Theorem Lower-character normalization and convolution on an ambient interval
✓ theorem proved

Lower-character normalization and convolution on an ambient interval

Fix a finite-generator Coxeter system and its polynomial coefficient algebra from one fixed real basis. Let \(I_b=[1,b]\), and let \(C_{x,c}\) denote the character of the actual boundary-sheaf stalk at \(x\) on \([1,c]\). For \(c\in I_b\), define the lower-character function \[ L_x^c(q)= \begin{cases} C_{x,c}(q),&x\le c,\\ 0,&x\not\le c. \end{cases} \] Then \(L_c^c=1\), and \(L_b^c=1\) for \(c=b\) and \(0\) otherwise.

For every \(x\in I_b\), \[ \sum_{z\in I_b}R_{x,z}L_z^c =\sum_{z\in[1,c]}R_{x,z}C_{z,c}. \] If \(x\not\le c\), this sum is zero.

Consequently, reciprocity for the boundary sheaf on \([1,c]\) implies \[ \operatorname{reflect}_{\delta(x,c)}L_x^c =\sum_{z\in I_b}R_{x,z}L_z^c \qquad(x\in I_b), \] where \(\delta(x,c)=\max\{\ell(c)-\ell(x),0\}\) is the natural-number rank difference.

Proof

Step 1. Normalize the top character

The normalized boundary top is the unit coefficient module, with one homogeneous basis vector in degree zero. Its character is \(q^0=1\). In the source, unitPolynomialBasis records that singleton basis, and char_unit computes its character.

Transport through the boundary's graded top identification gives \(C_{c,c}=1\), hence \(L_c^c=1\).

If \(c=b\), this gives \(L_b^c=1\). If \(c\ne b\), the two inequalities \(c\le b\) and \(b\le c\) would force equality. Thus \(b\not\le c\), and the support definition gives \(L_b^c=0\).

Step 2. Remove terms outside the smaller lower interval

Whenever \(z\not\le c\), the factor \(L_z^c\) is zero. Therefore \[ \sum_{z\in I_b}R_{x,z}L_z^c =\sum_{z\in I_b,\ z\le c}R_{x,z}L_z^c. \]

The source's lowerIntervalEquiv identifies \(\{z\in I_b:z\le c\}\) with \([1,c]\) by keeping the underlying group element. Its inverse uses transitivity \(z\le c\le b\) to regard that same element as a vertex of \(I_b\).

On this subset, \(L_z^c=C_{z,c}\). Reindexing by the bijection gives the displayed convolution identity. All characters use the same ambient coefficient algebra and basis.

Step 3. Prove vanishing outside the lower center’s support

Assume \(x\not\le c\). For any \(z\in I_b\), if \(z\not\le c\), then \(L_z^c=0\).

If \(z\le c\), the inequality \(x\le z\) would imply \(x\le c\), a contradiction. The triangular support of the R-family therefore gives \(R_{x,z}=0\).

Thus every summand is zero, proving the vanishing claim.

Step 4. Lift a smaller reciprocity identity

Assume reciprocity holds at every vertex of \([1,c]\). If \(x\le c\), apply it to that vertex: \[ \operatorname{reflect}_{\ell(c)-\ell(x)}C_{x,c} =\sum_{z\in[1,c]}R_{x,z}C_{z,c}. \] Use Step 2 and the definition of \(L_x^c\) to obtain the ambient formula.

If \(x\not\le c\), the lower character is zero, and polynomial reflection sends zero to zero. The right side is zero by Step 3. Hence the same formula holds for all \(x\in I_b\), including the incomparable case.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.Graded.ModuleData.unitPolynomialBasis: CharacterReciprocity.lean:2055.

OAI.KLInvariance.Graded.ModuleData.char_unit: CharacterReciprocity.lean:2068.

OAI.KLInvariance.BruhatGraph.stalkCharacter_top: CharacterReciprocity.lean:2091.

OAI.KLInvariance.BruhatGraph.lowerCharacter_diagonal: CharacterReciprocity.lean:2099.

OAI.KLInvariance.BruhatGraph.lowerCharacter_at_top: CharacterReciprocity.lean:2103.

OAI.KLInvariance.BruhatGraph.lowerIntervalEquiv: CharacterReciprocity.lean:2134.

OAI.KLInvariance.BruhatGraph.lowerCharacter_convolution: CharacterReciprocity.lean:2140.

OAI.KLInvariance.BruhatGraph.lowerCharacter_convolution_zero: CharacterReciprocity.lean:2166.

OAI.KLInvariance.BruhatGraph.lowerCharacter_reciprocity_lift: CharacterReciprocity.lean:2174.

Used by