟶ J d − 1 ⊂ J d ⊂ ⋯ ⊂ J n rk ( J i ) = i J n ∈ L ( W ) ⨁ St ( J d − 1 ) ⊗ St ( J d / J d − 1 ) ⊗ ⋯ ⊗ St ( J n / J n − 1 )
be the map induced by δ d − 1 \delta^{d-1} δ d − 1 . Let
δ t o t a l : St ω ( W ) ⟶ ⨁ J 1 ⊂ J 2 ⊂ ⋯ ⊂ J n rk ( J i ) = i J n ∈ L ( W ) St ( J 1 ) ⊗ St ( J 2 / J 1 ) ⊗ ⋯ ⊗ St ( J n / J n − 1 ) . \delta^{\mathrm{total}}:\operatorname{St}^{\omega}(W)\longrightarrow\bigoplus_{\substack{J_1\subset J_2\subset\cdots\subset J_n\\ \operatorname{rk}(J_i)=i\\ J_n\in\mathcal{L}(W)}} \operatorname{St}(J_1)\otimes\operatorname{St}(J_2/J_1)\otimes\cdots\otimes\operatorname{St}(J_n/J_{n-1}). δ total : St ω ( W ) ⟶ J 1 ⊂ J 2 ⊂ ⋯ ⊂ J n rk ( J i ) = i J n ∈ L ( W ) ⨁ St ( J 1 ) ⊗ St ( J 2 / J 1 ) ⊗ ⋯ ⊗ St ( J n / J n − 1 ) .
be the composition
δ t o t a l : = D 1 ∘ D 2 ∘ ⋯ ∘ D n − 1 ∘ δ ω . \delta^{\mathrm{total}}:=D^1\circ D^2\circ\cdots\circ D^{n-1}\circ\delta^\omega. δ total := D 1 ∘ D 2 ∘ ⋯ ∘ D n − 1 ∘ δ ω .
Since J 1 , J 2 / J 1 , … , J n / J n − 1 J_1,J_2/J_1,\ldots,J_n/J_{n-1} J 1 , J 2 / J 1 , … , J n / J n − 1 are rank one, their Steinberg modules are naturally isomorphic to k \mathbb{k} k . The indexing set of the direct sum is the set of full flags of isotropic summands of W W W . Thus, we may view δ t o t a l \delta^{\mathrm{total}} δ total as a map to the top cellular chains on the symplectic Tits building:
δ t o t a l : St ω ( W ) ⟶ C ~ n − 1 ( T ω ( W ) ) . \delta^{\mathrm{total}}:\operatorname{St}^{\omega}(W)\longrightarrow\widetilde{C}_{n-1}(\mathcal{T}^{\omega}(W)). δ total : St ω ( W ) ⟶ C n − 1 ( T ω ( W )) .
Here n n n is the genus of W W W .
Lemma 6.5. The map δ t o t a l : St ω ( W ) ⟶ C ~ n − 1 ( T ω ( W ) ) \delta^{\mathrm{total}}:\operatorname{St}^{\omega}(W)\longrightarrow\widetilde{C}_{n-1}(\mathcal{T}^{\omega}(W)) δ total : St ω ( W ) ⟶ C n − 1 ( T ω ( W )) agrees with the inclusion
H ~ n − 1 ( T ω ( W ) ) ⟶ C ~ n − 1 ( T ω ( W ) ) . \widetilde{H}_{n-1}(\mathcal{T}^{\omega}(W))\longrightarrow\widetilde{C}_{n-1}(\mathcal{T}^{\omega}(W)). H n − 1 ( T ω ( W )) ⟶ C n − 1 ( T ω ( W )) .
Proof. Let ( v ⃗ 1 , … , v ⃗ 2 n ) (\vec{v}_1,\ldots,\vec{v}_{2n}) ( v 1 , … , v 2 n ) be an ordered symplectic basis of W ⊗ R F W\otimes_R F W ⊗ R F . After unpacking the definition, we see that δ t o t a l ( [ v ⃗ 1 , … , v ⃗ 2 n ] ) = \delta^{\mathrm{total}}([\vec{v}_1,\ldots,\vec{v}_{2n}])= δ total ([ v 1 , … , v 2 n ]) =
⨁ σ ∈ S 2 n ω χ ( σ ) [ v ⃗ σ ( 1 ) ] ⊗ [ v ⃗ σ ( 3 ) ] ⊗ ⋯ ⊗ [ v ⃗ σ ( 2 n − 1 ) ] . \bigoplus_{\sigma\in\mathfrak{S}_{2n}^{\omega}}\chi(\sigma)[\vec{v}_{\sigma(1)}]\otimes[\vec{v}_{\sigma(3)}]\otimes\cdots\otimes[\vec{v}_{\sigma(2n-1)}]. σ ∈ S 2 n ω ⨁ χ ( σ ) [ v σ ( 1 ) ] ⊗ [ v σ ( 3 ) ] ⊗ ⋯ ⊗ [ v σ ( 2 n − 1 ) ] .
Here [ v ⃗ σ ( 1 ) ] ⊗ [ v ⃗ σ ( 3 ) ] ⊗ ⋯ ⊗ [ v ⃗ σ ( 2 n − 1 ) ] [\vec{v}_{\sigma(1)}]\otimes[\vec{v}_{\sigma(3)}]\otimes\cdots\otimes[\vec{v}_{\sigma(2n-1)}] [ v σ ( 1 ) ] ⊗ [ v σ ( 3 ) ] ⊗ ⋯ ⊗ [ v σ ( 2 n − 1 ) ] is in the summand associated to the flag
span ( v ⃗ σ ( 1 ) ) ⊂ span ( v ⃗ σ ( 1 ) , v ⃗ σ ( 3 ) ) ⊂ ⋯ ⊂ span ( v ⃗ σ ( 1 ) , … , v ⃗ σ ( 2 n − 1 ) ) . \operatorname{span}(\vec{v}_{\sigma(1)})\subset\operatorname{span}(\vec{v}_{\sigma(1)},\vec{v}_{\sigma(3)})\subset\cdots\subset\operatorname{span}(\vec{v}_{\sigma(1)},\ldots,\vec{v}_{\sigma(2n-1)}). span ( v σ ( 1 ) ) ⊂ span ( v σ ( 1 ) , v σ ( 3 ) ) ⊂ ⋯ ⊂ span ( v σ ( 1 ) , … , v σ ( 2 n − 1 ) ) .
