Search

Neuromorphic computing using spintronic devices, such as spin-torque oscillators (STOs), has been intensively studied for energy-efficient data processing. One of the critical issues in this application is stochasticity in magnetization dynamics, which limits the accuracy of computation. Such stochastic behavior, however, plays a key role in stochastic computing and machine learning. It is therefore important to develop methods for both suppressing and enhancing stochastic response in spintronic devices. We report on experimental investigations on control of stochastic quantity, such as the width of a distribution of transient time in magnetization dynamics in vortex-type STO. The spin-transfer effect can suppress stochasticity in transient dynamics from a non-oscillating to oscillating state, whereas an application of a radio-frequency magnetic field is effective in reducing stochasticity on the time evolution of the oscillating state.

Dzyaloshinskii–Moriya interaction (DMI) is considered as one of the most important energies for specific chiral textures such as magnetic skyrmions. The keys of generating DMI are the absence of structural inversion symmetry and exchange energy with spin–orbit coupling. Therefore, a vast majority of research activities about DMI are mainly limited to heavy metal/ferromagnet bilayer systems, only focusing on their interfaces. Here, we report an asymmetric band formation in a superlattices (SL) which arises from inversion symmetry breaking in stacking order of atomic layers, implying the role of bulk-like contribution. Such bulk DMI is more than 300% larger than simple sum of interfacial contribution. Moreover, the asymmetric band is largely affected by strong spin–orbit coupling, showing crucial role of a heavy metal even in the non-interfacial origin of DMI. Our work provides more degrees of freedom to design chiral magnets for spintronics applications.

Fully-epitaxial magnetoresistance (MR) devices with half-metallic Heusler alloys has attracted considerable attention for years due to their excellent spin-dependent transport properties such as high MR ratio and ultra-low resistance-area product. However, their poor manufacturability due to the necessity of epitaxial growth on a special single crystalline substrate such as MgO(001) hinders their practical applications. To overcome this issue, in this study, we fabricated Heusler-based epitaxial current-perpendicular-to-plane giant magnetoresistance (CPP-GMR) film grown on Si substrate and directly bonded it to poly-crystalline electrode wafer by using three-dimensional (3D) integration processes such as direct wafer bonding and removal of backside Si substrate. First we explored suitable seed/buffer layers for the (001)-oriented epitaxial growth of Heusler CPP-GMR on Si(001) substrate. Si-subs/NiAl/CoFe seed/buffer structure was found to induce (001)-oriented epitaxial growth and to suppress an Al diffusion from NiAl to the Heusler electrode even after annealing at 500 °C, resulting in large MR ratio comparable to that with MgO substrate. After direct wafer bonding process, the microstructure analysis revealed clean damage-free bonded interface between the epitaxial Heusler GMR and poly-crystalline electrode films with Au capping layers. After the 3D integration processes, high MR performance has been successfully reproduced. This unique processing method enables the integration of high performance Heusler-based epitaxial spintronic devices for various practical applications.

We study the dynamic switching properties of a nanomagnet under microwave electric field pumping. The periodic modulation of an anisotropy field induced by microwave electric field pumping efficiently excites the uniform magnetization oscillation, allowing for precise control of magnetization switching. Accurate shaping of the pumping voltage waveform also enables us to investigate the transient reaction of magnetization to the relative phase difference of the pumping signal. We demonstrate both experimentally and theoretically the existence of a dead angle in which the uniform oscillation of magnetization is inhibited even though the microwave frequency itself satisfies the conditions of parametric excitation. Our results provide an energy-efficient way of manipulating ultrafast magnetization dynamics in nanomagnetic devices.

The reservoir computing neural network architecture is widely used to test hardware systems for neuromorphic computing. One of the preferred tasks for bench-marking such devices is automatic speech recognition. However, this task requires acoustic transformations from sound waveforms with varying amplitudes to frequency domain maps that can be seen as feature extraction techniques. Depending on the conversion method, these may obscure the contribution of the neuromorphic hardware to the overall speech recognition performance. Here, we quantify and separate the contributions of the acoustic transformations and the neuromorphic hardware to the speech recognition success rate. We show that the non-linearity in the acoustic transformation plays a critical role in feature extraction. We compute the gain in word success rate provided by a reservoir computing device compared to the acoustic transformation only, and show that it is an appropriate benchmark for comparing different hardware. Finally, we experimentally and numerically quantify the impact of the different acoustic transformations for neuromorphic hardware based on magnetic nano-oscillators., 13 pages, 5 figures

