I was asked a question by someone the other day regarding the topology of pointwise convergence, and I can't seem to get anywhere with it. I was wondering if anyone could be of any assistance...

The question was: is $C[0,1]$, the set of continuous real-valued functions from $[0,1]$ to $\mathbb{R}$, with the topology of pointwise convergence (defined by the sub-basis $A_{a,x,b} :=\{f\in C[0,1] : a< f(x) < b\}$ for $x \in [0,1]$ and $a<b \in \mathbb{R}$) a normal topological space?

Wikipedia suggests that $C(\mathbb{R})$ is not normal, so I've tried to show this as a starting point, and then work out whether the same argument holds or fails when we restrict to $[0,1]$: but I can't really get anywhere with this either.

My gut instinct seems to tell me that $C[0,1]$ is not normal, and I have some rather vague hand-wavey intuition for why but I can't really pin anything down at all and have given myself a bit of a headache.

Could anyone point me in the right direction with a good hint or reference please? The question is quite interesting so I'd prefer to try and finish it myself than be given a full solution.


  • 1
    $\begingroup$ According to Problem 215 from Tkachuk: A $C_p$-Theory Problem Book for a metrizable space $X$ we havep $C_p(X$) is Lindelof $\Leftrightarrow$ $C_p(X)$ is normal $\Leftrightarrow$ $X$ is second countable. Solution is given on p.170, but it relies heavily on some problems done eralier in that book. $\endgroup$ – Martin Sleziak Jun 9 '12 at 16:19

Let $\mathscr{B}=\{B_n:n\in\omega\}$ be a countable base for $[0,1]$. For $m,n\in\omega$ let $$F(m,n)=\{f\in C_p[0,1]:f[B_m]\subseteq B_n\}\;.$$ Let $\mathscr{F}_0=\{F(m,n):m,n\in\omega\}$, and let $\mathscr{F}=\{\bigcap\mathscr{A}:\mathscr{A}\subseteq\mathscr{F}_0\text{ and }|\mathscr{A}|<\omega\}$; note that $\mathscr{F}$ is countable.

Suppose that $U$ is an open subset of $C_p[0,1]$, and $f\in U$. Then for some finite $\{x_1,\dots,x_n\}\subseteq[0,1]$ and $B_{k_1},\dots,B_{k_n}\in\mathscr{B}$ we have

$$f\in\Big\{g\in C_p[0,1]:g(x_i)\in B_{k_i}\text{ for }i=1,\dots,n\Big\}\subseteq U\;.$$

Since $f$ is continuous, for each $i=1,\dots,n$ there is a $B_{j_i}\in\mathscr{B}$ such that $x_i\in B_{j_i}$ and $f[B_{j_i}]\subseteq B_{k_i}$, i.e., $f\in F(j_i,k_i)$; clearly

$$f\in\bigcap_{i=1}^nF(j_i,k_i)\subseteq\Big\{g\in C_p[0,1]:g(x_i)\in B_{k_i}\text{ for }i=1,\dots,n\Big\}\subseteq U\;,$$ where $$\bigcap_{i=1}^nF(j_i,k_i)\in\mathscr{F}\;,$$ so $\mathscr{F}$ is a countable network for $C_p[0,1]$.

This is enough to ensure that $C_p[0,1]$ is Lindelöf. To see this, let $\mathscr{U}$ be an open cover of $C_p[0,1]$. For each $f\in C_p[0,1]$ there are a $U_f\in\mathscr{U}$ and an $F_f\in\mathscr{F}$ such that $f\in F_f\subseteq U_f$. Let $$\mathscr{F}_\mathscr{U}=\Big\{F_f:f\in C_p[0,1]\Big\}\;,$$ and for each $F\in\mathscr{F}_\mathscr{U}$ choose $U_F\in\mathscr{U}$ such that $F\subseteq U_F$; $\mathscr{F}_\mathscr{U}$ is countable, so $\{U_F:F\in\mathscr{F}_\mathscr{U}\}$ is a countable subcover of $\mathscr{U}$.

Finally, $C_p[0,1]$ is certainly $T_3$, and a Lindelöf $T_3$-space is normal, so $C_p[0,1]$ is normal.


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.