Intersection Cohomology Without Spaces

Basic Information

• Paper ID: 2510.09488
• Title: Intersection cohomology without spaces
• Authors: Tom Braden (University of Massachusetts), Nicholas Proudfoot (University of Oregon)
• Classification: math.AG (Algebraic Geometry), math.CO (Combinatorics), math.RT (Representation Theory)
• Publication Date: October 10, 2025 (arXiv preprint)
• Paper Link: https://arxiv.org/abs/2510.09488

Abstract

This paper surveys three important settings in which the dimensions of intersection cohomology groups of algebraic varieties provide profound combinatorial and representation-theoretic information, and these groups themselves can be computed via combinatorial sheaves on finite posets. The three settings are: (1) intersection cohomology of Schubert varieties, associated Kazhdan-Lusztig polynomials and their realization through matrix graph sheaves and Soergel bimodules; (2) intersection cohomology of toric varieties, associated g-polynomials of convex polytopes and their realization through intersection cohomology theory of fans; (3) intersection cohomology of arrangement Schubert varieties, associated Kazhdan-Lusztig polynomials of matroids and their realization through intersection cohomology of matroids. In all three settings, these constructions remain valid in more general cases where the varieties do not exist, leading to "intersection cohomology without spaces."

Research Background and Motivation

Core Problems

The core problems addressed in this paper are how to compute and understand the combinatorial and algebraic properties of intersection cohomology in the absence of geometric spaces. Specifically:

1. Unification of Geometry and Combinatorics: In special cases (such as Weyl groups, rational polytopes, realizable matroids), the corresponding KLS-polynomials can be interpreted through intersection cohomology of algebraic varieties, thereby proving the non-negativity of coefficients.
2. Generalization Challenges: For general Coxeter groups, arbitrary polytopes, and general matroids, no corresponding geometric objects exist, yet the properties of related polynomials must still be understood.
3. Unified Theoretical Framework: A unified algebraic framework is needed to handle these three seemingly different mathematical objects.

Significance

The importance of this problem is manifested in:

• Representation Theory Applications: Kazhdan-Lusztig polynomials play a key role in representation theory of Hecke algebras, Lie algebras, and algebraic groups
• Combinatorial Geometry: g-polynomials are fundamental invariants in convex polytope theory
• Matroid Theory: Kazhdan-Lusztig polynomials of matroids represent an important recent development in combinatorial mathematics

Limitations of Existing Methods

The limitations of traditional approaches lie in:

• Dependence on the existence of specific geometric structures
• Lack of unified algebraic treatment
• Difficulty in generalization to more general cases

Core Contributions

1. Unified Framework: Provides a unified framework for understanding intersection cohomology computations in three different settings
2. Algebraic Construction: Gives purely algebraic methods for constructing intersection cohomology without geometric spaces
3. Parallel Theory: Demonstrates deep parallels between theories of Schubert varieties, toric varieties, and arrangement Schubert varieties
4. Extended Applications: Extends applications of intersection cohomology to broader combinatorial and algebraic objects
5. Positive Characteristic Generalization: Discusses parity sheaf theory in positive characteristic and its applications

Detailed Methods

Task Definition

The core task of this paper is to construct graded vector spaces for a given finite-rank poset P and P-kernel κ such that their Poincaré polynomials equal the corresponding KLS-polynomials f_(t).

Theoretical Framework

Definition of KLS-Polynomials

Given a poset P with a strictly increasing rank function rk: P → ℤ, a P-kernel is a collection of polynomials {κ_(t) | x ≤ y} satisfying:

• κ_(t) = 1 for all x ∈ P
• deg κ_(t) ≤ r_ for all x ≤ y ∈ P
• ∑{x≤y≤z} t^{r} κ_(t^{-1})κ_(t) = 0 for all x < z ∈ P

The KLS-polynomials {f_(t)} are uniquely determined by:

• f_(t) = 1 for all x ∈ P
• deg f_(t) < r_/2 for all x < y ∈ P
• t^{r_} f_(t^{-1}) = ∑{x≤y≤z} κ(t)f_(t) for all x ≤ z ∈ P

From Topology to Sheaves on Posets

Torus Equivariance and Localization: The key idea is to exploit the action of a torus T so that equivariant intersection cohomology classes can be localized. For Schubert varieties or arrangement Schubert varieties, there is an injective map:

IH_T(X) ↪ ⊕_{p∈X^T} IH_T(X)_p

Equivariant Intersection Cohomology as Sheaves: Given a T-invariant stratification {S_x | x ∈ P}, define a presheaf:

Q ↦ F(Q) := IH_T(U_Q)

where U_Q = ⋃_{x∈Q} S_x.

Two key properties make this construction work:

1. Sheaf Property: F is actually a sheaf
2. Surjectivity: The restriction map IH_T(U_Q) → IH_T(U_{Q'}) is surjective

Local Computation

By results of Bernstein-Lunts, for an affine neighborhood N and a contraction subgroup C* ⊂ T, there is a short exact sequence:

0 → IH_T(N,N°) → IH_T(N) → IH_T(N°) → 0

where:

• IH_T(N) is a free R-module with generators in degrees 0 ≤ d < dim X
• IH_T(N,N°) is a free R-module with generators in degrees dim X < d ≤ 2 dim X
• IH_T(N) is the minimal free R-module surjecting onto IH_T(N°)

