How to deal with discontinuous path in brownian motion?

91 Views Asked by At

I'm having great difficulty trying to understand brownian motion. They require almost every path to be continuous, yet in proofs they assume all path are continuous. In quite a few questions I'm trying to prove, I need all path to be continuous too. Example: let $K\subseteq \mathbb{R}$ by closed, prove $T(\omega) = \inf\{t:B_t(\omega)\in K\}$ is a stopping time. My method is first let $U_\varepsilon = \bigcup_{x\in K} \mathcal{B}(x,\varepsilon)$, \begin{equation} \{T>t\}= \bigcap_{x\in[0,t]} B^{-1}_x(K^c) = \bigcup_n \bigcap_{q\in [0,t]} B^{-1}_q(U_{1/n}^c) \end{equation} where $q$ is rational. I need every $\omega$ to be continuous for the second equality to hold.

Discontinuous paths also defeat the purpose of stopping time, since they can leap across the enclosed boundary used to define $T$. This is not possible for discrete random walk $S_n$, no walk can leap across the boundary. What to do about the discontinuous path?

1

There are 1 best solutions below

0
On BEST ANSWER

Even though the trajectories of a Brownian motion $B$ are assumed to be a.s. continuous, i.e. there exists a measurable set $A$ of probability one such that on $A$ every trajectory is continuous, you can always define another Brownian motion $\tilde B$ to be equal to $B$ on $A$ and identically $0$ otherwise. This way, $\tilde B$ has everywhere continuous trajectories and $B=\tilde B$ a.s.

Now, $\tilde T$ (hitting time of $K$ for $\tilde B$) is a stopping time, and that is what you have proven. Since $T=\tilde T$ a.s., if the filtration is complete, then it readily follows that $T$ is also a stopping time.

Completeness is usually a key condition for hitting times to be stopping times (see the Debut theorem, e.g. here).