🔢 mathematics

The Davenport-Lewis-Schinzel problem on the reducibility of $f(X)-g(Y)$

This paper resolves the long-standing Davenport–Lewis–Schinzel problem on the reducibility of $f(X)-g(Y)$ , providing an almost-complete solution to the Hilbert–Siegel problem and yielding applications in arithmetic dynamics and functional equations.

Original authors: Angelot Behajaina, Joachim König, Danny Neftin

Published 2026-03-31

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Angelot Behajaina, Joachim König, Danny Neftin

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 two complex machines, let's call them Machine F and Machine G. You feed a number into Machine F, and it spits out a result. You do the same with Machine G.

The big question mathematicians have been asking since the 1950s is: Can you predict when the output of Machine F will exactly match the output of Machine G?

More specifically, if you write down the equation $F(X) = G(Y)$ , does this equation describe a single, smooth, connected path (like a single river), or does it break apart into several disconnected, smaller paths (like a river splitting into multiple tributaries)?

In math terms, if the equation "breaks apart," we say the polynomial is reducible. If it stays whole, it's irreducible.

For decades, mathematicians knew of a few specific, weird machines where the equation always broke apart. But they had no idea if there were any other hidden machines out there that did the same thing. It was like knowing that a specific type of lock only opens with a specific key, but wondering if there were other secret keys hidden in the universe that could open it too.

This paper, by Behajaina, König, and Neftin, is the ultimate lock-picking manual. They have finally solved the mystery. They proved that there are no other secret keys.

Here is how they did it, explained with some everyday analogies:

1. The "Lego" Analogy (Decomposing the Machines)

Think of these complex machines (polynomials) as being built out of smaller Lego bricks.

Some machines are just one giant, indivisible brick (called indecomposable).
Others are built by stacking smaller bricks on top of each other (called decomposable).

The authors realized that to understand if the whole machine breaks apart, you have to look at the smallest bricks at the very bottom of the stack. They proved that if the whole equation breaks apart, it's almost always because the two machines share a common bottom brick (a common factor).

2. The "DNA" of the Machines (Monodromy Groups)

To prove their point, the authors didn't just look at the machines; they looked at their DNA. In math, this DNA is called a "monodromy group." It's a way of describing how the machine's internal gears twist and turn when you spin the input knob.

The Solvable Case: For most machines, the DNA is "simple" (like a straight line or a simple loop). The authors used a clever trick: they showed that if the equation breaks apart, the DNA of the two machines must be so tightly intertwined that they are essentially the same machine, just dressed up differently.
The "Monster" Case: There were rumors of a "Cassels Monster"—a hypothetical, incredibly complex machine that could break apart the equation in a way no one had ever seen. The authors used a massive database of mathematical "creatures" (transitive groups) and showed that this Monster does not exist. The DNA simply doesn't allow for it.

3. The "Recipe" for Breaking Apart

After ruling out all the monsters and the unknowns, they found that there are only three specific recipes that allow two machines to break apart together:

The "Shared Parent" Recipe: Both machines were built using the exact same core component (a common factor).
The "Chebyshev Twins" Recipe: The machines are based on a special family of polynomials called Chebyshev polynomials (think of them as the "famous twins" of the math world). Specifically, one is the "positive" version and the other is the "negative" version.
The "Rare Species" Recipe: There are a handful of very specific, rare, high-degree machines (degrees 7, 11, 13, etc.) that act like special keys. These were known before, but the authors proved these are the only ones.

Why Does This Matter?

You might ask, "Who cares if a math equation breaks apart?"

This problem is actually a master key for many other fields:

Number Theory (The Hilbert-Siegel Problem): It helps us figure out which numbers can be "reached" by these machines. If the equation breaks apart, it means there are infinitely many "special" numbers that behave in a predictable way. This paper tells us exactly which numbers those are.
Arithmetic Dynamics (Stability): Imagine a machine that runs itself over and over again (like a computer program looping). This paper helps us predict when that loop will suddenly become unstable or break.
Functional Equations: It solves puzzles where we need to find functions that satisfy specific rules, which is crucial for understanding how complex systems evolve.

