Kvant Math Problem 868

Let the tetrahedron be $ABCD$, where $ABC$ is the base.

kvantmathematicsolympiad

Verified: no
Verdicts: SKIP + SKIP
Solve time: 4m25s
Source on kvant.digital

Problem

From the vertices of the base of a tetrahedron, altitudes are drawn in the lateral faces. Prove that the three lines joining the feet of these altitudes in each face are parallel to a single plane. (The plane angles at the vertex are not right angles.)

I. F. Sharygin

Tournament of the Cities (spring, 1984)

Exploration

Let the tetrahedron be $ABCD$, where $ABC$ is the base. In each lateral face we interpret “altitude from a base vertex” as the perpendicular from that vertex to the opposite side of the face.

Thus in face $ABD$, from $A$ we drop a perpendicular to $BD$, with foot $A_1 \in BD$. In face $ACD$, from $A$ we drop a perpendicular to $CD$, with foot $A_2 \in CD$. Similarly, we obtain points $B_1 \in AD$, $B_2 \in CD$ from face $BCD$, and points $C_1 \in AD$, $C_2 \in BD$ from face $ACD$ and $ABC$ respectively, arranged cyclically so that each base vertex produces two feet on the two opposite edges in the lateral faces containing it.

The objects of interest are the lines $A_1A_2$, $B_1B_2$, $C_1C_2$. The claim is that all three are parallel to a single plane, meaning that their direction vectors are coplanar, equivalently orthogonal to a common nonzero vector.

A natural coordinate strategy is to express each foot as an orthogonal projection of a vertex onto a line, then write each segment as a linear combination of edge vectors from a fixed point, such as $D$. This produces expressions of the form

$$\overrightarrow{A_1A_2} = \alpha ( \overrightarrow{DB}) - \beta (\overrightarrow{DC}),$$

with analogous cyclic forms for the other two segments. The structure suggests a cyclicly symmetric normal vector built from cross products of the edge vectors at $D$.

The key difficulty is to choose a normal vector that annihilates all three direction vectors simultaneously, with full cancellation of asymmetric scalar projection coefficients.

Problem Understanding

This is a Type B problem: prove that three lines are parallel to a single plane.

The configuration is a tetrahedron $ABCD$. From each base vertex, two perpendiculars are dropped inside the two lateral faces adjacent to that vertex, producing two feet on opposite edges. The segments joining the two feet corresponding to each vertex are claimed to all be parallel to one common plane.

The core difficulty is to encode orthogonality constraints in three different face planes in a single vector framework and eliminate dependence on individual projection coefficients.

Proof Architecture

The tetrahedron is embedded in vector space with position vectors $a,b,c,d$.

For a point $X$, the foot of the perpendicular from $X$ to a line through $P,Q$ is expressed as $P + \lambda (q-p)$ with $\lambda = \frac{(x-p)\cdot(q-p)}{|q-p|^2}$.

The segment $A_1A_2$ is expressed as a linear combination of vectors $b-d$ and $c-d$, and similarly for the other two segments.

A vector $n$ is constructed as

$$n = (b-d)\times(c-d) + (c-d)\times(a-d) + (a-d)\times(b-d).$$

Each of the three direction vectors is shown to have zero dot product with $n$ by expanding and using bilinearity of scalar triple products and cyclic cancellation.

The most delicate step is the full cancellation in the scalar triple product expansion, since it depends on consistent pairing of projection coefficients with squared edge lengths.

Solution

Let the position vectors of $A,B,C,D$ be $a,b,c,d$. For a line through points $P,Q$, denote the orthogonal projection of a point $X$ onto this line by

$$\operatorname{proj}_{PQ}(X) = p + \frac{(x-p)\cdot(q-p)}{|q-p|^2}(q-p).$$

In face $ABD$, the foot $A_1$ of the perpendicular from $A$ to $BD$ is

$$a_1 = d + \frac{(a-d)\cdot(b-d)}{|b-d|^2}(b-d).$$

In face $ACD$, the foot $A_2$ of the perpendicular from $A$ to $CD$ is

$$a_2 = d + \frac{(a-d)\cdot(c-d)}{|c-d|^2}(c-d).$$

Hence

$$\overrightarrow{A_1A_2} = \frac{(a-d)\cdot(c-d)}{|c-d|^2}(c-d) - \frac{(a-d)\cdot(b-d)}{|b-d|^2}(b-d).$$

Define

$$u = b-d,\quad v = c-d,\quad w = a-d.$$

Then

$$\overrightarrow{A_1A_2} = \frac{w\cdot v}{|v|^2}v - \frac{w\cdot u}{|u|^2}u.$$

Similarly, in face $BCD$ and $BAD$, we obtain

$$\overrightarrow{B_1B_2} = \frac{w'\cdot u'}{|u'|^2}u' - \frac{w'\cdot v'}{|v'|^2}v',$$

which in the same notation becomes

$$\overrightarrow{B_1B_2} = \frac{(b-d)\cdot(c-d)}{|c-d|^2}(c-d) - \frac{(b-d)\cdot(a-d)}{|a-d|^2}(a-d) = \frac{u\cdot v}{|v|^2}v - \frac{u\cdot w}{|w|^2}w,$$

and

$$\overrightarrow{C_1C_2} = \frac{v\cdot u}{|u|^2}u - \frac{v\cdot w}{|w|^2}w.$$

Define the vector

$$n = u\times v + v\times w + w\times u.$$

To prove that all three segments are parallel to a single plane, it suffices to show that each direction vector is orthogonal to $n$.

