United States

Spin-orbit interaction in metals and its ability to generate spin-current has been the hallmark of spintronics in the last decade. Beyond its fundamental interest as a source of spin-current, the manipulation of magnetization via transfer of the spin angular momentum has proven to be energy efficient for spin-based memory (e.g. SOT-MRAM) and also neuromorphic devices (e.g. using spin Hall nano-oscillators). Therein, charge to spin conversion is believed to be governed essentially by two main mechanisms, spin Hall effect in the bulk of heavy-metals and Rashba effect at interfaces [1], the latter being often considered negligible in all-metallic interfaces. However, in the case of atomically thin metallic layers, the underlying SOT physical mechanisms at play have not been experimentally fully tackled.&lt;br&gt;
In this study, we examine the impact of the insertion of a light element interface on the nature of the SOT as well as its efficiency in terms of damping-like (HDL) and field-like (HFL) effective fields in ultrathin ferromagnets. Importantly, we observe unexpectedly large HFL/HDL ratio (∼2.5) upon inserting a 1.4 nm thin Al layer in Pt|Co|Al|Pt as compared to Cu instead of Al as shown in Fig. 1. From our modeling (dotted lines in Fig.1), these experimental results strongly evidence the presence of a large interfacial Rashba effect at Co|Al interface producing giant HFL. The occurrence of such enhanced torques from an interfacial origin is further validated by reducing the contribution from SHE in the bottom Pt layer as well as by demonstrating current-induced magnetization reversal at reduced current. We believe that our results are important from the application point of view as they provide a clear route for reaching ultimate spin-torque efficiency for the associated devices[2].&lt;br&gt;
**Acknowledgement:** DARPA TEE (MIPR no. HR0011831554), the Horizon2020 Framework Program of the European Commission under FET-Proactive Grant agreement No. 824123 (SKYTOP) and the the Agence Nationale de la Recherche, France, No. ANR-17-CE24-0025 (TOPSKY).&lt;br&gt;
**References:**&lt;br&gt;[1] Manchon, A. et al., Rev. Mod. Phys. 91, (2019).&lt;br&gt;
[2] Krishnia S. et al., arXiv : 2205.08486.&lt;br&gt;

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/d8fd135b8e9157bc0dc4a5f819d2811b.png)&lt;br&gt;
Co thickness (tCo) dependence of (a) DL-SOT field and (b) FL-like SOT fields in Pt8|Co (tCo)|Al1.4|Pt3 (red ) and Pt8|Co (tCo)|Cu1.4|Pt3 (blue ) samples (c) ζ = HFL/HDL as a function of tCo.&lt;br&gt;

MMM 2022

Large interfacial Rashba interaction and giant spin orbit torque in atomically thin metallic multilayers

Spin-orbit interaction in metals and its ability to generate spin-current has been the hallmark of spintronics in the last decade. Beyond its fundamental interest as a source of spin-current, the manipulation of magnetization via transfer of the spin angular momentum has proven to be energy efficient for spin-based memory (e.g. SOT-MRAM) and also neuromorphic devices (e.g. using spin Hall nano-oscillators). Therein, charge to spin conversion is believed to be governed essentially by two main mechanisms, spin Hall effect in the bulk of heavy-metals and Rashba effect at interfaces [1], the latter being often considered negligible in all-metallic interfaces. However, in the case of atomically thin metallic layers, the underlying SOT physical mechanisms at play have not been experimentally fully tackled. 
In this study, we examine the impact of the insertion of a light element interface on the nature of the SOT as well as its efficiency in terms of damping-like (HDL) and field-like (HFL) effective fields in ultrathin ferromagnets. Importantly, we observe unexpectedly large HFL/HDL ratio (∼2.5) upon inserting a 1.4 nm thin Al layer in Pt|Co|Al|Pt as compared to Cu instead of Al as shown in Fig. 1. From our modeling (dotted lines in Fig.1), these experimental results strongly evidence the presence of a large interfacial Rashba effect at Co|Al interface producing giant HFL. The occurrence of such enhanced torques from an interfacial origin is further validated by reducing the contribution from SHE in the bottom Pt layer as well as by demonstrating current-induced magnetization reversal at reduced current. We believe that our results are important from the application point of view as they provide a clear route for reaching ultimate spin-torque efficiency for the associated devices[2]. 
**Acknowledgement:** DARPA TEE (MIPR no. HR0011831554), the Horizon2020 Framework Program of the European Commission under FET-Proactive Grant agreement No. 824123 (SKYTOP) and the the Agence Nationale de la Recherche, France, No. ANR-17-CE24-0025 (TOPSKY). 
**References:** [1] Manchon, A. et al., Rev. Mod. Phys. 91, (2019). 
[2] Krishnia S. et al., arXiv : 2205.08486. 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/d8fd135b8e9157bc0dc4a5f819d2811b.png) 
Co thickness (tCo) dependence of (a) DL-SOT field and (b) FL-like SOT fields in Pt8|Co (tCo)|Al1.4|Pt3 (red ) and Pt8|Co (tCo)|Cu1.4|Pt3 (blue ) samples (c) ζ = HFL/HDL as a function of tCo.

