No label defined (Q4395535)

Let \(B(t)\), \(t\geq 0\), be a real Brownian motion starting from 0 and let \(L^x_t\), \(t\geq 0\), \(x\in R\), denote its jointly continuous local time. The authors consider a continuous random process \[ H(t)=\int^t_0{ds\over B(s)}=\int^\infty_0(L^x_t-L^{-x}_t){dx\over x} \] defined as Cauchy's principal value. It is proved that \[ \limsup_{t\to\infty}H(t)/(2t\log\log t)^{1/2}=2\text{ almost surely}. \] The proof is based on a careful analysis of the sample path properties of Brownian local times and exact distribution of Brownian principal values. In the appendix the modification of Burkholder-Davis-Gundy inequality for \(H(t)\), \(t\geq 0\), is established. For the entire collection see [Zbl 0889.00015].