Hyperspectral perovskite X-ray detection method and system

Abstract

The invention discloses a hyperspectral perovskite X-ray detection method and system. The system comprises a multi-component perovskite array detector, a signal acquisition and amplification module, a spectrum unmixing module, a hyperspectral imaging module and a material identification and quantitative analysis module. The method comprises the following steps of: preparing perovskite array detectors with different components on a TFT / CMOS (Thin Film Transistor / Complementary Metal-Oxide-Semiconductor Transistor) plate by adopting a mask spraying process; processing a signal output by the perovskite array detector through a signal acquisition and amplification module to obtain a multi-component response vector corresponding to each pixel; carrying out unmixing processing on the multi-component response vector through a spectrum unmixing module, and reconstructing an incident energy spectrum; a'space * energy 'data cube is constructed through the hyperspectral imaging module, and a hyperspectral image is generated; material identification and quantitative analysis are realized based on a preset database through a material identification and quantitative analysis module; according to the invention, the problem that the resolution capability of a traditional detector is reduced in a wide-energy-band and high-flux scene can be solved.

Description

Technical Field

[0001] This invention belongs to the field of perovskite detection technology, and more specifically, relates to a hyperspectral perovskite X-ray detection method and system. Background Technology

[0002] X-ray energy spectroscopy is a core analytical tool in fields such as medical imaging (e.g., energy spectral CT), industrial non-destructive testing, mineral and metallurgical analysis, and archaeological research, placing stringent demands on the comprehensive performance of detectors. An ideal X-ray energy spectroscopy detector must simultaneously meet the following requirements: a wide energy range response capability of 2-100 keV, a high energy resolution level as low as 5%, high throughput and low noise characteristics, and the engineering potential to easily achieve large-area arrays.

[0003] In existing technologies, energy spectrum detection schemes mostly rely on direct conversion materials such as silicon (Si) or high-Z semiconductors (e.g., CdTe, CZT), achieving energy differentiation through pulse height analysis (PHA) or multi-threshold counting techniques. The working principle of these direct conversion materials is to utilize photons to generate electron-hole pairs within the bulk material and collect charges to form pulses. Their energy resolution performance is jointly determined by statistical fluctuations (Fano factor), electronic noise, leakage current, and carrier collection efficiency. To expand the detection energy range and improve energy identification capabilities, the engineering field often employs the following optimization methods: ① Hardware layering / filtering technology, using multilayer absorber K-edge filters to coarsely segment the spectral bands; ② Multi-threshold counting or pixel-level PHA technology, setting several thresholds and statistically counting to obtain an approximate energy spectrum; ③ Array imaging technology, achieving material differentiation through pixel-energy joint spatial methods. However, the above-mentioned technical solutions have inherent defects in practical applications: when faced with high-throughput and wide-energy-range detection scenarios, problems such as charge sharing, pulse stacking, threshold drift and poor consistency between pixels are prone to occur, resulting in a reduction in the number of available energy channels of the detector and a decrease in effective resolution. At the same time, it significantly increases the system calibration cost and the difficulty of long-term stability maintenance, and fails to achieve effective breakthroughs in the two core requirements of wide-spectrum continuous energy resolution and multi-material hyperspectral identification.

[0004] Perovskite materials (with the general chemical formula ABX3, where A is selected from one or more of Cs, FA, and MA, and X is selected from one or more of Cl, Br, and I) naturally possess the potential for engineered "tunable spectral response" due to their high linear decay coefficient, tunable bandgap, high carrier mobility-lifetime product (μτ), and the advantage of being prepared via low-temperature solution methods. The high atomic number (some containing elements such as Pb and Bi), high absorption coefficient, and tunable bandgap characteristics of these materials enable them to exhibit excellent absorption performance for low-energy to medium-high-energy X-rays. Through composition / halogen control and thickness engineering, perovskite materials with different formulations can exhibit differentiated quantum efficiencies, carrier lifetimes, and trap spectra at different energy bands. Furthermore, their low-temperature compatibility with CMOS / TFT supports large-area arrays and pixelated readout. However, current published research mainly focuses on the response characteristics and imaging verification of single-formulation perovskite materials or limited energy bands. A systematic technical solution has not yet been developed to address the need for calibrated "multi-material + multi-energy band" joint identification across a wide spectral range of 2-100 keV. Summary of the Invention

