In the following paper (http://www.ams.org/journals/jams/1991-04-04/S0894-0347-1991-1119199-X/), C.L. Stewart showed that there exist infinitely many integers $h$ such that the equation

(1) \begin{equation} xy(x+y) = h \end{equation}

has at least 18 solutions in co-prime integers $x$ and $y$. In the same paper, he also conjectured (his conjecture covers far more) that there exists positive numbers $r,c$ such that for all integers $h$ with $|h| \geq r$, the equation (1) has at most $c$ solutions in co-prime integers $x$ and $y$.

My question is, is it reasonable to conjecture that one may take $c = 18$? That is, does there exist infinitely many $h$ for which (1) has at least 19 solutions in co-prime integers $x$ and $y$?

  • 10
    $\begingroup$ You probably already know this, but letting $E_h$ be the elliptic curve (1), there is a bound $$\#\text{(coprime integer solutions)}\le C^{1+\text{rank }E_h(\mathbb Q)}$$ for an absolute constant $C$. So if the current conjecture that elliptic curve ranks are bounded is true, then Stewart's conjecture would follow. $\endgroup$ Commented Jan 10, 2018 at 1:12
  • $\begingroup$ Also that if you drop the coprimality condition then the number of solutions is unbounded. $\endgroup$ Commented Jan 11, 2018 at 2:30
  • $\begingroup$ @NoamD.Elkies As Noam says, without the gcd condition, the number is unbounded, but I believe that it is not known if for all $\epsilon>0$, the number of solutions can be greater than $O_\epsilon(|h|^{1-\epsilon})$ for infinitely many $h$. This seems to be a very interesting, and also very difficult, problem. $\endgroup$ Commented Jan 11, 2018 at 4:31
  • $\begingroup$ Do you mean $O_\epsilon(\log|h|)^{1-\epsilon}$? $\endgroup$ Commented Jan 11, 2018 at 4:36
  • $\begingroup$ (That's what one would get from elliptic curves of unbounded rank in this family.) $\endgroup$ Commented Jan 11, 2018 at 4:52

2 Answers 2


There seem to be lots of $h$ which have $20$ solutions. Among squarefree $47$-smooth numbers I found $$101623830, 363993630, 455885430, 1418488890, 15427730010, 31983962010, 322640788470, 1087394017710, 11406069164490, 304250263527210$$ with $20$ solutions and $903210$ and $23730036330$ with $24$.

EDIT: Oops: my program was counting only solutions with $x > 0$ (but $y$ of any sign). Those numbers $101623830, \ldots, 304250263527210$ actually have $30$ solutions each, and $903210$ and $23730036330$ have $36$.

  • $\begingroup$ Yes, the number of solutions should be a multiple of 6 since whenever $(x,y)$ is a solution then so are $(y,x), (y, - x - y), (-x-y, x), (-x-y, y), (x,-x-y)$, and these should typically be co-prime whenever $(x,y)$ are co-prime with only finitely many exceptions. $\endgroup$ Commented Jan 12, 2018 at 10:35

There are infinitely many such numbers with at least $12$ solutions. Here is the new proof:

If $(x,y)$ is a solution to



$$(y,x), (x,-x-y),(-x-y,x),(y,-x-y),(-x-y,y)$$

are also solutions to


In fact, one can take simultaneously $x=1$ and $x=2$ to get two positive integer solutions $(1,a)$ and $(2,b)$ provided


If $2a+1=k$ and $b+1=l$, one has:


a generalized Pell equation with infinitely many positive integer solutions $(k,l)$, and obviously $k$ must be odd. In fact, $h=\frac{k^2-1}{4}=2l^2-2$.

This brings up at least $2\cdot6=12$ solutions for infinitely many $h$.


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.