My questions concern uniform spaces in specific cases involving compact spaces.

The first one is about how a compact Hausdorff space is a uniform space. I'm aware of indirect proofs using pseudometrics, but I'd like to see a direct proof that the set of neighbourhoods of the diagonal $\Delta_K$ of a compact Hausdorff space $K$ forms a basis for a uniformity on $K$ that induces its topology. How can I find such a proof ? (the only hard part being that for all such $U$ there is such a $V$ with $V\circ V\subset U$, i.e. $K$ is divisible)

Once this is solved, I'm wondering about the following question: let $X$ be a uniform space, $Z$ its Cauchy completion (which is again a uniform space), and assume $Z$ is compact (i.e. $X$ is precompact). Let $K$ be a compact space, and $f:X\to K$ a continuous function. Is $f$ necessarily uniformly continuous ?

If the answer is yes, then it can be extended to a map $g: Z\to K$ by the usual construction. If the answer is no, does this last bit still hold: even if $f$ is not uniformly continuous, can it always be extended to the completion when the codomain is compact ? (here the completion is not assumed to be compact)

My thoughts on how to answer these questions:

for the first one, I've been stuck for quite a while and have not made much progress. I've tried using the fact that for an open set $U\subset K$, there is $V$ open with $\overline{V}\subset U$, but this hasn't led me anywhere. Similarly, finding that a neighbourhood of the diagonal contains another such neighbourhood of the form $\displaystyle\bigcup_{i=1}^n U_i\times U_i$, where each $U_i$ open hasn't given me anything.

For the second one, I think the answer is yes, but only because I think that the map can be extended, and since $Z$ is compact, then $g$ would be uniformly continuous, and thus its restriction to $X$ too. So I've tried actually answering the third question (i.e. not proving that $f$ is uniformly continuous, and not any assumptions on $Z$; simply proving that it can be extended), but I run into some complications when I try to prove that if $x_i \to x, y_j \to x$, and $f(x_i), f(y_j)$ both converge in $K$, then it's to the same limit, and haven't managed to show this (which would help)

  • 1
    $\begingroup$ Since these are two distinct questions, it would be better to ask them separately. $\endgroup$ – mathematics2x2life Mar 23 '18 at 21:03
  • $\begingroup$ The second question has a simple counterexample $\sqrt{x}$ as a function $(0,1)\to[0,1]$ with the usual metrics $\endgroup$ – Dap Mar 24 '18 at 9:58
  • $\begingroup$ @Dap : this will be uniformly continuous, and this follows from the fact that it has a continuous extension $[0,1]\to [0,1]$ $\endgroup$ – Max Mar 24 '18 at 10:18
  • $\begingroup$ @Max: oh, sorry, I was thinking of something else. $\sin 1/x$ then $\endgroup$ – Dap Mar 24 '18 at 10:19
  • $\begingroup$ @mathematics2x2life : you are probably right but I thought since the two questions are related I could ask them both at once $\endgroup$ – Max Mar 24 '18 at 10:19

For 1, by regularity any point is contained in some open $A$ with $\overline{A}\subset B$ with $B$ open and $B\times B\subset U.$ This means there is a finite open cover $\mathcal A=\{A_1,\dots,A_n\}$ where each $A_i$ comes with a corresponding open $B_i$ satisfying $\overline{A_i}\subset B_i$ and $B_i\times B_i\subset U.$ For each $i$ we have a two-element open cover $\mathcal A_i=\{K\setminus \overline{A_i}, B_i\},$ and the original open cover $\mathcal A.$ Let $\mathcal V$ be an open cover finer than all these covers, such as the family of sets of the form $X\cap X_1\cap \dots \cap X_n$ with $X\in \mathcal A$ and $X_i\in\mathcal A_i$ for each $i.$ Let $V=\bigcup_{X\in\mathcal V} X \times X.$ It's an open neighborhood of the diagonal. If $(x,y),(y,z)\in V$ then $x,y\in A_i$ for some $i,$ so $y,z\in B_i$ and $(x,z)\in B_i\times B_i\subset U$ as required.

  • $\begingroup$ thank you ! I'll accept your answer because you already answered the rest in the comments with $sin(1/x)$ which can't be continuously extended to $[0,1]$ (moreover as pointed out I should have asked separate questions) $\endgroup$ – Max Mar 24 '18 at 10:24

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.