The given sequence of functions is defined as $f_n(x) = \frac{x^n}{n+x^n}$ for $x\ge 0$ and $n = 1,2,\ldots$; let $f = \begin{cases} 0 &:0\le x \le 1\\ 1 &:x>1 \end{cases} $

My hypothesis is that $f_n\to f$ point-wise, but not uniformly (the function $f$ was not given, I have defined it). I have managed to show $f_n \to f$ point-wise on $[0,1]$ but am having difficulties proving convergence for $x>1$. My hunch is that the $N$ we choose must depend on $\epsilon$ and $x$ (which is why I also believe $f_n$ does not converge uniformly), but I can't seem to find a suitable $N$. Any hints would be appreciated.


Point-wise convergence for $x>1$ follows from $$\lim_{n\to+\infty} \frac{x^n}{n+x^n}=1,$$ since $x^n$ grows exponentially with base $x>1$, hence it dominates for large $n$: $$\frac{n+1}{x^{n+1}}\Big/ \frac{n}{x^n}= \frac{n+1}{nx}<1$$ for any $n$ such that $n>1/(x-1)$. Let $n_0$ the minimum $n$ such that $n>1/(x-1)$. Thus for any $n\geq n_0$ it holds $$\frac{n+1}{x^{n+1}}\Big/ \frac{n}{x^n}= \frac{n+1}{nx}<1,$$ hence $$\frac{n_0+m}{x^{n_0+m}}\Big/ \frac{n_0}{x^{n_0}}\leq \left(\frac{n_0+1}{n_0x}\right)^m.$$ Fix $\epsilon>0$: $$\frac{n_0+m}{x^{n_0+m}}\leq \frac{n_0}{x^{n_0}}\left(\frac{n_0+1}{n_0x}\right)^m<\epsilon$$ whenever $m>\log_{\frac{n_0+1}{n_0x}} (\epsilon \frac{x^{n_0}}{n_0})$.

You are also correct about non-uniform convergence: fix an arbitrary $\epsilon>0$, if uniform convergence holds, then there exists $N$ such that for any $n\geq N$, $\sup_x|f_n(x)-f(x)|<\epsilon$. However for any $n$ you have $$\lim_{x\to 1^+} 1-\frac{x^n}{n+x^n} = \lim_{x\to 1^+} \frac{n}{n+x^n} =\frac{n}{n+1}.$$

  • $\begingroup$ Yes, which is how I constructed the limit function $f$ in the first place. However, I am looking for a rigorous $\epsilon$ proof. $\endgroup$ Dec 3 '14 at 1:36
  • $\begingroup$ I am still a bit confused by how $\frac{n}{x^n} < \epsilon$ whenever $n>1/\epsilon$, this is the part I am stuck at in my own work. $\endgroup$ Dec 3 '14 at 1:51
  • $\begingroup$ Ok, answer edited. $\endgroup$
    – Milly
    Dec 3 '14 at 3:50

For $x>1$, we have $\frac{x^n}{x^n+n}=\frac{1}{1+(\frac{n^{1/n}}{x})^n} \in \left[\frac{1}{1+(\frac{1+(x-1)/2}{x})^n},1\right] = \left[\frac{1}{1+a^n},1\right]$ for sufficiently large $n$, where $a=\frac{1+(x-1)/2}{x}<1$. Further, by Squeeze theorem, $\frac{x^n}{x^n+n}$ converges to 1.

Then let $(x_n)=(n^{1/n})>(0)$. But $f_n(x_n)=(\frac{1}{2})$ does not converge to 1.

In fact, if you have already learnt that "continuous functions which converge uniformly must converge (also pointwise) to a continuous function", then you can conclude immediately that the convergence was not uniform, because $f$ is not continuous.


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.