Neuromorphic weighted sum with magnetic skyrmions
(2024) EHCI workshop of the Eindhoven University of Technology — Location: Eindhoven (Holland) (12.January.2024)
Files
No attached file found for this publication.
Details
- Authors
- Abstract
- Magnetic skyrmions are topological magnetic solitons that can be stabilized at room temperature in magnetic thin films or heterostructures with optimized magnetic anisotropy and Dzyaloshinskii-Moriya interaction. Recent experimental studies have shown, often separately, that magnetic skyrmions can be nucleated [1-4], moved [2,4], annihilated [3] and electrically detected using the anomalous Hall effect [5] or tunneling magnetoresistance [6-7]. Skyrmions have a wide span of appealing features that make them prime candidates for energy-efficient computing operations [8-10], such as stability at room temperature, deep sub-micron dimensions, non-volatility, particle-like behavior, and motion at low power. These characteristics align closely with the needs of neuromorphic computing, a discipline aiming to emulate neural network behaviors using in-memory computing to create energy-efficient, artificial intelligence (AI)-specialized hardware [11]. Recent reports have shown that skyrmions can serve various roles in neuromorphic circuits [8-12]. However, a fundamental neural network operation, the weighted sum of input neuron signals, is still missing in the context of skyrmions [11]. In our recent works [13-14], we propose an experimental proof of concept for a large scale, low energy consumption hardware weighted sum based on magnetic skyrmions. We demonstrate the precise electrical control of skyrmion nucleation and movement in specially designed magnetic tracks. The number of generated skyrmions is directly proportional to the applied current, allowing us to implement input multiplication by synaptic weights. Fine-tuning of the weights can be achieved through adjustments to the external magnetic field. Our design employs highly resistive Ta transverse electrodes connected to the edge of the track only, enabling the electrical detection of skyrmions via the anomalous Hall effect without adversely affecting their motion or providing sneak paths in crossbar arrays. We validate the weighted sum operation in a device featuring two parallel tracks intersected by a Hall elec- trode; the resulting Hall voltage corresponds to the combined number of skyrmions in both tracks [13]. This ensures efficient execution of the fundamental weighted sum operation, a cornerstone for neuromorphic computing. Additionally, we explore the integration of magneto-ionic effects for non-volatile and reversible control over local magnetic properties. We demonstrate voltage gating control of the magnetic anisotropy through the application of an electric field from an AlOx layer, allowing for the in-plane to out-of-plane anisotropy switch in the top Co layer [14]. These changes in the magnetic properties are found to be non-volatile and reversible. These results path the way towards gate voltage control of the skyrmion nucleation and motion (i.e., non-volatile tuning of the weighs). [1] W. Legrand et al., Nano Letters, vol. 17, pp. 2703, 2017. [2] A. Hrabec et al., Nature Communication, vol. 8, pp. 1-6, 2017. [3] S. Woo et al., Nature Electronics, vol. 1, pp. 288-296, 2018. [4] S. Woo et al., Nature Materials, vol. 15, pp. 501-506, 2016. [5] D. Maccariello et al., Nature Nanotechnology, vol. 13, pp. 233-237, 2018. [6] Hannekenet al., Nature Nanotechnology 10, 1039–1042, 2015. [7] J. U. Larrañaga et al., arXiv:2308.00445, 2023. [8] A. Fert, N. Reyren, and V. Cros, Nat Rev Mater 2, 17031, 2017. [9] G. Bourianoff et al., Aip Advances, vol. 8(5), pp. 055602, 2018. [10] K.M. Song et al., Nature Electronics, vol. 3.3, pp. 148-155, 2020. [11] J. Grollier et al., Nature Electronics 3, 360–370, 2020. [12] Y. Huang et al., Nanotechnology 28, 08LT02 (2017). [13] T. da Câmara Santa Clara Gomes et al., arXiv:2310.16909, 2023. [14] T. da Câmara Santa Clara Gomes et al., arXiv:2310.01623, 2023.
- Affiliations
Citations
da Câmara Santa Clara Gomes, T., Yanis Sassi, Dedalo Sanz-Hernandez, Sachin Krishnia, Marie-Blandine Martin, Pierre Seneor, Tanvi Bhatnagar-Schöffmann, Dafiné Ravelosona, Damien Querlioz, Liza Herrera-Diez, Vincent Cros, Julie Grollier, & Nicolas Reyren. (2024). Neuromorphic weighted sum with magnetic skyrmions. EHCI workshop of the Eindhoven University of Technology, Eindhoven (Holland). https://hdl.handle.net/2078.5/27556