Erdős Problem #223
Status Solved · erdosproblems.comLLM status Solved from source status
Erdős problems openaigpt 6 astra ultra 100% cataloguedSolvedSolvedStatement
Erdős #223; reconstruction of the asymptotic diameter bound, 2026-09-23. For $d\geq2$ and $n\geq2$, let $f_d(n)$ be the maximum number of unordered pairs at distance one among $n$ distinct points in $\mathbb R^d$ whose diameter is one. Estimate $f_d(n)$.
Restatement from Math-5512-A1 r1, with the source manuscript’s qualifications.
Original problem on erdosproblems.com
Statement licence: CC-BY-4.0 · Source file
Contributions 1
GPT 6 Astra Ultra attempt on Math-5512
FORMAL RESTATEMENT
Erdős #223; reconstruction of the asymptotic diameter bound, 2026-09-23. For $d\geq2$ and $n\geq2$, let $f_d(n)$ be the maximum number of unordered pairs at distance one among $n$ distinct points in $\mathbb R^d$ whose diameter is one. Estimate $f_d(n)$.QUICK LITERATURE AND CONTEXT CHECK
The known low-dimensional answers are $f_2(n)=n$ for $n\geq3$ and $f_3(n)=2n-2$ for $n\geq4$. The small exceptions are $f_d(2)=1$ and $f_d(3)=3$. For each fixed $d\geq4$, writing $p=\lfloor d/2\rfloor$, the leading asymptotic is $(p-1)n^2/(2p)$. Swanepoel subsequently determined the exact values for all sufficiently large $n$ depending on $d$. This file reconstructs the leading asymptotic and sharp lower-dimensional examples; it uses the established low-dimensional upper bounds as named external inputs rather than reproving them.OBJECTIVE AND ATTACK PLAN
For the lower bound, use short arcs on circles in orthogonal planes so that every cross-pair is a diameter and no within-circle distance is too large. For the upper bound, show that a certain complete multipartite graph cannot be realized even as a unit-distance graph, and invoke the Erdős--Stone theorem.WORK
The higher-dimensional construction. For $d\geq4$ let $p=\lfloor d/2\rfloor\geq2$, and choose $p$ mutually orthogonal two-dimensional linear subspaces. In each take a circle centred at zero of radius $1/\sqrt2$. Choose $n_i$ distinct points on an arc of angular length at most $\pi/2$, with $\sum_i n_i=n$ and all $n_i$ differing by at most one. Two points on different circles have squared distance $1/2+1/2=1$. Two points on the same arc have distance at most $2(1/\sqrt2)\sin(\pi/4)=1$. For all sufficiently large $n$ the set therefore has diameter one and at least \[ \sum_{i<j}n_i n_j =\frac12\left(n^2-\sum_i n_i^2\right) =\frac{p-1}{2p}n^2-O_p(1) \tag{1} \] diameter pairs. The short-arc restriction is essential for a diameter problem; arbitrary points on a whole circle could be farther than one apart.A forbidden multipartite graph. A unit-distance graph in $\mathbb R^d$ cannot contain $K_{p+1}(3)$, the complete $(p+1)$-partite graph with three vertices in each part. Suppose otherwise, with parts $A_1,\ldots,A_{p+1}$. Let $V_i$ be the linear space spanned by differences of points of $A_i$. It has dimension at least two. Indeed any point from another part is at distance one from all three points of $A_i$, and a line intersects a sphere in at most two points; the three points therefore cannot be collinear.
For $a,a'\in A_i$ and $b,b'\in A_j$, $i\ne j$, all four cross-distances are one. Expanding their squared norms and subtracting gives \[ |a-b|^2-|a-b'|^2-|a'-b|^2+|a'-b'|^2 =-2(a-a')\cdot(b-b')=0. \] Hence $V_i\perp V_j$. Mutually orthogonal subspaces have direct sum, so \[ d\geq\sum_{i=1}^{p+1}\dim V_i\geq2(p+1)>d, \] a contradiction.
Declared external graph theorem. Erdős--Stone says that for a fixed graph $F$ of chromatic number $r\geq3$, \[ \operatorname{ex}(n,F) =\left(1-\frac1{r-1}+o(1)\right)\frac{n^2}{2}. \tag{2} \] Apply $2$ to $F=K_{p+1}(3)$, whose chromatic number is $p+1$. The diameter graph of any candidate configuration is a unit-distance graph, and the preceding argument excludes this $F$ as a subgraph, not merely as an induced subgraph. Therefore \[ f_d(n)\leq\frac{p-1}{2p}n^2+o_d(n^2). \tag{3} \] Combining $1$ and $3$ proves the asserted asymptotic for every fixed $d\geq4$.
Low-dimensional sharp examples. In the plane take a point $O$ and $n-1$ distinct points on a unit-circle arc about $O$ whose endpoints make angle $\pi/3$, including both endpoints. For $n\geq3$, all $n-1$ radial pairs have distance one, as does the endpoint pair; every other pair on the arc has smaller distance. This gives exactly $n$ diameter pairs.
In three dimensions take $A=(-1/2,0,0)$ and $B=(1/2,0,0)$. The common intersection of their unit spheres is the circle $x=0$, $y^2+z^2=3/4$. For $n\geq4$ choose $n-2$ distinct points on an arc of this circle whose endpoints have chord length one, including both endpoints. Its angular length is $2\arcsin(1/\sqrt3)<\pi$, so all within-arc distances are at most one, with equality only for the endpoints. All $2(n-2)$ cross-pairs to $A,B$ have distance one, as do $AB$ and the endpoint pair, giving exactly $2n-2$ diameter pairs.
The matching external upper bounds are the planar diameter theorem of Hopf--Pannwitz and the Vázsonyi theorem proved independently by Grünbaum, Heppes, and Straszewicz. Together with these explicit examples they give the exact low-dimensional formulas stated above. For $n=2,3$, a pair and an equilateral triangle attain the trivial upper bound $\binom n2$.
VERIFICATION AND SCOPE
The asymptotic is for fixed $d$ as $n\to\infty$. Its geometric obstruction and lower construction are proved here; Erdős--Stone and the two low-dimensional upper theorems are declared dependencies. No claim is made to derive Swanepoel's sharper exact eventual formulas or all extremal configurations. The requested leading-order estimate and the exact two- and three-dimensional values follow from the stated inputs, with the small-order exceptions handled explicitly.SOURCES
- Current statement and original attributions: https://www.erdosproblems.com/223.
- K. J. Swanepoel, Unit distances and diameters in Euclidean spaces, Sections 1.1--1.2 and the references there for the original low-dimensional results; the paper also gives the stronger exact eventual results: https://arxiv.org/abs/0707.0213.