🔢 mathematics

A Bombieri-Vinogradov theorem for exponential sums over products of k primes

This paper establishes a Bombieri-Vinogradov type theorem for exponential sums over products of $k$ primes and applies it to prove a lower bound of $x^{1/6 - \varepsilon}$ for the supremum of such sums over integers with exactly $k$ prime factors.

Original authors: Pierre-Alexandre Bazin

Published 2026-07-17

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Pierre-Alexandre Bazin

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 detective trying to solve a mystery hidden inside the infinite sequence of numbers we call integers. Some of these numbers are the "atoms" of math, called prime numbers, which can only be divided by themselves and one. Others are "molecules," built by multiplying these primes together. For centuries, mathematicians have been obsessed with a specific question: How are these numbers distributed? Do they line up in neat rows, or do they scatter like popcorn in a hot pan?

To answer this, mathematicians use a special tool called an "exponential sum." Think of this as a magical radar that scans a group of numbers and checks if they are hiding a secret pattern. If the numbers are truly random, the radar's signal should cancel itself out, resulting in a flat line. But if there's a hidden rhythm or a conspiracy among the numbers, the radar will spike, revealing a wave. The big challenge in this field is that these patterns are incredibly stubborn. Sometimes, the numbers seem to hide their secrets so well that the radar only picks up a faint whisper, making it hard to tell if a pattern exists or if it's just noise. This paper dives into a specific, tricky corner of this mystery: what happens when we look at numbers made of exactly k prime factors, and how do they behave when we scan them with this magical radar?

The Story of the Prime Factor Detective

In this paper, the author, Pierre-Alexandre Bazin, acts like a master detective upgrading the tools used to hunt for patterns in numbers. Specifically, he is looking at numbers that are products of exactly k primes. For example, if k is 2, he's looking at numbers like 6 (2×3), 10 (2×5), or 15 (3×5). If k is 3, he's looking at 30 (2×3×5), and so on.

The goal is to understand how these numbers behave when we apply a "twist" to them using a special mathematical wave (represented by the symbol $\alpha$ ). In the world of math, this is called an "exponential sum." If the numbers were perfectly random, this sum would be very small. But if the numbers have a hidden structure, the sum could be surprisingly large.

The Problem with Previous Maps
Before this paper, other detectives (mathematicians) had drawn maps of where these patterns might hide. One famous map, called the Bombieri-Vinogradov theorem, was excellent at finding patterns in numbers up to a certain distance. However, when the author tried to use this map for his specific case (numbers with k prime factors), he hit a wall. A previous attempt by a mathematician named Yao tried to extend the map further, but the paper points out that Yao's proof had a crack in it. It worked for a short distance, but if you tried to go too far (specifically, if the range of numbers was larger than about $x^{0.33}$ ), the proof fell apart. It was like trying to drive a car across a bridge that looked solid but actually had a missing section in the middle.

The New Solution
Bazin's paper fixes this broken bridge. He proves a new, stronger version of the Bombieri-Vinogradov theorem that works for a wider range of numbers. He shows that even when we look at a huge range of numbers (up to $x^{1/3}$ , which is a cube root of the total count), we can still reliably predict how these "k-prime" numbers behave.

To do this, he didn't just patch the old map; he built a new engine. He used a technique called "Vaughan's identity," which is like breaking a complex machine down into smaller, simpler gears. He showed that the complicated function counting these k-prime numbers can be broken down into two types of simpler pieces: "Type I" (which are easy to handle) and "Type II" (which are a bit trickier but manageable). By proving that his new engine works for these simpler pieces, he proved it works for the whole machine.

The Big Discovery: The Numbers Are Never Quiet
The most exciting part of the paper is what happens after the map is fixed. The author uses this new, stronger tool to answer a question that had been hanging in the air: "How big can the hidden patterns get?"

He proves that no matter how you tune your radar (no matter what value $\alpha$ you choose), there is always a moment when the signal spikes. He establishes a "lower bound," which is a guarantee that the signal will never be smaller than a certain size. Specifically, he shows that the signal is at least as big as $x^{1/6-\varepsilon}$ .

To put this in perspective: If you have a billion numbers ( $x = 10^9$ ), the paper guarantees that the "noise" or "pattern" you find will be at least as loud as the 6th root of a billion (which is 1,000). This is a significant discovery because it proves that these numbers are never perfectly silent; they always have a rhythm that is strong enough to be heard, even in the most difficult cases.

Why This Matters
This isn't just about counting primes. Understanding how these numbers distribute themselves helps mathematicians solve other deep problems in number theory, like how primes are spaced out or how they interact with other mathematical structures. By proving that the "broken bridge" is actually solid and that the "hidden signals" are always loud enough to detect, Bazin has given the mathematical community a more reliable toolkit for exploring the infinite landscape of numbers.

In short, the paper says: "We fixed the map, we proved the old one was wrong in the middle, and we confirmed that these special numbers always have a heartbeat that we can measure."

Technical Summary of "A Bombieri-Vinogradov theorem for exponential sums over products of k primes"

Problem Statement
The paper addresses the distribution of exponential sums over integers with a fixed number of prime factors (counted with multiplicity). Specifically, it investigates the sum
$\Pi_k(x; \alpha) := \sum_{n \le x} f_k(n) e(\alpha n),$
where $f_k(n) = \mathbf{1}_{\Omega(n)=k}$ is the indicator function for integers with exactly $k$ prime factors, and $e(\beta) = e^{2\pi i \beta}$ .

