2
$\begingroup$

The de la Vallée Poussin criterion (which is often used in combination with the Dunford-Pettis theorem) is usually formulated for probability measures/finite measures, for example in [Bogachev: Measure Theory. (2007), Thm. 4.5.9.] the following version is given:

Let $(X, \mathfrak{A}, \mu)$ be a finite measure space. A family of functions $\Phi \subset L^1(X, \mu)$ is uniformly integrable if and only if there exists a nonnegative increasing function $G$ on $[0, + \infty)$ such that \begin{equation*} \lim\limits_{t \to +\infty} \frac{G(t)}{t} = \infty \quad \text{and} \quad \sup\limits_{f \in \Phi} \int_{X} G(|f(x)|) \, \mathrm{d}\mu (x) < \infty. \end{equation*} In such a case, one can choose a convex increasing function $G$.

The common definition of uniform integrability in the general case is:

The set $\Phi \subset L^1(X, \mu)$ is called uniformly integrable if $$ \forall \varepsilon > 0 \, \exists g \in L^1(X, \mu): \, \int_{\{|f| > g\}} |f| \, \mathrm{d} \mu \leq \varepsilon, \, \forall f \in \Phi.$$

But is there a version for infinite measures too? At least under certain conditions like $\sigma$-finite measures or atomless measures or at least for the (weighted) Lebesgue measure? Or is there an easy way to generalize it?

Would be very grateful for any references and help!

$\endgroup$

1 Answer 1

3
$\begingroup$

This result holds for any measure space.

Indeed, let $G$ be a Borel-measurable nonnegative function $G$ on $[0,\infty)$ such that $\lim_{t\to\infty}G(t)/t=\infty$ and hence $$\lim_{c\to\infty}h(c)=\infty,\text{ where }h(c):=\inf_{t>c}G(t)/t \tag{1}\label{1}$$ for real $c>0$. Assume also that $$S:=\sup_{f\in\Phi}\int_X d\mu\,G(|f|)<\infty. \tag{2}\label{2}$$ Then $\Phi$ is uniformly integrable.

Indeed, for each real $c>0$ $$\sup_{f\in\Phi}\int_A d\mu\,|f|\,1(|f|\le c) \le c\sup_{f\in\Phi}\int_A d\mu\to0 \tag{3}\label{3}$$ if $\mu(A)\to0$. Also, in view of \eqref{1} and \eqref{2}, for real $c>0$, $$\sup_{f\in\Phi}\int_A d\mu\,|f|\,1(|f|>c) \le\frac1{h(c)}\sup_{f\in\Phi}\int_X d\mu\,G(|f|)=\frac S{h(c)}\to0 \tag{4}\label{4}$$ as $c\to\infty$.

The desired result follows from \eqref{3} and \eqref{4}. $\quad\Box$

$\endgroup$
4
  • $\begingroup$ Could you maybe give a reference for this theorem and similar ones, if there is one? $\endgroup$
    – kumquat
    Commented May 4 at 21:30
  • $\begingroup$ @kumquat : I have not seen this result elsewhere. It is just a very straightforward generalization of the de la Vallée Poussin theorem. $\endgroup$ Commented May 5 at 1:25
  • $\begingroup$ Ok, I see. In Ekeland/Temam (Convex Analysis and Variational Problems, 1999, p.239, Thm. 1.3) I read a version for the Lebesgue measure (even with unbounded subsets), which uses the same assumptions as the Theorem for the case of finite measures. So I was wondering if the Theorem for the finite case would be applicable in the case of 𝜎-finite and atomless measures too? (Or with other more general assumptions.) At least to obtain uniformly absolutely continuous integrals? $\endgroup$
    – kumquat
    Commented May 5 at 14:13
  • $\begingroup$ @kumquat : The desired result actually holds for any measure $\mu$, as is now shown. $\endgroup$ Commented May 5 at 15:10

You must log in to answer this question.

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