United States

Artificial spin ices (ASIs) are lithographically patterned arrays of single domain magnetic nanoislands where the dipolar interactions between these Ising-like macrospins have served as the basis for investigating phenomena such as geometric frustration and emergent magnetic monopoles [1]. While the majority of ASI studies have utilized nanoislands composed of soft magnetic metals (e.g. permalloy), the use of complex oxides, such as La0.7Sr0.3MnO3 (LSMO), could permit room temperature studies of thermally-active ASIs (TC ~ 360 K) and serve as a platform for coupling the rich physics of both complex oxide heterostructures and ASI structures. In this work, we utilize x-ray photoemission electron microscopy (X-PEEM) to image LSMO-based brickwork ASI structures where a breakdown of the Ising model of individual nanoislands occurs at a critical coupling strength which balances intra- and inter-island magnetic interactions. For ASI structures with nanoislands of identical dimensions, the ASI lattice parameter controls the inter-island coupling strength, resulting in systems composed solely of single-domain magnetic nanoislands, or a mixture of single-domain and complex spin textures (CST, i.e. single and double vortex states). Analysis of magnetic ordering in X-PEEM images reveals that nucleation of CST in frustrated nanoislands is accompanied by a disruption in long-range ordering. While it is generally assumed that inter-island interactions in ASI systems only influence the ordering of the nanomagnetic ensemble, in this LSMO-based ASI system inter-island coupling can dictate both the spin texture within individual nanoislands and long-range magnetic ordering. These results suggest the potential to create tunable nanoisland domain states, providing a new direction in developing ASI systems with exotic phase transitions as well as novel nanomagnetic logic architectures.References: 1. Skjærvø, S.H., et al., Nature Reviews Physics., Vol. 2, p.13 (2020)

MMM 2020

The Formation of Complex Spin Textures in La0.7Sr0.3MnO3 Artificial Spin Ice Arrays

The 65th Annual Conference on Magnetism and Magnetic Materials (MMM 2020) was held in a virtual form. The COVID-19 pandemic, with its associated uncertainty about international travel and the ability to gather in large groups, meant that the Steering Committee decided to cancel the face-to-face meeting that had been planned in Palm Beach, Florida November 2-6, 2020.

We were reluctant to give up the benefits of meeting face-to-face, and know that shifting to a virtual conference is a significant change for this community. Nevertheless, we are first and foremost concerned about the health and safety of all our Conference participants and their loved ones on returning home after travelling.

The MMM 2020 Steering Committee was working exceptionally hard, along with our sponsors, the IEEE Magnetics Society and AIP Publishing, to make this event serve the needs of our participants, who live in time zones all around the globe, in the best possible way. 

You need a ticket to explore this event.

If you have already registered for MMM 2020, and you have already created an account, then you
just need to login (above).

If you have already registered for MMM 2020, but you have not yet created an account, then you
need to click on &quot;create an account&quot; (above) and then login.

Forgot your password? Go to
https://www.underline.io/log-in

If you have not yet registered for MMM 2020 but would like to do so, go to
https://magnetism.org/register

Still need help? Contact our team!