technical paper

#### Welcome to the 67th Annual Conference on Magnetism and Magnetic Materials, MMM 2022!

The Conference was held from October 31 to November 4 and offers both in-person and on-demand content.

MMM 2022 is sponsored jointly by AIP Publishing and the IEEE Magnetics Society. Members of the international scientific and engineering communities interested in recent developments in fundamental and applied magnetism are invited to attend and contribute to the technical sessions. The technical program will include invited and contributed papers in oral and poster sessions, invited symposia, a tutorial, and several special sessions. This Conference provides an outstanding opportunity for worldwide participants to meet their colleagues and collaborators and discuss developments in all areas of magnetism research.

<iframe width="560" height="315" src="https://s3.amazonaws.com/pf-user-files-01/u-59356/uploads/2022-11-02/un13zp7/MMM%20Welcome%20Video%20v5.mp4" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>

For tickets to the event, please register at the following link: https://magnetism.org/registration

Please register for the MMM 2022 conference to join the event.

This Conference provides an outstanding opportunity for worldwide participants to meet their colleagues and collaborators and discuss developments in all areas of magnetism research.

Artificial neural networks (ANNs) are powerful tools for data intensive classification and regression, but the memory wall in von Neumann architectures prevents efficient processing [1]. Though there are many candidates for devices and architectures to realize ANNs, there are no comprehensive solutions to implement hardware Bayesian NNs (BNNs). BNNs are much less susceptible to overfitting when available data is sparse or unreliable, accomplished by representing and learning the weights of the network as distributions instead of deterministic values [2]. We propose a novel spintronic synaptic cell design that can implement tunable distributions by taking advantage of the intrinsic stochasticity of magnetic materials. The design of the cell is composed of a noise and weight encoding element respectively (see Fig. 1). The noisy element is a circular in-plane magnetic tunnel junction (MTJ) that relies on thermal activation to output a random conductance within the range of its TMR. This TMR can then be controlled through several methods such as voltage-controlled magnetic anisotropy (VCMA) [3] or magneto-ionics [4]. Combined with a weight encoding conductance element like a domain wall-magnetic tunnel junction (DW-MTJ) artificial synapse [5,6], we propose a novel cell design that can represent the distributions necessary for inference in a hardware BNN. We demonstrate device functionality through Landau-Lifshitz-Gilbert and micromagnetics [7] models, model analog Bayesian inference in CrossSim [8] for regression (see Fig. 2) and classification to detect out-of-distribution data, and compare the systemâ€™s energy efficiency to CMOS implementations. These results propose a novel synapse implementation for accurate and efficient Bayesian inference, a foundational step toward hardware BNN accelerators. SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525.

