🔢 mathematics

Derived equivalence for the simple flop of type $G_2^{\dagger}$ via tilting bundles

This paper establishes the derived equivalence for the local model of the simple flop of type $G_2^{\dagger}$ , a unique case arising from a non-homogeneous roof, by constructing tilting bundles that yield a noncommutative crepant resolution.

Original authors: Wahei Hara

Published 2026-04-30

📖 4 min read🧠 Deep dive

CC0 1.0

Original authors: Wahei Hara

Original paper dedicated to the public domain under CC0 1.0 (http://creativecommons.org/publicdomain/zero/1.0/). ✨ 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 looking at a complex, twisted piece of origami. In the world of mathematics, specifically algebraic geometry, this "origami" represents a shape called a variety. Sometimes, mathematicians want to transform one shape into another without tearing or gluing, just by folding and unfolding in a specific way. This process is called a flop.

Think of a flop like a "magic trick" where you take a shape, fold a part of it inside out, and it emerges as a slightly different shape. Even though the two shapes look different on the outside, they are secretly "twins" in a deeper, invisible sense.

This paper, written by Wahei Hara, solves a long-standing puzzle about a very specific, tricky magic trick called the simple flop of type $G_2^\dagger$ .

Here is a breakdown of what the paper does, using simple analogies:

1. The Big Question: Are They the Same?

Mathematicians have a famous guess (called the DK conjecture) that says: If you can turn Shape A into Shape B using this "flop" magic trick, then their "DNA" (called derived categories) must be identical.

For most simple magic tricks, we already knew this was true. But there was one special, weird trick (the $G_2^\dagger$ flop) that was different from all the others. It came from a "non-homogeneous roof"—a fancy way of saying the structure holding the trick together wasn't a perfect, uniform crystal like the others; it was irregular. No one had proven that the "DNA" of the two shapes in this specific trick was actually the same.

2. The Solution: The "Tilting Bundle" Tool

To prove the shapes are twins, the author uses a special tool called a tilting bundle.

The Analogy: Imagine you have two different languages (Shape A and Shape B). To prove they are saying the same thing, you need a perfect translator. A "tilting bundle" is like a universal translator dictionary.
How it works: If you can build a dictionary that works perfectly for Shape A, and you can show that this same dictionary also works perfectly for Shape B, then you have proven that Shape A and Shape B are mathematically equivalent.

3. The Challenge: The Dictionary Was Broken

The author tried to use a famous, pre-existing dictionary (called Kapranov's bundle) that works for many other shapes. However, when he tried to use it for this specific $G_2^\dagger$ trick, the dictionary failed. It had a "glitch" (mathematically, a non-zero "Ext" group) that meant it couldn't translate the shapes correctly.

4. The Fix: Sewing the Glitch

Instead of throwing the dictionary away, the author found a clever repair.

He noticed that the glitch only happened between two specific pages in the dictionary.
He "stitched" those two pages together in a new way (creating a mathematical extension).
This repair fixed the glitch, turning the broken dictionary into a brand new, working tilting bundle.

5. The "Exchange Diagram": The Bridge

To prove that this new dictionary works for both shapes (the "before" and "after" of the flop), the author built a bridge between them.

He looked at the "roof" (the structure where the two shapes meet).
He found a special "key bundle" (a specific mathematical object) living on this roof.
He showed that if you push this key bundle down toward Shape A, it becomes the dictionary for Shape A. If you push it down toward Shape B, it becomes the dictionary for Shape B.
Because it's the same key bundle doing the work for both sides, the dictionaries are identical.

6. The Result

The paper proves that:

The Shapes are Twins: The two sides of this specific $G_2^\dagger$ flop are indeed derived equivalent. Their "DNA" matches perfectly.
A New Resolution: The author also found a "Noncommutative Crepant Resolution" (NCCR). Think of this as a secret, hidden blueprint that describes the singularity (the messy point where the shapes meet) in a way that is smoother and easier to understand than the original shapes. This blueprint is derived equivalent to both sides of the flop.

