High-throughput methods for ab-initio interface tribology

Autonomous screening of a large number of structures for beneficial properties using simulations, known as high-throughput computations, are becoming increasingly popular for material discovery. The large amount of data produced in such a way is then commonly analyzed with machine- learning algorithms that mine through the data and automatically group it. High-throughput methods for ab-initio Interface Tribology (HIT) will pioneer the use of these methods to find material pairings with beneficial frictional properties. To this end a suite of interoperable software tools for high-throughput computations of interfaces in sliding contact will be developed and a large amount of reference data on friction related properties will be produced. The developed software tools will individually be useful by themselves for a wide variety of tasks in surface and interface science, but together they will allow even an inexperienced user to perform self-correcting computations on interfacial friction without having to worry about computational parameters or other issues which usually require a specialist. At the same time we will construct a large, public, and searchable database containing relevant figures of merit for a large number of materials, starting with simple monoatomic crystals rubbing against a mirrored counterpart, to complex multiatomic systems (e.g. hard coatings or solid lubricants). Such a database does not exist to date, but would be very useful in choosing materials for novel applications, especially on the micro- and nano-scale. Furthermore it can be used as a source for reliable input parameters for large-scale continuum simulations which play an important role in mechanical engineering. Since the method is based on a quantum mechanical description of the electrons in the modeled materials, the accuracy of the simulations is very high and conclusions about the fundamental relations between the quantum properties of a crystal or interface and its mechanical attributes can be drawn. These include e.g. adhesion strength, resistance to sliding, and the likelihood of the production of stress-relieving lattice defects which can travel through the crystal. Such defects, called dislocations, are the major source of plasticity in metals and have to date not been connected well with friction studies on the nanoscale, although they certainly play a major role. With ASH, we will thus be able to judge how interfaces will perform and provide information for the design of novel layered meta-materials tailored to specific applications. The multitude of data produced in HIT will be analyzed with the help of artificial intelligence methods based on neural networks. This will enable us to not only find hidden correlations in the data, but will also allow us to make predictions about the sliding performance of interface systems that are not yet in the database or cannot be calculated due to computational restrictions on the system size.

In our project "High-throughput methods for ab-initio interface tribology (HIT)" we aimed to contribute to the scientific understanding of friction and wear, essential factors influencing various aspects of daily life, from machinery performance to medical implant reliability. Inherent limitations, mostly cost and time-related, exist with traditional experimental approaches to studying these phenomena. We sought to overcome these limitations through advanced computational methods. Our primary objective was to develop high-throughput techniques for efficiently characterizing the interfacial regions of crystalline materials using quantum-mechanical simulations. By doing so, we aimed to automate the computation of critical friction and wear properties of solid-solid material couplings, such as adhesion and shear strength. These properties are crucial for predicting and designing materials with tailored frictional behavior. A significant aspect of our approach was the creation of user-friendly software tools, making complex interface simulations accessible to a wider audience, including non-experts. By democratizing these techniques, we hope to have catalyzed advancements in computational tribology and paved the way for leveraging big data in this field. Our efforts yielded several important outcomes. We produced a fully autonomous workflow to compute surface energies and nanoparticle shapes of arbitrarily complex crystals and used it to produce a dataset of over 650 surface energies for 36 diverse materials. These data will serve as a valuable resource for researchers worldwide, providing a benchmark for new and fast approaches utilizing machine-learned potentials. We also developed a workflow to match arbitrary surfaces to form coherent interfaces and evaluate their characteristic properties under lateral sliding and a compressive load. All computations are automatic, streamlining essential tasks and accelerating scientific progress. This workflow can be combined with the surface energy code to produce stable interface structures with nothing but the bulk structures of the two materials as input. Furthermore, the project was part of a landmark study for the evaluation and verification of density functional theory codes. We and our collaborators expanded on the ideas of a previous study from 2016. We expanded the dataset to 960 elemental crystals and oxides that cover the entire periodic table. This effort established reference datasets and lays the groundwork for future verification studies in the field, enhancing the trustworthiness of computational approaches and contributing to advancements in materials science, benefiting various industries and technologies. Looking ahead, our work holds promise for various practical applications, from designing more efficient mechanical systems to improving battery technology and solar cells. By advancing our understanding of interfaces, we aim to contribute to a future where friction and wear are better managed, and multi-layered hetero-structures can be predicted in the computer before they are experimentally verified.