论文标题
光学神经网络的三元菱形Znin2s4中超敏感,超快和栅极可调的二维光电探测器
Ultrasensitive, Ultrafast and Gate-Tunable Two-Dimensional Photodetectors in Ternary Rhombohedral ZnIn2S4 for Optical Neural Networks
论文作者
论文摘要
对电子和光电学中高性能半导体的需求促使低维材料研究扩展到三元化合物。但是,基于2D三元材料的光电探测器通常会遭受大深色电流和缓慢的响应,这意味着增加功耗和性能降低。在这里,我们报告了一项系统的研究研究,该研究表现出良好的菱形Znin2S4(R-ZIS)纳米片的光电性能,该特性表现出极低的深色电流(7 pa偏差为7 pa)。以一系列参数表示的出色性能超过了大多数2D对应物。超高特异性检测率(1.8 x 10^14琼斯),相当短的响应时间(τ_rise=222μs,τ_decay=158μs)以及与高频操作(1000 Hz)的兼容性特别突出。此外,观察到栅极可调的特征,这归因于光子,并通过两个数量级来改善光响应。门控技术可以有效地调节从光导效应到显性光的光电生成机制。超高灵敏度,超快响应和高栅极可调节性的结合使R-ZIS光晶体管成为低能消耗和高频光电的理想设备,这进一步证明了它在光学神经网络中的出色性能以及在光学深度学习和计算中的潜在潜力。
The demand for high-performance semiconductors in electronics and optoelectronics has prompted the expansion of low-dimensional materials research to ternary compounds. However, photodetectors based on 2D ternary materials usually suffer from large dark currents and slow response, which means increased power consumption and reduced performance. Here we report a systematic study of the optoelectronic properties of well-characterized rhombohedral ZnIn2S4 (R-ZIS) nanosheets which exhibit an extremely low dark current (7 pA at 5 V bias). The superior performance represented by a series of parameters surpasses most 2D counterparts. The ultrahigh specific detectivity (1.8 x 10^14 Jones), comparably short response time (τ_rise = 222 μs, τ_decay = 158 μs) and compatibility with high-frequency operation (1000 Hz) are particularly prominent. Moreover, a gate-tunable characteristic is observed, which is attributed to photogating and improves the photoresponse by two orders of magnitude. Gating technique can effectively modulate the photocurrent-generation mechanism from photoconductive effect to dominant photogating. The combination of ultrahigh sensitivity, ultrafast response and high gate tunability makes the R-ZIS phototransistor an ideal device for low-energy-consumption and high-frequency optoelectronic applications, which is further demonstrated by its excellent performance in optical neural networks and promising potential in optical deep learning and computing.