Summary

In short, Wahei Hara took a very difficult, irregular mathematical shape transformation that no one had solved yet. He found a broken tool, fixed it by stitching two parts together, and used this repaired tool to prove that the two sides of the transformation are secretly the same. He also discovered a new, smoother way to look at the messy center of the transformation.

This doesn't just solve a puzzle; it confirms that even for the most irregular, "non-homogeneous" shapes in this family, the deep mathematical rules of equivalence still hold true.

1. Problem Statement and Context

The paper addresses a specific case of the DK Conjecture (named after Bondal and Orlov, and Kawamata), which posits that if two smooth algebraic varieties $X_+$ and $X_-$ are $K$ -equivalent (i.e., there exists a common resolution $\tilde{X}$ such that the pullbacks of their canonical divisors are linearly equivalent), then they are $D$ -equivalent (i.e., their bounded derived categories of coherent sheaves, $D^b(\text{coh } X_+)$ and $D^b(\text{coh } X_-)$ , are equivalent as triangulated categories).

While the conjecture is known for standard flops (like the Atiyah flop) and certain homogeneous cases, this paper focuses on an exceptional, non-homogeneous example:

The Object: A simple flop of type $G^\dagger_2$ .
The Geometry: This flop arises from a "roof" (a variety $W$ admitting two distinct $\mathbb{P}^2$ -bundle structures) that is not a rational homogeneous manifold. This distinguishes it from all previously studied simple flops, which were derived from homogeneous roofs.
The Local Model: The varieties $X_+$ and $X_-$ are total spaces of vector bundles over a 5-dimensional quadric $Q_5$ , specifically $X_\pm = \text{Tot}_{Q_5}(G_\pm^\vee(-1))$ , where $G_\pm$ are Ottaviani bundles (stable rank-3 bundles with Chern classes $(2,2,2)$ ).
The Challenge: Constructing an explicit derived equivalence $\Phi: D^b(\text{coh } X_+) \xrightarrow{\sim} D^b(\text{coh } X_-)$ that respects the base ring $R = H^0(X_\pm, \mathcal{O}_{X_\pm})$ (i.e., an $R$ -linear equivalence compatible with the flopping contractions).

2. Methodology

The author employs the method of tilting bundles to establish the derived equivalence. The core strategy involves:

Constructing Tilting Bundles: Finding a vector bundle $T_\pm$ on each side ( $X_+$ and $X_-$ ) that satisfies the tilting conditions (generating the derived category and having no higher self-extensions).
Endomorphism Ring Isomorphism: Proving that the endomorphism algebras $\Lambda_+ = \text{End}_{X_+}(T_+)$ and $\Lambda_- = \text{End}_{X_-}(T_-)$ are isomorphic as $R$ -algebras.
Exchange Diagram: Utilizing the geometry of the "roof" $W$ (the exceptional divisor of the blow-up) to relate the bundles on $X_+$ and $X_-$ via a "key bundle" $E$ defined over $W$ .

Key Technical Steps:

Borel–Bott–Weil Computations: The author performs detailed cohomological calculations on the 5-dimensional quadric $Q_5$ (a homogeneous space of type $B_3$ ) to determine the vanishing of Ext groups for various bundles, including Ottaviani bundles and spinor bundles.
Modification of Kapranov's Bundle: A standard tilting bundle on $Q_5$ (due to Kapranov) does not pull back to a tilting bundle on $X_\pm$ because it acquires non-vanishing $\text{Ext}^1$ . The author resolves this by taking a non-trivial extension of specific summands to cancel the obstruction, creating a new tilting bundle $T_\pm$ .
The Exchange Diagram: The proof relies on a specific vector bundle $E$ over the roof $W$ . By pulling $E$ back to the common resolution $\tilde{X}$ and pushing it forward to $X_\pm$ , the author shows that $E$ generates the specific extension bundles $P_\pm$ required for the tilting bundles on both sides.

