Solving a homogeneous quadratic equation in three variables over the integers.

TLDR; If there exists one nonzero solution, then there exist infinitely many solutions, paramatrized by $\Bbb{Z}^2$. Popular candidates to test are triplets $(0,y,z)$ where $y\mid c$ and $z\mid a$, though such solutions need not exist.


Every integral solution to your equation with $x\neq0$ yields a rational solution to \begin{equation}aY^2+bYZ+cZ^2=1,\end{equation} by setting $Y:=\tfrac yx$ and $Z:=\tfrac zx$. Conversely, every rational solution to this equation yields an integral solution to your equation by multiplying out the denominators. This also shows that multiplying an integral solution through by an integer yields another integral solution.

Given a nonzero rational solution $(Y_0,Z_0)$ to your equation, also $$(aY_0+bZ_0,-aZ_0),$$ is a rational solution, and moreover for every $k\in\Bbb{Z}$ also $$\left((aY_0+bZ_0)k^2+2cZ_0k-cY_0,-aZ_0k^2+2cY_0k+bY_0+cZ_0\right),$$ is a rational solution. These are in fact all rational solutions. This then in turn yields all integral solutions, except those with $x=0$. For these we have $$Y=\frac{-b\pm\sqrt{b^2-4ac}}{2a}Z,$$ so such solutions exist if and only if $\frac{-b\pm\sqrt{b^2-4ac}}{2a}$ are rational, i.e. if and only if $ax^2+bx+c$ has a rational root, which is easy to test.


Given,

$$ay^2 + by z + c z^2 = x^2\tag1$$

for the particular case when $z = 1$, and $a$ is not a square, then YES, given an initial positive integer solution, one can find an infinite more positive ones by solving a Pell equation. For your example,

$$202y^2 + 14y z + 9 z^2 = x^2\tag2$$

the solution is,

$$x = 15 + 418 p q + 6060 q^2$$

$$y = 1 + 30 p q + 418 q^2$$

$$z = 1$$

and $p,q$ solve,

$$p^2-202q^2 = 1$$

This is just a special case of a more general solution.


Alright, if $$ x = 87 u^2 + 116uv + 59 v^2, $$ also $$ y = uv + 4 v^2, $$ and $$ z = 29 u^2 + 34 u v + 3 v^2, $$ THEN $$ x^2 = 202 y^2 + 14 y z + 9 z^2. $$

This comes from a method in Fricke and Klein (1897) which is not widely known. By choosing $u,v$ positive you can get all three $x,y,z$ positive. It is also not so simple to find this sort of recipe. I wrote a program, here was the command:

 ./homothety_indef -1 202 9 14 0 0    0  7076 0 0 -7076 0   119

the output lines I used:

 87    116     59 transposed       87      0     29
  0      6      4 transposed      116      6     34
 29     34      3 transposed       59      4      3



? x = 87 * u^2 + 116 * u * v + 59 * v^2
%21 = 87*u^2 + 116*v*u + 59*v^2
? y = 6 * u * v + 4 * v^2
%22 = 6*v*u + 4*v^2
? z = 29 * u^2 + 34 * u * v + 3 * v^2
%23 = 29*u^2 + 34*v*u + 3*v^2
? 
? 202 * y^2 + 14 * y * z + 9 * z^2
%24 = 7569*u^4 + 20184*v*u^3 + 23722*v^2*u^2 + 13688*v^3*u + 3481*v^4
? x^2
%25 = 7569*u^4 + 20184*v*u^3 + 23722*v^2*u^2 + 13688*v^3*u + 3481*v^4
? 202 * y^2 + 14 * y * z + 9 * z^2 - x^2
%26 = 0
?