Before I continue asking my question, I have seen this exact same question posted numerous times, but I don't feel like I seen any answers that were satisfactory in answering my complaint. For some reason, this idea is just not "clicking".

First off, I know the definition of pointwise convergence and uniform convergence. I also know that uniform convergence implies continuity. But I am still having trouble with the deeper understanding of uniform continuity.

I will use an example from the book to illustrate my confusion: For each natural number n, define $f_n(x)=x^n$ for $0 \leq x \leq 1$. We know that $\lim_{n \to \infty} f_n(1)=1$ and that for $0\leq x < 1$, $\lim_{n \to \infty}=0$. So, $\{f_n\}$ converges pointwise on $[0,1]$ to $f$, where $f=1$ if $x=1$, and $f=0$ if $0 \leq x < 1$. I understand this completely. However, my book says it fails to be uniformly convergent. I understand this is because $f$ is not continuous and it fails at $x=1$, but why? The book I have uses this definition for uniform convergence: $\{f_n\}$ converges uniformly to $f$, provided that for each positive number $\epsilon$, there exists an $N \in \mathbb{N}$ such that $$|f(x)-f_n(x)|<\epsilon$$ for all $n\geq N$ and all points $x \in D$. However, hasn't this definition been satisfied? Given any $\epsilon$ you'd give me, $|f(x)-f_n(x)|<\epsilon$ for all points. I don't see how $x=1$ is a problem and I don't see how this connects to continuity at all.

  • 1
    $\begingroup$ If you pick $\epsilon=1/2$, for instance, then $|f(1/2)-f_1(1/2)|<\epsilon$ is false. So I don't know why you say that "given any $\epsilon$ you'd give me, $|f(x)-f_n(x)|<\epsilon$ for all points". $\endgroup$ – Eric Wofsey Mar 5 '18 at 21:09
  • $\begingroup$ If you are only considering the definition of uniform convergence, then looking at the point $x=1$ will not show that uniform convergence fails because we rather trivially have $1^n=1$ for every positive integer $n$. It is the neighborhoods of $1$ where the uniform convergence will fail because we won't be able to find a finite positive integer $N \in \mathbb{N}$ to satisfy the definition of uniform convergence. math.stackexchange.com/questions/2212486/equicontinuity-of-xn $\endgroup$ – Matt A Pelto Mar 5 '18 at 21:18

I like to draw a picture for this. Draw the "$\epsilon$-band" around the graph of the limit function $f$. So in this case the $\epsilon$-band will be $y=\pm\epsilon$, for $0\le x<1$, and at $x=1$ it will be two dots at $1\pm\epsilon$. Uniform convergence says that for all large $n$, the graph of $f_n$ is entirely contained in the $\epsilon$-band. [Draw pictures of this yourself!!]

Now consider $f_n(x)=x^n$. It is (on or) outside the $\epsilon$-band on the interval $\root n\of\epsilon\le x<1$. Thus, no matter how large we make $n$, the graph of $f_n$ is certainly is not contained in the $\epsilon$-band.


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.