🔢 mathematics

On $p$ -adic solubility of $Ax^\ell + By^m + Cz^n = 0$

This paper establishes that for generalized Fermat equations $Ax^\ell + By^m + Cz^n = 0$ , the probability of $p$ -adic solubility for almost all primes is governed by a rational function of specific greatest common divisors, leading to the conclusion that the proportion of such equations with solutions everywhere locally is positive if the exponents are pairwise coprime, but zero otherwise.

Original authors: Santiago Arango-Piñeros, Christopher Keyes, Andrew Kobin

Published 2026-08-20

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Santiago Arango-Piñeros, Christopher Keyes, Andrew Kobin

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

Mathematicians have long been fascinated by a specific type of puzzle involving whole numbers. Imagine an equation where you add together three terms, each consisting of a number multiplied by a variable raised to a power, and the total must equal zero. The challenge is to find whole number solutions for the variables. This is a classic problem in number theory, a field dedicated to understanding the hidden properties of integers. For centuries, mathematicians have known that for such an equation to have a solution in whole numbers, it must first pass a series of local tests. Think of these tests as checking if the equation works in different "neighborhoods" of the number system. One neighborhood is the real numbers, which we use for everyday measurements. Another set of neighborhoods involves prime numbers, where the rules of arithmetic shift slightly. If an equation fails to have a solution in even one of these prime-based neighborhoods, it is impossible for it to have a solution in whole numbers. This principle is a necessary condition, a gatekeeper that any potential solution must pass.

The question that has puzzled researchers is how often these equations actually pass all the local tests. If an equation passes every single local test, does it guarantee a whole number solution? Or are there equations that pass every local test but still fail to have a whole number solution? To answer this, one must understand the likelihood of an equation passing these local tests in the first place. This is where probability enters the picture. Instead of looking at one specific equation, mathematicians look at the entire family of these equations, varying the coefficients and asking: what is the chance that a randomly chosen equation will have a solution in the neighborhood of a specific prime number?

In a recent study, researchers Christopher Keyes and Andrew Kobin, with an appendix by Santiago Arango-Piñeros, tackled this question for a broad class of these equations. They focused on equations where the variables are raised to different powers, a setup known as a generalized Fermat equation. Their goal was to calculate the precise probability that such an equation has a solution in the p-adic numbers, a mathematical structure that captures the behavior of integers relative to a specific prime number. They discovered that for almost all prime numbers, this probability is not a random guess but follows a strict, predictable pattern. The probability can be described by a specific rational function, which is essentially a fraction made of polynomials. Remarkably, the exact form of this fraction depends only on the greatest common divisors of the powers in the equation and the prime number being tested. This means that if you know the exponents and the prime, you can calculate the exact chance of a local solution without needing to test every single possibility.

The researchers found that the behavior of these probabilities changes dramatically based on the relationship between the exponents. When the three exponents are pairwise coprime—meaning no two of them share a common factor other than one—the probability of having a local solution is always positive. In this scenario, there is a genuine, non-zero chance that a randomly chosen equation will pass the local test for every prime number. However, the study reveals a stark contrast when the exponents are not pairwise coprime. If any two exponents share a common factor, the proportion of equations that pass every local test drops to zero. In other words, if the exponents share a factor, it becomes statistically impossible to find a random equation that is locally soluble everywhere. This finding effectively rules out the possibility of such equations having a high density of local solutions.

To reach these conclusions, the team developed a method to break down the complex problem into smaller, manageable pieces. They analyzed the equation by looking at the divisibility of its coefficients by the prime number in question. They calculated the probability of a solution existing under various conditions, such as when the coefficients are not divisible by the prime, or when they are divisible by it once, twice, or more. By linking these conditional probabilities together, they constructed a complete picture of the overall likelihood. They proved that for the vast majority of primes, the answer is determined by a finite set of rational functions. They also provided a way to compute these functions explicitly, offering a clear algorithm that can be implemented on a computer. The paper includes detailed examples, such as when the exponents are 3, 3, and 2, showing exactly how the probability shifts depending on whether the prime number leaves a remainder of 1 or 2 when divided by 3.

The implications of these findings extend beyond just calculating numbers. The study connects the local solubility of these equations to the broader question of whether they have whole number solutions. When the exponents are pairwise coprime, the positive probability of local solubility suggests that there is a substantial set of equations that might have whole number solutions. The researchers estimate that for the specific case of exponents 2, 3, and 5, about 78.2 percent of all such equations are locally soluble everywhere. This is a significant finding because it provides a concrete lower bound for the density of equations that satisfy the necessary conditions for having a whole number solution. Conversely, for cases where the exponents share factors, the zero density result implies that such equations are extremely rare in terms of passing all local tests.

The work also touches on the geometry of these equations. The researchers viewed the solutions not just as numbers, but as points on a geometric object known as a stacky curve. This perspective allowed them to use tools from algebraic geometry to count the solutions more effectively. They showed that the behavior of these curves over different prime numbers is governed by how the prime splits in a specific type of number field extension. This geometric interpretation helps explain why the probabilities depend on the greatest common divisors of the exponents and the prime. The study confirms that the local solubility of these equations is a well-understood phenomenon that can be described with precision, provided one knows the relationship between the exponents.

Ultimately, this research provides a definitive map of the landscape of local solubility for generalized Fermat equations. It clarifies that the ability of these equations to have solutions in prime-based neighborhoods is not a matter of chance but a matter of arithmetic structure. When the exponents are independent, the door to local solubility remains open, allowing for a positive proportion of equations to pass. When they are linked by common factors, that door closes almost entirely. The paper offers a rigorous, algorithmic way to determine the exact probability for any given set of exponents and primes, turning a complex theoretical question into a calculable reality. This clarity helps mathematicians better understand the distribution of solutions and sets the stage for further investigations into the global solubility of these equations, bridging the gap between local behavior and the existence of whole number solutions.

Technical Summary: On $p$ -adic Solubility of $Ax^\ell + By^m + Cz^n = 0$

Problem Statement
The paper investigates the local solubility of the generalized Fermat equation $Ax^\ell + By^m + Cz^n = 0$ over the $p$ -adic integers $\mathbb{Z}_p$ . Specifically, the authors study the existence of primitive solutions, defined as triples $(x, y, z) \in \mathbb{Z}_p^3 \setminus (p\mathbb{Z}_p)^3$ . A necessary condition for the existence of primitive integer solutions is the existence of primitive $p$ -adic solutions for all primes $p$ . The central goal is to determine the probability $\rho_{\ell,m,n}(p)$ that a random triple of coefficients $(A, B, C) \in \mathbb{Z}_p^3$ yields an equation with a primitive $p$ -adic solution. Furthermore, the authors analyze the natural density of such equations that are everywhere locally soluble (solvable in $\mathbb{R}$ and $\mathbb{Z}_p$ for all $p$ ) within the space of all such equations, ordered by height.

Methodology
The approach combines arithmetic geometry, probability theory on $p$ -adic fields, and combinatorial analysis of valuations.

Geometric Framework: The authors interpret primitive solutions as points on a relative stacky curve $X_{A,B,C}$ inside a weighted projective stack $\mathbb{P}(w_0, w_1, w_\infty)$ , where weights are derived from the least common multiple of the exponents. This geometric perspective allows the use of tools like Hensel's Lemma and the Hasse–Weil bound on the reduction of the curve modulo $p$ .
Probabilistic Decomposition: The total probability $\rho_{\ell,m,n}(p)$ is decomposed into conditional probabilities based on the $p$ -adic valuations of the coefficients $A, B, C$ . The authors define auxiliary probabilities (e.g., $\rho_0, \rho_A, \rho_{AB}$ ) corresponding to cases where $p$ divides none, one, or two of the coefficients.
Recursive Relations: By analyzing the lifting of solutions from $\mathbb{F}_p$ to $\mathbb{Z}_p$ and utilizing lemmas regarding the solvability of diagonal forms modulo $p$ , the authors establish a system of linear equations relating these auxiliary probabilities. These relations depend on the greatest common divisors of the exponents and the prime $p$ (specifically $\gcd(p-1, \ell, m)$ , etc.).
Linear Algebra and Rational Functions: The system of equations is shown to be invertible for sufficiently large primes. This invertibility allows the authors to express the conditional probabilities, and consequently $\rho_{\ell,m,n}(p)$ , as rational functions in $p$ . The specific form of these functions depends only on the values of $\gcd(p-1, g_{\ell m})$ , $\gcd(p-1, g_{\ell n})$ , and $\gcd(p-1, g_{mn})$ , where $g_{xy} = \gcd(x,y)$ .
Asymptotic Density: To determine the global density of everywhere locally soluble equations, the authors apply the sieve of Ekedahl (following the framework of Browning, Brüdern, and Loughran). This involves verifying that the local densities satisfy specific convergence conditions (specifically, that $1 - \rho(p) = O(1/p^2)$ ) and analyzing the product of local densities over all primes.

