In trying to solve a certain [third-degree] Diophantine equation, I have used the quadratic equation to determine that $$c^4-72b^2c^2+320b^3c-432b^4$$ must be a positive integer square, where $c$ and $b$ are positive integers.

From [admittedly limited] maxima calculations, it appears that the complete solution set is $(c,b) = (27k,5k)$ for $k = 1,2,3,\dots$.

I've tried to prove this, but have had no luck so far. Any hints or links/references to useful papers and techniques would be greatly appreciated.

EDIT: The original equation I was trying to solve is $$y^2 = x^3 - 2. \tag{$\star$}$$ I am aware of the solution to this equation (see, for example, Conrad's paper, and about a dozen questions on this site), and so am not looking for a direct solution to ($\star$). Rather, I am interested in discovering techniques — especially elementary ones — by which one may attack the [type of] equation in the question.

  • 1
    $\begingroup$ For what reason do you withhold the full problem, and only give us this derived problem? People might be able to help with the original question... $\endgroup$ – Thomas Andrews Jun 11 '15 at 13:37
  • $\begingroup$ @ThomasAndrews: The original problem is a Mordell equation, for which there are known solutions. Ultimately, I am trying to find general techniques, rather than particular solutions, so I didn't want to bias the answer(s) I received. $\endgroup$ – Kieren MacMillan Jun 12 '15 at 0:40
  • $\begingroup$ But, for instance, the answer below took your equation and argues that you can convert it to an elliptic curve, but a Mordell equation is already an elliptic curve. So, by not giving the original equation, you've actually made it harder for people to help you. $\endgroup$ – Thomas Andrews Jun 12 '15 at 0:47
  • $\begingroup$ Yes, that's one answer. Hopefully there will be others. In the meantime, I have edited my original question by adding the Mordell equation I am studying. $\endgroup$ – Kieren MacMillan Jun 12 '15 at 1:41
  • 1
    $\begingroup$ @KierenMacMillan: I gave a elementary method which you can easily apply to other situations when you need to solve a quartic $F(x) = y^2$. $\endgroup$ – Tito Piezas III Jun 12 '15 at 15:16

In general, an equation,

$$F(x) = y^2$$

with an initial rational point $x$ and where $F(x)$ is a quartic polynomial can be transformed into an elliptic curve, hence there should be an infinite number of rational points $x$. It is the case for your particular curve,

$$x^4 - 72 x^2 + 320x - 432 = y^2$$


$$x_1 = c/b = 27/5$$

$$x_2 = c/b = 649193045949/48488608130$$

where $x_2$ is not reducible to $x_1$ and so on, ad infinitum.

(P.S. I've probably missed a smaller solution.)

$\color{blue}{Edit:}$ Here's an elementary method how to find more solutions $x_i$.

(I believe it is called the tangent method and goes back to Fermat.) Simply equate,

$$x^4 - 72 x^2 + 320x - 432 =(a_2x^2+a_1x+a_0)^2$$

collect powers of $x$ to get,

$$p_4x^4+p_3x^3+p_2x^2+p_1x+p_0 = 0\tag1$$

and for this particular quartic, your three unknowns $a_0,a_1,a_2$ are enough to set $p_4 = p_3 = p_2 = 0$ and making the $x^4, x^3, x^2$ vanish, leaving you with,

$$x = -p_0/p_1 = 27/5$$

To get more solutions, do the substitution $x = v+27/5$, to get,

$$v^4+108/5v^3+2574v^2/25+21532/125v+(171/25)^2 = (b_2v^2+b_1v+b_0)^2$$

which you assume to be a square in the RHS. Collect powers of $v$,

$$q_4v^4+q_3v^3+q_2v^2+q_1v+q_0 = 0\tag2$$

Your three unknowns $b_0,b_1,b_2$ are also enough to set the other end $q_0 = q_1 = q_2 = 0$ and making the $v^0, v^1, v^2$ vanish, leaving you with,

$$q_4v^4+q_3v^3 = (q_4v+q_3)v^3 = 0$$


$$x_2 = v+x_1 = -q_3/q_4+27/5 = 649193045949/48488608130$$

I'm sure with this easy method, you now have a technique to generate new solutions from old ones.

  • 1
    $\begingroup$ Ah yes, there is also $x = c/b = 5383/855$. $\endgroup$ – Tito Piezas III Jun 11 '15 at 14:49
  • $\begingroup$ Thanks for the lovely technique. I look forward to investigating it further. $\endgroup$ – Kieren MacMillan Jun 12 '15 at 15:36
  • $\begingroup$ Can this method be used to find all solutions? e.g., how does one find $c/b=5383/855$ in the first place? $\endgroup$ – Kieren MacMillan Jun 12 '15 at 20:50
  • 1
    $\begingroup$ @KierenMacMillan: You solve for the other end of $(2)$, $q_4 = q_3 = q_2 = 0$ hence $$x_3 = v +x_1 = -q_0/q_1+27/5 = 5383/855$$ I'm not sure if it can find all solutions but, for appropriate quartics, it does give an infinite number of them. $\endgroup$ – Tito Piezas III Jun 13 '15 at 1:15
  • 1
    $\begingroup$ See also this post $\endgroup$ – Tito Piezas III Mar 23 '16 at 2:14

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.