Torsion pairs and 3-fold flops

This paper classifies intermediate t-structures on the local derived category of a 3-fold flopping contraction and minimal resolutions of Kleinian singularities, thereby describing the complete lattice of torsion classes for associated modification and preprojective algebras and providing foundational steps for understanding spherical objects in derived categories of surfaces and 3-folds.

The Gist

Imagine you are an architect trying to understand the hidden structure of a very strange, twisted building. This building isn't made of bricks and mortar, but of pure mathematics. Specifically, it's a "3-fold flop," a shape that exists in higher dimensions where certain curves can be flipped inside out, changing the building's layout without changing its fundamental DNA.

This paper, written by Parth Shimpi, is a master key that unlocks the secrets of how to navigate this building. It answers a question that has puzzled mathematicians for decades: "If I look at this building through different mathematical 'lenses' (called t-structures), what are all the possible ways I can organize its contents?"

Here is the breakdown using everyday analogies:

1. The Building and the "Lenses"

Think of the 3-fold flop as a complex, multi-room mansion.

• The Rooms (Objects): Inside are various mathematical objects (like sheaves, which are like bundles of data or fabric draped over the curves of the building).
• The Lenses (t-structures): A mathematician can look at the mansion through different "lenses."
  • Lens A (Geometric): You see the rooms as they physically are. You see the walls, the floors, and the people (sheaves) walking around. This is the "geometric" view.
  • Lens B (Algebraic): You see the mansion as a giant spreadsheet or a circuit board. You don't see the walls; you see the connections, the numbers, and the logic gates. This is the "algebraic" view.
  • The Problem: For a long time, mathematicians knew about Lens A and Lens B. But they didn't know if there were other lenses in between, or if they had found every possible way to look at the building.

2. The "Heart" of the Matter

In this math world, every lens has a "Heart." The Heart is the specific list of items you consider "normal" or "positive" when looking through that lens.

• If you look through the Geometric Lens, the Heart is "Coherent Sheaves" (the physical fabric).
• If you look through the Algebraic Lens, the Heart is "Modules" (the spreadsheet data).

The paper's main goal was to classify all possible Hearts that sit somewhere between these two extremes.

3. The Three Types of Views (The Classification)

Shimpi discovered that there are only three types of ways to organize this building. It's like saying there are only three ways to arrange a library:

1. The Purely Geometric View: You are looking at a slightly different version of the building (a "birational model"). Maybe you flipped a curve inside out (a "flop"), and now the rooms are arranged differently, but you are still looking at the physical structure.
2. The Purely Algebraic View: You are looking at the building as a set of equations and algebraic rules. These are "mutations" of the standard algebraic view. It's like rearranging the spreadsheet formulas.
3. The Hybrid View (The "Semi-Geometric"): This is the most interesting one. Imagine you are looking at the building, but you decide to treat some parts of it as physical rooms and other parts as spreadsheet data.
   • Analogy: Imagine a house where the kitchen is treated as a physical room (you can walk in it), but the attic is treated as a list of numbers (you can only access it via a database). The paper proves that every possible way to look at the building is just a combination of these three types. You can't invent a fourth, weird way to look at it.

4. The "Heart Fan" (The Map)

To prove this, the author draws a massive map called a Heart Fan.

• Imagine a giant compass rose or a snowflake made of triangles.
• Each point on this map represents a specific "Heart" (a specific way of organizing the building).
• The paper shows that this map is perfectly complete. There are no hidden corners. If you are standing anywhere on this map, you are either looking at a geometric version, an algebraic version, or a mix of both.
• The map is built using "Dynkin diagrams," which are like family trees of shapes. The author shows that the rules for moving around this map are the same as the rules for flipping curves in the building.

5. Why Does This Matter? (The "Bricks")

The paper also solves a side puzzle: Classifying "Bricks."

