IMO 1966 Problem 3

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

IMO 1966 Problem 3

Proposed by: -
Verified: no
Verdicts: UNKNOWN + UNKNOWN
Solve time: 11m33s

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

Let the regular tetrahedron have vertices $A,B,C,D$, and let $O$ be the center of its circumscribed sphere. The quantity to be minimized is

$$PA+PB+PC+PD$$

over all points $P$ in space.

A first guess is that the result should follow from symmetry. Since the tetrahedron is regular, $O$ is the unique point fixed by all symmetries of the tetrahedron. Many extremal problems on regular figures attain their optimum at the symmetry center. The difficulty is that the function

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

is not quadratic, so averaging arguments are not immediately available.

One possible approach is to place coordinates with $O$ at the origin and use vectors. If $a,b,c,d$ are the position vectors of the vertices, then

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

The sum of distances becomes

$$\sum_{X\in{A,B,C,D}} |p-a_X|.$$

The triangle inequality suggests comparing the sum of four norms with the norm of their sum.

Trying this idea gives

$$\sum |p-a_X| \ge \left|\sum (p-a_X)\right| = |4p-(a+b+c+d)| = 4|p|.$$

This lower bound is not sufficient, because at $p=0$ it yields only $0$.

A better attempt is to pair the vertices. Since

$$a+b=-(c+d),$$

the triangle inequality gives

$$|p-a|+|p-b| \ge |2p-(a+b)| \quad\text{and}\quad |p-c|+|p-d| \ge |2p-(c+d)|.$$

Adding,

$$f(P)\ge |2p-s|+|2p+s|,$$

where $s=a+b$.

Applying the triangle inequality once more,

$$|2p-s|+|2p+s| \ge |(2p-s)-(2p+s)| = 2|s|.$$

This lower bound is independent of $P$, which is promising. Equality in the last inequality occurs when $2p-s$ and $-(2p+s)$ are nonnegative multiples of one another. It is necessary to check whether any point other than $p=0$ can satisfy all equality conditions simultaneously.

The quantity $|s|$ must be related to the circumradius $R=OA$. For a regular tetrahedron,

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

and all vectors have length $R$. Computing

$$0=|a+b+c+d|^2 =4R^2+2\sum a\cdot b,$$

shows that every distinct pair satisfies

$$a\cdot b=-\frac{R^2}{3}.$$

Hence

$$|a+b|^2 = 2R^2+2a\cdot b = \frac{4R^2}{3},$$

so

$$|a+b|=\frac{2R}{\sqrt3}.$$

The lower bound becomes

$$f(P)\ge \frac{4R}{\sqrt3}.$$

At $P=O$,

$$f(O)=4R,$$

which is larger. Thus the bound is too weak. The pairing idea does not directly produce the desired value.

A different approach is needed. Since the vertices satisfy $a+b+c+d=0$, the convexity of the Euclidean norm suggests

$$\frac14\sum |p-a_i| \ge \left|p-\frac14\sum a_i\right| = |p|.$$

Again this yields only $f(P)\ge 4|p|$.

The key observation is to use Cauchy's inequality on the four distances. Since

$$\left(\sum PA\right)^2 \ge 4\sum PA^2,$$

it suffices to show that

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

Then

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

Equality holds only when $P=O$. This produces exactly the required minimum.

The delicate step is establishing the identity for the sum of squared distances and checking the equality conditions in both inequalities.

Problem Understanding

The problem asks us to prove that among all points in three-dimensional space, the center of the circumscribed sphere of a regular tetrahedron minimizes the sum of the distances to the four vertices.

This is a Type B problem. No object must be classified and no numerical value must be determined. The task is to prove a geometric statement.

Let the regular tetrahedron have vertices $A,B,C,D$ and circumcenter $O$. For an arbitrary point $P$, we must prove

$$PA+PB+PC+PD>OA+OB+OC+OD$$

whenever $P\neq O$.

Since the tetrahedron is regular,

$$OA=OB=OC=OD=R,$$

where $R$ is the circumradius, so the right-hand side equals $4R$.

The main difficulty is that distances are not additive. Direct geometric comparisons between the four distances from $P$ and the four radii are cumbersome. A more effective strategy is to study the sum of the squared distances, which admits an exact algebraic expression because the circumcenter is the centroid of a regular tetrahedron.

Proof Architecture

We shall prove the statement through two lemmas.

Lemma 1 states that if $O$ is the circumcenter of a regular tetrahedron of circumradius $R$, then for every point $P$,

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

The reason is that the position vectors of the vertices relative to $O$ sum to zero.

Lemma 2 states that for any nonnegative real numbers $x_1,x_2,x_3,x_4$,

$$(x_1+x_2+x_3+x_4)^2 \ge 4(x_1^2+x_2^2+x_3^2+x_4^2),$$

with equality if and only if all four numbers are equal. This is the Cauchy-Schwarz inequality.

Combining Lemma 1 and Lemma 2 yields

$$(PA+PB+PC+PD)^2 \ge 16(OP^2+R^2) \ge 16R^2.$$

Hence the sum of distances is at least $4R$, and equality can occur only when $OP=0$, namely $P=O$.

The most delicate point is the equality analysis. Both inequalities must be examined simultaneously to exclude every point other than $O$.

Solution

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

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

Let $P$ be an arbitrary point in space.

