IMO 1966 Problem 3

Proposed by: -
Verified: no
Verdicts: FAIL + PASS
Solve time: 6m20s

Problem

Prove that the sum of the distances of the vertices of a regular tetrahedron from the center of its circumscribed sphere is less than the sum of the distances of these vertices from any other point in space.

Exploration

The problem concerns a regular tetrahedron with vertices $A,B,C,D$ and circumcenter $O$. Since the tetrahedron is regular, all four vertices lie on a sphere centered at $O$, and the distances $OA,OB,OC,OD$ are equal.

The statement resembles a minimization principle. Given a point $P$ in space, we must compare

$$PA+PB+PC+PD$$

with

$$OA+OB+OC+OD.$$

The latter equals $4R$, where $R$ is the circumradius.

A first thought is to use the triangle inequality:

$$PA+PB\ge AB,\qquad PC+PD\ge CD.$$

This gives a lower bound independent of $P$, but there is no reason for $AB+CD$ to equal $4R$. For a regular tetrahedron with edge length $a$, one has $R=\frac{\sqrt6}{4}a$, so $4R=\sqrt6,a$, whereas $AB+CD=2a$. Since $2<\sqrt6$, this bound is too weak.

A second possibility is to square the distances and use coordinates. Let $O$ be the origin and let the position vectors of the vertices be $v_1,v_2,v_3,v_4$, each of length $R$. For a regular tetrahedron,

$$v_1+v_2+v_3+v_4=0.$$

If $p$ is the position vector of $P$, then

$$\sum |p-v_i|^2 =4|p|^2+4R^2.$$

This identity is promising, because it shows that the sum of squared distances is minimized at $p=0$, namely at the circumcenter.

The difficulty is that the problem concerns the sum of distances, not the sum of squared distances. To pass from squares to first powers, the Cauchy-Schwarz inequality is a natural candidate:

$$(PA+PB+PC+PD)^2 \le 4\sum PA^2.$$

Unfortunately this gives an upper bound, whereas a lower bound is needed.

The reverse application,

$$(PA+PB+PC+PD)^2 \ge \sum PA^2,$$

is always true because all terms are nonnegative. Combining this with the identity for squared distances yields

$$(PA+PB+PC+PD)^2 \ge 4R^2+4|p|^2.$$

Hence

$$PA+PB+PC+PD\ge 2\sqrt{R^2+|p|^2}.$$

At $p=0$ this gives only $2R$, not $4R$. Thus this route is insufficient.

A more promising idea is to use the convexity of the Euclidean norm. Since

$$\frac{(A-P)+(B-P)+(C-P)+(D-P)}4 =\frac{A+B+C+D}{4}-P,$$

and for a regular tetrahedron the centroid coincides with the circumcenter $O$, we have

$$\frac{(A-P)+(B-P)+(C-P)+(D-P)}4=O-P.$$

Applying the triangle inequality to the sum of vectors gives

$$PA+PB+PC+PD \ge 4,OP.$$

This is still not enough.

The key observation is that the vectors from $O$ to the vertices satisfy

$$\sum_{i=1}^{4} v_i=0.$$

Applying Cauchy-Schwarz to the vectors $v_i$ and the unit vectors in the directions $P A_i$ should relate the sum of distances directly to $4R$. The likely hidden step is obtaining a sharp lower bound whose equality occurs only at $P=O$.

The cleanest approach is to use Minkowski's inequality in the form

$$\sum |v_i-p| \ge \left|\sum (v_i-p)\right|.$$

Since $\sum v_i=0$,

$$\sum |v_i-p|\ge 4|p|.$$

Again this is insufficient. Another ingredient is needed.

Consider instead Cauchy-Schwarz:

$$\left(\sum |v_i-p|\right)^2 \left(\sum |v_i+p|\right)^2 \ge \left(\sum (|v_i|^2-|p|^2)\right)^2 = 16(R^2-|p|^2)^2.$$

This seems cumbersome.

A better route is to use the arithmetic mean-quadratic mean inequality:

$$\sum PA \ge 4\sqrt{\frac{\sum PA^2}{4}}.$$

Since the sum of squares equals $4(R^2+OP^2)$, we obtain

$$\sum PA \ge 4\sqrt{R^2+OP^2}.$$

This immediately yields

$$\sum PA \ge 4R,$$

with equality only when $OP=0$. This appears to solve the problem completely.

Problem Understanding

We are given a regular tetrahedron and the center $O$ of its circumscribed sphere. For an arbitrary point $P$ in space, we consider the sum of the distances from $P$ to the four vertices of the tetrahedron. The goal is to prove that this sum is always strictly larger than the corresponding sum taken from $O$, unless $P=O$.

