Selective Photonic Synapses: Expanding the Horizon of Neuromorphic Computing

The human visual system is a highly complex network of neurons and synapses that enable us to perceive and process external information, with over 80% of this information being acquired through our visual sense. To mimic this sophisticated system, artificial visual systems have traditionally combined photodetectors and electrical artificial synapses. However, recent advances have led to the development of photonic synapses that possess the dual capability of light response and synaptic functions, allowing for the parallel processing of visual information. Photonic synapses seek to mimic the functionality of biological synapses using photonic (light-based) components. These synapses use light signals to transmit and process information, analogous to the way biological synapses transmit signals between neurons in the brain using chemical and electrical signals. Photonic synapses have the potential to enable faster and more energy-efficient information processing compared to traditional electronic synapses. Scientists have tried to construct photonic synapses using perovskites, metal oxides, two-dimensional materials, organic semiconductors, and even biomaterials, each with its unique properties. However, recently organic semiconductors have emerged as promising candidate due to their excellent tunable light absorption range, solution processability, low-cost manufacturing, mechanical flexibility, and biocompatibility. While various organic semiconductors have shown excellent performance in photonic synapses, selective response to different wavelengths, akin to the human ability to perceive colors, has remained a challenge. Typically, optical filters are used to detect specific wavelengths, but this approach complicates device architecture and limits performance, especially in flexible conformal artificial retinas.

Short-wavelength infrared light (SWIR, 1000–3000 nm) has numerous applications in optical communications, industrial manufacturing, and defense. However, SWIR light can cause irreversible damage to the human eye’s retinas and lenses. Therefore, there is a an urgent need for artificial photonic synapses that selectively respond to SWIR light, expanding the range of human vision without causing harm.

In a new study published in the Angewandte Chemie International Edition by Dr. Song Wang, Dr. Hao Chen, Dr. Tianhua Liu, Dr. Yanan Wei, Dr. Qijie Lin, Dr. Xiao Han, and Professor Hui Huang from the University of Chinese Academy of Sciences in collaboration with Dr. Guo Yao and Professor Chunfeng Zhang from Nanjing University   developed a novel narrow-band gap conjugated polymer (P1) and demonstrated its exceptional photoelectric properties in the context of photonic synapses, specifically focusing on its selective response to SWIR light. This innovation holds great promise for advancing neuromorphic computing and optical communication.

In their study, the research team designed and synthesized a narrow-band gap conjugated polymer named P1, characterized by a donor-acceptor-donor-acceptor (DA-DA’) structure. This structure incorporated key motifs, such as benzo[1,2-c;4,5-c’] bis[1,2,5]-thiadiazole (BBT), 4,4-bis(2-ethylhexyl)-4H-cyclopenta[2,1-b:3,4-b’]dithiophene (CPDT), and (E)-4,4′-bis(2-octyldodecyl)-[6,6′-bithieno[3,2-b]pyrrolylidene]-5,5′(4H,4’H)-dione (TIG). P1 exhibited excellent solubility in common organic solvents, thermal stability, and a narrow optical band gap (Egopt) of 0.69 eV, making it a suitable candidate for SWIR light detection. The highest occupied molecular orbital and lowest unoccupied molecular orbital energy levels were determined to be -4.86 eV and -4.17 eV, respectively.

According to the authors, the photoelectric device based on P1 has unique ability to selectively respond to SWIR light. When P1 was incorporated into photodiode devices, the researchers observed photonic synaptic characteristics in response to a broad spectrum of light, ranging from visible to SWIR. Remarkably, replacing P1 with P1:PC71BM (a blend with the fullerene derivative PC71BM) led to photonic synaptic characteristics with selective response to SWIR light. The mechanism underlying this selective SWIR response is attributed to the energy barriers between P1 and PC71BM and the unbalanced charge transport properties within the P1 material. Unlike in the visible to NIR region, where both P1 and PC71BM contribute to exciton dissociation, in the SWIR range, only P1 absorbs light and generates excitons. The type-I heterojunction formed between P1 and PC71BM prevents efficient exciton dissociation at the interface. This leads to the trapping and delayed de-trapping of charge carriers within the P1 phase, enabling selective SWIR response. This mechanism was corroborated by ultrafast transient absorption measurements and photoluminescence (PL) studies.

The authors demonstrated that P1-based photonic synapses exhibited a range of neuron and synaptic functions, including excitatory postsynaptic current response, paired-pulsed facilitation (PPF), short-time plasticity (STP), and long-time plasticity (LTP) behaviors. These functionalities are crucial for information processing and learning, and they were demonstrated with an ultra-low energy consumption of 2.85 femtojoules (fJ) per synaptic event, comparable to the energy efficiency of biological synapses.

Indeed, the new study by Professor Hui Huang and colleagues opens up exciting possibilities for a variety of applications. The development of photonic synapses with selective SWIR response can significantly enhance artificial visual systems, enabling safer and more advanced human-machine interfaces and prosthetic devices. Furthermore, it has potential applications in optical communication, data processing, and remote sensing, especially in environments where SWIR light is prevalent. Looking ahead, the exploration of other conjugated polymers with tailored energy levels and the incorporation of P1 into flexible, conformal artificial retinas hold promise for further advancing this technology. Additionally, integrating these photonic synapses with traditional silicon-based neuromorphic circuits could pave the way for comprehensive neuromorphic computing systems with enhanced sensory perception and information processing capabilities.