France



Artificial neural networks (NN) running on traditional computing hardware tend to be relatively slow and power-hungry. One possible solution to this issue is to create hardware with neurons that spike in a manner similar to biological neurons. For example, Intel [1] and IBM [2] have recently fabricated neuromorphic chips based on spiking NN. Still, neurons in these chips are fabricated with multiple silicon transistors, which limits power efficiency of such an approach. The development of neuromorphic chips would benefit from a nanoscale, single element synthetic neuron that behaves in a manner that closely resembles a biological neuron. Such a synthetic neuron can be fabricated with nanometer dimensions from an antiferromagnetic (AFM) material mounted on platinum [3]. These “AFM neurons” are spin Hall oscillators driven by a sub-threshold spin current that can produce ultra-short (~ 5 ps) spikes in response to a weak external stimulus.&lt;br/&gt;&lt;br/&gt;In traditional spiking NN, synapses that connect artificial neurons are formed by complex active circuits. A much simpler approach would be to use passive memristors as artificial synapses [4]. However, memristors create &lt;i&gt;symmetric coupling&lt;/i&gt; between neurons, which allows a spike produced by one neuron to travel to all neighboring neurons, both forwards and backwards. Unfortunately, this bi-directional signal propagation makes use of standard machine learning algorithms impossible. Therefore, it would be interesting to design NN in which unidirectional spike propagation is possible even with symmetrically coupled neurons.&lt;br/&gt;&lt;br/&gt;Here, we demonstrate that the unidirectional signal propagation is realizable in symmetrically-coupled NNs based on AFM neurons. This regime is possible due to an effective inertia that originates from exchange coupling between the two magnetic sublattices [3] and creates a relatively long refractory period - an interval of time after a neuron fires when it is unable to fire again. We show, that it is possible to adjust synapse weights such that each neuron is still in its refractory period when it receives a backward propagating signal from its neighbors. Then, a simple chain of neurons receiving an input at one end carry the signal to the opposite end without any reflections.&lt;br/&gt;&lt;br/&gt;However, for such a simple chain of AFM neurons, an input can occur at either end, and thus spikes would be able to travel in both directions. It would be beneficial to create an ‘isolator’ neuromorphic circuit, which would allow the propagation of spikes in only one direction. Figure 1 depicts the schematic of a possible ‘isolator’ circuit composed of AFM neurons. This figure uses circles to represent AFM neurons. The isolator is composed of neurons 4, 5, and 6. In the figure, double arrows represent symmetrical synapses with arrow width proportional to the coupling strength.&lt;br/&gt;&lt;br/&gt;This isolator allows spikes to be carried “forwards” from neuron 1 to 9, but stops signals carried “backwards” from neuron 9 to 1, as illustrated by the results of numerical simulations shown in Fig. 2. The isolator functions as follows. When the spike is initiated in neuron 1 (Fig. 2(a)), it travels along the chain to neuron 4. Then, the spike generated by neuron 4 travels to both neuron 5 and neuron 6. Neuron 5 also generates a spike, which is carried to neuron 6. The synaptic weights connecting to neuron 6 are small enough that a single input is not sufficient to induce a spike in the AFM neuron. However, when both spikes combine at neuron 6, they provide enough stimulus for neuron 6 to fire. The spike can then continue to propagate from neuron 6 to 9. When traveling backwards (Fig. 2(b)), a spike begins at neuron 9 and travels unidirectionally to neuron 6. Neuron 6 will generate a spike, but the synaptic weights from neuron 6 to 4/5 are insufficient to induce an additional spike. This, in effect, prevents the backward travel of spikes through the isolator.&lt;br/&gt;&lt;br/&gt;Using the proposed isolator NN, it is possible to design a number of other neuromorphic circuits supporting unidirectional information propagation, for example, neuromorphic combiners and unidirectional memory rings. The creation of such circuits with symmetric synapses shows hope for AFM neurons to be used to create more realistic spiking NN. The delay caused by the exchange inertia in the AFM neurons plays a key role in any neuromorphic circuit constructed with these neurons. These physically realistic neurons will increase the possibility for more functional spiking neural network hardware.

**References:**

[1] M. Davies et al., &lt;i&gt;Loihi: A Neuromorphic Manycore Processor with On-Chip Learning&lt;/i&gt;, in IEEE Micro, vol. 38, no. 1, pp. 82-99, January/February 2018, doi: 10.1109/MM.2018.112130359.&lt;br/&gt;[2] M. V. DeBole et al., &lt;i&gt;TrueNorth: Accelerating From Zero to 64 Million Neurons in 10 Years&lt;/i&gt;, in Computer, vol. 52, no. 5, pp. 20-29, May 2019, doi: 10.1109/MC.2019.2903009.&lt;br/&gt;[3] R. Khymyn, I. Lisenkov, J. Voorheis, O. Sulymenko, O. Prokopenko, V. Tiberkevich, J. Akerman, and A. Slavin, &lt;i&gt;Ultra-fast artificial neuron: generation of picosecond-duration spikes in a current-driven antiferromagnetic auto-oscillator&lt;/i&gt;, Scientific Reports 8, 15727 (2018).&lt;br/&gt;[4] Sung Hyun Jo, Ting Chang, Idongesit Ebong, Bhavitavya B. Bhadviya, Pinaki Mazumder, and Wei Lu &lt;i&gt;Nanoscale Memristor Device as Synapse in Neuromorphic System, Nano Lett. 2010, 10, 4, 1297–1301&lt;/i&gt;

**Images:**

![](https://assets.underline.io/uploads/markdown_image/1/image/eb70901938b9e7ce0e61457aefc96cde.jpg)
Fig 1: Chain of spiking neurons, configured as an isolator circuit. Gray circles represent neurons, and black arrows represent coupling between neurons. Thick black arrows represent symmetric synaptic weight tuned such that a spike cannot flow backwards. Thin black double arrows represent symmetric synaptic weight weak coupling.

![](https://assets.underline.io/uploads/markdown_image/1/image/7460602315b4c8ef8afb70fa1a01a721.jpg)
&lt;br /&gt;
Fig 2: Results of simulation of AFM neurons configured as an isolator circuit. (a) Initial spike is in neuron 1. The spike travels through the chain, to neuron 9. (b) The initial spike is at neuron 9. The spike travels to neuron 6, then is unable to continue.

INTERMAG 2021

 Antiferromagnetic Artificial Neural Networks with Symmetric Coupling

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.

