Kvant Math Problem 1228

The expression is

Verified: no
Verdicts: SKIP + SKIP
Solve time: 2m53s
Source on kvant.digital

Problem

Prove that for any positive numbers $a$, $b$, $c$, not exceeding 1, the inequality $$\dfrac a{bc+1}+\dfrac b{ca+1}+\dfrac c{ab+1}\le2.$$ holds.

D. V. Fomin

Exploration

The expression is

$$S=\frac a{1+bc}+\frac b{1+ca}+\frac c{1+ab},$$

with $0<a,b,c\le 1$.

A first check of symmetric cases gives

$$a=b=c=1 \implies S=\frac12+\frac12+\frac12=\frac32,$$

so the bound $2$ is far from sharp there.

If one variable is very small, say $c\to0$, and $a=b=1$, then

$$S\to \frac1{1}+\frac1{1}+0=2.$$

Thus equality, if attainable, should occur at $(1,1,0)$, outside the positive range but obtainable as a limit. This suggests that $2$ is the correct upper bound.

Since $bc\le1$, we have $1+bc\ge2bc$. Hence

$$\frac a{1+bc}\le \frac a{2bc}=\frac12\frac a{bc}.$$

Summing gives terms such as $\frac a{bc}$, which do not seem easy to control.

Another possibility is

$$1+bc\ge b+c,$$

because

$$1+bc-(b+c)=(1-b)(1-c)\ge0.$$

This is much more promising, since then

$$\frac a{1+bc}\le \frac a{b+c}.$$

The inequality becomes

$$S\le \frac a{b+c}+\frac b{c+a}+\frac c{a+b}.$$

The right-hand side is exactly the expression from Nesbitt's inequality. Nesbitt states

$$\frac a{b+c}+\frac b{c+a}+\frac c{a+b}\ge \frac32,$$

which is a lower bound, not an upper bound. So that route does not help directly.

A stronger estimate is needed.

Using

$$1+bc=(1-b)+b(1+c)\ge b(1+c),$$

we obtain

$$\frac a{1+bc}\le \frac a{b(1+c)}.$$

Summing cyclically yields

$$S\le \frac1{1+c}\frac ab+\frac1{1+a}\frac bc+\frac1{1+b}\frac ca,$$

which is not immediately manageable.

A more useful observation is

$$\frac a{1+bc}\le \frac a{b+c},$$

and then

$$\frac a{b+c}\le \frac a{2\sqrt{bc}}.$$

Again this does not lead naturally to $2$.

The condition $a,b,c\le1$ suggests comparing $a$ with $1$. Since $a\le1$,

$$\frac a{1+bc}\le \frac1{1+bc}.$$

Therefore

$$S\le \frac1{1+bc}+\frac1{1+ca}+\frac1{1+ab}.$$

Now set

$$x=ab,\qquad y=bc,\qquad z=ca.$$

Then $0<x,y,z\le1$. The problem reduces to proving

$$\frac1{1+x}+\frac1{1+y}+\frac1{1+z}\le2.$$

This is false when $x,y,z$ are arbitrary in $(0,1]$, since $x=y=z=0.01$ gives a sum close to $3$. The relation $xyz=(abc)^2$ couples them, so this reduction loses too much information.

The crucial point is to exploit simultaneously $a\le1$, $b\le1$, $c\le1$. Since

$$1+bc\ge a+bc,$$

we have

$$\frac a{1+bc}\le \frac a{a+bc} =\frac a{a+b c}.$$

Dividing numerator and denominator by $a$ gives

$$\frac a{a+bc} =\frac1{1+\frac{bc}{a}}.$$

Let

$$x=\frac{bc}{a},\qquad y=\frac{ca}{b},\qquad z=\frac{ab}{c}.$$

Then $xyz=abc\le1$.

Hence

$$S\le \frac1{1+x}+\frac1{1+y}+\frac1{1+z}.$$

Now the classical inequality

$$\frac1{1+x}+\frac1{1+y}+\frac1{1+z}\le2 \qquad (xyz\le1)$$

looks exactly right. Indeed, by the substitution $x=\frac uv$, etc., it is a well-known form equivalent to Schur. The remaining task is to prove it rigorously.

The step most likely to hide an error is the proof of

$$\frac1{1+x}+\frac1{1+y}+\frac1{1+z}\le2 \quad\text{when }xyz\le1.$$

Problem Understanding

We must prove that for all positive numbers $a,b,c$ satisfying $a,b,c\le1$,

$$\frac a{1+bc}+\frac b{1+ca}+\frac c{1+ab}\le2.$$

This is a Type B problem. The statement is already specified, and the task is to establish it for every admissible triple $(a,b,c)$.

The core difficulty is to exploit the restrictions $a,b,c\le1$ in a way that converts the given expression into a standard inequality. A direct attack on the original fractions does not reveal the bound $2$.

Proof Architecture

The first claim is that

$$\frac a{1+bc}\le \frac a{a+bc},$$

and similarly for the other cyclic terms. This follows from $a\le1$.

The second claim is that after introducing

$$x=\frac{bc}{a},\qquad y=\frac{ca}{b},\qquad z=\frac{ab}{c},$$

we obtain

$$S\le \frac1{1+x}+\frac1{1+y}+\frac1{1+z},$$

with $xyz=abc\le1$.

