Cosmic ray composition from direct measurements

The knowledge of the cosmic ray (CR) composition is a basis of models for their origin which is still an un-clarified question, and for the physics of the interactions at the highest energies(exceeding those attainable by artificial accelerators). This paper is mainly focussed to the results of experiments measuring directly the CR composition at the highest energies. Above the value of 10(12) eV of the total energy there are results from: the Chicago group experiment on Spacelab 2 [1], two Russian experiments (SOKOL [2] on satellite and MUBEE [3] - balloon-borne), the Japanese balloon [4] and the well known ballon collaboration JACEE [5]. The maximum total energies measured reach a few x 10(14) eV and the charge range reaches the iron group, starting from the lightest nuclei (apart from the Japanese experiment, where Z(min) approximate to 14, and the Chicago one with Z(min) = 6). The comparison of the energy spectra (for separate charge groups) from different experiments shows still large discrepancies (by a factor of 3, particularly for E > 100 GeV/n) [6]. But the relative charge contributions seem to be measured with a good accuracy up to some 10(13) eV for the total energy of the nucleus. The results for 10(12) - 10(13) eV are: protons: approximate to 40%, alpha particles: approximate to 25%, CNO: 10-12%, Z = 10 - 16: 10% the iron group (Z > 16): 10-15% (see also the compilations: ref. [7] and the rapporteur talk [6]). Above 10(13) eV the data seem to indicate a decrease of the proton contribution, down to about 20% at 10(1)4 eV. The relative proportions of the nuclei charge groups are consistent with being energy independent. There is no clear interpretation of these results. Applying the leaky-box propagation model leads to a source composition modulated by the first ionization potential [1] (as observed long time ago for the low energy CR) in the source medium with T approximate to 104 K. Calculations of the source spectra for different nuclei have been made by Berezhko et al. [8] on the basis of the model of CR acceleration by supernova remnants. The injection of differ ent elements has been carefully treated. However the predicted contributions do not fit well to the observed spectrum. In particular, the model predicts too many alpha particles (approximate to 50%) in the full energy region of 10(12) - 10(15) eV. But an interesting feature of the predictions is that the cut-off energy in the elemental spectra appears to be less than proportional to the charge Z: for the iron group the cut-off is expected at an energy only approximate to 4 times larger than that for protons. This may help to understand the rather sharp knee in the observed total particle spectrum. Nevertheless, as long as a satisfactory model description for E > 10(15) eV does not exist, all interpretations below this energy should be considered with reservation.