Perpendicular magnetic anisotropy (PMA) is becoming increasingly important in spintronics research, especially for high-density magnetoresistive random access memories (MRAMs). The PMA induced at an Fe/MgO interface is widely used in magnetic tunnel junctions. Here, we propose inserting an ultrathin LiF layer at the interface in an epitaxial Fe/MgO junction. With a 0.3 nm-thick LiF layer, a large intrinsic interface PMA energy, Ki,0, of 2.8 mJ/m2 was achieved. We also found that the LiF/MgO bilayer tunneling barrier exhibited a large tunnel magnetoresistance (TMR) effect, suggesting that a coherent spin-dependent tunneling process was maintained in the ultrathin LiF layer. Atomic-scale interface engineering using fluoride can further improve the PMA and TMR properties of spintronic devices.

We report on a systematic investigation of current-induced domain wall motion in TbFeCo ferrimagnetic thin films with a Pt underlayer. The Tb concentration of the alloy is varied to study the effect of the magnetization compensation on the current-induced motion of domain walls. We find that the direction in which domain walls move changes when the Tb concentration crosses the magnetization compensation point. Interestingly, the domain walls move along (against) the current flow for FeCo-rich (deficient) films. When the film composition is slightly Tb-rich than the magnetic compensation point, the domain wall moves along the current at a lower current but reverses its direction when the current is increased. These results suggest that two competing torques act on the domain walls in the Tb-based ferrimagnets (TbFeCo), causing the domain wall velocity to be more than one order of magnitude smaller than that of Gd-based ferrimagnets.

Modulation of a spin-torque oscillator (STO) signal based on a magnetic tunnel junction (MTJ) with perpendicularly magnetized free layer is investigated. Magnetic field inductive loop was created during MTJ fabrication process, which enables microwave field application during STO operation. The frequency modulation by the microwave magnetic field of up to 3 GHz is explored, showing a potential for application in high-data-rate communication technologies. Moreover, an inductive loop is used for self-synchronization of the STO signal, which after field-locking exhibits significant improvement of the linewidth and oscillation power., Comment: 4 pages, 5 figures

Improving the perpendicular magnetic anisotropy (PMA) and voltage-controlled magnetic anisotropy (VCMA) properties are fundamentally important for the development of voltage-controlled magnetoresistive random access memories (VC-MRAM). Recently, we reported on a large increase in PMA at an Fe/MgO interface brought about by inserting an ultrathin LiF layer at the interface. In this paper, we investigate the PMA, VCMA, and TMR properties in MTJs with an Ir-doped ultrathin ferromagnetic layer and a LiF/MgO hybrid tunneling barrier. We observed a clear increase in the interfacial PMA by a factor of 2.5 when an ultrathin 0.25 nm LiF layer was inserted. A large VCMA coefficient, exceeding −300 fJ/Vm, was also achieved while maintaining the high TMR ratio and high interfacial PMA. These results demonstrate the high potential of interface engineering using ultrathin LiF layers for spintronic devices.

To develop a high-quality Weyl magnet ${\mathrm{Mn}}_{3}\mathrm{Sn}$ thin film, polycrystalline Mn-Sn was sputter deposited onto an fcc or hexagonal close-packed metallic seed layer of Pt, Cu, Ir, Ru, Cr, or Al. Among the tested seed materials, Pt and Cu enabled fabrication of highly $c$-plane oriented textured $D{0}_{19}$-type ${\mathrm{Mn}}_{3}\mathrm{Sn}$ films with a smooth surface (mean roughness below 0.3 nm) after postannealing at $350{\phantom{\rule{0.16em}{0ex}}}^{\ensuremath{\circ}}\mathrm{C}$ or below. The ${\mathrm{Mn}}_{3}\mathrm{Sn}$ films exhibited substantial signals of the anomalous Nernst effect (ANE), which is one characteristic of a Weyl magnet. Microstructural analysis revealed that the Cu seed layer significantly diffused into the ${\mathrm{Mn}}_{3}\mathrm{Sn}$ layer during postannealing, while the Pt seed layer changed little. The results suggested that a small amount of Cu mixture (up to 2.5 at. %) did not degrade the ANE signal but effectively lowered the postannealing temperature to form the $D{0}_{19}$ structure. The developed films and techniques are beneficial for antiferromagnetic spintronics in terms of the smoothness and process consistency in a standard process for magnetic random-access memory.

