Kernels of upward edge conditions
Let \(B\) be a sheaf on an ordered graph, let \(x\) be a vertex, and let \(U_x=\{e:s(e)=x\}\) be its set of outgoing edges. For \(F\subseteq U_x\), define \[ K_x(F)=\{v\in B^x:\rho^-_e(v)=0\text{ for every }e\in F\}. \] This is the upward kernel for the selected edge conditions. Equivalently, \[ K_x(F)=\ker\!\left(B^x\longrightarrow\prod_{e\in F}B^e\right) =\bigcap_{e\in F}\ker\rho^-_e. \] The map into the product sends \(v\) to all its selected lower restrictions. The intersection is a submodule because every restriction is linear.
Ordering the conditions
With no equations, \(K_x(\varnothing)=B^x\). If \(F\subseteq G\), every vector satisfying all equations in \(G\) satisfies those in \(F\), so \[ K_x(G)\subseteq K_x(F). \] For any two sets, \[ K_x(F\cup G)=K_x(F)\cap K_x(G), \] because satisfying the equations in the union means satisfying both collections.
The full outgoing kernel is \(K_x=K_x(U_x)=\ker r_x\), where \(r_x\) is the stalk map into the boundary module. It is called the costalk in this blueprint. At a top vertex, \(U_x\) is empty and \(K_x=B^x\).
One edge and one additional condition
For the congruence sheaf over \(A=\mathbb R[\alpha]\) on \(x\to y\), \[ K_x(\varnothing)=A,\qquad K_x(\{e\})=\alpha A,\qquad K_y=A. \] The lower quotient map kills exactly the multiples of \(\alpha\).
More generally, restricting \(\rho^-_e\) to \(K_x(F)\) gives a map with kernel \(K_x(F\cup\{e\})\). Its image is the lower edge image, so \[ 0\longrightarrow K_x(F\cup\{e\}) \longrightarrow K_x(F) \longrightarrow \rho^-_e(K_x(F)) \longrightarrow0 \] is exact. The final arrow is onto its stated image by definition; its kernel is precisely the added vanishing equation.
Gradings and schedules
For a graded sheaf, each restriction preserves degree. If \(v=\sum_n v_n\) satisfies an outgoing equation, then the degree-\(n\) component of that equation is \(\rho^-_e(v_n)=0\). Thus every \(K_x(F)\) is a homogeneous submodule.
A schedule orders the outgoing edges. A suffix kernel takes \(F\) to be the edges remaining after an initial segment has been removed. Successive removals enlarge these submodules; the exact sequence above measures each enlargement by an edge image. The simultaneous induction proves freeness for the suffix kernels used in the comparison.
References
The source declarations below give the definitions and constructions used here.
Lean formalization
OAI.KLInvariance.MomentGraph.Sheaf.upwardKernel: GradedSheaves.lean:599.
Used by
- Freeness of all terminal kernels in a transported schedule
- Kernel Hilbert values and normalized states
- Lower and upper images in the same edge module
- The exact sequence for adding one outgoing condition
- The local intersection map and its matrix
- Local boundary models over a local coefficient ring
- Freeness of a saturated image over the edge quotient ring
- Scaled equality of embedded edge images
- Remaining conditions and the Hilbert-state vector
- The full outgoing-kernel freeness condition
- Stalk and outgoing-kernel characters of graded sheaves
- Scaled image equality in products and model sums
- The outgoing label product of a selected edge set
- Outgoing label products send stalks and edge modules into selected images
- Supported selected edges and their label product
- Selected kernels of a structure model are product submodules
- Flat base change of selected kernels and edge images
- The upper edge image is free over its quotient coefficient ring
- Boundary-sheaf pullback with complete outgoing edge lifts
- Minimal-boundary sheaves and their normalized stalk
- Structural properties and kernels of finite sheaf sums
- Outgoing kernels under sheaf isomorphisms, sums, and shifts
- Sheaf properties preserved by lower-ideal extension
- Selected kernels under outgoing-complete graph pullback
- Section and outgoing-kernel constraint maps
- Sections supported at a vertex and their stalk values
- Stalk characters and selected kernels under upper-interval restriction