Bionic tactile-visual-morphic system based on oxide photoelectronic neuromorphic transistor and triboelectric nanogenerator

Abstract

Recently, hardware based bionic perceptual systems have attracted great attention. However, most reported bionic perceptual systems only have a single perceptual function, making it difficult to mimic multi-sensory perception process in real environments. Here, a bionic tactile-visual-morphic system (BTVMS) is proposed by integrating an indium gallium zinc oxide (IGZO) photoelectronic neuromorphic transistor (PNT) and a PDMS-ZnO based triboelectric nanogenerator (TENG). The IGZO-PNT exhibits stable electrical performance and can sensitively perceive optical stimuli. The PDMS-ZnO based TENG can convert mechanical stimuli into electrical signals, exhibiting a high sensitivity of ~0.75 V/kPa and good durability. The proposed BTVMS exhibits information encryption and decryption functions based on Morse code strategy. In addition, it can simulate the tactile and visual dual cognition behavior of brain. Thus, post-traumatic stress disorder behavior has been mimicked successfully. The present BTVMS provides a valuable idea for intelligent prosthetics and humanoid robots to achieve efficient bionic visual and tactile perceptions.

Graphical abstract

Keywords

oxide photoelectronic neuromorphic transistor / triboelectric nanogenerator / bionic tactile-visual-morphic system (BTVMS)

1 Introduction

Artificial intelligence (AI) is undergoing a disruptive transformation [1−3]. Recently, an open-source AI model called DeepSeek quickly becomes popular worldwide due to its outstanding performance [4, 5]. As comparison, traditional central processing units (CPUs) are facing serious problems such as insufficient computing efficiency and rapidly increased power consumption, which makes them inadequate to meet the needs of the AI in the future [6, 7]. Unlike the conventional von Neumann architecture, our brain can efficiently process complex tasks with extremely low energy consumption [8, 9]. Relying on huge amounts of neurons and synapses, the brain can simultaneously process massive information. Additionally, our brain possesses powerful adaptive learning abilities. The strength of synaptic connections can be adjusted based on external stimuli, achieving efficient learning and memory activities. Therefore, developing neuromorphic devices that can mimic the computing mode of the brain is an effective way to overcome the bottleneck of AI. Thus, brain inspired neuromorphic devices have attracted great attention in neuromorphic engineering. Such devices include two terminal memristors [10−14] and multi terminal transistors [14−19]. Memristor based neuromorphic devices have a simple structure and exhibit priorities in three-dimensional integrations, enabling low-power information storage and computation. While the transistor based neuromorphic devices would execute learning and updating synaptic weight synchronously. Moreover, multi-gates can also be designed on a single neuromorphic transistor, exhibiting structural similarities to biological neurons. As implement the dendrite integration and neural algorithm [20−22].

On the other hand, brain synergetically integrates multi-modal sensory data to comprehensively perceive the outside world, such as vision, touch and audition [23, 24]. Thus, mimicking perceptual functions on neuromorphic devices is becoming a hot topic in neuromorphic electronics, exhibiting potentials in sensing-memory-computing integration for advanced intelligent portable applications. Such achievements include visual perception [25, 26], tactile perception [27−29], auditory perception [30−32], etc. Additionally, our daily activities usually require the synergy of multiple perceptions. Through multimodal perception, brain can form a comprehensive and accurate cognition of the environment. Fig.1(a) schematically shows the formation of visual and tactile perceptions. When external optical stimuli enter the eye, the photoreceptors on the retina convert optical signals into nerve impulses. These neural impulses are transmitted to subsequent neurons through synapses and then transmitted to brain via the optic nerve. The visual center of brain integrates and processes this information to form vision [33, 34]. Similarly, when pressure stimuli are applied on the skin, the tactile receptors within the skin are activated, converting the pressure stimuli into neural impulses. These neural impulses are transmitted along sensory neurons and arrive at the brain. Then, the somatosensory cortex of the brain integrates and processes the information to form tactile perception [35, 36].

For living organisms, continuous synergy between perception and action is crucial in a multisensory environment. Therefore, bionic perceptual systems with a single perceptual function show limitations in high-fidelity biomimicry. The simulation of the synergistic effect of vision and touch is thus significant for developing realistic and efficient humanoid robots. It is reported that indium gallium zinc oxide (IGZO) is highly sensitive to light illumination due to persistent photoconductivity (PPC) effect [37, 38]. Thus, photoelectronic neuromorphic transistors (PNT) using IGZO as channel can simulate visual activities by optical stimuli. In addition, low power consumption is highly appreciated for energy efficient neuromorphic electronics. Interestingly, triboelectric nanogenerator (TENG) couples frictional electrification and electrostatic induction to efficiently convert mechanical stimuli into electrical signals, showing great potentials in self-powered sensors and wearable devices [39−41]. Although single TENG unit can perceive pressure stimuli, the PNTs could integrate and amplify electrical signals, contributing accurate perception of external pressures. Therefore, integration of PNTs with TENGs is expected to simulate the synergistic effect of vision and touch, contributing to the high-fidelity biomimicry of perceptual activities. Here, a bionic tactile-visual-morphic system (BTVMS) is proposed by integrating an IGZO-PNT and a PDMS-ZnO based TENG, as schematically shown in Fig.1(b). The IGZO-PNT exhibits stable electrical performance and can sensitively perceive optical stimuli. The PDMS-ZnO based TENG can effectively convert pressure stimuli into electrical signals. The proposed BTVMS exhibits information encryption and decryption. In addition, it can synergistically respond to pressure and optic dual-modal stimuli. Thus, post-traumatic stress disorder (PTSD) behavior has been imitated. The present BTVMS provides an interesting idea for developing realistic humanoid robots.