technical paper

Progress has been made on emulating the neuron and the synapse using magnetic devices [1], [2] especially with three-terminal magnetic tunnel junctions (3T-MTJs) [3], [4]. Increasingly more accurate biological mimics are being developed with the 3T-MTJ like the Leaky-Integrate-Fire (LIF) [5] neuron displaying lateral inhibition, and the 3T-MTJ synapse model showing spike timing dependent plasticity (STDP) [6], [7]. However, more work must be done to achieve real-time machine learning. Using an earlier developed synapse circuit [8] and a 3T-MTJ spice model [9], we show the transient behavior of a 2 by 2 crossbar array (cartoon in Fig 1). There is STDP for different delay conditions between the pre- and post-synaptic neural spikes. The shorter the time difference between the onset of the pre- and post-synaptic neuron spike, the higher the current through the ferromagnet (Fig 2). Higher current leads to higher rate of change of the domain wall (DW) position per neural event and higher DW displacement. At t=0, the initial pair of presynaptic and postsynaptic pulses helps to randomly give the DWs a head start. DW for synapse S2 (red plot) rises faster because it has the least delay of 1 ns between its pre- and post-synaptic pulse despite its pulses lagging that of S1 whose DW had a head start. Scaling this up, the 3T MTJ synapse is robust in more complex systems such as a machine learning clustering task.References: [1] S. Lequeux et al., Sci. Rep., vol. 6, no. August, 2016, doi: 10.1038/srep31510. [2] J. Torrejon et al., Nature, vol. 547, no. 7664, pp. 428–431 (2017) [3] N. K. Upadhyay, H. Jiang, Z. Wang et al, Adv. Mater. Technol., vol. 1800589, pp. 1–13 (2019) [4] N. Hassan et al., J. Appl. Phys., vol. 124, no. 15, p. 152127 (2018) [5] G. Bi and M. Poo, Annu. Rev. Neurosci., vol. 24, no. 1, pp. 139–166 (2002) [6] H. Markram, W. Gerstner, and P. J. Sjöström, Front. Synaptic Neurosci., vol. 4, pp. 2–5 (2012) [7] O. Akinola, X. Hu, C. H. Bennett et al, J. Phys. D. Appl. Phys., vol. 52, no. 49 (2019) [8] X. Hu, A. Timm, W. H. Brigner et al, IEEE Trans. Elect. Dev., vol. 66, no. 6, pp. 2817–2821 (2019)

Online Training of Spiking Neural Networks Using Domain Wall Magnetic Tunnel Junction Synapses

Domain wall-magnetic tunnel junction (DW-MTJ) devices intrinsically emulate neuron [1] and synapse [2] functionalities for neuromorphic systems. While these devices efficiently emulate neurobiological components, significant CMOS overhead is required to integrate multiple perceptron layers [3]. This work proposes a four-terminal DW-MTJ neuron [4] that enables purely spintronic multilayer perceptrons that do not require CMOS. The proposed leaky integrate-and-fire neuron [5] in Fig. 1 has a ferromagnetic DW track between terminals T1 and T2 and an MTJ between terminals T3 and T4. The MTJ free layer is coupled to the DW track by an electrically insulated layer. Spin-transfer torque induced by the current through the T1-T2 path moves the DW from right to left to perform integration. DW leaking from left to right intrinsically occurs due to either dipolar magnetic field [1], anisotropy gradient along [6], or shape variation [7]. When the DW passes underneath the electrically-isolated MTJ, the free ferromagnet reorients itself to match the magnetization of the DW track due to dipolar coupling, thereby switching the MTJ between the anti-parallel resistive state and parallel conductive state. A constant voltage applied at the T3 terminal, fires by switching between low and high output currents. Additionally, by exploiting stray magnetic field interactions, these neurons perform lateral inhibition among a group of neighboring neurons [1]. In the multilayer perceptron of Fig. 2, each four-terminal DW-MTJ neuron is connected to the three-terminal analog DW-MTJ input synapses through the T1 terminal and is cascaded to the output synapses through the T4 terminal without any CMOS interface. Finally, an unsupervised learning algorithm results from feedback between the T4 terminals of the neurons and right terminals of the synapses, providing best results of 98.11% accuracy on the Wisconsin breast cancer clustering task [8].References: [1] N. Hassan, X. Hu, L. Jiang-Wei, et al., J. Appl. Phys., Vol. 124, no. 15, p. 152127 (2018) [2] O. Akinola, X. Hu, C. H. Bennett, et al., J. Phys. D: Appl. Phys., Vol. 52, no. 49, LT01 (2019) [3] A. Sengupta, Y. Shim, and K. Roy, IEEE TBioCAS, Vol. 10, no. 6, pp. 1152–1160 (2016) [4] W. H. Brigner, N. Hassan, X. Hu, et al., arXiv preprint arXiv: 2002.00862 (2020) [5] A. N. Burkitt, Biol. Cybern., Vol. 95, pp. 97–112 (2006) [6] W. H. Brigner, X. Hu, N. Hassan, et al., IEEE JXCDC, Vol. 5, no. 1, pp. 19–24 (2019) [7] W. H. Brigner, N. Hassan, L. Jiang-Wei, et al., IEEE TED, Vol. 66, no. 11, pp. 4970-4975 (2019) [8] A. Velasquez, C. H. Bennett, N. Hassan, et al., arXiv preprint arXiv: 2003.11120 (2020)

