🔢 mathematics

On the Thickness of Infinite Generalized Sidon Sets, II

The paper establishes an upper bound for the asymptotic lower density of infinite $B_h$ -sets for every even $h$ , proving that the limit inferior of their counting function normalized by $\sqrt[h]{n/\log n}$ does not exceed a specific constant involving $\pi$ , $\log 2$ , and Gamma functions.

Original authors: Kevin O'Bryant

Published 2026-07-28

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Kevin O'Bryant

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 about how numbers can hide from each other. In the world of mathematics, there is a special club called "Sidon sets." Think of these as exclusive parties where the guests (numbers) are so unique that if you pick any two of them and add their ages together, the result is a sum that no other pair of guests could possibly create. It's like a room full of people where every possible handshake creates a unique, unrepeatable sound. Mathematicians love these sets because they are incredibly efficient at packing numbers into a line without causing "noise" or collisions.

But what happens if we make the party bigger? What if instead of just two people shaking hands, we invite groups of three, four, or even ten people to combine their ages? This is where the concept of a " $B_h$ -set" comes in. It's a group where any combination of $h$ people (allowing for repeats) creates a sum that is totally unique. The big question mathematicians have been asking for decades is: How big can these parties get before they get too crowded? If you look at the first $n$ numbers on a number line, how many of them can you invite to this unique-sum party? This isn't just a game of logic; it's about understanding the fundamental limits of how numbers can be arranged, which has deep connections to cryptography, signal processing, and the very structure of mathematics itself.

Now, enter Kevin O'Bryant, a mathematician who has been investigating the "thickness" of these infinite parties. In a paper titled "On the Thickness of Infinite Generalized Sidon Sets, II," O'Bryant tackles the specific case where the group size, $h$ , is an even number (like 2, 4, 6, etc.). He isn't just asking if these sets can exist; he is trying to find the exact "speed limit" for how fast they can grow.

Imagine you are trying to fill a bucket with water, but the bucket has a tiny hole. You want to know the maximum rate at which you can pour water in before the hole drains it all away. O'Bryant's paper is about finding the precise size of that hole for these number sets. He proves that no matter how cleverly you try to pack these numbers together, there is a hard ceiling on their density. Specifically, he shows that if you look at a set of numbers up to a very large number $n$ , the count of numbers in your set cannot grow faster than a specific formula involving $n$ , the number of people in the group ( $h$ ), and some famous mathematical constants like $\pi$ and the Gamma function (which is just a fancy way of extending the idea of factorials to non-whole numbers).

The paper's main finding is a precise mathematical inequality. O'Bryant proves that for any even number $h$ , the ratio of the size of the set to the "growth limit" (which looks like the $h$ -th root of $n$ divided by the logarithm of $n$ ) must eventually drop below a specific constant. This constant is calculated using a complex-looking formula: $\left( \frac{\pi}{\log 2} \cdot \frac{\Gamma(1 + h/2)^2}{\Gamma(1 + 1/h)^h} \right)^{1/h}$ . In simpler terms, he has drawn a line in the sand and said, "No matter how you try to build this set, it cannot cross this line."

This result is a significant improvement over previous work. Thirty-five years ago, a mathematician named Chen proved that this limit was finite (meaning the set couldn't grow infinitely fast), but he didn't know the exact number. O'Bryant has now provided that exact number. He also clarifies that while his proof works perfectly for even numbers, the situation for odd numbers (like groups of 3 or 5 people) remains a bit of a mystery, though he suspects the same rule applies there too.

The paper doesn't just state a number; it uses a clever strategy involving "multisets" (groups where you can have the same number more than once) and a technique called "averaging over shifts." Imagine trying to find a pattern in a noisy crowd. Instead of looking at the crowd from one fixed angle, O'Bryant and his method look at the crowd from many different angles, shifting the view slightly each time, to smooth out the noise and reveal the underlying structure. By doing this, he was able to lower the "ceiling" of the growth rate, making the limit tighter and more precise than anyone had managed before.

It is important to note that this is a rigorous mathematical proof, not a guess or a simulation. O'Bryant has demonstrated with certainty that for even $h$ , the growth of these sets is bounded by his specific constant. He does not claim to have solved the problem for odd numbers, nor does he claim to have found the best possible set (the one that gets closest to the limit), only that no set can exceed the limit he has calculated. He suspects that the limit is actually zero for the ratio he is studying, meaning these sets might be even thinner than his current bound suggests, but that remains an open question.

In the end, this paper is like a cartographer drawing a more accurate map of a mathematical landscape. For years, explorers knew there was a mountain range (the limit of how big these sets can get), but they didn't know exactly how high the peaks were. O'Bryant has climbed the peak for even-numbered groups and measured its height with a new, precise instrument. While the journey for odd-numbered groups continues, this new measurement provides a solid foundation for future explorers to build upon, ensuring that anyone trying to pack numbers into these unique-sum sets knows exactly how much space they have to work with.

