|
|
@@ -0,0 +1,114 @@ |
|
|
|
|
|
\documentclass{lecture} |
|
|
|
|
|
|
|
|
|
|
|
\begin{document} |
|
|
|
|
|
|
|
|
|
|
|
\section{Real closures} |
|
|
|
|
|
|
|
|
|
|
|
\begin{satz} |
|
|
|
|
|
Let $k$ be a real field. Then there exists a real-closed |
|
|
|
|
|
algebraic orderable extension $k^{r}$ of $k$. |
|
|
|
|
|
\label{satz:existence-alg-closure} |
|
|
|
|
|
\end{satz} |
|
|
|
|
|
|
|
|
|
|
|
\begin{proof} |
|
|
|
|
|
Let $\overline{k}$ be an algebraic closure of $k$ and $E$ be the set of intermediate |
|
|
|
|
|
extensions $k \subseteq L \subseteq \overline{k}$ such that $L$ is real and algebraic over $k$. |
|
|
|
|
|
$E \neq \emptyset$ since $k \in E$. Define $L_1 < L_2$ on $E$ if and only if |
|
|
|
|
|
$L_1 \subseteq L_2$ and $L_2 / L_1$ is ordered, i.e. |
|
|
|
|
|
the order relation on $L_1$ coincides with the on induced by $L_2$. |
|
|
|
|
|
Then |
|
|
|
|
|
every totally ordered familiy $(E_i)_{i \in I}$ has an upper bound, namely |
|
|
|
|
|
$\bigcup_{i \in I} E_i$. By Zorn, $E$ has a maximal element, which we |
|
|
|
|
|
denote by $k^{r}$ and which is an algebraic extension of $k$. Such |
|
|
|
|
|
a $k^{r}$ is real-closed, because otherwise it would admit a proper real |
|
|
|
|
|
algebraic extension contradicting the maximality of $k^{r}$ as a real algebraic extension of $k$. |
|
|
|
|
|
\end{proof} |
|
|
|
|
|
|
|
|
|
|
|
\begin{definition}[] |
|
|
|
|
|
A real-closed real algebraic extension of a real field $k$ is called |
|
|
|
|
|
a \emph{real closure} of $k$. |
|
|
|
|
|
\end{definition} |
|
|
|
|
|
|
|
|
|
|
|
\begin{bem} |
|
|
|
|
|
By the construction in the proof of \ref{satz:existence-alg-closure}, |
|
|
|
|
|
a real closure of a real field $k$ can be chosen as a subfield |
|
|
|
|
|
$k^{r}$ of an algebraic closure of $\overline{k}$. |
|
|
|
|
|
Since $k^{r}[i]$ is algebraically closed and algebraic over $k^{r}$, so also over $k$, |
|
|
|
|
|
it follows $k^{r}[i] = \overline{k}$. |
|
|
|
|
|
\end{bem} |
|
|
|
|
|
|
|
|
|
|
|
\begin{satz} |
|
|
|
|
|
Let $k$ be a real field and $L$ be a real-closed extension of $k$. Let |
|
|
|
|
|
$\overline{k}^{L}$ be the relative algebraic closure of $k$ in $L$, i.e. |
|
|
|
|
|
\[ |
|
|
|
|
|
\overline{k}^{L} = \{ x \in L \mid x \text{ algebraic over } k\} |
|
|
|
|
|
.\] Then $\overline{k}^{L}$ is a real closure of $k$. |
|
|
|
|
|
\end{satz} |
|
|
|
|
|
|
|
|
|
|
|
\begin{proof} |
|
|
|
|
|
It is immediate that $\overline{k}^{L}$ is a real algebraic extension of $k$. Let |
|
|
|
|
|
$x \in \overline{k}^{L}$. Then $x$ or $-x$ is a square in $L$, since |
|
|
|
|
|
$L$ is real-closed. Without loss of generality, assume that |
|
|
|
|
|
$x \in L^{[2]}$. Then $t^2 - x \in \overline{k}^{L}[t]$ |
|
|
|
|
|
has a root in $L$. Since this root is algebraic over $\overline{k}^{L}$, hence over $k$, |
|
|
|
|
|
it belongs to $\overline{k}^{L}$. Thus $x$ is in fact a square in $\overline{k}^{L}$. By |
|
|
|
|
|
the same argument every polynomial of odd degree has a root in $\overline{k}^{L}$. |
|
|
|
|
|
\end{proof} |
|
|
|
|
|
|
|
|
|
|
|
\begin{bsp}[] |
|
|
|
|
|
\begin{enumerate}[(i)] |
|
|
|
|
|
