Passive LiNbO₃ Memristor With Multilevel States for Neuromorphic Computing

The implementation of multilevel conductance states is still difficult for passive memristors used in neuromorphic computing. Here, a passive single-crystalline LiNbO<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sub> (LN) memristor with multilevel states was proposed, which can be precisely programmed into multilevel target states (with a standard deviation below <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$0.008~\mu $ </tex-math></inline-formula>s). Moreover, 32 separated and reliable conductance states can be achieved. The pattern recognition simulation for different numbers of conductance states (<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${N}_{\textrm {G}}$ </tex-math></inline-formula>) is performed. As <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${N}_{\textrm {G}}$ </tex-math></inline-formula> increases, the inference accuracy rises and reaches 98.01% when <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${N}_{\textrm {G}}$ </tex-math></inline-formula> is 32. Even taking into account the conductance programming and drift error of the memristors, the accuracy can still reach 90.37%. The results validate the application potential of this passive memristor with 32 conductance states in neuromorphic computing.