Artificial neural networks (NN) running on traditional computing hardware tend to be relatively slow and power-hungry. One possible solution to this issue is to create hardware with neurons that spike in a manner similar to biological neurons. For example, Intel [1] and IBM [2] have recently fabricated neuromorphic chips based on spiking NN. Still, neurons in these chips are fabricated with multiple silicon transistors, which limits power efficiency of such an approach. The development of neuromorphic chips would benefit from a nanoscale, single element synthetic neuron that behaves in a manner that closely resembles a biological neuron. Such a synthetic neuron can be fabricated with nanometer dimensions from an antiferromagnetic (AFM) material mounted on platinum [3]. These “AFM neurons” are spin Hall oscillators driven by a sub-threshold spin current that can produce ultra-short (~ 5 ps) spikes in response to a weak external stimulus. In traditional spiking NN, synapses that connect artificial neurons are formed by complex active circuits. A much simpler approach would be to use passive memristors as artificial synapses [4]. However, memristors create symmetric coupling between neurons, which allows a spike produced by one neuron to travel to all neighboring neurons, both forwards and backwards. Unfortunately, this bi-directional signal propagation makes use of standard machine learning algorithms impossible. Therefore, it would be interesting to design NN in which unidirectional spike propagation is possible even with symmetrically coupled neurons. Here, we demonstrate that the unidirectional signal propagation is realizable in symmetrically-coupled NNs based on AFM neurons. This regime is possible due to an effective inertia that originates from exchange coupling between the two magnetic sublattices [3] and creates a relatively long refractory period - an interval of time after a neuron fires when it is unable to fire again. We show, that it is possible to adjust synapse weights such that each neuron is still in its refractory period when it receives a backward propagating signal from its neighbors. Then, a simple chain of neurons receiving an input at one end carry the signal to the opposite end without any reflections. However, for such a simple chain of AFM neurons, an input can occur at either end, and thus spikes would be able to travel in both directions. It would be beneficial to create an ‘isolator’ neuromorphic circuit, which would allow the propagation of spikes in only one direction. Figure 1 depicts the schematic of a possible ‘isolator’ circuit composed of AFM neurons. This figure uses circles to represent AFM neurons. The isolator is composed of neurons 4, 5, and 6. In the figure, double arrows represent symmetrical synapses with arrow width proportional to the coupling strength. This isolator allows spikes to be carried “forwards” from neuron 1 to 9, but stops signals carried “backwards” from neuron 9 to 1, as illustrated by the results of numerical simulations shown in Fig. 2. The isolator functions as follows. When the spike is initiated in neuron 1 (Fig. 2(a)), it travels along the chain to neuron 4. Then, the spike generated by neuron 4 travels to both neuron 5 and neuron 6. Neuron 5 also generates a spike, which is carried to neuron 6. The synaptic weights connecting to neuron 6 are small enough that a single input is not sufficient to induce a spike in the AFM neuron. However, when both spikes combine at neuron 6, they provide enough stimulus for neuron 6 to fire. The spike can then continue to propagate from neuron 6 to 9. When traveling backwards (Fig. 2(b)), a spike begins at neuron 9 and travels unidirectionally to neuron 6. Neuron 6 will generate a spike, but the synaptic weights from neuron 6 to 4/5 are insufficient to induce an additional spike. This, in effect, prevents the backward travel of spikes through the isolator. Using the proposed isolator NN, it is possible to design a number of other neuromorphic circuits supporting unidirectional information propagation, for example, neuromorphic combiners and unidirectional memory rings. The creation of such circuits with symmetric synapses shows hope for AFM neurons to be used to create more realistic spiking NN. The delay caused by the exchange inertia in the AFM neurons plays a key role in any neuromorphic circuit constructed with these neurons. These physically realistic neurons will increase the possibility for more functional spiking neural network hardware.

**References:**

[1] M. Davies et al., Loihi: A Neuromorphic Manycore Processor with On-Chip Learning, in IEEE Micro, vol. 38, no. 1, pp. 82-99, January/February 2018, doi: 10.1109/MM.2018.112130359. [2] M. V. DeBole et al., TrueNorth: Accelerating From Zero to 64 Million Neurons in 10 Years, in Computer, vol. 52, no. 5, pp. 20-29, May 2019, doi: 10.1109/MC.2019.2903009. [3] R. Khymyn, I. Lisenkov, J. Voorheis, O. Sulymenko, O. Prokopenko, V. Tiberkevich, J. Akerman, and A. Slavin, Ultra-fast artificial neuron: generation of picosecond-duration spikes in a current-driven antiferromagnetic auto-oscillator, Scientific Reports 8, 15727 (2018). [4] Sung Hyun Jo, Ting Chang, Idongesit Ebong, Bhavitavya B. Bhadviya, Pinaki Mazumder, and Wei Lu Nanoscale Memristor Device as Synapse in Neuromorphic System, Nano Lett. 2010, 10, 4, 1297–1301

**Images:**

![](https://assets.underline.io/uploads/markdown_image/1/image/eb70901938b9e7ce0e61457aefc96cde.jpg)
Fig 1: Chain of spiking neurons, configured as an isolator circuit. Gray circles represent neurons, and black arrows represent coupling between neurons. Thick black arrows represent symmetric synaptic weight tuned such that a spike cannot flow backwards. Thin black double arrows represent symmetric synaptic weight weak coupling.

![](https://assets.underline.io/uploads/markdown_image/1/image/7460602315b4c8ef8afb70fa1a01a721.jpg)
 
Fig 2: Results of simulation of AFM neurons configured as an isolator circuit. (a) Initial spike is in neuron 1. The spike travels through the chain, to neuron 9. (b) The initial spike is at neuron 9. The spike travels to neuron 6, then is unable to continue.

technical paper

We theoretically investigate the topological aspects of the magnon-phonon hybrid excitation in a simple two-dimensional (2D) square-lattice ferromagnet with perpendicular magnetic anisotropy [1]. Several distinguishing features of our model are as follows. Our model is optimized for atomically thin magnetic crystals, i.e., 2D magnets. The recent discovery of magnetism in 2D van der Waals materials opens huge opportunities for investigating unexplored rich physics and future spintronic devices in reduced dimensions [2,3]. In our 2D model, we ignore the nonlocal dipolar interaction, which is not a precondition for a finite Berry curvature in 2D magnets. Moreover, the Berry curvature we find requires neither a special spin asymmetry such as the Dzyaloshinskii-Moriya interaction (DMI) nor a special lattice symmetry: Our 2D model description is applicable for general thin-film ferromagnets. Therefore, we show that even without such long-range dipolar interaction, DMI, or special lattice symmetry, the nontrivial topology of a magnon-phonon hybrid can emerge by taking account of the well-known magnetoelastic interaction originates from the magnetocrystalline anisotropy. Because the magnetocrystalline anisotropy is ubiquitous in ferromagnetic thin-film structures, our result does not rely on specific preconditions and thus is quite generic. Furthermore, we show that the topological structures of the magnon-polaron bands can be manipulated by effective magnetic fields via topological phase transition. We uncover the origin of the nontrivial topological bands by mapping our model to the well-known two-band model for topological insulators [4], where the Chern numbers are read by counting the number of topological textures, called skyrmions, of a certain vector in momentum space. In this picture, the magnon-phonon hybridization induces the chiral texture of the momentum space vector. As an experimental probe for our theory, we propose the thermal Hall conductivity [Fig. 2].

**References:**

[1] G. Go et al., Topological Magnon-Phonon Hybrid Excitations in Two-Dimensional Ferromagnets with Tunable Chern Numbers, Phys. Rev. Lett. 123, 237207 (2019). [2] C. Gong et al., Discovery of intrinsic ferromagnetism in two-dimensional van der Waals crystals, Nature (London) 546, 265 (2017). [3] B. Huang et al., Layer-dependent ferromagnetism in a van der Waals crystal down to the monolayer limit, Nature (London) 546, 270 (2017). [4] B. A. Bernevig, T. L. Hughes, and S.-C. Zhang, Quantum spin Hall effect and topological phase transition in HgTe quantum wells, Science 314, 1757 (2006).

**Images:**

![](https://assets.underline.io/uploads/markdown_image/1/image/436ed988646e642bcb99ed25391293f6.jpg)
Fig. 1. (a) The schematic illustration of the magnon and phonon system. The ground state of the magnetization is given by the uniform spin state along the z-axis (red arrow). (b) The Chern number of our magnon-phonon hybrid system. H represents the effective magnetic field including the anisotropy field and the external magnetic field. Band dispersion (c) without the magnon-phonon interaction and (d) with the magnon-phonon interaction.

![](https://assets.underline.io/uploads/markdown_image/1/image/2bc284e0ccbc6784cc3e8d6d0786fed4.jpg)
FIG. 2. (a) Dependence of thermal Hall conductivity on the effective magnetic field Heff and with different temperatures T using the parameters in the main text. (b) Dependence of the Chern number on the effective magnetic field Heff using the parameters in the main text.

 Topological magnon-polaron in a two-dimensional ferromagnet

Acoustically-driven ferromagnetic resonance (ADFMR) gained interest in 2012 as Dreher et al. (1) developed it as a powerful and versatile method of examining complex magnetoacoustic interactions. While initial studies established the technique, the physical mechanisms behind the coupling must be thoroughly understood if ADFMR is to be exploited in next-generation microwave communications devices. In particular, the nonlinearity of ADFMR driven at high powers (2) is not well characterized, which may result in unwanted or unpredicted behavior when devices become optimized. To address this aspect, we produce ADFMR devices using curved interdigitated transducers (IDTs), in contrast to the straight IDTs that are typically used. Like an optical lens, these IDTs are engineered to focus acoustic energy towards a region smaller than the IDTs themselves. In this way, acoustic and magnetic amplitudes in a small region in the center of the device may be driven much higher than was previously possible, allowing examination of nonlinear processes such as Suhl instabilities. This may also produce an avenue for improved magnetoacoustic coupling, as well as further device miniaturization. Our devices are produced in much the same way as previously described (2), using aluminum split-finger IDTs and a 20-nm thin-film nickel layer grown on a y-cut lithium niobate (LiNbO3) substrate. We examine a device using a scanning acoustic-wave imaging interferometer similar to that used in Ref. (3). A 785-nm laser is split into signal and reference paths using an acousto-optic modulator operated at 40 MHz. The signal component is focused onto the device with a 10X objective lens. The beams are then recombined and the interference between them is recorded using a lock-in amplifier. Meanwhile, the ADFMR device is driven at its fundamental frequency 284 MHz or a higher odd harmonic. The DC photodiode signal is simultaneously recorded as a reference image (Fig. 1a), and follows the reflectivity of the scanned spot, showing the exact positions of the metallic nickel and aluminum surfaces. The vibration of the probed spot, downmixed to the range of the lock-in, is shown in Fig. 1b. Here, the gaussian-like curve of the focused SAWs is clearly visible as they converge to a high-amplitude region near the center of the device. Extracting a longitudinal segment of this signal allows us to see the oscillatory behavior in Fig. 1c. These oscillations indicate a standing wave with wavelength v/f = 12 µm (v = 3488 m/s is the Rayleigh velocity of SAWs in the LiNbO3 and f = 284 MHz is the driving frequency), as the output of the device is reflected back towards the input. To quantify the power convergence, transverse slices were taken near the center and fitted with the lorentzians shown in Fig. 1d, using the fitting parameters in Fig. 1e. Near the center (X ≈ 20.7 mm), the amplitude reaches a value of nearly 5 times greater than it is near the IDTs, as the width shrinks to a minimum of 22 µm, demonstrating that the focused IDTs significantly increase the energy density in a small region. ADFMR measurements on a focused IDT device show extremely power-dependent behavior. The linear regime is accessed with a -15 dBm driving field (Fig. 2a), where the shape is independent of power. However, with only +15 dBm of input power (Fig. 2b) the absorption lobes begin to shift towards lower field, and with a weaker maximum absorption. At +25 dBm of input power (Fig. 2c) the lobes shift even further and are much weaker in comparison. These results are a clear indication that a nonlinear power regime is accessed with modest input powers with the assistance of the curved IDTs, dramatically concentrating acoustic energy density. *This material is based on work supported by the Air Force Office of Scientific Research under project number FA9550-20RXCOR074

**References:**

1. L. Dreher, M. Weiler, M. Pernpeintner, H. Huebl, R. Gross, M. S. Brandt, and S. T. B. Goennenwein, “Surface acoustic wave driven ferromagnetic resonance in nickel thin films: Theory and experiment,” Physical Review B, vol. 86, no. 13, p. 134415, 2012. 2. D. Bas, P. Shah, M. McConney and M. Page, “Optimization of acoustically-driven ferromagnetic resonance devices,” Journal of Applied Physics, vol. 126, no. 11, p. 114501, 2019. 3. Z. Shen, X. Han, C. L. Zou, and H. X. Tang “Phase sensitive imaging of 10 GHz vibrations in an AlN microdisk resonator,” Review of Scientific Instruments, vol. 88, No. 12, p. 123709, 2017.

**Images:**

![](https://assets.underline.io/uploads/markdown_image/1/image/cf30be12a3755d154f36adeba3fe0211.jpg)
Figure 1. Interferometric imaging of microwave surface acoustic waves on a device. (a) Reflectivity map of the device in arbitrary units, scanned with 5-µm steps, highlighting the curved interdigitated transducers (IDTs), thin film of nickel (Ni), and substrate of lithium niobate (LiNbO3). (b) Interferometric acoustic wave imaging showing the amplitude of surface vibrations in arbitrary units, demodulated at the fundamental frequency of the IDTs (284 MHz) after downmixing. (c) Line scan indicated by the blue dashed line in (b), fitted with a sinusoid of period 12 µm, which corresponds to the period of standing waves in LiNbO3 at 284 MHz. (d) A series of lorentzian fits along the Y-direction for the data indicated by the dashed orange box in (b). (e) Fitting parameters width and height for the curves in (d).

![](https://assets.underline.io/uploads/markdown_image/1/image/43305569f2e8ce46063c1c55933a4975.jpg)
Figure 2. Normalized ADFMR signal in a focused IDT device. Driving field is (a) -15 dBm, (b) +15 dBm, and (c) +25 dBm at 864 MHz.

 Focused SAWs enhance nonlinearity in ADFMR

Magnetoelectric multiferroics, which simultaneously stabilize magnetic and polar static orders, have been predicted to present unusual dynamical magnetoelectric effects (ME). The relatively large magnitude of this dynamical ME response as well as its strong interac-tion with light is extremely attractive for potential applications among others are multifunctional spin-based memories or optical switches. The dynamical ME effects under consideration in this presentation, are characterized by the existence of an electro-active magnetic excitation called electromagnon. The electroactivity of electromagnons was first observed in type II multiferroics, where ferroelectricity is magnetically induced. In orthorhombic RMnO3 and RMn2O5(R=Tb, Y) families, both THz [1] and Raman [2] spectroscopies showed that this mode can be excited by means of the electric field of the light. The demonstration of the complementary magnetic nature of this excitation has been the prerogative of inelastic neutron scattering (INS) [3]. Sincethen, electromagnons have also been reported in Type I multiferroics hexagonal RMnO3 [4], in BiFeO3 and even in a paraelectric compound Ba2Mg2Fe12O22. To describe this unusual property, different mechanisms were proposed, in particular for the manganites families. These models basically rely on the very same magnetoelectric coupling which accounts for the static properties: lattice vibrations alter the exchange couplings or anisotropies which in turn affect the magnon spectrum. In helical magnetic systems for instance, where the multiferroicity originates from Dzyaloshinskii-Moriya Interac-tion (DMI), the same DMI has been proposed to explainthe origin of the electromagnon [5]. Although later refuted in DyMnO3, such DMI based scenario seems to be at play in CuO. Other propositions are based on Exchange-Striction (ES) [6], and couple a polarlattice mode to the magnon spectrum [1]. So far, these models have been consistent with experimental results in TbMn2O5 [3]. These interpretations tend to reduce the definition of the electromagnon to a hybrid magnon-phonon mode, or, as proposed in ErMnO3, to a mixing between a magnon and an electro-active rareearth crystal field transition [4]. In this presentation we question this issue and support the idea that electroactivity can be a purely electronic phenomenon. The RMn2O5 multiferroic compounds crystallize in the polar Pm space group [7] but present quasi-symmetries of the paraelectric Pbam space group. Along the c direction, the structure is composed of chains of Mn4+O6 octahedra, separated by layers of R3+ or Mn3+ ions. In the (a,b) plane, zig-zag chains of Mn4+O6 octahedra and Mn3+O5 square-pyramids run along the a axis and are stacked along b. GdMn2O5, like the other members of the family, shows an Incommensurate Magnetic ordering (ICM) belowTN=38 K, with a magnetic propagation wave vector k1= (½,0,0.28). Then, below Tc=32 K, the magnetic order becomes commensurate with a propagation wave vector k2= (½,0,0), accompanied by the largest electric polarization known in type II multiferroics [8]. This strong electric polarization has been ascribed to two cooperative ES mechanisms [6]. As all members of the series, the electric polarization is mainly oriented along the b axis [8] even if few studies have recently suggested the existence of small additional components along the a and even the c directions. The dynamical properties of this compound have been studied by means of IR and INS on high quality single crystals. In this presentation, we will mainly focus on the demonstration and characterization of an electromagnon present in the compound GdMn2O5. More precisely, the IR measurements allowed us to demonstrate the existence of an excitation below the TN, at an energy of 35cm-1 at 5K, which is electroactive because it is sensitive to the direction of the electric field of light (see fig. 1). The temperature tracking shows that this excitation disappears above the TN, which shows the magnetic origin of this excitation. Moreover, INS measurements have allowed us to measure the excitation in the center of the magnetic zone (½,0,0) (see fig. 2), but also to measure its dispersion, which we have been able to interpret through simulations based on the «Holstein-Primakov» representation, which shows that this excitation is an electroactive spin wave, and thus an electromagnon. Polarized INS measurements (see fig. 2) allowed us to discriminate between the nuclear and magnetic contributions, and to show that this excitation is purely magnetic, thus excluding the hypothesis of a coupling with a phonon. Finally, the presence of a crystalline field level was excluded by both INS and IR measurements, and by ab-initio calculations. Consequently, the electroactivity of this electromagnon is of a different nature than those observed in the past, and seems to have a purely electronic origin.

**References:**

a[1] A. B. Sushkov, R. V. Aguilar and S. W. Cheong, Phys. Rev. L 98, 3 (2007) [2] P. Rovillain, Cazayous and A. Sacuto, Phys. Rev. B 86, 014437 (2012) [3] S. Petit, V. Balédent and P. Foury-Leylekian, Phys. Rev. B 87, 1 (2013) [4] L. Chaix P. Roy and V. Simonet, Phys. Rev. L 112, 1 (2014) [5] H. Katsura, A. V. Balatsky and N. Nagaosa, Phys. Rev. L 98, 1 (2007) [6] V. Balédent, M. B. Lepetit, and P. Foury-Leylekian, Phys. Rev. B 97, 5 (2018) [7] V. Balédent, S. Chattopadhyay and P. Foury-Leylekian, Phys. Rev. L 114, 117601 (2015) [8] B. Khannanov, E. I. Golovenchits and V. A. Sanina, J.Phys.: Conf. Series 572, 12046 (2014)

**Images:**

![](https://assets.underline.io/uploads/markdown_image/1/image/bd2e503f7f47b740d4f6a1478408d52f.jpg)
Inelastic polarized neutron scattering at Q=(3/2,0,0) performed at THALES@ILL for 15 K. NSFx, SFx SFy and SFz stands for Non-Spin-Flip, Spin-flip with polarisation parallel to a, c and b crystallographic axis respectively.

![](https://assets.underline.io/uploads/markdown_image/1/image/bd2ea5b97909165cb1bda30c587ddab0.jpg)
Absorbance at 5 K with a reference at 40 K in Gd for e//a h//b & e//b & h//a (electric and magnetic field of the light) with respect to crystallographic axis.

 Study of a new type of excitation in the multiferroic magnetoelectric compound GdMn2O5, the electromagnon

Integrated magnetic devices show great potential for applications of radiofrequency and microwave integrated circuits, such as chip-scale magnetic inductors, bandpass filters, phase shifters, etc. [1, 2]. Recently, many researchers have been vastly concentrating on ferromagnetic/insulator sandwich structure for high-frequency applications. This thin film stack reduces the eddy-current as well as the out-of-plane anisotropy [3, 4], reducing the energy dissipation in form of heat. The FeGaB has soft magnetic nature, a large magnetostriction coefficient, and saturated at lower magnetic fields. Due to the low loss property of Al2O3 in microwave frequency, we have attempted to obtain low damping (α) and narrow linewidth through the FeGaB/Al2O3 thin film stack and characterize the ferromagnetic resonance (FMR) by utilizing Cryo-FMR. In this article, Ta (5)/ [FeGaB (15)/ Al2O3 (3)] n/Ta (3) / Al2O3 (2) thin-film stack are deposited by DC and RF sputtering on Si substrate, where n indicates the number of multilayer (n=1, 5, and 10) and parenthesis indicates thickness in nanometers. Fig.1 (a) shows normalized magnetic loop of the thin film stack, which reveals that the magnetic saturation moment has increased for increasing layer thickness. The coercive filed (Hc) has a low value below 3 Oe as indicated in Fig.1 (b). The dynamical magnetic properties have been studied using NanOsc Instruments Cryo-FMR in VersaLab system with temperature variation 300-100 K and exciting frequency from 2 to 20 GHz. The ferromagnetic absorption spectrum is shown in Fig.2 (a). The absorption spectrum contains the perpendicular standing spin wave (PSSW) modes for n=5 and n=10 in addition to FMR mode. The linewidth of absorption spectrum is fitted with dH= dH0 + αf/γ, (d=delta) to obtain damping factor (α), and inhomogeneous line width (dH0)[5]. The thin film stack with n=1 has a narrow linewidth and a low damping factor compared to other two thin film stacks, as shown in Fig.2 (b). When the temperature decreases, the damping factor has decreased for n=1, but increased for n=5 and n=10 due to defects arise for lowering the temperature. The Fig.2 (c) shows temperature dependent inhomogeneous linewidth for thin film stacks. As the temperature lowering to 100 K, the dH0 has increased for n=1 and n=10 thin film stacks, whereas decreased for n=5. The dH0 is lower for n=1 and larger value for n=5. The frequency dependence of FMR fields (Hres) plot provides effective magnetization (Meff) and magnetic anisotropy, when the plot has fitted to the Kittle equation. The Meff has a lower value for n=10 than other two stacks. The Meff has increased in all thin film stack as the temperature decreases. Since the absorption of the spectrum shows the PSSW modes in additional to FMR mode, the exchange magnetic fields (Hex) have been calculated by the subtraction method [6]. The plot of Hex and number of PSSW mode has provided the spin-wave exchange stiffness (D). The D values are 13.6615 and 10.7371 x10-18 Tm-2 for n=5, and n=10, respectively at 300 K. As the temperature decreases to100 K, the D has decreased for n= 5 thin film stack whereas increased for n=10. The temperature-dependent stiffness is following the T2 model in a linear form, and it suggests the existence of the electron–magnon interaction in this thin film stack. In order to calculate the constant of spin-wave exchange stiffness (A), saturation magnetization of thin film stack has multiplied with the exchange stiffness. At 300K, the A values are 7.1108, and 5.3213 pJ/m, respectively. In summary, we have deposited multilayer film stack of Ta (5)/ [FeGaB (15)/ Al2O3 (3)] n/Ta (3) / Al2O3 (2) onto Si substrates, and studied magnetodynamic (ferromagnetic resonance) properties with multilayer thickness and temperature dependence. The magnetic properties of thin film stack reveal that saturation magnetic moment has enhanced for increasing thickness of thin film stack, and Hc has obtained below 3 Oe. Therefore, the multilayer thin film stacks have the most significant for microwave applications. The dynamic magnetic properties suggest that the lowest damping and narrow linewidth are obtained for n=1 thin film stack compared to other thin film stacks. The FMR absorption spectrum of thin film stack with n=5, and n=10 show the PSSW modes in addition to the FMR mode. Consequently, the spin-wave exchange stiffness has been extracted. The temperature-dependent stiffness has followed the T2 model, which provides the strong presence of electron–magnon interaction in materials. The constant of spin-wave exchange stiffness has calculated with help of saturation magnetic moment and the exchange stiffness. Hence, the multilayer thin film stacks can be employed for microwave device applications.

