next up previous
Next: About this document ...

$ \mathbb{Z}_p$ , $ \mathbb{Q}_p$ , and the ring of Witt vectors

No.10:

% latex2html id marker 802
\fbox{The ring of Witt vectors when $A$\ is a ring of characteristic $p\neq 0$.}

DEFINITION 10.1   Let $ A$ be a commutative ring. For any $ a\in A$ , we denote by $ [a]$ the element of $ \mathcal W_1(A)$ defined as follows:

$\displaystyle [a]=(1-a T)
$

We call $ [a]$ the Teichmüller lift'' of $ a$

LEMMA 10.2   Let $ A$ be a commutative ring. Then:
  1. $ (\mathcal W_1(A),\boxplus,\boxtimes)$ is a commutative ring with the zero element $ [0]$ and the unity $ [1]$ .

  2. For any $ a,b \in A$ , we have

    $\displaystyle [a]\boxtimes [b]=[a b]
$

% latex2html id marker 1130
$ \qedsymbol$

PROPOSITION 10.3   Let $ p$ be a prime number. Let $ A$ be a ring of characteristic $ p$ . Then:
  1. If $ n$ is a positive integer which is not divisible by $ p$ , then $ n$ is invertible in $ \mathcal W_1(A)$ . To be more precise,

    $\displaystyle \frac{1}{n}\boxdot [1]= (1-T)^{\frac{1}{n}}=
1+\sum_{j=1}^\infty \binom{\frac{1}{n}}{j} (-T)^j.
$

  2. $ p \boxdot : \mathcal W_1(A)\to \mathcal W_1(A) $ is an injection.
  3. For any positive integer $ n$ which is not divisible by $ p$ , we define

    $\displaystyle e_n=\frac{1}{n}\boxdot (1-T^n).
$

    as an element of $ \mathcal W_1(A)$ .
  4. For any positive integer $ n$ , $ e_n$ is an idempotent. (That means, $ e_n^{\boxtimes 2}=e_n$ .)
  5. If $ n\vert m$ , then % latex2html id marker 947
$ e_n \succeq e_m$ in the order of idempotents. That means, $ e_n \boxtimes e_m=e_m$ .

PROOF.. (1) follows from the next lemma. % latex2html id marker 951
$ \qedsymbol$

LEMMA 10.4   Let $ n$ be a positive integer. Let $ k$ be a non negative integer. Then we have always

$\displaystyle \binom{\frac{1}{n}}{k}\in \mathbb{Z}\left[\frac{1}{n}\right].
$

PROOF..

      $\displaystyle \binom{\frac{1}{n}}{k}\in \mathbb{Z}\left[\frac{1}{n}\right]$
    $\displaystyle =$ $\displaystyle \frac{\frac{1}{n}(\frac{1}{n}-1)\cdots (\frac{1}{n}-(k-1))}{k!}$
    $\displaystyle =$ $\displaystyle \frac{1}{n^k} \frac{(1(1-n)(1-2n)\dots (1-(k-1)n)}{k!}$

So the result follows from the next sublemma. % latex2html id marker 965
$ \qedsymbol$

SUBLEMMA 10.5   Let $ n$ be a positive integer. Let $ k$ be a non negative integer. Let $ \{a_j\}_{j=1}^k\subset \mathbb{Z}$ be an arithmetic progression of common difference $ n$ . Then:
  1. For any positive integer $ m$ which is relatively prime to $ n$ , we have

    % latex2html id marker 990
$\displaystyle \char93 \{j;\ m \vert a_j\ \} \geq \left\lfloor \frac{k}{m} \right\rfloor
$

  2. For any prime $ p$ which does not divide $ n$ , let us define

    $\displaystyle c_{k,p}=
\sum_{i=1}^\infty \lfloor \frac{k}{p^i}\rfloor
$

    (which is evidently a finite sum in practice.) Then

    $\displaystyle p^{c_{k,p}} \vert \prod_{j=1}^k a_j
$

  3. % latex2html id marker 1000
$\displaystyle p^{c_{k,p}}\vert k!,\qquad p^{c_{k,p}+1} \not \vert k!
$

  4. $\displaystyle \frac{\prod_{j=1}^k a_j }{k!} \in \mathbb{Z}_{(p)}
$

PROOF.. (1) Let us put $ t=\lfloor\frac{k}{m}\rfloor$ . Then we divide the set of first $ kt$ -terms of the sequence $ \{a_j\}$ into disjoint sets in the following way.

      $\displaystyle S_0=\{a_1,a_2,\dots, a_m\},$
      $\displaystyle S_1=\{a_{m+1},a_{m+2},a_{m+m}\},$
      $\displaystyle S_2=\{a_{2m+1},a_{2m+2},a_{2m+m}\},$
      $\displaystyle \dots$
      $\displaystyle S_{t-1}=\{a_{(t-1)m+1},a_{(t-1)m+2},\dots, a_{(t-1)m+m}\}$

Since $ m$ is coprime to $ n$ , we see that each of the $ S_u$ gives a complete representative of $ \mathbb{Z}/n\mathbb{Z}$ .