Purely Spintronic Multilayer Perceptron Enabled by Four-Terminal Domain Wall-Magnetic Tunnel Junction Neuron

Magnetic materials are already widely used for long-term data storage, but ongoing work seeks to use them as both working memory and computing architectures. Particularly, devices based on magnetic domain walls (DWs) can perform logic operations and can readily store information [1]. However, the stochasticity of DW pinning limits the feasibility of creating technologically viable devices. Here, we demonstrate how stochasticity can be changed from a technologically inhibitive behaviour into an integral property of machine learning algorithms for specialised neuromorphic devices. We first present experimental measurements that demonstrate the tunability of stochastic processes by applying external stimuli. Focused magneto-optic Kerr effect measurements were used to probe the pinning of DWs at notch-shaped defect sites in 400 nm wide Permalloy nanowires (Fig. 1). DWs were injected into the nanowires and propagated to the defect sites using a field parallel to the nanowire (Hx). The probability of the DWs being pinned at the defect sites was found to depend sigmoidally on the magnitude of a field transverse to the nanowires (Hy). The stochastic behaviour of the nanowire were then integrated into machine learning models. We have developed feedforward neural networks with the nanowires acted as binary stochastic synapses (BSS) that can be sampled by repeating the network. It is found that for a single cycle of the network it can be trained to recognise handwritten digits with an error rate of 19%. Repeated sampling of the network decreases the error rate to 9% for 128 cycles - close to mean field performance (Fig. 2). Adding a hidden layer improves performance with extra pathways for the stochastic behaviour to be averaged over. Our work illustrates how the stochasticity of DW devices can be harnessed to provide tunable bespoke hardware for machine learning tasks.References: [1] R. L. Stamps et al., J. Phys. D. Appl. Phys. Vol. 47, p. 333001 (2014)

Harnessing Stochastic Domain Wall Pinning for Machine Learning

Lateral inhibition (LI) is an important neuromorphic network functionality, modeled after the biological neuron behavior that a firing neuron deactivates its neighbors and prevents their firing. In most neuromorphic hardware platforms LI is implemented by external circuitry, thereby decreasing the energy efficiency and increasing the area overhead. Recently, the domain wall – magnetic tunnel junction (DW-MTJ) neuron is shown in modeling to be intrinsically inhibitory1. LI in DW-MTJ neurons results from magnetostatic interaction between adjacent neurons (Fig.1): the two DWs, DWN and DWI, are driven by electrical current of densities JeI<JeN, so that the DW velocities vDWI<vDWN and the active Neuron N fires first. DWI falls behind DWN and is subjected to the stray field of Neuron N in the −z direction. It is demonstrated that such stray field can decrease DWI velocity and prevent Neuron I from firing, thus emulating LI. However, the LI mechanism in DW-MTJ neurons has not been studied thoroughly, leading to weak inhibition only in very closely spaced devices. In our work2 we model the current- and field-driven DW motion in two adjacent DW-MTJ neurons to approach these problems. LI is maximized by an optimal magnetic interaction strength, determined by tuning the spacing between the two neurons (Fig.2). The results are explained by the current- and field-driven DW velocity characteristics and are quantified by an analytical model. We further study the influence of device parameters and demonstrate the LI behavior in an array of 1000 neurons, which renders our proposed DW-MTJ array promising for implementing a k-winner-take-all layer. With strong LI achieved, a path towards competitive learning algorithms are made possible on the DW-MTJ neurons. This work is funded by Sandia National Lab. Sandia National Lab (SNL) is managed and operated by NTESS under DOE NNSA contract DE-NA0003525References: 1. N. Hassan et al. Journal of Applied Physics, Vol.124, p.152127 (2018) 2. C. Cui et al. Nanotechnology, Vol.31(29), p.294001 (2020)

