Near-infrared silicon-based micro-computing spectrometer based on time and space dual coding

Abstract

The invention discloses a near-infrared silicon-based micro-computing spectrometer based on time and space dual coding, which combines space coding and time coding, and introduces voltage, incident light angle and shading distance to code a device. Detection of unknown spectrums is realized by combining spectral response of the modulator in different coding states with a spectrum reconstruction algorithm. The micro spectrometer has extremely high spectral resolution, can be kept stable for a long time, and has the characteristics of low preparation cost, simple process, high repeatability and the like.

Description

Technical Field

[0001] The present invention relates to the field of micro-spectrometers, and in particular to a near-infrared silicon-based micro-computing spectrometer based on time and space dual encoding. Background Art

[0002] Spectroscopic analysis, a core tool for studying the interaction between matter and electromagnetic radiation, can precisely determine a substance's chemical composition, molecular structure, and physical properties by analyzing the spectral signatures emitted, absorbed, or scattered by the substance. Its fundamental principle stems from the unique characteristic spectra generated by atomic / molecular energy level transitions, forming a "fingerprint" for material identification. Spectrometers, instruments capable of measuring and analyzing spectral information, have been widely used in key areas such as environmental monitoring, biomedical diagnostics, food safety, materials science, and space exploration. However, traditional spectrometers are bulky, complex to operate, and rely on laboratory environments, making them difficult to use outdoors or in the field. Miniaturization of spectrometers holds significant application value in areas such as environmental monitoring, industrial online testing, and emergency response. Existing approaches to miniaturizing spectrometers often involve designing various patterns through micro-nanofabrication to replace the bulky dispersive optical components in the spectrometer. These methods include planar waveguides, photonic crystals, and metasurfaces. However, due to the reduced size, the resolution of miniaturized spectrometers achieved with these approaches is significantly reduced.

[0003] In recent years, with the rapid development of artificial intelligence algorithms, micro-computational spectrometers based on reconstruction algorithms have emerged. Unlike traditional spectrometers, computational spectrometers do not rely on spectroscopic elements such as gratings and narrowband filters. Instead, they adjust physical parameters such as material composition, micro-nanostructure, and electron energy levels to obtain a set of spectral curves with nonlinear correlations within a specific wavelength range. A linear system of equations is constructed using the current values ​​measured under the unknown spectrum and the device's spectral curve. This linear system is solved using an algorithm to reconstruct the intensity-wavelength relationship of the unknown spectrum. By eliminating the dispersive element, the resolution of the reconstruction spectrometer is no longer limited by the optical path length, further reducing the size of the spectrometer.

[0004] Based on the encoding method, micro-computational spectrometers can be roughly divided into two categories: spatially modulated spectral encoding and temporally modulated spectral encoding. Spatially modulated spectral encoding is characterized by integrating detectors with different spectral responses to achieve spectral encoding in a compact space. Yang Zongyin et al. (Single-nanowire spectrometers[J]. Science. 2019, 365(6457):1017-+) from the University of Cambridge developed a micro-spectrometer based on a single nanowire. The alloy composition of the nanowire varies along its length, causing the spectral response of the nanowire to also vary with its length. This device feature, combined with a reconstruction algorithm, enables the detection of both monochromatic and polychromatic light. Similar work also includes the use of printed perovskites with gradient bandgap widths, in-situ perovskite modulation, the Stark effect in black phosphorus, and superconducting nanowires with tunable quantum efficiency. Although spatially modulated spectral encoding micro-spectrometers can effectively reduce the size of the spectrometer, their resolution is still affected by the number of detectors, resulting in extremely high space and manufacturing costs for high-resolution spectrometers. Temporal modulation spectral encoding, on the other hand, utilizes conditions such as bias voltage to dynamically change the detector's spectral response function, enabling the detection of spectral information using only a single detector. Since the number of codes in the device no longer depends on the number of detectors, this type of spectrometer can achieve high spectral resolution with a very small size. However, most current temporal modulation-based micro-spectrometers use two-dimensional materials and perovskites, which inevitably leads to problems such as poor spectrometer stability and complex manufacturing processes.

[0005] Wang Li et al. from Hefei University of Technology (Wang L, Chen BH, Fang CY, He J, Wu CY, Zhang X, et al. Wavelength-Tunable Multispectral Photodetector With Both Ultraviolet and Near-Infrared Narrowband Detection Capability [J]. IEEE Transactions on Electron Devices. 2022, 69(6): 3258-3261.) proposed a silicon-based photodetector based on an Au / n-Si / Au back-to-back double Schottky junction structure. The spectral response of the device can be changed by voltage. The detector consists of only an n-type silicon wafer and two upper and lower Schottky electrodes. The Schottky electrode on the upper surface mainly collects short-wavelength light, while the electrode on the lower surface collects long-wavelength light, thereby achieving different spectral response characteristics of the two Schottky electrodes. This work provides a new idea for realizing spectrally tunable photodetectors using a single material. However, this structure requires double-sided lithography and special silicon wafer preparation, which is relatively expensive. In addition, it can only achieve time coding and has poor spectral reconstruction performance. Summary of the Invention

[0006] In view of the above-mentioned deficiencies in the prior art, the present invention provides a near-infrared silicon-based micro-computing spectrometer based on time and space dual encoding, aiming to reduce the size of the spectrometer while significantly improving the resolution of the spectrometer.

