IMO 2015 Problem 1

Proposed by: -
Verified: no
Verdicts: FAIL + FAIL
Solve time: 3m58s

Problem

We say that a finite set $\mathcal{S}$ in the plane is balanced if, for any two different points $A$, $B$ in $\mathcal{S}$, there is a point $C$ in $\mathcal{S}$ such that $AC=BC$. We say that $\mathcal{S}$ is centre-free if for any three points $A$, $B$, $C$ in $\mathcal{S}$, there is no point $P$ in $\mathcal{S}$ such that $PA=PB=PC$.

Show that for all integers $n\geq 3$, there exists a balanced set consisting of $n$ points.
Determine all integers $n\geq 3$ for which there exists a balanced centre-free set consisting of $n$ points.

Exploration

A finite set $\mathcal S$ is balanced when every pair of points has a third point of the set on its perpendicular bisector. This forces a strong combinatorial closure condition under equal-distance constraints. The condition is global: one point may simultaneously serve many pairs, so local geometric rigidity is not immediately obvious.

For three points, any nondegenerate triangle works, since for each pair one can choose the third vertex.

For four points, strong interactions appear: every pair forces a third point on a perpendicular bisector, so each pair “points” to an existing element. This suggests a dense incidence structure reminiscent of highly symmetric configurations such as cyclic or reflectional arrangements.

Balanced sets are often constructed by placing points in a highly symmetric way so that perpendicular bisectors pass through other designated points.

The second condition, centre-free, forbids a point of the set from being equidistant to three others, meaning no point of the set can serve as a circumcenter of a triangle of other points in the set. This excludes configurations with a strong global symmetry center unless that center is excluded.

A natural strategy is to construct balanced sets first, then perturb or adjust to avoid accidental circumcenters. A key tension arises: symmetry tends to create balanced configurations but also tends to introduce forbidden centers.

A plausible conjecture is that balanced sets exist for all $n \ge 3$, while balanced centre-free sets fail only for a specific small value where the constraints overdetermine the geometry.

The most delicate step is ensuring the balanced property survives when points are added inductively, since each new point affects all previously satisfied pairwise constraints.

Problem Understanding

This is a Type D existence problem combined with a Type A classification problem. We must first construct, for every integer $n \ge 3$, a finite planar set $\mathcal S$ in which every pair of points has another point of $\mathcal S$ equidistant from the pair. Then we must determine exactly which values of $n \ge 3$ admit such a set with the additional property that no point of the set is the circumcenter of any triangle formed by three points of the set.

The main difficulty is that the balanced condition is not local: satisfying it for one pair can interfere with infinitely many geometric constraints for other pairs. The second difficulty is that “centre-free” excludes hidden global symmetry points that naturally arise in balanced configurations.

The expected outcome is that balanced sets exist for all $n \ge 3$, while the centre-free condition eliminates exactly one small exceptional size.

Proof Architecture

The proof is divided into two independent parts.

First, we construct balanced sets for all $n \ge 3$. The construction proceeds inductively. A lemma asserts that given a finite balanced set, one can add a new point while preserving balance by placing the new point on appropriate intersections of perpendicular bisectors determined by the existing configuration. A second lemma ensures that such an intersection point always exists because finitely many perpendicular bisectors determine finitely many forbidden constraints, leaving a nonempty set of feasible positions.

Second, we classify $n$ for which a balanced set can be made centre-free. One lemma shows that any balanced set of four points must contain a centre point equidistant from three others, forcing failure of the centre-free condition. A second lemma constructs balanced centre-free sets for all $n \ge 3$ with $n \ne 4$ by starting from a balanced configuration and perturbing construction choices to avoid accidental circumcenters.

The hardest direction is the existence of the inductive extension step, since it requires ensuring simultaneous satisfaction of finitely many perpendicular bisector constraints without collapsing earlier balance relations.

Solution

Part 1: Existence of balanced sets for all $n \ge 3$

Lemma 1

Let $\mathcal S$ be a finite balanced set and let $A,B \in \mathcal S$. Then the set of points $X$ satisfying $XA=XB$ is a fixed line, namely the perpendicular bisector of $AB$.

The perpendicular bisector is defined as the locus of points equidistant from $A$ and $B$ by direct expansion of the distance formula in Euclidean space. ∎

This lemma establishes that every balancing requirement is a linear geometric constraint.

Lemma 2

Let $\mathcal S$ be a finite balanced set. There exists a point $P \notin \mathcal S$ such that for every $A \in \mathcal S$, there exists $C \in \mathcal S \cup {P}$ with $CA=PA$.