• In this math world, a "Brick" is a building block that cannot be broken down further. It's the simplest possible object in a specific view.
• The paper proves that every single "Brick" you can find in this building is either:
  • A simple point (like a single pixel on a screen).
  • A simple curve (like a single thread of fabric).
  • Or a specific algebraic number derived from the building's DNA.
• This is huge because "Bricks" are often used to build "Spherical Objects," which are the keys to understanding how the building can be transformed into itself (autoequivalences). If you know all the bricks, you know all the keys.

Summary

Parth Shimpi's paper is like a complete inventory list for a magical, shape-shifting building.

• Before: Mathematicians knew about the "Physical" view and the "Algebraic" view, but they were worried there might be weird, undiscovered views hiding in the shadows.
• Now: The paper proves that no shadows exist. Every possible view is either a physical version, an algebraic version, or a sensible mix of the two.
• The Result: We now have a complete map (the Heart Fan) and a complete list of building blocks (Bricks). This allows mathematicians to navigate these complex shapes with confidence, knowing exactly what tools they need to understand how these shapes twist, turn, and relate to one another.

It's the difference between wandering lost in a maze and having a perfect, 3D blueprint that shows every single corridor, dead end, and secret door.

Technical Summary

Here is a detailed technical summary of the paper "Torsion pairs and 3-fold flops" by Parth Shimpi.

1. Problem Statement and Context

The paper addresses the classification of t-structures (specifically their hearts) and torsion pairs within the bounded derived category of coherent sheaves, $D^bX$, supported on the exceptional fiber of a 3-fold flopping contraction $\pi: X \to Z$.

• Setting: $Z = \text{Spec}(R)$ is a complete local base with an isolated Compound Du Val (cDV) singularity. $X$ is a Gorenstein terminal 3-fold. The category of interest is $D^0X$, the full subcategory of $D^bX$ supported on the exceptional fiber $\pi^{-1}(\mathfrak{m})$.
• The Challenge: While the classification of "algebraic" hearts (module categories of finite-dimensional algebras) and "geometric" hearts (categories of coherent sheaves on birational models) was partially understood via the Homological Minimal Model Programme (HMMP) and Bridgeland-Chen flops, a complete classification of all intermediate t-structures (those lying between the perverse heart and its shift) was missing.
• Specific Gap: Previous work by Hara and Wemyss classified hearts in the "finite-type" null subcategory $C = \ker(R\pi_*)$. However, $D^0X$ exhibits "affine" behavior with an infinite silting fan, making standard mutation-based classification insufficient. The paper seeks to determine if there are "hidden" t-structures that do not fit into the known algebraic or geometric frameworks.

2. Methodology

The author employs a synthesis of homological algebra, representation theory, and convex geometry to tackle the problem.

• Heart Fans and Numerical Criteria: The paper utilizes the concept of the heart fan (introduced by Broomhead, Pauksztello, Ploog, and Woolf). This associates a cone in a real vector space (dual to the Grothendieck group) to each t-structure. The core strategy is to prove that every intermediate heart is numerical (i.e., its heart cone is non-zero), thereby allowing the classification of hearts to be reduced to the combinatorics of the heart fan.
• Mutation Combinatorics: The paper leverages the combinatorics of Dynkin diagrams (specifically affine ADE types) and mutation classes of modifying modules (Iyama-Wemyss theory). The set of birational models of $X$ is shown to be in bijection with chambers in an intersection arrangement derived from the affine root system.
• Semi-Geometric Decomposition: The author introduces the concept of semi-geometric hearts. These are constructed by gluing algebraic hearts (associated with partial contractions) and geometric hearts (associated with birational models) over an open cover of the space.
• Line Bundle Actions: A crucial technique involves the action of the Nef monoid (line bundles on birational models) on the poset of t-structures. By analyzing the orbits of the standard perverse heart under tensor products with nef line bundles, the author establishes bounds for arbitrary t-structures.
• Skyscraper Hunting: To prove that non-algebraic hearts are numerical, the paper develops an algorithm ("Skyscraper-hunting") that tracks the position of skyscraper sheaves $\mathcal{O}_p$ within the torsion pairs of a given heart. This forces any non-algebraic heart to be bounded by geometric hearts, proving its heart cone is non-zero.

