Group‐III Nitride‐Based Wide‐Spectrum Multifunctional Synapses for Encrypted Light Communication and Image Recognition

Why It Matters

Merging high‑speed photodetection with synaptic behavior accelerates neuromorphic hardware, opening low‑power, secure optical links and on‑chip AI inference.

Key Takeaways

• InGaN nanorods deliver 31.47 A/W responsivity at 810 nm
• Response times as fast as 190 µs under 5 V bias
• UV‑driven synaptic plasticity enables 89.12% image‑recognition accuracy
• Wavelength‑controlled switching supports encrypted dual‑band communication

Pulse Analysis

Neuromorphic computing has long sought components that can both sense and process information, reducing the latency inherent in separate photodetector‑processor pipelines. Traditional designs treat photodetection and synaptic modulation as mutually exclusive, forcing engineers to compromise on speed, sensitivity, or memory retention. The emergence of multifunctional photonic synapses promises to collapse this divide, delivering hardware that can detect light across a broad spectrum while simultaneously encoding synaptic weights for on‑chip learning.

The breakthrough reported by the Wiley team hinges on InGaN core‑shell nanorods featuring a graded indium composition and an ultrathin oxide interface. Density‑functional theory reveals that oxide‑induced trap states finely tune carrier relaxation, extending the device’s spectral reach from visible to infrared. Under 810 nm illumination the nanorods produce a record‑high responsivity of 31.47 A/W and respond within 190–240 µs at a modest 5 V bias, rivaling dedicated photodiodes. When illuminated with 365 nm UV light, the same structure exhibits stable, tunable synaptic plasticity, enabling analog weight updates without external circuitry.

These dual capabilities translate directly into real‑world applications. By encoding data in the wavelength domain, the synapses support encrypted light communication that is both fast and resistant to eavesdropping. Simultaneously, their synaptic response under UV light powers a handwritten‑digit classifier that reaches 89.12% accuracy, showcasing on‑chip image‑recognition potential. As the semiconductor industry pushes toward integrated AI accelerators, such wide‑spectrum, multifunctional devices could become the cornerstone of low‑power, secure, and ultra‑fast neuromorphic processors. Future research will likely explore scaling the nanorod arrays and integrating them with CMOS back‑ends to bring this technology to commercial chips.