The signed sum of these flags is exactly a symplectic apartment. □
Let W W W have genus n n n . The quotient map T ω ( W ) → T ω ( W ) / Sp ( W ) \mathcal{T}^{\omega}(W)\to\mathcal{T}^{\omega}(W)/\operatorname{Sp}(W) T ω ( W ) → T ω ( W ) / Sp ( W ) is invariant and hence induces a map
H 0 ( Sp ( W ) ; St ω ( W ) ) ⟶ H ~ n − 1 ( T ω ( W ) / Sp ( W ) ) . H_0(\operatorname{Sp}(W);\operatorname{St}^{\omega}(W))\longrightarrow\widetilde{H}_{n-1}(\mathcal{T}^{\omega}(W)/\operatorname{Sp}(W)). H 0 ( Sp ( W ) ; St ω ( W )) ⟶ H n − 1 ( T ω ( W ) / Sp ( W )) .
We now recall the following result of Brück–Himes [3 Theorem 1.2 and Section 3] .
Theorem 6.6 (Brück–Himes). Let R R R be a Dedekind domain with class number c c c and let W W W be a symplectic module with genus n n n . There is a map
π : H ~ n − 1 ( T ω ( W ) / Sp ( W ) ) ⟶ k ( c − 1 ) n \pi:\widetilde{H}_{n-1}(\mathcal{T}^{\omega}(W)/\operatorname{Sp}(W))\longrightarrow\mathbb{k}^{(c-1)n} π : H n − 1 ( T ω ( W ) / Sp ( W )) ⟶ k ( c − 1 ) n such that the composition
H 0 ( Sp ( W ) ; St ω ( W ) ) ⟶ H ~ n − 1 ( T ω ( W ) / Sp ( W ) ) ⟶ k ( c − 1 ) n H_0(\operatorname{Sp}(W);\operatorname{St}^{\omega}(W))\longrightarrow\widetilde{H}_{n-1}(\mathcal{T}^{\omega}(W)/\operatorname{Sp}(W))\longrightarrow\mathbb{k}^{(c-1)n} H 0 ( Sp ( W ) ; St ω ( W )) ⟶ H n − 1 ( T ω ( W ) / Sp ( W )) ⟶ k ( c − 1 ) n is surjective.
We obtain the following corollary on the image of the structure homomorphism.
Corollary 6.7. Let R R R be a Dedekind domain with class number c c c . Let k \mathbb{k} k be a field. There is a surjection
V n , 0 ↠ k ( c − 1 ) n . V_{n,0}\twoheadrightarrow\mathbb{k}^{(c-1)n}. V n , 0 ↠ k ( c − 1 ) n .
Proof. Lemma 6.5 implies that
St ω ( R 2 n ) → δ t o t a l C n − 1 ( T ω ( R 2 n ) ) → ∼ C ~ ∗ ( T ω ( R 2 n ) ) \operatorname{St}^{\omega}(R^{2n}) \xrightarrow{\delta^{\mathrm{total}}} C_{n-1}(\mathcal{T}^{\omega}(R^{2n})) \xrightarrow{\sim} \widetilde{C}_{*}(\mathcal{T}^{\omega}(R^{2n})) St ω ( R 2 n ) δ total C n − 1 ( T ω ( R 2 n )) ∼ C ∗ ( T ω ( R 2 n ))
is a (degree n − 1 n-1 n − 1 ) quasiisomorphism and thus induces an isomorphism
H 0 ( Sp 2 n ( R ) ; St ω ( F 2 n ) ) ⟶ H 0 ( Sp 2 n ( R ) ; C ~ ∗ ( T ω ( F 2 n ) ) ) . H_{0}(\operatorname{Sp}_{2n}(R);\operatorname{St}^{\omega}(F^{2n})) \longrightarrow H_{0}(\operatorname{Sp}_{2n}(R);\widetilde{C}_{*}(\mathcal{T}^{\omega}(F^{2n}))). H 0 ( Sp 2 n ( R ) ; St ω ( F 2 n )) ⟶ H 0 ( Sp 2 n ( R ) ; C ∗ ( T ω ( F 2 n ))) .
There is an equivariant map
C ~ ∗ ( T ω ( R 2 n ) ) ⟶ C ~ ∗ ( T ω ( R 2 n ) / Sp 2 n ( R ) ) \widetilde{C}_{*}(\mathcal{T}^{\omega}(R^{2n})) \longrightarrow\widetilde{C}_{*}(\mathcal{T}^{\omega}(R^{2n})/\operatorname{Sp}_{2n}(R)) C ∗ ( T ω ( R 2 n )) ⟶ C ∗ ( T ω ( R 2 n ) / Sp 2 n ( R ))
and C ~ ∗ ( T ω ( R 2 n ) / Sp 2 n ( R ) ) \widetilde{C}_{*}(\mathcal{T}^{\omega}(R^{2n})/\operatorname{Sp}_{2n}(R)) C ∗ ( T ω ( R 2 n ) / Sp 2 n ( R )) is quas isomorphic to H ~ n − 1 ( T ω ( R 2 n ) / Sp 2 n ( R ) ) \widetilde{H}_{n-1}(\mathcal{T}^{\omega}(R^{2n})/\operatorname{Sp}_{2n}(R)) H n − 1 ( T ω ( R 2 n ) / Sp 2 n ( R )) on which Sp 2 n ( R ) \operatorname{Sp}_{2n}(R) Sp 2 n ( R ) acts trivially. Theorem 6.6 implies that the composition
H 0 ( Sp 2 n ( R ) ; St ω ( F 2 n ) ) ⟶ H 0 ( Sp 2 n ( R ) ; H ~ n − 1 ( T ω ( R 2 n ) / Sp 2 n ( R ) ) ) = H ~ n − 1 ( T ω ( R 2 n ) / Sp 2 n ( R ) ) → π k ( c − 1 ) n H_{0}(\operatorname{Sp}_{2n}(R);\operatorname{St}^{\omega}(F^{2n})) \longrightarrow H_{0}(\operatorname{Sp}_{2n}(R);\widetilde{H}_{n-1}(\mathcal{T}^{\omega}(R^{2n})/\operatorname{Sp}_{2n}(R))) = \widetilde{H}_{n-1}(\mathcal{T}^{\omega}(R^{2n})/\operatorname{Sp}_{2n}(R)) \xrightarrow{\pi} k^{(c-1)^{n}} H 0 ( Sp 2 n ( R ) ; St ω ( F 2 n )) ⟶ H 0 ( Sp 2 n ( R ) ; H n − 1 ( T ω ( R 2 n ) / Sp 2 n ( R ))) = H n − 1 ( T ω ( R 2 n ) / Sp 2 n ( R )) π k ( c − 1 ) n
is a surjection.
Observe that the map
H 0 ( Sp 2 n ( R ) ; ⨁ L ∈ L ( R 2 n ) St ( L ) ) ⟶ ∼ ⨁ [ L ] , L ∈ L ( R 2 n ) H 0 ( Stab Sp 2 n ( R ) ( L ) ; St ( L ⊗ F ) ) ⟶ ⨁ [ L ] , rk ( L ) = n H 0 ( GL ( L ) ; St ( L ⊗ F ) ) H_{0}\left(\operatorname{Sp}_{2n}(R);\bigoplus_{L\in\mathcal{L}(R^{2n})}\operatorname{St}(L)\right) \overset{\sim}{\longrightarrow} \bigoplus_{[L],L\in\mathcal{L}(R^{2n})} H_{0}(\operatorname{Stab}_{\operatorname{Sp}_{2n}(R)}(L);\operatorname{St}(L\otimes F)) \longrightarrow\bigoplus_{[L],\operatorname{rk}(L)=n} H_{0}(\operatorname{GL}(L);\operatorname{St}(L\otimes F)) H 0 Sp 2 n ( R ) ; L ∈ L ( R 2 n ) ⨁ St ( L ) ⟶ ∼ [ L ] , L ∈ L ( R 2 n ) ⨁ H 0 ( Stab Sp 2 n ( R ) ( L ) ; St ( L ⊗ F )) ⟶ [ L ] , rk ( L ) = n ⨁ H 0 ( GL ( L ) ; St ( L ⊗ F ))
is an isomorphism because Stab Sp 2 n ( R ) ( L ) \operatorname{Stab}_{\operatorname{Sp}_{2n}(R)}(L) Stab Sp 2 n ( R ) ( L ) acts on St ( L ) \operatorname{St}(L) St ( L ) through the projection Stab Sp 2 n ( R ) ( L ) → GL ( L ) \operatorname{Stab}_{\operatorname{Sp}_{2n}(R)}(L)\to\operatorname{GL}(L) Stab Sp 2 n ( R ) ( L ) → GL ( L ) .
Thus because δ t o t a l \delta^{\mathrm{total}} δ total factors through
δ ω : St ω ( F 2 n ) ⟶ ⨁ L ∈ L ( R 2 n ) St ( L ) , \delta^{\omega}:\operatorname{St}^{\omega}(F^{2n})\longrightarrow\bigoplus_{L\in\mathcal{L}(R^{2n})}\operatorname{St}(L), δ ω : St ω ( F 2 n ) ⟶ L ∈ L ( R 2 n ) ⨁ St ( L ) ,
the surjective composition
A n , 0 ≅ H 0 ( Sp 2 n ( R ) ; St ω ( F 2 n ) ) ⟶ k ( c − 1 ) n A_{n,0}\cong H_{0}(\operatorname{Sp}_{2n}(R);\operatorname{St}^{\omega}(F^{2n}))\longrightarrow k^{(c-1)^{n}} A n , 0 ≅ H 0 ( Sp 2 n ( R ) ; St ω ( F 2 n )) ⟶ k ( c − 1 ) n
factors through δ : A n , 0 → B n , 0 \delta:A_{n,0}\to B_{n,0} δ : A n , 0 → B n , 0 and thus through V n , 0 V_{n,0} V n , 0 . □
Lower bounds parallel to the vcd
Using items (1), (2), (3) of the following theorem, we obtain Corollary 1.2 except in the case m = 0 m=0 m = 0 which is Brück–Himes [3 Theorem 1.1] .
Theorem 6.8. Let k k k be a field of characteristic zero. Let R R R be a number ring with field of fractions F F F and class number c c c . Let r 1 r_{1} r 1 be the number real embeddings of F F F and r 2 r_{2} r 2 pairs of complex embeddings of F F F . Let
S = { ( n , i ) ∣ i = ν 2 n ω − j with j ≤ 2 ( r 1 + 2 r 2 ) ( n − 1 ) − 1 } ∪ { ( 1 , i ) ∣ i > r 1 + r 2 − 1 } ⊂ N 0 2 . S=\{(n,i)\mid i=\nu_{2n}^{\omega}-j\text{ with }j\leq2(r_{1}+2r_{2})(n-1)-1\}\cup\{(1,i)\mid i>r_{1}+r_{2}-1\}\subset\mathbb{N}_{0}^{2}. S = {( n , i ) ∣ i = ν 2 n ω − j with j ≤ 2 ( r 1 + 2 r 2 ) ( n − 1 ) − 1 } ∪ {( 1 , i ) ∣ i > r 1 + r 2 − 1 } ⊂ N 0 2 . There is a surjection from A A A onto
( ⨁ n V n , 0 ) ⊗ ( ⨁ ( n , i ) ∈ S A n , i ) \left(\bigoplus_{n}V_{n,0}\right)\otimes\left(\bigoplus_{(n,i)\in S}A_{n,i}\right) ( n ⨁ V n , 0 ) ⊗ ( n , i ) ∈ S ⨁ A n , i dim k ( V n , 0 ) ≥ ( c − 1 ) n \dim_{k}(V_{n,0})\geq(c-1)^{n} dim k ( V n , 0 ) ≥ ( c − 1 ) n ,
(Borel–Serre) A n , ν 2 n ω ≅ k A_{n,\nu_{2n}^{\omega}}\cong k A n , ν 2 n ω ≅ k ,
(Borel–Serre, Sierra–Wahl) For i = ν 2 n ω − j i=\nu_{2n}^{\omega}-j i = ν 2 n ω − j with j ≤ 2 n − 8 3 j\leq\frac{2n-8}{3} j ≤ 3 2 n − 8 , A n , i ≅ H j ( Sp ∞ ( R ) ) A_{n,i}\cong H^{j}(\operatorname{Sp}_{\infty}(R)) A n , i ≅ H j ( Sp ∞ ( R )) ,
(Borel) H ∗ ( Sp ∞ ( R ) ) ≅ k [ { x i , j ∣ i ≥ 1 , 1 ≤ j ≤ r 1 } ] ⊗ Λ k [ { y a , b ∣ a ≥ 1 , 1 ≤ b ≤ r 2 } ] H^{*}(\operatorname{Sp}_{\infty}(R))\cong k[\{x_{i,j}\mid i\geq1,1\leq j\leq r_{1}\}]\otimes\Lambda_{k}[\{y_{a,b}\mid a\geq1,1\leq b\leq r_{2}\}] H ∗ ( Sp ∞ ( R )) ≅ k [{ x i , j ∣ i ≥ 1 , 1 ≤ j ≤ r 1 }] ⊗ Λ k [{ y a , b ∣ a ≥ 1 , 1 ≤ b ≤ r 2 }] with ∣ x i , j ∣ = 4 i − 2 |x_{i,j}|=4i-2 ∣ x i , j ∣ = 4 i − 2 and ∣ y a , b ∣ = 4 a − 1 |y_{a,b}|=4a-1 ∣ y a , b ∣ = 4 a − 1 .
