Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Problem: If $$ x^n+y^n=z^n$$ where $x,y $ and $ z $ are real numbers $\gt 0$ and n is any integer $ \neq 0$ then $$ \frac {x}{n} + y >z$$ assuming $x \lt y $

Background: While studying Fermat's theorem I saw that the following are the characteristics of $x,y$ and $z$ $$ x+y \gt z\gt y \gt x$$ But while studying further, I also saw the above to be true.

Request: I could prove this when $n \gt 0$ but couldn't do the same for $n \lt 0$. If anyone could prove or disprove both the cases or show me any site where this has been discussed. A detailed proof would be appreciated. I shall share my proof shortly for a review for the case $n \gt0$

share|cite|improve this question
up vote 2 down vote accepted

Here is a proof for $n \geq 1$, unfortunately for $n<0$ we get a weaker ienquality.

For all $n \notin(0,1)$ we have

$$(y+\frac{x}{n})^n = y^n(1+\frac{x}{ny})^n > y^n(1+\frac{x}{y})$$

with the last inequality following from Generalized Bernolli

For $n \geq 1$ you also have:

$$ y^n+xy^{n-1}> y^n+x^n =z^n\,,$$

and by combining these two you get the desired result.

Maybe someone can fix it in general.

Added: Inspired by only's answer.

If $n=-1$, your equation is

$$\frac{1}{x}+\frac{1}{y}=\frac{1}{z} \Rightarrow xz+yz+=xy \Rightarrow z= \frac{xy}{x+y}$$

You need $y-x >z$, or equivalently

$$y-x > \frac{xy}{y+x} \Leftrightarrow y^2 > x^2+xy$$

Note that $y=3, x=2$ is a counterexample to this equation, thus $y=3, x=2, z=\frac{6}{5}, n=-1$ is a counterexample to your claim.

In general, if $n=-m<0 $ you want to prove that

$$\frac{1}{x^m}+\frac{1}{y^m} =\frac{1}{z^m} \mbox{implies} \, y-\frac{x}{m} >z$$

but since

$$\frac{z^m}{y^m}+\frac{z^m}{x^m}=1\,,$$ if you set $\frac{z^m}{y^m} =\frac{1}{2}+\alpha$ and $\frac{z^m}{x^m} =\frac{1}{2}-\alpha$, the desired inequality becomes

$$\frac{1}{\sqrt[m]{\frac{1}{2}+\alpha}}-\frac{1}{m\sqrt[m]{\frac{1}{2}-\alpha}} >1$$ Since this is increasing in $\alpha$, is easy to see that this is true for all $0 < \alpha < \frac{1}{2}$ if and only if

$$\frac{m-1}{m\sqrt[m]{\frac{1}{2}}} =\lim_{\alpha \to 0^+} \frac{1}{\sqrt[m]{\frac{1}{2}+\alpha}}-\frac{1}{m\sqrt[m]{\frac{1}{2}-\alpha}} \geq 1$$

or equivalently

$$2 \geq (\frac{m}{m-1})^m $$.

But this is not possible, since the RHS is a sequence which is decreasing to $e$. This shows that for each $m$ there exists some $\alpha$ which makes the inequality fail....

share|cite|improve this answer
I couldn't get the last portion of your answer for n>0 . Is it possible to just fill up the last few lines? Thanks for your help – Barun Dasgupta Aug 11 '12 at 9:27
My answer for the case n>0 is as follows. Can you kindly verify and comment. Let $\frac{z}{x}=q$ and $\frac{y}{x}=p$. This shows $1=q^n - p^n \Leftrightarrow 1 = (q-p)(q^{n-1}+\cdots+p^{n-1}).$ Now since $q,p>1$,each term in the bracket $(q^{n-1}+\cdots+p^{n-1})$ is greater than $1$ and we have $n$ such terms in the bracket. So $(q^{n-1}+\cdots+p^{n-1})>n$ Which makes $$(q-p)<1/n$$$$\Leftrightarrow q<p+1/n$$ $$\Leftrightarrow z/x<y/x+1/n$$$$\Leftrightarrow z<y+x/n$$ – Barun Dasgupta Aug 11 '12 at 9:46
@BarunDasgupta $$(y+\frac{x}{n})^n > y^n(1+\frac{x}{y})=y^n+xy^{n-1}\geq y^n+x^n=z^n$$ – N. S. Aug 18 '12 at 6:45

$n = -1, x = y = 2, z = 1$ is a solution (I know you require $x < y$; just decrease x by a bit and increase y by a bit.)

Then, we can make $y-x$ arbitrarily close to 0, while z = 1, so this disproves your assertion for $n = -1.$

share|cite|improve this answer
But if you increase $y$ and decrease $x$ then it is no longer true that $y-x=0$. – Asaf Karagila Aug 9 '12 at 17:35
But we can make it arbitrarily small, and then it'll still be < 1. For example; y = 2.1, x = $\frac{21}{11}$ and $y-x = \frac{1}{11} + \frac{1}{10},$ which is still < 1. – only Aug 9 '12 at 17:58
True, but observe that this is still not $0$. – Asaf Karagila Aug 9 '12 at 18:13
Sorry; by "y-x = 0", I meant "in this case, y-x = 0, so if you require $x < y$ we can just change it a bit. I'll edit my answer to be more precise. – only Aug 9 '12 at 18:23

1) The estimates pertaining to fermat's equation have been extensively studied. One may see this paper

2) Your problem is a theorem of Grunert, and it admits a simple proof.

Hope these helpful.


share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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