Bimodal alteration of cognitive accuracy for spintronic artificial neural networks.

Spintronics-based artificial neural networks (ANNs) exhibiting nonvolatile, fast, and energy-efficient computing capabilities are promising neuromorphic hardware for performing complex cognitive tasks of artificial intelligence and machine learning. Early experimental efforts focused on multistate device concepts to enhance synaptic weight precisions, albeit compromising on cognitive accuracy due to their low magnetoresistance. Here, we propose a hybrid approach based on the tuning of tunnel magnetoresistance (TMR) and the number of states in the compound magnetic tunnel junctions (MTJs) to improve the cognitive performance of an all-spin ANN. A TMR variation of 33-78% is controlled by the free layer (FL) thickness wedge (1.6-2.6 nm) across the wafer. Meanwhile, the number of resistance states in the compound MTJ is manipulated by varying the number of constituent MTJ cells (n = 1-3), generating n + 1 states with a TMR difference between consecutive states of at least 21%. Using MNIST handwritten digit and fashion object databases, the test accuracy of the compound MTJ ANN is observed to increase with the number of intermediate states for a fixed FL thickness or TMR. Meanwhile, the test accuracy for a 1-cell MTJ increases linearly by 8.3% and 7.4% for handwritten digits and fashion objects, respectively, with increasing TMR. Interestingly, a multifarious TMR dependence of test accuracy is observed with the increasing synaptic complexity in the 2- and 3-cell MTJs. By leveraging on the bimodal tuning of multilevel and TMR, we establish viable paths for enhancing the cognitive performance of spintronic ANN for in-memory and neuromorphic computing.

Multistate Compound Magnetic Tunnel Junction Synapses for Digital Recognition.

The quest to mimic the multistate synapses for bioinspired computing has triggered nascent research that leverages the well-established magnetic tunnel junction (MTJ) technology. Early works on the spin transfer torque MTJ-based artificial neural network (ANN) are susceptible to poor thermal reliability, high latency, and high critical current densities. Meanwhile, work on spin-orbit torque (SOT) MTJ-based ANN mainly utilized domain wall motion, which yields negligibly small readout signals differentiating consecutive states and has designs that are incompatible with technological scale-up. Here, we propose a multistate device concept built upon a compound MTJ consisting of multiple SOT-MTJs (number of MTJs, n = 1-4) on a shared write channel, mimicking the spin-based ANN. The n + 1 resistance states representing varying synaptic weights can be tuned by varying the voltage pulses (±1.5-1.8 V), pulse duration (100-300 ns), and applied in-plane fields (5.5-10.5 mT). A large TMR difference of more than 13.6% is observed between two consecutive states for the 4-cell compound MTJ, a 4-fold improvement from reported state-of-the-art spin-based synaptic devices. The ANN built upon the compound MTJ shows high learning accuracy for digital recognition tasks with incremental states and retraining, achieving test accuracy as high as 95.75% in the 4-cell compound MTJ. These results provide an industry-compatible platform to integrate these multistate SOT-MTJ synapses directly into neuromorphic architecture for in-memory and unconventional computing applications.