I am currently trying to understand why $P(\tau_b < \infty)$ for $b \in \mathbb{R}$, where $\tau_b := \inf\{t \geq 0: B_t = b\}$. There was a little step in the argument below which I did not understand.

Let $(B_t)_{t \geq 0}$ be a one-dimensional Brownian motion. Let us write $$\sup_{t \geq 0}B_t \geq \sup_{n \in \mathbb{N}}B_n = \sup_{n \in \mathbb{N}} (\xi_1 +...+\xi_n)$$ where $\xi_j=B_j-B_{j-1}$ are i.i.d. standard normal random variables. If we further write $$S:=\sup_{n\in \mathbb{N}}(\xi_1 +...+\xi_n) = \xi_1 + \sup_{n \in \mathbb{N},n \geq 2}(\xi_2 +...+\xi_n) =: \xi_1 + S'$$ then $S$ and $S'$ have the same distribution and $\xi_1$ and $S'$ are independent. I need to somehow conclude, since $\xi_1$ is standard normal distributed, that then $S=\infty$ a.s. How do I see this?

Thanks a lot in advance!


By 0-1 law for exchangeable random variables $P\{|S|<\infty\}$ is $0$ or $1$. If it is $1$ then $Ee^{itS}Ee^{it\xi}=Ee^{itS}$ for all $t$ and $Ee^{itS}=Ee^{itS}$ because $S$ and $S'$ have the same distribution. This gives $Ee^{it\xi}=1$ at least for $|t|$ sufficiently small, which is a contradiction because$Ee^{it\xi}=e^{-t^{2}/2}$.Hence $P\{|S|=\infty\}=1$. By definition $S>-\infty$ almost surely so we get $S=\infty$ almost surely.

  • $\begingroup$ Doesn't it hold for arbitrary $t$ instead of sufficiently small $|t|$? $\endgroup$ – vaoy Aug 30 '18 at 9:26
  • $\begingroup$ I cannot cancel $Ee^{itS}$ unless I know it is not zero. For $|t|$ small it is close to $1$ and hence non-zero. $\endgroup$ – Kavi Rama Murthy Aug 30 '18 at 9:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.