United States

A key challenge of neuromorphic computing is to build hardware neural networks with dense connectivity between nanoscale synapses and neurons, approaching the brain’s thousands of synapses per neuron (&lt;i&gt;1&lt;/i&gt;). In this talk, I will show that spin diodes are microwave nano-synapses that naturally and densely interface with neurons made of spin-torque nano-oscillators (&lt;i&gt;2&lt;/i&gt;). I will describe the principle of these novel synapses and explain how their microwave features can be harnessed to achieve the multiply-and-accumulate operations at the core of neural networks, with over 500 synapses/neuron. I will show that they implement all-to-all connections as well as convolutional operations in a fully parallel way with high density and speed. I will describe our experiments highlighting these functionalities (&lt;i&gt;3&lt;/i&gt;), and show that they open the path to radio-frequency deep learning.

2022 Joint MMM-INTERMAG

Radio-Frequency Spintronic Neural Networks

INTERMAG 2021 Welcome Video



	
<figure class="video_container">
 <iframe src="https://screencast-o-matic.com/player/crflqFVnUpZ?v=6&ff=1&title=0&controls=1" width=640 height=360 frameborder="0" allowfullscreen="true"> </iframe>
</figure>

Want to buy a ticket? 

Register for INTERMAG 2021 here

[![](https://assets.underline.io/uploads/markdown_image/1/image/17ce76e7960db633258024eef8d6da7f.png)](https://intermag.org/register)

Got your ticket, but still can’t access the event?

Contact us and our team will look into it.

INTERMAG 2021

A key challenge of neuromorphic computing is to build hardware neural networks with dense connectivity between nanoscale synapses and neurons, approaching the brain’s thousands of synapses per neuron (1). In this talk, I will show that spin diodes are microwave nano-synapses that naturally and densely interface with neurons made of spin-torque nano-oscillators (2). I will describe the principle of these novel synapses and explain how their microwave features can be harnessed to achieve the multiply-and-accumulate operations at the core of neural networks, with over 500 synapses/neuron. I will show that they implement all-to-all connections as well as convolutional operations in a fully parallel way with high density and speed. I will describe our experiments highlighting these functionalities (3), and show that they open the path to radio-frequency deep learning.

symposium

The 15th Joint MMM-INTERMAG Conference (2022 Joint) 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 plenary session, and an evening session, with about 1500 presentations overall. This Conference provides an outstanding opportunity for worldwide participants to meet their colleagues and collaborators and discuss developments in all areas of magnetism research.

The 2022 Joint Conference will offer both in-person and prerecorded on-demand content.

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

Please register for the Joint MMM-Intermag 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.

The flux leakages at axial end and radial inner side on the flux-concentrating rotor are more likely to generate, as shown in Fig.1, and therefore reduces the motor performance and degrades the utilization efficiency of PMs [1]-[3]. According to Table I, the modeling of the machines are carried out. Fig. 2 is the 2-D model. Fig. 3 shows the leakage flux by 3-D spoke type model. In Fig. 4, the air gap flux density, back electromotive forces (EMFs) and torque from the 2-D and 3-D analyses indicate a difference of approximately 10.33%, 5.88 % and 9.45%, respectively. Fig. 5 shows the rotor prototype and the test platform. The errors are 9.78% and 5.21% compared with the experiment respectively which indicates that the 3-D model is closer to the actual result than the 2-D model because the 3-D model takes leakage flux into account. The motor with novel rotors are shown in Fig. 6 and Fig. 7 which consumes the same weight of PMs with spoke-type model. The principle of reduction axial end leakage flux in Model 1 is shown in Fig. 6 (b). The details on the rotor of Model 2 is shown in Fig. 7 (b). The model in Fig. 6 is called Model 1. The model in Fig.7 is Model 2. Fig. 8 and Fig. 9 compare the rotor-end and rotor-interior leakage flux of two new structures and spoke-type motor under no-load. In the spoke type model the leakage flux is large, up to 0.9 T and 1.0 T. But it is only 0.38 T and 0.29 T in Model 1 and Model 2. The electromagnetic characteristics are compared in Fig. 10. Model 2 can further improve the air-gap flux density, back-EMF and torque, up to 1.499 T, 512.4 V and 68.7 Nm. In this paper, a novel rotor with hybrid radial and axial flux concentration considering leakage flux is designed. The 3-D models of concentrated winding PM motors with different rotors are built to analyze the leakage flux and electromagnetic performance. A 12-slot 10-pole motor with spoke type PM rotor is manufactured and experimented to verfy the significance of 3-D FEA takeing into account flux leakages. The 3-D FEA results of novel flux-concentrating PM rotors shows the axial leakage flux and rotor-interior leakage flux are decreased by 58.1% and 61.7%. The electromagnetic torque can be improved by 5.48% with phase current of 30 A.

![](https://assets.underline.io/uploads/markdown_image/1/image/79a32f6edee8bf86e65a0ad4807a2502.jpg)

![](https://assets.underline.io/uploads/markdown_image/1/image/630d41826d47b1eabe8f216c0ae4e275.jpg)

A Novel Spoke-type PM Rotor with Hybrid Radial and Axial Flux Concentration for Reduction of Interpolar Leakage Flux

It is necessary to accurately estimate iron loss including magnetic hysteresis behavior to develop high-efficiency electric machines. In a previous paper, a novel magnetic circuit model incorporating the play model [1] and Cauer circuit theory [2] was proposed. The play model is one of phenomenological models of the dc hysteresis characteristic. The Cauer circuit theory can simply express the skin effect by formulating the frequency characteristic of the complex permeability. Its calculation accuracy was proved by comparing the measured and calculated results of hysteresis loops and iron loss under PWM excitation by using simple-shaped ring cores. This paper presents that the above model is extended a reluctance network analysis (RNA) model to analyze devices with more complicated shapes such as motors. The proposed method is experimentally validated by using an IPM motor, which is generally driven by a PWM converter, as an object of discussion. Fig. 1(a) and (b) show the proposed RNA model of the IPM motor. In the stator core, each reluctance expresses the dc hysteresis characteristic and classical and anomalous eddy current losses. The dc hysteresis characteristic is represented by the play model, the classical eddy current loss is represented by the Cauer circuit theory, and the anomalous eddy current loss is represented by a simple controlled source element, respectively. On the other hand, each reluctance in an air gap and a rotor is determined by dimensions of each element and permeability of the materials, and each MMF generated from winding currents and permanent magnets is concentrated as a form of controlled source. Fig. 2(a) - (c) show the measured and calculated characteristics of torque, iron loss, and efficiency, respectively. Here, the theoretically infinite number of stages of the ladder circuit part which denotes classical eddy current loss is truncated at 3rd stage for practical use. From these figures, it is understood that both results are in good agreement. Besides, as shown in Fig. 2(d), the proposed method can draw the local hysteresis loop in a certain divided element, which is generally difficult to measure and calculate, though its calculation accuracy should be experimentally verified.

![](https://assets.underline.io/uploads/markdown_image/1/image/ba72a3d383a76039104055b74a096d4a.jpg)


![](https://assets.underline.io/uploads/markdown_image/1/image/51d551699060515c03658bc9628a80f9.jpg)

Iron Loss Calculation Method of Interior Permanent Magnet Motor Taking Carrier Harmonics into Account Based on Reluctance Network Analysis Representing Dynamic Hysteresis Characteristics

I Introduction Reactors have been widely used to ensure the stability of the power system by limiting short-circuit current, suppressing harmonic and compensating reactive power. There are some methods realize the vibration reduction of the reactor from the angle of the material filled at the air gap[1-2]. However, in order to improve this serious noise problem of the reactor, it is necessary to study the vibration reduction from the source of reactor vibration. Therefore, an optimization method based on topology optimization and GA(genetic algorithm) is proposed in this paper to reduce the segmented core reactor's vibration. II Topology Optimization of the Reactor This paper adopts the method of active vibration reduction and proposes a global topology optimization algorithm to get a novel design of the reactor. First, the electromagnetic-mechanical coupling model of the reactor is established to analyze the vibration and deformation of the reactor, which are shown in Fig1.(a-c). Second, the optimization problem is defined to obtain optimal configuration that minimum the vibration displacement of the reactor under a minimum bound of the total volume. Then, by introducing a SIMP(Solid Isotropic Material with Penalization) function into the definition of the permeability of the reactor's air gap part materials, the distribution of ferromagnetic material is represented. Meanwhile, the inductance value of the reactor can be kept stable in the optimization process through the joint GA. Finally, a novel reactor structure is obtained as shown in Fig2.(a) and the finite element numerical method is used to calculate the optimized parameter values before and after the optimization to verify the excellence of the design. III Conclusion The calculation result at sampling point A is shown in the Fig.2(b). It can be seen that the vibration displacement after optimization is reduced by about 15%, which verifies the effectiveness of the proposed method in this paper.

