Polynomial R-convolution for word-stalk and kernel characters
Fix a finite-generator Coxeter system, a finite basis of its extended realization, and a reduced word \(\mathbf t\) with product \(b\). Put \(I_b=[1,b]\). For \(x\in I_b\), let \(d=\ell(b)-\ell(x)\). Let \(C_x^{\mathbf t}\) be the character of the polynomial word-sheaf stalk at \(x\), and \(K_x^{\mathbf t}\) the character of its full outgoing kernel, the intersection of all outgoing restriction kernels. Then \[ \operatorname{reflect}_d C_x^{\mathbf t}(q) =\sum_{z\in I_b}R_{x,z}(q)C_z^{\mathbf t}(q) =K_x^{\mathbf t}(q). \] These are identities in \(\mathbb Z[q]\). The degree bound \(\operatorname{natDegree}C_x^{\mathbf t}\le d\) makes the left polynomial equal to \(q^d C_x^{\mathbf t}(q^{-1})\) in the Laurent polynomial ring.
Relation to the OpenAI notes
Different intermediate objects.
OpenAI notes: The normalization discussion identifies indecomposable boundary-sheaf characters with KL polynomials by the cited character theorem. bg:normalization.
Lean (primary): The word-stalk character is proved to equal a positive Hecke-word coefficient. Its R-convolution is then proved before model decomposition and cancellation. The separate boundary-character identification is exposed as StalkCharacterObligation. IntegralPairings:289, CharacterReciprocity:1134, SheafMorphisms:219, BottSamelson:1924, BottSamelson:1945, IntegralPairings:104, IntegralPairings:120, CharacterReciprocity:915.
Role in the proof: The word formula and R-convolution describe the tensor word. A common decomposition into boundary models then supplies the character identities used in the cancellation argument.
Proof
Step 1. Work at the formal Laurent unit
Let \(\Lambda=\mathbb Z[q,q^{-1}]\), and let \(u\in\Lambda^\times\) be its distinguished variable unit. The embedding \[ j:\mathbb Z[q]\hookrightarrow\Lambda \] sends a polynomial \(F\) to \(F(u)\). Indeed, both descriptions send a monomial \(a q^n\) to \(a u^n\), and preserve finite sums.
The positive Hecke-word reciprocity theorem applies to this unit. Writing \(m=|\mathbf t|\), it gives \[ \sum_{z\le b}R_{x,z}(u)\,h_{\mathbf t}(z;u) =u^{-\ell(x)}u^m h_{\mathbf t}(x;u^{-1}). \] Reducedness is the support hypothesis that makes this a sum over the finite lower interval of \(b\).
Step 2. Substitute the word-character coefficients
The word-character theorem identifies the positive-word coefficient polynomials with \(C_z^{\mathbf t}\). Evaluation commutes with the positive-word recursion because it preserves addition, multiplication and integer coefficients. Therefore \[ h_{\mathbf t}(z;u)=C_z^{\mathbf t}(u),\qquad h_{\mathbf t}(x;u^{-1})=C_x^{\mathbf t}(u^{-1}). \] The ordinary R-polynomial evaluation theorem identifies the R-values in the Hecke identity with the displayed \(R_{x,z}(u)\).
Reducedness gives \(m=\ell(b)\). Since \(x\le b\), \(\ell(x)\le m\), and the unit powers simplify to \[ u^{-\ell(x)}u^m=u^{m-\ell(x)}=u^d. \]
Step 3. Recognize the reflected polynomial
The outgoing-kernel character theorem supplies \(\operatorname{natDegree}C_x^{\mathbf t}\le d\). Hence bounded polynomial reflection gives \[ u^d C_x^{\mathbf t}(u^{-1}) =j(\operatorname{reflect}_d C_x^{\mathbf t}). \] The finite sum on the other side is \[ j\left(\sum_{z\le b}R_{x,z}C_z^{\mathbf t}\right), \] because \(j\) is a ring homomorphism.
Step 4. Return to integer polynomials and interval vertices
Injectivity of \(j\) proves \[ \operatorname{reflect}_d C_x^{\mathbf t} =\sum_{z\le b}R_{x,z}C_z^{\mathbf t} \] in \(\mathbb Z[q]\).
Every group element below \(b\) lies above \(1\). Thus \(z\mapsto\langle z,1\le z,z\le b\rangle\) identifies this finite index set with \(I_b\), giving the stated sum over interval vertices.
Step 5. Insert the actual kernel character
The kernel-character theorem gives \(K_x^{\mathbf t}=\operatorname{reflect}_d C_x^{\mathbf t}\). Substituting into Step 4 proves the full R-convolution identity.
For one letter \(s\), the two stalk characters are \(1\). At the lower vertex, the sum is \(R_{1,1}+R_{1,s}=1+(q-1)=q\), which is the character of the kernel \(\alpha_s A\). At the top vertex, the sum and the kernel character are both \(1\).
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.TitsSpace.bottStalkCharacter_natDegree_le: CharacterReciprocity.lean:1088.
OAI.KLInvariance.TitsSpace.bottStalkCharacter_reciprocity: CharacterReciprocity.lean:1105.
OAI.KLInvariance.TitsSpace.bottKernelCharacter_rConvolution: CharacterReciprocity.lean:1134.
OAI.KLInvariance.HeckeQ.positiveWord_rValue: CharacterReciprocity.lean:997.
OAI.KLInvariance.HeckeQ.eval_positiveWord: CharacterReciprocity.lean:1027.