🔢 mathematics

Rickard's question on standard derived equivalences

This paper negatively answers Rickard's 1991 question by constructing a series of non-standard derived equivalences of algebras and providing a necessary and sufficient condition for such equivalences to be standard.

Original authors: Wei Hu, Changchang Xi, Jin Zhang

Published 2026-08-11

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Wei Hu, Changchang Xi, Jin Zhang

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 a vast, invisible library where every book represents a different way of building mathematical structures. In this library, called "algebra," mathematicians have developed a special set of rules to translate one book into another without losing any of the story. These translations are called "derived equivalences." Think of them as a magical dictionary that lets you speak two different mathematical languages fluently. For decades, mathematicians have wondered if every single one of these translations follows a strict, predictable recipe. This recipe, known as a "standard derived equivalence," is like a factory assembly line: you take a specific set of parts (complexes of modules), mix them in a specific way, and out pops the translation. It's reliable, clean, and easy to understand.

The big question, posed by a mathematician named Rickard in 1991, was simple but profound: Is every possible translation in this library built using that standard factory recipe? Or are there some translations that are "hand-crafted" in a weird, non-standard way that doesn't fit the assembly line? This matters because if everything is standard, we have a powerful, unified tool to solve problems and calculate hidden properties of these mathematical worlds. If there are exceptions, it means the library is more chaotic and mysterious than we thought, and our tools might need a major upgrade.

In this paper, Wei Hu, Changchang Xi, and Jin Zhang step into the library to investigate Rickard's question. They set out to see if the "assembly line" rule holds true for every single translation. After building a series of intricate mathematical traps and testing them in a very specific environment, they found the answer: No, the rule does not hold for everything. They successfully constructed a series of "non-standard" translations—mathematical transformations that are real and valid but simply cannot be built using the standard factory recipe.

To understand how they did this, imagine trying to build a bridge between two islands. The standard way is to use a pre-fabricated kit of beams and bolts. The authors found a way to build a bridge that works perfectly and connects the islands, but it's made of a strange, twisted material that doesn't fit the kit. They proved that these "twisted bridges" exist in infinite numbers, but only under very specific conditions. They discovered that these weird translations only appear when the mathematical "ground" they are built on has a special property related to the number 2 (specifically, in a world where the only numbers are 0 and 1, like a light switch being on or off).

The authors didn't just guess; they built a rigorous proof. They first created a checklist to determine if a translation is standard or not. Then, they used a special type of algebra (a "local commutative Frobenius algebra") over a field with two elements (0 and 1) to construct their counterexamples. They showed that in this specific setting, they could create a "pseudo-identity"—a translation that looks like it does nothing to the objects but secretly twists the connections between them in a way the standard recipe can't replicate.

The paper proves that Rickard's question has a negative answer: not all derived equivalences are standard. However, they also offer a glimmer of hope. They suspect that if you move to a mathematical world with a different number of basic elements (any number other than 2, like 3, 5, or 7), the standard recipe might actually work for everything. They haven't proven this yet, but they have a strong hunch. For now, they have opened a door to a new, weirder side of the mathematical library, showing us that the universe of algebraic translations is far more creative and less predictable than the "standard" model suggested.

Technical Summary: Rickard's Question on Standard Derived Equivalences

Problem Statement
The paper addresses a fundamental question in the representation theory of finite-dimensional algebras, originally posed by Rickard in 1991. Let $A$ and $B$ be finite-dimensional algebras over a field. A derived equivalence $F: D^b(A) \to D^b(B)$ is called standard if it is naturally equivalent to the derived tensor functor induced by a bounded complex of $B$ - $A$ -bimodules. While Rickard established that any derived equivalence agrees with a standard one on the subcategory of projective modules and that they are isomorphic as objects, it remained unknown whether they are naturally equivalent as triangle functors.

Question 1.1 (Rickard): Are all derived equivalences standard?

The paper investigates whether this question holds universally for finite-dimensional algebras. Previous work had confirmed a positive answer for specific classes of algebras, including hereditary algebras, triangular algebras, and derived-discrete algebras. However, the general case remained open.

