Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Consider an ordered ring $R$. One can define a function $$|\cdot|: R \rightarrow R$$ by $$ |x| = \left\{ \begin{align} x &, x \geq 0 \\ -x&, x < 0 \end{align} \right. $$ If we set $d(x,y) = |x - y|$, it can be shown that $d$ satisfies:

  1. $d(x,y) \geq 0$ and $d(x,y) = 0$ iff $x=y$
  2. $d(x,y) = d(y,x)$
  3. $d(x,z) < d(x,y) + d(y,z)$

Here, $d$ looks very much like a metric except for the fact that $d$ actually produces an element of $R$, not an element $\mathbb{R}$. Otherwise, $(R, d)$ seems to have all of the features one would expect of a metric space. So, is there a way in which $(R, d)$ can be considered a metric space? Using this "metric", we can define Cauchy sequences and other items of interest but since the "metric" doesn't actually produce an element $\mathbb{R}$, anything that depends on the output of $d$ being a number would fail.

share|cite|improve this question
Problem: $|\cdot|: R \rightarrow R$. $d$ has $R \times R$ for domain. – Thomas Andrews Dec 7 '11 at 19:19
@ThomasAndrews Thanks for pointing that out; it's been fixed. – ItsNotObvious Dec 7 '11 at 19:21
While technically not a metric space, of course, since the distance isn't a real value, it has some of the properties of a metric space - specifically, a topology on $R.$ Is this topology the same as the order topology on $R$? Are the maps: $+:R\times R\rightarrow R$ and $\times:R\times R\rightarrow R$ continuous under this topology? – Thomas Andrews Dec 7 '11 at 19:29
@Jack: You can get such if and only if there is a $u\in R$ such that $0<u<1$. Then $0<ux<x$. – Thomas Andrews Dec 7 '11 at 20:18
(Whoops, the "only if" part of that last statement isn't true if $u$ is a zero divisor, which is apparently possible.) – Thomas Andrews Dec 7 '11 at 20:30
up vote 11 down vote accepted

Added: The main results do not in fact depend on $R$ having no zero divisors, though I had to modify the arguments just a little in order to show this. I’ve made the necessary modifications below; significantly changed material is inset, like this note.

$R$ will be throughout a commutative ordered ring with identity, and $R^+\triangleq\{r\in R:r>0\}$. Define the $R$-norm $|\cdot|$ and $R$-metric $d$ as in the question. For $$x,r\in R$ let $B(x,r)=\{y\in R:d(x,y)<r\}\,.$$

Prop. 1: The $R$-metric topology on $R$ coincides with the order topology.

Proof: If $a<x<b$, let $r=\min\{x-a,b-x\}>0$; clearly $x\in B(x,r)\subseteq(a,b)$. Conversely, $B(x,r)=(x-r,x+r)$. $\dashv$

Henceforth I assume that $R$ bears this topology.

Prop. 2: $R$ is discrete iff $R^+$ has a minimum element, and in that case $R$ is of course metrizable. $\dashv$

Assume henceforth that $R$ is not discrete.

Prop. 3: For each $x\in R^+$ there is a $y\in R^+$ such that $2y\le x$.

Proof: Let $x\in R^+$. Since $R$ is not discrete, there is a positive $u<x$, and there is a positive $v<x-u$. Let $y=\min\{u,v\}$; then $2y\le u+v=x$. $\dashv$ (This argument no longer depends on multiplication.)

Theorem 4: If there is a strictly decreasing sequence $\langle r_n:n\in\omega\rangle$ in $R^+$ that is co-initial in $R^+$, then $R$ is metrizable.

Proof: For $n\in\omega$ let $\mathscr{B}_n=\{B(x,r_n):x\in R\}$. Suppose that $V$ is an open nbhd of some $x\in R$. By Prop. 1 there is some $r\in R^+$ such that $B(x,r)\subseteq V$. By Prop. 3 there is an $n\in\omega$ such that $2r_n\le r$, whence $\operatorname{st}(x,\mathscr{B}_n)\subseteq V$. Thus, $\{\mathscr{B}_n:n\in\omega\}$ is a development for $R$. It’s well known that every linearly ordered topological space is collectionwise normal, so $R$ is metrizable by the Moore metrization theorem (also sometimes called the Bing metrization theorem). $\dashv$

Added: It takes a bit more work, but one can show that in this case $R$ has a metric $d$ such that the $R$-Cauchy sequences are precisely the ordinary Cauchy sequences in $\langle R,d\rangle$. (This is trivially true in the discrete case.)