[0007] The present invention solves the technical problem by adopting the following technical solutions:

[0008] The near-infrared silicon-based micro-computing spectrometer based on time and space dual encoding provided by the present invention is characterized in that: the near-infrared silicon-based micro-computing spectrometer uses a silicon wafer as a substrate, a doped region is provided in the center of the silicon wafer to form a photogenerated carrier recombination gradient field, and p pairs of Schottky junctions are provided on the surface of the silicon wafer, each pair of Schottky junctions including two Schottky electrodes symmetrically arranged on the left and right sides of the doped region; a portion of the area between each pair of Schottky electrodes, including the doped region, is used as an illuminated area, and a non-illuminated area on the substrate surface is shielded from light;

[0009] The distance between the Schottky electrode and the illuminated area is defined as the electrode shielding distance, so that the shielding distances of the p Schottky electrodes located on the left side of the doped area are equal, and the shielding distances of the p Schottky electrodes located on the right side of the doped area vary in a gradient, forming p shielding distance variables. The spectral response characteristics of the device are controlled by changing the shielding distances.

[0010] The incident light is defined as 0° when it is perpendicular to the substrate surface, and the deflection toward the left side of the doped region is defined as positive, forming q incident light angle variables. The spectral response characteristics of the device can be controlled by changing the incident light angle.

[0011] Define the Schottky electrode on the left side of the doped region as the cathode and the Schottky electrode on the right side of the doped region as the anode. Apply an external bias voltage to each pair of Schottky electrodes and adjust the magnitude of the applied bias voltage to form r voltage variables. By changing the voltage, the spectral response characteristics of the device are controlled.

[0012] P shading distance variables are used for spatial encoding, and q incident light angle variables and r voltage variables are used for temporal encoding. The working states of the device under the three variables are combined to obtain p×q×r coding states. The detection of unknown spectra is achieved by modulating the spectral response of the device in different coding states and combining the spectral reconstruction algorithm.

[0013] Furthermore, the silicon wafer is an n-type lightly doped silicon wafer, and the doping element in the doped region is Au, and doping can be achieved by high-temperature diffusion or ion implantation. The energy level of deep-level impurities is often located near the center of the semiconductor band gap, which can effectively modulate the spectral response. It introduces energy levels that can capture electrons or holes, thereby promoting the recombination process between carriers and impurities. Gold, as a typical deep-level impurity, can accelerate the recombination rate of non-equilibrium carriers. Gold doping can change the collection of photogenerated carriers by the electrode, thereby affecting the spectral response. At the same time, for the same gold doping concentration, the minority carrier lifetime in p-type silicon is 1.9 times that in n-type silicon. The recombination effect of gold doping has a greater impact on the minority carrier lifetime in n-type silicon. Therefore, the present invention uses an n-type lightly doped silicon wafer as a substrate and Au as a doping element.

[0014] Furthermore, in a pair of Schottky junctions, the shading distance of the Schottky electrode located to the left of the doped region (hereinafter referred to as the left electrode) is greater than the shading distance of the Schottky electrode located to the right of the doped region (hereinafter referred to as the right electrode). The doped region is located at the center of the silicon wafer to form a composite gradient field, resulting in different recombination velocities at different horizontal positions. The distances between each electrode and the doped region are equal. A shading process is then used to vary the distances between the illuminated region and the two Schottky electrodes, resulting in different spectral response characteristics for the two Schottky junctions. Changing the illumination angle changes the distribution of photogenerated carriers in the silicon, thereby affecting the collection of photogenerated carriers by the Schottky electrodes.

[0015] In addition, since the two Schottky electrodes have different spectral response characteristics, their spectral response characteristics can be controlled by voltage. Due to the large resistance of the depletion region, an external bias will be applied to the depletion region. By applying a negative bias to the right electrode, its built-in electric field can be enhanced, making its depletion region wider, while the depletion region of the left electrode narrower, and vice versa. The increase in the depletion region can enhance the ability of the junction to capture carriers, so the carrier collection ability of the two electrodes can be changed by adjusting the bias. Since the spectral response of the device is jointly determined by the currents flowing in opposite directions collected by the two electrodes, and the distances of the illumination area from the two electrodes are different, resulting in the two Schottky electrodes having different spectral response characteristics, the spectral response characteristics of the device can be changed by voltage.

[0016] Furthermore, after the working state encoding is completed, the spectral response of the device in different encoding states is tested using a series of known monochromatic lights to obtain a spectral response function determined by the incident wavelength and the device encoding, and the spectral response function is converted into a spectral response matrix to complete the learning step. The unknown spectrum is reconstructed by testing the photocurrent of the device in different encoding states under the unknown spectrum. The reconstruction algorithm is at least one of any algorithms for solving ill-conditioned linear equations, such as an adaptive Tikhonov regularization algorithm, a regression algorithm, a deep neural network, a sparse algorithm, etc.