By utilizing Ta/Mo layered adhesion structures, thermally robust perpendicular magnetic anisotropy and voltage-controlled magnetic anisotropy (VCMA) effects were achieved in magnetic tunnel junctions (MTJs) with ultrathin CoFeB films grown on MgO. After annealing at 400 °C, MTJs with Ta/Mo layered adhesion exhibited VCMA coefficients of 48 fJ/Vm. The combination of Ta and Mo improved the crystalline orientation and flatness of the CoFeB/MgO tunneling barrier interfaces, as determined by cross-sectional scanning transmission electron microscopy. Additionally, we demonstrate that the small interdiffusion between Mo and CoFe enables effective scavenging of B from CoFeB by increasing the thickness of the B sink layer without impairing the device performance due to atomic diffusion.

$D{0}_{3}$-ordered ${\mathrm{Fe}}_{3}\mathrm{Ga}$ has been recently reported to show a giant anomalous Nernst effect due to a large Berry curvature along the interconnected nodal lines concentrated near the Fermi level. This peculiar band structure may be advantageous for spintronics device applications because one can expect a strong response to a gate-voltage application. This study evaluates the Gilbert damping constants, a key parameter governing the magnetization dynamics of epitaxial thin films of $D{0}_{3}\text{\ensuremath{-}}{\mathrm{Fe}}_{3}\mathrm{Ga}$ and bcc-Fe by ferromagnetic resonance measurements. We find that the Gilbert damping constant of $D{0}_{3}\text{\ensuremath{-}}{\mathrm{Fe}}_{3}\mathrm{Ga} [(6.0\ifmmode\pm\else\textpm\fi{}0.2)\ifmmode\times\else\texttimes\fi{}{10}^{\ensuremath{-}3}]$ is relatively low and comparable to that of Fe $[(2.3\ifmmode\pm\else\textpm\fi{}0.2)\ifmmode\times\else\texttimes\fi{}{10}^{\ensuremath{-}3}]$. The low Gilbert damping suggests that the interconnected nodal lines near the Fermi level do not hinder magnetization dynamics, making $D{0}_{3}\text{\ensuremath{-}}{\mathrm{Fe}}_{3}\mathrm{Ga}$ even more attractive as a building block of spintronics devices.

We study the perpendicular magnetic anisotropy (PMA) and the voltage-controlled magnetic anisotropy (VCMA) effect in MgO/Ta/CoFeB/MgO junctions. A few monolayers (3–5 A) of Ta acts as an adhesive layer, allowing for the formation of flat and ultrathin ( ∼ 10 A) CoFeB films with large PMA on top of MgO. In the MgO/Ta/CoFeB/MgO junctions, the diffusion of Ta to the upper CoFeB/MgO interface is effectively eliminated, and superior thermal stability up to 400 ∘ C is demonstrated. Detailed nanostructural analyses reveal the appearance of Co–Fe compositional variation in the CoFeB layer associated with the insertion of Ta layer, which explains the observed changes in PMA and VCMA. Our results will facilitate the development of VCMA-based memory devices and also pave the way toward controlling the magnetic and electric properties of CoFeB thin films through subnanometer-scale compositional modulation.

Present artificial intelligence algorithms require extensive computations to emulate the behavior of large neural networks, operating current computers near their limits, which leads to high energy costs. A possible solution to this problem is the development of new computing architectures, with nanoscale hardware components that use their physical properties to emulate the behavior of neurons. In spite of multiple theoretical proposals, there have been only a limited number of experimental demonstrations of brain-inspired computing with nanoscale neurons. Here we describe such demonstrations using nanoscale spin-torque oscillators, which exhibit key features of neurons, in a reservoir computing approach. This approach offers an interesting platform to test these components, because a single component can emulate a whole neural network. Using this method, we classify sine and square waveforms perfectly and achieve spoken-digit recognition with state of the art results. We illustrate optimization of the oscillator’s operating regime with sine/square classification.

We study the perpendicular magnetic anisotropy (PMA) in the MgO/CoFeB (CFB)/MgO junctions with an angstrom-thick Mo spacer layer separating the CFB layer. Perpendicularly magnetized CFB/Mo/CFB films are achieved for a wide range of CFB thicknesses, and a large PMA energy density of > 0.3 mJ m−2 is demonstrated by tuning the thickness ratio of the two CFB layers as well as the thickness of the Mo spacer layer. The PMA in the MgO/CFB/Mo/CFB/MgO is controlled by a voltage applied across the junction, and a sign inversion in the voltage-controlled magnetic anisotropy effect is clearly observed between the ‘top free’ and ‘bottom free’ magnetic tunnel junctions, in which the CFB/Mo/CFB layers are fabricated on top and bottom of an MgO barrier layer, respectively. Nanostructural analyses reveal the difference in the morphology of the top free and bottom free magnetic tunnel junctions and also suggest that the flatness of CFB/MgO interface is rather important for improving the efficiency of the voltage-controlled magnetic anisotropy.

