I tried to find the solution of this equation, or conclude there are none. This i what i found out:

I noticed that $x \neq -1$ mod $3$.

We can write $y^2 + 4 = x^3 - 8 = (x-2)(x^2+2x+4)$

I tried to find a prime divider from the righthand side so i could use legendre on the left hand side, but couldnt solve it. Can someone help me?

  • $\begingroup$ It is not true that $x$ odd implies $x^3-12\equiv 5$ mod($8$). For example, $5^3-12=113\equiv 1$ mod ($8$). $\endgroup$ – lulu Dec 11 '15 at 12:20
  • $\begingroup$ Oh thanks, you are right. Something went wrong there! $\endgroup$ – S. Ferwerda Dec 11 '15 at 12:22
  • 1
    $\begingroup$ If $x$ is even then you can divide through by $4$ to get the new Diophantine equation $y^2=2x^3-3$. That one is impossible (just check it mod($8$). So if there is a solution to the original equation, $x$ is odd. $\endgroup$ – lulu Dec 11 '15 at 12:25
  • $\begingroup$ Continuing from @lulu , for $x$ odd mod(8), $x^3\equiv x\pmod 8$. Since the squares mod 8 are $0$, $1$, and $4$, the only option is $x\equiv 5\pmod 8$ (all other possibilities do not leave you with squares, mod 8, on the RHS). $\endgroup$ – Michael Burr Dec 11 '15 at 12:30

As you've noticed, $y^2+4=(x-2)\left(x^2+2x+4\right)$.

If $x$ is even, then let $x=2k$. Then $y=2m$, so $m^2=2k^3-3$, impossible (use mod $8$, as has been suggested in the comments. $2k^3\equiv \{0,2,6\}\pmod{8}$, but then $m^2\equiv \{5,7,3\}\pmod{8}$, contradiction, because $5,7,3$ are not quadratic residues mod $8$).

If $x$ is odd, then: let $p$ be a prime divisor of either $x-2$ or $x^2+2x+4$. Then $p$ is odd and $p\mid y^2+4$, so $\left(y2^{-1}\right)^2\equiv -1\pmod{p}$, so $p=4t+1$ (by Quadratic Reciprocity).

Therefore all the prime divisors of $x-2$, $x^2+2x+4$ are of the form $4t+1$. Also $x^2+2x+4=(x+1)^2+3>0$ and $y^2+4>0$, so $x-2>0$, so $x-2\equiv 1\pmod{4}$ and $x^2+2x+4\equiv 1\pmod{4}$. The first congruence gives $x\equiv 3\pmod{4}$, but $3^2+2\cdot 3+4\not\equiv 1\pmod{4}$, contradiction.

  • $\begingroup$ I can't understand what you mean by $y2^{-1}$, it seems there is a typo (and I'm not able to correct it). $\endgroup$ – Tom-Tom Dec 11 '15 at 12:59
  • $\begingroup$ @Tom-Tom $2^{-1}$ denotes the modular inverse of $2$ mod $p$. See Wikipedia. $\endgroup$ – user236182 Dec 11 '15 at 13:00
  • $\begingroup$ That makes sense, then. Thanks. $\endgroup$ – Tom-Tom Dec 11 '15 at 13:03
  • $\begingroup$ If x-2 only has prime divisors of p = 1 mod 4, then x-2 = 1 mod 4? Thats good to know! This aint true for other numbers right? Like if we only have prime divisors p = 3 mod 4, $\endgroup$ – S. Ferwerda Dec 11 '15 at 15:19
  • $\begingroup$ @S.Ferwerda We also need to know $x-2>0$ (because e.g. the only prime divisors of $-5$ are of the form $4k+1$, but $-5\not\equiv 1\pmod{4}$). Also, if all the prime divisors of $m>0$ are of the form $4k+3$, then we could either have $m\equiv 1$ or $3\pmod{4}$, e.g. $3^2\equiv 1\pmod{4}$ and $3^3\equiv 3\pmod{4}$. $\endgroup$ – user236182 Dec 11 '15 at 15:26

A general problem considering the equations of the form $$ y^2 = x^3 + k$$ are known as Bachet equations. There is a known theorem as follows (which could be found in Richard Mollin's Algebraic Number Theory: theorem 4.2) :

Let $F=\mathbb{Q}(\sqrt{k})$ be a complex quadratic field with radicand $k< -1$ such that $k \neq 1 \pmod 4,$ and $h_{\mathfrak{D}_F} \neq 0\pmod 3.$ Then there are no solutions to the Batchet equation in integers $x,y$ except in the following cases: there exists an integer $u$ such that $$(k,x,y) = (\pm 1-3u^2, 4u^2 \mp 1, \epsilon \cdot u(3 \mp 8u^2) ),$$

where the $\pm$ signs correspond to the $\mp$ signs and $\epsilon = \pm 1$ is allowed in either case.

In this case, $k=-12<-1$ and $k\equiv0\pmod 4$, $h_{\mathfrak{D}_F}=1$, however, $12\pm 1\neq0\pmod 3$, hence there is no integer solutions.

  • $\begingroup$ Also called Mordell's equations. $\endgroup$ – user236182 Dec 11 '15 at 12:52
  • $\begingroup$ @user236182 Yes. One could check A054504 and A081121 to see for which $k$ has no integral solutions. $\endgroup$ – user297600 Dec 11 '15 at 13:11
  • $\begingroup$ $17$ years ago Mordell's equation was fully solved by Gebel for all $0\le |k|\le 10^4$. OEIS has the amounts of solutions for each of those $k$. Once you know the amount, you can found all the solutions using a program. $\endgroup$ – user236182 Dec 11 '15 at 13:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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