PhD Position: Deep learning for phase-contrast synchrotron X-ray tomography

PhD Position: Deep learning for phase-contrast synchrotron X-ray tomography

Helmholtz Zentrum Hereon operates an outstation at DESY in Hamburg, providing access to highly brilliant synchrotron radiation through its German Engineering Materials Science Center (GEMS). At the PETRA III synchrotron radiation source, Hereon jointly operates several beamlines. Our imaging beamlines offer advanced micro- and nano-tomography techniques with a strong focus on material science and life science applications. The micro‑tomography setups cover a wide energy range and offer unique phase contrast capabilities. The full-field nano‑tomography setup is among the fastest worldwide due to its unique geometry and beam characteristics. Additionally, the highly coherent beam enables advanced phase contrast methods such as near-field holography.

This PhD position is part of the ErUM‑Data project CmarT, which aims to develop a novel multi‑scale imaging approach at synchrotron radiation facilities, particularly PETRA III at DESY. Biological and materials science samples often exhibit hierarchical structures that determine their function. However, scanning an entire sample volume at the highest spatial resolution is not feasible. Therefore, we need an imaging scheme that captures relevant features at different length scales and integrates them into a single reconstruction volume.

This PhD project focuses on learning‑based phase retrieval in the weak holographic regime, bridging the gap between micro- and nano‑tomography. While propagation‑based phase‑contrast imaging enhances visualization of soft tissues and weakly attenuating samples, phase retrieval in this regime remains challenging, limiting multiscale imaging approaches in near‑field holotomography. To address this, the PhD project combines machine learning, high‑performance computing, and synchrotron‑radiation experiments. The goal is to explore physics‑informed self‑supervised learning approaches (e.g., deep image priors, GANs) and iterative methods, combined with multi‑scale tomography and local adaptive reconstruction to overcome these challenges.

• develop physics‑informed, self‑supervised learning approaches for phase retrieval
• implement reconstruction algorithms on HPC clusters for large‑scale hierarchical tomography
• apply algorithms to experimental tomographic data
• collaborate closely with beamline scientists and project partners
• publish and present results in international journals and conferences

Your profile

• Master’s degree in physics, mathematics, computer science, or a related field
• strong analytical skills and solid theoretical background
• experience in machine/deep learning and image processing/tomography
• advanced programming skills in Python
• excellent command of English (written and spoken)

Assets:

• team‑oriented mindset with strong communication and organizational skills
• experience with HPC clusters, parallel or distributed computing

We offer

• a cutting‑edge research environment at PETRA III imaging beamlines
• collaboration with leading institutions in imaging, mathematics, and data science
• access to high‑performance computing (HPC) resources and advanced experimental setups
• an exciting and varied job in a research centre with around 1,000 employees from more than 60 nations
• a well-connected research campus (public transport) and best networking opportunities, subsidy for the Deutschlandticket if certain conditions are met (job ticket)
• individual opportunities for further training social benefits according to the collective agreement of the public service and remuneration up to pay group 13 according to TV EntgO Bund
• an excellent technical infrastructure and modern workplace equipment
• 6 weeks holiday per year; company holidays between Christmas and New Year's Day
• very good compatibility of private and professional life; offers of mobile and flexible work
• free assistance program for employees (EAP)

Severely disabled persons and those equaling severely disabled persons who are equally suitable for the position will be considered preferentially within the framework of legal requirements.