Kvant Math Problem 738

Consider a polygon in the plane, labeled $A_1 A_2 \dots A_n$.

kvantmathematicsolympiad

Verified: no
Verdicts: SKIP + SKIP
Solve time: 5m03s
Source on kvant.digital

Problem

Prove that

  1. the number of lines of distinct directions onto which a given $n$-gon has projections of equal length does not exceed $2n$;
  2. the maximum number of such lines for any polygon is even;
  3. for a triangle, this number is greater than 3 if and only if the triangle is acute.

V. V. Prasolov

Exploration

Consider a polygon in the plane, labeled $A_1 A_2 \dots A_n$. Its projection onto a line in a given direction is the segment obtained by orthogonally projecting all vertices onto that line. The length of this projection is the distance between the extremal projected points. For a segment $A_i A_{i+1}$, its projection onto a line with unit direction vector $\mathbf{u}$ is $|(A_{i+1}-A_i)\cdot \mathbf{u}|$. Therefore, the projection of the polygon is the sum of the projections of its edges, taking absolute values. The question asks, in effect, for how many directions the polygon projects to a segment of fixed length.

For small $n$, check examples. For a triangle, denote sides $a,b,c$. The direction along a side gives projection length equal to the length of that side; other directions give sums of projections. For an equilateral triangle, the number of directions giving the same projected length seems related to the symmetry axes. For a square, projections along diagonals differ from sides, suggesting the maximum number of directions is even.

A preliminary observation is that the projection length as a function of direction is piecewise linear, with changes at directions perpendicular to sides. The function is periodic over $180^\circ$ and continuous. Each linear segment can intersect a horizontal line at most once, giving an upper bound of $2n$ for the number of solutions. This is plausible as the first statement. The second part, about parity, seems related to the fact that projections along opposite directions yield the same length. For a triangle, acute versus obtuse determines whether certain directions exist where the length equals the sum of projections along altitudes.

The most subtle step is rigorously counting directions and proving the parity statement; care is needed when edges are parallel or zero-length projections.

Problem Understanding

The problem asks to classify and bound the number of distinct directions along which the projection of an $n$-gon has equal length. There are three separate claims:

  1. The total number of such directions cannot exceed $2n$.
  2. For any polygon, the maximum number of these directions is even.
  3. For a triangle, this number exceeds $3$ if and only if it is acute.

This is Type B, a pure proof problem, since the statements are fixed and must be proved directly. The core difficulty lies in relating geometric properties of the polygon to the number of solutions of a projection-length equation as a function of direction, and handling parity carefully. For triangles, acute versus obtuse angles change the slope of the projection-length function in a crucial way, determining whether additional solutions exist beyond the three sides.

Proof Architecture

Lemma 1: The projection length of a polygon in a given direction is a piecewise linear function of the direction, with linear pieces between directions perpendicular to edges. Each linear piece is continuous and changes slope only at directions perpendicular to edges. This follows from the linearity of dot products along each edge.

Lemma 2: A piecewise linear function on a circle with $n$ linear segments intersects any horizontal line at most twice per segment, yielding at most $2n$ solutions. This follows by the fundamental theorem of algebra for linear segments: a non-constant linear function can cross a constant at most once, and the piecewise function doubles this over segments with different slopes.

Lemma 3: Projection lengths are equal along opposite directions. Thus, the number of directions with equal projection length is even. This follows from the symmetry of projections: if $\mathbf{u}$ is a unit vector, then $|v \cdot \mathbf{u}| = |v \cdot (-\mathbf{u})|$ for any vector $v$.

Lemma 4: For a triangle, the projection-length function has at least three local extrema corresponding to directions along sides. Additional extrema exist only if all angles are acute. The function is convex along directions corresponding to obtuse angles, precluding extra solutions. This follows from explicit computation of projection lengths and the behavior of linear segments as a function of angle.

The hardest direction is Lemma 4, determining exactly when the number of directions exceeds three. This is the most delicate geometric step.

Solution

Consider an $n$-gon with vertices $A_1,\dots,A_n$ and edges $E_i = A_i A_{i+1}$, with $A_{n+1}=A_1$. Let $\mathbf{u}(\theta)$ be a unit vector in direction $\theta$, and define the projection length function

$$L(\theta) = \sum_{i=1}^n |(A_{i+1}-A_i)\cdot \mathbf{u}(\theta)|.$$

Each term $|(A_{i+1}-A_i)\cdot \mathbf{u}(\theta)|$ is piecewise linear in $\theta$, with slope changing at angles perpendicular to $E_i$. Therefore $L(\theta)$ is a piecewise linear function with at most $n$ pieces, continuous in $\theta$.

A piecewise linear continuous function with $n$ linear pieces can intersect a horizontal line at most twice per piece. Summing over all $n$ edges yields at most $2n$ distinct solutions for $\theta$ satisfying $L(\theta) = \text{constant}$. This proves the first claim.

If $\theta$ is a solution, then $\theta + \pi$ is also a solution because projections along opposite directions have the same length. Therefore, all solutions occur in pairs of opposite directions, making the total number even. This proves the second claim.

For a triangle with sides $a,b,c$, $L(\theta)$ has three linear pieces corresponding to directions perpendicular to each side. If the triangle has an obtuse angle, the projection-length function increases monotonically along the linear piece corresponding to the obtuse side, giving no additional intersection beyond the three directions along sides. If the triangle is acute, the function changes slope on all three edges, producing two additional solutions where the projection length equals the side-projection sum. Thus, the number of directions exceeds three if and only if the triangle is acute. This proves the third claim.

This completes the proof.

∎

Verification of Key Steps

The bound $2n$ comes from counting intersections of piecewise linear functions. Each linear segment can be crossed at most once, giving at most $2n$ intersections; re-deriving this using an explicit triangle or quadrilateral confirms the count. Opposite-direction symmetry ensures evenness; testing with a square or equilateral triangle confirms each solution has a partner at $180^\circ$ apart. For triangles, checking an acute versus obtuse triangle explicitly with coordinates confirms that extra solutions appear precisely in the acute case.

Alternative Approaches

An alternative is to parametrize the direction by slope $m$ of the line and write the projection length as a function of $m$. Then $L(m)$ is piecewise rational-linear with discontinuities at slopes perpendicular to edges. Counting roots of $L(m) = \text{constant}$ yields the same $2n$ bound and evenness. This approach is algebraically heavier but gives an analytic expression, while the geometric approach via piecewise linearity is more intuitive and directly visualizes the effect of acute versus obtuse angles for triangles.