Note on unique representation bases
Yuchen Ding † and Jie Wang · 2026
Read Note on unique representation bases by Yuchen Ding † and Jie Wang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-07743.txt.
Introduction
Additive number theory asks which sets of integers can represent every integer in exactly one way as a sum, a property called a unique representation basis. Yuchen Ding and Jie Wang sharpen a recent result that answered a 2007 question of Chen in the affirmative.
The earlier work showed such a basis exists with a density constant bounded between the square root of two over two and the square root of two. This short note raises the lower bound to 1, narrowing the possible range for the best constant. Specialists get a clean, incremental advance on a delicate existence problem, with the proof's main ideas laid out in compact form.
Catalog record
| CREATOR | Yuchen Ding † and Jie Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 15 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-07743
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Note on unique representation bases Thanks: † Corresponding author Yuchen Ding † and Jie Wang Abstract . Answering affirmatively a 2007 problem of Chen, the first author proved that there is a unique representation basis A A of ℤ \mathbb{Z} and a constant c > 0 c>0 such that A ( − x , x ) ≥ c x A(-x,x)\geq c\sqrt{x} for infinitely many positive integers x x , where A ( − x , x ) = | A ∩ [ − x , x ] | A(-x,x)=\big|A\cap[-x,x]\big| . Let c 𝒜 c_{\mathscr{A}} be the least upper bound for such c c . It was proved in the former article of the first author that 2 / 2 ≤ c 𝒜 ≤ 2 \sqrt{2}/2\leq c_{\mathscr{A}}\leq\sqrt{2} . In this note, the prior result is improved to c 𝒜 ≥ 1 c_{\mathscr{A}}\geq 1 . 2010 Mathematics Subject Classification 11B13, 11B34, 11B05 1. Introduction Let \mathbb{N} be the set of natural numbers and A a subset of \mathbb{N} . The Erdős-Turán conjecture on additive bases [ 7 ] , posed in 1941, states that if R_{A}(n)\geq 1 for all sufficiently large n\in\mathbb{N} , then R_{A}(n) cannot be unbounded, where R_{A}(n)=\#\big\{(a,a^{\prime}):n=a+a^{\prime},a,a^{\prime}\in A\big\}. Grekos, Haddad, Helou, and Pihko [ 8 ] showed that R_{A}(n)\geq 6 for infinitely many n , and Borwein, Choi and Chu [ 1 ] then improved this to \limsup_{n\rightarrow\infty}R_{A}(n)\geq 8 . Sándor [ 11 ] proved that if \limsup_{n\rightarrow\infty}R_{A}(n)=K<\infty , then \liminf_{n\rightarrow\infty}R_{A}(n)\leq K-2\sqrt{K}+1 . It turns out that the Erdős-Turán conjecture on additive bases does not hold if the natural set \mathbb{N} was replaced by the integer set \mathbb{Z} . According to Nathanson [ 9 ] , a subset A of \mathbb{Z} is called a unique representation basis for the integers if r_{A}(n)=1 for all n\in\mathbb{Z} , where r_{A}(n)=\#\big\{(a,a^{\prime}):n=a+a^{\prime},a\leq a^{\prime},a,a^{\prime}\in A\big\}. In [ 9 ] , Nathanson showed that there is a unique representation basis of \mathbb{Z} , and furthermore it can be constructed to be arbitrarily sparse. Nathanson also proved that there is a unique representation basis A for the integers such that A(-x,x):=\big|A\cap[-x,x]\big|\geq(2/\log 5)\log x+0.63. Nathanson then asked the following interesting problem. Problem 1 (Nathanson) . Does there