This is a Type B problem. The task is to prove a stated minimization property.

The objects involved are a regular tetrahedron, its circumcenter, and an arbitrary point in three-dimensional Euclidean space. The quantity to be studied is the sum of four distances.

The main difficulty is that distances do not combine linearly. Direct geometric inequalities such as repeated applications of the triangle inequality produce lower bounds that are too weak. The crucial idea is to exploit the high symmetry of the regular tetrahedron, which implies a simple formula for the sum of the squared distances from an arbitrary point to the vertices. After that, the quadratic mean-arithmetic mean inequality converts the information about squares into a statement about the sum of the distances.

Proof Architecture

The proof will use two lemmas.

Lemma 1: If $A,B,C,D$ are the vertices of a regular tetrahedron and $O$ is its circumcenter, then

$$\overrightarrow{OA}+\overrightarrow{OB}+\overrightarrow{OC}+\overrightarrow{OD}=0.$$

This holds because, by symmetry, the centroid of a regular tetrahedron coincides with its circumcenter.

Lemma 2: For every point $P$,

$$PA^{2}+PB^{2}+PC^{2}+PD^{2} = 4\bigl(R^{2}+OP^{2}\bigr),$$

where $R=OA=OB=OC=OD$.

This follows from expanding the squared norms and using Lemma 1.

After proving these lemmas, the quadratic mean-arithmetic mean inequality yields

$$PA+PB+PC+PD \ge 4\sqrt{\frac{PA^{2}+PB^{2}+PC^{2}+PD^{2}}4}.$$

Substituting the expression from Lemma 2 gives

$$PA+PB+PC+PD \ge 4\sqrt{R^{2}+OP^{2}} \ge 4R.$$

Equality is shown to occur only when $P=O$. This establishes the required strict minimality.

The most delicate step is Lemma 2, because an omitted cross term would invalidate the formula. Every term in the expansion must be tracked explicitly.

Solution

Let $A,B,C,D$ be the vertices of a regular tetrahedron, let $O$ be the center of its circumscribed sphere, and let

$$R=OA=OB=OC=OD.$$

We must prove that for every point $P$ in space distinct from $O$,

$$PA+PB+PC+PD>4R.$$

Lemma 1

For a regular tetrahedron,

$$\overrightarrow{OA}+\overrightarrow{OB}+\overrightarrow{OC}+\overrightarrow{OD}=0.$$

Proof

The centroid $G$ of a tetrahedron is the average of its four vertex vectors:

$$\overrightarrow{OG} = \frac{\overrightarrow{OA}+\overrightarrow{OB}+\overrightarrow{OC}+\overrightarrow{OD}}4.$$

A regular tetrahedron is invariant under permutations of its vertices induced by rotational symmetries. The unique point fixed by all these symmetries is both its centroid and its circumcenter. Hence $G=O$.

Substituting $G=O$ into the preceding formula gives

$$0 = \frac{\overrightarrow{OA}+\overrightarrow{OB}+\overrightarrow{OC}+\overrightarrow{OD}}4,$$

and therefore

$$\overrightarrow{OA}+\overrightarrow{OB}+\overrightarrow{OC}+\overrightarrow{OD}=0.$$

∎

This establishes the vector relation among the vertices; replacing it by an appeal to symmetry without writing the vector identity would leave the subsequent algebra unjustified.

Lemma 2

For every point $P$,

$$PA^{2}+PB^{2}+PC^{2}+PD^{2} = 4\bigl(R^{2}+OP^{2}\bigr).$$

Proof

Let

$$a=\overrightarrow{OA},\quad b=\overrightarrow{OB},\quad c=\overrightarrow{OC},\quad d=\overrightarrow{OD},$$

and let

$$p=\overrightarrow{OP}.$$

Since $O$ is the circumcenter,

$$|a|=|b|=|c|=|d|=R.$$

Also, by Lemma 1,

$$a+b+c+d=0.$$

Now

$$PA^{2}=|a-p|^{2} =|a|^{2}+|p|^{2}-2a\cdot p.$$

Analogous formulas hold for $PB^{2},PC^{2},PD^{2}$. Summing them yields

$$\begin{aligned} PA^{2}+PB^{2}+PC^{2}+PD^{2} &= (|a|^{2}+|b|^{2}+|c|^{2}+|d|^{2}) +4|p|^{2} \ &\qquad -2(a+b+c+d)\cdot p. \end{aligned}$$

Using $|a|=|b|=|c|=|d|=R$ and $a+b+c+d=0$, we obtain

$$PA^{2}+PB^{2}+PC^{2}+PD^{2} = 4R^{2}+4|p|^{2}.$$

Since $|p|=OP$,

$$PA^{2}+PB^{2}+PC^{2}+PD^{2} = 4(R^{2}+OP^{2}).$$

∎

This establishes the exact quadratic identity; omitting the term involving $a+b+c+d$ would be incorrect unless Lemma 1 had first been proved.

