Skin plays a crucial role as the primary interface and protective barrier between the human body and its surrounding environment^[1−4]. The integration of the sensory capabilities and mechanical flexibility not only enables effortless perception and response to various external stimuli but also facilitates complex tasks in dynamic, unstructured settings^[5−7]. Thus developing an artificial neuromorphic skin system is indispensable for applications such as wearable electronics, health monitoring, intelligent robots, internet of things (IoT), artificial neural networks. In biological systems, ion transport serves as a fundamental mechanism for biological activity, encompassing the generation and transduction of efficient neural sensing signals. Ionogel exhibits high ionic conductivity and exceptional mechanical properties, while its unique three-dimensional spatial network structure facilitates ion diffusion and transfer, which is extensively used in sensing, flexible electronics, and other domains. He et al.^[8] proposes a photopolymerization-induced microphase separation strategy to prepare highly conductive and stretchable nanostructure (CSN) ionogels, and its high compatibility with DLP 3D printing enables the fabrication of complex ionogel microstructures with high resolution (up to 5 μm), enabling the fabrication of capacitive sensors with superior sensing performance. Similarly, Wang et al.^[9] reports a method to make two monomers with different solubility in ionic liquids randomly copolymerize, producing elastic and rigid domains of phase separation, resulting in an ultra-tough and stretchable ionogel. In addition to excellent mechanical properties, supertough copolymer ionogels also exhibit good self-healing properties, excellent self-healing properties and excellent shape memory properties^[10−12]. Therefore, By integrating with light-sensing materials, it is feasible for achieving artificial skin with visual perception capabilities.

The photothermoelectric effect (PTE) is a unique photocurrent mechanism that enables the conversion of energy between light, heat, and electricity. Unlike other photoelectric detection mechanisms, PTE does not require an external electric field to separate the electrons and hole pairs generated by light excitation^[13−15]. Instead, it utilizes thermal carrier diffusion within the device to establish a temperature gradient for generating electricity. The PTE has emerged as a novel photoelectric detection mechanism in recent years due to its advantages of zero-bias operation, wide spectrum response, and unrestricted band gap. Local surface plasmon resonance (LSPR) refers to the phenomenon where metal nanoparticles exhibit strong absorption of photon energy when the frequency of incident light matches their overall vibration frequency or that of metal conduction electrons^[16−20]. When the plasma metal nanoparticles are exposed to light, the internal conducting electrons undergo oscillation in response to the incident magnetic field under the dual influence of the applied electric field and the Coulomb force caused by electron displacement. This oscillatory behavior leads to electromagnetic field localization around the plasma metal nanoparticles, resulting in rapid generation of non-thermal electrons with exceedingly high energies within a short period^[21−24]. Subsequently, there is swift heating due to electron−electron scattering for a brief duration, followed by heat generation through electron−phonon scattering, which is then rapidly transferred to the surrounding environment^[25−27].

In this paper, we propose an ionogel synapse with the configuration of plasma silver nanoparticles (AgNP) doped Ionogel heterostructure. The synapse exhibits important synaptic behaviors such as excitatory postsynaptic current (EPSC), paired-pulse facilitated (PPF, 124%), learning and forgetting experiences as well as the transition from short-term potentiation (STP) to long-term potentiation (LTP), with the electrical signals are attributed to the temperature driven ions migration due to the LSPR of AgNP. The synapse also demonstrates exceptional mechanical flexibility and self-healing capability derived from the gel matrix. Finally, a neuromorphic visual skin array with 5 × 5 ionogel synapses was demonstrated, which is capable of autonomous and comprehensive recognizing, learning and memorizing of diverse patterns in both curved and flat configurations. The study provides a feasible route in building artificial visual sensing skin towards smart soft robotic systems.

Materials: Monomers of 4-hydroxybutyl acrylate (HBA) and 2−(2−ethoxyethoxy) ethylacrylate (EOEOEA), the solid ionic salt bis(trifluoromethane)sulfonimide lithium salt (LiTFSI), the photoinitiator 1−hydroxycyclohexyl phenyl ketone (HCPK) were purchased from sigma-aldrich. Sodium dodecyl sulfate (SDS) and sodium borohydride, silver nitrate was purchased from sinopharm chemical reagent Co. All reagents were used as received without further purification.

Synthesis of Ag nanoparticles (AgNPs): Firstly, 17 mg silver nitrate, 510 mg SDS were placed in 250 ml flask, 50 ml of deionized water was added, and 0.1 mg/ml of sodium borohydride was quickly injected when the oil bath was heated to 100 ℃. Finally, it was sucked out and placed in 4 ℃ of ice water with a rubber head dropper.