[0005] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a hyperspectral perovskite X-ray detection method and system. By constructing a multi-component perovskite array detector on a TFT / CMOS board and deeply collaborating with spectral unmixing algorithms, hyperspectral imaging, and material identification and analysis modules, it not only achieves continuous energy resolution over a wide energy range of 2–100 keV (FWHM ≤ 5% at 59.5 keV), but also significantly improves spectral line separation and image contrast at the same throughput. This effectively solves problems such as charge sharing, pulse stacking, threshold drift, and pixel inconsistency that easily occur in traditional Si or CdTe / CZT-based detection schemes in wide-energy, high-throughput scenarios. Furthermore, relying on the "space × energy" data cube and K-edge differential technology, it greatly improves the accuracy of multi-material (element / phase) identification. The invention achieves consistency in both accuracy and quantification, filling the gap in existing perovskite-related technologies by lacking a systematic "multi-material + multi-energy band" joint identification scheme. Simultaneously, through a multi-component material and algorithm synergy approach, it reduces hardware channels and calibration steps, lowering equipment size and manufacturing costs. Combined with the compatibility of perovskite material preparation via low-temperature solution processing and CMOS / TFT, it supports large-area array integration and mass production, improving production efficiency and operational stability. Furthermore, through physical prior constraints and drift self-calibration mechanisms on the algorithm side, it ensures long-term reliable online operation of the system, requiring a lower imaging dose. It achieves comprehensive optimization in detection quality, accuracy, efficiency, and ease of use, providing a superior X-ray detection solution for fields such as medical imaging, industrial non-destructive testing, mineral and metallurgical analysis, and archaeological research.

[0006] To achieve the above objectives, one aspect of the present invention provides a hyperspectral perovskite X-ray detection method, comprising the following steps: S1. Perovskite array detectors of different compositions are fabricated on a TFT / CMOS board using a mask spraying process; S2. The signal output from the perovskite array detector is processed by the signal acquisition and amplification module to obtain the multi-component response vector corresponding to each pixel; S3. The multi-component response vector is demixed using the spectral demixing module to reconstruct the incident energy spectrum; S4. Construct a "space × energy" data cube and generate hyperspectral images using the hyperspectral imaging module; S5. Material identification and quantitative analysis are performed based on a preset database through the material identification and quantitative analysis module.

[0007] Furthermore, the fabrication process of the multi-component perovskite array detector in step S1 includes the following steps: S11: Substrate and Pre-processing: a-SiTFT or CMOS APS / ROIC is used as the readout substrate, with a pixel pitch of 50-200μm, and the substrate surface is covered with SiN. x Planarization / passivation dielectric; after windowing, deposit Ti / Au pixel electrodes with a thickness of 50-200nm and arrange guard rings / isolation trenches; treat with UV-O3 for 1-5min to adjust the contact angle of the substrate surface to 20-50°; S12: Precursor and ink preparation: Prepare perovskite precursor ink according to stoichiometric ratio, wherein the perovskite is selected from at least two of CsPbBr3, FAPbBr3, MAPbCl3, and Cs2AgBiBr6, and the precursor ink is filtered twice through a 0.22μm filter head; S13: Mask alignment and spraying: Use a stainless steel / PI shadow mask with a thickness of 30-100μm, with a gap between the mask and the substrate of <50μm and an alignment error of ≤±5μm; perform multi-cycle spraying using an ultrasonic / pneumatic nozzle, with a nozzle-substrate distance of 5-15cm, substrate temperature controlled at 60-90℃, and spray each component for 10-200 cycles to form a perovskite layer with a thickness of 10-300μm. Pre-bake at 80-120℃ for 1-5min between cycles, and finally anneal at 100-150℃ for 10-60min. S14: Top electrode fabrication: A common bias electrode of Au with a thickness of 50-200 nm is deposited by mask evaporation, and an electron / hole blocking layer of 10-50 nm is deposited if necessary.

[0008] Furthermore, the formulation of the precursor and ink in step S12 satisfies any of the following conditions: Condition 1: In the CsPbBr3 precursor ink, the mass molar ratio of CsBr to PbBr2 is 1:1, and the solvent is a mixture of DMF and DMSO in a volume ratio of 7:3. Condition 2: In the FAPbBr3 precursor ink, the mass molar ratio of FABr to PbBr2 is 1.05:1.0, and 5-10% excess PbBr2 is added; Condition 3: Use MAPbCl3 or Cs2AgBiBr6 as the high-energy-range responsive perovskite precursor ink.

[0009] Furthermore, the working process of the signal acquisition and amplification module in step S2 includes: sequentially performing current-to-voltage conversion, low-noise amplification, basic shaping, and filtering on the weak current output by the multi-component perovskite array detector; Synchronous sampling and digitization are performed, and baseline / dark current correction and temperature drift compensation are combined to output multi-channel time series or frame data in a standardized format, forming a multi-component response vector corresponding to each pixel.

[0010] Furthermore, the working process of the spectral unmixing module in step S3 includes: establishing a response matrix and completing initial calibration using standard samples or characteristic spectral lines; During online operation, the multi-component response vector of each pixel is input into the demixing process, and a linear or nonlinear solution with physical prior constraints is used for demixing to separate the signal contribution of photons with different energies and reconstruct the incident energy spectrum; at the same time, gain drift and zero drift self-calibration are performed in combination with reference lines or built-in benchmarks.

[0011] Furthermore, the physical prior constraints mentioned in step S3 include nonnegativity constraints, smoothness constraints, and feature absorption edge preservation constraints; the linear or nonlinear solution method is selected from one or more of NMF, PCA preprocessing combined regression, and lightweight neural network residual correction. Furthermore, the working process of the hyperspectral imaging module in step S4 includes: fusing the incident energy spectrum reconstructed in step S3 with the pixel coordinates to construct a "space × energy" data cube; Based on the data cube, an energy window map, a difference map before and after the feature absorption edge, and a material comparison map are generated, and spatial / energy domain denoising and artifact suppression are performed.

[0012] Furthermore, the working process of the material identification and quantitative analysis module described in step S5 includes: Establish and maintain three types of core databases: end-member response library of multi-component perovskites, reference energy spectrum and absorption edge library of target materials, and equipment status and calibration history library; The system calls a preset database to perform library matching and unmixing on the reconstructed energy spectrum of each pixel, outputting the material category, endmember abundance and uncertainty, and generating element / phase distribution map, quantitative thermogram and quality control map. Record the complete analysis process and version information, and interface with medical, security inspection or industrial inspection systems in a standard data format.

[0013] A second aspect of the present invention provides a hyperspectral perovskite X-ray detection system for implementing the aforementioned hyperspectral perovskite X-ray detection method, comprising a multi-component perovskite array detector, a signal acquisition and amplification module, a spectral demixing module, a hyperspectral imaging module, and a material identification and quantitative analysis module; The multi-component perovskite array detector is fabricated on a TFT / CMOS board by a mask spraying process using perovskite arrays of different components. The signal acquisition and amplification module is used to convert, amplify, filter, sample and correct the weak current output by the multi-component perovskite array detector, and output the multi-component response vector corresponding to each pixel. The spectral unmixing module is used to reconstruct the incident energy spectrum based on the response differences of multi-component perovskites through response matrix calibration and unmixing algorithms. The hyperspectral imaging module is used to fuse the reconstructed energy spectrum with pixel coordinates to generate a hyperspectral image. The material identification and quantitative analysis module is used to perform material identification and quantitative analysis based on a preset database.

[0014] Furthermore, the preset database includes an endmember response library, a reference energy spectrum and absorption edge library, and a device status and calibration history library; the endmember response library is used to store response characteristic data of multi-component perovskites, and the reference energy spectrum and absorption edge library is used to store reference energy spectrum and absorption edge data of target materials.

[0015] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: (1) The hyperspectral perovskite X-ray detection method and system of the present invention achieves continuous energy resolution in a wide energy range of 2–100 keV through the hardware and software co-design of multi-component perovskite array and spectral unmixing algorithm. The energy resolution FWHM is ≤5% at 59.5 keV. Compared with the traditional direct conversion detection scheme based on Si or CdTe / CZT, it significantly improves the spectral line separation and image contrast under the same flux conditions. It effectively solves the problems of charge sharing, pulse stacking, threshold drift and pixel inconsistency that are easy to occur in the traditional scheme in wide energy range and high flux scenarios, and greatly increases the number of available energy channels and effective resolution.

[0016] (2) The hyperspectral perovskite X-ray detection method and system of the present invention, relying on the construction of "space × energy" data cube and K-edge differential technology, significantly improves the accuracy and quantitative consistency of multi-material (element / phase) identification, and can output pixel-level element distribution and abundance maps. It can fill the technical gap that existing perovskite-related detection technologies focus on single formulations or limited energy band responses and lack a systematic "multi-material + multi-energy band" joint identification scheme.

[0017] (3) The hyperspectral perovskite X-ray detection method and system of the present invention adopts a technical approach that combines multi-component perovskite materials with spectral unmixing algorithms. Compared with the traditional detection scheme that relies on multilayer filtering or high-end crystals, it reduces the number of hardware channels and system calibration steps, and reduces the equipment size and manufacturing cost. At the same time, the low-temperature solution preparation process of perovskite materials has good compatibility with CMOS / TFT substrates, supports large-area array integration and large-scale manufacturing, and can significantly improve production efficiency and system operation stability.

[0018] (4) The hyperspectral perovskite X-ray detection method and system of the present invention introduces physical prior constraints such as non-negative, smooth, and characteristic absorption edge preservation in the spectral unmixing algorithm, and combines them with gain drift and zero drift self-calibration mechanism to effectively offset the performance degradation caused by environmental changes and device aging, and ensure the long-term online reliable operation of the system in the full spectrum. Moreover, compared with the traditional scheme, the present invention requires a lower imaging dose, which is beneficial to energy consumption control and environmental safety, and realizes the comprehensive optimization of detection quality, accuracy, efficiency and ease of use.