Three Concrete Realizations

1. Coxeter Groups and Soergel Bimodules

• Matrix Graph Method: Using matrix graph sheaves of Bruhat graphs for computation
• Edge Modules: M_E ≅ M_v/α_E M_v
• Boundary Modules: M^∂v ≅ Im(M(Γ{>v}) → ⊕_E M_E)

2. Polytopes and Fans

• Rational Fan Case: Direct use of equivariant intersection cohomology of toric varieties
• General Fan Case: Definition via combinatorial methods, boundary modules M^∂_τ = F(Σ°_τ)
• Karu's Theorem: Proves a combinatorial version of the hard Lefschetz theorem

3. Matroids and Arrangement Schubert Varieties

• Realizable Case: Using intersection cohomology of arrangement Schubert varieties Y(L)
• General Case: Via algebraic construction, boundary modules M^∂_F := F(L°F) ⊗{A(L_F)} R
• Palindromic Property: Z-polynomials satisfy t^{rk_E} Z_L(t^{-1}) = Z_L(t)

Experimental Setup

Theoretical Verification

This paper is primarily a theoretical survey, with correctness of the theory verified through:

1. Concrete Examples:
   • The case of CP¹ (Example 7.1)
   • Arrangement Schubert varieties of 4-dimensional subspaces (Example 7.2)
   • Combinatorial intersection cohomology of simple fans (Examples 5.1, 5.2)
2. Comparison with Known Results:
   • Consistency with the original Kazhdan-Lusztig definition
   • Isomorphism with intersection cohomology in the geometric case
   • Conformity with known polynomial properties

Experimental Results

Main Theoretical Results

Non-negativity Proofs

In all three settings, non-negativity of KLS-polynomial coefficients is successfully established:

• Coxeter Groups: Proved by Elias-Williamson via Soergel bimodules
• Polytopes: Proved by Karu via intersection cohomology of fans
• Matroids: Proved by Braden et al. via matroid intersection cohomology

Inequality Results

1. Monotonicity: f_(t) ⪯ f_(t) when x ≤ y ≤ z
2. Kalai Inequality (polytope case): f_{σ0}(t) ⪰ f_{στ}(t)f_{τ0}(t)
3. Top-Heaviness: h_j ≤ h_k when j ≤ k ≤ d-j

Positive Characteristic Results

• Parity Sheaf Theory: Corresponding polynomials can still be defined in positive characteristic
• Modularity Criterion: Provides necessary and sufficient conditions for p-KLS polynomials to be trivial
• Representation Theory Applications: Important applications in modular representation theory

Related Work

Historical Development

1. Kazhdan-Lusztig (1979): Introduced Kazhdan-Lusztig polynomials
2. Stanley (1992): Generalized to general KLS-polynomials
3. Soergel (1990s): Developed bimodule theory
4. Karu (2004): Proved hard Lefschetz theorem for non-rational polytopes
5. Braden et al. (2020): Established intersection cohomology theory of matroids

Related Theories

• Decomposition Theorem: Foundational work of Beilinson-Bernstein-Deligne
• Equivariant Localization: Matrix graph theory of Goresky-Kottwitz-MacPherson
• Hodge Theory: Hard Lefschetz theorem and Hodge-Riemann bilinear relations

Conclusions and Discussion

Main Conclusions

1. Unification: Three seemingly different theories actually possess deep unified structure
2. Algebraization: Geometric intersection cohomology can be completely algebraized without dependence on specific spaces
3. Generalizability: Methods can be extended to more general combinatorial objects
4. Applied Value: Important applications in representation theory, combinatorial geometry, and matroid theory

Limitations

1. Computational Complexity: Although algorithms are provided, practical computation remains difficult
2. Hard Lefschetz Proof: The combinatorial version of the proof remains complex and technical
3. Incomplete Unification: The three theories still have differences in certain details

Future Directions

1. More General CW Spheres: Generalization to regular CW spheres that are not boundaries of convex polytopes
2. Combinatorial Invariance Conjecture: Whether Kazhdan-Lusztig polynomials depend only on abstract posets
3. Real Root Properties: Real root conjecture for matroid Kazhdan-Lusztig polynomials
4. Positive Characteristic Theory: Further development of theory in positive characteristic

In-Depth Evaluation

Strengths

1. Theoretical Depth: Unifies three important mathematical fields under one framework
2. Methodological Innovation: Develops the new concept of "intersection cohomology without spaces"
3. Broad Applications: Important applications across multiple mathematical branches
4. Clear Exposition: Parallel presentation of complex theories facilitates comparison and understanding

Weaknesses

1. High Technical Threshold: Requires deep background in algebraic geometry, representation theory, and combinatorics
2. Computational Practicality: Although theoretically complete, practical computation remains highly challenging
3. Missing Proofs: Some key results' proofs are too technical and can only be outlined in the paper

Impact

1. Academic Value: Provides unified perspective for three important mathematical fields
2. Theoretical Contribution: Advances development of intersection cohomology theory
3. Application Prospects: Broad application prospects in representation theory and combinatorial mathematics
4. Methodological Significance: Demonstrates how to completely algebraize geometric concepts

Applicable Audiences

• Mathematicians researching Kazhdan-Lusztig polynomials
• Researchers working on Hecke algebra and quantum group representation theory
• Experts in convex geometry and polytope theory
• Researchers in matroid theory and combinatorial geometry
• Scholars interested in applications of algebraic geometry to combinatorial mathematics

This is a high-quality survey paper that successfully unifies three important mathematical theories under the framework of "intersection cohomology without spaces," providing new perspectives and methods for research in related fields. Although highly technical, its theoretical value and application prospects are both very significant.