Proof. (1) This combines Corollary 5.6 and Lemma 6.2 .
(2) This is Corollary 6.7.
Borel–Serre duality Theorem 6.1 implies that
H 2 n ν ω ( Sp 2 n ( R ) ; St ω ( F 2 n ) ) ≅ H 0 ( Sp 2 n ( R ) ) ≅ k . H_{2n}^{\nu^\omega}(\operatorname{Sp}_{2n}(R); \operatorname{St}^{\omega}(F^{2n})) \cong H^0(\operatorname{Sp}_{2n}(R)) \cong\mathbb{k}. H 2 n ν ω ( Sp 2 n ( R ) ; St ω ( F 2 n )) ≅ H 0 ( Sp 2 n ( R )) ≅ k .
This combines Borel–Serre duality Theorem 6.1 and Sierra–Wahl’s homological stability result Theorem 6.3.
This is Borel’s calculation of the stable rational cohomology of symplectic groups Theorem 6.4.
Items (4) and (5) can be used to obtain more information.
Example 6.9. If R R R is the ring of integers in Q [ − 23 ] \mathbb{Q}[\sqrt{-23}] Q [ − 23 ] , then r 1 = 0 r_1=0 r 1 = 0 , r 2 = 1 r_2=1 r 2 = 1 , and c = 3 c=3 c = 3 ,
H 3 ( Sp 18 ( R ) ; Q ) ≅ H 3 ( Sp ∞ ( R ) ; Q ) ≅ Q , H^3(\operatorname{Sp}_{18}(R);\mathbb{Q}) \cong H^3(\operatorname{Sp}_{\infty}(R);\mathbb{Q}) \cong\mathbb{Q}, H 3 ( Sp 18 ( R ) ; Q ) ≅ H 3 ( Sp ∞ ( R ) ; Q ) ≅ Q ,
and ν 18 ω = 162 \nu_{18}^{\omega}=162 ν 18 ω = 162 . We conclude that
dim Q H 2 n 2 − 159 ( Sp 2 n ( R ) ; Q ) ≥ 2 n − 9 \dim_{\mathbb{Q}} H^{2n^2-159}(\operatorname{Sp}_{2n}(R);\mathbb{Q}) \geq2^{n-9} dim Q H 2 n 2 − 159 ( Sp 2 n ( R ) ; Q ) ≥ 2 n − 9
for n ≥ 9 n \geq9 n ≥ 9 .
Remark 6.10. The proof of Corollary 6.7 implies that there is a surjection V n , 0 → A n , 0 V_{n,0} \to A_{n,0} V n , 0 → A n , 0 . It would be interesting to investigate if one could use this observation and ideas from [13 ] to construct lower bounds for the rational cohomology of symplectic groups of number rings with class number equal to 1 that are not Euclidean.
Remark 6.11. Using Franke [7 Section 7] as in Speh–Venkataramana [17 Section 4] and Giansiracusa–Kupers–Tshishiku [8 Theorem 13] , one can describe the image of H ∗ ( Sp ∞ ( R ) ) H^*(\operatorname{Sp}_{\infty}(R)) H ∗ ( Sp ∞ ( R )) in H ∗ ( Sp 2 n ( R ) ) H^*(\operatorname{Sp}_{2n}(R)) H ∗ ( Sp 2 n ( R )) . Let us define
A n = k [ c 1 , … , c n ] / I A_n=\mathbb{k}[c_1,\ldots,c_n]/I A n = k [ c 1 , … , c n ] / I
where I I I is the ideal generated by the polynomials
∑ a + b = k ( − 1 ) b c a c b − δ k , 0 \sum_{a+b=k}(-1)^b c_a c_b-\delta_{k,0} a + b = k ∑ ( − 1 ) b c a c b − δ k , 0
for all k k k with the convention that c 0 = 1 c_0=1 c 0 = 1 and c i = 0 c_i=0 c i = 0 if i ∉ [ 0 , n ] i\notin[0,n] i ∈ / [ 0 , n ] . The ring map k [ x 1 , x 2 , … ] → A n \mathbb{k}[x_1,x_2,\ldots]\to A_n k [ x 1 , x 2 , … ] → A n sending x i x_i x i to c 2 i − 1 c_{2i-1} c 2 i − 1 is surjective. Identify ( A n ) ⊗ r 1 (A_n)^{\otimes r_1} ( A n ) ⊗ r 1 with the corresponding quotient of
k [ { c i , j ∣ 1 ≤ i ≤ n , 1 ≤ j ≤ r 1 } ] , \mathbb{k}[\{c_{i,j}\mid1\leq i\leq n,1\leq j\leq r_1\}], k [{ c i , j ∣ 1 ≤ i ≤ n , 1 ≤ j ≤ r 1 }] ,
then
k [ { x i , j ∣ 1 ≤ i , 1 ≤ j ≤ r 1 } ] ⟶ ( A n ) ⊗ r 1 \mathbb{k}[\{x_{i,j}\mid1\leq i,1\leq j\leq r_1\}]\longrightarrow(A_n)^{\otimes r_1} k [{ x i , j ∣ 1 ≤ i , 1 ≤ j ≤ r 1 }] ⟶ ( A n ) ⊗ r 1
given by x i , j x_{i,j} x i , j sent to c 2 i − 1 , j c_{2i-1,j} c 2 i − 1 , j is surjective and the image of H ∗ ( Sp ∞ ( R ) ) H^*(\operatorname{Sp}_{\infty}(R)) H ∗ ( Sp ∞ ( R )) in H ∗ ( Sp 2 n ( R ) ) H^*(\operatorname{Sp}_{2n}(R)) H ∗ ( Sp 2 n ( R )) is isomorphic to
( A n ) ⊗ r 1 ⊗ ⋀ k [ { y a , b ∣ 1 ≤ a ≤ n , 1 ≤ b ≤ r 2 } ] ( c n , 1 ⋯ c n , r 1 ⊗ y n , 1 ∧ ⋯ ∧ y n , r 2 ) . (A_n)^{\otimes r_1}\otimes\frac{\bigwedge_{\mathbb{k}}[\{y_{a,b}\mid1\leq a\leq n,1\leq b\leq r_2\}]}{(c_{n,1}\cdots c_{n,r_1}\otimes y_{n,1}\wedge\cdots\wedge y_{n,r_2})}. ( A n ) ⊗ r 1 ⊗ ( c n , 1 ⋯ c n , r 1 ⊗ y n , 1 ∧ ⋯ ∧ y n , r 2 ) ⋀ k [{ y a , b ∣ 1 ≤ a ≤ n , 1 ≤ b ≤ r 2 }] .
