Kvant Math Problem 1050

Let the chosen points be

Verified: no
Verdicts: SKIP + SKIP
Solve time: 2m50s
Source on kvant.digital

Problem

On the segment $[-1, 1]$, $k$ distinct points are chosen. For each point, the product of its distances to the remaining $k-1$ points is computed, and $S$ denotes the sum of the reciprocals of these $k$ products. Prove that

  1. $S\ge2$ for $k=3$;
  2. $S\ge4$ for $k=4$.

L. D. Kurlandchik

Exploration

Let the chosen points be

$$-1\le x_1<x_2<\cdots<x_k\le 1,$$

and let

$$P_i=\prod_{j\ne i}(x_i-x_j).$$

The required quantity is

$$S=\sum_{i=1}^k \frac1{|P_i|}.$$

For $k=3$, write

$$a=x_2-x_1,\qquad b=x_3-x_2.$$

Then $a,b>0$ and $a+b\le2$. The products are

$$|P_1|=a(a+b),\qquad |P_2|=ab,\qquad |P_3|=b(a+b),$$

hence

$$S=\frac1{a(a+b)}+\frac1{ab}+\frac1{b(a+b)} =\frac2{ab}.$$

Thus the problem becomes proving

$$\frac2{ab}\ge2,$$

or $ab\le1$. Since $a+b\le2$, the maximum of $ab$ is $1$, attained at $a=b=1$.

For $k=4$, write

$$a=x_2-x_1,\qquad b=x_3-x_2,\qquad c=x_4-x_3.$$

Then $a,b,c>0$ and $a+b+c\le2$. A direct computation gives

$$S= \frac1{a(a+b)(a+b+c)} +\frac1{ab(b+c)} +\frac1{bc(a+b)} +\frac1{c(b+c)(a+b+c)}.$$

The expression is homogeneous of degree $-3$. Since enlarging $a+b+c$ while keeping the ratios fixed decreases $S$, the minimum must occur when

$$a+b+c=2.$$

Substituting $d=a+b+c=2$,

$$S= \frac1{2a(a+b)} +\frac1{ab(b+c)} +\frac1{bc(a+b)} +\frac1{2c(b+c)}.$$

Trying the symmetric point $a=b=c=\frac23$ gives

$$S=\frac{81}{32}>4.$$

Trying the boundary tendency $b\to0$ makes $S\to\infty$. Hence the minimum should occur in the interior. The expression is symmetric under $a\leftrightarrow c$, suggesting that the minimum occurs when $a=c$.

Set $a=c=t$, $b=2-2t$. Then

$$S=\frac1{4t(2-t)}+\frac1{2t^2(1-t)}.$$

Differentiation yields a unique critical point at $t=\frac12$, where

$$S=4.$$

This suggests that equality occurs for

$$(a,b,c)=\left(\frac12,1,\frac12\right).$$

The step most likely to hide an error is proving rigorously that

$$S\ge4.$$

A direct algebraic estimate is needed.

Problem Understanding

We have $k$ distinct points on the segment $[-1,1]$. For each point, we form the product of its distances to all the other chosen points. The quantity $S$ is the sum of the reciprocals of these products.

We must prove two lower bounds:

$$S\ge2 \qquad (k=3),$$

and

$$S\ge4 \qquad (k=4).$$

This is a Type B problem. The statement is already given and must be proved.

The core difficulty is the case $k=4$. After expressing $S$ through the consecutive gaps between neighboring points, one must obtain a sharp lower bound under the constraint that the sum of the gaps does not exceed $2$.

Proof Architecture

Let $a,b,c$ denote consecutive gaps between neighboring points.

For $k=3$, express $S$ as $2/(ab)$ and use $(a+b)^2\ge4ab$ together with $a+b\le2$.

For $k=4$, first reduce to the case $a+b+c=2$ by homogeneity.

Next prove

$$\frac1{2a(a+b)}+\frac1{2c(b+c)} \ge \frac2{(a+c)(a+b+c)}.$$

This follows from the harmonic mean inequality.