**References:**

[1] Y. He, et.al,” IEEE Trans. on Magn., vol. 54, no. 9, pp. 1-4, 2018. [2] T. Wu, et.al, IEEE Magn. Lett., vol. 2, pp. 6000104-6000104, 2011. [3] X. Xing, et.al, IEEE Trans. on Magn., vol. 47, no. 10, pp. 3104-3107, 2011. [4] T. Wu, et.al, Appl. Phys. Lett., vol. 98, no. 1, pp. 012504, 2011. [5] K. Yadagiri, & T. Wu,J. Magn. Magn Mater., vol. 515, pp. 167277, 2020/12/01/, 2020. [6] K. Yadagiri, et.al, IEEE Trans. on Magn., vol. 57, no. 1, pp. 1-7, 2021.

**Images:**

![](https://assets.underline.io/uploads/markdown_image/1/image/958b8bb25594b09daba2dc26ecee751d.jpg)
Fig.1 (a) Normalized M-H loop for Ta (5)/ [FeGaB (15)/ Al2O3 (3)] n/Ta (3) / Al2O3 (2); n=1, n=5 and n=10, inset of Fig.1 (a) shows the Hc of thin film stacks, (b) coercive field HC (Oe) and Mr/Ms as a function of number of multilayer (n).

![](https://assets.underline.io/uploads/markdown_image/1/image/f01c6f6422c349af3f382d094dd43e8f.jpg)
Fig.2 (a) The absorption spectra of Ta (5)/ [FeGaB (15)/ Al2O3 (3)] n/Ta (3) / Al2O3 (2); n=10 at 300 K with exciting frequencies 2-20 GHz; inset of figure shows absorption spectra at 15 GHz with FMR mode and PSSW modes, (b) the damping factor (α) versus termeapture for thin film stack, and (c) the inhomogeneous linewidth (deltaH0) as a function of temperature for Ta (5)/ [FeGaB (15)/ Al2O3 (3)] n/Ta (3) / Al2O3 (2).

 Magnetodynamic properties of FeGaB/Al2O3 multilayer thin film stack for microwave applications

Magnetic materials are at the hearts of billions of high frequency communication devices and on-chip integration of magnetic microwave devices is of paramount importance for efficient utilization of the huge electromagnetic (EM) spectrum beyond 5G. Magnonics – a sub-field of spintronics – employs magnetic waves and oscillations in magnetic thin films and patterned nanostructures for processing GHz to sub-THz EM signal [1-2]. Magnetic ferrites are conventionally used in such magnetic microwave devices. However, due to the need for external bias magnetic field and incompatibility with complementary metal oxide semiconductor (CMOS) process, ferrites are not suitable for miniaturized microwave components. On the other hand, metallic magnetic thin films and nanostructures have great potential for efficient engineering of microwaves. Although, the need for bias field remains a bottleneck. We have been working on solutions where the requirement of the bias-field is absent. We have successfully demonstrated several strategies for achieving such device architecture by utilizing self-biased nanomagnets with reprogrammable remanent states which are associated with distinct magnetization dynamics [2-11]. To illustrate reconfigurable microwave properties, we have shown in Fig 1 as an example, a pair of dipolarly coupled rhomboid nanomagnets (RNMs). Two different remanent states (↑↑ and ↑↓) are obtained by utilizing a simple field initialization (HI) scheme: applying a saturating field along short and long axes of the RNMs with subsequent removal of the field (Fig. 1(a)). Due to different distributions of the stray fields for the two remanent states, the effective fields (Fig. 1(b)) are different which manifest in the shift of the microwave spectra for the two remanent states. Measurement of microwave properties from a single isolated nanomagnetic pair was obtained by using micro-focused Brillouin light scattering (micro-BLS) experiment (Fig. 1(c)) and the micro-BLS spectra are shown in Fig. 1(d). This strategy for reconfigurable magnonic responses is expanded to 2D arrays of the RNMs for different magnetic configurations: antiferromagnetic-type and ferrimagnetic-type artificial magnonic crystals. The maximum amount of the shift in the microwave responses is achieved by using multilayer isolated as well as dipolar-coupled networks of RNMs. Beyond, RNM shape, we have also demonstrated reconfigurable microwave operation based on other strategic shapes with multiple remanent states, such as arrow-, C-, L- and S-type nanomagnets and zigzag nanowires. Furthermore, we emphasize that such self-biased nanomagnets are also suitable for magnonic device applications where the propagation and gating of magnons – quanta for spin waves – are of primary interest. Conventionally, a large bias magnetic field is used in order to satisfy an orthogonal orientation between spin wave propagation and magnetization direction for efficient transmission characteristics (known as surface geometry or Damon-Eshbach mode). The need of such bias field hinders device integration on-chip for magnonic waveguides. Therefore, we have proposed and demonstrated a self-biased magnonic waveguide based on dipolar-coupled but physically separated chain of RNMs (Fig. 2). A prototype of such magnonic devices is shown in Fig. 2(a) where a microwave antenna (input) excites the spin waves and the spin waves are detected using micro-BLS technique with a focused laser spot (output). In order to gate the spin wave propagation, we have flipped a RNM in the waveguide and it is marked as ‘gate’ in the scanning electron microscopy (SEM) image (Fig. 2(b)). Such waveguide design enables one to have two different remanent states: ferromagnetically ordered (FO) and FO with a defect (FO*) when field initialized (same scheme as mentioned above) along the width and the length of the waveguide, respectively. FO and FO* remanent magnetic states are shown in Fig. 2(b) using magnetic force microscopy (MFM) images. Gating of the spin wave propagation has been shown by plotting 2D spatial profiles of the spin wave intensities using micro-BLS technique which clearly indicate gated magnon flow for the FO* state in comparison to the FO state. Thus, one can also achieve binary gating of output signal amplitude using such self-biased nanomagnetic waveguide which is remarkable as it also eliminates the need for bias field. In this presentation, we would like to emphasize the need for bias-free reconfigurable magnonics along with recent developments and future roadmap.

**References:**

[1] A. Haldar, C. Tian and A. O. Adeyeye, Sci. Adv. 3, e1700638 (2017). [2] A. Haldar, D. Kumar and A. O. Adeyeye, Nat. Nanotech. 11, 437 (2016). [3] A. Haldar and A. O. Adeyeye, ACS Nano. 10, 1690 (2016). [4] A. Haldar and A. O. Adeyeye, Appl. Phys. Lett. 108, 192404 (2016). [5] A. Haldar and A. O. Adeyeye, Appl. Phys. Lett. 108, 022405 (2016). [6] A. Haldar and A. O. Adeyeye, J. Appl. Phys. 123, 243901 (2018). [7] K. Begari and A. Haldar, J. Phys. D: Appl. Phys. 51, 275004 (2018). [8] K. Begari and A. Haldar, J. Phys. D: Appl. Phys. 52, 335003 (2019). [9] K. Begari and A. Haldar, J. Phys. D: Appl. Phys. 53, 455005 (2020). [10] A. Haldar and A.O. Adeyeye, Appl. Phys. Lett. 116, 162403 (2020). [11] A. Haldar and A. O. Adeyeye, J Appl. Phys. 128, 240902 (2020).

**Images:**

![](https://assets.underline.io/uploads/markdown_image/1/image/d84bb5df54b8e122952d84e6e7237690.jpg)
Figure 1: (a) Reprogrammable remanent states (X- and Y-state) of a RNM network. (b) Stray field distributions showing difference in the effective field for the two remanent states. (c) Measurement scheme from a single RNM network using micro-BLS technique. (d) Reconfigurable microwave responses for the two different remanent states.

![](https://assets.underline.io/uploads/markdown_image/1/image/408cd838fb91485f6c26689a83c74bc1.jpg)
Figure 2: (a) Propagating spin wave measurement scheme from a RNM chain using micro-BLS technique. (b) Logic device based on dipolar coupled RNM chain is shown in the SEM image. MFM images show the two different magnetic states (FO and FO*). Logic operation is demonstrated by using 2D spatial maps of the spin wave propagation for the FO and FO* states.

 Reconfigurable magnonics using self-biased reprogrammable nanomagnetic structures INVITED

Magnonics where magnons, the associated quasiparticles of spin waves are used to perform advanced logical or analogue functions represents a promising paradigm for new means of information processing for beyond complementary metal oxide semiconductor (CMOS) based architectures. Furthermore, in view of the emerging nano-magnonics which can eventually lead to a large-scale integration of magnonic devices new approaches for a compact, reconfigurable operation of magnonic device are required. Although such a reconfigurability is key to any efficient application such as magnonic filters, it has hitherto only theoretically [1] but not experimentally been shown on a sub-micron scale. On the other hand, multiferroics inherently offer compact means to implement voltage-control and reconfigurability in such microscopic magnonic systems [2]. In this work, we combine multiferroics with ferromagnets by employing a fully epitaxial heterostructure to achieve such voltage-controlled and reconfigurable magnonic system. Frequency filtering is most often based on magnonic crystals (MC) where a periodic modulation of the spin wave’s energy landscape results in frequency bandgaps in the magnon spectrum [3,4]. Here, we employ a thin-film multiferroic bismuth ferrite BiFeO3 (BFO)- lanthanum strontium manganite La2/3Sr1/3MnO3 (LSMO) heterostructure with a low Gilbert damping ~6x10-3 for the LSMO layer. Instead of modulating the sample shape, we create a MC by imprinting a periodic remnant electrical polarization in the multiferroic BFO layer which modulates the effective magnetic field in the ferromagnetic layer. We characterize the spin wave propagation spectra in this artificial, voltage-induced magnonic crystal and demonstrate the occurrence of a robust magnonic bandgap with &gt;20 dB rejection, i.e. filtering. The imprinted ferroelectric domains can be reconfigured repeatedly and are long-lived. Our results open a new path for the transduction and configurable filtering of spin waves within a single magnonic device. In general, bridging between between the scientific fields of functional oxides and magnonics, it opens new perspectives for the development of beyond CMOS based technologies [5].

**References:**

[1] Q. Wang , A. V. Chumak, L. Jin et al., Phys. Rev. B. 95, 134433 (2017) [2] N. Spaldin and R. Ramesh, Nat. Mat. 18, 203-212 (2019) [3] P. Frey , A.A. Nikitin, D. Bozhko et al., Commun Phys 3, 17 (2020) [4] A. V. Chumak, A. A. Serga and B. Hillebrands, J. Phys. D: Appl. Phys. 50, 244001 (2017) [5] H. Merbouche, I. Boventer, V. Haspot et al., submitted

**Images:**

![](https://assets.underline.io/uploads/markdown_image/1/image/243806faa5791ce2503db35d16d621e6.jpg)
Schematics of the experimental setup shown for the periodic ferroelectric domain state for the BFO layer.

![](https://assets.underline.io/uploads/markdown_image/1/image/4e0581eb0df6af24b4bcc64b618bde47.jpg)
Evidence for the bandgap for the BFO/LSMO heterostructure with periodic domains in the BFO in the amplitude representation in logarithmic scale. The data is normalized by the mean amplitude. At the bandgap frequency, the amplitude is dropping by >20 dB. A reconfiguration of the periodic domain state to the uniform down state erases the gap and reconstitutes the initial uniform state's response. Pn,m label the different domain states where n is the cycle number (1 cycle : down (virgin)-> up-> periodic-> ) and m the direction of the ferroelectric polarization.

 Voltage-Controlled Magnonic Crystal at the Sub-Micron Scale

In the last decades, the investigation of microwave spin waves (SW) in thin magnetic films has been an attractive field of research due to their very short wavelength reaching down to nanometers at GHz frequencies. SW also have a rich dispersion relation, that depends on their propagation direction with respect to a magnetic field. Hence, SW -based devices are promising candidates for microwave information processing, and eventually for overcoming the limitations encountered with CMOS-devices such as power consumption. In magnonic devices, a key challenge is to achieve long propagation distances of the spin waves. This requires to work with magnetic materials with small magnetic damping such as Yttrium Iron Garnet (YIG). (YIG), has by far the lowest magnetic damping which allows the spin waves propagation to be spread over millimeter-scale distances making it a reference material for spin-wave dynamics studies. Kajiwara and al. [1] confirmed that an electric current injection in a YIG film can be converted into spin-waves. In fact, when placing a non-magnetic metal with a high spin orbit coupling such as platinum on top of a magnetic insulator such as YIG, flow of charge current in the metal leads to its conversion into a pure spin current (via the Spin Hall Effect), which is then injected in the magnetic insulator where it adds up as an additional torque on the YIG’s magnetization. The above-mentioned mechanism of STT generation originating from pure spin-current obtained through the spin-orbit interaction of a heavy metal is called spin orbit torque (SOT). Recent studies [2], [3] with Platinum stripes placed on top of a 20 nm thick YIG waveguide (Pt/YIG) showed that SOT could generate a full compensation of the damping, leading to auto-oscillations of the magnetization above a critical injected current density. When excited with a microwave field, spin waves propagation length was increased by a factor of 10 in the bilayer. However, the possibility to achieve an amplification of propagating spin-waves was not observed yet, due to the onset of nonlinear dissipative processes above the critical current. More recently, materials having perpendicular magnetic anisotropy have demonstrated a full damping compensation along with SW emission, such behavior is accomplished by substituting Bismuth in YIG [4]. Here, we report on a study of SW propagation driven by SOT in a Pt/Bi-YIG waveguide where SW are detected using micro-focused Brillouin Light Scattering spectroscopy (BLS). The 20 nm thick and 21 µm long Bi-YIG waveguide is covered with 7 nm of Platinum. The experiments show the lossless propagation of a pulse of spin-waves excited by an rf-field generated by an antenna placed on top of the waveguide, under zero or positive damping during the course of 200 ns. Here, we show the full scale micromagnetic simulations done to emulate this experiment. The result shows a verification of the lossless propagation of SWs as seen in the experiment. By including a realistic Gilbert damping parameter and the Slonczewski torque to the Bi-YIG structure to compensate the magnetic losses, the results reproduce not only the steady state dynamics but also transitional dynamics when the rf-excitation is turned on and off at the ns time-scale. In summary, the inclusion of STT in the micromagnetic simulations showed an accordance with the experimental results, which verifies the ability to obtain a lossless propagation of SWs in the Bi-YIG waveguide and their non-suppression even after a hundred of nanoseconds after cutting off the excitation field.

