2

The following is a proposed solution to a fairly routine problem (which has in fact been asked on stackexchange Quotient ring of Gaussian integers, but not as a solution verification question), which I am having severe difficulty understanding. As there are many questions, partial answers (e.g. answering only questions 1, 5 for example) are very welcome! I tend to get confused easily when answers are compact, please consider spelling things out in full glory please!

Problem A: Let $a$ and $b$ be relatively prime integers, and $i^{2}=-1 .$ Show that the natural map $\mathbb{Z} \rightarrow \mathbb{Z}[i] /\langle a+b i\rangle$ induces an isomorphism of $\mathbb{Z}/\left(a^{2}+b^{2}\right) \mathbb{Z}$ with $\mathbb{Z}[i] /\langle a+b i\rangle$.

Proof: First, note that the map $\mathbb{Z}[X] \rightarrow \mathbb{Z}[i]$ taking $X$ to $i$ is surjective, and its kernel is the ideal generated by $X^{2}+1$.

$\textbf{Question 1:}$ I think that this is "obvious", I mean the two rings "look" the same! But how would I rigorously argue that the kernel is precisely the ideal generated by $X^{2}+1$?)

Therefore, $\mathbb{Z}[i]$ is isomorphic to $\mathbb{Z}[X] /\left\langle X^{2}+1\right\rangle,$ and $\mathbb{Z}[i] /\langle a+b i\rangle$ is isomorphic to $\mathbb{Z}[X] /\left\langle X^{2}+1, a+b X\right\rangle$. We therefore work with the latter.

$\textbf{Question 2:}$ Why is this? How can we take quotients simultaneously like that? And why is the $i$ transforming into the $X$; can anyone spell this out rigorously/pedantically please?)

Note that $a^{2}+b^{2}$ is in the kernel of the map from $\mathbb{Z}$ to $\mathbb{Z}[X] /\left\langle X^{2}+\right.$ $1, a+b X\rangle,$ as $a^{2}+b^{2}=(a+b X)(a-b X)+b^{2}\left(X^{2}+1\right)$. We need to show that the map is surjective, and that its kernel is exactly $a^{2}+b^{2}$.

For surjectivity, since $\mathbb{Z}[X] /\left\langle X^{2}+1\right\rangle$ is spanned by $1$ and $X$ it suffices to show that $1$ and $X$ are in the image of $\mathbb{Z} \rightarrow \mathbb{Z}[X] /\left\langle X^{2}+1, a+b X\right\rangle$.

$\textbf{Question 3:}$ Why does it suffice? I get that $\mathbb{Z}[X] /\left\langle X^{2}+1, a+b X\right\rangle$ is also spanned by $1,X$. But the natural map $\mathbb{Z} \rightarrow \mathbb{Z}[X] /\left\langle X^{2}+1, a+b X\right\rangle$ can only map to a constant term of degree $0$. So how do we "span using $X$" to get the desired conclusion?)

It is clear that $1$ is in the image. For $X,$ note that since $a$ and $b$ are relatively prime, so are $b$ and $a^{2}+b^{2}$. Write $1=m b+n\left(a^{2}+b^{2}\right)$. Then $X+m a = m(a+b X)+n(a+b X)(a-b X)+n b^{2}\left(X^{2}+1\right),$

$\textbf{Question 4:}$ How in the world does one come up with this identity)

and the right hand side is clearly contained in $\left\langle a+b X, X^{2}+1\right\rangle .$ Thus $X$ is congruent to $-m a$ modulo $\left\langle a+b X, X^{2}+1\right\rangle,$ so $X$ is in the image of $\mathbb{Z} \rightarrow \mathbb{Z}[X] /\left\langle X^{2}+1, a+b X\right\rangle$.

We thus have a surjective map of $\mathbb{Z} /\left(a^{2}+b^{2}\right) \mathbb{Z}$ onto $\mathbb{Z}[X] /\left\langle X^{2}+1, a+bX\right\rangle$ so it suffices to show that it is injective. We do this by showing that there are at least $a^{2}+b^{2}$ distinct elements of $\mathbb{Z}[X] /\left\langle X^{2}+1, a+bX\right\rangle .$

$\textbf{Question 5:}$ I have no idea what the next few lines of the proof means. In particular, I do not know what "free abelian group" of "rank two" is, and the argument using parallograms entirely goes over my head.)

Note that $\mathbb{Z}[X] /\left\langle X^{2}+1\right\rangle$ is a free abelian group of rank two, spanned by 1 and $X$ and the ideal inside this generated by $\langle a+bX\rangle$ is spanned by $a+bX$ and $X(a +bX)=b-aX .$ The number of elements in the quotient of these two abelian groups is equal to the area of the parallogram spanned by $(a, b)$ and $(b,-a),$ which is $a^{2}+b^{2}$.

Albert
  • 1,557
  • 1
    See here for a simple proof, and see here for motivation and generalizatons. – Bill Dubuque Nov 08 '20 at 12:27
  • 1
    Thanks. Those proofs are similar to other 'cleverer' proofs that I know of the above problem. I'm trying to understand the presented proof above, which seems to be a 'categorical method' of solving these problems. – Albert Nov 08 '20 at 13:26
  • 1
    Some hints:

    For question 1, first show that the ring homomorphism $\mathbb{Z}\left[X\right] \to \mathbb{Z}\left[i\right]$ sending $X$ to $i$ factors through $\mathbb{Z}\left[X\right] / \left(X^2+1\right)$; then show that the resulting map is injective (since $\mathbb{Z}\left[X\right] / \left(X^2+1\right)$ has a basis consisting of the projections of $1$ and $X$, and this basis gets sent to a linearly independent family).

    – darij grinberg Nov 08 '20 at 23:46
  • 1
    For question 2, what is being used here is that if $f : A \to B$ is a surjective ring homomorphism, and if $J$ is an ideal of $B$, then $A / f^{-1}\left(J\right) \cong B / J$. – darij grinberg Nov 08 '20 at 23:47
  • I don't understand question 3. The author does show below that $1$ and $X$ lie in the image. Since they span, this shows that the image is the entire target. – darij grinberg Nov 08 '20 at 23:49
  • 1
    For question 5, I consider the parallelogram argument unrigorous. Maybe there is an easy way to justify it, but it's not convincing to me as it stands. Sure, the area of a parallelogram can be approximated by the number of integer points, but if one wants to know for sure that they are equal, one needs some further arguments (I think it's possible to argue this using Ehrhart theory or asymptotics). It is much easier to ditch the picture and use the Smith normal form, but this is not likely to help you if you don't know about free modules. – darij grinberg Nov 08 '20 at 23:52
  • 1
    Question 4 is a good one. I see how one might come up with an identity like this (namely, you want to write $i+c$ as a product of $a+bi$ with another Gaussian integer), but when I do it I find the simpler identity that @MathGems found in https://math.stackexchange.com/a/373163/ , not the one you cited. – darij grinberg Nov 08 '20 at 23:55

0 Answers0