I am having a bit of trouble on this revision question.

To determine pointwise convergence: $\lim_{n\rightarrow\infty} = nx(1-x)^n $. For $x=0, x=1$, it's clear that the limit is $0$. How can I determine the limit for $0 \le x \le 1$? (My limit finding skills are rusty). I'm quite sure it is $0$ as well, but how do I go about showing this properly?

Supposing this is true, then to determine uniform convergence, let $d_n(x):= |f_n(x)-f(x)|= nx(1-x)^n - 0$

Then $$ d_n'(x)= n(1-x)^n - n^2(1-x)x = 0$$ if $x=1$. So the maximum of $f_n(x)$ occurs at $x=1$. It follows that:

$$0 \le d_n(x)= |f_n(x)-f(x)|< d_n(1) = 0$$ So $(f_n)$ converges uniformly on $[0,1]$. Would this be correct?

Thanks for the help in advance!

  • 1
    $\begingroup$ No, otherwise you could move limit inside the integral, but $\displaystyle\lim\limits_{n\to\infty}\int f_n=\lim\limits_{n\to\infty}\frac{n}{2n+2}=\frac{1}{2}\neq0=\int \lim\limits_{n\to\infty} f_n$ $\endgroup$ – OBDA May 27 '14 at 15:51

The plan is good, the execution isn't.

If you maximize $f_n(x)$ for fixed $n$ as a function of $x$ you get $x=1/(n+1)$. Uniform convergence then doesn't hold because

$$\sup_x f_n(x) = \frac{n}{n+1} \Bigl( 1 - \frac{1}{n+1}\Bigr)^n \to \frac{1}{e},$$

as $n\to\infty$.

  • 1
    $\begingroup$ You're missing the $n$ in front of $f_n$ $\endgroup$ – Omnomnomnom May 27 '14 at 15:42
  • $\begingroup$ noted, fixed... $\endgroup$ – JPi May 27 '14 at 19:06

You are correct in stating that the pointwise limit is zero.

Your differentiation, however, was incorrect. We have $$ d_n'(x)= n(1-x)^n - n^2(1-x)^{n-1}x = 0 \implies\\ n(1-x)^{n-1}((1-x) - nx) = 0 \implies\\ 1 - (n+1)x = 0 \implies\\ x = \frac 1{n+1} $$ So, the maximum difference is $$ n \cdot \frac{1}{n+1} \cdot \left( 1 - \frac{1}{n+1} \right)^n \not \to 0 $$ Which, in fact, shows that we do not have uniform convergence.

  • $\begingroup$ Thanks, I realised I made an error in differentiation. $\endgroup$ – Terrence J May 27 '14 at 22:57

To answer your first question about pointwise convergence for $0<x<1$, note that for a fixed value of $x$ there exists $y>0$ such that

$$(1-x) = \frac{1}{1+y}.$$


$$0<nx(1-x)^n = \frac{nx}{(1+y)^n}.$$

Applying the binomial theorem we see that

$$(1+y)^n = 1 + ny + \frac{n(n-1)}{2}y^2 + \ ... \ >\frac{n(n-1)}{2}y^2,$$


$$0<nx(1-x)^n < \frac{2x}{(n-1)y^2}.$$

The RHS of the above inequality can be made smaller than any $\epsilon>0$ for fixed values of x and y by choosing a sufficiently large value of $n$.

In particular, let $N$ be an integer such that

$$N > 1+\frac{2x}{\epsilon y^2}.$$

Then for all $n \geq N$ we have

$$0<nx(1-x)^n < \epsilon,$$

and the pointwise limit of the sequence is $0$.

  • $\begingroup$ Thanks, really helpful. Just a quick question, how did you derive your first line? $\endgroup$ – Terrence J May 27 '14 at 22:54
  • $\begingroup$ Is $y$ chosen, dependent on $x$ I suppose? Sorry, it just doesn't seem obvious to me. $\endgroup$ – Terrence J May 27 '14 at 23:06
  • 2
    $\begingroup$ This is a standard argument to show pointwise convergence. We consider $x$ as a fixed number between $0$ and $1$. It is true that $y$ depends on $x$, but it is independent of $n$ and must be greater than $0$ if $x$ is less than $1$. I'll show more detail in applying the binomial theorem to obtain the inequality. $\endgroup$ – RRL May 27 '14 at 23:29
  • 1
    $\begingroup$ After $$0<nx(1-x)^n < \frac{2x}{(n-1)y^2}.$$ Is the rest necessary? Could we not just say that since as $n \rightarrow \infty$ by squeeze it tends to $0$ also. So since for $x \in [0,1]$, $nx(1-x)^n$ tends to $0$, then we can conclude this is the pointwise limit? $\endgroup$ – Terrence J May 28 '14 at 2:28
  • 2
    $\begingroup$ @user153663: Yes you can apply the Squeeze Theorem to reach the same conclusion. $\endgroup$ – RRL May 28 '14 at 2:39

Observe that $$\sup_{x \in [0,1]} |f_n(x)-f(x)| \geq \left|f_n \left(\frac{1}{n} \right) -f\left( \frac{1}{n} \right) \right|=\left(1-\frac{1}{n} \right)^n \to \frac{1}{e}$$

as $n \to \infty$ .So that we can't have $\lim_{n \to \infty} \sup_{x \in [0,1]} |f_n(x)-f(x)|=0$. i.e. there is no uniform convergence.


There is a way, a bit redundant, because it uses the fact that $n^{1/n} \to 1$ when $n \to \infty$. Anyway, here is it.

For, $x=0$ or $x=1$, we don't have anything to do. Note that, for $0<x<1$, we have that $$f_n(x)^{1/n}=n^{1/n}x^{1/n}(1-x) \to 1-x,$$ when $n \to \infty$. So, for each $\epsilon>0$, there is $N \in \mathbb{N}$, such that $$ n>N \Rightarrow (1-x)-\epsilon<f_n(x)^{1/n}<(1-x)+\epsilon.$$ Take then $\epsilon>0$ such that $(1-x)+\epsilon <1$. It follows that $$ n>N \Rightarrow f_n(x)<[(1-x)+\epsilon]^n<1.$$ Passing $n \to \infty$, we have that $f_n(x) \to 0$.


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.