🔢 mathematics

Indecomposable extensions of perverse sheaves over a closed stratum

This paper establishes a categorical framework for extensions of perverse sheaves over a closed stratum by replacing topological homotopy assumptions with the semisimplicity of local systems, thereby generalizing the MacPherson–Vilonen description to construct a maximal extension functor and provide a structural classification of indecomposable perverse sheaves.

Original authors: Alessio Cipriani

Published 2026-07-13

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Alessio Cipriani

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 an architect trying to build a sturdy house (a mathematical object called a "perverse sheaf") on a piece of land that has a weird, jagged edge. This land is called a "stratified space," and the jagged edge is a "closed stratum" (let's call it the "Boundary").

For a long time, mathematicians had a very strict rulebook for building these houses. They said, "You can only build if the Boundary is perfectly smooth and has no holes or twists in its shape." This was like saying, "You can only build a house if the ground is perfectly flat and empty." This rule came from famous work by MacPherson and Vilonen in 1986.

But what if the ground isn't flat? What if it's a bumpy, twisted rock? The old rulebook said you couldn't build there.

The Big Discovery
Alessio Cipriani, the author of this paper, says, "Hold on! We don't need the ground to be perfectly flat. We just need the 'local systems' (think of these as the tiny, fundamental bricks or patterns on the Boundary) to be simple enough to sort out."

Cipriani proves that as long as you can easily separate and categorize these tiny bricks (a condition called "semisimple"), you can build your house on any bumpy, twisted Boundary. You don't need the ground to be perfectly smooth anymore.

The New Blueprint: Extension Pairs
So, how do you build on this bumpy ground? Cipriani introduces a clever new way to describe the house using a "two-part blueprint" called an Extension Pair.

Imagine you want to build a house that connects to the Boundary. Instead of trying to draw the whole messy house at once, you draw two simpler sketches:

Sketch A: A version of the house that has no part sticking out onto the Boundary (it's "clean" at the bottom).
Sketch B: A version of the house that has no part hanging off the top (it's "clean" at the top).

Cipriani shows that if you take these two sketches and snap them together in a very specific way (making sure a certain "glue" between them doesn't fall apart), you get a unique, indestructible house. He calls this a maximal extension.

This is a huge deal because it turns a messy, impossible-to-solve puzzle into a game of matching two simpler pieces. It's like saying, "To understand the whole complex machine, just look at the part that doesn't touch the floor and the part that doesn't touch the ceiling, and see how they fit."

What This Rules Out
The paper is very clear about what doesn't work.

It explicitly rejects the idea that you need the Boundary to be perfectly smooth (specifically, it doesn't need the second "hole" or twist in the shape to be zero, which was a requirement in the old 1986 rules).
It also rejects the idea that you can always build a house just by gluing things together randomly. If the "glue" (the mathematical connection between your two sketches) falls apart or splits into two separate pieces, you don't get a single, strong house. You get a pile of junk. The paper proves that a house is only "indecomposable" (a single, solid unit) if that glue holds tight and doesn't split.

How Sure Are We?
The author isn't just guessing or running a computer simulation. This is a proven mathematical fact.

The paper constructs a rigorous "equivalence of categories," which is a fancy way of saying, "We have built a perfect, two-way bridge between the messy world of building on bumpy ground and the clean world of matching two sketches."
The proof relies on solid logic (using things called "octahedral axioms" and "triangles" in a mathematical sense) to show that every valid house corresponds to exactly one pair of sketches, and vice versa.
The only "catch" is that the tiny bricks on the Boundary must be "semisimple." The paper explains that this is true in many common situations (like when the Boundary has a finite number of twists and you are using a specific type of number system), but if the bricks are too chaotic, the method doesn't apply.

The Final Takeaway
Before this paper, if you wanted to classify all the possible "indecomposable" (unbreakable) houses on a bumpy Boundary, you were stuck. Now, Cipriani gives you a checklist:

Look at the tiny bricks on the Boundary. If they are simple, you're good.
If you want to build a house, find a "clean-bottom" sketch and a "clean-top" sketch.
Check if they fit together without splitting.
If they fit, you have found a unique, unbreakable house.

This doesn't just solve one small problem; it extends a famous 1986 theory to a much wider, messier, and more realistic world, giving mathematicians a new, powerful tool to understand complex shapes.

