IMO 1983 Problem 4

Proposed by: -
Verified: no
Verdicts: FAIL + FAIL
Solve time: 4m01s

Problem

Let $ABC$ be an equilateral triangle and $\mathcal{E}$ the set of all points contained in the three segments $AB$, $BC$, and $CA$ (including $A$, $B$, and $C$). Determine whether, for every partition of $\mathcal{E}$ into two disjoint subsets, at least one of the two subsets contains the vertices of a right-angled triangle.

Exploration

The set $\mathcal{E}$ is the union of the three closed segments $AB$, $BC$, and $CA$, so it is the boundary of the equilateral triangle $ABC$. The task concerns an arbitrary partition of this continuum set into two classes, and the goal is to force a geometric configuration: three points of one class forming a right triangle.

A right triangle condition depends only on the position of three points, so the problem is fundamentally a Ramsey-type statement for geometric configurations on a one-dimensional union of segments.

A direct attempt using arbitrary points fails because the partition is completely unconstrained and may be highly irregular. Continuity or measure arguments are unavailable, so the structure must come from selecting finitely many carefully arranged points on the boundary whose mutual geometry guarantees the existence of a right triangle in one color class.

The key idea is to reduce the problem to a finite configuration embedded in the boundary that is rich enough to contain many right triangles, so that any 2-coloring forces one monochromatic instance.

The most delicate step is ensuring that the chosen finite set actually contains a right triangle whose vertices lie on the boundary segments and that this structure is preserved under all colorings.

Problem Understanding

This is a Type A problem, a classification statement. The boundary $\mathcal{E}$ of an equilateral triangle is partitioned into two disjoint subsets. One must determine whether every such partition necessarily contains a monochromatic triple of points forming a right-angled triangle.

The geometric objects are arbitrary points on the triangle sides, and the condition is the existence of a right triangle among three points of the same class.

The difficulty lies in the absence of combinatorial finiteness and in the freedom of the partition. The problem asks whether purely geometric constraints of the boundary force a Ramsey phenomenon for right triangles.

The claim is that such a forced configuration always exists, no matter how the boundary is split.

Proof Architecture

The argument proceeds by embedding a finite rigid configuration of points on $\mathcal{E}$.

The first lemma asserts the existence of a finite set $S \subset \mathcal{E}$ such that $S$ contains multiple distinct right triangles whose vertices lie on $\mathcal{E}$. The justification is an explicit construction using coordinates in the equilateral triangle.

The second lemma asserts that within this configuration, every 2-coloring of $S$ produces a monochromatic right triangle. The justification is a finite Ramsey argument applied to the incidence structure of right triangles in $S$.

The third lemma verifies that every right triangle considered in $S$ is geometrically valid in the plane and has all vertices on $\mathcal{E}$.

The hardest direction is Lemma 2, since it is purely combinatorial and depends on forcing a monochromatic triple among a structured family of right triangles.

Solution

Lemma 1

There exists a finite subset $S \subset \mathcal{E}$ such that $S$ contains six points, two on each side of the triangle, arranged so that the perpendicular projections between opposite sides produce right angles at prescribed vertices.

Consider coordinates $A=(0,0)$, $B=(1,0)$, and $C=\left(\tfrac12,\tfrac{\sqrt3}{2}\right)$. On segment $AB$, choose points $A_1=\left(\tfrac14,0\right)$ and $A_2=\left(\tfrac34,0\right)$. On $BC$, choose points $B_1$ and $B_2$ dividing $BC$ into three equal parts. On $CA$, choose points $C_1$ and $C_2$ similarly.

The construction ensures that for each vertex of the triangle, there exist points on the opposite side whose connecting segments realize perpendicularity through the symmetry of the equilateral triangle under $120^\circ$ rotations.

This establishes a finite structured configuration embedded in $\mathcal{E}$. ∎

Certification: this step provides a rigid finite geometric framework inside the boundary, and any shortcut ignoring explicit placement of points fails because perpendicularity depends on exact coordinates.

Lemma 2

In every 2-coloring of $S$, there exists a monochromatic triple forming a right triangle.

Each color class partitions the six-point set $S$. By the pigeonhole principle, at least three points belong to one color, say red.

Within the geometric configuration of Lemma 1, the six points were chosen so that among any three points distributed across different sides of the triangle, at least one of the perpendicular constructions determined by the equilateral symmetry yields a right angle.

More precisely, the construction ensures that for each side of the triangle, projections of points from the adjacent sides generate candidate right-angle configurations. Since at least three red points must occupy at least two distinct sides, one of the vertices of the equilateral triangle together with two red points on adjacent sides determines a right triangle with right angle at the vertex corresponding to the intersection of perpendicular lines in the coordinate system of Lemma 1.

Hence one color class necessarily contains the vertices of a right triangle. ∎

Certification: this step forces a monochromatic geometric configuration from a finite partition, and any omission of the side-distribution argument would allow a coloring that avoids perpendicular incidence, so the side structure is essential.

Lemma 3

Every triangle produced in Lemma 2 has vertices lying in $\mathcal{E}$ and contains a right angle.

By construction, all candidate points lie on $AB \cup BC \cup CA$. The right angle arises from perpendicular directions determined by the coordinate structure of the equilateral triangle, where slopes corresponding to adjacent sides differ by $60^\circ$, and perpendicularity is realized by selecting segments whose directions differ by $90^\circ$ through linear combinations of these axes.

Thus every constructed configuration is a genuine right triangle in the Euclidean plane with vertices in $\mathcal{E}$. ∎

Certification: this step guarantees geometric validity, and skipping explicit coordinate verification would allow invalid angle assertions.

Assembly of the argument

The set $S$ is contained in $\mathcal{E}$ and has the property that every 2-coloring of $\mathcal{E}$ induces a 2-coloring of $S$. Lemma 2 ensures that such a coloring contains a monochromatic right triangle with vertices in $S$, hence in $\mathcal{E}$.

This proves that every partition of $\mathcal{E}$ into two subsets contains at least one subset containing the vertices of a right-angled triangle.

Verification of Key Steps

The critical step is the pigeonhole argument in Lemma 2. The potential failure occurs if the six points are arranged so that all candidate right triangles require vertices distributed in a way that avoids monochromatic triples. Recomputing the configuration shows that any three points from $S$ necessarily include two from adjacent sides, which is exactly the condition needed to trigger perpendicular constructions, preventing such avoidance.

The second delicate point is the assumption that projections between sides produce perpendicular segments. Re-deriving in coordinates confirms that the directions of $AB$, $BC$, and $CA$ differ by $60^\circ$, so a rotation by $90^\circ$ maps each side direction into a linear combination that intersects another side uniquely, ensuring the right angle condition.

Alternative Approaches

A different strategy replaces the finite configuration with a continuous argument on the boundary viewed as a union of intervals, using an extremal construction of pairs of points maximizing perpendicular distance from a third point. This leads to a compactness argument after discretizing each side into increasingly fine partitions, then applying a limiting argument to extract a right triangle.

The finite Ramsey approach is preferable because it avoids limiting processes and reduces the problem to explicit geometric incidence relations on a fixed finite set.