[0019] (5) The hyperspectral perovskite X-ray detection method and system of the present invention improves the condition number and robustness of the response matrix through composition / halogen control, thickness engineering and bias / readout link optimization design, and achieves near pulse height spectrum level resolution performance with a limited number of channels. It successfully achieves a balance between core performance requirements such as wide energy range response, high energy resolution, high throughput and low noise and large area arraying, and provides a better X-ray detection technology solution for fields such as medical imaging, industrial non-destructive testing, mineral metallurgical analysis and archaeological research. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the structure of the hyperspectral perovskite X-ray detection system according to an embodiment of the present invention; Figure 2 This is a schematic flowchart of the hyperspectral perovskite X-ray detection method according to an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0022] The core idea of ​​this invention's hyperspectral detection is that "each pixel corresponds to a high-dimensional spectrum," which in the X-ray detection field translates to "each pixel corresponds to an energy-response vector." Spectral unmixing technology decomposes the observed signal into several "endmembers" and their "abundance," enabling the reconstruction of the true components or energy components. Typical algorithms include NMF, PCA, ICA, sparse coding, Bayesian / iterative constraint-based solutions based on physical priors, and end-to-end neural network regression. If the energy response function Ri(E) of perovskite materials with different compositions can be discernibly differentiated, the incident energy spectrum I(E) can be inverted by solving S=RI from the observed vector S=[S1,S2,…], achieving hardware-software co-operational energy reconstruction. For wide-spectrum applications ranging from 2-100keV, by combining multi-component perovskite materials and optimizing the thickness / bias / readout links, the condition number and robustness of the response matrix R can be improved, thereby achieving near-pulse hyperspectral resolution performance with a limited number of channels. Therefore, by utilizing the differences in spectral response of perovskites with different compositions and combining them with a spectral unmixing algorithm to achieve energy resolution within the range of 2-100 keV, it is possible to exert efforts on both the material and algorithm ends simultaneously, providing a new technical path for solving the challenges of wide-spectrum high resolution and multi-material identification.

[0023] Based on the above reasons, such as Figure 1As shown, one aspect of the present invention provides a hyperspectral perovskite X-ray detection system, including a multi-component perovskite array detector, a signal acquisition and amplification module, a spectral demixing module, a hyperspectral imaging module, and a material identification and quantitative analysis module; the multi-component perovskite array detector is fabricated on a TFT / CMOS board by a mask spraying process using perovskite arrays of different components; the signal acquisition and amplification module is used to convert, amplify, filter, sample, and correct the weak current output by the multi-component perovskite array detector, and output a multi-component response vector corresponding to each pixel; the spectral demixing module... The mixing module is used to reconstruct the incident energy spectrum based on the response differences of multi-component perovskites through response matrix calibration and demixing algorithms; the hyperspectral imaging module is used to fuse the reconstructed energy spectrum with pixel coordinates to generate a hyperspectral image; the material identification and quantitative analysis module is used to realize material identification and quantitative analysis based on a preset database; the preset database includes an endmember response library, a reference energy spectrum and absorption edge library, and a device status and calibration history library; the endmember response library is used to store the response characteristic data of multi-component perovskites, and the reference energy spectrum and absorption edge library is used to store the reference energy spectrum and absorption edge data of the target material.

[0024] Furthermore, the fabrication process of the multi-component perovskite array detector includes the following steps: S11: Substrate and Pre-processing: a-SiTFT or CMOS APS / ROIC is used as the readout substrate, with a pixel pitch of 50-200μm, and the substrate surface is covered with SiN. x Planarization / passivation dielectric; after windowing, deposit Ti / Au pixel electrodes with a thickness of 50-200nm, and arrange guard rings / isolation trenches to reduce sidewall leakage; treat with UV-O3 for 1-5 minutes to adjust the contact angle of the substrate surface to 20-50° to facilitate spray wetting but avoid excessive percolation; S12: Precursor and ink preparation: Prepare perovskite precursor ink according to stoichiometric ratio, wherein the perovskite is selected from at least two of CsPbBr3, FAPbBr3, MAPbCl3, and Cs2AgBiBr6, and the precursor ink is filtered twice through a 0.22μm filter head; S13: Mask alignment and spraying: Use a stainless steel / PI shadow mask with a thickness of 30-100μm, with a gap between the mask and the substrate of <50μm and an alignment error of ≤±5μm; perform multi-cycle spraying using an ultrasonic / pneumatic nozzle, with a nozzle-substrate distance of 5-15cm, substrate temperature controlled at 60-90℃, and spray each component for 10-200 cycles to form a perovskite layer with a thickness of 10-300μm. Pre-bake at 80-120℃ for 1-5min between cycles, and finally anneal at 100-150℃ for 10-60min. S14: Top electrode fabrication: Deposit a 50-200 nm thick Au common bias electrode using a mask, and if necessary, deposit a 10-50 nm electron / hole blocking layer (such as C). 60 NiO X (10-50nm), suppressing injection leakage.

