Can an unlabeled order determine a Kazhdan–Lusztig polynomial

← All digests

Family 168 · Combinatorics
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-11

1. The question

Some mathematical objects are built using detailed labels, yet their essential information may survive after the labels are erased. Kazhdan–Lusztig polynomials present a striking instance. Their usual definition uses generators of a group, and their geometric interpretation involves singular spaces. A conjecture asks whether a much simpler object—a finite partially ordered set—already determines each polynomial.

The manuscript in family 168 claims a positive answer in full equal-parameter generality. Whenever two Bruhat intervals have the same abstract order structure, their Kazhdan–Lusztig polynomials agree, even if they come from different Coxeter groups.

The point is not that the intervals retain hidden permutation names or geometric coordinates. The proposed theorem uses only which elements lie above which others. Understanding why this is surprising requires separating that order from the richer algebra used to define the polynomial.

2. Background and essential definitions

Coxeter groups and Bruhat intervals

A Coxeter group is generated by reflections $S$, subject to relations $s^2=1$ and specified relations $(st)^{m_{st}}=1$. An infinite value of $m_{st}$ means no finite-order relation is imposed for that pair. The group together with its chosen generators is a Coxeter system $(W,S)$.

The symmetric group of permutations is a standard example: adjacent swaps generate every permutation. The length $\ell(w)$ is the smallest number of generators needed to express $w$. Such a shortest expression is called reduced.

Bruhat order can be described by subwords: $u\le w$ if a reduced expression for $w$ contains a subword that is a reduced expression for $u$. For $u\le b$, the interval is

$$[u,b]=\{x\in W:u\le x\le b\}.$$

It is finite even when $W$ is infinite. Its rank is $\ell(b)-\ell(u)$. An abstract interval remembers the order relation but forgets the words, generators and reflection labels. A poset isomorphism is a bijection preserving the order in both directions.

Where the polynomials come from

The Hecke algebra is a deformation of the group algebra, with basis symbols $T_w$ and a parameter $q$. Products corresponding to length-additive group multiplication satisfy $T_xT_y=T_{xy}$. For a generator,

$$T_s^2=(q-1)T_s+q.$$

At $q=1$, this returns the relation $s^2=1$. There is a bar operation sending $q^{1/2}$ to $q^{-1/2}$ and $T_s$ to $T_s^{-1}$. A distinguished bar-invariant basis has expansions

$$C'_b=q^{-\ell(b)/2}\sum_{u\le b}P_{u,b}(q)T_u,$$

where $P_{b,b}=1$ and, for $u<b$,

$$\deg P_{u,b}<\frac{\ell(b)-\ell(u)}2.$$

These normalization and invariance conditions characterize the usual equal-parameter Kazhdan–Lusztig polynomials. The formal symbol $q$ records grading; it is not a probability. “Equal parameter” means the same parameter governs all generators.

In Weyl-group settings the coefficients also encode local intersection cohomology of Schubert varieties, a tool for measuring topology near singularities. The theorem can be understood without computing that cohomology: the question is whether the polynomial defined by the algebra is already determined by the smaller order structure.

3. Development history and previous partial results

Kazhdan and Lusztig introduced the polynomials in 1979, and their subsequent geometric work connected them with Schubert varieties. The combinatorial invariance conjecture is associated with Lusztig and Dyer.

Dyer showed how the interval order reconstructs the directed Bruhat graph, including reflection edges that are not just immediate order covers. This recovers more than the usual diagram of adjacent levels, but still does not recover the root labels used in geometric constructions.

A major positive case concerns principal lower intervals $[1,b]$, starting at the identity. Work by du Cloux, Brenti, Brenti–Caselli–Marietti and Delanoy established broad invariance results using special matchings and order-theoretic recursions. Delanoy’s paper treats arbitrary Coxeter systems in that setting.

The distinction is subtle. Knowing a larger interval $[1,b]$ and where $u$ sits inside it gives additional information. The full conjecture asks to recover $P_{u,b}$ from $[u,b]$ alone, whose bottom may be far from the identity.

Other progress handled short intervals, individual coefficients and special decompositions. Recent work by Barkley, Gaetz and Lam treats the coefficient of $q$ in arbitrary Coxeter systems and yields full invariance at low interval rank. Hypercube decomposition methods, including the machine-learning-guided work of Blundell and collaborators, supplied further formulas and conjectures in symmetric groups.

Moment-graph sheaves offer another route. Braden–MacPherson, Soergel, Fiebig, and Elias–Williamson developed algebraic and geometric models recovering the polynomials. Prior comparisons of labeled moment graphs could use compatible maps on root labels. The new manuscript’s task is to compare the relevant information when those label maps are absent.

