Spin-wave platform in hybrid PMA/YIG films

• Applicants: Professor Dr.-Ing. Markus Becherer; Dr. Carsten Dubs
• Subject Area: Electronic Semiconductors, Components and Circuits, Integrated Systems, Sensor Technology, Theoretical Electrical Engineering ; Synthesis and Properties of Functional Materials
• Term: since 2019
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 429656450

This proposal aims to design and experimentally realize analog signal processing devices in out-of-plane magnetized Yttrium-Iron-Garnet (YIG) thin films with robust perpendicular magnetic anisotropy (PMA). Combined with fine-tuned ferromagnetic bistable PMA nanomagnets for locally modifying the dispersion of the spin wave media, we propose a reconfigurable neuromorphic magnonic signal processing/computing platform. The PMA nanomagnets are switched via electrically controlled nanowire crossbar arrays, enabling device reconfigurability during runtime. We expect the array to be reprogrammed on timescales of two-digit nanoseconds per nanomagnet and microseconds in the whole array. Fully electrical reconfigurability is based on perpendicular nanomagnetic logic (pNML), where the magnets' engineered switching field distributions and fine-tuned stray-field generation are central. The nanomagnets prevent static power loss, which is beneficial compared to current-induced spin-wave steering concepts, as power is needed only for reconfiguration. Furthermore, we aim for highly optimized LPE-grown iron garnet PMA material as the spin-wave media. The lowest or even no required bias fields are targeted to ensure isotropic wave propagation in the forward-volume (FV) configuration. At the same time, we want to extend YIG fabrication methods to reach for the lowest damping and highest propagation velocities and simultaneously for a significant improvement in magneto-optical response to facilitate device characterization. The neuromorphic signal processing, performed directly in the spin-wave media, evolves on two-digit nanosecond timescales. This allows for neuromorphic reconfigurability such that neuromorphic learning epochs can be carried out in dedicated simulation engines (i.e., SpinTorch, a branch of PyTorch for micromagnetic platforms) and parallel on the developed spin-wave hardware. The fabricated devices will exhibit low-power characteristics simultaneously with high computational speed, a feat that is very hard to achieve with other technologies. The electrically reconfigurable spin wave devices developed will open new avenues for learning-in-hardware, inverse design, and neuromorphic computing for applications like Edge AI: low-power decision-making on end-user devices. Our device platform focuses on promising applications and will significantly exceed the state-of-the-art in magnonic technology.

• DFG Programme: Research Grants