High-Resolution On-Chip Spectral Reconstruction via Graded-Bandgap Perovskite Single-Crystal Heterojunctions

Spectral computing and reconstruction technology does not require the complex optical paths and optical components in traditional spectral detection, and can achieve the integration of the detection system. It has important application prospects in medical imaging, industrial inspection, remote sensing, and environmental monitoring. However, photodetectors (PDs) used for spectral reconstruction are often prepared with filters. Once the photodetectors under this scheme are prepared, they cannot be dynamically adjusted. Meanwhile, the extensive use of filters will also cause mutual constraints between spectral resolution and spatial resolution. Herein, we propose a multilayer organic-inorganic halide perovskite single-crystal heterojunction (MPSCH) for spectral reconstruction via solution-processed epitaxial growth, with a graded-bandgap structure of MAPbCl₃/MAPbBr_1.5Cl_1.5/MAPbBr₃/MAPbI_1.5Br_1.5/MAPbI₂Br to achieve wavelength-dependent photon absorption across the vertical stack. Instead of relying on static physical filters, we exploit the nonlinear extraction of photogenerated carriers under varying electric fields. This bias-tunable mechanism allows for the dynamic modulation of spectral response, enabling a number of detection channels that significantly exceed the limit of physical filter arrays, thus facilitating higher spectral resolution without sacrificing spatial resolution. The unknown spectrum is fitted by using the Gaussian function through the least-squares method based on this device. The final structure can achieve spectral calculation and reconstruction within the range of 420 nm-740 nm. Compared with the commonly used 9/16/25 channels in filter array-based devices, our device can achieve more than 60 reconstruction channels. Single-point scanning imaging of the picture can be achieved through this device. The minimum mean square error (MSE) order of magnitude of the reconstructed spectrum calculated by it, compared with commercial spectrometers, can reach 10^-4.