The Challenge: Electromagnetic systems exhibit complex frequency-dependent behavior governed by resonances. Characterizing these systems requires solving parameterdependent finite element models with millions of degrees of freedom across 10–20 input parameters—a computationally intractable problem for design exploration and optimization.

Modern surrogate modeling techniques (data-driven approximations trained on highdimensional datasets) offer a solution, but their success depends critically on the quality and structure of training data.

The Challenge: Shrinking mobility electronics with increasing functionality create critical EMC challenges. Complex multi-component systems demand innovative simulation approaches that combine classical field solvers with modern AI techniques.

The Opportunity: Join Bosch's EMC Expert Team and TU Darmstadt's TEMF Institute to develop AI-enhanced electromagnetic simulation methods for real eBike applications.

You will:

• Apply AI/ML techniques to accelerate and improve electromagnetic field simulations

• Model and simulate production-relevant eBike EMC radiated emission scenarios

• Master CST Studio Suite and AI frameworks in an industrial context

• Collaborate with experienced EMC engineers and researchers

To support post-processing of CST Studio Suite results from the Wakefield and Eigenmode solvers, an existing modular MATLAB workflow is extended with graphical user interfaces developed using MATLAB App Designer. The goal is to develop MATLAB GUIs (App Designer) that provide a unified and user-friendly interface for routine analysis (import, evaluation, visualization, export). The implementation will follow clean-code principles and will be complemented by an automated test suite that verifies selected outputs against analytically solvable benchmark cases to ensure long-term reliability and reproducibility.

Ray tracing is a central component of geometry-based propagation, as it provides the geometric information required to analyze visibility, intersection points, and reflection paths in complex environments. While the underlying idea is straightforward, the computational cost grows quickly with increasing scene complexity and ray counts, making CPU-based approaches a bottleneck for realistic meshes.

This work develops and validates a modular GPU-accelerated ray-tracing backend based on NVIDIA OptiX. The backend is designed as an independent software component with clean interfaces and a reproducible data pipeline. It exports SBR-ready geometric path data, including hit/miss information, hit positions, surface normals, primitive/material identifiers, and accumulated path lengths for multi-bounce traces.

Collimators are beamline components used to define the beam aperture and intercept halo particles in order to protect downstream equipment and reduce uncontrolled losses. In this project, electromagnetic wakefields in an existing beam collimator geometry will be studied using CST Studio Suite. The geometry is already available and will be imported into CST for numerical wakefield simulations. The focus is on extracting and analyzing key wakefield quantities such as longitudinal and transverse impedances extract loss and kick factors, shunt impedances, quality factors and resonance frequencies. The simulation outputs will be post-processed using the provided post-processing software in MATLAB.

The reduced magnetic vector potential (RMVP) formulation has recently been used to drastically speed up the simulation of accelerator magnets at CERN. The formulation decomposes the magnetic vector potential into several components, depending on the different subdomains of the computational domain. With the currents in a coil being known, it starts by computing a source magnetic vector potential with Biot-Savart's law and proceeds with the corrections towards the total solution. In this process, the magnetostatic Maxwell equations are used.

Extending the reduced magnetic vector potential formulation to the magnetoquaistatic Maxwell equations is not straightforward. Due to the eddy currents, the current distribution in conducting domains is not known beforehand and depends on the surrounding domain. This causes errors in the total solution. It is expected that an iterative scheme that iterates between the different domains can effectively account for the eddy current effects.

The operation of bulk Nb accelerator cavities is extremely expensive due to the low-temperature helium cooling. It is expected that Nb3Sn-coated cavities and cavities with superconductor-insulator-superconductor layered coatings can be operated with higher fields and higher temperatures, leading to a decrease in power consumption. Such coatings are applied to the cavity via sputtering. However, the sputtering distribution on a complex shape, such as a TESLA cavity, is typically inhomogeneous. This leads to the question of what the influence of inhomogeneous sputtering is on the quality of the cavity. Before attempting to answer this question, we want a way to represent this sputtering distribution in a simulation software standard in the field, namely CST.

Computational engineering and numerical simulations are essential in designing and analyzing accelerator magnets, especially as magnet geometries become more complex. During the early design phase, engineers often need to explore multiple magnet designs quickly for rough quality estimates. However, modeling each design in a sophisticated CAD software can be time-consuming and inefficient, as many ideas will not make it past the initial phase.

This proposal aims to accelerate the process by leveraging image recognition. You will develop a Python package that can interpret hand-drawn, 2D magnet geometries– recognizing the iron yoke shape and coil wire positions – and convert them into a format compatible with our in-house Biot-Savart-based solver [1] for magnetic field simulations.

This project focuses on the numerical characterization of sub-wavelength nanostructured photocathodes, which are a promising technology for improving photoemission quantum efficiency. By engineering the surface of the photocathode with a periodic array of nanogrooves, the structure can support a Surface-Plasmon-Polariton (SPP) mode. When coupled with an excitation laser pulse, this SPP mode enhances photon absorption, resulting in an increase in quantum efficiency (QE) and a reduction in the power requirements for the photocathode-laser system.

This research aims to contribute to the development of more efficient electron sources for future applications, particularly in the context of high-duty-cycle upgrades at facilities like the European XFEL (EuXFEL).

High-temperature superconductors (HTS) are essential for high-field applications like particle accelerators. Their nonlinear material properties and strong dependence on temperature and magnetic fields make simulations challenging. Accurate modeling is crucial for optimizing HTS conductor designs.

This thesis aims to implement an electromagnetic simulation of HTS conductors in COMSOL Multiphysics or CST Studio Suite using the T-A or H-ϕ formulation. The focus is on analyzing magnetic field and current distribution under realistic conditions. Depending on the scope, mechanical stress from Lorentz forces may also be investigated to assess coil stability.

A multitude of devices is required to accelerate charged particles on a closed orbit of a synchrotron ring. A beam pipe separates the vacuumized particle beam trajectory from all other equipment in the tunnel and consists classically of conducting material. Bending magnets apply a strong Lorentz force on the particles which must rise proportionally to the particles’ energy. According to Lenz’ law, unwanted eddy currents are induced in the beam pipe in order to resist the change of magnetic flux density.

This work aims to quantify this disadvantageous effect of beam pipes on the field quality of bending magnets.

Introduction: Soon the electromotive market will rapidly grow up. One aspect for this growth is the increase of the energy density of high voltage battery systems. This systems are built up with Li-Ion cells with a module voltage up to 850V. To increase the performance of such systems, the inner resistance of the module has to be as small as possible. As a result, the short circuit current of such High Voltage Batteries reaches up to 20kA. To interrupt such a high short circuit current within 2ms the Pyrotechnical Battery Disconnector was developed by Joyson Safety Systems Aschaffenburg GmbH.

Task: Development of an electrodynamic model of a pyrotechnical battery disconnector. Transient nonlinear electrodynamic FE simulations to improve the design of the Pyrotechnical Battery Disconnector.