15
$\begingroup$

I'm trying to get through a proof of Gauss' that certain primes can be written as the sum of two squares. An assumption is that

the order of $\mathbb{Z}[i]/(a+bi)$ is $a^2+b^2$.

I get that $(a+bi)(a-bi)=a^2+b^2$, so this places a bound on the order of integers with no imaginary part. But since $b$ isn't a unit, it doesn't seem like this finishes the proof.

Any hints?

$\endgroup$
5
  • 6
    $\begingroup$ Do you know what the ideal $(a + bi)$ looks like in the complex plane? $\endgroup$ Jul 20, 2011 at 1:10
  • 7
    $\begingroup$ Related: math.stackexchange.com/questions/23358/… . See math.stackexchange.com/questions/23358/… $\endgroup$
    – lhf
    Jul 20, 2011 at 1:19
  • $\begingroup$ Out of curiosity: how does this proof use the order of these quotients? The proof I'm thinking of doesn't need that at all. $\endgroup$ Jul 20, 2011 at 1:21
  • $\begingroup$ @Dylan: It is used to show that $\mathbb{Z}[i]$ is a Euclidean domain. $\endgroup$
    – Xodarap
    Jul 20, 2011 at 2:50
  • 1
    $\begingroup$ @Xodarap Interesting. Here's what I had in mind: to divide $a$ by $b \neq 0$ I need $q, r \in \mathbf{Z}[i]$ such that $a = qb + r$ and $|r| < |b|$. Then it's enough to find a $q$ such that $|a/b - q| < 1$. But the farthest any point in the complex plane can be from the lattice $\mathbf{Z}[i]$ is $1/\sqrt{2}$. $\endgroup$ Jul 20, 2011 at 3:10

3 Answers 3

13
$\begingroup$

HINT $\ $ If $\rm\ (a,b) = 1\ $ then $\rm\ x^2 = -1,\ a = b\:x\ \iff\ a^2 = -b^2,\ x = a/b\:.\ $ Therefore

$$\rm\mathbb Z[i]/(a-b\:i)\ \cong\ \mathbb Z[x]/(x^2+1,\ a-b\:x)\ \cong\ \mathbb Z[x]/(a^2+b^2,x-a/b)\ \cong\ \mathbb Z/(a^2+b^2) $$

$\endgroup$
4
  • 1
    $\begingroup$ I tried to prove your last isomorphism by the first isomorphism theorem, but I have difficulty showing surjectivity since I don't really know the elements of $\mathbb Z[x]/(a^2+b^2,x-a/b)$ - it doesn't seem like $\mathbb Z[x]/(x-a/b)\cong\mathbb Z$, for example. How can I attack this? $\endgroup$
    – Xodarap
    Jul 20, 2011 at 3:10
  • 2
    $\begingroup$ HINT $\ $ It's simply $\rm\:\mathbb Z[x]/(n,x-m)\ \cong\ \mathbb Z/n\ $ since $\rm\ (a,b)=1\ \Rightarrow\ (a^2+b^2,b)\ =\ (a^2,b)\ =\ 1\ \ $ hence $\rm\ 1/b\ $ exists $\rm\:(mod\ a^2+b^2)\:.\:$ So put $\rm\ n = a^2+b^2,\ \ a/b\equiv m\ (mod\ n)\:.$ $\endgroup$ Jul 20, 2011 at 3:17
  • 1
    $\begingroup$ It might be better to write this as two successive quotients, as $x - a/b$ isn't really an element of $\mathbf{Z}[x]$. Great answer! $\endgroup$ Jul 20, 2011 at 3:21
  • $\begingroup$ @Dylan $\rm\:a/b\:$ denotes an integer $\rm\:m\equiv a/b\ (mod\ n)\:$ so it is correct as written - see my prior comment. $\endgroup$ Jul 20, 2011 at 3:26
5
$\begingroup$

Expanding on Qiaochu's comment, but trying to leave you some room to think:

$\mathbb{Z}[i]$ is a square lattice in the complex plane, and the ideal $(a+bi)$ is a square sublattice. (Why?) The index of $(a+bi)$ in $\mathbb{Z}[i]$ is the number of elements of $\mathbb{Z}[i]$ inside a fundamental square of $(a+bi)$, including part but not all of its boundary. (Why? And what's with the boundary funny business?) This is the same as the ratio of the area of the fundamental square of $\mathbb{Z}[i]$ to the area of the fundamental square of the sublattice $(a+bi)$. (Why?)

$\mathbb{Z}[i]$'s fundamental square is a unit square, so the question is really about the area of $(a+bi)$'s fundamental square. The line from $0$ to $a+bi$ is a side of this square...

$\endgroup$
3
  • 1
    $\begingroup$ I recall you mentioning Artin in another question (math.stackexchange.com/questions/50757/…). Chapter 11 has a lot of stuff relevant to this question: section 2 has a diagram showing what $(2+i)$ looks like inside $\mathbb{Z}[i]$, as well as an expanded version of the proof that $\mathbb{Z}[i]$ is a Euclidean domain mentioned by Dylan Moreland; section 5 works through which primes are sums of squares; and section 10 relates the index of a sublattice to areas of fundamental regions as I have here (though he also leaves the proof as an exercise). $\endgroup$ Jul 20, 2011 at 15:39
  • $\begingroup$ Thanks Ben! Artin's geometric proof on pages 397-398 (similar to what you have here) is much easier to understand (for me) than the abstract nonsense argument. I'm still working through a couple things, but hopefully this will do it... $\endgroup$
    – Xodarap
    Jul 20, 2011 at 17:26
  • $\begingroup$ I had a question that was too long for a comment - could you take a look if you get a chance? $\endgroup$
    – Xodarap
    Jul 21, 2011 at 14:16
1
$\begingroup$

Hint: The Gaussian integers are a Euclidean domain with respect to the norm. Hence given $z\in \mathbb Z[i]$ there are $q$ and $r$ in $\mathbb Z[i]$ such that $z=q(a+bi)+r$, with $r=0$ or $N(r)<N(a+bi)$.

$\endgroup$
1
  • $\begingroup$ I actually wanted a proof of this question so that I could prove $N(a+bi)=a^2+b^2$ made the domain Euclidean :-) $\endgroup$
    – Xodarap
    Jul 20, 2011 at 2:47

Not the answer you're looking for? Browse other questions tagged or ask your own question.