The Bottom Line

For 60 years, mathematicians were looking for a needle in a haystack, wondering if there were hidden needles they missed. Behajaina, König, and Neftin didn't just find the needles; they proved that the haystack only contains the needles we already knew about.

They closed the book on this specific problem, showing that the universe of these polynomial equations is much more orderly and predictable than anyone thought. If you want to know when two polynomial machines will "collide" (break apart), you just need to check if they fit one of the three simple recipes they discovered. If they don't, they will never break apart.

1. The Problem

The paper addresses the Davenport–Lewis–Schinzel (DLS) problem, originally posed in the late 1950s. The core question is:

For which polynomials $f, g \in \mathbb{C}[X]$ of degree $\ge 2$ is the bivariate polynomial $f(X) - g(Y) \in \mathbb{C}[X, Y]$ reducible?

Context and Trivial Cases:

Trivial Reducibility: $f(X) - g(Y)$ is trivially reducible if $f$ and $g$ share a common left composition factor $h$ (i.e., $f = h \circ f_1, g = h \circ g_1$ ).
Substitution Reducibility: If $f_1(X) - g_1(Y)$ is reducible, then $f_1(f_2(X)) - g_1(g_2(Y))$ is also reducible.
Minimally Reducible Pairs: A pair $(f, g)$ is called minimally reducible if $f(X) - g(Y)$ is reducible, but no such non-trivial decomposition exists (i.e., it is not of the form above).
Nontrivial Minimally Reducible Pairs: These are pairs that are not reducible via common left factors or linear substitutions. The DLS problem seeks to classify all such nontrivial pairs.

Historical Background:
Prior to this work, the classification was known for indecomposable polynomials (those that cannot be written as a composition of lower-degree polynomials). Fried (1986) showed that for indecomposable $f, g$ , reducibility implies specific degrees (7, 11, 13, 15, 21, 31) arising from outer automorphisms of monodromy groups (e.g., $PSL_3(2)$ ). The case for decomposable polynomials remained largely open, including the famous "(2, 3)-problem" (reducibility of $g(Y)^2 = p(f(X))$ where $p$ is cubic).

2. Methodology

The authors employ a sophisticated combination of Galois theory, monodromy group theory, and group-theoretic constraints on decomposable polynomials.

A. Monodromy and Decomposition

The reducibility of $f(X) - g(Y)$ is equivalent to the non-linear disjointness of the root fields of $f(X)-t$ and $g(Y)-t$ over $\mathbb{C}(t)$ . This implies their Galois closures coincide. The authors analyze the monodromy groups ($Mon(f)$) of the polynomials.

They decompose $f$ and $g$ into indecomposable factors: $f = f_1 \circ \dots \circ f_r$ and $g = g_1 \circ \dots \circ g_s$ .
They study the block kernels (kernels of the projection of the monodromy group of the composition to the monodromy group of the left factors).

B. The "Diagonality" Argument (Solvable Case)

The core innovation lies in handling the solvable monodromy case (where indecomposable factors are related to $X^p$ or Chebyshev polynomials $T_p$ ).

Kernel Constraints: They prove that for a minimally reducible pair, the kernel $K$ of the map $Mon(f) \to Mon(h)$ (where $h$ is a left factor) must be highly constrained.
Diagonal Subgroups: They show that $K$ must be a diagonal subgroup of a wreath product (meaning its projections to the components are injective) or isomorphic to $C_2 \times C_2$ .
Largeness Properties: Using results from their companion paper [BKN26], they establish "largeness" properties for kernels in compositions of polynomials. Specifically, they prove that for compositions of length $\ge 3$ (or specific length-2 cases), the kernel contains large subdirect powers (e.g., $C_q^d$ or augmentation ideals).
Contradiction: They demonstrate that the "largeness" of the kernel (derived from the composition structure) contradicts the "diagonality" requirement (derived from minimal reducibility) unless the degrees are very small and specific (e.g., degrees 2, 4, 8, 16).
Low-Degree Verification: For the remaining low-degree cases (degrees 8 and 16), they use computational verification via Magma (using the database of transitive groups) to rule out further examples.

C. The Nonsolvable Case

For polynomials with nonsolvable monodromy, the authors utilize a simpler argument based on the classification of indecomposable polynomials (which relies on the Classification of Finite Simple Groups, CFSG). They show that if $f$ and $g$ have nonsolvable monodromy, they must essentially be indecomposable (or have solvable "tails"), reducing the problem to the known classification of indecomposable pairs.

3. Key Contributions and Results

Main Theorem (Theorem 1.1 & 6.1)

The paper provides a complete classification of nontrivial minimally reducible pairs $(f, g)$ over a field of characteristic 0. $f(X) - g(Y)$ is reducible if and only if:

Common Left Factor: $f$ and $g$ share a common left composition factor $h$ of degree $\ge 2$ .
Chebyshev Exception: $f = (\mu \circ T_4) \circ f_1$ and $g = (\mu \circ (-T_4)) \circ g_1$ (where $T_4$ is the 4th Chebyshev polynomial).
Dickson Exception: $f = (\mu \circ D_{4, \alpha}) \circ f_1$ and $g = (\mu \circ (-\frac{1}{4}D_{4, 2\alpha})) \circ g_1$ (Dickson polynomials).
Indecomposable Exceptions: $f$ and $g$ are composed of indecomposable polynomials of degrees 7, 11, 13, 15, 21, or 31 (specific pairs listed in [CNC99]), possibly composed with linear maps and a common left factor.

Significance: This result rules out the existence of a "Cassels monster" (a hypothetical minimally reducible pair with $\ge 4$ branch points) and solves the long-standing (2, 3)-problem (and generally the $(m, n)$ -problem) for all $m, n$ .

Application 1: Hilbert–Siegel Problem (Theorem 1.2)

The authors solve the Hilbert–Siegel problem for polynomials $f \in \mathbb{Q}[X]$ . This problem asks for the set of integers $a$ such that the fiber $f^{-1}(a)$ is reducible over $\mathbb{Q}$ .

Result: For $f$ that does not factor through an indecomposable of degree 2 or 4, the set of exceptional integers is exactly the union of the value sets of its left composition factors (plus a finite set).
This provides an "almost-complete" solution, with the only remaining ambiguity being polynomials factoring through degree 2 or 4 solvable components.

Application 2: Stability in Arithmetic Dynamics (Corollary 1.3)

They address the stability of iterates. A polynomial $f$ is stable if $f^{\circ n}(X) - a$ is irreducible for all $n$ .

Result: If $f$ does not factor through a degree 2 or 4 polynomial, then the set of integers $a$ for which $f^{\circ n}$ becomes newly reducible is finite. This resolves a major open question regarding the finiteness of "newly reducible" fibers in iterations.

Application 3: Functional Equations (Corollary 1.4)

They classify solutions to $f(X) = g(Y)$ in rational functions $X(z), Y(z) \in \mathbb{C}(z)$ .

Result: They explicitly describe the structure of $f$ and $g$ that admit such solutions, linking them to the Chebyshev/Dickson exceptions and the specific indecomposable pairs. This resolves the "reducible case" of the functional equation, which was previously poorly understood.

4. Significance

Completeness: This is the first complete solution to the DLS problem for decomposable polynomials, a case that had resisted solution for decades due to the complexity of monodromy groups in compositions.
Methodological Breakthrough: The introduction of "diagonality" constraints on block kernels combined with "largeness" properties of solvable monodromy groups provides a powerful new toolkit for studying reducibility in arithmetic geometry.
Unification: The paper unifies several disparate areas: Hilbert's Irreducibility Theorem, arithmetic dynamics (stability of iterates), and functional equations.
Resolution of Open Problems: It definitively answers Fried's (2, 3)-problem and the question of whether minimally reducible pairs can have unbounded degree (the answer is no).

In summary, Behajaina, König, and Neftin have closed a 60-year-old chapter in polynomial reducibility, providing a definitive classification that relies on deep structural properties of monodromy groups and has immediate, profound implications for number theory and dynamical systems.