Problem: Let s and t be odd integers with $ s > t \geq 1$ and $\gcd(s,t)=1$. Prove that the three numbers

$ st , \frac{s^2-t^2}{2}, $ and $\frac{s^2+t^2}{2} $

are pairwise relatively prime.

Proof: Let $s=2k_1+1$ and $t=2k_2+1$, where $\gcd(k_1,k_2)=1$.

$st=(2k_1+1)(2k_2+1)=4k_1k_2+2k_1+2k_2+1=2(2k_1k_2+k_1+k_2)+1$. So st is odd.

$\frac{s^2-t^2}{2}=\frac{(2k_1+1)^2-(2k_2+1)^2}{2}=\frac{4k_1^2+4k_1+1-4k_2^2-4k_2-1}{2}=2(k_1^2+k_1-k_2^2-k_2)$. So $\frac{s^2-t^2}{2}$ is even.

$\frac{s^2+t^2}{2}=\frac{(2k_1+1)^2+(2k_2+1)^2}{2}=\frac{4k_1^2+4k_1+1+4k_2^2+4k_2+1}{2}=2(k_1^2+k_1+k_2^2+k_2)+1$. So $\frac{s^2+t^2}{2}$ is odd.

So, $\gcd(st,\frac{s^2-t^2}{2})=1$, because st is odd and $\frac{s^2-t^2}{2}$ is even. So, $\gcd(\frac{s^2+t^2}{2},\frac{s^2-t^2}{2})=1$, because $\frac{s^2-t^2}{2}$ is odd and $\frac{s^2-t^2}{2}$ is even. But st and $\frac{s^2+t^2}{2}$ are both odd, $\dots$

This is where I am stuck can is it correct for me to choose $k_1$ and $k_2$ to be relatively prime to each other? If not can I use that fact to show that the $\gcd(st,\frac{s^2+t^2}{2})=1$?

  • $\begingroup$ You seem to think that the gcd of an odd number and an even number must be $1$. This is not the case $-$ for instance, $\gcd(3,6)=3$. $\endgroup$ – TonyK Oct 9 '15 at 11:40

If a prime $p\ne2$ divides into $\text{gcd}(st,\frac{s^2+t^2}2)$, then either $p$ divides $s$, or it divides $t$. It cannot divide into both of them, because $\text{gcd}(s,t) = 1$. Therefore $p$ cannot divide into $s^2+t^2$, and we arrive at a contradiction.

The case $p=2$ is slightly different. You know $st$ is odd. Show that only one of $\frac{s^2+t^2}2$ and $\frac{s^2-t^2}2$ can be even.

  • $\begingroup$ Just to clarify. Didn't I just show the part of p=2? $\endgroup$ – Username Unknown Dec 29 '13 at 19:42
  • $\begingroup$ Yes you did show the part $p=2$. $\endgroup$ – Stephen Montgomery-Smith Dec 29 '13 at 21:47
  • $\begingroup$ why do we have a case for p = 2 though? if s and t are odd then st is always odd, and p=2 cannot divide it, so st is relatively prime to the other two numbers? $\endgroup$ – Zeina Migeed Feb 9 '16 at 17:06
  • $\begingroup$ I think the question might have been edited since I posted my answer. I was explaining how to show $\text{gcd}(\frac{s^2+t^2}2,\frac{s^2-t^2}2) = 1$. $\endgroup$ – Stephen Montgomery-Smith Feb 10 '16 at 0:24

Note of $p$ divides any two of them, it should divide the third also. So just concentrate on the last two: $$ p |\frac{s^2-t^2}{2}, ~\hbox{and}~p |\frac{s^2+t^2}{2} \Rightarrow p | s^2, ~\hbox{and}~~ p |t^2$$ since if $p$ divides two numbers, it should divide their sum and difference.

But $s$ and $t$ are co-prime, so no such $p$ exists.


Suppose prime $\,p\,$ divides two of $\,(a,b,c) = (st,(s^2-t^2)/2,(s^2+t^2)/2).\,$ Then it divides the third too, by $\,a^2\!+b^2\!-c^2=0.\,$ By Bezout $\,(s,t)=1\, \Rightarrow\, \color{#c00}{is+jt=1}\,$ for some $\,i,j\in \Bbb Z.\,$ Hence

$$p\mid \color{#0a0}{a,b,c}\, \Rightarrow\, p\mid i^2(\color{#0a0}{b\!+\!c})+2ij\color{#0a0}a+j^2(\color{#0a0}{c\!-\!b})= i^2s^2\!+2ijst+j^2t^2 = (\color{#c00}{is\!+\!jt})^2 = \color{#c00}1\ \Rightarrow\Leftarrow$$

Remark $\ $ Or, by ideals $\,(p)\supseteq(a,b,c)\,\Rightarrow\, (p)\supseteq (b\!+\!c,a,c\!-\!b)=(s^2,st,t^2)=(s,t)^2 = (1)$


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.