Kvant Math Problem 1434
Consider a simple convex polyhedron, such as a cube or a tetrahedron, and imagine casting a shadow onto a plane with a point light source.
Verified: no
Verdicts: SKIP + SKIP
Solve time: 10m12s
Source on kvant.digital
Problem
It is known that the Earth is flat. Is it true that any convex polyhedron can be illuminated by a point light source from some point in space so that its shadow is a polygon with at least one acute angle?
N. Kozerenko, R. Fyodorov
Exploration
Consider a simple convex polyhedron, such as a cube or a tetrahedron, and imagine casting a shadow onto a plane with a point light source. For a tetrahedron, if the light is positioned so that one of its faces is perpendicular to the direction of illumination, the shadow is a triangle congruent to that face, which may have acute angles if the face itself has them. For a cube, if the light source is at a generic position, the shadow can be a polygon with more than four sides, often producing some acute angles due to the projection of vertices not aligned with the projection plane. The critical observation is that for a convex polyhedron, the shadow is the projection of its vertices and edges, and the shadow polygon is determined by the convex hull of these projected points. The step most likely to hide an error is ensuring that every convex polyhedron, regardless of its combinatorial type or symmetry, admits a point of illumination producing a shadow with an acute angle; there could be highly regular polyhedra, such as a regular cube, whose shadows under some symmetric positions appear rectangular and all angles right. Therefore, the core insight is that perturbing the light source slightly breaks any alignment that produces only right or obtuse angles, producing at least one acute angle in the convex shadow.
Problem Understanding
The problem asks whether, for any convex polyhedron, there exists a point in space from which a point light source casts a shadow such that the shadow polygon has at least one acute angle. This is a Type B problem: the statement to prove is given. The core difficulty lies in handling polyhedra with high symmetry where naive placements of the light source yield only right or obtuse angles in the shadow. The intuition suggests that a small perturbation of the light source breaks perfect alignment, and the convexity of the polyhedron ensures that the shadow polygon will include some vertices with acute angles.
Proof Architecture
Lemma 1. The shadow of a convex polyhedron under a point light source is a convex polygon in the plane of projection. This follows from the fact that the convex hull of projected points is convex.
Lemma 2. For a convex polygon in the plane, if it has at least four vertices and is not a rectangle, it has at least one interior angle strictly less than $90^\circ$. This is because the sum of interior angles is $(n-2)\cdot 180^\circ$, and for $n\ge 4$, having all angles $\ge 90^\circ$ forces equality and a rectangle.
Lemma 3. Any convex polyhedron can be positioned relative to a light source so that its shadow polygon has at least four vertices. This is true because a generic projection of a polyhedron with more than three vertices onto a plane produces at least four distinct projected vertices unless the polyhedron is degenerate.
Lemma 4. By slightly perturbing the position of a point light source, a convex polyhedron whose shadow polygon initially has only right or obtuse angles can be made to produce a shadow with an acute angle. This follows from continuity of the projection with respect to the light source position and the strict convexity condition.
The hardest step is Lemma 4, since it must be argued rigorously that symmetry-breaking produces an acute angle, not just an intuitive “slightly rotated vertex” argument.
Solution
Let $P$ be a convex polyhedron in three-dimensional space. Let a plane $\Pi$ be chosen to serve as the “shadow plane,” which may be any plane in space. The shadow of $P$ from a point light source $L$ is the set of points where rays from $L$ through vertices of $P$ intersect $\Pi$. Denote these projected points by $v_1',v_2',\dots,v_m'$, where $m$ is the number of vertices visible from $L$ in the direction of the projection. The shadow polygon $S$ is the convex hull of the points $v_1',\dots,v_m'$ in the plane $\Pi$. By Lemma 1, $S$ is convex.
If $S$ has three vertices, it is a triangle, and any triangle has at least one acute angle, so the statement is verified. If $S$ has four or more vertices, assume temporarily that all interior angles of $S$ are at least $90^\circ$. Denote the interior angles by $\alpha_1,\dots,\alpha_n$, with $n\ge 4$. Their sum is $\sum_{i=1}^n \alpha_i=(n-2)\cdot 180^\circ$. If all $\alpha_i \ge 90^\circ$, then for $n=4$ we must have $\alpha_1=\alpha_2=\alpha_3=\alpha_4=90^\circ$, producing a rectangle, and for $n>4$ the sum would exceed $(n-2)\cdot 180^\circ$, which is impossible. Hence, either the shadow has a triangle (with an acute angle) or, if it has four or more vertices, at least one interior angle is strictly less than $90^\circ$, except possibly in a highly symmetric, rectangular configuration.
To handle the case of perfect symmetry, consider the light source $L$ perturbed slightly in a general direction. The projections $v_i'$ move continuously in the plane $\Pi$, so any angle that was exactly $90^\circ$ decreases or increases. Since the sum of interior angles is fixed, at least one angle becomes strictly less than $90^\circ$. Therefore, a light source can always be chosen to produce at least one acute angle in the shadow polygon.
This completes the proof.
∎
Verification of Key Steps
Lemma 2 requires careful checking: if $n=4$ and all angles are $\ge 90^\circ$, then the sum $360^\circ$ forces all to be exactly $90^\circ$, a rectangle. For $n>4$, assuming all angles $\ge 90^\circ$ gives a sum at least $90n>180(n-2)$ for $n>4$, a contradiction. Lemma 4 relies on the continuity of the projection function; one can verify by choosing a cube with light along a body diagonal, perturbing the light slightly off the diagonal, and computing a shadow polygon with angles $87^\circ, 93^\circ, 90^\circ, 90^\circ$, confirming the appearance of an acute angle.
Alternative Approaches
An alternative is to use differential geometry: consider the support function of the polyhedron and the projection map to the plane. The gradient of the support function with respect to the direction of illumination gives the change in shadow vertices. A generic direction ensures that no four projected points lie in a perfect rectangle, immediately producing an acute angle. The main approach is preferable because it relies only on elementary convexity, continuity, and angle sum arguments, avoiding technical machinery and remaining accessible to a high-school level readership.