Masterarbeiten in der Astroteilchenphysik und Kosmologie (2020-2021)

 

Forecasts for the cosmological survey of the EUCLID satellite

(mainly supervised by Prof. Julien Lesgourgues and MSc. Nils Schoeneberg)

The EUCLID satellite will be launched in 2022 to map the large scale of the universe (that is, the three-dimensional distribution of galaxies and of the gravitational potential). The goal is to measure several cosmological parameters (density and property of components such as dark matter, neutrinos, dark energy, extensions of Einstein\u2019s gravity, etc.) The student working on this project will interact with the international EUCLID team and will work (as a theorist) on the preparation of numerical tools (simulation codes, likelihoods describing the probability of a model given the data, algorithm for fitting models to the data using Monte Carlo Markov Chains) that may be used ultimately in the final analysis.

Improving the efficiency and robustness of a Monte Carlo Markov Chain algorithm for cosmological parameter estimation

(mainly supervised by Dr. Jesus Torrado)

Monte Carlo Markov Chain (MCMC) methods are the usual tool of choice for finding the parameters of a cosmological model that better fits the data; but they require some fine-tunning to be efficient and less prone to failure. We will explore different avenues for automatically learning this kind of tuning in real time (i.e. while the algorithm is running), and, if successful, release our tools to the community in the context of a widely-used package for statistical analyses in cosmology.

Efficient implementation of massive neutrinos in N-body simulations

(mainly supervised by Dr. Christian Fidler)

Massive neutrinos leave a distinct signature in the matter power spectrum, allowing us to provide an upper bound for the neutrino masses using the upcoming EUCLID satellite mission. In order to optimally exploit the data, efficient numerical simulations of structure formation with massive neutrinos are crucial. The student will work on the implementation and improvement of a novel method that we have developed recently to include the impact of massive neutrinos in cosmological N-body simulations.

Explaining tensions in cosmological data with non-trivial Dark Matter models

(mainly supervised by Prof. Julien Lesgourgues and MSc. Nils Schoeneberg)

The student will play with existing cosmology codes and modify them in order to infer constraints from CMB and Large Scale Structure data on extended models of particle physics that feature non-minimal Dark Matter (feebly interacting, decaying, etc.) or a non-minimal Dark Sector (with several relic particles). These models have a potential to solve a few tensions in current cosmological data - such as the famous Hubble tension.

Boosting cosmological simulations and parameter estimation with Neural Networks

(mainly supervised by Dr. Christian Fidler, Prof. Julien Lesgourgues, MSc. Nils Schoeneberg, Dr. Jesus Torrado)

Two previous Masterarbeit students from our group have developed successfully some deep neural networks that are conditioned to reproduce the result of complex cosmological simulation codes in much less time. A third student would be welcome for working on the concrete implementation of these networks and for preparing their public release (together with a set of tools designed for re-training them). These tools will then used by a large community of cosmologists worldwide.

Fast and accurate theoretical predictions of reionization and the 21cm temperature

(mainly supervised by Prof. Julien Lesgourgues and MSc. Nils Schoeneberg)

Cosmology codes are very precise when they simulate some very well-known stages in the evolution of the universe, such as the formation of the Cosmic Microwave Background anisotropies. They are currently less precise when they simulate less-known stages called the \u201cDark Ages\u201d - like the epoch of cosmological reionization, when the first stars from and emit ionising photons in the inter-galactic medium. Since future experiments will become sensitive to the Dark Ages, we will try to improve current simulation codes in order to better take into account physical processes during this era (in collaboration with experts in Cambridge).