I have come across a problem in which it is to be determined for what values of $x$ the expression $x^2+9y^2 \pm y$ is a perfect square, i.e. $x^2+9y^2 \pm y=z^2$, for $x,y,z \in \mathbb N$
There are obvious trivial solutions, when $y=0$ and $y=x^2$, so I am interested in solutions where $0<y<x^2$. Looking at small numbers by trial and error calculation, there are some values of $x$ that give rise to no solutions, such as $x=5,10,12$. But many other values of $x$ give rise to one or more solutions, including $(x,y,z)=(4,3,10),(8,9,28),(20,23,72),(20,36,110),(20,57,172)$, etc.
I have included the tag Pythagorean triples because when $9y^2$ is viewed as $(3y)^2$ the equation resembles a variation of the Pythagorean equation, and some students of that relationship may have insights. I am unaware of any tools to analyze the equation of interest, $x^2+9y^2 \pm y=z^2$ with regard to what values of $x$ do or do not give rise to solutions.
My questions are: Are there methods to decide what values of $x$ will give rise to solutions? And, more specifically, is there a value of $x$, call it $x_m$, such that for all $x>x_m$, solutions can be found?
I am extremely embarrassed to have posted essentially the same question twice (see link in comments). The reason I find this question to be so important (and I arrived at it twice in my pursuits) is because the statement "There are infinitely many values of $x$ for which the expression $x^2+9y^2 \pm y$ cannot be a perfect square" (for $y$ within the range specified) is equivalent to the statement "The twin prime conjecture is true." The demonstration of that claim is lengthy and goes far afield beyond the specific question posed, but if proof thereof is demanded, I can append it in a further addition to this question.
To big for a comment
For $\quad x^2+9y^2-y=z^2\quad$ the pattern appears to be $\quad (x,y,z)=(x,x^2,3x^2).\quad$
This means there is no case where $\,y<x^2.\quad$ For example, a spreadsheet reveals these values
$$(x,y,z)\in\big\{(1,1,3),(2,4,12),(3,9,27), (4,16,48),(5,25,75),(7,49,147),\cdots\big\}$$
For $\quad x^2+9y^2+y=z^2\quad$ the nearest thing to a pattern, given limited examples, appears to be that $\,z=3y+c\,$ where $\,c\in\mathbb{N}$
For example:
$$(x,y,z)\in\big\{ (4,3,10),(6,7,22),(9,7,23),(9,16,49), (11,24,73),(13,15,47),(14,11,36), (14,39,118),(16,51,154),(19,15,49), (19,72,217),(20,3,22),(20,23,72), (20,36,110),(21,88,265),(24,7,32), (24,19,62),(24,52,158),(24,115,346) \big\}$$
These are in ascending order of $\,x.\quad$ Perhaps you can find a pattern among these values that I did not.