2 Experiment

Figure S1(a) schematically shows the flow chart of the IGZO-PNT. Firstly, hydrogel solution was obtained by dissolving 10 wt% fish gelatin (GEL) and 4 wt% carboxylated chitosan (CCs) in 86 wt% deionized water. The mixture was stirred at 70 °C with rate of 400 r·min⁻¹ in a magnetic stirrer. Next, the GEL-CCs solution was drop-coated onto a conductive substrate at the speed of 2000 r·min⁻¹ for 20 s. After dried at 50 °C for 20 min in atmosphere ambient, a solid-state GEL-CCs electrolyte film was obtained. Then, IGZO channels were deposited on the GEL-CCs film by sputtering IGZO target (In₂O₃:Ga₂O₃:ZnO = 1:1:1 at%) using a metal shadow mask in pure Ar ambient. During sputtering, RF power, sputtering pressure, and Ar flow rate were set to 80 W, 1.0 Pa and 57 sccm, respectively. Due to the reflection of IGZO nanoparticles at the mask edge, a “U” shaped IGZO channel layer can be formed on the GEL-CCs film. Afterwards, ITO electrodes were formed by sputtering ITO target (In₂O₃:SnO₂ = 90:10 wt%) using the same mask. The DC power, sputtering pressure, and Ar flow rate were set to 80 W, 0.8 Pa and 41 sccm, respectively. Figure S2(a) shows the photograph of the IGZO-PNT.

Figure S1(b) schematically shows the flow chart of the triboelectric nanogenerator (TENG). Firstly, polydimethylsiloxane (PDMS) elastomer and curing agent were mixed in a mass ratio of 1:10. Then, zinc oxide (ZnO) powder was added to the mixture at different mass ratio. After thoroughly stirring, the solutions were drop-coated on polyethylene terephthalate (PET) substrate and dried at 80 °C for 2 h. The peeled solid-state PDMS-ZnO membranes were then used for obtaining relative dielectric constant. The prepared solutions were also drop-coated on sanding papers and dried at 80 °C for 2 h. Next, the solid-state PDMS-ZnO membranes were peeled off from the sanding papers to act as friction layers. Copper tapes were attached to PET substrates to act as electrodes. Then, kapton membrane and PDMS-ZnO membrane were attached to the top and bottom electrodes, respectively. The kapton membrane was deemed as another friction layer. Two conducting wires could be drawn out from electrodes for testing. Here, adding a certain proportion of ZnO powder in PDMS could increase the dielectric constant of the friction layer and achieve more effective charge transferring between two friction layers, improving the performances [42, 43]. Figure S2(b) shows the photograph of the PDMS-ZnO based TENG.

The thickness of GEL-CCs based electrolyte film was characterized using scanning electron microscopy (SEM). Surface morphologies of kapton membrane and GEL-CCs film were characterized using atomic force microscopy (AFM). Frequency dependent capacitance of GEL-CCs based electrolyte film was characterized by using an IM3533 impedance analyzer. A linear motor (LPS1.2) was adopted to load pressures on TENG. Output voltages of TENG were recorded with an oscilloscope. Electrical performances of the IGZO-PNT and the BTVMS were measured with a semiconductor parameter analyzer (Keithley 4200A SCS). All the electrical characterizations were performed in a dark box with air relative humidity of ~50%.

3 Results and discussion

Fig.2(a) shows the cross-sectional SEM image of the GEL-CCs based electrolyte film. The film is uniform with thickness of ~4.6 μm. Fig.2(b) shows AFM surface morphology image of the GEL-CCs film. The surface root-mean-square roughness (rms) value is ~1.2 nm. Fig.2(c) shows frequency dependent capacitance of the GEL-CCs electrolyte film. A big specific capacitance of ~2.5 μF/cm² is obtained at 1 Hz. It can be attributed to the formation of electric-double-layer (EDL) at the GEL-CCs/IGZO interface. Fig.2(d) shows leakage current of the electrolyte film. It is quite small of <4 nA at the bias ranged between −1.5 V and 1.5 V, indicating good dielectric properties. Fig.2(e) shows output curves of the IGZO-PNT measured at V_gs altered from 0 V to 1.6 V with steps of 0.4 V under the dark. The results indicate good ohmic contact at the S/D electrode. Fig.2(f) shows ten repeated transfer curves of the IGZO-PNT at constant V_ds of 2 V under the dark. The curves approach with each other very well, indicating stable electrical performances. Anti-clockwise hysteresis window of ~0.6 V is observed, which can be attributed to the migration of protons within the GEL-CCs electrolyte. As is similar to those in mesoporous silica [44] and pectin/chitosan hybrid electrolyte [45]. The ON/OFF ratio, subthreshold swing, threshold voltage, and mobility are estimated to be ~2.7 × 10⁵, ~178 mV/dec, ~0.2 V, and ~1.1 cm²/(V⋅s), respectively. Basically, there is obvious persistent photoconductivity (PPC) effects in IGZO film [37, 38]. Its carrier concentration would increase significantly under light illumination, resulting in a persistent change in conductivity. Therefore, electrical performance of the IGZO-PNT maybe modulated by light illumination. Fig.2(g) shows the transfer curves taken under different optical power intensities at a fixed light wavelength of 365 nm. The transfer curves will drift upwards with the increased optical power intensity. The I_ds value increases from 6.2 × 10⁻⁸ A to 3.1×10⁻⁵ A at V_gs of 0 V with the increased optical power intensity from 0 to 26.1 mW/cm².

