Advanced Bioinspired Organic Sensors for Future‐Oriented Intelligent Applications

Bioinspired organic sensors with excellent flexibility, biocompatibility, sensing performance, and self‐adaptability have attracted considerable attention due to their fundamental role in next‐generation intelligent sensory systems with simultaneous sensing and processing functions, such as advanced robots, disease diagnosis, Internet of Things, human–robot interaction systems, and beyond. The design of advanced bioinspired organic sensors requires a deep understanding of the interplay between the unique material property, sensing mechanism as well as intelligent functionality. Here, the recent progress of bioinspired organic sensors for artificial perceptron covering visual, tactile, and other bionic sensors from the view of materials, sensing mechanisms, and functionality, is summarized. Furthermore, the intelligent applications of bioinspired organic sensors, including associative learning, information security, electronic skin, and integrated artificial sensory systems, are presented. Finally, the potential challenges and prospects of bioinspired organic sensors are discussed, providing a promising avenue toward advanced bioinspired organic sensors for future‐oriented intelligent applications.


Introduction
Sensors, acting as the bridge connecting the external environment and electrical signals, have been widely applied to diverse fields, such as image sensors, medical monitoring, robots, and manufacturing process monitoring. [1][2][3][4][5] In recent years, with the rapid development of big data, artificial intelligence (AI), and the Internet of Things (IoTs), sensor technologies have been confronted with new challenges from materials, sensing mechanisms, and functionalities to deal with the complex sensing tasks and high energy consumptions. [6][7][8][9] Note that, living organisms have the ability to perceive the changes in the external environment with ultrahigh sensitivity, rapid response, high energy-efficiency, and selfadaptability by the sensory organs containing vision, touch, hearing, olfaction, and gustation. The environmental information captured from complicated surroundings is preprocessed in the sensory organs and transmitted through afferent nerves to the brain in the form of bioelectric impulses for further analysis and decision-making. [10][11][12] Inspired by the distinctive materials and well-adapted structures of natural sensory systems, integrating organic functional materials with the unique device structure will lead to bioinspired sensors with intelligent sensing and processing functions, which promotes intelligent perception, artificial machine vision, edge computing, and other fields into a new era of development. [12][13][14] Organic materials, acting as the key components of bioinspired sensors, consisting of conjugated molecules assembled by weak van der Waals forces have attracted considerable research attention. [15][16][17][18][19][20] Compared to the inorganic counterparts, organic materials possess excellent physical and chemical properties to enable the design and manufacture of bioinspired sensors with superior sensing performance, biocompatibility, and elasticity. [20][21][22][23][24] In particular, organic materials can be directly fabricated on various soft and flexible substrates using low-cost deposition methods, such as spin-coating, blade coating, rollto-roll, and inkjet printing, which is suitable for the fabrication of flexible bionic devices. [19,[25][26][27][28] Notably, organic materials with electrical relaxation behavior are suitable for artificial synapses for high-efficiency neuromorphic computing beyond conventional computing architecture. These unique advantages make organic materials a preferred choice for the design of bioinspired sensory devices for the development of futureoriented intelligent applications. [20,[29][30][31] Furthermore, designing organic sensors by mimicking specific biological structures in nature will improve the device performance and endow sensors with new functions. For example, integrating organic optical materials with flexible and curved substrates leads to the fabrication of artificial visual sensors, which simulate the structure features of the human retina and eyes. [32,33] Besides that, the adaptive and stretchable structure of human skin has long inspired the surface nanostructure and device architecture of artificial tactile sensing and actuation systems. [34,35] Tailoring molecular structures, micro/nanofilm structures of functional materials, and the semiconductor/dielectric interface is a vital strategy for achieving better sensing capabilities. Therefore, benefit from breakthroughs in materials and structures, bioin-spired organic sensors can transfer perception to cognition via detecting, collecting, integrating, memorizing, and processing the data in multifunctional perception, which can reduce the redundant and unstructured sensing data as well as improve the energy-efficiency. [29,36] In the article, we provide a review of the state-of-the-art bioinspired organic sensors in emerging intelligent applications. First, an overview of the material properties, sensing mechanisms as well as intelligent functionalities of various bioinspired organic sensors is introduced including visual, tactile, and other bionic sensors. Subsequently, we review the advanced and intelligent applications of bioinspired organic sensors and integrated sensory systems. Finally, challenges and outlooks for developing more advanced organic sensory devices and systems for bionic applications are discussed (Figure 1).

Bioinspired Perception Based on Advanced Organic Sensors
The biomimetic strategy of bioinspired organic sensors takes inspiration from living organisms, especially human beings, to endow existing sensory devices with dramatically improved performance and novel functions. The bionic sensors mainly focus on the five conventionally acknowledged senses (vision, touch, hearing, olfaction, and gustation) of living organisms and mimic their behaviors through the combination of functional materials, the design of device structures, and sensing physics. [46][47][48][49][50] In this section, several types of advanced organic sensory devices, including visual, tactile, and other artificial sensors are discussed.

Organic Sensors for Visual Perception
The human brain can visually perceive more than 80% of the information from the environment. [51][52][53] The human eye efficiently captures visual scenes and converts them into neural electrical signals, which are further processed and recognized in the retina and cerebral cortex. [54][55][56][57] The structure and function of human vision system provides attractive design inspiration for artificial visual sensors and neuromorphic processing devices, which will be conducive to developing next-generation machine vision systems. In addition, organic materials have unique photoelectric properties, mechanical flexibility, and biocompatibility, making them excellent candidates for emerging visual sensors.
Here, bioinspired artificial visual sensors and neuromorphic image processing sensors are introduced. As briefly listed in Table 1, we summarized the organic sensors for visual perception based on the function, functional materials and working mechanism with the detailed discussions in the following sections.

