3

I tried going for a common denominator but then it turned the whole inequality into a big muddle... I also tried multiplying the brackets out but to not much avail...

I also browsed through some known inequalities such as Cauchy and AM-GM (I only know the simpler ones) but I couldn't find any known inequalities that I could use in this problem.

Also, I tried to look for other problems on this site with similar inequalities, but I couldn't find much relevant inequalities.

I'm legit stuck.

If $z=0$, then there is a divisibility by $0$ and the world explodes.

Here is the question again:

For $x, y > 0$, can you show that

$$\frac{x(2x-y)}{y(2z + x)} + \frac{y(2y-z)}{z(2x+y)}+\frac{z(2z-x)}{x(2y+z)}\geqslant 1$$

Multiplied out:

$$\frac{2x^{2}-xy}{2yz+xy} + \frac{2y^{2}-yz}{2xz+zy} + \frac{2z^{2}-xz}{2xy+xz}\geqslant 1$$

Common Denominator would be $xyz(2z+x)(2x+y)(2y+z)$

Any contribution will be deeply appreciated.

Please don't mark this one as 'unconstructive' or 'duplicate' because I really need an answer for an assignment. Help me out please?

Thanks a lot :)

mathsnoob
  • 667
  • 4
    You seemed okay with the question being closed yesterday. Also, this site is community-run; as you get more reputation points, you get more abilities to help out the moderators on the site. Anyone with more than 3000 points can vote to close a question; the people who closed your original question were not "admins". – Zev Chonoles Feb 24 '13 at 19:53
  • Yea I was okay with it yesterday because I thought I could solve this on my own, but then it turns out that trying to get a common denominator doesn't work... :/ Can anyone please help? I have to hand this in in about 12 hours... – mathsnoob Feb 24 '13 at 19:55
  • How did the common denominator not work? Should z>0 as well in this question? If z=0, the left side of the inequality is not computable. – JB King Feb 24 '13 at 19:58
  • 3
    @mathsnoob We are here to help you understand mathematics and answer questions about it. But that is not the same as being here to do your homework. As it stands, it looks like you are just asking us to solve this exercise for you. If you were to edit the post to include your thoughts on the question and what you've tried so far, people would be much more inclined to help you. – Alex Becker Feb 24 '13 at 19:59
  • @JBKing: Well when I tried a common denominator, the whole thing turned into an extremely complicated inequality... Help... Please... :'( – mathsnoob Feb 24 '13 at 19:59
  • Have you tried with some values, say see what happens if $x = y = z$? What if one is much larger than the others? – vonbrand Feb 24 '13 at 20:00
  • @mathsnoob Do you meant for $z$ to be greater than $0$ too? I suggest you write your computations on the question. You'll lose a few minutes and in return you'll probably get an answer. – Git Gud Feb 24 '13 at 20:01
  • Erm... Okay I'm sorry that I couldn't solve this on my own, but please someone help me? At least give me a clue on how to start? – mathsnoob Feb 24 '13 at 20:02
  • No, z doesn't have to be greater than 0 too – mathsnoob Feb 24 '13 at 20:03
  • 3
    @mathsnoob Look the best way to get an answer on this site is to do this. At the top of your question, write the statement, i.e. "Prove that [insert inequality] is true." Then below that write out your detailed scratch work, like "Begin with this assumption. Then this is true then this is true blah blah." Your attempt at the proof. Then, somewhere you'll get stuck. At that point, ask a specific question, like "why doesn't this work?" or "can I get a hint on this next step?" rather than "OK now solve this problem for me." The reason this will help you is that it will show us you cared – Alexander Gruber Feb 24 '13 at 20:16
  • 1
    enough to try - you showed 'research effort.' Sometimes you'll even find the answer to the problem yourself trying to do this. The general ideology is if we're going to take time to answer your question, we want to know you've taken as much time to try to answer it yourself. So showing your steps not only helps prove that, it helps us see where, specifically, you're confused. – Alexander Gruber Feb 24 '13 at 20:17
  • As for where to start: I would look at some other inequality questions on SE that have been answered, to see some methods for proving tricky inequalities like these. – Alex Zorn Feb 24 '13 at 20:19
  • Kinda edited it... is it any better? – mathsnoob Feb 24 '13 at 20:22
  • Don't use $$ in titles. It breaks the front page. – Asaf Karagila Feb 24 '13 at 20:23
  • So... how do I make those fractions then? – mathsnoob Feb 24 '13 at 20:24
  • @mathsnoob You still haven't done the most important thing: show us your attempts at cracking this. – Git Gud Feb 24 '13 at 20:25
  • Use one $ symbol. Like it is now. – Asaf Karagila Feb 24 '13 at 20:25
  • Okay thanks Asaf Karagila. @Git Gud: Sure, let me edit the problem a bit – mathsnoob Feb 24 '13 at 20:26
  • Edited the problem a bit, showing what I've tried. Any better now? – mathsnoob Feb 24 '13 at 20:32
  • @mathsnoob You just described what you tried, you haven't actually shown it to us.Type your calculations. – Git Gud Feb 24 '13 at 20:36
  • Wow really? Multiply the thing out??? That's gonna take AAAGGES to type – mathsnoob Feb 24 '13 at 20:55
  • 1
    @mathsnoob You got around 11 hours, I believe. At the moment I don't think anyone is willing to help you unless you do that. – Git Gud Feb 24 '13 at 20:57
  • 1
    However 6 hours of that 11 is bedtime :) I edited the problem, again – mathsnoob Feb 24 '13 at 20:59
  • 1
    For gods sakes you closed it again – mathsnoob Feb 24 '13 at 21:03
  • 6
    @mathsnoob If somebody solves it for you they're going to have to take AAAGGES to type it too. I don't see why you think we're some kind of on-demand solutions manual. It's not like we get paid for this or anything. – Alexander Gruber Feb 24 '13 at 21:46
  • 2
    Hmm. So you are in IMO training camp? Hopefully you tell the coach that you got help for some of the homework problems assigned to you another aspirants. I mean, it would not be fair, if you were selected to the team in place of a better problem solver just because you enlisted the help of hundreds of pros, would it? – Jyrki Lahtonen Feb 26 '13 at 18:51
  • Yea, he already knows, and we already discussed it. I did not hand in the problems that I got help with on this site. Thanks Jyrki Lahtonen – mathsnoob Feb 27 '13 at 15:35
  • 1
    @mathsnoob You will need to assume that $z>0$. Try to substitute $x=1$, $y=1$, $z=-{1\over 4}$, for instance, and the inequality will fail. – Per Erik Manne Feb 28 '13 at 14:48

