I haven't seen it stated as such, which is why I'm raising it here for confirmation.
In most references I'm seeing a homeomorphism $\phi$ between topological spaces $(W, \mathscr S)$ and $(X, \mathscr T)$ defined as:
- a bijection, $\phi: W \to T$
- which is continuous, i.e. $\phi^{-1} (T \in \mathscr T) \in \mathscr S$, where $\phi^{-1} (T \in \mathscr T)$ is the pre-image of $T$
- whose inverse $\phi^{-1}: T \to W$ is continuous.
It's fairly easy to show that a bijection $\phi: W \to T$ gives a bijection $\phi: \mathscr P (W) \to \mathscr P (T)$ between their powersets (with the inverse defined by the pre-image), so condition 3 becomes $\phi (S \in \mathscr S) \in \mathscr T$, i.e. $\phi $ is an "open map"
Recognizing that a topological space consists of a set and its collection of (open) subsets can't we then just say that a homeomorphism is mapping $\phi: W \to X, \phi: \mathscr S \to \mathscr T$ which is invertible in both cases.
Equivalently, since $\mathscr S , \mathscr T$ are also sets, a homeomorphism is a bijection between the underlying sets and between the corresponding topologies ?.