Sequence definition in Durrett Ex.1.1.2

97 Views Asked by At

I'm working through the exercises of Durrett, and I'm confused about the second part of this particular exercise.

We first show $\sigma(\mathcal{S}_d) \subseteq \mathcal{R}^d $ by showing that $\mathcal{S}_d \subseteq \mathcal{R}^d$. Let $A = (a_1, b_1] \times \cdots \times (a_d, b_d] \in \mathcal{S}_d$.

I've found a detailed solution here, (the different numbering of the exercise is due to different edition of the book), and I understand the first part of the exercise, but confused about the second part.

Let $A$ be an open set in $\mathcal{R}^d$. Let $A_{n, i} \in S_n$ be a sequence of all sets in $S_n$ that are contained in $A$. Let $B = \cup_{n,i} A_{n, i}$ then by definition of $A_{n, i}$, we have $\cup_{n, i} A_{n, i} \subseteq A$. We claim that all points in $A$ are contained in some $A_{n, i}$ to establish the reverse containment. Suppose $a \in A$. Since $A$ is an open set, there is a cube centered at $a$ that is contained in $A$. It is clear that there is an $n$ for which there exists a set $A_{n, i}$ that is contained in $A$. Therefore, $\cup_{n, i} A_{n, i} = A$. Since the left hand side is a countable union of sets in $\mathcal{S}_d$, $A \in \sigma(\mathcal{S}_d)$.

The introduction of such a sequence is very confusing for me, I understand what is n, but as we are talking about $ \mathbb{R}^d $, I'm confused about the need for the iterator i, and what it is exactly coding. I understand that n encodes for the granularity, but $i$ doesn't add any information on what the coefficients will be, at least for me.

I tried to step through that, but I found the entire explanation very convoluted, so I would appreciate an alternative explanation, and I would be also be interested in the background intuition for this problem.

1

There are 1 best solutions below

0
On

I finally understood your problem... (but be rigorous in your questions, because you neither defined $\mathcal S_d$ nor $\mathcal R^d$, and we are not supposed to guess them). Instead of doing the proof in $\mathbb R^n$, do it in $\mathbb R$ (the principle is exactly the same and easier to see).

We want to prove that $\sigma (\mathcal S)=\mathcal R$ where $$\mathcal S=\{(a,b]\mid -\infty <a<b<\infty \},$$ and $\mathcal R$ is the Borel $\sigma -$algebra.

As you understood (so I won't prove it), $\mathcal S\subset \mathcal R$, and thus $\sigma (\mathcal S)\subset \mathcal R$.


To show the converse, we want to show that open sets are in $\sigma (\mathcal S)$. Indeed if $\mathcal O=\{\text{open sets of }\mathbb R\}$ and $\mathcal O\subset \sigma (\mathcal S)$, then $$\mathcal R=\sigma (\mathcal O)\subset \sigma (\mathcal S),$$ what will prove the claim.


So, let $A$ be an open set of $\mathbb R$. What you want is to write $A$ as a countable union of elements of $\mathcal S$. So, let $$\mathcal U=\{(a,b]\mid (a,b]\subset A\text{ and } a,b\in \mathbb Q \}.$$ Notice that $\mathcal U\subset \mathcal S$ and that $\mathcal U$ is countable. Moreover, $$\bigcup_{U\in \mathcal U}U\subset A.$$

For the converse inclusion, let $a\in A$. Since $A$ is open, there exist $\delta >0$ s.t. $(a-\delta ,a+\delta )\subset A$. By density of $\mathbb Q$ in $\mathbb R$, we can find $c,d\in\mathbb Q$ s.t. $$a-\delta <c<a<d<a+\delta.$$ Therefore, $$a\in \underbrace{(c,d]}_{\in \mathcal U}\subset A,$$ i.e. $\displaystyle a\in \bigcup_{U\in \mathcal U}U$. Therefore, $\displaystyle A=\bigcup_{U\in \mathcal U}U$ and thus $A\in \sigma (\mathcal S) $ as wished.