Invited
HfO_x based flexible and synaptic memristor for brain-inspired computing

A reliable flexible HfO_x-based memristor is demonstrated by the insertion of an TaO_x layer to mimic synaptic characteristics. Such device exhibits analog switching behavior with low set/reset voltages (1 V/−1 V), good switching endurance, high retention at room and high temperature (~120 ^oC), improved variance for LRS and HRS, stable multilevel cell characteristic for both AC pulses and DC parameters, high conductance linearity and stable potentiation and depression epochs cycles for more than 10000 pulses making it favorable for synaptic application. In addition, the flexible memristor device has stable switching for bending of ~4 mm radius with more than 1000 bending cycles. The X-ray photoelectron spectroscopy analysis indicates that oxygen vacancy concentration is increased in inserted TaOx layer leading towards improved switching performance. The device mimics biological synaptic time-dependent plasticity (STDP) operation when presynaptic and postsynaptic pulses are applied on top and bottom electrodes, respectively. The relationship between nonlinearity coefficient and control parameters in STDP is derived and established. It achieves more that 96% accuracy only after 20 iterations for neuromorphic application when a system of 2500 synapses incorporating 50 × 50 pixel image for recognition is deployed.