MSc thesis topics
FAQ - Informationen zur Masterarbeit in der Arbeitsgruppe von Prof. Ferber
| Topic: | Search for Displaced Dark Higgs Decays from Dark Photon Higgs Strahlung at Belle II |
| Summary: | Belle II provides an opportunity to probe dark sector scenarios in which a dark photon couples to the Standard Model through kinetic mixing: Most searches concentrate on final states that are either fully visible or fully invisible. In this project you will study a different process in which a dark Higgs boson is emitted together with the dark photon. This mechanism is tied to the mass generation of the dark photon. If the dark Higgs boson is the lightest state in the dark sector, it is expected to have a long lifetime and to decay into Standard Model particles through mixing with the Higgs field. You will learn how to generate simulated samples with Monte Carlo tools and how to perform statistical analyses needed to evaluate the sensitivity of Belle II to this process. More details can be found in . |
| You learn: | Python programming, data analysis, statistics |
| Prerequisites: | introductory MSc course in particle physics (e.g. lecture and all exercises TP1), basic knowledge in Python programming (e.g. lecture and all exercises in "Rechnernutzung", also see our BSc/MSc FAQ), ideally MSc course on modern methods of data analysis |
| Supervisor: | Prof. Dr. Torben Ferber and Prof. Dr. Felix Kahlhoefer |
| Contact: | Dr. Giacomo De Pietro (Team Dark Physics) |
| Last update: | 28.11.2025 |
| Topic: | Differentiable Vertex Reconstruction using Machine Learning at Belle II |
| Summary: | This project explores a differentiable vertex fitting approach that formulates the reconstruction of particle decay vertices as an optimization problem, where the solution’s gradients are accessible via implicit differentiation and can be backpropagated through neural network architectures. The goal of this master’s thesis is to integrate this differentiable vertex fitting technique into the GNN-based track finding and fitting pipepline of Belle II. For an example of GNN-based track finding, see the CAT project at Belle II, for differentiable vertex fitting see this example. |
| You learn: | Differentiable programming for secondary vertex fitting, integration of optimization-based methods into ML architectures, interfacing with Belle II software, and evaluation of vertex reconstruction performance |
| Prerequisites: | Good proficiency in English is required. Solid knowledge of Python programming, for example from the lecture and all exercises in Rechnernutzung, is required (also see our BSc/MSc FAQ). The exam in ModEx 1+2 must be completed. Prior experience in software development projects is an advantage. Basic familiarity with machine learning frameworks such as TensorFlow or PyTorch is beneficial. Detailed understanding of charged particle motion in magnetic fields is required. |
| Supervisor: | Prof. Dr. Torben Ferber |
| Contact: | Lea Reuter (Team Tracking) |
| Last update: | 28.11.2025 |
| Topic: | Physics-Informed GNNs for Helical Track Reconstruction in the Belle II Drift Chamber |
| Summary: | Charged particle tracks in the Belle II detector follow helical trajectories in the uniform solenoidal magnetic field. While standard GNN-based tracking methods learn these patterns implicitly from simulated data, explicitly incorporating this known physical constraint can improve performance and training efficiency. This project investigates physics-informed GNN architectures that embed the helix model directly into the learning process. A particular challenge is the Belle II central drift chamber geometry, which uses tilted stereo wires to obtain z-coordinate information. However, this wire arrangement makes track-finding challenging since the exact hit position along a trajectory is only known after track parameter estimation. You will explore strategies such as helix-aware message-passing layers, and loss functions that penalise deviations from physically consistent trajectories. The aim is to produce a GNN that not only identifies track candidates with high efficiency but also outputs parameters consistent with the underlying helix model, even in the presence of background hits and geometric distortions from the stereo wires. Performance will be benchmarked against existing GNNs on simulated Belle II data, focusing on efficiency, fake rate suppression, and resolution in transverse and longitudinal parameters. For an example of GNN-based track finding, see the CAT project at Belle II. |
