4
$\begingroup$

Given a topological space $X$, by its sequential convergence structure I mean the full information about the convergent sequences in $X$ together with their limits. (I guess to be formal, it is the subset of $X^{\mathbb{N}} \times X$ consisting of all pairs $( ( x_n )_{n=0}^\infty , x )$ such that $x = \lim_{n \to \infty} x_n$.)

It is well-known that a topology is not in general characterized by its sequential convergence structure. For example, if you give an uncountable set $X$ both the discrete and the co-countable topologies, then the convergent sequences are exactly the eventually constant sequences which have the obvious limit.

Other examples can be found in the answers to the following question:

Recall that a topological space $X$ is called Polish if it is separable and completely metrizable. It is called perfect Polish if, in addition, it has no isolated points.

The discrete topology on $\mathbb{N}$ is a common example of a Polish space, however it is very far from being perfect Polish (since all points are isolated). This brings me to my question:

Question: Are there examples of topologies $\mathcal{O}_p$, $\mathcal{O}_n$ on a set $X$ such that

  1. $( X , \mathcal{O}_p )$ is a perfect Polish space, and
  2. $( X , \mathcal{O}_n )$ is not a perfect Polish space, and
  3. $( X , \mathcal{O}_p )$ and $( X , \mathcal{O}_n )$ have the same "sequential convergence structure"?
$\endgroup$
3
$\begingroup$

Consider the usual topology on $[0,1]$, and the topology $T$ which consists of the empty set together with all cocountable sets that are open in the usual topology. Obviously $T$ is not perfect Polish (it is not even Hausdorff). However, I claim that the convergent sequences for $T$ are the same as for the usual topology. Since $T$ is contained in the usual topology, one direction is trivial.

For the nontrivial direction, suppose $(x_n)$ does not converge to $x$ in the usual topology. Then there is some open interval $U$ around $x$ and some subsequence $(x_{n_k})$ of our sequence which is never in $U$. Passing to a further subsequence, we may assume that $(x_{n_k})$ converges to some point $y\in[0,1]$ in the usual topology. Now let $A=\{y\}\cup\{x_{n_k}:k\in\mathbb{N}\}$ and $V=[0,1]\setminus A$. Then $V\in T$, and infinitely many terms of $(x_n)$ are not in $V$. Since $x\in V$, this means $(x_n)$ does not converge to $x$ with respect to $T$.

(Of course, there's nothing special about $[0,1]$ here: you could do the same thing starting with any nonempty perfect Polish space $X$ in place of $[0,1]$. If $X$ is not compact, then $(x_{n_k})$ may not have a convergent subsequence, but then you can just take $A=\{x_{n_k}\}$.)

$\endgroup$
4
$\begingroup$

The answer to your question is that yes, there are such spaces, but I don't know any easy example.

Consider the space $\ell^1$ with its usual topology and with the weak topology. The former is perfect Polish, while the latter isn't metrizable, however $\ell^1$ has the so called Schur's property, meaning that a sequence converges weakly iff it converges in norm (for a proof see Conway's A course in functional analysis, theorem V.5.2, this is a nontrivial result)

$\endgroup$

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.