The Stacks project

Lemma 10.138.7. Let $R \to S$ be a ring map. Let $P \to S$ be a surjective $R$-algebra map from a polynomial ring $P$ onto $S$. Denote $J \subset P$ the kernel. Then $R \to S$ is formally smooth if and only if the sequence

\[ 0 \to J/J^2 \to \Omega _{P/R} \otimes _ P S \to \Omega _{S/R} \to 0 \]

of Lemma 10.131.9 is a split exact sequence.

Proof. Assume $S$ is formally smooth over $R$. By Lemma 10.138.5 this means there exists an $R$-algebra map $S \to P/J^2$ which is a right inverse to the canonical map $P/J^2 \to S$. By Lemma 10.131.11 we have $\Omega _{P/R} \otimes _ P S = \Omega _{(P/J^2)/R} \otimes _{P/J^2} S$. By Lemma 10.131.10 the sequence is split.

Assume the exact sequence of the lemma is split exact. Choose a splitting $\sigma : \Omega _{S/R} \to \Omega _{P/R} \otimes _ P S$. For each $\lambda \in S$ choose $x_\lambda \in P$ which maps to $\lambda $. Next, for each $\lambda \in S$ choose $f_\lambda \in J$ such that

\[ \text{d}f_\lambda = \text{d}x_\lambda - \sigma (\text{d}\lambda ) \]

in the middle term of the exact sequence. We claim that $s : \lambda \mapsto x_\lambda - f_\lambda \mod J^2$ is an $R$-algebra homomorphism $s : S \to P/J^2$. To prove this we will repeatedly use that if $h \in J$ and $\text{d}h = 0$ in $\Omega _{P/R} \otimes _ R S$, then $h \in J^2$. Let $\lambda , \mu \in S$. Then $\sigma (\text{d}\lambda + \text{d}\mu - \text{d}(\lambda + \mu )) = 0$. This implies

\[ \text{d}(x_\lambda + x_\mu - x_{\lambda + \mu } - f_\lambda - f_\mu + f_{\lambda + \mu }) = 0 \]

which means that $x_\lambda + x_\mu - x_{\lambda + \mu } - f_\lambda - f_\mu + f_{\lambda + \mu } \in J^2$, which in turn means that $s(\lambda ) + s(\mu ) = s(\lambda + \mu )$. Similarly, we have $\sigma (\lambda \text{d}\mu + \mu \text{d}\lambda - \text{d}\lambda \mu ) = 0$ which implies that

\[ \mu (\text{d}x_\lambda - \text{d}f_\lambda ) + \lambda (\text{d}x_\mu - \text{d}f_\mu ) - \text{d}x_{\lambda \mu } + \text{d}f_{\lambda \mu } = 0 \]

in the middle term of the exact sequence. Moreover we have

\[ \text{d}(x_\lambda x_\mu ) = x_\lambda \text{d}x_\mu + x_\mu \text{d}x_\lambda = \lambda \text{d}x_\mu + \mu \text{d} x_\lambda \]

in the middle term again. Combined these equations mean that $x_\lambda x_\mu - x_{\lambda \mu } - \mu f_\lambda - \lambda f_\mu + f_{\lambda \mu } \in J^2$, hence $(x_\lambda - f_\lambda )(x_\mu - f_\mu ) - (x_{\lambda \mu } - f_{\lambda \mu }) \in J^2$ as $f_\lambda f_\mu \in J^2$, which means that $s(\lambda )s(\mu ) = s(\lambda \mu )$. If $\lambda \in R$, then $\text{d}\lambda = 0$ and we see that $\text{d}f_\lambda = \text{d}x_\lambda $, hence $\lambda - x_\lambda + f_\lambda \in J^2$ and hence $s(\lambda ) = \lambda $ as desired. At this point we can apply Lemma 10.138.5 to conclude that $S/R$ is formally smooth. $\square$


Comments (7)

Comment #1808 by Peter Johnson on

Around line 4 of the proof, P/J^2 is a direct sum of R-modules, but in general will only be a split extension of R-algebras.

Comment #3206 by Dario Weißmann on

Typo: The map is a right inverse to .

And why not invoke lemma 10.130.10 from there?

Comment #3310 by on

A rare change which shortens the Stacks project. Thanks very much. Change is here.

Comment #4297 by Bogdan on

A formula should read

Comment #6689 by WhatJiaranEatsTonight on

TThe first paragraph need to cite 10.131.11 to show .

There are also:

  • 5 comment(s) on Section 10.138: Formally smooth maps

Post a comment

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.




In order to prevent bots from posting comments, we would like you to prove that you are human. You can do this by filling in the name of the current tag in the following input field. As a reminder, this is tag 031I. Beware of the difference between the letter 'O' and the digit '0'.