8
$\begingroup$

How do I prove that the equation $x^4 - y^4 = 2 z^2$ has no solutions using the fact that the equations $x^4 + y^4 = z^2$ and $x^4 - y^4 = z^2$ have no solutions. I cant think of a method of reducing the above equation to one of these forms.

$\endgroup$
8
  • $\begingroup$ Are $x,y,z$ integers? $\endgroup$
    – Jemmy
    Mar 22, 2013 at 18:55
  • $\begingroup$ Can you substitute $z = \frac{z'}{\sqrt{2}}$ for $z$ in the first equation? Then you have an equation of the form $x^4+y^4=z^2$. Apologies if this is in error - I can't remember if this is valid. $\endgroup$
    – SSumner
    Mar 22, 2013 at 18:55
  • $\begingroup$ apparently $x,y,z$ are integers. $\endgroup$
    – user63181
    Mar 22, 2013 at 18:56
  • 1
    $\begingroup$ I assume $x$ and $y$ must be distinct, or you have a solution. $\endgroup$ Mar 22, 2013 at 19:00
  • $\begingroup$ yah x,y,z are integers $\endgroup$
    – noddy
    Mar 22, 2013 at 19:01

2 Answers 2

14
$\begingroup$

I think you meant nonexistence of positive solutions. Suppose that there exist some positive solutions(meaning that all of $x$, $y$, $z$ are positive). Then there is a positive solution $(x_0,y_0,z_0)$ with $x_0$ smallest.

First, notice that parity of $x_0$ and $y_0$ cannot be different. So, either both even or both odd. Both even case is not possible, because otherwise $(x_0/2,y_0/2,z_0/4)$ is a positive solution solution with smaller $x_0$.

Similarly with odd prime $p$, suppose $p|gcd(x_0,y_0)$, then you also get smaller solution $(x_0/p, y_0/p, z_0/p^2)$. Thus we can assume $(x_0,y_0)=1$.

Thus we can now assume that $x_0$ and $y_0$ are both odd, and coprime. Then by looking at the expression $$ \frac{x_0^2-y_0^2}{2}\cdot \frac{x_0^2+y_0^2}{2}=\frac{z_0^2}{2} $$

The right side must be an integer, so it must be $2Z^2$ , and any prime $p$ cannot divide both $\frac{x_0^2-y_0^2}{2}$ and $\frac{x_0^2+y_0^2}{2}$(otherwise we would have some prime $p$ dividing both $x_0$ and $y_0$. Then we have $$ x_0^2-y_0^2 = u^2$$ , and $$x_0^2+y_0^2 = 2v^2$$ for some positive integers $u,v$.

Now we solve for the Pythagorian triple in the first equation. $$ x_0=s^2+t^2\\ y_0=s^2-t^2\\ u=2st $$ for some positive integers $s,t$.

Then $x_0^2+y_0^2=2(s^4+t^4)=2v^2$. Hence we obtain $s^4+t^4=v^2$. However, this cannot have positive integer solution.

$\endgroup$
4
  • 8
    $\begingroup$ Fermat would be proud of you. $\endgroup$ Mar 22, 2013 at 23:18
  • $\begingroup$ Why do we have $x_0^2-y_0^2 = u^2$? Couldn't we have $x_0^2-y_0^2 = 2u^2$ instead? $\endgroup$
    – user940
    Mar 23, 2013 at 3:47
  • $\begingroup$ We started from both $x_0$ and $y_0$ odd. So for all prime power divisor of $(x_0^2+y_0^2)/2$ is odd, and the powers are even in $2Z^2$. $\endgroup$ Mar 23, 2013 at 4:53
  • $\begingroup$ Oh I see. Using modulo 4, we see that $(x_0^2+y_0^2)/2$ is the odd factor, and $(x_0^2-y_0^2)/2$ is the even one. $\endgroup$
    – user940
    Mar 23, 2013 at 14:41
2
$\begingroup$

IDENTITY--The identity $(X^4 - Y^4)^2 + 4(XY)^4 = (X^4 + Y^4)^2 $ solves the proposer's question.

The proposer assumes that there is a non-zero triple of integers X,Y,Z

such that $(X^4 - Y^4 ) = 2(Z^2)$. WOLOG we may assume $(X,Y,Z) =1$.

We substitute " $2(Z^2)$ " for $(X^4 -Y^4)$ in the above IDENTITY.

This translates to $(2Z^2)^2 + 4(XY)^4 = (2B)^2$ where $X^4 + Y^4 = 2B$. ( Note that X,Y, are both odd as $X^4 = Y^4 = 2(Z^2)$ by hypothesis. Hence we may write $X^4 + Y^4) = 2B$ for some nonzero integer $B$ in the above IDENTITY.)

Dividing both sides by $4$ yields

$Z^4 + (XY)^4 = B^2$, which Fermat proved is impossible in non-zero integers $Z$,$(XY)$, and $B$ and is the connection that the proposer was seeking.

Many thanks to JW Tanner and Tinlyx for improving the formatting of this proof. I wanted to share the above IDENTITY equation with the readership because it is an important tool to shorten and simplify Fermat-type equations of exponent $4$ which may or may not have a "$2$" as a coefficient. The first answer given to this problem above is correct, but this solution is also very elegant, and simpler as well.

$\endgroup$
1
  • 1
    $\begingroup$ As it’s currently written, your answer is unclear. Please edit to add additional details that will help others understand how this addresses the question asked. You can find more information on how to write good answers in the help center. $\endgroup$
    – Community Bot
    Jan 6 at 4:49

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .