7
$\begingroup$

Poonen, Schaefer, & Stoll give the primitive solutions to $x^2+y^3=z^7$:

$$ (±1, −1, 0), (±1, 0, 1), ±(0, 1, 1), (±3, −2, 1), (±71, −17, 2),\\ (±2213459, 1414, 65), (±15312283, 9262, 113), (±21063928, −76271, 17). $$

I'm looking for all the solutions with $1\le z\le\ell$ for some fixed $\ell$.

Clearly the primitive solutions yield infinitely many imprimitive solutions via multiplication by $\operatorname{lcm}(2,3,7)=42$nd powers, but this doesn't find them all. For example, $250^2+25^3 = 5^7$ and $832^2+112^3 = 8^7$.

How can I find all the solutions? My first instinct was to deal with each special case on its own but I'm afraid I'll miss cases that way.

[1] Bjorn Poonen, Edward F. Schaefer, and Michael Stoll, Twists of X(7) and primitive solutions to x^2+y^3=z^7, Duke Math. J. 137:1 (2007), pp. 103-158.

$\endgroup$
1
  • 1
    $\begingroup$ Poonen, Schaefer and Stoll prove a hard theorem by some hard work. If you allow the gcd of x,y,z to be some number $d>1$ then dividing the equation by $d^7$ and tidyinggives a different equation with primitive solutions $x/d$, $y/d$, $z/d$, and to find all primitive solutions to this will be more hard work. And as far as I can see this hard work is different for every $d$ and most of it has not been done. So I think that you're just going to have to do a computer search and hope $\ell$ is not too big; I think the preceding comments are in some sense a proof that the PSS result cannot help you. $\endgroup$
    – znt
    Jun 7, 2016 at 18:01

1 Answer 1

14
$\begingroup$

You can find the solutions for any given $z$ by looking for the integral points on the elliptic curve $$x^2 = (-y)^3 + z^7$$ (which would usually be written $y^2 = x^3 + z^7$). The curve is isomorphic to the curve obtained by replacing $z^7$ with $z$, so the computation is feasible for reasonable values of $z$.

Magma (for example, but also SAGE) has an implementation of a procedure that finds all these points (there are finitely many in each case) in quite reasonable time. For example, there are exactly 990 solutions (up to a sign change in $x$) for $1 \le z \le 1000$. Here is the Magma code (replace ell by $\ell$):

list := [];
for z := 1 to ell do
  pts := IntegralPoints(EllipticCurve([0, z^7]));
  list cat:= [<Abs(pt[2]), -pt[1], z> : pt in pts];
end for;
list;

If you do not have access to Magma, you can try it out with the online Magma calculator. With ell=100, it takes about 13 seconds there. The computation for $\ell = 1000$ on my computer took about as long as writing this answer.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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