Intermediate infrared focal plane array spectral imager and spectral imaging method thereof

Abstract

The invention relates to the field of mid-infrared focal plane array spectral imagers, and discloses a mid-infrared focal plane array spectral imager and a spectral imaging method thereof.The mid-infrared focal plane array spectral imager comprises a focal plane array, the focal plane array comprises a plurality of pixels, and the pixels are arranged in the focal plane array. Wherein each pixel comprises a PIN heterojunction and a CdS layer arranged at the interface of the PIN heterojunction, and the imaging method comprises the following steps: carrying out spectrum training on each pixel in a focal plane array of the mid-infrared focal plane array spectrum imager, and establishing an independent spectrum model for each pixel so as to form a spectrum model cube; performing spectrum reconstruction on each pixel in the focal plane array based on the spectrum model cube; and combining the reconstructed spectrums of all the pixel points to generate a spectral imaging result. The invention solves the problem that the prior art lacks a spectral imager for the mid-infrared focal plane, and has the characteristics of low cost, chip and simple process.

Description

Technical Field

[0001] The present invention relates to the field of mid-infrared focal plane array spectral imagers, and more specifically, to a mid-infrared focal plane array spectral imager and a spectral imaging method thereof. Background Art

[0002] Mid-infrared spectroscopic imagers utilize the differential absorption characteristics of infrared radiation corresponding to chemical bonds in molecules (such as CH, OH, C=O, and NH) in different mid-infrared bands to simultaneously capture spatial thermal and spectral information. They are widely used for qualitative identification, quantitative analysis, and molecular structure research. Based on the principle of spectral resolution, mid-infrared spectroscopic imagers can be divided into four types: filter-based, grating dispersion-based, Fourier interferometry-based, and computational spectroscopy-based. The first three types combine large mechanical filter elements with a cooled HgCdTe focal plane imager for high-spectral-resolution imaging to achieve scanning spectral resolution. Their high detection rate, low signal-to-noise ratio, and high mechanical scanning spectral resolution have led to their widespread use in the military and defense sectors. However, their large size (Dewar flask packaging, scanning spectral cavity), high energy consumption (liquid nitrogen cycle cooling), and slow response time (step-by-step scanning) have limited their further development. Computational spectral imaging can perform initial training on all response spectra while operating under the operation of an uncooled thermal imager. The trained accuracy replaces the resolution of the scanned spectrum, thereby reconstructing the spectrum of the target thermal image. This significantly reduces the cost of large-volume filter components and cooling equipment, and greatly improves the resolution of the computed spectrum, thus offering significant advantages in spectral imaging using computational spectral imaging.

[0003] Currently, uncooled computational spectroscopy imagers, both domestically and internationally, still rely on micro- and nanostructured filter spectroscopy devices, including resonant filters and broadband filters. The former utilizes a specific periodic structure to achieve resonant filtering, while the latter employs different filtering units within a single filter. However, the fabrication of micro- and nanostructured filters requires complex semiconductor processes (e.g., photolithography, development, etching, ion implantation, etc.), which increases manufacturing costs. In response, a pioneering effort abroad has proposed utilizing a tunable heterojunction without a filter to achieve filter-free, single-pixel mid-infrared computational spectroscopy imaging reconstruction. This approach eliminates the need for complex and expensive filtering components, reduces the cost and size of the imaging system, and increases spectral tuning flexibility. However, this approach also faces challenges, including the need to tune the device interface band structure and integrate it with the metal oxide semiconductor (CMOS) readout circuitry, which have become key research areas.

[0004] The heterojunction interface within a mid-infrared detector undergoes band bending under varying bias voltages, resulting in varying degrees of electron transport performance and, in turn, a wavelength shift in the detector's response spectrum with voltage. A greater degree of this response spectrum shift improves the accuracy of the trained spectrum, leading to higher resolution in the reconstructed spectrum. This shift is primarily related to the built-in electric field at the heterojunction interface: the stronger the built-in electric field, the greater the response spectrum shift.

[0005] The study found that different CdS loadings optimized the heterojunction interface in the PbS / PbSe / ZnO heterojunction, increasing the heterojunction's built-in electric field. This resulted in a larger response spectral shift (4.45-4.55 µm) in the single-point pixel spectrometer under the action of bias voltage. Furthermore, the reconstructed spectra, trained and reconstructed using a regularized algorithm, exhibited ultra-high mid-infrared spectral resolution (3.2 nm). Furthermore, the spectral training and reconstruction of single-point pixels in the spectrometer's heterogeneous integrated readout circuit must account for both pixel current and temperature non-uniformity.