Bioinspired Artificial Vision Sensors
The human eye consists of a curved retina with photoreceptor cells, a lens with a gradient refractive index, and an adjustable iris.
Visible light forms an image on the human eye's retina through a refractive system. The human eye produces efficient image perception through simple and ingenious optical structures, including high-quality imaging, color resolution, and photopic or scotopic adaptation. [54,66,67] Therefore, it is essential to prepare electronic devices based on organic materials to simulate the curved surface structure and functional advantages of the human eye to complete efficient image acquisition, simplifying the structure of traditional planar imaging modules. The curved retina is one of the prominent features of the human eyes, which brings about a wide field of view (FOV) (150°-160°), aberration compensation of curved focal plane, and high-resolution imaging. [46] To imitate the above features, it is necessary to prepare a hemispherical image sensor array. Organic materials are an indispensable part for the fabrication of curved bionic visual sensors because of their adjustable spectral response, excellent flexibility, low temperature, low cost, and biocompatibility. A common method for fabricating curved image sensors is to prepare thin and deformable planar devices and then transfer and attach them to the target surface. [46,68,69] Wang et al. reported the water-assisted glass strategy to prepare a conformal organic surface array to simulate retinal structures (Figure 2a). [68] The device consisted of an organic heterojunction-based photosensitive voltage divider and an organic electrochemical synapse, which were responsible for light perception and neural morphology readout, respectively. However, there are high-density sensory cells in the retina of ≈10 million per square centimeter, with an average spacing of 3 μm. [70] At present, only the most advanced planar commercial charge-coupled device and complementarymetal-oxide-semiconductor sensors can achieve the equivalent high imaging resolution, which undoubtedly requires high manufacturing technology for curved surface sensors.
Moreover, the traditional patterning process requires photolithography, transfer, and other techniques, which makes organic materials vulnerable to damage caused by ultraviolet light, heat, and organic solvents. Therefore, it is a critical problem for organic array devices to reach the same sensing performance and pixel as silicon-based optoelectronic devices without sacrificing the electrical performance of sensors. In this regard, Kim et al.  [68] Copyright 2018, Wiley-VCH. b) Schematic of an organic hemispherical image sensor array. Reproduced with permission. [46] Copyright 2018, Wiley-VCH. c) Schematic of a spherical biomimetic electrochemical eye. Reproduced with permission. [32] Copyright 2020, The authors, published by Springer Nature. d) Schematic and performance of 3D-printed polymer photodetectors on a hemispherical surface. Reproduced with permission. [33] Copyright 2018, Wiley-VCH. prepared a hemispherical organic image sensor array with a density of 308 pixels per square centimeter through new lithography and transfer technologies ( Figure 2b). [46] The mechanically peeled organic sensor array was transferred to the elastomer by plasma bonding to make the flexible array, which protected the organic semiconductor. The design of high modulus rigid island SiN x ensured the flatness of the photoelectric storage transistors, thus realizing the uniform light response of the entire array. The organic hemispherical image sensor had linear light response (light intensity range, from 1 to 50 W m −2 ) and a response rate of 1.6 A W −1 (wavelength = 465 nm) at a dark current of 0.24 A m −2 (drain voltage = −1 V). The combination of a hemispherical organic sensor and a single central lens successfully captured and reproduced the image "X". It provided a reference for manufacturing high-resolution 3D organic semiconductor arrays in the future.
Another strategy for manufacturing hemispherical image sensor arrays is direct 3D manufacturing, including direct growth of materials on 3D surfaces and 3D printing. [32,33] Gu et al. fabricated a spherical biomimetic electrochemical eye, in which the hemispherical retina was obtained by direct vapor-phase growth in a template by a high-density array of nanowires ( Figure 2c). [32] The bionic electrochemical eye was composed of a lens, organic ionic liquid, a hemispherical photoelectric sensor array, and liquid-metal wires, which respectively imitated the lens, vitreous humor, retina, and nerve fibers behind the retina of the human eye. The appropriate concentration of organic ionic liquid contributed to the rapid response of the photoelectric sensor, even shorter than the response and recovery time of the human eye. The bionic electrochemical eye was able to capture different imaging letters and 100.1°diagonal FOV. Due to the proper isolation between each pixel, the imaging had high contrast and clear edges. Noted that the spacing of single-crystal nanowires is 500 nm. If the diameter limitation of liquid metal wires can be solved, it is possible to realize photoreceptors similar to human eye density. Park et al. demonstrated a polymer photodetector array that was 3D printed on a hemispherical surface layer by layer ( Figure 2d). [33] The 3D-printed hemispherical photodetector was able to maintain the same performance as the 3D-printed planar device. It successfully reconstructed the cross pattern projected on its surface from different angles, which showed the application prospect of the directly 3D-printed hemispherical curved surface image sensor in the field of bionic eyes.
Shape-morphing polymers provide a new strategy for the development of curved image sensors. They change shape in response to external stimuli, such as temperature or pH, which can adaptively contact various curvature surfaces. [71] Du and co-workers fabricated a flexible microelectrode array based on  [59] Copyright 2021, The authors, published by Springer Nature. b) Schematic of an organic optical sensing inverter with RGB broadband photoresponse. Reproduced under the terms of Creative Commons Attribution 4.0 international (CC BY 4.0). [60] Copyright 2021, The authors, published by Wiley-VCH.
shape-morphing polymers, which can control the deformation to adapt to the curved surface. [58] At the physiological temperature, the flexible array with 126 circled microelectrodes could be selfunfolding from a minisized tube (2 mm) to thin films (10 × 10 mm 2 ), which was attached to the eyeball-sized spherical template. Microelectrode arrays provided high-resolution retinal stimulation and wide field of vision, which were expected to be applied in the field of visual prosthesis and implantable electronics in combination with flexible sensors.
In the human retina, there are two kinds of photoreceptors, cone cells and rod cells, which determine the adaptive ability of the human eye to light intensity and color discrimination. The cone cells can be excited by intense light, while the rod cells are sensitive to dark light. They complement each other to generate an appropriate output signal level according to the environment's light intensity so that the eyes can adapt to external lighting. [51,53] Imitating the light adaptability of the human eye requires combining photoelectric effect and charge trapping of organic materials to achieve light intensity-dependent transient response and dynamic adaptation. [59,72] He et al. designed an or-ganic transistor containing two bulk heterojunctions to obtain tunable optical adaptability. [59] The photovoltaic effect induced light excitation in the two organic heterostructures coupled with light suppression dominated by light intensity dependent electron capture at the poly(3-hexylthiophene-2,5-diyl) (P3HT): [6,6]phenyl-C61-butyric acid methyl ester (PCBM)/polyvinyl alcohol interface, realizing the modulation of carrier concentration with light intensity in the conductive channel (Figure 3a). The I DS /I 0 of a curved 3 × 3 transistor array could display the "T" image of the stimulus lamp in a dark background and can produce a low contrast image with adaptive light intensity within 2 s under a strong background. The luminance-dependent active adaptation capability in a single device significantly reduced the power consumption and biological operation times of the image sensor array.
In addition, the cone cells of the human eye can be divided into three types, which contain three kinds of visual pigments that absorb red, green, and blue light. They can perceive color information from different wavelengths of incident visible light. [55,57] The combination of organic materials with different wavelength selectivity in the active layer makes it possible to www.advancedsciencenews.com www.advsensorres.com identify the color by bioinspired artificial vision. [37,52,60,73,74] Hung et al. fabricated an organic optical sensing inverter similar to cone cells, which had high red/green/blue (RGB) broadband photoresponse, fast response speed (<300 ms), and low energy consumption (Figure 3b). [60] Perylene-diimide (PDI)sol and 2,9-didecyldinaphtho[2,3-b:2′,3′-f ]thieno[3,2-b]thiophene (C 10 -DNTT) have dual functions of charge-trapping (conjugated rod) and tunneling (insulating coil), while n-type bis(2-phenylethyl)-perylene tetracarboxylic diimide (BPE-PTCDI) and ptype dinaphtho[2,3-b:2′,3′-f ]thieno[3,2-b]thiophene (DNTT) were used as corresponding transporting layers. DNTT films were sensitive to shorter wavelengths, while PDI film has broader absorption. Therefore, the photodetector could display multilevel status under red/green/blue light. The output voltage of the hemispherical device array reproduced three patterns with different RGB colors, confirming the color image capture capability of the organic optical sensing inverter.

Neuromorphic Image Processing Sensors
In the human visual system, massive visual information is processed in the first stage through the retina after being efficiently collected by the human eyes, and then transmitted to the visual cortex of the human brain by the optic nerve for recognition processing. [62,69] Inspired by the human vision system, emerging neuromorphic devices have been developed for efficient image processing, which improves the machine vision efficiency of intelligent systems such as automatic driving and intelligent robots. Here, we will introduce image preprocessing sensors based on organic optoelectronic synapses, as well as image recognition sensors combined with neuromorphic computing.
The sensory layer of the human retina can not only detect light stimuli but also preprocess the visual information to reduce redundant information and improve the sensory quality. [75,76] Optoelectronic synaptic devices that mimic human eyes for image preprocessing have been widely studied. Similar to biology, artificial optoelectronic synapses can respond to optical stimuli, and conduct real-time preprocessing and short-term memory of imaging information, which can improve the efficiency and accuracy of subsequent image recognition tasks. The image preprocessing functions of optoelectronic synapse are realized through lightintensity-dependent and illumination-time-dependent plasticity, including image noise reduction, contrast enhancement, and sharpening. Choi et al. proposed bent neuromorphic phototransistor arrays based on MoS 2 light-sensitive layer and poly(1,3,5trimethyl-1,3,5-trivinyl cyclotrisiloxane) (pV3D3) dielectric layer, which completed image acquisition and preprocessing by a single readout of the electrical output (Figure 4a). [69] Due to the large exciton binding energy of MoS 2 -pV3D3 heterostructures and the trapping of photogenerated carriers at the MoS 2 -pV3D3 interface, phototransistors exhibited phototriggered synaptic plasticity. The imaging system based on bending neuromorphic devices could collect the preprocessed image from a large number of noisy light inputs to reach noise reduction and contrast enhancement without repetitive data storage, processing, and communications as well as complicated optics, required in conventional imaging and recognition systems. The proposed curved neuromorphic image sensor array can substantially improve the efficiency of the im-age acquisition and recognition process, enabling a step forward to the next generation of machine vision.
In addition, in order to further imitate the intelligent processing of the retina, it is important to fabricate self-powered artificial optoelectronic synapses through appropriate material and device structure design. [61,62,77] Hao et al. demonstrated that a self-powered artificial optoelectronic synapse based on organic asymmetric heterostructures was able to perform image sharpening preprocessing with zero-powered synaptic plasticity (Figure 4b). [62] Semivertical organic asymmetric heterojunction devices based on dioctylbenzothienobenzothiophene (C 8 -BTBT) and hexadecafluorophthalocyanine (F 16 CuPc) exhibited typical photovoltaic responses. Due to the charge-trapping effect at the C 8 -BTBT/poly(methyl methacrylate) (PMMA) interface, the devices presented short-term plasticity (STP), including excitatory postsynaptic current (EPSC), paired-pulse facilitation, etc. The self-powered artificial synapses had a frequency-dependent gain of light pulses, which was applied to the high-pass filtering properties of simulated synaptic information processing to reach the sharpening of the flowers in the image. Wang et al. prepared a self-powered organic optoelectronic synapse with asymmetric Schottky contacts by constructing asymmetric ultrathin C 8 -BTBT molecular layers in a single device (Figure 4c). [61] Self-powered optoelectronic synapses not only got image sharpening by the high-pass filter function of the spike-rate-dependent plasticity but also achieved image denoising through spike-amplitudedependent plasticity.
Image recognition sensors based on neuromorphic computing efficiently store and calculate information parallelly by simulating the way of information processing in the human brain to bring out high-accuracy image recognition. [7,[63][64][65]78] By simulating the long-term potentiation (LTP) and long-term depression (LTD) of biological synapses, neuromorphic organic synapses can build a neural network. And the synaptic weights of synapses can be dynamically tuned by the change of light or electrical pulses. For example, Lv et al. fabricated organic electrophotoactive synaptic transistors with a carbon dots (CDs)/silk protein (silk) blend, which can be used as a light-tunable charge trapping medium (Figure 5a). [63] By adjusting the input light pulse to form an unstable and stable charge accumulation behavior in the charge-trapping layer, the channel current exhibited temporary and permanent changes, leading to STP and LTP characteristics of synapses. The synaptic-like behaviors resulted from the photogating effect induced by trapped photogenerated electrons in the hybrid CDs/silk films. Furthermore, the artificial neural network (ANN) with a single perception layer was emulated with 73% accuracy for pattern recognition. The optical stimulated synaptic transistors provide promising building blocks for applications of bioinspired photonic computing. In addition, in practical intelligent applications, input data sometimes lack labels, so image recognition under unsupervised learning is particularly important. Wang et al. reported an autoencoder based on LTP and LTD of organic photoelectric synapses to simulate unsupervised ANN for image recognition (Figure 5b). [78] After 50 training cycles, the accuracy of the image recognition reached 81.37%.
Another emerging neural network is the spike neural network (SNN), in which neurons communicate with each other using spikes via synapses connecting the neurons with adjustable  [69] Copyright 2022, The authors, published by Springer Nature. b) An artificial optoelectronic synapse based on organic asymmetric heterostructures performing image sharpening preprocessing with zero-powered synaptic plasticity. Reproduced under the terms of Creative Commons Attribution 4.0 international License (CC BY 4.0). [62] Copyright 2022, The authors, published by Wiley-VCH. c) A self-powered organic optoelectronic synapse with asymmetric Schottky contacts realizing image sharpening and denoising. Reproduced with permission. [61] Copyright 2022, American Chemical Society. weight values. SNN updates synaptic weights based on localized learning rules using spatiotemporal information, leading to high computing efficiency and low energy consumption. Mu et al. fabricated an artificial sensory neuron system consisting of a leaky integrate and fire (LIF) neuron based on an Ag/SiO 2 /Ag memristor and an optoelectrical organic synaptic transistor. [7] The nearinfrared ray (NIR) light signals were encoded as electrical spikes using the sensory neuron system. The spiking rate was effectively modulated by changing NIR light power density. A twolayer SNN based on the LIF neurons was constructed to verify the computation ability of the artificial neural system. According to the confusion matrix of classification results, recognition accu-racy of 63.21% was achieved in the Modified National Institute of Standards and Technology digit classification. Pradhan et al. used unsupervised SNN to complete facial recognition ( Figure 5c). [64] The external sensor converted the optical signal of each pixel of the human face image into a presynaptic spike signal, which is then transferred to the optoelectronic synapse and postsynaptic neurons based on organic-inorganic hybrid perovskite quantum dots for processing to reach the facial feature extraction.
Reservoir computing (RC) is an efficient brain-like algorithm for processing complex time-related data with low training costs. RC includes an input layer, a dynamic reservoir, and an output layer, which maps input signals into higher dimensional Figure 5. The implementations of neural networks based on neuromorphic image processing sensors. a) Organic synaptic transistor based on carbon dots/silk protein blend mixture and the structure of ANN with the single perception layer. Reproduced with permission. [63] Copyright 2019, Wiley-VCH. b) Schematic of unsupervised ANN-based image recognition. Reproduced with permission. [78] Copyright 2021, American Chemical Society. c) Schematic of unsupervised SNN-based facial recognition. Reproduced under the terms of Creative Commons Attribution NonCommercial License 4.0 (CC BY-NC). [64] Copyright 2020, The authors, published by AAAS. d) Schematic of a conceptual RC system and dynamic vehicle flow analysis. Reproduced under the terms of Creative Commons License 4.0 (CC BY 4.0). [65] Copyright 2022, The authors, published by Wiley-VCH.
computational spaces through the reservoir. RC only needs to train the weight of the connection between the dynamic reservoir and output layer, which reduces the cost of network training. For example, Lao et al. proposed an ultralow-power in-sensor RC based on self-powered optoelectronic synapses. [65] The nonlinear and spatiotemporal-linked EPSC triggered by optical stimuli demonstrated the ability of the device to process complex temporal information. In the RC systems, the network size and training cost for recognition and classification can be greatly reduced by encoding the space information to time-series signals (Figure 5d). Both static face classification and dynamic vehicle flow analysis were performed by the in-sensor RC and demonstrated high recognition accuracies comparable with the simulation results.
Artificial vision sensors based on organic optoelectronic neuromorphic devices have made important progress in recent years, but there are still many difficulties and challenges. First, in terms of material synthesis, the organic optoelectronic neuromorphic sensors need to be improved for long-term environmental stability. This point requires the development of organic optoelectronic materials that are stable in long-term exposure to air or biological fluids as well as electronic packaging materials that can protect devices. It is also important to better understand the molecular design rules and develop new organic semiconductor materials with excellent optoelectronic properties (broad spectrum and high response), mechanical properties, and biocompatibility. [79] Second, in terms of devices and array architectures, most of the organic optoelectronic neuromorphic sensors that have been reported so far are single devices, and rarely involve hardware networks containing multiple devices. They are mainly limited by the following aspects. 1) The uniformity between devices and the stability of individual devices are insufficient, which affects the accuracy of neuromorphic functional simulation and makes it difficult to achieve complex applications. It is necessary to further optimize the preparation and processing technologies [80] of the device to improve performance.
2) The high energy consumption of devices restricts the density of device integration, which is necessary to develop novel self-powered and ultralow energy consumption devices. [39,61,62,81,82] Finally, in terms of functional application, the complex functions such as image recognition realized by organic optoelectronic neuromorphic sensors still require the help of traditional circuits and software algorithms. Therefore, it is necessary to develop a new artificial vision system based on the integration of sensing, storage, and computing.

