Bachelor- und Masterarbeitsthemen

Bei Interesse bitte einfach bei uns in der Abteilung vorbeischauen, oder bei Brian Moser melden. Bei den Themen zur Higgs-Boson Paarproduktion arbeiten wir eng mit der AG Jakobs zusammen.

Learning goals:

• Detailed understanding of particle physics and in particular the Higgs boson 
• Fundamentals of a physics analysis: event selection (also utilizing neural networks and deep learning methods), background estimation, statistical data analysis
• Learning software development (C++, Python, machine learning software like Tensorflow and PyTorch, bash, version control and continuous integration with git)
• How to perform collaborative research within a global scientific environment
• Research communication

Mögliche Arbeitsthemen:

Optimization and development of searches for Higgs boson pair production 

The observed Higgs boson holds a key position within the Standard Model. The electroweak symmetry breaking, from which the Higgs boson is born, creates dynamically the masses of the W and Z bosons, while keeping the photon massless. In addition, fermion masses can be parametrized via their interaction strength to the Higgs boson. The production of two Higgs bosons would allow for a direct measurement of the shape of the Higgs potential. Despite tremendous progress on both experiment and theory in the last decade, many properties of the Higgs boson remain unknown. The central role of the Higgs boson in the Standard Model, paired with the blurry picture that we currently have of it, make it a prime candidate for experimental exploration. One of the least explored properties is the potential associated with its underlying quantum field, which can be studied via multi-Higgs production. The most prominent channel to look for Higgs boson pair production in the bbττ decay channel, which our groups in Freiburg are investigating. There are several opportunities to further develop these analyses with novel methods and explore new avenues towards the planned High-Luminosity Upgrade of the LHC (HL-LHC). 

Possible topics:

• Development of triggers for HH→bbττ that will work in the high pile-up environment of the HL-LHC.
• Development of a combined b+τ identification algorithm.

Machine learning – “Deep Learning”

Machine learning means that an algorithm can learn certain properties in (simulated) data, using complex mathematical models. “Deep Neural Network” (DNN) algorithms use many layers in their architecture in order to extract high-level properties from data. Generally speaking, “deep learning” methods are often superior to traditional methods. Bachelor or Master candidates can train and optimize DNNs (or other types of networks) for a certain task of an ATLAS data analysis. They will use modern software for training and evaluating the performance of the developed DNN. Machine learning can be used to separate signal from background processes, to classify different signal categories, to learn kinematic properties, and many more. These modern technologies have the potential to pave new ways for many physics analyses, e.g., precision Higgs boson measurement, searches for Higgs boson pair production, and more. 

Possible topics:

• Development of a parameterized neural network (PNN) or usage of Simulation-based Inference methods to enhance the Higgs boson self-coupling constraint from HH→bbττ.
• Development of an algorithm that learns to emulate τ-decay kinematics (τ-embedding).
• Study of advanced machine learning architectures such as diffusion models and their usage for the Run 3 HH→bbττ analysis.

The Higgs boson at future colliders

Ultimate precision on the Higgs boson properties is promised by an e⁺e^– collider that is planned to follow the HL-LHC. The Future Circular Collider (FCC-ee) is one of the proposals of such a “Higgs factory”, that would be located at CERN in a 91 km tunnel and achieve e⁺e^– collisions at centre-of-mass energies of up to 365 GeV and very high luminosities. Feasibility studies of such a machine are currently being performed. In parallel to these feasibility studies, it is important to understand the potential physics performance of such a machine and the proposed detectors. For this, Monte Carlo simulations of collision events are passed through reference detectors and subsequently analyzed. Bachelor or Master candidates can perform such performance studies and provide valuable inputs towards deciding on the next big project after the LHC.

Possible topics:

• Sensitivity estimates for Higgs boson decay channels that are either still uncovered or only poorly understood.
• Comparison of jet clustering algorithms → How should we cluster hadronic signatures to get the best performance?
• Constraining the Higgs boson self-coupling from higher order electroweak corrections to single Higgs boson production.