0
$\begingroup$

I have included some definitions right before the theorem, for convenience.

$\newcommand{\cA}{ {\cal A} }$ $\newcommand{\bR}{ {\mathbb{R}} }$ $\newcommand{\cP}{ {\cal P} }$ $\newcommand{\cM}{ {\cal M} }$

Definition. Let $\cA$ be a set of subsets of $\bR^n$.
Then $\cA$ is a field of subsets of $\bR^n$ when it has the following properties:

(FS-1) The empty set $\emptyset$ belongs to $\cA$.

(FS-2) If $A,B \in \cA$, then the set-difference $A \setminus B$ still belongs to $\cA$.

(FS-3) If $A_1, \ldots , A_k \in \cA$, then the union $A_1 \cup \cdots \cup A_k$ and the intersection $A_1 \cap \cdots \cap A_k$

Definition. Volume function $\mbox{vol}_o : \cP_n \to [0, \infty )$, where for $P = ( a_1, b_1 ] \times \cdots \times ( a_n, b_n ] \in \cP_n$, $\mbox{vol}_o (P) := (b_1 -a_1) \cdots (b_n -a_n)$.

Definition. Half-open rectangle in $\bR^n$ is a set of the form $P = ( a_1, b_1 ] \times \cdots \times ( a_n, b_n ] \subseteq \bR^n$, with $a_1 < b_1, \ldots , a_n < b_n$ in $\bR$. The collection of all half-open rectangles in $\bR^n$ is denoted by $\cP_n$.

Definition. Let $A \subseteq \bR^n$ be a Jordan measurable set. Consider the real non-negative numbers $V^{-} (A)$ and $V^{+} (A)$ defined as follows: $ V^{-}(A) := \sup \Bigl\{ \sum_{i=1}^r \mbox{vol}_o (P_i) \ \begin{array}{cc} \lvert & P_1, \ldots , P_r \in \cP_n, \mbox{ with } \cup_{i=1}^r P_i \subseteq A \\ \lvert & \mbox{and with $P_i \cap P_j = \emptyset$ for $i \neq j$} \end{array} \Bigr\} .$

$V^{+}(A) := \inf \Bigl\{ \sum_{j=1}^s \mbox{vol}_o (Q_j) \mid Q_1, \ldots , Q_s \in \cP_n, \mbox{ with } \cup_{j=1}^s Q_j \supseteq A \Bigr\} . $

Then $V^{-}(A) = V^{+}(A)$. The volume of $A$ is defined as $ \mbox{vol} (A) := V^{-}(A) = V^{+}(A). $

Definition. Let $\cA$ be a field of subsets of $\bR^n$. A function $\mu : \cA \to [0, \infty )$ is said to be an additive set-function when it has the property that: $ \left( \begin{array}{c} A_1, \ldots , A_k \in \cA \\ \mbox{with $A_i \cap A_j = \emptyset$ for $i \neq j$} \end{array} \right) \ \Longrightarrow \ \Bigl( \, \mu (A_1 \cup \cdots \cup A_k) = \sum_{i=1}^k \mu (A_i) \, \Bigr) . $

Theorem. Consider the function $\mbox{vol} : \cM_n \to [0, \infty )$, as defined above, where $\cM_n$ is the collection of all Jordan measurable sets on $\bR^n$. Prove that $\mbox{vol} : \cM_n \to [0, \infty )$ is an additive set-function.

I want to prove the theorem stated above. Please let me know if my approach is OK. I'm concerned that it is not rigorous enough.

Let $A_1, \dots, A_k\in \cM_n$, with $A_i \cap A_j = \emptyset$ if $i\ne j$. Then $V^-(A_1\cup \dots \cup A_k)=\sup\left\{ \sum\limits_{i=1}^r\mbox{vol}_0(P_i): P_1, \dots, P_r\in \cP_n, \text{ with } \bigcup\limits_{i=1}^r P_i \subset A_1\cup \dots\cup A_k \text{ and } P_i\cap P_j = \emptyset \text{ for } i\ne j \right\}$.

Since $P_i$ are mutually disjoint and $A_m$ are mutually disjoint, if $x\in P_i$ for some $1\le i \le r$, $x\in A_m$ for some $1\le m\le k$. Reorder the indices of the sets $P_i$ in such a way as to have $\bigcup\limits_{i=1}^{r_1} P_{i} \subset A_1$, $\bigcup\limits_{i=r_1+1}^{r_2} P_{i} \subset A_2$, $\dots$, $\bigcup\limits_{i=r_s}^{r} P_{i} \subset A_k$ and consider the following:

$\sup\left\{ \sum\limits_{i=1}^r\mbox{vol}_0(P_i): P_1, \dots, P_r\in \cP_n, \text{ with } \bigcup\limits_{i=1}^r P_i \subset A_1\cup \dots\cup A_k \text{ and } P_i\cap P_j = \emptyset \text{ for } i\ne j \right\}\\ = \sup\left\{ \sum\limits_{i=1}^{r_1}\mbox{vol}_0(P_i)+\sum\limits_{i=r_1+1}^{r_2}\mbox{vol}_0(P_i) + \dots + \sum\limits_{i=r_s}^r\mbox{vol}_0(P_i): P_1, \dots, P_r\in \cP_n, \text{ with } \bigcup\limits_{i=1}^{r_1}\subset A_1,\dots, \bigcup\limits_{i=r_s}^{r}\subset A_k, \text{ and } P_i\cap P_j = \emptyset \text{ for } i\ne j \right\}\\ =\sup\left\{ \sum\limits_{i=1}^{r_1}\mbox{vol}_0(P_i): P_1, \dots, P_{r_1}\in \cP_n, \text{ with } \bigcup\limits_{i=1}^{r_1} P_i \subset A_1 \text{ and } P_i\cap P_j = \emptyset \text{ for } i\ne j \right\}\\ +\dots+\sup\left\{ \sum\limits_{i=r_s}^{r}\mbox{vol}_0(P_i): P_{r_s}, \dots, P_{r}\in \cP_n, \text{ with } \bigcup\limits_{i=r_s}^{r} P_i \subset A_k \text{ and } P_i\cap P_j = \emptyset \text{ for } i\ne j \right\}\\ =\mbox{vol}(A_1)+\dots+\mbox{vol}(A_k)$.

Q.E.D.

$\endgroup$
  • $\begingroup$ What is a "field of subsets" ? $\endgroup$ – Duchamp Gérard H. E. Jul 17 '17 at 6:22
  • $\begingroup$ Do you mean "closed under (finite) intersections, unions and complements" ? $\endgroup$ – Duchamp Gérard H. E. Jul 17 '17 at 6:25
  • $\begingroup$ @DuchampGérardH.E. I've just added this definition at the top of the post. $\endgroup$ – sequence Jul 17 '17 at 6:38
  • $\begingroup$ Do you suppose that $\mathbb{R}^n$ is in $\mathcal{A}$ (that was inside my previous question). $\endgroup$ – Duchamp Gérard H. E. Jul 17 '17 at 6:41
  • $\begingroup$ No, that is not supposed. $\endgroup$ – sequence Jul 17 '17 at 7:01
1
$\begingroup$

$\newcommand{\vol}{\operatorname{vol}}\newcommand{\P}{\mathcal{P}}\newcommand{\M}{\mathcal{M}}$ Instead of using those long formulae with unproven equalities of suprema, use $\varepsilon$-arguments instead:

Take $A_1,\ldots A_k \in \M$ pairwise disjoint and let $A = \cup_i A_i$.
Suppose $\varepsilon>0$. Then for each fixed $i$ ($1 \le i \le k$) we have some finite pairwise disjoint family $\mathcal{P}_i = \{P^i_1, P^i_2, \ldots, P^i_{n_i}\} \subseteq \P$, such that

$$ \sum_{j=1}^{n_i} \vol_0(P^i_j) > \vol^-(A_i) - \frac{\varepsilon}{k}$$

which can be done by the definition of $\vol^-(A_i)$ as a supremum of such sums.

Then note that $\P' = \cup_{i=1}^k \P_i$ is also a pairwise disjoint subfamily of $\P$ (as the $A_i$ are) and

$$\sum_{P \in \P'} \vol_0(P) = \sum_{i=1}^k \sum_{j=1}^{n_i} vol_0(P^i_j) > \sum_{i=1}^k vol^-(A_i) - \varepsilon$$.

As this holds for all $\varepsilon >0$, and $\P'$ is a valid family in the definition of $\vol^-(A)$ we have

$$\vol^-(A) \ge \sum_{i=1}^k vol^{-}(A_i)$$ and the other inequality should be clear as well:

Suppose $\P'= \{P_1, \ldots P_N\}$ is a pairwise disjoint subfamily of $\P$ such that $\cup \P' \subseteq A$. Then we can split ( gap here!) each $P \in \P'$ into a part inside $A_i$ for each $i$ and use the $\vol^-(A_i)$ upperbounds to have a bound for $\vol^-(A)$... I think the gap can be filled with ideas from Tao's post on these matters, working with properties of what he calls elementary sets, i.e. unions of half-open boxes, which form the smallest algebra of sets containing $\P$, and first extend $\vol_0$ to these...

Just some ideas to get you started, you see it's all quite subtle.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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