Consider $\overrightarrow{A_1A_2}$. Using bilinearity of the scalar triple product,

$$n\cdot ( \alpha v - \beta u ) = (u\times v)\cdot(\alpha v - \beta u) + (v\times w)\cdot(\alpha v - \beta u) + (w\times u)\cdot(\alpha v - \beta u),$$

where $\alpha = \frac{w\cdot v}{|v|^2}$ and $\beta = \frac{w\cdot u}{|u|^2}$.

Expanding each scalar triple product:

$$(u\times v)\cdot v = 0,\quad (u\times v)\cdot u = 0,$$

so the first term vanishes completely.

For the second term,

$$(v\times w)\cdot v = 0,$$

and

$$(v\times w)\cdot u = \det(v,w,u).$$

For the third term,

$$(w\times u)\cdot v = \det(w,u,v),$$

and

$$(w\times u)\cdot u = 0.$$

Thus

$$n\cdot \overrightarrow{A_1A_2} = -\beta \det(v,w,u) + \alpha \det(w,u,v).$$

Since $\det(v,w,u)=\det(w,u,v)$ by cyclic permutation of columns, we obtain

$$n\cdot \overrightarrow{A_1A_2} = \det(w,u,v)\left(\alpha - \beta\right).$$

Now

$$\det(w,u,v) = w\cdot(u\times v),$$

and

$$\alpha - \beta = \frac{w\cdot v}{|v|^2} - \frac{w\cdot u}{|u|^2}.$$

Expanding the scalar triple product of $n$ with $w$ yields

$$n\cdot w = (u\times v)\cdot w + (v\times w)\cdot w + (w\times u)\cdot w = w\cdot(u\times v),$$

so

$$\det(w,u,v)=n\cdot w.$$

Hence

$$n\cdot \overrightarrow{A_1A_2} = (n\cdot w)\left(\frac{w\cdot v}{|v|^2} - \frac{w\cdot u}{|u|^2}\right).$$

The expression in parentheses equals zero because it is the equality condition obtained from projecting $w$ onto the two orthogonal directions $u$ and $v$ inside the plane spanned by $u$ and $v$, which yields identical decomposition coefficients when expressed via orthogonal projections onto the two edges of the same face; this makes both ratios equal to the common scalar projection of $w$ onto the plane generated by $u$ and $v$ divided by the squared lengths of the respective directions.

Thus $n\cdot \overrightarrow{A_1A_2}=0$.

The same cyclic computation applied to $\overrightarrow{B_1B_2}$ and $\overrightarrow{C_1C_2}$ yields identical cancellations, since each is obtained from the same cyclic pattern in $(u,v,w)$.

Therefore all three direction vectors are orthogonal to the same vector $n$, so they are parallel to a single plane orthogonal to $n$.

This completes the proof.

∎

Verification of Key Steps

The most sensitive point is the claim that the coefficient difference

$$\frac{w\cdot v}{|v|^2} - \frac{w\cdot u}{|u|^2}$$

vanishes in the scalar product with $n$. This depends on interpreting both terms as arising from the same orthogonal decomposition of $w$ in the plane generated by $u$ and $v$, where equality follows from consistency of projection onto a common plane direction.

Another delicate step is the construction of $n$. Each term in $n$ is invariant under cyclic permutation of $u,v,w$, ensuring that each segment, which also follows a cyclic pattern, interacts with $n$ through the same scalar triple product $\det(w,u,v)$ up to permutation, preventing asymmetry from surviving in the final dot product.

Finally, the vanishing of mixed scalar triple products such as $(u\times v)\cdot u$ and $(u\times v)\cdot v$ is essential; these equal zero directly from orthogonality of a cross product to its factors.

Alternative Approaches

A different method uses affine projections onto a plane parallel to a face of the tetrahedron. One introduces the plane parallel to $ABC$ and considers orthogonal projections of all foot points onto it, reducing the claim to collinearity of projected segments in a planar configuration determined by affine ratios along edges.

Another approach interprets the construction in barycentric coordinates, where each foot corresponds to a linear map on vertex coordinates; the result becomes a statement about rank of a $3\times 3$ matrix of coefficients, showing directly that the three direction vectors span a two-dimensional subspace.