The inequality above seems very compelling for the pqr-method.
So this was my attempt-
$$ LHS = (a^2 + 1)(b^2 + 1)(c^2 + 1) = 1 + a^2 + b^2 + c^2 + a^2b^2 + b^2c^2 + c^2a^2 + a^2b^2c^2 $$
Now substituting $p = a+b+c$ , $q = ab+bc+ca$ and $r = abc$.
$$ LHS = 1 + p^2 - 2q + q^2 - 2pr + r^2 \geq 2q \Rightarrow 1 + p^2 + q^2 + r^2 \geq 4q + 2pr $$
It's quite well-known that $p^2\geq 3q$ and $q^2\geq 3pr$. So,
$$ 1 + 3q + 3pr + r^2 \geq 4q + 2pr \Rightarrow 1 + pr + r^2 \geq q $$
But I don't know how to prove it. It can also be seen that $a\ge b\ge c$, but I can't exploit symmetry.
Any help is thankfully welcome.
- 1,506
-
since there is no equality condition for $a = b = c,$ using inequalities that requires this condition will not be useful for you. In other words, $p^2 = 3q$ if and only if $a = b= c.$ I would just write the whole thing as a quadratic in $a$ find it's minimum in terms of $b$ and $c$, then show that minimum is non-negative. – dezdichado Aug 25 '20 at 04:30
-
@dezdichado Yes, I destroyed the equality case. Wouldn't you just write it in a quadratic of $a$ and show the discriminant is negative? Well, there is a sixth degree expression in the$ LHS$. – Book Of Flames Aug 25 '20 at 04:37
3 Answers
If you write LHS - RHS as a quadratic in $a$, then the discriminant is: $$D = 4(b+c)^2 - 4((b^2+1)(c^2+1)-2bc)(b^2+1)(c^2+1).$$ But $(b^2+1)(1+c^2)\geq (b+c)^2$ by C-S and $(b^2+1)(c^2+1)-2bc = b^2c^2+(b-c)^2+1\geq 1,$ so the discriminant is non-positive.
- 13,888
-
-
there is an equality case only when the quadratic has a real root. However, you can do more with the discriminant: the second inequality used above is equality only when $b = c = 0$, while the first C-S attains equality only when $b = c = 1.$ So the discriminant is actually always negative. – dezdichado Aug 25 '20 at 05:05
-
I think the equality case actually exists. I checked my question twice, and it is correct. There you see $(1+a^2)(1+b^2)(1+c^2)\geq2(ab+bc+ca)$. The sign is not a strict inequality, but greater than or equal to. This says that equality condition holds at some $a,b,c$. So where does it hold? – Book Of Flames Aug 26 '20 at 02:46
-
I literally proved to you in the previous comment that that was impossible. Choosing to ignore a completely valid proof is not gonna help you get better at mathematics. Besides, $\geq$ includes the possibility of $>$, so you cannot logically deduct the equality must be attained. I can write $6\geq 5$ and it will be completely fine. – dezdichado Aug 26 '20 at 02:57
Proceeding along your approach:
It suffices to prove that $1 + p^2 - 2q + q^2 - 2pr + r^2 \ge 2q$.
Since $p^2 \ge 3q, q^2 \ge 3pr$, it suffices to prove that
$1 + 3q - 2q + q^2 - \frac{2}{3}q^2 \ge 2q$
or $1 - q + \frac{q^2}{3} \ge 0$
or $\frac{1}{3}(q - \frac{3}{2})^2 + \frac{1}{4} \ge 0$ which is true. We are done.
- 37,323
-
How did you get the second inequality? Please explain little more. – Book Of Flames Aug 25 '20 at 12:17
-
@BookOfFlames Do you mean $p^2 - 3q = a^2 + b^2 + c^2 - ab - bc - ca \ge 0$? – River Li Aug 25 '20 at 12:42
-
I meant how did you reach $1 + 3q - 2q + q^2 - \frac{2}{3}q^2 \ge 2q$ from $1 + p^2 - 2q + q^2 - 2pr + r^2 \ge 2q$ ? – Book Of Flames Aug 25 '20 at 12:43
-
@BookOfFlames Since $p^2 \ge 3q$, $-2pr \ge -\frac{2}{3}q^2$ and $r^2 \ge 0$, we have $1 + p^2 - 2q + q^2 - 2pr + r^2 \ge 1 + 3q - 2q + q^2 - \frac{2}{3}q^2$. – River Li Aug 25 '20 at 13:16
Because $2(ab + bc + ca) \leqslant 2(|a||b| + |b||c| + |c||a|)$ and $$(a^2 + 1)(b ^2 + 1)(c ^2 + 1) = (|a|^2 + 1)(|b| ^2 + 1)(|c| ^2 + 1),$$ so we need to prove the inequality when $ a,\,b,\,c \geqslant 0.$
Indeed, easy to check $3t^2 \geqslant 3t-1.$ Now, using the AM-GM we have $$(a^2 + 1)(b ^2 + 1)(c ^2 + 1) \geqslant a^2 + b^2 + c^2 + 1 + a^2b^2 + b^2c^2 + c^2a^2$$ $$ \geqslant a^2+b^2+c^2+1+3\sqrt[3]{(abc)^4}$$ $$ \geqslant a^2+b^2+c^2+3\sqrt[3]{(abc)^2}.$$ Therefore we will show that $$a^2+b^2+c^2+3\sqrt[3]{(abc)^2} \geqslant 2(ab+bc+ca).$$ Which is very known (here, here).
- 4,569
- 9
- 16
-
How did $1+3\sqrt[3]{(abc)^4}\geq 3\sqrt[3]{(abc)^2}$ by AM-GM? – Book Of Flames Aug 25 '20 at 04:44
-
-
@nguyenhuyen_ag For those of us not so experienced with inequalities, it would be instructive of you to supply a proof for the final inequality (the one you say is well-known). – ΑΘΩ Aug 25 '20 at 05:03
-
the last inequality is called Popovicu's inequality and while it's certainly well-known in contest math community, it is not as elementary as, say, Cauchy-Schwarz. – dezdichado Aug 25 '20 at 05:06
-
1@ΑΘΩ But it's indeed, well known. For the proof use Schur and Muirhead. – Michael Rozenberg Aug 25 '20 at 05:07
-
-
1
-
@ΑΘΩ Thank you, I was updated the proof. Moreover it's called Shleifer's inequality. – nguyenhuyenag Aug 25 '20 at 06:34