4. What the new manuscripts claim

M1’s Theorem 1.1 takes arbitrary Coxeter systems $(W,S)$ and $(W',S')$. If

$$[u,b]\cong[u',b']$$

as abstract posets, it asserts

$$P^W_{u,b}(q)=P^{W'}_{u',b'}(q).$$

The systems may be infinite or noncrystallographic, meaning they need not arise from the integral root systems familiar in Lie theory. No preservation of generator labels or root coordinates is required.

Restricting the poset isomorphism to every subinterval gives equality of all corresponding polynomials inside the interval. The source also recovers the associated $R$-polynomials through standard reciprocity relations. These are related Hecke-algebra polynomials entering the inversion formulas behind the canonical basis.

The claim is specifically for equal parameters. It does not state an unequal-parameter version, where different generators carry different weights. It is a determination theorem: identical order data imply identical polynomials. It does not by itself promise a fast algorithm using only the diagram of the interval.

5. Why the advance matters

Importance: 5/5. Full combinatorial invariance would settle a central question connecting algebraic combinatorics, representation theory and geometry. It would identify precisely how much information can be discarded while preserving these important polynomials.

The result’s breadth matters. A statement only for symmetric groups, only for intervals starting at the identity, or only for a few coefficients would leave substantial parts of the conjecture unresolved. The manuscript claims all coefficients for arbitrary intervals across arbitrary Coxeter systems.

The theorem would also clarify a conceptual puzzle: root labels are useful in constructing the invariant, yet the final answer would depend only on order. A successful argument must explain that independence without assuming the labels can somehow be reconstructed unchanged.

6. The main idea

Keep the two label systems separate

A moment-graph sheaf assigns algebraic modules to vertices and edges, with maps recording compatibility across edges. A graded module is organized by degree, and its graded size can be summarized by a polynomial or series. In the relevant sheaf, the module at a vertex records a Kazhdan–Lusztig polynomial.

The source studies vectors whose restrictions vanish along selected upward edges. With no edge conditions, one has the full vertex module. With all upward edges imposed, a known duality formula describes the opposite extreme. Intermediate choices are harder because their modules must also have the necessary freeness and graded behavior.

Choose a reflection order in the second Coxeter system and transport its edge schedule across the poset isomorphism. It need not become a reflection order for the first system. The argument therefore retains both realizations rather than pretending there is one common set of root labels.

Compare one edge at a time

At an edge from $x$ to $y$, impose the conditions coming from later edges and compare the images from its two endpoints. Localizing at a plane spanned by root labels reduces the comparison to dihedral, or two-reflection, models.

The order structure controls the relevant relative dihedral intervals and the counts of later edges. The source obtains an inclusion between graded images with a precisely determined degree shift. This gives an inequality on their Hilbert series, which count dimensions in each degree.

Removing edge conditions in the transported schedule accumulates these inequalities. Dyer’s reflection-path formula identifies the resulting product with an $R$-polynomial matrix in the other system.

Reciprocal comparisons force equality

The argument applies the comparison in both directions and uses the reciprocity relations linking $R$- and Kazhdan–Lusztig polynomials. Because the intermediate terms are nonnegative in the relevant range, equality of the global quantities forces the individual inequalities to be sharp.

This also establishes the freeness needed for the next step of the induction. Polynomial equality and module structure are proved together, in increasing interval rank. The mechanism therefore avoids assuming at the start the intermediate properties it needs later.

The broad idea is that local order-controlled inequalities, constrained by a two-way algebraic symmetry, leave no room for the richer labels to change the final polynomial.

7. Scope, limitations, and open questions

The theorem concerns the usual equal-parameter polynomials. Weighted Hecke algebras, other canonical bases and characteristic-dependent analogues are not included merely because their notation looks similar.

Combinatorial invariance of these polynomials does not mean the entire geometric spaces associated with two intervals are identical. A shared numerical invariant may capture only part of their topology or singularity structure.

The source supplies a structural proof through sheaves, localization and reciprocity. Finding a short direct combinatorial formula, or an efficient label-free computation for arbitrary intervals, remains a different goal. Determination by a poset is weaker than a practical procedure with a controlled running time.

The claim is broad enough to compare intervals from different Coxeter systems, but it still requires an actual order isomorphism. Matching only rank sizes or the total number of interval elements would discard more information than the theorem allows.

8. Sources and further reading

M1’s Theorem 1.1 states full invariance; Corollary 6.1 records the associated consequences. Sections 2–3 introduce the realizations, sheaves and transported orders. Sections 4–6 give the local comparison and simultaneous induction.

The original and partial-result papers linked in Section 3 trace the problem’s development. Wikipedia’s Coxeter group and Kazhdan–Lusztig polynomial pages provide optional examples; the algebra and order conventions needed for the claim are stated above.

Source manuscripts

• M1. OpenAI, Combinatorial invariance of Kazhdan–Lusztig polynomials (September 24, 2026).

Sources consulted on 11 October 2026 at repository revision fd4aeeb. Download the manuscript citations.