**References:**

[1] Y. Kajiwara et al., “Transmission of electrical signals by spin-wave interconversion in a magnetic insulator,” Nature, vol. 464, no. 7286, Art. no. 7286, Mar. 2010 [2] M. Collet et al., “Generation of coherent spin-wave modes in yttrium iron garnet microdiscs by spin–orbit torque,” Nat. Commun., vol. 7, no. 1, Art. no. 1, Jan. 2016 [3] M. Evelt et al., “High-efficiency control of spin-wave propagation in ultra-thin yttrium iron garnet by the spin-orbit torque,” Appl. Phys. Lett., vol. 108, no. 17, p. 172406, Apr. 2016 [4] L. Soumah et al., “Ultra-low damping insulating magnetic thin films get perpendicular,” Nat. Commun., vol. 9, no. 1, Art. no. 1, Aug. 2018

**Images:**

![](https://assets.underline.io/uploads/markdown_image/1/image/2a26b6d8f9040dc1bef07c68aad030d7.jpg)
Fig.1: Two dimensional colormap showing the spin waves intensity in the simulated Bi-YIG waveguide after 80 ns from turning the rf-excitation on.

![](https://assets.underline.io/uploads/markdown_image/1/image/ea7d4882c8bc05ce1b677bad5a334f51.jpg)
Fig.2: Two dimensional colormap showing a lossless propagation of the SW intensity under SOT amplification, 100 ns after emission. The extracted group velocity is 55 m/s. The intensity of z-component of the magnetization is identical to that of Fig.1 recorded at a timestamp of 80 ns.

 Micromagnetic Simulations of Spin Waves propagation by SOT in a Bi-YIG waveguide

Conversion of heat into spin current has recently drawn a lot of attention due to its potential usage on heat recovery via the spin Seebeck effect (SSE) [1] as well as fundamental interests including generation of spin super currents [2] and Bose-Einstein condensation of magnons [3]. The underlying mechansim for SSE is understood as a combination of interfacial and bulk (intrinsic) contriubtion. The interfacial SSE is driven by the tempearture drop at the interface between metal and magnetic layer while the bulk SSE is driven by diffusive thermal magnons. In the nonlocal spin transport measurements, the long range exponential decay of thermally generated spin signal was explained solely by the intrinsic SSE [4] while there are several studies showing that there is another decay mechanism in subhundred nanometer scale [5-7]. Here, we study the spatial decay of thermally excited magnon currents inside a thin magnetic insulator by focusing on the short-range behavior in the nonlocal geometry. We compare SSE signal (Δ1,2) on the same device before and after adding a nonmagnetic Al capping layer. The Al capping alters the thermal profile significantly near the heat source, which results in that the nonequilibrium thermal magnon profile deviates from an exponential decay and shows two sign reversals within 1 μm from the heat source. From simulated temperature profiles, we find the vertical temperature gradient also reverses twice within the same length scale in the Al deposited case. The correlation suggests that thermal magnons behave locally at this length scale. Using a phenomenological spin transport model, we calculate the spatial profile of nonequilibrium spin density. The observed short-range behavior can be accounted if the relavant decay length is on the order of subhundred nanometers. Our result shows the existance of the shorter decay length in the nonlocal magnon transport experiments. This suggests that the long-range SSE signal in previous measurements should be attributed to magnons with lower energy.

**References:**

[1] S. R. Boona, R. C. Myers, and J. P. Heremans, Energy & Environmental Science, 7, 885 (2014). [2] D. A. Bozhko, A. A. Serga, P. Clausen, et. al., Nature Physics, 12, 1057 (2016). [3] Y. Tserkovnyak,S. A. Bender, R. A. Duine, et. al., Physical Review B, 93, 100402. (2016). [4] L. J. Cornelissen, J. Liu, R. A. Duine, et. al., Nature Physics, 11, 1022 (2015). [5] A. Kehlberger, U. Ritzmann, D. Hinzke, et. al., Physical Review Letters, 115, 096602. (2015). [6] A. Prakash, B. Flebus, J. Brangham, et. al., Physical Review B, 97, 020408. (2018). [7] J. S. Jamison, Z. Yang, B. L. Giles, et. al., Physical Review B, 100, 134402. (2019).

**Images:**

![](https://assets.underline.io/uploads/markdown_image/1/image/71fcbac32cc3cda92f6d18b1228e05ae.jpg)
Fig.1: Comparison of the experimentally measured local and non-local SSE signals (Δ1,2) at 1 and 2 mA injection and the numerically calculated chemical potential profiles with different input values of decay length λ for YIG/Pt/Si3N4 (case A) on a symmetric log scale. The insets show the schematic of the sample structure (bottom left) and temperature profile (top right).

![](https://assets.underline.io/uploads/markdown_image/1/image/898460b298a266d784aeaeab1dbbca16.jpg)
Fig.2: Same plots as Fig.1 for YIG/Pt/Si3N4/Al (case B). The result with λ = 0.1 µm reproduces the measured double crossing as shown, revealing the short range thermal magnon diffusion. The second crossing does not appear anymore when increasing λ to 0.5 or 1 µm. The insets show the schematic of the sample structure (bottom left) and temperature profile (top right).

 Short-range Thermal Magnon Diffusion in Magnetic Garnet

Spin waves are collective excitations of a spin system, where the magnetic moments precess with a spatially periodic phase shift that determines the wavelength [1]. From a technological point of view, the use of spin waves as signal carriers in future spintronic logic and memory devices has the advantages of a potentially lower power consumption and improved miniaturization prospects compared to the present charge-based CMOS technology [2]. For achieving nanoscale integration, topological spin textures such as magnetic vortex cores can be driven by alternating magnetic fields to excite coherent spin waves with short, sub-micrometer wavelengths, as shown recently for both synthetic ferrimagnetic systems and single ferromagnetic layers [3, 4]. In this contribution, we will demonstrate the versatility of this excitation mechanism by showing that the application of a static magnetic bias field makes the vortex core a manipulable spin-wave source that can be continuously translated in space without altering the resulting spin-wave dispersion relation [5]. We employ time-resolved scanning transmission x-ray microscopy (TR-STXM) to directly image vortex-core spin textures as well as the emission and propagation of spin waves from them. By applying a static in-plane magnetic bias field, we can continuously displace the vortex core from its equilibrium position in the center of a structured element. For a synthetic ferrimagnet (Co/Ru/Ni81Fe19), it is possible to shift the core relatively close to the edge while maintaining a radial, isotropic spin-wave emission pattern (see Figure 1). The wavelengths of the excited spin waves are not noticeably changed by the applied magnetic bias field, but remain tunable by the driving frequency in the same way as without bias field. For a single Ni81Fe19 layer with a thickness of the order of 100 nm, we observe a similar behavior for relatively small magnetic bias fields but when the vortex core approaches the edge of the disk, it also expands (see Figure 2). This expansion can be seen as a transformation from a zero-dimensional point source to a one-dimensional curved object where its shape and extent is controlled by the magnitude of the applied magnetic bias field. Such an expansion leads to a directional emission of the spin waves that propagate away from the core in mainly two opposite directions rather than radially as observed for non-expanded cores. For certain combinations of magnetic bias fields and driving frequencies, focusing effects appear. Such effects provide promising means for a controlled and directional propagation of spin waves without the need for additional patterning or waveguides.

