I was recently trying to play with mean inequalities and Jensen inequality. My question is, if the following relation holds for any positive real numbers $a$ and $b$ $$(1+a)(1+a+b)\geq\sqrt{27ab}$$ and if it does, then how to prove it. By AM-GM inequality we could obtain $1+a\geq\sqrt{4a}$ and $1+a+b\geq\sqrt[3]{27ab}$, so by putting this together we obtain$$(1+a)(1+a+b)\geq\sqrt{4a}\sqrt[3]{27ab},$$but this is slightly different from what I wanted. It's probably true that for all positive $a$ and $b$ there is $\sqrt{4a}\sqrt[3]{27ab}\geq\sqrt{27ab}$, but could there be equality?

Thanks a lot.


Use AM-GM: $$\frac{1}{2}+\frac{1}{2}+a\ge3\sqrt[3]{\frac{a}{4}}\\ \frac{1}{2}+\frac{1}{2}+a+\frac{b}{3}+\frac{b}{3}+\frac{b}{3}\ge6\sqrt[6]{\frac{ab^3}{108}}\\\therefore(1+a)(1+a+b)\ge\left(3\sqrt[3]{\frac{a}{4}}\right)\left(6\sqrt[6]{\frac{ab^3}{108}}\right)=\sqrt{27ab}$$

Equality holds iff $\frac{1}{2}=a=\frac{b}{3}$.

  • $\begingroup$ That is something for I was originally looking for. Very elegant solution! Thank you. $\endgroup$ – PrimeLord May 8 '16 at 23:43
  • 1
    $\begingroup$ This is just multiplying two ordinary AM-GM inequalities, not using weights? $\endgroup$ – M.M May 9 '16 at 1:22

The two variables function defined on $[0,+\infty)^2$ $$ f(x,y)=(1+x)(1+x+y)-\sqrt{27xy} $$ has minimum $0$ in $P_0=\left(\frac{1}{2},\frac{3}{2}\right)$.

Let us check the borders of the domain. We have $$ f(x,0)=(1+x)^2>0\\ f(0,y)=1+y>0 $$ moreover$$ \lim_{x\to+\infty}f(x,y)=+\infty\quad\forall y>0\\ \lim_{y\to+\infty}f(x,y)=+\infty\quad\forall x>0 $$ we can conclude that the inequality is strictly satisfied on the whole domain, with the exception of $P_0$, where the equal sign holds.

  • $\begingroup$ I had never thought about it that way, thanks for interesting contribution! $\endgroup$ – PrimeLord May 8 '16 at 23:44

It's probably true that for all positive $a$ and $b$ there is $\sqrt{4a}\sqrt[3]{27ab}\geq\sqrt{27ab}$

Since these quantities are all positive, raise both sides to the sixth power. Hence the given inequality is equivalent to $$4^3a^327^2a^2b^2\ge 27^3a^3b^3$$ which in turn is equivalent to $$4^3a^2\ge 27 b$$

Unfortunately, this need not be true.

  • $\begingroup$ Ou, you're right, thank you for correcting my mistake. $\endgroup$ – PrimeLord May 8 '16 at 16:18

We can begin by homogenizing the inequality to the equivalent inequality


with $a,b,c\gt0$, which can then be rewritten as


with $b,s\gt0$ and $0\lt c\lt s$. It's easy to see that, for any fixed $s\gt0$, the cubic $c^2(s-c)$, which is $0$ at $c=0$ and $c=s$, has a maximum at $c={2\over3}s$, and thus


as desired.

  • $\begingroup$ Is there a more general or explicit definition of "homogenize" as used here? $\endgroup$ – Ovi May 9 '16 at 6:53
  • $\begingroup$ @Ovi, all I'm doing is giving both $a$ and $b$ a denominator, $c$, then clearing it out. $\endgroup$ – Barry Cipra May 9 '16 at 11:17
  • $\begingroup$ Hm ok, i tought you just set c=1 $\endgroup$ – Ovi May 9 '16 at 15:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.