Weak convergence of the past and future of Brownian motion given the present

Abstract

In this paper, we show that for t > 0, the joint distribution of the past {W _t−s : 0 ≤ s ≤ t} and the future {W _{t + s} :s ≥ 0} of a d-dimensional standard Brownian motion (W _s ), conditioned on {W _t ∈ U}, where U is a bounded open set in ℝ^d, converges weakly in C[0,∞)×C[0,∞) as t→∞. The limiting distribution is that of a pair of coupled processes Y + B ¹,Y + B ² where Y,B ¹,B ² are independent, Y is uniformly distributed on U and B ¹,B ² are standard d-dimensional Brownian motions. Let σ _t ,d _t be respectively, the last entrance time before time t into the set U and the first exit time after t from U. When the boundary of U is regular, we use the continuous mapping theorem to show that the limiting distribution as t → ∞ of the four dimensional vector with components $(W_{\sigma _{t}},t-\sigma _{t},W_{d_{t}},d_{t}-t)$, conditioned on {W _t ∈U}, is the same as that of the four dimensional vector whose components are the place and time of first exit from U of the processes Y + B ¹ and Y + B ² respectively.