And one can find a basis given by
∏ c i , j α i , j ⊗ ∏ y a , b β a , b \prod c_{i,j}^{\alpha_{i,j}}\otimes\prod y_{a,b}^{\beta_{a,b}} ∏ c i , j α i , j ⊗ ∏ y a , b β a , b
with α i , j , β a , b ∈ { 0 , 1 } \alpha_{i,j},\beta_{a,b}\in\{0,1\} α i , j , β a , b ∈ { 0 , 1 } such that the product α n , 1 ⋯ α n , r 1 β n , 1 ⋯ β n , r 2 = 0 \alpha_{n,1}\cdots\alpha_{n,r_1}\beta_{n,1}\cdots\beta_{n,r_2}=0 α n , 1 ⋯ α n , r 1 β n , 1 ⋯ β n , r 2 = 0 . This gives a lower bound for H ∗ ( Sp 2 n ( R ) ) H^*(\operatorname{Sp}_{2n}(R)) H ∗ ( Sp 2 n ( R )) , many of which are in the range
∗ ≥ ν 2 n ω − r 1 ( 2 n − 3 ) − r 2 ( 4 n − 5 ) * \geq\nu_{2n}^{\omega}-r_1(2n-3)-r_2(4n-5) ∗ ≥ ν 2 n ω − r 1 ( 2 n − 3 ) − r 2 ( 4 n − 5 )
but outside the stable range and thus improves the lower bounds established in Theorem 6.8 (4) and (5).
Furthermore, one can adapt [2 Proof of Theorem 1.2 in Section 4.2] to show that all these images of stable classes are indecomposable, which would improve Lemma 6.2 .
Consequences for cohomology of congruence subgroups
In this section, we establish lower bounds for the cohomology of congruence subgroups of Sp 2 n ( Z ) \operatorname{Sp}_{2n}(\mathbb{Z}) Sp 2 n ( Z ) .
Definition 7.1. Let W W W be a symplectic R R R -module and I I I an ideal of R R R . Let Sp ( W , I ) \operatorname{Sp}(W,I) Sp ( W , I ) denote the kernel of
Sp ( W ) → Sp ( W / I ) . \operatorname{Sp}(W) \to\operatorname{Sp}(W/I). Sp ( W ) → Sp ( W / I ) . For W = R 2 n W=R^{2n} W = R 2 n , we denote Sp ( R 2 n , I ) \operatorname{Sp}(R^{2n},I) Sp ( R 2 n , I ) by Sp 2 n ( I ) \operatorname{Sp}_{2n}(I) Sp 2 n ( I ) . Similarly defined congruence subgroups of general linear groups.
Definition 7.2. Let U U U be an m m m -dimensional vector space over a field F F F . A ± \pm ± -orientation on U U U is a choice of basis of the one-dimensional F F F -vector space ⋀ F m U \bigwedge_F^m U ⋀ F m U up to multiplication by ± 1 \pm1 ± 1 .
If ( b ⃗ 1 , … , b ⃗ n ) (\vec{b}_1,\ldots,\vec{b}_n) ( b 1 , … , b n ) is an ordered basis of a free Z \mathbb{Z} Z -module N N N , then b ⃗ 1 ∧ ⋯ ∧ b ⃗ n \vec{b}_1\wedge\cdots\wedge\vec{b}_n b 1 ∧ ⋯ ∧ b n gives a ± \pm ± -orientation on the F p \mathbb{F}_p F p -vector space N / p N/p N / p . If T : N → N T:N\to N T : N → N is a Z \mathbb{Z} Z -linear isomorphism, then
b ⃗ 1 ∧ ⋯ ∧ b ⃗ n and T ( b ⃗ 1 ) ∧ ⋯ ∧ T ( b ⃗ n ) \vec{b}_1\wedge\cdots\wedge\vec{b}_n \quad\text{and}\quad T(\vec{b}_1)\wedge\cdots\wedge T(\vec{b}_n) b 1 ∧ ⋯ ∧ b n and T ( b 1 ) ∧ ⋯ ∧ T ( b n )
differ by the ratio det ( T ) = ± 1 \det(T)=\pm1 det ( T ) = ± 1 . Thus, β \beta β and β ′ \beta' β ′ induce the same ± \pm ± -orientation on N / p N/p N / p . In this way, N / p N/p N / p has a natural ± \pm ± -orientation induced by the Z \mathbb{Z} Z -module N N N .
The following is a corollary of Capovilla-Searle [6 Proposition 4.11] .
Theorem 7.3 (Capovilla-Searle). Let W W W be a symplectic Z \mathbb{Z} Z -module. Let J 1 J_1 J 1 and J 2 J_2 J 2 be isotropic summands of W W W . Then J 1 J_1 J 1 and J 2 J_2 J 2 are in the same orbit of Sp ( p W ) \operatorname{Sp}(pW) Sp ( p W ) if and only if J 1 / p = J 2 / p J_1/p=J_2/p J 1 / p = J 2 / p and the induced ± \pm ± -orientations agree. Additionally, every ± \pm ± -oriented isotropic subspace of W / p W/p W / p lifts to an isotropic subspace of W W W .
The following is Borel–Serre dual to Corollary 1.5. Note that since Sp 2 n ( p Z ) \operatorname{Sp}_{2n}(p\mathbb{Z}) Sp 2 n ( p Z ) is torsion free for p ≥ 3 p\geq3 p ≥ 3 , Borel–Serre duality applies without restrictions on k k k .
Theorem 7.4. Let k k k be a field and p p p be an odd prime. Then
dim k H i ( Sp 2 n + 2 ( p Z ) ; St 2 n + 2 ω ( Q ) ) ≥ p 2 n + 1 ( p − 1 2 ) dim k H i ( Sp 2 n ( p Z ) ; St 2 n ω ( Q ) ) . \dim_k H_i(\operatorname{Sp}_{2n+2}(p\mathbb{Z});\operatorname{St}^{\omega}_{2n+2}(\mathbb{Q})) \geq p^{2n+1}\left(\frac{p-1}{2}\right)\dim_k H_i(\operatorname{Sp}_{2n}(p\mathbb{Z});\operatorname{St}^{\omega}_{2n}(\mathbb{Q})). dim k H i ( Sp 2 n + 2 ( p Z ) ; St 2 n + 2 ω ( Q )) ≥ p 2 n + 1 ( 2 p − 1 ) dim k H i ( Sp 2 n ( p Z ) ; St 2 n ω ( Q )) .
