19

In Folland's Real analysis, two of properties of measures are stated as follows:

Let $(X,\mathcal{M}, \mu)$ be a measure space.

Continuity from below: If $\{E_j\}_1^{\infty} \subset \mathcal{M}$ and $E_1 \subset E_2 \subset \cdots$, then $\mu(\bigcup_1^{\infty} E_j) = \lim_{j \to \infty} \mu(E_j)$

Similarly,

Continuity from above: If $\{E_j\}_1^{\infty} \subset \mathcal{M}$, $E_1 \supset E_2 \supset \cdots$ and $\mu(E_1) < \infty$. Then $\mu(\bigcap_1^{\infty} E_j) = \lim_{j \to \infty} \mu(E_j)$

I do not understand the point of these statements. How they are related to continuity for example?

  • 1
    It's reasonable for a continuous increasing function $f$ with real values, defined on the real line, to have $f(t_n)\uparrow f(t)$ when $t_n\uparrow t$ and $f(t_n)\downarrow f(t)$ when $t_n\downarrow t$. As a measure can take possibly infinite values, we have to be careful in this context, but that's the idea. – Davide Giraudo Nov 10 '12 at 17:48
  • Aside the word "continuity", what else exactly are you not understanding in these statements? – T. Eskin Nov 10 '12 at 19:21
  • 2
    Roughtly, a reason to call it a continuity property is that a phenomenon such $\lim_n g(x_n)=g(\lim_n x_n)$ is stated to be true. – Rgkpdx Feb 07 '16 at 16:11

2 Answers2

25

It's reasonable for a continuous increasing function $f$ with real values, defined on the real line, to have $f(t_n)\uparrow f(t)$ when $t_n\uparrow t$ and $f(t_n)\downarrow f(t)$ when $t_n\downarrow t$. As a measure can take possibly infinite values, we have to be careful in this context, but that's the idea.

The set functions $\mu\colon\cal M\to \overline{\Bbb R}_+$ is increasing in the sense that if $A\subset B$ then $\mu(A)\subset \mu(B)$. If $A_n\uparrow A$, then $\mu(A_n)\uparrow \mu(A)$.

We have to assumed that at least a set in $\{B_n\}$ is of finite measure in the case of decreasing sequence, to avoid counter-examples likes $X=\Bbb R$ with Lebesgue measure and $B_n=[n,+\infty)$ (like if we had a function defined on the real line with values in $\overline{\Bbb R}_+$ and such that $f(x)=0$ if $x\leqslant 0$, $f(x)=+\infty$ otherwise).

Davide Giraudo
  • 172,925
  • 1
    really nice concrete example. –  Feb 10 '15 at 11:35
  • I was wondering do we need the condition of ${B_n}$ to be finite in both continuity from above and below ? or only from above ?? –  Feb 10 '15 at 11:47
  • @user111750 We only need the condition of $B_n$ to be finite in the continuity from above. – Sam Wong Nov 28 '20 at 14:18
  • @DavideGiraudo Hi, could you explain again the counter example in the answer? sorry for my stupidity, but could you elaborate more why $B_n = [n, +\infty)$ is a counter example? – john_w May 04 '21 at 20:05
  • @john_w Because the intersection of all $B_n$s is empty, the measure of all these intersections is zero while the limit of $\mu(B_n)$ is $\infty$ as $n$ tends to $\infty$. – Apoorv Potnis Aug 01 '21 at 03:00
  • @ApoorvPotnis Hi could you explain why limit of $\mu(B_n)$ is $\infnty$ when n tends to infinity? I thought the lebesgue measure of $B_n$ (i.e. $\mu (B_n)$) would be decreasing if the set is decreasing? is it by definition that $\mu(B_n)$ is $\infnty$ when n tends to $\infnty$ even thought the set seems to become smaller and smaller? (so it is by definition right?) – john_w Aug 01 '21 at 18:20
  • @john_w We are using the counting measure on $B_n$, which is equal to the number of objects contained in the set if the set is finite, and infinite otherwise. Thus the measure of each of the $B_n$s is infinite as each $B_n$ contains infinitely many natural numbers. Thus the limit of measures of $B_n$s is not a real number. – Apoorv Potnis Aug 01 '21 at 21:27
  • @john_w "I thought the lebesgue measure of Bn (i.e. μ(Bn)) would be decreasing if the set is decreasing?" This not true. The set is indeed getting smaller intituitively as there are "less" numbers contained in $B_{n+1}$ compared to $B_n$. But, the number of elements in each $B_n$ is still infinite. – Apoorv Potnis Aug 01 '21 at 21:30
  • @john_w See https://en.wikipedia.org/wiki/Counting_measure – Apoorv Potnis Aug 01 '21 at 21:30
  • @ApoorvPotnis Hi, so in the notation $(X, M, \mu)$ of the question here, the $\mu$ here means Counting Measure? But just wondering there are also other "types" of measure right? But in this case, it is Counting Measure, is it correct? thank you – john_w Aug 02 '21 at 07:46
  • @john_w yes, the counting measure is generally the usual measure on countable measurable spaces. – Apoorv Potnis Aug 02 '21 at 11:55
1

