🔢 mathematics

Non-commutative crepant resolutions, an overview

This paper provides a survey of results concerning non-commutative crepant resolutions (NCCRs), which serve as non-commutative analogues to the classical crepant resolutions found in algebraic geometry.

Original authors: Michel Van den Bergh

Published 2026-02-16

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Michel Van den Bergh

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 fix a broken building. In the world of algebraic geometry, "buildings" are shapes defined by equations, and sometimes these shapes have sharp corners, crinkles, or singularities where the geometry breaks down.

The goal of this paper is to explore two different ways to "fix" these broken shapes so they become smooth and usable for calculations. The author, Michel Van den Bergh, introduces a new, unconventional tool called a Non-Commutative Crepant Resolution (NCCR).

Here is the breakdown of the paper using simple analogies:

1. The Problem: The Broken Building

Imagine a beautiful sculpture that has a sharp, jagged point where it shouldn't. In math, this is a singularity.

The Old Way (Crepant Resolution): Traditionally, mathematicians tried to fix this by "blowing up" the sharp point. They would replace the jagged spot with a smooth curve or a small sphere. This is like sanding down the rough edge until it's perfectly smooth.
- The Catch: Sometimes, you can't sand it down. Or, if you can, there might be two different ways to sand it, resulting in two different smooth shapes.
The Mystery: Even though these two different smooth shapes look different, they are secretly "twins." They share the same hidden DNA (mathematically, their "derived categories" are equivalent). This is a deep mystery in geometry.

2. The New Tool: The "Non-Commutative" Fix

Instead of physically sanding the building (which might be impossible), the author suggests building a virtual blueprint that acts like a smooth building, even if the physical one is still broken.

The Analogy: Imagine you have a broken, jagged rock. You can't smooth the rock itself. But, you can create a complex, non-physical "shadow" or "map" of the rock using a special language (non-commutative algebra).
Why "Non-Commutative"? In normal math, $A \times B = B \times A$ (order doesn't matter). In this new language, order does matter ( $A \times B \neq B \times A$ ). This extra flexibility allows the "shadow" to be perfectly smooth, even when the original rock is jagged.
The Result: This "shadow" (the NCCR) behaves exactly like a smooth building. It allows mathematicians to do calculations that would be impossible on the broken rock.

3. The "Flop" and the "Tiling"

The paper discusses a specific scenario called a Flop.

The Analogy: Imagine a bridge that can be flipped upside down. The two sides look different, but they connect the same two points.
The Magic: The author shows that if you have two different ways to fix the broken building (two different smooth bridges), they are both secretly connected to the same non-commutative "shadow."
The Tiling: To build this shadow, mathematicians use something called a Tilting Bundle. Think of this as a set of Lego bricks. If you arrange these bricks just right on the broken building, they snap together to form a perfect, smooth structure (the non-commutative ring) that reveals the building's true, hidden smoothness.

4. Symmetry and Groups (The "Covariants")

A huge part of the paper deals with shapes created by symmetry groups (like rotating a snowflake).

The Analogy: Imagine a kaleidoscope. You have a pattern, and you rotate it. The result is a complex, symmetric shape. Sometimes this shape has a bad center.
The Solution: The paper provides a recipe to build the "shadow" (NCCR) for these symmetric shapes. It involves picking specific "ingredients" (mathematical modules) that fit together perfectly.
The "Window" Concept: The authors use a concept called "Windows." Imagine looking at a complex 3D object through a window. Depending on where you stand (the angle of the window), you see a different 2D picture. The paper proves that all these different pictures are actually just different views of the same underlying "shadow" (the NCCR).

5. The "Stringy" Counting Game

Finally, the paper touches on a counting problem.

The Analogy: If you have a broken building, how many "rooms" does it effectively have?
The Stringy Euler Characteristic: Mathematicians have a special way of counting rooms in broken buildings called the "Stringy Euler Characteristic." It's like counting the rooms in the idealized version of the building, ignoring the cracks.
The Big Question: The author asks: "If we build our non-commutative shadow, does the number of 'rooms' in the shadow match the 'stringy' count of the broken building?"
The Answer: In many cases, yes! The number of building blocks in the shadow matches the magical count of the broken shape. This suggests the shadow isn't just a trick; it captures the true essence of the object. However, the paper also admits there are some weird exceptions where this doesn't work, showing the mystery is still not fully solved.