Maximized Lateral Inhibition in Paired Magnetic Domain Wall Racetracks for Neuromorphic Computing

Neuromorphic computing attempts to mimic the neurons and synapses in a human brain in order to provide significant improvements in the computation of unstructured, real-world data. In past research, we have proposed three separate leaky integrate-and-fire (LIF) neurons that provide the leaking, integrating, and firing characteristics without the use of any additional circuitry [1]-[3]. These neurons are therefore able to significantly reduce the area and energy overhead of neuromorphic circuits. To increase the biological accuracy of these neurons and to better match the neurons to different neuromorphic schema and algorithms, it is desirable for neuron leaking to implement specific mathematical functions in addition to exhibiting the three basic LIF neuronal functionalities. By varying the shape of the devices, it is possible to implement a variety of leaking characteristics. In this work, we discuss the implementation of linear leaking characteristics. In the neurons of [1]-[3], the leaking DWs accelerate towards the ends of the tracks. To counter this, it is necessary to alter these devices. In fact, by simply introducing slight exponential variations in the sides of the neuron from [3], as in Figure 1, we are capable of constructing neurons with considerably more linear leaking characteristics than those discussed in [1]-[3]. As the exponential variations become more pronounced, the linearity of the neuron’s leaking increases as well. Beyond a certain point, however, continued increases in the curvature causes a regression in the leaking functionality’s linearity. Figure 2 demonstrates the effect of the curvature of the DW track on the device’s leaking functionality, starting with no curvature and progressing to more pronounced curvature.References: [1] N. Hassan, X. Hu, L. Jiang-Wei et al., “Magnetic domain wall neuron with lateral inhibition,” J. Appl. Phys., vol. 124, no. 15, p. 152127, 2018. [2] W. H. Brigner, X. Hu, N. Hassan, et al., “Graded-Anisotropy-Induced Magnetic Domain Wall Drift for an Artificial Spintronic Leaky Integrate-and-Fire Neuron,” IEEE Journal on Exploratory Solid-State Computational Devices and Circuits, vol. 5, no. 1, pp. 19–24, 2019. [3] W. H. Brigner, N. Hassan, L. Jiang-Wei, et al., “Shape-Based Magnetic Domain Wall Drift for an Artificial Spintronic Leaky Integrate-and-Fire Neuron,” IEEE Transactions on Electron Devices, vol. 66, no. 11, pp. 4970-4975, 2019.

Linear Intrinsic Leaking in a Domain-Wall Magnetic Tunnel Junction Neuron

The relaxation kinetics and emerging correlations of arrays of interacting Ising-like nanomagnets are governed by the switching rates of the individual magnets. These rates depend, via the Arrhenius law, on the energy barrier for moment reversal in the nanomagnets. In this work, we consider how the switching behaviour of a nanomagnet in archetypical artificial square ice is modified by the interaction with its six nearest neighbours. We find that due to two effects the transition barriers can be significantly lowered compared to an approximation commonly used for kinetic Monte Carlo simulations: (1) Based on a simplified point-dipole model, we find that the barriers for clockwise and counter-clockwise uniform rotation can differ significantly due to the interactions with neighbours. For a subset of environmental microstates this leads to a significant enhancement of the switching rate, which is especially marked in the limit of rare events. (2) Micromagnetic simulations employing the string method to model the moment reversal in an extended nanomagnets reveal that, in addition to the clockwise/counter-clockwise splitting, the environments promote non-uniform reversal leading to further reductions of the switching barriers. These results highlight that local microstates can significantly enhance moment reversals already in ideal systems, without defects or additional anisotropies, and are particularly pronounced for fully-magnetised configurations. Thus, we expect that the initial stages of demagnetisation will proceed faster than assumed from previous models. These findings are of integral importance to achieve realistic kinetic Monte Carlo simulations of emergent correlations in extended artificial spin systems or the evolution of small-scale nanomagnetic logic circuits.