[0025] Furthermore, the formulation of the precursor and ink in step S12 satisfies any of the following conditions: Condition 1: In the CsPbBr3 precursor ink, the mass molar ratio of CsBr to PbBr2 is 1:1, and the solvent is a mixture of DMF and DMSO in a volume ratio of 7:3. Condition 2: In the FAPbBr3 precursor ink, the mass molar ratio of FABr to PbBr2 is 1.05:1.0, and 5-10% excess PbBr2 is added; Condition 3: Use MAPbCl3 or Cs2AgBiBr6 as the high-energy-range responsive perovskite precursor ink; Furthermore, the working process of the signal acquisition and amplification module includes: sequentially performing current-to-voltage conversion, low-noise amplification, basic shaping and filtering on the weak current output by the multi-component perovskite array detector; then performing synchronous sampling and digitization, and combining baseline / dark current correction and temperature drift compensation to output multi-channel time series or frame data in a standardized format to form a multi-component response vector corresponding to each pixel; The working process of the spectral unmixing module includes: establishing a response matrix and completing initial calibration using standard samples or characteristic spectral lines; during online operation, inputting the multi-component response vector of each pixel into the unmixing process, using a linear or nonlinear solution method with physical prior constraints to unmix, separating the signal contributions of photons of different energies and reconstructing the incident energy spectrum; and simultaneously performing gain drift and zero-point drift self-calibration in conjunction with a reference line or built-in benchmark.

[0026] The physical prior constraints include nonnegativity constraints, smoothness constraints, and feature absorption edge preservation constraints; the linear or nonlinear solution method is selected from one or more of NMF, PCA preprocessing combined with regression, and lightweight neural network residual correction.

[0027] The hyperspectral imaging module operates by fusing the reconstructed incident energy spectrum with pixel coordinates to construct a "space × energy" data cube; generating an energy window map, a difference map before and after feature absorption edges, and a material comparison map based on the data cube, and performing spatial / energy domain denoising and artifact suppression; and supporting data registration and intensity normalization for static and dynamic scenes.

[0028] The working process of the material identification and quantitative analysis module includes: calling a preset database, performing library matching and unmixing on the reconstructed energy spectrum of each pixel, outputting the material category, endmember abundance and uncertainty, generating element / phase distribution map, quantitative thermogram and quality control map; recording the analysis process and version information, and connecting to medical, security inspection or industrial inspection systems in a standard data format.

[0029] This invention achieves continuous energy resolution over a wide energy range of 2–100 keV through the synergistic effect of a multi-component perovskite array and a spectral unmixing algorithm. At 59.5 keV, the energy resolution FWHM is ≤5%. Under the same throughput conditions, the spectral line separation and image contrast are significantly improved compared with traditional schemes, which can effectively solve the problem of reduced resolution of traditional detectors in wide energy range and high throughput scenarios.

