🔢 mathematics

$\mathrm{D}-\mathrm{mod}(\mathrm{Bun}_G^\mathrm{I})$ is Compactly Generated

This paper establishes that the category of D-modules on the algebraic stack of principal $G$ -bundles equipped with Iwahori level structure at one or more points on a curve is compactly generated, extending the foundational result of Drinfeld and Gaitsgory for the case without level structure.

Original authors: Taeuk Nam

Published 2026-08-12

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Taeuk Nam

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

Imagine you are trying to organize a massive, chaotic library where the books are constantly rearranging themselves, and the shelves are stretching out infinitely in every direction. In the world of modern mathematics, specifically a field called algebraic geometry, researchers study "stacks," which are like these super-complex, shape-shifting libraries that classify all possible ways to bundle things together (like wrapping a ribbon around a sphere in every conceivable way). To make sense of these infinite, messy structures, mathematicians use a tool called "D-modules," which acts like a high-powered flashlight, illuminating the hidden patterns and rules within the chaos.

The big question mathematicians have been asking is: Can we break these infinite libraries down into a manageable set of "building blocks"? If we can find a finite collection of special, compact objects that can be combined to build any other object in the library, we say the library is "compactly generated." This is a crucial property because it turns an impossible, infinite puzzle into a solvable one. It allows mathematicians to understand the whole library just by studying the small, well-behaved pieces. This concept is the backbone of a massive, ongoing theory called the Geometric Langlands Program, which attempts to connect two completely different worlds of mathematics: the geometry of shapes and the algebra of equations. If this connection holds, it could unlock deep secrets about the fundamental structure of the universe, much like finding a universal translator between two alien languages.

Now, enter a specific type of library called $\text{Bun}_G$ , which organizes bundles on a curve. A few years ago, mathematicians Drinfeld and Gaitsgory proved that the standard version of this library (where the bundles are smooth and unbroken) is indeed compactly generated. But what happens if we introduce a "kink" or a specific twist at a single point on the curve? This is called "Iwahori level structure." It's like taking that smooth ribbon and pinning it down tightly at one specific spot, forcing it to behave in a more rigid, complicated way. This creates a new, more twisted library called $\text{Bun}_I^G$ .

The paper you are reading, written by Taeuk Nam, tackles the question of whether this new, twisted library is also compactly generated. The author proves that, yes, it is. Even with the added complexity of the "pin" at the point, the library can still be broken down into a manageable set of building blocks. The proof is a bit like showing that even if you add a complex knot to a rope, you can still untangle the whole thing by looking at specific, smaller sections of the rope that are easy to handle.

To do this, the author uses a clever strategy involving "stratification." Imagine the library isn't just one big room, but a series of nested rooms, some small and tidy, others huge and wild. The author shows that if you look at the "wild" parts of the library (the parts that stretch out infinitely), they have a special property: they are "contractive." Think of this like a magical funnel or a black hole that pulls everything in that direction toward a single, manageable point. Because these wild sections can be "contracted" or pulled back into a safe, finite zone, the entire infinite library can be tamed.

The paper doesn't just stop at one pin. It also shows that this logic holds even if you pin the ribbon down at multiple points along the curve. The author proves that no matter how many points you choose to pin the bundle, as long as you follow the right mathematical rules, the library remains compactly generated. This is a significant step forward because it confirms that the tools used to understand the smooth, unramified world can be successfully extended to the more complex, "ramified" world where these special pins exist. It's a proof that the universe of these mathematical bundles is more orderly than it first appears, even when you add the most complicated knots imaginable.

Technical Summary: $D\text{-mod}(\text{Bun}^I_G)$ is Compactly Generated

Problem Statement
The paper addresses the question of whether the DG category of D-modules, denoted $D\text{-mod}(\text{Bun}^I_G)$ , is compactly generated. Here, $\text{Bun}^I_G$ is the algebraic stack of principal $G$ -bundles on a smooth projective curve $X$ equipped with an Iwahori level structure at a fixed point $x \in X$ .

This problem arises as a natural generalization of the unramified geometric Langlands program. In the unramified case, Drinfeld and Gaitsgory proved in [DG15] that $D\text{-mod}(\text{Bun}_G)$ is compactly generated. This property is a necessary condition for the geometric Langlands conjecture, which posits an equivalence between $D\text{-mod}(\text{Bun}_G)$ and the category of ind-coherent sheaves on the stack of local systems. While the unramified case was settled, the ramified case (specifically with Iwahori level structure, representing tame ramification) remained an open question regarding compact generation.

