Let $k$ be an ordered field of cofinality $cf(k)$ whose Cauchy $cf(k)$-sequences are convergent.$^{(1)}$

Let $\mathcal{R}(k)$ be its real closure. As an algebraic extension of $k$, it has the same cofinality.$^{(2)}$

I wonder if $\mathcal{R}(k)$ has the same Cauchy completeness property as $k$, and if so, if this is true of any algebraic (ordered field) extension of $k$?

$^{(1)}$A map $u: cf(k) \rightarrow k$ is Cauchy if $\forall \varepsilon >_k 0(\exists \alpha \in cf(k)(\forall \beta,\gamma > \alpha(|u(\beta) - u(\gamma)| <_k \varepsilon)))$. It is convergent to $l \in k$ if $\forall \varepsilon >_k 0(\exists \alpha \in cf(k)(\forall \beta > \alpha(|u(\beta) -l| <_k \varepsilon)))$.

$^{(2)}$ If $(K,\rho)$ is an extension of $k$ then an induction on $d$ yields: $\forall d \in \mathbb{N}^*(\forall x \in K(\exists P \in \rho(k)_d[X](P(x) \leq 1 \longrightarrow [x;+\infty[ \cap \rho(k) \neq \varnothing)))$ so if $K / k$ is algebraic, $\rho(k)$ is cofinal. Therefore, $cf(K) \leq cf(k)$.

Now, given $E,E'$ cofinal in $k,K$ of order type $cf(k),cf(K)$, the range of the map $f:\alpha \in cf(K) \mapsto \min(\{\rho(x) \in \rho(E) \ | \ \rho(x) > E'^{\alpha}\})$ is cofinal in $\rho(E)$ so $|f(cf(K))| = cf(k)$, and $cf(K) \geq cf(k)$.


Note that for a nonarchimedean ordered field, this notion of completeness is equivalent to its completeness as a valued field under the natural valuation induced by the order (that is, with valuation ring $\{a\in k:\exists n\in\mathbb N\,(-n\le a\le n)\}$).

Finite algebraic extensions of complete discretely valued fields are complete, but infinite extensions typically are not. The canonical example is the complete field of $p$-adic numbers $\mathbb Q_p$, whose algebraic closure is not complete. This is not an ordered field, but it’s easy to cook up ordered counterexamples as well.

For example, let $k$ be the field of formal Laurent series $\mathbb Q(\!(x)\!)$, ordered so that $x$ is positive infinitesimal, and $K$ be its real closure. Let $R\subseteq\mathbb R$ be the real closure of $\mathbb Q$. Since the field of Puiseux series $R\langle\!\langle x\rangle\!\rangle$ is real-closed, $K\subseteq R\langle\!\langle x\rangle\!\rangle$. In fact, since the residue field of a finite extension of $k$ is a finite extension of $\mathbb Q$, $K$ is even smaller: $$K\subseteq\bigcup\{F\langle\!\langle x\rangle\!\rangle:F\supseteq\mathbb Q\text{ finite ordered}\}\subsetneq R\langle\!\langle x\rangle\!\rangle.$$ On the other hand, since Puiseux series can be approximated by Puiseux polynomials over $R$, which are algebraic over $k$, the completion of $K$ includes all of $R\langle\!\langle x\rangle\!\rangle$, and by the same argument also the larger field of those Hahn series $f\in R[\![x^{\mathbb Q}]\!]$ whose support is either finite or tends to $+\infty$. So, $K$ is incomplete.

It might be worth mentioning that all becomes well-behaved if you swap the roles of algebra and topology: that is, when $k$ is a real-closed field, the completion of $k$ is also real-closed.

  • $\begingroup$ (I had been trying to prove that $\mathcal{R}(\widetilde{k}) = \widetilde{\mathcal{R}(k)}$ (where $\widetilde{.}$ denotes the completion or an ordered field) but I could only get dense embeddings $\mathcal{R}(k) \rightarrow \mathcal{R}(\widetilde{k})$ or $\mathcal{R}(\widetilde{k}) \rightarrow \widetilde{\mathcal{R}(k)}$, that's why I was interested in $\mathcal{R}(\widetilde{k})$ being complete.) $\endgroup$ – nombre Jun 28 '15 at 23:37
  • $\begingroup$ Right. Using these two operators, we can construct from any ordered field $k$ the four fields $\mathcal R(k)$, $\tilde k$, $\mathcal R(\tilde k)$, and $\widetilde{\mathcal R(k)}=\widetilde{\mathcal R(\tilde k)}=\mathcal R(\widetilde{\mathcal R(k)})$. There are natural embeddings $k\to\tilde k\to\mathcal R(\tilde k)$, $k\to\mathcal R(k)\to\mathcal R(\tilde k)$, and $\mathcal R(\tilde k)\to\widetilde{\mathcal R(k)}$, but in general all these may be strict extensions, and there are no other embeddings. $\endgroup$ – Emil Jeřábek 3.0 Jun 29 '15 at 9:05

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.