Theorem Structural properties and kernels of finite sheaf sums
✓ theorem proved

Structural properties and kernels of finite sheaf sums

Let \(R\) be a commutative ring, \(G\) a labelled ordered graph, and \((B_i)\) a finite family of sheaves on \(G\). Their finite sum preserves free stalks, finite free stalks, global generation, flabbiness and the upper-quotient rule whenever the summands have those properties.

For every vertex \(x\), \[ K_x\!\left(\bigoplus_iB_i\right)=\prod_iK_x(B_i) \] under the componentwise stalk identification. For any ideal \(J\subseteq R\), the inclusions \(K_x(B_i)\subseteq JB_i^x\) imply \[ K_x\!\left(\bigoplus_iB_i\right) \subseteq J\left(\bigoplus_iB_i^x\right). \] Thus the result applies both to a local maximal ideal and to the positive ideal of a graded coefficient algebra.

For graded sheaves, finite graded sums and nonnegative grading shifts also preserve the corresponding degree conditions. The positive-ideal case gives graded kernel minimality.

Proof

Step 1. Read compatibility one component at a time

The restriction equation for an assignment \(m\) on an edge is an equality of tuples: \[ (\rho^{i,-}_e(m_{s(e),i}))_i =(\rho^{i,+}_e(m_{t(e),i}))_i. \] It holds exactly when every coordinate equation holds.

For global generation, realize each prescribed coordinate \(v_i\in B_i^x\) by a global section of \(B_i\), then take their tuple. For flabbiness, extend each component of a compatible assignment on an upper set, then assemble the resulting global sections. The coordinate equations verify compatibility and agreement on the upper set.

Step 2. Check the upper-quotient property

For an edge \(e:x\to y\), componentwise surjectivity makes \(\prod_i B_i^y\to\prod_i B_i^e\) surjective.

A tuple \((v_i)\) lies in its kernel exactly when \(v_i\in\alpha_e B_i^y\) for every \(i\). Choose \(w_i\) with \(v_i=\alpha_e w_i\). Then \[ (v_i)_i=\alpha_e(w_i)_i. \] Conversely, every such scalar multiple maps to zero in each edge component. Thus the kernel is \(\alpha_e\prod_i B_i^y\).

Step 3. Identify the outgoing kernels

A stalk tuple has zero restriction on every outgoing edge exactly when each component does. This gives the displayed product description of the full outgoing kernel. The same argument applies to any selected set of outgoing edges.

Step 4. Assemble membership in an ideal multiple

For modules \(M_i\) indexed by a finite set, suppose \(v_i\in JM_i\) for every \(i\). Each \(v_i\) is a finite sum of terms \(a m\) with \(a\in J\). Insert each of these vectors into coordinate \(i\), putting zero in all other coordinates.

Every inserted term belongs to \(J\prod_i M_i\), because coordinate insertion is linear. Summing first within each coordinate and then over the finite index set gives \[ (v_i)_i\in J\prod_iM_i. \] Apply this to \(M_i=B_i^x\) and the kernel tuple from Step 3. The component minimality assumptions yield minimality of the sum.

Step 5. Check freeness and shifts

Choose a free basis in each stalk summand. The coordinate insertions of these basis vectors form a basis of the finite product: coordinatewise expansions prove spanning, and coordinatewise independence proves independence. If the stalk bases are finite, their finite union is finite, giving finite free stalks.

A grading shift changes the degrees but leaves the underlying modules and restrictions unchanged. Therefore it preserves the section, quotient, and kernel conditions. Its nonnegative shift also retains the grading requirements.

The compatibility, generation, flabbiness, quotient, and kernel arguments are componentwise even for arbitrary products of ungraded sheaves. The finite-family assumption supplies the ideal-coordinate sum in Step 4 and, with finite free inputs, the finite basis in Step 5.

References

The local source declarations below give the constructions and proofs.

Lean formalization

OAI.KLInvariance.MomentGraph.Sheaf.pi_compatible: GradedSheaves.lean:825.

OAI.KLInvariance.MomentGraph.Sheaf.pi_generated: GradedSheaves.lean:835.

OAI.KLInvariance.MomentGraph.Sheaf.pi_flabby: GradedSheaves.lean:842.

OAI.KLInvariance.MomentGraph.Sheaf.pi_upperQuotient: MinimalExtensions.lean:315.

OAI.KLInvariance.MomentGraph.Sheaf.pi_kernel_iff: GradedSheaves.lean:849.

OAI.KLInvariance.LocalCover.pi_mem_ideal_smul: GradedSheaves.lean:874.

OAI.KLInvariance.MomentGraph.Sheaf.pi_minimal: GradedSheaves.lean:893.

Used by