Why $\{A_n \quad a.a.\}:=\bigcup_{n=1}^\infty \bigcap_{k=n}^\infty A_k$ is not defined as $\lim_{n\rightarrow\infty} \bigcap_{k=n}^\infty A_k$

probability

Given a probability tripple $(\Omega, F, P)$ and infinite many events $A_1, A_2, \dots, \in F$, a new event "almost always" in $F$ is normally defined as:

$$\{A_n \quad a.a.\}:=\bigcup_{n=1}^\infty \bigcap_{k=n}^\infty A_k$$

I'm wondering, why it is not defined as
$$\{A_n \quad a.a.\}:=\lim_{n\rightarrow\infty} \bigcap_{k=n}^\infty A_k$$
? The two definitions look equivalent to me and the latter is easier to understand.

Best Answer

The first definition is easiest to understand in my opinion. It describes all those sets for which only finitely many $A_n$ are not in $F$.

Regarding your suggestion the set-theoretic limit is defined as follows. Let the limit infinum and limit supremum be defined as \begin{equation*} \liminf_{n\rightarrow \infty}A_n = \bigcup_{n \geq 1} \bigcap_{j \geq n} A_j \qquad \limsup_{n\rightarrow \infty}A_n = \bigcap_{n \geq 1} \bigcup_{j \geq n} A_j. \end{equation*} If the two coincide the set theoretic limit exists and is equal to the resulting set.

You can see that the first definition for "almost always in $F$" is equal to $\liminf_{n\rightarrow \infty}A_n$. As pointed out in the comments $B_n = \bigcap_{j \geq n} A_j$ is an increasing sequence and thus $$\lim_{n\rightarrow \infty} B_n = \liminf_{n\rightarrow \infty}B_n = \bigcup_{n \geq 1} \bigcap_{j \geq n} B_j = \bigcup_{n \geq 1} B_n = \bigcup_{n \geq 1} \bigcap_{j \geq n} A_j= \liminf_{n\rightarrow \infty}A_n.$$

Related Question