Charmed mesons may be produced when a primary cosmic ray or the leading hadron in an air shower collide with an atmospheric nucleon. At energies \ge 10^8 GeV their decay length becomes larger than 10 km, which implies that they tend to interact in the air instead of decaying. We study the collisions of long-lived charmed hadrons in the atmosphere. We show that (\Lambda_c,D)-proton diffractive processes and partonic collisions of any q^2 where the charm quark is an spectator have lower inelasticity than (p,\pi)-proton collisions. In particular, we find that a D meson deposits in each interaction just around 55% of the energy deposited by a pion. On the other hand, collisions involving the valence c quark (its annihilation with a sea cbar quark in the target or c-quark exchange in the t channel) may deposit most of D meson energy, but their frequency is low (below 0.1% of inelastic interactions). As a consequence, very energetic charmed hadrons may keep a significant fraction of their initial energy after several hadronic interactions, reaching much deeper in the atmosphere than pions or protons of similar energy.

We analyze the possibility that the cosmic ray knee appears at an energy threshold where the proton-dark matter cross section becomes large due to new TeV physics. It has been shown that such interactions could break the proton and produce a diffuse gamma ray flux consistent with MILAGRO observations. We argue that this hypothesis implies knees that scale with the atomic mass for the different nuclei, as KASKADE data seem to indicate. We find that to explain the change in the spectral index in the flux from E^{-2.7} to E^{-3.1} the cross section must grow like E^{0.4+\beta} above the knee, where \beta=0.3-0.6 parametrizes the energy dependence of the age (\tau \propto E^{-\beta}) of the cosmic rays reaching the Earth. The hypothesis also requires mbarn cross sections (that could be modelled with TeV gravity) and large densities of dark matter (that could be clumped around the sources of cosmic rays). We argue that neutrinos would also exhibit a threshold at E=(m_\chi/m_p)E_{knee}\approx 10^8 GeV where their interaction with a nucleon becomes strong. Therefore, the observation at ICECUBE or ANITA of standard neutrino events above this threshold would disprove the scenario.