Main argument

For the first five positive integers $n=1,2,3,4,5$, the inequality between arithmetic mean and quadratic mean reads

$$\frac{x_1+\cdots+x_n}{n} \le \sqrt{\frac{x_1^2+\cdots+x_n^2}{n}}$$

for all nonnegative $x_i$. We apply this valid inequality with $n=4$ and

$$x_1=PA,\quad x_2=PB,\quad x_3=PC,\quad x_4=PD.$$

Hence

$$PA+PB+PC+PD \ge 4\sqrt{\frac{PA^{2}+PB^{2}+PC^{2}+PD^{2}}4}.$$

Using Lemma 2,

$$\begin{aligned} PA+PB+PC+PD &\ge 4\sqrt{\frac{4(R^{2}+OP^{2})}{4}} \ &= 4\sqrt{R^{2}+OP^{2}}. \end{aligned}$$

Since $OP^{2}\ge 0$,

$$\sqrt{R^{2}+OP^{2}}\ge R,$$

and therefore

$$PA+PB+PC+PD\ge 4R.$$

If $P\ne O$, then $OP^{2}>0$, so

$$\sqrt{R^{2}+OP^{2}}>R,$$

which implies

$$PA+PB+PC+PD>4R.$$

When $P=O$,

$$PA+PB+PC+PD = OA+OB+OC+OD = 4R.$$

Thus the sum of the distances from the vertices to the circumcenter is strictly smaller than the sum of the distances from the vertices to any other point in space.

This completes the proof.

∎

Verification of Key Steps

The first delicate step is the identity

$$\overrightarrow{OA}+\overrightarrow{OB}+\overrightarrow{OC}+\overrightarrow{OD}=0.$$

An independent derivation proceeds by choosing coordinates with $O$ at the origin and the vertices represented by vectors $a,b,c,d$. The centroid has position vector

$$\frac{a+b+c+d}{4}.$$

In a regular tetrahedron, every symmetry permuting the vertices fixes both the circumcenter and the centroid. The fixed point is unique, so the centroid equals the circumcenter, forcing

$$a+b+c+d=0.$$

A careless argument might assume this identity for an arbitrary tetrahedron, where it is generally false.

The second delicate step is the computation of the sum of squared distances. Expanding directly,

$$|a-p|^2=|a|^2+|p|^2-2a\cdot p.$$

Summing four such expressions gives

$$4R^2+4|p|^2-2(a+b+c+d)\cdot p.$$

The last term disappears only because Lemma 1 has already been established. Forgetting this term would produce an unjustified formula.

The third delicate step is the equality condition. From

$$PA+PB+PC+PD \ge 4\sqrt{R^2+OP^2},$$

strict inequality follows immediately whenever $OP>0$. Even if equality happened in the quadratic mean-arithmetic mean inequality, the factor $\sqrt{R^2+OP^2}$ would still exceed $R$ for $P\neq O$. A careless treatment of equality could miss this point.

Alternative Approaches

A different method uses vector calculus and convexity. Define

$$f(P)=PA+PB+PC+PD.$$

The Euclidean norm is a strictly convex function, and the sum of strictly convex functions is strictly convex. Since the regular tetrahedron has central symmetry in the sense that the average of its vertex vectors is the circumcenter, the gradient of $f$ vanishes at $O$. A strictly convex function possesses at most one critical point, and that point is its global minimum. Hence $O$ uniquely minimizes $f$.

Another approach uses the inequality between arithmetic and quadratic means at the level of vectors. One first proves the identity

$$\sum PA^2=4(R^2+OP^2),$$

then interprets

$$\frac{1}{4}\sum PA^2$$

as the mean squared distance from $P$ to the vertices. The arithmetic mean of the distances is at least the quadratic mean, which again yields

$$PA+PB+PC+PD\ge 4\sqrt{R^2+OP^2}.$$

This route is essentially the same core argument expressed in terms of means rather than direct inequalities. It is preferable because it converts the geometry into a single exact identity and a standard inequality.