Alphabeta Math
LemmaStatement: Literature-sourcedProof: AI-adaptedPipeline-generatedprecheck pass
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

The Lefschetz operator factors through deleted Bott-Samelson subwords, with positive rescalings

Facts & Assumptions

Given: An expression x‾=s1⋯sm in simple reflections, the Bott-Samelson bimodule BS(x‾) with its subexpression right basis {cε}, top element ctop, trace Tr⁡ and intersection form ⟨−,−⟩ of The trace, the invariant intersection form on Bott-Samelson bimodules and its reduction, a strictly dominant ρ with associated operator L0=ρ⋅(−) (The reduced Soergel module and the dominant Lefschetz operator), and the auxiliary maps Bri(b1⋯bm):=b1⋯(bicsi)⋯bm, φi(b1⋯bm):=b1⋯μ(bi)⋯bm and χi(b1⋯bi^⋯bm):=b1⋯csi⋯bm, where μ:Bsi→S(1) is the multiplication map f⊗g↦fg and x‾^i is the word with letter si deleted. The maps Bri have internal degree two; their factorizations use degree-zero bimodule maps φi and χi into the displayed shifts. Thus Bri:BS(x‾)→BS(x‾) and φi:BS(x‾)→BS(x‾^i)(1), χi:BS(x‾^i)→BS(x‾)(1) with Bri=χi∘φi, and Tr⁡(χi(γ))=Tr⁡i(γ) for all γ∈BS(x‾^i), where Tr⁡i is the trace of BS(x‾^i).

[F1]

In each rank-one factor the elements cid,cs form a right S-basis with cidcid=cid, cidcs=cs, cscs=csαs and the left-action rule r⋅cid=cid⋅sr+∂s(r)cs, r⋅cs=cs⋅r, for r∈S (The trace, the invariant intersection form on Bott-Samelson bimodules and its reduction, The Soergel bimodule Bi of a simple reflection).

[F2]

The trace is right S-linear and the intersection form is symmetric and invariant (The trace, the invariant intersection form on Bott-Samelson bimodules and its reduction).

[F3]

Descent detection. For every w and simple s, (wρ)(αs∨)>0 if and only if sw>w (The reduced Soergel module and the dominant Lefschetz operator).

[F4]

The positive Rouquier generator complex is Fi=[Bsi→  S(1)] with Bsi in cohomological degree 0; the Rouquier complex of a word is the iterated signed tensor totalization of the Fi (The positive and negative Rouquier generator complexes, The Rouquier complex of a braid word).

Statement

Let x‾=s1⋯sm be an expression, BS(x‾) its real Bott-Samelson bimodule, Bri=χi∘φi the Demazure operation at the i-th letter, and Tr⁡,⟨−,−⟩ the trace and intersection form of The trace, the invariant intersection form on Bott-Samelson bimodules and its reduction. Then:

(a) as endomorphisms of BS(x‾) one has ρ⋅(−)=∑i=1m(si−1⋯s1ρ)(αsi∨) χi∘φi+(−)⋅x−1ρ (EW Lemma 3.4), where (−)⋅x−1ρ denotes right multiplication by the polynomial x−1ρ;

(b) with the rescaled forms ⟨−,−⟩i′:=(si−1⋯s1ρ)(αsi∨)⟨−,−⟩i on BS(x‾^i) and ⟨−,−⟩′ their direct sum, one has ⟨b,ρb′⟩BS(x‾)‾=⟨φ(b),φ(b′)⟩′ for all b,b′ in the reduced module, where φ=∑iφi:BS(x‾)→⨁iBS(x‾^i)(1) (EW Lemmas 6.14 and 6.15);

(c) if x‾ is reduced then every scaling factor γi:=(si−1⋯s1ρ)(αsi∨) is strictly positive, so the rescalings change no signs in the HR statements, and the unscattered term (−)⋅x−1ρ vanishes after reduction to the Soergel module;

(d) for a reduced expression, consequently left multiplication by ρ on BS(x‾)‾ factors, up to positive rescalings, as the composite BS(x‾)→⨁iBS(x‾^i)(1)→BS(x‾)(2), and this composite is the first differential of the Rouquier complex Fs1⋯Fsm followed by its dual map, up to the same positive rescalings.

Proof

technique · direct; a rank-one computation followed by iteration over the letters
1.1givenF1

The single-factor identity. Let s be a simple reflection with root α=αs, let Br(b):=bcs on Bs, and let σ∈S. For every b∈Bs one has σ⋅b=b⋅(sσ)+∂s(σ) Br(b). Indeed both sides are right S-linear in b, so it suffices to check the two basis elements: for b=cid the identity is the rule r⋅cid=cid⋅sr+∂s(r)cs of [F1] with r=σ, and for b=cs it reads σ⋅cs=cs⋅sσ+∂s(σ)csαs, which is csσ=cs(sσ+αs∂s(σ)) and holds because σ−sσ=αs∂s(σ) and cs commutes with S.

