Neurons in the brain behave as non-linear oscillators, which develop rhythmic activity and interact to process information1. Taking inspiration from this behavior to realize high density, low power neuromorphic computing will require huge numbers of nanoscale non-linear oscillators. Indeed, a simple estimation indicates that, in order to fit a hundred million oscillators organized in a two-dimensional array inside a chip the size of a thumb, their lateral dimensions must be smaller than one micrometer. However, despite multiple theoretical proposals2–5, and several candidates such as memristive6 or superconducting7 oscillators, there is no proof of concept today of neuromorphic computing with nano-oscillators. Indeed, nanoscale devices tend to be noisy and to lack the stability required to process data in a reliable way. Here, we show experimentally that a nanoscale spintronic oscillator8,9 can achieve spoken digit recognition with accuracies similar to state of the art neural networks. We pinpoint the regime of magnetization dynamics leading to highest performance. These results, combined with the exceptional ability of these spintronic oscillators to interact together, their long lifetime, and low energy consumption, open the path to fast, parallel, on-chip computation based on networks of oscillators.
Spin-based electronics has evolved into a major field of research that broadly encompasses different classes of materials, magnetic systems, and devices. This review describes recent advances in spintronics that have the potential to impact key areas of information technology and microelectronics. We identify four main axes of research: nonvolatile memories, magnetic sensors, microwave devices, and beyond-CMOS logic. We discuss state-of-the-art developments in these areas as well as opportunities and challenges that will have to be met, both at the device and system level, in order to integrate novel spintronic functionalities and materials in mainstream microelectronic platforms.Conventional information processing and communication devices work by controlling the flow of electric charges in integrated circuits. Such circuits are based on nonmagnetic semiconductors, in Technologies based on GMR and MTJ devices are now firmly established and compatible with CMOS fab processes. Yet, in order to meet the increasing demand for high-speed, high-density, and low power electronic components, the design of materials, processes, and spintronic circuits needs to be continuously innovated. Further, recent breakthroughs in basic research brought forward novel phenomena that allow for the generation and interconversion of charge, spin, heat, and optical signals.Many of these phenomena are based on non-equilibrium spin-orbit interaction effects, such as the spin Hall and Rashba-Edelstein effects 6,8,23 or their thermal 24 and optical 25,26 analogues. Spin-orbit torques (SOT), for example, can excite any type of magnetic materials, ranging from metals to semiconductors and insulators, in both ferromagnetic and antiferromagnetic configurations 6 . This versatility allows for the switching of single layer ferromagnets, ferrimagnets, and antiferromagnets, as well as for the excitation of spin waves and auto-oscillations in both planar and vertical device geometries 10,11 . Charge-spin conversion effects open novel pathways for information processing using Boolean logic, as well as promising avenues for implementing unconventional neuromorphic 27,28,29 and probabilistic 30 computing schemes. Finally, spintronic devices cover a broad bandwidth ranging from DC to THz 31,32 , leading to exciting opportunities for the on-chip generation and detection of high frequency signals.
