🔢 mathematics

Dynamical Mordell-Lang problem for automorphisms of surfaces in positive characteristic

This paper establishes the solution to the dynamical Mordell-Lang problem for automorphisms of projective surfaces in positive characteristic.

Original authors: Junyi Xie, She Yang

Published 2026-05-04

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Junyi Xie, She Yang

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 have a magical, infinite dance floor (a mathematical "surface") and a dancer (an "automorphism") who moves around it according to a strict, unchanging set of rules. Every time the dancer moves, they take a step forward in time.

Now, imagine you paint a specific pattern on the floor (a "subvariety"). You want to know: When will the dancer step exactly on that painted pattern?

In the world of mathematics, this is called the Dynamical Mordell–Lang problem. It asks if the list of times the dancer hits the target follows a simple, predictable pattern (like counting by 2s: 2, 4, 6, 8...) or if the list is a chaotic mess with no rhyme or reason.

The Setting: A Different Kind of Math

Most of the time, mathematicians study this dance on a floor made of "standard" numbers (like real or complex numbers). In that world, the answer is usually "yes, it's predictable."

However, this paper by Junyi Xie and She Yang studies the dance on a floor made of positive characteristic numbers. Think of this as a dance floor with a weird, repeating rhythm (like a clock that resets every $p$ hours). In this specific rhythm, the dancer's path can get incredibly messy and complicated. For a long time, mathematicians thought predicting the dancer's steps on this weird floor was impossible for certain types of dancers.

The Breakthrough: Solving the "Surface" Puzzle

The authors solved this problem specifically for surfaces (2D dance floors). They proved that even on this weird, rhythmic floor, the dancer's return times are not chaotic. They are almost always a simple list of patterns, with just one tiny, specific exception.

Here is the simple breakdown of their findings:

1. The Three Types of Dancers
The authors realized that the dancer's behavior depends on their "energy" or speed. They categorized them into three types:

The Bounded Dancer (Elliptic): This dancer moves in a small, contained circle. Their path is very predictable.
The Hyperbolic Dancer: This dancer speeds up exponentially, flying off to infinity very fast. Their path is also predictable in a specific way.
The Parabolic Dancer: This is the tricky one. They move at a steady, medium pace that doesn't speed up or slow down, but they don't stay in a small circle either. This is the main focus of the paper.

2. The Secret Weapon: The "Fibred" Floor
For the tricky "Parabolic" dancers, the authors used a famous theorem (Gizatullin's Theorem) to realize something amazing: The floor isn't just a flat sheet; it's actually a stack of loops (like a stack of pancakes or a spiral staircase).

The dancer moves along these loops in a very regular way. By realizing the floor has this "stacked" structure, the authors could use a mathematical tool called "Height" (think of it as a ruler measuring how far the dancer has traveled).

3. The Two Speeds Trick
To prove the pattern, they needed to find two different "speeds" of growth.

Imagine the dancer's position on the stack of loops grows like a square ( $n^2$ ).
But their position along the loop grows like a cube ( $n^3$ ).
Because these two speeds grow at different rates, they can't accidentally line up with the painted pattern forever unless they are following a strict arithmetic rule. This "mismatch" in growth rates is what forces the pattern to be simple.

The Result: What the Return List Looks Like

The authors proved that the list of times the dancer hits the target is always a finite union of arithmetic progressions.

Normal Pattern: Like counting by 5s: $5, 10, 15, 20...$
The "Weird" Exception: In this positive characteristic world, there is one special type of pattern that can appear, but only if the dancer is the "Bounded" type. This pattern looks like a geometric progression involving powers of the clock's reset number ( $p$ $p$ ).
- Analogy: Instead of $5, 10, 15$, you might see a list like $1, p, p^2, p^3...$ (where $p$ is the size of the clock cycle).

Crucially, for the "Parabolic" dancers (the main focus), this weird "power of $p$ " pattern never happens. Their return times are always simple arithmetic progressions.

The "Converse" Question

The paper also asks the reverse question: "Can we build a dance floor and a dancer to create any pattern we want?"
They found that while you can create simple arithmetic patterns, you cannot create just any random list. For example, you cannot create a list that is just "powers of $p$ " ( $p, p^2, p^3...$ ) unless the dancer is the specific "Bounded" type. The patterns must be "complete" and follow the rules of the dance floor's geometry.

Summary

In simple terms, Xie and Yang showed that even in a mathematical world where things usually get chaotic and unpredictable, the movement of a dancer on a 2D surface is strictly ordered. If you know the rules of the dance, you can predict exactly when the dancer will step on a specific spot, and that prediction will always follow a simple, repeating rhythm.

1. Problem Statement

The paper addresses the Dynamical Mordell–Lang (DML) conjecture in the context of positive characteristic ( $p > 0$ ).

The Conjecture: For a variety $X$ over a field $K$ and an endomorphism $f$ , the "return set" $S = \{n \in \mathbb{Z} \mid f^n(x) \in V(K)\}$ (where $x \in X(K)$ and $V \subseteq X$ is a closed subvariety) should be a finite union of arithmetic progressions.
The Challenge: While the conjecture holds in characteristic 0 for many cases (e.g., étale endomorphisms, $\mathbb{A}^2$ ), it fails in positive characteristic. In characteristic $p$ , return sets can be much more complex, involving sets of the form $\{c_0 + c_1 \frac{q^m}{q-1} \mid m \in \mathbb{N}\}$ where $q$ is a power of $p$ .
Specific Goal: The authors aim to solve the DML problem for automorphisms of projective surfaces in positive characteristic, characterizing the exact structure of these return sets.

