🔢 mathematics

A proof of Dolbeault geometric Langlands for $\mathrm{GL}_2$ with reduced spectral curves

This paper establishes the Dolbeault geometric Langlands correspondence for $\mathrm{GL}_2$ over the locus of reduced spectral curves by utilizing limit categories to handle the non-quasi-compact nature of the relevant moduli stacks, thereby providing a foundational step and strategic framework for proving the correspondence in greater generality.

Original authors: Yukinobu Toda

Published 2026-06-12

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Yukinobu Toda

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: A Cosmic Translation Machine

Imagine you are trying to translate a complex book written in a language of pure geometry (shapes, curves, and bundles) into a language of pure physics (waves, particles, and forces). This is the goal of the Geometric Langlands Correspondence. It's like a universal translator that claims two completely different mathematical worlds are actually the same thing, just looking at it from different angles.

For a long time, mathematicians could only prove this translation worked in "perfect" conditions—where the shapes involved were smooth, unbroken, and tidy. But in the real world of math, things get messy. Curves can break, split, or have sharp corners.

This paper is a major breakthrough because it proves the translation works even when the shapes are messy (specifically, when they are "reduced" but not necessarily smooth). It's like proving your universal translator still works even if the book has torn pages or ink blots, as long as the words are still readable.

The Main Characters

To understand the paper, we need to meet three key players:

The Higgs Bundle (The Shape): Think of this as a complex, multi-layered geometric object. In this paper, the author focuses on a specific type called GL2, which is like a two-dimensional shape with some extra "twist" attached to it.
The Spectral Curve (The Map): Every Higgs bundle has a hidden "map" or "shadow" called a spectral curve.
- Smooth Curves: These are like a single, unbroken rubber band. Previous proofs only worked here.
- Reduced Curves: These are like a rubber band that has snapped into two pieces but is still held together at the knots. They are "reduced" (the pieces are distinct) but not "irreducible" (they aren't one single piece). This is the "messy" territory this paper explores.
The Limit Category (The Safety Net): When shapes get messy (like having infinitely many broken pieces), standard mathematical tools break down. They become too big or too chaotic to handle. The author uses a special tool called a "Limit Category."
- Analogy: Imagine trying to count grains of sand on a beach. If you try to count them one by one, you'll go crazy. But if you use a "limit" approach—grouping them into buckets and counting the buckets—you can handle the infinity. The Limit Category is this "bucket system" that allows mathematicians to organize the chaos of broken shapes.

The Problem: The "Non-Compact" Nightmare

In the "perfect" world (smooth curves), the collection of all possible shapes is compact. Think of it like a closed box; everything fits inside, and you can easily check every item.

However, when the curves are "reduced" (snapped into pieces), the collection of shapes becomes non-compact.

Analogy: Imagine a box that has a hole in the bottom. As you try to put shapes inside, they keep falling out into an infinite abyss. There are infinitely many ways a shape can break, so you can't just "count" them all. Standard math tools fail here because the "box" is too big and open.

The Solution: The "Whittaker Normalization"

The author proves the translation works by using a clever strategy involving a "Whittaker Normalization."

The Metaphor: Imagine you have two different maps of the same territory. One map is drawn by a cartographer who only knows smooth roads. The other is drawn by a cartographer who knows about broken bridges and dirt paths.
To prove the maps match, you need a Reference Point. The author uses a specific, simple shape (the "Hitchin section") as a reference point.
They show that if you take a simple, standard shape and apply a specific "filter" (the Arinkin Sheaf), it transforms perfectly into the reference point on the other side.
The "Arinkin Sheaf": Think of this as a special lens or a magical prism. If you shine a light (a mathematical object) through it, it refracts the light perfectly, turning a messy, broken shape into a clean, organized one. The paper proves this prism works even when the input shape is broken.

How the Proof Works (Step-by-Step)

Building the Prism: The author constructs a specific mathematical tool (the Arinkin sheaf) that acts as a bridge between the "messy" world of broken curves and the "clean" world of the translation.
Testing the Translation: They check if this bridge respects the "rules of the game." In this math world, there are special operators called Wilson and Hecke operators.
- Analogy: These are like "checkpoints" or "quality control tests." If you translate a word, does it still pass the grammar test? The paper proves that if you translate a shape using the Arinkin prism, it still passes all the grammar tests (Wilson/Hecke compatibility).
The "Safety Net" (Limit Categories): Because the shapes are broken and infinite, the author uses the "Limit Category" (the bucket system) to ensure the translation doesn't fall into the abyss. This is the first time this specific tool has been used to solve this problem outside of the "perfect" world.
The Final Check (Whittaker Normalization): The author proves that the simplest shape (the "vacuum" state) translates exactly to the reference point. Since the simplest shape translates correctly, and the translation rules (operators) are consistent, the entire system must be correct.

The Result

The paper successfully proves that the Dolbeault Geometric Langlands Correspondence holds true for GL2 (two-dimensional shapes) even when the underlying curves are reduced (broken into pieces).

Why it matters: This is the first time this correspondence has been proven in a situation where the shapes are not "compact" (where they don't fit in a neat box). It shows that the "Limit Category" is the right tool to handle the infinite complexity of broken geometric shapes.
The Limitation: The proof currently works specifically for GL2 (two-dimensional shapes). The author hints that this strategy could be used for larger, more complex shapes (GL3, GL4, etc.), but that is a job for future papers.

In a Nutshell

This paper is like a master carpenter proving that a specific type of joint (the Langlands correspondence) holds strong even when the wood is cracked and splintered (reduced spectral curves). They didn't just patch the wood; they invented a new way of measuring the wood (Limit Categories) and a special glue (Arinkin Sheaf) that ensures the structure remains solid, even when it's falling apart.