Technical Summary: On the Thickness of Infinite Generalized Sidon Sets, II

Problem Statement
The paper investigates the asymptotic density of infinite $B_h$ -sets. A set $A$ of nonnegative integers is defined as a $B_h$ -set if all sums of $h$ elements (allowing repetition, $a_1 + \dots + a_h$ with $a_1 \le \dots \le a_h$ ) are distinct. A $B_2$ -set is the classical Sidon set. The primary object of study is the counting function $A(n) = |A \cap [0, n)|$ .

The central question addressed is the upper bound on the lower limit of the normalized counting function. Specifically, the paper seeks to determine the behavior of:
$\liminf_{n \to \infty} \frac{A(n)}{(n / \log n)^{1/h}}$
While Chen [1] established 35 years ago that this limit inferior is finite for every even integer $h$ , the specific constant governing this bound remained unspecified. Previous work by the author (Part I) addressed a different generalization and the $h=2$ case, but a unified explicit constant for general even $h$ was lacking.

Methodology
The proof strategy synthesizes combinatorial number theory with analytic techniques, building upon the framework established by Jia [9] and the author's previous work on Golomb rulers. The methodology proceeds through the following steps:

Multiset Formalism: The paper utilizes multiset notation to rigorously handle the $B_h$ property. A key insight is that for a $B_h$ -set $A$ , the map from the multiset of $p$ elements ( $1 \le p \le h$ ) to their sum is a bijection onto the $p$ -fold sumset $pA$. This allows the authors to treat sumsets and difference sets with precise combinatorial control.
Energy and Block Counting: Following Erdős, the authors analyze the intersection of the set with intervals of the form $[(\ell-1)N, \ell N)$ . They define "block counts" $F_\ell$ representing the number of elements in these intervals. The core of the argument involves bounding the "energy" of these blocks, specifically the sum of squares $\sum \binom{F_\ell}{2}$ .
Lower Bound on Sumsets (Lemma 8): A novel contribution is Lemma 8, which provides a nontrivial lower bound on the size of the $k$ -fold sumset $S = kA$ (where $h=2k$ ) under the hypothesis that $A(n)$ grows at a specific rate. This is achieved by transforming the counting problem into a lattice point counting problem in $\mathbb{R}^k$ , approximated via Dirichlet integrals.
Difference Set Control (Lemma 9 & 11): To establish an upper bound, the authors control the number of pairs in the sumset $S$ with small differences. They employ a weighted Cauchy-Schwarz inequality and a quantitative version of Jia's lemma regarding the size of the set $\Phi(r, p, q; x, L)$ , which counts tuples of elements with specific sum-difference properties. Crucially, the authors introduce a parameter $M = \lfloor N / \log^3 N \rfloor$ (separating the number of intervals from their width) to avoid the need for Jia's growth regularity hypothesis ( $A(N^2) = O(A(N)^2)$ ), thereby allowing for explicit constant derivation.
Contradiction Argument: The proof assumes the existence of a constant $\tau$ larger than the proposed bound. This assumption yields a lower bound on the density of the sumset $S$ (via Lemma 8). Simultaneously, the block counting argument (via Lemma 11 and Lemma 6) yields an upper bound on the density of $S$ . Comparing these bounds leads to a contradiction, thereby proving the upper limit on $\tau$ .

Key Contributions and Results
The primary result is Theorem 1, which establishes an explicit upper bound for the lim inf of the normalized counting function for any $B_h$ -set where $h$ is an even positive integer:

$\liminf_{n \to \infty} \frac{A(n)}{(n / \log n)^{1/h}} \le \left( \frac{\pi}{\log 2} \cdot \frac{\Gamma(1 + h/2)^2}{\Gamma(1 + 1/h)^h} \right)^{1/h}$

Corollary 2 translates this result into a lower bound for the growth of the elements $a_n$ of the set:
$\limsup_{n \to \infty} \frac{a_n}{n^h \log n} \ge \frac{\log 2}{\pi} \cdot \frac{h \Gamma(1 + 1/h)^h}{\Gamma(1 + h/2)^2}$

The paper notes that for $h=2$ , this result reduces to a special case of the author's previous work. The constant provided is explicit, involving the Gamma function, and the paper includes a figure illustrating the behavior of this constant for varying $h$ , noting a minimum at $h=4$ .

Significance and Claims
The authors state that their contribution is the provision of the explicit constant in the bound, a detail absent from Chen's earlier proof of finiteness. The paper is modest in its scope, noting that:

The result is proven specifically for even $h$ . While every $B_h$ -set is also a $B_{h-1}$ -set, allowing the result to extend to odd $h$ in an "artificial" manner, the authors acknowledge that no result similar to Theorem 1 is currently known for odd $h$ directly, though Jia [9] has conjectured one.
The paper does not claim to improve the known lower bounds for the existence of such sets (e.g., Cilleruelo's construction), but rather refines the upper bound on their density.
The authors suspect that the lim inf is actually 0 for all $h$ (even or odd), but this paper only establishes the finite upper bound with the specific constant derived.

The work relies heavily on the interplay between the combinatorial structure of sumsets and analytic estimates, utilizing tools like Dirichlet integrals and multiset algebra to bridge the gap between discrete counting and continuous asymptotic analysis.

More like this