Summary

This paper is a guidebook for a new way of thinking about broken geometric shapes. Instead of trying to physically fix the cracks (which is often impossible), it suggests building a perfect, non-commutative "shadow" that acts as a smooth substitute.

Old Way: Sand the rock until it's smooth.
New Way (NCCR): Build a perfect, virtual map of the rock that behaves as if it were smooth.
Why it matters: This map allows mathematicians to solve problems that were previously stuck, and it reveals that different ways of fixing a shape are all connected to the same hidden truth.

The author, Michel Van den Bergh, is essentially saying: "Don't just look at the cracks. Look at the shadow they cast, and you'll find the shape was smooth all along."

Technical Summary: Non-commutative crepant resolutions, an overview

Author: Michel Van den Bergh
Source: Proceedings of the International Congress of Mathematicians 2022 (arXiv:2207.09703v2)

1. Problem Statement and Context

The paper addresses the problem of resolving singularities in algebraic geometry, specifically focusing on crepant resolutions. A crepant resolution $\pi: Y \to X$ of a normal variety $X$ with Gorenstein singularities is a resolution where the canonical bundle is preserved ( $\pi^* \omega_X = \omega_Y$ ).

Key Challenges:

Existence: Crepant resolutions do not always exist (e.g., certain 3D hypersurface singularities).
Uniqueness: When they exist, they are generally not unique. Different crepant resolutions are related by "flops" (birational transformations).
Categorical Equivalence: The Bridgeland-Kawamata Conjecture posits that all crepant resolutions of a given variety $X$ are derived equivalent (i.e., their bounded derived categories of coherent sheaves, $D(Y)$ , are equivalent as triangulated categories).
Non-Commutative Substitutes: In cases where commutative crepant resolutions do not exist, or to unify the study of different resolutions, the paper explores Non-Commutative Crepant Resolutions (NCCRs). These are non-commutative algebras $\Lambda$ that serve as "smooth" substitutes for the singular variety $X$ , preserving key homological properties.

2. Methodology and Definitions

2.1. Core Definitions

The paper establishes rigorous definitions for NCCRs based on the properties of the endomorphism ring of a reflexive module $M$ over a normal Gorenstein domain $R$ :

Non-Commutative Resolution (NCR): An algebra $\Lambda = \text{End}_R(M)$ where $\Lambda$ has finite global dimension.
Non-Commutative Crepant Resolution (NCCR): An NCR where $\Lambda$ is additionally a Cohen-Macaulay $R$ -module.
Twisted NCCR: A generalization where $\Lambda$ is a reflexive Azumaya algebra (locally matrix algebra) with finite global dimension.

2.2. Theoretical Framework

The methodology relies on the interplay between:

Tilting Theory: Using tilting complexes/bundles $T$ on a resolution $Y$ to establish derived equivalences $D(Y) \simeq D(\text{End}_Y(T))$ .
Minimal Model Program (MMP): Utilizing the fact that different crepant resolutions are connected by flops.
Representation Theory: Constructing NCCRs via modules of covariants for reductive group actions.
Mutation Theory: Using "mutations" (analogous to flops) to generate infinite families of NCCRs from a single one.
Hodge Theory: Investigating the relationship between NCCRs and stringy invariants (stringy $E$ -functions).

3. Key Contributions and Results

3.1. Construction of NCCRs

The paper surveys several methods for constructing NCCRs:

Quotient Singularities (Finite Groups): For a finite group $G \subset SL(n)$ acting on $W$ , the skew group ring $\text{Sym}(W)\#G$ is an NCCR. This generalizes the classical McKay correspondence.
Quotient Singularities (Reductive Groups): A major contribution is the construction of NCCRs for $X//G$ $X // G$ where $G$ $G$ is a reductive group.
- The author and Špela Špenko define strongly critical weights and construct modules of covariants $M(U_\delta)$ such that $\Lambda = \text{End}_{S^G}(M(U_\delta))$ is an NCCR.
- This applies to quasi-symmetric representations (where weights sum to zero on lines through the origin), covering determinantal varieties (matrices of rank $\leq l$ ) and other GIT quotients.
Toric Varieties:
- 3D Case: Broomhead's result (via dimer models) proves that 3D Gorenstein affine toric varieties admit toric NCCRs.
- Higher Dimensions: The paper presents constructions for higher-dimensional toric varieties that may not admit toric NCCRs but do admit general NCCRs.
Crepant Resolutions with Tilting Bundles: If a commutative crepant resolution $Y \to X$ admits a tilting bundle $T$ , then $\text{End}_Y(T)$ is an NCCR. The paper discusses examples like cotangent bundles of flag varieties ( $T^*(G/P)$ ) and determinantal varieties.

