Research Topics of the Department of Theoretical Solid State Physics

The Dreßler group investigates condensed matter systems using molecular dynamics simulations and develops multiscale models to describe ion transport in energy materials.

Molecular dynamics simulations of condensed matter

The Dreßler group focuses on the investigation of the structure and dynamics of condensed matter systems using ab initio classical and machine learned force field simulations. Our research aims to achieve a detailed microscopic understanding of complex materials and interfaces. In particular, we study chemical charge transport in energy related materials, water semiconductor interfaces, the behavior of binary liquids confined in nanopores, and the solvent dependence of chemical reaction barriers.

Machine learned force field simulations

Machine learned force field based simulations enable us to investigate aspects of ion transport that are challenging to capture with traditional ab initio molecular dynamics. Our work includes studies of proton transport in fuel cell membranes [1], hydroxide diffusion in aqueous solution [2], lithium ion mobility in crystalline LixSiy [3], and sulfur vacancy migration in MoS₂ [4]. Multi nanosecond machine learned force field simulations provide converged diffusion coefficients, accurate activation energies, and mechanistic explanations for experimental observations such as defect formation and memristive behavior.

[1] Grunert, Großmann, Hänseroth, Flötotto, Oumard, Wolf, ..., DreßlerModeling Complex Proton Transport Phenomena - Exploring the Limits of Fine-Tuning and Transferability of Foundational Machine-Learned Force Fields.J. Phys. Chem. C (2025) 129(21), 9662-9669. https://doi.org/10.1021/acs.jpcc.5c02064

[2] Hänseroth, DreßlerOptimizing machine learning interatomic potentials for hydroxide transport: Surprising efficiency of single-concentration training.J. Chem. Phys. (2025) 163(8): 084118. https://doi.org/10.1063/5.0284063

[3] Qaisrani, Kirsch, Flötotto, Hänseroth, Oumard, Sebastiani, DreßlerBridging Atomistic and Mesoscale Lithium Transport via Machine-Learned Force Fields and Markov Modelshttps://doi.org/10.48550/arXiv.2511.20863

 

[4] Flötotto, Spetzler, von Stackelberg, Ziegler, Runge, DreßlerLarge-scale cooperative sulfur vacancy dynamics in two-dimensional MoS₂ from machine-learning interatomic potentialshttps://doi.org/10.48550/arXiv.2508.13790

How to create accurate machine-learned force fields

Machine learned force fields enable molecular dynamics simulations with near ab initio accuracy while reaching time scales comparable to those of classical molecular dynamics simulations.

Foundational machine learned force field models are typically trained on structures from large databases and are designed to enable simulations of a wide range of compounds without further modification. However, their accuracy can often be significantly improved by fine tuning these models using a small amount of compound specific reference data. In a benchmark study, we demonstrated this concept in our work titled Fine Tuning Unifies Foundational Machine Learned Interatomic Potential Architectures at ab initio Accuracy [1]. This study shows that each state of the art graph neural network architecture can be fine tuned to achieve highly accurate models.

During our own fine tuning efforts, we experienced the fragmentation of current software ecosystems for machine learned force fields first hand. This motivated the development of our Python based aMACEing toolkit [2], which provides tools for the semi automatic fine tuning of foundational machine learned force fields. The toolkit supports the generation of training data, the fine tuning process itself, and the execution of machine learned force field based molecular dynamics simulations across multiple software packages.

TAB project "KapMemLyse"

The Dreßler Group is a member of the TAB research group KapMemLyse, whose aim is the investigation and development of capillary-supported  alkaline anion exchange membrane electrolysis for the production of green hydrogen. Within the framework of this project, the physical fundamentals in the area of ion and mass transport are being established, and a scalable concept for components and cells is being developed. This new technology is intended to combine the advantages of the capillary electrolysis according to Wallace et al. with the innovative anion exchange membrane water electrolysis technology. In a paper published in Nature in 2022, it was already demonstrated that efficiencies of 98% can be achieved at a current density of 0.5 A/cm².

In this project, the Dreßler Group is particularly focused on predicting the hydroxide conductivity and gas permeability of the membrane materials using molecular dynamics simulations. This is achieved through the use of classical, ab initio, and machine-learned force-field molecular dynamics.

Multiscale approaches for the simulation of long-range proton transfer

Bild MultiskalenansatzChristian Dreßler

The Dressler group has developed a multiscale simulation approach for describing proton long range transfer on the micrometer length and millisecond time scales.[1] We have shown that our approach can be used to calculate the proton conductivity of nanoporous CsH2PO4 systems as a function of porosity.[2] The multiscale method is based on a combination of molecular dynamics simulation and probabilistic propagation of protons by a Monte Carlo approach. Utilizing this method, we are able to condense the information concerning the proton dynamics of an entire molecular dynamics trajectory into a single transition matrix, which allows for the simulation of much larger systems.[3] Through careful mathematical analysis of this matrix, qualitative properties of the model and information about the mechanism of proton conduction can be derived. In the future, we plan to extend the multiscale approach to the transport of other types of ions. In general, we are interested in developing Markov models for the description of any microscopic dynamical processes.

Why do we need multiscale approaches?

Ab initio molecular dynamics simulations can be used to simulate proton transfer in bulk-phase homogeneous materials, but their application is limited to small system sizes of several hundred atoms and simulation times of less than a nanosecond. These accessible time and length scales are far too small for describing realistic, non-ideal, inhomogeneous, and nanostructured proton conducting materials.

[1] Dreßler, Kabbe, Brehm, Sebastiani. Exploring non-equilibrium molecular dynamics of mobile protons in the solid acid CsH2PO4 at the micrometer and microsecond scale. The Journal of Chemical Physics, 152, 164110, 2020.

[2] Wagner, Dreßler, Lohmann-Richters, Hanus, Sebastiani, Varga, Abel. Mechanism of ion conductivity through polymer-stabilized CsH2PO4 nanoparticular layers from experiment and theory. Journal of Materials Chemistry A , 7, 27367–27376, 2019.

[3] Dreßler, Kabbe, Brehm, Sebastiani. Dynamical matrix propagator scheme for large-scale proton dynamics simulations, The Journal of Chemical Physics, 152, 114114, 2020.

 

Efficient representations of the linear density-density response function

The static linear density-density response function and, in particular, its frequency-dependent pendant are complex and theoretically rich objects that have applications in many fields. The Dreßler group is investigating low-dimensional representations of the static linear density-density response function for the efficient calculation of intermolecular electrostatic polarization interactions.[1,2] If the response function is known, molecular density responses to arbitrary perturbations can be calculated at a fraction of the computational cost of a first principle calculation.[3] In the future, we plan to develop a polarizable force field based on the efficient representation of the static linear density-density response function. We also plan to extend the efficient representation to the dynamical variant of the response function.

[1] Dreßler, Scherrer, Ahlert, Sebastiani. Efficient representation of the linear density-density response function, Journal of Computational Chemistry, 40, 2712–2721, 2019.

[2] Dreßler, Sebastiani, Reduced eigensystem representation of the linear density-density response function, International Journal of Quantum Chemistry, 120, e26085, 2020.

[3] Dreßler, Sebastiani, Polarization Energies from Efficient Representation of the Linear Density-Density Response Function, Advanced Theory and Simulation, 4, 2000260, 2021.

 
Densityresponce_1Christian Dreßler