9

$a,b,c >0$, and $ab+bc+ca=3$, prove $$(a^ab^bc^c)^{\frac{3}{a+b+c}} \geqslant \sqrt[3]{\frac{a^3+b^3+c^3}{3}}$$

I think the equality is only achieve when $a=b=c=1$. The condition $ab+bc+ca=3$ is necessary. I used the estimation $x^x \geqslant \frac12 (x^2+1)$ but cannot proceed further.

HN_NH
  • 4,361

2 Answers2

1

Firstly, we'll prove that: $$\left(a^ab^bc^c\right)^{\frac{3}{a+b+c}}\geq\left(\frac{\sqrt{a^3}+\sqrt{b^3}+\sqrt{c^3}}{3}\right)^2.$$ Indeed, after checking that this inequality is true for $a\rightarrow0^+$ and for $a\rightarrow+\infty$ and since continuous on a compact function gets a minimum

(let it happens in the point (a,b,c)), it's enough to check, what happens for $\frac{\partial f}{\partial a}=\frac{\partial f}{\partial b}=\frac{\partial f}{\partial c}=0,$ where $$f(a,b,c)=a\ln{a}+b\ln{b}+c\ln{c}-\frac{a+b+c}{1.5}\ln\frac{\sqrt{a^3}+\sqrt{b^3}+\sqrt{c^3}}{3}.$$ Thus, $$\ln{a}+1-\frac{1}{1.5}\ln\frac{\sqrt{a^3}+\sqrt{b^3}+\sqrt{c^3}}{3}-\frac{(a+b+c)\sqrt{a}}{\sqrt{a^3}+\sqrt{b^3}+\sqrt{c^3}}=0,$$ $$\ln{b}+1-\frac{1}{1.5}\ln\frac{\sqrt{a^3}+\sqrt{b^3}+\sqrt{c^3}}{3}-\frac{(a+b+c)\sqrt{b}}{\sqrt{a^3}+\sqrt{b^3}+\sqrt{c^3}}=0$$ and $$\ln{c}+1-\frac{1}{1.5}\ln\frac{\sqrt{a^3}+\sqrt{b^3}+\sqrt{c^3}}{3}-\frac{(a+b+c)\sqrt{c}}{\sqrt{a^3}+\sqrt{b^3}+\sqrt{c^3}}=0.$$ We'll prove that it's possible for $a=b=c$ only.

Indeed, let it not be so and $a\leq b<c$.

If $a=b$ so let $a=b=xc$ and we need to prove that: $$(x^x)^2\geq\left(\frac{2\sqrt{x^3}+1}{3}\right)^{\frac{2x+1}{1.5}}$$ or $$x\geq\left(\frac{2\sqrt{x^3}+1}{3}\right)^{\frac{2x+1}{3x}}$$ $$x^2\geq\left(\frac{2x^3+1}{3}\right)^{\frac{2x^2+1}{3x^2}}$$ or $g(x)\geq0,$ where $$g(x)=2\ln{x}-\frac{2x^2+1}{3x^2}\ln\frac{2x^3+1}{3}.$$ which easy to show.

Indeed,$$g'(x)=\frac{2}{x}+\frac{2}{3x^3}\ln\frac{2x^3+1}{3}-\frac{2(2x^2+1)}{2x^3+1}=\frac{2}{3x^3}\left(\ln\frac{2x^3+1}{3}-\frac{3x^2(x-1)}{2x^3+1}\right),$$ $$\left(\ln\frac{2x^3+1}{3}-\frac{3x^2(x-1)}{2x^3+1}\right)_{x=1}=0$$ and $$\left(\ln\tfrac{2x^3+1}{3}-\tfrac{3x^2(x-1)}{2x^3+1}\right)'=\tfrac{3x(4x^4-2x^3-x+2)}{2x^3+1}=\tfrac{3x(2x^4+x+2(x-1)^2(x^2+x+1))}{2x^3+1}>0,$$ which says that $1$ is an unique root of $g'$ and since $$\lim_{x\rightarrow+\infty}\left(\ln\frac{2x^3+1}{3}-\frac{3x^2(x-1)}{2x^3+1}\right)=+\infty,$$ we obtain: $$g(x)\geq g(1)=0$$ for any $x>0$.

Now, let $a<b<c$.

Thus, we see that: $$\ln{b}-\ln{a}=\frac{(a+b+c)(\sqrt{b}-\sqrt{a})}{\sqrt{a^3}+\sqrt{b^3}+\sqrt{c^3}},$$ which gives $$\frac{\ln{b}-\ln{a}}{\sqrt{b}-\sqrt{a}}=\frac{\ln{c}-\ln{b}}{\sqrt{c}-\sqrt{b}}=\frac{\ln{c}-\ln{a}}{\sqrt{c}-\sqrt{a}}$$ or $$\frac{\ln\sqrt{b}-\ln\sqrt{a}}{\sqrt{b}-\sqrt{a}}=\frac{\ln\sqrt{c}-\ln\sqrt{b}}{\sqrt{c}-\sqrt{b}}=\frac{\ln\sqrt{c}-\ln\sqrt{a}}{\sqrt{c}-\sqrt{a}},$$ which is impossible by the Lagrange's mead value theorem.

Id est, it's enough to check $a=b=c$, which gives an equality.

