4
$\begingroup$

Let $k$ be a field, and let $f : k[t_1,\dots, t_r]\to k[x_1,\dots, x_n]$ be the $k$-algebra map defined by $f(t_i) = f_i\in k[x_1,\dots, x_n].$ Suppose that the $f_i$ are algebraically independent over $k,$ so that the map $f$ is an injection. To show that $f$ is flat, it suffices to show that $f_\mathfrak{m} : k[t_1,\dots, t_r]\to k[x_1,\dots, x_n]_\mathfrak{m}$ is flat for $\mathfrak{m}\subseteq k[x_1,\dots, x_n]$ an arbitrary prime ideal. If $(f_1,\dots, f_r)\subseteq\mathfrak{m},$ then $f_\mathfrak{m}$ is flat if and only if the $f_i$ form a regular sequence in $k[x_1,\dots, x_n]_\mathfrak{m}$ (c.f. Eisenbud's Commutative Algebra, exercise 18.18).

In general, we cannot expect that the $f_i$ form a regular sequence in every $k[x_1,\dots, x_n]_\mathfrak{m}.$ However, suppose that if they do fail to form a regular sequence, they fail in a "silly" way: by some of the $f_i$ becoming units in $k[x_1,\dots, x_n]_\mathfrak{m}.$ My question is, does flatness still hold if some of these $f_i$ become units? In particular, can we generalize exercise 18.18 of Eisenbud's Commutative Algebra to say:

Let $(R,\mathfrak{m})$ be a Noetherian local ring containing a field $k,$ and let $x_1,\dots, x_r\in\mathfrak{m}$ and $u_1,\dots, u_s\in R^\times$ be two sequences of elements. Then $i : k[x_1,\dots, x_r, u_1, \dots, u_s]\to R$ is flat if $x_1,\dots, x_r$ is a regular sequence.

Edit: Above I am happy to interpret $k[x_1,\dots, x_r, u_1, \dots, u_s]$ as either the $k$-subalgebra of $R$ generated by the $x_i$ and $u_i$ or as a polynomial ring over $k$ in the $x_i$ and $u_i$ (mapping to $R$ by sending each variable to the corresponding element). In the situation I am concerned with, these coincide, so I would be interested in results about either situation.

Edit 2: As noted in the comments to Eric's answer below, this problem could also be stated as follows:

Suppose that $k$ is a field, and that $(R,\mathfrak{m})$ is a Noetherian local ring containing $A := k[t_1^{\pm1},\dots, t_s^{\pm1}].$ Let $x_1,\dots, x_r$ be a sequence of elements in $\mathfrak{m}.$ Then $i : A[x_1,\dots, x_r]\to R$ is flat if $x_1,\dots, x_r$ is a regular sequence.

The map $i$ above will be flat if and only if $A[x_1,\dots, x_n]_\mathfrak{n}\to R$ is flat, where $\mathfrak{n} = i^{-1}(\mathfrak{m}).$ Moreover, assuming that $R$ is Cohen-Macauley (which I am happy to do), this map is flat if and only if $\dim R = \dim A[x_1,\dots, x_r]_\mathfrak{n} + \dim R/\mathfrak{n}R,$ but I'm not sure how to get a handle on $A[x_1,\dots, x_r]_{\mathfrak{n}}$ (or $\dim R/\mathfrak{n}R$) in general. (The proof of the exercise mentioned above does not generalize immediately to this situation: $\mathfrak{n}$ need not be simply $(x_1,\dots, x_n)A[x_1,\dots, x_r];$ there could be contributions coming from the $t_i.$)

As an example, we might consider the injection $k[u]\to k[x]_{(x)},$ where $u = 1/(1+x).$ In this case the preimage of $(x)$ is $(1 - u),$ so $k[u]\to k[x]_{(x)}$ is flat because $k[u]_{(1 - u)}\to k[x]_{(x)}$ is an isomorphism. The result is true in this very simple example, but I'm not sure that it suggests a more general approach.

$\endgroup$
3
  • $\begingroup$ What do you mean by $k[x,u]$ with $u\in R^{\times }$? $\endgroup$
    – abx
    Oct 22, 2019 at 12:47
  • $\begingroup$ @abx: I guess it's the k-subalgebra of R generated by x and u. For example, if $R=k[[x, y]]$, $i$ could be $i: k[x,1/(1-y)] \to k[[x,y]]$. $\endgroup$
    – tj_
    Oct 22, 2019 at 14:53
  • $\begingroup$ I mean on the left hand side that that $x$ and $u$ are formal variables mapping to the specified elements of $R,$ although I would also be happy interpreting the left hand side as the $k$-subalgebra of $R$ generated by those elements. In the cases I'm interested in, the subalgebra generated by the elements will be a polynomial algebra on those elements. $\endgroup$
    – Stahl
    Oct 22, 2019 at 21:31

1 Answer 1

2
$\begingroup$

Let's call $A = k[x_1, \dots, x_r; u_1, \dots, u_s]$. If $u_1, \dots, u_s \in R^\times$, then $i: A \to R$ factors through the localization $A_u$ with respect to $u = u_1 \cdots u_s$. Localization is flat, so I believe the condition reduces to the statement of the exercise, as stated in Eisenbud.

A bit more precisely, suppose $0 \to N' \to N \to N'' \to 0$ is an exact $A$-sequence, and we want to check that applying $-\otimes_A R$ leaves this sequence exact. As stated above, the $u_i$ are units in $R$ so $-\otimes_A R = (-\otimes_A A_u) \otimes_{A_u}R$. Therefore, we can first localize $0 \to N'_u \to N_u \to N''_u \to 0$, and now $-\otimes_{A_u} R$ will leave the sequence exact iff the $x_j$ form a regular sequence.

$\endgroup$
6
  • $\begingroup$ This might work! But, I have to think about whether the proof of the exercise will apply to $A_u\to R,$ as $A_u$ is not necessarily going to be of the form $k'[x_1,\dots, x_r]$ for some $k'$ contained in $R.$ Is this obvious? $\endgroup$
    – Stahl
    Oct 23, 2019 at 6:03
  • $\begingroup$ @Stahl: If I'm not mistaken, $A_u \cong k[u_i, u_i^{-1}][x_j]$. $\endgroup$
    – tj_
    Oct 23, 2019 at 8:24
  • $\begingroup$ @tj_ that's true, but $k[x,x^{-1}]$ is not the same as $k(x)$ (and similarly for more variables): the latter contains things like $1/(x-1)$, but the former does not. $\endgroup$
    – Stahl
    Oct 23, 2019 at 15:03
  • $\begingroup$ @Stahl: That's why I have written $k[u_i^\pm]$ and not $k(u_i)$. But maybe Eisenbud can be adjusted for $k[u_i^\pm]$ in place of a field (I don't know if it can). $\endgroup$
    – tj_
    Oct 23, 2019 at 20:59
  • $\begingroup$ @tj_ Ah, I see. Yes, I've been wondering this as well. It doesn't seem unreasonable that if $R$ is a local noetherian ring containing $A_u$ then the map $A_u[\{x_j\}_j]\to R$ will be flat iff the $x_j$ form a regular sequence in $R,$ but I haven't been able to write down a proof of that statement. $\endgroup$
    – Stahl
    Oct 23, 2019 at 21:15

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.