How they are related to continuity for example?

As stated in the comments continuity of a function $\mu$ can be interpreted as a property that allows for the calculation $$\mu(\lim_{j \to \infty} E_j) = \lim_{j \to \infty}\mu(E_j).\tag{1}$$

$\mu$ is a measure, i.e. it is a mapping into a field of numbers, therefore the right-hand side of (1) makes sense (if the sequence converges).

Convergence for a sequence $(E_j)_{j \in \mathbb{N}}$ of sets

We want to give a meaning to the expression $\lim_{j \to \infty} E_j$. One way, would be to define a topology on a suitable collection of sets that includes the sets $(E_j)$ and check, whether convergence in this topological space implies (1).

A more pragmatical approach is the following

Definition (limes superior, limes inferior and limit).

Let $(E_j)_{j \in \mathbb{N}}$ a sequence of subsets of $X$. Then we call $$ \limsup_{j \to \infty} E_j := \bigcap_{j = 1}^\infty \; \bigcup_{j = k}^\infty A_j $$ and $$ \liminf_{j \to \infty} E_j := \bigcup_{j = 1}^\infty \; \bigcap_{j = k}^\infty A_j $$ the limes superior and limes inferior of $(E_j)$, respectively.

If $\limsup_{j \to \infty} E_j = \liminf_{j \to \infty} E_j$, we say that $(E_j)$ converges and define the limit $$ \lim_{j \to \infty} E_j := \limsup_{j \to \infty} E_j = \liminf_{j \to \infty} E_j. $$

Definition and Lemma (increasing and decreasing sequences). Let $(E_j)_{j \in \mathbb{N}}$ a sequence of subsets of $X$.

If $E_j \subset E_{j + 1}$ for all $j \in \mathbb{N}$ we call the sequence $(E_j)$ increasing. Every increasing sequence converges with limit $$ \lim_{j \to \infty} E_j = \bigcup_{j =1}^\infty E_j. $$

If $E_j \supset E_{j + 1}$ for all $j \in \mathbb{N}$ we call the sequence $(E_j)$ decreasing. Every decreasing sequence converges with limit $$ \lim_{j \to \infty} E_j = \bigcap_{j =1}^\infty E_j. $$

Interpretation of (1) in the sense of set-convergence

Using the notion of continuity, OP's reference may be read as follows:

Let $(X,\mathcal{M}, \mu)$ be a measure space.

Continuity from below: If $\{E_j\}_1^{\infty} \subset \mathcal{M}$ and $E_1 \subset E_2 \subset \cdots$, then $\mu(\lim_{j \to \infty} E_j) = \lim_{j \to \infty} \mu(E_j)$

Similarly,

Continuity from above: If $\{E_j\}_1^{\infty} \subset \mathcal{M}$, $E_1 \supset E_2 \supset \cdots$ and $\mu(E_1) < \infty$. Then $\mu(\lim_{j \to \infty} E_j) = \lim_{j \to \infty} \mu(E_j)$

el_tenedor
  • 3,530