[0017] Compared with the prior art, the beneficial effects of the present invention are embodied in:

[0018] 1. The deep-level doped back-to-back double Schottky junction photodetection unit of the present invention can regulate its spectral response by voltage, incident light angle and shading distance. In order to increase the richness of the spectral response curve during the encoding process, the present invention combines spatial encoding and temporal encoding, and introduces voltage, incident light angle and shading distance to encode the device to improve the performance of the spectrometer. Subsequently, a silicon-based micro-spectrometer was prepared and the performance of the spectrometer was studied. The micro-spectrometer successfully identified monochromatic light and polychromatic light: the average detection error of the spectral peak of monochromatic light was 0.29nm, and the minimum could reach 0.01nm. At the same time, the detection resolution of polychromatic light reached 5nm, covering the spectral range of 800nm ​​to 1200nm. In addition, thanks to the good linear dynamic range and stability of the device, the spectrometer can perfectly identify monochromatic light sources of different powers and maintain stable performance for a long time. Afterwards, the present invention also explored the impact of different numbers of codes on the performance of the spectrometer. The results showed that increasing the number of spectral codes can effectively improve the spectral resolution of the micro-spectrometer.

[0019] 2. The near-infrared silicon-based micro-computing spectrometer of the present invention can realize the spectrometer function without the need for a detector array and a filter array. The modulation method is easy to implement, the device structure is simple, and there is no limit on the device size, so an extremely small micro-spectrometer can be realized.

[0020] 3. The near-infrared silicon-based micro-computing spectrometer of the present invention has the characteristics of low preparation cost, simple process and easy implementation.

[0021] 4. Compared with other micro-spectrometers made of nano, two-dimensional and perovskite materials, the silicon-based micro-computing spectrometer of the present invention has better stability, a wide spectral response range and high resolution. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 (a) is a schematic structural diagram of the back-to-back double Schottky junction photodetection unit, and (b) is the IV curve of the back-to-back double Schottky junction photodetection unit under different incident light angles.

[0023] Figure 2 Spectral response curve of the back-to-back double Schottky junction photodetection unit under different incident light angles.

[0024] Figure 3 Spectral response curves of the back-to-back double Schottky junction photodetection unit at different voltages.

[0025] Figure 4 Schematic diagram of the structure of a single Schottky structure photodetector when the distance between the illumination area and the Schottky electrode is (a) 0 and (b) d, as well as the (c) spectral response curve and (d) normalized curve of the device when the Schottky junction is at different distances from the illumination area.

[0026] Figure 5 This is the spectral response curve of the back-to-back double Schottky junction photodetection unit at -0.1V when the illumination area is at different distances from the Schottky on the right.

[0027] Figure 6 (a) is the IV curve of the back-to-back double Schottky junction photodetection unit under 1100nm light of different powers, and (b) is the curve of the photocurrent changing with the light power when the voltage is -0.1V, 0V and 0.1V. The inset in the figure is a local enlarged view.

[0028] Figure 7 (a) is the -3dB bandwidth of devices with different shading distances under 1100nm light. Figure 7 (b), (c), and (d) are the response times when the shading distance is 0 μm, 100 μm, and 200 μm at the cutoff frequency, respectively.

[0029] Figure 8 Schematic diagram of the structure of a silicon-based microcomputing spectrometer.

[0030] Figure 9The operation process of the near-infrared silicon-based micro-spectrometer: (a) encoding process; (b) learning process; (c) testing process; (d) reconstruction process.

[0031] Figure 10 Color scale diagram of the spectral response matrix of the silicon-based microcomputing spectrometer.

[0032] Figure 11 (a) shows the reconstruction results of a series of quasi-monochromatic spectra in the 800-1200 nm band by the spectrometer. The solid line is the reconstruction result of the silicon-based spectrometer, and the dotted line is the control result of the commercial spectrometer test. Figure 11 (b) is the reconstruction result of the spectrometer for a double-peak spectrum with two peaks separated by 5 nm. The inset in the upper left corner is the result of local magnification.

[0033] Figure 12 Comparison between the reconstruction results of (a) near-infrared LED light source and (b) deuterium lamp by silicon-based micro-spectrometer and the results measured by commercial spectrometer.

[0034] Figure 13 (a) Comparison between the reconstruction results of a narrow-band spectrum with a spectral range of 900-1000 nm and a step size of 5 nm by a silicon-based micro-computing spectrometer and the test results of a commercial spectrometer. Figure 13 (b) The peak wavelength difference between the reconstructed spectrum and the reference spectrum (top figure) and the relationship between the peak wavelength of the reconstructed spectrum and the measured spectrum and the input wavelength (bottom figure).

[0035] Figure 14 (a) The spectrometer reconstructs narrowband light at 1100 nm, with light intensities ranging from 43.3 to 3000 μW / cm 2 ; Figure 14 (b) As the light intensity increases from 43.3 to 3000 μW / cm 2 , the changes in the peak position and peak intensity of the reconstructed narrowband spectrum (at 1100 nm); Figure 14 (c) Comparison of dark current changes at different voltages after one month; Figure 14 (d) Comparison of the reconstruction results of the spectrometer for 1100nm monochromatic light with the reconstruction results after one month without recalibration.