Organic Sensors for Tactile Perception
Human beings perceive the surrounding environment through the skin, which is the largest tactile organ. [83] Tactile receptors in the skin can sense multiple external stimuli, including mechanical force, temperature, etc. [84,85] In recent years, tactile sensors that simulate the tactile properties of human skin have gradually emerged with the rapid development of sensing technology and flexible devices based on organic materials. [9,35,[86][87][88][89] Tactile sensors sense and quantify stimuli by converting external stimuli into electrical signals. [28,[90][91][92][93] The perception techniques of different stimuli originate from the research on different effects and materials. Emerging tactile sensors have been widely used in electronic skin, intelligent prosthetics, robots, human-robot interaction, medical detection, and other fields. [20,94,95] In this section, tactile sensors are introduced into four categories: pressure, temperature, proximity, and multimode as summarized in Table 2.

Force/Pressure Sensing
Force/pressure sensors convert the mechanical deformation or pressure generated by the local force into electrical signals, and extract the magnitude and/or direction of the force/pressure through the change of the signal. Force/pressure sensors based on organic materials are attractive because the characteristics of organic materials can be chemically adjusted, and their relatively low modulus of elasticity makes them compressive. In addition, the micro-nanostructured design of the organic active layer makes force/pressure sensors high sensitivity, fast response, and long durability. Organic force/pressure sensors contain a variety of conduction mechanisms, each based on different materials and structures. The latest progress in resistive, capacitive, piezoelectric, and triboelectric organic force/pressure sensors is presented.
Resistive sensors detect the external force/pressure stimulus by measuring the change of resistance caused by structural deformation, including the resistance of the active material and contact resistance. Resistive sensors are promising candidates for low-force/pressure sensing due to their high sensitivity and simple structure. [34,47,96,[105][106][107][108][109][110] Lai et al. reported a contact resistance tactile sensor based on selfbulged silver nanowires (AgNW)/poly(ethylene terephthalate) and poly(dimethylsiloxane) (PDMS), which had a high sensitivity of 1.04 × 10 4 -6.57 × 10 6 kPa −1 in a low range pressure of Figure 6. Force/pressure sensors with organic materials for tactile perception. a) Schematic of triaxial resistive sensor. Reproduced with permission. [96] Copyright 2009, The Royal Society of Chemistry. b) A capacitive sensor to integrate PDMS membrane based on a pyramid structure with microhair structure. Reproduced with permission. [97] Copyright 2014, Wiley-VCH. c) A dual organic transistor containing a capacitive pressure sensor and a synaptic OFET. Reproduced with permission. [114] Copyright 2017, Wiley-VCH. d) Performance test of the flexible piezoelectric tactile sensor. Reproduced with permission. [22] Copyright 2020, Elsevier Ltd. e) Schematic of the triboelectric smart finger that can identify the material type and roughness. Reproduced under the terms of Creative Commons Attribution NonCommercial License 4.0 (CC BY-NC). [115] Copyright 2022, The authors, published by AAAS.
<3 kPa. [47] The sensors can detect various mechanical forces, including throat muscle movement and hand movement. In addition, in resistive sensors, the sensing layer with micronanostructure provides a small contact area with the electrode, which makes the contact resistance and conductive path significantly change when the device is subjected to external force, resulting in high sensitivity. Zhang et al. fabricated a highly sensitive resistive sensor with triaxial sensing by combining micropyramid PDMS and reduced graphene oxide (rGO) film. [96] The rGO/PDMS pattern was deformed by force, resulting in the change of the contact area between the rGO film and the electrode. Hence, the device showed high sensitivity of −1.71 kPa −1 in the low-pressure of 0-225 Pa. In addition, the magnitude and direction of a spatial force could be calculated by decoupling the relationship between the applied spatial force and the change of output current, which showed the ability of sensors to detect 3D force (Figure 6a).
Capacitive sensors usually detect the change of force/pressure stimulus according to the change in dielectric layer capacitance. [111][112][113] The capacitance C is obtained from the equation of C = 0 r A/d, where 0 is the dielectric constant. The change of electrode spacing d is used to reflect the normal force, the change of electrode positive area A can be applied to detect the shear force, and r is used to measure the force on certain special materials. High-mechanical sensitivity requires materials with low mechanical modulus such as PDMS and polyurethane (PU). [112] The design of micro-nanostructure patterns of capacitive sensors can also enhance the compressibility and variations of the effective dielectric constant value. [5,97,112] Mannsfeld et al. employed the pyramid-structured PDMS film in the capacitive sensor for the first time, which greatly improved the sensitivity and response time of the film compared with unstructured elastic film. [5] The patterned PDMS film as a dielectric layer was integrated into organic field-effect transistors (OFETs) array with the same excellent performance. Pang et al. fabricated a capacitive sensor, which integrates pyramid-structured PDMS film with a microhair structure (Figure 6b). [97] The microhair structure could accurately measure the microsignal from the skin surface and improve the signal-to-noise ratio. In addition, the combination of capacitive sensors and synaptic devices makes dynamic tactile perception come true. Zhang et al. prepared a dual organic transistor containing a capacitive pressure sensor and a synaptic OFET (Figure 6c). [114] The capacitance of the dielectric layer changed with the pressure so that the transistor exhibited variable electrical currents, which led to the delivery of continuous presynaptic spikes to the gate of synaptic OFET. The output current from this system extracted not only the pressure intensity but also the frequency and duration of the dynamic pressure to produce a spatial resolution image of perceived information.
Triboelectric and piezoelectric sensors are very sensitive to mechanical motion by touch and are suitable for detecting microvibrations. The external force causes the deformation of the piezoelectric material of the sensor, which leads to the charge polarization inside the materials and generates charges on the material surface. The triboelectric sensors contain two materials with different electronegativities and generate electrical voltage through electrostatic induction during contact and separation between them. Compared with resistance and capacitive sensors, triboelectric and piezoelectric sensors have a better dynamic response and lower power consumption, [4,8,22,98,[115][116][117][118][119][120][121][122][123][124][125][126] and they can also improve their sensing performance through micro-nanostructured functional layers. Piezoelectric sensors with micro-nanostructure possess high sensitivity due to the improvement of compressibility and strain, as well as reduce the mutual interference between pressure sensing sites. For the micro-nanostructure of triboelectric sensor, it improves the effective contact area to increase the friction charge density of the friction surface. Wang et al. reported a flexible piezoelectric tactile sensor by combining the piezoelectric capability of -phase polyvinylidene fluoride (PVDF) nanorod arrays and the signal amplification capability of OFETs ( Figure 6d). [22] PVDF nanorod arrays deformed under pressure, and the polarization voltage generated in this process could be used as the gate voltage of OFETs to turn on the transistor, realizing the conversion and amplification of pressure signals. Wu et al. further fabricated a self-powered neuromorphic tactile sensor based on triboelectric sensing. [98] The Al electrode layer and polyimide (PI):rGO hybrid layer constituted a single electrode vertical contact-separation triboelectric sensor. The sensor generated signals of different amplitudes through previous pressure stimulations. In addition, triboelectric sensors can also identify different materials combined with machine learning. [115] Qu et al. demonstrated a smart finger with four triboelectric sensors based on different organic friction layers, which could accurately identify the material type and roughness ( Figure 6e). [115] Since each material had a different ability to gain and lose electrons, the output signal of the sensor would be diverse when the triboelectric finger was contacted with different materials. Machine learning made it possible to balance the number of sensors and the prediction accuracy of smart fingers. With only four sensors, the accuracy of material recognition reached 96.8%. Moreover, the implantable tactile sensors based on the triboelectric nanogenerator are expected to help tactile recovery. [127] Shlomy et al. implanted a self-powered triboelectric sensor based on cellulose acetate butyrate and PDMS under the skin of rats to restore tactile sensation for injured rats. PDMS and fibrin glue were used as biocompatible isolation materials to prevent the device from contacting the surrounding physiological environment. The triboelectric sensor converted tactile pressure into electric potential, which was transmitted through cuff electrodes to stimulate healthy sensory nerves, resulting in tactile sensation.

Temperature Sensing
Temperature sensors can be used to determine the surface temperature of objects (or people) and provide rich information on thermal characteristics. [89] Traditional commercial temperature sensors are difficult to compatible with tactile sensors due to their inherent rigidity of materials. However, organic temperature sensors have good flexibility, excellent extensibility, flexible spatial resolution, and convenient integration, which are promising to artificial skin and multisensory human-computer interaction systems. [128,129] At present, organic temperature sensors can be mainly divided into the thermal type and thermoelectric type.
Thermal temperature sensors measure the temperature through the change of resistance with temperature. Temperature sensing elements can be integrated with FET to amplify signals, forming a temperature sensor system with higher sensitivities and reliabilities. [130,131] Hong et al. fabricated a thermal resistor-type temperature sensor based on polyaniline nanofiber, exhibiting ultrahigh linearity (R 2 = 0.998) and low hysteresis. [131] The temperature sensor and single-walled carbon nanotubes TFT array were combined to form a stretchable sensor array, which had mechanical stability under 30% biaxial tension, and the resulting space temperature mapping did not show any mechanical or electrical degradation (Figure 7a). In addition, thermalresponsive nanocomposites can be used for temperature sensing in a wider temperature range. [13,15] Ren et al. constructed a two-terminal thermistor organic flexible sensor by embedding a layer of discontinuous silver nanoparticles (AgNPs) in the pentacene film, which could measure the temperature in the range of 20-100°C. The conductivity and sensitivity of the thermistor could be controlled by adjusting the silver nanoparticle. [15] Temperature sensors and OFET were connected in series one by one to form a 16 × 16 temperature sensor array, which displayed 2D temperature information of objects with irregular contact. [15] Furthermore, printing can achieve low-cost and rapid device fabrication and integration, which provides a good platform for developing feasible wearable devices. Yamamoto et al. reported a temperature sensor printed from a mixture of the carbon nanotube (CNT) ink and poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS) solution. [13] The high sensitivity of the temperature sensor (0.89% per°C) was attributed to the electronic hopping at the interface between the CNT and PEDOT:PSS. The combination of a printed temperature sensor, three-axis acceleration sensor, and electrocardiogram (ECG) sensor could simultaneously detect skin temperature, heart rate, ultraviolet radiation, and body movement, which has great potential in health monitoring and other biomedical applications.
However, two-terminal thermistor sensors are easily limited by external interferences such as pressure and humidity, which results from a single parameter output (current or voltage) when measuring the temperature. Temperature sensors based on OFETs can measure temperature through multiparameter output (I DS , V DS , V T , and ), so as to effectively distinguish the impact of temperature and other environmental factors. [99,[132][133][134] Trung et al. developed a transparent and stretchable temperature sensor using rGO/PU composite materials as a temperatureresponsive active layer. [134] The temperature sensor based on OFETs showed a higher temperature response (1.34% per°C) than the resistive organic sensor (0.9% per°C). By integrating the temperature sensor vertically with the strain sensor array and attaching it to the human throat, the changes in stress and temperature could be simultaneously detected when the human drank the hot water (Figure 7b). Moreover, the thermal sensitivity of organic semiconductors can also be adjusted by barrier  [131] Copyright 2015, Wiley-VCH. b) The organic sensor monitoring images of human neck skin during drinking hot water, including temperature changes and muscle movements. Reproduced with permission. [134] Copyright 2015, Wiley-VCH. c) Energy band model diagrams at the grain boundary of polycrystalline semiconductors and applications of the thermosensitive OFET for temperature sensing. Reproduced under the terms of Creative Commons Attribution 4.0 International License (CC BY 4.0). [99] Copyright 2021, The authors, published by Springer Nature. d) Schematic, optic images, and electrical properties of tascPLA-based OFET temperature sensor array. Reproduced with permission. [133] Copyright 2015, Wiley-VCH. e) Optic images of the self-powered temperature sensor array and corresponding temperature distribution. Reproduced under the terms of creative commons attribution-noncommercial 3.0 unported (CC BY-NC 3.0). [100] Copyright 2018, The authors, published by The Royal Society of Chemistry. f) The structure of the ionic thermoelectric gating organic transistor. Reproduced under the terms of Creative Commons Attribution 4.0 International License (CC BY 4.0). [137] Copyright 2017, The authors, published by Springer Nature.
engineering. The charge carriers around the local bent band at the grain boundary of organic semiconductor polycrystalline films need to hop across the potential barrier through thermionic emission, so the effective height of the potential barrier plays a key role in the thermal emission charge transport (Figure 7c). Huang et al. reported that the effective height of the grain boundary barrier was tuned by precisely adjusting the grain size, thus realizing a strong temperature dependence of charge transport in organic semiconductors. [99] When effective potential barrier height reached a maximum at a grain size close to twice the Debye length, the temperature dependence of charge transport was the strongest. Through the above strategy, DNTT-based OFETs could be used as a high thermal sensitive temperature sensor to identify human touch and environmental thermal radiation (Figure 7c). In addition to the thermistor characteristics of the semiconductor layer, the thermal sensitivity of OFETs can also be significantly enhanced due to the charge-trapping effect by polar groups in the dielectric/semiconductor interface. Three-arm stereocomplex polylactide (tascPLA) with strong polar groups as a dielectric layer could significantly improve the thermal sensitivity of various OFETs above room temperature. [133] This was because the strong polar groups (carbonyl functional groups) in the polylactide layer induced multiple traps of different energy levels at the dielectric/semiconductor interface. With the increase in temperature, the deep traps at the high energy level slowly release charge carriers, and the shallow traps at the low energy level capture charge carriers for less time, so the charge carrier transport rate increases (Figure 7d).
In addition, organic thermoelectric temperature sensors are also widely studied because of their self-powered characteristics.  [102] Copyright 2020, The authors, published by AAAS. c) Fabrication and structure of the AgNWs-BC/PDMS capacitive optical fiber sensor and optical image of proximity piano. Reproduced with permission. [103] Copyright 2020, American Chemical Society. d) Screen-printed stretchable magnetoresistive sensor for touchless interactive electronics. Reproduced under the terms of creative commons attribution 4.0 international (CC BY 4.0). [104] Copyright 2021, The authors, published by Wiley-VCH.
Organic thermoelectric materials can reversibly obtain the direct conversion between heat and electricity through Seebeck and Peltier effects. [135] PEDOT:PSS is one of the most promising organic thermoelectric materials. [100,136] Jung et al. fabricated a wearable self-powered temperature sensor by printing PEDOT:PSS and AgNPs conductive inks. PEDOT:PSS, as a ptype thermoelectric material, was patterned and overlapped with the n-type thermoelectric material of AgNPs, which contributed to highly sensitive temperature detection. And the p-n junction of thermoelectric materials can generate the voltage according to temperature difference. [100] When the finger contacted the temperature sensor array printed on the stretchable fabric locally, the mapping array can be measured in real time to reflect the temperature change without an additional power supply (Figure 7e). Furthermore, the combination of thermoelectric devices and transistors can realize the signal amplification of temperature sensing, which further improves the practicability of thermoelectric materials. [137] Zhao et al. constructed a thermoelectric-gated organic transistor, which used a polyethyleneoxide (PEO)-NaOHbased ionic thermoelectric supercapacitors as the gate of the P3HT-based transistor. [137] The device converted the temperature fluctuation into a significant change of drain current, thereby increasing the sensitivity by more than an order of magnitude, which will have a key application in electronic skin and health detection. (Figure 7f).

