Lemma 10.138.12. Let R \to S be a ring map. Let I \subset R be an ideal. Assume
I^2 = 0,
R \to S is flat, and
R/I \to S/IS is formally smooth.
Then R \to S is formally smooth.
Lemma 10.138.12. Let R \to S be a ring map. Let I \subset R be an ideal. Assume
I^2 = 0,
R \to S is flat, and
R/I \to S/IS is formally smooth.
Then R \to S is formally smooth.
Proof. Assume (1), (2) and (3). Let P = R[\{ x_ t\} _{t \in T}] \to S be a surjection of R-algebras with kernel J. Thus 0 \to J \to P \to S \to 0 is a short exact sequence of flat R-modules. This implies that I \otimes _ R S = IS, I \otimes _ R P = IP and I \otimes _ R J = IJ as well as J \cap IP = IJ. We will use throughout the proof that
and similarly for P (see Lemma 10.131.12). By Lemma 10.138.7 the sequence
is split exact. Of course the middle term is \bigoplus _{t \in T} S/IS \text{d}x_ t. Choose a splitting \sigma : \Omega _{P/R} \otimes _ P S/IS \to J/(IJ + J^2). For each t \in T choose an element f_ t \in J which maps to \sigma (\text{d}x_ t) in J/(IJ + J^2). This determines a unique S-module map
with the property that \tilde\sigma (\text{d}x_ t) = f_ t. As \sigma is a section to \text{d} the difference
is a self map J/J^2 \to J/J^2 whose image is contained in (IJ + J^2)/J^2. In particular \Delta ((IJ + J^2)/J^2) = 0 because I^2 = 0. This means that \Delta factors as
where \overline{\Delta } is a S/IS-module map. Using again that the sequence (10.138.12.1) is split, we can find a S/IS-module map \overline{\delta } : \Omega _{P/R} \otimes _ P S/IS \to (IJ + J^2)/J^2 such that \overline{\delta } \circ d is equal to \overline{\Delta }. In the same manner as above the map \overline{\delta } determines an S-module map \delta : \Omega _{P/R} \otimes _ P S \to J/J^2. After replacing \tilde\sigma by \tilde\sigma + \delta a simple computation shows that \Delta = 0. In other words \tilde\sigma is a section of J/J^2 \to \Omega _{P/R} \otimes _ P S. By Lemma 10.138.7 we conclude that R \to S is formally smooth. \square
Comments (3)
Comment #7696 by Peter Fleischmann on
Comment #7697 by Stacks Project on
Comment #7698 by Peter Fleischmann on
There are also: