An analog magnon adder for all-magnonic neurons

TL;DR

This paper introduces an analog magnon adder that integrates spin-wave pulses using a resonator and parametric amplifier, serving as a foundational component for all-magnonic neurons in neural network applications.

Contribution

It presents the first implementation of an analog magnon adder using a resonator and parametric amplifier, enabling integration of spin-wave signals for magnonic neural networks.

Findings

Successful demonstration of spin-wave pulse integration

Resonator with parametric amplifier compensates losses effectively

Potential for building all-magnonic neural network components

Abstract

Spin-waves are excellent data carriers with a perspective use in neuronal networks: Their lifetime gives the spin-wave system an intrinsic memory, they feature strong nonlinearity, and they can be guided and steered through extended magnonic networks. In this work, we present a magnon adder that integrates over incoming spin-wave pulses in an analog fashion. Such an adder is a linear prequel to a magnonic neuron, which would integrate over the incoming pulses until a certain nonlinearity is reached. In this work, the adder is realized by a resonator in combination with a parametric amplifier which is just compensating the resonator losses.