![](https://assets.underline.io/uploads/markdown_image/1/image/11bed137395190e0772f919055bce856.jpg)

Fig.1. (a) Schematic diagram of the reactor,(b) Magnetic flux density modulus distribution nephogram,(c) Displacement distribution nephogram.

![](https://assets.underline.io/uploads/markdown_image/1/image/834fe5bd196aaa212c5cc13e1ff70c28.jpg)

Fig.2. (a) Value distribution of control variable p,(b) Comparison of vibration displacement at sampling point a.

Reactor Vibration Reduction Using Global Topology Optimization Algorithms

Background and problem considered Eddy Current Testing (ECT) is one of the Non-Destructive Techniques and is widely used in manufacturing processes of long metal parts for surface inspection of metals. However, ECT creates a bottleneck preventing the speeding up of processes because the vibration of the inspection sample caused by the high speed of the manufacturing processes has a negative impact on ECT [1]. In previous studies, signals caused by vibration are removed from the detection signal by signal processing [2]-[4]. However, these processes are limited in their inspection speed by the speed of the digital signal processors performing the signal processing. We have proposed a new ECT probe that is more effective in suppressing the vibration component in the detection signal than that of the conventional self-induction probe does not require signal processing such as the multi-frequency method [5]. In this paper, the principle and effect of our proposed probe that can suppress vibration signals in detection signals are discussed. The Principle and Vibration Signal Suppression Effect in Dynamic Environment A photo of a used experimental setup is shown in Fig. 1. A brass rod that has two cracks was used as the test specimen. A vibration generator, consisting of a pulley and motor, can apply vibration to the test specimen. The proposed probe is an encircling probe with four coils and three induction modes, the self, mutual and dual induction modes. This probe is characterized by the fact that the vibration signals detected in the self and mutual induction modes cancel each other out, as in noise-canceling headphones when the dual induction mode is used. Fig. 2 shows measured results of the self and dual induction modes when vibration was applied to the brass rod. The amplitude of the dual induction mode was lower than the amplitude of the self induction mode. However, crack signals were not suppressed. The detail of the principle, the experimental method and comparison between the proposed method and the multi frequency method will be presented in our presentation.

![](https://assets.underline.io/uploads/markdown_image/1/image/fc796e26a65458a6407b5fb627e4e763.jpg)

Fig. 1. The experimental setup.

![](https://assets.underline.io/uploads/markdown_image/1/image/9fd0f25e30c8f398cb05cd462ad3f6e1.jpg)

Fig. 2. Measured results in dynamic environment.

Vibration Signal Suppression Effect of Dual Induction Eddy Current Probe in Dynamic Environment

In the last years, a new field has emerged at the intersection between Condensed Matter and Quantum Optics, denominated “Quantum Magnonics”. This field strives to control the elementary excitations of magnetic materials, denominated magnons, to the level of the single quanta, and to interface them coherently to other elementary excitations such as photons or phonons. The recent developments in this field, with proof of concept experiments such as a single-magnon detector, have opened the door for hybrid quantum systems based on magnetic materials. This can allow us to explore magnetism in new ways and regimes, has the potential of unraveling quantum phenomena at unprecedented scales, and could lead to breakthroughs for quantum technologies. A predominant role in these developments is played by cavity magnonic systems, where an electromagnetic cavity, either in the optical or microwave regime, is used to enhance and control the interaction between photons and magnons. This Tutorial Session will cover the basic concepts behind Quantum Magnonics, as well as the most recent developments in the field.