[0006] In view of the problem that the existing technology currently lacks a spectroscopic imager for mid-infrared focal plane, how to invent an uncooled PbSe mid-infrared focal plane spectroscopic imager based on CdS is a technical problem that urgently needs to be solved in this technical field. Summary of the Invention

[0007] In order to solve the problem that the prior art lacks a spectral imager for mid-infrared focal planes, the present invention provides a mid-infrared focal plane array spectral imager and a spectral imaging method thereof, which have the characteristics of low cost, chip-based, and simple process.

[0008] In order to achieve the above-mentioned purpose of the present invention, the technical solutions adopted are as follows: A mid-infrared focal plane array spectroscopic imager includes a focal plane array, wherein the focal plane array includes a plurality of pixels, wherein each pixel includes a PIN heterojunction and a CdS layer arranged at the interface of the PIN heterojunction, and the PIN heterojunction includes a PbS hole transport layer, a Sn-doped PbSe photosensitive layer, and a ZnO electron transport layer in sequence.

[0009] Preferably, the CdS layer is formed at the interface of the PIN heterojunction of the mid-infrared focal plane array spectroscopic imager by ion deposition, and is used to control the built-in electric field of the heterojunction by regulating the deposition amount.

[0010] A spectral imaging method for a mid-infrared focal plane array spectral imager comprises the following specific steps: Performing spectral training on each pixel in the focal plane array of the mid-infrared focal plane array spectral imager, and establishing an independent spectral model for each pixel, thereby forming a spectral model cube; Based on the spectral model cube, the spectrum of each pixel in the focal plane array is reconstructed; The reconstructed spectra of all pixels are combined to generate spectral imaging results.

[0011] Preferably, performing spectral training on each pixel in the pixel array and establishing an independent spectral model for each pixel includes the following specific steps: Using a monochromatic light source to sequentially illuminate the pixels of the mid-infrared focal plane array spectroscopic imager at a plurality of predetermined wavelengths; At each of the predetermined wavelengths, a plurality of different bias voltages V are applied to the PIN heterostructure, and corresponding photocurrent values ​​I(λ, V) are measured synchronously; The photocurrent values ​​measured at all known wavelengths and bias voltages are correlated with the optical power P(λ) of the reference monochromatic light source at the corresponding wavelength to construct the responsivity matrix R(λ, V); A regularization algorithm is used, and the response rate matrix and standard spectral data at known wavelengths are used as input and output respectively to train the spectral model.

[0012] Furthermore, the reference monochromatic light source is a quantum cascade laser, and its output wavelength range covers 4.35µm to 4.65µm.

[0013] Furthermore, a regularization algorithm is used to train a spectral model for reconstructing the spectrum from the photocurrent response, taking the response rate matrix and the standard spectral data at a known wavelength as input and output, respectively. The specific steps are as follows: The response rate matrix R(λ, V) measured for each known wavelength λ is constructed as an input feature vector, and the standard spectrum at the wavelength λ is constructed as a target output vector to build a basic spectrum model; The basic spectral model is optimized and trained using a cost function including a regularization term; the cost function is intended to minimize the difference between the spectrum predicted by the model based on the input feature vector and the target output vector, while penalizing the complexity of the model; The internal parameters of the model are adjusted through an iterative optimization algorithm until the cost function converges, thereby obtaining a trained spectral model.

[0014] Furthermore, the spectrum of each pixel in the focal plane array is reconstructed. The specific steps are as follows: irradiating pixels of the mid-infrared focal plane array spectroscopic imager using a light source to be measured; Applying a preset bias voltage corresponding to the training spectral model to the PIN heterojunction, and measuring to obtain a response rate matrix to be measured; The response rate matrix to be measured is input into the trained spectral model to obtain the reconstructed gas spectrum.

[0015] Furthermore, the light source to be measured is specifically a light source to be measured obtained by passing mid-infrared broadband light generated by a blackbody light source through two independent gas chambers.

[0016] Furthermore, before performing spectral training, a step of performing non-uniformity calibration on the pixel array is also included. This step measures the response difference of each pixel under a uniform radiation source and calculates a correction coefficient to compensate for the inherent response inconsistency between each pixel.

[0017] Furthermore, the spectral imaging results are used to identify the gas components in the target scene and perform spatial positioning and concentration analysis on them by analyzing the spectral information that matches the characteristic absorption peaks of specific gas molecules.