Methodology
The paper employs the framework established by Drinfeld and Gaitsgory for proving compact generation of D-modules on non-quasicompact stacks. The core strategy involves demonstrating that the stack $\text{Bun}^I_G$ is truncatable. A stack is truncatable if it can be covered by a cofinal system of quasicompact open substacks that are co-truncative.

The methodology proceeds through the following steps:

Reduction to Co-truncativity: Since $\text{Bun}^I_G$ is locally QCA (quasicompact with affine automorphism groups) but not quasicompact, the author aims to express it as a union of quasicompact open substacks $\text{Bun}^I_{G, (\leq \theta)}$ . The goal is to show that for sufficiently large $\theta$ , the complement of these substacks is truncative.
Stratification: The author constructs a stratification of $\text{Bun}^I_G$ $Bun_{G}^{I}$ analogous to the Harder-Narasimhan stratification used in the unramified case. This involves:
- Stratifying by the Harder-Narasimhan type $\lambda$ of the underlying principal $G$ -bundle.
- Further stratifying by the relative position $w$ between the Iwahori level structure (a $B$ -reduction) and the parabolic reduction ( $P$ -reduction) associated with the stratum.
Contractiveness Criterion: To prove that the complements of the open substacks are truncative, the author utilizes the "contractive substack" criterion. A substack is contractive if it admits an $\mathbb{A}^1$ -action that contracts the substack to a section, satisfying specific geometric conditions (smoothness, surjectivity, and open embeddings in a specific diagram).
Global Generation Estimates: A critical technical component involves proving that certain associated vector bundles are globally generated. Specifically, for a Levi subgroup $M$ and a parabolic $P$ , the author shows that the bundle $\mathfrak{n}(P)F_M$ (associated with the nilradical of $P$ ) is globally generated under specific numerical bounds on the degree of the bundle. This relies on vanishing theorems for $H^1$ of twisted bundles.

Key Contributions and Results

Main Theorem: The paper proves that $D\text{-mod}(\text{Bun}^I_G)$ is compactly generated (Theorem 0.2.3).
Generalization: The result is extended to the case of Iwahori level structures at a finite collection of points $x_1, \dots, x_k \in X$ , proving that $D\text{-mod}(\text{Bun}^{(I; x_1, \dots, x_k)}_G)$ is also compactly generated (Theorem 0.2.5).
Refinement of Bounds: A significant technical contribution is the adjustment of the numerical bound required for the stratification to be co-truncative.
- In the unramified case [DG15], the bound for the pairing $\langle \theta, \check{\alpha}_i \rangle$ was $2g - 2$ .
- In the single-point Iwahori case, the author proves the bound must be raised to $2g - 1$ .
- In the $k$ -point case, the bound is shown to be $2g - 2 + k$ .
  This shift is necessitated by the requirement for $k$ -global generation of the relevant vector bundles to ensure the contractiveness of the stratification complements.
Toy Model: Section 2 provides a detailed, self-contained proof for the case $G = SL_2$ , illustrating the conceptual machinery (fiber squares, surjectivity of maps, and contractive actions) without the full combinatorial complexity of the general case.

Significance
The paper claims that establishing the compact generation of $D\text{-mod}(\text{Bun}^I_G)$ is a necessary step toward formulating and proving the Iwahori ramified geometric Langlands conjecture.

The significance is twofold:

Prerequisite for Equivalence: As noted in the introduction, the right-hand side of the proposed Langlands equivalence (ind-coherent sheaves on the stack of local systems) is a priori compactly generated. Therefore, the compact generation of the left-hand side ( $D\text{-mod}(\text{Bun}^I_G)$ ) is a necessary condition for such an equivalence to exist.
Methodological Extension: The work demonstrates that the combinatorial and geometric techniques developed by Drinfeld and Gaitsgory for the unramified case can be successfully adapted to the ramified setting. The author notes that while an unpublished sketch by Gaitsgory and Arkhipov [AG09] existed, it contained combinatorial deficiencies fixed in [DG15] for the unramified case; this paper provides the necessary modifications to fix the argument for the Iwahori ramified case.

The paper does not claim to prove the full Langlands equivalence for the ramified case, but rather establishes the foundational finiteness property required for such a program to proceed.

Technical Summary: D-mod(BunGI)D\text{-mod}(\text{Bun}^I_G)D-mod(BunGI​) is Compactly Generated

More like this

Technical Summary: $D\text{-mod}(\text{Bun}^I_G)$ is Compactly Generated