The Impact of ⁶⁰Co γ -Ray Irradiation on the Linearity of Conductance Update and the Learning Precision of Al:HfO<i>x</i>/TiO<i>x</i> Heterojunction Memristive Synapses

Good linearity and radiation resistance of conductance updates in memristive synapses are crucial for their use in neuromorphic computing chips in space. Here, the Au/Al:HfOx/TiOx/Ti memristive synapses with good linearity are prepared. The effects of irradiation on the memristive performances are studied through 60Co <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\gamma $ </tex-math></inline-formula>-ray irradiation experiments. The effect of irradiation on the image recognition accuracy of the neural network based on the memristive synapse crossbar array is predicted. The results show that the devices maintained good conductance update linearity under irradiation. When the 60Co <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\gamma $ </tex-math></inline-formula>-ray irradiation reaches 1 Mrad(Si), the nonlinearity (NL) of conductance update only increases from 0.096 to 0.163, and the image recognition accuracy of the memristor neural network decreases by about 2% (from 91% to 89%). The initial value of oxygen defect content in HfOx resistive materials after Al doping modification is relatively high (36%), which can explain why irradiation has little effect on conductance update linearity and recognition accuracy. These results suggest that the Au/Al:HfOx/TiOx/Ti memristors show good radiation resistance under 60Co <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\gamma $ </tex-math></inline-formula>-ray irradiation, which makes them promising candidates for artificial synapses in space applications.