A problem of Erdős about rich distances

Abstract.

An old question posed by Erdős asked whether there exists a set of $n$ points such that $c\cdot n$ distances occur more than $n$ times. We provide an affirmative answer to this question, showing that there exists a set of $n$ points such that $\lfloor\frac{n}{4}\rfloor$ distances occur more than $n$ times. We also present a generalized version, finding a set of $n$ points where $c_{m}\cdot n$ distances occurring more than $n+m$ times.

1. Introduction

In a 1997 paper, Erdős [1] asked the following question:

Problem (Erdős).

For a set of $n$ points in a plane, can $c\cdot n$ of the distances occur more than $n$ times?

We provide an affirmative answer to the question by proving the following theorem:

Theorem 1.1.

For all $n\in\mathbb{N}$, there exists a set of $n$ points such that $\lfloor\frac{n}{4}\rfloor$ distances occur at least $n+1$ times.

We also show the following generalization of Theorem 1.1, indicating that $c_{m}$ distances can occur $n+m$ times.

Theorem 1.2.

For all $n\in\mathbb{N}$, there exist a set of $n$ points such that at least $\big{\lfloor}\frac{n}{2(m+1)}\big{\rfloor}$ distances occur at least $n+m$ times.

2. Proof of Theorem 1.1

Figure 1. A set of 9 points with 2 distances appearing 10 times.

Figure 2. A set of 11 points with 2 distances appearing at least 12 times.

We start with proving the following simple claim:

Claim 2.1.

In a regular $m-$gon, $\lfloor\frac{m-1}{2}\rfloor$ distances appear $m$ times.

Proof.

Observe that in a regular $m-$gon $v_{1}\dots v_{m}$, the distances $\|v_{i}-v_{(i+k)}\|$ and $\|v_{j}-v_{(j+k)}\|$ are equal for all $i,j\in[m]$ and some $k\in[\lfloor\frac{m-1}{2}\rfloor]$. Thus, we conclude $\lfloor\frac{m-1}{2}\rfloor$ of the distances are repeating $m$ times. ∎

Proof of Theorem  1.1.

For $n\leq 4$ the statement is vacuously true. Hence, we will assume $n\geq 4$. We will consider two cases: Case 1 for $n$ odd and Case 2 for $n$ even.

Case 1 Since $n$ is odd, let $n=2m+1$. Consider an $(m+1)-$gon $v_{1}\dots v_{m+1}$. From Claim 2.1, $\lfloor\frac{m}{2}\rfloor$ of the distances are repeated $m+1$ times. Now, rotate the $(m+1)-$gon around vertex $v_{1}$ to get a new $(m+1)-$gon $v_{1}r_{2}\dots r_{m+1}$. Again, $\lfloor\frac{m}{2}\rfloor$ of the distances are repeating $m+1$ times in the new $(m+1)-$gon. Since $v_{1}$ is the only common vertex between the two $(m+1)-$gons, the total number of vertices in the two $(m+1)-$gons is

$$2\cdot(m+1)-1=2m+1=n.$$

Also observe that as $n$ is odd, $\lfloor\frac{m}{2}\rfloor=\lfloor\frac{n-1}{4}\rfloor=\lfloor\frac{n}{4}\rfloor$ of the distances occur

$$2\cdot(m+1)=2m+2=n+1$$

times. Hence, we get a set of $n$ points $\{v_{1},\dots,v_{m+1},r_{2},\dots,r_{m+1}\}$ where $\lfloor\frac{n}{4}\rfloor$ distances occur at least $n+1$ times. This concludes Case 1.

Figure 3. A set of 10 points with 2 distance appearing at least 11 times.

Case 2 The proof is similar to that of Case 1. However, instead of rotating a regular polygon around one of its vertices, we will reflect a regular polygon on one of its edges.

