MathOverflow is a question and answer site for professional mathematicians. 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

Let {X,T} be a topology, T the set of open subsets of X.

Definition: Three points x, y, z of X are in relation N (Nxyz, read "x is nearer to y than to z") iff

  1. there is a basis B of T and b in B such that x and y are in b but z is not and
  2. there is no basis C of T and c in C such that x and z are in c but not y.

For some topologies there are no points x, y, z in relation N, for example if T = {Ø,X} or T = P(X), but for others there are (e.g. for ones induced by a metric [my claim]).

Definition: A topology has property M1 iff

(x)(y) ((z) (z ≠ x & z ≠ y) → Nxyz) → x = y

(This is an analogue of d(xy) = 0 → x = y, the best one I can imagine).

Definition: A topology has property M2 iff

(x)(y)(z) Nxyz & Nyzx → Nzyx

(This is a kind of an analogue of d(xy) = d(yx), the best one I can imagine)

First (bunch of) question(s):

  1. Properties M1 and M2 do not capture the whole of the corresponding conditions of a metric. Can anyone figure out "better" definitions (e.g. an analogon of x = y → d(xy) = 0)?

  2. Can anyone figure out a property M3 that is an analogue of the triangle equality?

If it can be shown that no such property M3 is definable, the following becomes obsolete.

If such a definition can be made, we define:

Definition: A topology has property M (read "induces a metric") iff it has properties M1, M2, M3.

Second question:

Which topologies have property M, i.e. induce a metric? Are these "accidentally" exactly those that are induced by a metric?

share|cite|improve this question
I don't have an answer to your questions, but, more conventionally, there is a standard theorem about which topological spaces admit a metric. It's beautiful, though not too easy -- it's one of the key points of a Point-Set topology course. Look in Munkres's Topology, or here: Also, if you like these kinds of questions, I highly recommend you read either Bourbaki's "General Topology" or Kelley's "General Topology". Both have a few generalizations of metrics that you might enjoy (in particular, uniformities). – Ilya Grigoriev Dec 20 '09 at 0:39
Usage comments: "analogon" is much less common than the synonomous "analogue". "Has property X" or "is in relation N" are poor choices. Better is "Given a topology T on X, say that x is closer to y than to z if....". – Theo Johnson-Freyd Dec 20 '09 at 3:16
You might just want to learn… . – Qiaochu Yuan Dec 20 '09 at 4:46
All of the important information of a metric can be recovered by its induced uniform space. This carries both the topological information as well as the uniform information (uniform convergence, completeness, et cetera). I don't believe you can get all of the information you want with just a topology. I am sure though, that one can only recover a metric up to topological equivalence. – Harry Gindi Dec 20 '09 at 4:59
Given the revisions at…, I'm voting to close this question as no longer relevant. – Theo Johnson-Freyd Dec 27 '09 at 16:30
up vote 12 down vote accepted

Your condition 1 is satisfied for all triples $x,y,z\in X$ such that $z\not\in\{x,y\}$ if the space is $T_1$.

Maybe reading a bit about uniform spaces and the corresponding metrizability results will be of help.

share|cite|improve this answer
Picking up Aaron's answer: Might there be a way to sensibly restrict the class of "allowed" (= "sensible") bases. What is T<sub>1</sub>? – Hans Stricker Dec 20 '09 at 1:09
There is no more reference to bases anymore in your (edited) answer. My follow-up question remains (sorry): what space is T_1= – Hans Stricker Dec 20 '09 at 1:13
I found the term T_1 linked and will follow this. Thanks! – Hans Stricker Dec 20 '09 at 1:34
In particular, HP Stricker's claim is patently false. – Theo Johnson-Freyd Dec 20 '09 at 3:19
+1 for mentioning uniform spaces – Harry Gindi Dec 20 '09 at 8:53

I'm not sure your "nearness" relation is the right definition to make. It makes sense on the face of it, but for example I think you can generate the usual topology on the plane by tubular neighborhoods of half-circles. For three points x,y,z on a line in that order, I don't think you get any relationship. Using the usual disk basis you'd satisfy condition (1) for Nxyz, but you could violate condition (2) using the other basis I just suggested, if the tubular neighborhood is sufficiently small.

share|cite|improve this answer
However, I too have wondered which topologies can be induced by metrics. I think the first step in this direction would be to look for a topology which is <i>not</i> induced by a metric... – Aaron Mazel-Gee Dec 20 '09 at 0:25
I thank you for this. Maybe the look for "unusual" bases should be sharpened (for beginners like me). Coming the standard way I thought only of disk bases. But maybe my line of thought can be "rescued"? If not: maybe it helps understanding the difference between topologies and metric spaces. – Hans Stricker Dec 20 '09 at 0:30
I don't know much about this stuff, but it looks like people who have commented on the original question suggest a wealth of resources for you to browse through... – Aaron Mazel-Gee Dec 23 '09 at 1:47

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.