The title pretty much is the question (by "closed interval" I mean to include possibly "degenerate" closed intervals consisting of a single point). The plausibility argument (I wouldn't quite call it a "proof") that leads me to think the answer is "yes" goes something like this:
We know the open set $O$ is a countable union of disjoint open intervals, call them $o_i$. It seems the closure of $O$ can, in general, introduce two kinds of limit points not already members of $O$:
- Any endpoint of an $o_i$ (thus each $o_i$ "becomes" a closed interval in $\overline{O}$)
- A point which is some finite distance from every $o_i$, but is the limit to which some (infinite) subset of the $o_i$ converges. It seems plausible (though I don't have a proof) that there would be at most countably many such limit points.
By this reasoning, the closure of any open subset of $\mathbb{R}$ would be equal to some countable union of closed (possibly "degenerate") intervals.
Is this right? Or am I missing some case where it fails to hold?