Theorem The global inclusion for actual ambient boundary-sheaf edge images
✓ theorem proved

The global inclusion for actual ambient boundary-sheaf edge images

Fix finite-generator real Coxeter systems \(W,W'\), an interval order isomorphism \(\varphi:[u,b]_W\simeq[u',b']_{W'}\), and a finite real basis of the source extended realization. Let \(A\) be the polynomial coefficient algebra and \(\mathcal B_b\) the underlying \(A\)-module boundary sheaf on \([1,b]_W\).

Choose an actual edge \(e:x\to_Wy\) with \(u\le_Wx\). Let \(a\) be its positive source root and \(\alpha'_{\varphi,e}\) its mapped edge root. Write \(\lambda(a)\) for the root polynomial. Put \[ T=T_{\varphi,\alpha'_{\varphi,e}},\qquad F=T(x),\qquad G=T(y), \] where \(T\) is the ambient later-edge set and \(T(x),T(y)\) are its outgoing selections.

Inside \(N=\mathcal B_b^e\), define \[ I=\rho_e^-(K_x^{\mathcal B_b}(F)),\qquad J=\rho_e^+(K_y^{\mathcal B_b}(G)),\qquad f=F_{x,a}\in A. \] The polynomial \(F_{x,a}\) is the fixed correction lift. Assume the upper selected kernel \(K_y^{\mathcal B_b}(G)\) is free over \(A\). Then \[ I\subseteq fJ. \]

Relation to the OpenAI notes

Prime divisors of one multiplier organize the local-to-global step.

OpenAI notes: The notes use membership in a free target at height-one primes. Primes avoiding all reduced root labels give the full localized edge module; root primes use the dihedral local comparison. edge:lattice, edge:bound.

Lean (primary): The proof fixes one global multiplier, the correction lift times the outgoing label product. A prime dividing its reduction is associated to a reduced root and receives the corresponding plane witness. A prime not dividing it uses the global multiplier itself. The free-target prime-removal theorem then gives membership. TerminalKernels:992, TerminalKernels:1045.

Role in the proof: The Blueprint follows these two prime cases and checks quotient-ring freeness, torsion-free cancellation, nonzero global multiplication and every local witness before concluding the inclusion.

Proof

Step 1. Establish the coefficient action and the free target

Set \(C=A/(\lambda(a))\), and write \(S=fJ\subseteq N\). The upper-image freeness theorem gives a compatible \(C\)-action on \(N\), makes \(N\) free over \(C\), and makes the actual upper image \(J_C\) free over \(C\) using the assumed freeness of the selected upper kernel.

The ring \(C\) is a domain, so its free module \(N\) is torsion-free. The correction-action theorem gives injectivity of multiplication by the fixed lift \(f\) on \(N\). The scalar-image freeness theorem therefore makes \(S_C=(fJ)_C\) free over \(C\).

Step 2. Construct a nonzero global multiplier

Let \[ g=\prod_{h\in G}\lambda(a_h),\qquad c=fg, \] the outgoing label product followed by the correction factor.

The boundary sheaf satisfies label annihilation and has surjective upper restrictions. The outgoing-product image theorem gives \(gN\subseteq J\). Hence for every \(z\in N\), \[ cz=f(gz)\in fJ=S. \] The root-factor divisibility theorem gives \(\overline c\ne0\) in \(C\).

Step 3. Classify the prime divisors of this multiplier

The same root-factor theorem applies to the actual edge \(e\) and outgoing selection \(G\). Every prime \(\pi\mid\overline c\) is associated to \(\overline{\lambda(\beta)}\) for a positive root \(\beta\ne a\).

This verifies the root-factor hypothesis of plane-denominator membership for the particular multiplier \(c\) constructed in Step 2.

Step 4. Obtain a multiplier outside each plane prime

Fix \(z\in I\), and let \(\beta\ne a\) be any positive source root. Put \[ P=\operatorname{span}\{a,\beta\},\qquad R_P=A_{\mathfrak p_P}. \] Write \(\eta_P:A\to R_P\) for localization. The actual ambient plane-comparison theorem applies to the same edge, threshold and endpoint selections. It gives equality of the localized lower image with \(\eta_P(f)\) times the localized upper image.

Apply the scaled-image denominator theorem with original ring \(A\), localization \(R_P\), multiplicative set \(A\setminus\mathfrak p_P\), scalar \(f\), and the chosen lower-image vector \(z\). It supplies \[ s\notin\mathfrak p_P,\qquad sz\in fJ=S. \] This works for every positive \(\beta\ne a\).

Step 5. Remove all denominators for the chosen lower-image vector

Apply plane-denominator membership to \(N,S,z,c\). Step 1 supplies the compatible actions, torsion-free ambient module and free target. Step 2 supplies a nonzero reduced multiplier and \(cz\in S\). Step 3 classifies its prime divisors, and Step 4 supplies all plane witnesses.

The theorem gives \(z\in S\). Since the choice \(z\in I\) was arbitrary, every vector in the lower image belongs to \(fJ\), proving \(I\subseteq fJ\).

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.BruhatGraph.boundary_edge_image_le: TerminalKernels.lean:1045.

Used by