The Chemical composition of cosmic rays is a powerful tool for identifying the mechanisms of their acceleration. In addition to protons and He ions, recent AMS-02 observations also provide the high-precision rigidity spectra of C and O. They widen the window into a complicated selection process whereby collisionless shocks, such as supernova remnant (SNR) blast waves, extract different chemical elements from an interstellar mix. We investigate the particle injection into the diffusive shock acceleration using one- and two-dimensional self-consistent hybrid simulations. Our 1D simulations prove that an SNR shock can modify the chemical composition of accelerated cosmic rays by preferentially extracting them from a homogeneous background plasma without additional, largely untestable assumptions.
Our results show that selection rate of different ion species increases with mass-to-charge ratio (A/Z), saturates, and peaks as a function of the shock Mach number, confirming the earlier theoretical predictions. The 2D simulations also evidence a deviation from the linear injection efficiency vs A/Z trend, pointing towards a saturation for higher A/Z. The integrated SNR rigidity spectrum for the proton-to-helium ratio, obtained by a convolution of the time-dependent injection rates of protons and helium ions with a decreasing shock strength over the active lives of SNRs compares well with the AMS-02 and PAMELA data.