This is a follow up to a question from yesterday regarding the representation of integers by binary quadratic forms with integer coefficients. OEIS A031363 lists the positive integers of the form $x^2+xy-y^2$ which turn out to be the same as those of the form $5x^2-y^2.$ Yesterday's question was how to show the two sets are the same, but I notice that on the OEIS page, it says, under "Formula:
Consists exactly of numbers in which primes == 2 or 3 mod 5 occur with even exponents.
I've been wondering what it takes to prove this, in particular whether it can be shown by elementary methods, since I know no algebraic number theory at all.
I think it's likely to be much easier to prove that numbers not of this form can't be represented than to prove the converse, so I'm starting with that.
For $p=2$ it says, no $n \equiv 2 \pmod 4$ can be so represented, and it's trivial to check that $5x^2-y^2$ cannot be of this form, since the only squares $\pmod 4$ are $0 \text{ and }1.$
For a general attack, I'm thinking about trying to show that $n$ is representable if and only if each of its prime factors is representable. If true, this would reduce the problem to representability of primes. I have some confidence in the "if" part (at least I know that the sum of two squares times the sum of two squares is again the sum of two squares) but I have no feeling for the "only if" part at all. Since $k^2n$ is representable if and only if $n$ is representable (from the form $5x^2-y^2),$ and $1$ is we don't need to consider powers.
This isn't something I need for school or work. I'm retired and sometimes do math for recreation; I want to know if this is a suitable problem for me to work on, that is, one that offers a reasonable chance of success, or at least progress. I'm not asking for a proof; I just want to have some idea of the level of difficulty of the problem.