On the distribution of ranked heights of excursions of a Brownian bridge. (Q1872189)

Let \(B\) be a \(\beta\)-selfsimilar strong Markov process, \(B^{\text{br}}(u)\), \(u\in [0,1]\), its ``standard bridge'', \(B^{\text{ex}}\) its standard excursion, \(F\) a \(\gamma\)-homogeneous functional of \(e_{t}\), \(t\in [0,V_{e}]\), i.e. \(F=V_{e}^{\gamma }F(V_{e}^{-\beta }e_{\cdot V_{e}})\). If \(E(F(B^{\text{ex}})^{\alpha /\gamma }) <\infty\), where \(\alpha\) is the index of the inverse \(\tau_{a}\) of the local time of \(B\), as a stable subordinator, then the authors prove that the \(e\) with \(F(e)>0\), where \(e\) runs over the excursions of \(B^{\text{br}}\), can be arranged in a decreasing sequence \(F_{j}^{\text{br}}\) and that the joint distribution of \(\mu (\lambda^{-\gamma }\Gamma_{\alpha }^{\gamma }F_{j}^{\text{br}})\) is the same as that of \(\sum_{i\leq j}\varepsilon_{i}/\varepsilon_{0}\): \(\Gamma_{a}\) is independent of \(B^{\text{br}}\) with a \(\gamma(\alpha )\) distribution, \[ \mu (x)= \int_{_{0}}^{^{\infty }}\alpha\lambda^{-\alpha }\Gamma (1-\alpha )^{-1}t^{- \alpha -1}e^{-\lambda t}P(F(B^{\text{ex}})>xt^{-\gamma })dt \] and the \(\varepsilon\)'s are independent standard exponential. The sequence is inhomogeneous Markovian and expresses the distributions of its terms and the transition probabilities. The authors deduce that, if \(N\) is standard Gaussian, independent of the Brownian bridge \(D^{\text{br}}\), then \(| N| \), multiplied by the \(j\)th value \(M_{j}\) of the maxima of the excursions of \(| D^{\text{br}}| \) when arranged in decreasing order, is Markovian, with distribution function \(1-(1-\tanh x)^{j}\) and transition distribution functions (of \(x\)) \((\tanh x/\tanh y)^{j}\). Furthermore, they show that the distribution function of \(M_{j}\) is \(1-2^{j}\sum_{n\geq 0}(_{n}^{-j})\exp(-2(n+j)^{2}x^{2})\) and that of the \(j\)th positive excursion of \(D^{\text{br}}\) is \(1-\exp(-2j^{2}x^{2})\). An application to Bessel processes and Bessel bridges follows. If \({\mathcal F}_{a}\) is the set of values of \(F(e)\), where \(e\) runs over the excursions of \(B\) before \(\tau_{a}\), then the authors compute the intensity measure of the Poisson point process \({\mathcal F}_{a}\) and show that if \(E(F(B^{\text{ex}})^{\alpha /\gamma })=\infty\), then \({\mathcal F}_{a}\) is dense in \((0,\infty)\), while in the contrary case it is a descending sequence \(F_{j}\), identical in law with \(((\sum_{i\leq j}\varepsilon_{i})^{-\gamma /\alpha })_{j\geq 1}\), multiplied by a constant depending on \(F\), \(\alpha\), \(a\), \(\gamma\). The paper finishes with a calculation aiming to determine the joint law of \(\tau_{1}\) and \(F_{j}\) for \(a=1\). The paper contains also other examples and remarks and an alternative formula for \(\mu\), which leads to a generalization of a result by \textit{Ph. Biane} and \textit{M. Yor} [ Bull. Sci. Math., II. Sér. 111, 23--101 (1987; Zbl 0619.60072)].