In biological system, once excitatory neurotransmitters bind to the receptors on postsynaptic membrane, excitatory postsynaptic current (EPSC) will be triggered [46, 47]. Such EPSC response is of great significance in the transmission of neural signals. Interestingly, GEL-CCs electrolyte gated IGZO-PNT can mimic the EPSC response due to the proton related EDL effect. Fig.2(h) shows typical EPSC responses triggered by spikes with different amplitudes detected with a constant V_ds of 0.5 V. The EPSC peak value increases from ~0.1 μA to ~1.3 μA with the increased spike amplitude from 0.5 V to 2.0 V. Fig.2(i) shows EPSC responses triggered by 10 consecutive spikes (1 V, 10 ms) under different optical power intensities. The frequency of spikes and the light duration are 10 Hz and 1 s, respectively. The EPSC peak value is only ~0.5 μA under dark condition. While it gradually increases with the increased optical power intensity. Under the optical power intensity of 26.1 mW/cm², the EPSC peak value reaches ~1.32 μA. The results hint the synergetic effects between the gate coupling and the photo coupling.

Interestingly, due to the PPC effect of IGZO, typical EPSC response can also be triggered by optical stimuli. Fig.3(a) schematically shows the mechanism of PPC effect in IGZO. With light illumination, oxygen vacancy (V_O) will ionize into shallow donor states (V_O⁺, V_O²⁺), causing an increase in electron density in IGZO [38, 48]. Therefore, the conductivity of the channel will increase, triggering an EPSC response. When the optical spike ends, V_O²⁺ captures electrons and restores back to V_O. However, this process requires additional activation energy and proceeds slowly, resulting in a slow decay of EPSC back to its initial state. Fig.3(b) shows typical EPSC response triggered by an optical spike (3.4 mW/cm², 10 ms). The EPSC suddenly increases to ~63 nA and slowly decays to the initial current level in ~15 s. Fig.3(c) shows optical spike intensity dependent EPSC responses. The spike width is fixed at 10 ms. The EPSC value increases with the increased optical intensity. The EPSC peak value is only ~63 nA under the optical intensity of ~3.4 mW/cm². While it reaches ~169 nA when the optical intensity increases to ~26.1 mW/cm². Figure S3(a) shows EPSC peak values under single optical spike with different power intensities. The EPSC peak value increases linearly with the increased optical power intensity. Here, an optical sensitivity can be defined by the following equation: SE_O = ΔI/ΔP_O, where I and P_O are the EPSC peak value and optical power intensity, respectively. Thus, SE_O value is ~4.7 nA·cm²/mW. Optical spike width also greatly affects the EPSC response, as shown in Fig.3(d). With the fixed optical intensity at 3.4 mW/cm², the EPSC response gradually increases with the increased spike width. The EPSC peak value reaches ~244 nA with the spike width of 200 ms.

Fig.3(e) shows typical EPSC response triggered by paired optical spikes (3.4 mW/cm², 10 ms) with an interval time (ΔT) of 40 ms. The second absolute EPSC value (A₂) is higher than the first one (A₁). Thus, biological paired-pulse facilitation (PPF) effects have been mimicked. Here, PPF index is defined as A₂/A₁ × 100%. Thus, PPF index is ~165% when ΔT is 40 ms. Fig.3(f) shows PPF index as a function of ΔT. For ΔT at 20 ms, the PPF index is ~178%. It drops sharply with the increased ∆T value. Then, it gradually deceases to ~115% at higher ΔT values. The curve could be fitted by a double exponential decay function:

(1)\mathrm{P}\mathrm{P}\mathrm{F}\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}=C_0{+C}_1\cdot \mathrm{e}\mathrm{x}\mathrm{p}(-\mathrm{\Delta }T/{\tau }_1{)+C}_2\cdot \mathrm{e}\mathrm{x}\mathrm{p}(-\mathrm{\Delta }T/{\tau }_2),

where C₁/C₂ and {\tau }_1/{\tau }_2 are initial facilitation magnitudes and characteristic relaxation times of the two decay phases, respectively. The fitting gives C₀, C₁, C₂, {\tau }_1 and {\tau }_2 values of ~1.15, ~0.54, ~0.24, ~62.2 ms and ~825.6 ms, respectively. Fig.3(g) shows spike number dependent EPSC responses with optical spikes (3.4 mW/cm², 10 ms). As spike number increases, the EPSC increases accordingly. The EPSC peak value triggered by 50 consecutive spikes reaches ~479 nA. In addition, the frequency of spikes also has a significant influence on EPSC response. Figure S3(b) shows the EPSC response triggered by 10 consecutive optical spikes (3.4 mW/cm², 10 ms) at different frequencies. The EPSC gain is defined as A₁₀/A₁ × 100%. As shown in Fig. S3(c), the gain value increases with the increased spike frequency. The gain value is ~170% at the spike frequency of 1 Hz. While it reaches ~270% at the spike frequency of 50 Hz. Due to the PPC effects of IGZO, not only the optical spike duration and intensity but also the optical spike history would affect the EPSC decay kinetics. As results in the imitation of “Learning-Forgetting-Relearning” activities by optical spikes, as shown in Fig.3(h). A learning threshold is set at 870 nA. At the first learning stage, 50 consecutive optical spikes (26.1 mW/cm², 50 ms) with spike interval of 200 ms will induce EPSC peak value of ~876 nA, indicating the acquisition of the learning level. In 60 s after the optical spikes terminate, the EPSC value decreases to ~470 nA. At the second learning stage, 20 consecutive spikes are applied to the channel. Only 11 spikes are enough to let EPSC value reach the learning threshold. In 60 s after the optical spikes terminate, the EPSC value decreases to ~570 nA. At the third learning stage, 10 consecutive spikes are applied to the channel. Only 4 spikes are enough to let EPSC value reach the learning threshold. These results indicate that previous learning activities are helpful for the following learning activities, as is quite similar to the “Learning−Forgetting−Relearning” activities in our brain.

