Single-shot time-resolved measurements of nanosecond-scale spin-transfer induced switching: stochastic versus deterministic aspects

Emerging memristive artificial neuron and synapse devices for the neuromorphic electronics era

Growth of data eases the way to access the world but requires increasing amounts of energy to store and process. Neuromorphic electronics has emerged in the last decade, inspired by biological neurons and synapses, with in-memory computing ability, extenuating the 'von Neumann bottleneck' between the memory and processor and offering a promising solution to reduce the efforts both in data storage and processing, thanks to their multi-bit non-volatility, biology-emulated characteristics, and silicon compatibility. This work reviews the recent advances in emerging memristive devices for artificial neuron and synapse applications, including memory and data-processing ability: the physics and characteristics are discussed first, i.e., valence changing, electrochemical metallization, phase changing, interfaced-controlling, charge-trapping, ferroelectric tunnelling, and spin-transfer torquing. Next, we propose a universal benchmark for the artificial synapse and neuron devices on spiking energy consumption, standby power consumption, and spike timing. Based on the benchmark, we address the challenges, suggest the guidelines for intra-device and inter-device design, and provide an outlook for the neuromorphic applications of resistive switching-based artificial neuron and synapse devices.

Dynamical Anyon Generation in Kitaev Honeycomb Non-Abelian Spin Liquids

Relativistic Mott insulators known as "Kitaev materials" potentially realize spin liquids hosting non-Abelian anyons. Motivated by fault-tolerant quantum-computing applications in this setting, we introduce a dynamical anyon-generation protocol that exploits universal edge physics. The setup features holes in the spin liquid, which define energetically cheap locations for non-Abelian anyons, connected by a narrow bridge that can be tuned between spin liquid and topologically trivial phases. We show that modulating the bridge from trivial to spin liquid over intermediate time scales-quantified by analytics and extensive simulations-deposits non-Abelian anyons into the holes with O(1) probability. The required bridge manipulations can be implemented by integrating the Kitaev material into magnetic tunnel junction arrays that engender locally tunable exchange fields. Combined with existing readout strategies, our protocol reveals a path to topological qubit experiments in Kitaev materials at zero applied magnetic field.

Antiferromagnetic spintronics: An overview and outlook

Over the past few decades, the diversified development of antiferromagnetic spintronics has made antiferromagnets (AFMs) interesting and very useful. After tough challenges, the applications of AFMs in electronic devices have transitioned from focusing on the interface coupling features to achieving the manipulation and detection of AFMs. As AFMs are internally magnetic, taking full use of AFMs for information storage has been the main target of research. In this paper, we provide a comprehensive description of AFM spintronics applications from the interface coupling, read-out operations, and writing manipulations perspective. We examine the early use of AFMs in magnetic recordings and conventional magnetoresistive random-access memory (MRAM), and review the latest mechanisms of the manipulation and detection of AFMs. Finally, based on exchange bias (EB) manipulation, a high-performance EB-MRAM is introduced as the next generation of AFM-based memories, which provides an effective method for read-out and writing of AFMs and opens a new era for AFM spintronics.

Forecasting the outcome of spintronic experiments with Neural Ordinary Differential Equations

Deep learning has an increasing impact to assist research, allowing, for example, the discovery of novel materials. Until now, however, these artificial intelligence techniques have fallen short of discovering the full differential equation of an experimental physical system. Here we show that a dynamical neural network, trained on a minimal amount of data, can predict the behavior of spintronic devices with high accuracy and an extremely efficient simulation time, compared to the micromagnetic simulations that are usually employed to model them. For this purpose, we re-frame the formalism of Neural Ordinary Differential Equations to the constraints of spintronics: few measured outputs, multiple inputs and internal parameters. We demonstrate with Neural Ordinary Differential Equations an acceleration factor over 200 compared to micromagnetic simulations for a complex problem - the simulation of a reservoir computer made of magnetic skyrmions (20 minutes compared to three days). In a second realization, we show that we can predict the noisy response of experimental spintronic nano-oscillators to varying inputs after training Neural Ordinary Differential Equations on five milliseconds of their measured response to a different set of inputs. Neural Ordinary Differential Equations can therefore constitute a disruptive tool for developing spintronic applications in complement to micromagnetic simulations, which are time-consuming and cannot fit experiments when noise or imperfections are present. Our approach can also be generalized to other electronic devices involving dynamics.

Emerging Artificial Neuron Devices for Probabilistic Computing

In recent decades, artificial intelligence has been successively employed in the fields of finance, commerce, and other industries. However, imitating high-level brain functions, such as imagination and inference, pose several challenges as they are relevant to a particular type of noise in a biological neuron network. Probabilistic computing algorithms based on restricted Boltzmann machine and Bayesian inference that use silicon electronics have progressed significantly in terms of mimicking probabilistic inference. However, the quasi-random noise generated from additional circuits or algorithms presents a major challenge for silicon electronics to realize the true stochasticity of biological neuron systems. Artificial neurons based on emerging devices, such as memristors and ferroelectric field-effect transistors with inherent stochasticity can produce uncertain non-linear output spikes, which may be the key to make machine learning closer to the human brain. In this article, we present a comprehensive review of the recent advances in the emerging stochastic artificial neurons (SANs) in terms of probabilistic computing. We briefly introduce the biological neurons, neuron models, and silicon neurons before presenting the detailed working mechanisms of various SANs. Finally, the merits and demerits of silicon-based and emerging neurons are discussed, and the outlook for SANs is presented.

Real-time Hall-effect detection of current-induced magnetization dynamics in ferrimagnets

Measurements of the transverse Hall resistance are widely used to investigate electron transport, magnetization phenomena, and topological quantum states. Owing to the difficulty of probing transient changes of the transverse resistance, the vast majority of Hall effect experiments are carried out in stationary conditions using either dc or ac. Here we present an approach to perform time-resolved measurements of the transient Hall resistance during current-pulse injection with sub-nanosecond temporal resolution. We apply this technique to investigate in real-time the magnetization reversal caused by spin-orbit torques in ferrimagnetic GdFeCo dots. Single-shot Hall effect measurements show that the current-induced switching of GdFeCo is widely distributed in time and characterized by significant activation delays, which limit the total switching speed despite the high domain-wall velocity typical of ferrimagnets. Our method applies to a broad range of current-induced phenomena and can be combined with non-electrical excitations to perform pump-probe Hall effect measurements.

Efficient spin excitation via ultrafast damping-like torques in antiferromagnets

Damping effects form the core of many emerging concepts for high-speed spintronic applications. Important characteristics such as device switching times and magnetic domain-wall velocities depend critically on the damping rate. While the implications of spin damping for relaxation processes are intensively studied, damping effects during impulsive spin excitations are assumed to be negligible because of the shortness of the excitation process. Herein we show that, unlike in ferromagnets, ultrafast damping plays a crucial role in antiferromagnets because of their strongly elliptical spin precession. In time-resolved measurements, we find that ultrafast damping results in an immediate spin canting along the short precession axis. The interplay between antiferromagnetic exchange and magnetic anisotropy amplifies this canting by several orders of magnitude towards large-amplitude modulations of the antiferromagnetic order parameter. This leverage effect discloses a highly efficient route towards the ultrafast manipulation of magnetism in antiferromagnetic spintronics.

Spin damping plays a fundamental role in many areas of spintronics, however, it is typically assumed that damping is irrelevant for impulsive spin excitations due to their rapid timescale. Here, the authors demonstrate that damping leads to a large and immediate spin-canting in anti-ferromagnets.

Immunity of nanoscale magnetic tunnel junctions with perpendicular magnetic anisotropy to ionizing radiation

Spin transfer torque magnetic random access memory (STT-MRAM) is a promising candidate for next generation memory as it is non-volatile, fast, and has unlimited endurance. Another important aspect of STT-MRAM is that its core component, the nanoscale magnetic tunneling junction (MTJ), is thought to be radiation hard, making it attractive for space and nuclear technology applications. However, studies on the effects of ionizing radiation on the STT-MRAM writing process are lacking for MTJs with perpendicular magnetic anisotropy (pMTJs) required for scalable applications. Particularly, the question of the impact of extreme total ionizing dose on perpendicular magnetic anisotropy, which plays a crucial role on thermal stability and critical writing current, remains open. Here we report measurements of the impact of high doses of gamma and neutron radiation on nanoscale pMTJs used in STT-MRAM. We characterize the tunneling magnetoresistance, the magnetic field switching, and the current-induced switching before and after irradiation. Our results demonstrate that all these key properties of nanoscale MTJs relevant to STT-MRAM applications are robust against ionizing radiation. Additionally, we perform experiments on thermally driven stochastic switching in the gamma ray environment. These results indicate that nanoscale MTJs are promising building blocks for radiation-hard non-von Neumann computing.

