OctoInformBeta

Combinatorial invariance of Kazhdan–Lusztig polynomials

Let (W,S)(W,S) be a Coxeter system, with length function ℓ\ell and Bruhat order ≤\le. For u≤bu\le b in WW, the Bruhat interval [u,b][u,b] is the set of x∈Wx\in W with u≤x≤bu\le x\le b, ordered by ≤\le. It is a finite graded poset of rank ℓ(b)−ℓ(u)\ell(b)-\ell(u), even when WW is infinite. Write Pu,b(q)P_{u,b}(q) for the equal-parameter Kazhdan-Lusztig polynomial of u≤bu\le b, normalized by Pu,u=1P_{u,u}=1 and deg⁡Pu,b<(ℓ(b)−ℓ(u))/2\deg P_{u,b}<(\ell(b)-\ell(u))/2 for u<bu<b. A second Coxeter system (W′,S′)(W',S'), with u′≤b′u'\le b' in W′W', has its polynomials marked with a prime, as in Pu′,b′′(q)P'_{u',b'}(q). An isomorphism ι:[u,b]→[u′,b′]\iota:[u,b]\to[u',b'] is a bijection that preserves and reflects order, with no reflection labels or root data attached. The combinatorial invariance conjecture asks whether Pu,b=Pu′,b′′P_{u,b}=P'_{u',b'} whenever such an isomorphism exists.

The conjecture is attributed to Lusztig and Dyer. Dyer’s published formulation of it is for finite Coxeter systems. The directed Bruhat graph of [u,b][u,b] is the directed graph on [u,b][u,b] with an edge x→yx\to y whenever y=tx>xy=tx>x for a reflection tt, whether or not yy covers xx. For finite Coxeter systems, Dyer reconstructed the directed Bruhat graph of an interval from its order. A principal lower interval is an interval [1,b][1,b], where 11 is the identity of WW. For principal lower intervals, du Cloux proved that an isomorphism ι:[1,b]→[1,b′]\iota:[1,b]\to[1,b'] gives Px,y=Pι(x),ι(y)′P_{x,y}=P'_{\iota(x),\iota(y)} for all x≤yx\le y in [1,b][1,b] when one of the two groups belongs to a class that includes every finite and every affine Coxeter group. Brenti proved this for symmetric groups. Brenti, Caselli and Marietti proved it for arbitrary Coxeter systems. Delanoy also proved it for arbitrary Coxeter systems. These results use the positions of xx and yy inside [1,b][1,b], and not only the interval [x,y][x,y]. For intervals with an arbitrary bottom element, Patimo proved that in finite simply-laced Weyl groups an isomorphism preserves the coefficient of qq in Pu,bP_{u,b}. Barkley, Gaetz and Lam proved the same for arbitrary Coxeter systems. They deduced the conjecture for intervals with ℓ(b)−ℓ(u)≤6\ell(b)-\ell(u)\le 6. Esposito, Marietti and Stella proved the conjecture for ℓ(b)−ℓ(u)≤10\ell(b)-\ell(u)\le 10 in finite Weyl groups. They proved it for ℓ(b)−ℓ(u)≤12\ell(b)-\ell(u)\le 12 in finite type AA.

The preprint proves the conjecture for arbitrary Coxeter systems, finite or infinite, crystallographic or not: every isomorphism ι:[u,b]→[u′,b′]\iota:[u,b]\to[u',b'] gives Pu,b(q)=Pu′,b′′(q)P_{u,b}(q)=P'_{u',b'}(q) (Theorem 1.1). For arbitrary Coxeter systems, it also proves that ι\iota preserves and reflects the directed Bruhat graphs (Theorem 3.4).

The Lean library KazhdanLusztig proves Theorem 1.1 for Coxeter systems of any rank, with no further hypothesis. Lean has checked the proof of every result of the Lean library, so the Lean library is the reference. Each result on these pages is stated as the Lean library states it and linked to the Lean declaration behind it, with the preprint’s statement beside it where the two differ. The preprint is a guide only: it gives the results their order and their names. The Lean library constructs, on the directed Bruhat graph of each interval [1,b][1,b], the Braden-MacPherson sheaf that the preprint takes from Braden and MacPherson and from Fiebig. A geometric reflection order lists the positive roots α\alpha by increasing f(α)/h(α)f(\alpha)/h(\alpha), where hh is a linear form positive on the simple roots and ff is a linear form that makes these ratios distinct. The Lean library proves for geometric reflection orders, the ones the preprint uses, the path formula that the preprint takes from Dyer. The Lean library states, but does not prove, that the stalk at each x∈[1,b]x\in[1,b] of the sheaf on [1,b][1,b] has graded rank Px,b(q)P_{x,b}(q). The preprint derives this identity from results of Fiebig and the character theorem of Elias and Williamson. Theorem 1.1 in the Lean library does not depend on this identity.

The Lean library: KazhdanLusztig

KazhdanLusztig is a Lean library: 30 items, in six groups. Each has a page written by OctoInform, listed below.

§1. Introduction 2 items
  1. Bruhat intervals and the equal-parameter polynomials Bruhat intervals are finite and Kazhdan-Lusztig polynomials well-defined. §1
  2. Theorem 1.1 Isomorphic Bruhat intervals have equal Kazhdan-Lusztig polynomials.
§2. Real realizations and moment-graph sheaves 7 items
  1. Reduction to finite rank Every Bruhat interval has a model of finite rank. §2.1
  2. Lemma 2.1 A nondegenerate reflection-faithful realization of least dimension.
  3. Geometric reflection orders and normalized R-polynomials Slopes of positive roots and the normalized R-polynomial. §2.3
  4. Theorem 2.2 Dyer’s path formula for geometric orders.
  5. The Bruhat moment graph, sheaves and upward kernels The Bruhat moment graph of an interval and its sheaves. §2.4
  6. Theorem 2.3 Minimal moment-graph sheaves exist on every Bruhat interval.
  7. Proposition 2.4 Stalk characters of the boundary sheaf satisfy reciprocity.
§3. Dihedral intervals and transported reflection orders 8 items
  1. Lemma 3.1 Reflections with roots in a plane generate a dihedral group.
  2. Lemma 3.2 Length decides Bruhat order and Bruhat graph in dihedral groups.
  3. Theorem 3.4 Order isomorphisms of Bruhat intervals preserve the Bruhat graph.
  4. Proposition 3.5 Interval isomorphisms transport dihedral intervals of length at least two.
  5. Lemma 3.6 Root slopes order plane dihedral reflections by signed index.
  6. Transported schedule, terminal portions and k_D(e) Edges of a Bruhat interval ordered through an interval isomorphism. §3.4
  7. Lemma 3.7 The count of later edges drops by the plane weight.
  8. Lemma 3.8 Root planes split the length gap of a Bruhat edge.
§4. Localization at a root plane 5 items
  1. Lemma 4.1 Kazhdan-Lusztig polynomials of dihedral groups are trivial.
  2. Structure sheaves and the local ring Structure sheaves and the local ring of a root plane. §4
  3. Lemma 4.2 Localized structure sheaves on dihedral cosets are local models.
  4. Lemma 4.3 Sheaves with the quotient rule stay flabby on a localized coset.
  5. Proposition 4.4 Localized on a plane coset, B(b)B(b) splits into structure sheaves.
§5. Comparison of edge images 2 items
  1. Lemma 5.1 A free submodule as the intersection of its height-one localizations.
  2. Proposition 5.2 A free upper kernel bounds the lower edge image.
§6. Reciprocity and the simultaneous induction 6 items
  1. The simultaneous statement Isomorphic Bruhat intervals have equal R̃\tilde R and free terminal kernels. §6
  2. A nonnegative comparison (6.1) to (6.3) Free terminal kernels bound stalk characters through normalized R-polynomials. §6
  3. Polynomial equality Two-sided bounds force equal polynomials in the bottom row. §6
  4. Equality forces all edge images (6.6) to (6.8) Endpoint equality forces every edge image to its bound. §6
  5. Freeness closes the induction Matching edge images along a schedule make suffix kernels free. §6
  6. Corollary 6.1 Bruhat interval isomorphisms preserve all normalized R-polynomials in finite rank.

No item matches.