Related to a nice problem I met yesterday, a question arises:

Suppose $\{f_n\}$ is a sequence of mappings from a connected complete metric space $X$ to a metric space $Y$. Given $f\colon X\to Y$ such that $f_n(x_n)\to f(x)$ whenever a real sequence $x_n\to x$. Must $\{f_n\}$ converge uniformly?

According to Canez's hint, it's true that $f_n\to f$ pointwise and $f$ is continuous (modifying my ugly answer, replace $+1/N_k$ with $+0$) even if $X$ is an arbitrary metric space (without assuming that $X$ is connected or complete).

If it's too abstract, we can consider a concrete one, such as $X=Y=\mathbb R$ or $X=Y=\mathbb C$.


If $X$ is compact, it's true that $f_n\to f$ uniformly.

Prove by contradiction. Suppose $\exists\epsilon_0>0,\forall N,\exists n\ge N$ such that $$\sup_{x\in X}d_Y(f_n(x),f(x))>\epsilon_0$$

We can choose a increasing sequence $\{m_k\}\subset\mathbb N$, and a sequence $\{x_{m_k}\}\subset X$, such that $d_Y(f_{m_k}(x_{m_k}),f(x_{m_k}))>\epsilon_0$ for all $k$. Since $X$ is compact, there's a subsequence $\{n_k\}$ of $\{m_k\}$ such that $x_{n_k}\to x$ for some $x\in X$, then we can show that $f_{n_k}(x_{n_k})\to f(x)$ by constructing a sequence with duplicated $x_{n_k}$'s, just like this. On the other hand, since $f$ is continuous, we have $f(x_{n_k})\to f(x)$ contradicting to $d_Y(f_n(x),f(x))>\epsilon_0$.

In contrary to the preceding statement, if $X$ isn't connected (or at least, with finite connected components), even if $X$ is complete, $f_n\to f$ needn't be uniform, such as $X=\mathbb Z$.


I don't know whether $X$ could be replaced with an arbitrary connected metric space, or even $X$, $Y$ are just Hausdorff spaces. I don't know whether the preceding compact metric space $X$ could be replaced with a compact Hausdorff space. It's out of my ability and I can't discuss more.

  • $\begingroup$ For $X=Y=\Bbb R$, the sequence $(f_n)$ where $f_n=\chi_{[n,n+1]}$ would be a counterexample. (Note in the Explanations section, you might only have a subsequence of $(f_n)$ that satisfies your $\sup$ condition. But the argument can still be carried through.) $\endgroup$ – David Mitra Feb 23 '13 at 5:16
  • $\begingroup$ @DavidMitra You're right. Could you pleased post an answer and I'll accept yours? $\endgroup$ – Yai0Phah Feb 23 '13 at 5:49
  • $\begingroup$ Continuous convergence is equivalent to continuity of $f$ and uniform convergence $f_n \to f$ on compact subsets. See here: math.stackexchange.com/q/182223 $\endgroup$ – Martin Feb 23 '13 at 12:31

Note in the Explanations section, you might only have a subsequence of $(f_n)$ that satisfies your sup condition. But the argument still carries through. The result holds for compact sets.

For $X=Y=R$, the sequence $(f_n)$ where $f_n=\chi_{[n,n+1]}$ would be a counterexample.

The result also fails for general bounded sets, as the sequence $(x^n)$ on $(0,1)$ witnesses.


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.