Abstract:The recent passing of the petaflop per second landmark by the Roadrunner project at the Los Alamos National Laboratory marks a preliminary peak of an impressive world-wide development in the high-performance scientific computing sector. Also, purely academic state-of-the-art supercomputers such as the IBM Blue Gene/P at Forschungszentrum Jülich allow us nowadays to investigate large systems of the order of 10(3) spin polarized transition metal atoms by means of density functional theory. Three applications will be presented where large-scale ab initio calculations contribute to the understanding of key properties emerging from a close interrelation between structure and magnetism. The first two examples discuss the size dependent evolution of equilibrium structural motifs in elementary iron and binary Fe-Pt and Co-Pt transition metal nanoparticles, which are currently discussed as promising candidates for ultra-high-density magnetic data storage media. However, the preference for multiply twinned morphologies at smaller cluster sizes counteracts the formation of a single-crystalline L1(0) phase, which alone provides the required hard magnetic properties. The third application is concerned with the magnetic shape memory effect in the Ni-Mn-Ga Heusler alloy, which is a technologically relevant candidate for magnetomechanical actuators and sensors. In this material strains of up to 10% can be induced by external magnetic fields due to the field induced shifting of martensitic twin boundaries, requiring an extremely high mobility of the martensitic twin boundaries, but also the selection of the appropriate martensitic structure from the rich phase diagram.

Deep-learning electronic-structure calculation of magnetic superstructures

A deep-learning method for studying magnetic superstructures

Accelerating the electronic-structure calculation of magnetic systems by equivariant neural networks

Equivariant Neural Network Force Fields for Magnetic Materials

A deep equivariant neural network approach for efficient hybrid density functional calculations

Deep Learning Illuminates Spin and Lattice Interaction in Magnetic Materials

Efficient hybrid density functional calculation by deep learning

Super-resolution of magnetic systems using deep learning

Deep-learning density functional theory Hamiltonian for efficient ab initio electronic-structure calculation

DeepH-2: Enhancing deep-learning electronic structure via an equivariant local-coordinate transformer

Magnetic Hamiltonian Parameter Estimation Using Deep Learning Techniques

Application of deep learning methods to the study of magnetic phenomena

Improving density matrix electronic structure method by deep learning

General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian

Machine learning magnetism classifiers from atomic coordinates

Deep learning density functional theory Hamiltonian in real space

Neural network representation of electronic structure from $ab$ $initio$ molecular dynamics

Complex Spin Hamiltonian Represented by an Artificial Neural Network

Simulating functional magnetic materials on supercomputers