OctoInformBeta

OpenAI preprint, September 2026 · §2. Real realizations and moment-graph sheaves · §2.4

The Bruhat moment graph, sheaves and upward kernels: An edge for every upward Bruhat step, covers or not, labelled by a reflection and a root; sections, flabbiness and upward kernels Lx(F)L_x(F)

The Bruhat moment graph of an interval and its sheaves.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Moment-graph sheaves in the sense of Braden-MacPherson and Fiebig live on a graph whose vertices are partially ordered and whose edges carry linear forms of a polynomial ring; in Kazhdan-Lusztig theory the graph is the Bruhat graph of an interval and the forms are roots. The library defines the Bruhat moment graph of [u,b][u,b], with its edges labelled through the Tits realization of WW; sheaves on ordered graphs, with their sections, flabbiness and generation by global sections; and the upward kernel of a stalk of a graded sheaf along a set of upward edges.

Let (W,S)(W,S) be a Coxeter system with Coxeter matrix M=(mij)i,j∈IM = (m_{ij})_{i,j\in I} on a finite index set II, with S={si:i∈I}S = \{s_i : i\in I\}. An upward edge x→yx\to y of the Bruhat graph of WW is a pair of elements xx and yy of WW with y=txy = tx for a reflection tt of WW and ℓ(x)<ℓ(y)\ell(x)<\ell(y), where ℓ\ell is the length function of (W,S)(W,S).✦ The Bruhat order ≤\le on WW is the reflexive transitive closure of the relation given by these edges✦, and for uu and bb in WW the Bruhat interval [u,b][u,b] is the set of x∈Wx\in W with u≤x≤bu\le x\le b.✦

Roots and the Extended Realization

Let UU be a real vector space with a bilinear form BB and vectors αi\alpha_i, i∈Ii\in I, such that B(αi,αj)=−2cos⁡(π/mij)B(\alpha_i,\alpha_j) = -2\cos(\pi/m_{ij}), with π/mij\pi/m_{ij} read as 00 when mij=∞m_{ij}=\infty. The geometric simple reflection σi\sigma_i is the linear automorphism x↦x−B(αi,x)αix\mapsto x - B(\alpha_i,x)\,\alpha_i of UU✦, and the geometric action of WW on UU is the homomorphism from WW to the linear automorphisms of UU that sends sis_i to σi\sigma_i.✦ The extended realization space of (U,B)(U,B) is U×(ker⁡B)*U\times(\ker B)^*, where ker⁡B\ker B is the space of vectors of UU that pair to zero under BB with every vector of UU.✦

The Tits realization takes U=ℝIU = \mathbb{R}^I, the simple root αi\alpha_i the coordinate vector of ii✦, and the Tits form B(x,y)=∑i∈Ixini(y),ni(y)=∑j∈Igijyj,gij=−2cos⁡(π/mij),B(x,y) = \sum_{i\in I} x_i\,n_i(y),\qquad n_i(y) = \sum_{j\in I} g_{ij}\,y_j,\qquad g_{ij} = -2\cos(\pi/m_{ij}),(1)✦ built from the Coxeter Gram entries gijg_{ij} and the normal functionals nin_i. The geometric action of WW on ℝI\mathbb{R}^I for the Tits realization is the Tits action of WW, written a↦w(a)a\mapsto w(a) for w∈Ww\in W.✦ A real root is a vector a=w(αi)a = w(\alpha_i) of ℝI\mathbb{R}^I with w∈Ww\in W and i∈Ii\in I✦; it represents the reflection tt when a=w(αi)a = w(\alpha_i) and t=wsiw−1t = ws_iw^{-1} for one and the same pair (w,i)(w,i).✦ A positive root is a real root whose coordinates are all nonnegative.✦ The positive root αt\alpha_t of a reflection tt is the positive root that represents tt✦, and the reflection of a positive root is the reflection it represents.✦

The extended Tits space V=ℝI×(ker⁡B)*V = \mathbb{R}^I\times(\ker B)^* is the extended realization space of the Tits form✦, and ℝI\mathbb{R}^I embeds in it by a↦(a,0)a\mapsto(a,0).✦ Fix a basis (ej)j∈J(e_j)_{j\in J} of VV. The coefficient ring A=ℝ[Xj:j∈J]A = \mathbb{R}[X_j : j\in J] is the real polynomial ring on this basis, identified with the symmetric algebra of VV by ej↦Xje_j\mapsto X_j, so that the vector ∑jcjej\sum_j c_je_j of VV is the linear form ∑jcjXj\sum_j c_jX_j of AA.✦