Thus, it's enough to prove that: $$\left(\frac{\sqrt{a^3}+\sqrt{b^3}+\sqrt{c^3}}{3}\right)^2\geq\sqrt[3]{\frac{a^3+b^3+c^3}{3}}.$$ Now, let $a^3+b^3+c^3=3u$.

Thus, by Holder and AM-GM we obtain: $$\left(\frac{\sqrt{a^3}+\sqrt{b^3}+\sqrt{c^3}}{3}\right)^2=\frac{3u+2\sum\limits_{cyc}\sqrt{a^3b^3}}{9}=\frac{3u+2\sqrt{\frac{(1+1+1)\left(\sum\limits_{cyc}\sqrt{a^3b^3}\right)^2}{3}}}{9}\geq$$$$\geq\frac{3u+2\sqrt{\frac{(ab+ac+bc)^3}{3}}}{9}=\frac{u+2}{3}\geq\sqrt[3]{u}=\sqrt[3]{\frac{a^3+b^3+c^3}{3}}.$$

  • 1
    maybe it would be useful for the reader to complete the parts "which easy to show" and "which is smooth" – Thomas Jun 29 '21 at 13:49
  • ( did not down vote by the way ) – Thomas Jun 29 '21 at 17:54
  • @Thomas I just think that for you it's indeed easy and smooth. Otherwise, I am ready to explain. – Michael Rozenberg Jun 29 '21 at 18:14
  • Ok. Please proceed with the explanations :) . For example for the last one, CS or Holder have the wrong direction to be applied straightforwardly. Using a^2b^2+a^2c^2+b^2c^2 \le a^4+b^4+c^4 seems not useful. Bunching looks scaring with these high order products so for me it would be a problem by itself trying to prove it. It may be that in the future also other readers come with the same doubts, for me at the moment without any guideline it is not obvious how to proceed, which is ok for me if this was the text of a problem ... – Thomas Jun 30 '21 at 07:33
  • @Thomas I found something nice. See now. – Michael Rozenberg Jun 30 '21 at 08:22
  • Thanks it is indeed nice and clear. I guess that you "knew" in advance that the last inequality would have been true, and that you had a brute force proof of the previous formulation (I may be wrong). Did you check it before using some computer calculation ? – Thomas Jun 30 '21 at 10:08
  • 1
    Yes, I.checket it before by WA. After this I found a proof by SOS, which turned out ugly enough, but it's a proof by hand. My second proof is better. – Michael Rozenberg Jun 30 '21 at 11:49
  • 1
    Checking with WA looks indeed like a nice trick to drive experimentations :)... I guess the SOS solution would look very different than the second one :D . Anyway congrats as usual ... – Thomas Jun 30 '21 at 14:12
  • Why is this question so well received? And why did you answer it? I see no substantial effort on OP's part in this question. And yet I get suspended for answering low-effort questions... – Ritam_Dasgupta Jul 02 '21 at 05:55
  • @Ritam_Dasgupta Do you think that it's an easy question? It was unsolved five years. – Michael Rozenberg Jul 02 '21 at 06:02
  • Does that matter? I thought it was about adding context. The difficulty of the question is irrelevant. – Ritam_Dasgupta Jul 02 '21 at 06:55
0

It is not an answer, but rather some results which can be helpful for the further investigations.

  1. $abc \leq 1$
    $\rhd$ $$\frac{ab +bc + ac}{3} \geq \sqrt[3]{a^2b^2c^2}\Rightarrow\sqrt[3]{a^2b^2c^2}\leq 1$$ $\lhd$

  2. One can prove that $$\forall a_i > 0,\forall p_i>0: \sum\limits_{1}^{n}p_i = 1$$ the following inequality is fulfilled $$\sum\limits_{i =1}^{n}p_i a_i \geq \prod\limits_{i = 1}^{n}a^{p_i}_{i}.$$ Let us denote \begin{align} \frac{a}{a+b+c} &= \alpha\\ \frac{b}{a+b+c} &= \beta,\\ \frac{c}{a+b+c} &= \gamma, \end{align} then $\alpha+\beta+\gamma = 1$ and, according to the previous formula ($a_i = \frac1a, \frac1b, \frac1c$), $$(a^ab^bc^c)^{\frac{3}{a+b+c}} =(a^\alpha b^\beta c^\gamma)^3 \geq \Big(\frac{a+b+c}{3}\Big)^3 = \Big(\frac{a^3 + b^3 + c^3 + 9(a+b+c) - 3abc}{27}\Big)\geq \Big(\frac{a^3 + b^3 + c^3}{27} + \frac{a+b+c}{3} - \frac19\Big),$$ where the last inequality uses 1.

LRDPRDX
  • 1,288
  • Sorry, I did not see @barto's comment. Of course he (she) said about the same. But I think weights are $\frac{a}{a+b+c}$ and so on. – LRDPRDX Jan 25 '17 at 23:45
  • $\Big(\frac{a+b+c}{3}\Big)^3 \ge \sqrt[3]{\frac{a^3+b^3+c^3}{3}}$ does not hold, right? Please check $b = c = \frac{23}{36}$ and $a = 3359/1656$. – River Li Jun 07 '20 at 13:22
  • @RiverLi, Yes you're right. It seems that the result I present is not useful. – LRDPRDX Jun 08 '20 at 07:15