A cloudy Vlasov solution

We integrate the Vlasov–Poisson equations giving the evolution of a dynamical system in phase space using a continuous set of local basis functions. In practice, the method decomposes the density in phase space into small smooth units having compact support. We call these small units ‘clouds’ and choose them to be Gaussians of elliptical support. Fortunately, the evolution of these clouds in the local potential has an analytical solution that can be used to evolve the whole system during a significant fraction of dynamical time. In the process, the clouds, initially round, change shape and become elongated. At some point, the system needs to be remapped on round clouds once again. This remapping can be performed optimally using a small number of Lucy iterations. The remapped solution can be evolved again with the cloud method, and the process can be iterated a large number of times without showing significant diffusion. Our numerical experiments show that it is possible to follow the two‐dimensional phase‐space distribution during a large number of dynamical times with excellent accuracy. The main limitation to this accuracy is the finite size of the clouds, which results in coarse‐graining the structures smaller than the clouds and induces small aliasing effects at these scales. However, it is shown in this paper that this method is consistent with an adaptive refinement algorithm which allows one to track the evolution of the finer structure in phase space. It is also shown that the generalization of the cloud method to four‐dimensional and six‐dimensional phase space is quite natural.