🔢 mathematics

Sums of three powerful numbers

This paper establishes upper bounds with power-saving over the trivial bound for the number of primitive positive Campana points on a specific orbifold, which correspond to solutions of $a+b=c$ where $a$ , $b$ , and $c$ are $p$ -full, $q$ -full, and $r$ -full numbers respectively, by utilizing uniform estimates for primitive integral points on generalized Fermat surfaces.

Original authors: Tim Browning, Matteo Verzobio

Published 2026-08-26

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Tim Browning, Matteo Verzobio

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

Mathematics has long been fascinated by the hidden architecture of whole numbers, particularly how they can be built from smaller pieces. One classic puzzle asks how many ways a number can be written as the sum of two other numbers, a question that leads to deep theories about the distribution of primes and the nature of equations. A more recent and subtle variation of this puzzle involves "powerful" numbers. These are special integers that are built from prime factors in a very specific way: if a prime number divides them, it must divide them at least a certain number of times. For instance, a number that is "square-full" cannot be divided by a prime just once; that prime must appear at least twice in its makeup. This restriction makes powerful numbers much rarer than ordinary integers, creating a sparse landscape where finding solutions to simple equations becomes a difficult hunt.

The question at the heart of this research is how many solutions exist when you add two powerful numbers together to get a third, where each number in the equation follows a different rule for how "powerful" it must be. If the rules are too strict, mathematicians suspect that there might be only a handful of solutions, or perhaps none at all, once the numbers get large enough. This idea connects to a broader set of conjectures about the behavior of numbers, suggesting that if the rules are tight enough, the universe of solutions should be finite. However, proving that a list of solutions is truly finite is often beyond the reach of current mathematical tools. Instead, researchers focus on counting how many solutions exist up to a certain size, hoping to show that the number of solutions grows so slowly that it effectively stops, or at least grows much slower than a simple guess would predict.

In this work, the authors tackle the problem of counting these specific sums of three powerful numbers. They consider a scenario where the first number must be divisible by a prime at least $p$ times, the second at least $q$ times, and the resulting sum at least $r$ times. The researchers are interested in the case where these rules are strict enough that the total "weight" of the rules suggests there should be very few solutions. They begin with a simple, obvious estimate: if you just look at how many powerful numbers exist up to a certain size, you can guess an upper limit for how many sums you might find. This guess, known as the trivial bound, is based on the idea that you are simply picking from two separate lists of numbers. The goal of the paper is to prove that the actual number of solutions is significantly smaller than this simple guess, a result known as a power saving.

To achieve this, the authors translate the problem of counting sums into a problem of counting points on a geometric surface. Imagine a three-dimensional space where every point represents a potential solution. The equation linking the three numbers defines a specific shape within this space. The researchers need to count how many points with whole-number coordinates lie on this shape, but with a catch: the points must be "primitive," meaning the numbers share no common factors. The shape they are studying is a generalization of a famous type of surface called a Fermat surface. The difficulty lies in the fact that the surface is "lopsided," meaning the rules for the three numbers are different, and the researchers need a method that works uniformly regardless of the specific coefficients in the equation.

The authors develop a new, uniform method to count these points by breaking the problem down into smaller, more manageable slices. They use a technique that involves projecting the three-dimensional points onto two-dimensional curves and then analyzing the properties of those curves. A key part of their strategy involves determining whether certain polynomials, which describe the geometry of the problem, can be broken down into simpler pieces. If a polynomial cannot be broken down, it behaves in a very rigid way that limits the number of solutions. If it can be broken down, the researchers show that the geometry of the resulting pieces still forces the number of solutions to be small. By carefully combining these geometric insights with advanced counting techniques, they are able to prove that the number of solutions is indeed much smaller than the trivial guess.

The main result of the paper is a precise upper bound on the number of solutions. They show that for a wide range of rules, the number of solutions grows at a rate that is strictly slower than the simple guess would suggest. This difference, though it might seem small in the abstract, is mathematically significant because it represents a "power saving," meaning the number of solutions is suppressed by a factor that grows as the numbers get larger. The authors demonstrate that this saving holds true even when the rules for the three numbers are different, which is the most difficult case. They provide a formula that tells you exactly how much smaller the number of solutions is, depending on the specific rules chosen. In many cases, including when the rules are identical for all three numbers, their method proves that the number of solutions is far fewer than previously known.

This work does not claim to prove that there are only a finite number of solutions for every possible set of rules, a result that would require tools that do not yet exist. Instead, it provides strong quantitative evidence that the number of solutions is very small. The authors show that their method works particularly well when the rules for the three numbers are similar in size, but it also succeeds in cases where the rules are quite different. By establishing these tighter bounds, they move the field closer to understanding the true nature of these powerful numbers. Their findings suggest that the universe of solutions is indeed sparse, consistent with the broader mathematical intuition that such restrictive equations should have very few answers. The paper stands as a rigorous demonstration that even in a landscape of infinite numbers, the strict rules of powerful numbers create a structure so tight that solutions become exceptionally rare.

Technical Summary: Sums of Three Powerful Numbers

Problem Statement
This paper investigates the arithmetic of Campana points on the projective line $\mathbb{P}^1$ equipped with a specific $\mathbb{Q}$ -divisor $\Delta_{p,q,r} = (1-1/p)[0] + (1-1/q)[1] + (1-1/r)[\infty]$ , where $p, q, r \geqslant 2$ . Primitive Campana points on this orbifold correspond bijectively to primitive positive integer solutions $(a, b, c)$ of the equation $a + b = c$ where $a$ is $p$ -full, $b$ is $q$ -full, and $c$ is $r$ -full. An integer $n$ is defined as $m$ -full if for every prime $\ell$ , the valuation $v_\ell(n)$ is either $0$ or $\geqslant m$ .

The authors focus on the "log-general type" range, defined by the condition $1/p + 1/q + 1/r < 1$ . In this regime, conjectures by Campana and the $abc$ conjecture suggest that the set of such points is not Zariski dense, implying a finiteness result ( $N(B) = O(1)$ ) for the number of solutions with height at most $B$ . However, since a proof of finiteness is currently out of reach, the paper aims to establish quantitative upper bounds for the counting function $N(B)$ that improve upon the trivial bound.

Methodology
The core of the analysis relies on a uniform estimate for primitive integral points on generalized Fermat surfaces within rectangular boxes. The authors define the surface by the equation $f(x, y, z) = a_1 x^p + a_2 y^q + a_3 z^r = 0$ and seek to bound the number of primitive solutions in the box $|x| \leqslant X, |y| \leqslant Y, |z| \leqslant Z$ .

The proof strategy combines two primary analytic tools:

Salberger's Determinant Method: This reduces the problem of counting points on the surface to counting points on curves defined by the intersection of the surface with auxiliary polynomials $f_j(x, y, z)$ .
Function Field Diophantine Geometry: The authors analyze the irreducibility of the polynomial $F(T) = a_3 T^r + a_1 x^p + a_2 y^q$ $F (T) = a_{3} T^{r} + a_{1} x^{p} + a_{2} y^{q}$ over the function field $K = \mathbb{Q}(C)$ $K = Q (C)$ , where $C$ $C$ is the curve defined by an auxiliary polynomial $c(x, y) = 0$ $c (x, y) = 0$ .
- They utilize the Brownawell–Masser theorem (the function field analogue of the $abc$ conjecture) and the Mason–Stothers theorem to bound the degrees of polynomials involved in vanishing sums of $S$ -units.
- Capelli's criterion is employed to determine when binomials are reducible.
- Bombieri–Pila and Heath-Brown estimates are applied to count integral points on the resulting curves, particularly handling cases where the curves are lines, conics, or higher-degree polynomials.

A critical aspect of the methodology is the uniformity of the bounds with respect to the coefficients $a_1, a_2, a_3$ of the Fermat surface equation.

Key Contributions and Results

Uniform Bound for Generalized Fermat Surfaces (Theorem 1.2):
The authors establish a uniform upper bound for the number of primitive integral points $S(X, Y, Z, f)$ on the surface $a_1 x^p + a_2 y^q + a_3 z^r = 0$ . The bound is given by:
$\#S(X, Y, Z, f) \ll_{\varepsilon, p, q, r} (XYZ)^\varepsilon \left( W^2 + W \max\{X, Y\}^{\frac{2}{\sqrt{\max\{p, q, 36\}}}} + W \max\{X, Y\}^{\frac{1}{r}} \right)$
where $W = \exp\left(\sqrt{\log X \log Y / r}\right)$ . This result is significant because the implied constant is independent of the coefficients $a_i$ , and the exponent of the height bound tends to zero as the exponents $p, q, r$ grow large.
Power-Saving for Sums of Powerful Numbers (Theorem 1.1):
By transferring the uniform bounds from the Fermat surface to the counting function $N(B)$ , the authors prove that for $p = r + u$ and $q = r + v$ (with $u \geqslant v \geqslant 0$ ), there exists an explicit $\eta_{u,v}(r) > 0$ such that:
$N(B) \ll_{\varepsilon, u, v, r} B^{1/p + 1/q - \eta_{u,v}(r) + \varepsilon}$
where $\eta_{u,v}(r) = \frac{1}{r^2} + O_{u,v}(r^{-5/2})$ . This represents a power-saving over the trivial bound $N(B) \ll B^{1/p + 1/q}$ .
Handling Non-Symmetric Cases:
The results are particularly effective in non-symmetric cases where $p, q, r$ are of similar size. For instance, in the case $(p, q, r) = (r+2, r+1, r)$ , the method yields a non-trivial bound for all $r \geqslant 2$ .

Significance and Claims
The paper claims to provide quantitative evidence supporting the conjecture that Campana points in the log-general type range are not Zariski dense. While the finiteness of $N(B)$ remains unproven, the establishment of upper bounds with a power-saving over the trivial estimate constitutes a significant step forward.

The authors emphasize that their work offers a uniform estimate that does not depend on the specific coefficients of the Fermat equation, distinguishing it from previous non-uniform results (such as those by Darmon–Granville or Beukers). Furthermore, the paper provides a refined analysis of the "lopsided" ranges of exponents, which had previously been difficult to control. The result yields a uniform power-saving even in the fully symmetric case ( $p=q=r$ ), although it does not yet reach the conjectured balanced exponent suggested by the geometry of the orbifold.

The work is presented as a technical advancement in the arithmetic of Campana points, bridging the gap between the study of rational points and integral points, and demonstrating the efficacy of combining the determinant method with function field Diophantine geometry in this context.

More like this