1
$\begingroup$

I'm hoping to make a generalization of the answer to this question. Let's say that instead that we're composing two uniformly continuous function sequences, does this composition converge uniformly to $f \circ g$.

Here's why I think it does.

Consider $\left( f _ { k } \right) _ { k = 1 } ^ { \infty }$ and $( g_k ) \stackrel { \infty } { k } = 1$ to be sequences of continuous functions $[ 0,1 ] \rightarrow [ 0,1 ]$ converging uniformly to $f$ and $g : [ 0,1 ] \rightarrow \mathbb { R }$ respectively.

By applying the Weierstrass M-Test, $ \exists \quad M > 0$ so that $|f_k(x)| \leq M \quad \forall x \in [0,1]$ and $k \in \mathbb{N}$

Now restrict the domain of $(g_k)$ to $[ - M , M ]$. Since the codomain of $(f_k)$ is $[0,1]$, this $[ - M , M ]$ will be within that domain. Therefore, there exists $N \in \mathbb{N}$ such that $(g_k)$ converges uniformly within this restricted domain.

Therefore

\begin{equation} \left\| g _ { k } ( f_k(x )) - g(f ( x )) \right\| < \varepsilon \text { for all } x \in S \text { and } k \geq N \end{equation} so $f _ { k } \circ g _ { k }$ converge uniformly to $f \circ g$.

Is it that simple, or am I missing something?

$\endgroup$
3
  • $\begingroup$ It is not that simple: math.stackexchange.com/questions/1680370/… $\endgroup$ – Matt A Pelto Dec 4 '18 at 0:35
  • $\begingroup$ How do I show Lipshitz then? Your link doesn't make sense... $\endgroup$ – Andrew Hardy Dec 4 '18 at 0:40
  • $\begingroup$ See zhw.'s response, since $f$ is continuous it is uniformly continuous on a compact subset. There are some errors in your post that make it confusing to read. $\endgroup$ – Matt A Pelto Dec 4 '18 at 0:51
0
$\begingroup$

Suppose $g_n: [0,1] \longrightarrow [0,1]$, $\,f_n:[0,1] \longrightarrow \mathbb R$ $(n=1,2,\ldots)$ are two sequences of continuous functions that converge uniformly to $g$ and $f$, respectively.

Let $\varepsilon>0$ be given. By the uniform limit theorem and by the Heine-Cantor theorem, $f$ is uniformly continuous on $[0,1]$. Thus there is $\delta>0$ so that $|\,f(u)-f(v)|<\frac12 \varepsilon$ whenever $|u-v|<\delta$ and $u,v \in [0,1]$. Since $g_n \to g$ uniformly, there is a positive integer $N_1$ so that for all $x \in [0,1]$ we have $$|g_n(x)-g(x)|<\delta \text{ whenever } n \geq N_1.$$ Since $f_n \to f$ uniformly, there is a positive integer $N_2$ so that for all $x \in [0,1]$ we have $$|\,f_n(x)-f(x)|<\frac{\varepsilon}{2} \text{ whenever} n \geq N_2.$$ Set $N=\max\{N_1,N_2\}$. So if $n \geq N$ and $x \in [0,1]$, then we have \begin{equation} \begin{split} | \,f_n(g_n(x)) - f(g(x))| &= |\,f_n(g_n(x)) - f(g_n(x))+f(g_n(x)) - f(g(x)) |\\ & \leq |\,f_n(g_n(x)) - f(g_n(x))|+|\,f(g_n(x)) - f(g(x))| \\ & <\frac{\varepsilon}{2} +\frac{\varepsilon}{2}\\ & = \varepsilon. \end{split}\end{equation}

Therefore, $f_n \circ g_n \to f \circ g$ uniformly on $[0,1]$ (definition of uniform convergence).

$\endgroup$
3
  • 1
    $\begingroup$ OK let me paraphrase to see if I follow every step. Boundedness of the domain gives us uniform continuity of f, Then we use uniform converge to restrict g small enough so that it fits into the epsilon requirement of f. Then using an e/2 argument & triangle inequality, you show the condition for uniform converge? $\endgroup$ – Andrew Hardy Dec 4 '18 at 1:57
  • $\begingroup$ Precisely, the only thing I would correct from your comment is that we need the domain of $f$ to be closed and bounded. Here is a proof of the proposition that I referred to as the Heine-Cantor theorem: math.stackexchange.com/questions/2757071/… $\endgroup$ – Matt A Pelto Dec 4 '18 at 2:06
  • $\begingroup$ For the record, having every function in the sequence of functions $\{g_n\}_{n=1}^\infty$ map into the same codomain $[0,1]$ is an important assumption unless we want to just initially assume that $f$ is uniformly continuous. $\endgroup$ – Matt A Pelto Dec 4 '18 at 3:06

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.