[0018] The beneficial effects of the present invention are as follows: The present invention discloses a spectral reconstruction method for a mid-infrared focal plane array spectrometer. This method achieves spectral resolution by regulating the detector's own spectral response through voltage and combining it with a computational method. This method eliminates the need for expensive, bulky external filters or spectroscopic elements, and also eliminates the need for deep refrigeration equipment. This method greatly simplifies the system structure, reduces manufacturing costs and equipment volume, and paves the way for the realization of chip-scale spectrometers. The interface electric field control method proposed in the present invention significantly enhances the built-in electric field of the PIN heterojunction by introducing a CdS layer, significantly increasing the degree of voltage shift in its spectral response. A larger spectral shift provides richer response information for computational reconstruction, thereby enabling higher spectral resolution. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The present invention is a schematic diagram of a specific process of a spectral imaging method of a mid-infrared focal plane array spectral imager.

[0020] Figure 2 This is a voltage-controlled PIN heterojunction response rate-wavelength curve in Example 2 of the present invention.

[0021] Figure 3 This is a comparison chart of the reconstructed CO gas and N2O gas spectra in Example 2 of the present invention and the results of a commercial spectrometer.

[0022] Figure 4 This is a schematic diagram of the spectral imaging results of CO gas and N2O gas using an infrared focal plane array in Example 3 of the present invention. DETAILED DESCRIPTION

[0023] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] Example 1 A mid-infrared focal plane array spectroscopic imager includes a focal plane array, wherein the focal plane array includes a plurality of pixels, wherein each pixel includes a PIN heterojunction and a CdS layer arranged at the interface of the PIN heterojunction, and the PIN heterojunction includes a PbS hole transport layer, a Sn-doped PbSe photosensitive layer, and a ZnO electron transport layer in sequence.

[0025] In a specific embodiment, the CdS layer is formed at the interface of the PIN heterojunction of the mid-infrared focal plane array spectrometer by ion deposition, and is used to control the built-in electric field of the heterojunction by regulating the deposition amount.

[0026] In this embodiment, the deposition amount is regulated to enhance the built-in electric field of the heterojunction. The specific process is as follows: First, the electrodes at the upper and lower ends of the CdS-deposited PIN heterojunction PbS / PbSe / ZnO were connected to the two ends of the probe, which were respectively connected to the positive and negative terminals of a Keithley 4200 source meter instrument. The source meter was set to a sweep current curve and the voltage was set to +0.5 V and -0.5 V. Secondly, a mid-infrared QCL laser (Daylight Solution, MIRcat-QT, spot size: 2 mm) was irradiated onto the PIN heterojunction, and the different excitation wavelengths emitted by the mid-infrared QCL laser were adjusted from 4.35 to 4.65 μm with a step size of 0.01 μm. Then, a mid-infrared HgCdTe power meter was used to record the response power P at each interval wavelength, and a source meter was used to record the photocurrent in the dark state and under illumination at each interval wavelength, thereby obtaining the response rate curve (R(λ) = I(λ) / P(λ)) at +0.5 V for each interval wavelength. The magnitude of the interfacial electric field is obtained by calculating the integer ratio of the photocurrent values ​​of the PIN heterojunction at +0.5 V and −0.5 V; In this example, when the CdS deposition level changes from 15% to 30%, the interface electric field increases from 16 to 1603, and the corresponding voltage response wavelength shift changes from no shift to a 100 nm shift (4.45-4.55 µm). The CdS deposition level at the PIN heterojunction was determined using EDS mapping.

[0027] Example 2 like Figure 1 As shown, a spectral imaging method of a mid-infrared focal plane array spectral imager is characterized by comprising the following specific steps: Performing spectral training on each pixel in the focal plane array of the mid-infrared focal plane array spectral imager, and establishing an independent spectral model for each pixel, thereby forming a spectral model cube; Based on the spectral model cube, the spectrum of each pixel in the focal plane array is reconstructed; The reconstructed spectra of all pixels are combined to generate spectral imaging results.

[0028] In this embodiment, the spectral model training method of the mid-infrared focal plane array spectroscopic imager includes the following specific steps: Using a monochromatic light source to sequentially illuminate the pixels of the mid-infrared focal plane array spectroscopic imager at a plurality of predetermined wavelengths; At each of the predetermined wavelengths, a plurality of different bias voltages V are applied to the PIN heterostructure, and corresponding photocurrent values ​​I(λ, V) are measured synchronously; The photocurrent values ​​measured at all known wavelengths and bias voltages are correlated with the optical power P(λ) of the reference monochromatic light source at the corresponding wavelength to construct the responsivity matrix R(λ, V); In this embodiment, the voltage-controlled PIN heterojunction responsivity-wavelength curve is as follows: Figure 2 As shown; In this embodiment, mid-infrared spectroscopy training is based on voltage-controlled CdS deposition PIN heterojunction PbS / PbSe / ZnO response wavelength. The Keithley 4200 source meter is set to scan the IV curve with a voltage sweep range of -1 to 1 V and an interval voltage of 0.02 V. The responsivity matrix (R(λ, V) = I(λ, V) / P(λ)) of the PIN heterojunction is tested at different bias voltages and different wavelength intervals. In this embodiment, the standard spectrum (P(λ)) of the QCL laser at each wavelength interval (4.35 to 4.65 μm, with an interval step of 0.01 μm) is measured using Yokogawa's mid-infrared spectrometer AQ6377E. A regularization algorithm is used, with the response rate matrix and standard spectral data at known wavelengths as input and output, respectively, to train a spectral model for reconstructing the spectrum from the photocurrent response.

