Speaker
Description
Properties of ultra-high-energy cosmic-ray sources can be inferred in the approximation of identical sources spread over cosmological volumes. By assuming a parametric form for the injection spectrum, e.g., a power law with a cutoff, a combined description of the observed energy spectrum and mass composition is possible. However, the inferred injection parameters are highly degenerate with each other, their theoretical interpretation is limited, and in some cases, a consistent description of the data cannot be reached. We develop a nonparametric reconstruction of the injection spectrum using Information Field Theory. We model the injection spectrum shape with a Gaussian random field, allowing it to flexibly fit the data. By comparing the propagated spectrum and mass composition to the observations by the Pierre Auger Observatory, we sample the posterior distribution and reconstruct the injection spectrum shape along with its uncertainty. The reconstruction reveals additional structure not captured by the usual parametric form: a soft low-energy tail and a very hard spectrum segment confined to roughly half a decade in rigidity. This region is responsible for the often-quoted negative values of spectral index. By varying prior assumptions, hadronic interaction models, and experimental systematics treatment, we verify the stability of the reconstruction and our qualitative conclusions. We find that for the EPOS.LHC model, the tension between the spectrum and composition cannot be fully resolved by our flexible injection model, confirming its inconsistency with the observations. Finally, we outline future directions for extending the analysis to higher-dimensional signals such as 3D source distribution.