I. Core Research Summary

• Low-cost, high-purity carbon nanomaterial ink system

Our in-house‑developed arc‑discharge–derived carbon nanotube raw material costs only one‑tenth of commercial products. After PFIID sorting, it is free of metallic carbon impurities, with a D/G ratio as low as 0.06 and a carrier mobility of 10–25 cm² V⁻¹ s⁻¹. It requires no high‑temperature acid treatment and can be directly aerosol‑printed.

• Ultra-wide spectral photoelectric synaptic response from 275 to 1050 nm

PbS quantum dots, subjected to a brief EDT ligand exchange, exhibit a substantial enhancement in photogenerated carrier transport efficiency. The resulting devices generate postsynaptic currents across the ultraviolet, visible, and near-infrared spectral ranges, thereby overcoming the wavelength‑band limitations of conventional devices.

• Ultra-fast response + atomic-level ultra-low power consumption

The optical signal rise time is only 3.14 ms, and the energy consumption per single near-infrared pulse at 1050 nm is as low as 0.1 fJ—far below the power consumption of biological neurons—thus meeting the requirements of miniature passive visual terminals.

• Complete biomimetic synaptic behavior

Stably achieving the conversion of EPSC, paired-pulse facilitation (PPF), and short-term plasticity (STP) into long-term potentiation (LTP); enabling optogenetically controlled Pavlovian conditioning; and realizing image perception, memory, and contrast preprocessing with an 8×8 pixel array that mimics the ultraviolet early-warning system of Arctic reindeer and the infrared‑adaptive visual mechanisms of salmon.

• Flexible, long-lasting, air-stable device

After being bent with an 18 mm radius, the device exhibits only slight performance degradation; stored in ambient air at room temperature for four months, its optoelectronic properties show no significant deterioration. The entire fabrication process is compatible with low‑temperature PI flexible films, making it well suited for integration into curved wearable devices.

II. Preparation Process Flow and Core Technological Advantages of the Complete Aerosol Spray Device Set

(1) Complete Fabrication Process for Flexible Optoelectronic Synapses

• PI substrate cleaning, spin-coating of a PVP buffer layer, thermal evaporation of the aluminum gate electrode, and ALD deposition of AlO₃ as the insulating dielectric;
• In-house development of sc‑SWCNT toluene ink via ultrasonic atomization, enabling precise aerosol‑based printing of transistor active channels, followed by toluene washing to remove excess polymer.
• Curing of carbon nanotube films by low-temperature baking at 120°C;
• Spin-coat oleic acid–capped PbS quantum dots, followed by ligand exchange with an EDT solution to shorten the surface organic chains.
• Aerosol-printed silver nanowire source–drain electrodes, followed by low-temperature annealing to fabricate a fully functional three-terminal synaptic transistor.
• A test platform for optoelectronic devices has been established, enabling multi-wavelength pulse testing, bending‑induced aging tests, and biomimetic visual‑simulation evaluations.

(II) Five Core Process Advantages (Comparison of Lithography, CVD, and Inkjet)

• All‑low‑temperature, maskless additive manufacturing with zero damage to flexible substrates.

The maximum thermal treatment temperature throughout the process is 120°C, eliminating the need for vacuum coating or photolithography and etching. The ultra-thin PI film remains deformation-free, enabling direct printing on curved wearable substrates.

• Precise site-specific deposition of carbon nanomaterials, reducing consumption of high-value nanomaterials.

The aerosol beam is controllable, enabling the deposition of sc‑SWCNTs exclusively in the transistor channel region while leaving large areas of the substrate free of material, thereby significantly reducing the cost of carbon nanotube usage.

• Multi-material rapid-switching molding

The single-device platform is compatible with both carbon‑nanotube and silver‑nanoparticle conductive inks, and, when paired with a spin‑coated quantum‑dot photosensitive layer, enables the one‑step fabrication of multilayer structures for active synaptic devices, thereby simplifying the manufacturing process.

• Digital Rapid Iteration Array Layout

CAD enables free adjustment of pixel size and array configuration; an 8×8 neuromorphic photodetector array can be prototyped in just a few hours, without the need for custom lithography masks, reducing the R&D cycle by 80%.

• Excellent uniformity across large-area arrays

Aerosol-based film deposition is performed at a uniform rate, yielding devices within the same batch with exceptionally low variability in threshold voltage and on/off ratio, making it well suited for large-scale mass production of flexible sensor arrays.

III. Five Key Technological Innovations

Innovation 1: A Low-Cost PFIID Sorting Scheme for Aerosol-Printed Carbon Nanotubes

Innovation 2: EDT Ligand Exchange Broadens the Ultra-Wide Spectral Response Mechanism

Innovative 3-femtojoule-class ultra-low-power optically controlled synaptic architecture

Innovative 4‑Multi‑Biological Visual Behavior Hardware Simulation Platform

Innovation 5: A bendable, long-term air-stable, flexible synaptic process

IV. Existing Process Challenges and Corresponding Solutions for Overcoming Them

Challenge 1: Commercial semiconductor carbon nanotubes are expensive, and their purification process is complex and can easily damage the nanotubes.

Challenge 2: Conventional PbS quantum dots are passivated with oleic acid ligands, which insulate the particles and hinder the transport of photogenerated carriers, resulting in weak infrared response.

Challenge 3: Traditional optoelectronic synapses are sensitive to only a single wavelength band, making it impossible to achieve all‑day, multispectral recognition.

Challenge 4: Neuromorphic devices generally consume high power, limiting the battery life of miniature wearable devices.

Challenge 5: Lithography and CVD processes are costly, making it impossible to fabricate large-area flexible synaptic arrays.

