New rare earth lean permanent magnets from computational design and the challenge of the 4f electrons

Computational design has become a powerful tool to tailor materials properties such as the magnetocrystalline anisotropy (MCA) of permanent magnets. Fe-rich REFe_12−xZ_x (RE = rare earth) phases are promising candidates for future permanent magnets. However, accurate predictions of magnetic properties of 4f systems are challenging since the localization of the 4f states varies with the RE element and its chemical environment and different degrees of localization might call for different theoretical modelling. Aiming to predict new high-performance permanent magnets we used NdFe₁₁Ti as a prototype for an 1:12 magnet and investigated the dependence of the magnetic properties of the electronic structure approach. While for Sm a core treatment is sufficient [1], the conic MCA of NdFe11Ti needs at least an intermediate sized Hubbard U correction [2]. Fully localized Nd 4f states wrongly produce a uniaxial MCA at low temperatures. Using the Nd_1−xY_xFe_12−yTi_y as a test case we investigated how far the strong dependence of the magnetic properties on the description of the Nd 4 f states influences the prediction of new phases. The magnetic properties were determined in a full-potential LMTO framework (RSPt) while for structural relaxation VASP was used. Results were compared to single crystal measurements.