**References:**

[1] A.G. Gurevich and G.A. Melkov, Magnetization Oscillations and Waves. Vol. 1, CRC Press (1996) [2] A.V. Chumak et al., Nat. Phys. Vol.11, p.453-461 (2015) [3] S. Wintz et al. Nat. Nanotechnol. Vol.11, p.948-953 (2016) [4] G. Dieterle et al., Phys. Rev. Lett. Vol.122, p.117202 (2019) [5] S. Mayr et al., Nano Lett. accepted (2021)

**Images:**

![](https://assets.underline.io/uploads/markdown_image/1/image/1ab356e578b624eb1b59b9aeeb5b4f0a.jpg)
Fig. 1: TR-STXM snapshots of a synthetic ferrimagnet, taken with out-of-plane magnetic sensitivity at the Fe L3 edge at a magnetic bias field of 5 mT. (a) Absolute absorption image showing the black vortex core shifted from the center of the structure (green circle). (b) Normalized contrast displaying the radial spin-wave emission from the core for a 1 GHz alternating field excitation.

![](https://assets.underline.io/uploads/markdown_image/1/image/2d254de94235b1183f441f680c4fed5a.jpg)
Fig. 2: TR-STXM images with out-of-plane sensitivity displaying the response of the vortex in a Ni81Fe19 layer to a magnetic bias field of 22.5 mT. (a) Absolute absorption image showing the expansion of the vortex core (encircled in green). (b) Normalized image displaying only the temporal changes with respect to the average magnetization, showing the spin waves emitted from the expanded vortex core in two directions for an excitation frequency of 8.6 GHz.

 Spin-wave emission from vortex cores under static magnetic bias fields

Synthetic antiferromagnets (SyAFs) are an excellent platform to explore novel spintronic and magnonic concepts with coupled magnetic moments[1,2]. Unlike a homogeneously magnetised single-layer ferromagnet, coupled magnetic layers are able to offer rich magnetic states where the competition between interlayer exchange coupling, external-field-induced Zeeman interaction as well as other magnetic anisotropy terms plays a role. This can expand into their dynamic regimes as the coupled moment nature inherently provides two eigenmodes, acoustic and optical modes[3-7], where time-dependent components of two coupled moments are oscillating in-phase (acoustic) and out-of-phase (optical)in a spin-flop (canted) regime. When we excite spin-waves by oscillating magnetic fields, the spatial symmetry/profile of microwave excitation determines which spin-wave modes are excited. In the simplest case, a uniform distribution of microwave excitation fields across a magnet can excite the uniform spin-wave mode (wavevector k = 0) as well as higher-order spin-wave resonance modes (k &gt; 0) with odd index numbers since the spatial profile of the microwave excitation and spin-wave mode amplitude (with phase) matches to each other in terms of symmetry. This in turn suggests that it should be possible to control the selection rules of spin-wave excitations by designing the spatial profile of excitation fields. Accessing hidden spin-wave states as well as tailoring spin-wave excitation efficiency by this approach has not been much explored in the past. We report in this work the current-induced-torque excitation of acoustic and optical modes in Ta/NiFe/Ru/NiFe/Ta synthetic antiferromagnet stacks grown on SiO2/Si substrates(Fig 1(a)). The two Ta layers serve as spin torque sources with the opposite polarisations both in spin currents and Oersted fields acting on their adjacent NiFe layers. This can create the odd symmetry of spatial spin torque distribution across the growth direction, allowing us to observe different spinwave excitation efficiency from synthetic antiferromagnets excited by homogeneous torques. We analyse the torque symmetry by in-plane angular dependence of symmetric and anti-symmetric lineshape amplitudes for their resonance (as shown in Fig 2(a-b)) and confirm that the parallel (perpendicular) pumping nature for the acoustic (optical) modes in our devices, which is in stark difference from the modes excited by spatially homogeneous torques. The terms parallel (perpendicular) pumping here signify that the torque (hence spin-wave excitation efficiency) is maximised when the oscillating fields and dc magnetic field are collinear(perpendicular) to each other respectively. This is evident from Fig 2(c-d) that strongly suggests the torque symmetry for the acoustic(optical) mode excitation being of form sin Φ(cos Φ) . We also generalize macrospin model for this particular spin-torque excitation geometry, which excellently supports our experimental observation. Our results offer capability of controlling spin-wave excitations by local spin-torque sources and we can explore further spin-wave control schemes based on this concept. We envisage that the control of spin-wave excitations in STT nanodevices will be useful for future spintronic and magnonic nano-devices.

**References:**

[1] R. A. Duine, K.-J. Lee, S. S. P. Parkin, and M. D. Stiles, Nat. Phys. 14, 217 (2018). [2] A. V. Chumak, V. I. Vasyuchka, A. A. Serga, and B. Hillebrands, Nat. Phys. 11, 453461 (2015). [3] F. Keffer and C. Kittel, Phys. Rev. 85, 329 (1952). [4] J. J. Krebs, P. Lubitz, A. Chaiken, and G. A. Prinz, J. Appl. Phys. 67, 5920 (1990). [5] S. M. Rezende, A. Azevedo, and R. L. Rodrguez-Surez, J. Appl. Phys. 126, 151101 (2019). [6] A. Sud, Y. Koike, S. Iihama, C. Zollitsch, S. Mizukami, H. Kurebayashi, arXiv:2012.12999 . [7] A. Sud, C. W. Zollitsch, A. Kamimaki, T. Dion, S. Khan, S. Iihama, S. Mizukami, and H. Kurebayashi, Phys. Rev. B 102, 100403(R) (2020).

**Images:**

![](https://assets.underline.io/uploads/markdown_image/1/image/b406e148e73969fbd07f012bcc683fa7.jpg)
FIG. 1: (a) A schematic of the sample geometry used for STT-FMR measurements in our study. (b) V obtained at different frequencies for acoustic and optical modes in our device for Φ = 55ο (the values for some frequencies have been scaled to show them properly). (c) A 2D colorplot of V as a function of applied field and frequency, measured for Φ = 55ο . (d) Numerical results calculated for the experimental conditions as in (c).

![](https://assets.underline.io/uploads/markdown_image/1/image/aaeaf82e3f75f0759e29b390a7923bff.jpg)
FIG. 2: (a-b) Angular dependence of symmetric (Vsym) and antisymmetric (Vasy) components of voltage for (a) acoustic mode measured at 8 GHz and (b) optical mode measured at 8.5 GHz. The solid curves are obtained using theoretical model. (c-d) The symmetry of torques obtained by dividing the Voltage by (c) sin 2Φ for acoustic mode and (d) cos 2Φ for optical mode. The dominant sin Φ (cos Φ) dependence of torques confirms the parallel(perpendicular) pumping configuration.

Antiferromagnetic Artificial Neural Networks with Symmetric Coupling

Downloads

Next from INTERMAG 2021

Topological magnon-polaron in a two-dimensional ferromagnet

Similar lecture

Field Computation in Media Exhibiting Hysteresis Using Hopfield Neural Networks

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Antiferromagnetic Artificial Neural Networks with Symmetric Coupling

.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 INTERMAG 2021

Topological magnon-polaron in a two-dimensional ferromagnet

Similar lecture

Field Computation in Media Exhibiting Hysteresis Using Hopfield Neural Networks

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads