Coordinate polynomials and a polynomial family of functionals
Let \(I\) be a finite coordinate set of size \(N\), and choose an enumeration
\(\nu:I\simeq\{0,\ldots,N-1\}\). A vector \(v\in\mathbb R^I\) has coordinates \(v_i\).
Its coordinate polynomial is
\[
P_v(X)=\sum_{i\in I}v_i X^{\nu(i)}\in\mathbb R[X].
\]
Lean calls this ReflectionSchedules.coordinatePolynomial and uses its chosen finite enumeration of \(I\).
Recovery of the vector
Different coordinates have different exponents. The coefficient of \(X^{\nu(j)}\) in \(P_v\) is therefore exactly \(v_j\): the \(j\)-term contributes this coefficient, and all other monomials have different degrees. Consequently \(P_v=P_w\) implies \(v_j=w_j\) for each \(j\), hence \(v=w\). In particular, \[ P_v=0\Longleftrightarrow v=0. \] This gives the coefficient and injectivity lemmas used in the functional constructions.
Evaluation as a linear functional
For a real parameter \(\tau\), define
\[
m_\tau(v)=\sum_{i\in I}v_i\tau^{\nu(i)}=P_v(\tau).
\]
The source calls \(m_\tau\) the momentFunctional. At fixed \(\tau\), it is linear in \(v\): distributing the coordinate sum gives
\[
m_\tau(v+w)=m_\tau(v)+m_\tau(w),\qquad
m_\tau(c v)=c\,m_\tau(v).
\]
Thus \(\tau\mapsto m_\tau\) is a one-parameter family of real linear functionals.
A prescribed functional with polynomial perturbation
For a fixed real linear functional \(f_0\), put \[ f_\tau=f_0+\tau m_\tau. \] At a fixed vector \(v\), \[ f_\tau(v)=f_0(v)+\tau P_v(\tau). \] The right side is a polynomial in \(\tau\), so it is continuous. This continuity lets finitely many strict positivity conditions persist near \(\tau=0\).
If \(v\ne0\), the polynomial \[ d_v(X)=f_0(v)+X P_v(X) \] is nonzero. Indeed, \(d_v=0\) would give \(f_0(v)=d_v(0)=0\), then \(XP_v=0\), then \(P_v=0\), contradicting \(v\ne0\). It therefore has only finitely many real roots. The prescribed-chamber existence theorem applies this observation to differences of normalized roots.
For example, with two coordinates and \(v=(2,-1)\), one has \[ P_v(X)=2-X,\qquad m_\tau(v)=2-\tau. \] If \(f_0(v)=3\), then \(f_\tau(v)=3+2\tau-\tau^2\). This example illustrates the distinction between linearity in the vector and polynomial dependence on the parameter.
Relation to the OpenAI notes
An actual chamber numerator and the inversion-sign order.
OpenAI notes: The notes cite the reflection-order path formula and construct separating ratios of linear functionals for the geometric orders used in comparison. bg:paths, sec:background.
Lean (primary): The path proof chooses a numerator positive on the transformed simple roots of a prescribed chamber. It implies the inversion-sign condition and supplies a least simple descent. Existence uses height composed with the inverse action, a polynomial perturbation preserving finitely many strict inequalities, and avoidance of countably many root-tie polynomials. DyerPaths:1240, DyerPaths:1198, DyerPaths:1242, DyerPaths:1220, DyerPaths:1266, ReflectionSchedules:18, ReflectionSchedules:39.
Role in the proof: The Blueprint distinguishes the Chamber and Adapted predicates and explains the coordinate-polynomial family used to satisfy genericity and chamber positivity together.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.ReflectionSchedules.coordinatePolynomial: ReflectionSchedules.lean:18.
OAI.KLInvariance.ReflectionSchedules.coordinatePolynomial_coeff: ReflectionSchedules.lean:21.
OAI.KLInvariance.ReflectionSchedules.coordinatePolynomial_injective: ReflectionSchedules.lean:33.
OAI.KLInvariance.ReflectionSchedules.momentFunctional: ReflectionSchedules.lean:39.
OAI.KLInvariance.ReflectionSchedules.coordinatePolynomial_eval: ReflectionSchedules.lean:44.