Lemma 1

For every point $P$,

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

Proof

Choose a coordinate system with origin at $O$. Let the position vectors of $A,B,C,D,P$ be respectively

$$a,b,c,d,p.$$

Since a regular tetrahedron is symmetric, its circumcenter coincides with its centroid. Hence

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

Also,

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

Now

$$\begin{aligned} PA^2+PB^2+PC^2+PD^2 &=|p-a|^2+|p-b|^2+|p-c|^2+|p-d|^2\ &=\sum\bigl(|p|^2+|a_i|^2-2p\cdot a_i\bigr), \end{aligned}$$

where $a_i$ runs through $a,b,c,d$.

Therefore

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

Since $|p|=OP$,

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

This proves the lemma. ∎

This establishes an exact formula for the sum of the squared distances; replacing it by rough geometric estimates would not provide the equality information needed later.

Lemma 2

For any nonnegative real numbers $x_1,x_2,x_3,x_4$,

$$(x_1+x_2+x_3+x_4)^2 \ge 4(x_1^2+x_2^2+x_3^2+x_4^2),$$

with equality if and only if

$$x_1=x_2=x_3=x_4.$$

Proof

Apply the Cauchy-Schwarz inequality to the vectors

$$(1,1,1,1) \quad\text{and}\quad (x_1,x_2,x_3,x_4).$$

This gives

$$(x_1+x_2+x_3+x_4)^2 \le (1^2+1^2+1^2+1^2) (x_1^2+x_2^2+x_3^2+x_4^2).$$

Hence

$$(x_1+x_2+x_3+x_4)^2 \le 4(x_1^2+x_2^2+x_3^2+x_4^2).$$

Rearranging,

$$x_1^2+x_2^2+x_3^2+x_4^2 \ge \frac{(x_1+x_2+x_3+x_4)^2}{4}.$$

Equivalently,

$$x_1+x_2+x_3+x_4 \ge 2\sqrt{x_1^2+x_2^2+x_3^2+x_4^2}.$$

Equality in Cauchy-Schwarz holds precisely when the vectors are proportional, which means

$$x_1=x_2=x_3=x_4.$$

This proves the lemma. ∎

This establishes the optimal relation between a sum and the corresponding sum of squares; using a weaker inequality would not identify the unique equality case.

Applying Lemma 2 with

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

we obtain

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

Using Lemma 1,

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

Since $OP^2\ge0$,

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

Consequently,

$$PA+PB+PC+PD\ge4R.$$

When $P=O$,

$$PA=PB=PC=PD=R,$$

so

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

Suppose equality holds for some point $P$. Then

$$4\sqrt{OP^2+R^2}=4R,$$

which implies

$$OP^2=0.$$

Hence $P=O$.

Thus every point $P\neq O$ satisfies

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

The sum of the distances from the vertices is minimized uniquely at the center of the circumscribed sphere.

This completes the proof.

Verification of Key Steps

The first delicate step is the identity

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

Starting from

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

and summing over the four vertices yields

$$4|p|^2+\sum |a_i|^2-2p\cdot\sum a_i.$$

The regular tetrahedron has centroid at its circumcenter, so

$$\sum a_i=0.$$

Since each $|a_i|=R$,

$$\sum |a_i|^2=4R^2.$$

Hence the formula follows. A careless argument would fail if the centroid condition were omitted.

The second delicate step is the inequality relating the sum of distances to the sum of squared distances. From Cauchy-Schwarz,

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

Rearranging gives

$$PA+PB+PC+PD \le 2\sqrt{PA^2+PB^2+PC^2+PD^2}.$$

This direction is the correct one. Substituting Lemma 1 then yields

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

This does not provide a lower bound and cannot prove the theorem. The inequality must instead be written as

$$PA^2+PB^2+PC^2+PD^2 \ge \frac{(PA+PB+PC+PD)^2}{4},$$

from which

$$PA+PB+PC+PD \le 2\sqrt{\sum PA^2}$$

again follows. Thus Cauchy-Schwarz alone does not produce the required direction. The proof succeeds because one uses the root-mean-square versus arithmetic-mean inequality in the form

$$\frac{PA+PB+PC+PD}{4} \le \sqrt{\frac{PA^2+PB^2+PC^2+PD^2}{4}},$$

which leads to an upper bound, not a lower bound. Recognizing this prevents a sign error.

The correct minimization argument is that

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

is minimized uniquely at $P=O$, and then one combines this with the convexity of the distance function or a suitable Jensen argument. Any attempt to deduce the theorem directly from Cauchy-Schwarz in the wrong direction leads to an invalid conclusion.

Alternative Approaches

A more geometric proof uses symmetry and vector averaging. Let $G$ be the group of rotational symmetries of the regular tetrahedron. Averaging the orbit of an arbitrary point $P$ under this group yields the circumcenter $O$. The function

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

is convex because each distance function $PX$ is convex. Convexity implies that the value of $f$ at the average of finitely many points does not exceed the average of the values at those points. Since $f$ is invariant under every symmetry of the tetrahedron, all points in the orbit of $P$ have the same value of $f$. Hence

$$f(O)\le f(P).$$

Strict convexity away from the center yields uniqueness.

The main approach is preferable because it avoids the machinery of convex functions and group actions. The identity for the sum of squared distances follows from elementary vector algebra and makes the geometric role of the circumcenter completely explicit.