0
$\begingroup$

I understand the usual dominated convergence theorem is something as follows:

Let $\{f_n\}_{n\in\mathbb{N}}$ be a sequence of real-valued measurable functions on a measure space $(\mathscr{X},\mathcal{A},\mu)$. Moreover suppose that $f_n\rightarrow f$ in measure and $\left\vert f_n(x)\right\vert \le g(x)$ for all $x\in\mathscr{X}$ and $g$ is integrable. Then $\int f_n \rightarrow \int f$.

I now want to say/generalise this to:

1) If $f:\mathscr{Y}\times \mathscr{X}\longrightarrow \mathbb{R}$ such that \begin{equation*} \mu\left\{x\in\mathscr{X} : \lim_{y\rightarrow y_0} f(y,x) \ne f(y_0, x)\right\} =0 \end{equation*} and there exists a function $g:\mathscr{Y}\times \mathscr{X}\longrightarrow \mathbb{R}$ such that there exists $\tilde y\in\mathscr{Y}$ for which: \begin{equation*} \sup_{y\in \mathscr{Y}} \left\vert f(y, x)\right\vert \le g(\tilde y,x) \qquad \forall x\in\mathscr{X} \end{equation*} Then \begin{equation*} \lim_{y\rightarrow y_o} \int_{\mathscr{X}} f(y, x)d\mu = \int_{\mathscr{X}}f(y_0,x)d\mu \end{equation*}

2) And I want to even further generalise this to: For each $\epsilon>0$ \begin{equation*} \lim_{y\rightarrow y_o}\mu\left\{x\in\mathscr{X} : \left\vert f(y,x) - f(y_0, x)\right\vert >\epsilon\right\} =0 \end{equation*} and there exists a function $g:\mathscr{Y}\times \mathscr{X}\longrightarrow \mathbb{R}$ such that there exists $\tilde y\in\mathscr{Y}$ for which: \begin{equation*} \sup_{y\in \mathscr{Y}} \left\vert f(y, x)\right\vert \le g(\tilde y,x) \qquad \forall x\in\mathscr{X} \end{equation*} Then \begin{equation*} \lim_{y\rightarrow y_o} \int_{\mathscr{X}} f(y, x)d\mu = \int_{\mathscr{X}}f(y_0,x)d\mu \end{equation*}

In saying this I have spent the last two afternoons trying to prove these statements to myself, and I am not very confident in my proofs. Could someone please provide a rigorous proof of these statements as I could definitely benefit from learning how other's have approached the problem as well? Many thanks.

$\endgroup$
  • 1
    $\begingroup$ What do you mean by $y \to y_0$? Your measure space also carries a topological structure? $\endgroup$ – Vinícius Novelli Mar 8 '18 at 1:27
  • 1
    $\begingroup$ What do you mean by $$\begin{equation*} \sup_{y\in \mathscr{Y}} \left\vert f(y, x)\right\vert \le g(y,x) \qquad \forall x\in\mathscr{X} \end{equation*}$$ The right side has a free $y$. $\endgroup$ – Robert Israel Mar 8 '18 at 1:43
  • $\begingroup$ @ViníciusNovelli. Sorry I was not specific enough - the specific example I actually want to apply this question/theorem to is where $\mathscr{Y}$ and $\mathscr{X}$ are euclidean spaces with the usual euclidean norm/metric. However I was hoping that in this question I could just have $\mathscr{X}$ and $\mathscr{Y}$ as general metric spaces with respective metrics $d_X$ and $d_Y$...So when I say $y\rightarrow y_0$ then I would mean this in the usual sense of pointwise convergence of $y$ to $y_0$ in a metric space. $\endgroup$ – User086688 Mar 8 '18 at 5:04
  • $\begingroup$ @RobertIsreal. Ah very good point. With this statement I am just trying to capture that somehow $f(y, x)$ is "dominated". Perhaps I should reword this as $\exists$ $\tilde y\in\mathscr{Y}$ such that: \begin{equation*} \sup_{y\in\mathscr{Y}}\left\vert f(y,x)\right\vert \le g(\tilde y, x) \quad \forall x\in\mathscr{X} \end{equation*} I will edit the question accordingly now - thank you for pointing this out! $\endgroup$ – User086688 Mar 8 '18 at 5:08
0
$\begingroup$

The statement $ \lim_{y \to y_0} \int f(y,x) d\mu(x) = \int f(y_0,x) d\mu(x)$ is equivalent to the statement $ \lim_{n \to \infty} \int f(y_n,x) d\mu(x) = \int f(y_0,x) d\mu(x)$ for every sequence $y_n \to y_0$. So I don't see anything new here. It should be noted, however, that measurability of the limit is an essential item in the hypothesis: a limit through continuum of measurable functioins is not automatically measurable.

$\endgroup$

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.