Multilayer spintronic neural networks with radiofrequency connections.

Spintronic nano-synapses and nano-neurons perform neural network operations with high accuracy thanks to their rich, reproducible and controllable magnetization dynamics. These dynamical nanodevices could transform artificial intelligence hardware, provided they implement state-of-the-art deep neural networks. However, there is today no scalable way to connect them in multilayers. Here we show that the flagship nano-components of spintronics, magnetic tunnel junctions, can be connected into multilayer neural networks where they implement both synapses and neurons thanks to their magnetization dynamics, and communicate by processing, transmitting and receiving radiofrequency signals. We build a hardware spintronic neural network composed of nine magnetic tunnel junctions connected in two layers, and show that it natively classifies nonlinearly separable radiofrequency inputs with an accuracy of 97.7%. Using physical simulations, we demonstrate that a large network of nanoscale junctions can achieve state-of-the-art identification of drones from their radiofrequency transmissions, without digitization and consuming only a few milliwatts, which constitutes a gain of several orders of magnitude in power consumption compared to currently used techniques. This study lays the foundation for deep, dynamical, spintronic neural networks.

Bimodal ionic photomemristor based on a high-temperature oxide superconductor/semiconductor junction.

Memristors, a cornerstone for neuromorphic electronics, respond to the history of electrical stimuli by varying their electrical resistance across a continuum of states. Much effort has been recently devoted to developing an analogous response to optical excitation. Here we realize a novel tunnelling photo-memristor whose behaviour is bimodal: its resistance is determined by the dual electrical-optical history. This is obtained in a device of ultimate simplicity: an interface between a high-temperature superconductor and a transparent semiconductor. The exploited mechanism is a reversible nanoscale redox reaction between both materials, whose oxygen content determines the electron tunnelling rate across their interface. The redox reaction is optically driven via an interplay between electrochemistry, photovoltaic effects and photo-assisted ion migration. Besides their fundamental interest, the unveiled electro-optic memory effects have considerable technological potential. Especially in combination with high-temperature superconductivity which, in addition to facilitating low-dissipation connectivity, brings photo-memristive effects to the realm of superconducting electronics.

An Oxygen Vacancy Memristor Ruled by Electron Correlations.

Resistive switching effects offer new opportunities in the field of conventional memories as well as in the booming area of neuromorphic computing. Here the authors demonstrate memristive switching effects produced by a redox-driven oxygen exchange in tunnel junctions based on NdNiO3 , a strongly correlated electron system characterized by the presence of a metal-to-insulator transition (MIT). Strikingly, a strong interplay exists between the MIT and the redox mechanism, which on the one hand modifies the MIT itself, and on the other hand radically affects the tunnel resistance switching and the resistance states' lifetime. That results in a very unique temperature behavior and endows the junctions with multiple degrees of freedom. The obtained results bring up fundamental questions on the interplay between electronic correlations and the creation and mobility of oxygen vacancies in nickelates, opening a new avenue toward mimicking neuromorphic functions by exploiting the electric-field control of correlated states.

Evaluation of Self-Field Effects in Magnetometers Based on Meander-Shaped Arrays of Josephson Junctions or SQUIDs Connected in Series.

Arrays of superconducting quantum interference devices (SQUIDs) are highly sensitive magnetometers that can operate without a flux-locked loop, as opposed to single SQUID magnetometers. They have no source of ambiguity and benefit from a larger bandwidth. They can be used to measure absolute magnetic fields with a dynamic range scaling as the number of SQUIDs they contain. A very common arrangement for a series array of SQUIDs is with meanders as it uses the substrate area efficiently. As for most layouts with long arrays, this layout breaks the symmetry required for the elimination of adverse self-field effects. We investigate the scaling behavior of series arrays of SQUIDs, taking into account the self-field generated by the bias current flowing along the meander. We propose a design for the partial compensation of this self-field. In addition, we provide a comparison with the case of series arrays of long Josephson junctions, using the Fraunhofer pattern for applications in magnetometry. We find that compensation is required for arrays of the larger size and that, depending on the technology, arrays of long Josephson junctions may have better performance than arrays of SQUIDs.

Imaging the itinerant-to-localized transmutation of electrons across the metal-to-insulator transition in V2O3.

