Theorem The stalk surjects onto the boundary image
✓ theorem proved

The stalk surjects onto the boundary image

Let \(B\) be a sheaf over a commutative ring \(R\) on an ordered graph. Suppose it is flabby and generated by global sections.

Fix a vertex \(x\). Write \[ \partial B_x=\prod_{e:\,s(e)=x}B^e \] for the boundary module, and define \[ r_x:B^x\to\partial B_x,\qquad r_x(v)_e=\rho^-_e(v), \] \[ b_x:\Gamma(B,\{y:x<y\})\to\partial B_x,\qquad b_x(f)_e=\rho^+_e(f_{t(e)}). \] The boundary image is \(B^{\delta x}=\operatorname{im}b_x\). Then \[ \operatorname{im}r_x=B^{\delta x}. \] Consequently \(r_x\), with codomain restricted to \(B^{\delta x}\), is a surjective linear map from the stalk onto the boundary image.

Proof

Step 1. Global generation gives the first containment

Take \(v\in B^x\). Global generation supplies a global section \(g\) with \(g_x=v\). Its assignment is compatible on the strict upper set \(U_x=\{y:x<y\}\), so it defines an element of \(\Gamma(B,U_x)\).

For each outgoing edge \(e:x\to y\), global compatibility says \[ r_x(v)_e=\rho^-_e(g_x)=\rho^+_e(g_y)=b_x(g)_e. \] Equality at every edge coordinate gives \(r_x(v)=b_x(g)\). Thus \(\operatorname{im}r_x\subseteq\operatorname{im}b_x\).

Step 2. Flabbiness gives the reverse containment

Take \(w\in\operatorname{im}b_x\), and choose \(f\in\Gamma(B,U_x)\) with \(b_x(f)=w\).

The set \(U_x\) is upper: if \(x<y\le z\), then \(x<z\). Flabbiness therefore supplies a global section \(g\) with \(g_y=f_y\) for every \(y\in U_x\).

The target \(t(e)\) of every outgoing edge is strictly above \(x\), so \[ r_x(g_x)_e =\rho^-_e(g_x) =\rho^+_e(g_{t(e)}) =\rho^+_e(f_{t(e)}) =b_x(f)_e. \] Thus \(r_x(g_x)=w\), giving \(\operatorname{im}b_x\subseteq\operatorname{im}r_x\).

Step 3. Restrict the codomain

The equality of images proves that every \(r_x(v)\) belongs to \(B^{\delta x}\), defining a linear map \(B^x\to B^{\delta x}\). For any vector of this boundary image, Step 2 supplies a preimage \(g_x\). Hence this map is surjective.

For the one-edge congruence sheaf, both boundary maps are \(A\to A/(\alpha)\), and the common image is \(A/(\alpha)\). For the boundary construction on a larger graph, this theorem identifies the precise image that the chosen free stalk must cover.

References

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

Lean formalization

OAI.KLInvariance.MomentGraph.Sheaf.range_stalkMap_eq_boundaryImage: Sheaves.lean:199.

OAI.KLInvariance.MomentGraph.Sheaf.stalkCover: Sheaves.lean:230.

OAI.KLInvariance.MomentGraph.Sheaf.stalkCover_surjective: Sheaves.lean:236.

Used by