Proof. Let S \mathcal{S} S denote the set of genus-1 symplectic summands of Z 2 n + 2 \mathbb{Z}^{2n+2} Z 2 n + 2 . Let J \mathcal{J} J denote the set of rank-1 isotropic summands of Z 2 n + 2 \mathbb{Z}^{2n+2} Z 2 n + 2 . Consider the composition of the following restrictions of the product and the coproduct
⨁ U ∈ S St ω ( U ) ⊗ St ω ( U ⊥ ) ⟶ St ω ( Z 2 n + 2 ) ⟶ ⨁ J ∈ J St ( J ) ⊗ St ω ( J ⊥ / J ) . \bigoplus_{U\in\mathcal{S}}\operatorname{St}^{\omega}(U)\otimes\operatorname{St}^{\omega}(U^\perp) \longrightarrow\operatorname{St}^{\omega}(\mathbb{Z}^{2n+2}) \longrightarrow\bigoplus_{J\in\mathcal{J}}\operatorname{St}(J)\otimes\operatorname{St}^{\omega}(J^\perp/J). U ∈ S ⨁ St ω ( U ) ⊗ St ω ( U ⊥ ) ⟶ St ω ( Z 2 n + 2 ) ⟶ J ∈ J ⨁ St ( J ) ⊗ St ω ( J ⊥ / J ) .
Applying the functor H i ( Sp 2 n + 2 ( p Z ) ; − ) H_i(\operatorname{Sp}_{2n+2}(p\mathbb{Z});-) H i ( Sp 2 n + 2 ( p Z ) ; − ) and Shapiro’s lemma gives maps
⨁ [ U ] ∈ S / Sp 2 n + 2 ( p Z ) H i ( Stab Sp 2 n + 2 ( p Z ) ( U ) ; St ω ( U ) ⊗ St ω ( U ⊥ ) ) ⟶ H i ( Sp 2 n + 2 ( p Z ) ; St ω ( Z 2 n + 2 ) ) \bigoplus_{[U]\in\mathcal{S}/\operatorname{Sp}_{2n+2}(p\mathbb{Z})} H_i(\operatorname{Stab}_{\operatorname{Sp}_{2n+2}(p\mathbb{Z})}(U);\operatorname{St}^{\omega}(U)\otimes\operatorname{St}^{\omega}(U^{\perp}))
\longrightarrow H_i(\operatorname{Sp}_{2n+2}(p\mathbb{Z});\operatorname{St}^{\omega}(\mathbb{Z}^{2n+2})) [ U ] ∈ S / Sp 2 n + 2 ( p Z ) ⨁ H i ( Stab Sp 2 n + 2 ( p Z ) ( U ) ; St ω ( U ) ⊗ St ω ( U ⊥ )) ⟶ H i ( Sp 2 n + 2 ( p Z ) ; St ω ( Z 2 n + 2 ))
⟶ ⨁ [ J ] ∈ J / Sp 2 n + 2 ( p Z ) H i ( Stab Sp 2 n + 2 ( p Z ) ( J ) ; St ( J ) ⊗ St ω ( J ⊥ / J ) ) . \longrightarrow\bigoplus_{[J]\in\mathcal{J}/\operatorname{Sp}_{2n+2}(p\mathbb{Z})} H_i(\operatorname{Stab}_{\operatorname{Sp}_{2n+2}(p\mathbb{Z})}(J);\operatorname{St}(J)\otimes\operatorname{St}^{\omega}(J^{\perp}/J)). ⟶ [ J ] ∈ J / Sp 2 n + 2 ( p Z ) ⨁ H i ( Stab Sp 2 n + 2 ( p Z ) ( J ) ; St ( J ) ⊗ St ω ( J ⊥ / J )) .
There is a natural isomorphism:
Stab Sp 2 n + 2 ( p Z ) ( U ) ≅ Sp ( U , p Z ) × Sp ( U ⊥ , p Z ) . \operatorname{Stab}_{\operatorname{Sp}_{2n+2}(p\mathbb{Z})}(U)\cong\operatorname{Sp}(U,p\mathbb{Z})\times\operatorname{Sp}(U^{\perp},p\mathbb{Z}). Stab Sp 2 n + 2 ( p Z ) ( U ) ≅ Sp ( U , p Z ) × Sp ( U ⊥ , p Z ) .
The projection from the parabolic to the Levi gives a natural surjective map:
Stab Sp 2 n + 2 ( p Z ) ( J ) ⟶ GL ( J , p Z ) × Sp ( J ⊥ / J , p Z ) . \operatorname{Stab}_{\operatorname{Sp}_{2n+2}(p\mathbb{Z})}(J)\longrightarrow\operatorname{GL}(J,p\mathbb{Z})\times\operatorname{Sp}(J^{\perp}/J,p\mathbb{Z}). Stab Sp 2 n + 2 ( p Z ) ( J ) ⟶ GL ( J , p Z ) × Sp ( J ⊥ / J , p Z ) .
After applying these maps and the Künneth isomorphism, we obtain maps
⨁ [ U ] ∈ S / Sp 2 n + 2 ( p Z ) H 0 ( Sp ( U , p Z ) ; St ω ( U ) ) ⊗ H i ( Sp ( U ⊥ , p Z ) ; St ω ( U ⊥ ) ) ⟶ H i ( Sp 2 n + 2 ( p Z ) ; St ω ( Z 2 n + 2 ) ) \bigoplus_{[U]\in\mathcal{S}/\operatorname{Sp}_{2n+2}(p\mathbb{Z})} H_0(\operatorname{Sp}(U,p\mathbb{Z});\operatorname{St}^{\omega}(U))\otimes H_i(\operatorname{Sp}(U^{\perp},p\mathbb{Z});\operatorname{St}^{\omega}(U^{\perp}))
\longrightarrow H_i(\operatorname{Sp}_{2n+2}(p\mathbb{Z});\operatorname{St}^{\omega}(\mathbb{Z}^{2n+2})) [ U ] ∈ S / Sp 2 n + 2 ( p Z ) ⨁ H 0 ( Sp ( U , p Z ) ; St ω ( U )) ⊗ H i ( Sp ( U ⊥ , p Z ) ; St ω ( U ⊥ )) ⟶ H i ( Sp 2 n + 2 ( p Z ) ; St ω ( Z 2 n + 2 ))
⟶ ⨁ [ J ] ∈ J / Sp 2 n + 2 ( p Z ) H 0 ( GL ( J , p Z ) ; St ( J ) ) ⊗ H i ( Sp ( J ⊥ / J , p Z ) ; St ω ( J ⊥ / J ) ) . \longrightarrow\bigoplus_{[J]\in\mathcal{J}/\operatorname{Sp}_{2n+2}(p\mathbb{Z})} H_0(\operatorname{GL}(J,p\mathbb{Z});\operatorname{St}(J))\otimes H_i(\operatorname{Sp}(J^{\perp}/J,p\mathbb{Z});\operatorname{St}^{\omega}(J^{\perp}/J)). ⟶ [ J ] ∈ J / Sp 2 n + 2 ( p Z ) ⨁ H 0 ( GL ( J , p Z ) ; St ( J )) ⊗ H i ( Sp ( J ⊥ / J , p Z ) ; St ω ( J ⊥ / J )) .
