# A question about sigma-algebras and generators

Suppose you are given a set $\Omega$ and a collection $\mathcal{G}$ of subsets of $\Omega$. Assume further that $A \subset \Omega$. Now let $\sigma_{\Omega} (\mathcal{G})$ denote the smallest sigma-algebra on $\Omega$ containing $\mathcal{G}$, and let $\sigma_{A}(\mathcal{G} \ \cap A)$ denote the smallest sigma-algebra on A containing the collection $\mathcal{G} \ \cap A$.

Is it true that $\sigma_{A}(\mathcal{G} \ \cap A) = \sigma_{\Omega} (\mathcal{G}) \cap A$ ?

The inclusion " $\subset$ " is clear, since if $\mathcal{H}$ is a sigma-algebra on $\Omega$ containing $\mathcal{G}$, then $\mathcal{H} \cap A$ is a sigma-algebra on $A$ containing $\sigma_{A}(\mathcal{G} \ \cap A)$. But what about the other inclusion?

Thanks for your help! Regards, Si

-
I don't understand the notation $\mathcal{G}\cap A$, since $\mathcal{G}$ is the set of subsets of $\Omega$ while $A$ is the subset of $\Omega$. – Jack Nov 17 '11 at 15:15
@Jack: I think that's why Si wrote "the family $\mathcal{G} \ \cap A$", to make clear that every set in $\mathcal{G}$ is being intersected with $A$, much like when we write $gH$ for a coset of a subgroup. – joriki Nov 17 '11 at 15:21

Let $$\mathcal{B} = \left\{B \subset X | B \cap A \in \sigma_A(\mathcal{G} \cap A)\right\}.$$ Notice that $\mathcal{G} \subset \mathcal{B}$. It is easy to see that $\mathcal{B}$ is a $\sigma$-algebra. For example, if $B_j \in \mathcal{B}$, then $$\left(\bigcup B_j\right) \cap A = \bigcup (B_j \cap A) \in \sigma_A(\mathcal{G} \cap A),$$ because $B_j \cap A \in \sigma_A(\mathcal{G} \cap A)$.
Therefore, since $\mathcal{B}$ is a $\sigma$-algebra containing $\mathcal{G}$, we can conclude that $\sigma(\mathcal{G}) \subset \mathcal{B}$. So, for every $B \in \sigma(\mathcal{G})$, $B \cap A \in \sigma_A(\mathcal{G} \cap A)$. In other words, $$\sigma(\mathcal{G}) \cap A \subset \sigma_A(\mathcal{G} \cap A).$$
+1, nice! To put it in words: It doesn't matter what happens outside $A$; the elements of $\mathcal G$ happen to have certain parts outside $A$, and these may interact in complicated ways to determine $\sigma_\Omega(\mathcal{G})$, but when we cut everything down to $A$ none of that matters; we might as well consider all possible extensions outside $A$ of the elements of $\sigma_A(\mathcal G\cap A)$ (as you did in $\mathcal B$); then the result is clearly a sigma algebra, and yet we don't get more than $\sigma_A(\mathcal G\cap A)$ when we cut it down to $A$. – joriki Nov 17 '11 at 15:57