In solids, strong repulsion between electrons can inhibit their movement and result in a "Mott" metal-to-insulator transition (MIT), a fundamental phenomenon whose understanding has remained a challenge for over 50 years. A key issue is how the wave-like itinerant electrons change into a localized-like state due to increased interactions. However, observing the MIT in terms of the energy- and momentum-resolved electronic structure of the system, the only direct way to probe both itinerant and localized states, has been elusive. Here we show, using angle-resolved photoemission spectroscopy (ARPES), that in V2O3, the temperature-induced MIT is characterized by the progressive disappearance of its itinerant conduction band, without any change in its energy-momentum dispersion, and the simultaneous shift to larger binding energies of a quasi-localized state initially located near the Fermi level.

A quantum material spintronic resonator.

In a spintronic resonator a radio-frequency signal excites spin dynamics that can be detected by the spin-diode effect. Such resonators are generally based on ferromagnetic metals and their responses to spin torques. New and richer functionalities can potentially be achieved with quantum materials, specifically with transition metal oxides that have phase transitions that can endow a spintronic resonator with hysteresis and memory. Here we present the spin torque ferromagnetic resonance characteristics of a hybrid metal-insulator-transition oxide/ ferromagnetic metal nanoconstriction. Our samples incorporate [Formula: see text], with Ni, Permalloy ([Formula: see text]) and Pt layers patterned into a nanoconstriction geometry. The first order phase transition in [Formula: see text] is shown to lead to systematic changes in the resonance response and hysteretic current control of the ferromagnetic resonance frequency. Further, the output signal can be systematically varied by locally changing the state of the [Formula: see text] with a dc current. These results demonstrate new spintronic resonator functionalities of interest for neuromorphic computing.

Subthreshold firing in Mott nanodevices.

Resistive switching, a phenomenon in which the resistance of a device can be modified by applying an electric field1-5, is at the core of emerging technologies such as neuromorphic computing and resistive memories6-9. Among the different types of resistive switching, threshold firing10-14 is one of the most promising, as it may enable the implementation of artificial spiking neurons7,13,14. Threshold firing is observed in Mott insulators featuring an insulator-to-metal transition15,16, which can be triggered by applying an external voltage: the material becomes conducting ('fires') if a threshold voltage is exceeded7,10-12. The dynamics of this induced transition have been thoroughly studied, and its underlying mechanism and characteristic time are well documented10,12,17,18. By contrast, there is little knowledge regarding the opposite transition: the process by which the system returns to the insulating state after the voltage is removed. Here we show that Mott nanodevices retain a memory of previous resistive switching events long after the insulating resistance has recovered. We demonstrate that, although the device returns to its insulating state within 50 to 150 nanoseconds, it is possible to re-trigger the insulator-to-metal transition by using subthreshold voltages for a much longer time (up to several milliseconds). We find that the intrinsic metastability of first-order phase transitions is the origin of this phenomenon, and so it is potentially present in all Mott systems. This effect constitutes a new type of volatile memory in Mott-based devices, with potential applications in resistive memories, solid-state frequency discriminators and neuromorphic circuits.

Robust Coupling between Structural and Electronic Transitions in a Mott Material.

The interdependences of different phase transitions in Mott materials are fundamental to the understanding of the mechanisms behind them. One of the most important relations is between the ubiquitous structural and electronic transitions. Using IR spectroscopy, optical reflectivity, and x-ray diffraction, we show that the metal-insulator transition is coupled to the structural phase transition in V_{2}O_{3} films. This coupling persists even in films with widely varying transition temperatures and strains. Our findings are in contrast to recent experimental findings and theoretical predictions. Using V_{2}O_{3} as a model system, we discuss the pitfalls in measurements of the electronic and structural states of Mott materials in general, calling for a critical examination of previous work in this field. Our findings also have important implications for the performance of Mott materials in next-generation neuromorphic computing technology.

Vowel recognition with four coupled spin-torque nano-oscillators.

In recent years, artificial neural networks have become the flagship algorithm of artificial intelligence1. 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 synchronization2-6, for solving complex problems with small networks7-11. This approach is especially interesting for hardware implementations, because emerging nanoelectronic devices can provide compact and energy-efficient nonlinear auto-oscillators that mimic the periodic spiking activity of biological neurons12-16. The dynamical couplings between oscillators can then be used to mediate the synaptic communication between the artificial neurons. One challenge for using nanodevices in this way is to achieve learning, which requires fine control and tuning of their coupled oscillations17; the dynamical features of nanodevices can be difficult to control and prone to noise and variability18. Here we show that the outstanding tunability of spintronic nano-oscillators-that is, the possibility of accurately controlling their frequency across a wide range, through electrical current and magnetic field-can be used to address 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 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 nonlinear dynamical features such as oscillations and synchronization.