4
$\begingroup$

I originally asked this on Stack Exchange but with no luck. Upon doing research with some noncommutative rings, I thought of a curious question. Does there exist a noncommutative unital ring $R$ and left ideal $I$ such that $R/I\cong R/\text{Ann}_R(R/I)$ as left $R$-modules but $I\neq \text{Ann}_R(R/I)$? I suspect that such an $R$ and $I$ do exist but are awkward to construct. Note that

$$\text{Ann}_R(R/I)=\{r\in R:\forall x\in R,\ rx\in I\}\subseteq I$$

and that $\text{Ann}_R(R/I)$ is always a two-sided ideal. As such, if $I$ is not two-sided then we already get $I\neq \text{Ann}_R(R/I)$. Any ideas would be appreciated.

My question is equivalent to this question with the additional constraint that $J$ must be two-sided.

$\endgroup$
2
  • $\begingroup$ How about taking the ring freely generated by two non-commuting variables $X$ and $Y$. Set $I=RX$. Then the annihilator is zero and the map $w\mapsto wY$ is an isomorphism form $R$ to $R/I$. $\endgroup$
    – user130903
    Sep 18, 2020 at 7:23
  • $\begingroup$ That map is not surjective is it? What would map to the coset $1+I$? I was thinking about using free rings as well, but couldn't think of a way to make it work. Maybe there's a way though. $\endgroup$
    – Anonymous
    Sep 18, 2020 at 8:26

1 Answer 1

4
$\begingroup$

You might as well assume that $\mathrm{Ann}_R(R/I)=0$ since if $S=R/\mathrm{Ann}_R(R/I)$ then $R/I\cong S$ as $R$-modules if and only if $S/(I/\mathrm{Ann}_R(R/I))\cong S$ as $S$-modules.

So now the question is whether you can have a ring $R$ and a non-zero left ideal $I$ such that $R/I\cong R$. Equivalently can there be $r\in R$ such that (i) $\mathrm{ann}_R(r)=I\neq 0$ and (ii) there is some $s\in R$ with $sr=1$; if $\varphi\colon R/I\to R$ is the isomorphism then take $r=\varphi(1+I)$ and $\varphi(s+I)=1$. To further rephrase we want an element $r$ of $R$ that is a right zero-divisor and a right unit.

This is indeed possible. Here is one example: let $R$ be the $\mathbb{C}$-linear endomorphisms of the polynomial ring $\mathbb{C}[X]$ and let $r$ be the element of $R$ given by multiplication by $X$. It is easy to construct a left inverse for $r$ (in fact one can easily find infinitely many). But if $x$ is the element of $R$ that sends a polynomial to its constant term viewed as the constant polynomial then $xr=0$.

The punchline If you just wanted to pull a rabbit out of a hat you could simply say that for $R=\mathrm{End}_{\mathbb{C}}(\mathbb{C}[X])$ and $r\in R$ given by $r(f)=Xf$, $I=\mathrm{ann}_R(r)\neq 0$ and there is an isomorphism of left $R$-modules $R/I\to R$ given by $x+I\mapsto xr$. It is surjective since there is $s\in R$ with $sr=1$ (Moreover $\mathrm{Ann}_R(R/I)=0$).

$\endgroup$
3
  • $\begingroup$ That is not an isomorphism. It is not injective. $\endgroup$
    – Anonymous
    Sep 18, 2020 at 9:58
  • 1
    $\begingroup$ Sorry. You are right. I think it is all correct except the punchline doesn't correctly reinterpret the previous discussion. I'll edit. $\endgroup$ Sep 18, 2020 at 10:10
  • $\begingroup$ Thanks! This is really useful! $\endgroup$
    – Anonymous
    Sep 18, 2020 at 10:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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