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

Suppose the countable subspace $D$ is dense in the separable Tychonoff space $X$ and $f$ is a continous function from $D$ to the closed unit interval. What are some conditions on $X$ or $D$, which make $f$ continuously extendable over $X$?

share|cite|improve this question
Are you aware of Tietze's Extension Theorem? It's related, but with a $T_4$ space $X$ rather than $T_{3.5}$. See: – David White Oct 20 '11 at 4:12
But here $D$ is dense, so Tietze applies only if $D=X$... – Pietro Majer Oct 20 '11 at 8:30

A relevant paper is

Taĭmanov, A. D. On extension of continuous mappings of topological spaces. (Russian) Mat. Sbornik N.S. 31(73), (1952). 459–463. 56.0X

The MR of this paper is:

Let $S$ be a $T_1$-space, $A$ a dense subspace of $S$, and $R$ a compact Hausdorff space. Let $f$ be a continuous mapping of $A$ into $R$. Then $f$ admits a continuous extension over $S$ if and only if for all disjoint closed subsets $A_1,A_2$ of $R$, the relation $(f^{-1}(A_1))^-\cap(f^{-1}(A_2))^-=0$ obtains (closure in $S$). From this result, a theorem of Yu. M. Smirnov [Uspehi Matem. Nauk 6, no. 4(44), 204--206 (1951)] is easily proved, as well as a theorem of Vulih [Mat. Sbornik N.S. 30(72), 167--170 (1952); MR0048790 (14,70c)]. A final corollary is a special case of a theorem widely known and recently published by Katětov [Fund. Math. 38, 85--91 (1951); MR0050264 (14,304a)].

I remembered this because as a graduate student I used it to give a (I think new at the time) proof that every compact Hausdorff space is a continuous image of a compact totally disconnected compact Hausdorff space (which, in turn, I use these days to reduce proving the Riesz representation theorem for $C(K)$ to the case where the compact space $K$ is totally disconnected).

share|cite|improve this answer
O,thanks! it is very beautiful! – Paul Oct 21 '11 at 0:43

The criterion for "EVERY continuous map from $D$ to $[0, 1]$ has a continuous extension to $X$" is that any two disjoint zerosets in $D$ have disjoint closures in $X$. You can find this in Chapter 6 of Gillman and Jerison's classic "Rings of Continuous Functions". They also consider the "local problem" of continuously extending a single map at length in some of the exercises, e.g. given $f:D\rightarrow Y$ (not necessarily $Y=[0, 1]$) Exercise 6G characterizes the largest subspace of $X$ to which $f$ can be continuously extended in terms of $z$-filters.

share|cite|improve this answer
Thank you very much. But I can't find the book of "Rings of Continuous Functions". Could you tell me where i can find this book? – Paul Oct 21 '11 at 0:29
You'll probably have to go through some sort of interlibrary loan, although Amazon lists some used copies available for purchase. – Todd Eisworth Oct 21 '11 at 19:43

The situation is analogous to the particular case of $X$ a metric space, for any Tychonoff space $X$ is uniformisable, and a real valued function $f$ on a dense subset $D$ of a uniform space $X$ is certainly continuously extendable to $X$ provided it is uniformly continuous. This is also a necessary condition if $X$ is compact, for any continuous function on a compact uniform space is always uniformly continuous.

share|cite|improve this answer
Is "Any Tychonoff space X is uniformisable" means that every Tychonoff space is a uniform space? – Paul Oct 20 '11 at 9:34
Pietro provided a link to Wikipedia in his answer. But the answer is essentially yes; the topology on $X$ will be compatible with a construction that makes $X$ into a uniform space. – Christopher A. Wong Oct 20 '11 at 9:47
Yes, it admits a uniform structure (not unique in general). So the complete answer is: $f$ is uniformly continuous on $D$ wrto one such uniform structures. – Pietro Majer Oct 20 '11 at 10:31
To be specific, $f$ is extendable if and only if it is uniformly continuous wrt (the restriction to $D$ of) the fine uniformity on $X$. – Emil Jeřábek Oct 20 '11 at 13:35
With respect to – Richard Rast Oct 21 '11 at 1:38

At least in the case of a metric space $X$, such a function $f$ extends from $D$ to all of $X$ if and only if $f$ maps Cauchy sequences to Cauchy sequences (note that this is a weaker condition than uniform continuity).

As mentioned by Pietro, for your general Tychonoff space $X$, you make it a uniform space, so I think you can generalize my statement above to the following: $f$ extends if and only if $f$ maps Cauchy nets to Cauchy nets.

share|cite|improve this answer
Also note that a function maps Cauchy nets to Cauchy nets if and only if it is uniformly continuous. – Pietro Majer Oct 20 '11 at 10:32
@Pietro: Are you sure? For example, every continuous function from $\mathbb R$ to any uniform space also maps Cauchy nets to Cauchy nets. (Let $\{x_a\}_{a\in D}$ be a Cauchy net. There is $a_0$ such that $|x_a-x_{a_0}|\le1$ for every $a\ge a_0$, hence all such $x_a$ are confined to the compact interval $I=[x_{a_0}-1,x_{a_0}+1]$. Then $f$ is uniformly continuous on $I$, which implies that $\{f(x_a)\}_{a\in D}$ is Cauchy.) – Emil Jeřábek Oct 20 '11 at 13:24
OTOH, Christopher’s condition is clearly not necessary, even in the metric case. For instance, if $D=X$, then every continuous $f$ trivially extends, but in general does not map Cauchy sequences to Cauchy sequences (e.g., take $X=\mathbb Q$, $f(x)=0$ for $x<\pi$, $f(x)=1$ for $x>\pi$). – Emil Jeřábek Oct 20 '11 at 13:31
@Emil: oh yes, my distraction! – Pietro Majer Oct 20 '11 at 14:46
@ Emil: In the metric case, I suppose I was thinking of the process of completion as uniquely determined by Cauchy sequences. So perhaps what I really want is that $f$ maps Cauchy sequences to Cauchy sequences, wherever these Cauchy sequences converge to a point $p \notin D$, $p \in X$. – Christopher A. Wong Oct 20 '11 at 16:27

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.