5

If $abc=1$ for positive $a,b,c$, then $\sum\limits_{cyc}^{}{\dfrac{1}{b(a+b)}}\ge \dfrac{3}{2}$

I have tried the following,in decreasing order of success:

1)AM-GM:$a+b+c\ge 3$ and $ab+bc+ca\ge 3$

2)Substituting $1=abc$ yields nothing

3)Substituting $a=\frac{x}{y},b=\frac{y}{z},c=\frac{z}{x}$ yields something weird

4)Rearrangement inequality on the sequences $\frac{1}{a},\frac{1}{b},\frac{1}{c}$ and $\frac{1}{a+b},\frac{1}{b+c},\frac{1}{c+a}$

5)Titu's lemma doesn't help

A little nudge in the right direction would help.

rah4927
  • 3,864

4 Answers4

8

(I) By rearrangement inequality:

$\displaystyle \begin{align} &\sum_{cyc} \frac{1}{b(a+b)} \ge \sum\limits_{cyc} \frac{1}{b(a+c)} \tag{1} \\ \iff & \sum_{cyc} \frac{1}{b(a+b)} \ge \sum_{cyc} \frac{1}{2}\left(\frac{1}{b(a+b)} + \frac{1}{b(a+c)}\right) = \sum_{cyc} \frac{1}{2}\left(\frac{1}{b(a+b)} + \frac{1}{c(a+b)}\right) \\ \iff & \sum_{cyc} \frac{1}{b(a+b)} \ge \frac{1}{2}\sum_{cyc} \frac{b+c}{bc(a+b)} \end{align}$

By Am-Gm Inequality :

$$\sum_{cyc} \frac{b+c}{bc(a+b)} \ge 3\sqrt[3]{\prod\limits_{cyc} \frac{b+c}{bc(a+b)}} = 3$$

Thus establishing desired inequality.

Note: $(1)$ can be viewed as a consequence of CS as well.

$$\sum_{cyc} \left(\frac{1}{b(a+b)} - \frac{1}{b(a+c)}\right) \ge 0 \iff \sum_{cyc} \frac{c-b}{b(a+b)(a+c)} \ge 0 \\ \iff \sum_{cyc} \frac{c^2-b^2}{b} \ge 0 \iff \sum_{cyc} \frac{c^2}{b} \ge \sum_{cyc} b$$

(II) Substituting $\displaystyle a=\frac{x}{y},b=\frac{y}{z},c=\frac{z}{x}$:

The inequality required to prove becomes:

$$\sum\limits_{cyc} \frac{x^2}{z^2+xy} \ge \frac{3}{2}$$

We may rewrite LHS as $\displaystyle \sum\limits_{cyc} \frac{x^4}{x^2z^2+x^3y}$ and apply Cauchy-Schwarz Inequality:

$$\sum\limits_{cyc} \frac{x^4}{x^2z^2+x^3y} \ge \frac{(x^2+y^2+z^2)^2}{\sum\limits_{cyc} x^2z^2 + \sum\limits_{cyc} x^3y}$$

So it suffices to prove that: $$2(x^2+y^2+z^2)^2 \ge 3\sum\limits_{cyc} x^2z^2 + 3\sum\limits_{cyc} x^3y \\ \iff (x^4+y^4+z^4) + \sum\limits_{cyc} (x^4 + x^2y^2) \ge 3\sum\limits_{cyc} x^3y$$

This is the consequence of adding the following:

(i) The Rearrangement Inequality: $x^4+y^4+z^4 \ge x^3y+y^3z+z^3x$

(ii) The Am-Gm Inequality: $\displaystyle \sum\limits_{cyc} (x^4 + x^2y^2) \ge 2\sum\limits_{cyc} \sqrt{x^6y^2} = 2\sum\limits_{cyc} x^3y$

r9m
  • 17,938
1

Hint:

Indeed, as @Macavity pointed out, the sum $s(a,b,c)$ only has circular symmetry. In fact it is not hard to show ( the same method as below) that if $a\le b\le c$ then $s(a,b,c)\le s(c,b,a)$.

Here is a method amenable to a computer algebra system. As suggested, substitute $a=u/v$, $b=v/w$, $c=w/u$. Get an equivalent inequality homogeneous of degree $6$ in $u$,$v$, $w\ge 0$

$$2 u^4 v^2-3 u^3 v^3+2 u v^5+2 u^5 w-3 u^4 v w+4 u^2 v^3 w- 3 u v^4 w+4 u^3 v w^2-\\-6 u^2 v^2 w^2+2 v^4 w^2-3 u^3 w^3+4 u v^2 w^3-3 v^3 w^3+2 u^2 w^4-3 u v w^4+2 v w^5\ge 0$$

with circular symmetry. It is enough to consider the cases $u\le v \le w$ and $u \ge v \ge w$.