3. Key Contributions and Results

A. Construction of Tilting Bundles

The paper explicitly constructs the tilting bundles $T_\pm$ on $X_\pm$ . Let $S_\pm$ be the spinor bundle on $Q_5$ and $P_\pm$ be the unique non-trivial extension defined by:
$0 \to \mathcal{O}_{X_\pm}(-2) \to P_\pm \to S_{X_\pm}^\vee \to 0$
The tilting bundle is given by:
$T_\pm = \mathcal{O}_{X_\pm}(-2) \oplus \mathcal{O}_{X_\pm}(-1) \oplus \mathcal{O}_{X_\pm} \oplus \mathcal{O}_{X_\pm}(1) \oplus \mathcal{O}_{X_\pm}(2) \oplus P_\pm$
The author proves (Theorem 3.5) that $T_\pm$ generates $D^b(\text{coh } X_\pm)$ and satisfies $\text{Ext}^i(T_\pm, T_\pm) = 0$ for $i > 0$ .

B. The Main Theorem (Derived Equivalence)

Theorem 1.2: There exists an exact equivalence of $R$ -linear triangulated categories:
$\Phi: D^b(\text{coh } X_+) \xrightarrow{\sim} D^b(\text{coh } X_-)$
such that $Rf_{+*} \simeq Rf_{-*} \circ \Phi$ .

Significance: This confirms the DK conjecture for the first known non-homogeneous simple flop.
Mechanism: The equivalence is induced by the isomorphism of endomorphism rings $\Lambda \cong \text{End}_{X_+}(T_+) \cong \text{End}_{X_-}(T_-)$ .

C. Non-Commutative Crepant Resolutions (NCCR)

Corollary 3.9: The algebra $\Lambda$ constructed above is a Non-Commutative Crepant Resolution (NCCR) of the singularity $\text{Spec } R$ .

This provides a concrete example where a singular affine scheme admits an NCCR that is derived equivalent to both of its projective crepant resolutions ( $X_+$ and $X_-$ ).
The modules $f_{+\ast}T_+$ and $f_{-\ast}T_-$ are shown to be Cohen-Macaulay, satisfying the criteria for NCCRs.

D. Geometric Insights

The Roof $W$ : The paper clarifies the geometry of the roof $W$ (a smooth intersection of divisors in $OGr(3,8)$), identifying it as the only known non-homogeneous roof.
K3 Surfaces: The flop induces a derived equivalence between two non-isomorphic K3 surfaces of degree 12 (zero loci of sections of the twisted Ottaviani bundles), offering a new perspective on their D-equivalence.

4. Significance and Impact

Breaking Homogeneity: Prior to this work, derived equivalences for simple flops were largely restricted to cases where the exceptional divisor (the roof) was a rational homogeneous manifold. This paper proves the DK conjecture for a non-homogeneous roof, significantly expanding the scope of known $D$ -equivalences in birational geometry.
Methodological Advancement: The technique of modifying a known tilting bundle by taking a specific non-trivial extension (to kill $\text{Ext}^1$ ) provides a robust template for constructing tilting bundles in more complex, non-homogeneous settings.
NCCR Construction: The explicit construction of the NCCR $\Lambda$ for the $G^\dagger_2$ singularity contributes to the "Craw–Ishii type" problem, linking geometric resolutions to moduli spaces of modules over non-commutative algebras.
Alternative to Mutation: While other works (e.g., by Kanemitsu or Toda) use mutations of semi-orthogonal decompositions (SODs) to prove such equivalences, this paper uses tilting bundles. This approach yields a more explicit description of the equivalence and ensures the $R$ -linearity (compatibility with the flop contraction) more directly, which can be difficult to verify with mutation methods.

In summary, Wahei Hara's paper successfully resolves the derived equivalence for the exceptional $G^\dagger_2$ simple flop by constructing explicit tilting bundles and identifying the associated non-commutative crepant resolution, thereby bridging a gap between homogeneous and non-homogeneous birational geometry.