Nowadays the development of machine vision is oriented toward real-time applications such as autonomous driving. This demands a hardware solution with low latency, high energy efficiency, and good reliability. Here, we demonstrate a robust and self-powered in-sensor computing paradigm with a ferroelectric photosensor network (FE-PS-NET). The FE-PS-NET, constituted by ferroelectric photosensors (FE-PSs) with tunable photoresponsivities, is capable of simultaneously capturing and processing images. In each FE-PS, self-powered photovoltaic responses, modulated by remanent polarization of an epitaxial ferroelectric Pb(Zr0.2Ti0.8)O3 layer, show not only multiple nonvolatile levels but also sign reversibility, enabling the representation of a signed weight in a single device and hence reducing the hardware overhead for network construction. With multiple FE-PSs wired together, the FE-PS-NET acts on its own as an artificial neural network. In situ multiply-accumulate operation between an input image and a stored photoresponsivity matrix is demonstrated in the FE-PS-NET. Moreover, the FE-PS-NET is faultlessly competent for real-time image processing functionalities, including binary classification between ‘X’ and ‘T’ patterns with 100% accuracy and edge detection for an arrow sign with an F-Measure of 1 (under 365 nm ultraviolet light). This study highlights the great potential of ferroelectric photovoltaics as the hardware basis of real-time machine vision.

Robust and reversible polar topological center domains were found in BiFeO3 nanodots, which are individually controllable.

Achieving high power conversion efficiencies (PCEs) in ferroelectric photovoltaics (PVs) is a longstanding challenge. Although recently ferroelectric thick films, composite films, and bulk crystals have all been demonstrated to exhibit PCEs >1%, these systems still suffer from severe recombination because of the fundamentally low conductivities of ferroelectrics. Further improvement of PCEs may therefore rely on thickness reduction if the reduced recombination could overcompensate for the loss in light absorption. Here, a PCE of up to 2.49% (under 365-nm ultraviolet illumination) was demonstrated in a 12-nm Pb(Zr 0.2 Ti 0.8)O 3 (PZT) ultrathin film. The strategy to realize such a high PCE consists of reducing the film thickness to be comparable with the depletion width, which can simultaneously suppress recombination and lower the series resistance. The basis of our strategy lies in the fact that the PV effect originates from the interfacial Schottky barriers, which is revealed by measuring and modeling the thickness-dependent PV characteristics. In addition, the Schottky barrier parameters (particularly the depletion width) are evaluated by investigating the thickness-dependent ferroelectric, dielectric and conduction properties. Our study therefore provides an effective strategy to obtain high-efficiency ferroelectric PVs and demonstrates the great potential of ferroelectrics for use in ultrathin-film PV devices.

away quickly when the light stimuli are removed. [3][4][5] In other words, traditional photodetectors can detect the images like a retina, but they lack the memory function owned by the visual cortex (see schematic illustration of human visual system in Figure 1a). To realize both detection and memory functions so as to better imitate the human visual system, researchers have attempted to integrate the photodetectors with the nonvolatile memory devices. [6][7][8][9] For example, a bioinspired visual system comprising an In 2 O 3 nanowire photodetector connected in series with an Al 2 O 3 memristor was fabricated recently, which could capture a butterflyshaped image and store it for more than 1 week. [9] In such integrated devices, however, the units responsible for detection, processing, and storage of optical information are physically separated, resulting in high power consumption for data transfer between different units (just as the Von Neumann bottleneck [10,11] ).A simple yet effective humanoid optoelectronic device emerging recently is the artificial optoelectronic synapse based on the photoelectric memristor, which can co-locate the detection and memory functions in a single unit. As the name suggests, a photoelectric memristor can continuously change its resistance upon light stimuli, resembling the physiological The rapid development of artificial intelligence technology has led to the urge for artificial optoelectronic synapses with visual perception and memory capabilities. A new type of artificial optoelectronic synapse, namely a photo electric memcapacitor, is proposed and demonstrated. This photoelectric memcapacitor, with a planar Au/La 1.875 Sr 0.125 NiO 4 /Au metal-semiconductormetal structure, displays a complementary optical and electrical modulation of capacitance, which can be attributed to the charge trapping/detrapping induced Schottky barrier variation. It further exhibits versatile synaptic functions, such as photonic potentiation/electric depression, pairedpulse facilitation, short/longterm memory, and "learningexperience" behavior. Moreover, the photoplasticity of the memcapacitor can be modulated by varying the frequency of applied AC voltage, thus enabling selfadaptive optical signal detection and mimicry of interestmodulated human visual memory. Therefore, it represents a new paradigm for artificial optoelectronic synapses and opens up opportunities for developing lowpower humanoid optoelectronic devices.

A wealth of fascinating phenomena have been discovered at the BiFeO domain walls, examples such as domain wall conductivity, photovoltaic effects, and magnetoelectric coupling. Thus, the ability to precisely control the domain structures and accurately study their switching behaviors is critical to realize the next generation of novel devices based on domain wall functionalities. In this work, the introduction of a dielectric layer leads to the tunability of the depolarization field both in the multilayers and superlattices, which provides a novel approach to control the domain patterns of BiFeO films. Moreover, we are able to study the switching behavior of the first time obtained periodic 109° stripe domains with a thick bottom electrode. Besides, the precise controlling of pure 71° and 109° periodic stripe domain walls enable us to make a clear demonstration that the exchange bias in the ferromagnet/BiFeO system originates from 109° domain walls. Our findings provide future directions to study the room temperature electric field control of exchange bias and open a new pathway to explore the room temperature multiferroic vortices in the BiFeO system.