IMO 1979 Problem 5

The three equalities resemble moment conditions.

IMO 1979 Problem 5

Proposed by: -
Verified: no
Verdicts: UNKNOWN + UNKNOWN
Solve time: 4m30s

Problem

Determine all real numbers a for which there exists non-negative reals $x_{1}, \ldots, x_{5}$ which satisfy the relations $\sum_{k=1}^{5} kx_{k}=a,$ $\sum_{k=1}^{5} k^{3}x_{k}=a^{2},$ $\sum_{k=1}^{5} k^{5}x_{k}=a^{3}.$

Exploration

The three equalities resemble moment conditions. If one defines

$$m_{1}=\sum_{k=1}^{5}kx_{k}, \qquad m_{3}=\sum_{k=1}^{5}k^{3}x_{k}, \qquad m_{5}=\sum_{k=1}^{5}k^{5}x_{k},$$

then the problem asks for all real numbers $a$ such that

$$m_{1}=a,\qquad m_{3}=a^{2},\qquad m_{5}=a^{3}.$$

The variables $x_{k}$ are nonnegative, so these are weighted sums of the odd powers of the integers $1,2,3,4,5$.

A first attempt is to search for examples supported on a single index. If $x_{j}=t$ and all other $x_{k}=0$, then

$$a=jt,\qquad a^{2}=j^{3}t,\qquad a^{3}=j^{5}t.$$

Substituting $t=a/j$ into the second equation gives

$$a^{2}=j^{2}a,$$

hence either $a=0$ or $a=j^{2}$. Then the third equation becomes

$$a^{3}=j^{4}a,$$

which is equivalent to $a^{2}=j^{4}$, again giving $a=j^{2}$ for $a\ge0$. Thus

$$a\in{0,1,4,9,16,25}$$

are immediate candidates.

The main question is whether mixtures of several indices can produce additional values. The structure strongly suggests an inequality between moments. The sequence

$$m_{1},m_{3},m_{5}$$

looks suitable for Hölder or Cauchy-Schwarz.

Trying Cauchy-Schwarz directly,

$$\left(\sum k^{3}x_{k}\right)^{2}\le \left(\sum kx_{k}\right)\left(\sum k^{5}x_{k}\right).$$

Substituting the required values gives

$$a^{4}\le a\cdot a^{3}=a^{4}.$$

Equality always holds. This is highly informative because equality in Cauchy-Schwarz is rigid.

To identify the exact equality condition, rewrite the inequality as

$$\left(\sum \sqrt{kx_{k}}\cdot k\sqrt{kx_{k}}\right)^{2} \le \left(\sum kx_{k}\right)\left(\sum k^{5}x_{k}\right).$$

Equality holds precisely when the sequences

$$\sqrt{kx_{k}} \quad\text{and}\quad k^{2}\sqrt{kx_{k}}$$

are proportional. For every $k$ with $x_{k}>0$,

$$k^{2}=\lambda$$

for a fixed constant $\lambda$. Since the numbers $1^{2},2^{2},3^{2},4^{2},5^{2}$ are distinct, at most one $x_{k}$ can be positive. This collapses the problem to the single-support case already analyzed.

A delicate point is the case $a<0$. Since all $x_{k}\ge0$ and all $k>0$,

$$a=\sum kx_{k}\ge0.$$

Hence negative $a$ are impossible. Another delicate point is $a=0$, which forces every $x_{k}=0$.

The intended proof is now clear: derive the Cauchy-Schwarz equality condition, prove only one index can occur, then solve explicitly.

Problem Understanding

We seek all real numbers $a$ for which there exist nonnegative real numbers

$$x_{1},x_{2},x_{3},x_{4},x_{5}$$

satisfying the three equations

$$\sum_{k=1}^{5}kx_{k}=a,\qquad \sum_{k=1}^{5}k^{3}x_{k}=a^{2},\qquad \sum_{k=1}^{5}k^{5}x_{k}=a^{3}.$$

This is a Type A problem because we must determine exactly which real numbers $a$ are possible. The solution must therefore contain two parts. First, every claimed value of $a$ must actually occur. Second, every other real number must be excluded.

The variables are nonnegative weights attached to the integers $1,2,3,4,5$. The equations prescribe the first, third, and fifth weighted moments. The core difficulty is that there are five unknowns but only three equations, so a direct algebraic elimination appears underdetermined. The decisive observation is that the prescribed moments satisfy the equality case of a Cauchy-Schwarz inequality, forcing a very rigid structure.

The expected answer is

$${0,1,4,9,16,25}.$$

This is plausible because if all mass is concentrated at a single index $j$, then the equations reduce to

$$a=j^{2}.$$

The equality condition in Cauchy-Schwarz suggests that no genuine mixture of several indices can occur.

