Experimental study demonstrates pulse-engineered synaptic linearity in MoO3/TiO2 bilayer memristors, indicating enhanced accuracy for neuromorphic image recognition.
Key Points
To optimize pulse protocols and switching characteristics in a MoO3/TiO2 bilayer memristor for simultaneous non-volatile digital memory and linear analog synaptic operation.
Fabricated and characterized Au/MoO3/TiO2/FTO bilayer-oxide memristive devices.
Evaluated switching endurance (25,000 cycles), retention (1000 s), thermal stability (up to 60°C), and operational durability over 70 weeks.
Applied an optimized combinational voltage pulse protocol to tune potentiation–depression linearity and evaluated neural network performance on handwritten digit recognition.
The memristor exhibited robust non-volatile bipolar switching with endurance of 25,000 cycles, retention of 1000 s, and stability over 70 weeks up to 60°C.
The RESET process displayed quantized conductance steps, evidencing atomic-scale filament constriction.
Optimized pulse schemes established highly linear potentiation–depression transitions, yielding a ~20% improvement in artificial neural network inference accuracy for handwritten digit recognition.
Cite This Study
Chandrashekar et al. (2026) studied this question.