Key Contributions and Results

Explicit Rational Functions (Theorem 1.2): For all but finitely many primes $p$ , the probability $\rho_{\ell,m,n}(p)$ is given by an explicit rational function $R_{i,j,k}(p)$ . The specific function depends only on the triple $(i, j, k)$ , where $i = \gcd(p-1, \gcd(\ell, m))$ , $j = \gcd(p-1, \gcd(\ell, n))$ , and $k = \gcd(p-1, \gcd(m, n))$ . The paper provides an effective algorithm (implemented in Sage code) to compute these functions for any given exponents.
Global Density Dichotomy (Theorem 1.5): The paper establishes a sharp dichotomy regarding the proportion of equations that are everywhere locally soluble:
- Pairwise Coprime Exponents: If $\ell, m, n$ are pairwise coprime, the natural density of everywhere locally soluble equations is positive. It is given by the convergent product $\prod_p \rho_{\ell,m,n}(p)$ .
- Non-Pairwise Coprime Exponents: If the exponents are not pairwise coprime, the natural density is zero.
Explicit Examples: The authors provide detailed computations for various cases, including:
- $(\ell, m, n) = (3, 3, 2)$ , yielding different rational functions for $p \equiv 1 \pmod 3$ and $p \equiv 2 \pmod 3$ .
- $(\ell, m, n) = (2, 2, n)$ , recovering and extending previous results for diagonal hypersurfaces.
- $(\ell, m, n) = (2, 3, 5)$ , where the authors estimate the density to be approximately $78.2\%$ .
Generalization to Number Fields: The results are extended to arbitrary number fields $K$ . The local probability depends on the splitting behavior of prime ideals in the cyclotomic extension $K(\zeta)/K$ , where $\zeta$ is a root of unity determined by the exponents.

Significance and Claims
The paper claims to provide a complete and effective description of the local solubility probability for generalized Fermat equations, moving beyond specific cases to a general framework dependent on the arithmetic of the exponents and the prime $p$ .

The primary significance lies in the dichotomy of global solubility. The authors demonstrate that the pairwise coprimality of the exponents is the critical factor determining whether a positive proportion of these equations satisfy the Hasse principle locally. When exponents share common factors, the local obstructions become too frequent (density zero), whereas pairwise coprimality ensures a positive density of locally soluble equations.

The work connects the arithmetic of Diophantine equations to the geometry of stacky curves and fibrations, offering a concrete realization of local solubility probabilities in terms of rational functions. The authors note that their results recover known findings for diagonal hypersurfaces (e.g., Hirakawa and Kanamura) and extend recent work on specific cases (e.g., Duque-Rosero et al.). They position Theorem 1.5 as a partial answer to local solubility questions for fibrations of stacky curves, providing a foundation for further study of global solubility in these families.

Technical Summary: On ppp-adic Solubility of Axℓ+Bym+Czn=0Ax^\ell + By^m + Cz^n = 0Axℓ+Bym+Czn=0

More like this

Technical Summary: On $p$ -adic Solubility of $Ax^\ell + By^m + Cz^n = 0$