Multifunctional BiFeO3 Thin Film-Based Memristor Device as an Efficient Synapse: Potential for Beyond von Neumann Computing in Neuromorphic Systems Multifunctional BiFeO3 Thin Film-Based Memristor Device as an Efficient Synapse: Potential for Beyond von Neumann Computing in Neuromorphic Systems AITranslate

Basic Information:

DOI：https://doi.org/10.1021/acsaelm.2c00901

Chinese Library Classification Number:

Citation Information:

We report the potential of low-cost sol–gel-synthesized phase pure BiFeO3 thin films as an active material with Cu as the top contact and fluorine-doped tin oxide (FTO) as the bottom contact, that is, Cu/BiFeO3/FTO-based devices for analogue resistive switching characteristics, suitable for beyond von Neumann computing. The Cu/BiFeO3/FTO device showed robust repeatability of bipolar resistive switching up to 3000 cycles. Moreover, retention measurements suggest that the device is stable for 17,000 s at 0.2 V, and endurance measurements suggest reliability for 5000 cycles at 1 and −1 V. Moreover, the Cu/BiFeO3/FTO device showed synaptic behavior, including short- and long-term plasticity and change in ionic conductance with time-dependent pulses suitable for neural network applications. The pulse paired facilitation (PPF) measurements demonstrated a reduction in the PPF value from 9 to 1% with the increasing pulse width from 10 to 400 ms. The strengthening in signal transmission is shown by increasing the synaptic weight from 20 to 22.5 approximately, with the increasing scan rate from 100 to 400 mV/s. The increase in the scan rate resulted in the synaptic weight saturation, and a change in ionic conductance is observed at 0.7 and 0.2 V with varying pulse widths ∼10, 20, and 50 ms, having a 10 ms pulse interval. The spike time-dependent plasticity measurement is carried out for the symmetric and asymmetric Hebbian rule with 4 s pulse width at 1.2 and −1.2 V pulse amplitude. The device also showed ∼96% recognition characteristics, as estimated from retention loss characteristics. Thus, the present work emphasizes the utility of BiFeO3-based analogue devices for memristive neuromorphic computing devices.