2.1step 1.1F1

Iteration over the letters proves (a). For m=0 both outer actions are multiplication by ρ on S. For m>0, put σ0=ρ and σi=siσi−1. Apply step 1.1 in slot i to the remainder b1⋯bi−1(σi−1bi)bi+1⋯bm. This produces ∂si(σi−1)Bri(b) and a remainder with σi acting on the right of slot i. Balancing moves that polynomial to the left of slot i+1; at the last slot it is the outer right action by x−1ρ. Each σi−1 is linear, so ∂si(σi−1)=σi−1(αsi∨)=γi. Also bcs=μ(b)cs in one factor: on cid and cs this follows from μ(cid)=1, μ(cs)=αs and cs2=csαs. Thus Bri=χiφi with the appropriate shift of χi, and the stated degree-two identity holds on every tensor monomial and hence on the whole bimodule.

3.1step 1.1step 2.1givenF1

The covariance identity. For all b,b′∈BS(x‾) and all i one has ⟨b,Brib′⟩=⟨φib,φib′⟩i. It suffices to check on tensor monomials b=b1⋯bm, b′=b1′⋯bm′: the left side is Tr⁡((b1b1′)⋯(bibi′csi)⋯(bmbm′))=Tr⁡((b1b1′)⋯μ(bi)μ(bi′)csi⋯(bmbm′)), using bicsi=μ(bi)csi from step 2.1, so this equals Tr⁡(χi((b1b1′)⋯μ(bi)μ(bi′)⋯(bmbm′)))=Tr⁡i((b1b1′)⋯μ(bi)μ(bi′)⋯(bmbm′))=⟨φib,φib′⟩i, using Tr⁡∘χi=Tr⁡i and the multiplicativity of the ring structure.

4.1step 2.1step 3.1F1algebra

The rescaled form identity (b). With φ=∑iφi and ⟨−,−⟩′=⨁i⟨−,−⟩i′, the direct sum decomposition is orthogonal, so for all b,b′ one has ⟨φ(b),φ(b′)⟩′=∑iγi⟨b,Brib′⟩; substituting step 3.1 gives ∑iγi⟨b,Brib′⟩=⟨b,∑iγiBrib′⟩, and substituting the factorization of step 2.1 gives ⟨b,ρb′⟩−⟨b,b′⋅x−1ρ⟩. The second summand is −⟨b,b′⟩ x−1ρ by right S-linearity of the form, and it vanishes after reduction because x−1ρ has positive degree; hence ⟨b,ρb′⟩BS(x‾)‾=⟨φ(b),φ(b′)⟩′ for b,b′ in the reduced module, as claimed in (b).

5.1step 4.1F1F2F3

Positivity for reduced words and the vanished remainder (c). Let x‾ be reduced. For each i, the prefix wi:=si−1⋯s1 has length i−1 and siwi has length i; the descent criterion [F3] therefore gives γi=(wiρ)(αsi∨)>0. All rescalings are by strictly positive real numbers, so no sign in a definite form changes; and the polynomial x−1ρ lies in S2⊂S+ whose right multiplication lands in BS(x‾)S+, which becomes zero after reduction, so the remainder vanishes on the reduced module. This proves (c).

6.1step 2.1step 4.1step 5.1F2F4∎

Identification with the Rouquier differential (d). By [F4] the first differential of the Rouquier complex is the sum, over the letters, of the maps φi from the cohomological degree 0 term BS(x‾) to the degree 1 term ⨁iBS(x‾^i)(1), and its dual maps decompose the generator csi back, i.e. are the maps χi with the displayed internal shifts. Composing the differential with the rescaled dual maps γiχi and summing over i gives ∑iγiBri, which by step 2.1 equals left multiplication by ρ modulo the right-multiplication remainder that vanishes on the reduced module by step 5.1. Hence left multiplication by ρ on the reduced Bott-Samelson module factors, up to the positive rescalings γi, as the composite BS(x‾)→φ⨁iBS(x‾^i)(1)→ γχ BS(x‾)(2) built from the first Rouquier differential and its dual; all first-differential Koszul signs are +1, because every preceding tensor factor is in cohomological degree zero, so this is exactly ρ⋅(−) after right reduction. This proves (d) and completes the lemma; no positivity of characters and no character conjecture is used.

Remarks

  • The remainder term is genuinely present before reduction: right multiplication by x−1ρ does not vanish on the raw Bott-Samelson bimodule, and the formula (2) is stated on the reduced module, where it does.
  • The rescalings γi are positive precisely for reduced x‾; for nonreduced words the factorization (1) still holds as an identity of endomorphisms, but no positivity claim about the scalings is made.

Depends on

Used by

Dependency tree · two levels

20 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources