Intersection cohomology groups of instanton moduli spaces and cotangent bundles of affine flag varieties

This paper presents a conjectural characterization of the equivariant costalk of the intersection cohomology complex of the Coulomb branch of a quiver gauge theory at a torus fixed point in terms of a geometric Satake correspondence for Kac-Moody settings, with a proof sketch provided for the affine type A case.

The Gist

The Big Picture: Two Different Maps to the Same Treasure

Imagine you are trying to describe a complex, beautiful landscape. You have two different maps:

1. Map A is drawn by looking at the landscape from the ground up (geometry and physics).
2. Map B is drawn by looking at the landscape from a high-level, abstract bird's-eye view (algebra and representation theory).

For a long time, mathematicians knew these two maps described the same territory, but the connection was a bit fuzzy. This paper, by Hiraku Nakajima (based on work with Dinakar Muthiah), is about sharpening the connection between these two maps, specifically for a very complex type of landscape called an "Affine Flag Variety" and its cousins.

The author is essentially saying: "We know these two maps are related. Now, let's prove exactly how they match up, even in the most complicated, infinite-dimensional versions of these landscapes."

---

Part 1: The Original Connection (The "Ground" vs. The "Sky")

The paper starts by recalling a famous result from 2004 (by Arkhipov, Bezrukavnikov, and Ginzburg).

• The Ground (Geometry): Imagine a bundle of strings hanging from a pole. This represents the "cotangent bundle of a flag variety." It's a physical, geometric space where you can count sections (like counting how many ways you can tie a knot).
• The Sky (Topology): Imagine an infinite, swirling cloud of points called the "Affine Grassmannian." This is a massive, abstract space. Inside it, there are specific "islands" (called Schubert varieties).

The Discovery: The 2004 result showed that if you count the knots on the ground (Map A), you get the exact same numbers as if you count the holes and shapes in the islands in the sky (Map B). It's like saying, "The number of ways to arrange books on a shelf is exactly the same as the number of ways to arrange stars in a specific galaxy."

Part 2: The Physics Twist (Singular Monopoles)

The paper then introduces a "physics" perspective to make this more concrete.

• The Analogy: Imagine a magnetic monopole (a particle with only a North pole, no South pole) floating in 3D space.
• The Twist: Usually, these particles are smooth. But here, the author considers "singular" monopoles—particles that have a tiny, sharp "kink" or "singularity" at the center, like a needle point.
• The Connection: The author explains that the "islands" in the sky (from Part 1) are actually the same as the "moduli space" (the collection of all possible shapes) of these singular magnetic particles.
  • If you change the "kink" in the particle, you move to a different island in the sky.
  • This bridges the gap between abstract math and the physics of magnetic fields.

Part 3: The "Coulomb Branch" (The Machine That Builds the Map)

The paper introduces a modern tool called the Coulomb Branch. Think of this as a 3D printing machine.

• How it works: You feed the machine a set of instructions (a "quiver," which is just a diagram of dots and arrows representing a gauge theory).
• The Output: The machine prints out a geometric shape.
• The Result: The author shows that if you feed the right instructions into this machine, it prints out the exact same "islands" (singular monopole spaces) we discussed earlier. This is a powerful way to generate these complex shapes using algebraic rules.

Part 4: The New Challenge (Infinite Dimensions)

So far, everything works for "finite" groups (like standard rotations in 3D space). But the author wants to go further into Kac-Moody Lie algebras.

• The Problem: Think of finite groups as a finite Lego set. Kac-Moody groups are like an infinite Lego set. The rules get much more complicated, and the "islands" in the sky become harder to define.
• The Proposal: The author and his collaborators proposed a new version of the "Geometric Satake Correspondence" (the rule that links the ground map to the sky map) for these infinite sets. They suggested that even in this infinite world, the "Coulomb Branch" machine still prints the correct shapes, and the math still holds up.

Part 5: The Current Work (The "Proof in Progress")

The final section of the paper is where the author is currently working with his colleague. They are trying to prove a very specific, delicate detail about the connection between the maps.

