# If $p\equiv1\mod{4}$, then $p=a^{2}+b^{2}$ where a and b are uniquely determined

From Ireland,Rosen's Number Theory book, ex.8.12.

Let $p\equiv1\mod{4}$, then we can write $p=a^{2}+b^{2}$, where $a,b\in\mathbb{Z}$. Now, let $a$ be odd and $b$ be even and $a,b>0$. How can we show that $a$ and $b$ are uniquely determined? The hint says to use the fact that $\mathbb{Z[i]}$ is a UFD and $p=a^{2}+b^{2}\implies a+bi$ is prime in $\mathbb{Z[i]}$

-
Assume they aren't unique (up to sign change), and show that $p$ has two nonequivalent factorizations in $\mathbb{Z}[i]$ (impossible because it's a UFD). – anon Feb 23 '12 at 5:32
Remember, also, that the units in $\mathbf Z[i]$ are just $\{\pm1, \pm i\}$. – Dylan Moreland Feb 23 '12 at 5:35

We have a natural norm $N$ on $\mathbb{Z}[i]$ given by $N(a+bi) = a^2 + b^2$, for all $a+bi \in \mathbb{Z}[x]$. One can show that w.r.t. this norm, $\mathbb{Z}[i]$ is a Euclidean domain, hence a PID, hence a UFD.

Let $p \in \mathbb{N}$ be a prime such that $p = a^2 + b^2$ for some $a,b \in \mathbb{Z}$. Now suppose $\exists c,d \in \mathbb{Z}$ such that $p = c^2 + d^2$. Then in $\mathbb{Z}[i]$ we have $N(a+bi) = p = N(c+di)$.

Suppose $a+bi = (e+fi)(e' +f'i)$ in $\mathbb{Z}[i]$. Then taking norms (our norm $N$ is multiplicative) we have $a^2+b^2 = p = (e^2+f^2)(e'^2 + f'^2)$, and so without loss of generality one can say that $e^2 + f^2 = p$. Then $e'^2 + f'^2 = 1$, and the only elements with norm 1 in $\mathbb{Z}[i]$ are $\{1,-1,i,-i\}$, which are also units. Thus, $e'+f'i$ is a unit in $\mathbb{Z}[i]$, which shows that $a+bi$ is irreducible (actually one needs to use the fact that every unit of $\mathbb{Z}[i]$ has norm 1 to conclude that $a+bi$ is not a unit in the first place, but this is easy). Since, $N(c+bi)=p$, by a similar argument $c+di$ is irreducible in $\mathbb{Z}[i]$.

Then in $Z[i]$ we have $(a+bi)(a-bi) = p = (c+di)(c-di)$ is a factorization of $p$ into irreducibles. By unique factorization, say without loss of generality that $c+di = unit \times a+bi$ (here the unit is a unit of $\mathbb{Z}[i]$). Then you are basically done (why?).

-
Oops, Arturo must have posted an answer while I was typing. – Rankeya Feb 23 '12 at 5:55
Welcome to the club. – Gerry Myerson Feb 23 '12 at 6:25

If $p=a^2+b^2$, then $a+bi$ is a prime in $\mathbb{Z}[i]$. If we also have $p=c^2+d^2 = (c+id)(c-id)$, then $a+bi$ divides $(c+id)(c-id)$, hence $a+ib$ divides either $c+id$ or $c-id$. But, symmetrically, $c+id$ and $c-id$ are both prime, hence irreducible, so their only divisors are units and associates. Since $a+ib$ is not a unit, then it is an associate of, say, $c+id$. Hence either $a+ib = c+id$, $a+ib = -(c+id)$, $a+ib = i(c+id)$, or $a+ib = -i(c+id)$. Either way, $\{c^2,d^2\}=\{a^2,b^2\}$, so the two expressions of $p$ as sums of squares are actually identical (up to order of the summands).

-