Let $x,y,z\in \mathbb R^+$ prove that:

$$\sqrt {(x^2y+ y^2z+z^2x)(y^2x+z^2y+x^2z)} \ge xyz + \sqrt[3]{xyz(x^2+yz)(y^2+xz)(z^2+xy)}$$

The inequality $\sqrt {(x^2y+ y^2z+z^2x)(y^2x+z^2y+x^2z)} \overset{C-S}{\ge} 3xyz$ will not help because $3xyz \le RHS$. How to prove above inequality ? It is too hard for me.


The inequality is homogeneous, so we may assume that $xyz=1$. Then if we expand left and right hand side, we see that we have to show that $$\sqrt{x^3 + y^3 + z^3 + \frac{1}{x^3} + \frac{1}{y^3} + \frac{1}{z^3} + 3} \ge 1 + \sqrt[3]{x^3 + y^3 + z^3 + \frac{1}{x^3} + \frac{1}{y^3} + \frac{1}{z^3} + 2}.$$ Let $$u = \sqrt[3]{x^3 + y^3 + z^3 + \frac{1}{x^3} + \frac{1}{y^3} + \frac{1}{z^3} + 2}.$$ If we square both sides of the expanded inequality, we get the equivalent inequality $$u^3 \ge 2u + u^2,$$ or, $$u^2 - u - 2 \ge 0.$$ Therefore it is enough to show that $u \ge 2$. But this is certainly true since $x^3 + \frac{1}{x^3}, y^3 + \frac{1}{y^3}, z^3 + \frac{1}{z^3} \ge 2$.

  • $\begingroup$ $x=y=z=2$ violates your assumption, no? $\endgroup$ – response May 30 '13 at 15:41
  • $\begingroup$ @response what do you mean? I'm allowed to do that assumption because I'm allowed to multiply all of the variables $x,y,z$ by a constant of my choicing (because of the form of the inequality). In particular I'm allowed to divide all of them by $2$ before I start. $\endgroup$ – J. J. May 30 '13 at 15:44
  • $\begingroup$ @response What he means is the following: let $x'=\frac{x}{\sqrt[3]{xyz}}$ and so on. Then the inequality for $x',y',z'$ is exactly the same.... $\endgroup$ – N. S. May 30 '13 at 15:44
  • $\begingroup$ I see it now. Somehow I glossed over the part where you talk about it being a 'homogeneous' system. $\endgroup$ – response May 30 '13 at 15:45

$$\sqrt {(x^2y+ y^2z+z^2x)(y^2x+z^2y+x^2z)} \overset{AM-GM}{\ge} 3xyz$$

DOES help, since you can also prove:

$$2\sqrt {(x^2y+ y^2z+z^2x)(y^2x+z^2y+x^2z)} \geq 3 \sqrt[3]{xyz(x^2+yz)(y^2+xz)(z^2+xy)}$$

Note that this is the same as

$$2^6(x^2y+ y^2z+z^2x)^3(y^2x+z^2y+x^2z)^3 \geq 3^6x^2y^2z^2(x^2+yz)^2(y^2+xz)^2(z^2+xy)^2 \,.$$

which after long computations and many tears can be reduced to this all junk being positive. If you look to the few negative terms , you can get easely rid of them by invoking the AM-GM inequality.

Now, all this suggests that the inequality

$$2\sqrt {(x^2y+ y^2z+z^2x)(y^2x+z^2y+x^2z)} \geq 3 \sqrt[3]{xyz(x^2+yz)(y^2+xz)(z^2+xy)}$$

is true, all you have to do is find a neater solution :)

  • $\begingroup$ heh, +1 for the work :) $\endgroup$ – J. J. May 30 '13 at 15:58
  • $\begingroup$ @J.J. You should give a +1 to WA then :D $\endgroup$ – N. S. May 30 '13 at 15:59
  • $\begingroup$ I assume you at least checked that you really can bunch the terms. :) $\endgroup$ – J. J. May 30 '13 at 16:00
  • $\begingroup$ @J.J. Yea, but since the first alternate form is a product of huge polynomials, and I only saw 1 minus, in only took me few seconds.... $\endgroup$ – N. S. May 30 '13 at 16:01
  • $\begingroup$ Oh, you're right indeed. I stared a while at the second form. :P $\endgroup$ – J. J. May 30 '13 at 16:02

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.