Since $n$ is even, let $n=2m$. Consider an $(m+1)-$gon $v_{1}\dots v_{m+1}$. From Claim 2.1, $\lfloor\frac{m}{2}\rfloor$ of the distances are repeated $m+1$ times. Now, reflect the $(m+1)-$gon over the edge $v_{1}v_{2}$ to get a new $(m+1)-$gon $v_{1}v_{2}v^{\prime}_{3}\dots v^{\prime}_{m+1}$. Again, $\lfloor\frac{m}{2}\rfloor$ of the distances are repeated $m+1$ times in the new $(m+1)-$gon. Since $v_{1}$ and $v_{2}$ are the only common vertices between the two $(m+1)-$gons, the total number of vertices in the union of the two $(m+1)-$gons is

$$2\cdot(m+1)-2=2m=n.$$

Also, observe that the only distance common between the two $(m+1)-$gons is $\|v_{1}-v_{2}\|$, and repeating only for the edge $v_{1}v_{2}$. Thus, $\lfloor\frac{m}{2}\rfloor=\lfloor\frac{n}{4}\rfloor$ of the distances occur at least

$$2\cdot(m+1)-1=2m+1=n+1$$

times. Hence, we get a set of $n$ points $\{v_{1},\dots,v_{m+1},v^{\prime}_{3},\dots,v^{\prime}_{m+1}\}$ where $\lfloor\frac{n}{4}\rfloor$ many of the distances occur at least $n+1$ times. This concludes Case 2 and proves the theorem.

∎

3. Proof of Theorem  1.2

Figure 4. A set of 8 points with 1 distance appearing at least 11 times. The diagram consists of two rotations and one reflection of the triangle $v_{1}v_{2}v_{3}$.

Proof of Theorem  1.2.

For $n\leq m+3$ the statement is vacuously true hence, we will assume $n\geq m+3$. Let $n=(m+1)k+r$ for some $r\in[2,m+2]$. To prove this theorem, we start with a regular $(k+2)-$gon $v_{1}\dots v_{k+2}$. Fix a vertex, say $v_{1}$, and take $(r-2)$ arbitrary rotations of the $(k+2)-$gon around $v_{1}$, resulting in a total of $(r-1)$ regular $(k+2)-$gons with a common vertex $v_{1}$. Now iteratively reflect the $(k+2)-$gon over an edge $v_{i},v_{i+1}$ for some $i\in[k]$, then chose another edge of any the $(k+2)-$gon and reflect again with a total of $(m+2-r)$ reflections. Hence, the total number of points is

$$(k+2)+(r-2)(k+1)+(m+2-r)(k)=(m+1)k+r=n.$$

Observe that from Claim 2.1 $\lfloor\frac{k+1}{2}\rfloor$ of the distances repeat $k+2$ times in each $(k+2)-$gon, with only repetition of one edge for each reflection. Since there are $m+2-r$ reflections in total, $\lfloor\frac{k+1}{2}\rfloor$ distances appear at least

$$(k+2)(m+1)-(m+2-r)=[(m+1)k+r]+m=n+m$$

times. Finally, since $n=(m+1)k+r$ we have,

$$\Big{\lfloor}\frac{k+1}{2}\Big{\rfloor}\geq\Big{\lfloor}\frac{n}{2(m+1)}\Big{\rfloor},$$

and we conclude that at least $\big{\lfloor}\frac{n}{2(m+1)}\big{\rfloor}$ of the distances appear at least $n+m$ times. ∎

4. Further Research

In [1], the main problem of this paper was mentioned in the context of the previously conjectured Erdős’ distinct distance problem [2] from 1946.

Problem (Erdős’ Distinct Distance Problem).

Does every set of $n$ distinct points in $\mathbb{R}^{2}$ determine $\gg n/\sqrt{\log n}$ many distinct distances?

Erdős’ distinct distance problem was almost settled (with a remaining gap of $\sqrt{\log{n}}$) by Guth and Katz [3]. In the same paper [1] Erdős also mentioned another question of himself and Pach.

Problem (Erdős and Pach).

Let $A\subset\mathbb{R}^{2}$ be a set of $n$ points. Must there be two distances which occur at least once but between at most $n$ pairs of points?

Pannwitz and Hopf [4] proved that the largest distance between points of $A$ can occur at most $n$ times, but it remains unknown whether a second such distance must occur. Erdős and Pach believe that such a distance exists.

Another popular distance problem of Erdős mentioned alongside the distinct distance problem [2] is known as Erdős’ unit distance problem.

Problem (Erdős’ Unit Distance Problem).

Does every set of $n$ distinct points in $\mathbb{R}^{2}$ contain at most $n^{1+O(1/\log\log n)}$ pairs which are distance $1$ apart?

This bound would be best possible as it is achievable for the integer lattice. The best known upper bound is $O(n^{4/3})$, due to Spencer, Szemerédi, and Trotter [5].