Proximity Sensing
In addition to accurate tactile perception through direct contact, proximity perception is also important in tactile perception. In general, vision is easy to be blocked in near-environment detection, so we need proximity sensors to replace the vision system to locate and track the contacted objects. Moreover, proximity sensing can also prevent humans or robots from contact with unnecessary hazards to effectively improve safety. [138] Organic proximity sensors based on thermal radiation, humidity, electrostatic, and magnetic effects are introduced.
Thermal radiation and humidity effects are two critical sensing mechanisms for detecting the proximity of objects by measuring the temperature or humidity, such as human fingers. [101,102,139] Cao et al. fabricated a thermal radiation proximity sensor by incorporating Ti 3 C 2 T x NPs-lamellae hybrid networks into PDMS substrates. [101] PDMS substrate expanded when the external temperature rose, causing the conductive network to stretch and the Ti 3 C 2 T x NPs-lamellae hybrid networks to slip, which changed the resistance of the device. The sensor could detect the proximity of fingers within 9 cm, as well as the proximity of vials with different temperatures (Figure 8a). Kang et al. reported a 3D noncontact color sensing display based on the block copolymer photonic crystal and hygroscopic ionic liquids doping. [102] When the finger was close to the device surface, the capacitance of the humidity sensor changed, and the position of the human finger could be effectively visualized (Figure 8b).
The proximity sensor based on the electrostatic coupling effect alters the electrical signals via the change of device capacitance caused by an object approaching. [103,[140][141][142][143][144][145] Guan et al. prepared a capacitive optical fiber sensor based on the AgNWs-bacterial cellulose (BC) fiber electrode, realizing the pressure and proximity sensing (Figure 8c). [103] AgNWs-BC electrode and PDMS dielectric formed a core-shell structure. The proximity piano was constructed and played based on the 1 × 7 orthogonal optical fiber sensor array, where the intersection of each fiber was a note of a piano. In addition, compared with other proximity sensing effects, the magnetic effect has higher reliability because it is not affected by the environment (such as temperature, humidity, etc.) and nonmagnetic objects. [104,[146][147][148][149] Ha et al. constructed a stretchable magnetoresistive sensor by screen printing. [104] The use of the elastomer triblock copolymer as a binder for magnetosensitive paste enhanced the mechanical properties of materials and sensitivity under small magnetic fields. The position change of the magnetoresistive sensor connected to the human fingertip relative to the magnet could be converted into an electrical signal change (Figure 8d). Finger with the sensor successfully controlled the scrolling of electronic documents and the zooming of maps remotely, which showed the potential of the magnetoresistive sensor in the field of human-robot interface based on noncontact interaction.

Multimodal Tactile Sensing
Human skin can sense multiple external stimuli at the same time. Therefore, it is very important to develop multimodal tactile sensing systems that can simultaneously process multiple stimuli. On the one hand, multimodal tactile sensors can be realized via a single tactile sensor that can respond to multiple stimuli. [44,150,151] Tactile sensors with a single sensing mechanism can achieve multimodality by identifying the signal differences generated by different stimuli. [150,[152][153][154] Park et al. reported multimodal tactile sensors based on the ferroelectric rGO/PVDF composite film, which could detect pressure, temperature, and microvibration. [150] The interlocked microdome arrays on the film surface enhanced the piezoresistive, piezoelectric, and pyroelectric sensing of the film, improving the sensitivity of the sensor (Figure 9a). Lee et al. used machine learning algorithms to effectively decouple mixed pressure and temperature stimuli detected by the organic cross-reactive sensor matrix. It could simplify the complexity of the device architecture and improve environmental adaptability (Figure 9b). [151] However, due to the interference between multiple signals, it is difficult for multimodal tactile sensors based on a single sensing mechanism to detect different stimuli accurately and sensitively. [45,[155][156][157] When a single sensor has independent sensing effects on different stimuli, decoupling analysis is no longer necessary. Zhang et al. fabricated a self-powered organic thermoelectric sensor based on a microstructure framework, which had a dual-parameter sensing mechanism. [157] Sensors could decouple temperature and pressure stimuli into voltage and current signals through thermoelectric and piezoresistive effects, respectively. Electronic skin based on a dual-parameter sensor array was attached to the artificial hand to obtain spatially resolved images of temperature and pressure when in contact with the external world (Figure 9c). On the other hand, tactile sensors with different responses can be integrated into a single platform, which makes it easy to distinguish mixed signals. [158][159][160][161] Hua et al. proposed 3D high-density integrated multimodal tactile sensors by scalable integrated multiple tactile sensor units. [158] The multimodal tactile sensors were based on a structured polyimide network with eight sensing units integrated through multiple layers of stacking, and the sensors could respond to three stimuli simultaneously (Figure 9d). However, high-density interconnection brings technological difficulties, which still need to be further studied and solved.

Auditory Perception
The basilar membrane is an important part of the biological ear. For living organisms, the ability to distinguish acoustic vibrations with different frequencies is dependent on the mechanical properties of the basilar membrane with a trapeziform geometry in the cochlear. [162] The narrow part of the basilar membrane can respond to high-frequency stimuli, while the wide part responds to low-frequency stimuli. [42,48] Inspired by the structure of the biological basilar membrane, Jang et al. designed a triboelectric-based artificial basilar membrane (TEABM) based on a beam structure formed by the stacking of Kapton films and aluminum foils (Figure 10a). [42] Acoustic-to-electric transduction was realized through triboelectrification. The TEABM exhibited sensitivity to sound in the range of 1.74-13.1 mV Pa −1 and frequency selectivity in the range of 294.8-2311 Hz. By connecting a deafened guinea pig with the TEABM, the auditory brainstem response was successfully evoked, demonstrating the validity of the artificial auditory system. In order to further enhance frequency discrimination, Gong et al. designed a soft resistive artificial basilar membrane based on a suspended pointcracked vertically aligned gold nanowire film (Figure 10b). [48] The artificial basilar membrane was able to respond to sound pressure with 0.48 Pa -1 and discriminate the sound with frequencies up to 3000 Hz, demonstrating superior acoustic frequency discrimination and extremely high sensitivity. The eardrum in the biological ear has the capability to absorb acoustic oscillation and transmit signals to the brain. Yang et al. developed a selfpowered eardrum-inspire sensor based on a multilayered polytetrafluoroethylene (PTFE)/nylon/ITO/poly-ethylene terephthalate structure. [163] Artificial eardrum showed ultrawide frequency ranges from 0.1 to 3.2 kHz, a rapid response speed of 6 ms, and superior sensitivity of 51 mV Pa -1 .

Olfactory Perception
Olfactory is an important function of living organisms, contributing to recognizing different smells, perceiving dangerous gas, and avoiding a toxic atmosphere. [164] For the future development of bionic robots, the realization of biological olfactory-like perception in emerging electronics is necessary. Wang et al. fabricated a Figure 9. Multimodal tactile sensors with organic materials for tactile perception. a) Multimodal tactile sensors for simultaneous detection of temperature and pressure. Reproduced under the terms of creative commons attribution-noncommercial license (CC BY-NC 4.0). [150] Copyright 2015, The authors, published by AAAS. b) Multimodal tactile sensors based on machine learning algorithm decoupling mixed stimuli. Reproduced with permission. [151] Copyright 2020, Wiley-VCH. c) Schematic diagram and performance of self-powered dual-parameter organic sensors. Reproduced under the terms of Creative Commons Attribution 4.0 International License (CC BY 4.0). [157] Copyright 2015, The authors, published by Springer Nature. d) Illustrative schematic of 3D integrated multimodal tactile sensor array and performances of devices with three stimuli. Reproduced under the terms of Creative Commons Attribution 4.0 International License (CC BY 4.0). [158] Copyright 2018, The authors, published by Springer Nature.
highly sensitive gas sensor based on polyaniline/multiwalled carbon nanotubes (MWCNTs) composite thin film, which emulated the biological olfactory. The sensor can accurately detect ammonia (NH 3 ) concentration at room temperature (Figure 10c). [165] It exhibited a response of 10% at an extremely low concentration of 0.01 ppm NH 3 and an ultrahigh response of 255% at 100 ppm NH 3 . The results indicated that the sensor had an extremely low detecting threshold and high sensitivity. Despite the impressive progress in the property exploration of gas sensors, further performance and structure optimization of sensors still urgently needs to be carried out. Xue et al. designed a flexible-smelling electronic skin based on sandwich polyaniline (PANI)/PTFE/PANI nanostructure. [40] The movement of PTFE film between two PANI films contributed to electricity generation under the effect of gas flow. Meanwhile, PANI chains could be changed by volatile organic compounds, which resulted in the change in triboelectric charges on the PANI films. Through the coupling of triboelectrification/gas-sensing, the smelling electronic skin was able to rapidly detect various volatile organic compounds in the air without any external electrical power. Moreover, the excellent flexibility allowed the smelling electronic skin to operate under different bending status. The work opens up a promising direction for the development of novel olfactory perception systems. Figure 10. Auditory, olfactory, and gustatory sensors with organic materials for tactile perception. a) Schematic of the TEABM, and animal testing using the sensor. Reproduced with permission. [42] Copyright 2016, Wiley-VCH. b) Schematic of the fabrication process of the resistive acoustic sensor based on a suspended point-cracked vertically aligned gold nanowire film and the resistance response. Reproduced with permission. [48] Copyright 2020, Wiley-VCH. c) Output voltage and output current of triboelectric nanogenerator based on different PTFE films. Reproduced with permission. [165] Copyright 2018, Elsevier Ltd. d) Tasting beverages based on the artificial sensory-substitution system. Reproduced with permission. [168] Copyright 2018, Elsevier Ltd.