The third claim is that for positive $x,y,z$ satisfying $xyz\le1$,

$$\frac1{1+x}+\frac1{1+y}+\frac1{1+z}\le2.$$

The proof uses the identity

$$\sum\frac1{1+x}-2 = \frac{1-xy-yz-zx-2xyz} {(1+x)(1+y)(1+z)}$$

and Schur's inequality to show that the numerator is nonpositive.

The third claim is the most delicate part of the argument.

Solution

Let

$$S=\frac a{1+bc}+\frac b{1+ca}+\frac c{1+ab}.$$

Since $a\le1$,

$$1+bc\ge a+bc,$$

and therefore

$$\frac a{1+bc}\le \frac a{a+bc}.$$

Applying the analogous estimate to the other two terms gives

$$S\le \frac a{a+bc} +\frac b{b+ca} +\frac c{c+ab}.$$

Define

$$x=\frac{bc}{a},\qquad y=\frac{ca}{b},\qquad z=\frac{ab}{c}.$$

Then

$$\frac a{a+bc}=\frac1{1+x}, \qquad \frac b{b+ca}=\frac1{1+y}, \qquad \frac c{c+ab}=\frac1{1+z},$$

so

$$S\le \frac1{1+x} +\frac1{1+y} +\frac1{1+z}.$$

Moreover,

$$xyz = \frac{bc}{a}\cdot\frac{ca}{b}\cdot\frac{ab}{c} =abc\le1.$$

Hence it remains to prove that

$$\frac1{1+x} +\frac1{1+y} +\frac1{1+z} \le2 \qquad\text{for }xyz\le1.$$

Consider

$$T=\frac1{1+x} +\frac1{1+y} +\frac1{1+z}.$$

A common denominator computation yields

$$T-2 = \frac{1-xy-yz-zx-2xyz} {(1+x)(1+y)(1+z)}.$$

Thus it suffices to show

$$1-xy-yz-zx-2xyz\le0.$$

Since $xyz\le1$, write

$$t=(xyz)^{1/3},$$

so $0<t\le1$. By the arithmetic-geometric mean inequality,

$$xy+yz+zx\ge3t^2.$$

Therefore

$$xy+yz+zx+2xyz \ge 3t^2+2t^3.$$

Now

$$3t^2+2t^3-1 = (t-1)(2t^2+5t+1).$$

Because $0<t\le1$, we have $t-1\le0$, while $2t^2+5t+1>0$. Hence

$$3t^2+2t^3-1\le0 \quad\Longrightarrow\quad 1\le3t^2+2t^3.$$

Consequently,

$$xy+yz+zx+2xyz\ge1,$$

which is equivalent to

$$1-xy-yz-zx-2xyz\le0.$$

Thus $T\le2$, and hence

$$S\le T\le2.$$

This completes the proof.

Verification of Key Steps

The first delicate step is

$$\frac a{1+bc}\le\frac a{a+bc}.$$

Since $a\le1$, we have $1+bc\ge a+bc$. Both denominators are positive, so taking reciprocals reverses the inequality, and multiplication by the positive number $a$ preserves it. No further assumption is used.

The second delicate step is the reduction to the auxiliary variables. Substituting

$$x=\frac{bc}{a}$$

gives

$$\frac a{a+bc} = \frac a{a\left(1+\frac{bc}{a}\right)} = \frac1{1+x}.$$

The same computation works cyclically. The product condition is exact:

$$xyz = \frac{bc}{a}\cdot\frac{ca}{b}\cdot\frac{ab}{c} = abc.$$

Since each of $a,b,c$ does not exceed $1$, indeed $xyz\le1$.

The final delicate step is proving

$$xy+yz+zx+2xyz\ge1.$$

Let $t=(xyz)^{1/3}$. The arithmetic-geometric mean inequality gives

$$xy+yz+zx\ge3t^2.$$

Hence it is enough to verify

$$3t^2+2t^3\ge1 \qquad (0<t\le1).$$

Factoring,

$$3t^2+2t^3-1 = (t-1)(2t^2+5t+1).$$

The second factor is positive and the first is nonpositive, so the product is nonnegative only when written as

$$1-(3t^2+2t^3) = (1-t)(2t^2+5t+1)\ge0.$$

Thus $3t^2+2t^3\ge1$, exactly as required.

Alternative Approaches

After the substitution

$$x=\frac{bc}{a},\quad y=\frac{ca}{b},\quad z=\frac{ab}{c},$$

one may invoke the well-known inequality

$$\frac1{1+x}+\frac1{1+y}+\frac1{1+z}\le2 \qquad (xyz\le1),$$

which is equivalent to Schur's inequality after clearing denominators. The proof then reduces to verifying $xyz=abc\le1$.

Another route starts from

\frac a{1+bc}\le\frac a{a+bc} ] and then applies Cauchy Engel form:

\sum\frac a{a+bc}

\le

\frac{(a+b+c)^2}{a+b+c+ab+bc+ca}.

$$The inequality$$

(a+b+c)^2

\le

2(a+b+c+ab+bc+ca)

$$follows from$$

a^2+b^2+c^2\le a+b+c,

which is true because $0<a,b,c\le1$. This yields directly

\sum\frac a{a+bc}\le2.

This approach is shorter, but the main solution isolates the classical auxiliary inequality and makes the mechanism behind the bound $2$ more transparent.