On $\mathbb R$, every archimedean absolute value $|\cdot|$, such that $(\mathbb R, |\cdot|)$ is complete, is equivalent to the usual absolute value defined for $x\in \mathbb R$ by: $$\mathbb |x|_{\mathbb R}= \begin{cases} x & \text{if } x\ge 0 \\ -x & \text{if } x\le 0 \end{cases}$$

Now, I am trying to show that $\forall x, |x|=\mathbb |x|^a_{\mathbb R}$, for some $a>0$. Using the same steps as in the proof of the Ostrowski Theorem for $\mathbb Q$ in the archimedean case, one knows that there exists an $a>0$ such that $|r|=\mathbb |r|^a_{\mathbb R}, \forall r\in \mathbb Q$. Now I want to show that this "$a$" does work for all reals, not just the rationals. So, for an arbitrary $x\in \mathbb R$, I was thinking to start with a sequence of rationals $(r_i)$, such that $r_i\rightarrow x$ with respect to "usual" absolute value. Now, since $|r|=\mathbb |r|^a_{\mathbb R}$, it means that $(r_i)$ is Cauchy w.r.t. $|\cdot|$, hence converges to some $t\in \mathbb R$, because of completeness asumption. Now I want to show that $t=x$, but I am not able to do that.

Note: I use the definition of absolute value from Neukirch, chapter 2.

  • $\begingroup$ might be a good idea to use the archimedean property $\endgroup$ – Mikhail Katz Nov 6 '16 at 14:59
  • $\begingroup$ Can you suggest me how exactly to do that? $\endgroup$ – user223794 Nov 6 '16 at 16:17
  • $\begingroup$ What I was pointing out is that the procedure you outlined works for all valuations, whether p-adic or archimedean, so it couldn't yield a solution. There is only one Euclidean valuation on Q so it must be the one yielding the real numbers as completion. $\endgroup$ – Mikhail Katz Nov 6 '16 at 16:20
  • $\begingroup$ @MikhailKatz: It is fair to say that the archimedean property is already used for showing that the restriction to $\mathbb{Q}$ is equivalent to the usual value. However, I do not see how to prove the statement without also using the order structure on $\mathbb{R}$. $\endgroup$ – Torsten Schoeneberg Jan 3 '18 at 1:03

It seems that we really have to make use of more extra structure on $\mathbb{R}$, namely, the usual order $<$.

So let $t \in \mathbb{R}$, wlog $t \ge 0$, and $r_i$ a sequence of rationals that converge to $t$ with respect to $|\cdot|$. By replacing $r_i$ with $2t-r_i$ if necessary, we can make sure that

(*) for all indices $i\in \Bbb{N}$, we have $r_i \le t$ for odd $i$ and $r_i\ge t$ for even $i$; in particular, $t$ lies between $r_i$ and $r_{i+1}$.

I think you already got the following, and it does not need (*): $r_i$ is a Cauchy sequence w.r.t. $|\cdot|$. Then the sequence of values $|r_i| = |r_i|_\Bbb{R}^a$ is a Cauchy sequence in $\mathbb{R}$ w.r.t. the usual value (! -- this is how the metric is defined), hence converges to some $x_0 \in \mathbb{R}_{\ge 0}$. So $|t| = x_0$ by continuity of the value. On the other hand, if $|r_i| = |r_i|_{\mathbb{R}}^a \to x_0$, then by continuity of $a$-exponentiation, $|r_i|_{\mathbb{R}} \to x_0^{1/a}$. In other words, with respect to the usual value, $r_i$ converges to $x := x_0^{1/a}$. So $|t| = x^a = |x|_\mathbb{R}^a$.

So what is left to show is $x=t$. But by (*), we have $t\in \bigcap_{n\in \Bbb{N}} [r_{2n-1}, r_{2n}]$, and because the sequence is Cauchy w.r.t the usual value, that intersection is actually a singleton, hence $\{t\}$. But one also sees (going to monotonous subsequences in the odd resp. even indices) that $x\in \bigcap_{n\in \Bbb{N}} [r_{2n-1}, r_{2n}]$. Hence $x=t$ and $|x|=|x|_\mathbb{R}^a$.

  • $\begingroup$ The only problem with this solution is the second sentence. How do we know that every real number $r$ can be written as a limit of rational numbers with respect to the (unknown) absolute value $|\cdot|$ (I don't mean the usual one $|\cdot|_{\mathbb R}$)? The rest of the solution would of course be fine. $\endgroup$ – user223794 Feb 5 '18 at 23:13
  • $\begingroup$ Good point. I admit I'm not sure how to circumvent this. Maybe a) one can start with rationals converging to $x$ wrt the usual value, as in your attempt, and then possibly reorder them similar to (*); or b) (approach very different from my answer) one could show that the completion of $\Bbb Q$ wrt the new value is isomorphic to $\Bbb R$, and by the universal property of completion we get an injection of that $\Bbb R$ into the usual $\Bbb R$ which then has to be the identity. $\endgroup$ – Torsten Schoeneberg Feb 7 '18 at 22:41

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.