Let $(X,d_X)$ and $(Y,d_Y)$ be metric spaces. $f_n: X\to Y$ with $n\in\mathbb{N}$ and $f:X\to Y$ functions. $f_n$ is continuous for every $n$ and $f_n\stackrel{n\to\infty}{\longrightarrow} f$ uniformly.

Then $f$ is continuous.

Here is my proof:

Let $x\in X$ and $\epsilon >0$ be arbitrary. Since $f_n\to f$ uniformly it exists $N\in\mathbb{N}$ such that $d_Y(f_N(x), f(x))<\epsilon/3$ for every $x\in X$.

As $f_N$ is continuous we have for $x_0\in X$ and $\delta > 0$, that $d_Y(f_N(x),f_N(x_0))<\epsilon/3$ if $d_X(x,x_0)<\delta$.

This gives us:

$$\begin{align}d_Y(f(x),f(x_0)&\leq d_Y(f(x), f_N(x))+d_Y(f_N(x), f(x_0))\\ &\leq d_Y(f(x), f_N(x))+d_Y(f_N(x), f_N(x_0))+d_Y(f_N(x_0), f(x_0))\\ &< \epsilon/3+\epsilon/3+\epsilon/3=\epsilon\end{align}$$

Thanks in advance for your correction.

  • 1
    $\begingroup$ I see no need to correct whatever (other than a typo). $\endgroup$ – José Carlos Santos Jul 17 '18 at 18:41
  • 1
    $\begingroup$ Thanks for the confirmation. Better safe then sorry. $\endgroup$ – Cornman Jul 17 '18 at 18:42

The idea of this proof is entirely correct, and it's very well executed.

There is some minor and entirely inconsequential stuff, like $d_Y(f(x), f(x_0)$ missing a bracket. The biggest thing for me, however, when I read this is that it is not really clear to me where exactly $\delta$ comes from. Furthermore, you haven't really made sure that $d_Y(f_N(x_0), f(x_0))<\epsilon/3$.

So I would change it to something like this:

Let $x\in X$ and $\epsilon >0$ be arbitrary. Since $f_n\to f$ uniformly, it exists $N\in \Bbb N$ such that $d_Y(f_N(x_1),f(x_1))<\epsilon/3$ for all $x_1\in X$.

As $f_N$ is continuous, there is a $\delta>0$ such that for any $x_0$ with $d_X(x, x_0)<\delta$ we have $d_Y(f_N(x), f(x_0))<\epsilon/3$.

Then from here do as you have already done, except you have to put in at least one $<$ somewhere instead of $\leq$.

  • $\begingroup$ Thank you. I edited that the estimation $d_Y(f_N(x), f(x))<\epsilon/3$ holds for every $x\in X$. Espacially for $x_0$. This is given by the uniform convergenc. $\endgroup$ – Cornman Jul 17 '18 at 18:55
  • $\begingroup$ @Cornman That is good. However, I would use a different letter, since you have already fixed an (arbitrary) $x$ and using the same name for two different things is not considered good practice. $\endgroup$ – Arthur Jul 17 '18 at 18:57
  • $\begingroup$ Ok, I will keep that in mind. :) $\endgroup$ – Cornman Jul 17 '18 at 18:58

You are close, but in my opinion in one detail you are not sufficiently precise enough.

The problem lies in the inequality $$ d_Y(f_N(x_0),f(x_0)) \leq \varepsilon/3, $$ which you only showed for $x$, not for $x_0$.

In my opinion you should replace the first line of the proof with something like

Let $\varepsilon>0$ be arbitrary. Since $f_n\to f$ uniformly there exists $N\in\mathbb N$ such that $d_Y(f_N(z),f(z))<\varepsilon/3$ for all $z\in X$.

If I would read your first line, it sounds like it would also be true if $f_n$ converges pointwise and not uniformly. However, this fact is very important for the proof.

  • $\begingroup$ Thanks for this answer. I forgot to mention, when recapping the definition of uniform convergenc, that this holds for every $x\in X$. Espacially for $x_0$. This is were the estimation of $d_Y(f_N(x_0), f(x_0))$ came from. $\endgroup$ – Cornman Jul 17 '18 at 18:54

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.