The Stacks project

Lemma 60.2.7. Let $(B, I, \gamma ) \to (B', I', \gamma ')$ be a homomorphism of divided power rings. Let $I \subset J \subset B$ and $I' \subset J' \subset B'$ be ideals. Assume

  1. $B/I \to B'/I'$ is flat, and

  2. $J' = JB' + I'$.

Then the canonical map

\[ D_{B, \gamma }(J) \otimes _ B B' \longrightarrow D_{B', \gamma '}(J') \]

is an isomorphism.

Proof. Set $D = D_{B, \gamma }(J)$. Choose elements $f_ t \in J$ which generate $J/I$. Set $\mathcal{R} = \{ (r_0, r_ t) \in I \oplus \bigoplus \nolimits _{t \in T} B \mid \sum r_ t f_ t = r_0 \text{ in }B\} $ as in the proof of Lemma 60.2.4. This lemma shows that

\[ D = B\langle x_ t \rangle / K \]

where $K$ is generated by the elements $x_ t - f_ t$ and $\delta _ n(\sum r_ t x_ t - r_0)$ for $(r_0, r_ t) \in \mathcal{R}$. Thus we see that

60.2.7.1
\begin{equation} \label{crystalline-equation-base-change} D \otimes _ B B' = B'\langle x_ t \rangle /K' \end{equation}

where $K'$ is generated by the images in $B'\langle x_ t \rangle $ of the generators of $K$ listed above. Let $f'_ t \in B'$ be the image of $f_ t$. By assumption (1) we see that the elements $f'_ t \in J'$ generate $J'/I'$ and we see that $x_ t - f'_ t \in K'$. Set

\[ \mathcal{R}' = \{ (r'_0, r'_ t) \in I' \oplus \bigoplus \nolimits _{t \in T} B' \mid \sum r'_ t f'_ t = r'_0 \text{ in }B'\} \]

To finish the proof we have to show that $\delta '_ n(\sum r'_ t x_ t - r'_0) \in K'$ for $(r'_0, r'_ t) \in \mathcal{R}'$, because then the presentation (60.2.7.1) of $D \otimes _ B B'$ is identical to the presentation of $D_{B', \gamma '}(J')$ obtain in Lemma 60.2.4 from the generators $f'_ t$. Suppose that $(r'_0, r'_ t) \in \mathcal{R}'$. Then $\sum r'_ t f'_ t = 0$ in $B'/I'$. As $B/I \to B'/I'$ is flat by assumption (1) we can apply the equational criterion of flatness (Algebra, Lemma 10.39.11) to see that there exist an $m > 0$ and $r_{jt} \in B$ and $c_ j \in B'$, $j = 1, \ldots , m$ such that

\[ r_{j0} = \sum \nolimits _ t r_{jt} f_ t \in I \text{ for } j = 1, \ldots , m \]

and

\[ i'_ t = r'_ t - \sum \nolimits _ j c_ j r_{jt} \in I' \text{ for all }t \]

Note that this also implies that $r'_0 = \sum _ t i'_ t f_ t + \sum _ j c_ j r_{j0}$. Then we have

\begin{align*} \delta '_ n(\sum \nolimits _ t r'_ t x_ t - r'_0) & = \delta '_ n( \sum \nolimits _ t i'_ t x_ t + \sum \nolimits _{t, j} c_ j r_{jt} x_ t - \sum \nolimits _ t i'_ t f_ t - \sum \nolimits _ j c_ j r_{j0}) \\ & = \delta '_ n( \sum \nolimits _ t i'_ t(x_ t - f_ t) + \sum \nolimits _ j c_ j (\sum \nolimits _ t r_{jt} x_ t - r_{j0})) \end{align*}

Since $\delta _ n(a + b) = \sum _{m = 0, \ldots , n} \delta _ m(a) \delta _{n - m}(b)$ and since $\delta _ m(\sum i'_ t(x_ t - f_ t))$ is in the ideal generated by $x_ t - f_ t \in K'$ for $m > 0$, it suffices to prove that $\delta _ n(\sum c_ j (\sum r_{jt} x_ t - r_{j0}))$ is in $K'$. For this we use

\[ \delta _ n(\sum \nolimits _ j c_ j (\sum \nolimits _ t r_{jt} x_ t - r_{j0})) = \sum c_1^{n_1} \ldots c_ m^{n_ m} \delta _{n_1}(\sum r_{1t} x_ t - r_{10}) \ldots \delta _{n_ m}(\sum r_{mt} x_ t - r_{m0}) \]

where the sum is over $n_1 + \ldots + n_ m = n$. This proves what we want. $\square$

Comments (3)

Comment #4286 by Pavel Čoupek on

1) It seems that the first paragraph of the proof is redundant: the rest of the proof computes that a presentation of agrees with that of given by Lemma 07HC.

2) After using the equational criterion for flatness, the equality should a priori hold only in , so where As a result, one gets , and the result of the computation at the end of the proof should contain some terms of the form This should be still OK, since comes from when .

Comment #4450 by on

Good catch. Luckily it seems that in my paper it is correct and does the thing you suggest. Phew. Anyway, I have fixed it exactly as how you suggested! See here.

There are also:

  • 4 comment(s) on Section 60.2: Divided power envelope

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 07HE. Beware of the difference between the letter 'O' and the digit '0'.



View Lemma 60.2.7 as pdf