# Are all finite sets closed?

say $X=\{a,b\}$ be a set. The following is a topology on $X$.

$\tau=\{\{ a\}, \{a,b\}, \varnothing\}$

Then $b$ is a limit point of $a$, as all open sets $(\{a,b\})$ intersect $\{a\}$ at points other than $b$. Then how is it that all finite sets are closed? What I am doing wrong here!

• Note that $\{\phi\}$ is not a subset of $X$, whereas $\varnothing$ is a subset of $X$. Apr 23, 2015 at 11:44
• @AsafKaragila: Unless one of $a,b$ is $\emptyset$. ;) Apr 23, 2015 at 22:06
• @tomasz: Sure, but it wouldn't be a topology, and in case $\phi$ is a Greek letter, not the symbol for the empty set! :-) Apr 23, 2015 at 22:13

All finite sets in a $T_1$ space are closed. Spaces like $\Bbb R^n$ are Hausdorff and therefore $T_1$. But the Sierpinski space (which you suggest) is not a $T_1$ space, and so not every finite set there is closed.
Note that if all finite sets were closed, universally, then every topology on a finite set would be the discrete topology. But then the trivial topology, $\{X,\varnothing\}$ wouldn't be a topology on any non-singleton finite set.
Since $\{b\}$ is not in the topology, by definition it is not open, and hence its complement, $\{a\}$, is not closed.
Are you asking "how are all finite sets closed" in this particular case? Or are you providing a counter example as to why all finite sets are not closed. Under the right conditions. I know it is sufficient that you require each pair of point in a space $X$ have open neighborhoods that don't contain the other (i.e., $X$ is $T_1$ as was mentioned by Asaf Karagila). But for your example this is not the case (as $b$ has no open set that does not contain $a$), and so we can not assume that all finite sets will be open.