3

The following is a standard result of real-valued measurable functions:

Theorem: let $\{f_n\}$ be a sequence of measurable functions $(X,\Sigma_X)\to(\mathbb{R},\mathcal{B}_{\mathbb{R}})$, where $\mathcal{B}_{\mathbb{R}}$ is the Borel $\sigma$-algebra generated by the Euclidean topology on $\mathbb{R}$. Then $$f_n\to f \text{ (pointwise)} \implies f \text{ is measurable}.$$


More generally, given a measurable function $(X,\Sigma_X)\to(Y,\Sigma_Y)$, if the codomain measure space is Borel we can already talk (because there is a topology on $Y$) about limits of sequences i.e. questions such as $$\text{Does $f_n(x)$ have any limits as $n\to\infty$?}$$ make sense. If the topology in question is Hausdorff, the limits -if they exist- must be unique, and we can talk about the function $$f:X\to Y:x\mapsto \lim_{n\to \infty}f_n(x).$$

With all that in mind, I'm wondering if the following, more general result holds:

Theorem (?): let $\{f_n\}$ be a sequence of measurable functions $(X,\Sigma_X)\to(Y,\mathcal{B}_{Y})$, where $\mathcal{B}_Y$ is the Borel $\sigma$-algebra generated by some Hausdorff topology $\tau$ on $Y$. Then $$f_n\to f \text{ (pointwise)} \implies f \text{ is measurable}.$$


I know two proofs of the theorem involving real-valued functions: one which uses $\liminf f$ and $\limsup f$, and another which works with $\pi$-systems and the fact that $f$ is measurable if and only if $$\{f\le c\} := \{r\in\mathbb{R} : f(r) \le c\}\in\Sigma_X \ \text{for any } c\in\mathbb{R}.$$ Neither proof seems generalizable.


In this post the OP asks whether the above result is true for first countable Hausdorff spaces, and it seems they recieve no solid answer on the matter.

I'm asking whether it holds for any Hausdorff space.

Sam
  • 5,208
  • the change is only euclidean to hausdorff? – BCLC Nov 12 '22 at 12:43
  • 1
    @BCLC and that the topology need no longer be on $\mathbb{R}$. To be clearer, the codomain is now $(Y,\mathcal{B}_Y)$ where $Y$ is an arbitrary non-empty set and $\mathcal{B}_Y=\sigma(\tau)$ where $(Y,\tau)$ is an arbitrary Hausdorff topological space. – Sam Nov 12 '22 at 12:47
  • 2
    The result may be false if topological space $(Y,\mathcal{B}_Y)$ is not metrizable. See this posting for example which discusses a counter example by Dudley. – Mittens Nov 12 '22 at 17:15
  • 2
  • @Ramiro that is a very enlightening post, but I do not think it answers my questions. I'm asking specifically whether the result holds for general Hausdorff topological spaces, while that post shows it holds for regular spaces where each closed set has a countable basis. – Sam Nov 14 '22 at 12:29
  • The result holds for spaces more general than metric spaces (as discussed in https://math.stackexchange.com/questions/1343860/limit-of-measurable-functions-is-measurable ). But, for sure, the result does not hold for general Hausdorff topological spaces, as clearly shown in @OliverDíaz answer in https://math.stackexchange.com/questions/4396439/is-fa-b-times-omega-to-e-measurable – Ramiro Nov 29 '22 at 01:44

0 Answers0