An actual ambient edge has coordinates in a least plane coset
Let \(W\) be a finite-generator real Coxeter system, \(P\) a root plane with a plane basis \(B\), and \(D\) its induced dihedral subgroup. Let \(e:x\to_Wy\) be an edge of the ambient interval graph on \([1,b]_W\), and suppose its positive root \(a_e\) lies in \(P\).
There are a least coset representative \(a_0\), vertices \(g,h\) of \[ K=\{d\in D:da_0\le_Wb\}, \] and an intrinsic edge \(d:g\to_Dh\) of \(K\) such that \[ ga_0=x,\qquad ha_0=y,\qquad j_E(d)=e. \] Here \(j_E\) is the coset graph map, and least means \[ \ell_W(a_0)\le\ell_W(za_0)\qquad(z\in D). \]
Relation to the OpenAI notes
The full ambient edge set and its relative-coset restriction.
OpenAI notes: The local comparison uses the plane-subgroup coset through the current ambient edge and restricts the transported later-edge conditions to that coset. dih:plane, edge:bound.
Lean (primary): ambientLater is an edge set on the full graph [1,b], with an explicit source-above-u witness. scheduledOutgoing_eq is its inverse-image identity on the coset graph. Least coset coordinates for the source and multiplication by the actual edge reflection construct the relative edge, and both mapped-root definitions evaluate the same target edge. PlaneLocalization:1215, PlaneLocalization:1237, PlaneLocalization:1700, PlaneLocalization:1715, PlaneLocalization:1744.
Role in the proof: The Blueprint defines the ambient selection before transferring the two endpoint kernels, and constructs the relative coordinates with both endpoint bounds and the exact equality of edges.
Proof
Step 1. Choose least coset coordinates for the source
The least-coset representative theorem, applied to the source \(x\), gives \(a_0\) and \(g\in D\) such that \[ x=ga_0,\qquad \ell_W(a_0)\le\ell_W(za_0)\quad(z\in D). \] The ambient bound \(x\le_Wb\) makes \(g\) a vertex of \(K\).
Step 2. Use the edge reflection to construct the target coordinate
The left reflection of \(e\) is \(t=yx^{-1}\), represented by the positive root \(a_e\). Since \(a_e\in P\), the plane-reflection membership theorem gives \(t\in D\).
Set \(h=tg\in D\). Then \[ ha_0=tga_0=tx=y. \] The bound \(y\le_Wb\) makes \(h\) a vertex of \(K\). In the source's full CosetVertex data, the additional inequalities \(1\le_Dg\) and \(1\le_Dh\) hold because the identity is below every element.
Step 3. Recover the intrinsic edge
The ambient edge \(x\to_Wy\) can now be written \(ga_0\to_Wha_0\). The representative \(a_0\) is least, so the coset graph theorem gives \[ g\to_Dh. \] Together with the two vertex memberships, this defines the relative coset edge \(d\).
Step 4. Identify the actual ambient edge
The graph map sends \(d\) to the ambient edge with source \(ga_0=x\) and target \(ha_0=y\). Ambient Bruhat graph edges are endpoint pairs satisfying the reflection-step condition. Both endpoints therefore agree with those of \(e\), giving \(j_E(d)=e\).
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.TitsSpace.PlaneBasis.exists_coset_edge: PlaneLocalization.lean:1715.