The local intersection map and its matrix
Fix a finite-generator Coxeter system and a finite real basis of its extended realization. Let \(A=\mathbb R[X_1,\ldots,X_N]\) be the resulting polynomial coefficient algebra, with variables in degree one, and let \(B=B(b)\) be the chosen normalized boundary sheaf on \([1,b]\).
At \(x\in[1,b]\), write \[ M_x=B^x,\qquad M_x^*=\operatorname{Hom}_A(M_x,A), \] \[ K_x=\{v\in M_x:\rho^-_e(v)=0 \text{ for every outgoing edge }e:x\to y\}. \] Thus \(K_x\) is the kernel of the combined outgoing restriction map \(\partial_x\).
The constructed symmetric section pairing has shift \(\ell(b)\), and its section duality supplies \[ \kappa_x:K_x\xrightarrow{\sim}M_x^*. \] The local intersection map is \[ I_x:M_x^*\longrightarrow M_x,\qquad I_x=\iota_x\circ\kappa_x^{-1}, \] where \(\iota_x:K_x\hookrightarrow M_x\) is the inclusion. A functional first determines its unique outgoing-kernel vector through the duality, and the inclusion regards that vector as a stalk element.
The coefficient-valued local form
The same map defines an \(A\)-valued bilinear form on the stalk dual: \[ b_x(\lambda,\mu)=\lambda(I_x\mu). \] Linearity of \(I_x\) and of the functionals proves linearity in each argument. The stalk–costalk symmetry theorem gives \[ b_x(\lambda,\mu)=b_x(\mu,\lambda). \] The local intersection properties theorem proves that \(I_x\) is injective and has image \(K_x\).
Coordinates in a homogeneous stalk basis
Choose a finite homogeneous \(A\)-basis \((m_i)\) of the stalk, with degrees \(d_i\), and let \(\lambda_i(m_j)=\delta_{ij}\) be its dual basis. Set \[ a_{ij}=\lambda_i(I_x\lambda_j). \] Then \[ I_x\lambda_j=\sum_i a_{ij}m_i. \] Thus the \(j\)-th matrix column consists of the coordinates of the outgoing-kernel vector representing \(\lambda_j\). The symmetric form has the same matrix: \(b_x(\lambda_i,\lambda_j)=a_{ij}\).
For \(d=\ell(b)-\ell(x)\), the local intersection properties theorem gives \[ a_{ij}\in A_{d-d_i-d_j},\qquad a_{ij}=a_{ji}. \] At a proper vertex \(x<b\), every \(a_{ij}\) also belongs to \(A_+\). These are the inputs to the strict diagonal-degree theorem.
Relation to the OpenAI notes
Different intermediate degree statement.
OpenAI notes: The normalization proposition identifies stalk graded ranks with canonical KL polynomials using the cited character theorem and a conversion from degree-two to degree-one labels. bg:normalization, eq:bg-ranks.
Lean (primary): The actual local intersection map is constructed from integral stalk–costalk duality. Boundary minimality puts its image in the positive ideal, and the strict degree theorem assumes that the specified diagonal coefficient is nonzero. PositiveTransfer:132, PositiveTransfer:160, PositiveTransfer:186.
Role in the proof: The local calculation gives \(2d_i<\ell(b)-\ell(x)\) for the basis generator named in that nonzero-diagonal hypothesis. Actual-character reciprocity supplies the endpoint identity in the comparison route.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.BruhatGraph.ActualBoundary: PositiveTransfer.lean:94.
OAI.KLInvariance.BruhatGraph.actualSymmetricPairing: PositiveTransfer.lean:96.
OAI.KLInvariance.BruhatGraph.actualSectionDuality: PositiveTransfer.lean:100.
OAI.KLInvariance.BruhatGraph.localIntersection: PositiveTransfer.lean:132.