A study of the ultrahigh-energy cosmic ray mass composition with the MACRO and EAS-TOP experiments

Two components of cosmic-ray-induced air showers are measured simultaneously at the Gran Sasso Laboratory: the electromagnetic shower at the ground surface by the EAS-TOP extensive air shower array, and the deep-underground muons by the MACRO experiment. The two independent data sets collected during 96.3 days of simultaneous running are combined, and underground muon multiplicity distributions are obtained for anticoincident events (no surface trigger) and high-energy, coincident events. These categories correspond to ranges in primary energy from about 2 x 10^3 GeV to a few times 10^5 GeV, and from about 1.5 x 10^5 GeV to about 10^7 GeV, respectively. The experimental shower size and muon multiplicity distributions, as well as the distribution of mean muon multiplicity as a function of shower size (N_µ - log(N_e) relation), are compared to the ones obtained with detailed Monte Carlo calculations (with a generator based on recent hadronic accelerator data) using various trial compositions as input. This is done in an effort to discriminate between these models of primary cosmic-ray mass composition at and above the "knee" in the all-particle spectrum, where contradictory experimental evidence exists and where a knowledge of the composition would bear upon possible mechanisms for cosmic-ray acceleration and propagation. Detailed studies of simulated anticoincident event rates (which arise from a region of primary energy where the composition has been measured directly by satellite and balloon experiments) uncover problems with the generator used, with between 25 and 40% too few high-energy muons created. This, combined with the dependence of absolute event rates on the assumed differential primary energy spectra, hampers the interpretation in terms of composition of underground muon or surface air shower data taken separately. However, the (N_µ - log(N_e)relation is independent of the spectra or overall Monte Carlo normalization problems. The simulated (N_µ - log(N_e) relation for coincident events is found to be inconsistent with the possibility that the cosmic ray flux becomes proton-dominated at and above the knee.