0

Is there a way to obtain (enumerate) the integer solutions $(x,y,z)$ of the following quadratic Diophantine equation

$z^2=a^2+bx^2+cy^2$

where $a$ is an integer and $b, c$ are positive integers? I have checked the literature on Diophantine equations but could not find anything useful. Any help or suggestions would be greatly appreciated.

Bart Michels
  • 26,355
Daniel
  • 1
  • Did you see Legendre's theorem – Elaqqad Apr 09 '15 at 13:44
  • Of course you can solve it. The only cumbersome formula. And decisions are determined by Pell equations. For example for some simple cases, you can write. http://math.stackexchange.com/questions/351491/integral-solutions-of-hyperboloid-x2y2-z2-1/709219#709219 http://math.stackexchange.com/questions/298053/quadratic-diophantine-equations/710766#710766 http://math.stackexchange.com/questions/74931/integral-solutions-of-x2y21-z2/789972#789972 – individ Apr 09 '15 at 14:23

2 Answers2

0

I think you can at first solve the well known equation $w^{2} + a^{2} = z^{2}$ for convenient values of a and go after to the equation $w^2 = bx^2 + cy^2$ which is solved in the book of L. J. Mordell “Diophantine Equations” (Academic Press. London and New York, 1969 page 44).

Itoz Darbien
  • 466
  • 16
Piquito
  • 29,594
0

We have at least two, $(0,0,a)$ and $(0,0,-a).$

We have infinitely many solutions to $$ b x^2 + c y^2 - z^2 = -a^2. $$

It is possible, however, that no solutions need be primitive: under some choices of $a$ it may be that all solutions have $x,y,z$ all divisible by $a.$ An early example of this (not quite the pattern your form takes) is due to C. L. Siegel, see pages 168 and 253 in CASSELS.

Actually, an easy example is given by $$ 3 x^2 + 3 y^2 - z^2 = -9. $$ We see immediately that $z^2$ is divisible by $3,$ therefore $z$ is divisible by $3.$ As a result, $3x^2 + 3 y^2$ is divisible by $9,$ so that $x^2 + y^2$ is divisible by $3.$ Since Legendre symbol $(-1|3) = -1,$ we see that $x,y$ are also both divisible by $3.$

Find a solution ( the smallest one with $u,v > 0$ is best) to $u^2 - b v^2 = 1.$ We can now replace any of the solutions we have found so far with $$(x,y,z) \mapsto (ux+vz,y, bvx+uz). $$

Find a solution ( the smallest one with $p,q > 0$ is best) to $p^2 - c q^2 = 1.$ We can replace any of the solutions we have found so far with $$(x,y,z) \mapsto (x,py+qz, cqy+pz). $$

Will Jagy
  • 139,541