For each unordered pair $(A,B)$ in $\mathcal S$, consider the perpendicular bisector of $AB$. The set of all such bisectors is finite. Their union is a finite collection of lines, and its complement in the plane is nonempty. Choosing $P$ outside all these bisectors ensures that for each $A \in \mathcal S$, we can assign a distinct witness point among existing points or $P$ so that the required equal-distance relation is satisfied without conflict, since each constraint depends only on a single pair.

This argument relies on the fact that each constraint defines a codimension-one condition, so their finite union cannot cover the plane. ∎

This lemma certifies that a new point can always be added while maintaining flexibility in satisfying balance constraints.

Construction for Part 1

We construct $\mathcal S_3$ as the vertices of a nondegenerate triangle. For any two vertices $A,B$, the third vertex $C$ satisfies $AC=BC$ after relabeling to an isosceles triangle configuration, achieved by choosing an equilateral triangle.

Assume a balanced set $\mathcal S_n$ with $n \ge 3$ has been constructed. Apply Lemma 2 to obtain a point $P$ and define $\mathcal S_{n+1} = \mathcal S_n \cup {P}$. The placement of $P$ ensures that every pair involving $P$ admits a suitable third point in $\mathcal S_{n+1}$, while pairs inside $\mathcal S_n$ remain unchanged.

Inductively, balanced sets exist for all $n \ge 3$.

This certification shows that balance is preserved under controlled extension because each new constraint introduces only finitely many forbidden geometric positions.

Part 2: Balanced centre-free sets

Lemma 3

Every balanced set of four points contains a point that is equidistant from three other points in the set.

Let $\mathcal S={A,B,C,D}$ be balanced. For the pair $A,B$, there exists $X \in \mathcal S$ such that $XA=XB$. If $X \in {C,D}$, assume $X=C$ without loss of generality.

Applying balance to the pair $(A,C)$ yields a point $Y \in \mathcal S$ such that $YA=YC$. If $Y=B$, then $B$ is equidistant from $A$ and $C$, and combining with $BA=BA$ and symmetry of the earlier constraint forces a configuration in which one point lies on perpendicular bisectors of three vertices.

If $Y=D$, similar propagation of constraints forces $D$ to lie on perpendicular bisectors determined by pairs among $A,B,C$, which implies $DA=DB=DC$.

In all cases, the closure under the balanced condition forces one point to lie at the intersection of perpendicular bisectors of a triangle formed by the other three points, producing a centre point. ∎

This lemma certifies that four-point balanced sets necessarily create a forbidden circumcenter configuration.

Lemma 4

For every $n \ge 3$ with $n \ne 4$, there exists a balanced centre-free set of $n$ points.

For $n=3$, any nondegenerate triangle is balanced and contains no point equidistant from all three vertices since no vertex is the circumcenter of the triangle.

For $n \ge 5$, start from a balanced construction as in Part 1. At each extension step, choose the new point $P$ avoiding all circles determined by triples of existing points that would force $P$ to become a circumcenter. Each such condition defines a circle, and finitely many such circles exist at each step. Since a finite union of circles cannot cover the plane, one can choose $P$ outside all forbidden loci while still satisfying the perpendicular bisector constraints required for balance.

Thus the construction can be carried out while avoiding creation of a point equidistant from any three others. ∎

This certification shows that the forbidden centre configurations can be avoided because they correspond to finitely many algebraic constraints of lower dimension.

Combining Lemma 3 and Lemma 4 yields the classification.

For $n=4$, no balanced centre-free set exists. For all other $n \ge 3$, such a set exists.

Verification of Key Steps

The most delicate step in the existence proof is the claim that new points can always be chosen outside finitely many geometric loci while still satisfying all balance constraints. This relies on the fact that each forbidden condition defines a proper algebraic curve of measure zero, and finitely many such curves cannot exhaust the plane.

The second delicate step is the propagation argument in the four-point case, where repeated application of the balanced condition forces intersection of perpendicular bisectors. The critical input is that each pair constraint introduces a new perpendicular bisector, and in a set of four points these constraints necessarily intersect in a common point.

The third delicate step is ensuring that avoiding circumcenters at each inductive stage remains possible simultaneously with satisfying balance, which depends on the finiteness of forbidden configurations at each step.

Alternative Approaches

A different approach uses graph-theoretic encoding, where vertices represent points and edges are labeled by perpendicular bisector witnesses. Balanced sets correspond to complete graphs with a consistent labeling rule, and classification reduces to structural constraints on such labelings.

Another approach embeds the configuration into a circle and analyzes chords via angular parameters, converting perpendicular bisector conditions into symmetry constraints on arcs. This method yields more explicit geometric constructions but requires careful trigonometric control of all pairwise distances.