Magnetism at interfaces: from quantum to reality

Permanent magnets are a key technology for modern society with applications in air conditioning, mobility or power generation. The quality of a permanent magnet is given by the strength to withstand an external applied field. At a certain external field, the coercive field, the magnet loses its magnetic state. The measured coercive fields in modern permanent magnets reach only a small fraction of theoretical values. A series of experimental studies have shown that discontinuities and misalignment at the at omic scale significantly affect the coercivity. Understanding defects at the atomic scale, and their relation to the quality of the magnet, helps to modify and improve the production process of permanent magnets and to obtain stronger permanent magnets. Since fabrication and experiments are expensive and time consuming, the best way to investigate magnets is modelling and simulation. In this project, we develop a quantitative model of coercivity, taking into account the local atomic structure, the spatial variation of the magnetic properties close to the atomic defects, and the physical microstructure of the magnet. To achieve this goal we bridge different length scales, from sub-nanometer features up to a few micrometers. Usually it is difficult to pass information between different length scales. With the help of modern computer systems and sophisticated algorithms we bridge the different length scales to obtain the coercivity of a permanent magnet based on its smallest entity, the atomic structure. The developed model is guided by well described magnetic materials to validate the system throughout the progress of the project. There are material systems, which are examined experimentally and theoretically for many years, like for example Iron-Nickel (FeNi) and Iron- Platinum (FePt). Here we start to develop and validate our multiscale approach. Afterwards we use the developed model to analyse and understand more complex material systems. We invite international experts for each material system to cooperate closely between experiments and our developed model. Throughout the project we gather information from experiments and from simulations. We apply data assimilation, which is a machine learning model, to adjust the simulation model according to coercivity measurements by correcting systematic errors. Using the developed multiscale model and the machine leaning model, we will be able to describe magnetic material systems based on its atomistic structure and to develop stronger permanent magnets.