Sections commute with finite flat scalar extension
Let \(R\to S\) be a homomorphism of commutative rings, and let \(B\) be an \(R\)-linear sheaf on an ordered graph with finite vertex and edge sets. Suppose \(S\) is flat as an \(R\)-module: tensoring with \(S\) preserves exact sequences.
For a vertex subset \(\Omega\), write \(E_\Omega=\{e:s(e),t(e)\in\Omega\}\). The natural map \[ \Theta_\Omega:S\otimes_R\Gamma_\Omega(B) \xrightarrow{\sim}\Gamma_\Omega(B_S) \] is an \(S\)-linear isomorphism. Its pure-tensor formula is \[ \Theta_\Omega(s\otimes f)_x=s\otimes f_x. \] The injectivity assertion requires finiteness only of the vertex set; the surjectivity proof also uses the finite edge set.
Relation to the OpenAI notes
Explicit finite-flat construction.
OpenAI notes: The localization discussion uses flat extension of kernels and images and then decomposes the localized sheaf over the local ring on a dihedral coset. loc:models, loc:decomposition.
Lean (primary): The scalar-extension lemmas first treat general commutative rings and graph sheaves. Finite product maps identify tensor constraints; flat exactness gives section and selected-kernel identities. Image commutation is proved separately by tensor generators. MinimalExtensions:1157, MinimalExtensions:1162, TerminalKernels:279, TerminalKernels:360.
Role in the proof: These algebraic results precede the root-plane and coset specialization. The later pullback argument controls the surviving edges, while both localized edge images retain their common ambient module.
Proof
Step 1. Commute the finite assignment product with tensoring
Put \(P=\prod_{x\in V}B^x\). The assignment map \[ \Phi:S\otimes_R P\to\prod_{x\in V}(S\otimes_R B^x) \] is bijective. Its inverse is the finite sum of the tensor maps induced by the coordinate inclusions \(\iota_x:B^x\to P\): \[ (z_x)_x\longmapsto\sum_x(1\otimes\iota_x)(z_x). \]
Projection onto a coordinate returns that coordinate, since the other coordinate inclusions project to zero. In the other direction, on \(s\otimes f\), the composite gives \[ \sum_xs\otimes\iota_x(f_x) =s\otimes\left(\sum_x\iota_x(f_x)\right) =s\otimes f. \] Pure tensors generate the tensor product, so these are inverse linear maps.
The same construction for the finite set \(E_\Omega\) gives \[ \Psi:S\otimes_R\prod_{e\in E_\Omega}B^e \xrightarrow{\sim}\prod_{e\in E_\Omega}(S\otimes_R B^e). \]
Step 2. Compare the two constraint maps
Let \(\delta\) be the section constraint for \(B\), and \(\delta_S\) the constraint for \(B_S\). They satisfy \[ \delta_S\Phi=\Psi(1\otimes\delta). \] On a pure tensor \(s\otimes f\), the \(e\)-coordinate of the left side is \[ s\otimes\rho^-_e(f_{s(e)}) -s\otimes\rho^+_e(f_{t(e)}) =s\otimes\bigl(\rho^-_e(f_{s(e)})-\rho^+_e(f_{t(e)})\bigr), \] which is the right-hand coordinate. Additivity proves the identity on all tensors.
Step 3. Tensor the kernel sequence
Let \(i:\Gamma_\Omega(B)\hookrightarrow P\) be the section inclusion. The constraint-kernel identity gives the exact sequence \[ 0\longrightarrow\Gamma_\Omega(B)\xrightarrow{i}P\xrightarrow{\delta} \prod_{e\in E_\Omega}B^e. \] Exactness here says that \(i\) is injective and its image is \(\ker\delta\).
Flatness gives injectivity of \(1\otimes i\) and \[ \operatorname{im}(1\otimes i)=\ker(1\otimes\delta). \]
Step 4. Lift an arbitrary section after base change
Take \(f\in\Gamma_\Omega(B_S)\). By Step 1 there is a tensor \(z\in S\otimes_R P\) with \(\Phi(z)=f\) as assignments.
Compatibility gives \(\delta_S(f)=0\). Step 2 then gives \(\Psi((1\otimes\delta)z)=0\). Since \(\Psi\) is injective, \((1\otimes\delta)z=0\).
Step 3 supplies \(w\in S\otimes_R\Gamma_\Omega(B)\) with \((1\otimes i)w=z\). Therefore \[ \Theta_\Omega(w)=\Phi((1\otimes i)w)=\Phi(z)=f. \] This proves surjectivity.
Step 5. Prove injectivity and assemble the isomorphism
If \(\Theta_\Omega(w)=\Theta_\Omega(w')\), equality of their underlying assignments gives \[ \Phi((1\otimes i)w)=\Phi((1\otimes i)w'). \] Injectivity of \(\Phi\) gives equality after \(1\otimes i\), and flatness makes that map injective. Hence \(w=w'\).
This step uses the finite vertex-product map but not the edge-product map. Combining both directions gives tensorSectionsEquiv.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.MomentGraph.Sheaf.tensorAssignments_bijective: MinimalExtensions.lean:1102.
OAI.KLInvariance.MomentGraph.Sheaf.constraint_tensorAssignments: MinimalExtensions.lean:1107.
OAI.KLInvariance.MomentGraph.Sheaf.tensorSections_surjective: MinimalExtensions.lean:1121.
OAI.KLInvariance.MomentGraph.Sheaf.tensorSections_injective: MinimalExtensions.lean:1147.
OAI.KLInvariance.MomentGraph.Sheaf.tensorSectionsEquiv: MinimalExtensions.lean:1157.