🔢 mathematics

The Deligne-Simpson Problem

This paper confirms the first author's conjecture regarding the Deligne-Simpson problem by proving the remaining implication that establishes the existence of matrices with no common invariant subspace and a product equal to the identity within given similarity classes, based on an associated root system.

Original authors: William Crawley-Boevey, Andrew Hubery

Published 2026-04-16

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: William Crawley-Boevey, Andrew Hubery

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 a master architect trying to build a complex machine. You have a set of specific blueprints (called conjugacy classes) that dictate exactly what kind of gears and levers you must use. Each blueprint says, "Your gear must look like this when you spin it," but it doesn't tell you exactly how big the gear is or how it connects to the others.

Your goal is to assemble these gears into a single, working engine. There are two rules for your engine:

The Loop Rule: When you turn all the gears in sequence, they must return to their starting position perfectly (mathematically, their product is the identity).
The Independence Rule: The gears must work together as a single, unified team. They cannot be split into two separate, independent sub-teams that don't talk to each other. If they can be split, the machine is "reducible" (boring and broken). If they can't be split, it's "irreducible" (a true masterpiece).

This puzzle is known as the Deligne–Simpson Problem. For years, mathematicians wondered: Given a specific set of blueprints, is it possible to build a working, unified machine?

The Big Idea: A Map to the Solution

In this paper, William Crawley-Boevey and Andrew Hubery finally solve the puzzle. They prove that you can build this machine if and only if the blueprints satisfy a very specific, hidden geometric pattern.

To understand this pattern, the authors use a clever metaphor involving Root Systems. Think of a root system as a giant, invisible crystal lattice or a map of possible shapes.

Some shapes on this map are "real" (solid, stable).
Some are "imaginary" (fuzzy, flexible).
The authors show that your blueprints must correspond to a specific point on this map (a "positive root").
Furthermore, the "energy" of your blueprints (a calculation involving the numbers in the blueprints) must equal exactly 1.
Finally, the shape of your blueprint must be "stronger" than any way you could try to break it into smaller pieces.

If your blueprints pass these three tests, a working machine exists. If they fail even one, no matter how hard you try, you cannot build a unified engine.

The Journey: From Gears to Gardens

The paper is a long journey through different mathematical landscapes to prove this. Here is how they did it, using simple analogies:

1. The Parabolic Garden (Weighted Projective Lines)
Instead of just looking at the gears (matrices), the authors imagine them as plants in a special garden.

The garden has a main path (the projective line) and some special marked spots (the singularities).
The "weights" in the problem are like different types of soil at these spots.
A "parabolic bundle" is like a plant that has a specific structure of branches and leaves at these special spots.
The "connection" is like a wind blowing through the garden. The wind must blow in a way that respects the specific shape of the branches at the marked spots.

2. The Tubular Case (The Donut Problem)
Sometimes, the garden looks like a donut (a torus). In this "tubular" case, the plants grow in neat, repeating tubes.

The authors prove that if you try to grow a plant that is essentially "two copies" of a basic tube stuck together (multiplicity $h \ge 2$ ), it will inevitably fall apart. It will split into two separate, non-interacting plants.
This proves that for these specific shapes, you can never build a unified machine if the blueprint is "too big."

3. The Extended Tubular Case (The Donut with a Tail)
What if the garden is a donut with a long tail attached?

The authors show that even with this extra tail, if you try to build a machine using a blueprint that is a "tail plus two copies of a tube," it still falls apart. The tail and the tubes just don't mix well enough to stay unified.

4. The Mirror World (Reflections)
Finally, to prove the rule works for every possible blueprint, they use a "mirror trick."

Imagine you have a blueprint that looks impossible. The authors show you can reflect it in a mirror (a mathematical operation called a "reflection functor") to turn it into a different blueprint.
They prove that if the original blueprint was impossible, the reflected one is also impossible.
By bouncing these blueprints around in the mirror, they eventually turn every "impossible" blueprint into one of the simple cases (the donut or the donut-with-tail) that they already proved were impossible.

The Conclusion

The paper confirms a conjecture made by the first author years ago. It tells us that the universe of these matrix machines is not random. It is governed by a strict, beautiful geometry.

If your blueprint fits the geometric pattern: You can build a unified, irreducible machine.
If your blueprint doesn't fit: No matter how you twist and turn the gears, the machine will always fall apart into smaller, independent pieces.

In short, the authors have drawn the ultimate map for building these mathematical engines, showing us exactly which blueprints lead to success and which lead to failure. They didn't just find a solution; they proved that this is the only way the solution can exist.

1. Problem Statement

The Deligne–Simpson Problem asks for a criterion to determine the existence of an irreducible solution to the matrix equation:
$A_1 A_2 \cdots A_k = I$
where each $A_i$ belongs to a prescribed conjugacy class $C_i$ in $GL_n(\mathbb{C})$ .

Irreducibility: The solution is irreducible if the matrices $A_1, \dots, A_k$ share no non-trivial common invariant subspace.
Context: This problem arises in the classification of systems of linear ordinary differential equations (ODEs) on the complex plane, specifically relating local monodromies to global properties.

The conjugacy classes $C_i$ are parameterized by:

A weight sequence $w = (w_1, \dots, w_k)$ .
A collection of non-zero complex numbers $\xi = (\xi_{ip})$ (eigenvalues).
Integer sequences $n_{ip}$ defining the ranks of products of $(A_i - \xi_{ip})$ .

2. Methodology and Framework

The authors resolve the problem by translating the matrix equation into the language of representation theory of quivers and geometry of weighted projective lines.

A. Quiver and Root System

The data $(w, \xi, n_{ip})$ is encoded into a star-shaped quiver $Q_w$ with a central vertex $\ast$ and $k$ arms corresponding to the conjugacy classes.

The problem is mapped to the existence of a simple module over a multiplicative preprojective algebra $\Lambda_q(Q_w)$ .
The dimension vector $\alpha$ of the module corresponds to the rank data of the matrices.
The parameters $\xi$ determine a multiplicative parameter $q \in (\mathbb{C}^*)^I$ for the algebra.

B. Geometric Translation (Parabolic Bundles)

The authors utilize the category of parabolic bundles on a weighted projective line $X$ .

They establish an equivalence between representations of the fundamental group of $\mathbb{P}^1 \setminus D$ (where $D$ are marked points) and logarithmic connections on vector bundles with prescribed residues.
Riemann-Hilbert Correspondence: They use this correspondence to link the existence of irreducible matrix solutions to the existence of irreducible parabolic bundles equipped with a $\zeta$ -connection (where $\zeta$ are logarithms of $\xi$ ).

C. Key Technical Tools

Reflection Functors: The authors employ multiplicative reflection functors (analogous to Coxeter reflections) to transform the quiver and parameters while preserving the existence of simple modules. This allows them to reduce complex cases to standard forms.
Tubular and Extended Tubular Types: The analysis is divided based on the type of the underlying quiver $Q_w$ $Q_{w}$ :
- Tubular: The quiver is an extended Dynkin diagram (e.g., $\tilde{D}_4, \tilde{E}_6$ , etc.).
- Extended Tubular: The quiver is "over-extended" (extended Dynkin with an extra vertex).
Perpendicular Categories: They analyze subcategories orthogonal to specific stable bundles to reduce the problem to representations of smaller, well-understood quivers.

3. Key Contributions and Results

The paper confirms a conjecture by the first author (Crawley-Boevey) regarding the necessary and sufficient conditions for the existence of an irreducible solution.

The Main Theorem (Theorem 1.1)

An irreducible solution exists if and only if the associated dimension vector $\alpha = \alpha_C$ satisfies three conditions:

Root Condition: $\alpha$ is a positive root of the root system associated with $Q_w$ .
Product Condition: $\xi[\alpha] = 1$ , where $\xi[\alpha] = \prod \xi_{ip}^{n_{ip-1} - n_{ip}}$ .
Indecomposability Condition (The "Strict" Condition): For any non-trivial decomposition $\alpha = \beta + \gamma + \dots$ into positive roots such that $\xi[\beta] = \xi[\gamma] = \dots = 1$ , the inequality holds:
$p_w(\alpha) > p_w(\beta) + p_w(\gamma) + \dots$
where $p_w(\alpha) = 1 - \langle \alpha, \alpha \rangle$ is the defect function.

Specific Contributions of this Paper

Proof of Necessity: While the sufficiency of these conditions was proven in previous work (Crawley-Boevey and Shaw), this paper provides the rigorous proof of the necessity direction.
Resolution of Tubular Cases: The authors prove that if the quiver is of tubular type and the dimension vector is a multiple of the minimal imaginary root ( $\alpha = h\delta$ with $h \ge 2$ ), no irreducible solution exists. This is done by analyzing the structure of connections on tubes and showing that such bundles always admit invariant subbundles.
Resolution of Extended Tubular Cases: Similarly, they prove non-existence for over-extended Dynkin quivers under specific conditions (Theorem 5.2).
Multiplicative Preprojective Algebras: The paper establishes a deep connection between the Deligne-Simpson problem and the representation theory of multiplicative preprojective algebras, proving that simple modules only exist for specific "rigid" dimension vectors (Theorem 7.5).

4. Significance

Completeness: This work completes the solution to the Deligne-Simpson problem, providing a complete combinatorial and geometric characterization of when irreducible monodromy representations exist.
Unification: It unifies several areas of mathematics:
- Linear Algebra: Matrix equations and conjugacy classes.
- Algebraic Geometry: Weighted projective lines, parabolic bundles, and logarithmic connections.
- Representation Theory: Quivers, root systems, and preprojective algebras.
Methodological Advancement: The paper demonstrates the power of using weighted projective lines and reflection functors to solve problems that are difficult to approach via direct linear algebraic methods. It clarifies the role of "rigid" representations in the context of differential equations.
Independence: The authors note that while a parallel proof was announced by Cheng Shu, this paper offers a direct, self-contained approach based on the 2018 draft, avoiding the complexities of parabolic bundles in general position arguments used in other approaches.

In summary, the paper establishes that the existence of irreducible solutions to the Deligne-Simpson problem is governed entirely by the combinatorics of the associated root system and the specific values of the eigenvalues, confirming that "generic" solutions exist if and only if the dimension vector is a root that cannot be decomposed into smaller roots with the same product of eigenvalues.