Chosen roots for the supported plane factors
Fix a finite-generator real Coxeter system \(W\), its original root space \(V_0\), the extended realization \(V\), and the embedding \(i_0:V_0\to V\). Fix \(x\in W\) and a positive root \(a\).
Let \(k_P=k_P(x,a)\) be the geometric plane weights and \[ \mathcal P(x,a)=\{P:k_P\ne0\} \] their finite support. The weights count half-roots grouped by the plane they span with \(a\), so \(k_P\ge0\).
For each \(P\in\mathcal P(x,a)\), choose a positive root \(b_P\) satisfying \[ b_P\ne a,\qquad \operatorname{span}_{\mathbb R}\{a,b_P\}=P. \] The source calls this chosen factor representative root planeRoot.
Existence of the representative
A nonzero coefficient in the grouped plane-weight function comes from at least one root in the support of the half-root indicator. Such a root is different from \(a\), and the grouping map sends it to its span with \(a\).
This gives a witness \(b\ne a\) with \(\operatorname{span}\{a,b\}=P\). The definition planeRoot chooses a witness to precisely these two properties. The theorems planeRoot_ne and planeRoot_plane retain them for every supported plane.
Reduction to the edge coefficient ring
Write \[ S=\operatorname{Sym}_{\mathbb R}(V),\qquad \iota:V\to S,\qquad E_a=S/(\iota(i_0a)), \] and let \(q_a:S\to E_a\) be the quotient map. The reduced plane root attached to \(P\) is \[ r_P=q_a(\iota(i_0b_P)). \] This is reducedPlaneRoot.
Since distinct positive roots are independent, \(i_0b_P\) is not in the line \(\mathbb R i_0a\). The linear-quotient identification \[ E_a\simeq\operatorname{Sym}_{\mathbb R}(V/\mathbb R i_0a) \] therefore sends \(r_P\) to the linear generator of a nonzero quotient vector. The edge-factor properties theorem proves that each \(r_P\) is nonzero and has degree one in quotient polynomial coordinates.
These finite representative data are the factors used in the reduced-root product and in its polynomial lift.
Relation to the OpenAI notes
The same root product with explicit weight and coordinate constructions.
OpenAI notes: The notes assign each plane its half-excess relative dihedral length, choose an independent root whenever that weight is positive, and multiply its reduced linear form to that power in the edge quotient. edge:bound.
Lean (primary): The weights are first constructed by grouping half-root indicators. planeWeight_coset_edge identifies them with the relative dihedral length formula. planeRoot chooses a positive root with the required plane, edgeFactor is a product in the symmetric quotient, and edgeFactorLift is its explicit product before the polynomial quotient. The definitions and nonvanishing hold for every x; noninversion is used for the length-degree formula. ReflectionSchedules:908, ReflectionSchedules:944, ReflectionSchedules:985, ReflectionSchedules:993, PlaneLocalization:1389, PlaneLocalization:1413, PlaneLocalization:1096.
Role in the proof: The Blueprint keeps the chosen representatives, all scalar maps and the natural/integer exponent conversion explicit. The quotient of the polynomial lift is exactly the constructed edge factor.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.TitsSpace.planeWeight_support_root: ReflectionSchedules.lean:895.
OAI.KLInvariance.TitsSpace.planeRoot: ReflectionSchedules.lean:908.
OAI.KLInvariance.TitsSpace.planeRoot_ne: ReflectionSchedules.lean:912.
OAI.KLInvariance.TitsSpace.planeRoot_plane: ReflectionSchedules.lean:916.
OAI.KLInvariance.TitsSpace.reducedPlaneRoot: ReflectionSchedules.lean:933.