Spin-waves (SWs) are low-loss, collective excitation of electromagnetic waves in magnetic films, e.g. permalloy, yttrium iron garnet (YIG), bismuth substituted lutetium iron garnet (BLIG) etc. Shape, size and defects of the structure, magnetic material properties, and external saturation magnetic field can influence SWs characteristics greatly.

In a magnetic thin film, SWs can be excited in different methods e.g. optical excitation, acoustic wave, microwave-antenna excitation etc. In our experiment, we focus on magneto-static surface SWs (MSSWs) that propagate perpendicular to the external saturation field. With the all-electrical spectroscopy method, we have detected MSSWs of frequencies lying between 2 to 2.5 GHz in a BLIG thin film. We are trying to build an MSSW based delay-line oscillator. Such delay-line oscillators are a promising candidate for physical reservoir computing.

However, electrical spectroscopy is unable to give the spatio-temporal distribution of SWs. Here, the micromagnetic simulation method comes into the picture. We have studied SW spatio-temporal dynamics in simulated magnonic systems of nm, and μm size, but magnetic films of a few mm long are used in our experiment. Simulation of an mm-sized film is memory as well as time-consuming. So, data-driven nonlinear dynamics predicting algorithm based on sparse regression, i.e. SINDy algorithm is being used to circumvent this issue. We will build a SINDy model by training it with magnetization data acquired from micromagnetic simulation of a magnonic system that runs for a brief period of time. This model will be able to predict the future dynamics of the system in a resource-efficient manner.