2
$\begingroup$

In Chapter 4.9 of the book "Measures of noncompactness and condensing operators" (Vol. 55 of *Operator theory: advances and applications), the authors mention the property "compactness in measure". They say

Here compactness in measure means compactness in the normed space $S$ of all measurable, almost everywhere finite functions $x$, equipped with the norm

$$\|x\| = \inf_{s>0} \big\{s + \text{mes} \{t \, : \, |x(t)| \geq s\}\big\}$$

Where "$\operatorname{mes} D$" means the measure of the set $D$.

My questions are: Does this property go by any other names? And are there good sources in English which mention it?

The only sources I can find which use it are papers by N. A. Erzakova, not all of which have been translated into English, and possibly a paper in Russian by P. P. Zabreiko.

$\endgroup$
6
  • $\begingroup$ Is the inf supposed to be over $s$? Are we working on any particular measure space, or a general one? I guess it has to be an infinite measure space? $\endgroup$ Nov 27, 2018 at 15:36
  • $\begingroup$ Yes, the inf is over $s$; I've fixed that. This topic comes up in the section on $L^p(V)$, where $V$ is a set of finite Lebesgue measure in a finite-dimensional space. $\endgroup$
    – Matt R.
    Nov 27, 2018 at 15:49
  • 3
    $\begingroup$ Then I'm confused how this is a norm. If $V$ has measure, say, 3, then no function could have norm greater than 3 (taking $s \to 0$). So it can't be homogeneous. Am I missing something? $\endgroup$ Nov 27, 2018 at 16:22
  • $\begingroup$ I guess the authors (who are they anyway?) are talking about the space $S$, also known as $L_0(\mu)$, equipped with the topology of convergence in measure, which can be generated by the metric $d(f,g)=\inf\{\varepsilon>0: \mu\{|f-g|\ge\varepsilon\}\le\varepsilon\}$. Reference to this topology is often made by saying ``in measure'', e.g., one can say that the unit ball of $L_1(\mu)$ is compact in this topology by saying it's compact in measure. $\endgroup$ Nov 27, 2018 at 22:45
  • $\begingroup$ In my previous comment I meant to say closed, not compact. There are subspaces of $L_1$ with a unit ball that is compact in measure; see G. Godefroy, N. Kalton, D. Li, J. Reine Angew. Math. 471, 43-75 (1996). $\endgroup$ Nov 28, 2018 at 6:13

1 Answer 1

2
$\begingroup$

Probably the authors refer to the space $L_0(\mu)$ of all (equivalence classes) of measurable functions. This is a complete metric space in the metric I have mentioned in a comment above or, which is closer to the quotation in the question, for the equivalent metric given by $d'(f,g)=\inf_{s>0} \{s+\mu\{|f-g|\ge s\}\}$, i.e., $d'(f,g)=\|f-g\|$ in the notation of the question. But note that $\|\,.\,\|$ is not a norm; $L_0(\mu)$ is a non-locally convex complete metric space. This topology on $L_0$ is called topology of convergence in measure, and one uses the epithet "in measure" to refer to this topology, hence the phrases "closed in measure" or "compact in measure". A fundamental reference for sets closed in measure is A.V. Bukhvalov, G.Ya. Lozanovskij, On sets closed in measure in spaces of measurable functions. Trans. Mosc. Math. Soc. 34, 127-148 (1978) (Russion version online). As for sets compact in measure see G. Godefroy, N. Kalton, D. Li, On subspaces of $L_1$ which embed into $\ell_1$. J. Reine Angew. Math. 471, 43-75 (1996).

Precise reference: Definition of equivalent metric for convergence in measure, §III.2 of Dunford & Schwartz, Linear Operators, Part I: General Theory (1958). For a theorem actually characterizing compactness in measure see Thm.IV.11.1 in Dunford & Schwartz.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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