2. Methodology

The authors employ a combination of algebraic geometry, dynamical systems theory, and arithmetic height machinery. The proof strategy involves classifying surface automorphisms based on their dynamical degrees and applying specific tools to each class.

A. Classification of Automorphisms

The authors classify the automorphism $f$ of a surface $X$ based on its dynamical degree $\lambda_1(f)$ and the growth of the intersection numbers $((f^n)^*L \cdot L)$ for a big and nef line bundle $L$ :

Bounded-degree (Elliptic): The sequence is bounded.
Parabolic: $\lambda_1(f) = 1$ , but the sequence is unbounded.
Hyperbolic: $\lambda_1(f) > 1$ .

Note: Cases (i) and (iii) were previously resolved in the literature. The core contribution of this paper is the proof for Case (ii): Parabolic automorphisms.

B. Key Tools

Gizatullin's Theorem: A crucial structural result stating that any parabolic automorphism of a smooth projective surface preserves a fibration $\pi: X \to C$ (where $C$ is a curve). This allows the dynamics to be analyzed fiber-wise.
Weil Height Machinery: The authors utilize height functions to measure the "size" of points. They establish that for parabolic automorphisms, different components of the orbit grow at different polynomial rates.
- By finding two different speeds of growth (e.g., linear vs. quadratic, or quadratic vs. quartic), they can prove that the intersection with a subvariety must be finite unless the orbit is trapped in a specific algebraic structure.
Albanese Map Arguments: For surfaces where the Albanese map is surjective, the problem is reduced to the study of automorphisms on abelian varieties or elliptic curves.
Reduction to Abelian Surfaces: The case of abelian surfaces is treated explicitly by decomposing the automorphism into translations and group automorphisms, utilizing the specific structure of elliptic curves.

3. Key Contributions and Results

Main Theorem (Theorem 1.1)

For a projective surface $X$ , an automorphism $f$ , a closed subvariety $V$ , and a point $x$ , the return set $S = \{n \in \mathbb{Z} \mid f^n(x) \in V(K)\}$ is:

A finite union of arithmetic progressions.
Plus finitely many sets of the form $\{c_0 + c_1 \frac{q^m}{q-1} \mid m \in \mathbb{N}\}$ , where $q$ is a power of $p$ and $c_0, c_1$ are integers satisfying $q-1 \mid c_0 + c_1$ .
Crucial Condition: These "p-sets" (the second type) only appear if $f$ is of bounded-degree. If $f$ is parabolic or hyperbolic, the return set is strictly a finite union of arithmetic progressions.

Theorem 1.3 (Parabolic Case)

The authors prove that if $f$ is a parabolic automorphism of a smooth projective surface, the return set is exactly a finite union of arithmetic progressions (no "p-sets" appear).

Proof Strategy: They use the Gizatullin fibration $\pi: X \to C$ $π : X \to C$ .
- If the base curve $C$ has genus $>1$ , orbits are not dense.
- If $C$ is $\mathbb{P}^1$ or an elliptic curve, they construct height functions showing that the growth of points along the fiber and along the base curve are incompatible with infinite intersection unless the orbit is periodic or the intersection is finite.

Corollary 1.2 (Higher Dimensions)

The result extends to cohomologically hyperbolic automorphisms of projective threefolds when intersecting with a closed subcurve, reducing the problem to the surface case via the hyperbolicity condition.

Theorem 6.1 (Converse Problem)

The authors determine which sets can actually be realized as return sets.

If the orbit is not dense, the set is a finite union of arithmetic progressions.
If the orbit is dense, the set can be a finite union of arithmetic progressions plus the specific "p-sets" described above.
They prove that every set of this form can be realized by a surface automorphism, showing the classification is complete.

4. Significance

Resolution of a Hard Case: The DML problem in positive characteristic is notoriously difficult due to the existence of "p-sets." This paper completely resolves the problem for surface automorphisms, a fundamental class of dynamical systems.
Structural Insight: It establishes a sharp dichotomy: "p-sets" (the exotic return sets unique to positive characteristic) are strictly associated with bounded-degree (elliptic) dynamics. For parabolic and hyperbolic dynamics, the behavior reverts to the "classical" arithmetic progression structure, despite the positive characteristic setting.
Methodological Advancement: The paper successfully adapts the height argument to zero-entropy (parabolic) systems. Typically, height arguments require exponential growth (hyperbolic) to separate orbits. The authors overcome this by leveraging the Gizatullin fibration to find distinct polynomial growth rates in different directions of the surface.
Completeness: By solving the converse problem, the authors show that the classification of return sets is exhaustive; no other types of sets can occur for surface automorphisms.

Summary of Proof Flow

Reduction: Reduce to smooth surfaces and classify $f$ as Elliptic, Parabolic, or Hyperbolic.
Known Cases: Cite existing literature for Elliptic and Hyperbolic cases.
Parabolic Case (Core):
- Apply Gizatullin's Theorem to get a fibration $\pi: X \to C$ .
- Analyze the base curve $C$ $C$ :
  - If $g(C) > 1$ , trivial.
  - If $g(C) = 0$ or $1$, use Height Machinery.
- Construct height functions $h_L$ on $X$ and $h_H$ on $C$ .
- Show that $h_L(f^n(x))$ grows quadratically (or quartically in specific abelian cases) while $h_H(\pi(f^n(x)))$ grows linearly (or quadratically).
- Use the Northcott property (finite number of points with bounded height) to prove that the intersection with any subvariety $V$ must be finite unless the orbit is trapped in a specific algebraic structure (which implies arithmetic progressions).
Converse: Construct explicit examples on $\mathbb{P}^2$ and semi-abelian varieties to show all predicted forms are achievable.