Boundary values supplied by upper sections
For a vertex \(x\), let \(E_x^+=\{e:s(e)=x\}\) be its outgoing edges. Define the boundary module \[ \partial_xB=\prod_{e\in E_x^+}B^e. \] For a finite graph this is a finite product, equivalently a finite direct sum.
The outgoing stalk map is \[ r_x:B^x\to\partial_xB,\qquad r_x(v)=(\rho^-_e(v))_{e\in E_x^+}. \] Its kernel is the full outgoing kernel.
The upper-section map
Let \(U_x=\{y:x<y\}\). Every outgoing edge has target in \(U_x\). A section compatible on \(U_x\) supplies the tuple \[ b_x(m)=(\rho^+_e(m_{t(e)}))_{e\in E_x^+}. \] This defines a linear map \[ b_x:\Gamma(U_x,B)\to\partial_xB. \] Its image is the boundary image \[ B^{\delta x}=\operatorname{im}b_x. \]
In the ambient section convention, \(m\) also has coordinates outside \(U_x\). The displayed formula reads only its upper-endpoint coordinates.
A labelled diamond
Take vertices \(x<y,z<t\), edges \(x\to y,x\to z,y\to t,z\to t\), and modules \(A\) at vertices and \(A/(\alpha)\) at edges, with quotient restrictions.
Above \(x\), compatibility gives \(m_y\equiv m_t\) and \(m_z\equiv m_t\) modulo \(\alpha\). Therefore \[ B^{\delta x}=\{(\overline a,\overline a):\overline a\in A/(\alpha)\} \subseteq (A/(\alpha))^2. \] Every diagonal tuple occurs by choosing equal upper values lifting \(\overline a\). Thus the upper edge equations determine the boundary image explicitly.
A free cover \(A\to B^{\delta x}\), \(a\mapsto(\overline a,\overline a)\), supplies the new stalk at \(x\). Its two coordinate maps are the outgoing restrictions.
Graded boundary data
For finite vertex and edge sets, all restrictions have degree zero, so \(b_x\) is a homogeneous map between the section and boundary modules. Over the Noetherian polynomial coefficient ring, its image is finite graded module data.
The boundary-sheaf construction takes a minimal graded free cover of this image. Global generation and flabbiness later identify the image of \(r_x\) with the same boundary image.
References
The local definitions and constructions are listed below.
Lean formalization
OAI.KLInvariance.MomentGraph.Sheaf.Boundary: Sheaves.lean:180.
OAI.KLInvariance.MomentGraph.Sheaf.stalkMap: Sheaves.lean:183.
OAI.KLInvariance.MomentGraph.Sheaf.boundaryMap: Sheaves.lean:189.
OAI.KLInvariance.MomentGraph.GradedSheaf.boundaryHom: GradedSheaves.lean:74.
OAI.KLInvariance.MomentGraph.GradedSheaf.boundaryImage: GradedSheaves.lean:81.