[0036] Figure 15 Comparison of the reconstruction results of 900nm monochromatic light using 99 and 33 coding units with the test results of a commercial spectrometer.

[0037] Figure 16Results of reconstructing a bimodal spectrum with two peaks separated by (a) 20 nm, (c) 10 nm, and (e) 5 nm using 33 coding units, and reconstructing a bimodal spectrum with two peaks separated by (b) 20 nm, (d) 10 nm, and (f) 5 nm using 99 coding units. DETAILED DESCRIPTION

[0038] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process. However, the protection scope of the present invention is not limited to the following embodiment.

[0039] 1. Modulation effect of different parameters on device spectral response characteristics

[0040] To verify the modulation effect of different parameters on the device's spectral response characteristics, a back-to-back dual Schottky junction photodetector unit (i.e., a device containing only one pair of Schottky junctions) was prepared according to the following steps:

[0041] (1) First, a 4-inch n-type (100) lightly doped single-polished silicon wafer (thickness 500 μm, resistivity 1-10 Ω·cm) was cut into 1 cm × 1 cm substrates. The cut silicon substrates were then immersed in a prepared BOE solution (hydrofluoric acid (mL): ammonium fluoride (g): deionized water (ml) = 3:6:10) for 2 minutes to remove the natural oxide layer on the substrate surface. The substrates were then cleaned in an ultrasonic cleaner with acetone, alcohol, and deionized water for 15 minutes each, and finally dried with nitrogen for later use.

[0042] (2) A 1 mm × 5 mm matrix window was defined in the center of the polished surface of the silicon substrate using a metal mask, and a 2 nm thick gold film was sputtered using an ion sputtering apparatus.

[0043] (3) The sputtered substrate was transferred to an annealing furnace and annealed at 500°C for 30 minutes under a nitrogen atmosphere to allow the gold atoms to diffuse into the silicon. After doping, the substrate was immersed in a solution of deionized water (mL): potassium iodide (mg): iodine (mg) = 3:30:10 for 10 minutes to remove any gold particles remaining on the surface.

[0044] (4) A 1.5 mm × 5 mm window was defined on the substrate surface at 300 μm from the left and right sides of the doped region using a metal mask, and then a 50 nm gold electrode was evaporated using an electron beam evaporation system for Schottky contact.

[0045] (5) Finally, the prepared device was fixed on a custom PCB board, and the electrodes were led out using conductive silver paste and silver wire to facilitate subsequent testing. Then, insulating black tape was used to shield the area outside the illuminated area.

[0046] 1. Regulation of the spectral response characteristics of the device by the incident light angle

[0047] When the distance between the illuminated area and the left Schottky electrode (junction 2) and the right Schottky electrode (junction 1) (i.e., the light shielding distance) is 300 μm and 0 μm respectively (e.g., Figure 1 (a) shows that the device has an optical power of 1mW / cm 2 The IV curves of the device under different angles of 1100nm light. The results show that as the incident light angle increases from 0° to 60° (the light angle changes towards the junction 2 as positive), the bias voltage at the transition point where the current in the device IV curve changes from negative to positive gradually decreases from 0.06V to -0.02V (as shown in Figure 2). Figure 1 (b)). This is because the device's spectral response is determined by the currents flowing in opposite directions collected by junctions 1 and 2. Light with a wavelength of 1100nm can penetrate deep into silicon, and the light absorption rate at different depths in silicon is relatively uniform. As the illumination angle increases, the photogenerated carriers generated by the 1100nm light move closer to junction 2, increasing the number of carriers collected by junction 2 and the generated photocurrent. However, junction 1's ability to collect photogenerated carriers generated by 1100nm light weakens, causing the current in junction 2 to gradually dominate. When the illumination angle is 45°, the transition point where the current changes from negative to positive is 0V, which also means that the current collected by the two electrodes is balanced. This phenomenon shows that by changing the illumination angle, the distribution of photogenerated carriers generated by light of different wavelengths in silicon can be changed, thereby changing the number of carriers collected by the two Schottky electrodes.

[0048] Further testing of the spectral response curve of the device at different incident light angles (without applying additional voltage) shows the following results: Figure 2 As shown in the figure, the device's spectral response curve changes significantly with changes in illumination angle. As the illumination angle increases from 0° to 60°, the device's spectral response in the 800-1200nm band gradually decreases, with the decrease in spectral response occurring more significantly for longer wavelengths (1000-1200nm). This further demonstrates that changes in illumination angle can alter the distribution of photogenerated carriers in silicon, thereby adjusting the device's spectral response.

[0049] 2. Voltage regulation of device spectral response characteristics

[0050] Further study the effect of voltage on the spectral response of the device, Figure 3The device's spectral response curves at different voltages when illuminated vertically are shown. As the applied voltage increases from -0.1V to 0.1V, the response to short-wavelength light gradually weakens, while the response to long-wavelength light shifts from positive to negative. This is because junction 2 is farther from the illuminated region, and short-wavelength light is captured by the recombination center before being collected by the electrode, exhibiting distinct narrowband absorption characteristics. Junction 1, on the other hand, is closer to the illuminated region and can simultaneously collect photogenerated carriers of different wavelengths, resulting in a wide spectral response. Because the two electrodes exhibit different spectral response characteristics, the intensity of their responses can be manipulated by an external bias, thereby affecting the spectral response of the entire device.