Synthesis of ionogel and AgNPs-doped ionogel: The ionogel was synthesized through a one-step co-photopolymerization of HBA and EOEOEA monomers. In a typical procedure, the monomers HBA and EOEOEA, and crosslinker ETPTA were mixed at a volume ratio of 1 : 4 : 0.05. With the addition of photoinitiator HCPK (1% mass ratio concentration to the mixed solution) and solid ionic salt LiTFSI (0.5 M mole ratio concentration to the mixed solution), the mixed solution was stirred at 500 rpm for 30 min to obtain a uniform precursor ink, from which 2 mL of the precursor ink was poured into a polytetrafluoroethylene (PTFE) mould and exposed to 365 nm UV light at 10 W for 2 min to produce a transparent ionogel. To synthesize AgNPs-doped ionogel, the same procedure was followed, but with the addition of AgNPs to the precursor ink.

Preparation of ionogel heterojunction: The ionogel heterojunction was prepared using a PTFE mould with a thin PTFE clapboard in the middle of mould. Firstly, the precursor ink without AgNPs was added into one part of the mould, and expose it to 365 nm UV light at 10 W for 2 min to obtain a transparent ionogel. Next, with removal of clapboard, another ink containing AgNPs was injected into the other part of mould. After irradiation with the same UV source for 4 min, a yellow AgNPs-doped ionogel was solidified, and finally achieve the formation of a heterojunction due to the tight contact between the two ionogels.

Fabrication of two-terminal optoelectronic synaptic device: prior to the polymerization of the ionogel heterojunction, the two copper strips were respectively polymerized in situ inside the ionic liquid. This process led to the fabrication of ionogel heterojunction based optosynaptic devices.

Characterization: He UV−vision−infrared absorption spectra of AgNPs were recorded by PE UV−1750 spectrophotometer. All photoelectric measurements are made using the keithley 4200 semiconductor parameter analyzer. Light pulses with adjustable wavelength, intensity and frequency come from LED drivers (THORLABS, DC 2200 terminal). A range of LED light sources are used, covering wavelengths from ultraviolet to visible and near-infrared. All thermal images were taken using an infrared camera (FLIR Ti 100). The optical power density was measured by an optical power meter (CEL−NP 2000). The light source system is housed in a keithley 4200 semiconductor shield box to prevent interference from external light signals. All electrical and optical measurements are made at atmospheric pressure and room temperature.

Artificial synapses, as a crucial element of human neural systems, have been extensively investigated in recent decades. Current studies on neuromorphic devices primarily focus on electrically controlled synapses, which exhibit significant disparities in power consumption and functional applications compared to the human brain, thereby constraining the advancement of neuromorphic synaptic devices. Emerging neuromorphic synaptic devices not only demonstrate remarkable progress in optical and electrical performance and biocompatibility but also offer distinct advantages in reducing device power consumption and expanding application domains^[28−32]. Fig. 1(a) depicts an artificial flexible visual skin array comprising 25 individual optoelectronic synapse units, fabricated on a PET substrate with copper electrodes deposited in an electrode pattern. Each unit features a heterostructure with two distinct ends, one is ionogel doped with silver nanoparticles, while the other end remained untreated. The synthesized ionogel demonstrates excellent transparency and flexibility, with (lithium bis(trifluoromethylsulfonyl imide)) provide free lithium ions. The photo-thermoelectric effect induced by the local surface plasmon resonance of doped Ag nanoparticles (AgNP) would drive the ions to migrate spontaneously within the gel under illumination, thus enable the synaptic unit self-power capability and light perception. Impressively, the doped Ag nanoparticles expands the absorption range of the gel, resulting in prominent absorption across the UV−visible−infrared spectrum, while minimally impacting its flexibility and transparency. The optical image of an ionogel array was displayed in Fig. 1(d). The whole array possesses small dimensions, with a maximum side length of only 5 cm. Each unit functions as a skin sensor, ensuring high sensitivity and consistent performance across the entire device surface. Given its straightforward fabrication, cost-effectiveness, and convenient transferability, it presents an attractive prospect for future large-scale production.