Then prove

$$\frac1{ab(b+c)}+\frac1{bc(a+b)} = \frac{a+b+c}{bc(a+b)(b+c)} = \frac2{bc(a+b)(b+c)}.$$

Combining these estimates yields

$$S\ge \frac1{a+c} +\frac2{bc(a+b)(b+c)}.$$

Put $t=a+c$. Since $a+b+c=2$, we have $b=2-t$. Use

$$(a+b)(b+c)\le\left(\frac{(a+b)+(b+c)}2\right)^2 =\left(1+\frac b2\right)^2 =\left(2-\frac t2\right)^2.$$

This gives a lower bound depending only on $t$:

$$S\ge \frac1t+\frac8{(2-t)t(4-t)^2}.$$

Finally show

$$\frac1t+\frac8{(2-t)t(4-t)^2}-4 = \frac{(2t-1)^2(t-2)^2}{(2-t)t(4-t)^2} \ge0.$$

The hardest part is obtaining a sharp one-variable estimate after reducing the expression to the gap variables.

Solution

Let

$$-1\le x_1<x_2<\cdots<x_k\le1,$$

and define

$$P_i=\prod_{j\ne i}(x_i-x_j).$$

Since $S$ uses distances,

$$S=\sum_{i=1}^k\frac1{|P_i|}.$$

Case $k=3$

Put

$$a=x_2-x_1,\qquad b=x_3-x_2.$$

Then

$$a>0,\qquad b>0,\qquad a+b=x_3-x_1\le2.$$

The three distance products are

$$|P_1|=a(a+b),$$

$$|P_2|=ab,$$

$$|P_3|=b(a+b).$$

Hence

$$S= \frac1{a(a+b)} +\frac1{ab} +\frac1{b(a+b)} = \frac{a+b}{ab(a+b)}+\frac1{ab} = \frac2{ab}.$$

Since

$$(a+b)^2\ge4ab,$$

and $a+b\le2$, we obtain

$$ab\le\frac{(a+b)^2}{4}\le1.$$

Therefore

$$S=\frac2{ab}\ge2.$$

This proves the first statement.

Case $k=4$

Put

$$a=x_2-x_1,\qquad b=x_3-x_2,\qquad c=x_4-x_3.$$

Then

$$a,b,c>0, \qquad a+b+c=x_4-x_1\le2.$$

The four distance products are

$$|P_1|=a(a+b)(a+b+c),$$

$$|P_2|=ab(b+c),$$

$$|P_3|=bc(a+b),$$

$$|P_4|=c(b+c)(a+b+c).$$

Thus

$$S= \frac1{a(a+b)(a+b+c)} +\frac1{ab(b+c)} +\frac1{bc(a+b)} +\frac1{c(b+c)(a+b+c)}.$$

The expression is homogeneous of degree $-3$. If all gaps are multiplied by a factor $\lambda>1$, then $S$ is divided by $\lambda^3$. Consequently, for fixed ratios of $a,b,c$, increasing $a+b+c$ decreases $S$. Hence the minimum under the constraint $a+b+c\le2$ is attained when

$$a+b+c=2.$$

Substituting this relation,

$$S= \frac1{2a(a+b)} +\frac1{ab(b+c)} +\frac1{bc(a+b)} +\frac1{2c(b+c)}.$$

Using

$$\frac1u+\frac1v\ge\frac4{u+v},$$

with

$$u=2a(a+b),\qquad v=2c(b+c),$$

we obtain

$$\frac1{2a(a+b)} +\frac1{2c(b+c)} \ge \frac4{2a(a+b)+2c(b+c)}.$$

Because $a+b+c=2$,

$$a(a+b)+c(b+c) = a^2+ab+bc+c^2 = (a+c)(a+b+c) = 2(a+c).$$

Hence

$$\frac1{2a(a+b)} +\frac1{2c(b+c)} \ge \frac1{a+c}.$$

For the middle two terms,

$$\frac1{ab(b+c)} +\frac1{bc(a+b)} = \frac{a+b+c}{bc(a+b)(b+c)} = \frac2{bc(a+b)(b+c)}.$$

Therefore

$$S\ge \frac1{a+c} +\frac2{bc(a+b)(b+c)}.$$

Let

$$t=a+c.$$

Then $0<t<2$ and $b=2-t$.