Single-shot dynamics of spin-orbit torque and spin transfer torque switching in three-terminal magnetic tunnel junctions

Current-induced spin-transfer torques (STT) and spin-orbit torques (SOT) enable the electrical switching of magnetic tunnel junctions (MTJs) in non-volatile magnetic random access memories. To develop faster memory devices, an improvement in the timescales that underlie the current-driven magnetization dynamics is required. Here we report all-electrical time-resolved measurements of magnetization reversal driven by SOT in a three-terminal MTJ device. Single-shot measurements of the MTJ resistance during current injection reveal that SOT switching involves a stochastic two-step process that consists of a domain nucleation time and propagation time, which have different genesis, timescales and statistical distributions compared to STT switching. We further show that the combination of SOT, STT and the voltage control of magnetic anisotropy leads to reproducible subnanosecond switching with the spread of the cumulative switching time smaller than 0.2 ns. Our measurements unravel the combined impact of SOT, STT and the voltage control of magnetic anisotropy in determining the switching speed and efficiency of MTJ devices.

A cryogenic spin-torque memory element with precessional magnetization dynamics

We present a study of precessional magnetization switching in orthogonal spin-torque spin-valve devices at low temperatures. The samples consist of a spin-polarizing layer that is magnetized out-of-the film plane and an in-plane magnetized free and reference magnetic layer separated by non-magnetic metallic layers. We find coherent oscillations in the switching probability, characterized by high speed switching (~200 ps), error rates as low as 10^-5 and decoherence effects at longer timescales (~1 ns). Our study, which is conducted over a wide range of parameter space (pulse amplitude and duration) with deep statistics, demonstrates that the switching dynamics are likely dominated by the action of the out-of-plane spin polarization, in contrast to in-plane spin-torque from the reference layer, as has been the case in most previous studies. Our results demonstrate that precessional spin-torque devices are well suited to a cryogenic environment, while at room temperature they have so far not exhibited coherent or reliable switching.

Resonant magneto-acoustic switching: influence of Rayleigh wave frequency and wavevector

We show on in-plane magnetized thin films that magnetization can be switched efficiently by 180 degrees using large amplitude Rayleigh waves travelling along the hard or easy magnetic axis. Large characteristic filament-like domains are formed in the latter case. Micromagnetic simulations clearly confirm that this multi-domain configuration is compatible with a resonant precessional mechanism. The reversed domains are in both geometries several hundreds of [Formula: see text], much larger than has been shown using spin transfer torque- or field-driven precessional switching. We show that surface acoustic waves can travel at least 1 mm before addressing a given area, and can interfere to create magnetic stripes that can be positioned with a sub-micronic precision.

Generation and annihilation time of magnetic droplet solitons

Magnetic droplet solitons were first predicted to occur in materials with uniaxial magnetic anisotropy due to a long-range attractive interaction between elementary magnetic excitations, magnons. A non-equilibrium magnon population provided by a spin-polarized current in nanocontacts enables their creation and there is now clear experimental evidence for their formation, including direct images obtained with scanning x-ray transmission microscopy. Interest in magnetic droplets is associated with their unique magnetic dynamics that can lead to new types of high frequency nanometer scale oscillators of interest for information processing, including in neuromorphic computing. However, there are no direct measurements of the time required to nucleate droplet solitons or their lifetime-experiments to date only probe their steady-state characteristics, their response to dc spin-currents. Here we determine the timescales for droplet annihilation and generation using current pulses. Annihilation occurs in a few nanoseconds while generation can take several nanoseconds to a microsecond depending on the pulse amplitude. Micromagnetic simulations show that there is an incubation time for droplet generation that depends sensitively on the initial magnetic state of the nanocontact. An understanding of these processes is essential to utilizing the unique characteristics of magnetic droplet solitons oscillators, including their high frequency, tunable and hysteretic response.

Spatially and time-resolved magnetization dynamics driven by spin-orbit torques

Current-induced spin-orbit torques are one of the most effective ways to manipulate the magnetization in spintronic devices, and hold promise for fast switching applications in non-volatile memory and logic units. Here, we report the direct observation of spin-orbit-torque-driven magnetization dynamics in Pt/Co/AlO_x dots during current pulse injection. Time-resolved X-ray images with 25 nm spatial and 100 ps temporal resolution reveal that switching is achieved within the duration of a subnanosecond current pulse by the fast nucleation of an inverted domain at the edge of the dot and propagation of a tilted domain wall across the dot. The nucleation point is deterministic and alternates between the four dot quadrants depending on the sign of the magnetization, current and external field. Our measurements reveal how the magnetic symmetry is broken by the concerted action of the damping-like and field-like spin-orbit torques and the Dzyaloshinskii-Moriya interaction, and show that reproducible switching events can be obtained for over 10¹² reversal cycles.