The Definition

Definition 1. An ordered graph 𝒢\mathcal{G} on a partially ordered set 𝒱\mathcal{V} is a set ℰ\mathcal{E} of edges with two maps from ℰ\mathcal{E} to 𝒱\mathcal{V}, the source and the target, such that, writing E:x→yE\colon x\to y for an edge with source xx and target yy, x<yfor every edge E:x→y.x<y\qquad\text{for every edge } E\colon x\to y.(2)✦ The edge set of the interval [u,b][u,b] is ℰ(u,b)={(x,y)∈[u,b]×[u,b]:x→y is an upward edge of the Bruhat graph of W}.\mathcal{E}(u,b) = \{(x,y)\in[u,b]\times[u,b] : x\to y\text{ is an upward edge of the Bruhat graph of } W\}.(3)✦ The Bruhat moment graph of [u,b][u,b] is the ordered graph on [u,b][u,b], ordered by the Bruhat order, with edge set ℰ(u,b)\mathcal{E}(u,b): 𝒢(u,b)=([u,b],ℰ(u,b)),(x,y):x→y.\mathcal{G}(u,b) = \bigl([u,b],\,\mathcal{E}(u,b)\bigr),\qquad (x,y)\colon x\to y.(4)✦ An edge E:x→yE\colon x\to y of 𝒢(u,b)\mathcal{G}(u,b) carries the reflection label tE=yx−1,t_E = yx^{-1},(5)✦ its root is the positive root αtE\alpha_{t_E} of tEt_E✦, and its polynomial label is the linear form of AA given by the embedded root (αtE,0)(\alpha_{t_E},0) of VV: αE=∑j∈JcjXj,where (αtE,0)=∑j∈Jcjej.\alpha_E = \sum_{j\in J}c_jX_j,\qquad\text{where } (\alpha_{t_E},0) = \sum_{j\in J}c_je_j.(6)✦

A sheaf ℳ\mathscr{M} on an ordered graph 𝒢\mathcal{G} over a ring RR consists of an RR-module ℳx\mathscr{M}^x for every vertex xx, the stalk at xx, an RR-module ℳE\mathscr{M}^E for every edge EE, the edge module, and for every edge E:x→yE\colon x\to y two RR-linear restriction maps ρx,E:ℳx→ℳE,ρy,E:ℳy→ℳE.\rho_{x,E}\colon\mathscr{M}^x\to\mathscr{M}^E,\qquad \rho_{y,E}\colon\mathscr{M}^y\to\mathscr{M}^E.(7)✦ The sections of ℳ\mathscr{M} over a set Ω\Omega of vertices are the families m=(mx)x∈𝒱m=(m_x)_{x\in\mathcal{V}}, with mx∈ℳxm_x\in\mathscr{M}^x at every vertex, that are compatible on Ω\Omega: Γ(Ω,ℳ)={m∈∏x∈𝒱ℳx:ρx,E(mx)=ρy,E(my) for every edge E:x→y with x,y∈Ω}.\Gamma(\Omega,\mathscr{M}) = \Bigl\{m\in\prod_{x\in\mathcal{V}}\mathscr{M}^x : \rho_{x,E}(m_x) = \rho_{y,E}(m_y)\text{ for every edge } E\colon x\to y\text{ with } x,y\in\Omega\Bigr\}.(8)✦ The global sections are the sections over 𝒱\mathcal{V}, forming Γ(ℳ)=Γ(𝒱,ℳ)\Gamma(\mathscr{M}) = \Gamma(\mathcal{V},\mathscr{M}).✦ The sheaf ℳ\mathscr{M} is flabby if for every upper set Ω⊆𝒱\Omega\subseteq\mathcal{V} and every section mm over Ω\Omega there is a global section gg with gx=mxfor every x∈Ω.g_x = m_x\qquad\text{for every } x\in\Omega.(9)✦ The sheaf ℳ\mathscr{M} is generated by global sections if for every vertex xx Γ(ℳ)→ℳx,g↦gx,is surjective.\Gamma(\mathscr{M})\to\mathscr{M}^x,\qquad g\mapsto g_x,\qquad\text{is surjective}.(10)✦

