Here is a fiendishly clever lemma.
Proof.
We choose a presentation
\bar C = R[x_1, \ldots , x_ n]/(f_1, \ldots , f_ m)
We also denote I = (f_1, \ldots , f_ m) and \bar I the image of I in R/\pi ^2R[x_1, \ldots , x_ n]. Since R is Noetherian, so is \bar C. Hence the smooth locus of R/\pi ^2 R \to \bar C is quasi-compact, see Topology, Lemma 5.9.2. Applying Lemma 16.2.2 we may choose a finite list of elements a_1, \ldots , a_ r \in R[x_1, \ldots , x_ n] such that
the union of the open subspaces \mathop{\mathrm{Spec}}(\bar C_{a_ k}) \subset \mathop{\mathrm{Spec}}(\bar C) cover the smooth locus of R/\pi ^2 R \to \bar C, and
for each k = 1, \ldots , r there exists a finite subset E_ k \subset \{ 1, \ldots , m\} such that (\bar I/\bar I^2)_{a_ k} is freely generated by the classes of f_ j, j \in E_ k.
Set I_ k = (f_ j, j \in E_ k) \subset I and denote \bar I_ k the image of I_ k in R/\pi ^2R[x_1, \ldots , x_ n]. By (2) and Nakayama's lemma we see that (\bar I/\bar I_ k)_{a_ k} is annihilated by 1 + b'_ k for some b'_ k \in \bar I_{a_ k}. Suppose b'_ k is the image of b_ k/(a_ k)^ N for some b_ k \in I and some integer N. After replacing a_ k by a_ kb_ k we get
(\bar I_ k)_{a_ k} = (\bar I)_{a_ k}.
Thus, after possibly replacing a_ k by a high power, we may write
a_ k f_\ell = \sum \nolimits _{j \in E_ k} h_{k, \ell }^ jf_ j + \pi ^2 g_{k, \ell }
for any \ell \in \{ 1, \ldots , m\} and some h_{i, \ell }^ j, g_{i, \ell } \in R[x_1, \ldots , x_ n]. If \ell \in E_ k we choose h_{k, \ell }^ j = a_ k\delta _{\ell , j} (Kronecker delta) and g_{k, \ell } = 0. Set
D = R[x_1, \ldots , x_ n, z_1, \ldots , z_ m]/ (f_ j - \pi z_ j, p_{k, \ell }).
Here j \in \{ 1, \ldots , m\} , k \in \{ 1, \ldots , r\} , \ell \in \{ 1, \ldots , m\} , and
p_{k, \ell } = a_ k z_\ell - \sum \nolimits _{j \in E_ k} h_{k, \ell }^ j z_ j - \pi g_{k, \ell }.
Note that for \ell \in E_ k we have p_{k, \ell } = 0 by our choices above.
The map R \to D is the given one. Say \bar C \to \Lambda /\pi ^2\Lambda maps x_ i to the class of \lambda _ i modulo \pi ^2. For an element f \in R[x_1, \ldots , x_ n] we denote f(\lambda ) \in \Lambda the result of substituting \lambda _ i for x_ i. Then we know that f_ j(\lambda ) = \pi ^2 \mu _ j for some \mu _ j \in \Lambda . Define D \to \Lambda by the rules x_ i \mapsto \lambda _ i and z_ j \mapsto \pi \mu _ j. This is well defined because
\begin{align*} p_{k, \ell } & \mapsto a_ k(\lambda ) \pi \mu _\ell - \sum \nolimits _{j \in E_ k} h_{k, \ell }^ j(\lambda ) \pi \mu _ j - \pi g_{k, \ell }(\lambda ) \\ & = \pi \left(a_ k(\lambda ) \mu _\ell - \sum \nolimits _{j \in E_ k} h_{k, \ell }^ j(\lambda ) \mu _ j - g_{k, \ell }(\lambda )\right) \end{align*}
Substituting x_ i = \lambda _ i in (4) above we see that the expression inside the brackets is annihilated by \pi ^2, hence it is annihilated by \pi as we have assumed \text{Ann}_\Lambda (\pi ) = \text{Ann}_\Lambda (\pi ^2). The map \bar C \to D/\pi D is determined by x_ i \mapsto x_ i (clearly well defined). Thus we are done if we can prove (b), (c), and (d).
Using (4) we obtain the following key equality
\begin{align*} \pi p_{k, \ell } & = \pi a_ k z_\ell - \sum \nolimits _{j \in E_ k} \pi h_{k, \ell }^ jz_ j - \pi ^2 g_{k, \ell } \\ & = - a_ k (f_\ell - \pi z_\ell ) + a_ k f_\ell + \sum \nolimits _{j \in E_ k} h_{k, \ell }^ j (f_ j - \pi z_ j) - \sum \nolimits _{j \in E_ k} h_{k, \ell }^ j f_ j - \pi ^2 g_{k, \ell } \\ & = -a_ k(f_\ell - \pi z_\ell ) + \sum \nolimits _{j \in E_ k} h_{k, \ell }^ j(f_ j - \pi z_ j) \end{align*}
The end result is an element of the ideal generated by f_ j - \pi z_ j. In particular, we see that D[1/\pi ] is isomorphic to R[1/\pi ][x_1, \ldots , x_ n, z_1, \ldots , z_ m]/(f_ j - \pi z_ j) which is isomorphic to R[1/\pi ][x_1, \ldots , x_ n] hence smooth over R. This proves (b).