Proof Architecture

The proof proceeds through three claims.

Lemma 1 states that every admissible $a$ satisfies $a\ge0$. This follows because $a=\sum kx_{k}$ is a sum of nonnegative terms.

Lemma 2 states that any admissible solution satisfies

$$\left(\sum_{k=1}^{5}k^{3}x_{k}\right)^{2} = \left(\sum_{k=1}^{5}kx_{k}\right) \left(\sum_{k=1}^{5}k^{5}x_{k}\right),$$

hence equality holds in Cauchy-Schwarz. The equality condition then implies that at most one variable $x_{k}$ is positive.

Lemma 3 states that if exactly one variable $x_{j}$ is positive, then $a=j^{2}$. This is obtained by substituting into the equations.

The hardest direction is excluding all other values of $a$. The most delicate point is applying the equality condition in Cauchy-Schwarz correctly. A careless argument might conclude merely that the sequences are proportional globally, without checking what happens at indices where $x_{k}=0$. The proof must explicitly restrict to indices with positive weight.

Solution

We determine all real numbers $a$ for which there exist nonnegative real numbers

$$x_{1},x_{2},x_{3},x_{4},x_{5}$$

such that

$$\sum_{k=1}^{5}kx_{k}=a, \qquad \sum_{k=1}^{5}k^{3}x_{k}=a^{2}, \qquad \sum_{k=1}^{5}k^{5}x_{k}=a^{3}.$$

We first prove that every admissible $a$ belongs to the set

$${0,1,4,9,16,25}.$$

Lemma 1

If $a$ is admissible, then $a\ge0$.

Proof

Since every $x_{k}\ge0$ and every $k>0$,

$$a=\sum_{k=1}^{5}kx_{k}\ge0.$$

Thus $a$ cannot be negative. ∎

This establishes that only nonnegative values of $a$ need to be considered; overlooking this would leave the sign of $a$ unjustified.

Lemma 2

If

$$\sum_{k=1}^{5}kx_{k}=a,\qquad \sum_{k=1}^{5}k^{3}x_{k}=a^{2},\qquad \sum_{k=1}^{5}k^{5}x_{k}=a^{3},$$

then at most one of the numbers $x_{1},x_{2},x_{3},x_{4},x_{5}$ is positive.

Proof

Consider the sequences

$$u_{k}=\sqrt{kx_{k}}, \qquad v_{k}=k^{2}\sqrt{kx_{k}}$$

for $k=1,2,3,4,5$.

Then

$$\sum_{k=1}^{5}u_{k}^{2} = \sum_{k=1}^{5}kx_{k} = a,$$

$$\sum_{k=1}^{5}v_{k}^{2} = \sum_{k=1}^{5}k^{5}x_{k} = a^{3},$$

and

$$\sum_{k=1}^{5}u_{k}v_{k} = \sum_{k=1}^{5}k^{3}x_{k} = a^{2}.$$

By the Cauchy-Schwarz inequality,

$$\left(\sum_{k=1}^{5}u_{k}v_{k}\right)^{2} \le \left(\sum_{k=1}^{5}u_{k}^{2}\right) \left(\sum_{k=1}^{5}v_{k}^{2}\right).$$

Substituting the above expressions yields

$$(a^{2})^{2}\le a\cdot a^{3}.$$

Hence

$$a^{4}\le a^{4},$$

so equality holds.

Equality in the Cauchy-Schwarz inequality holds precisely when the sequences $(u_{k})$ and $(v_{k})$ are proportional. Therefore there exists a real number $\lambda$ such that

$$v_{k}=\lambda u_{k}$$

for every $k$.

For every index $k$ with $x_{k}>0$, we have $u_{k}>0$. Dividing by $u_{k}$ gives

$$k^{2}=\lambda.$$

Thus every index with positive $x_{k}$ must have the same value of $k^{2}$.

Since

$$1^{2},2^{2},3^{2},4^{2},5^{2}$$

are pairwise distinct, there can be at most one index $k$ such that $x_{k}>0$. ∎

This establishes the rigidity forced by equality in Cauchy-Schwarz; merely applying the inequality without the equality condition would not determine the structure of the solution.

Lemma 3

If exactly one variable $x_{j}$ is positive, then

$$a=j^{2}.$$

Proof

Assume

$$x_{j}>0$$

for some $j\in{1,2,3,4,5}$, and

$$x_{k}=0$$

for all $k\ne j$.

