Brownian motion dyadic construction convergence

brownian motionprobabilityprobability theory

Say $B$ is a Brownian motion and $B^n$ is the function that is equal to $B$ at all numbers of the form $2^{-n}\mathbb{N}$ and linear on the intervals $[k2^{-n},(k+1)2^{-n}]$. I am trying to show that $\lim_{n\to\infty} P(\sup_{[0,1]}|B-B^n|\leq\epsilon )=1$. I was thinking that $|B-B^n|$ can be bounded by $|B_{\frac{k}{2^n}}-B_{\frac{k+1}{2^n}}|$ which is $\mathcal{N}(0,\frac{1}{2^n})$ distributed. Is this correct? Any hint would be appreciated

Best Answer

First, the process $W_t^{n,k}:=(B_t-B_t^n)$, $t\in[k2^{-n},(k+1)2^{-n}]$ is a Brownian bridge. Therefore, using for example this result, \begin{align*} &\mathsf{P}\!\left(\sup_{t\in [k2^{-n}, (k+1)2^{-n}]}|W_t^{n,k}|>\epsilon\right) \\ &\qquad\le 2\mathsf{P}\!\left(\sup_{t\in [k2^{-n}, (k+1)2^{-n}]}W_t^{n,k}>\epsilon\right)=2e^{-2^{n+1}\epsilon^2}, \end{align*} and so, \begin{align*} \mathsf{P}\!\left(\sup_{t\in[0,1]}|B_t-B_t^n|> \epsilon\right)&= \mathsf{P}\!\left(\max_{0\le k\le 2^n-1}\sup_{t\in [k2^{-n}, (k+1)2^{-n}]}|W_t^{n,k}|> \epsilon\right) \\ &\le \sum_{k=0}^{2^n-1}\mathsf{P}\!\left(\sup_{t\in [k2^{-n}, (k+1)2^{-n}]}|W_t^{n,k}|> \epsilon\right) \\ &\le 2^{n+1}e^{-2^{n+1}\epsilon^2}\to 0 \quad\text{as}\quad n\to\infty. \end{align*}

Related Question