In this proofwiki page, it is written without any justification that $\cap \mathcal{B}_x = \{x\}$ where $\mathcal{B}_x$ is some countable basis of a point $x$ in the uncountable set $S$ endowed with the cofinite topology. Why is this true? Indeed, I just proved that it cannot be true in an uncountable space with the cofinite topology. Is this proof flawed? This steps seems crucial in obtaining a contradiction.

What I am trying to do is prove that if $X$ is an uncountable set with the cofinite topology, then it cannot be first countable. I read this, but I couldn't see how to use Brian Scott's hints. Here is what he wrote:

To prove that $X$ is not first countable, you must show that some point of $X$ does not have a countable local base. All points of $X$ ‘look alike’ in the cofinite topology, so it doesn’t matter what point we pick, so let $x\in X$ be any point. Suppose that $\mathscr{B}=\{B_n:n\in\Bbb N\}$ is a countable local base of open sets at the point $x$, meaning that if $U$ is any open nbhd of $x$, then $x\in B_n\subseteq U$ for some $n\in\Bbb N$. For each $n\in\Bbb N$ let $F_n=X\setminus B_n$: $B_n$ is open, so by definition $F_n$ is finite. Let $F=\{x\}\cup\bigcup_{n\in\Bbb N}F_n$; $F$ is the union of countably many finite sets, so $F$ is countable. $X$ is uncountable, so there is some $y\in X\setminus F$. Let $U=X\setminus\{y\}$.

  • Is $U$ an open nbhd of $x$?
  • Is there any $n\in\Bbb N$ such that $x\in B_n\subseteq U$?

I would answer the first question affirmatively; and to answer the question, which is presumably "no", I would suppose that there is such an $n \in \Bbb{N}$. But I am having trouble identifying the contradiction.

Before I even sought help from google, I tried the following (and several variations on it). Let $x \in X$, and let $y \in X - \{x\}$. Then $X-\{y\}$ is open and therefore there is an open set $U_y$ such that $x \in U_y \subseteq X - \{y\}$. Thus we have a mapping $y \mapsto U_y$. I tried to show this mapping is injective, but I couldn't. If I had succeeded, then I believe this would have shown that if there were any basis at $x$, it would have to be uncountable, which means that $X$ is not first countable. But, as you may have noticed, I abjectly failed in this task.

Is there a way to get any of these to proves to work, especially the one I proposed?

  • $\begingroup$ Doesn't that statement follow from the contradiction hypothesis? That is, the proof begins by assuming the existence of a countable neighborhood basis, then proceeds to arrive at a contradiction. $\endgroup$ – Xander Henderson Oct 25 '17 at 13:11
  • $\begingroup$ Since no point has a countable neighbourhood basis, it is vacuously true. It's probably just badly expressed. Point is, the cofinite topology is $T_1$, so if $\mathcal{B}_x$ is a neighbourhood basis of $x$, then $\bigcap \mathcal{B}_x = \{x\}$. Since the intersection of countably many neighbourhoods of $x$ is larger than $\{x\}$, it follows that $\mathcal{B}_x$ cannot be countable. $\endgroup$ – Daniel Fischer Oct 25 '17 at 13:12
  • $\begingroup$ @XanderHenderson If $\mathcal{B}$ being a countable basis for $x$ implies $\bigcap \mathcal{B} = \{x\}$, then yes. But I don't see why this implication is true. $\endgroup$ – user193319 Oct 25 '17 at 13:13
  • $\begingroup$ Because in a $T_1$ space, that is the case for a neighbourhood basis. $\endgroup$ – Daniel Fischer Oct 25 '17 at 13:15

Here is a positive proof, which encapsulates the same basic idea, but framing them in a different way.

Let $U_n$, for $n\in\Bbb N$ be any countable number of distinct open sets such that $x\in U_n$ for all $n$. Then for every $n$, there is some finite set $y_n$ such that $U_n=X\setminus y_n$. Therefore, by DeMorgan's law, $$X\setminus\bigcap U_n=\bigcup y_n\quad\text{and so}\quad \bigcap U_n=X\setminus\bigcup y_n.$$

But now $y_n$ are all finite sets, so we have a countable union of finite sets. Therefore this union is countable. So $X\setminus\bigcup y_n=\bigcap U_n$ is a co-countable set, which cannot be a singleton since $X$ is uncountable. So $\bigcap U_n\neq\{x\}$.

In particular, it is impossible that $x$ has a countable neighborhood basis. $\square$


The key point here is that when intersecting a countable number of open sets, the complement is a countable union of finite sets, and therefore cannot be equal to $X$ minus a single point.

Finally, your proof idea is flawed. For any $x\in\Bbb R$ consider $U_y=\Bbb R\setminus\{y\}$ as your open set for all $y\neq x$. These are all open sets such that $x\in U_y$. So there are uncountably many of them, since if $y\neq y'$ we have $y\in U_{y'}\setminus U_y$ and $y'\in U_y\setminus U_{y'}$. But nonetheless, $\Bbb R$ is first countable, and in fact second countable.


The proof just uses the following general fact:

If $X$ is $T_1$ (or equivalently all sets of the form $\{x\}$ are closed) and $x \in X$ and $\mathcal{B}$ is any local base at $x$, then $\bigcap_{B \in \mathcal{B}} B = \{x\}$.

Proof: Of course $x$ is that intersection, so that's a trivial inclusion. Suppose that $p \in \bigcap_{B \in \mathcal{B}} B$ and that $p \neq x$. Then $\{p\}$ is closed and so $x \in X\setminus\{p\}$ which is open so (as we have a local base) for some $B_p \in \mathcal{B}$ we have $x \in B_p \subseteq X\setminus\{p\}$, so $p \notin B_p$, which contradicts $p \in \bigcap_{B \in \mathcal{B}} B$, so $p$ must equal $x$, which proves the other inclusion.

The cocountable topology is $T_1$ as one point sets are countable so closed.


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.