Lets $X$ be a compact metrizable space and $f:X\to Y$ be a quotient map such that $Y$ equipped with the quotient topology is Hausdorff. Thus $Y$ is metrizable. Lets $\sim$ be an equivalence relation on $X$ such that $x\sim y$ if $f(x)=f(y)$. Then for any metric $d$ compatible with the topology of $X$ one can build a (pseudo)metric $d_\sim$ on $Y$ with: \begin{equation} d_\sim(a,b) = \inf\{d(p_1,q_1) + \cdots+ d(p_n,q_n);[p_1] = a,[q_i] = [p_{i+1}],[q_n] = b\}.(1) \end{equation} Where the $\inf$ is taken over all finite chains of points $\{p_i\}_{i=1}^n$, $\{q_i\}_{i=1}^n$ between $a$ and $b$.

As in this question which has not been fully answered (Quotient of metric spaces) can we show that $d_\sim$ is a metric compatible with the quotient topology in $Y$ ?

If not, what would be a sufficient condition on the quotient map in order to have the result ?

Here is an attempt:

In Herman 1968, Quotient of metric spaces, in theorem 4.8, is stated the following :

THEOREM: Let $f$ be a function from a pseudo-metrizable space $X$ to a topological space $Y$, and suppose that $Y$ has the quotient topology relative to $f$, then the following are equivalent:
1) $Y$ is pseudo-metrizable
2) There exists a pseudo-metric $\rho$ compatible with the topology in $X$ such that the quotient pseudo-metric $\rho_\sim$, defined as in (1), is in fact compatible with the quotient topology of $Y$ (The definition of the quotient pseudo-metric by Herman should be equivalent to the one introduced earlier).

If $X$ is in fact metrizable, then it is pseudo-metrizable and $Y$ is also pseudo-metrizable. Therefore, from the theorem there exists a pseudo-metric $d^*$ compatible with the topology in $X$ (it is a metric as $X$ is Hausdorff) such that the quotient pseudo-metric $d^*_\sim$ is compatible with the topology in $Y$ (it is also a metric because $Y$ is Hausdorff too).

Is it possible to show that any quotient (pseudo)metric from an arbitrary metric $d$ is topologically equivalent to $d^*_\sim$ ?

We know that $X$ is metrizable and compact, thus there is a unique uniform structure in it and all metrics compatible with the topology are uniformly equivalent. Therefore any metric $d$ compatible with the topology of $X$ is uniformly equivalent to the metric $d^*$.

From uniform equivalent metrics, maybe there is a relation between their corresponding quotient pseudo-metrics but I am stucked here, do you have any idea/theorem/reference that would help me ?


  • $\begingroup$ According to the first line of your post, I think $Y$ is always metrizable, provided it is Haussdorf. (No quotion topology is needed for its metrizability). am I mistaken? $\endgroup$ – Ali Taghavi Apr 5 at 13:38

For the Cantor starcase function $f:C\to[0,1]$ from the standard ternary Cantor set $C$ onto the interval $[0,1]$ and for the standard Euclidean metric $d$ on $C$ the quotient pseudometric $d_\sim$ is constant zero (this follows from the fact that the Cantor set $C$ has length zero). So, the pseudometric $d_\sim$ is not necessarily a metric.

This example should be known but I cannot mention a suitable reference at the moment.

Added in Edit. Essentially the same counterexample is discussed in the answer of Wlodzimierz Holsztynski to this MO-question.

  • $\begingroup$ Thank for the answer ! Indeed it is the same counter-example than in the question I have quoted. However, I have realised that I need to deal with path-connected spaces so that quotient space is path-connected in the quotient pseudo-metric. Is there a known example that does not use the cantor set ? $\endgroup$ – VMrcel Jun 25 '18 at 9:44
  • $\begingroup$ @VMrcel You can extend the Cantor starcase function to a continuous function on the closed interval and then you will get a continuous function between closed intervals, for which the quotient pseudometric still is zero. By the way, the quotient space is path-connected in the quotient metric (since it determines the anti-discrete topology). So, maybe some more precise question should be asked (but a good question is a half of an answer). $\endgroup$ – Taras Banakh Jun 25 '18 at 10:53
  • $\begingroup$ Oh you are right, I'll think about it, thank you ! $\endgroup$ – VMrcel Jun 26 '18 at 12:08

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.