TENG can couple frictional electrification and electrostatic induction to convert various forms of mechanical energy into electrical energy [39−41, 49]. Here, PDMS-ZnO based TENG is fabricated. Fig.4(a) shows the operating mechanism of the PDMS-ZnO TENG. When an external pressure is loaded to the TENG, the upper and lower friction layers will come into contact. Due to the electrostatic induction and frictional electrification effects, the PDMS-ZnO membrane with electronegativity will gain electrons, while the kapton surface will carry positive charges. Once the external pressure is removed, the upper and lower friction layers separate. Thus, a potential difference will be formed between the top and bottom electrodes. In order to balance this potential difference, electrons flow from the bottom electrode to the top electrode, forming a current in external circuit. When external pressures cause the upper and lower friction layers to approach again, the potential difference between the two electrodes decreases, causing electrons to flow back. An opposite current is thus generated in external circuit. Therefore, an alternating voltage can be generated by alternatively applying and releasing pressures to the TENG.

Figure S4(a) shows AFM surface morphology of the kapton membrane. The surface rms value is ~2.0 nm. Fig.4(b) shows SEM image of the ZnO powder used in TENG. The ZnO powder exhibits a certain degree of agglomeration and a relatively uniform particle size distribution. Fig.4(c) shows the SEM surface image of the sanding paper mold. The surface is tightly arranged with irregular particles. Fig.4(d) shows the SEM surface image of the PDMS-ZnO membrane with ZnO/PDMS mass ratio of 1:10. The microstructure has a smaller scale than that of the surface of sanding paper. Figure S4(b) shows energy dispersive spectrometer elemental distribution mapping of the PDMS-ZnO membrane with ZnO/PDMS mass ratio of 1:10. The elements, i.e., Si, C, Zn, O, are uniformly distributed, indicating that the distribution of ZnO powder in PDMS is relatively uniform. Fig.4(e) shows the SEM surface image of the PDMS-ZnO membrane after impacts with pressure of 70 kPa for 10000 times. There are no significant changes after impacts, indicating good stabilities. Here, relative dielectric constant (ε_r) of PDMS-ZnO membranes with different ZnO/PDMS mass ratio were investigated, as shown in Fig. S4(c). The ε_r value increases from ~2.59 to ~3.18 with the increased ZnO/PDMS mass ratio from 0:1 to 1:10. While it decreases slightly from 3.18 to 3.04 with the increased ZnO/PDMS mass ratio from 1:10 to 1:5. These results indicate that adding a certain proportion of ZnO to PDMS can increase the ε_r value of the PDMS-ZnO membranes, as may improve the output performance of the TENG. Figure S4(d) also shows the short-circuit current (I_SC) of the TENG under the same pressure (70 kPa, 3.5 Hz). The I_SC value increases from ~1.5 μA to ~4.5 μA with the increased ZnO/PDMS mass ratio from 0:1 to 1:10. While it decreases from ~4.5 μA to ~2.2 μA with the increased ZnO/PDMS mass ratio from 1:10 to 1:5. The results are quite similar to those in Ref. [50]. The results indicate that the TENG adopts the PDMS-ZnO membrane with a ZnO/PDMS mass ratio of 1:10 would have the best charge transfer efficiency. Thus, we use the PDMS-ZnO membrane with a ZnO/PDMS mass ratio of 1:10 for the following experiments.

Fig.4(f) shows the open-circuit voltage (V_OC) generated by 20 consecutive pressures with fixed frequency at 3.5 Hz. With the increased pressures from 4 kPa to 70 kPa, the V_OC value increases from ~6.4 V to ~42.8 V. Additionally, the V_OC peak value is quite stable at each pressure level. Figure S5(a) shows an enlarged curve of the V_OC waveform under pressure of 9 kPa. The waveform is quite stable. The results indicate good stabilities. The pressure frequency also affects the V_OC of TENG. Figure S5(b) shows the V_OC generated by pressures at different frequencies. The V_OC generated by the pressure (9 kPa, 2 Hz) is ~9.2 V, slightly lower than that by the pressure (9 kPa, 3.5 Hz) of ~10.5 V. The blue dots in Fig.4(g) shows the positive V_OC values under different pressures. The V_OC value gradually increases with the increased pressure. At larger pressures, the growth rate of V_OC value slows down. Here, a sensitivity can be defined by the following equation: SE = ΔV_OC/ΔP, where P is the loaded pressure. When the pressure is less than 24 kPa, the SE value is estimated to be ~0.75 V/kPa. When the pressure exceeds 24 kPa, the SE value is estimated to be ~0.49 V/kPa. Even after 10000 impacts with pressure of 70 kPa, the positive V_OC values did not change significantly, as shown in the red dots in Fig.4(g). Figure S5(c) shows the V_OC values of the TENG with/without microstructures under different pressures. The V_OC values of TENG with microstructure are larger, indicating the positive effects of microstructures on the PDMS-ZnO membrane. Fig.4(h) shows V_OC generated by 6000 consecutive impacts (70 kPa, 3.5 Hz). The V_OC waveform is quite stable, indicating that the TENG has good durability. Figure S5(d) also shows the V_OC of the TENG under light illumination with different optical power intensities. Fortunately, the light illumination has no obvious effects on the V_OC. The present PDMS-ZnO TENG can also be used as a power supply for electronic devices, as shown in Fig.4(i). After full-wave rectification, the output power of TENG can light 44 LEDs with different light wavelength, indicating that the PDMS-ZnO TENG has the ability to effectively convert mechanical energy into electrical energy.

