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

In the definition of a metric space, replace the triangle inequality by the weaker inequality

d (x, z) ≤ C max {d (x, y), d (y, z)},

where C is a positive constant (depending on the "metric", but not on the points x, y, z). Had structures like this ever been studied?

One can associate a more or less natural topology to this "metric", calling a set open if every point belonging to the set has a ball of positive radius, centered at this point and contained in the set. But I cannot say much on this topology. For instance, it is not obvious (and may be not true) that a ball is an open set. Neither could I prove that this topology is Hausdorff.

Any information, reference, etc. would be welcome.

share|cite|improve this question
When $C=1$ I think this might hold for $p$-adic numbers... in which case the condition is stronger than the usual metric space condition, and makes $p$-adic analysis a bit easier. – Dylan Wilson Oct 16 '10 at 23:51
Yes, for C=1, this is just the so-called ultrametric Though why it is called (or should be called) ultrametric is not fully clear to me. – Suvrit Oct 17 '10 at 8:30
up vote 3 down vote accepted

Yes, they were introduced in valuation theory by Emil Artin and remain present in many contemporary treatments, including mine: see

especially Section 1.2 and

[Added: as RW has justly pointed out, my answer here makes sense in the context of valuation theory only. The procedure that I give from passing from a "weak metric" to a metric is not going to work in general, I think, but only in the presence of some additional algebraic structure. If I were the OP, I might not choose this as the accepted answer, or at least not yet.]

The basic idea here is to consider two such guys equivalent if one can be obtained from the other by the operation of raising $d$ to some positive real number power $\alpha$. In such a way, one can make the constant $C \leq 2$ in which case one gets an actual metric. The topology one gets in this way is easily seen to be independent of the choice of $\alpha$.

As it happens, when writing up these notes for a course I taught last spring I also thought a little bit about trying to define "balls" with respect to such a weak metric (i.e., without first renormalizing). I didn't get anywhere with this either.

Finally, I should say that in valuation theory at least, these "weak metrics" (in the valuation theoretic context I called them "Artin absolute values" and then after the course was over decided to change to the terminology to just "absolute values", while retaining the word "norm" for such a guy which was actually a metric) come up as a useful tool but are not really studied on their own or in any deep way: I have yet to see a text on valuation theory where weak norms appear after page 20 or so. Whether they may have wider applicability in some other context, I couldn't say...

share|cite|improve this answer
The arguments you refer to (that for $C\le 2$ one gets an actual metric) heavily depend on an additional algebraic structure. I don't see how they would work in the general case. – R W Oct 17 '10 at 1:14
@RW: That may well be. I have only encountered the phenomenon in the setting of valuation theory, and I don't mean to imply that what I say works in the general case. Perhaps this should not be the accepted answer... – Pete L. Clark Oct 17 '10 at 2:56

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.