🔢 mathematics

A note on Azumaya algebras and one-forms

This paper demonstrates that for a smooth variety possessing a non-closed global one-form, the Azumaya algebra of crystalline differential operators fails to split even on a degree one cover of the Frobenius twist, thereby resolving a question posed by Sasha Petrov.

Original authors: Siqing Zhang

Published 2026-05-29

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Siqing 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 you are a mathematician trying to solve a puzzle about the hidden shapes of the universe. In this specific paper, the author, Siqing Zhang, is looking at a very strange, high-dimensional landscape called a "smooth variety" (think of it as a perfectly smooth, multi-dimensional surface) that exists in a world where the rules of arithmetic are different—specifically, a world where numbers wrap around after a certain point (characteristic $p > 0$ ).

Here is the story of the paper, broken down into simple concepts:

1. The Mystery Object: The "Locked Box"

In this mathematical world, there is a special object called an Azumaya algebra. You can think of this as a complex, locked box that sits on top of a map of the landscape (specifically, on the "cotangent bundle," which is like a map showing all possible directions and speeds you could move).

Usually, when mathematicians look at this box in certain special spots (called "spectral varieties"), they find that the lock is easy to pick. The box "splits," meaning it opens up and reveals a simple, understandable structure inside. This has been true for many famous shapes, like curves or torus-like shapes (abelian varieties).

2. The Question: Can We Find a Box That Won't Open?

A mathematician named Sasha Petrov asked a simple but tricky question:

"Is there any shape (a smooth, proper variety) and any specific path on its map where this locked box refuses to open? Is there a place where the lock is truly stuck?"

For a long time, people weren't sure if such a stubborn box existed.

3. The Key to the Lock: The "One-Form"

Zhang's paper provides the answer: Yes, such a box exists.

The key to finding this stubborn box lies in something called a global one-form.

The Analogy: Imagine walking across a landscape and measuring the wind. A "closed" one-form is like a wind that flows in a perfect circle or a straight line without ever swirling or getting tangled. It's predictable.
The Twist: Zhang looks for a landscape where the wind does swirl or get tangled. In math terms, he looks for a shape where there is a "non-closed" global one-form. This is a wind pattern that cannot be smoothed out into a simple circle or line.

4. The Discovery

Zhang proves a clever rule:

If your landscape has this "swirly, tangled wind" (a non-closed one-form), then there is a specific path you can draw on the map.
If you take the "locked box" (the Azumaya algebra) and move it to sit exactly on top of this path, it will not open.
The lock remains stuck. The box stays complex and un-simplified.

This answers Sasha's question with a definitive "Yes."

5. Real-World Examples (The "Surfaces")

The paper doesn't just say this is possible in theory; it points to actual shapes where this happens:

Mumford's Surfaces: These are specific 2D shapes (surfaces) discovered decades ago that have this "swirly wind" property.
Takeda's Surfaces: Newer, slightly more complex surfaces that also have this property.
Building Bigger Shapes: You can even take these surfaces and combine them with other shapes (like adding a line or a plane to them) to create bigger, 3D or 4D shapes that still have the stubborn, un-openable box.

6. The Catch (The "Witt Vectors" Limit)

The paper ends with a small warning. These special shapes that have the "swirly wind" are very fragile. They are so weird that they cannot be "lifted" into a slightly different, more standard mathematical world (specifically, they don't work with a system called "second Witt vectors"). This is a technical way of saying: "These shapes are unique to this specific, weird type of math world and don't exist in the 'standard' versions of math we usually use."

Summary

In short, Siqing Zhang solved a puzzle by showing that if a mathematical shape has a specific kind of "tangled flow" (non-closed one-form), you can find a spot on its map where a complex mathematical structure (the Azumaya algebra) refuses to simplify. This proves that these "un-openable boxes" definitely exist in the world of positive characteristic geometry.

Technical Summary: A Note on Azumaya Algebras and One-Forms

Problem Statement
In the context of algebraic geometry over an algebraically closed field $k$ of characteristic $p > 0$ , the sheaf of crystalline differential operators on a smooth proper variety $X$ defines an Azumaya algebra, denoted $A_X$ , on the cotangent bundle $T^*X'$ of the Frobenius twist $X'$ . While it is known that $A_X$ splits when restricted to certain spectral varieties (a phenomenon utilized in characteristic- $p$ geometric Langlands and non-abelian Hodge theory), a specific question remained open regarding the existence of varieties where this splitting fails.

