High-throughput methods for ab-initio interface tribology

To date, most questions about the tribology are answered by experiments. Simulations however, offer large advantages in cost and also control, by allowing to monitor buried interfaces directly. Due to increased computer power and high accuracy, ab-intio based tribology simulations have become more common in the last decades. However, only a couple of dozens of interesting systems have been analyzed so far on the quantum level, hindering the discovery of universal trends and generalization of results.
The project High-throughput methods for ab-initio Interface Tribology (HIT) is concerned with introducing advanced high-throughput methods to interface science with special regard to tribology. Flexible and inter-operable software modules based on density functional theory and the Automated Interactive Infrastructure and Database for Computational Science (AiiDA), will be developed to autonomously compute key tribological figures of merit for nearly arbitrary solid-state interfaces. From an accurate description of the generalized stacking fault energy surface obtained by ab-initio methods, we will compute e.g. adhesion- and shear-strength, dislocation core structure and energy, and the stress needed to move the dislocation over the lattice. Since our calculation are rooted in quantum mechanical description of the materials, we will also analyze the fundamental connection between the mechanical and tribological properties of an interface with its electronic structure. By investigating an extensive amount of different, technologically relevant interfaces (including metal/metal, metal/ceramic, hard coatings/solid lubricants,...) we will be able to find high level descriptors describing the basic tribological properties of these systems. These can then be used to make predictions for tailoring interfaces for specific use cases, like heterostructures and low friction coatings.
HIT will make all data publicly available in a searchable database which will be automatically created by the AiiDA code we will produce. This will be the first time a comprehensive database for tribological figures of merit produced with ab-initio methods will be available. Machine learning algorithms (supervised and unsupervised learning) will be used on the large number of raw data to refine and curate it, discover trends within it, and suggest further interesting pairings that will show beneficial properties.
The work packages will be implemented by a small, focused team of a post doctoral researcher (the principal investigator or PI) and a PhD student. The PI, Michael Wolloch, has extensive experience with DFT-based nano-tribology and high throughput methods and has recently been involved in the development of the direct precursors to HIT. The PhD student will be selected after advertising the position internationally, where candidates with previous experience in machine learning will be strongly favored. HIT has also secured the collaboration of international experts Prof. M. Clelia Righi, and Prof. Petr Lazar, who will support the project with their expertise in computational tribology (Prof. Righi) and theory of dislocations (Prof. Lazar).