In perpendicularly magnetized heavy-metal/ferromagnetic heterostructures, the spin–orbit torque-induced multilevel switching of magnetic moment offers promising potential for nonvolatile multilevel information storage. Compared with traditional binary memory architectures, this approach provides improved data density and lower power consumption. However, the necessity of an additional in-plane magnetic field to break magnetic symmetry in perpendicularly magnetized heterostructures impedes the integration of spin–orbit torque-driven magnetization switching devices. This work investigates field-free spin–orbit torque switching in obliquely sputtered Pt/Co/Ta heterostructures, demonstrating that the oblique deposition angle determines the tilt of magnetic moments, thereby enabling field-free switching. By adjusting the width and amplitude of current pulses, we implemented multiple resistance states, mimicking synaptic behaviors such as long-term potentiation/long-term depression and excitatory/inhibitory postsynaptic potentials. Besides, linear optimization of magnetization switching curves revealed their compatibility with the rectified linear unit activation function, further highlighting the versatility of this approach for application in artificial neural networks. Finally, a fully connected convolutional neural network utilizing field-free spin–orbit torque devices achieved an impressive classification accuracy of 93.38% on the CIFAR-10 data set, demonstrating the practical feasibility of this technology in advanced machine learning tasks.
Guo et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: