I want to prove that every closed subset of $\mathbb R$ is finite or countable or continuum.
I know that for arbitrary subset we can not make similar statements - because of continuum hypothesis.
It looks like I need some machinery to prove mentioned fact. I feel I need somehow characterize all closed subsets of $\mathbb{R}$. For open subsets I know that they are union of disjoint intervals. For closed I do not know anything like that. Intution tells me that some deep unknown for me fact should be used here.
Assume that $F$, the closed set, is uncountable. Then, we can show that there exist two disjoint closed intervals $I_0=[a_0,b_0]$ and $I_1=[a_1,b_1]$ such that $$ \lvert F\cap I_0\rvert>\aleph_0 \,\,\,\&\,\,\,\lvert F\cap I_1\rvert>\aleph_0. $$ In order to to this, we first pick an $x\in F$, such that $\lvert(x-\varepsilon,x+\varepsilon)\cap F\rvert>\aleph_0$, for all $\varepsilon>0$. If no such $x$ existed, then we could cover $F$ with countably many open intervals (since $F$ is separable), each of which intersects $F$ at countably many points, and hence $F$ would be countable. Clearly, this point $x$ can not be unique, for otherwise $F\!\smallsetminus\!\{x\}$ would similarly be covered by countably many open intervals, each intersecting $F$ at countably many points. If $x<y$ are two such points, and $3d=y-x$, then we can set $$ I_0=[x-d,x+d]\quad\text{and}\quad I_1=[y-d,y+d]. $$
Repeating the same argument, with $F\cap I_0$ and $F\cap I_1$ in the place of $F$, we next find four disjoint closed intervals $I_{0,0}$, $I_{0,1}$, $I_{1,0}$ and $I_{1,1}$, such that $I_{0,0},\, I_{0,1}\subset I_0$ and $I_{1,0},\, I_{1,1}\subset I_1$, and $$ \lvert F\cap I_{i,j}\rvert>\aleph_0, \quad i,j=0,1. $$ Continuing this procedure, for every sequence $j\in 2^{\mathbb N}=\{0,1\}^{\mathbb N}$ we have a shrinking to zero sequence of closed intervals $$ I_{j(1)}\supset I_{j(1),j(2)}\supset I_{j(1),j(2),j(3)}\supset I_{j(1),\ldots,j(n)}\supset\cdots $$ each of which has as an intersection a singleton $\{x_j\}$, and clearly $x_j\in F$, as $F$ is closed, for every $j\in 2^{\mathbb N}$. But, by construction of these closed intervals, if $j\ne j'$, then $x_j\ne x_{j'}$. Thus we have constructed an injection $\varphi :2^{\mathbb N}\to F$, and hence $$ \lvert F\rvert\ge 2^{\aleph_0}. $$