Summary
A fully funded PhD position is available in the group of Prof. Tristan Bereau at Heidelberg University. The project focuses on developing generative machine learning methods (specifically diffusion models) to construct amorphous molecular thin films and estimate free energies at coarse-grained resolutions. The position is embedded in the SIMPLAIX research initiative funded by the Klaus Tschira Foundation and conducted in collaboration with the group of Prof. Ullrich Köthe.
Fully Funded PhD Position: Generative Machine Learning for Molecular Thin Films, Tristan Bereau, Germany
Designation
PhD Student (Fully Funded, 3 years at 75% E13 TV-L)
Research Area
- Deep Generative Modeling (Diffusion Models, Score Learning)
- Statistical Mechanics & Thermodynamics
- Molecular Simulation & Coarse-Graining
- Amorphous Molecular Thin Films / Organic-Electronic Materials
Location
Institute for Theoretical Physics, Heidelberg University, Heidelberg, Germany
Eligibility / Qualification
- Required Degree: A degree in Physics is required for admission to the Heidelberg Graduate School for Physics.
- Core Background: Solid grounding in statistical mechanics.
- Technical Skills: Strong Python and PyTorch skills along with experience in machine learning and/or molecular simulation.
- Mindset: Genuine interest in method development and comfort with mathematical formalism.
Job Description
- Develop generative machine learning methods for amorphous molecular thin films, which govern the performance of organic-electronic materials.
- Address the computational bottleneck of brute-force molecular dynamics by building coarse-grained representations using diffusion models.
- Formulate models that generate not only realistic supramolecular structures but also compute free energies.
- Collaborate closely with the research group of Prof. Ullrich Köthe within the SIMPLAIX initiative.
Key Relevant Publications
| Title | Journal | Volume / Year | DOI |
|---|---|---|---|
| Data-efficient multidimensional free energy estimation via physics-informed score learning | The Journal of Chemical Physics | Vol. 165 (2026) | 10.1063/5.0327436 |
| Split-Flows: Measure Transport and Information Loss Across Molecular Resolutions | AISTATS Proceedings | Vol. 267 (2026) | 10.48550/arXiv.2511.01464 |
| Fokker–Planck Score Learning: Efficient Free-Energy Estimation under Periodic Boundary Conditions | The Journal of Physical Chemistry B | Vol. 129 (2025) | 10.1021/acs.jpcb.5c04579 |
| Solvation free energies from neural thermodynamic integration | The Journal of Chemical Physics | Vol. 162 (2025) | 10.1063/5.0251736 |
| Neural Thermodynamic Integration: Free Energies from Energy-Based Diffusion Models | The Journal of Physical Chemistry Letters | Vol. 15 (2024) | 10.1021/acs.jpclett.4c01958 |
How to Apply
To apply, submit the following materials directly via email to Prof. Tristan Bereau at bereau@thphys.uni-heidelberg.de:
- Curriculum Vitae (CV)
- Short Statement of Interest
- Contact details / names of two academic referees
Last Date to Apply
Until the position is filled.