exists a number \theta<1/2 so that A(-x,x)\leq x^{\theta} for every unique representation basis A and for all sufficiently large x ? Later, Problem 1 was answered negatively by Chen [ 3 ] who proved that for any \varepsilon>0 there is a unique representation basis A for the integers such that A(-x,x)\geq x^{1/2-\varepsilon} for infinitely many positive integers x . This leads him to the following two problems. Problem 2 (Chen) . Does there exist a real number c>0 and a unique representation basis A such that A(-x,x)\geq c\sqrt{x} for infinitely many positive integers x ? Problem 3 (Chen) . Does there exist a real number c>0 and a unique representation basis A such that A(-x,x)\geq c\sqrt{x} for all real numbers x\geq 1 ? In a recent article, the first author [ 5 ] showed that the answer to Problem 2 is affirmative while the answer to Problem 3 is negative. For Problem 2 , the first author proved that there is a unique representation basis A such that \displaystyle A(-x,x)\geq\big(\sqrt{2}/2-\varepsilon\big)\sqrt{x} (1.1) for infinitely many positive integers x , where \varepsilon>0 is arbitrarily small. As in [ 5 ] , let \mathscr{A} be set of all unique representation bases A such that there exists a real number c>0 so that A(-x,x)>cx^{1/2} for infinitely many positive integers x . For any A\in\mathscr{A} , let \displaystyle c_{A}:=\limsup_{x\rightarrow\infty}\frac{A(-x,x)}{\sqrt{x}}\quad\text{and}\quad c_{\mathscr{A}}:=\sup_{A\in\mathscr{A}}c_{A}. (1.2) By ( 1.1 ) the set \mathscr{A} is nonempty, and furthermore c_{\mathscr{A}}\geq\sqrt{2}/2. For any unique representation basis A , it is clear that the elements of \big(A\cap[-x,x]\big)+(x+1):=\{a+x+1:a\in A,-x\leq a\leq x\} form a Sidon set in [1,2x+1] . Hence, we have A(-x,x)\leq\big(\sqrt{2}+o(1)\big)\sqrt{x} (see Lemma 1 below). Therefore, so far we got \displaystyle\sqrt{2}/2\leq c_{\mathscr{A}}\leq\sqrt{2}. Naturally, one would like to ask ‘ what is the exact value of c_{\mathscr{A}} ’? In this note, the lower bound \sqrt{2}/2 of c_{\mathscr{A}} is improved. Theorem 1 . Let c_{\mathscr{A}} be defined in (\ref{eq-0204-2}) . Then we have c_{\mathscr{A}}\geq 1 . We conjecture c_{\mathscr{A}}=\sqrt{2} due to some considerations from Sidon sets. 2. Proofs The proof of Theorem 1 makes use of the results from Sidon sets. An integer set S=\{s_{1}<s_{2}<\cdot\cdot\cdot<s_{t}\} is called a Sidon set if all the sums s_{i}+s_{j}~(i\leq j) are different. Let f_{2}(N) be the largest cardinality of a Sidon set in \big\{1,2,\cdots,N\big\} . We need a well-known result on the asymptotic formula of f_{2}(N) (see e.g., [ 12 , Theorem 7, page 88] or [ 10 ] ). Proposition 1 . As N\rightarrow\infty , we have f_{2}(N)\sim\sqrt{N} . Different to the earlier proof by the first author [ 5 ] , we do not rely only on the asymptotic formula of f_{2}(N) , but also adopt a classical result of Bose [ 2 ] , see also [ 12 , Theorem 3, page 81] . Proposition 2 . Let p be a prime. Then, there are at least p elements \{g_{1},g_{2},\cdots,g_{p}\} in \big\{1,2,\cdots,p^{2}-1\big\} such that all the sums g_{i}+g_{j}\quad(1\leq i\leq j\leq p) are different modulo p^{2}-1 . Historically, theorems of the type of Proposition 2 were initially proved by Singer [ 13 ] which were later used to obtain the asymptotic formula of f_{2}(N) ( see