Reused writeup on Math-6036 (gpt pro 5.2)

Erdős Problem 982 --- collection record

PROVENANCE AND STATUS

Official problem: https://www.erdosproblems.com/982

Saved collaborative-database status: falsifiable. Snapshot checked: 2026-09-23T17:22:24Z. Latest status and formalization links: https://teorth.github.io/erdosproblems/

This is an attributed collection of existing writeups. The model-folder names below identify their original attribution; placement in GPT_6_Astra_Ultra does not attribute their mathematical work to that model. All locally available source versions selected for this problem are reproduced below, without editing their text.

Proof claims, computations, literature statements, and completion estimates in the reused text have not been independently verified in this collection pass. A source claiming a solution does not change the saved upstream status.

Primary source (latest numbered version in the preferred model folder): attacks/erdos/gpt_pro_5.2/982.tex.

REUSED SOURCE 1

Original attribution: gpt_pro_5.2. Original file: attacks/erdos/gpt_pro_5.2/982.tex.

1) FORMAL RESTATEMENT

Let $V=\{v_1,\dots,v_n\}\subset\mathbb R^2$ be the vertex set of a (strictly) convex polygon with $n\ge 3$ vertices. For a vertex $v\in V$, define \[ D(v):=\#\left\{\|v-w\|: w\in V\setminus\{v\}\right\}, \] i.e. the number of distinct Euclidean distances from $v$ to the other vertices. Define \[ f(n):=\min_{\text{convex }n\text{-gons }V}\;\max_{v\in V} D(v). \] The problem states the conjecture that $f(n)=\lfloor n/2\rfloor$, i.e. every convex $n$-gon has a vertex with at least $\lfloor n/2\rfloor$ distinct distances to the other vertices. It also notes that a regular $n$-gon has $D(v)=\lfloor n/2\rfloor$ for every vertex, so $f(n)\le \lfloor n/2\rfloor$.

2) QUICK LITERATURE/CONTEXT CHECK

The provided statement lists known lower bounds for $f(n)$ (Moser; Erdős--Fishburn; Dumitrescu; Nivasch--Pach--Pinchasi--Zerbib). A quick web search $2026-01-16$ did not surface an explicit proof of the conjectured value $\lfloor n/2\rfloor$ beyond what is stated there.

3) ATTACK PLAN

Proof track: relate small $D(v)$ to many repeated distances from $v$, hence many isosceles triangles with apex $v$. Upper bound the total number of isosceles triangles determined by vertices in convex position; then deduce a lower bound on $\max_v D(v)$. Disproof track: construct convex polygons with unusually many repeated distances from each vertex.

I will carry out the standard isosceles-triangle counting argument yielding Moser's $\lceil n/3\rceil$ bound, and verify the regular polygon upper bound.

4) WORK

FAST REALITY CHECK

- $n=3$: an equilateral triangle has $D(v)=1$ for each vertex, so $f(3)=1=\lfloor 3/2\rfloor$. - $n=4$: a square has $D(v)=2$ for each vertex, so $f(4)\le 2=\lfloor 4/2\rfloor$. - Regular $n$-gons have $D(v)=\lfloor n/2\rfloor$ (proved below).

Lemma 4.1 (regular $n$-gon achieves $\lfloor n/2\rfloor$ distinct distances). In a regular $n$-gon, for each vertex $v$, \[ D(v)=\lfloor n/2\rfloor. \]

Proof. Label vertices $v_0,v_1,\dots,v_{n-1}$ around the circle. The distance from $v_0$ to $v_j$ depends only on $\min\{j,n-j\}$, because chords with central angles $2\pi j/n$ and $2\pi(n-j)/n$ have the same length. As $j$ ranges from $1$ to $n-1$, the values of $\min\{j,n-j\}$ range exactly over $1,2,\dots,\lfloor n/2\rfloor$, each occurring at least once. Distinct such values give distinct chord lengths. Hence there are exactly $\lfloor n/2\rfloor$ distinct distances from $v_0$ to the other vertices; by symmetry the same holds for any vertex. $\square$

Lemma 4.2 (at most two isosceles apices per base in convex position). Let $V$ be the vertex set of a strictly convex polygon. Fix an unordered pair of distinct vertices $\{a,b\}\subset V$. Then there are at most two vertices $c\in V\setminus\{a,b\}$ such that $\|c-a\|=\|c-b\|$. Equivalently, for each base segment $ab$, there are at most two isosceles triangles $\triangle abc$ with vertices in $V$ having $ab$ as the base.