[0051] 3. Control of shading distance on device spectral response characteristics

[0052] In order to further study the influence of different shading distances on the spectral response characteristics of the Schottky junction, a group of single Schottky structure photodetectors were designed based on the original structure, such as Figure 4 As shown in (a) and (b), gold is doped in the center of the n-type silicon wafer. The doping conditions are the same as the original structure. The size of the doped area is 1mm×5mm, and the distance from the left and right electrodes is 0.5mm. The left ohmic electrode (titanium) is shielded from light. The spectral response of the device is tested by changing the distance d between the Schottky electrode (gold) and the illuminated area. The test results are shown in Figure 1. Figure 4 As shown in (c), it can be found that as the distance between the Schottky electrode and the illuminated area gradually increases from 0μm to 500μm, the device's response to shorter wavelength light is gradually suppressed, and the device's peak spectral response size decreases from 57.4mA / W to 16.5mA / W. Through the normalized spectral curve of the device, it can be found that when the distance between the Schottky electrode and the illuminated area gradually increases from 0μm to 500μm, the device's spectral response peak gradually redshifts from 1085nm to 1105nm, and the half-height width of the spectral response decreases from 244nm to 101nm. This is because the photogenerated carriers generated by short-wavelength light are greatly affected by the surface recombination center. As the shielding distance increases, they are recombined before being collected by the Schottky electrode, so the device exhibits a selective response to long wavelengths.

[0053] Based on the above analysis, the spectral response curve of the back-to-back double Schottky junction photodetection unit at -0.1V under different shading conditions of the right Schottky electrode was further tested. Figure 5 As the shielding distance increases, the response to short-wavelength light gradually weakens and the response peak red-shifts, which is consistent with the above conclusion.

[0054] 4. Characterization and analysis of device optoelectronic performance

[0055] The photoelectric performance of the back-to-back double Schottky junction photodetector unit under different shading conditions is characterized and analyzed. First, the IV curves of the device under different power illumination with a wavelength of 1100nm are tested when the distance between the illumination area and the right Schottky electrode is 0μm. Figure 6 (a) As shown in the figure, it can be found that when the voltage range is from -0.1V to 0.1V, the device photocurrent increases with the increase of light power. When the voltage is 0V, as the light power increases from 70.7μW / cm 2 Increased to 6220μW / cm 2 , the photocurrent of the device increased from 0.081μA to 3.68μA. Figure 6 (b) The curves of the photocurrent changing with the optical power at -0.1V, 0V, and 0.1V are extracted. It can be found that the photocurrent of the device shows a good linear relationship with the optical power at different voltages. The linear dynamic range (LDR) is usually used to evaluate the range of light intensities within which a photodetector can maintain a linear response. A larger LDR value means that the detector can operate accurately over a wider range of light intensities and will not distort or saturate when processing strong and weak light signals. Its calculation formula is as follows:

[0056]

[0057] Among them I light Indicates the maximum photocurrent of the device within the linear range; I dark Represents the dark current of the device. The formula shows that when the applied voltage is 0V, the LDR of the device is about 82dB, showing excellent linear working ability.

[0058] In addition, the present invention also studies the high-frequency characteristics of the device. -3dB bandwidth is usually used to evaluate the high-frequency characteristics of the device. A higher -3dB bandwidth means that the photodetector can process higher frequency signals. In this experiment, the present invention uses a function generator to adjust the frequency of the 1100nm wavelength light emitted by the Anyang laser, and uses an oscilloscope to measure the response amplitude of the device under different frequency light. -3dB bandwidth refers to the frequency corresponding to when the device response amplitude drops to 0.707 times the response at 1Hz frequency. The test results are as follows: Figure 7 (a) As shown. When the shading distance is 0, 100 and 200 μm, the -3dB bandwidth of the device is 5.3kHz, 5kHz and 3.8kHz respectively. As the shading distance increases, the -3dB bandwidth of the device gradually decreases. This may be due to the fact that the distance between the photogenerated carriers and the electrode increases with the increase in the shading distance. Response time refers to the device's ability to respond to rapidly changing light signals and is usually used as one of the important parameters for evaluating the performance of photodetectors. Response time is generally measured by the rise time (T r ) and fall time (Tf ) are characterized by their rise time. Rise time is the time it takes for a signal to rise from 10% to 90% of its maximum value, while fall time is the time it takes for a signal to fall from 90% to 10% of its maximum value. Shorter rise and fall times indicate that the device can respond more quickly to changes in the optical signal, reflecting its sensitivity to high-frequency signals. Figure 7 Figures (b), (c), and (d) show the rise and fall times of the three devices at different shading distances at the cutoff frequency. Under 1100nm illumination, the rise and fall times of the device are 39.5μs and 67.5μs, respectively, when the shading distance is 0μm; 22μs and 87μs, respectively, when the shading distance is 100μm; and 99.5μs and 110μs, respectively, when the shading distance is 200μm. As the distance between the illuminated area and the Schottky junction increases, the device's response speed decreases, but it still exhibits excellent fast response characteristics.