Gustatory Perception
Gustatory is one of the most important physiological senses in living organisms. Various gustatory receptors are mainly located on the surface of the tongue, responsible for the reception and recognition of different smell information. [43,166,167] In order to emulate biological gustatory perception, Shimizu et al. proposed a microfluidic electronic tongue composed of a single piece of PDMS, five parallel microwires coated with metal films, and one perpendicular microwire responsible to pump the sample solutions. [50] Each film-coated microwire represented a sensing unit of the electronic tongue. Compared with conventional electronic tongues, the microfluidic electronic tongue exhibited global selectivity. According to the change in electrical impedance during testing, the chemical composition in the sample solutions could be easily classified through a single measurement without recalibration. Fu et al. designed a self-powered artificial sensorysubstitution system based on a composite film consisting of Cu, PDMS, and polymerizing polypyrrole. [168] The system could re-place the gustatory organ of the sensory-handicap patient to perceive smell information. According to the difference in output current against different drinks, the taste of the beverage was effectively classified by the system (Figure 10d).

Future-Oriented Intelligent Applications Based on Organic Sensors
In the previous section, we mentioned that bioinspired organic sensors had made considerable breakthroughs, providing machines with excellent sensory capabilities. In recent years, with the development of IoTs and AI, sensors need to collect multiple information and process intelligently with high efficiency and low power consumption when facing complex scenes, which puts higher requirements on organic sensors. In this section, we provide examples of organic sensors for future-oriented intelligent applications, including associative learning, information security, electronic skin, and integrated artificial sensory systems.  [39] Copyright 2021, The authors, published by Science and Technology Review Publishing House. b) The image of the integrated optoelectronic sensor array and the demonstration of the unrecoverable erasure induced by light illumination. Reproduced with permission. [181] Copyright 2016, Wiley-VCH. c) Schematic of the phototransistor array for photoinduced encryption and decryption. Reproduced with permission. [182] Copyright 2019, American Chemical Society.

Intelligent Applications Based on Organic Optoelectronic Sensors
Organic optoelectronic sensors have the advantages of high adjustability, reliability, and sensitivity. The development and application of organic photoelectric sensors will significantly help to improve the future intelligent infrastructure and system. Here, the applications of organic optoelectronic sensors in associative learning and information security are introduced.

Associative Learning
Living organisms enhance their cognition abilities and connections with the environment by continuous learning during growth. Therefore, simulating the human brain's learning process is vital to achieving artificial intelligent applications. Associative learning is a crucial learning principle that combines discrete ideas with perception, which is significant for individuals to adapt quickly to the environment. [169][170][171] Pavlovian dog experiment is the best-known classical associative learning. [172] In this experiment, food and the bell ring are regarded as the unconditioned and neutral stimuli. Only when dogs were trained in neutral and unconditioned stimulus simultaneously could they produce saliva (unconditional response) to respond to the neutral stimulus. It has been reported that combining organic optoelectronic sensors and synaptic devices can achieve associative learning, which is conducive to simplifying complex circuit design.
Pavlovian associative learning can be demonstrated in optoelectronic devices by coupling electrical signals and optical signals. [39,52,173,174] Li et al. simulated associative learning using an organic optoelectronic synapse based on the ferroelectric dielectric of poly(vinylidene fluoride-cotrifluoroethylene) (P(VDF-TrFE)). [173] The synaptic device exhibited STP property under electrical stimuli. With the assistance of light pulses, the transition from short-term memory to long-term memory occurred due to ferroelectric polarization, which facilitated the establishment of an effective relationship between electrical and light signals. In order to achieve associate learning, the electrical pulse applied to the device was considered as a neutral stimulus, and the light pulse was considered as an unconditioned stimulus. Associative unconditioned stimulus and neutral stimulus learning were realized by coupling electrical and optical signals. The neutral and unconditioned stimulus could trigger a conditioned reflex, leading to Pavlovian associative learning. Furthermore, the highly adjustable associative learning in a single device can better simulate the adaptive ability of biology. Recently, Pei et al. demonstrated a smarter Pavlovian dog based on an organic ferroelectric neuromem with the ultrathin P(VDF-TrFE) film, showing adjustable associative learning by exploiting the photo-ferroelectric coupling (Figure 11a). [39] During the training, ferroelectric materials presented different polarization states under voltage modulation, contributing to a high tunability conductance state. Pavlovian dog, modulated by electrical signals, was successfully simulated after thirteen training epochs. When UV light illuminated the device, the photogate effect induced by interfacial traps stabilized the ferroelectric polarization, resulting in the rapid increase of channel current. Therefore, the training pulses were reduced from 13 to 2 under the light. Meanwhile, the channel current of the device exhibited a distinguishable difference over 10 3 between the unconditional response and conditioned response, similar to the all-or-nothing behavior in biological sensory neurons.
In addition to the electrical and optical stimulus, associative learning modulated by different physics signals is beneficial for coping with more complex external scenes. Ji et al. designed an associative learning circuit modulated by light and pressure inputs. [174] The circuit integrated a pressure sensor, a photoresistor, a volatile organic electrochemical transistor (OECT) based on PEDOT:tosylate, and a nonvolatile OECT based on polytetrahydrofuran. Pressure input and light input were regarded as unconditioned stimulus and conditioned stimulus, respectively. When unconditioned or conditioned stimulus was applied to the associative learning circuit alone, the nonvolatile OECT could not be triggered, resulting in rapid decay of the transient current because most of the gate voltage was applied to the volatile devices. The gate voltage applied on the nonvolatile OECT was sufficient to induce memory property and trigger an unconditional response only when unconditioned and conditioned stimuli were applied together. After five training epochs, the unconditional response could be triggered by a conditioned stimulus without the assistance of the unconditioned stimulus. Therefore, the Pavlovian dog modulated by two physical signals (pressure and light) was successfully demonstrated in the circuit.

Information Security
With the wide application of information technology in all walks of life, it has become urgent to store and protect information securely. Because optical information processing has high speed, parallelism, multidimensional properties, and low cost, it has natural advantages compared with electronic means. [175][176][177][178] Here we will introduce the application of organic optoelectronic sensors in the field of information security.
In conventional memory devices, even if the information in the memory is erased electronically, it could also be restored. Organic optoelectronic sensors combined with chemical destruction methods can effectively erase stored data and make the data unrecoverable, which provides absolute security for the stored information. [3,179,180] Lee et al. designed an ultrathin acid-destructible optoelectronic sensor coated with multidye-sensitized upconverting nanoparticles (UCNPs)/photoacid-generators (PAGs)/PEO matrix for information security applications. [181] Different types of sensitizers bounded to the surface of the UCNPs contributed to a broad absorption spectrum extended from the NIR to the visible range. Under the light illumination (from vis to NIR), the UCNPs could produce the emission of ultraviolet (UV) light due to the UV/blueemitting Tm 3+ ions doping. The PAGs immediately absorbed the energy of the emitting UV photons and generated photoacid, which resulted in the rapid chemical destruction of data stored in the flexible sensor array under the illumination of 800 nm light. The ultrathin and deformable hardware systems presented excellent information security, indicating the application potential in wearable security electronics (Figure 11b).
Under extreme conditions, chemical destruction methods can prevent important information from being stolen. However, it is also important to achieve reversible information encryption and decryption of electronic devices in information security. Recently, Xu et al. designed hybrid organic phototransistors based on the C 8 -BTBT single-crystal array coated with CH 3 NH 3 PbI 3 NPs, achieving information encryption and decryption based on wavelength selectivity (Figure 11c). [182] The C 8 -BTBT films dominated the light response of the device under UV light illumination, and NPs played a dominant role in visible light illumination due to ultrahigh light-harvesting efficiency. Under UV light irritation, the patterned C 8 -BTBT film absorbed the photons and reduced the light penetration. The array can appear the "SOS" character due to the photocurrent reduction in the covered area. By contrast, when visible light irritated the array, the "SOS" character became invisible due to ultrahigh transparent (>90%) of the C 8 -BTBT film in the visible range, indicating the successful demonstration of secure communication.

