Symmetry-guided and AI-accelerated design of intercalated transition metal dichalcogenides for antiferromagnetic spintronics

Abstract

The advancement of antiferromagnetic spintronics depends on quantum materials with target symmetry-dictated functionalities, however, their systematic discovery is hindered by the immense configurational complexity of the available material space. Here, we introduce a symmetry-guided, AI-accelerated framework incorporating graph neural networks with high generalization ability to overcome this bottleneck. Based on fully intercalated transition metal dichalcogenides (iTMDs) and using only 200 relaxed partially intercalated structures for transfer learning, our model effectively explores more than 100,000 partially intercalated configurations and identifies 35 altermagnetic and 20 $T\tau$-antiferromagnetic ground-state candidates. Interestingly, we show that tuning spin-group symmetry through intercalant arrangement or magnetic ordering realizes a series of d-wave altermagnets in these hexagonal systems with high spin-charge conversion efficiency. Furthermore, we reveal plentiful $T\tau$-antiferromagnets enabling efficient N\'eel spin-orbit torque switching, driven by giant $T$-odd spin Edelstein susceptibilities. These results establish iTMDs as a versatile platform for spintronics and provide a general strategy for the accelerated design of symmetry-enforced quantum materials.