Recently, bionic tactile perceptual system (BTPS) integrating TENGs and neuromorphic transistors has been received widespread attention [27, 28, 51]. It can be used for identifying and analyzing the characteristics of pressures, helping humanoid robots to interact with environment better. With the signal amplification and regulation functions of neuromorphic transistors, the system accurately simulates human tactile perception and can be widely adopted in fields such as intelligent prosthetics and wearable devices. Fig.5(a) shows the schematic diagram of the present BTPS. The operation mechanism is as follows. The PDMS-ZnO based TENG generates alternating voltage under pressures based on frictional electrification and electrostatic induction effects. The voltage signal of TENG is loaded on the gate of the IGZO-PNT after half-wave rectification with a diode. Therefore, protons within the GEL-CCs based electrolyte will migrate, inducing an electric-double-layer (EDL) at the electrolyte/IGZO interface. Thus, the carrier concentration in the channel will increase, triggering EPSC responses. As enables the BTPS to simulate synaptic activities under pressures. Fig.5(b) shows EPSC responses triggered by single pressures with different amplitudes in two consecutive tests. The EPSC peak value increases with the increased pressure. It is ~0.3 μA under the pressure of 3 kPa. When the pressure increases to 13 kPa, it reaches ~4.4 μA. Fig.5(c) shows EPSC peak values extracted from Fig.5(b). Here, a sensitivity can be defined by the following equation: SE_P = ΔI/ΔP, where I is the peak EPSC value. Interestingly, the SE_P value remains stable at ~0.43 μA/kPa for the two consecutive tests. Table S1 shows the comparison between the present BTPS and those for previously reported works. The table indicates that the present BTPS has higher V_oc and higher sensitivity. Fig.5(d) shows typical EPSC response triggered by paired pressures (4 kPa) with an interval time (ΔT) of 2 s. The second absolute EPSC value (A₂) is higher than the first one (A₁), i.e., ~1.23 μA vs. ~0.86 μA. Thus, PPF index is estimated to be ~143.1%. Fig.5(e) shows PPF index as a function of ΔT. As the ΔT value increases gradually, the PPF index approaches ~105%. Fig.5(f) shows pressure pulse number dependent EPSC responses triggered by pressure pulses (4 kPa, 2 Hz). Fig.5(g) shows the extracted EPSC peak values from Fig.5(f). The EPSC peak value increases linearly with the increased pressure pulse numbers. For a single pressure pulse, the EPSC peak value is ~0.9 μA. While for 50 pressure pulses, the EPSC peak value increases to ~7.1 μA. Fig.5(h) shows EPSC response triggered by 50 consecutive pressures with amplitudes of 4 kPa and 7 kPa. The EPSC peak value is ~7.1 μA under the pressure of 4 kPa. While, it reaches ~12.5 μA when the pressure is 7 kPa. All these results indicate the high sensitivity of the present BTPS to external tactile stimuli.

In order to simplify the structure of the perceptual system, electrical signals of TENG are not rectified with a diode, as schematically shown in Fig.1(b). Fig.6(a) shows EPSC responses on a pressure with amplitude of 4 kPa with and without rectification. TENG without rectification will have a positive V_OC peak and a negative V_OC peak. In this case, the EPSC rapidly increases when the pressure is loaded and quickly returns to the initial state once the pressure is released. As comparison, it slowly returns back to the initial state for TENG with rectification. The EPSC peak value without rectification is ~1.1 μA, which is slightly larger than the EPSC peak value of ~0.9 μA with rectification under the same pressure. As is quite normal since the V_OC of TENG decreases after rectification, as shown in Fig. S5(e). Due to the fast response characteristics of the EPSC triggered by pressures, the system has the potential in the field of communication. Morse code communication uses dots (·) and dashes (-) to represent letters, numbers, or marks. Information can be sent or received through various media such as radio, light, and pressure. In our case, information is encrypted with pressures of 7 kPa and 4 kPa, i.e., “-” and “·” signals, respectively. In information decryption stage, a “-” signal is decrypted if the EPSC value exceeds 3 μA. While a “·” signal is decrypted if the EPSC value is below 3 μA. As shown in Fig.6(b), by applying a series of pressures to TENG, signals can be decoded into alphabets “N”, “B” and “U”.

Fig.6(c) shows the EPSC responses of the BTVMS triggered by 20 consecutive pressures (4 kPa, 2 Hz) under light illumination at different optical power intensities. In the dark, the EPSC value does not increase under the pressure trains. The EPSC gain, defined as A₂₀/A₁, is ~0.95. At the light power of 6.8 mW/cm², the EPSC value slowly increases under the same pressure trains. The EPSC peak value is ~2.6 μA. The EPSC gain is ~1.93. While at the light power of 41.9 mW/cm², the EPSC peak value increases to ~13 μA. The EPSC gain increases to ~5.26. The results indicate that the present BTVMS can synergetically perceive pressure and optical stimuli. Such synergetic effects greatly enhance the responses in tactile perception. In another word, the visual and tactile perception would endow the bionic system to respond to external stimuli more effectively. Based on the performance, post-traumatic stress disorder (PTSD) behavior has been imitated successfully. The PTSD behavior typically occurs after an individual experiences a severe traumatic event, which is usually composed of multiple intense stimuli [52, 53]. In our case, the simultaneous occurrence of pressures and optical stimuli is defined as a traumatic event, as schematically shown in Fig.6(d). Single pressure stimuli or single optical stimuli do not induce a fear response before experiencing a traumatic event. Fig.6(e) shows the EPSC responses triggered by pressure trains or light stimulus. The fear threshold is defined at 3 μA. The EPSC peak value triggered by 30 consecutive pressure stimuli (4 kPa, 2 Hz) is only ~1.7 μA. Meanwhile, the EPSC peak value triggered by the optical spike (12.7 mW/cm²) with duration of 15 s is only ~1.6 μA. These EPSC peak values do not exceed 3 μA, indicating that applying pressures or light stimulus only will not induce a fear response. However, when pressure stimuli and optical stimuli occur simultaneously, the fear response can be triggered in ~4.4 s, as shown in Fig.6(f). When the mixed stimuli end, the EPSC peak value reaches ~4.9 μA. Interestingly, the fear response can still be triggered by pressure stimuli in ~135 s. In another word, the PTSD behavior can be mimicked by the present BTVMS.

