4
$\begingroup$

Let $\mathbf Z_K$ be the ring of integers of an algebraic number field $K$. It is well known that $\mathbf Z_K$ has infinitely many non-associated atoms (and hence is not a Cohen-Kaplansky domain).

Q. Is there a slick proof of this result? If so, could you provide a reference?

I feel the answer should be yes, but the only proof I knew until a few days ago is based on proving something much stronger and is sensibly more complicated than what I'm keen to consider a ``slick proof'': that $\mathbf Z_K$ has infinitely many prime ideals, and each of these contains a prime element.

$\endgroup$
4
  • 1
    $\begingroup$ Do you mean, there are infinitely many irreducible elements that don't differ by a unit? If so, why can't we pick the irreducible inductively? If we have $a_1,\dotsc,a_k$, then pick a prime $\mathfrak{p}$ not containing any of the $a_i$, pick an element $b \in \mathfrak{p}$, and decompose until we get $a_{k+1} \in \mathfrak{p}$ irreducible. $\endgroup$ Nov 7, 2019 at 4:31
  • $\begingroup$ @DongryulKim Maybe I'm being slow, but how do you know in the first place that such a prime ideal đť”­ does exist? Are you basically trying to reproduce Euclid's proof of the infinitude of rational primes? Something like, ``Assume to the contrary that there are only finitely many pairwise non-associated atoms $a_1, \ldots, a_n$ and look at the factorizations of $a_1 \cdots a_n+1$". But the problem with this argument is that one should guarantee that $a_1\cdots a_n+1$ isn't going to be a unit. $\endgroup$ Nov 7, 2019 at 4:48
  • $\begingroup$ (1) For the existence of $\mathfrak{p}$, here's what I'm doing. Given any nonzero number $a$, there exist only finitely many prime ideals containing $a$. So given a finite number of $a_i$, I can find a prime not containing any of them. (2) I guess the spirit is similar to Euclid's proof, but it's slightly different. I already know that there are infinitely many prime ideals, and I'm using it to run the argument. $\endgroup$ Nov 7, 2019 at 5:10
  • $\begingroup$ @DongryulKim I see, and this is slicker than what I had in mind, since the existence of infinitely many prime ideals in $\mathbf Z_K$ has a rather straightforward proof (somehow I had convinced myself that this wouldn't have been enough and I would have also needed to know that each prime ideal contains a prime element). Thanks. $\endgroup$ Nov 7, 2019 at 6:08

1 Answer 1

6
$\begingroup$

Fact 1. The ring $\textbf{Z}_K$ has infinitely many prime ideals and has Krull dimension $1$.

Proof. We have a surjective morphism $\mathrm{Spec}\,\textbf{Z}_K\rightarrow \mathrm{Spec}\mathbb{Z}$. This follows from the fact that the inclusion $\mathbb{Z}\subseteq \textbf{Z}_K$ is an integral extension of rings. So it suffices to invoke the going-up theorem of Cohen-Seidenberg. There is at least one prime ideal of $\textbf{Z}_K$ in each fiber of the morphism.

Fact 2. Let $A$ be a Noetherian domain. Then every nonzero and non-invertible element $a\in A$ can be decomposed as

$$a = f_1\cdot f_2\cdot ...\cdot f_n$$

where $f_1,...,f_n\in A$ are irreducible.

Proof. Fix $a\in A$ as in the statement. Let $\mathcal{F}$ be a family of proper principial ideals of $A$ that contain $a$. Since $A$ is Noetherian, in $\mathcal{F}$ there exists a maximal element with respect to inclusion. Say that this maximal element is generated by some $f_1\in A$. Then $a = f_1\cdot a_1$. If $a_1$ is non-invertible apply this argument to $a_1$ and obtain $f_2$ and $a_2$. In other words, continue by induction. It is clear that

  1. All $f_1,f_2,...$ are irreducible (otherwise ideals generate by them won't be maximal among principial ideals).

  2. This procedure must stop at some point. Indeed, $a_{n+1}$ must be invertible for some $n\in \mathbb{N}$, since otherwise you will obtain infinite and increasing chain of ideals $Aa_1\subseteq Aa_2\subseteq ...\subseteq Aa_n\subseteq ...$ in the Noetherian ring $A$.

This implies that $a = f_1\cdot f_2\cdot ...\cdot f_n$.

Remark. The decomposition in Fact 2 is by no means unique.

Corollary. All nonzero prime ideals of $\textbf{Z}_K$ are maximal.

Now you can construct infinite sequence of irreducible elements of $\textbf{Z}_K$ as follows. Pick $\{\mathfrak{p}_n\}_{n\in \mathbb{N}}$ a sequence of pairwise distinct maximal ideals in $\textbf{Z}_K$. This is possible by Fact 1 and Corollary. Prime avoidance shows that

$$\mathfrak{p}_{n+1}\subsetneq \bigcup_{k=0}^n\mathfrak{p}_k$$

Hence for every $n\in \mathbb{N}$ there exists $$r_{n+1}\in \mathfrak{p}_{n + 1}\setminus \left(\bigcup_{k=0}^n\mathfrak{p}_k\right)$$

and by virtue of Fact 2 you may pick irreducible divisor $f_{n+1}$ of $r_{n+1}$ in $\mathfrak{p}_{n+1}$. You still have that $$f_{n+1}\in \mathfrak{p}_{n + 1}\setminus \left(\bigcup_{k=0}^n\mathfrak{p}_k\right)$$ and hence $\{f_n\}_{n\in \mathbb{N}}$ are irreducible elements of $\textbf{Z}_K$ such that no two differ by unit.

This actually proves the following.

Proposition. Let $A$ be a Noetherian domain with infinitely many maximal ideals, then $A$ contains an infinite set of irreducible elements such that no two differ by unit.

Remark. If $A$ is a Dedekind ring with finitely many prime ideals, then one can prove that $A$ is a UFD and hence the proposition above does not hold.

$\endgroup$

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.