Electronic Skin Based on Organic Tactile Sensors
Electronic skin is an advanced bioinspired sensor committed to simulating the characteristics of human skin. Electronic skin not only has the ability of the organic tactile sensors to sense multiple external stimuli but also has the ability of flexibility, extensibility, and self-healing similar to biological skin. [44,183,184] Electronic skin is expected to become the interface between electronic devices and the environment, playing an important role in intelligent robots, bionic artificial limbs, and human-computer interaction. [185][186][187][188] In addition, electronic skin combined with neuromorphic devices can realize brain-like signal processing and learning behavior. [189] Here, we focus on the applications of electronic skin based on bioinspired organic tactile sensors.
The practical application of electronic skin needs to rely on solution-processed organic materials with diverse chemical and physical properties to achieve large-area, flexibility, elasticity, and self-healing. However, it is difficult to obtain conformal contact between electronic skin and 3D irregular surface. [190][191][192] Direct spraying on the 3D surface is a promising strategy to fabricate the large-area 3D device. Park and co-workers prepared porous MWCNTs-PDMS-based 3D tactile sensors on different 3D irregular surfaces by the solution-based spray-coating method, highly suitable for application to electronic skin. [190] The stretchability of the electronic skin can be increased by patterns such as honeycomb grid to achieve conformal contact. Takahashi et al. used the laser to cut the polymer substrate into a honeycomb mesh structure so that the sensors had good stretchability and adhesion. [191] The implantable ultraflexible multielectrode arrays (MEAs) are used to establish conformal contact with soft biological organs. Lee et al. fabricated the electronic skin based on active MEAs with the honeycomb grid substrate to directly contact the dynamically beating heart of a rat (Figure 12a). [193] PMC3A coating provided antithrombotic properties of the device, ensuring effective ECG monitoring even with capillary bleeding. Furthermore, emerging 3D printing technology could dynamically adapt to the shape of the target in real time to convenient the fabrication of electronic skin. Zhu et al. developed an adaptive 3D printing method that  [193] Copyright 2018, The authors, published by AAAS. b) Illustrative schematic of adaptive 3D printing on a 3D surface and a moving human hand. Reproduced with permission. [192] Copyright 2018, Wiley-VCH. c) Diagram of the self-healing composite and demonstrations of integrating self-healing electronic skin into a humanoid mannequin. Reproduced with permission. [194] Copyright 2012, Springer Nature. d) Demonstrations of self-healable electronic skin with strain sensing. Reproduced with permission. [25] Copyright 2017, Wiley-VCH. e) Dynamic reconstruction of conductive nanowire network. Reproduced with permission. [195] Copyright 2018, Springer Nature.
can produce multifunction organic sensors on moving curved surfaces (Figure 12b). [192] The changes of states in the 3D printing workspace in terms of the geometries and motions of target surfaces can be perceived by an integrated robotic system aided by computer vision. 3D printing can print the graphics of various parts of the device at the same time, providing new intelligent manufacturing.
Another important feature of electronic skin is that it can be repaired in case of mechanical damage under repeated mechanical stretching, similar to the self-healing function of natural skin. This feature can improve the durability of electronic skin. Selfhealing can be achieved via the inherent reversibility of molecular interactions of polymer materials. [25,[194][195][196] Tee et al. used organic-inorganic composite materials to fabricate an electronic skin that can be repeatedly healed at room temperature (Figure 12c). [194] The polymer chains at the fracture interface of the film could be rearranged, wetted, and diffused via the action of hydrogen bonds. Electronic skin could be integrated into a humanoid mannequin to detect pressure and bending of the prosthesis. Wang et al. prepared the self-healable electronic skin with strain and pressure sensing abilities based on PANI, polyacrylic acid, and phytic acid ternary composite materials (Figure 12d). [25] Due to the contribution of hydrogen bonds and electrostatic interaction, fractured conductive polymer film recovered 99% of its electrical and mechanical properties within 24 h. Son et al. embedded the self-healing polymer into conductive nanowires to fabricate electronic skin (Figure 12e). [195] The nanowire network could be reconnected as the polymers dynamically bind, allowing the electronic skin to recover conductivity and mechanical properties.
Electronic skin based on the tactile sensor can enable robots or prostheses to perceive the surrounding environment, manipulate objects, and interact with the outside world, which is crucial for the future development of medical and industrial fields. [197][198][199][200] Kim et al. constructed a PDMS-based multifunctional electronic skin prosthesis. [197] It could detect pressure, strain, humidity, temperature, and other stimuli, including handshake signals, typing, surface temperature, and moisture of contact objects (Figure 13a). Integrating the humidity sensor and resistance heater can adjust skin humidity and body temperature. Stretchable nanoelectrodes were connected to the nervous system to transmit electrical signals generated by stimulation. Gerratt et al. fabricated a soft electronic skin based on an array of organic capacitive sensors with multipoint touch and dynamic detection. [198] The electronic skin combined with a human-in-the-loop system with visual feedback enables the manipulator to grasp the object. Furthermore, to grasp and manipulate objects flexibly, robots and prosthetics need to detect the slip and direction of forces accurately. The electronic skin based on PU with 3D pyramid microstructures completed real-time detection of normal force and shear force. [199] This delicate tactile signal extraction avoided damaging fragile fruit when the manipulator touches it (Figure 13b). This accurate perception of force facilitated the more flexible operation of the robot arm. Reproduced with permission. [197] Copyright 2014, Springer Nature. b) Experimental setup of a robot arm attached with electronic skin and operation diagram of the mechanical arm with or without tactile feedback. Reproduced with permission. [199] Copyright 2018, AAAS.
Combining electronic skin and the artificial neural system provides prosthetics and robots with intelligent capabilities of biological signal processing and perception. [11,34,[201][202][203] Wan et al. fabricated an artificial sensory electronic skin consisting of a CNTs/PDMS-based pressure sensor, a polyvinyl alcohol-based ionic cable, and an organic synaptic transistor (Figure 14a). [202] The artificial neuron recognized and learned the tactile pattern by extracting tactile information. The mechanical receptors in the simulated organism generated spike signals, which were expected to interact directly with the nervous system. Tee et al. used an organic ring oscillator to convert the pressure signal generated by the piezoresistive sensor into a spike signal (Figure 14b). [34] This system with digital output characteristics was integrated into a prosthetic device. Tan et al. constructed an optoelectronic spiking afferent nerve. [11] Flexible organic pressure sensors were coupled with light-emitting diodes and analog-to-digital converters to convert pressure signals into optical spikes (Figure 14c). Then, the spike signals were transmitted to the photosensitive synapse for processing. Combined with a neural network algorithm, the artificial neural system can recognize Morse code, braille, object movement, and handwritten words. [203] Osborn et al. fabricated a multilayer electronic skin by imitating the layered structure of tactile receptors in human skin. The neuromorphic sensor converted the pressure signal generated in the electronic skin into spike signals. The electronic skin could help amputees feel tactile stimula-tion and pain through the prosthesis to avoid further injury (Figure 14d). [203]

Integrated Organic Sensors for Artificial Sensory Systems
Human beings can detect, collect, integrate, and process information from different senses in real time and react quickly to cope with a complex environment. Therefore, it is of great significance to use electronic devices to simulate multifunctional perception and feedback regulation capabilities for developing artificial perception systems for soft biological robots, humancomputer interfaces, and prosthetics. Here, we will introduce the integrated organic sensor based on multifunction perception and feedback functions for artificial sensory systems.

Integrated Organic Sensors for Multifunctional Perception
Multifunctional perception of biology is to integrate the perceptive information from different sensory organs into a comprehensive consciousness to eliminate the distinction of external stimuli and improve responsiveness. [204][205][206][207] In addition, interference from the external environment may affect the accuracy of a single perception. The combination of multiple senses can help better detect complex environments. Therefore, integrated organic Figure 14. Electronic skin prosthesis with neuromorphic devices. a) Schematic diagram of the neuromorphic tactile processing system (NeuTap) imitating sensory neurons. Reproduced with permission. [202] Copyright 2018, Wiley-VCH. b) Organic digital mechanoreceptor consisting of a pressuresensitive tactile element and an organic ring oscillator. Reproduced with permission. [34] Copyright 2015,AAAS. c) Schematic diagram of the biological and artificial optoelectronic spiking afferent nerve systems. Reproduced under the terms of Creative Commons Attribution 4.0 International License (CC BY 4.0). [11] Copyright 2020, The authors, published by Springer Nature. d) Diagram of prosthesis system applied with neuromorphic multilayered electronic skin. Reproduced with permission. [203] Copyright 2018, AAAS.
sensors for multifunctional perception into intelligent bionic applications are crucial for implementing more complex tasks.
When multiple sensory organs synchronously receive external stimuli, the brain of living organisms can weaken the influence aroused by noise from the external environment, contributing to a higher sensitivity than a single sensory organ. Similarly, multifunctional perception sensors can improve the sensitivity and accuracy of perception. Wu et al. proposed an artificial multisensory integration nervous (AMIN) system, exhibiting tactile and visual sensory modalities (Figure 15a). [45] By combining the PDMS-based triboelectric nanogenerator with the organic optoelectronic synapse, external tactile,and optical stimuli were effectively fused and converted into electrical signals. Under the regulation of multiple sensory stimuli, the system showed superior synaptic plasticity. The recognition accuracy based on the system was significantly improved from 68.02% to 80.55%. In addition, under dark or bright background, the object could not be clearly recognized with a single vision sensor. With the assistance of the tactile sensor, the image contrast obtained significant improvement. The environment-adaptable perception behavior further demonstrated the superiority of the AMIN system for object recognition.
Even though multifunctional perception integration of multiple information can enhance recognition accuracy and precision, the separation between sensors inevitably incurs computing time and energy costs due to signal transmission. Developing a single device with multifunctional perception is a promising method to overcome the issue. Liu et al. proposed a P3HT/PEO NWs-based intrinsically stretchable organic neuromorphic transistor (NISNT), which integrated tactile sensing, visual perception, and synaptic information processing functions (Figure 15b). [208] Because of the stretchable feature, the increased channel length changed the electrical characteristics when the device was stretched. Therefore, the neuromorphic devices could sense and recognize different gestures. The gesture datasets were input into a neural network, and a recognition rate of ≈92.6% was attained. In addition, the NISNT with the light-sensitive feature could also implement optical sensing. By combining tactile and visual information, the accuracy of gesture recognition was further improved to 96.3%, which showed that integrated organic sensors for multifunctional perception were significant in future intelligent bionic applications.

