Assuming the axiom of choice, is there a well-ordering of the reals such that every initial segment is closed for the usual topology? If the continuum hypothesis helps, we can also assume it.

An initial segment is a set of the form $\{x : x<y\}$ for some $y$, according to the well-ordering $<$.


3 Answers 3


No. There cannot be a strictly increasing $\omega_1$-sequence of closed sets in a topological space with a countable base. Their complements would be unions of open sets from the basis, and it is impossible to drop elements of a countable set uncountably many times.

  • 3
    $\begingroup$ The key property is that the reals are hereditarily separable HS. Every second countable space is HS but not the other way around. $\endgroup$ Nov 27, 2012 at 0:15

The answer is no. Assume there is such a well-order; then some downward-closed subset $D$ (which might be an initial segment, or might be all of $\mathbb{R}$) will be of order type $2^{\aleph_0}$, i.e., the least ordinal of continuum size. Being a subset of $\mathbb{R}$, $D$ contains a countable subset that is dense in $D$ (see for example here if you are in need of proof), say $x_1 \lt x_2 \lt \ldots$ according to how the elements appear in the well-order. Since there is no countable cofinal sequence in the order type $2^{\aleph_0}$, it must be that some proper initial segment contains all the $x_i$. This initial segment is both closed and dense, hence is all of $D$. Contradiction.

(Edited after Ramiro pointed out a glitch. Hopefully okay now.)

  • $\begingroup$ Very nice. Separability is weaker than second countability. $\endgroup$
    – Goldstern
    Nov 27, 2012 at 0:05
  • 1
    $\begingroup$ Todd, there are well orderings of the reals with a countable cofinal sequence. It might be for instance that all the rationals form the tail of the well ordering. $\endgroup$ Nov 27, 2012 at 0:10
  • $\begingroup$ you can even put a singleton at the tail ! $\endgroup$
    – Denis
    Nov 27, 2012 at 0:50
  • 1
    $\begingroup$ Yes, thanks both of you. The relevance of hereditary separability, pointed out by Ramiro under Goldstern's answer, makes itself felt here as well. $\endgroup$
    – Todd Trimble
    Nov 27, 2012 at 1:01


If there was such a well order, let $r$ be the first real for which the initial segment $I$ bounded by $r$ is uncountable. Since $I$ is separable (being a subspace of the reals) we can fix a countable $D \subseteq I$ dense in $I$. But $I$ has order type $\omega_1$ which is a regular cardinal (usinge some choice), so there would be an $s \in I$ that bounds $D$. This is already a contradiction because then the clousure of $D$ would be contained in the initial segment determined by $s$, which is a proper subset of $I$.

  • $\begingroup$ This r does not exist if we assume the continuum hypothesis. $\endgroup$
    – Denis
    Nov 27, 2012 at 0:14
  • 2
    $\begingroup$ You are right, if we assume CH this $r$ might not exist (IF the given well order happens to have type exactly $\omega_1$); in that case just ignore $r$ and take $I$ as the whole real line. $\endgroup$ Nov 27, 2012 at 0:19

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.