Denote by $D$ the set of all continuous, increasing functions $f: \mathbb{R} \to [0, \infty)$ such that $\lim \limits_{x \to - \infty} f(x) = 0 $ and $\lim \limits_{x \to + \infty} f(x) = 1 $. If $f_n,f \in D$ for all $n \geq 1$ and $f_n \to f$ pointwisely on $\mathbb{R}$, prove that $f_n \to f$ uniformly on $\mathbb{R}$.

My take on it:

NTS: $f_n\to f$ uniformly on $\mathbb{R}$ if $M_n = \sup|f_n(x) - f(x)|$ exists for all sufficiently large n, and $\lim\limits_{n \to \infty} M_n = 0, x \in \mathbb{R}$.

$\lim\limits_{x \to \infty} f(x) = 1$ $\iff$ $\forall \epsilon > 0$ given $\exists M \gt0$ such that $|f(x) - 1| \lt \epsilon$,

$\lim\limits_{x \to - \infty}f(x) = 0 \iff \forall \epsilon \gt 0$ given $\exists M \gt 0$ such that $|f(x)| \lt \epsilon$

$f_n \to f$ pointwisely in $\mathbb{R} \iff \lim\limits_{n\to\infty}f_n(x)=f(x)$ $\forall x \in \mathbb{R}$.

$\forall \epsilon \gt 0, \exists M \gt0$ such that $|f_n(x) - f(x)|\lt\epsilon$ $\forall n \geq M, \forall x \in \mathbb{R}$

$\implies \lim\limits_{n \to \infty}|f_n(x)-f(x)|=0$

$\implies \lim\limits_{n \to \infty}\sup{|f_n(x)-f(x)|:x\in\mathbb{R}} = 0$.

If someone can please help me and let me know if I am going in the right direction and how to proceed forward.

Thanks in advance.

  • $\begingroup$ Remark: $D$ is the collection of functions that correspond to the distribution function (CDF) of continuous random variables. $\endgroup$ – Sean Roberson Nov 27 '18 at 19:52

Since $f\to 1$ and $f\to 0$ as $x\to \infty$ and $x\to -\infty$ respectively, for fixed $\epsilon>0$ there exists $M$ s.t. $|x|\ge M$ implies either $f(x)>1-\epsilon$ or $f(x)<\epsilon$. By the convergence of $f_n$, there exists $N$ s.t. $n\ge N$ implies $f_n(M)>1-\epsilon$ and $f_n(-M)<\epsilon$. But each $f_n$ is increasing, so $|x|>M$ implies either $f_n(x)>1-\epsilon$ or $f_n(x)<\epsilon$. Therefore, for $x$ outside $[-M,M]$, $|f_n(x)-f(x)|<2\epsilon$ by the Triangle Inequality.

By the continuity of $f$ on the compact $[-M,M]$, there exists $\delta>0$ s.t. $|x-y|<\delta$ implies $|f(x)-f(y)|<\epsilon$. Partition $[-M,M]$ into intervals $A_i=[t_i,t_{i+1})$ of length shorter than $\delta$. For each $t_i$ there exists $N_i$ where $n\ge N_i$ implies $|f_n(t_i)-f(t_i)|<\epsilon$. Fix $T = \max\{N_i, N\}$ and pick $x\in A_i$ and $n\ge T$. It follows that $f_n(t_{i+1})-f_n(x)\le f_n(t_{i+1})-f_n(t_i)\le (f(t_{i+1})+\epsilon)-(f(t_i)-\epsilon)\le 3\epsilon$. By the Triangle Ineq. $$\begin{align}|f_n(x)-f(x)|&\le |f_n(x)-f_n(t_{i+1})|+|f_n(t_{i+1})-f(t_{i+1})| + |f(t_{i+1})-f(t_i)| \\ &\le 5\epsilon \end{align}$$ But $A_i$ cover $[-M,M]$, so this holds for all $x \in [-M,M]$. Thus, $n\ge T$ forces $|f_n(x)-f(x)|\le 5\epsilon$ and the convergence is uniform.

  • $\begingroup$ Nice (full) answer (+1) $\endgroup$ – Robert Z Nov 28 '18 at 6:44

No, the last step does not work: pointwise convergence does not imply uniform convergence. You are not using the increasing property of such functions.

Hint. Use the second theorem given HERE on the compact set $[-R,R]$ (it is a variant of Dini's Theorem) and use the assumption about the limits at $\pm\infty$ in order to have the uniform convergence also on the complement of $[-R,R]$.

  • 1
    $\begingroup$ The version of Dini's Theorem you link to requires $f_n$ to converge monotonically to $f$ but in this problem, that need not be the case. Is there a more general version of the theorem? $\endgroup$ – Guacho Perez Nov 27 '18 at 21:41
  • $\begingroup$ @GuachoPerez Thanks for pointing out. I edited my answer. $\endgroup$ – Robert Z Nov 28 '18 at 6:42

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.