Switching barriers in artificial square ice

The dynamics involving directed separation of a monopole-antimonopole pair in two- or three-dimensional spin ices is conventionally defined as magnetic current or magnetricity. While the feasibility of magnetricity in artificial spin ices has been shown theoretically and experimentally [1,2], the controllability of magnetic currents is still under debate. Particularly, it has been theoretically shown that maximization of Dirac string’s tension is a necessary condition for stabilization of directed magnetic currents [1]. Experimentally, however, magnetic currents are often of only short duration, because magnetic monopoles created at one sample edge do not reach the other sample edge, but rather stick in the middle of the sample. Here, we explore theoretically the role of the termination of arrays and the higher order magnetic interactions for the directionality of magnetic currents in two-dimensional artificial spin-ice. Our Monte-Carlo calculations show that arrays of point dipoles and identical array with realistic multipole moments give qualitatively similar results with respect to the end states and with respect to the averaged propagation of the monopoles. However, the Monte-Carlo calculations using the higher order contributions reflect much better formation of domains with phase shifted Dirac strings, found in recent experiments [3]. Our calculations permit us to elaborate a phase diagram predicting the probability of directed propagation of monopole current as a function of termination type of a realistic artificial spin-ice array.References: [1] E. Y. Vedmedenko, Dynamics of Bound Monopoles in Artificial Spin Ice: How to Store Energy in Dirac Strings. Physical Review Letters 116, 077202 (2016). [2] A. Farhan et al., Direct Observation of Thermal Relaxation in Artificial Spin Ice. Phys. Rev. Lett. 111, 057204 (2013). [3] A. Hanu, E. Y. Vedmedenko, J. Cuia, J. Vijayakumard, A. Kleibertd, L.J. Heyderman, Control of emergent magnetic monopole currents in artificial spin ice, submitted.

Role of Higher-Order Magnetostatic Interactions for the Directional Injection of Monopole Currents

1D and 2D arrays of coupled magnetic nanoislands, also known as mesospins, have recently attracted much attention for offering the possibility to mimic exotic physical systems [1]. In analyzing these coupled mesospin arrays, the individual building blocks are often assumed to consist of simple dipoles, where any inner structure is ignored. Such an approximation of a mesospin however, leads to discrepancies when temperature becomes an important parameter [2]. Temperature excites thermal magnons in a ferromagnet, which is the basis for temperature dependence of the magnetization as expressed by the Bloch law. Here, we present a framework for studying temperature induced magnetization fluctuations in terms of magnonic excitations. We use MuMax3 to numerically subject a single stadium shaped mesospin (450 nm x 150 nm x 5 nm) to a highly inhomogeneous stochastic thermal field [3], and study the emerging spectrum by Fourier transforming the magnetization. The length, width and thickness of the mesospin element are varied to study the resulting effect of size variation on the spectra. The spectra are found to be gapped, and the spectral peaks pertain to specific edge- and center oscillation modes. Finally, we study the effect of temperature variation and we find that the magnon density of states plays an important role on the overall shape of the spectra. The results of this study are not only useful from the perspective of thermal switching of nanoislands, but also for the design of magnonic crystals based on mesospins.References: 1. S. H. Skjærvø, et al., Nature Reviews Physics, Vol. 2, p. 13–28 (2020) 2. M. Pohlit, G. Muscas, I. A. Chioar, et al., Physical Review B, Vol. 101, p. 134404 (2020) 3. J. Leliaert, J. Mulkers, J. De Clercq, et al., AIP Advances, Vol. 7, p. 125010 (2017)

