Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

It is well known that $\mathbb{Z}[a_1, \dots, a_n]/(a)$ is a finite ring if each $a_i$ is an algebraic integer and $a \neq 0.$

I suppose this statement becomes wrong if we just require those $a_i$ to be arbitrary algebraic numbers, but unfortunately I failed in finding any counterexample until now. I would appreciate it if somebody could name one. I am particularly interested in the case where $a$ is irreducible.

Thank you in advance!

share|improve this question
1  
Undoubtedly you know that any algebraic number can be written in the form "an algebraic integer divide by a rational integer". Somehow this suggests to me that you could learn something by first looking at the toy case $n=1$, $a_1=1/2$. May be you already did study it? –  Jyrki Lahtonen Aug 3 '13 at 17:49
    
@JyrkiLahtonen: Thank you for this hint! I already had similar thoughts and considered some quotients of say $\mathbb{Z}[\frac{1}{2}], \mathbb{Z}[\frac{\sqrt{2}}{3}]$ and so on - but without success. –  Dune Aug 3 '13 at 17:53
    
@Dune Let me point something out, that may be helpful. The fact that $\mathbb{Z}[a_1,\ldots,a_n]/(a)$ is always a finite ring relies heavily upon the fact that $\mathbb{Z}[a_1,\ldots,a_n]$ is a finitely generated group. Because then, $(a)$, being just $a\mathbb{Z}[a_1,\ldots,a_n]$ has the same finite rank as $\mathbb{Z}[a_1,\ldots,a_n]$. Thus, their quotient is a finitely generate, rank 0 abelian group, which forces it to be finite. Thus, when you start moving to non-integers, so tha the adjunction ring is no longer finite rank/free that's where you're problem lies. –  Alex Youcis Aug 3 '13 at 18:22
2  
If $R$ is a ring, $a\in R$, and $S\subset R$ is multiplicatively closed, then $S^{-1}R/aS^{-1}R\cong S^{-1}\left(R/aR\right)$ (where the $S$ on the right hand side really means the image of $S$ in $R/aR$). In particular, if $R/(a)$ is finite, the any localization of it is also finite, which means $S^{-1}R/aS^{-1}R$ is finite. This implies any localization of a ring of the form $\mathbb{Z}[a_1,\ldots,a_n]$ (with the $a_i$ algebraic integers) has the property that the quotient by a non-zero ideal is finite. –  Julian Rosen Aug 3 '13 at 18:41
1  
@Dune If a ring $R=\mathbb{Z}[a_1,\ldots,a_n]$ with the $a_i$ algebraic numbers is a countexample (i.e., contains a non-zero element such that the quotient by that element is infinite), then $R$ cannot be a localization of an algebra generated over $\mathbb{Z}$ by algebraic integers. In particular, the rings $\mathbb{Z}[1/2]$ and $\mathbb{Z}[\sqrt{3}/2]=\mathbb{Z}[\sqrt{3}][1/2]$ have the property that the quotient by a non-zero ideal is finite. –  Julian Rosen Aug 3 '13 at 19:59

1 Answer 1

up vote 5 down vote accepted

Every ring of the form $\mathbb{Z}[a_1,\ldots,a_n]$ with the $a_i$ algebraic numbers has the property that the quotient by a non-zero ideal is finite.

Fix an algebraic closure $\overline{\mathbb{Q}}$ of $\mathbb{Q}$, and let $a_1,\ldots,a_n\in \overline{\mathbb{Q}}$. Set $R=\mathbb{Z}[a_1,\ldots,a_n]$. One can check that $R$ is a $1$-dimensional Noetherian domain. This means that for $0\neq a\in R$, $R/(a)$ is $0$-dimensional, and therefore Artinian.

Artinian rings are finite products of Artin local rings, and each Artin local factor of $R/(a)$ has the property that its residue field is finitely-generated as a $\mathbb{Z}$-algebra, and hence finite. Artin local rings with finite residue field are finite, so this completes the proof.

EDIT: I've added a proof that $\mathbb{Z}[\alpha_1,\ldots,\alpha_n]$ is $1$-dimensional.

Lemma: Suppose $S$ is a Noetherian domain of Krull dimension at most $1$. Let $K$ be an algebraic closure of the fraction field of $S$, and take $s\in K$. Then $S[s]$ is a Noetherian domain of dimension at most $1$.

Proof: Let $x$ be an indeterminate, and consider the surjection of $S$-algebras $\varphi:S[x]\to S[s]$, $x\mapsto s$, inducing an isomorphism $S[x]/\ker(\varphi)\cong S[s]$.

Now, $S[x]$ has dimension at most 2. Because $s$ is algebraic over the fraction field of $S$, $\varphi$ is not injective. This means that $\ker(\varphi)\neq (0)$, so $S[x]/\ker(\varphi)$ has dimension strictly smaller than the dimension of $S[x]$, i.e., $S[x]/\ker(\varphi)\cong S[s]$ has dimension at most $1$.

Finally, $S[s]$ is a domain because it is a subring of a field, and is Noetherian, as it is finitely generated over a Noetherian ring.

By induction, this lemma implies that for $S$ a $1$-dimensional Noetherian domain and $s_1,\ldots,s_k$ in an algebraic closure of the fraction field of $S$, $S[s_1,\ldots,s_k]$ is a Noetherian domain of dimension at most $1$.

Note that $R=\mathbb{Z}[a_1,\ldots,a_n]$ cannot be $0$-dimensional (i.e., a field) because $a_1,\ldots,a_n$ have some common denominator $N\in\mathbb{Z}$, and $R[1/N]$ is integral over the $1$-dimensional ring $\mathbb{Z}[1/N]$.

share|improve this answer
    
That is surprising! I would be happy if you add a hint how to show $\dim(R)=1$. :) –  Dune Aug 3 '13 at 20:15
    
@Dune If $S$ is a 1-dimensional Noetherian domain, and $f(x)\neq 0\in S[x]$ is prime, then $S[x]$ is 2-dimensional, and $S[x]/f(x)$ is a 1-dimensional Noetherian domain (because $(f(x))$ is a height 1 prime). Using this and induction on the number of generators of $R$ shows that $\dim(R)=1$. I bet there's a simpler argument, though. –  Julian Rosen Aug 3 '13 at 20:24
    
Great, thank you very much! –  Dune Aug 3 '13 at 20:30
    
@PinkElephants: Thank you for this detailed explanation! But can't we simply deduce that $R$ is one dimensional since $R[\frac{1}{N}]$ is integral over $\mathbb{Z}[\frac{1}{N}]$ and therefore one dimensional by the going up theorem? –  Dune Aug 3 '13 at 23:48

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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