4 Conclusion

In summary, a bionic tactile-visual-morphic system (BTVMS) is proposed by integrating an indium gallium zinc oxide (IGZO) photoelectronic neuromorphic transistor (PNT) and a PDMS-ZnO based triboelectric nanogenerator (TENG). As the core unit of visual perception, the IGZO-PNT can mimic typical synaptic activities under optical stimuli. The PDMS-ZnO based TENG adopts a friction layer with unique microstructures. It can efficiently convert pressure stimuli into electrical signals, exhibiting high sensitivity of ~0.75 V/kPa and good durability. In addition, the IGZO-PNT can amplify and integrate the electrical signals generated by TENG. The BTVMS exhibits information encryption and decryption functions based on Morse code strategy. In addition, it can simulate the tactile and visual dual cognition behavior of brain. Thus, post-traumatic stress disorder behavior has been successfully mimicked. The present BTVMS is expected to provide interesting ideas for fields such as intelligent prosthetics and humanoid robots.

References

[1] H. Wang, T. Fu, Y. Du, W. Gao, K. Huang, Z. Liu, P. Chandak, S. Liu, P. Van Katwyk, A. Deac, A. Anandkumar, K. Bergen, C. P. Gomes, S. Ho, P. Kohli, J. Lasenby, J. Leskovec, T. Y. Liu, A. Manrai, D. Marks, B. Ramsundar, L. Song, J. Sun, J. Tang, P. Veličković, M. Welling, L. Zhang, C. W. Coley, Y. Bengio, and M. Zitnik, Scientific discovery in the age of artificial intelligence, Nature 620(7972), 47 (2023)

[2] M. Musib, F. Wang, M. A. Tarselli, R. Yoho, K. H. Yu, R. M. Andrés, N. F. Greenwald, X. Pan, C. H. Lee, J. Zhang, K. Dutton-Regester, J. W. Johnston, and I. M. Sharafeldin, Artificial intelligence in research, Science 357(6346), 28 (2017)

[3] B. Haibe-Kains, G. A. Adam, A. Hosny, F. Khodakarami, T. Shraddha, R. Kusko, S. A. Sansone, W. Tong, R. D. Wolfinger, C. E. Mason, W. Jones, J. Dopazo, C. Furlanello, L. Waldron, B. Wang, C. McIntosh, A. Goldenberg, A. Kundaje, C. S. Greene, T. Broderick, M. M. Hoffman, J. T. Leek, K. Korthauer, W. Huber, A. Brazma, J. Pineau, R. Tibshirani, T. Hastie, J. P. A. Ioannidis, J. Quackenbush, and H. J. W. L. Aerts, Transparency and reproducibility in artificial intelligence, Nature 586(7829), E14 (2020)

[4] J. Dreyer, China made waves with Deepseek, but its real ambition is AI-driven industrial innovation, Nature 638(8051), 609 (2025)

[5] G. Conroy and S. Mallapaty, How China created AI model DeepSeek and shocked the world, Nature 638(8050), 300 (2025)

[6] D. Shin and H. J. Yoo, The heterogeneous deep neural network processor with a non-Von Neumann architecture, Proc. IEEE 108(8), 1245 (2020)

[7] S. Dutta, H. Jeong, Y. Yang, V. Cadambe, T. M. Low, and P. Grover, Addressing unreliability in emerging devices and non-Von Veumann architectures using coded computing, Proc. IEEE 108(8), 1219 (2020)

[8] T. J. Prescott and S. P. Wilson, Understanding brain functional architecture through robotics, Sci. Robot. 8(78), eadg6014 (2023)

[9] H. Kuai, J. Chen, X. Tao, L. Cai, K. Imamura, H. Matsumoto, P. Liang, and N. Zhong, Never-ending learning for explainable brain computing, Adv. Sci. (Weinh.) 11(24), 2307647 (2024)

[10] Y. Lu, A. Alvarez, C. H. Kao, J. S. Bow, S. Y. Chen, and I. W. Chen, An electronic silicon-based memristor with a high switching uniformity, Nat. Electron. 2(2), 66 (2019)

[11] X. Zhao, J. Xu, D. Xie, Z. Wang, H. Xu, Y. Lin, J. Hu, and Y. Liu, Natural acidic polysaccharide-based memristors for transient electronics: Highly controllable quantized conductance for integrated memory and nonvolatile logic applications, Adv. Mater. 33(52), 2104023 (2021)

[12] Z. Li, J. Wang, L. Xu, L. Wang, H. Shang, H. Ying, Y. Zhao, L. Wen, C. Guo, and X. Zheng, Achieving reliable and ultrafast memristors via artificial filaments in silk fibroin, Adv. Mater. 36(4), 2308843 (2024)

[13] J. Shi, Y. Lin, Z. Wang, X. Shan, Y. Tao, X. Zhao, H. Xu, and Y. Liu, Adaptive processing enabled by sodium alginate based complementary memristor for neuromorphic sensory system, Adv. Mater. 36(32), 2314156 (2024)

[14] S. Lee, J. Kim, and S. Kim, Self-aligned TiO_x-based 3D vertical memristor for a high-density synaptic array, Front. Phys. (Beijing) 19(6), 63203 (2024)