Then the three equations become

$$jx_{j}=a, \qquad j^{3}x_{j}=a^{2}, \qquad j^{5}x_{j}=a^{3}.$$

From the first equation,

$$x_{j}=\frac{a}{j}.$$

Substituting into the second equation gives

$$j^{3}\cdot \frac{a}{j}=a^{2},$$

hence

$$j^{2}a=a^{2}.$$

By Lemma 1, $a\ge0$. If $a=0$, then $x_{j}=0$, contradicting $x_{j}>0$. Therefore $a>0$, and dividing by $a$ yields

$$a=j^{2}.$$

This identifies all admissible values arising from the one-support structure; skipping the case $a=0$ would make the division by $a$ invalid.

We now complete the classification.

Suppose $a$ is admissible. By Lemma 2, at most one variable $x_{j}$ is positive. If all variables are zero, then

$$a=\sum_{k=1}^{5}kx_{k}=0.$$

Otherwise exactly one variable is positive, and Lemma 3 gives

$$a=j^{2}$$

for some $j\in{1,2,3,4,5}$. Hence every admissible value belongs to

$${0,1,4,9,16,25}.$$

It remains to verify that every number in this set is indeed admissible.

For $a=0$, choose

$$x_{1}=x_{2}=x_{3}=x_{4}=x_{5}=0.$$

Then all three equations hold.

For $a=j^{2}$ with $j\in{1,2,3,4,5}$, choose

$$x_{j}=j, \qquad x_{k}=0\quad (k\ne j).$$

Then

$$\sum_{k=1}^{5}kx_{k}=j\cdot j=j^{2}=a,$$

$$\sum_{k=1}^{5}k^{3}x_{k}=j^{3}\cdot j=j^{4}=a^{2},$$

and

$$\sum_{k=1}^{5}k^{5}x_{k}=j^{5}\cdot j=j^{6}=a^{3}.$$

Thus each of these six values is admissible.

Therefore the complete set of all admissible real numbers is

$$\boxed{{0,1,4,9,16,25}}.$$

Verification of Key Steps

The most delicate step is the equality condition in Cauchy-Schwarz. Starting independently, define

$$u_{k}=\sqrt{kx_{k}},\qquad v_{k}=k^{2}\sqrt{kx_{k}}.$$

Then

$$\left(\sum u_{k}v_{k}\right)^{2} = \left(\sum k^{3}x_{k}\right)^{2} = a^{4},$$

while

$$\left(\sum u_{k}^{2}\right)\left(\sum v_{k}^{2}\right) = \left(\sum kx_{k}\right)\left(\sum k^{5}x_{k}\right) = a\cdot a^{3} = a^{4}.$$

Hence equality holds. The equality condition states that there exists $\lambda$ such that

$$v_{k}=\lambda u_{k}$$

for every $k$. If one divides by $u_{k}$ without checking whether $u_{k}=0$, the argument becomes invalid. Restricting to indices with $x_{k}>0$ avoids this error and gives

$$k^{2}=\lambda.$$

Another delicate step is deriving $a=j^{2}$ from

$$j^{2}a=a^{2}.$$

A careless proof might divide by $a$ immediately. However $a=0$ must first be excluded. Since

$$a=0$$

implies

$$x_{j}=a/j=0,$$

this contradicts the assumption that $x_{j}>0$. Only after this check may one divide by $a$.

A third point requiring verification is the construction for each admissible value. For $a=j^{2}$, taking $x_{j}=j$ gives

$$jx_{j}=j^{2}, \qquad j^{3}x_{j}=j^{4}, \qquad j^{5}x_{j}=j^{6}.$$

Since

$$(j^{2})^{2}=j^{4}, \qquad (j^{2})^{3}=j^{6},$$

all three equations are satisfied exactly.

Alternative Approaches

Another approach uses Hölder's inequality in the form

$$\left(\sum k^{3}x_{k}\right)^{2} \le \left(\sum kx_{k}\right)\left(\sum k^{5}x_{k}\right).$$

After substitution this again becomes an equality case. The equality condition for Hölder yields the same conclusion, namely that all positive weights must occur at a single index. This method is essentially equivalent to the Cauchy-Schwarz argument, but the equality condition is less transparent.

A different strategy is to interpret

$$p_{k}=\frac{kx_{k}}{a}$$

for $a>0$ as a probability distribution on ${1,2,3,4,5}$. Then

$$\sum p_{k}=1, \qquad \sum k^{2}p_{k}=a, \qquad \sum k^{4}p_{k}=a^{2}.$$

The variance of the random variable $k^{2}$ becomes

$$\sum k^{4}p_{k}-\left(\sum k^{2}p_{k}\right)^{2} = a^{2}-a^{2} = 0.$$

Hence $k^{2}$ is constant on the support of the distribution, so the support consists of a single index. This proof is elegant but introduces probabilistic language that is not necessary. The Cauchy-Schwarz approach remains more direct and entirely elementary.