1

I have a request for an alternative proof of a fact in ring theory. Given an extension of rings $R \supseteq A$, we say that $x\in R$ is integral over $A$ if there is a monic polynomial $p$ with coefficients from $A$ so that $p(x) = 0$.

In pg. 2 of Ash's "A Course in Algebraic Number Theory," Ash presents a theorem of the form "TFAE" for integrality over a ring. I am interested in an alternative proof of the following:

Theorem: If $x$ belongs to a subring $B \subseteq R$ which is a finitely-generated $A$-module so that $A \subseteq B$, then $x$ is integral over $A$.

Proof: Suppose that $B$ is $A$-generated by $\{\beta_1,\dots,\beta_n\}$. Then for each $i$:

$$x\beta_i = \sum_{i=1}^n c_{ij}\beta_j$$

for some $A$-coefficients $c_{ij}$. Writing $\overline{\beta}$ as the vector with elements $\beta_i$ and $C = [c_{ij}]$, it then follows that $(xI - C)\overline{\beta} = \overline{0}$. Recalling that $\text{adj}(A)A = \det(A)I$, it follows that $\det(xI-C)I\overline{\beta} = \overline{0}$, hence for every $b \in B$ it follows that $\det(xI - C)b = 0$. Setting $b=1$ gives the polynomial we sought. $\square$

Here is my question: is there a more "natural" proof of this fact? To avoid ambiguity, by "natural" I mean ring-theoretic. Since we have a foot in the world of modules, it's not unnatural to use some linear algebra, but when I tried to show this myself, it didn't even occur to me to use the determinant. So I suppose an alternative question I have is: "is there a nice proof here that doesn't use the determinant?"

While I Am
  • 2,380
  • This so-called "determinant trick" should seem natural if you are familiar with linear algebra, viz. as in Cayley-Hamilton, we compute a (monic characteristic) polynomial having $x$ as root (via determinant of linear map $\beta \to x\beta$), e.g. as here. For noncommutative generalizations we can employ Nakayama's Lemma, e.g. see here. – Bill Dubuque Jun 17 '22 at 01:47

1 Answers1

3

(I'll use $r$ instead of $x$ to avoid confusion with the polynomial algebra.)

We'll say is that an algebra $R$ over $A$ is finite over $A$ if it is finitely generated as an $A$ module.

It's easy (and "natural") to see that $A[r]$ is finite over $A$ if and only if $r$ is integral over $A$, so we'll take that for granted.

The question now becomes: how can we conclude that $A[r]$ is finite over $A$ if $r$ belongs to a subring $B$ which is finite over $A$?

Well, the obvious observation is that $A[r]$ is a submodule of $B$, so if $A$ is Noetherian, we are done. From there, we can seek for a way to reduce the general case to the Noetherian case, which is indeed possible.

Suppose that $B$ is $A$-generated by $\beta_1, \dots, \beta_n$. Write $$1=\sum_{i=1}^na_i \beta_i$$ $$\beta_i \beta_j= \sum_{k=1}^n c_{ijk} \beta_k$$ and $$r\beta_i=\sum_{j=1}^nd_{ij}\beta_j$$ For $a_i,c_{ijk},d_{ij} \in A$

Now let $A'=\Bbb Z[a_i,d_{ij},c_{ijk}] \subset A$ be the $\Bbb Z$-subalgebra generated by all these coefficients. Let $B'\subset B$ be the module $A'$-generated by $\beta_1, \dots, \beta_n$. The way we chose our coefficients implies that $B'$ is actually closed under multiplication, so it is a subalgebra and $r \in B'$. Thus we are in exactly the same situation as before except that $A$ was replaced by $A'$ and $B$ by $B'$. But $A'$ is Noetherian by the Hilbert basis theorem, so we win.

Lukas Heger
  • 20,801