9
$\begingroup$

I've been reading some papers on Igusa zeta functions, and they seem to be implicitly using a "quantitative version" of Hensel's Lemma, which also asserts the number of lifts of a $\mathbb{Z}/p\mathbb{Z}$-point to a $\mathbb{Z}/p^k\mathbb{Z}$-point. I'm looking for something like the following:

Let $X$ be a smooth irreducible separated scheme of finite type of relative dimension $n$ over the ring of $p$-adic integers $\mathbb{Z}_p$. Then for any $k>0$ do we have $$\# X(\mathbb{Z}/p^k\mathbb{Z}) = p^{n(k-1)}\# X(\mathbb{Z}/p\mathbb{Z}) \quad ?$$

I'm looking for either a proof or a reference where I can find a proof.

$\endgroup$
  • 1
    $\begingroup$ Searching, as I'm sure you did. shows that Hensel's-lemma questions are quite popular here. I wonder if Wanderer's answer, on the implicit-function theorem, to PeteL.Clark's old question has any relevance? $\endgroup$ – LSpice Nov 27 '16 at 12:27
  • 3
    $\begingroup$ Check out N. Bourbaki, Commutative Algebra, III, S4.5, Corollary 3. I think the uniqueness in the corollary should be sufficient to prove that $\#X(\mathbb Z/p^k\mathbb Z)/\# X(\mathbb Z/p^{k-1}\mathbb Z)=p^n$, for all $k>1$, from which the statement follows (at least in the case where $X$ is affine). $\endgroup$ – kneidell Nov 27 '16 at 12:28
  • 2
    $\begingroup$ It's smooth, so work with the completed local rings and then it's easy, no? $\endgroup$ – znt Nov 27 '16 at 14:46
  • 7
    $\begingroup$ Smooth over Z_p means the completed local ring at a Z/pZ-point in the special fibre is Z_p[[X_1,X_2,...,X_n]] with n the rel dim of the morphism. Now count how many maps from that ring to Z/p^kZ. Any map from Spec(Z/p^k) to the scheme whose image is this point induces a map on the local rings and because Z/p^k is Artinian it will factor through the completion. Isn't this enough? Did I miss something? $\endgroup$ – znt Nov 27 '16 at 16:08
  • 4
    $\begingroup$ Roughly speaking, since $X$ is smooth, the subset of $X({\mathbb Q}_p)$ consisting of points reducing mod $p$ to some specified point in $X({\mathbb F}_p)$ is analytically isomorphic to the corresponding set for affine $n$-space, for which the statement is pretty clear. (This is just an attempt at giving a geometric version of znt's previous comment.) $\endgroup$ – Michael Stoll Nov 27 '16 at 18:45
9
$\begingroup$

Let me try a very explicit proof. Consider the problem of taking a solution modulo $p^k$ and lifting to solutions modulo $p^{k+1}$. For this, we may assume that $X \subset \mathbb{A}^m$ is smooth and affine of dimension $n$, given by polynomials $f_1, \dotsc, f_r \in \mathbb{Z}_p[x_1, \dotsc, x_m]$. Let $\mathbf{a} \in \mathbb{Z}_p^m$ satisfy $f_1(\mathbf{a}) \equiv \dotsb \equiv f_r(\mathbf{a}) \equiv 0 \pmod{p^k}$.

To find points modulo $p^{k+1}$ that lift $\mathbf{a}$, you put $\mathbf{x} = \mathbf{a} + p^k \mathbf{y}$ and look for vectors $\mathbf{y}$ modulo $p$ giving solutions to $f_i(\mathbf{x}) \equiv 0 \pmod{p^{k+1}}$. Taylor expansion gives

$(f_1(\mathbf{x}), \dotsc, f_r(\mathbf{x})) = (f_1(\mathbf{a}, \dotsc, f_r(\mathbf{a})) + p^k \mathbf{J}(\mathbf{a}) \mathbf{y} + O(p^{k+1})$

where $\mathbf{J}$ is the Jacobian matrix $(\partial(f_i)/\partial x_j)$. Dividing by $p^k$ and reducing modulo $p$ gives an inhomogeneous linear equation over $\mathbf{F}_p$:

$\mathbf{J}(\mathbf{a}) \mathbf{y} \equiv -p^{-k} (f_1(\mathbf{a}, \dotsc, f_r(\mathbf{a})) \pmod{p}$ .

That $X$ is smooth over $\mathbb{Z}_p$ at $\mathbf{a}$ implies that $\mathbf{J}(\mathbf{a})$, when reduced modulo $p$, has rank $m-n$. The space of solutions $\mathbf{y}$ is non-empty, by Hensel's Lemma, so is a linear space of dimension $n$ over $\mathbb{F}_p$. In this way you see that every solution modulo $p^k$ lifts to precisely $p^n$ solutions modulo $p^{k+1}$, and you get your formula by induction.

$\endgroup$
  • 1
    $\begingroup$ I guess you can make this a bit neater by using that X is a local complete intersection, so you can take $r=m-n$. Then the Jacobian is surjective, and you don't need to appeal to Hensel's Lemma to show existence of a solution. $\endgroup$ – Martin Bright Nov 28 '16 at 13:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.