Technical Summary: Indecomposable Extensions of Perverse Sheaves Over a Closed Stratum

Problem Statement
The paper addresses the classification of indecomposable perverse sheaves on a topologically stratified space $X$ . While the category of perverse sheaves, $p\text{Perv}(X)$ , is known to be Krull–Schmidt (decomposing into finite direct sums of indecomposables) under certain conditions, a general structural classification of these indecomposable objects remains difficult. Specifically, the problem lies in understanding how perverse sheaves defined on the open complement $U = X \setminus S$ of a closed stratum $S$ can be extended to the whole space $X$ . Previous work by MacPherson and Vilonen [MV86] provided a framework for such extensions under restrictive homotopical assumptions ( $\pi_1(S) = \pi_2(S) = 0$ ) and for a specific perversity (the middle perversity). The goal of this work is to generalize this extension theory to arbitrary perversities and more general stratified spaces by replacing homotopical constraints with categorical ones.

Methodology
The author develops a categorical framework for extensions of a fixed perverse sheaf $F \in p\text{Perv}(U)$ over a closed stratum $S \subset X$ . The methodology relies on the following structural components:

Categorical Assumptions: The paper replaces the homotopical assumptions of [MV86] with the assumption that the category of local systems on the closed stratum $S$ , denoted $\text{Loc}(S)$ , is semisimple. This condition is satisfied, for instance, when $S$ has a finite fundamental group and the coefficient field characteristic does not divide the group order.
Projection Functors: Two additive functors, $P_!$ and $P_*$ , are defined on $p\text{Perv}(X)$ as the images of the natural morphisms $p\jmath_! p\jmath^* \to \text{id}$ and $\text{id} \to p\jmath_* p\jmath^*$ , respectively. These functors isolate the "part" of a perverse sheaf supported on the complement $U$ .
S-Small Extensions: The author introduces the notion of an $S$ -small extension, defined as an object $E \in p\text{Perv}(X)$ in the kernel of the intermediate restriction functor $p\iota_{!*}$ . These are characterized as extensions having no direct summands supported on $S$ .
Extension Pairs: A new auxiliary category, $\text{EP}(X \setminus S)$ , is introduced. Its objects are pairs $(A, B)$ of perverse sheaves on $X$ extending $F$ , where $A$ has no quotient supported on $S$ and $B$ has no sub-object supported on $S$ , subject to a vanishing condition on a specific product of extension classes in the derived category.

Key Contributions and Results

Equivalence of Categories (Theorem A & B): The primary result establishes a bijection, and subsequently an equivalence of additive categories, between the category of $S$ -small extensions of $F$ (denoted $\text{SES}(F)$ ) and the category of extension pairs relative to $F$ (denoted $\text{EP}(X \setminus S)$ ).
- The correspondence is given by the assignment $E \mapsto (P_!E, P_*E)$ .
- The inverse construction utilizes the octahedral axiom in the constructible derived category to construct an extension from a pair $(A, B)$ , relying on the vanishing of a specific $\text{Ext}^2$ class derived from the product of two $\text{Ext}^1$ classes.
- This equivalence holds under the sole assumption that $\text{Loc}(S)$ is semisimple, removing the need for $\pi_2(S)=0$ or specific perversity choices.
Maximal Extension Functor (Definition 3.18): The paper defines a generalized maximal extension functor, $p\Xi$ , which maps a perverse sheaf $F$ on $U$ to a specific $S$ -small extension on $X$ . This generalizes Beilinson's maximal extension (originally defined for algebraic varieties with middle perversity) to arbitrary topologically stratified spaces and perversities, provided $\text{Loc}(S)$ is semisimple.
Classification of Indecomposables (Theorem C): The paper provides a structural classification of all indecomposable perverse sheaves on $X$ . Every indecomposable object is either:
1. An extension by zero of an indecomposable local system on the closed stratum $S$ ; or
2. An $S$ -small extension arising from an extension pair $(A, B)$ where the canonical morphism $A \to B$ is indecomposable (cannot be written as a direct sum of morphisms).

Significance and Claims
The paper claims to extend the MacPherson–Vilonen description of perverse sheaves to a strictly more general setting. By shifting the focus from topological constraints on the stratum (vanishing of $\pi_1$ and $\pi_2$ ) to the algebraic property of semisimplicity of local systems, the framework applies to a wider class of stratified spaces and arbitrary perversity functions.

The work offers a "reduction result" for the classification problem: it reduces the study of indecomposable perverse sheaves on $X$ to the study of:

Indecomposable local systems on the strata.
Extension pairs of perverse sheaves on the complement, subject to a non-splitting condition on their canonical morphism.

The author notes that while the results generalize Beilinson's construction and the MacPherson–Vilonen theorem, the framework does not assume that $p\text{Perv}(X)$ is equivalent to a module category over a finite-dimensional algebra, nor does it require the Krull–Schmidt property to hold globally, relying only on the semisimplicity of the local systems on the specific closed stratum $S$ . The paper concludes with examples (including $\mathbb{CP}^1$ , $S^1$ , and $\mathbb{CP}^2$ ) illustrating how the classification works in specific two-strata scenarios, including cases where the maximal extension is not unique or where the standard conditions for extension pairs fail.

More like this