Anomalous spin-orbit torque switching due to field-like torque-assisted domain wall reflection

Spin-orbit torques (SOTs) allow the electrical control of magnetic states. Current-induced SOT switching of the perpendicular magnetization is of particular technological importance. The SOT consists of damping-like and field-like torques, and understanding the combined effects of these two torque components is required for efficient SOT switching. Previous quasi-static measurements have reported an increased switching probability with the width of current pulses, as predicted considering the damping-like torque alone. We report a decreased switching probability at longer pulse widths, based on time-resolved measurements. Micromagnetic analysis reveals that this anomalous SOT switching results from domain wall reflections at sample edges. The domain wall reflection was found to strongly depend on the field-like torque and its relative sign to the damping-like torque. Our result demonstrates a key role of the field-like torque in deterministic SOT switching and the importance of the sign correlation of the two torque components, which may shed light on the SOT switching mechanism.

Spintronic Nanodevices for Bioinspired Computing

Bioinspired hardware holds the promise of low-energy, intelligent, and highly adaptable computing systems. Applications span from automatic classification for big data management, through unmanned vehicle control, to control for biomedical prosthesis. However, one of the major challenges of fabricating bioinspired hardware is building ultra-high-density networks out of complex processing units interlinked by tunable connections. Nanometer-scale devices exploiting spin electronics (or spintronics) can be a key technology in this context. In particular, magnetic tunnel junctions (MTJs) are well suited for this purpose because of their multiple tunable functionalities. One such functionality, non-volatile memory, can provide massive embedded memory in unconventional circuits, thus escaping the von-Neumann bottleneck arising when memory and processors are located separately. Other features of spintronic devices that could be beneficial for bioinspired computing include tunable fast nonlinear dynamics, controlled stochasticity, and the ability of single devices to change functions in different operating conditions. Large networks of interacting spintronic nanodevices can have their interactions tuned to induce complex dynamics such as synchronization, chaos, soliton diffusion, phase transitions, criticality, and convergence to multiple metastable states. A number of groups have recently proposed bioinspired architectures that include one or several types of spintronic nanodevices. In this paper, we show how spintronics can be used for bioinspired computing. We review the different approaches that have been proposed, the recent advances in this direction, and the challenges toward fully integrated spintronics complementary metal-oxide-semiconductor (CMOS) bioinspired hardware.

Fast Magnetic Domain-Wall Motion in a Ring-Shaped Nanowire Driven by a Voltage

Magnetic domain-wall motion driven by a voltage dissipates much less heat than by a current, but none of the existing reports have achieved speeds exceeding 100 m/s. Here phase-field and finite-element simulations were combined to study the dynamics of strain-mediated voltage-driven magnetic domain-wall motion in curved nanowires. Using a ring-shaped, rough-edged magnetic nanowire on top of a piezoelectric disk, we demonstrate a fast voltage-driven magnetic domain-wall motion with average velocity up to 550 m/s, which is comparable to current-driven wall velocity. An analytical theory is derived to describe the strain dependence of average magnetic domain-wall velocity. Moreover, one 180° domain-wall cycle around the ring dissipates an ultrasmall amount of heat, as small as 0.2 fJ, approximately 3 orders of magnitude smaller than those in current-driven cases. These findings suggest a new route toward developing high-speed, low-power-dissipation domain-wall spintronics.

Spin-transfer torque magnetic memory as a stochastic memristive synapse for neuromorphic systems

Spin-transfer torque magnetic memory (STT-MRAM) is currently under intense academic and industrial development, since it features non-volatility, high write and read speed and high endurance. In this work, we show that when used in a non-conventional regime, it can additionally act as a stochastic memristive device, appropriate to implement a "synaptic" function. We introduce basic concepts relating to spin-transfer torque magnetic tunnel junction (STT-MTJ, the STT-MRAM cell) behavior and its possible use to implement learning-capable synapses. Three programming regimes (low, intermediate and high current) are identified and compared. System-level simulations on a task of vehicle counting highlight the potential of the technology for learning systems. Monte Carlo simulations show its robustness to device variations. The simulations also allow comparing system operation when the different programming regimes of STT-MTJs are used. In comparison to the high and low current regimes, the intermediate current regime allows minimization of energy consumption, while retaining a high robustness to device variations. These results open the way for unexplored applications of STT-MTJs in robust, low power, cognitive-type systems.