While the asymptotic behavior of the count $\Pi_k(x) = \sum_{n \le x} f_k(n)$ is well-understood via the Sathe-Selberg theorem for $k \le (2-\varepsilon)\log\log x$ , the distribution of these sums in arithmetic progressions and the behavior of the exponential sums $\Pi_k(x; \alpha)$ for irrational $\alpha$ are more complex. The paper aims to establish a Bombieri-Vinogradov type theorem for these exponential sums, providing a bound on the average error when approximating the sum over arithmetic progressions by its expected main term. A secondary goal is to use this estimate to derive a non-trivial lower bound for the supremum of these exponential sums.

Methodology
The proof strategy relies on reducing the problem to estimates of character sums and then applying a generalized version of Vaughan's method.

Reduction to Character Sums:
The author first expresses the exponential sum over an arithmetic progression $a/q$ in terms of Dirichlet characters. Using a decomposition of periodic functions, the sum is separated into a main term (related to the gcd structure) and an error term involving sums over non-principal characters. This reduces the Bombieri-Vinogradov estimate to bounding the quantity:
$\Xi(f, x, Q, \theta) := \sum_{q \le Q} \frac{q}{\phi(q)} \sum_{\chi \pmod q}^* \max_{|\lambda| \le \theta} \left| \sum_{n \le x} f(n) \chi(n) e(\lambda n) \right|.$
The paper notes a critical distinction from the classical setting: for exponential sums, the reduction yields a factor of $Q^{-1/2}$ rather than $Q^{-1}$ , limiting the range of $Q$ to $x^{1/3-\varepsilon}$ rather than the classical $x^{1/2-\varepsilon}$ .
Generalized Vaughan Estimate:
To bound $\Xi(f_k, x, Q, \theta)$ , the paper proves a general theorem (Theorem 8) for functions $f$ that can be expressed as a Dirichlet convolution of "Type I" and "Type II" functions.
- Type I: Functions with good character sum bounds uniformly in the modulus.
- Type II: Functions supported on a specific range (here, $[1, x^{2/3}]$ ).
  The proof utilizes Vaughan's method, splitting the sum based on the size of the factors in the convolution. It employs Lemma 9 (estimating Type I sums with characters) and Lemma 10 (estimating Type II sums using Cauchy-Schwarz and zero-density estimates).
Decomposition of $f_k$ :
The core technical contribution is expressing the indicator function $f_k$ (restricted to $[1, x]$ ) as a linear combination of convolutions of Type I and Type II functions. This is achieved by:
- Splitting integers into "rough" and "friable" parts based on the size of their prime factors ( $x^{1/2}$ ).
- Using the identity $1 * (\mu f^-) = f^+$ to relate these parts.
- Further decomposing the resulting terms to ensure support within $[1, x^{2/3}]$ and separating variables using integral representations (Lemma 13.11 from [6]) to handle the condition $P^-(a) \ge P^+(b)$ .
Lower Bound Application:
The established Bombieri-Vinogradov estimate is combined with a result from the author's previous work [2] (Theorem 11). This theorem relates the average error in arithmetic progressions to a lower bound on the $L^2$ -norm of the exponential sums over short intervals. The author analyzes the main term coefficients $g_{k,x}(q)$ derived from the Sathe-Selberg estimates to verify the necessary conditions for Theorem 11.

Key Results

Theorem 2 (Bombieri-Vinogradov Estimate):
For $Q \le x^{1/3}(\log x)^{-B}$ and $\theta \le Q^{-3}(\log x)^{-B}$ , and $k \le (2-\varepsilon)\log\log x$ , the paper proves:
$\sum_{q \le Q} \max_{(a,q)=1} \max_{|\lambda| \le \theta} \left| \sum_{n \le x} f_k(n) e\left(\left(\frac{a}{q} + \lambda\right)n\right) - \frac{1}{\phi(q)} \sum_{st=q} \mu(s)t \sum_{n \le x, t|n} f_k(n) e(\lambda n) \right| \ll \frac{x}{(\log x)^C}.$
This result improves upon previous work by Yao [17], which failed for $Q \ge x^{c_0}$ with $c_0 \approx 0.3294$ , by extending the valid range up to $x^{1/3-\varepsilon}$ .
Theorem 3 (Lower Bound):
As an application, the paper establishes a uniform lower bound for the exponential sums:
$\sup_{n \le x} |\Pi_k(n; \alpha)| \gg x^{1/6-\varepsilon}$
for $2 \le k \le (2-\varepsilon)\log\log x$ and any $\alpha \in \mathbb{R}$ . The bound is refined to include a factor depending on the deviation of $k$ from $\log\log x$ .

Significance and Claims
The paper claims to resolve a gap in the literature regarding Bombieri-Vinogradov estimates for exponential sums over products of $k$ primes. Specifically, it corrects a limitation in Yao's [17] proof, which relied on zero-density estimates that were insufficient for the range $Q \in [x^{c_0}, x^{1/3}]$ . By adapting Vaughan's character sum estimates (previously used by Liu and Zhan [7] for the von Mangoldt function $\Lambda$ ) to the function $f_k$ , the author extends the range of $Q$ to the optimal $x^{1/3-\varepsilon}$ for this setting.

The work also demonstrates that the techniques used for primes (the case $k=1$ ) can be successfully generalized to products of $k$ primes, provided the function is decomposed appropriately into Type I and Type II components. The paper notes that these results can be easily adapted to the function $h_k = \mathbf{1}_{\omega(n)=k}$ (integers with $k$ distinct prime factors) in a larger range of $k$ .

The significance of the lower bound (Theorem 3) is that it provides a non-trivial estimate for the maximum size of these exponential sums, showing they do not vanish uniformly, even for irrational $\alpha$ . This is derived directly from the quality of the distribution in arithmetic progressions established in Theorem 2.

The Story of the Prime Factor Detective

More like this