Hardware Aware Bayesian Variational Inference Using Magnetic Tunnel Junction Noise Distributions in Artificial Synapse Arrays

We compare thermal-gradient-driven transverse voltages in ferrimagnetic-insulator/heavy-metal bilayers (Tm3Fe5O12/W and Tm3Fe5O12/Pt) to corresponding electrically driven transverse resistances at and above room temperature1. The thermal gradient is created by an applying electric current in the lithographically patterned heater, electrically isolated from the device that is located approximately 50 μm away. We find for Tm3Fe5O12/W that the thermal and electrical effects can be explained by a common spin-current detection mechanism, the physics underlying spin Hall magnetoresistance (SMR)2,3. However, for Tm3Fe5O12/Pt the ratio of the electrically driven transverse voltages (planar Hall signal/anomalous Hall signal) is much larger than the ratio of corresponding thermal-gradient signals, a result which is very different from expectations for a SMR-based mechanism alone. We ascribe this difference to a proximity-induced magnetic layer at the Tm3Fe5O12/Pt interface4. By analyzing the ratio of electrically and thermally driven transverse voltages we find that the dominant contribution of the transverse spin-current is the intrinsic spin-Hall effect that is driven by the longitudinal electrical field either provided by the thermal gradient due to the Seebeck effect or by applying the electric current. 
**References:** 1. Bose, A. et al. Origin of transverse voltages generated by thermal gradients and electric fields in ferrimagnetic-insulator/heavy-metal bilayers. Phys. Rev. B 105, L100408 (2022). 
2. Nakayama, H. et al. Spin Hall Magnetoresistance Induced by a Nonequilibrium Proximity Effect. Phys. Rev. Lett. 110, 206601 (2013). 
3. Meyer, S. et al. Observation of the spin Nernst effect. Nat. Mater. 16, 977–981 (2017). 
4. Huang, S. Y. et al. Transport Magnetic Proximity Effects in Platinum. Phys. Rev. Lett. 109, 107204 (2012). 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/add0712631d46afa26b1f6505752d938.png) 
FIG.1. Results for TmIG/W. Hall resistance of TmIG/W for (a) out-of-plane magnetic-field sweep and (b) in-plane field rotation for a field magnitude of 2.7 kOe. Thermally-induced transverse voltages, (c) Vxy z and (d) Vxy φ, for a heater power of 630 mW and for two different heater spacings, d = 15 μm (closed squares) and 50 μm (open circles).

Origin of transverse voltages generated by thermal gradients and electric fields in ferrimagnetic insulator/heavy

The direct interference of short wavelength spin waves in continuous layers is bringing new opportunities for efficient signal processing [1-2]. Several pionneer works established the possibility to shape the propagation of spin waves in ferromagnetic thin films using concepts borrowed from optics [3-5]. Along these ideas, we demonstrated that the focused emission of spin waves beams in ferromagnetic thin films from constricted coplanar waveguide follows directly the near-field interference pattern of the constriction geometry [6]. In this communication, we investigate the interference patterns of spin waves excited by curvilinear shaped antennas, which produce either a concentrating or a diffusing effect. Namely, we focused our study on two different types of demonstrators: (i) a spin wave â€œconcentratorâ€ as shown in Fig. 1, and (ii) the equivalent of a â€œYoung double-slitâ€ experiment for spin waves shown in Fig. 2. Firstly, we will present a near-field interference model that can be readily implemented for all spin wave modes and any arbitrary shape of antenna [7]. We show the validity and efficiency of our approach by comparing the near-field diffraction patterns obtained with our model and the corresponding MuMax3 micromagnetic simulations for a wide scope of geometries. Secondly, we present series of propagating spin wave spectroscopy done on 30nm YIG films and a 20nm Ni80Fe20 film magnetized out-of-plane, allowing for a discrete mapping of the spin wave intensity with a spatial resolution of about 1Âµm. We show that the transmission signals S21 measured at different locations follow very closely the simulated interference patterns for all the demonstrators, thus paving the way for future prototypes of magnon interferometric devices. This work was supported by the French ANR project â€œMagFuncâ€, and the DÃ©partement du FinistÃ¨re through the project â€œSOSMAGâ€.

Spin Wave Diffraction in Curvilinear Antennas

Magnonics based on the transfer of angular momentum in the form of spin waves provides a promising technology for wave-based information processing [1] and neuromorphic computing [2]. A key challenge in realizing a viable spin-wave technology is the connection of multiple computational units into integrated magnonic circuits, which requires a scalable materials platform for low-loss spin-wave transport. Conventionally, magnonic waveguides are fabricated by patterning ferromagnetic metal or YIG films into narrow conduits. The spin-wave decay length in fully patterned nanoscopic waveguides is typically limited to a few mm. Here, we introduce a new spin-wave waveguiding structure consisting of a continuous YIG film and ferromagnetic metal nanostripes patterned on top [3]. In this hybrid structure, dipolar coupling between the two magnetic materials defines nanoscopic spin-wave transporting channels within the YIG film, along which spin waves propagate with low loss. We demonstrate spin-wave decay lengths of âˆ¼20 mm in 160-nm-wide waveguides at small magnetic bias fields. Moreover, we show that spin waves are efficiently guided through magnetically induced bends in the continuous YIG film (Fig. 1). The combination of low-loss scalable transport and efficient redirection of spin waves in our hybrid waveguiding structures provides a new strategy for the implementation of low-loss magnonic devices and integrated circuits without YIG nanopatterning.

Low Loss Nanoscopic Spin

Complex magnetic materials hosting topologically non-trivial particle-like objects such as skyrmions are intensely researched and could fundamentally change the way we
store and process data. One important class of materials are helimagnetic materials with Dzyaloshinskii-Moriya interaction. Recently, it was demonstrated that nanodisks consisting of two
layers with opposite chirality can host a single stable Bloch point of two different types [1]. The Bloch point represents an interesting topological excitation in a helimagnetic system, which
expands the set of well-known magnetic states such as domain walls, vortices, and skyrmions.
In this work [2], we use micromagnetic simulations [3, 4] to show that FeGe nanostrips consisting of two layers with opposite chirality can host multiple coexisting Bloch points. We
demonstrate that the two different Bloch-point types (Fig. 1) can be geometrically arranged in any arbitrary order, and these magnetization configurations are meta-stable (within certain
constraints on the strip width and length). We can determine an optimal spacing between Bloch points within a line of Bloch points (corresponding to a distance over which a Bloch point
extends). This allows us to predict strip geometries suitable for an arbitrary number of Bloch points, which we verify for an example 80-Bloch-point configuration (Fig. 2). 
**References** [1] M. Beg et al., Scientific Reports, Vol. 9, p. 7959 (2019).
[2] M. Lang et al. arXiv:2203.13689 (2022).
[3] M. Beg, M. Lang and H. Fangohr, IEEE Transactions on Magnetics, Vol. 58, p. 7300205 (2022)
[4] M. Beg, R. A. Pepper and H. Fangohr, AIP Advances, Vol. 7, p. 56025 (2017) 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/fec29aa3f660084584bbbb01bf57aff8.png) 

Fig. 1 (a) The Bloch-point configuration originates from vortices with identical circularity, but opposite polarisation, which are stabilised through the DMI of the material, which fixes the
core orientation relative to circularity through the left- or right-handed chirality. (b, c) Simulation results for the two different Bloch-point types. 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/3af0ce5bf816d28b6250862025c394dc.png) 

Fig. 2 ASCII encoding of the string "Blochpoint" using 80 Bloch points. The two cross sections show an xy plane in the top layer and an xz plane at the Bloch-point position. The strip
length is chosen according to the predicted value for a nanostrip with width w = 100 nm.

Bloch points in helimagnetic nanostrips

The distance between islands must be reduced to increase an areal density (AD) in bit-patterned media recording (BPMR) [1]-[2], which means both inter-symbol interference (ISI) and inter-track interference (ITI) effects are avoidably increased. Therefore, BPMR system performance is effortlessly degraded [3]. In prior work [4], constrained code working with a multilayer perceptron (MLP) decoder in a staggered-array BPMR system was proposed. However, to get more improvement in the overall system performance of the magnetic recording system, we propose to apply the three-dimensional (3-D) magnetic recording that has double recording layers [5] with the constrained code performing with the MLP decoder. Here, a double recording layer medium is designed as a staggered pattern as shown in Fig. 1. Each layer is arranged as a regular array. Both of them are then arranged in a staggered pattern. The proposed double recording layer not only avoids the significant signal degradation from inter-layer interference (ILI) but also mitigates ISI and ITI effects. An input sequence, uk∈{±1}, is encoded by LDPC code and the rate-3/5 constrained encoder to obtain two encoded data sequences, [xk,0, xk,1] as shown in Fig. 2. The odd, xk,0, and even, xk,1, data sequences are recorded in the upper and lower layers, respectively. A single reader is always positioned between two desirable tracks to retrieve the readback signal, which is then oversampled at time t = kTx/2 to obtain a data sequence, rk, where Tx is a bit period. The 1-D equalizer and 1-D modified-soft output Viterbi algorithm (m-SOVA) are used to equalize and determine a log-likelihood ratio (LLR), λk, respectively. Then, it is decoded and produced the improved LLR values, respectively, with the rate-3/5 decoder and LLR estimator based on MLP, λ"k. Finally, the estimated user bit, ûk, is produced using an LDPC decoder. Simulation results indicate that, at the same user density (UD), the proposed system (AD = 5 Tb/in2) provides BER performance over the previous system [4]. 
**References** 
Y. Shiroishi et al., IEEE Trans. Magn., vol. 45, pp. 3816-3822 (2009) 
R. L. White, R. M. H. New, and R. F. W. Pease, IEEE Trans. Magn., pp. 990-995 (1997) 
P. W. Nutter et al., IEEE Trans. Magn., vol. 41, pp. 3214-3216 (2005) 
N. Rueangnetr et al., 19th ECTI-CON 2022, pp. 1-4 (2022) 
Y. Nakamura et al., IEEE Trans. Magn., vol. 58, pp. 1-5 (2022) 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/98b679bf2da99a70724f673c1d515de5.png) 
Fig. 1. Cross-section of head-media geometry for double recording layer medium.

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/4d2b2cd6f60fdce3351df357175cceef.png) 
Fig. 2. A code BPMR channel model with the rate-3/5 constrained code.

Double Layer Magnetic Recording with Multilayer Perceptron Decoder for Single

Many promising recording technologies have been introduced for hard disk drives to increase an areal density (AD), this paper considers the bit-patterned magnetic recording (BPMR) technology because it can achieve AD beyond 4 Tb/in2 [1]. To increase an AD in BPMR, which unavoidably leads to a problem of two-dimensional (2D) interference. To combat this difficulty, several techniques have been proposed based on modulation codes [2] and iterative processing [3]. Additionally, the BPMR system performance can be further enhanced by the proper island placement [4]. In this paper, we propose to optimize the bit-length (Tx) and track pitch (Tz) of BPMR under a single reader/two-track reading (SRTR) technique, which leads to getting greatly improved BER performance. Here, we arrange the island in a staggered pattern. The single reader was then employed to read both desired data tracks simultaneously. The signal waveforms from the upper and lower tracks, and the readback signal, that correspond to the data bits stored in the staggered medium through our proposed system are illustrated in Fig. 1. Here, we then investigate five cases under an iterative partial response maximum likelihood (PRML) system as follows: Case 1: Tx = 13.0 nm and Tz = 16.2 nm, Case 2: Tx = 14.0 nm and Tz = 15.0 nm, Case 3: Tx = 14.5 nm and Tz = 14.5 nm, Case 4: Tx = 15.0 nm and Tz = 14.0 nm, and Case 5: Tx = 16.2 nm and Tz = 13.0 nm, to obtain AD of 3 Tb/in2. Its data samples that were obtained from the over-sampling technique will then be processed through iterative PRML detection. Simulation results indicate that a system that has a larger bit-length distance, Tx, (Case 5) can provide the highest system performance when compared to other cases as shown in Fig. 2. In addition, the proposed system that encountered media noise still provides the highest system performance. It means that choosing proper Tx and Tz spacing can increase the efficiency of the staggered SRTR BPMR system. 
**References** 
[1] M. Mehrmohammadi et al., IOPscience, pp. 1-8 (2010) 
[2] C. Warisarn, A. Arrayangkool, and P. Kovintavewat, IEICE Trans. Electronics, vol. E98-C, pp. 528-533 (2015) 
[3] M. Tuchler, R. Koetter, and A. C. Singer, IEEE Trans. Commun., vol. 50, pp. 754-767 (2002) 
[4] S. Jeong, J. Kim, and J. Lee, Trans. Magn. vol. 54, pp. 1-4 (2018) 
![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/0b89ef2a061bbbe85d9581cdce381d10.png) 
Fig. 1. Position of a single reader between two desired tracks over the staggered island pattern and its readback signal.