| You learn: | Physics-informed machine learning, detector geometry effects in track reconstruction, advanced GNN architecture design, integration of physical models into deep learning, high-precision parameter estimation in particle tracking |
| Prerequisites: | Good proficiency in English is required. Solid knowledge of Python programming, for example from the lecture and all exercises in Rechnernutzung, is required (also see our BSc/MSc FAQ). The exam in ModEx 1+2 must be completed. Prior experience in software development projects is an advantage. Basic familiarity with machine learning frameworks such as TensorFlow or PyTorch is beneficial. Detailed understanding of charged particle motion in magnetic fields is required. |
| Supervisor: | Prof. Dr. Torben Ferber |
| Contact: | Lea Reuter (Team Tracking) |
| Last update: | 28.11.2025 |
| Topic: | Domain Adaptation for Robust GNN-Based Track Reconstruction at Belle II |
| Summary: | Graph Neural Networks (GNNs) for charged particle track reconstruction at Belle II can show performance degradation when applied to data with input feature distributions or background conditions different from those seen during training. Such differences can arise from detector conditions, background levels, or other simulation-to-data discrepancies. This project investigates domain adaptation techniques to make GNN-based reconstruction more robust and less sensitive to these variations. The approach will augment the training dataset with tracks reconstructed using the standard Belle II baseline reconstruction algorithm, blending them with simulated truth-level training data. You will develop strategies for incorporating this additional domain-specific data into the GNN training process, aiming to improve generalisation and stability across different running conditions. The work will include designing suitable preprocessing and feature harmonisation methods, implementing domain adaptation loss functions or training schedules, and evaluating performance under varied detector and background scenarios. Results will be validated both on simulated and reconstructed data to quantify gains in reconstruction efficiency and resilience. For an example of GNN-based track finding, see the CAT project at Belle II. |
| You learn: | Domain adaptation in machine learning, robust GNN training techniques, detector simulation and reconstruction workflows, advanced data augmentation strategies, integration of ML methods into high-energy physics reconstruction software |
| Prerequisites: | Good proficiency in English is required. Solid knowledge of Python programming, for example from the lecture and all exercises in Rechnernutzung, is required (also see our BSc/MSc FAQ). The exam in ModEx 1+2 must be completed. Prior experience in software development projects is an advantage. Basic familiarity with machine learning frameworks such as TensorFlow or PyTorch is beneficial. Experience with track reconstruction algorithms is helpful but not required. |
| Supervisor: | Prof. Dr. Torben Ferber |
| Contact: | Lea Reuter (Team Tracking) |
| Last update: | 28.11.2025 |
| Topic: | GPU-accelerated track reconstruction using ACTS and Graph Neural Networks at Belle II and the FCC-ee |
| Summary: | Modern collider experiments like Belle II and the proposed FCC-ee generate large amounts of tracking data. Graph Neural Networks (GNNs) are increasingly used to find track candidates from raw hits. In this project, you will take the output of a GNN-based track finder and integrate it into ACTS (A Common Tracking Software), a modern C++ toolkit for track fitting. The focus is on implementing this connection and optimizing it for execution on GPUs. You will work in C++ and use accelerator frameworks like CUDA or SYCL to build a fast and scalable solution. The setup should be modular and suitable for both current (Belle II) and future (FCC-ee) detectors. For an example of GNN-based track finding, see the CAT project at Belle II. |
| You learn: | ACTS track reconstruction, GPU programming in C++ (CUDA/SYCL), integration of machine learning output, performance tuning and benchmarking, software development in particle physics |
| Prerequisites: | C++, introductory MSc course in particle physics (e.g. lecture and all exercises TP1), good knowledge in Python and/or C++ programming (e.g. lecture and all exercises in "Rechnernutzung"), ideally experience in own software projects, ideally MSc course on modern methods of data analysis, good communication skills (English). ❤️ We love nerds: If you have very good C++ skills or experience with CUDA or SYCL, please reach out even if you have no prior courses in particle physics. |
| Supervisor: | Prof. Dr. Torben Ferber |
| Contact: | Marc Neu (Team Real-Time Computing) |
| Last update: | 28.11.2025 |
| Topic: | FPGA-accelerated tracking using AMD Alveo V80 accelerator cards |
| Summary: | Real-time tracking in high-rate environments like Belle II or FCC-ee requires serious compute power, but maybe GPUs aren't always the answer. In this project, you'll explore an alternative: running core parts of the track reconstruction pipeline on AMD Alveo V80 FPGA cards, built for low-latency acceleration in data centers. Using high-level synthesis (HLS) and streaming I/O, you’ll implement modules such as hit filtering and candidate selection on the FPGA, in close collaboration with FPGA experts at ITIV. |
| You learn: | FPGA programming with HLS (e.g. Vitis), low-latency data processing, particle tracking logic, hardware/software co-design, performance tuning, resource optimization |
| Prerequisites: | C++, basic knowledge of digital logic and FPGAs, interest in pushing code into hardware, ideally some experience with HLS tools or RTL (e.g. from a lab or project), curiosity about data acquisition and real-time systems. ❤️ We love nerds: If you have very good C++ skills or experience with HLS, please reach out even if you have no prior courses in particle physics. |
| Supervisor: | Prof. Dr. Torben Ferber |
| Contact: | Marc Neu (Team Real-Time Computing) |
| Last update: | 28.11.2025 |
| Topic: | Next generation AI for the upgrade of the Belle II track trigger |
| Summary: | The Belle II experiment relies on its Central Drift Chamber (CDC) for precise charged particle tracking and momentum measurement. As the experiment prepares for future upgrades and increased luminosity, improving the CDC track reconstruction algorithms becomes a crucial challenge. The candidate will work within the interdisciplinary real-time computing team at KIT ETP and ITIV to develop real-time machine learning-based tracking methods. Special attention will be given to handling background noise and missing hits, as well as optimizing inference on cutting-edge real-time hardware, such as AMD Xilinx Versal AI Edge Series Gen 2 platforms. You will have access to dedicated hardware at KIT for testing machine learning models on real-time architectures, ensuring that proposed solutions can be validated under experimental conditions. Close collaboration with electrical engineering students at ITIV will be necessary to co-design efficient hardware-software interfaces for real-time inference. |
| You learn: | Python programming, machine learning, track reconstruction, algorithm optimization |
| Prerequisites: | introductory MSc course in particle physics (e.g. lecture and all exercises TP1), good knowledge in Python and/or C++ programming (e.g. lecture and all exercises in "Rechnernutzung", ideally experience in own software projects), ideally MSc course on modern methods of data analysis, good communication skills (english) |
| Supervisor: | Prof. Dr. Torben Ferber |
| Contact: | Prof. Dr. Torben Ferber (Team Real-Time Computing) |
| Last update: | 28.11.2025 |
| Topic: | First search for B0 → p anti-n ℓ ν at Belle II |
| Summary: | The persistent disagreement between inclusive and exclusive determinations of the Cabibbo-Kobayashi-Maskawa (CKM) matrix element |Vub| limits the possibility to overconstrain the CKM unitarity triangle in efforts to test the Standard Model (SM). A well-established strategy to determine |Vub| is to use measurements of semileptonic B-meson decays with b → uℓν transitions. These relatively abundant decays offer theoretically clean avenues to perform precise measurements of SM parameters, due to the factorization of the leptonic and hadronic final states. However, a major challenge for determinations of |Vub| is suppressing the CKM-favoured B → Xcℓν