V. Future Diversified Industrial Application Scenarios

1 Lightweight wearable brain-inspired visual device

2. Autonomous Night-Vision Perception System for Micro Unmanned Aerial Vehicles

3 Industrial All-Weather Vision Inspection Equipment

4 Polar/Field-Specific Monitoring Instruments

5. Brain-inspired Computing Laboratory Rapid Prototyping Platform for Universities

Figure 1. a) As‑prepared SWCNT electrode sample and single‑walled carbon nanotube (sc‑SWCNT) ink sorted by PFIID; b) Absorption spectra of sc‑SWCNT inks at various weight ratios of PFIID to as‑prepared SWCNTs; c) Raman spectra of as‑prepared SWCNTs and sorted sc‑SWCNTs; d) SEM image of the network structure of as‑prepared SWCNTs; e) SEM image of the network structure of sc‑SWCNTs sorted by PFIID; f) Typical transfer curve of a flexible printed sc‑SWCNT thin‑film transistor (TFT), with an inset showing the device architecture of the sc‑SWCNT TFT.

Figure 2. a) The channels of both bottom-contact (B-type) and top-contact (T-type) sc‑SWCNT TFTs are doped with lead sulfide quantum dots. b) Typical transfer curves for sc‑SWCNT TFTs with and without lead sulfide quantum dot doping: 1—T-type sc‑SWCNT TFT; 2—T-type sc‑SWCNT TFT with lead sulfide quantum dots in the channel; 3—T-type sc‑SWCNT TFT with lead sulfide quantum dots in the channel, following EDT ligand exchange (designated PbS‑EDT); 4—T-type lead sulfide TFT. c) Photoresponse characteristics of T‑type and B‑type sc‑SWCNT TFTs modified with lead sulfide quantum dots, under 900 nm pulsed illumination, before and after EDT ligand exchange. d) SEM image of single-walled carbon nanotubes modified with PbS‑EDT. e) Pb 4f XPS spectra and C 1s XPS spectra of single-walled carbon nanotube films, with and without lead sulfide quantum dots and PbS‑EDT, or without lead sulfide quantum dots.

Figure 3. Characteristics of the printed SWCNT/PbS‑EDT thin‑film transistor as a photoelectronic synapse. a) Schematic illustration of neural impulse transmission in the biological nervous system upon receipt of pulsed light signals; b) Schematic diagram of signal transduction in the printed SWCNT/PbS‑EDT thin‑film transistor; c) PPF behavior induced by pulsed light at wavelengths of 365, 440, 520, 660, 780, 900, and 1050 nm (3 mW·cm⁻², 0.1 s); d) PPF behavior triggered by two consecutive photon pulses (1050 nm, 3 mW·cm⁻², 0.02 s) with a Δt of 0.02 s; e) Dependence of the PPF index on Δt for photon pulses at 365, 660, and 1050 nm (3 mW·cm⁻², 0.02 s); f) In the printed SWCNT/PbS‑EDT thin‑film transistor, reducing the stimulation frequency from 50 Hz to 1 Hz (at a wavelength of 1050 nm) can induce a plasticity transition from STP to LTP.

Figure 4. a) Progressive modulation of the IDS in a PbS‑EDT/sc‑SWCNT thin-film transistor via 16 consecutive optical pulses (wavelength 1050 nm, power 1.5 mW·cm⁻², duration 0.02 s); the inset shows distinguishable resistance-state IDS curves extracted from these 16 pulses. b) LTP/LTD curves at wavelengths of 365, 660, and 1050 nm. c) UPS spectra of single-walled carbon nanotubes, lead sulfide, and PbS‑EDT. d) Energy band diagrams for PbS‑EDT and single-walled carbon nanotubes. e) Charge storage and release mechanisms.

Figure 5. a) Typical response speed of printed PbS‑EDT/sc‑SWCNT thin-film transistors under single optical pulses of different wavelengths; b) Power consumption; c) Transfer characteristics of the device before and after bending with a 18‑mm diameter curvature; d) EPSC response curves upon photon‑pulse stimulation (365 nm, 3 mW·cm⁻², 1 s), comparing measurements taken before and after four months of room‑temperature storage; e) Relationship between EPSC gain and pulse frequency induced by a series of photon‑pulse training stimuli (1050 nm); inset: schematic illustration of the device’s simulated high‑pass filtering behavior; f) EPSC response curve of an sc‑SWCNT/PbS‑EDT thin‑film transistor under a 275‑nm pulsed light (0.3 mW·cm⁻²) at VDS = −0.1 V.

Table 1. Performance comparison of sc‑SWCNT/PbS‑EDT thin-film transistors in optoelectronic synaptic transistors, focusing on power consumption, response time, operating wavelength, flexibility, and stability.

Figure 6. Biomimetic detection mechanism. a–c) Changes in normalized postsynaptic current (NPSC) responses under different conditions, used to simulate the variation in salmon’s near-infrared (NIR) sensing capabilities across environments such as transitions from seawater to freshwater; d–f) Temporal dynamics of NPSCs under white-light (3 mW·cm⁻²) and ultraviolet (3 mW·cm⁻²) illumination, modeling the detection process when Arctic wolves absorb ultraviolet light. (a, b, e, and f are schematic diagrams.)

Figure 7. a) A biomimetic human eye with a broad spectral response spanning from the ultraviolet to the near-infrared (NIR) region; b) Schematic illustration of image sensing and processing implemented using an 8×8 flexible optoelectronic synaptic TFT array, which exhibits image preprocessing functions analogous to those of the biomimetic human retina; c) Corresponding patterns of evoked postsynaptic currents (EPSCs) generated upon stimulation with light at wavelengths of 365, 660, and 1050 nm under different irradiation durations.