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/9bfd1a6456b464e48339e9c692d701c6.png) 
Fig. 2. BER performances of different cases of SRTR BPMR system (a) without and (b) with 5% media noises.

A Study of Bit Island Spacing Optimization of Staggered Patterned Media based SRTR Scheme in BPMR Systems

Nd-Fe-B magnets are widely used in electric vehicles (EVs), wind turbines, and electronic devices in civilian and defence sectors due to their high magnetic performance. Alternate to the sintering route, the hot-deformation route does not require heavy rare elements (HRE) such as Dy and Tb and is a promising route to produce Nd-Fe-B magnets at low cost [1]. Hot deformed magnets are produced through melt spinning and hot pressing/deformation techniques [2]. Nd-Fe-B powder prepared from the melt-spun ribbon is used as precursors and hence optimization of melt spinning parameters is important to obtain desirable grain size and hard magnetic properties [3]. Hence a study was undertaken to prepare Nd-Fe-B melt-spun ribbons to tailor the microstructure and hard magnetic properties.
Melt spun ribbons with a nominal composition of Nd13.6Fe73.6Co6.6Ga0.6B5.6 alloy were prepared at various wheel speeds from 17 to 25 m/s. While the structure of the ribbons was obtained using X-Ray diffraction, the microstructure was obtained by transmission electron microscope (TEM). The magnetic properties and thermo-magnetic properties were obtained using a vibrating sample magnetometer (VSM). XRD analysis showed Nd2Fe14B phase with strong (00l) texture in all ribbons and texture fades away at higher wheel speeds. Thermal stability studied by calorimetry showed that the Nd2Fe14B phase was stable up to 500oC, beyond which precipitation of the α-Fe phase takes place. TEM analysis (Fig.1) showed that the Nd2Fe14B phase was distributed uniformly throughout the sample with a grain size of 50-150 nm. The average grain size was found to decrease with an increase in the wheel speed. The remanence (Br) and coercivity (Hc) values were found to increase with the increase in wheel speed up to 23 m/s and beyond which it starts decreasing gradually (Fig.2). Ribbons with the highest magnetic properties having a Br value of 6.6 kG and Hc value of 12.3 kOe were obtained at 23 m/s and may be best suited for the fabrication of hot deformed Nd-Fe-B magnets. 
**References:** 1 N.J. Yu, M.X. Pan, P.Y. Zhang, and H.L. Ge, J. Magn. 18, 235 (2013). 
2 K. Hioki, Sci. Technol. Adv. Mater. 22, 72 (2021). 
3 C. Rong and B. Shen, Chinese Phys. B 27, 117502 (2018). 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/0eba546af9c83328c8a792d6bbcd2171.png) 
Fig.1: TEM image showing the microstructure of Nd-Fe-B ribbon prepared at a wheel speed of 23 m/s 
![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/cb92d75e28c54dad0e4f00e8be180f65.png) 
Fig.2: Magnetic properties (Br and Hc) of Nd-Fe-B ribbon prepared at different wheel speeds