[0059] These test results demonstrate that the back-to-back dual Schottky junction photodetector unit can control the device's spectral response curve by varying the illumination angle, voltage, and shielding distance. The device's photocurrent exhibits a good linear relationship with optical power, and the device exhibits excellent fast response characteristics.

[0060] 2. Design of a near-infrared silicon-based microcomputing spectrometer based on temporal and spatial dual encoding

[0061] 1. Design and preparation of spectrometer

[0062] The silicon-based photoelectric unit based on the above-mentioned dual Schottky structure can regulate its spectral response through voltage, incident light angle and shading distance. In order to increase the richness of the spectral response curve during the encoding process, the present invention combines spatial encoding and temporal encoding and uses voltage, incident light angle and shading distance to encode the device to improve the performance of the spectrometer. Based on the above principle, the present invention designs a near-infrared silicon-based micro-computing spectrometer based on time and space dual encoding. The spectrometer only contains p pairs of Schottky junctions, and uses shading distance (taking 3 pairs of Schottky junctions as an example to form 3 shading distance variables) for spatial encoding. At the same time, it uses incident light angle (setting q = 3 incident light angle variables) and voltage (setting r = 11 voltage variables) for time encoding to obtain 99 spectral codes. The device structure is as follows: Figure 8 The device contains six gold electrodes, labeled J1-J6, and a doped region is set in the central illumination area of ​​the device. The doped region size is 800×200μm 2, the distance between the doped region and the left and right electrodes is 300μm. The PMMA and metal mask on the top of the device are used for insulation and shading respectively, forming three pairs of electrodes with different shading effects, which are defined as J1-J2, J3-J4, and J5-J6. Among them, the distance between the illuminated area and J2, J4, and J6 is 300μm, while the distance from J1, J3, and J5 is 0μm, 100μm, and 200μm, respectively. Shading can make the distance between the illuminated area and the Schottky electrodes on both sides different, thereby affecting the collection of photogenerated carriers by the electrodes and making the two Schottky electrodes have different spectral response characteristics. In this way, the spectral response of the device can be controlled by voltage.

[0063] The detailed preparation process of the silicon-based micro-computational spectrometer is as follows:

[0064] (1) First, a 4-inch n-type (100) lightly doped single-polished silicon wafer (thickness 500 μm, resistivity 1-10 Ω·cm) was cut into substrates of 1.2 cm × 1.2 cm. The cut silicon substrate was then immersed in a prepared BOE solution (hydrofluoric acid (ml): ammonium fluoride (g): deionized water (ml) = 3:6:10) for 2 minutes to remove the natural oxide layer on the substrate surface. The substrate was then cleaned in an ultrasonic cleaner with acetone, alcohol, and deionized water for 15 minutes respectively, and finally dried with nitrogen gas for later use.

[0065] (2) Use a metal mask to define a 200×800μm area in the center of the polished surface of the silicon substrate. 2 A 2 nm thick gold film was sputtered on the matrix window using an ion sputtering instrument.

[0066] (3) The sputtered substrate was transferred to an annealing furnace and annealed at 500°C for 30 minutes under a nitrogen atmosphere to allow the gold atoms to diffuse into the silicon. After doping, the substrate was immersed in a solution of deionized water (mL): potassium iodide (mg): iodine (mg) = 3:30:10 for 10 minutes to remove any gold particles remaining on the surface.

[0067] (4) A metal mask is used to define the position and size of six electrodes on the silicon surface, and then an electron beam evaporation system is used to evaporate 50nm of gold electrodes for Schottky contacts.

[0068] (5) Finally, the device was shielded from light using a metal mask that had been immersed in PMMA solution for 5 minutes.

[0069] 2. Working principle of spectrometer

[0070] Figure 9The workflow of the silicon-based micro-computation spectrometer is presented, which includes four steps: (1) encoding, (2) learning, (3) testing, and (4) reconstruction. The silicon-based photodetector element based on the Au / n-Si / Au dual Schottky structure can control its spectral response by voltage, incident light angle, and shielding distance. In order to facilitate subsequent spectral reconstruction, the different working states of the device are combined and encoded in sequence, such as Figure 9 After the working state coding is completed, a series of known monochromatic light is used to test the spectral response of the device in different coding states, and the spectral response function R(S,λ) determined by the incident wavelength λ and the device code S is obtained, and it is converted into a spectral response matrix, as shown in Figure 2. Figure 9 After completing the learning step, the silicon-based micro-spectrometer of the present invention can detect unknown spectra. Figure 9 (c) shows the test process, testing the device's photocurrent under unknown illumination conditions and different encodings, and obtaining the photocurrent vector I S In fact, the current obtained during the test is obtained by integrating the spectrum function F(λ) of the spectrum and the spectral response of the device within the wavelength range of the spectrum function. By converting the integral equation of the photocurrent into a linear equation system and solving it using the constrained least squares method, the unknown incident spectrum F(λ) ( Figure 9 (d)).

[0071] 3. Spectrometer performance characterization and analysis:

[0072] In this work, the silicon-based spectrometer operates in the 800-1200 nm wavelength range, determined by the device's spectral tunability within this range. The spectrum is sampled with a 5 nm sampling pitch, 82 wavelengths, and 99 codes, including three angle variables, three shielding distance variables, and 11 voltage variables. Figure 10 A color scale diagram of the spectral response matrix R(S,λ) is shown. It can be seen from the figure that the size of the device's spectral response is determined by both the wavelength and the device coding state, and changes significantly with the changes in wavelength and device coding state. This feature is also key information for designing micro-computing spectrometers.

[0073] Next, we evaluated the performance of the spectrometer. First, we used a silicon-based micro-computational spectrometer to measure a series of quasi-monochromatic spectra in the 800-1200nm range and compared them with reference spectra measured by a commercial spectrometer. The results showed that the spectrometer's measurement results in this band were highly consistent with the reference spectra, verifying its detection capability over a wide spectral range ( Figure 11 (a)). In addition, the present invention measured a complex double-peak spectrum to evaluate the resolution of the spectrometer. The results showed that the spectrometer can distinguish peaks at 1100nm and 1105nm, which are 5nm apart ( Figure 11(b) It is worth noting that during the learning process, the step size of the spectral response matrix is ​​5 nm. Therefore, the minimum learning step size limits the spectrometer's resolution to no less than 5 nm. By reducing the learning step size, the spectrometer's resolution can be further improved.

[0074] In order to verify the ability of silicon-based spectrometer to identify complex color spectra, the present invention uses silicon-based micro-spectrometer to test near-infrared LED light source and deuterium lamp light source respectively and compares and analyzes them with the reference spectrum measured by commercial near-infrared spectrometer (NIR25s). Figure 12 (a) and Figure 12 As shown in (b), the reconstructed spectrum is highly consistent with the characteristic peaks of the measured reference spectrum. The mean square error (MSE) of the near-infrared LED and deuterium lamp light sources identified by the silicon-based computational spectrometer is 0.0027 and 0.0038, respectively. These results demonstrate that the silicon-based computational spectrometer can not only effectively identify monochromatic spectra, but also complex, broad-spectrum light sources with minimal error.

[0075] In addition, the peak wavelength recognition accuracy is also an important indicator for judging the performance of the spectrometer. In order to verify the wavelength recognition accuracy of the silicon-based micro-spectrometer for monochromatic light, the present invention tested the spectrometer's recognition ability for monochromatic light with a wavelength range of 900-1000nm and a step size of 5nm. The reconstruction results are as follows: Figure 13 (a) shows the spectrometer's excellent performance in narrowband spectral recognition. The average peak wavelength difference (Δλ) between the reconstructed spectrum and the reference spectrum is approximately 0.29 nm, with the minimum peak wavelength difference being only 0.01 nm. Compared to other common micro-spectrometers, the silicon-based micro-spectrometer of this invention offers significant advantages in wavelength recognition accuracy, providing higher precision and smaller errors.

[0076] The linear dynamic range of the spectrometer is also an important indicator for evaluating its performance. To this end, the present invention tests the reconstruction effect of the silicon-based spectrometer on different powers of 1100nm laser ( Figure 14 (a) and Figure 14 (b)). The results show that the spectrometer has a laser power range of 43.3μW / cm at 1100nm. 2 Up to 3000μW / cm 2 When the power of the incident light changes, the peak intensity changes linearly with the incident light power, and the peak error is less than 1nm. In addition, thanks to the stable chemical properties of silicon materials, the spectrometer can maintain stable performance within 1 month and can still maintain the recognition accuracy of monochromatic light within 1nm without recalibration. Figure 14 (c) and Figure 14 (d)).

[0077] In order to improve the resolution, accuracy and operation speed of micro-computational spectrometers, various strategies can be considered. These strategies include: (1) increasing the size of the dataset, by reducing the learning step size or increasing the number of codes to construct a higher density spectral response matrix, thereby improving the reconstruction accuracy; (2) designing a spectral response matrix with a smaller correlation coefficient to reduce the mutual interference between matrix elements and enhance the stability and accuracy of the reconstruction process; (3) optimizing the reconstruction algorithm, such as using more advanced regularization methods to suppress errors and disturbances, or using convolutional neural networks to improve the accuracy of spectral reconstruction. Densifying the sampling density of the spectral response matrix is ​​a direct and effective way to improve the performance of the spectrometer. The following will specifically discuss the impact of the spectral response matrix dataset size on the performance of the micro-computational spectrometer.

[0078] By testing the device's ability to recognize spectra under different numbers of codes, the effect of the number of codes on the performance of the micro-computing spectrometer was explored. The present invention uses the spectral responses of three pairs of Schottky junctions, J1-J2, J3-J4, and J5-J6, under different illumination angles and voltages to construct a spectral response matrix to obtain a micro-spectrometer with a spectral resolution of 5nm. Here, the number of codes in the process of constructing the spectral response matrix is ​​reduced: only the spectral responses of the J1-J2 pair of Schottky junctions under different illumination angles and voltages are used to construct the spectral response matrix and perform a spectral reconstruction test. First, 900nm monochromatic light is reconstructed using 99 and 33 coding units, and the results are as follows. Figure 15 As shown in the figure, the noise of the 900nm monochromatic light reconstructed using 33 coding units is significantly greater than that reconstructed using 99 coding units. In addition to the peak at 900nm, there are also error peaks at 895nm and 910nm. This is because when only the J1-J2 Schottky pair is used for encoding, the spectral response changes relatively uniformly within the 800nm-950nm spectral range, which means that the reconstruction result is more susceptible to noise.

