MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $(\Omega,\Sigma, \mu)$ be a probability measure and $X$ a Banach space. I am interested in subsets $F\subseteq L_\infty (\mu,X)$ that satisfy these two compactness conditions:

  1. $F$ is a norm-compact subset of $L_1(\mu, X)$; and
  2. For any sequence $f_n$ in $F$ there exists a set $E\in \Sigma$, $\mu(E)=1$, such that each $L_1$-norm convergent subsequence of $f_n$ converges pointwise on $E$.

An example of such a set is any compact subset $F$ of $L_\infty (\mu,X)$. To see this suppose that $f_n$ is a sequence in $F$. There is $E\in \Sigma$, $\mu(E)=1$, such that for any $m,n$ we have $\|f_m(\omega) - f_n(\omega)\|\leq \|f_m-f_n\|_\infty$ for all $\omega\in E$. Now if $f_m$ is an $L_1$-convergent subsequence of $f_n$, then it must be converging in $L_\infty$, and converging pointwise on $E$.

More generally, the compactness condition is satisfied by

  • $\star$ $F\subseteq L_\infty (\mu,X)$ and for every $\epsilon>0$ there exists $E\in \Sigma$, $\mu(E)>1-\epsilon$ with $$ \{f\chi_E: f\in F\} $$ is compact in $L_\infty (\mu,X)$.

An example of an $L_1$-compact set that does not satisfy the compactness conditions 2 or $\star$ is the set of monotone step functions $f\colon [0,1]\to \{0,1\}$.

Now for my question: Does there exist a set satisfying 1 and 2 but not satisfying the $L_\infty$ compactness condition $\star$?

I don't have an answer to this for the case $X=R$ and $\Omega=[0,1]$.

share|cite|improve this question
Notice that if $F$ is a set of functions that is compact and metrizable in the topology of pointwise convergence, then for every $\epsilon >0$ there exists $E$, $\mu(E)>1-\epsilon$ such that $$ \\{f\chi_E: f\in F\\} $$ is compact in the topology of uniform convergence. The proof of this is in "Markov Chains" D. Revuz (Proposition 5.5) – Rabee Tourky Oct 6 '12 at 12:10
The is (more or less) Egorov's theorem. – Jochen Wengenroth Oct 9 '12 at 9:11
@Jochen Well, yes sure it more or less smells like Egorov's theorem. But I've tried and I can't prove that 1 and 2 implies $\star*$. – Rabee Tourky Oct 9 '12 at 12:47
Thanks Jochen. – Rabee Tourky Oct 15 '12 at 22:52
up vote 3 down vote accepted

The answer is that (1) and (2) implies $\star$ (this resolves a nice problem in a game theory paper I'm working on though the final unresolved problem has to do with decomposable Banach spaces, which I'll ask in a different question).

Assume (1) and (2). Let $P=\{f_n\}$ be a sequence (actual versions of equivalence classes) in $X$ such that each $f\in X$ is the $L_1$ limit of some subsequence of $f_n$. Let $E$ be the measurable set in (2).

By (2), $P$ has compact closure $\overline{P}^p$ in the product topology $X^{E}$. Notice that $X= \overline{P}^{L_1}=\overline{P}^p$ also by (2) taken as equivalence classes. Thus, we can now quickly show by means of Egorov's theorem that for every $\epsilon>0$ there is $F\in \Sigma$, $\mu(F)>1-\epsilon$ such that $\overline{P}^p$ is compact in the topology of uniform convergence (on $F$). We have the result.

share|cite|improve this answer
For the record the affirmative answer can be obtained from known techniques: On pointwise convergence, compactness and equicontinuity in the lifting topology. by Ionescu Tulcea (1973) – Rabee Tourky Oct 18 '12 at 4:25

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.