Methodology and Characterization
The authors approach the problem by analyzing the structure of derived auto-equivalences, specifically focusing on pseudo-identities. A derived auto-equivalence $(F, \xi)$ is a pseudo-identity if it fixes all objects and all subcategories $A\text{-mod}[i]$ (or $A\text{-proj}[i]$ ).

Reduction to Pseudo-identities: The authors establish that Rickard's question holds for an algebra $A$ if and only if every derived auto-equivalence of $D^b(A)$ is standard. By decomposing any derived equivalence into a standard part and a pseudo-identity, the problem reduces to determining whether every pseudo-identity is naturally isomorphic to the identity functor.
Truncated Triangles: A key technical innovation is the introduction of truncated triangles. The authors define a truncated triangle $\Delta_i(X^\bullet)$ associated with a complex $X^\bullet$ and an integer $i$ . They prove a characterization theorem (Theorem 1.2) stating that Rickard's question holds for $A$ if and only if every pseudo-identity $(F, \text{id})$ that fixes all truncated triangles is naturally isomorphic to the identity. This significantly reduces the scope of verification, as one need only check pseudo-identities that preserve these specific structural triangles.
Construction of Counterexamples: To answer the question negatively, the authors construct a series of non-standard derived equivalences. They utilize commutative local Frobenius algebras over a field $k$ of characteristic 2. They define a natural transformation $\xi(a): \text{id}[1] \to [1]\text{id}$ using an element $a \in A$ satisfying specific trace conditions involving a Frobenius form $\omega$ . Specifically, they require $\omega(x^2) = \omega(ax)$ for all $x \in A$ and $a^2 \neq 0$ .

Key Results

Theorem 1.2 (Characterization): Rickard's question holds for a finite-dimensional algebra $A$ if and only if every pseudo-identity $(F, \text{id})$ on $K^b(A\text{-proj})$ that fixes all truncated triangles is naturally isomorphic to the identity triangle functor.
Theorem 1.3 (Negative Answer): There exist infinitely many non-standard derived equivalences between finite-dimensional algebras over fields.
Theorem 4.4: Under the conditions that $k = \mathbb{Z}/2\mathbb{Z}$ , $A$ is a finite-dimensional local commutative Frobenius algebra, and there exists $a \in A$ such that $a^2 \neq 0$ and $\omega(x^2) = \omega(ax)$ for all $x \in A$ , the functor $(\text{id}, \xi(a))$ is a non-standard derived equivalence.
Proposition 4.5: The property of admitting a non-standard derived equivalence is preserved under tensor products. If $A$ and $B$ satisfy the conditions of Theorem 4.4, then $A \otimes_k B$ also admits a non-standard derived equivalence.
Example 4.6: The authors provide explicit counterexamples using algebras of the form $k[x_1, \dots, x_m]/(x_1^{2n_1+1}, \dots, x_m^{2n_m+1})$ over a field of characteristic 2.

Significance and Claims
The paper claims to provide a definitive negative answer to Rickard's 1991 question, demonstrating that not all derived equivalences are standard. The significance of the work lies in:

Resolution of an Open Problem: It settles a long-standing question in the theory of derived categories of algebras.
New Characterization: It provides a necessary and sufficient condition (Theorem 1.2) for the validity of Rickard's question, shifting the focus to the behavior of pseudo-identities on truncated triangles. This characterization is used to give short proofs for known positive cases (e.g., hereditary and derived-discrete algebras).
Dependence on Characteristic: The counterexamples are constructed specifically over fields of characteristic 2. The authors explicitly conjecture that Rickard's question may hold for fields of characteristic $p \neq 2$ , noting that the specific algebraic condition $\omega(x^2) = \omega(ax)$ required for their construction forces the field to be $\mathbb{Z}/2\mathbb{Z}$ .

The paper does not claim to resolve the question for all characteristics but establishes that the general statement "all derived equivalences are standard" is false. The authors emphasize that their counterexamples arise from a specific obstruction related to the interaction between the Frobenius form and the square of elements in characteristic 2.

More like this