Let $X$ be a completely metrizable space and let $A\subset X$ be a co-meager subset.

Since $X$ is Baire, we know that there is some dense $G_\delta$ set contained in $A$, say $G\subset A$. But since $X$ is complete and $G$ is $G_\delta$, then $G$ is complete. Hence, as a complete subspace of another complete subspace, $G$ is closed.

But since $G$ is dense, $G=X$ which implies that $A=X$. So we conclude that any co-meager subset of a completely metrizable space, has to be the space itself. This is not true but I can't see where I made a mistake.

Can somebody help me? Thanks


Your mistake is "since $X$ is complete and $G$ is $G_\delta$, then $G$ is complete, hence . . . $G$ is closed." Nope. $\mathbb R$ is a complete metric space, and $(0,1)$ being an open set is a $G_\delta,$ but $(0,1)$ is not closed in $\mathbb R.$ You see, $(0,1)$ is completely metrizABLE (being homeomorphic to $\mathbb R$), but it is not complete with respect to the usual metric of $\mathbb R.$ A $G_\delta$ subset $G$ of a complete metric space $X$ is completely metrizable, there is a complete metric on $G$ which induces the topology which $G$ inherits as a subspace of $X,$ but in general it's not the same as the metric in $X.$ It can't be, unless $G$ happens to be closed in $X.$

  • $\begingroup$ Thanks to you it is now very clear the silly (but for me, subtle) mistake I made in my reasoning. $\endgroup$ – valls Sep 25 '16 at 8:17

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.