3. Key Contributions and Results

A. Classification of Intermediate Hearts (Theorem A & C)

The main result is a complete classification of all t-structures $K$ on $D^0X$ such that $K \subseteq \text{per}(X/Z)[-1, 0]$.

• Structure: Every such heart $K$ is determined by a birational model $W$ of $X$ and a partial contraction $\tau: W \to Y$.
• Decomposition: On the locus where $\tau$ is an isomorphism, $K$ restricts to coherent sheaves on $W$ (possibly tilted in skyscraper sheaves). On the exceptional fibers of $\tau$, $K$ restricts to an algebraic mutation of the perverse heart associated with the local flopping contraction.
• The Heart Fan: The set of all such hearts forms a complete lattice. The "heart fan" of the perverse heart $\text{per}(X/Z)$ is identified with the intersection arrangement associated with the restricted affine root system.
  • Full-dimensional cones correspond to algebraic hearts (mutations of the perverse heart).
  • Cones on the hyperplane $\{\delta = 0\}$ correspond to geometric and semi-geometric hearts (categories of coherent sheaves on birational models or partial contractions thereof).
  • Crucial Finding: There are no "hidden" hearts. Every intermediate heart is either algebraic, geometric, or a semi-geometric combination of the two.

B. Classification of Bricks and Semibricks (Theorem B & 5.25)

The paper classifies bricks (objects with 1-dimensional endomorphism algebras) in the perverse heart.

• Result: Every brick is either:
  1. A simple module over a modification algebra (algebraic type).
  2. A skyscraper sheaf on a birational model (geometric type).
• K-theory: The class of any brick in the Grothendieck group is a primitive restricted root of the associated affine Dynkin data. This generalizes Crawley-Boevey's result for preprojective algebras to the 3-fold setting.

C. The Heart Fan Structure

The paper proves that the heart fan of $\text{per}(X/Z)$ is the J-cone arrangement (a subfan of the affine Weyl arrangement).

• Numerical Completeness: It is proven that the trivial cone $\{0\}$ is not the heart cone of any intermediate heart. This implies that the "affine curse" (infinite paths in the silting fan) does not prevent a complete classification; the geometry of the heart fan is sufficiently rigid to capture all t-structures.
• Lattice Isomorphism: The poset of algebraic hearts is isomorphic to the weak order on chambers of the intersection arrangement. The poset of all intermediate hearts is a complete lattice where geometric hearts act as suprema/infima of chains of algebraic hearts.

D. Analogy with Surfaces

The paper establishes a rigorous correspondence between the 3-fold case and the 2-dimensional case (Kleinian singularities/preprojective algebras). It shows that the results for 3-folds restrict naturally to surfaces via hyperplane sections, providing a unified framework for understanding torsion pairs in these derived categories.

4. Significance and Impact

• Resolution of the "Affine Curse": The paper resolves the difficulty of classifying t-structures in affine-type derived categories (which have infinite silting fans) by showing that the heart fan is complete and that every heart is numerical.
• Bridge between Geometry and Algebra: It provides a precise dictionary between geometric operations (flopping curves, partial contractions) and algebraic operations (mutations of modifying modules, tilting in torsion pairs).
• Foundation for Future Work: This classification is a "first step" towards describing all t-structures and spherical objects in derived categories of surfaces and 3-folds. It provides the necessary tools to study autoequivalences and the structure of the derived category beyond the known "finite" or "purely geometric" cases.
• Application to Spherical Objects: The classification of bricks directly leads to a classification of spherical modules, which are central to the study of autoequivalences (e.g., spherical twists) in these categories.

In summary, Shimpi's work provides a complete, exhaustive classification of intermediate t-structures for 3-fold flops, demonstrating that the landscape of these categories is entirely governed by the interplay between the combinatorics of affine root systems (algebraic mutations) and the geometry of the Minimal Model Program (birational models).