Scaled polynomial evaluations and KL vectors
For an integral polynomial \(p\), a natural number \(d\), and a real parameter \(t\), define \[ S_d(p;t)=t^{-d}p(t^2). \] Integer coefficients are evaluated in \(\mathbb R\). This is Comparison.scaledPolynomial in the source.
For a Coxeter system and a parameter \(t\) in any field \(C\), the source also defines \[ p_{x,b}(t) =t^{\ell(x)-\ell(b)}P_{x,b}(t^2), \] using the selected canonical KL polynomial. This is scaledKL. Its exponent is the integer difference of lengths.
The source formulas use field integer powers at every \(t\); their reciprocal identities assume \(t\ne0\). For \(x\le b\) and real \(t\ne0\), the two definitions agree with \(d=d(x,b)\).
The exponent of each coefficient
If \(p(q)=\sum_i a_iq^i\) and \(t\ne0\), then \[ S_d(p;t)=\sum_i a_i t^{2i-d}. \] Thus twice the polynomial degree appears in the exponent. The strict bound \(2\,\operatorname{natDegree}(p)<d\) puts every nonzero term at a negative exponent.
Under parameter inversion, \[ S_d(p;t^{-1})=t^dp(t^{-2}) =\sum_i a_i t^{d-2i}. \] It reverses the signs of the exponents.
For \(p=1+2q\) and \(d=3\), \[ S_3(p;t)=t^{-3}+2t^{-1},\qquad S_3(p;t^{-1})=t^3+2t. \] For the rank-one KL value \(P_{1,s}=1\), the scaled value is \(t^{-1}\), while the top value \(p_{s,s}\) is one.
Linearity at a fixed rank
For fixed \(d,t\), evaluation and multiplication by \(t^{-d}\) preserve sums and differences: \[ S_d(p-p';t)=S_d(p;t)-S_d(p';t). \] The reciprocal-difference rigidity theorem uses this identity to pass from equality of two scaled differences to equality of their underlying integral polynomials.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.Comparison.scaledPolynomial: SimultaneousInvariance.lean:325.
OAI.KLInvariance.scaledKL: Dihedral.lean:730.