The polynomial grading of the ring k[Xj:j∈J]k[X_j : j\in J] over a field kk in variables XjX_j indexed by a set JJ puts every variable in degree one: for n∈ℤn\in\mathbb{Z}, with exponents νj\nu_j nonnegative integers, all but finitely many zero, k[Xj:j∈J]n=span⁡k{∏jXjνj:∑jνj=n},k[X_j : j\in J]_n = \operatorname{span}_k\Bigl\{\prod_{j}X_j^{\nu_j} : \sum_j\nu_j = n\Bigr\},(11)✦ and AA is graded this way, with k=ℝk=\mathbb{R}.

Let kk be a field and Λ=⨁n∈ℤΛn\Lambda = \bigoplus_{n\in\mathbb{Z}}\Lambda_n a kk-algebra graded by kk-subspaces, as AA is by (11). A graded Λ\Lambda-module is a finitely generated Λ\Lambda-module NN with a decomposition N=⨁n∈ℤNnN=\bigoplus_{n\in\mathbb{Z}}N_n into kk-subspaces such that ΛmNn⊆Nm+n\Lambda_mN_n\subseteq N_{m+n} for all mm, nn and Nn=0N_n = 0 for n<0n<0✦; a morphism of graded Λ\Lambda-modules N→N′N\to N' is a Λ\Lambda-linear map ff with f(Nn)⊆Nn′f(N_n)\subseteq N'_n for every nn.✦ A graded sheaf on 𝒢\mathcal{G} is a sheaf over Λ\Lambda whose stalks and edge modules are graded Λ\Lambda-modules and whose restriction maps are morphisms of graded Λ\Lambda-modules✦; forgetting the gradings leaves its underlying sheaf of Λ\Lambda-modules.✦

At a vertex xx, the upward edges are the edges with source xx✦, and the stalk map sends m∈ℳxm\in\mathscr{M}^x to (ρx,E(m))E(\rho_{x,E}(m))_E in the product of the edge modules ℳE\mathscr{M}^E over the upward edges EE at xx.✦ For a set FF of upward edges at xx, the upward kernel of a sheaf over RR is the RR-submodule of the stalk on which the components of the stalk map indexed by FF vanish✦, and for a graded sheaf ℳ\mathscr{M} it is the upward kernel of the underlying sheaf, the Λ\Lambda-submodule Lx(F)={m∈ℳx:ρx,E(m)=0 for every E∈F}.L_x(F) = \{m\in\mathscr{M}^x : \rho_{x,E}(m) = 0\text{ for every } E\in F\}.(12)✦

What Is Checked

The Bruhat moment graph, its edges and their reflection labels are defined for a Coxeter system (W,S)(W,S) with Coxeter matrix MM on a finite index set II and elements uu, bb of WW, the polynomial labels also for a basis of the extended Tits space VV; ordered graphs for a partially ordered set; sheaves, their sections, flabbiness and generation by global sections for a ring RR; the polynomial grading for a field kk and a set of variables; and graded modules, graded sheaves and upward kernels for a field kk and a kk-algebra graded by kk-subspaces.

The library proves that the Tits form takes the values B(αi,αj)=−2cos⁡(π/mij)B(\alpha_i,\alpha_j) = -2\cos(\pi/m_{ij}) on the simple roots✦, and that for any form with these values B(αi,αi)=2B(\alpha_i,\alpha_i)=2✦ and the geometric simple reflections satisfy the Coxeter relations (σiσj)mij=1(\sigma_i\sigma_j)^{m_{ij}} = 1 for finite mijm_{ij}.✦ It proves that the positive root of a reflection and the reflection of a positive root are inverse bijections between the reflections of WW and the positive roots✦, that the label tEt_E of every edge of 𝒢(u,b)\mathcal{G}(u,b) is a reflection✦, and that x<yx<y in the Bruhat order for every edge x→yx\to y of 𝒢(u,b)\mathcal{G}(u,b). It also proves that the sections over a set of vertices form a submodule of ∏xℳx\prod_x\mathscr{M}^x and the upward kernel a submodule of the stalk.