Bachelor Thesis With Prof. Krämer in 2023
Dark matter annihilation into antideuterons with machine learning
Antideuterons are a promising probe to indirectly detect the annihilation of dark matter particles, as the astrophysical background of antideuterons is very small. The upcoming GAPS experiment will have sufficient sensitivity to detect antideuterons and probe parts of the parameter space of interesting dark matter models. Hence, an important contribution to this effort is to provide simulations of antideuteron spectra from dark matter annihilation to probe the parameter space. However, the production of antideuterons from dark matter is a complex process that is difficult to model and is computationally expensive. The goal of this bachelor project is to understand the production mechanism of antideuterons and to use machine learning techniques to improve the predictions of antideuteron production from dark matter.
Main supervisor: Lena Rathmann
LISA Gravitational Wave Data Analysis with Machine Learning
Since the first measurement of a gravitational wave (GW) from a black hole binary merger in 2015, experimental and theoretical efforts to study GWs have increased significantly. One of the upcoming space-based GW detectors, the Laser Interferometer Space Antenna (LISA) will be able to cover wave frequencies in the millihertz regime.
This will enable us to probe a vast range of different signals, such as the merger of binary systems over a large mass range, or even the stochastic gravitational wave background, and allows for an independent measurement of the expansion of the universe. The LISA collaboration has released a data challenge (LDC) to introduce new researchers to GW data analysis. The content of the bachelor thesis will involve studying the theoretical landscape of GW physics, getting familiar with the mock data released with the LDC and applying a state-of-the-art machine learning approach to determine the relation between physical parameters and detected signals.
Main supervisor: Kathrin Nippel
Searching for new physics at the LHC with machine learning
The Large Hadron Collider (LHC) has confirmed the Standard Model of particle physics by countless measurements. In addition, there are many searches for models with specific additional particles or interactions. So far, all these searches have failed to uncover such new physics. Modern machine learning techniques do not only improve such traditional searches but also offer the possibility to find new physics in a more model independent way not relying on specific signal simulation. Some anomaly detection methods are completely unsupervised and try to order LHC events from ordinary and expected to anomalous and exciting. Others are weakly supervised and optimize the search for anomalies by comparing data in a signal region to a well-defined control region where only standard model physics is expected. Our group has successfully worked on both unsupervised as well as weakly supervised anomaly detection methods. We will continue to do so and the bachelor thesis will be embedded in our current research. You will learn about modern machine learning concepts and apply them to LHC physics. As a prerequisite, an affinity to computing and some python skills are helpful. Prior knowledge of machine learning is not required but of course very welcome.
Main supervisor: Alexander Mück