[0029] In a specific embodiment, the reference monochromatic light source is a quantum cascade laser, and its output wavelength range covers 4.35µm to 4.65µm.

[0030] In a specific embodiment, a regularization algorithm is used to train a spectral model for reconstructing a spectrum from a photocurrent response, using a response rate matrix and standard spectral data at a known wavelength as input and output, respectively. The specific steps are as follows: The response rate matrix R(λ, V) measured for each known wavelength λ is constructed as an input feature vector, and the standard spectrum at the wavelength λ is constructed as a target output vector to build a basic spectrum model; The basic spectral model is optimized and trained using a cost function including a regularization term; the cost function is intended to minimize the difference between the spectrum predicted by the model based on the input feature vector and the target output vector, while penalizing the complexity of the model; The internal parameters of the model are adjusted through an iterative optimization algorithm until the cost function converges, thereby obtaining a trained spectral model.

[0031] In this embodiment, when the CdS deposition amount changes from 15% to 30%, the mid-infrared spectrum resolution increases from 3.2 nm to 1.6 nm.

[0032] In this embodiment, the spectrum reconstruction method of the mid-infrared focal plane array spectroscopic imager is based on the spectrum model training method, and includes the following specific steps: irradiating pixels of the mid-infrared focal plane array spectroscopic imager using a light source to be measured; Applying a preset bias voltage corresponding to the training of the spectral model to the PIN heterojunction of the mid-infrared focal plane array spectroscopic imager, and measuring to obtain a response rate matrix to be measured; The response rate matrix to be measured is input into the trained spectral model to obtain the reconstructed gas spectrum.

[0033] In a specific embodiment, the light source to be measured is specifically a light source to be measured obtained by passing mid-infrared broadband light generated by a blackbody light source through two independent CO gas and N2O gas chambers.

[0034] In this embodiment, Figure 3 As shown in the figure, mid-infrared broadband light generated by a blackbody light source was tested using Yokogawa's AQ6377E mid-infrared spectrometer. The spectrum was compared using two separate gas chambers (CO and N₂O). The study found that the gas reconstructed spectrum of the PIN heterojunction exhibited 95% spectral overlap with that of the Yokogawa AQ6377E mid-infrared spectrometer.

[0035] Example 3 A spectral imaging method for a mid-infrared focal plane array spectral imager comprises the following specific steps: Performing spectral training on each pixel in the focal plane array of the mid-infrared focal plane array spectral imager, and establishing an independent spectral model for each pixel, thereby forming a spectral model cube; In this embodiment, the readout circuit of the focal plane detector is used to simultaneously read out the photocurrent of 10*10 array pixels at different bias voltages. The voltage setting range is 0.1~0.7V, and the interval voltage is 0.02V. The spectral model cube data (R xy (λ, V) = I xy (λ, V) / P(λ)); the response cube of the focal plane spectrometer and the standard spectrum (P(λ)) of the QCL laser tested by Yokogawa's mid-infrared spectrometer AQ6377E at each wavelength interval (4.35 ~ 4.65 µm, with an interval step of 0.01 µm) are used as the input and output of the regularization algorithm for spectral model cube training; Based on the spectral model cube, the spectrum of each pixel in the focal plane array is reconstructed; In this embodiment, mid-infrared broadband light generated by a blackbody light source is incident on a PIN heterojunction through two independent separated gas chambers (CO gas and N2O gas). The responsivity of the PIN heterojunction is input into the trained spectral model cube to obtain reconstructed gas focal plane spectral imaging. The reconstructed spectra of all pixels are combined to finally generate a spectral imaging result containing spatial dimension information and spectral dimension information. Figure 4 As shown, the spectral imaging results are consistent with the positions of the characteristic absorption peaks of CO gas and N2O gas.

[0036] In a specific embodiment, before performing spectral training on each pixel in the pixel array, a step of performing non-uniformity calibration on the pixel array is also included. This step measures the response difference of each pixel under a uniform radiation source and calculates a correction coefficient to compensate for the inherent response inconsistency between the pixels.

