Definition Local boundary models over a local coefficient ring
✓ definition formalized

Local boundary models over a local coefficient ring

Let \(R\) be a commutative local ring, with maximal ideal \(\mathfrak m\), and let \(G\) be an ordered graph with labels \(\alpha_e\in R\). A family of local boundary models consists of a sheaf \(O_c\) for every vertex \(c\), together with the following data and conditions.

Lean field Mathematical content
sheaf The family \(c\mapsto O_c\), including both endpoint restrictions on every edge.
free Every vertex stalk \(O_c^x\) is a free \(R\)-module.
finite Every vertex stalk is finitely generated. Together with free, this makes it finite free.
quotient Each upper map \(O_c^{t(e)}\to O_c^e\) is surjective and has kernel \(\alpha_eO_c^{t(e)}\).
generated Evaluation of global sections onto each stalk is surjective.
flabby Every compatible section on an upper vertex subset extends to a global section.
support The stalk \(O_c^x\) is zero whenever \(x\not\le c\).
top A chosen \(R\)-linear isomorphism \(O_c^c\simeq R\).
minimal For \(x\ne c\), the full outgoing kernel is contained in \(\mathfrak mO_c^x\).

The source packages these fields as LocalModels. The chosen top coordinate is part of the data. All the modules and maps here are ungraded.

The outgoing map and local minimality

At a vertex \(x\), collect the lower restrictions into \[ d_x:O_c^x\longrightarrow\prod_{e:s(e)=x}O_c^e, \qquad v\longmapsto(\rho^-_e(v))_e. \] Its kernel consists of vectors whose restrictions vanish on every outgoing edge. The minimal field is the inclusion \[ \ker d_x\subseteq\mathfrak mO_c^x\qquad(x\ne c). \]

At a vertex \(x\ne c\), let \(N_x=\operatorname{im}d_x\). The induced surjection \(\pi_x:O_c^x\to N_x\) becomes an isomorphism after reduction modulo \(\mathfrak m\). Indeed surjectivity persists, and \[ \pi_x^{-1}(\mathfrak mN_x) =\mathfrak mO_c^x+\ker\pi_x =\mathfrak mO_c^x. \] For the first equality, write an element of \(\mathfrak mN_x\) as a finite sum of ideal scalars times images and lift each image; the difference from this lifted sum lies in the kernel. The second equality is the minimality condition.

Thus the residue classes of a stalk basis provide exactly the generators needed for its outgoing image. This is the local-ring minimal-cover condition used in the decomposition.

A one-edge example

Let \(G\) consist of \(x<c\) with one edge labelled by \(\alpha\in\mathfrak m\). The structure model centered at \(c\) has stalks \(R,R\), edge module \(R/(\alpha)\), and quotient restrictions. At \(x\), \[ \ker d_x=\alpha R\subseteq\mathfrak mR. \] Its outgoing image is \(R/(\alpha)\), and reducing either the stalk or this image modulo \(\mathfrak m\) gives the residue field \(R/\mathfrak m\).

At the center \(c\), there are no outgoing edges and the outgoing kernel is \(R\). The minimality condition is imposed only at \(x\ne c\); the top condition instead specifies the free rank-one stalk at \(c\).

For the actual localized coset graph, the local structure-model theorem constructs these fields from the structure sheaves, their flabbiness, the maximal-ideal labels and the existence of outgoing edges below every larger center. Finite ungraded model sums assemble copies of these sheaves for the later decomposition theorem.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.MomentGraph.Sheaf.LocalModels: CharacterReciprocity.lean:352.

Used by