A Scalable Perovskite Platform With Multi-State Photoresponsivity for In-Sensor Saliency Detection

Area of Science:

• Materials Science
• Computer Engineering
• Artificial Intelligence

Background:

• Artificial vision systems are crucial for edge intelligence but face challenges with data latency and energy consumption due to separate sensors and processors.
• In-sensor computing (ISC) integrates sensing and computation to overcome these limitations, but existing ISC devices struggle with scalability, uniformity, and processability.

Purpose of the Study:

• To address the limitations of current ISC devices by developing a reconfigurable perovskite-photovoltaic platform.
• To enable facile solution processing, precise photoresponsivity tuning, and fabrication on diverse substrates.

Main Methods:

• Developed a reconfigurable perovskite-photovoltaic platform processed from solutions.
• Demonstrated precise photoresponsivity tuning with ultra-low variability.
• Fabricated devices on rigid and flexible substrates, including a 32x32 array for scalability testing.

Main Results:

• Achieved high photoresponsivity (up to ±1120 mA W⁻¹) and 1000 programmable states with excellent air stability and thermal reliability.
• Demonstrated a scalable 32x32 array with excellent uniformity and programmability.
• Utilized the ISC array for energy-efficient face detection (95.2% sensitivity, 4.51x speedup), image feature extraction, and MNIST digit recognition (96.97% accuracy).

Conclusions:

• Established a practical, scalable, and stable perovskite-based ISC platform.
• The platform offers enhanced tunability and fabrication capabilities compared to previous ISC implementations.
• Opens new avenues for intelligent computing in robotics, wearable electronics, and neuromorphic vision.