For the first time, inverse design has been used to engineer specific functionalities into a universal spin-wave-based device. It was created by Andrii Chumak and colleagues at Austria’s University of Vienna, who hope that their magnonic device could pave the way for substantial improvements to the energy efficiency of data processing techniques.
Inverse design is a fast-growing technique for developing new materials and devices that are specialized for highly specific uses. Starting from a desired functionality, inverse-design algorithms work backwards to find the best system or structure to achieve that functionality.
“Inverse design has a lot of potential because all we have to do is create a highly reconfigurable medium, and give it control over a computer,” Chumak explains. “It will use algorithms to get any functionality we want with the same device.”
One area where inverse design could be useful is creating systems for encoding and processing data using quantized spin waves called magnons. These quasiparticles are collective excitations that propagate in magnetic materials. Information can be encoded in the amplitude, phase, and frequency of magnons – which interact with radio-frequency (RF) signals.
Collective rotation
A magnon propagates by the collective rotation of stationary spins (no particles move) so it offers a highly energy-efficient way to transfer and process information. So far, however, such magnonics has been limited by existing approaches to the design of RF devices.
“Usually we use direct design – where we know how the spin waves behave in each component, and put the components together to get a working device,” Chumak explains. “But this sometimes takes years, and only works for one functionality.”
Recently, two theoretical studies considered how inverse design could be used to create magnonic devices. These took the physics of magnetic materials as a starting point to engineer a neural-network device.
Building on these results, Chumak’s team set out to show how that approach could be realized in the lab using a 7×7 array of independently-controlled current loops, each generating a small magnetic field.
Thin magnetic film
The team attached the array to a thin magnetic film of yttrium iron garnet. As RF spin waves propagated through the film, differences in the strengths of magnetic fields generated by the loops induced a variety of effects: including phase shifts, interference, and scattering. This in turn created complex patterns that could be tuned in real time by adjusting the current in each individual loop.
To make these adjustments, the researchers developed a pair of feedback-loop algorithms. These took a desired functionality as an input, and iteratively adjusted the current in each loop to optimize the spin wave propagation in the film for specific tasks.
This approach enabled them to engineer two specific signal-processing functionalities in their device. These are a notch filter, which blocks a specific range of frequencies while allowing others to pass through; and a demultiplexer, which separates a combined signal into its distinct component signals. “These RF applications could potentially be used for applications including cellular communications, WiFi, and GPS,” says Chumak.
While the device is a success in terms of functionality, it has several drawbacks, explains Chumak. “The demonstrator is big and consumes a lot of energy, but it was important to understand whether this idea works or not. And we proved that it did.”
Through their future research, the team will now aim to reduce these energy requirements, and will also explore how inverse design could be applied more universally – perhaps paving the way for ultra-efficient magnonic logic gates.
The research is described in Nature Electronics.
The post Inverse design configures magnon-based signal processor appeared first on Physics World.
