I'm trying to prove that Brownian motion absorbed at the origin is a Markov process with respect to the original filtration $\{\mathcal{F}_{t}\}$. To be more specific, let $(B_{t},\mathcal{F}_{t})_{t \geq 0}$ be a 1-d brownian motion such that $B_{0}=x,\ a.s. \mathbb{P}\ $ for some $x>0.\ $ Let $T_{0}=\inf\{ t \geq 0, \ B_{t}=0\}$ be the hitting time of $0$, and define the Brownian motion absorbed at origin to be the process $\{B_{t \land T_{0}}, \mathcal{F}_{t} \}$. I want to show that such a process is Markov w.r.t. $\mathcal{F_{t}}, \ $ i.e. $$\mathbb{P}[B_{t \land T_{0}} \in \Gamma | \mathcal{F}_{s}]=\mathbb{P}[B_{t \land T_{0}} \in \Gamma |B_{s \land T_{0}}]$$ for every pair of $0<s<t$. And further more it has transition density given by: $$\mathbb{P}[B_{(t+s) \land T_{0}} \in dy |B_{s \land T_{0}}=z]=\mathbb{P}^{z}[B_{t} \in dy, \ T_{0} >t], \ \forall y,z >0,$$ where $\mathbb{P}^{z}$ is the probability measure under which $B$ starts from $z$ almost surely.

I tried to use the strong Markov property for Brownian motion to derive these equalities, but I'm having a hard time dealing with $T_{0}$. For example, if $0 \notin \Gamma$, then$\{B_{t \land T_{0}} \in \Gamma\}=\{B_{t} \in \Gamma, T_{0} >t\}$. But how to apply strong Markov property to compute $\mathbb{P}[B_{t} \in \Gamma, T_{0} >t\ | \mathcal{F}_{s} ]\ $?

Any answer or comment is greatly appreciated, thanks!

up vote 4 down vote accepted

For every $r$ let $\theta_r$ denote the time-shift by $r$. Then, $$ [B_t\in\Gamma,T_0\gt t]=[T_0\gt s,B_s+(B_t-B_s)\in\Gamma,T_0\circ\theta_s\gt t-s], $$ where $[T_0\gt s]$ and $B_s$ are $\mathcal F_s$-measurable and $B_t-B_s$ is independent of $\mathcal F_s$ and distributed like $B_{t-s}$. Hence, $$ \mathbb P^x(B_t\in\Gamma,T_0\gt t\mid\mathcal F_s)=\mathbf 1_{T_0\gt s}\cdot u_{t-s}(B_s), $$ where, for every $r$ and $y$, $$ u_r(y)=\mathbb P^y(B_r\in\Gamma, T_0\gt r). $$ Likewise, $$ [B_{t\wedge T_0}\in\Gamma]=[B_t\in\Gamma,T_0\gt t]\cup[0\in\Gamma,T_0\lt t], $$ hence $$ \mathbb P^x(B_{t\wedge T_0}\in\Gamma\mid\mathcal F_s)=\mathbf 1_{T_0\gt s}\cdot u_{t-s}(B_s)+\mathbf 1_{0\in\Gamma}\cdot(1-\mathbf 1_{T_0\gt s}\cdot v_{t-s}(B_s)). $$ where, for every $r$ and $y$, $$ v_r(y)=\mathbb P^y(T_0\gt r). $$

  • Thank you very much! The shift operator is the thing that I missed. – user7762 Aug 20 '12 at 0:18

Your Answer


By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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