Dzyaloshinskii-Moriya interaction (DMI) is considered as one of the most important energy for specific chiral texture such as magnetic skyrmions. The key of generating DMI is absence of structural inversion symmetry and exchange energy with spin-orbit coupling. Therefore, a vast majority of researches about DMI is mainly limited to heavy metal/ferromagnet bilayer systems, only focusing on their interfaces. Here, we report that asymmetric band formation in an artificial superlattice arises from inversion symmetry breaking in stacking order of atomic layers, resulting in bulk DMI. Such bulk DMI is more than 300% larger than simple sum of interfacial contribution. Moreover, the asymmetric band is largely affected by strong spin-orbit coupling, showing crucial role of a heavy metal even in the non-interfacial origin of DMI. Such Rashba superlattices can be a new class of material design for spintronics applications.

We study the influence of inverse biases applied before and after a write pulse on the write-error rate (WER) of voltage-driven magnetization switching in magnetic tunnel junctions. It is demonstrated that the inverse bias applied after the write pulse is effective not only for improving the minimum value of the WER but also for broadening the operating pulse-width window. In contrast, an inverse bias applied before the write pulse hardly improves the WER or even narrows the operating window. Numerical simulations based on the macrospin approximation reproduce these experimental results well and a detailed analysis of the magnetization trajectory reveals that the observed write-error increase is due to a change in the magnetization angle before writing. These experimental results, as well as the numerical analysis, should facilitate the development of writing schemes in voltage-driven magnetic random-access memories.

The brain naturally binds events from different sources in unique concepts. It is hypothesized that this process occurs through the transient mutual synchronization of neurons located in different regions of the brain when the stimulus is presented. This mechanism of ‘binding through synchronization’ can be directly implemented in neural networks composed of coupled oscillators. To do so, the oscillators must be able to mutually synchronize for the range of inputs corresponding to a single class, and otherwise remain desynchronized. Here we show that the outstanding ability of spintronic nano-oscillators to mutually synchronize and the possibility to precisely control the occurrence of mutual synchronization by tuning the oscillator frequencies over wide ranges allows pattern recognition. We demonstrate experimentally on a simple task that three spintronic nano-oscillators can bind consecutive events and thus recognize and distinguish temporal sequences. This work is a step forward in the construction of neural networks that exploit the non-linear dynamic properties of their components to perform brain-inspired computations.

The correlation of phase fluctuations in any type of oscillator fundamentally defines its spectral shape. However, in nonlinear oscillators, such as spin torque nano-oscillators, the frequency spectrum can become particularly complex. This is specifically true when not only considering thermal but also colored 1/f flicker noise processes, which are crucial in the context of the oscillator’s long term stability. In this study, we address the frequency spectrum of spin torque oscillators in the regime of large-amplitude steady oscillations experimentally and as well theoretically. We particularly take both thermal and flicker noise into account. We perform a series of measurements of the phase noise and the spectrum on spin torque vortex oscillators, notably varying the measurement time duration. Furthermore, we develop the modelling of thermal and flicker noise in Thiele equation based simulations. We also derive the complete phase variance in the framework of the nonlinear auto-oscillator theory and deduce the actual frequency spectrum. We investigate its dependence on the measurement time duration and compare with the experimental results. Long term stability is important in several of the recent applicative developments of spin torque oscillators. This study brings some insights on how to better address this issue.