A Magnon Approach to Thermal Fluctuations in Ising-like Mesospins

The extent of magnetic ordering in highly-frustrated thermally-active artificial kagome spin ice is limited by the lowest achievable blocking temperature, as the moments freeze in before ordering is achieved [1]. A more substantial degree of magnetic ordering can be achieved by lowering the blocking temperature of the individual nanomagnets or by increasing the critical transition temperature for the system. We pursue an original approach of introducing interfacial Dzyaloshinskii-Moriya interactions, which lowers the blocking temperature while keeping the transition temperature unchanged [2]. Using this approach, we demonstrate that a seven-ring kagome structure consisting of 30 nanomagnets can be thermally annealed into its ground state. Furthermore, the spin-ice correlations extracted from extended kagome lattices are found to exhibit the quantitative signatures of long-range charge-order, thereby giving experimental evidence for the theoretically predicted continuous transition to a charge-ordered state. We also find that slight modifications of the nanomagnets at a vertex in the artificial kagome ice lead to an increase in the critical temperatures, and allow us to control the energy landscape. Our results provide the foundations for the tuning of magnetic ordering in several different artificial spin systems.References: [1] S.H. Skjærvø, C.H. Marrows, R.L. Stamps, L.J. Heyderman. “Advances in artificial spin ice.” Nature Reviews Physics 2, 13-28 (2020). [2] K. Hofhuis et al., under review.

Long-range Magnetic Ordering in Artificial Kagome Spin Ice

Novel computational paradigms in combination with proper hardware solutions are required to overcome the limitations of our state-of-the-art computer technology, in particular regarding energy consumption. Due to the inherent complex and non-linear nature, spintronics offers the possibility towards energy-efficient, non-volatile hardware solutions for various unconventional computing schemes.[1-3] In this talk, I will address the potential of topologically stabilized magnetic whirls – so-called skyrmions – for in particular two unconventional computing schemes – reservoir computing and stochastic computing. Reservoir computing is a computational scheme that allows to drastically simplify spatial-temporal recognition tasks. We have shown that random skyrmion fabrics provide a suitable physical implementation of the reservoir [4,5] and allow to classify patterns via their complex resistance responses either by tracing the signal over time or by a single spatially resolved measurement. [6] Stochastic computing is a computational paradigm that allows to speed up a calculation while trading for numerical precision. Information is encoded in terms of bit-streams as a probability. A key requirement and simultaneously a challenge is that the incoming bitstreams are uncorrelated. The Brownian motion of magnetic skyrmions allows creating a device that reshuffles the bit-streams. [7,8]References: [1] J. Grollier, D. Querlioz, K. Y. Camsari, KES, S. Fukami, M. D. Stiles, Nature Electronics (2020) [2] E. Vedmedenko, R. Kawakami, D. Sheka, ..., KES, et al., J. of Phys. D (2020) [3] G. Finocchio, M. Di Ventra, K.Y. Camsari, KES, P. K. Amiri and Z. Zeng, arXiv:1907.04601 [4] D. Prychynenko, M. Sitte, et al, KES, Phys. Rev. Appl. 9, 014034 (2018) [5] G. Bourianoff, D. Pinna, M. Sitte and KES, AIP Adv. 8, 055602 (2018) [6] D. Pinna, G. Bourianoff and KES, arXiv:1811.12623 [7] D. Pinna, F. Abreu Araujo, J.-V. Kim, et al, Phys. Rev. Appl. 9, 064018 (2018) [8] J. Zazvorka, F. Jakobs, D. Heinze, ..., KES, et al., Nat. Nanotech. 14, 658 (2019)

Downloads

Next from MMM 2020

Online Training of Spiking Neural Networks Using Domain Wall Magnetic Tunnel Junction Synapses

Similar lecture

Nickel nanotubes with low resistivity and coherent magnetization dynamics for 3D spintronics and magnonics prepared by atomic layer deposition

.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 MMM 2020

Online Training of Spiking Neural Networks Using Domain Wall Magnetic Tunnel Junction Synapses

Similar lecture

Nickel nanotubes with low resistivity and coherent magnetization dynamics for 3D spintronics and magnonics prepared by atomic layer deposition

Downloads