s21-algebra-1/inputs/algnumbertheory.tex
2022-09-23 13:03:54 +02:00

153 lines
7.6 KiB
TeX

% Alg 1 2022 Blomer
\section{Introduction to algebraic number theory}
\subsection{Number Fields and rings of integers}
\begin{definition}[Number field]
A \vocab{number field} $K$ is a finite (hence algebraic) extension of $\Q$.
Its \vocab{ring of integers} $\cO_k$ is the integral closure of $\Z$ in $K$.
\end{definition}
\begin{remark}
\begin{enumerate}[(a)]
\item It is not obvious that this is a ring, but we have shown this in % TODO REF
\item For $K = \Q(\i)$ we have $\cO_K = \Z[\i]$ by % Blomer (3.5) TODO
\item Every $x \in K$ of the form $\frac{y}{n}$ with $y \in \cO_K$, $n \in \Z \setminus \{0\}$ since $\exists f \in \Z[X]: f(x) = 0$ with leading coefficient $0 \neq c \in \Z$.
Then $c^{n-1}f(x) = 0$ implies that $cx$ is integral over $\Z$.
\item An element $x \in K$ is in $\cO_k$ if and only if its minimal polynomial is in $\Z[X]$.% see Blomer 3.5 TODO
\item Since $\char(\Q) = 0$, every extension is separable.
\end{enumerate}
\end{remark}
\begin{definition} % and Lemma TODO
For $a \in K$consider the $\Q$-vectorspace-homomorphism $T_a : K \to K, x \mapsto ax$ and define the \vocab{norm} and \vocab{trace}
of $a$ by
\[
\Tr(a) \coloneqq \Tr(T_a)
\]
and \[
\Nr(a) = \det(T_a)
\]
\end{definition}
\begin{lemma}
$\Tr : k \to \Q$ and $\Nr : K^\ast \to \Q^\ast$ are homomorphisms.
\end{lemma}
\begin{proof}
$\Nr(T_a)$ and $\Tr(T_a)$ are coefficients of the characteristic polynomial of $T_a$,
hence in $\Q$.
Since $T_a + T_b = T_{a+b}$ and $T_{ab} = T_a T_b$, we see that $\Tr$ and $\Nr$ are homomorphisms.
\end{proof}
\begin{theorem}
\begin{enumerate}[(a)]
\item Let $K$ be a number field of degree $n$ and denote by $\sigma$ the $n$ distinct homomorphisms
from $K$ to $\overline{\Q}$.
Then
\[
\Tr(a) = \sum_{\sigma} \sigma a, \Nr(a) = \prod_{\sigma} \sigma a
\]
\item If $a \in \cO_K$ then $\Tr(a), \Nr(a) \in \Z$
\end{enumerate}
\end{theorem}
\begin{proof}
\begin{enumerate}[(a)]
\item
If $a $ is of degree $m$ over $\Q$, then $(1, a, a^2, \ldots, a^{m-1})$ is a $\Q$- basis of $\Q(a) / \Q$.
Let $f_a(X) = X^n + c_1 X^{m-1} + \ldots+ c_m \in \Q[X]$ be the minimal polynomial of $a$.
Let $n = md$ and $(b_1, \ldots, b_d)$ a basis of $K / \Q(a)$.
Then $(b_1, b_1a, b_1a^2, \ldots, b_1a^{m-1}, \ldots, b_d, b_da, \ldots, b_da^{m-1})$ is a basis of $K / \Q$.
Then the matrix of $T_a$ with respect to basis has block structure with blocks:
\[
\begin{pmatrix} & & & & -c_m \\ 1& & & & -c_{m-1} \\ & 1 & & -c_{m-2} \\ & & \ddots & \\ & & & 1 & -c_1 \end{pmatrix}
\]
The charactersitic polynomial is $f_a$ for every block, hence $f_a^d$ for the entire matrix.
On the set $\Hom(K, \overline{Q})$ we introduce an equivalence relation by $\sigma \sim \tau(a)$ if $\sigma(a) = \tau(a)$.
Then the set of equivalence classes is in bijection with $\Hom(\Q(a), \overline{\Q})$ via
\[
[\sigma] \mapsto \sigma \defon{\Q(a)}
\]
Since $\{\sigma(a) | \simga \in \Hom(\Q(a), \overline{\Q})\}$ is the set of roots of $f_a$, by Vietas theorem we obtain
$\Tr_K(a) = d\Tr_{\Q(a)}(T_a) = d \sum_{\sigma \in \Hom(\Q(a), \overline{\Q}} \sigma(a) = \sum_{\delta \in \Hom(K, \overline{\Q}} \sigma(a)$.
Basically the same proof works for $\Nr(a)$.
\item If $a$ is integral over $\Z$, then all $\simga(a)$ are integral over $\Z$ (since they are roots of the same poynomial), so $\Nr(a)$ and $\Tr(a) \in \Q$ are integral over $\Z$ by % TODO Blomer 3.3 and a)
and so $\Nr(a), \Tr(a) \in \Z$.
\end{enumerate}
\end{proof}
\begin{definiton}
Let $K$ be a humber field with $\Q = \basis(\alpha_1, \ldots, \alpha_n)$. We define the \vocab{discriminant}
\[
D(\alpha_1, \ldots, \alphan) = \det \left( \left( \delta_i \alpha_j \right)_{\substack{ \le j \le n \\ \delta_i \in \Hom(K, \overline{\Q})\} \right)^{2}
\]
\end{definition}
\begin{lemma}
\begin{enumerate}[(a)]
\item We hae $D(\alpha_1, \ldots, \alpha_n) = \det(\Tr(\alpha_i \alpha_j)_{i,j}) \in \Q$
\item The map $(x,y) \mapsto \Tr(xy)$ is a non-degenerate bilinear form
\item We have $D(\alpha_1, \ldots, \alpha_n) \neq 0$
\end{enumerate}
\end{lemma}
\begin{proof}
\begin{enumerate}[(a)]
\item We have $\Tr(\alpha_i \alpha_j) = \sum_\sigma \sigma(\alpha_i \alpha_j ) = \sum_\sigma \sigma(\alpha_i) \sigma(\alpha_j)$,
so $\left( \Tr(\alpha_i \alpha_j) \right) = (\sigma \alpha_i)^T_{\sigma, i} \cdot (\sigma \alpha_j)_{\sigma, j}$.
\item By additivity of the trace, it's clear that this is a bilinear form.
The associated symmetric matrix is $M = \Tr(\alpha_i \alpha_j)_{i,j}$ where $\{\alpha_i\}$ is a basis.
We can assume that $K = \Q(\theta)$, so $(1, \theta, \theta^2, .., \theta^{n-1})$ is a basis.
So $(\sigma \alpha_i)_{\sigma, i}$ is a Vandermonde matrix % TODO?
, hence it has non-vanishing determinant and so $\det M \neq 0$.
\item follows from (a) and (b).
\end{enumerate}
\end{proof}
\begin{theorem}
Every ideal $\{0\} \neq \mathcal a \subseteq \cO_K$ (in particular $\cO_k$ itself) is a free $\Z$-module of rank $n = [K : \Q]$.
In particular, every ideal of $\cO_K$ is finitely generated as a $\Z$-module, in particular as an $\cO_K$-module.
Hence $\cO_K$ is noetherian.
\end{theorem}
\begin{proof}
Let $\alpha_1, \ldots, \alpha_n$ be a basis of $K / \Q$. W.l.o.g.~ $\alpha_j \in \cO_K$, with discriminant $D \in \Z$.
Let $x = \sum a_j \alpha_j \in \cO_K$ for some $a_j \in \Q$.
Then $\alpha_i x = \sum \alpha_i \alpha_j \cdot a_j$.
This implies $\Z \ni \Tr(\alpha_i x) = \sum_j \Tr(\alpha_i \alpha_j) a_j$ for all $i$.
We have $\Tr(\alpha_i \alpha_j) \in \Z$. This is an $n \times n$ linear system of equations in $a_j$.
By Cramer's rule, $a_j \in \frac{1}{D} \Z$. Hence $Dx \in \sum \Z \alpha_j$, so $D \cO_K \subseteq \sum \Z \alpha_j$ is a free module as a submodule of a ree $\Z$-module (by % TODO Blomer 0.9
), obviously of rank $n$.
Now let $\{0\} \neq \mathcal{a} \subseteq \cO_k$ be an ideal.
This is again a $\Z$-submodule of $\cO_k$, hence a free module. If $0 \neq a \in \mathcal{a}$, then $a \cO_K \subseteq \mathcal{a}$,
hence the $\Z$-rank of $\mathcal{a}$ is $n$.
\end{proof}
\begin{example}
A $\Z$-basis of $\Z[\i]$ is $\{1,\i\}$ and the discriminant is
\[
\det \begin{pmatrix} 1 & \i \\ 1 & -\i \end{pmatrix} ^{2} = (-2\i)^{2} = -4
\]
\end{example}
\begin{corollary}
\begin{enumerate}[(a)]
\item A $\Z$-module basis of $\cO_K$ is called an \vocab{integral basis} of $K$. The discriminant with respect to this basis is \vocab{the discriminant} $D_K$ of $K$.
It is indpendent of the basis.
\item Every ideal $\{0\} \neq \mathcal a \subseteq \cO_K$ has finite index.
We denote by $N \mathcla a = (\cO_k : \mathcal a) = \# \cO_K / \mathcal a$, \vocab{the norm} of $\mathcal a$.
If $(\apha_1, \ldots, \alpha_n)$ is a $\Z$-basis of $\mathcal a$, denn $D(\alpha_1, \ldots, \alpha_n) = (N \mathcal a)^2 D_K$
\end{enumerate}
\end{corollary}
\begin{proof}
\begin{enumerate}[(a)]
\item A change of basis in $\Z$ corresponds to multiplying with a matrix of determinant $\pm 1$.
\item Since $\mathcal a$ is a submodule of $\cO_K$ of full rank, we can choose compatible bases
$(\alha_1, \ldots, \alpha_n) \subseteq \cO_K$, $(d_1 \alpha_1, \ldots, d_n \alpha_n) \subseteq \mathcal a$.
$(\cO_k, \mathcal a) = d_1 \cdot \ldots \cdot d_n$.
\end{enumerate}
\end{proof}