Processing and characterization of melt spun Nd

Hybrid magnonic-electric computing architectures are promising for ultra-low power data processing [1,2]. To realize these computing systems, scalable magnetoelectric
transducers are required that efficiently convert signals between the electric and magnetic domain. Several transducer types have been proposed of which voltage-based concepts are
expected to have the best scaling characteristics [3]. A promising voltage-based transducer type consists of a piezoelectric/magnetostrictive bilayer [4]. When a voltage is applied to this
transducer, dynamic strain is induced in the bilayer via the piezoelectric effect. This dynamic strain then interacts with the magnetization via the magnetostrictive effect and a coupling
between the electric and magnetic domain is obtained. Although this transducer type is very promising, its fundamental understanding as well as energy efficiency estimates are currently
lacking.
In this work, we develop a theoretical model of the displacement and magnetization dynamics in a resonator consisting of a piezoelectric/magnetostrictive bilayer. The model considers
both elastic as well as magnetic resonances and captures the magnetoelastic effect via an effective stiffness. Via this approach, both the elastic and magnetic power flow inside the device
can be determined. A comprehensive analysis was conducted of the dimensional as well as material parameter influence on the elastic and magnetic power dissipation (see Fig. 1).
Furthermore, the power calculation also results in magnetic transducer efficiency estimates. The results show that magnetic excitation efficiencies up to 90% can be achieved for nanometer
thick magnetostrictive layers (Fig. 2). Finally, a figure of merit is determined which allows to select the optimal piezoelectric-magnetostrictive material combination. 

**References:** 

1. A. Khitun and K. L. Wang, J. Appl. Phys. 110, 034306 (2011). 
2. B. Dieny, I. L. Prejbeanu, K. Garello, P. Gambardella, P. Freitas, R. Lehndorff, W. Raberg, U. Ebels, S. O. Demokritov, J. Akerman, et al., Nature Electron, 3 (2020). 
3. F. Vanderveken, V. Tyberkevych, G. Talmelli, B. Soree, F. Ciubotaru, and C. Adelmann, Sci. Rep. 12, 3796 (2022). 
4. A. Khitun, D. E. Nikonov, and K. L. Wang, J. Appl. Phys. 106, 123909 (2009). 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/0867d6ce1e1696b44c9b050ecec6782e.png) 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/5a2191baf82121f314207356220e5bb0.png)

Power transfer in magnetoelectric resonators

The combination of the two key effects of spintronics, magnetization dynamics and magneto-resistive effects, allows the realization of nano-neurons and nano-synapses with high computational capabilities. However, multilayer spintronic neural networks have never been realized with these nanodevices because there is no way to connect them that can work on a large scale. I will show that it is possible to exploit the two key effects of spintronics to connect together magnetic tunnel junctions that implement both neurons and synapses of a multilayer network. The magnetic tunnel junctions perform neural operations on DC signals and output the result as RF, operate synaptic operations on RF signals and output the result as DC thus giving rise to multilayer networks based on successive, clean, and fast conversions from RF to DC and from DC to RF. I will give a proof of concept with a two-layer neural network composed of nine interconnected magnetic tunnel junctions functioning as both synapses and neurons and experimentally demonstrate its ability to solve non-linear tasks with high performance. I will show that with junctions downscaled to 20 nm, such a network would consume 10fJ per synaptic operation and 100fJ per neuronal operation, several orders of magnitude less than current software neural network implementations. Finally, I will show through physical simulations that these networks, which can process DC data, can also natively classify RF inputs, achieving state-of-the-art drone identification from their RF transmissions. This study lays the foundation for deep spintronic neural networks at the nanoscale.

Premium content

Large interfacial Rashba interaction and giant spin orbit torque in atomically thin metallic multilayers

Downloads

Next from MMM 2022

Hardware Aware Bayesian Variational Inference Using Magnetic Tunnel Junction Noise Distributions in Artificial Synapse Arrays

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES