1

I am allowed to assume that $x^4-y^4=2z^2$ already has one solution in $\mathbb{N}$. From this I have shown that there exists a solution in $\mathbb{N}$ with $x,y$ both odd.

I am now trying to show that there exists a solution with $\gcd(x,y,z)=1$. My guess is that if $x,y,z$ already have a common divisor, then I can find another solution by cancelling/dividing $x$ by something dependent on $d$, and similar for $y,z$.

I started by assuming that $d$ is a common divisor of $x,y,z$. That is $x=da,y=db,z=dc$. Then $d^4(a^4-b^4)=2d^2c^2 \implies d^2$ divides $2c^2$. Here is where I am stuck. It feels like the natural thing to do here is to show that $\gcd(2,c^2)=1$, so that in fact $d^2$ divides $c^2$, but I cannot. And even then I don't know where to go from that point.

1 Answers1

0

As you stated, you've shown there exists a solution with $x$ and $y$ both odd. Thus, you have that the common factor of $d$ must be odd as well. This means when you have $d^2 \mid 2c^2$, that $\gcd(2,d^2) = 1$, so $d^2 \mid c^2 \implies d \mid c$ (e.g., this is explained at If $a^2$ divides $b^2$, then $a$ divides $b$). As such, for some integer $e \ge 1$, you have $c = ed$, which means $z = (ed)d = ed^2$. When you plug this into your equation you get

$$x^4 + y^4 = 2z^2 \implies a^4d^4 + b^4d^4 = 2e^2d^4 \tag{1}\label{eq1A}$$

You can divide both sides by $d^4$ to get

$$a^4 + b^4 = 2e^2 \tag{2}\label{eq2A}$$

If $\gcd(a,b) = f \gt 1$, then $f \mid 2e^2 \implies f \mid e^2$. However, this means at least one prime factor of $f$ divides $e$, so it would have divided $x$, $y$ and $z$ initially, but that's not possible since the gcd value of $d$ was divided out. As such, $f = 1$, so $a$, $b$ and $e$ are relatively prime to each other, i.e., $\gcd(a,b,e) = 1$.

John Omielan
  • 47,976
  • Thanks, this was very helpful. Instead of everything in the final paragraph, is it not sufficient to say that if our 'new' solution $(\frac{x}{d},\frac{y}{d},\frac{z}{d})$ has some other common divisor $d' \neq 1$, then we can just derive a new solution by the exact same method? And that we can repeat this for all common divisors, until the only remaining common divisor is $1$, or is this not rigorous? – Azamat Bagatov May 07 '20 at 21:26
  • 1
    @AzamatBagatov You're welcome. Yes, an alternative is what you suggest, i.e., that you can repeat the procedure over and over again to end up at an eventual contradiction. It's rigorous and I have used that same procedure here to answer other questions, plus I've seen others do so as well. FYI, it's an example of Proof by infinite descent. I instead proceeded as I did above to continue from what you did since with the gcd removed, you can just end it in one step as I showed above. – John Omielan May 07 '20 at 21:30