Definition 4.8.1 (Hausdorff). Let $X$ be a topological space, then the following are equivalent:

  1. For any $x, y \in X$ with $x \ne y$, there exists $U \in \cn(x)$ and $V \in \cn(y)$ such that $U \cap V = \emptyset$.

  2. For every $x \in X$, $\bracs{x}= \bigcap_{U \in \cn(x)}\overline{U}$.

  3. Every convergent filter in $X$ has exactly one cluster point.

  4. Every filter in $X$ converges to at most one point.

  5. For any index set $I$, the diagonal $\Delta$ is closed in $X^{I}$.

  6. The diagonal $\Delta$ is closed in $X \times X$.

If the above holds, then $X$ is a T2/Hausdorff space.

Proof. $(1) \Rightarrow (2)$: Let $y \in X \setminus \bracs{x}$, then there exists $U \in \cn(x)$ and $V \in \cn(y)$ such that $U \cap V = \emptyset$. By (2) of Definition 4.5.2, $y \not\in \overline{U}\subset \bigcap_{U \in \cn(x)}\overline{U}$.

$(2) \Rightarrow (3)$: Let $\fF \subset 2^{X}$ be a filter and $x \in X$ such that $\cn(x) \subset \fF$, then

\[\bracs{x}= \bigcap_{U \in \cn(x)}\ol{U}\supset \bigcap_{U \in \cn(x)}\ol{U}\supset \bracs{x}\]

so $x$ is the only cluster point of $\fF$.

$(3) \Rightarrow (4)$: Let $\fF \subset 2^{X}$ be a filter. If $\fF$ converges to $x \in X$, then $x$ is a cluster point of $\fF$. As $\fF$ admits only one cluster point, $x$ is the only limit point of $\fF$.

$(4) \Rightarrow (5)$: Let $x \in \overline{\Delta}$, then by (4) of Definition 4.5.2, there exists $\fF \subset 2^{\Delta}$ converging to $x$. Let $i, j \in I$, then for any $y \in \Delta$, $\pi_{i}(y) = \pi_{j}(y)$. Thus $\pi_{i}(\fF) = \pi_{j}(\fF)$. By Proposition 4.7.2, $\pi_{i}(\fF)$ converges to $\pi_{i}(x)$ and $\pi_{j}(\fF)$ converges to $\pi_{j}(x)$. By assumption, $\pi_{i}(x) = \pi_{j}(x)$. Since this holds for all pairs $i, j \in I$, $x \in \Delta$.

$(5) \Rightarrow (6)$: Take $I = \bracs{1, 2}$.

$(6) \Rightarrow (1)$: Let $x, y \in X$ with $x \ne y$, then $\Delta^{c} \in \cn(x, y)$, and there exists $U \in \cn(x)$ and $V \in \cn(y)$ such that $(x, y) \in \pi_{1}^{-1}(U) \cap \pi_{2}^{-1}(V) \subset \Delta^{c}$. Thus $U \cap V = \emptyset$ contain the desired neighbourhoods.$\square$