For fixed k \in \{ 1, \ldots , r\} consider the ring
D_ k = R[x_1, \ldots , x_ n, z_1, \ldots , z_ m]/ (f_ j - \pi z_ j, j \in E_ k, p_{k, \ell })
The number of equations is m = |E_ k| + (m - |E_ k|) as p_{k, \ell } is zero if \ell \in E_ k. Also, note that
\begin{align*} (D_ k/\pi D_ k)_{a_ k} & = R/\pi R[x_1, \ldots , x_ n, 1/a_ k, z_1, \ldots , z_ m]/ (f_ j, j \in E_ k, p_{k, \ell }) \\ & = (\bar C/\pi \bar C)_{a_ k}[z_1, \ldots , z_ m]/ (a_ kz_\ell - \sum \nolimits _{j \in E_ k} h_{k, \ell }^ j z_ j) \\ & \cong (\bar C/\pi \bar C)_{a_ k}[z_ j, j \in E_ k] \end{align*}
In particular (D_ k/\pi D_ k)_{a_ k} is smooth over (\bar C/\pi \bar C)_{a_ k}. By our choice of a_ k we have that (\bar C/\pi \bar C)_{a_ k} is smooth over R/\pi R of relative dimension n - |E_ k|, see (2). Hence for a prime \mathfrak q_ k \subset D_ k containing \pi and lying over \mathop{\mathrm{Spec}}(\bar C_{a_ k}) the fibre ring of R \to D_ k is smooth at \mathfrak q_ k of dimension n. Thus R \to D_ k is syntomic at \mathfrak q_ k by our count of the number of equations above, see Algebra, Lemma 10.136.10. Hence R \to D_ k is smooth at \mathfrak q_ k, see Algebra, Lemma 10.137.17.
To finish the proof, let \mathfrak q \subset D be a prime containing \pi lying over a prime where R/\pi ^2 R \to \bar C is smooth. Then a_ k \not\in \mathfrak q for some k by (1). We will show that the surjection D_ k \to D induces an isomorphism on local rings at \mathfrak q. Since we know that the ring maps \bar C/\pi \bar C \to D_ k/\pi D_ k and R \to D_ k are smooth at the corresponding prime \mathfrak q_ k by the preceding paragraph this will prove (c) and (d) and thus finish the proof.
First, note that for any \ell the equation \pi p_{k, \ell } = -a_ k(f_\ell - \pi z_\ell ) + \sum _{j \in E_ k} h_{k, \ell }^ j (f_ j - \pi z_ j) proved above shows that f_\ell - \pi z_\ell maps to zero in (D_ k)_{a_ k} and in particular in (D_ k)_{\mathfrak q_ k}. The relations (4) imply that a_ k f_\ell = \sum _{j \in E_ k} h_{k, \ell }^ j f_ j in I/I^2. Since (\bar I_ k/\bar I_ k^2)_{a_ k} is free on f_ j, j \in E_ k we see that
a_{k'} h_{k, \ell }^ j - \sum \nolimits _{j' \in E_{k'}} h_{k', \ell }^{j'} h_{k, j'}^ j
is zero in \bar C_{a_ k} for every k, k', \ell and j \in E_ k. Hence we can find a large integer N such that
a_ k^ N\left( a_{k'} h_{k, \ell }^ j - \sum \nolimits _{j' \in E_{k'}} h_{k', \ell }^{j'} h_{k, j'}^ j \right)
is in I_ k + \pi ^2R[x_1, \ldots , x_ n]. Computing modulo \pi we have
\begin{align*} & a_ kp_{k', \ell } - a_{k'}p_{k, \ell } + \sum h_{k', \ell }^{j'} p_{k, j'} \\ & = - a_ k \sum h_{k', \ell }^{j'} z_{j'} + a_{k'} \sum h_{k, \ell }^ j z_ j + \sum h_{k', \ell }^{j'} a_ k z_{j'} - \sum \sum h_{k', \ell }^{j'} h_{k, j'}^ j z_ j \\ & = \sum \left( a_{k'} h_{k, \ell }^ j - \sum h_{k', \ell }^{j'} h_{k, j'}^ j \right) z_ j \end{align*}
with Einstein summation convention. Combining with the above we see a_ k^{N + 1} p_{k', \ell } is contained in the ideal generated by I_ k and \pi in R[x_1, \ldots , x_ n, z_1, \ldots , z_ m]. Thus p_{k', \ell } maps into \pi (D_ k)_{a_ k}. On the other hand, the equation
\pi p_{k', \ell } = -a_{k'} (f_\ell - \pi z_\ell ) + \sum \nolimits _{j' \in E_{k'}} h_{k', \ell }^{j'}(f_{j'} - \pi z_{j'})
shows that \pi p_{k', \ell } is zero in (D_ k)_{a_ k}. Since we have assumed that \text{Ann}_ R(\pi ) = \text{Ann}_ R(\pi ^2) and since (D_ k)_{\mathfrak q_ k} is smooth hence flat over R we see that \text{Ann}_{(D_ k)_{\mathfrak q_ k}}(\pi ) = \text{Ann}_{(D_ k)_{\mathfrak q_ k}}(\pi ^2). We conclude that p_{k', \ell } maps to zero as well, hence D_{\mathfrak q} = (D_ k)_{\mathfrak q_ k} and we win.
\square
Comments (0)