[0079] In addition, the present invention also tests the spectral resolution when using 99 and 33 coding units for spectrum reconstruction. Using these two cases, a double-peak spectrum with two intervals of 20nm, 10nm and 5nm is reconstructed. The results are as follows Figure 16As shown in the figure, it can be seen that using 33 coding units can identify double-peak spectra separated by 20nm and 10nm. However, for the more complex double-peak spectrum separated by 5nm, the recognition effect is significantly reduced: although there are characteristic peaks at 1100nm and 1105nm, there is a clear error peak at 1125nm. Using 99 coding units can perfectly identify these three spectra with only a small error. By increasing the number of codes, the resolution of the spectrometer is improved from 10nm to 5nm. This result shows that by increasing the number of spectral codes, the spectral resolution of the micro-spectrometer can be effectively improved.

[0080] In summary, in order to increase the richness of the spectral response curve during the encoding process, the present invention combines spatial encoding and temporal encoding and simultaneously introduces voltage, incident light angle, and shading distance to encode the device. By increasing the number of spectral codes, the spectral resolution of the micro-spectrometer is effectively improved. The micro-spectrometer successfully identifies monochromatic light and polychromatic light: the average detection error of the spectral peak of monochromatic light is 0.29nm, and the minimum can reach 0.01nm. At the same time, the detection resolution of polychromatic light reaches 5nm, covering the spectral range of 800nm ​​to 1200nm. In addition, thanks to the good linear dynamic range and stability of the device, the spectrometer can perfectly identify monochromatic light sources of different powers and can maintain stable performance within 1 month.

[0081] The specific embodiments of the present invention described above do not limit the scope of protection of the present invention. Any other corresponding changes and modifications made based on the technical concept of the present invention should be included in the scope of protection of the claims of the present invention.

Claims

1. A near-infrared silicon-based micro-computing spectrometer based on time and space dual encoding, characterized by: The near-infrared silicon-based micro-computing spectrometer uses a silicon wafer as a substrate, a doped region is provided in the center of the silicon wafer to form a photogenerated carrier recombination gradient field, and p pairs of Schottky junctions are provided on the surface of the silicon wafer, each pair of Schottky junctions including two Schottky electrodes symmetrically arranged on the left and right sides of the doped region; a portion of the area between each pair of Schottky electrodes, including the doped region, is used as an illumination area, and a non-illumination area on the substrate surface is shielded from light; The distance between the Schottky electrode and the illuminated area is defined as the electrode shielding distance, so that the shielding distances of the p Schottky electrodes located on the left side of the doped area are equal, and the shielding distances of the p Schottky electrodes located on the right side of the doped area vary in a gradient, forming p shielding distance variables. The spectral response characteristics of the device are controlled by changing the shielding distances. The incident light is defined as 0° when it is perpendicular to the substrate surface, and the deflection toward the left side of the doped region is defined as positive, forming q incident light angle variables. The spectral response characteristics of the device can be controlled by changing the incident light angle. Define the Schottky electrode on the left side of the doped region as the cathode and the Schottky electrode on the right side of the doped region as the anode. Apply an external bias voltage to each pair of Schottky electrodes and adjust the magnitude of the applied bias voltage to form r voltage variables. By changing the voltage, the spectral response characteristics of the device are controlled. P shading distance variables are used for spatial encoding, and q incident light angle variables and r voltage variables are used for temporal encoding. The working states of the device under the three variables are combined to obtain p×q×r coding states. The detection of unknown spectra is achieved by modulating the spectral response of the device in different coding states and combining the spectral reconstruction algorithm.

2. The near-infrared silicon-based micro-computing spectrometer based on time and space dual encoding according to claim 1, characterized in that: The silicon wafer is an n-type lightly doped silicon wafer, and the doping element of the doped region is Au.

3. The near-infrared silicon-based micro-computing spectrometer based on time and space dual encoding according to claim 1 or 2, characterized in that: The Schottky electrode is an Au electrode.

4. The near-infrared silicon-based micro-computing spectrometer based on time and space dual encoding according to claim 1, characterized in that: In a pair of Schottky junctions, the light shielding distance of the Schottky electrode located on the left side of the doped region is greater than the light shielding distance of the Schottky electrode located on the right side of the doped region.

5. The near-infrared silicon-based micro-computing spectrometer based on time and space dual encoding according to claim 1, characterized in that: After completing the working state encoding, a series of known monochromatic light tests are used to measure the spectral response of the device in different coding states. The spectral response function determined by the incident wavelength and device coding is obtained and converted into a spectral response matrix to complete the learning step. The unknown spectrum is reconstructed by testing the photocurrent of the device in different coding states under the irradiation of an unknown spectrum.

6. The near-infrared silicon-based micro-computing spectrometer based on time and space dual encoding according to claim 1, characterized in that: The working wavelength range of the near-infrared silicon-based micro-computing spectrometer is 800-1200 nm.

Images