Experimental demonstration of skyrmionic magnetic tunnel junction and its on-chip learning application

Chiral magnetic skyrmions are topological swirling spin textures that hold promise for future information technology. The electrical nucleation and motion of skyrmions have been experimentally demonstrated in the last decade, while electrical detection compatible with semiconductor processes is still considered one of the most crucial gaps regarding the use of skyrmions in real applications. In this study, we report the direct observation of nanoscale skyrmions in CoFeB/MgO-based magnetic tunnel junction devices at room temperature by using high- resolution magnetic force microscopy (MFM) imaging and tunneling magnetoresistance (TMR) measurements. This skyrmionic magnetic tunnel junction (SK-MTJ) shows a stable nonlinear multilevel resistance thanks to its topological nature and tunable density of skyrmions under current pulse excitation, which can represent the variations in the synaptic weights in the neuromorphic computing. Our chip-level simulation shows that such Sk-MTJ synapse array could be utilized to implement on-chip learning to solve image recognition task such as with CIFAR-10.