2

I want to find/describe all the ideals in the quotient ring $A:=\mathbb{Z}[x]/(x^2-1)$. I know from the correspondence theorem that the ideals are the images under quotient map of ideals of $\mathbb{Z}[x]$ that contain $(x^2-1)$. Since $x^2-1\in (x\pm 1)$ we can say that these are two ideals which contain this. There are of course more ideals like $(2,x\pm 1)$ ... . I have little idea on how to proceed with such computations. Any help will be appreciated.

Thanks.

hm2020
  • 10,015
tassle
  • 456
  • Yeah, its $x^2-1$. There are 4 units in the quotient ring right, $\pm1,\pm x$? but $\mathbb{Z}$ has only 2 units, so I dont think they are isomorphic ... – tassle Sep 25 '21 at 02:18
  • 1
    At least we know what not to do, see this post, or this one. We cannot apply the CRT first and then look for ideals. – Dietrich Burde Sep 25 '21 at 08:33
  • @tassle - below I construct all prime ideals in $A$ and products of these give a large class of non-trivial ideals. There is also a relation with ideals in $\mathbb{Z}[\frac{1}{2}]\oplus \mathbb{Z}[\frac{1}{2}]$ - maybe this can be helpful. – hm2020 Sep 30 '21 at 15:35
  • @tassle - you may also realize $A$ as a subring of the direct sum $\mathbb{Z}\oplus \mathbb{Z}$ - this gives another classification of all ideals. $A \cong\mathbb{Z}{u,v}$ is a free $\mathbb{Z}$-module of rank 2 on the generators $u:=(1,1),v:=(1,-1)$. – hm2020 Oct 01 '21 at 09:01
  • 2
    See this mother thread and a child for a description of prime ideals of $\Bbb{Z}[x]$. The preimage of a prime ideal of $A$ is one of those by the correspondence principle. – Jyrki Lahtonen Oct 02 '21 at 11:48

1 Answers1

3

By the correspondence theorem, it suffices to analyze ideals $I$ of $\Bbb Z[x]$ containing $(x^2-1)$. Write such an ideal as $I=(x^2-1,f_1,\cdots,f_m)$ (if there are no $f_i$, then $I=(x^2-1)$). Then up to subtracting multiples of $x^2-1$ from the $f_i$ and flipping signs if necessary, we may assume that all the $f_i$ are either constant or linear with positive leading coefficient. Now we can break in to two cases: write $I\cap\Bbb Z=(n)$ for $n\geq 0$, and either $n=0$ or $n>0$.

When $n=0$, then it must be the case that all the $f_i$ are either all multiples of $x-1$ or all multiples of $x+1$: if there is some linear form $ax+b$ with $|a|\neq |b|$, then $(ax+b)(ax-b)=a^2(x^2-1)+(a^2-b^2)$ gives rise to a nonzero element of $I\cap\Bbb Z$, in contradiction to our assumption. By taking the smallest positive multiple of $x\pm1$, we can write $I=(x^2-1,ax\pm a)$ for some $a\geq 0$. When $a=1$, we get $(x-1)$ and $(x+1)$.

When $n\neq 0$, things get a little more interesting. If all the coefficients of the linear $f_i$ are multiples of $n$, then $I=(x^2-1,n)$. When this doesn't occur, then we may write $I=(x^2-1,ax+b,n)$ for $a,n>0$ and $a\mid n$. But not every choice of $a,b,n$ satisfies $(x^2-1,ax+b,n)\cap\Bbb Z=(n)$: given any $p(ax+b)+qn\in I$ and any $cx+d\in \Bbb Z[x]$, we need to make sure that if $(p(ax+b)+qn)(cx+d)$ has no $x$ term, then it has remainder divisible by $n$ after division by $x^2-1$. In order for this product to have no $x$ term, we must have $(c,d)=\frac1r(pa,-pb-qn)$ for some $r$ dividing both $pa$ and $pb+qn$, which gives us $$ \frac{p^2a^2}{r}(x^2-1) + \frac{p^2a^2-(pb+qn)^2}{r}. $$ So $n$ must divide $\frac{p^2a^2-(pb+qn)^2}{r}$ for any choice of $p,q,r$. It's possible that there's a nice simplification of this condition based on just $a$ and $b$, but I'm not seeing it yet and I figured it would be better to post this and get suggestions on how to finish the last bit rather than sit on it.

In summary: the ideals of $\Bbb Z[x]/(x^2-1)$ are of the form $0$; $(n)$ for $n\in\Bbb Z_{>0}$; $(ax\pm a)$ for $a>0$; and $(ax+b,n)$ with $a,n>0$, $a\mid n$, and subject to the above condition.

Hank Scorpio
  • 2,951
  • "When this doesn't occur, then we may write $I=(x^2-1,ax+b,n)$..." How did you get rid of the other possible generators of degree one? – user26857 Oct 02 '21 at 18:47
  • @user26857 the set of leading coefficients of linear generators forms a subgroup of $\Bbb Z$, so we can take the smallest positive such leading coefficient as our $a$. From there, any two $ax+b$ and $ax+b'$ differ by a multiple of $n$, so we can just pick any $ax+b$. – Hank Scorpio Oct 03 '21 at 01:21
  • I am not sure how a finite set of leading coefficients can form a subgroup, but I would interpret this as follows: one considers the subgroup of $\mathbb Z$ generated by the leading coefficients of all degree one polynomials belonging to $I$, and let $a$ be the positive generator of this subgroup. This leads to another condition: $a\mid b$ since $bx+a\in I$. – user26857 Oct 03 '21 at 08:04
  • Furthermore, the condition you are looking for looks too complicated to me. Why need $qn$? This already belongs to $I$. This condition seems to be the following: if $ad+bc=0$ then $n\mid ac+bd$, and this leads to $na\mid a^2-b^2$. I didn't check carefully, but I think this condition is also sufficient for $I\cap\mathbb Z=(n)$. – user26857 Oct 03 '21 at 08:04