background, which exhibits a similar experimental signature and is O(100) times more abundant than B → Xuℓν decays. The signal extraction process is further complicated by the known resonant states comprising only a third of the total inclusive branching fraction, while non-resonant contributions of B → Xuℓν remaining poorly understood. For these reasons, the B → Xuℓν modelling uncertainty is hard to quantify and becomes dominant for both inclusive and exclusive studies. Decays involving baryons have remained largely unmeasured and could potentially constitute a sizeable component of the total inclusive B → Xuℓν branching. A first measurement of B0 → p anti-nℓν decays, a mode only accessible at Belle II, would provide much needed input to reduce modelling uncertainties in future B → Xuℓν studies, increase our current understanding of b → uℓν transitions, and shed light on the possible reasons for the infamous inclusive/exclusive puzzle. |
| You learn: | Python programming, machine learning, track reconstruction, algorithm optimization |
| Prerequisites: | introductory MSc course in particle physics (e.g. lecture and all exercises TP1), basic knowledge in Python programming (e.g. lecture and all exercises in "Rechnernutzung", also see our BSc/MSc FAQ), ideally MSc course on modern methods of data analysis, good communication skills (english) |
| Supervisor: | Prof. Dr. Torben Ferber |
| Contact: | Dr. Raynette Van Tonder (Team Flavour Physics) |
| Last update: | 28.11.2025 |
| Topic: | Search for dark photons and axionlike particles in single-photon events at Belle II |
| Summary: | Belle II is a world-leading experiment in the search for dark sector candidates in the GeV range, such as dark photons or axion-like particles (ALPs). Invisibly decaying or long-lived dark photons and ALPs give rise to events with a single, monochromatic photon in the event, a unique and very challenging experimental signature. You will join our team our team to work on this search, which is currently focused on on-shell dark photons, with the aim of extending it to so-called off-shell (heavy) dark photons and ALPs. You will learn how to use tools for generating simulated datasets with Monte Carlo techniques and for statistical analysis. |
| You learn: | Python programming, data analysis, statistics |
| Prerequisites: | introductory MSc course in particle physics (e.g. lecture and all exercises TP1), basic knowledge in Python programming (e.g. lecture and all exercises in "Rechnernutzung", also see our BSc/MSc FAQ), ideally MSc course on modern methods of data analysis |
| Supervisor: | Prof. Dr. Torben Ferber |
| Contact: | Dr. Giacomo De Pietro (Team Dark Physics) |
| Last update: | 28.11.2025 |
| Topic: | Search for long-lived dark photons decaying into visible final states at Belle II |
| Summary: | Among all the dark sector searches, the Belle II experiment has a unique sensitivity to a weakly coupled dark photon whose decays have a displaced vertex. You will start a new search for long-lived dark photons decaying into a pair of muons, pions, kaons and, more challenging, electrons, with the aim of exploring previously unexplored parameters’ space. You will learn how to characterise the detector performance for the reconstruction of displaced vertices and use tools for statistical analysis. |
| You learn: | Python programming, data analysis, statistics |
| Prerequisites: | introductory MSc course in particle physics (e.g. lecture and all exercises TP1), basic knowledge in Python programming (e.g. lecture and all exercises in "Rechnernutzung", also see our BSc/MSc FAQ), ideally MSc course on modern methods of data analysis |
| Supervisor: | Prof. Dr. Torben Ferber |
| Contact: | Dr. Giacomo De Pietro (Team Dark Physics) |
| Last update: | 28.11.2025 |
| Topic: | Search for B+→l nu gamma at Belle II |
| Summary: | The radiative leptonic decay B+ -> l+nu gamma yields important information for the theoretical predictions of non-leptonic B meson decays into light-meson pairs. The emission of the photon probes the first inverse moment λ_B of the light-cone distribution amplitude (LCDA) of the B meson. This parameter is a vital input to QCD factorisation schemes for the non-perturbative calculation of non-leptonic B meson decays. This project aims to observe the B+→l+nu gamma decay for the first time and set an improved limit on λ_B using modern machine learning analysis techniques. |
