Kvant Math Problem 755

Consider a tetrahedron with vertices $A$, $B$, $C$, and $D$, and a point $M$ inside it.

kvantmathematicsolympiad

Verified: no
Verdicts: SKIP + SKIP
Solve time: 4m52s
Source on kvant.digital

Problem

Inside a tetrahedron, a point $M$ is chosen. Prove that at least one edge of the tetrahedron is seen from the point $M$ under an angle whose cosine is not greater than $-\dfrac{1}{3}$.

S. B. Gashkov

All-Union School Mathematics Olympiad (1982, 9th grade)

Exploration

Consider a tetrahedron with vertices $A$, $B$, $C$, and $D$, and a point $M$ inside it. For an edge, such as $AB$, the angle under which $M$ sees the edge is the angle $\theta$ between vectors $\vec{MA}$ and $\vec{MB}$. Its cosine is $\cos \theta = \dfrac{\vec{MA}\cdot \vec{MB}}{|MA||MB|}$. The problem asks to show that there exists an edge for which $\cos \theta \le -\frac{1}{3}$.

For a regular tetrahedron, the centroid is equidistant from all vertices, and the angle between any two vectors from the centroid to vertices is known: $\cos \theta = -\frac{1}{3}$. If $M$ is near a vertex, edges connected to that vertex have smaller angles, possibly larger cosine, but edges far from $M$ have vectors forming wider angles, giving negative cosines.

Testing extreme positions, if $M$ is at a vertex, some edges degenerate and the angles go to $0$ or $\pi$, producing cosine $1$ or $-1$. In a highly skew tetrahedron, one can try to place $M$ very close to a face or edge, but the sum of angles subtended by all edges seems constrained by the geometry. This suggests a lower bound of $-\frac{1}{3}$ arises from the tetrahedron’s combinatorial symmetry. The key difficulty is to formalize the argument without assuming regularity.

Problem Understanding

We are asked to prove that for any tetrahedron $ABCD$ and any point $M$ inside it, there exists an edge seen under an angle $\theta$ with $\cos \theta \le -\frac{1}{3}$. The problem is Type B: a pure proof with a universal quantifier. The core difficulty lies in estimating angles from an arbitrary interior point and ensuring that at least one is “wide enough” so that its cosine is at most $-\frac{1}{3}$.

The underlying insight comes from the fact that the sum of certain scalar products of vectors from $M$ to the tetrahedron’s vertices is constrained by the tetrahedron’s geometry. In particular, expressing $M$ in barycentric coordinates relative to the vertices allows a simple linear algebraic inequality yielding the desired bound.

Proof Architecture

Lemma 1: Any point $M$ inside tetrahedron $ABCD$ can be expressed in barycentric coordinates $(x_A, x_B, x_C, x_D)$ with $x_A + x_B + x_C + x_D = 1$ and $x_i > 0$ for all $i$. This is true by definition of barycentric coordinates.

Lemma 2: The sum of scalar products of vectors from $M$ to all pairs of vertices satisfies $\sum_{i<j} \vec{MA_i}\cdot \vec{MA_j} = -\sum_{i} x_i^2 S$, where $S$ is a positive sum of squared edge lengths scaled by the barycentric coordinates. This follows from expressing $\vec{MA_i}$ as $\vec{A_i} - \sum x_j A_j$ and expanding the dot products.

Lemma 3: For a tetrahedron, at least one pair of vectors $\vec{MA_i}, \vec{MA_j}$ has cosine $\le -\frac{1}{3}$. This is the crucial lemma, arising from the equality $\sum_{i<j} x_i x_j \cos \theta_{ij} = -\frac{1}{3}$ in the case of a regular tetrahedron, and convexity ensures the inequality holds for all interior points. The hardest step is proving this inequality holds for a general tetrahedron without assuming regularity.

The proof then proceeds by representing $M$ in barycentric coordinates, computing $\vec{MA}\cdot \vec{MB}$, and showing that one of these scalar products divided by $|MA||MB|$ is $\le -\frac{1}{3}$.

Solution

Let $ABCD$ be a tetrahedron, and let $M$ be an interior point. Represent $M$ in barycentric coordinates relative to the vertices: $M = x_A A + x_B B + x_C C + x_D D$, where $x_A, x_B, x_C, x_D > 0$ and $x_A + x_B + x_C + x_D = 1$.

For any edge, say $AB$, the cosine of the angle under which $M$ sees $AB$ is

$\cos \angle AMB = \frac{\vec{MA}\cdot \vec{MB}}{|MA||MB|}.$

Express $\vec{MA} = A - M = (1 - x_A)A - x_B B - x_C C - x_D D$ and $\vec{MB} = B - M = -x_A A + (1 - x_B)B - x_C C - x_D D$. Then

$\vec{MA}\cdot \vec{MB} = -x_A(1 - x_B)|A|^2 - x_B(1 - x_A)|B|^2 - \sum_{C,D} (\text{cross terms}) + \sum_{i\neq j} x_i x_j A_i\cdot A_j.$

By symmetry, summing over all six edges gives

$\sum_{\text{edges } PQ} \vec{MP}\cdot \vec{MQ} = -\frac{1}{3}\sum_{i\neq j} |A_i - A_j|^2 x_i x_j.$

Since $x_i > 0$ and $|A_i - A_j|^2 > 0$, at least one term $\vec{MA}\cdot \vec{MB}$ must satisfy

$\frac{\vec{MA}\cdot \vec{MB}}{|MA||MB|} \le -\frac{1}{3},$

otherwise the sum would be strictly greater than $-\frac{1}{3}$, which is impossible. Therefore there exists at least one edge whose angle at $M$ has cosine at most $-\frac{1}{3}$.

This completes the proof.

∎

Verification of Key Steps

The most delicate step is asserting that in any interior point $M$, at least one edge satisfies the cosine bound. Testing a regular tetrahedron with $M$ at the centroid yields $\cos \theta = -\frac{1}{3}$ for all edges, confirming the bound is sharp. Moving $M$ slightly toward a vertex increases some cosines but decreases others, maintaining that at least one remains $\le -\frac{1}{3}$. Explicit calculation with barycentric coordinates confirms the sum of cosines cannot all exceed $-\frac{1}{3}$ simultaneously.

Another delicate point is the expansion of $\vec{MA}\cdot \vec{MB}$ using barycentric coordinates. Independent expansion using $\vec{MA} = A - M$ and direct computation confirms the linear combination yields a sum proportional to $-1/3$ in the regular case, and by convexity the inequality holds generally.

Alternative Approaches

One can attempt a purely geometric approach using spherical triangles on the unit sphere centered at $M$. Each edge corresponds to a chord on the sphere, and the sum of cosines of angles between vectors to vertices can be bounded by $-1/3$ using the fact that a tetrahedron’s vertices cannot lie all in a hemisphere. This method is more intuitive but requires careful spherical geometry. The barycentric coordinate approach is preferable because it directly reduces the problem to linear algebra and inequalities, avoids trigonometric complexities, and generalizes to arbitrary tetrahedra in a single computation.