11
$\begingroup$

I'm studying Brownian motion from Durrett. I'm trying to understand what Blumenthal's 0-1 law really says about what Durrett calls the germ field, $\mathcal{F}_0^+$.

Let $\mathcal{F}_t^+ = \cap_{s > t} \mathcal{F}_s^0$ and $\mathcal{F}_t^0 = \sigma(B_s,s\le t)$ where $B_t$ is Brownian motion. Also let $P_x$ denote the usual measure on $\mathcal{C}[0,\infty)$ making $B_t(\omega) = \omega(t)$ a Brownian motion starting at $x$.

Blumenthal's 0-1 law: If $A \in \mathcal{F}_0^+$ then for all $x \in \mathbb{R}$, $P_x(A) \in \{0,1\}$.

I have no trouble understanding this proof, but I'm not sure I understand why this is surprising/significant.

Obviously, $\mathcal{F}_0^0$ is trivial. Is there any intuitive way to explain why we might not expect $\mathcal{F}_0^+$ to be trivial as well? Even though $\mathcal{F}_0^+$ seems to involve a tiny bit of information into the future past $t=0$, since Brownian motion is continuous it would seem that we ought to expect no difference between the two fields.

$\endgroup$

1 Answer 1

24
$\begingroup$

To help hone your intuition, it might help to think about the following simple example, which shows that the Blumenthal law is using much more than just the continuity of Brownian motion.

Let $Z$ be a random variable with $P(Z=1) = P(Z=-1) = 1/2$, i.e. a coin flip. Set $X_t = tZ$, so that the process $X_t$ just moves with constant speed $\pm 1$ depending on the value of $Z$. Note that $X_t$ is continuous in $t$. Let $\mathcal{F}_t^0 = \sigma(X_s : 0 \le s \le t)$ be the natural filtration.

Now $X_0 = 0$ so $\mathcal{F}_0^0$ is trivial. On the other hand, for every $t > 0$ we have $\mathcal{F}_t^0 = \sigma(Z)$. So $\mathcal{F}_0^+$ is not trivial; it really contains more information than $\mathcal{F}_0^0$. For instance, it contains the nontrivial events $\{Z=1\}$ and $\{Z = -1\}$; it contains enough information to see how the coin flip came out.

Intuitively, by observing the process at time 0 you don't learn anything. But if you get to observe it at some time after time 0, no matter how short, you can see whether it is above or below 0, and thus deduce the value of $Z$.

$\endgroup$
3
  • $\begingroup$ Dang. This is such a simple yet effective device to explain this idea. Thanks! $\endgroup$
    – gogurt
    Commented Jul 26, 2015 at 18:27
  • $\begingroup$ Could you relate this back to the statement of the law? Since $X_t$ is a BM, the law is saying that any event in $\mathcal{F}_t^+$ has probability zero or one. But here, the two events in this $\sigma$-algebra $\{ Z = 1 \}, \{Z = -1\}$ do NOT have probability zero or one. Doesn't this contradict the law? What am I confusing here? Sorry for the zombie comment. $\endgroup$ Commented Mar 29, 2023 at 23:49
  • 2
    $\begingroup$ @TanishqKumar: $X_t$ in my example is not a Brownian motion, and this example shows that it does not satisfy the zero-one law. (It might be better to call it a "zero-one property" as it holds for some processes and not for others; it's not a "law" in the sense that all processes must satisfy it.) The original question suggested that the zero-one law was a direct consequence of continuity alone. By giving an example of a continuous process that doesn't satisfy it, it shows that in order to prove the Blumenthal 0-1 law for Brownian motion, you will have to use its other properties too. $\endgroup$ Commented Mar 29, 2023 at 23:53

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .