Structural properties of a model sum and minimality away from its centers
Let \((O_c)\) be a boundary-model family and let \(C\) be a model sum centered in \(U\). Then \(C\) has free stalks, the prescribed upper-quotient rule, global generation, and flabbiness.
At any vertex \(x\notin U\), \[ K_x(C)\subseteq JC^x. \] If \(U\subseteq T\), the same construction is also a model sum centered in \(T\).
Proof
Step 1. Verify the zero constructor
The zero sheaf has free stalks of rank zero. Its upper restrictions \(0\to0\) are surjective with kernel \(0=\alpha_e0\).
Its unique assignment is global, realizes the only vector at every stalk, and extends its restriction from every upper set. Every outgoing kernel is zero and lies in \(J0\). Thus all assertions hold for the initial constructor.
Step 2. Pass through an added block
Suppose the assertions hold for the preceding sum \(C\), and add \[ S=\bigoplus_{j=1}^mO_c(d_j),\qquad c\in U. \] Every model has free stalks, upper quotients, generation, and flabbiness. Nonnegative shifts retain those properties, and the finite-sum theorem transfers them to \(S\), then to \(C\oplus S\).
Step 3. Preserve minimality outside U
Fix \(x\notin U\). Since \(c\in U\), we have \(x\ne c\). The model-family minimality condition gives \(K_x(O_c)\subseteq JO_c^x\). It holds for every shift \(O_c(d_j)\), since the underlying maps are unchanged.
The finite-sum kernel and ideal-multiple theorem gives minimality of \(S\) at \(x\). The induction hypothesis gives it for \(C\). Applying the same theorem to their product gives \(K_x(C\oplus S)\subseteq J(C\oplus S)^x\).
Step 4. Enlarge the allowed center set
Every center used by the construction belongs to \(U\). If \(U\subseteq T\), it also belongs to \(T\). The zero constructor is valid for either set, and every added block remains valid with the enlarged center condition. Induction on the construction proves the final assertion.
References
The local source declarations below give the constructions and proofs.
Lean formalization
OAI.KLInvariance.MomentGraph.GradedSheaf.ModelSum.mono: MinimalExtensions.lean:683.
OAI.KLInvariance.MomentGraph.GradedSheaf.ModelSum.free: MinimalExtensions.lean:689.
OAI.KLInvariance.MomentGraph.GradedSheaf.ModelSum.quotient: MinimalExtensions.lean:698.
OAI.KLInvariance.MomentGraph.GradedSheaf.ModelSum.generated: MinimalExtensions.lean:707.
OAI.KLInvariance.MomentGraph.GradedSheaf.ModelSum.flabby: MinimalExtensions.lean:715.
OAI.KLInvariance.MomentGraph.GradedSheaf.ModelSum.minimal: MinimalExtensions.lean:720.