Research at the Theory Department

Multiscale Modeling and AI in Materials, Energy and Catalysis

The urgently required transition to a sustainable and circular energy system for our society critically relies on energy conversion and storage devices with significantly improved functionality and longevity. Catalysts for the production of base chemicals, green hydrogen, or synthetic fuels need to work in less energy-intensive (e.g. low temperature) processes, while simultaneously being more active, selective towards the desired product and ideally featuring rejuvenation capabilities for extended durability. Batteries must exhibit higher capacity, faster charging and enhanced cyclability, at reduced content of toxic or critical components and minimized CO₂ footprint in their production. To this end, any advances or rational design in terms of the employed materials hinge ultimately on a quantitative understanding of the relation between the targeted macroscopic functionality and the underlying atomic-scale mechanisms. Activity and selectivity are governed by the capability of specific atomic sites at the surface of the catalyst to make and break chemical bonds; capacity and charging derives from mobility and intercalation of charge carriers like lithium in the energy material; durability from the active components’ ability to regenerate their functional state.

The research at the Theory Department contributes to this context by developing and applying computational simulation and artificial intelligence (AI) approaches. In order to bridge the multiple length and time scales between macroscopic functionality and atomic-scale mechanism, we pursue both bottom-up and top-down strategies. In bottom-up multiscale modeling we advance hierarchical frameworks that integrate various levels of theory and machine learning (ML) all the way from the quantum mechanical description of individual elementary reaction steps at the molecular level up to the full reactor or chemical cell – with the objective to understand the function of existing devices, identify bottlenecks and rationally predict novel or improved materials that overcome the current limitations. This is complemented by top-down inference, where we pioneer AI for experiment planning and control, in particular within the context of emerging automated and self-driving laboratories. We develop AI/ML algorithms and agentic tools that drive combined simulation and experiment platforms for autonomous discovery and multi-objective optimization of catalytic and chemical kinetics, long-term stability, and sustainability/life-cycle analysis (LCA) descriptors.

Topically, current activities focus primarily on heterogeneous (thermal and electro) catalysis and solid-state batteries. This includes novel hybrid forms with combined stimuli like magneto-electrocatalysis or solar batteries, where magnetic fields and solar light are applied on top of the regular electric bias, respectively. A prevailing theme is the strong operando evolution of the involved solid-gas, solid-liquid and solid-solid interfaces in the corresponding devices. This means that either driven by the target functionality or even as one of its key enablers , these interfaces undergo continuous and massive changes in their structure, composition and morphology. This complexity challenges existing modeling, simulation, ML and AI approaches, and is a central motivation for the extensive and multi-faceted method development that has become a characteristic trademark for our work. This work is carried out in different working groups, that address the named objectives either more from the angle of individual application areas or more from a methodological frontier within an overall matrix-like and highly interactive structure. 

We invite you to browse the pages of the different working groups for more details, while the following provides a (non-exhaustive) list of presently targeted focus areas:

• Energy Conversion Processes 
  • (Photo-)Electrocatalysis
  • Photovoltaics and Charge Transfer
  • (Solid-State) Batteries
  • Solar Batteries
     
• Heterogeneous Catalysis 
  • Oxidation Catalysis
  • Syngas Reactions
     
• Surface Nanotechnology
  • Graphene Synthesis
     
• Multiscale Modeling and Data Sciences 
  • Multiscale Modeling Methodology
  • Kinetic Monte Carlo
  • Global Geometry Optimization and Crystal Structure Prediction
  • Machine Learned Potentials
  • Graph Theoretical Approach to Reaction Networks
  • Data Mining and Computational Materials Design
  • Uncertainty and Inverse Problems
  • Computational Fluid Dynamics and Reduced Order Reactor Modelling and Simulation
  • Automatic Process Exploration
  • Machine Learning of Experimental Observables
     
• First-Principles Method Development

  • Heat and Charge Transport

  • Time-Dependent Density Functional Theory

  • Machine-Learning Enhanced Semiempirical Electronic Structure
     

Further information about our research approach can be found in our following reviews on

• Role of the human-in-the-loop in emerging self-driving laboratories for heterogeneous catalysis (2025)

• Exploring catalytic reaction networks with machine learning (2023)

• Human- and machine-centered design of molecules and materials (2022)

• Implicit solvation methods for catalysis at electrified interfaces (2022)

• Adsorption enthalpies for catalysis modeling through machine-learned descriptors (2021)

• Mapping materials and molecules (2020)

• First-principles based multiscale modeling (2019)

• A practical guide to surface kinetic Monte Carlo simulations (2019)

• Recent developments in computational surface catalysis (2018)

• Charge transport in molecular materials (2017)

• Ab initio thermodynamics and first-principles microkinetics (2016)

• O2 adsorption dynamics (2013)

• First-principles kinetic modeling in heterogeneous catalysis (2012)

• First-principles kinetic Monte Carlo methodology (2011)

For further information don't hesitate to contact us, or browse through our recent publications.