Processing math: 100%

The Stacks project

Lemma 10.140.2. Let k be an algebraically closed field. Let S be a finite type k-algebra. Let \mathfrak m \subset S be a maximal ideal. The following are equivalent:

  1. The ring S_{\mathfrak m} is a regular local ring.

  2. We have \dim _{\kappa (\mathfrak m)} \Omega _{S/k} \otimes _ S \kappa (\mathfrak m) \leq \dim (S_{\mathfrak m}).

  3. We have \dim _{\kappa (\mathfrak m)} \Omega _{S/k} \otimes _ S \kappa (\mathfrak m) = \dim (S_{\mathfrak m}).

  4. There exists a g \in S, g \not\in \mathfrak m such that S_ g is smooth over k. In other words S/k is smooth at \mathfrak m.

Proof. Note that (1), (2) and (3) are equivalent by Lemma 10.140.1 and Definition 10.110.7.

Assume that S is smooth at \mathfrak m. By Lemma 10.137.10 we see that S_ g is standard smooth over k for a suitable g \in S, g \not\in \mathfrak m. Hence by Lemma 10.137.7 we see that \Omega _{S_ g/k} is free of rank \dim (S_ g). Hence by Lemma 10.140.1 we see that \dim (S_{\mathfrak m}) = \dim (\mathfrak m/\mathfrak m^2) in other words S_\mathfrak m is regular.

Conversely, suppose that S_{\mathfrak m} is regular. Let d = \dim (S_{\mathfrak m}) = \dim \mathfrak m/\mathfrak m^2. Choose a presentation S = k[x_1, \ldots , x_ n]/I such that x_ i maps to an element of \mathfrak m for all i. In other words, \mathfrak m'' = (x_1, \ldots , x_ n) is the corresponding maximal ideal of k[x_1, \ldots , x_ n]. Note that we have a short exact sequence

I/\mathfrak m''I \to \mathfrak m''/(\mathfrak m'')^2 \to \mathfrak m/(\mathfrak m)^2 \to 0

Pick c = n - d elements f_1, \ldots , f_ c \in I such that their images in \mathfrak m''/(\mathfrak m'')^2 span the kernel of the map to \mathfrak m/\mathfrak m^2. This is clearly possible. Denote J = (f_1, \ldots , f_ c). So J \subset I. Denote S' = k[x_1, \ldots , x_ n]/J so there is a surjection S' \to S. Denote \mathfrak m' = \mathfrak m''S' the corresponding maximal ideal of S'. Hence we have

\xymatrix{ k[x_1, \ldots , x_ n] \ar[r] & S' \ar[r] & S \\ \mathfrak m'' \ar[u] \ar[r] & \mathfrak m' \ar[r] \ar[u] & \mathfrak m \ar[u] }

By our choice of J the exact sequence

J/\mathfrak m''J \to \mathfrak m''/(\mathfrak m'')^2 \to \mathfrak m'/(\mathfrak m')^2 \to 0

shows that \dim ( \mathfrak m'/(\mathfrak m')^2 ) = d. Since S'_{\mathfrak m'} surjects onto S_{\mathfrak m} we see that \dim (S_{\mathfrak m'}) \geq d. Hence by the discussion preceding Definition 10.60.10 we conclude that S'_{\mathfrak m'} is regular of dimension d as well. Because S' was cut out by c = n - d equations we conclude that there exists a g' \in S', g' \not\in \mathfrak m' such that S'_{g'} is a global complete intersection over k, see Lemma 10.135.4. Also the map S'_{\mathfrak m'} \to S_{\mathfrak m} is a surjection of Noetherian local domains of the same dimension and hence an isomorphism. Hence S' \to S is surjective with finitely generated kernel and becomes an isomorphism after localizing at \mathfrak m'. Thus we can find g' \in S', g \not\in \mathfrak m' such that S'_{g'} \to S_{g'} is an isomorphism. All in all we conclude that after replacing S by a principal localization we may assume that S is a global complete intersection.

At this point we may write S = k[x_1, \ldots , x_ n]/(f_1, \ldots , f_ c) with \dim S = n - c. Recall that the naive cotangent complex of this algebra is given by

\bigoplus S \cdot f_ j \to \bigoplus S \cdot \text{d}x_ i

see Lemma 10.136.12. By Lemma 10.137.16 in order to show that S is smooth at \mathfrak m we have to show that one of the c \times c minors g_ I of the matrix β€œA” giving the map above does not vanish at \mathfrak m. By Lemma 10.140.1 the matrix A \bmod \mathfrak m has rank c. Thus we win. \square


Comments (1)

Comment #725 by Keenan Kidwell on

For the description of the naive cotangent complex in the last paragraph, I think the correct tag is (3) of 00SV. This reference is cited for the same reason in the proof of 00TE, but that's not where it is proved. For "smooth at iff some minor is non-zero modulo ," 00TE is the right tag.


Your email address will not be published. Required fields are marked.

In your comment you can use Markdown and LaTeX style mathematics (enclose it like $\pi$). A preview option is available if you wish to see how it works out (just click on the eye in the toolbar).

Unfortunately JavaScript is disabled in your browser, so the comment preview function will not work.

All contributions are licensed under the GNU Free Documentation License.