Strategic Scientific Workshop Aarhus Bayreuth
The goal of this workshop is to provide a forum for the in-depth exchange of ideas and insights within the broad domain of data-driven materials modelling, particularly using cutting edge machine learning (ML) and artificial intelligence (AI) methods. In this context, the closely related topics of materials design and structure prediction are of particular interest. The former is often considered a holy grail in materials modelling, since the reliable computational prediction of improved materials would avoid time- and resource-consuming experimental trials, which are currently impeding a fast transition towards more sustainable technologies in energy generation, catalysis and electronics, to name just a few. Similarly, predicting the structures of novel materials with atomistic precision is essential for understanding their functionality and rationally improving their properties.
With support from the UBT Humboldt Centre, this workshop brings the groups of Prof. Mie Andersen and Prof. Bjørk Hammer (both from Aarhus University), along with a distinguished set of invited speakers to Bayreuth.
The workshop is presented by the Research Center for AI in Science and Society.
Venue: SWO Tagungszentrum, Universitätsstraße 30, 95447 Bayreuth
Monday June 30th:
18:00-19:00 | Tutorial | Martin Vondrák et al.: “Equivariant neural networks from scratch” Background theory and setting up coding environments |
19:00- | Dinner | Pizza at the Venue |
Tuesday July 1st:
9:00-12:30 | Tutorial w. coffee break | Martin Vondrák et al.: “Equivariant neural networks from scratch” Hands-on exercises |
12:30-13:30 | Lunch | Mensa “Frischraum” |
Session topic: Data-driven catalysis modelling | ||
13:30-14:30 | Invited talk | Nongnuch Artrith: "(Machine) learning what makes a catalyst good" |
14:30-15:30 | 3 contributed talks | Raffaele Cheula: “Modeling and design of catalyst materials with graph models and machine learning potentials” Luuk Kempen: „Data-efficient modeling of metal-oxide interfaces using global optimization and machine learning“ Marius Juul Nielsen: „Prediction of adsorption energies at metal-oxide interfaces with interpretable machine learning“ |
15:30-16:00 | Coffee break | |
Session topic: Machine learning potentials & foundational models | ||
16:00-17:00 | Invited talk | Venkat Kapil: "Universal MLIPs: Towards Quantum Accuracy in Molecular Materials" |
17:00-18:00 | 3 contributed talks | Mads-Peter Verner Christensen: “Gibbs free energy optimization enabled by graph neural networks” Maciej Baradyn: “Catalysis at a Solid-Liquid Interface” Karlo Sovic: “Surface Adsorbate Structure Search” |
20:00- | Dinner | Oskar (https://www.oskar-bayreuth.de/) |
Wednesday July 2nd:
Session topic: Machine learning potentials, training and application | ||
9:00-10:00 | Invited talk | Arghya Bhowmik: “GNNs for Forward and Generative Modelling of Molecules and Materials” |
10:00-10:40 | 2 contributed talks | Martin Vondrák: “Electrostatics in ML Potentials” Konstantin Jakob: “Predicting and Modeling Disorder in Solids” |
10:40-11:10 | Coffee break | |
11:10-12:10 | Invited talk | Milica Todorovic: “Active Learning for Materials Optimisation” |
12:10-12:50 | 2 contributed talks | Nina Buckova: "Dynamics of water: First principles van der Waals forces in machine-learning interatomic potentials" Nils Gönnheimer: “How good are foundation models for vibrational properties and free energies?” |
12:50-14:00 | Lunch | Mensa “Frischraum” |
14:30- | Outing | Hike to Eremitage, including a guided tour. |
19:00 | Dinner | Manns-Bräu (https://www.mannsbraeu.de/) |
Thursday July 3rd:
Session topic: Machine learning potentials, dynamics and structure determination | ||
9:00-10:00 | Invited talk | Philipp Schienbein: “IR Spectroscopy and Electric Field Simulations Enabled by the Atomic Polar Tensor Neural Network” |
10:00-10:40 | 2 contributed talks | Joe Pitfield: “Accelerating Structure Prediction with Active Learning and Universal Potentials” Elisabeth Keller: “Testing foundation models for oxide solid- solid interfaces” |
10:40-11:10 | Coffee break | |
11:10-12:10 | Invited talk | Georg Madsen: “Taming inhomogeneous oxide surfaces using NNFFs” |
12:10-12:50 | 2 contributed talks | Emanuele Telari: "Phase diagrams of Ag nanoparticles from grand canonical replica exchange basin-hopping and ML potentials" Hyunwook Jung: “ML for automated transition state searching” |
12:50-14:00 | Lunch | Mensa “Frischraum” |
14:00- 15:00 | Invited talk | Harald Oberhofer: “Explainable ML Highlights Structural Influences on Perovskite Elasticities” |
15:00-18:30 | Tutorial w. coffee break | Mads-Peter Verner Christensen, Luuk Kempen: “Exploring reinforcement learning in materials science” Hands-on exercises |
20:00- | Dinner | Liebesbier (https://www.liebesbier.de/en/) |