A joint integral test for the locations of extrema for Brownian motion (Q867086)

Let \(\mu^+(t)=\inf\{0\leq s\leq t:\,W(s)=\sup_{0\leq u\leq t}W(u)\}\) be the location of the maximum of the standard Brownian motion \(\{W(t)\}_{t\geq0}\) over \([0,t]\), respectively, \(\mu^-(t)\) the location of the minimum with \(\sup\) replaced by \(\inf\). The author proves an integral criterion for a 0-1-law of \((\mu^+(t),\mu^-(t))\) in the sense of Lévy, according to which for two positive non-increasing functions \(a(t)\), \(b(t)\) such that \(ta(t)\) and \(tb(t)\) are increasing, the probability of \(\mu^+(t)<ta(t)\) and \(\mu^-(t)<tb(t)\) simultaneously infinitely often is either zero or one, if an integral depending on \(a\) and \(b\) converges, respectively, diverges. This generalizes a well known result of Chung and Erdős taking into account \(\mu^+(t)\) only. As an immediate consequence, a law of the iterated logarithm result for \(\max\{\mu^+(t),\mu^-(t)\}\) is established, similar to a corresponding result of \textit{E. Csáki, A. Földes} and \textit{P. Révész} [Probab. Theory Relat. Fields 76, 477--497 (1987; Zbl 0611.60077)] for the location maximum of a reflected Brownian motion. The proof relies on sharp estimates for the joint distribution function of \((\mu^+(1),\mu^-(1))\) and uses techniques following \textit{E. Csáki} [Z. Wahrscheinlichkeitstheor. Verw. Geb. 43, 205--221 (1978; Zbl 0372.60113)].