[15] H. Shim, F. Ershad, S. Patel, Y. Zhang, B. H. Wang, Z. H. Chen, T. J. Marks, A. Facchetti, and C. J. Yu, An elastic and reconfigurable synaptic transistor based on a stretchable bilayer semiconductor, Nat. Electron. 5(10), 660 (2022)

[16] L. Merces, L. M. M. Ferro, A. Nawaz, and P. Sonar, Advanced neuromorphic applications enabled by synaptic ion-gating vertical transistors, Adv. Sci. (Weinh.) 11(27), 2470162 (2024)

[17] C. Y. Wang, Y. S. Bian, K. Liu, M. C. Qin, F. Zhang, M. L. Zhu, W. K. Shi, M. C. Shao, S. C. Shang, J. X. Hong, Z. H. Zhu, Z. Y. Zhao, Y. Q. Liu, and Y. L. Guo, Strain-insensitive viscoelastic perovskite film for intrinsically stretchable neuromorphic vision-adaptive transistors, Nat. Commun. 15(1), 3123 (2024)

[18] W. S. Wang, Z. W. Shi, X. L. Chen, Y. Li, H. Xiao, Y. H. Zeng, X. D. Pi, and L. Q. Zhu, Biodegradable oxide photoelectric transistors for neuromorphic computing and anxiety disorder emulation, ACS Appl. Mater. Interfaces 15(40), 47640 (2023)

[19] Q. Chen, L. R. Wu, J. Q. Xu, S. W. Xin, D. L. Ge, M. Y. Wei, Y. B. Qin, Z. Liu, Y. Xi, and F. Y. Wang, Tunable oxygen vacancy defects for high-performance electrolyte-gated synaptic transistors, Front. Phys. (Beijing) 20(6), 064202 (2025)

[20] Y. Xu, G. Zhang, W. Liu, C. Jin, Y. Nie, J. Sun, and J. Yang, Flexible multiterminal photoelectronic neurotransistors based on self-assembled rubber semiconductors for spatiotemporal information processing, SmartMat 4(2), e1162 (2023)

[21] J. Guo, Y. Liu, F. Zhou, F. Li, Y. Li, and F. Huang, Linear classification function emulated by pectin-based polysaccharide-gated multiterminal neuron transistors, Adv. Funct. Mater. 31(33), 2102015 (2021)

[22] Y. J. Huang, L. F. Wu, J. K. Di, X. Huang, W. S. Wang, S. Y. Zhou, B. C. Gong, and L. Q. Zhu, Solidum alginate gated oxide dendritic transistor for spatiotemporal arithmetic application, Front. Phys. (Beijing) 20(3), 034205 (2025)

[23] H. Cai, Z. Ao, C. Tian, Z. Wu, H. Liu, J. Tchieu, M. Gu, K. Mackie, and F. Guo, Brain organoid reservoir computing for artificial intelligence, Nat. Electron. 6(12), 1032 (2023)

[24] K. Deisseroth, J. E. Haley, and A. R. Mehta, Form and function in the brain, Lancet Neurol. 20(7), 508 (2021)

[25] Y. K. Cheng, Z. Z. Li, Y. Lin, Z. Q. Wang, X. Y. Shan, Y. Tao, X. N. Zhao, H. Y. Xu, and Y. C. Liu, Color recognition achieved in multiwavelength controlled plasmonic optoelectronic memristor for neuromorphic visual system, Adv. Funct. Mater. 35(5), 2414404 (2025)

[26] C. Han, X. Han, J. Han, M. He, S. Peng, C. Zhang, X. Liu, J. Gou, and J. Wang, Light-stimulated synaptic transistor with high PPF feature for artificial visual perception system application, Adv. Funct. Mater. 32(22), 2113053 (2022)

[27] H. Lei, Z. Y. Yin, P. Huang, X. Gao, C. Zhao, Z. Wen, X. Sun, and S. D. Wang, Intelligent tribotronic transistors toward tactile near-sensor computing, Adv. Funct. Mater. 35(21), 2401913 (2025)

[28] Y. J. Park, Y. G. Ro, Y. E. Shin, C. Park, S. Na, Y. Chang, and H. Ko, Multi-layered triboelectric nanogenerators with controllable multiple spikes for low-power artificial synaptic devices, Adv. Sci. (Weinh.) 10(36), 2304598 (2023)

[29] Y. J. Huang, J. K. Di, W. S. Wang, X. Huang, S. Y. Zhou, B. C. Gong, Z. Q. Zhao, and L. Q. Zhu, Artificial tactile perceptual system based on capacitive tactile sensor and oxide neuromorphic transistor, Appl. Mater. Today 41, 102521 (2024)

[30] Y. Liu, E. Li, X. Wang, Q. Chen, Y. Zhou, Y. Hu, G. Chen, H. P. Chen, and T. L. Guo, Self-powered artificial auditory pathway for intelligent neuromorphic computing and sound detection, Nano Energy 78, 105403 (2020)

[31] Y. He, S. Nie, R. Liu, S. Jiang, Y. Shi, and Q. Wan, Spatiotemporal information processing emulated by multiterminal neuro‐transistor networks, Adv. Mater. 31(21), 1900903 (2019)

[32] S. Y. Zhou, W. S. Wang, X. Huang, B. C. Gong, J. K. Di, Y. J. Huang, and L. Q. Zhu, Sound frequency sensitive triboelectric nanogenerator for multi-functional auditory perceptual system applications, Nano Res. 18(6), 94907378 (2024)

[33] S. Najafian,E. Koch,K. L. Teh,J. Jin,H. Rahimi-Nasrabadi,Q. Zaidi,J. Kremkow,J. M. Alonso, A theory of cortical map formation in the visual brain, Nat. Commun. 13(1), 2303 (2022)

