2

The Mordell-equation $$x^2=y^3+k$$ with integer $k\ne 0$ is known to have only finite many integer solutions $(x/y)$

Given a positive integer $n$, can I construct an integer $k$ such that there are at least $n$ integer solutions to the above equation (with $x\ge0$) ?

An example with (at least) $24$ solutions for $y$ :

? n=54225;print(n);q=0;for(y=-1000,10^8,if(issquare(n+y^3)==1,q=q+1;print1(y," "
)));print;print(q)
54225
-36 -30 -24 -20 4 10 15 30 75 90 96 120 174 235 636 1180 1494 1830 3775 5266 607
0 9000 10860 95295
24
?
Peter
  • 84,454

1 Answers1

1

I imagine this is probably a duplicate of some other question but I cannot quite find it via the search.

Yes, you can construct a Mordell curve with arbitrarily many integer points if there are no restrictions to $k$. I will rotate the usage of $x$ and $y$ and consider the curve as $$ E: y^2 = x^3 + k $$ instead, so we want $y\geq 0$.

The idea is to pick a rank 1 curve, generate $n$ rational points using the generator $G$ and transform them into integer points. Looking at denominators of the points, let $D=LCM(${denominators}$)$. Now multiply by $D^6$ to transform the curve into

$$ E: (D^3y)^2 = (D^2 x)^3 + (D^6k) \Leftrightarrow Y^2 = X^3 + D^6k $$ Then, by definition of $D$, each of the $n$ points $(D^2x_i,D^3y_i)$ must now be integral.

Example:
Consider the Mordell curve over $\mathbb Q$ $$E: y^2 = x^3 + 2$$ It has rank one and the only torsion point is the neutral point $\mathcal O$. The generator is $G=(-1,1)$, we can find the first 5 rational points as $$ (x_k,y_k) = [k]G $$ for $1\leq k\leq 5$, where $[\cdot]$ denotes the point multiplication operation. This gives the following points

$$ \begin{align*} G &= (-1,1) \\ [2]G &= (17/4,-71/8) \\ [3]G &= (127/441 , 13175/9261) \\ [4]G &= (66113/80656 , -36583777/22906304) \\ [5]G &= (108305279/48846121 , 1226178094681/341385539669) \end{align*} $$ Next we find the LCM: $$ D = LCM(4,8,441,9261,80656,22906304,48846121,341385539669) = 72419917504456587635136 $$ Finally we apply the transformation $(x',y') = (D^2x,D^3y)$: $$ \begin{align*} G &= (-5244644451352297667755999615909647085061738496, 379816718507139378146913182774552967922429082711157728996445493395456)\\ [2]G &= (22289738918247265087962998367616000111512388608, -3370873376750861981053854497124157590311558109061524844843453753884672)\\ [3]G &= (1510362461047033568718847961951304262591475712, 540339624914324728116356892674088689383220296373988022840748231884800)\\ [4]G &= (4298988030800615648040473152730540787228311808, -606607252341406100174229116883478323397886401063198574917734250057728) \\ [5]G &= (11628818602801899585790299342807326836436987904, 1364214022300501406044776062737876197523992448609719798004173622738944) \end{align*} $$ and we check that each $(x,y)$ satisfies the new curve $$ E': Y^2 = X^3 + 2D^6 $$

Yong Hao Ng
  • 4,795
  • 19
  • 26
  • I think I remember that I asked this question, but forgot it in the meantime that I did it. Anyway, thank you for the answer. – Peter Jun 19 '18 at 16:14
  • @Peter If you meant this one it doesn't feel like the same for me. I was thinking more of "Are there elliptic curves with infinitely many integer points" kind of question. – Yong Hao Ng Jun 20 '18 at 02:13