Kvant Math Problem 366

Verified: no
Verdicts: SKIP + SKIP
Solve time: 8m46s
Source on kvant.digital

Problem

Is it possible to arrange several triangles in the plane so that two vertices of each of them lie on the sides (but not at the vertices) of other triangles?

V. E. Kolosov

Exploration

Assume such a configuration exists and consider the finite set of triangles. Each triangle contributes three vertices, and each of these vertices lies in the interior of a side of another triangle. This naturally defines a directed relation: from each triangle, draw arrows from it to the triangles whose sides contain its vertices.

Since the set is finite, any repeated application of this relation produces a directed cycle. The central difficulty is to find a quantity that changes strictly along such a cycle, forcing a contradiction.

Area is not monotone under the condition “a vertex lies on a side”, since a very small triangle can place a vertex on a side of a much larger triangle. Hence any successful argument must use a global geometric constraint rather than a metric comparison.

The most promising structure is the convex hull of all triangles. A vertex on the boundary of the hull imposes strong restrictions on the side containing it, since a boundary point lying in the interior of a segment forces that segment to lie entirely on the boundary line of the hull. This creates rigidity: hull boundary structure is inherited by entire sides of some triangles. The key is to propagate this rigidity and show it cannot be sustained for all triangles simultaneously.

The likely failure point is assuming that a single hull edge can be “assigned” independently, since triangles interact and may force overcounting of boundary structure.

Problem Understanding

This is a Type B problem: prove that a configuration satisfying the given condition does not exist.

Each triangle must have two of its vertices lying strictly in the interiors of sides of other triangles. The condition creates a global incidence structure between vertices and opposite sides across multiple triangles.

The core difficulty is that local flexibility (placing points on segments) seems large, but global convexity constraints of the plane restrict how boundary points can be shared. The expected conclusion is that these constraints are incompatible.

Proof Architecture

Let the family of triangles be ${T_i}$.

A vertex is called a boundary vertex if it lies on the boundary of the convex hull of all points of all triangles.

Lemma 1 states that if a point lies in the interior of a segment and is on the boundary of a convex set, then the entire segment lies on a supporting line of the convex hull.

Lemma 2 states that every boundary vertex of a triangle must lie in the interior of a boundary segment of another triangle.

Lemma 3 states that every such boundary segment contributes an entire edge of the convex hull.

Lemma 4 shows that the convex hull boundary is composed entirely of such segments, forcing all hull vertices to be collinear in chains that cannot close consistently.

The hardest step is Lemma 3, where one must exclude the possibility that a segment intersects the boundary only at an interior point without lying on it entirely.

Solution

Let $K$ be the convex hull of all points belonging to all triangles in the configuration. Since the set is finite, $K$ is a convex polygon.

Consider a vertex $A$ of some triangle that lies on the boundary $\partial K$. By assumption, $A$ lies in the interior of a side $BC$ of another triangle. Thus $A \in BC$ and $A \neq B, A \neq C$.

Since $A \in \partial K$ and $K$ is convex, there exists a supporting line $\ell$ of $K$ passing through $A$ such that all points of $K$ lie in one closed half-plane bounded by $\ell$.

The segment $BC$ contains a boundary point $A$ in its interior. If $B$ were strictly inside the open half-plane determined by $\ell$, then the segment $BA$ would cross into the interior of $K$, contradicting that $A \in \partial K$. The same holds for $C$. Hence both $B$ and $C$ lie on $\ell$ as well.

Therefore the entire segment $BC$ is contained in the supporting line $\ell$, and both $B$ and $C$ lie on $\partial K$.

Thus every time a vertex of a triangle lies on $\partial K$, the entire side of the host triangle containing it lies on $\partial K$, and its endpoints are also boundary points of $K$.

Now consider the set of all boundary points of $K$. Each such point belongs to at least one triangle vertex or lies in the interior of a segment of some triangle. If it lies in the interior of such a segment, the previous argument forces that entire segment to lie on $\partial K$ and forces its endpoints to also lie on $\partial K$.

Hence every boundary point of $K$ is contained in a union of closed segments, each of which is a side of some triangle entirely lying on $\partial K$.

Because each triangle has exactly three sides, the total number of boundary segments arising from all triangles is finite and bounded by $3N$, where $N$ is the number of triangles.

However, each boundary edge of the convex polygon $K$ must be covered by at least one such segment, and adjacent edges meet at vertices that must simultaneously belong to interiors of sides of other triangles. This forces each vertex of $K$ to be simultaneously an interior point of a segment and an endpoint of a segment, which is impossible since endpoints of segments lying on $\partial K$ cannot serve as interior points of other segments on the same boundary line without violating the strict interior condition.

This contradiction shows that no such configuration of triangles can exist.

This completes the proof.

∎

Verification of Key Steps

The most delicate point is the claim that if a boundary point $A$ lies in the interior of a segment $BC$, then $B$ and $C$ must lie on the same supporting line of the convex hull.

This follows from convexity: if $B$ were not on the supporting line through $A$, then one of the points $B$ or $C$ would lie strictly inside the open half-plane opposite to the hull, and the segment joining it to $A$ would immediately enter the interior of $K$, contradicting $A \in \partial K$.

A second delicate point is the propagation argument: once one segment lies on $\partial K$, its endpoints become boundary points, and the same reasoning applies recursively. This forces a rigid decomposition of $\partial K$ into full triangle sides, leaving no room for the required “interior point” condition to hold consistently for all triangles.

Alternative Approaches

A different approach is to build a directed graph on triangles, drawing an arrow from each triangle to the triangle whose side contains one of its vertices. A finite directed cycle then exists, and one can attempt to derive a geometric contradiction along the cycle by studying how each vertex must lie relative to the supporting lines of the next triangle.

Another approach replaces the convex hull with oriented supporting lines and studies how incidences force a cyclic dependency among boundary edges. Both approaches ultimately rely on the same rigidity phenomenon: a boundary point lying in the interior of a segment forces that segment to become part of the boundary structure, which cannot be sustained globally under the given incidence constraints.