Call the composition of the above two maps Φ \Phi Φ and let Φ [ J ] \Phi_{[J]} Φ [ J ] be the projection onto the summand associated to [ J ] ∈ J / Sp 2 n + 2 ( p Z ) [J]\in\mathcal{J}/\operatorname{Sp}_{2n+2}(p\mathbb{Z}) [ J ] ∈ J / Sp 2 n + 2 ( p Z ) .
Fix U ∈ S U\in\mathcal{S} U ∈ S and consider J 1 , J 2 ⊂ U J_1,J_2\subset U J 1 , J 2 ⊂ U rank-1 isotropic summands of U U U that are not in the same orbit of Sp 2 n + 2 ( p Z ) \operatorname{Sp}_{2n+2}(p\mathbb{Z}) Sp 2 n + 2 ( p Z ) and let J J J be another rank-1 isotropic summand of U U U . For i = 1 , 2 i=1,2 i = 1 , 2 , let j ⃗ i ∈ J i ⊗ Q \vec{j}_i\in J_i\otimes\mathbb{Q} j i ∈ J i ⊗ Q be a basis of J i J_i J i . Let α ∈ H i ( Sp ( U ⊥ , p Z ) ; St ω ( U ⊥ ) ) \alpha\in H_i(\operatorname{Sp}(U^{\perp},p\mathbb{Z});\operatorname{St}^{\omega}(U^{\perp})) α ∈ H i ( Sp ( U ⊥ , p Z ) ; St ω ( U ⊥ )) . Theorem 4.10 implies
Φ J ( [ j ⃗ 1 , j ⃗ 2 ] ⊗ α ) = [ j ⃗ 1 ] ⊗ α if [ J ] = [ J 1 ] \Phi_J([\vec{j}_1,\vec{j}_2]\otimes\alpha)=[\vec{j}_1]\otimes\alpha\quad\text{if }[J]=[J_1] Φ J ([ j 1 , j 2 ] ⊗ α ) = [ j 1 ] ⊗ α if [ J ] = [ J 1 ]
Φ J ( [ j ⃗ 1 , j ⃗ 2 ] ⊗ α ) = − [ j ⃗ 2 ] ⊗ α if [ J ] = [ J 2 ] \Phi_J([\vec{j}_1,\vec{j}_2]\otimes\alpha)=-[\vec{j}_2]\otimes\alpha\quad\text{if }[J]=[J_2] Φ J ([ j 1 , j 2 ] ⊗ α ) = − [ j 2 ] ⊗ α if [ J ] = [ J 2 ]
Φ J ( [ j ⃗ 1 , j ⃗ 2 ] ⊗ α ) = 0 otherwise. \Phi_J([\vec{j}_1,\vec{j}_2]\otimes\alpha)=0\quad\text{otherwise.} Φ J ([ j 1 , j 2 ] ⊗ α ) = 0 otherwise.
Here we use the natural isomorphisms
J ⊥ / J ≅ U ⊥ J^{\perp}/J\cong U^{\perp} J ⊥ / J ≅ U ⊥
to identify the homology groups
H i ( Sp ( J ⊥ / J , p Z ) ; St ω ( J ⊥ / J ) ) ≅ H i ( Sp ( U ⊥ , p Z ) ; St ω ( U ⊥ ) ) H_i(\operatorname{Sp}(J^{\perp}/J,p\mathbb{Z});\operatorname{St}^{\omega}(J^{\perp}/J))\cong H_i(\operatorname{Sp}(U^{\perp},p\mathbb{Z});\operatorname{St}^{\omega}(U^{\perp})) H i ( Sp ( J ⊥ / J , p Z ) ; St ω ( J ⊥ / J )) ≅ H i ( Sp ( U ⊥ , p Z ) ; St ω ( U ⊥ ))
and will view α \alpha α as an element of either group.
Let v ⃗ 1 , … , v ⃗ 2 n + 2 \vec{v}_1,\ldots,\vec{v}_{2n+2} v 1 , … , v 2 n + 2 be the standard symplectic basis of F p 2 n + 2 \mathbb{F}_p^{2n+2} F p 2 n + 2 . Let J ′ \mathcal{J}' J ′ be the set of isotropic summands J J J of Z 2 n + 2 \mathbb{Z}^{2n+2} Z 2 n + 2 with
J / p = span ( v ⃗ 2 + ∑ k ≠ 2 x k v ⃗ k ) . J/p=\operatorname{span}\left(\vec{v}_2+\sum_{k\ne2}x_k\vec{v}_k\right). J / p = span v 2 + k = 2 ∑ x k v k .
By Theorem 7.3 , there are p 2 n + 1 ( p − 1 ) / 2 p^{2n+1}(p-1)/2 p 2 n + 1 ( p − 1 ) /2 many elements in the image of J ′ \mathcal{J}' J ′ in J / Sp 2 n + 2 ( p Z ) \mathcal{J}/\operatorname{Sp}_{2n+2}(p\mathbb{Z}) J / Sp 2 n + 2 ( p Z ) . The factor of p 2 n + 1 p^{2n+1} p 2 n + 1 comes from the choices of x i ∈ F p x_i\in\mathbb{F}_p x i ∈ F p and the factor of ( p − 1 ) / 2 (p-1)/2 ( p − 1 ) /2 comes from the choice of orientation.
