Yuewei Yin1,Xiaoguang Li1
University of Science and Technology of China1
Yuewei Yin1,Xiaoguang Li1
University of Science and Technology of China1
The rapid development of information technology requires memory devices with fast speed, low power consumption, high density, <i>etc</i>. HfO<sub>2</sub>-based ferroelectric thin films are promising for the advanced information memory devices, especially due to the good silicon compatibility and intrinsic advantages of ferroeelctricity. Recently, we used atomic layer deposition to construct high-quality Hf<sub>1-<i>x</i></sub>Zr<i><sub>x</sub></i>O<sub>2</sub> (HZO) ferroelectric thin film heterostructures and investigated corresponding prototype memory devices, including ferroelectric capacitors for FeRAM and ferroelectric tunnel junctions for neural network computing. The ferroelectric endurance properties of TiN/HZO/TiN capacitor were studied, and it was found that the ferroelectric fatigue severity increases first and then decreases with increasing pulse voltage or width. Accordingly, a recovery method by introducing wake-up effect was utilized to realize a weaker fatigue, a sufficient switched polarization (7–12 μC/cm<sup>2</sup>) by 100 ns voltage pulses, and especially an enhanced endurance of >1.01×10<sup>12</sup> (>5.0×10<sup>13</sup> in expectation). On the basis of growing high-quality HZO heterostructure, ferroelectric tunnel junctions (FTJs) based on a SiO<sub>2</sub>/HZO<sub> </sub>composite barrier and both conducting electrodes were designed and fabricated on Si substrates. The FTJ achieves the fastest write speed of 500 ps under 5 V or 10 ns speed at a low voltage of 1.5 V, low write current density ~1.3×10<sup>4</sup> A/cm<sup>2</sup>, and 8 discrete states. Interestingly, in the FTJ-based memristive synapse, gradually tunable nonvolatile conductance states (128 states) with high linearity were obtained, and by inputting the corresponding experimental results into an online neural network simulation, a high accuracy in recognizing fashion product images was achieved. These results indicate that silicon-compatible HfO<sub>2</sub>-based ferroelectric heterostructures are important candidates for high-performance nonvolatile memory and artificial synaptic devices.