While playing around with some basic general topology, I have thought of some problems whose solutions are not so obvious (at least to me), and surprisingly I do not remember having seen these anywhere.

Disclaimer #1 : I have already posted those problems elsewhere, but I was unlucky and received no answer.

Disclaimer #2 : The French and English notions of compact sets are notoriously different. I will try to write it right (and it might be actually easier in English).

So, let us take any topological space $(X, T)$. As a matter of prophylaxy, I will assume that $X$ is not empty. A subset $K$ of $X$ is said to be compact if, from any covering of $K$ by open sets one can extract a finite sub-cover.

My first question is: When are the compact subsets for the topology $T$ the closed subsets for another topology $T^*$ (modulo the whole space $X$)?

Trivially, any finite union of compact subsets is compact. If any intersection of compact subsets is compact, then the set of compact subsets plus $X$ are the closed subsets for some topology $T^*$ on $X$. One can see the open subsets of $(X,T^*)$ as the open neighborhoods of infinity for the Alexandroff compactification of $(X, T)$, without said infinity. If $(X,T)$ is Hausdorff then any intersection of compact sets is compact (this is because the compact sets are then closed). However, this condition is not necessary: with the coarse or cofinite topology, any subset is compact, so that any intersection of compact subsets is compact. Do the topological spaces for which this property holds have a name, or a characterization?

My second question is: Assuming that the property discussed above is satisfied, is there anything interesting to be said about the operation $T \to T^*$?

For instance, if $T$ has the property that any intersection of compact subsets is compact, then does $T^*$ have the same property? This would be important as it would allow us to iterate the operation $T \to T^*$. In addition, it looks like $T^*$ can be seen, in some cases, as a dual topology on $X$. If $T$ is the coarse topology on $X$, then $T^*$ is the discrete topology, and $T^{**}$ is the cofinite topology. Then the sequence stabilizes, as $T^{***}$ is again the discrete topology. In a similar way, If $T$ is the usual topology on $\mathbb{R}^d$, I think (I have not written down the argument properly) that $T=T^{**}$. Are those special cases, or is there a more general pattern?

  • $\begingroup$ KC-space is a space in which every compact subset is closed. So every KC-space has this property. (But as your example with indiscrete spaces shows, the two properties are not equivalent.) $\endgroup$ – Martin Sleziak Dec 1 '11 at 7:19

I don’t have an answer to the first question, but I can show that the desired property isn’t necessarily preserved by the $\tau\mapsto\tau^*$ operation. (Barring mental hiccups, anyway: keeping track of these topologies is a bit of a pain.)

Here’s an example of a non-Hausdorff KC-space $\langle X,\tau\rangle$ such that $\langle X,\tau^*\rangle$ is also a non-Hausdorff KC-space, but $\langle X,\tau^{**}\rangle$ does not have the desired property: $\langle X,\tau^{**}\rangle$ is the space obtained by splitting the point $\omega_1$ in two in the compact Hausdorff space $\omega_1+1$ (with the order topology), as is done in constructing the line with two origins. The space $\langle X,\tau\rangle$ was originally constructed by Bill Fleissner as an example of a KC-space whose topology contains no minimal KC-topology.

Let $p$ and $q$ be two points not in $\omega_1$, and let $X=\omega_1\cup\{p,q\}$. Let $\tau_0$ be the order topology on $\omega_1$. Let $I=\{0\}\cup\{\xi+1:\xi\in\omega_1\}$, the set of isolated points of $\omega_1$. For $\alpha<\omega_1$ let $$B_\alpha(p)=\{p\}\cup I\setminus\alpha\text{ and }B_\alpha(q)=\{q\}\cup I\setminus\alpha\;,$$ and let $\tau$ be the topology on $X$ generated by the base $$\tau_0\cup\{B_\alpha(p):\alpha<\omega_1\}\cup\{B_\alpha(q):\alpha<\omega_1\}\;.$$ Let $\mathscr{K}_0$ be the set of $\tau_0$-compact subsets of $\omega_1$; $\mathscr{K}_0$ is simply the set of countable $\tau_0$-closed subsets of $\omega_1$. Let $\mathscr{K}\;$ be the set of $\tau$-compact subsets of $X$; clearly $$\mathscr{K}=\big\{K\cup F:K\in\mathscr{K}_0\land F\subseteq \{p,q\}\big\}\;,$$ which is precisely the set of countable $\tau$-closed subsets of $X$, and it follows that $$\tau^*\triangleq\{X\setminus K:K\in\mathscr{K}\;\;\}\cup\{\varnothing\}$$ is a topology on $X$. Let $\tau_0^*=\tau^*\cap\wp(\omega_1)\setminus\{\varnothing\}$; $\tau_0^*$ is the set of co-countable $\tau_0$-open subsets of $\omega_1$, and $$\tau^*=\big\{V\cup F:V\in\tau_0^*\land F\subseteq \{p,q\}\big\}\cup\{\varnothing\}\;.$$

Let $\mathscr{K}\;^*$ be the set of $\tau^*$-compact subsets of $X$; $$\mathscr{K}\;^*=\mathscr{K}\cup\big\{C\cup F:C\subseteq\omega_1\text{ is a }\tau_0\text{-cub} \land F\subseteq \{p,q\}\big\}\;.$$

Clearly $\tau^{**}\triangleq \{X\setminus K:K\in\mathscr{K}\;^*\}$ is a topology on $X$, and it’s not hard to see that $$\tau^{**}=\big\{V\cup F:V\in\tau_0\land F\subseteq\{p,q\}\big\}\subsetneqq\tau\;.$$

The subspaces $\omega_1\cup\{p\}$ and $\omega_1\cup\{q\}$ of $\langle X,\tau^{**}\rangle$ are homeomorphic in an obvious way to the compact Hausdorff space $\omega_1+1$ (with the order topology), and $\langle X,\tau^*\rangle$ is therefore $\omega_1+1$ with the point $\omega_1$ split into $p$ and $q$, as mentioned above.

Finally, let $\mathscr{K}\;^{**}$ be the set of $\tau^{**}$-compact subsets of $X$; $$\mathscr{K}\;^{**}=\mathscr{K}\cup\big\{C\cup F:C\subseteq\omega_1\text{ is a }\tau_0\text{-cub} \land\varnothing\ne F\subseteq \{p,q\}\big\}$$ and is not closed under intersections, since $X\setminus\{p\},X\setminus\{q\}\in\mathscr{K}\;^{**}$, but $X\setminus\{p,q\}\notin\mathscr{K}\;^{**}$.

  • $\begingroup$ Thank you for your answer (that I've only remarked now). It may take me some time to understand it, but it looks nice. $\endgroup$ – D. Thomine Jan 4 '12 at 23:26

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.