[0030] like Figure 2 As shown, a second aspect of the present invention provides a hyperspectral perovskite X-ray detection method, specifically including the following steps: S1. Fabrication of a multi-component perovskite array detector: A perovskite array detector with different compositions is fabricated on a TFT / CMOS board using a mask spraying process; including the following steps: S11: Substrate and Pre-processing: a-SiTFT or CMOS APS / ROIC is used as the readout substrate, with a pixel pitch of 50-200μm, and the substrate surface is covered with SiN. x Planarization / passivation of the dielectric improves surface smoothness and reduces leakage current. Ti / Au pixel electrodes are deposited in the windowed area of ​​the substrate, with the electrode thickness controlled at 50–200 nm. Guard ring / isolation trench structures are also arranged to effectively reduce the risk of sidewall leakage current. The substrate is treated with UV-O3 for 1–5 minutes to adjust the substrate surface contact angle to 20–50°, ensuring both effective ink wetting during subsequent spraying and preventing uneven film formation due to excessive percolation. S12: Precursor and ink formulation: Formulate various perovskite precursor inks according to near stoichiometric mass molar ratios, wherein the perovskite is selected from at least two of CsPbBr3, FAPbBr3, MAPbCl3, and Cs2AgBiBr6; The specific formula includes: CsPbBr3 precursor ink: The mass molar ratio of CsBr to PbBr2 is 1:1, and the solvent is a mixture of DMF and DMSO in a volume ratio of 7:3. FAPbBr3 precursor ink: The mass molar ratio of FABr to PbBr2 is 1.05:1.0, and 5–10% excess PbBr2 is added to promote film densification. High-energy-range responsive precursor inks: MAPbCl3 or Cs2AgBiBr6 are selected; All precursor inks are filtered twice through a 0.22μm filter to remove impurity particles and ensure ink purity and uniformity. S13: Mask Alignment and Spraying: A stainless steel / PI shadow mask with a thickness of 30-100μm is used. The gap between the mask and the substrate is controlled within 50μm, and the alignment error is ≤±5μm to ensure that the perovskite material is accurately deposited in the target pixel area. Multi-cycle spraying is performed using an ultrasonic / pneumatic nozzle. The nozzle-substrate distance is 5-15cm, and the substrate temperature is controlled at 60-90℃. Solvent evaporation promotes ink setting. Each component of perovskite material is sprayed in a multi-cycle manner, with each component sprayed for 10-200 cycles to form a perovskite layer with a thickness of 10-300μm. Between adjacent spraying cycles, pre-baking is performed at 80-120℃ for 1-5min to remove residual solvent. After all components are sprayed, a final annealing treatment is performed at a temperature of 100-150℃ for 10-60min to further improve the crystal quality and stability of the perovskite film. S14: Top Electrode Fabrication: Deposit a 50–200 nm thick Au common bias electrode using a mask evaporation process. If necessary, deposit a 10–50 nm electron / hole blocking layer (such as C) between the perovskite layer and the top electrode. 60 NiO X This effectively suppresses carrier injection leakage and improves the stability of the detector's electrical performance. S2. Signal Acquisition and Amplification: The signal output from the perovskite array detector is processed by the signal acquisition and amplification module to obtain a multi-component response vector for each pixel; specifically including: The pixel electrodes of the multi-component perovskite array are connected pixel-by-pixel to the readout board. The weak current signal output by the detector first enters the front-end processing circuit, where current-to-voltage conversion and low-noise amplification are performed. Subsequently, basic shaping and filtering are conducted to suppress out-of-band noise and impulse interference. The processed signal is then synchronously sampled and digitized. Combined with baseline / dark current correction and temperature drift compensation techniques, signal distortion caused by environmental factors and the device's own characteristics is eliminated, resulting in stable and comparable multi-channel time series or frame data. This step also provides a standardized energy line / flat field / dark field acquisition process and data format to ensure data consistency and reliability. Finally, the "multi-component response vector" corresponding to each pixel is output as input data for subsequent spectral demixing. S3. Spectral demixing and energy spectrum reconstruction: The incident energy spectrum is reconstructed by unmixing the multi-component response vectors using a spectral unmixing module; specifically, this includes: Based on the response differences of multi-component perovskite materials, a "response matrix" is first established using standard samples or characteristic spectral lines, and initial calibration is completed to clarify the response characteristics of different perovskite components to X-rays of various energies. During online operation, the multi-component response vector of each pixel is input into the unmixing process. Linear or nonlinear solutions with physical prior constraints (including non-negativity constraints, smoothness constraints, and characteristic absorption edge preservation constraints) are preferentially employed. Specifically, one or more of the following can be used: NMF algorithm, PCA preprocessing combined with regression, and lightweight neural network residual correction. This separates the contribution of photons of different energies to the signal and reconstructs the incident energy spectrum. Simultaneously, gain drift and zero-point drift self-calibration is performed using reference lines or built-in benchmarks to continuously evaluate energy resolution performance and ensure stable interpretation across the entire spectrum. S4, Hyperspectral Imaging: A "space × energy" data cube is constructed using a hyperspectral imaging module to generate hyperspectral images; specifically including: The incident energy spectrum and pixel coordinates obtained from step S3 are fused to construct a "space × energy" data cube, achieving deep coupling of spatial and energy information. Based on this data cube, energy window maps, feature absorption edge difference maps, and material comparison maps are generated according to actual application requirements. Spatial / energy domain denoising and artifact suppression algorithms are used to improve image visualization quality. This step supports data registration and intensity normalization for static and dynamic scenes, ensuring that comparable hyperspectral images are obtained under different times and throughput conditions, providing a unified data entry point for subsequent material identification and quantitative analysis. S5. Material Identification and Quantitative Analysis: The material identification and quantitative analysis module performs material identification and quantitative analysis based on a preset database; specifically, it includes: Three core databases are established and maintained: an endmember response library for multi-component perovskites, a reference energy spectrum and absorption edge library for target materials (elements / compounds), and a database of equipment status and calibration history. These databases are then used to perform library matching and demixing on the reconstructed energy spectrum of each pixel, outputting the material category, endmember (element / phase) abundance, and uncertainty, as well as generating element / phase distribution maps, quantitative thermograms, and quality control charts. Simultaneously, the complete analysis process and version information are recorded to facilitate result traceability and consistency comparison across devices / batch levels. The data is then integrated with medical, security, or industrial inspection systems in a standard data format to meet the integration needs of different application scenarios.

[0031] To make the technical solution of the present invention clearer and more explicit, the present invention will be described in detail below with reference to specific embodiments.

[0032] Example 1: Fabrication of a multi-component perovskite array detector A CMOS APS with a pixel pitch of 100μm was selected as the readout substrate, and the substrate surface was coated with SiN.X Passivation medium; after photolithography windowing, Ti / Au pixel electrodes are deposited using electron beam evaporation process, with a Ti layer thickness of 50nm and an Au layer thickness of 100nm, and a retaining ring structure is prepared at the same time; the substrate is placed in a UV-O3 cleaning equipment for 3min and the surface contact angle is measured to be 35°. Formulate three perovskite precursor inks: CsPbBr3 ink: CsBr (0.1 mol) and PbBr2 (0.1 mol) are dissolved in a mixed solvent of 7 mL LDMF and 3 mL DMSO, stirred evenly, and then filtered twice through a 0.22 μm filter. FAPbBr3 ink: FABr (0.105 mol) and PbBr2 (0.1 mol) are dissolved in the above mixed solvent, 5% excess PbBr2 is added, and the mixture is stirred, filtered, and set aside for later use; Cs2AgBiBr6 ink: Weigh CsBr, AgBr and BiBr3 according to the stoichiometric ratio, dissolve them in DMF solvent, stir and filter for later use; A 50μm thick stainless steel shadow mask was selected, aligned with the substrate, and fixed. The gap between the mask and the substrate was controlled at 30μm, with an alignment error ≤±3μm. An ultrasonic nozzle was used, with the nozzle-to-substrate distance set at 10cm, and the substrate temperature controlled at 80℃. The CsPbBr3 layer was sprayed for 50 cycles, with a thickness controlled at 100μm, and pre-baked at 100℃ for 3min between cycles. The FAPbBr3 layer was sprayed for 60 cycles, with a thickness controlled at 120μm. The Cs2AgBiBr6 layer was sprayed for 40 cycles, with a thickness controlled at 80μm. After all components were sprayed, the layers were annealed at 120℃ for 30min.

[0033] The detector was fabricated by using a thermal evaporation process to deposit a 100 nm thick Au top electrode through a mask, and then depositing a 30 nm thick C60 electron blocking layer between the Cs2AgBiBr6 layer and the Au electrode.

[0034] Example 2: Detection Method of Hyperspectral Perovskite X-ray Detection System Signal acquisition: The fabricated multi-component perovskite array detector is connected to a low-noise readout circuit. The detector is irradiated with continuous energy spectrum X-rays of 2-100keV from an X-ray source. The weak current signal output by the detector is converted into a voltage signal by the front-end circuit, amplified 1000 times by a low-noise amplifier, filtered by an 8th-order Butterworth filter (cutoff frequency 1MHz), and then synchronously sampled by a 16-bit ADC at a sampling frequency of 10MHz. Baseline correction and temperature drift compensation are performed simultaneously, and three channels (corresponding to three perovskite components) of time-series data are output to form a multi-component response vector for each pixel. Spectral unmixing: The system was initially calibrated using a standard Cs-137 radioactive source (characteristic energy 662keV) and an Am-241 radioactive source (characteristic energy 59.5keV) to establish the response matrix. During online operation, the NMF algorithm was used in combination with non-negativity constraints and characteristic absorption edge preservation constraints to unmix the multi-component response vector of each pixel and reconstruct the incident energy spectrum. At the same time, drift self-calibration was performed using the built-in reference signal, and calibration was performed every 10 minutes. Hyperspectral imaging: The reconstructed energy spectrum is fused with pixel coordinates to construct a "space × energy" data cube, generating energy window maps with energy windows of 10-30keV, 30-60keV, and 60-100keV, as well as difference maps before and after the Fe element K-edge (7.11keV). Spatial denoising is performed by median filtering to suppress artifacts. Material identification and quantitative analysis: By calling the end-member response library, reference energy spectra and absorption edge libraries of elements such as Fe, Cu, and Al, the energy spectrum of the imaging region is matched and unmixed, the abundance distribution of each element is output, and a quantitative heat map is generated. The quantitative error of Fe element is ≤3%. The data is output in DICOM format and connected to the medical imaging system. The detector prepared in Example 1 of this invention was compared with a conventional CdTe detector in terms of performance. The test conditions were: X-ray flux 10 6 Photons / second, detection energy range 2-100keV, the test results are shown in Table 1 below: Table 1 - Test results of the detector prepared in Example 1 The test results show that the detector of this invention is superior to the traditional CdTe detector in terms of core indicators such as energy resolution, number of channels, and multi-material recognition accuracy. It also has a longer calibration cycle and lower manufacturing cost, and has significant technical advantages and application prospects.

[0035] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A hyperspectral perovskite X-ray detection method, characterized in that, Includes the following steps: S1. Perovskite array detectors of different compositions are fabricated on a TFT / CMOS board using a mask spraying process; S2. The signal output from the perovskite array detector is processed by the signal acquisition and amplification module to obtain the multi-component response vector corresponding to each pixel; S3. The multi-component response vector is demixed using the spectral demixing module to reconstruct the incident energy spectrum; S4. Construct a "space × energy" data cube and generate hyperspectral images using the hyperspectral imaging module; S5. Material identification and quantitative analysis are performed based on a preset database through the material identification and quantitative analysis module.

2. The hyperspectral perovskite X-ray detection method according to claim 1, characterized in that, The fabrication process of the multi-component perovskite array detector in step S1 includes the following steps: S11: Substrate and Pre-processing: a-SiTFT or CMOS APS / ROIC is used as the readout substrate, with a pixel pitch of 50-200μm, and the substrate surface is covered with SiN. x Planarization / passivation dielectric; after windowing, deposit Ti / Au pixel electrodes with a thickness of 50-200nm and arrange guard rings / isolation trenches; treat with UV-O3 for 1-5min to adjust the contact angle of the substrate surface to 20-50°; S12: Precursor and ink preparation: Prepare perovskite precursor ink according to stoichiometric ratio, wherein the perovskite is selected from at least two of CsPbBr3, FAPbBr3, MAPbCl3, and Cs2AgBiBr6, and the precursor ink is filtered twice through a 0.22μm filter head; S13: Mask alignment and spraying: Use a stainless steel / PI shadow mask with a thickness of 30-100μm, with a gap between the mask and the substrate of <50μm and an alignment error of ≤±5μm; perform multi-cycle spraying using an ultrasonic / pneumatic nozzle, with a nozzle-substrate distance of 5-15cm, substrate temperature controlled at 60-90℃, and spray each component for 10-200 cycles to form a perovskite layer with a thickness of 10-300μm. Pre-bake at 80-120℃ for 1-5min between cycles, and finally anneal at 100-150℃ for 10-60min. S14: Top electrode fabrication: A common bias electrode of Au with a thickness of 50-200 nm is deposited by mask evaporation, and an electron / hole blocking layer of 10-50 nm is deposited if necessary.

