Anisotropies in the stochastic background of gravitational waves (jointly supervised by Prof. Lesgourgues)

Context: Gravitational waves from the merger of compact binaries are now routinely detected by the LIGO-VIRGO network of detectors. Neutron stars and supernova explosions are other sources that the next generation of detectors will be sensitive to. Sources which are too faint to be detected individually contribute to the stochastic astrophysical gravitational wave background (AGWB). The anisotropies of this background contain valuable information on the astrophysics of those sources and on cosmology. However, in order to leverage this potential, the different contributions to the AGWB need to be disentangled and differentiated from instrumental noise.

Projects: The project will largely consist of two parts: The simulation of the AGWB and the decomposition of the simulated AGWB into its components. For the first part, you will write a code that allows generating samples of the AGWB for a given set of astrophysical and cosmological parameters. For the second part, you will make use of component separation methods available from CMB physics and adapt them to the AGWB. Eventually, the component separation method will be tested on the simulations. If successful, your method will be considered in future analyses of the AGWB.

Requirements: A solid background in general relativity and good coding skills in C or C++ and python

The diffuse emission of the Galaxy at high energies

Context: High energy cosmic rays interact with the gas and radiation in the Galaxy creating secondary gamma-rays and neutrinos. These contain valuable information on the acceleration and transport of high-energy particles elsewhere in the Galaxy where we cannot observed cosmic rays directly. However, existing models are in conflict with data from the Fermi-LAT satellite experiment, thus pointing to some flaws in the models. Resolving these issues is urgently needed in light of data from a number of experiments to be presented in the coming years.

Project: You will carefully study existing models of the diffuse high-energy emission in gamma-rays and neutrinos. Following hints from the inconsistencies with Fermi-LAT data, we will investigate which assumptions need to be relaxed in order to improve the agreement with data. A particular focus will be on the reevaluation of the microphysics of interactions of cosmic rays with turbulent magnetic fields. Our models will be confronted with data and we will make predictions that will be tested in the future by experiments like IceCube, HAWC and LHAASO.

Requirements: A keen interest in high-energy astrophysics and familiarity with coding in python.

Non linear transport of the cosmic-ray spectra from stochastic sources

Context: Supernova remnants are the most likely candidate for sources of Galactic cosmic rays. Interestingly, it has been suggested that cosmic rays escaping from these sources excite magnetic turbulence which, in turn, regulates their transport close to the acceleration sites. This highly non-linear problem has been addressed for individual sources by numerically solving the coupled differential equations describing the evolution escaping particles and magnetic turbulence. But the collective effect of this self-confinement on the Galactic cosmic ray spectra, especially in the MeV to GeV energy range, has not been discussed in the literature.

Project: We will first revisit the problem of non-linear transport to better describe the time evolution of cosmic ray diffusion coefficient in the vicinity of supernova remnants. Once the diffusion coefficient has been estimated, we shall incorporate that into a model for Galactic cosmic ray transport to obtain the cosmic ray spectra together with their theoretical uncertainties due to the discrete nature of sources.

Requirements: A keen interest in cosmic-ray physics and familiarity with coding in C++.

Finding the sources of cosmic rays with machine learning (jointly supervised by Prof. Krämer)

Context: Finding the sources of cosmic rays is one of the most pressing questions in theoretical astroparticle physics. The spectra of high energy electrons and positrons are extremely sensitive to the ages and distances of young and nearby sources and can thus be used for identifying sources. However, to fully exploit existing data, the full liklihood function including correlations between different energy bins is required for which no closed form expression exists.

Project: The idea is to employ a specific neural network architecture to emulate the full likelihood function. The neural network will be trained on data that you will generated by performing Monte Carlo simulations of the flux of high energy electrons and positrons from a statistical ensemble of sources. We will quantify the performance of the new approach and compare to a traditional approach for the likelihood function. Eventually, this will applied to existing data from cosmic ray experiments in space.

Requirements: Some practical familiarity with machine learning (e.g. from Prof. Krämer's of Prof. Erdmann's lecture courses)