By the arithmetic mean-geometric mean inequality,

$$(a+b)(b+c) \le \left(\frac{(a+b)+(b+c)}2\right)^2.$$

Since

$$(a+b)+(b+c)=a+c+2b=t+2(2-t)=4-t,$$

we obtain

$$(a+b)(b+c)\le\left(\frac{4-t}{2}\right)^2.$$

Also,

$$bc=(2-t)c\le(2-t)\frac{a+c}{2} =(2-t)\frac t2.$$

Multiplying the last two inequalities gives

$$bc(a+b)(b+c) \le \frac{(2-t)t(4-t)^2}{8}.$$

Hence

$$\frac2{bc(a+b)(b+c)} \ge \frac{16}{(2-t)t(4-t)^2}.$$

Consequently

$$S\ge \frac1t+\frac{16}{(2-t)t(4-t)^2}.$$

A direct computation yields

$$\frac1t+\frac{16}{(2-t)t(4-t)^2}-4 = \frac{(2t-1)^2(2-t)^2}{(2-t)t(4-t)^2}.$$

The denominator is positive for $0<t<2$, so the right-hand side is nonnegative. Therefore

$$S\ge4.$$

Equality holds when all inequalities used above are equalities. This requires

$$a=c,\qquad (a+b)=(b+c),\qquad c=\frac{a+c}{2},$$

hence $a=c$, and together with $a+b+c=2$ and $2t-1=0$ we get

$$a=c=\frac12,\qquad b=1.$$

Thus $S=4$ is attained, for example, at the points

$$-!1,,-\frac12,,\frac12,,1.$$

This completes the proof.

Verification of Key Steps

The first delicate step is the reduction to $a+b+c=2$. Since every denominator in the definition of $S$ is a product of three distances, replacing $(a,b,c)$ by $(\lambda a,\lambda b,\lambda c)$ multiplies every denominator by $\lambda^3$ and therefore replaces $S$ by $S/\lambda^3$. If $a+b+c<2$, choosing

$$\lambda=\frac2{a+b+c}>1$$

produces another admissible configuration with a smaller value of $S$. Hence no minimizer can satisfy $a+b+c<2$.

The second delicate step is the estimate

$$a(a+b)+c(b+c)=2(a+c).$$

Expanding gives

$$a^2+ab+bc+c^2.$$

Factoring,

$$a^2+ab+bc+c^2=(a+c)(a+b+c).$$

After imposing $a+b+c=2$, this becomes $2(a+c)$. Missing this factorization destroys the sharpness of the argument.

The third delicate step is the final one-variable inequality. Substituting $t=\frac12$ gives

$$\frac1t+\frac{16}{(2-t)t(4-t)^2}=4,$$

so the bound is attainable. Substituting $t=1$ gives

$$1+\frac{16}{9}>2,$$

showing that the expression is not identically equal to $4$. The factorization

$$\frac1t+\frac{16}{(2-t)t(4-t)^2}-4 = \frac{(2t-1)^2(2-t)^2}{(2-t)t(4-t)^2}$$

proves nonnegativity exactly and identifies the equality case.

Alternative Approaches

A different approach for $k=4$ is to introduce

$$u=a+b,\qquad v=b+c.$$

Then

$$u+v=4-(a+c),$$

and the expression can be rewritten in terms of $u$, $v$, and $a+c$. Applying the arithmetic mean-geometric mean inequality first to $u$ and $v$, and then to $a$ and $c$, reduces the problem to a single variable. The resulting calculation is equivalent to the proof above but uses different intermediate quantities.

Another possibility is to apply Cauchy-Schwarz to the four reciprocals and estimate the sum of the denominators. This also leads to a sharp bound, but the algebra is longer and the equality case is less transparent. The gap-variable method keeps the geometry of the configuration visible and produces the extremal configuration naturally.