Let $f_n:A\subseteq\mathbb{R}\to\mathbb{R}$

$f_n$ is uniformly Cauchy $\implies$ $\exists f:A\to\mathbb{R}$ : $f_n\xrightarrow{u}f$ in A


$\forall\varepsilon>0$ $\exists\nu$ : $\forall n,m>\nu$ $\sup_{x\in A}|f_n(x)-f_m(x)|<\varepsilon$

I don't know why we start like that, I know by definition that a uniformly Cauchy is the second step of this proof

$\implies \forall\varepsilon>0$ $\exists\nu\in\mathbb{N}$ : $\forall n,m>\nu$ $\forall x\in A$ $|f_n(x)-f_m(x)|<\varepsilon$

Where is sup now?

So if $m\to+\infty$

$\implies \forall\varepsilon>0$ $\exists\nu\in\mathbb{N}$ : $\forall n>\nu$ $\forall x\in A$ $|f_n(x)-f(x)|\leq\varepsilon$

Ok it's clear why m disappears but I don't understand why $<\varepsilon$ becomes $\leq\varepsilon$

$\implies \forall\varepsilon>0$ $\exists\nu\in\mathbb{N}$ : $\forall n>\nu$ $\sup_{x\in A}|f_n(x)-f(x)|\leq\varepsilon$

This is not clear, where does sup come from?

$\implies\lim_{n\to\infty}sup_{x\in A}|f_n(x)-f(x)|=0$

This is directly from the definition of limit, and this means that

$\implies f_n\xrightarrow{u}f$ in $A$

  • $\begingroup$ Please consider accepting my answer if it has helped :) $\endgroup$ – user667 Nov 19 '18 at 14:11

Recall the definition of supremum (it is the least upper bound). Thus, if we can choose $v$ such that $\sup_{x\in A}|f_n(x)-f_m(x)|<\epsilon$ whenever $n,m>v$ then we are guaranteed that $\forall x\in A, |f_n(x)-f_m(x)|<\epsilon$ whenever $n,m>v$. Going the other way around, strict inequality could become non strict since the supremum is an upper bound itself. Thus if $|f_n(x)-f(x)|<\epsilon$ for all $x\in A$, we have that $\epsilon$ is an upper bound over all $x$. In particular, it could be the least upper bound (supremum). Thus we have to write $\sup_{x \in A}|f_n(x)-f|\leq\epsilon$ since equality could hold.

I agree that this proof is somewhat unclear. To be completely rigorous, we use the triangle inequality. Choose arbitrary $x \in A$. We have, $$|f_n(x)-f(x)|\leq|f_n(x)-f_m(x)|+|f_m(x)-f(x)|$$where $f(x)$ is the pointwise limit of the sequence $\{f_n(x)\}$. We know that there is a pointwise limit because $\mathbb{R}$ is complete. By definition of pointwise convergence, there exists some $N_1$ such that if $m>N_1$ we have $|f_m(x)-f(x)|<\frac{\epsilon}{2}$. By assumption, we can choose $N$ such that if $n,m>N$ we have $|f_n(x)-f_m(x)|<\frac{\epsilon}{2}$. Now set $m>\max\{N_1,N\}$ (this step is a rigorous way of saying $m \rightarrow \infty$). Putting it all together $$|f_n(x)-f(x)|\leq|f_n(x)-f_m(x)|+|f_m(x)-f(x)|<\frac{\epsilon}{2}+\frac{\epsilon}{2}=\epsilon$$Because $x$ was an arbitrary point of $A$, we have found an $N$ (independent of $x\in A$) such that, $$\forall x\in A, n>N, |f_n(x)-f(x)|<\epsilon$$


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.