Chongfan Technology

News

All Categories >

Home page

2025

Light | Broad spectrum optoelectronic synapse: gas adsorption assisted persistent photoconductivity

Author:

Introduction

Computer Visionis a core technology for emerging applications such as autonomous driving, intelligent robots, and intelligent manufacturing. Traditional visual systems use serial processing, generating a large amount of redundant data and frequent transmission during operation, leading to increased power consumption and information delays. In contrast, neuromorphic vision systems mimic the neuronal and synaptic structures of the human brain, using parallel processing to handle multiple streams of information simultaneously, significantly reducing power consumption and improving data processing speed. Integrating visual perception, data storage, and information processing functions into a single device further enhances these advantages (as shown in Figure 1). However, current integrated neuromorphic devices still need improvement in broadband detection, low-light detection, and data retention, especially in the near-infrared band, where performance is still limited. Therefore, developing optoelectronic synapses with broadband detection, high-sensitivity photoresponse, and stable data retention is of great significance for realizing high-performance computer vision systems.

Figure 1: Schematic diagram of an integrated neuromorphic device

Recently,Chinese Academy of SciencesResearcher Li Shaojian and Researcher Li Dabing from Changchun Institute of Optics, Fine Mechanics and Physics, CAS, and their collaboratorshave made a breakthrough in this field. The team proposed a photoelectric synaptic transistor based on Ta₂NiSe₅/SnS₂heterostructure, revealing the core mechanism of broadband (ultraviolet-near-infrared) high photoelectric conversion efficiency and non-volatility from the perspective of photogain and carrier lifetime control. This device, through a gas adsorption-assisted strategy combined with heterostructure band engineering, exhibits excellent photoelectric detection and synaptic performance in a wide spectral range (375-1310 nm) from ultraviolet to near-infrared, and simulates the perception and recognition function of human retinal cells to multispectral signals.

The relevant results were published inPhysisorption-assistant optoelectronic synaptic transistors based on Ta₂NiSe₅/SnS₂heterojunction from ultraviolet to near-infraredinLight: Science & ApplicationsThe first author of the paper is Tan Fan, a doctoral student at the Changchun Institute of Optics, Fine Mechanics and Physics, CAS,doctoral student Tan Fanand the corresponding authors are Li ShaojianResearcher Li ShaojianandAssociate Researcher Zhang Nan。

The research team, based on the photogain theory (G=τ/t_L，where τ is the carrier lifetime,t_Lis the transit time), by constructing a Ta₂NiSe₅/SnS₂heterostructure, innovatively combines the gas adsorption effect with the carrier photogating modulation mechanism, achieving a breakthrough improvement in broadband photoelectric conversion efficiency and non-volatility. On the one hand, tin sulfide (SnS₂) is a two-dimensional semiconductor material widely used in gas detection, with a large number of intrinsic sulfur vacancies on its surface. These sulfur vacancies can bind with gas molecules, providing sufficient defect states for synaptic behavior. On the other hand, tantalum nickel selenide (Ta₂NiSe₅) has a direct narrow band gap structure, and this material exhibits exciton insulator characteristics at room temperature, effectively absorbing photons and is widely used in infrared detection. Based on this, the type-II band structure formed at the heterojunction interface, photogenerated electrons migrate rapidly through the interface under reverse bias, while the interface barrier confines holes in the Ta₂NiSe₅layer, forming a persistent photogating effect, allowing electrons to achieve multi-cycle transport in the channel, and finally obtaining an ultra-high responsivity of 5.6×10³A/W in the visible light band (405 nm), with a quantum efficiency exceeding 1.7×10⁶%。Further combined with light pulse modulation technology, this device exhibits long-term data retention capability in a wide spectral range of 375-1310 nm, and successfully simulates the excitatory postsynaptic current (EPSC), paired-pulse facilitation (PPF index up to 158%), and short/long-term plasticity dynamic conversion of retinal cells, providing a solution with both high sensitivity and bionic functions for the hardware implementation of multispectral neuromorphic vision systems.

Figure 2: Photoelectric detection performance characterization of Ta₂NiSe₅/SnS₂ heterojunction

As shown in Figures 3a-b, paired-pulse facilitation (PPF) is a short-term synaptic plasticity enhancement (STP) phenomenon, referring to the fact that after a neuron receives two consecutive pulses, the excitatory postsynaptic potential (EPSC) caused by the second pulse is much larger than that caused by the first pulse. PPF behavior can be measured by the PPF index, which is a key indicator affecting the accuracy of visual information processing in neural networks. A larger PPF index will lead to better processing results. Under the synergistic effect of gas adsorption and persistent photoconductivity, by increasing the concentration of oxygen in the gas environment, the PPF index of the device is significantly improved, and the power consumption of completing a PPF behavior is also significantly reduced (as shown in Figure 3c).

Long-term plasticity (LTP) refers to the persistent change in synaptic connection strength under the action of continuous multiple pulse stimuli. LTP behavior is an important basis for learning and memory in the nervous system. The number of states contained in long-term plasticity can be measured by the dynamic range (DR), and the size of the dynamic range will significantly affect the training effect of the neural network. As shown in Figure 3, the dynamic range of the device is also significantly improved by increasing the concentration of oxygen. Therefore, the gas adsorption-assisted persistent photoconductivity effect will significantly improve the performance of the synapse.

Figure 3: Impact of gaseous environment on synaptic performance

In addition, most existing integrated neuromorphic devices can only process single-wavelength information, which seriously affects their accuracy in image recognition in complex scenarios, especially in noisy environments. Therefore, expanding the wavelength response range is the key to improving image recognition capabilities. As shown in Figures 4a-c, the device exhibits typical synaptic behaviors, including STP, PPF, and LTP, in response to optical pulses from 375-1310 nm. Based on the significant modulation of synaptic plasticity by optical pulses of different wavelengths and powers, the device achieves recognition and memory storage functions for RGB primary colors within the light intensity range of 2-13 μW (as shown in Figures 4d-e). The broadened wavelength response range will help extract image features and improve the accuracy of recognition tasks.

Figure 4: Spectral response of synaptic behavior and color recognition function

Summary and Outlook

The research team proposed a photoelectric synaptic transistor based on gas adsorption assistance. Through experiments and theoretical analysis, the synergistic mechanism between gas adsorption and persistent photoelectric effects was confirmed, and the synaptic transistor device successfully simulated the perception and recognition functions of the human retina to multispectral signals. This research provides new ideas for the development of optoelectronics and neuromorphic computer vision, and is expected to be applied in the field of artificial intelligence in the future.

Paper Information

Tan， F.， Chang， C.， Zhang， N. et al.₂NiSe₅/SnS₂Physically adsorbed auxiliary photoelectric synaptic transistors based on Ta heterostructures, from ultraviolet to near-infrared. Light Science Applications 14, 122 (2025).

https://doi.org/10.1038/s41377-025-01792-3