• The Delicate Difference: There are two slightly different ways to measure the "holes" in these shapes (mathematically called $i^!$ and $\Phi$). They are like two different rulers. They usually give the same length, but they measure slightly different things.
• The Goal: The author wants to prove that if you use the "Coulomb Branch" machine to generate the shape, and then measure it with the "Sky" ruler, it matches perfectly with the "Ground" ruler, even in the infinite case.
• The Strategy:
  1. Zoom Out: First, they prove the match works if you ignore the tiny, messy details (localization).
  2. Zoom In: Then, they check the messy details. They use a "Dynamical Weyl Group" (a symmetry tool) to show that if the match works for a simple piece (like a 2D slice), it works for the whole infinite structure.
  3. The Final Hurdle: For the most complex infinite cases (Affine Type A), they have to deal with a specific "imaginary" symmetry. They plan to solve this by relating it to a "Hilbert Scheme" (a space that counts points on a surface), which is a known, well-understood object.

Summary

In simple terms, this paper is a bridge-building project.

1. It connects Geometry (shapes of magnetic particles) with Algebra (representations of infinite groups).
2. It uses Physics (monopoles) and Machine Learning-style construction (Coulomb branches) to visualize these abstract shapes.
3. The author is currently writing the final proof to show that this bridge is solid, even when the structures become infinitely complex.

The paper doesn't claim to cure diseases or build new technology; it is purely about proving that two very different ways of looking at the mathematical universe are actually describing the same reality.

Technical Summary

Technical Summary: Intersection Cohomology Groups of Instanton Moduli Spaces and Cotangent Bundles of Affine Flag Varieties

Problem Statement
The paper addresses the geometric realization of representations of Kac-Moody Lie algebras, specifically extending the classical Geometric Satake correspondence to the affine setting. The central problem is to establish an isomorphism between the intersection cohomology of specific moduli spaces (Coulomb branches of quiver gauge theories or Uhlenbeck spaces) and the weight spaces of representations of the Langlands dual group $G$.

In the finite-dimensional case, Arkhipov-Bezrukavnikov-Ginzburg [ABG04] established an isomorphism between the global sections of line bundles on the cotangent bundle of the flag variety, $H^0(T^*(G/B), \mathcal{O}(\mu))$, and a direct sum involving the intersection cohomology of affine Grassmannian slices, $H^*(i_\mu^! IC(\text{Gr}_{G^\vee}^\lambda))$. The paper seeks to generalize this relationship to Kac-Moody groups, where the role of the affine Grassmannian is taken by Coulomb branches of 3D $\mathcal{N}=4$ supersymmetric gauge theories, and the role of the cotangent bundle is taken by resolutions of nilpotent cones in the Kac-Moody context. A specific technical hurdle identified is the distinction between the costalk of the intersection cohomology complex ($i_\mu^!$) and the hyperbolic restriction functor ($\Phi = \iota^* j_!$), which are known to yield isomorphic vector spaces in the finite case but require careful handling in the Kac-Moody setting.

Methodology
The methodology relies on the framework of Coulomb branches of 3D $\mathcal{N}=4$ gauge theories developed by Braverman-Finkelberg-Nakajima (BFN). The authors utilize the following structural components:

