Published: 2014
Total Pages: 166
Get eBook
Since the discovery of cosmic rays over a century ago, evidence of their origins has remained elusive. Deflected by galactic magnetic fields, the only direct evidence of their origin and propagation remain encoded in their energy distribution and chemical composition. Current models of galactic cosmic rays predict variations of the energy distribution of individual elements in an energy region around 3 x 1015 eV known as the knee. This work presents a method to measure the energy distribution of individual elemental groups in the knee region and its application to a year of data from the IceCube detector. The method uses cosmic rays detected by both IceTop, the surface-array component, and the deep-ice component of IceCube during the 2009-2010 operation of the IC-59 detector. IceTop is used to measure the energy and the relative likelihood of the mass composition using the signal from the cosmic-ray induced extensive air shower reaching the surface. IceCube, 1.5 km below the surface, measures the energy of the high-energy bundle of muons created in the very first interactions after the cosmic ray enters the atmosphere. These event distributions are fit by a constrained model derived from detailed simulations of cosmic rays representing five chemical elements. The results of this analysis are evaluated in terms of the theoretical uncertainties in cosmic-ray interactions and seasonal variations in the atmosphere. The improvements in high-energy cosmic ray hadronic-interaction models informed by this analysis, combined with increased data from subsequent operation of the IceCube detector, could provide crucial limits on the origin of cosmic rays and their propagation through the galaxy. In the course of developing this method, a number of analysis and statistical techniques were developed to deal with the difficulties inherent in this type of measurement. These include a composition-sensitive air shower reconstruction technique, a method to model simulated event distributions with limited statistics, and a method to optimize and estimate the error on a regularized fit.