In the first case, make the substitution $u=p$, $v=p+q$, $w=p+q+r$, with $p$, $q$, $r\ge 0$. One gets $$8 p^4 q^2+18 p^3 q^3+15 p^2 q^4+6 p q^5+q^6+8 p^4 q r+31 p^3 q^2 r +38 p^2 q^3 r+22 p q^4 r+\\+5 q^5 r+8 p^4 r^2+41 p^3 q r^2 +66 p^2 q^2 r^2+46 p q^3 r^2+13 q^4 r^2+14 p^3 r^3+43 p^2 q r^3+\\ +43 p q^2 r^3+17 q^3 r^3+9 p^2 r^4+17 p q r^4+10 q^2 r^4+2 p r^5+2 q r^5$$

clearly positive, with equality when $q=r=0$, that is, the $a$,$b$,$c$ are equal. The other case is treated similarly, giving the positive expression $$8 p^4 q^2+18 p^3 q^3+15 p^2 q^4+6 p q^5+q^6+8 p^4 q r+23 p^3 q^2 r+22 p^2 q^3 r+8 p q^4 r+q^5 r+8 p^4 r^2+33 p^3 q r^2+42 p^2 q^2 r^2+18 p q^3 r^2+3 q^4 r^2+14 p^3 r^3+35 p^2 q r^3+23 p q^2 r^3+5 q^3 r^3+9 p^2 r^4+11 p q r^4+2 q^2 r^4+2 p r^5$$

orangeskid
  • 53,909
  • Can you please tell me what you mean by circular symmetry?Can you elaborate on the hint?I find it unclear. – rah4927 Feb 10 '15 at 09:31
  • the expression $s(a,b,c)$ is unchanged under circular permutations of $a$, $b$,$c$, that is $s(a,b,c) = s(b,c,a) = s(c,a,b)$ – orangeskid Feb 10 '15 at 09:42
  • For three numbers on the circle with an orientation, either the largest is in front of the smallest one, or the smallest one is in front of the largest one. Hence, for expression in $u$, $v$, $w$ with a circular symmetry, it is enough to assume $u\le u \le w$ or $u\ge v \ge w$. If $u \le v \le w$, let $u=p$, $q=v-u$, $r= w-v$, all positive. Express $u$, $v$, $w$ in terms of $p$, $q$, $r$ and substitute. It so happens that we get an expression in $p$, $q$, $r$ with positive coefficients and thus positive. – orangeskid Feb 10 '15 at 09:46
-1

Look at $\frac{1}{b(a+b)}$ and $\frac{1}{a(a+b)}$. Adding together yields $\frac{1}{ab}$. So the sum is just $$\frac{1}{ab} + \frac{1}{bc} + \frac{1}{ac} = \frac{a + b + c}{abc}$$ The denominator is 1 and the numerator is at least 3 by AM-GM for positive $a,b,c$.

Chris Jones
  • 1,429
-1

I think the inequality is false. Currently I am searching for a triplet which disproves it. But here is my argument- Substitute $a=x/y....$. Now write your expression in terms of $x,y,z$ and then apply Titu's lemma. You should get $$\sum_{cyc}\frac{x^2}{xy+z^2}\geq\frac{(x+y+z)^2}{x^2+y^2+z^2+xy+yz+zx}.$$ This means after opening it and making it greater than $1/2$, you get a positive quantity less than $0$. Not possible right. Please tell me if I am wrong anywhere and also please edit this answer(I don't know what is wrong in my syntax).

Carol
  • 426
  • 3
    Not sure what you mean when you say greater than $1/2$, it should be greater than $3/2$. Further even if you get that to be negative, it just shows Titu's lemma in that form is not perhaps strong enough to prove the inequality, it does not disprove the inequality :( – Macavity Feb 10 '15 at 09:12
  • 1
    @orangeskid Check http://www-stat.wharton.upenn.edu/~steele/Publications/Books/CSMC/New%20Problems/CSNewProof/CauchySchwarzInequalityProof.pdf Titu's lemma is a particular form of Cauchy-Schwarz inequality. – Macavity Feb 10 '15 at 09:22
  • @Macavity: Oh, I see, thanks for the pointer. What about the book Cauchy-Schwarz master class? Oh, it's by the same author! Wow, neat. – orangeskid Feb 10 '15 at 09:50