27

I am given two spaces $X$ and $Y$, both Hausdorff. I have defined a uniformly continuous function on a dense set $D$ of $X$ that goes to $Y$. So once you have defined a function on a dense subset, if you would like to extend this function to the whole set in a continuous manner, then there is at most one way to do this. For the extension to be sequentially continuous, all decisions are "already made" by determining the value of the function on the dense subset. The value of $f(x)$ simply must be $\lim f(x_{n})$ where $x_{n}$ are in $D$ and converge to $x$. I proved well defined and unique.

I am having trouble proving this is continuous, any help is appreciated.

Thanks.. And this is embarrassing, but can someone tell me how to accept an answer. Sorry...

Srivatsan
  • 26,311
alice
  • 1,033
  • 4
    I'm slightly confused. The way you are discussing this would seem to suggest we are in metric spaces, yet you say Hausdorff. Even if you mean $X,Y$ to be uniform spaces, we already know they are Hausdorff. – Alex Youcis Oct 29 '11 at 03:13
  • @Srivatsan She clearly stated that he/she does not know how to accept answers. – Alex Youcis Oct 29 '11 at 03:22
  • @alice: See this explanation of how accepting an answer works. Nothing to be embarrassed about, this site can be pretty confusing sometimes! – Zev Chonoles Oct 29 '11 at 03:25
  • @Alex I apologize. I noticed that only while making the edit and I removed the comment soon after. alice, Apologies to you as well. – Srivatsan Oct 29 '11 at 03:26

2 Answers2

19

In the light of Alex Youcis's comment, I will assume that $X$ and $Y$ are metric spaces. The following approach has been taken verbatim (apart from notation changes) from John Erdman, A ProblemText in Advanced Calculus (Chapter 24, pp. 146-147).

  1. If $f : X \to Y$ is a uniformly continuous map between two metric spaces and $(x_n)$ is a Cauchy sequence in $X$, then $f(x_n)$ is a Cauchy sequence in $Y$.

  2. Let $X$ and $Y$ be metric spaces, $S \subseteq X$, and $f : S \to Y$ be uniformly continuous. If two sequences $(x_n)$ and $(y_n)$ in $S$ converge to the same limit in $X$ and if the sequence $f(x_n)$ converges, then the sequence $f(y_n)$ converges and $\lim f(x_n) = \lim f(y_n)$.

Proof Hint. Consider the "interlaced" sequence $(x_1, y_1, x_2, y_2, x_3, y_3, \ldots)$.

Now, the main theorem.

Theorem. Let $X$ and $Y$ be metric spaces, $S$ a subset of $X$, and $f : S \to Y$. If $f$ is uniformly continuous and $Y$ is complete, then there exists a unique continuous extension of $f$ to $\overline S$. Furthermore, this extension is uniformly continuous.

Proof Hint. Define $g : \overline S \to Y$ by $g(a) = \lim f(x_n)$ where $(x_n)$ is a sequence in $S$ converging to $a$. First show that $g$ is well defined. To this end you must show that

  • $\lim f(x_n)$ does exist, and
  • the value assigned to $g$ at $a$ does not depend on the particular sequence $(x_n)$ chosen. That is, if $x_n \to a$ and $y_n \to a$, then $\lim f(x_n) = \lim f(y_n)$.

Next show that $g$ is an extension of $f$.

To establish the uniform continuity of $g$, let $a$ and $b$ be points in $\overline S$. If $(x_n)$ is a sequence in $S$ converging to $a$, then $f(x_n) \to g(a)$. This implies that both $d(x_j , a)$ and $d (f(x_j ), g(a))$ can be made as small as we please by choosing $j$ sufficiently large. A similar remark holds for a sequence $(y_n)$ in $S$ which converges to $b$. From this show that $x_j$ is arbitrarily close to $y_k$ (for large $j$ and $k$) provided we assume that $a$ is sufficiently close to $b$. Use this in turn to show that $g(a)$ is arbitrarily close to $g(b)$ when $a$ and $b$ are sufficiently close.

The uniqueness argument is very easy.

Srivatsan
  • 26,311
  • This is a sweet explanation. Thanks. – uncookedfalcon Jan 16 '12 at 05:46
  • @uncookedfalcon You are welcome. As I mentioned in the post, this answer is verbatim copied from Erdman's ProblemText. In case you like the explanation, you might want to browse through the book; I enjoyed working through it. – Srivatsan Jan 16 '12 at 05:54
7

(I also assume that we work with metric spaces.)
If you only need continuity of $g$, then continuity (not uniform) of $f$ is enough*. Let a sequence $x_n \in X$ converge to some point $x$. We will show that $g(x_n)$ converges to $g(x)$. I will write '(#)' in places, where we use the definition of $g$ for better readibility.

For every $x_n$ ($n$ is fixed) there exists a sequence $(w_n^m)_{m=1}^\infty \in D$ that converges to $x_n$. Since $f(w_n^m) \ \to g(x_n)$ (#) and $w_n^m \to x_n$, we may choose such $m_n$ that $d_X(w_n^{m_n}, x_n)$ and $d_Y(f(w_n^{m_n}), g(x_n))$ are smaller then $\frac{1}{n}$.
Now (from $d_X(w_n^{m_n}, x_n) < \frac{1}{n}$) you have: $\lim_{n \to \infty} w_n^{m_n} = x $, so (#):
$\lim_{n \to \infty} f(w_n^{m_n}) = g(x) $,
but we have also $d_Y(f(w_n^{m_n}),g(x_n)) < \frac{1}{n}$, so:
$\lim_{n \to \infty} g(x_n) = g(x) $, which gives us the continuity of $g$.

*However you still need uniform continuity to show that $g$ is uniquely defined, so my approach is not better, it is just different.

Saikat
  • 1,583
savick01
  • 4,499
  • As @SrivatsanNarayanan wrote, you need completeness of $Y$ to start doing anything. – savick01 Oct 29 '11 at 10:24
  • If $g$ is not uniformly continuous, I didn't understand how we defined $g(x)$. If we take a sequence converging to $x$, its image will not necessarily be a Cauchy sequence in $Y$ no? – edamondo Jun 20 '21 at 11:19