Polynomial reflection in a fixed degree
Let \(d\in\mathbb N\) and \(p(q)=\sum_{i\ge0}c_iq^i\in\mathbb Z[q]\), with only finitely many nonzero coefficients. Polynomial reflection in degree \(d\) reverses the coefficient positions from zero through \(d\) and fixes the positions above \(d\): \[ \operatorname{reflect}_dp =\sum_{i=0}^d c_iq^{d-i}+\sum_{i>d}c_iq^i. \] Equivalently its coefficient at \(n\) is \[ (\operatorname{reflect}_dp)_n= \begin{cases} c_{d-n},&n\le d,\\ c_n,&n>d. \end{cases} \] These are the two coefficient cases used in Lean's lower-half solution proof.
For example, \[ \operatorname{reflect}_2(2+3q)=3q+2q^2, \] \[ \operatorname{reflect}_2(2+3q+5q^3)=3q+2q^2+5q^3. \] The number \(d\) is part of the operation.
The bounded reciprocal formula
If \(\operatorname{natDegree}(p)\le d\), the second sum is zero. At any invertible parameter \(q_0\) in a commutative coefficient ring, \[ (\operatorname{reflect}_dp)(q_0)=q_0^dp(q_0^{-1}). \] The coefficient map from \(\mathbb Z\) is understood. A term \(c_iq_0^dq_0^{-i}\) equals \(c_iq_0^{d-i}\), so summing proves the equality.
Lean uses \(\operatorname{natDegree}(0)=0\). A bound by \(d\ge0\) thus includes the zero polynomial and every nonzero polynomial supported in degrees at most \(d\).
Linearity and involutivity
Reflection permutes coefficient positions. It therefore preserves sums, differences and scalar multiples. Applying it twice sends \(i\le d\) to \(d-i\) and back to \(i\), while positions \(i>d\) are fixed both times. Hence \[ \operatorname{reflect}_d(\operatorname{reflect}_dp)=p. \]
If \(\deg p\le a\) and \(\deg r\le b\), then \[ \operatorname{reflect}_{a+b}(pr) =(\operatorname{reflect}_ap)(\operatorname{reflect}_br). \] Indeed, a product monomial in degree \(i+j\) moves to \(a+b-i-j=(a-i)+(b-j)\). Summing the coefficient products proves the identity. These degree hypotheses are the ones checked before reflecting a product in the KL recursion.
Reciprocal and anti-reciprocal polynomials
A polynomial is reciprocal in degree \(d\) if \(\operatorname{reflect}_dp=p\), and anti-reciprocal in degree \(d\) if \(\operatorname{reflect}_dp=-p\).
For an anti-reciprocal integral polynomial, every coefficient above \(d\) satisfies \(c_i=-c_i\), so \(2c_i=0\) and \(c_i=0\). Its degree is therefore at most \(d\). Within the range, \(c_{d-i}=-c_i\). When \(d=2m\), the middle coefficient also vanishes, since \(c_m=-c_m\).
For instance \(q^3+2q^2-2q-1\) is anti-reciprocal in degree three. Its low coefficients determine the high coefficients by the displayed sign relation.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.eval₂_reflect_unit: Hecke.lean:1101.
OAI.KLInvariance.reflect_finset_sum: OrdinaryR.lean:236.
OAI.KLInvariance.CanonicalP.lowPart_solution: CanonicalKL.lean:43.
Used by
- Character reflection from a homogeneous dual equivalence
- Rigidity of a low-degree polynomial under reciprocal evaluation
- The grading shift in stalk–costalk reciprocity
- The reciprocal convolution identity for R-polynomials
- Polynomial R-convolution for word-stalk and kernel characters
- R-reciprocity for the actual boundary-sheaf character
- Lower-character normalization and convolution on an ambient interval
- Scaled reciprocity and its restriction to an upper interval
- The strict lower-half truncation
- The normalization data for ordinary R and KL polynomials
- Normalized KL reciprocity in the square parameter
- Anti-reciprocity of the recursively constructed defect
- A reciprocal polynomial strictly below the midpoint is zero
- Solving an anti-reciprocal defect by its lower half
- The reversed R-kernel with its bottom term removed