Experimental Demonstration of a Spin-Wave Lens Designed With Machine Learning

We present the design and experimental realization of a device that acts like spin-wave lens i.e., it focuses spin waves to specified location. The structure does not resemble any conventional design, is nonintuitive pattern produced by machine learning algorithm. As tool, we used our custom micromagnetic solver "SpinTorch" has built-in automatic gradient calculation can perform backpropagation through time for propagation. training itself performed with saturation magnetization YIG film as variable parameter, goal guide predefined verified operation in widely mumax3 solver, realization. For implementation, developed technique create effective saturation-magnetization landscapes direct focused-ion-beam irradiation. This allows us rapidly transfer nanoscale patterns medium, without patterning material etching. measured corresponding FIB dose levels advance this mapping translate designed scatterer required levels. Our demonstration serves proof concept workflow be realize more sophisticated devices complex functionality, e.g., signal processors, or neuromorphic devices.