Brownian motion on volume preserving diffeomorphisms group and existence of global solutions of 2D stochastic Euler equation (Q859651): Difference between revisions

In this paper, a geometric Brownian motion performs a continuous time infinitesimal perturbation of the state of the system; this perturbation conserves the energy; exact expression for the energy transfer towards high modes is obtained ensuring existence for all time of the solution. The authors follow the Arnold point of view on considering Euler equation as the equation of geodesic on the `infinite-dimensional Riemannian manifold \(G\)' constituted by the volume preserving diffeomorphisms of the 2-dimensional torus. Using the resources of infinite-dimensional stochastic analysis the authors construct a re\-gu\-la\-ri\-zed Brownian motion \(g_x\) on \(G\) such that the geodesic flow can be obtained from the process \(g_x\) by a Girsanov subordination. The key point for constructing \(g_x\) is an estimation of the energy transfer from small modes to large modes and vice-versa; these key estimates are obtained through a machinery recently developed by H.~Airault and P.~Malliavin in the context of the Virasoro algebra of mathematical physics. This noise adjunction has a conceptual content essentially equivalent to an elliptic re\-gu\-la\-ri\-za\-ti\-on at the level of path space over \(G\). As in the elliptic re\-gu\-la\-ri\-za\-ti\-on the authors suppose that the noise has an arbitrarily small amplitude. The authors do not add any stabilizing viscosity: the energy is strictly preserved during the evolution; the tool is an a priori majorization of the remainder term in the spectral expansion series of the solution. The paper contains full proof of the existence of all introduced objects. The originality of this paper relies on the following two main points: (1) The construction of the noise which does not use any energy input. The noise can be described as an error propagating in time on the state of the system. It has some similitude with noise on the initial value of the deterministic system. But in the case this error on the initial data is distributed on the whole trajectory. (2) The variance induced by the noise can be exactly computed through the energy transfer matrix. This means that it would be possible to obtain positive results as done in this paper where the convergence of the approximation procedure is shown as well as negative results showing the explosion almost sure of solutions.