6
$\begingroup$

Let $R$ be a commutative ring. Clearly the kernel of $h$ is a prime ideal whenever $h : R \rightarrow \mathbb{Z}$ is a ring homomorphism. But is the converse true: does every prime ideal arise as kernel of a homomorphism into $\mathbb{Z}$?

$\endgroup$
  • 3
    $\begingroup$ Since the quotient group of an ideal is an integral domain iff the ideal is prime, this translates to "Given an integral domain $R$, is there always a homomorphism into $\Bbb Z$ with trivial kernel?" $\endgroup$ – Arthur Oct 1 '12 at 8:02
  • $\begingroup$ @Arthur, yes, I had commutative rings in mind. I've edited the question accordingly. $\endgroup$ – Martin Berger Oct 1 '12 at 8:08
  • 2
    $\begingroup$ As @Arthur's comment shows, is this wrong, since a homomorhpism $\mathbb Z[X] \to \mathbb Z$ can't have trivial kernel ... so the ideal $(Y) \subseteq \mathbb Z[X,Y]$ is prime but arises not as a kernel of a homomorphism to $\mathbb Z$. $\endgroup$ – martini Oct 1 '12 at 8:09
14
$\begingroup$

Let $R$ be a field of uncountable cardinality. For a minimal counterexample, let $R = \mathbb{F}_2$.

The correct salvage is that every prime ideal arises as the kernel of a homomorphism into some integral domain (in fact, into some field). It shouldn't be possible to say anything stronger than this.

$\endgroup$
  • 2
    $\begingroup$ "Into some integral domain" and "into some field" are equivalent statements, since any integral domain is contained in its field of fractions. Just as a side comment. $\endgroup$ – Arthur Oct 1 '12 at 8:13
  • 1
    $\begingroup$ Yes, but they're not identical statements; it takes a tiny bit of work to show that they're equivalent, and if I had said "surjective homomorphism" then they wouldn't be equivalent. I only mentioned integral domains because $\mathbb{Z}$ is an integral domain, not a field. $\endgroup$ – Qiaochu Yuan Oct 1 '12 at 8:14
  • $\begingroup$ Because $R/\mathfrak p$ is an integral domain into which there is a surjective homomorphism with exactly $\mathfrak p$ as kernel, but it's a field only if $\mathfrak p$ is maximal, right? $\endgroup$ – Arthur Oct 1 '12 at 8:17
  • 2
    $\begingroup$ @Marc: if something doesn't exist for cardinality reasons, then it really doesn't exist. It's a fun proof technique to use because it rarely applies, but when it does, it really applies. $\endgroup$ – Qiaochu Yuan Oct 1 '12 at 8:24
  • 5
    $\begingroup$ Only a minor comment: The first two sentences being in the same line makes them seem related, but they are not. Do you mind separating them (e.g. by an 'or')? $\endgroup$ – Gregor Botero Oct 1 '12 at 9:04
12
$\begingroup$

The converse is very, very false. For instance if $R$ is a ring of nonzero characteristic (think $\mathbb Z/n\mathbb Z$ with $n>0$), there aren't any homomorphismes $R\to\mathbb Z$ at all, yet $R$ will always (assuming the axiom of choice) have at least one prime ideal. Simple counterexample $\mathbb Z/2\mathbb Z$ and its zero ideal, or (if you don't like zero ideals) $\mathbb Z/4\mathbb Z$ and its ideal $\{\overline0,\overline2\}$.

$\endgroup$
  • 2
    $\begingroup$ @HagenvonEitzen Below "false" there is "wrong", "really wrong" and then "not even wrong" :P $\endgroup$ – rschwieb Oct 1 '12 at 12:46
1
$\begingroup$

If $h: R \rightarrow \mathbb{Z}$ is a ring homomorphism, then $h$ is also a homomorphism of (additive) abelian groups. If $P$ is the kernel of $h$, then $R/P$ is isomorphic to the image of $h$, which is a subgroup of $\mathbb{Z}$. And a subgroup of $\mathbb{Z}$ is either trivial or isomorphic to $\mathbb{Z}$.

So if $P$ is a prime ideal of $R$ (prime ideals are by definition proper, so $R/P$ is never trivial) such that $R/P$ is not isomorphic to $\mathbb{Z}$ as an abelian group, then $P$ cannot possibly be the kernel of a homomorphism $R \rightarrow \mathbb{Z}$. There are plenty of prime ideals which fit this description. If $R$ is $\mathbb{Z}$ itself, and $P = \mathbb{Z} p$ for some prime number $p$, then $R/P$ is finite!

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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