2 Answers2

3

I believe that the deadline of the assignment is probably over, so let me post an idea.

Note that

$$\sum \frac{x^2}{2yz + xy} \ge \frac{(\sum x)^2}{3(xy+yz+zx)} \ge 1$$

by Cauchy-Schwarz. So it suffices to show that $\displaystyle \sum \frac{x^2-xy}{2yz+xy} \ge 0$. Note that this is equivalent to

$$\begin{align} &\sum_{cyc} (x^2-xy) \left( \frac{1}{2yz+xy} - \frac{1}{x^2+xy+xz} \right) \ge 0 \\ &\Leftrightarrow \sum (x^2-xy) \frac{x^2 - xz + 2xz - 2yz}{(2yz+xy)(x^2+xy+xz)} \ge 0 \\ &\Leftrightarrow \sum \frac{x^2(x-y)(x-z)}{(2yz+xy)(x^2+xy+xz)} + \sum 2xz\frac{(x-y)^2}{(2yz+xy)(x^2+xy+xz)} \ge 0 \end{align}$$

It suffices to show that the first sum is non-negative. We further simplify it,

$$\sum \frac{x^2(x-y)(x-z)}{(2yz+xy)(x^2+xy+xz)} = \frac{1}{x+y+z} \sum \frac{x(x-y)(x-z)}{(2yz+xy)} $$

and it now suffices to show that the inner sum is non-negative. Since the sum is cyclic, WLOG assume that $x$ is the largest.

If $x \ge y \ge z$, note that among the three summands, only the $y(y-z)(y-x)$ term is negative. But note that when we consider it with the $x(x-y)(x-z)$ term together, it is $$\begin{align} &\frac{x(x-y)(x-z)}{(2yz+xy)} + \frac{y(y-z)(y-x)}{(2zx+yz)} \\ &\ge \frac{x(x-y)(y-z)}{(2yz+xy)} + \frac{y(y-z)(y-x)}{(2zx+yz)} \\ &= (x-y)(y-z) \left(\frac{x}{2yz+xy} - \frac{y}{2zx+yz}\right) \\ \end{align}$$ The denominator of the bracket is $2x^2z + xyz - 2y^2z - xyz = 2z(x^2-y^2) \ge 0$, so we are done for this case.

If $x \ge z \ge y$, note that among the three summands, only the $z(z-x)(z-y)$ term is negative. But note that when we consider it with the $x(x-y)(x-z)$ term together, it is $$\begin{align} &\frac{x(x-y)(x-z)}{(2yz+xy)} + \frac{z(z-x)(z-y)}{(2xy+zx)} \\ &\ge \frac{x(z-y)(x-z)}{(2yz+xy)} + \frac{z(z-x)(z-y)}{(2xy+zx)} \\ &= (z-y)(x-z) \left(\frac{x}{2yz+xy} - \frac{z}{2xy+zx}\right) \\ \end{align}$$ The denominator of the bracket is $2x^2y + x^2z - 2yz^2 - xyz = 2y(x^2-z^2) + xz(z-y) \ge 0$, so we are also done for this case.

0

A full expanding gives: $$\sum_{cyc}(x^4z^2+2x^4yz-x^3y^2z-x^3z^2y-x^2y^2z^2)\geq0$$ or $$xyz\sum_{cyc}(x^3-x^2y-x^2z+xyz)+\sum_{cyc}(x^4z^2-x^2y^2z^2)+xyz\sum_{cyc}(x^3-xyz)\geq0,$$ which is Muirhead and AM-GM.

Done!