Cavity Magnonics

Hybrid magnonic systems incorporating superconducting systems have recently emerged as a new promising direction for coherent information processing by exploiting magnon excitations and high quality factor superconducting resonators [1]. Magnons are fundamental excitations of magnetic systems with a frequency band which can be adapted in microwave superconducting quantum circuits. In particular, magnons possess large dipolar magnon-photon coupling and wide frequency tunability, which may find new potentials in quantum information. Superconducting resonators serve as coherent data buses of many quantum circuits with long microwave coherence time. Superconducting-circuit-based cavity magnonics may provide a convenient circuit platform that bridges magnonics and quantum information. In this tutorial, we will discuss recent progresses of cavity magnonics built on superconducting circuits. We will start from the original idea of coupling spin ensembles with superconducting resonators, and how magnetic spins are involved to replace spin ensembles. Then we will examine different magnetic materials and superconducting resonator designs for building superconducting cavity magnonics. Last, we will discuss future opportunities and challenges in scaling up such hybrid magnonic systems on superconducting circuits, with an emphasis on maintaining coherence time of magnons and photons, as well as utilizing superconducting properties such as vortices for empowering magnonics.

Emerging Opportunities for Coherent Information Processing at the Intersection of Superconductivity and Magnetism

The past years have witnessed an increasing interest in the field of quantum materials (QM), not only thanks to their fascinating properties as a macroscopic manifestation of quantum mechanics (high-temperature superconductivity, quantum Hall effect, giant magnetoelectric effect, etc...), but also for the opportunities that these materials can offer towards the next generation of electronic devices. Understanding the physics behind quantum materials is thus a primary goal of condensed matter physics, both from a fundamental point of view as well for gaining control over the QM functional properties. Among all the experimental observables, excitations in QMs can provide unique information about the key-interactions at play in the system (magnetic interactions, electron-phonon coupling, spin-orbit coupling, topology, multiferroicity, etc...) and the nature of the excitations themselves (magnon, phonon, intra-band electronic excitations, orbiton, exotic modes, etc...). In this talk, I will focus on soft Resonant Inelastic X-ray Scattering (RIXS), as an emergent technique for the study of excitations and their momentum dependence. I will present recent results achieved on ferromagnetic thin films to investigate the evolution of the spin dynamics as a function of thickness [1]. Our work reveals an asymmetric evolution of the spin dynamics between the in-plane and out-of-plane directions: while the spin excitations remain invariant along the in-plane direction, in the out-of-plane direction a dampening is observed as a function of thickness. Such result is interpreted as an effect of confinement, enabling the use of the thickness as a “knob” to control the spin excitation energies.

Tuning Spin Excitations in Magnetic Films by Confinement

Significant attention has recently been attracted by phases that spontaneously break real-space symmetries in addition to spin, phase, or gauge symmetries. These phases include, for example, the charge-density wave or stripe phases in the cuprate superconductors, the Ising-nematic phases in the iron pnictides, and the valence-bond-solids in certain quantum magnets. We discuss the interplay between the broken real-space symmetries and vacancies, impurities and other types of quenched disorder that are inevitable in real materials. Specifically, we demonstrate that spinless impurities in a frustrated Ising magnet give rise to a random-field mechanism that can destroy the stripe-ordered phase. The strength of the emerging random fields is governed by the spatial impurity distribution. Moreover, the mechanism can be tuned very efficiently by weak exchange anisotropies that explicitly break the real-space symmetry, providing a way of controlling the phase diagram of this many-particle system. This work was supported in part by the National Science Foundation under Grant Nos. DMR-1828489 and OAC-1919789.

![](https://assets.underline.io/uploads/markdown_image/1/image/28c2747899a50f243b196002129a9945.jpg)

Domains of horizontal and vertical stripes in the diluted J1-J2 Ising model.