3. The hyperspectral perovskite X-ray detection method according to claim 2, characterized in that, The preparation of the precursor and ink in step S12 satisfies any of the following conditions: Condition 1: In the CsPbBr3 precursor ink, the mass molar ratio of CsBr to PbBr2 is 1:1, and the solvent is a mixture of DMF and DMSO in a volume ratio of 7:

3. Condition 2: In the FAPbBr3 precursor ink, the mass molar ratio of FABr to PbBr2 is 1.05:1.0, and 5-10% excess PbBr2 is added; Condition 3: Use MAPbCl3 or Cs2AgBiBr6 as the high-energy-range responsive perovskite precursor ink.

4. A hyperspectral perovskite X-ray detection method according to any one of claims 1-3, characterized in that, The working process of the signal acquisition and amplification module in step S2 includes: sequentially performing current-to-voltage conversion, low-noise amplification, basic shaping and filtering on the weak current output by the multi-component perovskite array detector; Synchronous sampling and digitization are performed, and baseline / dark current correction and temperature drift compensation are combined to output multi-channel time series or frame data in a standardized format, forming a multi-component response vector corresponding to each pixel.

5. A hyperspectral perovskite X-ray detection method according to any one of claims 1-3, characterized in that, The working process of the spectral unmixing module in step S3 includes: establishing a response matrix and completing initial calibration using standard samples or characteristic spectral lines; During online operation, the multi-component response vector of each pixel is input into the demixing process, and a linear or nonlinear solution with physical prior constraints is used for demixing to separate the signal contribution of photons with different energies and reconstruct the incident energy spectrum; at the same time, gain drift and zero drift self-calibration are performed in combination with reference lines or built-in benchmarks.

6. The hyperspectral perovskite X-ray detection method according to claim 5, characterized in that, The physical prior constraints mentioned in step S3 include nonnegativity constraints, smoothness constraints, and feature absorption edge preservation constraints; the linear or nonlinear solution method is selected from one or more of NMF, PCA preprocessing combined regression, and lightweight neural network residual correction.

7. A hyperspectral perovskite X-ray detection method according to any one of claims 1-3, characterized in that, The working process of the hyperspectral imaging module in step S4 includes: fusing the incident energy spectrum obtained from step S3 reconstruction with the pixel coordinates to construct a "space × energy" data cube; Based on the data cube, an energy window map, a difference map before and after the feature absorption edge, and a material comparison map are generated, and spatial / energy domain denoising and artifact suppression are performed.

8. A hyperspectral perovskite X-ray detection method according to any one of claims 1-3, characterized in that, The working process of the material identification and quantitative analysis module in step S5 includes: Establish and maintain three types of core databases: end-member response library of multi-component perovskites, reference energy spectrum and absorption edge library of target materials, and equipment status and calibration history library; The system calls a preset database to perform library matching and unmixing on the reconstructed energy spectrum of each pixel, outputting the material category, endmember abundance and uncertainty, and generating element / phase distribution map, quantitative thermogram and quality control map. Record the complete analysis process and version information, and interface with medical, security inspection or industrial inspection systems in a standard data format.

9. A hyperspectral perovskite X-ray detection system, characterized in that, The method for implementing the hyperspectral perovskite X-ray detection method as described in any one of claims 1-8 includes a multi-component perovskite array detector, a signal acquisition and amplification module, a spectral demixing module, a hyperspectral imaging module, and a material identification and quantitative analysis module; The multi-component perovskite array detector is fabricated on a TFT / CMOS board by a mask spraying process using perovskite arrays of different components. The signal acquisition and amplification module is used to convert, amplify, filter, sample and correct the weak current output by the multi-component perovskite array detector, and output the multi-component response vector corresponding to each pixel. The spectral unmixing module is used to reconstruct the incident energy spectrum based on the response differences of multi-component perovskites through response matrix calibration and unmixing algorithms. The hyperspectral imaging module is used to fuse the reconstructed energy spectrum with pixel coordinates to generate a hyperspectral image. The material identification and quantitative analysis module is used to perform material identification and quantitative analysis based on a preset database.

10. The hyperspectral perovskite X-ray detection method according to claim 9, characterized in that, The preset database includes an endmember response library, a reference energy spectrum and absorption edge library, and a device status and calibration history library; the endmember response library is used to store response characteristic data of multi-component perovskites, and the reference energy spectrum and absorption edge library is used to store reference energy spectrum and absorption edge data of target materials.

Images