1
$\begingroup$

Is there a way to solve this Diophantine equation in $a,b,c,d$? $$19a^3-33a^2b+3a^2c+30a^2d+21ab^2+24abc-12abd-15ac^2-54acd-30ad^2+ $$ $$2b^3-12b^2c-6b^2d+42bc^2+108bcd+60bd^2-7c^3-51c^2d-99cd^2-56d^3=0$$

Wolfram Alpha unfortunately cannot understand my input whenever I input this equation, and I know no strategy to solve these kinds of equations. So basically, I'm stuck...


I want just all possible values for $a,b,c,d$ just like what Wolfram Alpha does.

Note: $a\neq b\neq c\neq d\neq 0$

$\endgroup$
  • $\begingroup$ did you try to group by factorization and simplify the structure ? $\endgroup$ – G Cab Sep 29 '16 at 0:07
  • $\begingroup$ @G Cab Um... No? Is that possible? $\endgroup$ – Frank Sep 29 '16 at 0:17
  • 1
    $\begingroup$ what produced this polynomial? Where did you get it, exactly? $\endgroup$ – Will Jagy Sep 29 '16 at 0:21
  • $\begingroup$ oh, well. If it factors, it is a homogeneous linear times a quadratic form. If the quadratic form is definite, the set of solutions is an integral lattice. Just one possibility. $\endgroup$ – Will Jagy Sep 29 '16 at 0:26
  • 1
    $\begingroup$ It doesn't factor. $\endgroup$ – Robert Israel Sep 29 '16 at 0:27
2
$\begingroup$

Your equation is homogeneous of degree $3$, so if $(a,b,c,d)$ is a solution so is $(ta,tb,tc,td)$ for any $t$. Thus it makes sense to look for primitive solutions, which are solutions with greatest common divisor $1$. I found some $288$ primitive solutions with $a,b,c \in [-20\ldots -1, 1\ldots 20]$ and $d \in [1 \ldots 20]$. I don't see an obvious pattern. Here are some of those solutions:

$$ \matrix{ a & b & c & d\cr -4 & -1 & -8 & 1\cr -4 & 1 & -8 & 3\cr -3 & 1 & -7 & 3\cr -3 & 1 & 3 & 1\cr -3 & 5 & -1 & 4\cr -2 & 1 & -6 & 3\cr -2 & 1 & -4 & 2\cr -2 & 2 & -4 & 3\cr -2 & 3 & -4 & 4\cr -2 & 3 & 2 & 1\cr -2 & 4 & -4 & 5\cr -2 & 5 & -4 & 6\cr -1 & -1 & -5 & 4\cr -1 & 1 & -5 & 3\cr -1 & 2 & -3 & 3\cr -1 & 5 & 1 & 1\cr 1 & -2 & -1 & 2\cr 1 & -1 & -5 & 2\cr 1 & 1 & -3 & 3\cr 1 & 2 & -7 & 6\cr 2 & 1 & -2 & 3\cr 2 & 2 & 4 & 1\cr 2 & 3 & 4 & 2\cr 2 & 4 & 4 & 3\cr 2 & 5 & 4 & 4\cr 2 & 6 & 4 & 5\cr 3 & -4 & -7 & 1\cr 3 & 1 & -1 & 3\cr 3 & 2 & -5 & 6\cr 3 & 4 & -3 & 4\cr 4 & 3 & 8 & 1\cr 5 & 1 & 1 & 3\cr 5 & 2 & -3 & 6\cr 6 & 1 & 2 & 3\cr 7 & 1 & 3 & 3\cr 7 & 2 & -1 & 6\cr 8 & 1 & 4 & 3\cr }$$

Did those coefficients come from somewhere in particular or are they just arbitrary?

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Robert, could it be the determinant of a 3 by 3 matrix with entries linear forms in $a,b,c,d?$ Not sure $\endgroup$ – Will Jagy Sep 29 '16 at 0:30
  • $\begingroup$ I don't know. How would one check that? $\endgroup$ – Robert Israel Sep 29 '16 at 0:35
  • $\begingroup$ Good point; it is just a possible source, the OP has not said. I do wonder whether we can join two integer solutions by a straight line and find another. Just musing. $\endgroup$ – Will Jagy Sep 29 '16 at 0:37
  • $\begingroup$ Hey; to answer your question, yes. The coefficients do come from somewhere! $\endgroup$ – Frank Sep 29 '16 at 0:40
  • 2
    $\begingroup$ @RobertIsrael: The substitution $a=u+v,\;c=-u+v$ reduces the OP's cubic equation to just a quadratic in $v$. Its discriminant is a quartic polynomial to be made a square, hence an elliptic curve is involved as you pointed out. $\endgroup$ – Tito Piezas III Sep 29 '16 at 1:40
3
$\begingroup$

As the OP pointed out in the comments that the equation is equivalent to,

$$(a + 2c - 2b + 3d)^3 + (2a - b + 2c + 4d)^3 - (a + b + 2c + 2d)^3 - (3a - 2b+ c + 3d)^3 = 0$$

then by equating terms, its complete solution is,

$$\begin{aligned} a= -2 x_1 - 7 x_2 - 5 x_3 - 8 x_4\\ b=-6 x_1 + 5 x_2 - 2 x_3 + 2 x_4\\ c=9 x_1 - 14 x_2 - 10 x_3 - 3 x_4\\ d=-5 x_1 + 15 x_2 + 7 x_3 + 6 x_4 \end{aligned}$$

where,

$$x_1^3+x_2^3+x_3^3+x_4^3 = 0$$

The complete rational solution by Euler is well-known.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ You wouldn't mind if you proved your formula: $n^3+(3n^2+2n+1)^3+(3n^3+3n^2+2n)^3=(3n^3+3n^2+2n+1)^3$? I read that in a pdf document you wrote and I'm interested to see your proof... $\endgroup$ – Frank Sep 29 '16 at 20:42
  • $\begingroup$ I'm wondering where did you get the starting grouped-equivalent from: I do not see any in the post, nor I got any reply to my comment asking exactly about that. Do you have a priviliged channel? or is it my browser's problem? $\endgroup$ – G Cab Sep 29 '16 at 21:02
  • $\begingroup$ @GCab No, I don't think the equation can be factored into your form. $\endgroup$ – Frank Sep 29 '16 at 21:32
  • $\begingroup$ @Frank: That formula in $n$? With elementary identities, the most direct way is just to expand and compare both sides. $\endgroup$ – Tito Piezas III Sep 30 '16 at 1:40
  • $\begingroup$ @GCab: The OP remarked about the $x_1^3+x_2^3+x_3^3+x_4^3=0$ form in this comment. $\endgroup$ – Tito Piezas III Sep 30 '16 at 1:44

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.