Up to here, then, $R$-metrics don’t really give us anything new. However, there are non-metrizable ordered rings. The easiest example that comes to mind is a countable ultrapower of $\mathbb{Q}$.

Example 5: Let $X=\mathbb{Q}^\omega$, with addition and multiplication defined coordinatewise, and let $\mathscr{U}$ be a free ultrafilter on $\omega$. Let $I=\big\{x\in X:\{n\in\omega:x(n)=0\}\in\mathscr{U}\big\}$; $I$ is a maximal ideal in $X$, so $R=X/I$ is a field. For $[x],[y]\in R$ define $[x]\le[y]$ iff $\{n\in\omega:x(n)\le y(n)\}\in\mathscr{U}$; the fact that $\mathscr{U}$ is an ultrafilter ensures that this is a linear order on $R$, and it’s easily seen to be compatible with the ring structure on $R$.

Now suppose that $\{x_n:n\in\omega\}\subseteq R^+$.

Now suppose that $\{[x_n]:n\in\omega\}\subseteq R^+$; I’ll show that there is an $x\in R^+$ such that $x<x_n$ for every $n\in\omega$, thereby showing that $R$ is not first countable (and hence certainly not metrizable).

Let $U_0=\{k\in\omega:x_0(k)>0\}$. Given $U_n$ for some $n\in\omega$, let $$U_{n+1}=U_n\cap\{k>n:x_{n+1}(k)>0\}.\,\tag{1}$$ Since $[x_n]>[0]$ for each $n\in\omega$, each $U_n\in\mathscr{U}$. Clearly $U_0\supseteq U_1\supseteq\dots$, and $\bigcap\limits_{n\in\omega}U_n=\varnothing$. Now define $x\in X$ as follows:

$$x(k)=\begin{cases} 0,&\text{if }k\in\omega\setminus U_0\\\\ \frac12\min\{x_m(k):m\le n\},&\text{if }k\in U_n\setminus U_{n+1}\;. \end{cases}$$

It follows from $(1)$ that if $k\in U_n\setminus U_{n+1}$, $x_m(k)>0\,$ for every $m\le n$ and hence that $x(k)>0$. Thus, $x(k)>0$ for every $k\in U_0\in\mathscr{U}$, and hence $[x]>[0]$. On the other hand, for each $n\in\omega$ we have $x(k)<x_n(k)$ for every $k\in U_n\in\mathscr{U}$, so $[x]<[x_n]$. $\dashv$

Definition 6: A sequence $\langle x_n:n\in\omega\rangle$ is $R$-Cauchy iff for each $r\in R^+$ there is an $n_r\in\omega$ such that $d(x_n,x_m)<r$ whenever $n,m\ge n_r$.

Obviously if $R$ is discrete, $R$-Cauchy sequences are eventually constant, while in the non-discrete metrizable case they behave like ordinary Cauchy sequences in metric spaces.

Prop. 7: If $R$ is not first countable, a sequence in $R$ is $R$-Cauchy iff it is eventually constant.

Proof: Clearly $R$ has a nested base at $0$, and this base cannot be countable: if it were, $R$ would be first countable. Let $\langle x_n:n\in\omega\rangle$ be an $R$-Cauchy sequence in $R$, and let $$D=\{d(x_n,x_m):n,m\in\omega\}\cap R^+,$$ the set of non-zero distances between terms of the sequence. $D$ is countable, so there is an $r\in R^+$ such that $r<d\,$ for every $d\in D$. Fix $n_r\in\omega$ such that $d(x_n,x_m)<r$ whenever $n,m\ge n_r$; clearly $d(x_n,x_m)=0$ and $x_n=x_m$ whenever $n,m\ge n_r$, i.e., the sequence is eventually constant. $\dashv$

Cor. 8: If $R$ is not discrete, $R$ is metrizable iff it has an $R$-Cauchy sequence that is not eventually constant. $\dashv$

share|cite|improve this answer
Thank you for the very nice and comprehensive answer, though it will take me a bit to fully digest it. – ItsNotObvious Dec 9 '11 at 15:24
So, in summary... yes. – Pete L. Clark Dec 9 '11 at 19:56

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.