| You learn: | python programming, data analysis |
| Prerequisites: | introductory MSc course in particle physics (e.g. lecture and all exercises TP1), basic knowledge in Python programming (e.g. lecture and all exercises in "Rechnernutzung", also see our BSc/MSc FAQ), ideally MSc course on modern methods of data analysis |
| Supervisor: | Prof. Dr. Torben Ferber |
| Contact: | Dr. Pablo Goldenzweig |
| Last update: | 28.11.2025 |
| Topic: | Search for Bs→φπ0 decays at Belle |
| Summary: | The Belle experiment, which concluded in 2011, collected a unique sample of Υ(5S) decays to Bs meson pairs in the clean e+e- collision environment. This dataset has yet to be fully exploited. The rare decay Bs→φπ0 is strongly suppressed in the Standard Model and has yet to be observed. However, in a theoretical analysis motivated by the Kπ CP-puzzle, models with modified or additional Z bosons allow for an increase of the branching fraction by an order of magnitude without inconsistencies with other measurements. The Kπ CP-puzzle consists of an unexpectedly large direct CP asymmetry in the decays B± → K±π0 and B0 → K±π∓. These decays are dominated by isospin-conserving processes, but have a small contribution from isospin-violating penguin processes as well. In the isospin-violating decay Bs→φπ0 the penguin processes dominate, which means that potential NP contributions can have a much larger relative effect. If these contributions exist, an observation of the Bs→φπ0 decay may be possible with the Belle Υ(5S) dataset. |
| You learn: | python programming, data analysis |
| Prerequisites: | introductory MSc course in particle physics (e.g. lecture and all exercises TP1), basic knowledge in Python programming (e.g. lecture and all exercises in "Rechnernutzung", also see our BSc/MSc FAQ), ideally MSc course on modern methods of data analysis |
| Supervisor: | Prof. Dr. Torben Ferber |
| Contact: | Dr. Pablo Goldenzweig (Team Flavour Physics) |
| Last update: | 28.11.2025 |
| Topic: | Bs→K+ l- nu with Y(5S) using machine learning-based Full Event Interpretation |
| Summary: | There are two conceptually different methods to determine the CKM matrix element Vub, a factor in the amplitude of transitions from a b-quark to a u-quark and a W boson. As quarks cannot be detected directly, one has to handle the hadronisation to infer Vub, even in the simplest case, where the W boson decays into a charged lepton and a neutrino. One method, which is easier to calculate but is experimentally more challenging, takes into account all possible final states that typically occur during the hadronisation process (inclusive measurement). The other involves calculating the hadronisation effects and searching for only a single final state, e.g., B0->pi-l+nu (exclusive measurement). Both approaches have been pursued and the result is a tantalizing 3sigma discrepancy between the measurements of the same quantity Vub. By measuring Bs->K-l+nu we can shed light on the possible reasons or sources for this, as we replace the spectator d quark in B0->pi-l+nu with an s quark. Since the calculation of the hadronisation in the exclusive measurement relies on Lattice QCD, the kaon in the final state simplifies the calculation significantly. A result close to the existing exclusive measurement would strengthen the trust in that measurement and could hint to new particles beyond the Standard Model. A result closer to the inclusive measurement would more likely be interpreted as a hint that the calculations for the exclusive measurement are unreliable and that the inclusive measurement gives the real value of Vub. This will constitute a first measurement of the rare Bs→K-l+nu decay by employing a modern machine learning-based algorithm to reconstruct the full Y(5S) event for the first time. |
| You learn: | machine learning, python programming, data analysis |
| Prerequisites: | introductory MSc course in particle physics (e.g. lecture and all exercises TP1), basic knowledge in Python programming (e.g. lecture and all exercises in "Rechnernutzung", also see our BSc/MSc FAQ), ideally MSc course on modern methods of data analysis |
| Supervisor: | Prof. Dr. Torben Ferber |
| Contact: | Dr. Pablo Goldenzweig (Team Flavour Physics) |
| Last update: | 28.11.2025 |