Proof. The locus of points $x\in\mathbb R^2$ satisfying $\|x-a\|=\|x-b\|$ is the perpendicular bisector line $\ell$ of segment $ab$. Thus any vertex $c$ with $\|c-a\|=\|c-b\|$ must lie on $\ell$. Because $V$ is the vertex set of a strictly convex polygon, no three vertices are collinear. Hence $\ell$ can contain at most two vertices of $V$. $\square$

Lemma 4.3 (Moser-type lower bound $f(n)\ge \lceil n/3\rceil$). For every strictly convex $n$-gon with vertex set $V$, there exists a vertex $v\in V$ with \[ D(v)\ge \left\lceil\frac{n}{3}\right\rceil. \] In particular, $f(n)\ge \lceil n/3\rceil$.

Proof. For a vertex $v\in V$, consider the multiset of the $n-1$ distances $\|v-w\|$ for $w\in V\setminus\{v\}$. Group the $n-1$ other vertices into classes according to equal distance from $v$. If these class sizes are $s_1,\dots,s_t$ (so $t=D(v)$ and $\sum_i s_i=n-1$), then the number of isosceles triangles with apex $v$ (i.e. triangles $\triangle v w_1 w_2$ with $\|v-w_1\|=\|v-w_2\|$) equals \[ I(v)=\sum_{i=1}^t \binom{s_i}{2}. \] Using $\binom{s}{2}=(s^2-s)/2$ and Cauchy--Schwarz, \[ \sum_{i=1}^t s_i^2\;\ge\;\frac{(\sum_i s_i)^2}{t}=\frac{(n-1)^2}{D(v)}. \] Therefore \[ I(v)=\frac12\left(\sum_i s_i^2-(n-1)\right)\ge \frac12\left(\frac{(n-1)^2}{D(v)}-(n-1)\right). \] Summing over all vertices $v\in V$ gives a lower bound on the total number $I$ of isosceles triangles determined by $V$: \[ I\;=\;\sum_{v\in V} I(v)\;\ge\;\frac{n}{2}\left(\frac{(n-1)^2}{D_{\max}}-(n-1)\right), \] where $D_{\max}:=\max_{v\in V}D(v)$. On the other hand, by Lemma 4.2, each unordered base pair $\{a,b\}$ supports at most two isosceles triangles with base $ab$, hence \[ I\le 2\binom{n}{2}=n(n-1). \] Combining the inequalities and dividing by $n>0$: \[ \frac12\left(\frac{(n-1)^2}{D_{\max}}-(n-1)\right)\le (n-1). \] Rearrange: \[ \frac{(n-1)^2}{D_{\max}}\le 3(n-1)\quad\Rightarrow\quad D_{\max}\ge \frac{n-1}{3}. \] Since $D_{\max}$ is an integer, $D_{\max}\ge \lceil (n-1)/3\rceil=\lceil n/3\rceil$. This shows that some vertex has at least $\lceil n/3\rceil$ distinct distances. $\square$

5) VERIFICATION

- Lemma 4.2 uses the "strictly convex" hypothesis to exclude collinear triples of vertices. - In Lemma 4.3 the upper bound $I\le 2\binom{n}{2}$ counts isosceles triangles by their base; equilateral triangles (if they occur) may be counted multiple times, but this only strengthens the inequality. - For $n=3$, Lemma 4.3 gives $\lceil n/3\rceil=1$, consistent with the equilateral triangle example.

6) FINAL

UNRESOLVED (i) Strongest proved partial result: $f(n)\le \lfloor n/2\rfloor$ by the regular $n$-gon (Lemma 4.1), and $f(n)\ge \lceil n/3\rceil$ by the isosceles-triangle counting argument (Lemma 4.3). (ii) First gap (crisp): improve the general lower bound from $\lceil n/3\rceil$ to $\lfloor n/2\rfloor$ (or exhibit a convex $n$-gon in which every vertex determines fewer than $\lfloor n/2\rfloor$ distinct distances). (iii) Top 3 next moves: 1. Sharpen Lemma 4.2 by proving a stronger upper bound on isosceles triangles in convex position that accounts for geometric constraints beyond collinearity, potentially lowering the factor $2$. 2. Use additive-combinatorial structure of distance sets from a vertex on a convex polygon to show that many repeated distances force a forbidden configuration (e.g. too many equal chords), improving the Cauchy--Schwarz step. 3. Computation/experiments: search (numerically) for convex polygons with many repeated distances from each vertex to identify plausible extremal constructions beyond the regular polygon. (iv) Minimal counterexample structure: a polygon contradicting the conjecture would have to force, for every vertex $v$, at least two other vertices at the same distance from $v$ among more than $n-1-\lfloor n/2\rfloor$ neighbors; equivalently it would need many isosceles triangles with each vertex serving as an apex, while keeping the total number of isosceles triangles from blowing up.

COMPLETION ESTIMATE

The primary source supplies no extractable numerical completion estimate. Its mathematical progress is not numerically assessed here.

Newly verified progress in this collection pass: $0\%$. This measures new verification work, not the amount of mathematics achieved in the reused text.