(2): Apply (1) to the cases where $ m=p,p^2,p^3,\dots$ and count the powers of $ p$ which appear in $ \prod a_j$ .

(3): Easy. (4) is a direct consequence of (2),(3). % latex2html id marker 1004
$ \qedsymbol$

PROPOSITION 10.6   Let $ p$ be a prime. Let $ A$ be an integral domain of characteristic $ p$ . Let us define an idempotent $ f$ of $ \mathcal W_1(A)$ as follows.

$\displaystyle f=
\bigvee
_{\substack{
n>1\\
p\not \vert n
}}
e_n
(=[1]\boxmi...
...{^\boxtimes }
\prod_
{\substack
{p \not \vert n\\
n>1
}}
([1]\boxminus e_n))
$

Then $ f$ defines a direct product decomposition

$\displaystyle \mathcal W_1(A)
\cong
\left (
f \boxtimes \mathcal W_1(A)
\right )
\times
\left(
(1\boxminus f)\boxtimes \mathcal W_1(A)
\right).
$

Furthermore, the factor algebra $ (1\boxminus f)\boxtimes \mathcal W_1(A)$ is isomorphic to the ring $ \mathcal W^{(p)}(A)$ of $ p$ -adic Witt vectors.

The following proposition tells us the importance of the ring of $ p$ -adic Witt vectors.

PROPOSITION 10.7   Let $ p$ be a prime. Let $ A$ be a commutative ring of characteristic $ p$ . For each positive integer $ k$ which is not divisible by $ p$ , let us define an idempotent $ f_k$ of $ \mathcal W_1(A)$ as follows.

$\displaystyle f_k=\bigvee_{
\substack{p\not \vert n  n >1}} e_{k n }
(=e_k\bo...
...\boxtimes }\prod
_{\substack{p\not \vert n  n >1}}
(e_k\boxminus e_{k n}))
$

Then $ f_k$ defines a direct product decomposition

$\displaystyle e_k\mathcal W_1(A)
\cong
\left (
f_k \boxtimes \mathcal W_1(A)
\right )
\times
\left(
(1\boxminus f_k)\boxtimes \mathcal W_1(A)
\right).
$

Furthermore, the factor algebra $ (1\boxminus f_k)\boxtimes \mathcal W_1(A)$ is isomorphic to the ring $ \mathcal W^{(p)}(A)$ of $ p$ -adic Witt vectors. Thus we have a direct product decomposition

$\displaystyle \mathcal W_1(A) \cong \mathcal W^{(p)}(A)^{\mathbb{N}}.
$

To understand the mechanism which appears in the proposition above, it would be better to prove the following

LEMMA 10.8   Let $ p$ be a prime number. Let $ A$ be a ring of characteristic $ p$ . Then for any $ n$ which is not divisible by $ p$ , a map

$\displaystyle \frac{1}{n} \boxdot V_n :
(\mathcal W_1(A),\boxplus,\boxtimes)
\to
(\mathcal W_1(A),\boxplus,\boxtimes)
$

is a ring homomorphism. Its image is equal to the range of the idempotent $ e_n$ . That means,

$\displaystyle \operatorname{Image}(\frac{1}{n}\boxdot V_n)
=e_n \boxtimes \mathcal W_1(A)
=\{ \sideset{}{^\boxplus}\sum_j (1-y_j T^{n j}); y_j \in A\}.
$

PROOF.. $ V_n$ is already shown to be additive. The following calculation shows that $ \frac{1}{n} \cdot V_n$ preserves the $ \boxtimes$ -multiplication.

      $\displaystyle (\frac{1}{n}\boxdot V_n(1-x T^a)) \boxtimes (\frac{1}{n}\boxdot V_n(1-y T^b))$
    $\displaystyle =$ $\displaystyle (\frac{1}{n}\boxdot (1-x T^{a n})) \boxtimes (\frac{1}{n}\boxdot (1-y T^{b n}))$
    $\displaystyle =$ $\displaystyle \frac{1}{n^2}\boxdot(1-x^{m/a}y^{m/b} T^{nm} )^d$
    $\displaystyle =$ $\displaystyle \frac{1}{n} \boxdot((1-x T^a )\boxtimes (1-y T^b))$

% latex2html id marker 1115
$ \qedsymbol$

In preparing from No.7 to No.10 of this lecture, the following reference (especially its appendix) has been useful:

http://www.math.upenn.edu/~chai/course_notes/cartier_12_2004.pdf ARRAY(0x8b0c098)


next up previous
Next: About this document ...
2008-07-08