Miniaturized spectrometer with intrinsic long-term image memory

Miniaturized spectrometer with intrinsic long-term image memory

23 January 2024 | Gang Wu, Mohamed Abid, Mohamed Zerara, Jiung Cho, Miri Choi, Cormac Ó Coileáin, Kuan-Ming Hung, Ching-Ray Chang, Igor V. Shvets & Han-Chun Wu
A single $ SnS_{2}/ReSe_{2} $ van der Waals heterostructure is used to create a miniaturized spectrometer with photodetection, spectrum reconstruction, spectral imaging, and long-term image memory capabilities. This device, with a footprint of 19 $ \mu $ m, operates in the visible range (400–800 nm) and has a spectral resolution of 5 nm and a $ >10^{4} $ s long-term image memory. The device utilizes interface trap states to induce a gate-tunable and wavelength-dependent photogating effect and a non-volatile optoelectronic memory effect. The spectrometer demonstrates a high specific detectivity of $ 3.4\times10^{12}\:cm\:Hz^{1/2}\:W^{-1} $ and a high photoresponsivity of approximately 200 A/W. The device's photocurrent increases linearly with incident light power in the low power density range and is influenced by gate voltage. The device's photocurrent is dominated by the interface between $ SnS_{2} $ and $ ReSe_{2} $ and is enhanced by the photogating effect. The device's long-term image memory is attributed to the trapping of holes at defect states, which sustain the photocurrent even after the light is turned off. The device's spectral reconstruction is achieved through compressed sensing and robust principal component analysis, enabling the reconstruction of unknown spectra from a small number of measurements. The device's spectral imaging capabilities are demonstrated through the reconstruction of reflected spectra from colored images. The device's neuromorphic computing capabilities are demonstrated through its ability to function as both a memory and a synapse, enabling the classification of handwritten digits using a two-layer artificial neural network. The device's performance is compared to other miniaturized spectrometers, showing its advantages in terms of footprint, spectral range, and resolution. The device's long-term image memory and neuromorphic computing capabilities make it a promising candidate for applications in artificial synapses, neural networks, machine vision, and imaging processing.A single $ SnS_{2}/ReSe_{2} $ van der Waals heterostructure is used to create a miniaturized spectrometer with photodetection, spectrum reconstruction, spectral imaging, and long-term image memory capabilities. This device, with a footprint of 19 $ \mu $ m, operates in the visible range (400–800 nm) and has a spectral resolution of 5 nm and a $ >10^{4} $ s long-term image memory. The device utilizes interface trap states to induce a gate-tunable and wavelength-dependent photogating effect and a non-volatile optoelectronic memory effect. The spectrometer demonstrates a high specific detectivity of $ 3.4\times10^{12}\:cm\:Hz^{1/2}\:W^{-1} $ and a high photoresponsivity of approximately 200 A/W. The device's photocurrent increases linearly with incident light power in the low power density range and is influenced by gate voltage. The device's photocurrent is dominated by the interface between $ SnS_{2} $ and $ ReSe_{2} $ and is enhanced by the photogating effect. The device's long-term image memory is attributed to the trapping of holes at defect states, which sustain the photocurrent even after the light is turned off. The device's spectral reconstruction is achieved through compressed sensing and robust principal component analysis, enabling the reconstruction of unknown spectra from a small number of measurements. The device's spectral imaging capabilities are demonstrated through the reconstruction of reflected spectra from colored images. The device's neuromorphic computing capabilities are demonstrated through its ability to function as both a memory and a synapse, enabling the classification of handwritten digits using a two-layer artificial neural network. The device's performance is compared to other miniaturized spectrometers, showing its advantages in terms of footprint, spectral range, and resolution. The device's long-term image memory and neuromorphic computing capabilities make it a promising candidate for applications in artificial synapses, neural networks, machine vision, and imaging processing.
Reach us at info@study.space