Electrical manipulation of emergent phenomena due to nontrivial band topology is a key to realize next-generation technology using topological protection. A Weyl semimetal is a three-dimensional gapless system that hosts Weyl fermions as low-energy quasiparticles. It exhibits various exotic phenomena such as large anomalous Hall effect (AHE) and chiral anomaly, which have robust properties due to the topologically protected Weyl nodes. To manipulate such phenomena, the magnetic version of Weyl semimetals would be useful as a magnetic texture may provide a handle for controlling the locations of Weyl nodes in the Brillouin zone. Moreover, given the prospects of antiferromagnetic (AF) spintronics for realizing high-density devices with ultrafast operation, it would be ideal if one could electrically manipulate an AF Weyl metal. However, no report has appeared on the electrical manipulation of a Weyl metal. Here we demonstrate the electrical switching of a topological AF state and its detection by AHE at room temperature. In particular, we employ a polycrystalline thin film of the AF Weyl metal Mn$_3$Sn, which exhibits zero-field AHE. Using the bilayer device of Mn$_3$Sn and nonmagnetic metals (NMs), we find that an electrical current density of $\sim 10^{10}$-$10^{11}$ A/m$^2$ in NMs induces the magnetic switching with a large change in Hall voltage, and besides, the current polarity along a bias field and the sign of the spin Hall angle $\theta_{\rm SH}$ of NMs [Pt ($\theta_{\rm SH} > 0$), Cu($\theta_{\rm SH} \sim 0$), W ($\theta_{\rm SH} < 0$)] determines the sign of the Hall voltage. Notably, the electrical switching in the antiferromagnet is made using the same protocol as the one used for ferromagnetic metals. Our observation may well lead to another leap in science and technology for topological magnetism and AF spintronics., Comment: 49 pages, 14 figures

Synchronized nonlinear oscillators networks are at the core of numerous families of applications including phased array wave generators and neuromorphic pattern matching systems. In these devices, stable synchronization between large numbers of nanoscale oscillators is a key issue that remains to be demonstrated. Here, we show experimentally that synchronized spin-torque oscillator networks can be scaled up. By increasing the number of synchronized oscillators up to eight, we obtain that the emitted power and the quality factor increase linearly with the number of oscillators. Even more importantly, we demonstrate that the stability of synchronization in time exceeds 1.6 milliseconds corresponding to 105 periods of oscillation. Our study demonstrates that spin-torque oscillators are suitable for applications based on synchronized networks of oscillators.

The anomalous Hall effect in ferromagnets is a well-known phenomenon in which electromotive force is generated in the direction perpendicular to both the electric current and the magnetization. With nonmagnetic heavy metals, the generation of spin current by the spin Hall effect, as well as excitation of magnetization dynamics in an adjacent ferromagnet by spin-transfer torque, has been reported. However, the generation of a spin current in ferromagnets by the anomalous Hall effect has so far not been measured. Here we report the observation of spin current generation by the anomalous Hall effect through spin-transfer torque excitation in CoFeB/Cu/NiFe trilayer films. In this spin anomalous Hall effect, a spin current generated in the CoFeB by the anomalous Hall effect is injected into the NiFe and excites spin-transfer torque in the NiFe. This leads to a measurable modulation in the ferromagnetic resonance linewidth of the NiFe. The spin anomalous Hall angle in the CoFeB is estimated to be −0.14 ± 0.05, which is comparable to the spin Hall angle reported in heavy metals. The generation of a spin current by the anomalous Hall effect is observed by measuring the modulation in the ferromagnetic resonance linewidth of CoFeB/Cu/NiFe trilayer films.

International audience; The recent demonstration of neuromorphic computing with spin-torque nano-oscillators has opened a path to energy efficient data processing. The success of this demonstration hinged on the intrinsic short-term memory of the oscillators. In this study, we extend the memory of the spin-torque nano-oscillators through time-delayed feedback. We leverage this extrinsic memory to increase the efficiency of solving pattern recognition tasks that require memory to discriminate different inputs. The large tunability of these non-linear oscillators allows us to control and optimize the delayed feedback memory using different operating conditions of applied current and magnetic field.

We have developed a random-number-generator (RNG) named “spin dice,” which employs the stochastic nature of spin-torque switching (STS) in a magnetic tunnel junction. The principle of the idea is that the switching probability first tuned around 0.5 is varied linearly with the applied current. After that, the switching results are converted into binary random numbers. We fabricated several types of “spin dice” by combining magnetic tunnel junctions and single-board microcomputer, and achieved generation speed of random numbers up to several hundred kbit/sec. Because STS is scalable and magnetic tunnel junctions have compatibility to semiconductor fabrication process, “spin dice” can be considered as a promising candidate for truly random-number-generator (TRNG) for security applications.

International audience; Fabricating powerful neuromorphic chips the size of a thumb requires miniaturizing their basic units: synapses and neurons. The challenge for neurons is to scale them down to submicrometer diameters while maintaining the properties that allow for reliable information processing: high signal to noise ratio, endurance, stability, reproducibility. In this work, we show that compact spin-torque nano-oscillators can naturally implement such neurons, and quantify their ability to realize an actual cognitive task. In particular, we show that they can naturally implement reservoir computing with high performance and detail the recipes for this capability.