Let $R \subseteq S$ be two Noetherian local rings, not necessarily regular, which are integral domains, with $m_RS=m_S$, namely, the ideal in $S$ generated by $m_R$ (= the maximal ideal of $R$) is $m_S$ (= the maximal ideal of $S$).

Further assume that $R$ and $S$ are $\mathbb{C}$-algebras, $R \subseteq S$ is flat and algebraic but not integral, where algebraic non-integral means: Every element of $S$ satisfies a polynomial with coefficients in $R$, with non-invertible (in $R$) leading coefficient.

Question: Could one find an example of such rings?

Unfortunately, the examples I find are integral, for example: $R=\mathbb{C}[x(x-1)]_{(x(x-1))}$, $S=\mathbb{C}[x]_{(x)}$.


(i) I am interested in both cases where $R$ and $S$ have the same fields of fractions or different fields of fractions.

(ii) Recall the following results, which are not applicable here, since I assume that $R \subseteq S$ is non-integral: If $A \subseteq B$ is integral and flat, then $A \subseteq B$ is faithfully flat, and if in addition $Q(A)=Q(B)$ (same fields of fractions), then $A=B$.

Relevant questions: a, b and c. Also asked in MSE.

Any hints and comments are welcome; thank you.

  • $\begingroup$ I assume you meant "not necessarily invertible leading coefficient"? $\endgroup$
    – Will Chen
    Apr 14, 2021 at 1:38
  • $\begingroup$ @WillChen, thank you for your comment. Actually, I meant 'necessarily, not invertible leading coefficient', since I want the extension to be algebraic but not integral. For an integral extension I have found an example (localizations of polynomial rings). I wish something like $\mathbb{Z} \subset \mathbb{Z}[\frac{1}{2}]$ which is flat, algebraic, non-integral, because $f(T)=2T-1 \in \mathbb{Z}[T]$ is such that $f(\frac{1}{2})=0$. But here the rings are requiered to be $\mathbb{C}$-algebras (I wished to exclude such cases as $\mathbb{Z} \subset \mathbb{Z}[\frac{1}{2}]$). $\endgroup$
    – user237522
    Apr 14, 2021 at 1:52
  • $\begingroup$ Well, the rings $\mathbb{Z}$ and $\mathbb{Z}[\frac{1}{2}]$ are not local, but we can consider localizations of them. . $\endgroup$
    – user237522
    Apr 14, 2021 at 2:03
  • $\begingroup$ Every element of $R$ satisfies a polynomial with coefficients in $R$ with non-invertible leading coefficient. E.g., $r \in R$ satisfies $2T - 2r\in R[T]$. I think what you mean to say is that every element of $S$ satisfies a polynomial in $R[T]$, but not every element in $S$ satisfies a monic polynomial in $R[T]$. $\endgroup$
    – Will Chen
    Apr 14, 2021 at 3:31
  • $\begingroup$ @WillChen, yes, you are right, I meant that every element of $S$ satisfies a polynomial in $R[T]$, bot not every element of $S$ satisfies a monic polynomial in $R[T]$. $\endgroup$
    – user237522
    Apr 14, 2021 at 13:19

1 Answer 1


In general if $R$ is a local ring, then its henselization $R^h$ is flat and "algebraic" over $R$, but rarely integral. The intuition is that the henselization is built out of localizations of etale extensions of $R$. Both localizations and etale extensions are flat and "algebraic" in your sense, but localizations are rarely integral.

To make a precise statement, first note that it suffices to find etale ring extensions $R\rightarrow R'\rightarrow S$ with $R$ normal, $S/R$ an etale local homomorphism of local rings, and $S/R'$ not integral.

Indeed, if we have found $R,R',S$ as above, then $S/R'$ non-integral implies $S/R$ non-integral. Moreover, etale morphisms preserve normality, so $S$ must be normal, hence a domain since $S$ is local so $\text{Spec }S$ is connected. Finally, etale ring maps are locally standard etale, so every element of $S$ is "algebraic" over $R$.

Here's a sketch of how to produce an example: Start with a connected normal scheme $Y$ over $\mathbb{C}$ and a finite flat map $f : X\rightarrow Y$ with $X$ irreducible. In characteristic 0, $f$ is generically etale, so let $y\in Y$ be a point above which $f$ is etale. Let $R := \mathcal{O}_{Y,y}$, let $X_R := X\times_Y\text{Spec }\mathcal{O}_{Y,y}$, and write $X_R = \text{Spec }R'$. Then $R'/R$ is finite etale. On the other hand, $R'$ is connected, since the generic point of $X$ maps to the generic point of $Y$ which lies in $\text{Spec }R$, so the generic point of $X$ lies in $X_R$ and specializes to every closed point of $X_R$, so $R'$ is a normal domain. However, if $X$ has multiple closed points lying over $y$, then $R'$ is not local, in which case let $\mathfrak{m}$ be a maximal ideal of $R'$, and let $S := R'_{\mathfrak{m}}$. Then $S/R$ becomes an etale local homomorphism of local rings with $S/R'$ non-integral, as desired.

For an explicit example, you can take $Y = \text{Spec }\mathbb{C}[y]$, and $$X := \text{Spec }\mathbb{C}[x,y]/(y^2-(x-a)(x-b)(x-c))$$ where $a,b,c\in\mathbb{C}$ are distinct. Then take $y\in Y$ to be the point $(y = 0)$. In particular, this shows that the henselization of $\mathbb{C}[y]_{(y)}$ also satisfies your conditions.

  • $\begingroup$ Thank you very much! Interesting. Please, what if $S$ is a simple extension of $R$, namely, $S=R[w]$, for some $w \in S$? (perhaps your answer also answers this; I do not know..). $\endgroup$
    – user237522
    Apr 14, 2021 at 13:23
  • $\begingroup$ @user237522 I don't understand your question $\endgroup$
    – Will Chen
    Apr 14, 2021 at 18:10
  • $\begingroup$ Please, I meant, same question and in addition to the conditions of the question we further assume that $S=R[w]$, for some $w \in S$. $\endgroup$
    – user237522
    Apr 14, 2021 at 18:56

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.