The self-powered photoresponse of a single photosynapse was investigated by observing the variation in ionic current under illumination (Fig. 2(a)). When the synapse is exposed to a UV pulse (14.4 μW∙mm⁻², 365 nm), a significant increase in the device current with prolonged exposure time was observed at a 0 V bias. Upon removal of the light source, the device current gradually decreases and returns to its initial state (Fig. 2(b)). The synaptic unit exhibits a rise time of 5.03 s and a fall time of 16.93 s, with the delayed photocurrent attributed to the slow ion flux dynamics, thus could simulate various functions of biological synapses. The ionogel synapse demonstrated wide-band response thanks to the incorporation of AgNP (Fig. 2(c)). The original AgNP aqueous solution exhibited a broad exciton absorption and a sharp exciton absorption peak at 414 nm. Upon doping AgNPs into pure ionogel, the resulting AgNPs-ionogel displayed a more prominent broad absorption peak at 400 nm compared to the pure gel, thus the absorption band of the ionogel synapse is further widened to the ultraviolet region (Fig. 2(d)). The ionogel synapse also exhibits roubust cyclic light response at various light power densities (Fig. 2(e)), with its photocurrent escalates from 1.33 to 4.06 nA response to the increased light power density (2.88 to 14.4 μW∙mm⁻²). Additionally, the photoresponse performance of the ionogel synapse hinges closely on its cross-sectional area. As the cross-sectional area increases, the responsive photoelectric current decreases from 3.19 to 2.94 nA, and the response time prolonged from 15.68 to 85.33 s (Fig. 2(f)). Consequently, smaller areas can generate higher photoelectric currents and faster response speeds compared to larger areas, allowing for increased array density and high-resolution imaging.

The intrinsic PPC of ionogel synapse enables the capability of mimicking fundamental learning and memory functions in human brain^[33−35]. The photo synaptic features were studied by measuring the EPSC under different wavelengths (Fig. 3(a)). The ionogel photosynapse demonstrated stable response across ultraviolet to visible light spectrum, with the device displayed superior sensitivity towards ultraviolet 365 nm light. Paired-pulse facilitation (PPF) serves as a crucial metric for assessing the synaptic strength between presynaptic and postsynaptic membranes in the biological nervous system, representing an important aspect of short-term plasticity. We effectively replicated this function with our ionogel synapse (inset in Fig. 3(b)) by employing two identical continuous pulses under a 365 nm ultraviolet light source with a power density of 14.4 μW∙mm⁻² and an intermediate interval time (DT) of 1 s. The figure illustrates that the peak of the second pulse surpasses that of the first response peak due to prolonged ion migration caused by the initial pulse, leading to additional ion migration induced by the second pulse and resulting in ion accumulation. The PPF index (A₂/A₁ × 100%) is proposed to comprehensively evaluate the PPF characteristics of our device, where A₁ and A₂ represent photocurrent values triggered by the first and second light pulses respectively. Our synaptic device achieved a maximum PPF index of 124% with and DT of 100 ms, which exponentially decreased as DT increased (Fig. 3(b)). The conversion of short-term potentiation (STP) to long-term potentiation (LTP) is essential for memory and forgetting models in biological brains. In a simulation where different numbers of UV pulse trains with the same pulse frequency were used to stimulate synapses (Fig. 3(c)), increasing the number of UV pulse trains from 2 to 11 extended the decay time from the original 23 to 59 s, indicating a transition from STP to LTP. This transition process is critical for understanding the key mechanisms of memory maintenance and enhancement in simulating biological brains. Similarly, with the increase of illumination time, the decay time after light removal is significantly extended, which also realizes this transformation process (Fig. 3(d)). Furthermore, the EPSC of the synaptic device can also be adjusted by manipulating the intensity and numbers of light pulses (Fig. 3(e)). Fig. 3(f) illustrates the mimicking of biological learning experience behavior based on our device. Initially, the EPSC shows a significant increase in the first 20 pulses (365 nm, 14.4 μW∙mm^-2, 1 Hz) followed by spontaneous decay upon removal of light, reflecting the process of learning and forgetting. Subsequent learning processes demonstrate that the EPSC returns to its initial level after each cycle, with only 8 light pulses required for the second process and 6 for the third. The results indicate that after three learning cycles, the time required for re-learning is significantly reduced compared to the initial learning period. Importantly, following multiple learning sessions, there is a notable increase in forgetting time for this synapse from an initial period of 18 to finally 40 s. This suggests that continuous repetition of learning stimuli can enhance the device retention ability, a feature consistent with ebbinghaus' theory on biological learning and relearning principles.

The skin, as the largest human organ, not only acts as a physical barrier to protect the body, but also serves as a crucial interface for communication with the external environment. Its surface and interior are abundant in various types of receptors. An ideal artificial skin should possess sensory abilities comparable to those of human skin and exhibit similar properties such as softness and repairability. Furthermore, it should be capable of maintaining its original performance after deformation and healing itself after injury^[36−38]. Therefore, researching its regenerative properties and flexibility is immensely significant^[39−41]. To assess the self-healing capability of the ionogel synapse, we cleaved it along the interface between two distinct gels, i.e., the interface between pure gel and AgNP doped gel. Subsequently, the two gels can rapidly heal at 60 °C without the addition of any external solvents or materials. It is noteworthy that upon retesting the synaptic performance of the healed gel, we observed a very similar EPSC response in comparison to the synaptic gel before cutting, further confirming its excellent self-healing performance (Fig. 4(a)). The mechanical flexibility of the ionogel synapse was also investigated by transferring it into various circular molds with differing radii (r = 2, 1.5, 1, 0.75 mm) and assessed its synaptic performance (Fig. 4(b)). The findings indicate that even under significant bending deformation (bending radius of 0.75 mm), the device exhibits robust synaptic plasticity across different bending states.