[0037] In a specific embodiment, the spectral imaging results are used to identify the gas components in the target scene and perform spatial positioning and concentration analysis on them by analyzing the spectral information that matches the characteristic absorption peaks of specific gas molecules.

[0038] Obviously, the above embodiments of the present invention are merely examples for the purpose of illustrating the present invention, and are not intended to limit the embodiments of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the claims of the present invention.

Claims

1. A mid-infrared focal plane array spectral imager, characterized in that: The invention comprises a focal plane array, wherein the focal plane array comprises a plurality of pixels, wherein each pixel comprises a PIN heterojunction and a CdS layer arranged at the interface of the PIN heterojunction, and the PIN heterojunction comprises a PbS hole transport layer, a Sn-doped PbSe photosensitive layer, and a ZnO electron transport layer in sequence.

2. The mid-infrared focal plane array spectroscopic imager according to claim 1, wherein: The CdS layer is formed at the interface of the PIN heterojunction of the mid-infrared focal plane array spectroscopic imager by an ion deposition method, and is used to regulate the built-in electric field of the heterojunction by regulating the deposition amount.

3. A spectral imaging method for a mid-infrared focal plane array spectral imager, characterized by: The specific steps include: Performing spectral training on each pixel in the focal plane array of the mid-infrared focal plane array spectral imager, and establishing an independent spectral model for each pixel, thereby forming a spectral model cube; Based on the spectral model cube, the spectrum of each pixel in the focal plane array is reconstructed; The reconstructed spectra of all pixels are combined to generate spectral imaging results.

4. The spectral imaging method according to claim 3, wherein: Performing spectral training on each pixel in the pixel array and establishing an independent spectral model for each pixel includes the following specific steps: Using a monochromatic light source to sequentially illuminate the pixels of the mid-infrared focal plane array spectroscopic imager at a plurality of predetermined wavelengths; At each of the predetermined wavelengths, a plurality of different bias voltages V are applied to the PIN heterostructure, and corresponding photocurrent values ​​I(λ, V) are measured synchronously; The photocurrent values ​​measured at all known wavelengths and bias voltages are correlated with the optical power P(λ) of the reference monochromatic light source at the corresponding wavelength to construct the responsivity matrix R(λ, V); A regularization algorithm is used, and the response rate matrix and standard spectral data at known wavelengths are used as input and output respectively to train the spectral model.

5. The spectral imaging method according to claim 4, characterized in that: The reference monochromatic light source is a quantum cascade laser, and its output wavelength range covers 4.35µm to 4.65µm.

6. The spectral imaging method according to claim 4, characterized in that: A regularization algorithm is used, with the response rate matrix and standard spectral data at known wavelengths as input and output, respectively, to train a spectral model for reconstructing the spectrum from the photocurrent response. The specific steps are as follows: The response rate matrix R(λ, V) measured for each known wavelength λ is constructed as an input feature vector, and the standard spectrum at the wavelength λ is constructed as a target output vector to build a basic spectrum model; The basic spectral model is optimized and trained using a cost function including a regularization term; the cost function is intended to minimize the difference between the spectrum predicted by the model based on the input feature vector and the target output vector, while penalizing the complexity of the model; The internal parameters of the model are adjusted through an iterative optimization algorithm until the cost function converges, thereby obtaining a trained spectral model.

7. The spectral imaging method according to claim 3, wherein: Spectral reconstruction is performed on each pixel in the focal plane array. The specific steps are as follows: irradiating pixels of the mid-infrared focal plane array spectroscopic imager using a light source to be measured; Applying a preset bias voltage corresponding to the training spectral model to the PIN heterojunction, and measuring to obtain a response rate matrix to be measured; The response rate matrix to be measured is input into the trained spectral model to obtain the reconstructed gas spectrum.

8. The spectral imaging method according to claim 7, characterized in that: The light source to be measured is specifically a light source to be measured obtained by passing mid-infrared broadband light generated by a blackbody light source through two independent gas chambers.

9. The spectral imaging method according to claim 3, wherein: Before performing spectral training, a step of performing non-uniformity calibration on the pixel array is also included. This step measures the response difference of each pixel under a uniform radiation source and calculates a correction coefficient to compensate for the inherent response inconsistency between each pixel.

10. The spectral imaging method according to claim 3, characterized in that: The spectral imaging results are used to identify the gas components in the target scene and perform spatial positioning and concentration analysis on them by analyzing the spectral information that coincides with the characteristic absorption peaks of specific gas molecules.

Images