3.2. Relationship Between Commutative and Non-Commutative Resolutions

Derived Equivalence: The paper confirms that in dimension 3, all NCCRs are derived equivalent to each other and to any commutative crepant resolution (if one exists). This is proven using the mutation theory of Iyama and Wemyss.
Mutation: The paper details the mutation procedure (left and right) which transforms one NCCR into another. For 3D terminal singularities, these mutations correspond to wall-crossings in a hyperplane arrangement, generating a finite set of NCCRs (often more than the number of commutative crepant resolutions).
Categorical Resolutions: NCCRs are conjectured to be examples of "strongly crepant categorical resolutions" (in the sense of Kuznetsov). The paper proves this for dimensions $\leq 3$ .

3.3. Connection to GIT and Hyperplane Arrangements

A significant portion of the work (joint with Špenko and Halpern-Leistner) connects NCCRs to Geometric Invariant Theory (GIT):

The "Window" Principle: NCCRs can be realized as endomorphism rings of tilting bundles on specific GIT quotients (Deligne-Mumford stacks) $X_{ss, \mu} // G$ .
Hyperplane Arrangements: The choice of the NCCR corresponds to a chamber in a hyperplane arrangement on the character lattice of $G$ . Moving between chambers corresponds to derived equivalences (flops).
SKMS (Stringy Kähler Moduli Space): The derived equivalences between different NCCRs are interpreted as monodromy actions of the fundamental group of the SKMS on the derived category. This provides a geometric interpretation of the non-canonical nature of the equivalences.

3.4. Stringy Invariants and Conjectures

The paper investigates the relationship between NCCRs and stringy $E$ -functions (motivic invariants of singularities):

Conjecture: If $X$ has an NCCR, its stringy $E$ -function is a polynomial (and the stringy Euler characteristic is an integer).
Evidence: The conjecture holds for finite quotient singularities and 3D toric varieties.
Counterexamples: The paper notes that the conjecture fails for twisted NCCRs and for certain GIT quotients (e.g., cones over Grassmannians where $\gcd(d,n) \neq 1$ ), highlighting the necessity of the "crepant" (non-twisted) condition.
Conjecture 5.14: The number of indecomposable summands in the module defining an NCCR equals the stringy Euler characteristic of the singularity. This is verified for several cases, including determinantal varieties and Grassmannian cones.

4. Significance and Impact

Unification of Singularities: The concept of NCCR provides a unified framework to study singularities that lack commutative crepant resolutions, extending the scope of the Minimal Model Program to non-commutative geometry.
Bridge Between Fields: The work deeply connects algebraic geometry, representation theory (modules of covariants), and non-commutative algebra (tilting theory, mutation).
Computational Tools: The introduction of hyperplane arrangements and "windows" provides a combinatorial toolkit for constructing NCCRs for complex GIT quotients, including determinantal varieties and symmetric/skew-symmetric matrix varieties.
Mirror Symmetry: The interpretation of derived equivalences via the SKMS and local systems strengthens the link between algebraic geometry and mirror symmetry, offering a concrete realization of the "stringy Kähler moduli space."
Open Problems: The paper clarifies the boundaries of current knowledge, particularly regarding the existence of NCCRs for general singularities and the precise relationship between NCCRs and stringy invariants, setting the stage for future research.

In summary, Van den Bergh's overview synthesizes a decade of research (much of it with Špenko) to demonstrate that non-commutative algebras are not just abstract substitutes but powerful, computable tools that reveal the hidden structure of singular algebraic varieties, generalizing the classical theory of crepant resolutions.