Physics-based generative materials design
Contact person: Anders Malthe-S?renssen
Keywords: Generative models, Neural networks, Sustainability, Materials science, Physics-based models
Research group: Njord Center
Department of Physics, Department of Geoscience
There is an urgent need for new materials for sustainable energy applications: materials that reduce friction or wear, light but strong materials, degradable materials with controlled lifetimes or materials to adsorb contaminants. But how can we design materials with specific properties? Recent advances in atomic-scale simulations allow properties of specific structures to be calculated. However, such simulations are computationally demanding and there are too many possible structures to search through them all. This has inspired the interdisciplinary field of generative materials design where physics-based models are used to calculate properties of a set of structures, which are then used to train either discriminative machine learning models that allow rapid classification of candidate structures or generative machine learning methods that generate structures with given properties. Research proposals in this field may span several methodological approaches within this field as well as a number of application domains.
Methodological research topics:
- Fundamentally new modeling approaches
- Methods for constrained generative modeling
- Efficient atomic scale modeling using machine learning
- Explainable models and model representations
Topics from natural sciences or technology:
- Geological materials that open for scalable applications
- Sustainable materials design
- Engineering of e.g. desalination or nano-plastic filtration systems
- Materials with tuneable mechanical properties and self-healing materials
Mentoring and internship will be offered by a relevant external partner.