Sasha Petrov posed the following question (Question 1 in the paper): Does there exist a smooth proper variety $X$ and a smooth subvariety $Z \subset T^*X'$ , finite over $X'$ , such that the restriction $A_X|_Z$ does not split?

Methodology
The paper addresses this question by analyzing the relationship between the splitting of the Azumaya algebra and the cohomological properties of global one-forms on $X$ . The methodology proceeds through the following logical steps:

Milne's Exact Sequence: The author utilizes Milne's exact sequence of étale sheaves on $X'$ , which relates the multiplicative group, the Frobenius pushforward of the multiplicative group, closed one-forms, and the sheaf of all one-forms. This sequence involves the Cartier operator $C_X$ and the $d\log$ map.
Obstruction Homomorphism: By splicing the exact sequence, a connecting morphism is defined, yielding an obstruction homomorphism:
$ob_{X'}: H^0(X', \Omega^1_{X'/k}) \to H^2_{\text{ét}}(X', \mathcal{O}^*_{X'})$
This map measures the obstruction to splitting the algebra.
Gerbe Identification: Citing Ogus–Vologodsky, the paper identifies $ob_{X'}(\omega)$ with the class of the $\mathbb{G}_m$ -gerbe of line bundles on $X$ with integrable connection and $p$ -curvature $\omega$ . Crucially, this gerbe is also identified as the gerbe of splittings of $A_X$ restricted to the section $i_\omega: X' \to T^*X'$ corresponding to a global one-form $\omega$ .
Dimensional Analysis of Picard Schemes: To determine when $ob_{X'}$ is non-zero, the paper analyzes the rigidified Picard scheme of line bundles with integrable connection, denoted $G$ . Using Cartier descent and properties of the $p$ -curvature morphism $\psi: G \to V$ (where $V$ is the vector group of global one-forms), the author compares the dimensions of the image of $\psi$ and the space of global one-forms.

Key Contributions and Results
The paper establishes the following main results:

Lemma 2: For a smooth variety $X$ and a global one-form $\omega \in H^0(X', \Omega^1_{X'/k})$ , the class of the restricted Azumaya algebra $[i_\omega^* A_X]$ in $H^2_{\text{ét}}(X', \mathcal{O}^*_{X'})$ is exactly $ob_{X'}(\omega)$ . Consequently, if $ob_{X'}(\omega) \neq 0$ , the algebra does not split on the graph $\Gamma_\omega$ .
Proposition 3: If $X$ is a smooth proper connected variety such that not every global one-form is closed (i.e., $H^0(X, Z^1_{X/k}) \subsetneq H^0(X, \Omega^1_{X/k})$ ), then the obstruction map $ob_{X'}$ is non-zero. The proof demonstrates that the image of the $p$ -curvature morphism is strictly smaller than the space of all global one-forms due to dimensional constraints on the Picard scheme components.
Theorem 4 (Main Result): If $X$ possesses a non-closed global one-form, there exists a smooth subvariety $Z \subset T^*X'$ (specifically, the graph of a non-closed form) which is finite over $X'$ (in fact, an isomorphism) such that $A_X|_Z$ does not split.

Examples and Significance
The paper provides affirmative examples in every positive characteristic:

Surface Examples: It references Mumford's construction of smooth projective surfaces with non-closed global one-forms. It also notes that generalized Raynaud surfaces constructed by Takeda (originally for $p>2$ ) also satisfy this condition for $p=2$ .
Higher Dimensions: Products of these surfaces with projective spaces $X \times \mathbb{P}^n_k$ yield higher-dimensional examples.

Significance and Limitations
The paper answers Question 1 affirmatively, demonstrating that the splitting of the Azumaya algebra of crystalline differential operators is not guaranteed even on finite covers (specifically, degree one covers) of the base if the variety admits non-closed global one-forms.

The author notes a direct consequence regarding lifting: by the Deligne–Illusie theorem, if the Hodge-to-de Rham spectral sequence degenerates at $E_1$ , all global one-forms must be closed. Therefore, the varieties constructed in this paper (which have non-closed forms) cannot lift to the ring of second Witt vectors $W_2(k)$ . The paper remains modest, framing its contribution as a resolution to a specific question posed by Petrov without proposing broader applications or future experimental proposals beyond the immediate geometric context.