Author：Weizhen Liu, Xuhui Yang, Zhongqiang Wang, Yuanzheng Li, Jixiu Li, Qiushi Feng, Xiuhua Xie, Wei Xin, Haiyang Xu & Yichun Liu 

Introduction

Visual perception, a vital sensing functionality for human beings and other vertebrates, contributes more than 80% of the perceptual information from the ambient environments to the brain. With the rapid development of artificial intelligence, artificial visual systems are demanded to be able to mimic the visual perception capabilities of biological systems. Among them, an important functionality is visual adaptation, which can automatically adjust the response to stimuli according to different light environments. For instance, the retina of the eye can reduce the levels of light dynamically to make the human-eye work normally when the illumination is from very dark to very bright. However, the existing efforts on mimicking visual adaptation functions have been trapped in complicated hardware and algorithms that typically reduce operating efficiency. Ideally, a new generation of artificial visual systems with adaptation functions should have a simpler structure and lesser logic operations. Recently, several cutting-edge artificial visual systems based on a single device have been designed with built-in visual adaptation, such as organic transistor, bilayer MoS2 phototransistor, and heterojunction phototransistor, which exhibit dynamic adaptation well to external light stimuli and show the potential to emulate human visual adaptation. However, their main mechanisms are still restricted to modulation of carrier trapping or ion migration, which are inadequate for the future development of visual adaptive devices and artificial visual systems. Hence, it is highly desired to explore more working mechanisms to serve the visual adaptation function of artificial visual systems that possess simple device architectures.

Visual adaptive devices based on pyroelectric materials may supply a new option for mimicking visual adaptation to constant light stimuli. That is because, upon light irradiation, the pyroelectric materials with noncentral-symmetric structure can convert time-dependent temperature fluctuation induced by the photothermal effect into instantaneous potential, which results from the changed polarization intensity with temperature causing interfacial charge release. Such behavior of the instantaneous potential is similar to photosensitivity reduction over time in human vision systems when constant light is applied. It follows that pyroelectric materials show great possibilities in simulating human visual adaptation and need to be further explored. γ-InSe, as an emerging two-dimensional (2D) layered III?VI semiconductor, has typical low-symmetry crystal structures and belongs to R3m space groups, similar to such typical pyroelectric materials as α-In2Se3, α-GeTe and GeSe. More importantly, unlike transition-metal dichalcogenides (TMDs), γ-InSe does not require the construction of hybrid structures to enhance photon absorption because of its direct bandgap structure in multilayer regime, which would undoubtedly simplify the device geometry. Together with ultrahigh mobility and broad spectral response, multilayer γ-InSe is considered to have great potential in developing high-performance photoelectric conversion devices.

In this work, a two-terminal opto-sensor based on multilayer γ-InSe flake is experimentally demonstrated to emulate the human visual adaptation ranging from ultraviolet to near-infrared light without a bias voltage. When exposed to soft light, the device requires little adaptation process, but when the device is exposed from soft (or dim) to bright light, the response jumps initially and then declines to reach equilibrium, well simulating the self-adaptation process of human eyes to light stimuli. The main working mechanism for the dynamic adaptation is confirmed to be the synergy of photo-pyroelectric effect and photo-thermoelectric effect. Under bright light illumination, the photo-pyroelectric effect mainly causes the initial rapid adaptation process whereas the subsequent slow adaptation process is primarily induced by the photo-thermoelectric effect. Moreover, several important visual adaptation functions have been mimicked, including light-sensing image adaptation, photosensitive recoverability, and synergetic visual adaptation.