1. Coulomb Branches as Moduli Spaces: The authors define the Coulomb branch $M(\lambda, \mu)$ as the spectrum of the equivariant Borel-Moore homology of the variety of BFN triples (principal bundles, framings, and sections). For finite type, these are identified with slices $W_{G^\vee, \mu}^\lambda$ in the affine Grassmannian. For affine types, they are identified with Uhlenbeck partial compactifications or "bow varieties."
2. Ring Objects and Convolution: The paper treats the intersection cohomology complexes as ring objects in the derived category of equivariant perverse sheaves. The convolution product on these spaces corresponds to the tensor product in the representation category, generalizing the Peter-Weyl theorem.
3. Hyperbolic Restriction: To extract weight spaces $V(\lambda)_\mu$ from the geometry, the authors employ the hyperbolic restriction functor $\Phi = \iota^* j_!$ associated with a $\mathbb{C}^\times$-action (defined by a regular cocharacter $\rho$). This functor isolates the fixed point set, which is conjectured to be a single point $z_\mu$ if the weight space is non-zero.
4. Equivariant Cohomology and Localization: The analysis is conducted within the realm of $T^\vee$-equivariant cohomology. The authors utilize the localization theorem to analyze the behavior of maps over the localization of the polynomial ring $H^*_{T^\vee}(\text{pt})$ at coroot hyperplanes.
5. Dynamical Weyl Group: To handle the extension of isomorphisms across coroot hyperplanes (specifically real coroots), the authors employ the dynamical Weyl group introduced by Etingof-Varchenko. This allows the reduction of the problem to Levi subgroups (essentially $SL_2$).
6. Heisenberg Subgroup and Hilbert Schemes: For the affine case, the treatment of the imaginary coroot (the "null root") involves analyzing the Heisenberg subgroup and relating the fixed point sets to symmetric powers of $S_\ell = \mathbb{C}^2/(\mathbb{Z}/\ell\mathbb{Z})$ via Hilbert schemes of points.

Key Contributions and Results
The paper presents a work in progress (joint with Dinakar Muthiah) that refines the conjectural Geometric Satake correspondence for Kac-Moody Lie algebras.

• Refinement of the Geometric Satake Correspondence: The paper proposes a natural isomorphism (Equation 5.3) between the zeroth cohomology of the hyperbolic restriction of the intersection cohomology complex of the Coulomb branch, $H^0(\Phi(IC(M(\lambda, \mu))))$, and the weight space $V(\lambda)_\mu$. This formulation is valid for arbitrary weights $\mu$, unlike previous formulations restricted to dominant weights.
• Main Conjecture (Equation 6.3): The core result is a commutative diagram relating three objects:
  1. The algebraic side: $(V(\lambda) \otimes \mathbb{C}[(\mathfrak{g}/\mathfrak{u})^*] \otimes \mathbb{C}_{-\mu})^B$, representing sections of line bundles on a resolution of the nilpotent cone.
  2. The geometric side (costalk): $H^*_{T^\vee}(i_\mu^! IC_\mu^\lambda)$.
  3. The geometric side (hyperbolic restriction): $H^*_{T^\vee}(\Phi(IC_\mu^\lambda))$.
     The conjecture asserts the existence of an isomorphism $\heartsuit$ of $H^*_{T^\vee}(\text{pt})$-modules that makes this diagram commute, effectively unifying the costalk and hyperbolic restriction perspectives in the Kac-Moody setting.
• Proof Strategy for Affine Type A: The paper outlines a proof strategy for the affine type $A$ case. It demonstrates that the isomorphism holds after localization at coroots. The extension across real coroots is handled via the dynamical Weyl group and reduction to $SL_2$. The extension across the imaginary coroot is handled by relating the geometry to the Heisenberg subgroup and using the construction of Hilbert schemes on the minimal resolution of $S_\ell$.

Significance and Claims
The paper claims to provide a unified geometric framework for understanding the representation theory of Kac-Moody algebras through the topology of instanton moduli spaces and Coulomb branches.

• Unification: It bridges the gap between the finite-dimensional Geometric Satake correspondence (Arkhipov-Bezrukavnikov-Ginzburg) and the affine setting, utilizing the language of 3D gauge theories (BFN).
• Clarification of Technical Distinctions: It explicitly addresses the subtle difference between the costalk ($i_\mu^!$) and the hyperbolic restriction ($\Phi$), showing how they relate via the Ginzburg-Riche framework in the Kac-Moody context.
• Generalization: The results generalize previous checks for type $A$ (Nakajima [Nak09]) and provide a roadmap for general affine types, relying on the geometric properties of the Coulomb branches $M(\lambda, \mu)$.
• Modesty: The authors characterize the work as "in progress." While the strategy for affine type $A$ is detailed and the main conjecture is stated, the full proof for general Kac-Moody types relies on verifying specific geometric properties of the Coulomb branches, which are noted as the primary remaining difficulty. The paper does not claim to have solved the general case but rather to have established the framework and proven the strategy for the affine type $A$ case.