Controlling the Stripe Order in a Diluted Frustrated Magnet

Microwave detectors based on the spin-torque diode effect are among the key emerging spintronic devices (see ref.[1] and references therein). By utilizing the spin of electrons in addition to their charge, they have the potential to overcome the theoretical performance limits of their semiconductor (Schottky) counterparts. Those devices realized with magnetic tunnel junctions and in active regime exhibit high-detection sensitivity &gt;200kV/W at room temperature and zero bias field, thanks to the injection locking mechanism, and for low-input power (micro-Watts or lower). Those performance can be improved with the bolometric effect or when coupled properly in arrays. Here, I will discuss our recent results in the field of microwave detectors based on spin diodes and possible implementations of THz detectors based on antiferromagnets working in linear and parametric excitation regime. I will point out how those devices can impact the IoT market and can be used for implement hardware multiplication by using the concept of degree of rectification and as dual band detectors in frequency shifting keying demodulation schemes in which the information is coded in two discrete values of the frequency in a carrier signal. In particular, the frequency response of spin-torque diodes and their current tunability can be also used as building blocks of the hardware realization of neurons and synapses in neuromorphic applications. Another application of spintronic diodes, when they have a broadband frequency response, is as electromagnetic energy harvesting, which offers an attractive energy source for applications in self-powered portable electronics in the “internet of things” era. Here I will show the development of a bias-field-free spin-torque diodes based on a magnetic tunnel junction having a canted magnetization in the free layer, and demonstrate that those devices could be an efficient harvester of broadband ambient RF radiation, capable to efficiently harvest microwave powers of microWatt and below and to power a black phosphorous nanodevice. I will also review recent idea on spintronic detectors in presence of nonuniform ground states such as skyrmions.

Challenges in microwave and THz spintronic diodes

There is a great interest to generate electricity using ambient RF energy. We addressed this using two different approaches. We first use the intrinsic quantum property of the Weyl semimetal to convert the RF energy into electricity [1]. The nonlinear Hall effect (NLHE) can produce a dc transverse voltage in zero magnetic field. However, it requires NLHE to be detected at room temperature. Despite of several studies of NLHE in topological materials and 2D materials, the room-temperature NLHE has not yet been achieved. We find the room-temperature NLHE in a type-II Weyl semimetal TaIrTe4, which hosts a robust NLHE due to broken inversion symmetry and large band overlapping at the Fermi level. The NLHE is present only when the electric field is applied along the axis of the mirror plane (a-axis of the crystal). Based on the observed room-temperature NLHE in TaIrTe4 we demonstrate the wireless RF rectification with zero external bias and magnetic field. An incident electric field along the a-axis (Ea) generates a rectified voltage along the b-axis (Vb), whereas, there is no rectification when the incident electric field is along the b-axis (Eb) in consistent with the symmetry constraints of NLHE. This work opens a door to realizing room temperature rectification applications based on the NLHE in Weyl semimetals. The other approach is to use the spin-torque diode effect of spin-torque oscillators (STOs) [2]. For the application in rectification and energy harvesting, the spin-diode effect is demonstrated by measuring the rectified dc voltage. Due to the canted anisotropy, the free layer of the individual STOs shows a large RF sensitivity as a result of non-linear dynamics. We demonstrate series connections have an advantage for rectification due to the additive effect of the diode voltages from STOs. Using eight oscillators in series, the rectified voltage is enhanced and we have achieved a rectification efficiency of 6% at -20 dBm, which outperforms the Schottky diode capability at sub µW power. By integrating the electrically connected eight STOs with conventional electronics, we demonstrate the battery-free energy-harvesting system by utilizing the wireless RF energy to power electronic devices such as LEDs. We also demonstrate the capability of this energy harvesting system in holding dc power using a time-varying signal, which is useful in harvesting energy from discrete commercial sources such as a WiFi router. Our results highlight the significance of electrical topology (series vs. parallel) while designing an on-chip STOs system.

Radio-Frequency Spintronic Neural Networks

Downloads

Next from 2022 Joint MMM-INTERMAG

A Novel Spoke-type PM Rotor with Hybrid Radial and Axial Flux Concentration for Reduction of Interpolar Leakage Flux

Similar lecture

Design of a Coplanar-Waveguide-Based Microwave-to-Spin-Wave Transducer

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Radio-Frequency Spintronic Neural Networks

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from 2022 Joint MMM-INTERMAG

A Novel Spoke-type PM Rotor with Hybrid Radial and Axial Flux Concentration for Reduction of Interpolar Leakage Flux

Similar lecture

Design of a Coplanar-Waveguide-Based Microwave-to-Spin-Wave Transducer

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads