Speaker
Description
Sources of ultra-high-energy cosmic rays (UHECRs) remain uncertain and continue to challenge the astroparticle physics community. We approach the problem of constraining their properties as that of Bayesian signal reconstruction within the framework of Information Field Theory. We reconstruct the injection energy spectrum and the source density evolution with distance without assuming a predefined parametric form. The joint spectral and spatial distribution is modeled as a Gaussian random field, and we assume an a priori correlation structure to ensure the physical interpretability of the reconstruction. This enables inference of continuous signals from highly indirect measurements, while also quantifying the associated uncertainty. The model accounts for the relevant effects during UHECR propagation (adiabatic losses, photomeson and pair production on photon fields in the Universe) and reproduces the energy spectrum and composition (proxied by distribution of shower depth maxima $X_\mathrm{max}$) at Earth as measured by the Pierre Auger Observatory. We benchmark the method using both synthetic and real data in well-established scenarios: the dip model for the ankle feature and flux suppression, and the combined fit of UHECR data above the ankle. The reconstruction performs well on synthetic data and shows qualitative agreement with results reported in the literature. In particular, the combined fit analysis by the Pierre Auger Collaboration favors very hard injection spectra, with a power-law index of $\gamma \approx −1.5$; however, this value strongly depends on the assumed cutoff shape. Our reconstruction constrains the very hard part of the spectrum to only half a decade in energy, indicating that it need not represent a global property of sources. Finally, we outline future directions for extending the framework, including additional observables and messengers, as well as higher-dimensional signals such as 3D source distribution.