In recent years, artificial neural networks have become the flagship algorithm of artificial intelligence 1 . In these systems, neuron activation functions are static and computing is achieved through standard arithmetic operations. By contrast, a prominent branch of neuroinspired computing embraces the dynamical nature of the brain and proposes to endow each component of a neural network with dynamical functionality, such as oscillations, and to rely on emergent physical phenomena, such as synchronization 2-7 , for computing complex problems with small size networks [7][8][9][10][11] . This approach is especially interesting for hardware implementations, as emerging nanoelectronic devices can provide highly compact and energy-efficient non-linear auto-oscillators that mimic the periodic spiking activity of biological neurons [12][13][14][15][16] . The dynamical couplings between oscillators can then be used to mediate the synaptic communication between neurons. However, one major challenge towards implementing these models with nano-devices is to achieve learning, which requires finely controlling and tuning their coupled oscillations 17 . The dynamical features of nanodevices can indeed be difficult to control, and prone to noise and variability 18 . In this work, we show that the outstanding tunability of spintronic nano-oscillators, i.e. the possibility to widely and accurately control their frequency through electrical current and magnetic field, can solve this challenge. We successfully train a hardware network of four spin-torque nano-oscillators to recognize spoken vowels by tuning their frequencies according to an automatic real-time learning rule. We show that the high experimental recognition rates stem from the outstanding ability of these oscillators to synchronize. Our results demonstrate that non-trivial pattern classification tasks can be achieved with small hardware neural networks by endowing them with non-linear dynamical features: here, oscillations and synchronization. This demonstration of real-time learning with an array of four spin-torque nano-oscillators is a milestone for spintronics-based neuromorphic computing.Spin-torque nano-oscillators are natural candidates for building hardware neural networks made of coupled nanoscale oscillators [8][9][10]13,15,18,19 . These nanoscale magnetic tunnel junctions emit microwave
In recent years, spin–orbit effects have been widely used to produce and detect spin currents in spintronic devices. The peculiar symmetry of the spin Hall effect allows creation of a spin accumulation at the interface between a metal with strong spin–orbit interaction and a magnetic insulator, which can lead to a net pure spin current flowing from the metal into the insulator. This spin current applies a torque on the magnetization, which can eventually be driven into steady motion. Tailoring this experiment on extended films has proven to be elusive, probably due to mode competition. This requires the reduction of both the thickness and lateral size to reach full damping compensation. Here we show clear evidence of coherent spin–orbit torque-induced auto-oscillation in micron-sized yttrium iron garnet discs of thickness 20 nm. Our results emphasize the key role of quasi-degenerate spin-wave modes, which increase the threshold current.
Wave control in the solid state has opened new avenues in modern information technology. Surface-acoustic-wave-based devices are found as mass market products in 100 millions of cellular phones. Spin waves (magnons) would offer a boost in today's data handling and security implementations, i.e., image processing and speech recognition. However, nanomagnonic devices realized so far suffer from the relatively short damping length in the metallic ferromagnets amounting to a few 10 micrometers typically. Here we demonstrate that nm-thick YIG films overcome the damping chasm. Using a conventional coplanar waveguide we excite a large series of short-wavelength spin waves (SWs). From the data we estimate a macroscopic of damping length of about 600 micrometers. The intrinsic damping parameter suggests even a record value about 1 mm allowing for magnonics-based nanotechnology with ultra-low damping. In addition, SWs at large wave vector are found to exhibit the non-reciprocal properties relevant for new concepts in nanoscale SW-based logics. We expect our results to provide the basis for coherent data processing with SWs at GHz rates and in large arrays of cellular magnetic arrays, thereby boosting the envisioned image processing and speech recognition.
Seven decades after the discovery of collective spin excitations in microwave-irradiated ferromagnets, there has been a rebirth of magnonics. However, magnetic nanodevices will enable smart GHz-to-THz devices at low power consumption only, if such spin waves (magnons) are generated and manipulated on the sub-100 nm scale. Here we show how magnons with a wavelength of a few 10 nm are exploited by combining the functionality of insulating yttrium iron garnet and nanodisks from different ferromagnets. We demonstrate magnonic devices at wavelengths of 88 nm written/read by conventional coplanar waveguides. Our microwave-to-magnon transducers are reconfigurable and thereby provide additional functionalities. The results pave the way for a multi-functional GHz technology with unprecedented miniaturization exploiting nanoscale wavelengths that are otherwise relevant for soft X-rays. Nanomagnonics integrated with broadband microwave circuitry offer applications that are wide ranging, from nanoscale microwave components to nonlinear data processing, image reconstruction and wave-based logic.
We investigate experimentally and analytically the impact of thermal noise on the sustained gyrotropic mode of vortex magnetization in spin transfer nano-oscillators and its consequence on the linewidth broadening due to the different nonlinear contributions. Performing some time domain measurements, we are able to extract separately the phase noise and the amplitude noise at room temperature for several values of dc current and perpendicular field. For a theoretical description of the experiments, we extend the general model of nonlinear auto-oscillators to the case of vortex core dynamics and provide some analytical expressions of the response-to-noise of the system as the coupling coefficient between the phase and the amplitude of the vortex core dynamics due to the nonlinearities. From the analysis of our experimental results, we demonstrate the major role of the amplitude-to-phase noise conversion on the linewidth broadening, and propose some solutions to improve the spectral coherence of vortex based spin transfer nano-oscillators.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.