Fig. 3: Performance of the FE-PS.
From: In situ training of an in-sensor artificial neural network based on ferroelectric photosensors
a Light intensity dependence and b long-term stability of photocurrents of the device in the States I to IV. c Transient current responses to illumination for the device in the full Pup (upper panel) and Pdown (lower panel) states. d Photoresponsivities of the device in the full Pup and Pdown states after different endurance cycles. e Photoresponsivity as a function of write pulse width. The device is preset into a fully Pup (Pdown) state before each application of a + 10 V (–10 V) write pulse with a varied width. f LTP/LTD characteristics measured with an amplitude-increasing pulse scheme. g Performance comparison between our FE-PS and other emerging programmable photosensors for in-sensor computing. The “0” and “1” on the “self-powered” axis represents “not self-powered” and “self-powered”, respectively. The “0” and “1” on the “Iph-Ilight relationship” represents “nonlinear” and “linear” relationships between photocurrent and light intensity, respectively. The “–1” on the “photoresponse time” and “write energy” axes refers to that the value is not reported.
