Magnet Circularity Evaluator MCE
Design your desired magnetic materials and let AI predict their most important properties. Compare their performance with that of standard materials from our database of calculated crystal structures.
With the magnetic materials designer, we show how we create databases of materials and their properties and use machine learning to facilitate the search for new components. As an example, this tool examines a new hard magnetic phase (known as the 1-12-X phase, e.g., NdFe12N), which, with its lower rare earth content, represents a promising alternative to commercially available hard magnets.
The calculation tool answers the question of the chemical composition in this phase, i.e., which elements make up the optimal magnet with the highest magnetic energy product and the highest anisotropy. Instead of having to simulate each selected option individually using quantum mechanics, AI is used to display the properties directly and compare them with common standard materials.
Properties are determined for the 1-12-X phase. This means that one can select one rare earth atom, 12 transition metals, and one light element. If a total of 13 or more transition metal atoms are specified in the button (with the plus sign continuing to increase), it is indicated that this material is not compatible. The calculable properties are:
Gain insight into the effects of impurities from alloy components on magnetic properties. Determine the tolerable amount of particular elements to achieve your desired materials properties.
The calculation tool shows how changes in chemical composition affect the magnetic properties of materials and provides insights into the tolerable amount of specified elements. Our model is based on neodymium-iron-boron magnets. These account for approximately half of all permanent magnets used worldwide. The other half are almost exclusively ferrites, which are less critical due to the absence of rare earths. Because Nd2Fe14B contains neodymium rare earths and is currently the strongest permanent magnet on the market (also widely used in wind turbines and electric motors), recycling is very important here.
The properties of hard magnetic compounds are calculated using the quantum mechanical method TB-LMTO-ASA [1]. However, performing these calculations for all possible compositions is computationally inefficient. To solve this problem, we calculated the properties for a specific subgroup of compositions, all of which are present in the same phase of Nd2Fe14B1 (also known as the 2-14-1 phase), and trained a machine learning model to predict the properties of other compositions. Our focus is on predicting the two key properties of magnetization and anisotropy. The general framework of our machine learning model is derived from previous work on the 1-12-X phase [2]. In addition, estimates of the carbon footprint can be made.
[1] Drebov, N.; Gercsi, Z.; Delczeg-Czirjak, E. K.; Bergqvist, L.; Nordström, L.; Eriksson, O.; Vitos, L., Ab initio screening methodology applied to the search for new permanent magnetic materials, New Journal of Physics 15 (2013) Art. 125023 Link
[2] Möller, J. J.; Schäfer, R.; Körner, W.; Kruk, R.; Hahn, H., Compositional optimization of hard-magnetic phases with machine-learning models, Acta Materialia 153 (2018) 53–61 Link
Calculate how hydrogen moves through iron and chromium alloys under variable conditions. Determine the relative hydrogen concentration as a function of time.
With this interactive tool, you can investigate how hydrogen penetrates iron and chromium alloys under different environmental conditions. The diffusivity of hydrogen has been calculated at the atomic level and tabulated, allowing the tool to simulate real-time simulations of the charging and discharging of a component containing hydrogen.
The question is how the hydrogen concentration in the materials changes over time when the alloy composition or external parameters such as hydrogen partial pressure and temperature change. The tabulated values are based on atomistic calculations of the energy barriers that hydrogen atoms must overcome for diffusion and a simulation of the macroscopic properties in the diffusion network, i.e., in the possible diffusion paths of the materials.
This method can also be applied to other complex systems, such as ferritic materials, i.e., alloys with elements that occur in the same crystal structure. In principle, all structures with a clearly defined diffusion network are possible. However, the calculation method must be adapted for this purpose. In the current version, the calculation tool is adapted to ferritic materials such as steels, as this is the most common area of application for materials in contact with hydrogen.
The variable (operating) conditions can be used in the calculation tool to simulate the practical situation (sample thickness, hydrogen partial pressure, chromium content, temperature).
Fraunhofer Institute for Mechanics of Materials IWM