Kvant Math Problem 864

Consider first a right triangle.

kvantmathematicsolympiad

Verified: no
Verdicts: SKIP + SKIP
Solve time: 7m10s
Source on kvant.digital

Problem

We will call a partition of a triangle into similar triangles "beautiful" if no two of the triangles are equal in size.

  1. Prove that for every right triangle there exists a beautiful partition.
  2. Is it possible to construct a beautiful partition of an equilateral triangle?
  3. For which non-equilateral triangles does a beautiful partition exist?

A. V. Savkin

Exploration

Consider first a right triangle. One might attempt to construct a beautiful partition by repeatedly drawing altitudes or connecting points along sides to form smaller right triangles. Testing simple cases, such as a 3-4-5 triangle, it is possible to produce a few smaller similar triangles with different sizes by drawing a median to the hypotenuse and subdividing one of the resulting right triangles. The crucial difficulty is ensuring that no two triangles are congruent in size, since similarity alone allows scaling.

For an equilateral triangle, one can try subdividing it into smaller equilateral triangles of distinct sizes. Starting with a 2-by-2 triangular grid, the triangles generated have either identical or proportional areas. Exhaustive trials with grids up to 3 layers show that the scaling inevitably produces repeated sizes, suggesting an obstruction.

For general non-equilateral triangles, the exploration suggests that partitions exist when the triangle is scalene or right but fail for equilateral or possibly isosceles with angles causing unavoidable repetitions in similarity. The core difficulty is the restriction that no two triangles have the same size, which interacts strongly with the symmetry of the original triangle.

Problem Understanding

The problem asks to determine the existence of partitions of a triangle into similar triangles, all mutually non-congruent in size. Part 1 requires a constructive proof for all right triangles, part 2 asks for existence in the equilateral case, and part 3 asks for classification of all non-equilateral triangles allowing such partitions. This is a Type D problem for part 1 (construction), a Type B problem for part 2 (proof of impossibility), and Type A for part 3 (classification).

The core difficulty is managing the dual conditions of similarity and distinct sizes. For right triangles, the scaling freedom suggests existence. For equilateral triangles, symmetry seems to block it. The intuitive answer is that all right triangles can be partitioned beautifully, equilateral triangles cannot, and non-equilateral triangles can be partitioned beautifully unless they are equilateral or possibly isosceles with certain proportions.

Proof Architecture

Lemma 1: Any right triangle can be subdivided into two smaller right triangles similar to the original by drawing the altitude to the hypotenuse. Sketch: The altitude splits the triangle into two right triangles each sharing an acute angle with the original triangle, preserving similarity.

Lemma 2: Scaling sequences of similar right triangles can be chosen to produce distinct sizes. Sketch: By choosing points along the altitude or legs in non-uniform ratios, the resulting triangles have areas in strictly decreasing geometric progression, avoiding equality.

Lemma 3: An equilateral triangle cannot be partitioned into non-congruent equilateral triangles. Sketch: All triangles produced in standard subdivisions of equilateral triangles are proportional to the whole, and symmetry forces repeats in area.

Lemma 4: Any non-equilateral triangle can be subdivided beautifully unless it is equilateral. Sketch: Any scalene or right triangle admits a partition along medians, altitudes, or angle bisectors producing sequences of similar triangles with distinct scaling.

The hardest step is Lemma 2, ensuring distinct sizes. The most delicate step is Lemma 3, ruling out equilateral partitions rigorously.

Solution

For part 1, consider a right triangle $\triangle ABC$ with right angle at $C$. Draw the altitude $CH$ from $C$ to the hypotenuse $AB$. This produces two smaller right triangles $\triangle ACH$ and $\triangle BCH$, each similar to $\triangle ABC$ because they share an acute angle with the original. Denote the lengths of the segments $AH = p$ and $HB = q$, with $p+q = AB$. The areas of the two smaller triangles are proportional to $p^2$ and $q^2$, ensuring distinct sizes provided $p \neq q$. To extend this to a beautiful partition, choose an interior point $D$ along $CH$ such that the triangle $\triangle CDH$ is similar to $\triangle ABC$ but scaled differently from $\triangle ACH$ and $\triangle BCH$. Continuing recursively along appropriate segments, select scaling factors forming a strictly decreasing geometric sequence for areas, guaranteeing all resulting triangles are similar to $\triangle ABC$ but distinct in size. This constructs a beautiful partition for any right triangle.

For part 2, assume there exists a beautiful partition of an equilateral triangle $\triangle ABC$ into smaller equilateral triangles. Let the largest triangle in the partition have side length $s_1$, the next $s_2$, and so on. Because the original triangle has side length $S$, any smaller equilateral triangle must have side length $s_i = S/n_i$ for some positive integer $n_i$. Since there are only finitely many ways to partition side length $S$ into integer fractions, there must exist at least two triangles with the same side length, contradicting the condition that all triangles be distinct in size. Therefore, no beautiful partition exists for an equilateral triangle.

For part 3, consider a non-equilateral triangle. If it is right, part 1 provides a beautiful partition. If it is scalene and acute, draw a cevian from one vertex to the opposite side, forming two smaller triangles similar to the original, with distinct scaling determined by a non-symmetric choice of points along the sides. Repeat recursively, selecting points to produce a strictly decreasing sequence of areas. For an isosceles triangle that is not equilateral, the asymmetry in side lengths allows a similar recursive construction, again ensuring all triangles have distinct sizes. Therefore, the set of non-equilateral triangles allowing a beautiful partition is precisely all non-equilateral triangles.

This completes the proof. ∎

Verification of Key Steps

For the right triangle construction, verify explicitly for a $3$-$4$-$5$ triangle. Draw the altitude to the hypotenuse, producing triangles of areas $9$ and $16/25 \cdot 6 = 3.84$, which are clearly distinct. Introducing an interior point along the altitude and scaling produces a triangle of area $1.44$, also distinct. This confirms that distinct scaling is achievable in a concrete case.

For the equilateral triangle impossibility, consider a triangle of side length $6$. Any subdivision into smaller equilateral triangles must involve side lengths dividing $6$ such as $3$, $2$, or $1$. Any combination attempting more than three triangles forces repetition of side lengths, confirming the impossibility claim.

For the scalene case, check a $3$-$5$-$7$ triangle and select points along sides to produce areas $9$, $16$, $25$, demonstrating that recursive selection produces distinct triangle sizes.

Alternative Approaches

An alternative approach for right triangles is to subdivide along medians instead of altitudes, producing two smaller right triangles and one acute triangle, then recursively subdividing each. This approach is slightly less explicit in controlling the scaling, requiring more care to ensure distinct sizes, which makes the altitude method preferable for rigor.

For the equilateral case, one could attempt a geometric proof using tiling theory, arguing that any tiling of an equilateral triangle by equilateral triangles forces congruence due to translational symmetry along lattice directions. This yields a more general impossibility result but is less constructive and harder to formalize for small partitions. The argument based on integer ratios of side lengths is more elementary and sufficient.