We discuss that hadron-induced atmospheric air showers from ultra-high energy cosmic rays are sensitive to QCD interactions at very small momentum fractions x where nonlinear effects should become important. The leading partons from the projectile acquire large random transverse momenta as they pass through the strong field of the target nucleus, which breaks up their coherence. This leads to a steeper x_F-distribution of leading hadrons as compared to low energy collisions, which in turn reduces the position of the shower maximum Xmax. We argue that high-energy hadronic interaction models should account for this effect, caused by the approach to the black-body limit, which may shift fits of the composition of the cosmic ray spectrum near the GZK cutoff towards lighter elements. We further show that present data on Xmax(E) exclude that the rapid ~ 1/x^0.3 growth of the saturation boundary (which is compatible with RHIC and HERA data) persists up to GZK cutoff energies. Measurements of pA collisions at LHC could further test the small-x regime and advance our understanding of high density QCD significantly.

We discuss particle production in the high-energy, small-x limit of QCD where the gluon density of hadrons is expected to become nonperturbatively large. Strong modifications of the phase-space distribution of produced particles as compared to leading-twist models are predicted which reflect in the properties of cosmic ray induced air showers in the atmosphere. Assuming hadronic primaries, our results suggest a light composition near GZK cutoff energies. We also show that cosmic ray data discriminate among various QCD evolution scenarios for the rate of increase of the gluon density at small x, such as fixed-coupling and running-coupling BFKL evolution. There are clear indications for a slower growth of the gluon density as compared to RHIC and HERA, due e.g. to running-coupling effects.

In ultrarelativistic heavy-ion collisions the local three-volume measure is expanding in the longitudinal and transverse directions. This is similar to the Hubble-expansion of the universe, except that the former is not locally isotropic. As an example the expansion rate is calculated assuming that the energy-momentum tensor in the central region is that of an ideal fluid, undergoing Bjorken flow in longitudinal direction, and with initial conditions as expected for BNL-RHIC energy. While the longitudinal expansion of three-volume is independent of the energy density of the fluid, in case of 3+1 dimensional expansion the form of the hydrodynamical solution (rarefaction wave or deflagration shock) affects the three-volume expansion rate on the hadronization hypersurface. As a consequence the average expansion rate on that surface depends on the transverse size of the system. This may reflect in an impact-parameter dependence of the formation probability of light nuclei and of the freeze-out temperature of the strong interactions in the system.