Speaker
Description
Magnonic devices exploit spin waves - the collective excitations of ordered magnets - for wave-based information processing without charge flow. Their performance hinges on a high-dimensional design space spanning geometry, local material parameters, magnetic field landscape, excitation, and nonlinearity, much of which is inaccessible to manual parameter studies.
Inverse design reverses the conventional workflow: the desired functionality is specified as an objective, and an algorithm finds the optimal structure automatically. The field has advanced rapidly since its founding works in 2021 [Wang et al., Nat. Commun 12, 2636 (2021); Papp et al., Nat. Commun. 12, 6422 (2021)], encompassing topology optimisation of magnonic demultiplexers and filters, ion-irradiation-based gradient-index lenses, experimentally realised reconfigurable radio-frequency devices and Boolean logic gates, and the development of dedicated differentiable micromagnetic solvers that bring the infrastructure of machine learning to the Landau-Lifshitz-Gilbert equation.
In this talk, I give an overview of inverse-design magnonics, organising the emerging literature along two axes - the design degrees of freedom that can be optimised, and the algorithmic toolbox available to do so - and lay out the principal bottlenecks that currently constrain the field. Building on this, I outline what I view as the most promising open frontiers - such as input shaping, nonlinear design and integrated amplification - with the long-term vision of a universal, software-defined magnonic device.