\item $\overline{\Q}^{\R} = \overline{\Q}^{\mathbb{C}} \cap \R$ |
|
|
|
|
|
is a real closure of $\Q$. In particular, $\overline{\Q}^{\mathbb{C}} |
|
|
|
|
|
= \overline{\Q}^{\R}[i]$ as subfields of $\mathbb{C}$. |
|
|
|
|
|
\item Consider the real field $k = \R(t)$ and the real-closed extension |
|
|
|
|
|
\begin{salign*} |
|
|
|
|
|
\widehat{\R(t)} = |
|
|
|
|
|
\bigcup_{q > 0} \R((t ^{t/q})) |
|
|
|
|
|
.\end{salign*} Then the subfield |
|
|
|
|
|
$\overline{\R(t)}^{\widehat{\R(t)}}$, consisting of all those real |
|
|
|
|
|
Puiseux series that are algebraic over $\R(t)$, is a real closure of $\R(t)$. |
|
|
|
|
|
|
|
|
|
|
|
The field of real Puiseux series itself is a real closure of the field $\R((t))$ |
|
|
|
|
|
of real formal Laurent series. |
|
|
|
|
|
\end{enumerate} |
|
|
|
|
|
\end{bsp} |
|
|
|
|
|
|
|
|
|
|
|
Real-closed fields $L$ admit a canonical structure of ordered field, where $x \ge 0$ |
|
|
|
|
|
in $L$, if and only if $x$ is a square. In particular, |
|
|
|
|
|
if $k$ is a real field and $k^{r}$ is a real closure of $k$, then |
|
|
|
|
|
$k$ inherits an ordering from $k^{r}$. However, different real closures may induce |
|
|
|
|
|
different orderings on $k$, as the next example shows. |
|
|
|
|
|
|
|
|
|
|
|
\begin{bsp}[] |
|
|
|
|
|
Let $k = \Q(t)$. This is a real field, since $\Q$ is real. Since $\pi$ |
|
|
|
|
|
is transcendental over $\Q$, we can embed $\Q(t)$ in $\R$ by sending $t$ to $\pi$. |
|
|
|
|
|
\[ |
|
|
|
|
|
i_1\colon \Q(t) \xhookrightarrow{\simeq} \Q(\pi) \subseteq \R |
|
|
|
|
|
.\] Since $\R$ is real-closed, the relative algebraic closure |
|
|
|
|
|
$i_1(\Q(t))^{\R}$ is a real closure of $i_1(\Q(t))$. |
|
|
|
|
|
|
|
|
|
|
|
We can also embed $\Q(t)$ in the field $\widehat{\R(t)}$ of real Puiseux series via |
|
|
|
|
|
a homomorphism $i_2$ and then |
|
|
|
|
|
$\overline{i_2(\Q(t))}^{\widehat{\R(t)}}$ is a real closure of $i_2(\Q(t))$. |
|
|
|
|
|
However, the ordering on $\overline{i_1(\Q(t))}^{\R}$ |
|
|
|
|
|
is Archimedean, because it is a subfield of $\R$, |
|
|
|
|
|
while the ordering on $\overline{i_2(\Q(t))}^{\widehat{\R(t)}}$ |
|
|
|
|
|
is not Archimedean (it contains infinitesimal elements, such as $t$ for instance). |
|
|
|
|
|
|
|
|
|
|
|
The fields $\overline{i_1(\Q(t))}^{\R}$ |
|
|
|
|
|
and $\overline{i_2(\Q(t))}^{\widehat{\R(t)}}$ cannot be isomorphic as fields. |
|
|
|
|
|
Indeed, when two real-closed fields $L_1, L_2$ are isomorphic as fields, |
|
|
|
|
|
then they are isomorphic as ordered fields, since positivity on a real |
|
|
|
|
|
closed field is defined by the condition of being a square, which is preserved |
|
|
|
|
|
under isomorphisms of fields. |
|
|
|
|
|
\end{bsp} |
|
|
|
|
|
|
|
|
|
|
|
The next result will be proved later on. |
|
|
|
|
|
|
|
|
|
|
|
\begin{lemma}[] |
|
|
|
|
|
Let $(k, \le )$ be an ordered field and $P \in k[t]$ be an irreducible polynomial. |
|
|
|
|
|
Let $L_1, L_2$ be real-closed extensions of $k$ that are compatible with the ordering of $k$. |
|
|
|
|
|
Then $P$ has the same number of roots in $L_1$ as in $L_2$. |
|
|
|
|
|
\end{lemma} |
|
|
|
|
|
|
|
|
|
|
|
\end{document} |