1
$\begingroup$

Let $z(x):=\int_{Y} f(x,y)d\mu(y)$ for $x \in \mathbb R$ be an integral function where $\mu$ is a finite(!) Borel measure on $Y$ and $x \mapsto f(x,y)$ is continuous for every $y.$

Moreover, we know that $\int_{ Y} \left\lvert f(x,y) \right\rvert d\mu(y)$ is uniformly bounded in $x.$

My question is whether this implies that $z$ is continuous as well?

What is the obstacle: It seems that the dominated convergence theorem does not apply so easily as $f(x,y)$ is not known to be uniformly bounded in both $x$ and $y.$

$\endgroup$
1
  • 1
    $\begingroup$ This is, in disguise, the fact that pointwise convergence plus $L^1$ bounded does not imply $L^1$ convergence. $\endgroup$ Commented Apr 22, 2018 at 17:09

1 Answer 1

2
$\begingroup$

$\newcommand{\de}{\delta} \newcommand{\De}{\Delta} \newcommand{\ep}{\epsilon} \newcommand{\ga}{\gamma} \newcommand{\Ga}{\Gamma} \newcommand{\la}{\lambda} \newcommand{\Si}{\Sigma} \newcommand{\thh}{\theta} \newcommand{\R}{\mathbb{R}} \newcommand{\F}{\mathcal{F}} \newcommand{\E}{\operatorname{\mathsf E}} \newcommand{\PP}{\operatorname{\mathsf P}} \newcommand{\ii}[1]{\operatorname{\mathsf I}\{#1\}}$

The answer is no. E.g., let $\mu(dy)=g(y)dy$ and $$f(x,y)=\frac1{x^2}g\Big(\frac{x-y}{x^2}\Big)$$ if $x\ne0$, and $f(0,y)=0$ for all real $x$ and $y$, where $g$ is the standard normal density: $g(y):=\frac1{\sqrt{2\pi}}\,e^{-y^2/2}$. Then $\mu$ is a probability measure on $Y=\R$, $x \mapsto f(x,y)$ is continuous for every $y$, and \begin{equation} z(x)=\int_\R |f(x,y)| \mu(dy) \end{equation} is $1$ for $x\ne0$ and $0$ for $x=0$.

$\endgroup$
2
  • $\begingroup$ Other example, with $\mu =$ Lebesgue on $[-1,1]$: mathoverflow.net/a/215865 $\endgroup$ Commented Apr 22, 2018 at 15:21
  • $\begingroup$ @FrancoisZiegler : That is a nice example. $\endgroup$ Commented Apr 22, 2018 at 15:31

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .