In-sensor computing using a MoS₂ photodetector with programmable spectral responsivity

Area of Science:

• Photonics
• Materials Science
• Sensor Technology

Background:

• Optical spectroscopy is crucial across scientific and industrial fields.
• Traditional methods require post-acquisition spectral data processing.
• Existing techniques often involve complex external computational steps.

Purpose of the Study:

• To design and realize a smart photodetector for direct spectral information processing.
• To enable in-sensor computation at the physical level of photon detection.
• To develop a miniaturized and energy-efficient optical sensing solution.

Main Methods:

• Tailoring the spectral responsivity of photodetectors to specific tasks.
• Implementing in-sensor computation directly within the sensor system.
• Utilizing an ensemble of optical cavity-enhanced MoS2 photodetectors with tunable properties.

Main Results:

• Demonstrated a smart photodetector performing direct spectral analysis.
• Showcased the device's ability to handle spectral mixture analysis.
• Validated the potential for programming the device for various spectral regression and classification tasks.

Conclusions:

• The developed smart photodetector integrates computation into photon detection.
• This approach significantly advances miniaturized and energy-efficient optical sensing.
• The technology is applicable to algorithms representable by linear operators.