Hausdorff dimension of the contours of symmetric additive Lévy processes (Q2464671): Difference between revisions

An additive Lévy process is a multiparameter process \(\mathfrak X(t)\), \(t\in\mathbb R_+^N\), which is the sum \[ \mathfrak X(t) = X(t_1)+\cdots+X(t_N),\quad t=(t_1,\dots,t_N) \] of \(N\) independent Lévy processes \(X(t_j), t_j\geq 0\). These processes have been studied by Khoshnevisan, Xiao and various co-workers since 2000, see also the monograph [\textit{D.\ Khoshnevisan}, Multiparameter Processes. An introduction to random fields. Springer Monographs in Mathematics. (New York), NY: Springer. (2002; Zbl 1005.60005)], and they have proved to be both interesting examples of a new kind of multiparameter processes and instrumental to deduce many formulae for classical Lévy processes. Their potential theory has been systematically developed in the above-mentioned monograph by Khoshnevisan. In the present paper the authors study exact formulae for the Hausdorff dimension of level sets and traces of level sets on non-random Borel sets \(F\subset (0,\infty)^N\). The main results of this paper are as follows: under the assumption that the characteristic functions of the Lévy processes \(X_j(t_j)\), \(e^{-t_j\psi_j(\xi)}\), are real-valued and integrable in \(\xi\), they prove almost surely on \(\{\mathfrak X^{-1}(\{0\}) \neq \emptyset\}\) [(i)] a closed expression for the Hausdorff dimension of the set \(\mathfrak X^{-1}(\{0\})\) in form of an integral test involving the \(L^1\)-norms of \(e^{-t_j\psi_j}\) (i.e.\ the density function of the law of \(\mathfrak X(t)\) at the origin); [(ii)] a closed expression for the Hausdorff dimension of the trace \(\mathfrak X^{-1}(\{0\})\cap B\) for any Borel set \(B\subset (0,\infty)^N\) in terms of an energy-type integral for measures carried on \(B\). \noindent These results generalize earlier results of two of the present authors [\textit{D. Khosnevisan, Y. Xiao}, Ann. Probab 30, 62--100 (2002; Zbl 1019.60049)] where bounds of the Hausdorff dimension of the level sets were shown. The new results include alternative proofs of known formulae for the dimension of level sets of Lévy processes due to \textit{J. Horowitz} [Isr. J. Math. 6, 176--182 (1968; Zbl 0186.50001)] and \textit{J. Hawkes} [J. Lond. Math. Soc. 8, 517--525 (1974; Zbl 0325.60074)]. Note, however, that the techniques are necessarily completely different as there are no simple covering results available in the multidimensional case. The classical covering arguments give, in general, only upper and lower bounds and it is only rarely possible to get the same bounds, i.e.\ exact results. This difficulty is elegantly overcome by the authors by introducing additional random sets (defined by auxiliary additive Lévy processes). The proof of (ii) is quite technical and goes on for more than half of the paper. On the way a few new additions to the potential theory of multiparameter (additive Lévy) processes are made. That (ii) implies (i) is then a routine application of methods developed in [\textit{D. Khoshnevisan, Y. Xiao}, Lévy processes: capacity and Hausdorff dimensions, Ann. Probab. 33, 841--878 (2005; Zbl 1072.60040)].