Mouvement brownien et espaces de besov

Author:

Abstract
Let be a linear brownian motion. Traditionally, this brownian motion is considered as a process which belongs a.s. to the space of continuous functions defined on [01], vanishing at the origin and endowed with the topology of uniform convergence. Let now be the besov space of the functions defined on [01], . They are spaces of sequences and an important fact is therefore that these Besov spaces are isomorphic to suitable Banach spaces of numerical sequences. This property is particularly well adapted to the study of brownian motion considered as a process which belongs a.s.to the spaces In this article we illustrate this in the following situations:Z. Ciesielski fias shown recently [C3] that brownian motion belongs a.s. to the space There we characterize in this first paragraph the set indexes αp, qsuch that brownian motion belongs a.s. to The classical theory of large deviations gives the behaviour, as ,of where Γ is a Borel set of the uniform topology. In the second paragraph we show that if Γ are chosen to be balls of the Besov topology, one obtains an equivalent of which is better than an equivalent of By the inequalities of Burkholder-Davis-Gundy (B.D.G.) the moments of stopping times T can be compared with the moments of We show in paragraph 3 that it is possible to obtain BDG-type inequalities by replacing in the expression above the sup-norm by the Besov-norm. Moreover we show that similar upper bounds can be obtained for the local time, generalizing the results obtained for Hölder-norms by Barlow M. T., Yor M. and Bass R. F. (see [BY] and [B]). In the fourth paragraph we study the belonging of the stochastic integral to the Besov spaces (IV.A). We show that, under suitable conditions on the integrand, the stochastic integral is defined in a deterministic manner and that it is a continuous functional (IV.B)