Plasticity in memristive devices for spiking neural networks

Memristive devices present a new device technology allowing for the realization of compact non-volatile memories. Some of them are already in the process of industrialization. Additionally, they exhibit complex multilevel and plastic behaviors, which make them good candidates for the implementation of artificial synapses in neuromorphic engineering. However, memristive effects rely on diverse physical mechanisms, and their plastic behaviors differ strongly from one technology to another. Here, we present measurements performed on different memristive devices and the opportunities that they provide. We show that they can be used to implement different learning rules whose properties emerge directly from device physics: real time or accelerated operation, deterministic or stochastic behavior, long term or short term plasticity. We then discuss how such devices might be integrated into a complete architecture. These results highlight that there is no unique way to exploit memristive devices in neuromorphic systems. Understanding and embracing device physics is the key for their optimal use.

High domain wall velocities via spin transfer torque using vertical current injection

Domain walls, nanoscale transition regions separating oppositely oriented ferromagnetic domains, have significant promise for use in spintronic devices for data storage and memristive applications. The state of these devices is related to the wall position and thus rapid operation will require a controllable onset of domain wall motion and high speed wall displacement. These processes are traditionally driven by spin transfer torque due to lateral injection of spin polarized current through a ferromagnetic nanostrip. However, this geometry is often hampered by low maximum wall velocities and/or a need for prohibitively high current densities. Here, using time-resolved magnetotransport measurements, we show that vertical injection of spin currents through a magnetic tunnel junction can drive domain walls over hundreds of nanometers at ~500 m/s using current densities on the order of 6 MA/cm(2). Moreover, these measurements provide information about the stochastic and deterministic aspects of current driven domain wall mediated switching.

Current-induced torques in magnetic materials

The magnetization of a magnetic material can be reversed by using electric currents that transport spin angular momentum. In the reciprocal process a changing magnetization orientation produces currents that transport spin angular momentum. Understanding how these processes occur reveals the intricate connection between magnetization and spin transport, and can transform technologies that generate, store or process information via the magnetization direction. Here we explain how currents can generate torques that affect the magnetic orientation and the reciprocal effect in a wide variety of magnetic materials and structures. We also discuss recent state-of-the-art demonstrations of current-induced torque devices that show great promise for enhancing the functionality of semiconductor devices.

Spin-transfer torque in nanoscale magnetic devices

We discuss recent highlights from research at Cornell University, Ithaca, New York, regarding the use of spin-transfer torques to control magnetic moments in nanoscale ferromagnetic devices. We highlight progress on reducing the critical currents necessary to produce spin-torque-driven magnetic switching, quantitative measurements of the magnitude and direction of the spin torque in magnetic tunnel junctions, and single-shot measurements of the magnetic dynamics generated during thermally assisted spin-torque switching.

A compact model for magnetic tunnel junction (MTJ) switched by thermally assisted Spin transfer torque (TAS + STT)

Thermally assisted spin transfer torque [TAS + STT] is a new switching approach for magnetic tunnel junction [MTJ] nanopillars that represents the best trade-off between data reliability, power efficiency and density. In this paper, we present a compact model for MTJ switched by this approach, which integrates a number of physical models such as temperature evaluation and STT dynamic switching models. Many experimental parameters are included directly to improve the simulation accuracy. It is programmed in the Verilog-A language and compatible with the standard IC CAD tools, providing an easy parameter configuration interface and allowing high-speed co-simulation of hybrid MTJ/CMOS circuits.

Nonadiabatic stochastic resonance of a nanomagnet excited by spin torque

We report microwave-frequency magnetization dynamics coexcited by alternating spin torque and thermal fluctuations. In these dynamics, temperature strongly enhances the amplitude of magnetization precession and enables excitation of nonlinear dynamic states of magnetization by weak alternating spin torque. We explain these thermally-activated dynamics in terms of nonadiabatic stochastic resonance of magnetization driven by spin torque.

Single-shot time-domain studies of spin-torque-driven switching in magnetic tunnel junctions

We report single-shot measurements of resistance versus time for thermally assisted spin-torque switching in magnetic tunnel junctions. We achieve the sensitivity to resolve the magnetic dynamics prior to as well as during switching, yielding detailed views of switching modes and variations between events. Analyses of individual traces allow measurements of coherence times, nonequilibrium excitation spectra, and variations in magnetization precession amplitude. We find that with a small in-plane hard-axis magnetic field the switching dynamics are more spatially coherent than for a zero field.

Biased quasiballistic spin torque magnetization reversal

We explore the ultrafast limit of spin torque magnetization reversal time. Spin torque precession during a spin torque current pulse and free magnetization ringing after the pulse is detected by time-resolved magnetotransport. Adapting the duration of the pulse to the precession period allows coherent control of the final orientation of the magnetization. In the presence of a hard axis bias field, we find optimum quasiballistic spin torque magnetization reversal by a single precessional turn directly from the initial to the reversed equilibrium state.