Integrated Organic Sensors with Feedback Functions
Biological somatosensory systems can respond to external stimuli in real time, dependent on the transduction of external stimuli into bioelectricity signals by sensory organs, signal transmission through neurons, and signal processing and decision-making by the brain. These feedback functions provide living organisms with the ability to shape their interactions with complex external environments. [17,20,[209][210][211][212] Therefore, the feedback functions are greatly significant for intelligent bionic applications to realize independent decision-making.
The biological reflex arc allows the cooperation of neurons and muscles to complete the reflex actions, which is the representation of the feedback function. Artificial reflex arc becomes a typical application in bioinspired sensory electronics with feedback functions. Lee et al. reported an artificial optoelectronic sensorimotor nervous system based on a stretchable organic nanowire synaptic transistor (s-ONWST), achieving the artificial reflex arc (Figure 16a). [17] The self-powered organic photodetector system converted optical signals into electrical signals and triggered the Figure 15. Integrated organic sensors for multifunctional perception. a) Simulation of the multisensory integration nervous system for object recognition in normal or extreme environments. Reproduced with permission. [45] Copyright 2021, Elsevier Ltd. b) Schematic of gesture recognition using the multisensory integration system, including visual and tactile perception. Reproduced with permission. [208] Copyright 2022, American Chemical Society.
EPSC to operate the actuator. The organic photodetector exhibited ultrafast response time (<1 ms), which eliminated temporal mismatch between optical inputs and electrical outputs, similar to the biological sensory organs. Furthermore, the nervous system demonstrated light-interactive actuation by connecting artificial muscles based on a polymer actuator with the system by a transimpedance circuit. The s-ONWST generated a small output voltage (≈1 V) without light illumination, leading to a slight contraction of the polymer actuator. With the continuous application of optical pulses, the increased output voltage resulted in the obvious displacement of the actuator. The optical-actuated artificial muscle demonstrates that the sensorimotor nervous system integrating optical sensor, processor, and actuator is an ideal platform to develop next-generation biomimetic robotics.
In addition to optical stimuli, the somatosensory system of living organisms can respond to tactile stimuli. Kim et al. designed a flexible organic artificial afferent nerve to fabricate an organic artificial reflex arc (Figure 16b). [20] The organic afferent nerve integrating pressure sensors, ring oscillators, and artificial synapses could transform tactile stimuli into voltage signals to trigger EPSC. The synaptic device with multiple electrodes could combine voltage signals from different ring oscillators connecting with multiple pressure sensors. The artificial afferent nerves were utilized to identify braille characters that were pressed on a 3 × 2 sensory array. The peak frequencies of EPSC well reproduced the input braille characters. Besides, the artificial afferent nerve was connected with the efferent nerves of a discoid cockroach to complete a hybrid bioelectronic reflex arc. The tactile information from pressure sensors flowed through the artificial afferent nerve to deliver postsynaptic signals into efferent nerves in a detached cockroach leg. The tibial extensor muscle in the leg was successfully actuated. The artificial sensory system endowed with smart feedback functions affords promising applications in human-robot interaction. Chen and co-workers proposed a human-robot interaction paradigm based on an artificial somatic reflex system integrating an organic tactile sensor based on PDMS film, a threshold controlling unit, and a carbon nanotube-based electrochemical actuator (Figure 17a). [213] The system based on thresholddependent behavior simplified complex circuit design. Only when pressure stimuli reached the threshold could the system generate an action potential to operate the electrochemical actuator without abundant computation, similar to the all-or-none law in biological neurons. The system was incorporated into a 3D-printed robot to emulate the infant grasp reflex. By applying touch stimuli on the robot's chest, the fingers equipped with the actuator strips could complete the grasp reflex. The artificial somatic reflex arc system brings different insights into the  [17] Copyright 2018, The authors, published by AAAS. b) Artificial afferent nerve to actuate the tibial extensor muscle in the leg of a cockroach. Reproduced with permission. [20] Copyright 2018, AAAS. simplification of autonomous control and presents promising applications in human-robot interaction.
Currently, most studies have achieved a single mode of sensory feedback functions. Realizing synchronous multisensory perception and feedback in a simple electrical hardware system is conducive to the practical application of future humanrobot interaction. [38,41,214] The kirigami technique enables onestep programmable customization of diverse electric device prototypes. Wang et al. designed organic somatosensory light-driven robots (SLiRs) with embedded multiple sensory systems by using kirigami technique, exhibiting great potential in human-robot interaction (Figure 17b). [214] Due to the photothermal effect, the actuator could translate the real-time light stimulus to a piezoresistive signal, leading to actuation displacement, thus providing feedback to external stimuli. An SLiR anthropomorphic hand with five fingers representing somatosensory receptors could perceive the temperature, softness/hardness of object, and complete specific movements. Moreover, the antenna of centipede was designed to interact with the environment and human interventions. An untethered SLiRs centipede could implement multifunctional tasks, such as autonomous walking, adaptive locomotion, and various sensing. This SLiR with perception and feedback exhibits promising opportunities for developing humanrobot interaction and closed-loop control of the actuation system.

Conclusion
The development and research of new-generation bioinspired organic sensory devices require a deeper understanding on the material properties, device structures as well as sensing mechanisms, which guide the optimizations of electrical performances and advanced intelligent functionalities of bioinspired organic sensors. In recent years, considerable progress has been obtained in the development of bioinspired organic sensory devices. The intelligent applications based on bioinspired organic sensors also have risen as a vital strategic research area with enormous potential. In this review, we summarize recent trends in research on bioinspired organic sensors: i) the materials, sensing mechanisms, and functionalities of bioinspired organic visual sensors, tactile sensors as well as other sensors for auditory, olfactory, and gustatory perception; ii) the future-oriented intelligent applications of bioinspired organic sensors including associative learning, information security, electronic skin, and intelligent multisensory systems. The integration of both the front-end multiple signal perception and back-end data processing functions in a bioinspired organic sensory system can significantly improve performance and efficiency. The comprehensive understanding of bioinspired organic sensors and intelligent processing applications are a significant step toward next-generation electronics. Figure 17. Integrated organic sensors with feedback functions for human-robot interaction. a) Photograph of a baby presenting the grasp reflex and a robot combined with the artificial somatic reflex arc systems to emulate the grasp reflex. Reproduced with permission. [213] Copyright 2019, Wiley-VCH. b) Photograph of an SLiR hand with the different bending states under the stimulations of the external environment and a schematic of an SLiR centipede based on wireless antenna sensing. Reproduced with permission. [214] Copyright 2020, Wiley-VCH.
Despite recent progress in bioinspired organic sensors, there are still many technological challenges and opportunities. First, the practical application of visual systems inspired by human retinas and eyes is still challenging due to the trade-offs between the high resolution and high curvature of the optical sensor array. The emerging ultrathin organic films, especially 2D films have become prosperous due to high uniform morphology, scalable solution-based preparation, and intrinsic electrical performance, which enabling fabrications of the optical sensor array with excellent curvature and resolution. Second, the stability and endurance of organic sensory devices, which determine the practicability and sensitivity of the bionic system, have become fundamental merits for the real-world application and product commercialization. More stable organic functional materials and lowcost encapsulation strategies without sacrificing the mechanical flexibility are highly needed. Third, despite the fast development of neuromorphic sensory devices that integrates the sensing and processing capabilities, most of the devices still need electrical modulations for learning and training by using computers. In addition, the poor linearity and symmetry of artificial sensory synapses may result in lower learning efficiency and computing accuracy. The use of heterojunction neuromorphic devices with strong optoelectronic coupling properties has emerged as a promising strategy to improve the performance of neuromorphic devices, which deserves more in-depth study and applica-tions. Finally, the integrated system comprising separated parts could incur high energy consumption in the data transmission. Therefore, it is crucial to develop simplified integrated system with multisensory and neuromorphic functions for low-power artificial sensory systems via advanced structure designs and methods.
Therefore, developing bioinspired organic sensors integrating both sensing and processing functionalities is emerging as an important direction for highly efficient and intelligent perceptive systems. To date, bioinspired organic sensory devices have achieved encouraging achievements by finely optimizing materials, device structures as well as utilization of new sensing mechanisms. These important understandings and strategies will provide guidance for the design of next-generation bioinspired organic sensors for various advanced applications. Yun Li is a professor at the School of Electronic Science and Engineering, Nanjing University. He has undergraduate and Ph.D. degrees in physics from Nanjing University. From 2010 to 2012, he worked as a postdoctoral researcher at National Institute for Materials Science (NIMS), Japan. His current research concentrates on organic electronics, including the growth and device physics of 2D molecular crystals, organic ferroelectric memories, and low-power organic optoelectronic devices.