Technical Summary: A Proof of Dolbeault Geometric Langlands for $GL_2$ with Reduced Spectral Curves

Problem Statement
The paper addresses the Dolbeault geometric Langlands correspondence, a conjectured equivalence between the derived category of D-modules on the moduli stack of $G$ -bundles and the derived category of coherent sheaves on the moduli stack of $G$ -Higgs bundles. Specifically, it focuses on the "classical limit" formulation proposed by the author and Tudor P˘adurariu [PTb], which involves categories of coherent sheaves with nilpotent singular support.

Prior to this work, the correspondence was established over the elliptic locus ( $B_{ell}$ ) of the Hitchin base, where spectral curves are integral (irreducible and reduced). In this regime, the moduli stacks of Higgs bundles are quasi-compact, and the correspondence follows from classical Fourier–Mukai transforms on dual abelian fibrations. However, beyond the elliptic locus—specifically over the locus of reduced spectral curves ( $B_{red}$ )—the moduli stacks are no longer quasi-compact due to the presence of infinitely many Harder–Narasimhan strata. In this non-quasi-compact setting, standard categories of (quasi- or ind-)coherent sheaves are ill-behaved (e.g., failing to be compactly generated), rendering the naive formulation of the correspondence inapplicable.

Methodology
The author employs the framework of limit categories, introduced in [PTb], which are designed to provide a well-behaved "classical limit" for moduli stacks that are not quasi-compact. The proof strategy for $G=GL_2$ over $B_{red}$ proceeds through the following key technical steps:

Construction of the Functor: A Fourier–Mukai functor $\Phi_P$ is constructed using the Arinkin sheaf (a Cohen–Macaulay extension of the Poincaré bundle) as the kernel. This functor maps the category of semistable Higgs bundles to the category of generically regular Higgs bundles.
Compatibility with Operators: The paper proves that $\Phi_P$ intertwines Wilson operators (defined via the universal Higgs bundle) and Hecke operators (defined via Hecke correspondences). This compatibility is established by utilizing an explicit description of the Arinkin sheaf via Haiman's isospectral Hilbert scheme [Hai01] and reducing the problem to the case of integral curves where the Hecke-eigen property is known [Gro13].
Whittaker Normalization: A crucial step involves identifying the image of the structure sheaf of the semistable locus, $\mathcal{O}_{Higgs^{ss}}$ , under the functor $\Phi$ . The paper conjectures (and proves for $GL_2$ ) that this image corresponds to $s!\mathcal{O}_B$ , where $s$ is the Hitchin section and $s!$ is a left adjoint to the pullback $s^*$ . The existence of $s!$ is a specific feature of the limit category framework, as such an adjoint does not exist in the standard setting of quasi-coherent sheaves on non-proper stacks.
Compact Generation and Conservativity: The proof relies on showing that the limit categories are compactly generated by iterated applications of Wilson operators (on the Higgs side) and Hecke operators (on the spectral side). The author establishes the conservativity of the right adjoints of these operators. For the Hecke operators, this requires a specific assumption on the singularities of the spectral curves (Assumption 4.18), which is automatically satisfied for $GL_2$ over reduced curves.
Resolution of the Arinkin Sheaf: To prove the Whittaker normalization for $GL_2$ , the author constructs an explicit resolution of the Arinkin sheaf by vector bundles. This involves analyzing the behavior of the sheaf over reducible spectral curves (nodal or with higher singularities) and using perturbation of stability conditions to reduce the problem to the case where semistability coincides with stability, allowing the application of results from [MRV19] on compactified Jacobians.

Key Contributions and Results

Theorem 1.2: The paper proves the Dolbeault geometric Langlands correspondence for $G=GL_2$ over the locus $B_{red} \subset B$ of reduced spectral curves. This is the first non-trivial case where the correspondence is established beyond the elliptic locus, specifically in a setting where the relevant moduli stacks are not quasi-compact.
Limit Categories as Essential Framework: The work demonstrates that the notion of limit categories is not merely a technical fix but is essential for both the formulation and the proof of the correspondence in non-compact settings. It allows for the definition of the left adjoint $s!$ and ensures the categories are compactly generated.
Wilson/Hecke Compatibility: Theorem 3.3 establishes the compatibility of the Fourier–Mukai transform with Wilson and Hecke operators over the reduced locus, a result that relies on the explicit construction of the Arinkin sheaf via isospectral Hilbert schemes.
Whittaker Normalization for $GL_2$ : The paper provides a proof of the Whittaker normalization conjecture (Conjecture 4.4) specifically for $GL_2$ (Theorem 1.7), showing that the Fourier–Mukai transform maps the structure sheaf to the object $s!\mathcal{O}_B$ .

Significance
The paper claims significance as the first proof of the Dolbeault geometric Langlands correspondence in a regime where the moduli stacks are not quasi-compact. It validates the "limit category" approach proposed in [PTb] as the correct framework for extending geometric Langlands beyond the elliptic locus. The results outline a strategy for proving the correspondence for general $GL_r$ and identify the remaining obstructions, primarily the Whittaker normalization and the generation of categories by Hecke operators in more general singular settings. The work also highlights the deep connection between categorical Donaldson–Thomas theory and the geometric Langlands correspondence.

The author notes that while the proof for $GL_2$ relies on specific features of the rank-two case (particularly the structure of singularities and the resolution of the Arinkin sheaf), the outlined strategy provides a roadmap for the general case. The paper does not claim to solve the correspondence for all groups or all spectral curves, but rather establishes a foundational non-trivial case that confirms the viability of the limit category formulation.