Wavelength‐Dependent Negative/Positive Photoresponse and Infrared Polarization Sensitivity of Two‐Dimensional PdSe2/NbSe2 Heterojunction‐Based Visual Synapse

Why It Matters

The dual‑polarity, low‑energy photonic response enables neuromorphic processors to handle complex infrared signals with brain‑like excitatory‑inhibitory dynamics, a key step toward efficient AI hardware for imaging and sensing.

Key Takeaways

• PdSe2/NbSe2 heterojunction switches between negative and positive photoresponse
• Works across 808‑2200 nm, enabling broadband infrared detection
• Polarization ratio reaches 12.67 at 1550 nm, showing strong anisotropy
• Runs at 1 mV bias, consuming 0.298 pJ per pulse
• Enables excitatory‑inhibitory dual‑state synaptic functions for neuromorphic computing

Pulse Analysis

Two‑dimensional materials have reshaped photodetector design, offering atomically thin channels, tunable bandgaps, and seamless integration with flexible electronics. In this study, anisotropic PdSe2 provides a narrow‑bandgap semiconductor that absorbs mid‑infrared photons, while metallic NbSe2 supplies a high‑conductivity contact, forming a heterojunction that leverages built‑in electric fields for efficient charge separation. The resulting device demonstrates a rare ability to invert its photocurrent polarity simply by adjusting the applied bias, a behavior that mirrors the excitatory and inhibitory signaling found in biological synapses.

Beyond polarity switching, the heterojunction operates at an ultra‑low bias of 1 mV and consumes sub‑picojoule energy per optical pulse. Such energy efficiency is critical for scaling neuromorphic hardware, where billions of artificial synapses must process data continuously without overheating. Compared with conventional infrared photodiodes that require tens of volts and milliwatt‑level power, this platform reduces power budgets by several orders of magnitude, opening pathways for edge AI devices that can run on battery or harvested energy.

The device’s pronounced polarization sensitivity—exhibiting a ratio of 12.67 at 1550 nm—adds a spatial‑orientation dimension to infrared sensing. High polarization contrast enables more accurate target discrimination in low‑visibility environments, benefiting applications ranging from autonomous navigation to security surveillance. As the market for infrared imaging and neuromorphic processors expands, integrating such multi‑functional visual synapses could accelerate the development of compact, brain‑inspired cameras that process visual information directly at the sensor level, reducing latency and computational load for downstream AI algorithms.