Let Φ ′ \Phi' Φ ′ be postcomposition of Φ \Phi Φ with the projection onto the summands coming from elements of J / Sp 2 n + 2 ( p Z ) \mathcal{J}/\operatorname{Sp}_{2n+2}(p\mathbb{Z}) J / Sp 2 n + 2 ( p Z ) in the image of J ′ \mathcal{J}' J ′ . The codomain of Φ ′ \Phi' Φ ′ is
⨁ [ J ] ∈ J / Sp 2 n + 2 ( p Z ) with J ∈ J ′ H 0 ( GL ( J , p Z ) ; St ( J ) ) ⊗ H i ( Sp ( J ⊥ / J , p Z ) ; St ω ( J ⊥ / J ) ) . \bigoplus_{[J]\in\mathcal{J}/\operatorname{Sp}_{2n+2}(p\mathbb{Z})\text{ with }J\in\mathcal{J}'} H_0(\operatorname{GL}(J,p\mathbb{Z});\operatorname{St}(J))\otimes H_i(\operatorname{Sp}(J^\perp/J,p\mathbb{Z});\operatorname{St}^{\omega}(J^\perp/J)). [ J ] ∈ J / Sp 2 n + 2 ( p Z ) with J ∈ J ′ ⨁ H 0 ( GL ( J , p Z ) ; St ( J )) ⊗ H i ( Sp ( J ⊥ / J , p Z ) ; St ω ( J ⊥ / J )) .
and has dimension
p 2 n + 1 ( p − 1 2 ) dim k H 2 n − i ν − i ( Sp 2 n ( p Z ) ) . p^{2n+1}\left(\frac{p-1}{2}\right)\dim_k H^{\nu-i}_{2n-i}(\operatorname{Sp}_{2n}(p\mathbb{Z})). p 2 n + 1 ( 2 p − 1 ) dim k H 2 n − i ν − i ( Sp 2 n ( p Z )) .
Since Φ ′ \Phi' Φ ′ factors through H ν + 2 n + 2 − i ( Sp 2 n + 2 ( p Z ) ) H^{\nu+2n+2-i}(\operatorname{Sp}_{2n+2}(p\mathbb{Z})) H ν + 2 n + 2 − i ( Sp 2 n + 2 ( p Z )) , it suffices to show Φ ′ \Phi' Φ ′ is surjective to establish our lower bounds for the dimension of H ν + 2 n + 2 − i ( Sp 2 n + 2 ( p Z ) ) H^{\nu+2n+2-i}(\operatorname{Sp}_{2n+2}(p\mathbb{Z})) H ν + 2 n + 2 − i ( Sp 2 n + 2 ( p Z )) .
We now prove that Φ ′ \Phi' Φ ′ is surjective. Fix J ∈ J ′ J\in\mathcal{J}' J ∈ J ′ and α ∈ H i ( Sp ( J ⊥ / J , p Z ) ; St ω ( J ⊥ / J ) ) \alpha\in H_i(\operatorname{Sp}(J^\perp/J,p\mathbb{Z});\operatorname{St}^{\omega}(J^\perp/J)) α ∈ H i ( Sp ( J ⊥ / J , p Z ) ; St ω ( J ⊥ / J )) . Pick j ⃗ ∈ Q 2 n + 2 \vec{j}\in\mathbb{Q}^{2n+2} j ∈ Q 2 n + 2 a vector with span Q ( j ⃗ ) = J ⊗ Q \operatorname{span}_{\mathbb{Q}}(\vec{j})=J\otimes\mathbb{Q} span Q ( j ) = J ⊗ Q . View [ j ⃗ ] [\vec{j}] [ j ] as the generator of St ( J ) \operatorname{St}(J) St ( J ) . Elements of the form
[ j ⃗ ] ⊗ α ∈ H 0 ( GL ( J , p Z ) ; St ( J ) ) ⊗ H i ( Sp ( J ⊥ / J , p Z ) ; St ω ( J ⊥ / J ) ) [\vec{j}]\otimes\alpha\in H_0(\operatorname{GL}(J,p\mathbb{Z});\operatorname{St}(J))\otimes H_i(\operatorname{Sp}(J^\perp/J,p\mathbb{Z});\operatorname{St}^{\omega}(J^\perp/J)) [ j ] ⊗ α ∈ H 0 ( GL ( J , p Z ) ; St ( J )) ⊗ H i ( Sp ( J ⊥ / J , p Z ) ; St ω ( J ⊥ / J ))
generate the codomain of Φ ′ \Phi' Φ ′ .
Using Capovilla-Searle [6 Lemma 4.5] , we can find a symplectic summand U ⊆ Z 2 n + 2 U\subseteq\mathbb{Z}^{2n+2} U ⊆ Z 2 n + 2 containing J J J with
U / p = span F p ( v ‾ 1 ) ⊕ J / p . U/p=\operatorname{span}_{\mathbb{F}_p}(\overline{v}_1)\oplus J/p. U / p = span F p ( v 1 ) ⊕ J / p .
Let k ⃗ ∈ U ⊗ Q \vec{k}\in U\otimes\mathbb{Q} k ∈ U ⊗ Q be a vector with
( span Q ( k ⃗ ) ∩ Z 2 n + 2 ) / p = span ( v ‾ 1 ) . \left(\operatorname{span}_{\mathbb{Q}}(\vec{k})\cap\mathbb{Z}^{2n+2}\right)/p=\operatorname{span}(\overline{v}_1). ( span Q ( k ) ∩ Z 2 n + 2 ) / p = span ( v 1 ) .
Since U ⊥ ≅ J ⊥ / J U^\perp\cong J^\perp/J U ⊥ ≅ J ⊥ / J , we may view α \alpha α as an element of H i ( Sp ( U ⊥ , p Z ) ; St ω ( U ⊥ ) ) H_i(\operatorname{Sp}(U^\perp,p\mathbb{Z});\operatorname{St}^{\omega}(U^\perp)) H i ( Sp ( U ⊥ , p Z ) ; St ω ( U ⊥ )) . Since U ∈ S U\in\mathcal{S} U ∈ S and span Q ( k ⃗ ) ∩ Z 2 n + 2 ∉ J ′ \operatorname{span}_{\mathbb{Q}}(\vec{k})\cap\mathbb{Z}^{2n+2}\notin\mathcal{J}' span Q ( k ) ∩ Z 2 n + 2 ∈ / J ′ , Equation 1 implies
Φ ′ ( [ j ⃗ , k ⃗ ] ⊗ α ) = [ j ⃗ ] ⊗ α . \Phi'([\vec{j},\vec{k}]\otimes\alpha)=[\vec{j}]\otimes\alpha. Φ ′ ([ j , k ] ⊗ α ) = [ j ] ⊗ α .
Thus Φ ′ \Phi' Φ ′ is surjective.
References
Email address: jeremykmiller@purdue.edu
Purdue University, West Lafayette IN 47907, USA
2074 East Hall, 530 Church Street, Ann Arbor, MI 48109, USA
Email address: urshita@umich.edu
University of Oklahoma, 601 Elm Avenue, Norman, OK 73019, USA
Email address: ppatzt@ou.edu