[34] T. D. Price and R. Khan, Evolution of visual processing in the human retina, Trends Ecol. Evol. 32(11), 810 (2017)

[35] C. Schwarz, The slip hypothesis: Tactile perception and its neuronal bases, Trends Neurosci. 39(7), 449 (2016)

[36] E. W. McCleskey, A mechanism for touch, Nature 573(7773), 199 (2019)

[37] M. Lee, M. Kim, J. W. Jo, S. K. Park, and Y. H. Kim, Suppression of persistent photo-conductance in solution-processed amorphous oxide thin-film transistors, Appl. Phys. Lett. 112(5), 052103 (2018)

[38] S. Yoo, D. S. Kim, W. K. Hong, J. I. Yoo, F. Huang, H. C. Ko, J. H. Park, and J. Yoon, Enhanced ultraviolet photoresponse characteristics of indium gallium zinc oxide photo-thin-film transistors enabled by surface functionalization of biomaterials for real-time ultraviolet monitoring, ACS Appl. Mater. Interfaces 13(40), 47784 (2021)

[39] F. R. Fan, Z. Q. Tian, and Z. L. Wang, Flexible triboelectric generator, Nano Energy 1(2), 328 (2012)

[40] X. R. Liu, Z. H. Zhao, Y. K. Gao, Y. Nan, Y. X. Hu, Z. T. Guo, W. Y. Qiao, J. Wang, L. L. Zhou, Z. L. Wang, and J. Wang, Triboelectric nanogenerators exhibiting ultrahigh charge density and energy density, Energy Environ. Sci. 17(11), 3819 (2024)

[41] F. Chen,Y. Wu,Z. Ding,X. Xia,S. Li,H. Zheng,C. Diao,G. Yue,Y. Zi, A novel triboelectric nanogenerator based on electrospun polyvinylidene fluoride nanofibers for effective acoustic energy harvesting and self-powered multifunctional sensing, Nano Energy 56, 241 (2019)

[42] J. Liu,J. Xu,Y. Wang,Z. Li,M. Li,N. Cui,F. Zhao,L. Meng,L. Gu, A human-skin inspired self-healing, anti-bacterial and high performance triboelectric nanogenerator for self-powered multifunctional electronic skin, Chem. Eng. J. 495, 153601 (2024)

[43] X. Suo, B. Li, H. Ji, S. Mei, S. Miao, M. Gu, Y. Yang, D. Jiang, S. Cui, L. Chen, G. Chen, Z. Wen, and H. Huang, Dielectric layer doping for enhanced triboelectric nanogenerators, Nano Energy 114, 108651 (2023)

[44] Z. W. Shi, W. S. Wang, L. Ai, Y. Li, X. L. Chen, H. Xiao, Y. H. Zeng, and L. Q. Zhu, Non-associative learning behavior in mixed proton and electron conductor hybrid pseudo-diode, J. Mater. Sci. Technol. 160, 204 (2023)

[45] Y. Li, Y. J. Huang, X. L. Chen, W. S. Wang, X. Huang, H. Xiao, and L. Q. Zhu, Multi-terminal pectin/chitosan hybrid electrolyte gated oxide neuromorphic transistor with multi-mode cognitive activities, Front. Phys. (Beijing) 19(5), 53204 (2024)

[46] L. Yao, T. Grand, J. E. Hanson, P. Paoletti, and Q. Zhou, Higher ambient synaptic glutamate at inhibitory versus excitatory neurons differentially impacts NMDA receptor activity, Nat. Commun. 9(1), 4000 (2018)

[47] H. Luo, H. Z. Liu, W. W. Zhang, M. Matsuda, N. Lv, G. Chen, Z. Z. Xu, and Y. Q. Zhang, Interleukin-17 regulates neuron-glial communications, synaptic transmission, and neuropathic pain after chemotherapy, Cell Rep. 29(8), 2384 (2019)

[48] M. K. Dai, Y. R. Liou, J. T. Lian, T. Y. Lin, and Y. F. Chen, Multifunctionality of giant and long-lasting persistent photoconductivity: Semiconductor–conductor transition in graphene nanosheets and amorphous InGaZnO hybrids, ACS Photonics 2(8), 1057 (2015)

[49] W. Qiao,L. Zhou,J. Zhang,D. Liu,Y. Gao,X. Liu,Z. Zhao,Z. Guo,X. Li,B. Zhang,Z. L. Wang,J. Wang, A highly-sensitive omnidirectional acoustic sensor for enhanced human–machine interaction, Adv. Mater. 36(48), 2413086 (2024)

[50] J. Yang, X. Wang, N. Hu, J. Chen, X. Yu, and W. Zhu, Effect of ZnO nanoparticle size on the output performance of triboelectric nanogenerator, J. Mater. Sci. Mater. Electron. 35(18), 1232 (2024)

[51] Y. Liu, W. Yang, Y. Yan, X. Wu, X. Wang, Y. Zhou, Y. Hu, H. P. Chen, and T. L. Guo, Self-powered high-sensitivity sensory memory actuated by triboelectric sensory receptor for real-time neuromorphic computing, Nano Energy 75, 104930 (2020)

[52] O. Perl, O. Duek, K. R. Kulkarni, C. Gordon, J. H. Krystal, I. Levy, I. Harpaz-Rotem, and D. Schiller, Neural patterns differentiate traumatic from sad autobiographical memories in PTSD, Nat. Neurosci. 26(12), 2226 (2023)

[53] R. K. Pitman, A. M. Rasmusson, K. C. Koenen, L. M. Shin, S. P. Orr, M. W. Gilbertson, M. R. Milad, and I. Liberzon, Biological studies of post-traumatic stress disorder, Nat. Rev. Neurosci. 13(11), 769 (2012)

RIGHTS & PERMISSIONS

Higher Education Press