As a type of noble metal nanoparticles, AgNP exhibits strong light absorption in the visible spectrum due to the surface plasmon resonance (SPR) effect, which arises from the collective oscillation of free electrons on their surfaces in response to incident light. When the incident light frequency matches its SPR frequency, the effective absorption of light energy would lead to an increase in the surface temperature of the particle. Under illumination, the temperature gradient formed within the ionogel heterojunction due to the increase in particle surface temperature in the doped region, further facilitates ion movement. However, the disparate migration rates of cations and anions (Li and TFSI species) result in an internal imbalance of ion concentration, leading to a greater infiltration of lithium ions into the pure gel region, establishing an internal electric field within the ionogel heterojunction and generates ion current. This internal electric field also hinders the diffusion of lithium ions while promoting the diffusion of TFSI ions until reaching equilibrium. Upon removal of the light stimulus, the photothermal effect of the nanoparticles dissipated, causing the temperature of the doped region to gradually return to room temperature, and the ions within the ionogel also reverted back to their original positions.

The dependence of device temperature and photocurrent on illumination time was also studied under a fixed light intensity, with the ultraviolet light cover the whole ionogel heterojunction to allow for comparison of the temperature difference between the two ends. The ionogel heterojunction exhibits a rapid increase in temperature and tends to reach saturation under ultraviolet irradiation (Fig. 5(b)). The maximum temperature difference between the doped region and the pure gel portion reaches 8 °C. This substantial temperature gradient drives Li⁺ ions to migrate towards lower temperatures region (pure gel part), resulting in an ionic current of 4 nA. These pronounced variations in both temperature and current highlight the exceptional photothermoelectric conversion properties of AgNP doped ionogel heterojunction. The variation of the internal electric field strength ΔU of the ionogel heterojunction with temperature ΔT was also recorded (Fig. 5(c)). The fitting data demonstrated a linear enhancement in ΔU as ΔT increased, which is consistent with the Soret effect, S = ΔU/ΔT, where S represents the Soret coefficient that typically characterizes thermal diffusion in ionogels. Furthermore, we conducted a comparative analysis on the photoelectric performance of pure ionogel, AgNP doped ionogel, and AgNP doped ionogel heterojunction (Fig. 5(d)). In comparison to doped and pure ionogel samples (<0.5 nA, which may originate from light-induced ionization of defects)^[42], the ionogel heterojunction exhibited exceptional photoelectric response (~2.5 nA), providing evidence that photocurrent is primarily governed by ion current induced by the photothermoelectric effect of doped AgNPs.

Excellent light perception and synaptic plasticity enable the ionogel heterojuncion for efficient optical signal processing and visual system imitation^[43−45]. The diagram in Fig. 6(a) illustrates an artificial neuromorphic visual skin array, comprising 25 ionogel synapses arranged in a 5 × 5 visual perception array, each of which functions autonomously and demonstrates excellent synaptic performance in both flat and curved situations (Fig. 6(b)). This skin array can be affixed to a surface and exhibit diverse images based on the lighting conditions. We affix it onto the dorsal side of a human hand seamlessly, and the skin array showcasing exceptional neuromorphic properties even during hand gestures such as making a fist or an open palm (Fig. 6(c)). When illuminating the skin array through a 'E'-shaped mask, it registers an elevated current at its 'E' position compared to its surrounding dark conditions—demonstrating exceptional precision in pattern recognition. The EPSC associated with this 'E'-shaped configuration increases proportionally with each additional light pulse applied, resulting in gradual refinement of its visual representation. Upon cessation of illumination pulses, there is a gradual decline in EPSC for this 'E'-pattern leading to its eventual fading away; during this decay phase there is an initial rapid decline followed by slower attenuation until complete disappearance occurs. Notably, this distinct photoelectric current phenomenon persists for some time post-illumination removal—a testament to outstanding data retention capabilities within our device (Fig. 6(d)).

In summary, we propose a self-healing, bendable, and self-powered photosynapse based on AgNPs and ionogels. The synapse demonstrates broad spectral response, biological synaptic plasticity, exceptional self-repair capability, and mechanical flexibility. Experimental analysis reveals that the synaptic features originate from the photothermal effect of LSPR and temperature gradient induced ion migration within the ionogel heterojunction. Furthermore, we have successfully demonstrated an artificial visual perception skin array comprising 25 ionogel synapses, achieving perceptual memory for various shapes in both curved and flat configurations. Our work lays a foundation for intelligent electronic skins and neural-inspired visual systems.