The two induction stages of combinatorial invariance
The main argument proves two types of statements together at the R-polynomial stage: invariance of the normalized R-data and freeness of the selected kernels that occur in schedule suffixes. Smaller intervals supply the upper-kernel freeness needed for comparison at the current rank.
The forward comparison gives inequalities between the relevant polynomial expressions. The argument in the opposite direction and the smaller-interval identities isolate the remaining R-entry and force equality. Once the R-data agree, character reciprocity makes the comparison endpoints exact. Numerical equality then becomes image equality, and reverse traversal recovers the suffix-kernel freeness at the same rank.
A separate induction then treats the canonical KL family. The common R-data give the same reciprocity equation; the strict degree bound eliminates a remaining reciprocal difference. Finite-support reduction transfers the finite-generator conclusion to arbitrary Coxeter systems. The reading route keeps the simultaneous R/freeness induction and the later canonical-polynomial induction distinct.
Reading sequence
- Freeness of all terminal kernels in a transported schedule — Define the freeness statement included in the induction.
- The normalized R-family as bar coefficients — Define the normalized R-family being compared.
- The R-tilde induction step from smaller intervals — Read the rank step and its smaller data.
- Extracting the bottom R-tilde row from character comparison — Isolate the relevant matrix comparison.
- R-matrix equality makes the comparison endpoints exact — Make the endpoints exact from common R-data.
- Simultaneous R-tilde invariance and terminal-kernel freeness — Complete the simultaneous induction.
- KL invariance for finite-generator Coxeter systems — Run the canonical KL induction.
- Combinatorial invariance of Kazhdan–Lusztig polynomials — Pass to arbitrary Coxeter systems.