Spectrometer-Less Remote Sensing Image Classification Based on Gate-Tunable van der Waals Heterostructures.

Area of Science:

• Materials Science
• Optoelectronics
• Remote Sensing

Background:

• Conventional remote sensing relies on spectrometers for hyperspectral imaging, facing challenges in miniaturization due to complex optics.
• Spectrometers capture reflectance spectra for classification and unmixing, crucial for environmental and agricultural applications.

Purpose of the Study:

• To develop a miniaturized, cost-effective photodetector for remote sensing classification.
• To leverage two-dimensional van der Waals heterojunctions (2D-vdWHs) for wide-spectral photoresponse and gate-tunability.

Main Methods:

• Fabrication of a 2D-vdWH photodetector using m-phase GaTe_0.5Se_0.5 and WSe₂ with a wide spectral response (250-1064 nm).
• Utilizing multi-photoresponses under varying gate voltages combined with deep learning algorithms for classification.
• Testing the device on six hyperspectral datasets to evaluate classification accuracy.

Main Results:

• The gate-tunable photodetector achieved an average classification accuracy of 87.00% on hyperspectral datasets.
• This accuracy is competitive with traditional hyperspectral data (88.72%) and significantly higher than non-tunable devices (71.17%).
• The device demonstrated high accuracy without requiring original hyperspectral reflectance data.

Conclusions:

• Gate-tunable wide-spectral 2D-vdWHs photodetectors offer a viable alternative to conventional spectrometers for remote sensing.
• This approach facilitates the development of miniaturized, cost-effective, and high-accuracy remote sensing classification technologies.
• The GaTe_0.5Se_0.5/WSe₂ heterojunction shows promise for advanced remote sensing applications.