Definition Balanced perfect duality on global sections
✓ definition formalized

Balanced perfect duality on global sections

Let \(B\) be a free sheaf on a Bruhat interval, with its polynomial labels and finitely many edges. Write \(\Gamma=\Gamma(B)\), let \(p\) be the product of all edge labels, and define \[ C_xm=p\,\delta_x(m_x). \] The upper-quotient condition ensures these assignments are global sections.

A section duality is an \(A\)-linear isomorphism \[ D:\Gamma\xrightarrow{\sim}\operatorname{Hom}_A(\Gamma,A) \] satisfying cut balance: \[ \langle C_xm,n\rangle=\langle m,C_xn\rangle \quad\text{for every }x,m,n, \qquad \langle m,n\rangle=D(m)(n). \] The isomorphism supplies a unique section representing each functional. Balance makes each cut self-adjoint: applying the cut in either argument has the same effect.

The two recorded fields

The pairing field records \(D\), including its inverse and \(A\)-linearity. The balance field records the displayed equation for the cuts made with the actual edge-label product.

The boundary-sheaf version specializes this same construction to a boundary sheaf. In the free-sheaf version the objects may also be word sheaves, because the cut construction uses their upper quotients and the later stalk-duality argument uses their free stalks and global generation.

A perfect graded section pairing supplies \(D\) and, by Support orthogonality makes integral cuts balanced, satisfies cut balance. Its separate homogeneity equation then controls the degrees of the resulting stalk-duality maps. The section-duality structure itself records the pairing and balance.

Recovering a functional from a supported section

For \(m\) supported at \(x\), the cut identities give \(C_xm=pm\). If \(n_x=0\), then \(C_xn=0\). Balance yields \[ p\langle m,n\rangle =\langle C_xm,n\rangle =\langle m,C_xn\rangle=0. \] The label product is nonzero, and \(A\) is a domain, so \(\langle m,n\rangle=0\). Consequently pairing with \(m\) depends only on \(n_x\). Global generation supplies a section with any prescribed \(x\)-value, so this defines a functional on \(B^x\).

Conversely, a stalk functional \(\lambda\) gives the global functional \(\lambda\circ\pi_x\). The inverse of \(D\) represents it by a section. Applying the balance equation at every other vertex proves that representative is supported at \(x\), as shown in Vertex support equals the annihilator of an evaluation kernel and Supported sections are dual to a generated stalk.

Relation to the OpenAI notes

Different construction of the stalk–costalk relation.

OpenAI notes: The normalization proof describes a lower-support layer, relates it to the outgoing kernel using flabbiness, and obtains its graded rank through the cited moment-graph and Soergel-module results. It temporarily doubles degrees and uses the shift convention M(k)^j = M^{j+k}. bg:normalization, eq:bg-upward-rank.

Lean (primary): The local construction starts with a free sheaf and a specified balanced perfect section pairing. Integral cuts identify vertex support with an evaluation-kernel annihilator. The incoming label product identifies that support module with the outgoing kernel, giving an explicit linear equivalence with the stalk dual. Support:521, CharacterReciprocity:150, CharacterReciprocity:185.

Role in the proof: The Blueprint follows these integral maps. The pairing construction and its homogeneity equation supply the later graded reciprocity statement. Degree shifts use the Lean convention M(d)n = M{n-d}.

References

The source declarations below give the definitions and constructions used here.

Lean formalization

In both declarations, pairing is the linear equivalence with the dual and balanced is the cut-balance equation.

OAI.KLInvariance.BruhatGraph.FreeSheaf.SectionDuality: CharacterReciprocity.lean:170.

OAI.KLInvariance.BruhatGraph.SectionDuality: Support.lean:1432.

Used by