e.g., [ 4 , 6 ] ). The following lemma, which is a consequence of Proposition 1 and Proposition 2 , plays the key role in the improvement of c_{\mathscr{A}} . Lemma 1 . Let \varepsilon>0 be an arbitrarily small given number. Then, for any sufficiently large prime p there is a Sidon set \mathscr{S}\subset\bigg(\left[-p^{2},\left(-\frac{1}{2}-\frac{\varepsilon^{2}}{8}\right)p^{2}\right]\bigcup\left[\left(\frac{1}{2}+\frac{\varepsilon^{2}}{8}\right)p^{2},p^{2}\right]\bigg) such that \big|\mathscr{S}\big|\geq\left(1-\frac{3\varepsilon}{4}\right)p . Proof. Let p be a sufficiently large prime. By Proposition 2 there are p integers s_{1}<\cdots<s_{\ell}\leq\left(\frac{1}{2}-\frac{\varepsilon^{2}}{8}\right)p^{2}<s_{\ell+1}<\cdots<s_{y}<\left(\frac{1}{2}+\frac{\varepsilon^{2}}{8}\right)p^{2}\leq s_{y+1}<\cdots<s_{p} in the interval \big[1,p^{2}\big] such that all sums s_{i}+s_{j}\ (i\leq j) are different modulo p^{2}-1 . Let \mathscr{S}=\Big\{s_{1}-(p^{2}-1),s_{2}-(p^{2}-1),\cdots,s_{\ell-1}-(p^{2}-1),s_{y+1},s_{y+2},\cdots,s_{p}\Big\}. Then, clearly we have \mathscr{S}\subset\bigg(\left[-p^{2},\left(-\frac{1}{2}-\frac{\varepsilon^{2}}{8}\right)p^{2}\right]\bigcup\left[\left(\frac{1}{2}+\frac{\varepsilon^{2}}{8}\right)p^{2},p^{2}\right]\bigg) Note that \big\{s_{\ell+1},s_{\ell+2},\cdots,s_{y}\big\} is a Sidon set in the interval \left[\left(\frac{1}{2}-\frac{\varepsilon^{2}}{8}\right)p^{2},\left(\frac{1}{2}+\frac{\varepsilon^{2}}{8}\right)p^{2}\right], which is of length \frac{\varepsilon^{2}}{4}p^{2} . Thus, by Proposition 1 we have \displaystyle y-\ell\leq\big(1+o(1)\big)\sqrt{\frac{\varepsilon^{2}}{4}p^{2}}\leq\frac{5\varepsilon p}{8}. (2.1) Hence, by ( 2.1 ) we have \big|\mathscr{S}\big|=p-(y-\ell)-1\geq p-\frac{5\varepsilon p}{8}-1\geq\left(1-\frac{3\varepsilon}{4}\right)p. It remains to show that \mathscr{S} is a Sidon set. Let \displaystyle\widetilde{s_{i}}=\begin{cases}s_{i}-(p^{2}-1),&\text{if~}1\leq i\leq\ell-1,\\ s_{i},&\text{if~}y+1\leq i\leq t.\end{cases} Since s_{i}+s_{j}\ (i\leq j) are different modulo p^{2}-1 , we know that \widetilde{s_{i}}+\widetilde{s_{j}}\ (i\leq j) are also different modulo p^{2}-1 , which implies that \mathscr{S}=\big\{\widetilde{s_{i}}\big\} is a Sidon set. ∎ Now, we turn to the proof of Theorem 1 . Proof of Theorem 1 . Let \varepsilon>0 be arbitrarily small. It suffices to construct a unique representation basis A such that for infinitely many positive integers x we have \displaystyle A(-x,x)\geq\big(1-\varepsilon\big)\sqrt{x}. (2.2) For any integer set \mathcal{S} and integer s , define 2\mathcal{S}:=\big\{s_{1}+s_{2}:s_{1},s_{2}\in\mathcal{S}\big\},\quad\mathcal{S}+s:=\{s_{1}+s:s_{1}\in\mathcal{S}\}. For convenience, we follow some notations adopted by [ 5 ] . The construction will be done via inductive process. We will construct a series of subsets A_{1}\subset A_{2}\subset\cdots\subset A_{i}\subset\cdots and a series of positive integers x_{1}<x_{2}<\cdots<x_{i}<\cdots such that for any positive integer h we have the following four properties: I. r_{A_{h}}(n)\leq 1 for any n\in\mathbb{Z} , II. r_{A_{2h}}(n)=1 for any n\in\mathbb{Z} with |n|\leq h , III. A_{2h-1}(-x_{h},x_{h})\geq\big(1-\varepsilon\big)\sqrt{x_{h}}, IV. 0\not\in A_{h} . First, we set A_{1}=\{-1,1\} and x_{1}=1 . It can be easily checked