5

Show that each integer of the form $a^2+b^2$ has all the factors of this form, where $(a, b)$ are distinct integers and relatively prime

Progress

If $a^2+b^2$ is prime then it is already proved, since every prime is a factor of its self and $1=0^2+1^2$, but I don't know how to prove for the rest of the case. Thanks!

user18724
  • 647
  • 1
    Your reasoning is erroneous: 3 is not a sum of squares ... – Nicky Hekster Aug 14 '14 at 20:38
  • $3$ is not also the sum of two distinct square numbers. $3^2+3^2$ cannot be accepted if that is the case you are talking about, since $(a, b)$ are distinct integers. @NickyHekster – user18724 Aug 14 '14 at 20:44
  • Do you mean prime factors in the title? Are you allowed to use Gaussian integers? – Conifold Aug 14 '14 at 20:50
  • Sorry, I don't know much about Gaussian integers but if they are natural positive integers then yes you're allowed to use them. Thanks. @Conifold – user18724 Aug 14 '14 at 20:57
  • I mean all the factor including prime factors @Conifold – user18724 Aug 14 '14 at 21:05
  • The question body should not be a continuation of the title; rather, it should be readable on its own. –  Aug 14 '14 at 21:47

1 Answers1

2

Assuming that you are talking about numbers of the form $n=a^2+b^2$ with $\gcd(a,b)=1$, then every odd prime $p$ that divides $n$ has the property: $$\left(\frac{-1}{p}\right)=1,\tag{1}$$ since $(1)$ is a consequence of $a^2+b^2\equiv 0\pmod{p}$. However, $(1)$ is equivalent to $$ p\equiv 1\pmod{4},\tag{2}$$ since the Legendre symbol $\left(\frac{-1}{p}\right)$ equals $(-1)^{\frac{p-1}{2}}$.

Now it is possible to prove that any prime $\equiv 1\pmod{4}$ is the sum of two coprime squares - see, for instance, this similar question in which I proved that every prime $p\equiv 1\pmod{3}$ is represented by the quadratic form $a^2+3b^2$ through Fermat's descent - and since the numbers represented by $a^2+b^2$ are a semigroup, due to the Lagrange's identity: $$(a^2+b^2)(c^2+d^2)=(ac-bd)^2+(ad+bc)^2\tag{3},$$ we have that every divisor of $n$ is represented by the quadratic form $a^2+b^2$. Notice that $(3)$ is just equivalent to the multiplicativity of the norm on the ring $\mathbb{Z}[i]$ (the gaussian integers).

Jack D'Aurizio
  • 353,855
  • Thanks for your proof, though I expected a rather simpler and elementary proof if there was one! @Jack – user18724 Aug 14 '14 at 21:12
  • All this argument can be condensated in a single statement, if you like: the gaussian integers are an euclidean domain, hence a UFD. – Jack D'Aurizio Aug 14 '14 at 21:17