Diffusive propagation of cosmic rays from supernova remnants in the Galaxy. II: anisotropy

We investigate the effects of stochasticity in the spatial and temporal distribution of supernova remnants on the anisotropy of cosmic rays observed at Earth. The calculations are carried out for different choices of the diffusion coefficient D(E) for propagation in the Galaxy. The propagation and spallation of nuclei are taken into account. At high energies we assume that $D(E)\sim(E/Z)^{\delta}$, with $\delta=1/3$ and $\delta=0.6$ being the reference scenarios. The large scale distribution of supernova remnants in the Galaxy is modeled following the distribution of pulsars with and without accounting for the spiral structure of the Galaxy. Our calculations allow us to determine the contribution to anisotropy resulting from both the large scale distribution of SNRs in the Galaxy and the random distribution of the nearest remnants. The naive expectation that the anisotropy amplitude scales as D(E) is shown to be an oversimplification which does not reflect in the predicted anisotropy for any realistic distribution of the sources. The fluctuations in the anisotropy pattern are dominated by nearby sources, so that predicting or explaining the observed anisotropy amplitude and phase becomes close to impossible. We find however that the very weak energy dependence of the anisotropy amplitude below $10^{5}$ GeV and the rise at higher energies, can best be explained if the diffusion coefficient is $D(E)\sim E^{1/3}$. Faster diffusion, for instance with $\delta=0.6$, leads in general to an exceedingly large anisotropy amplitude. The spiral structure introduces interesting trends in the energy dependence of the anisotropy pattern, which qualitatively reflect the trend seen in the data. For large values of the halo size we find that the anisotropy becomes dominated by the large scale regular structure of the source distribution, leading indeed to a monotonic increase of $\delta_A$ with energy.