Filter film thickness detection device and detection method

Abstract

The invention relates to the technical field of filter films, and discloses a filter film thickness detection device and a detection method. The method comprises the following steps: acquiring a first parallel beam, a second parallel beam and a third parallel beam; fixing a to-be-measured filter film sample on a sample table, and measuring the to-be-measured filter film sample to obtain a reflection spectrum and an initial refractive index parameter; irradiating the to-be-measured filter film sample by using the first parallel light beam, the second parallel light beam and the third parallel light beam, dividing the measurement area into a plurality of sub-areas according to the initial refractive index parameter, and fitting by using a least square method to obtain the optimal thickness value and refractive index value of each sub-area; based on the optimal thickness value, the refractive index value and the reflection spectrum, a mapping relation between the spectral reflectivity and the thickness is established, and a target thickness distribution diagram is obtained through iterative algorithm solving and error compensation calculation. According to the invention, the problem of uneven measurement of the thickness of the large-area filter film is solved, and the precision, efficiency and reliability of thickness detection of the filter film are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of filter films, and particularly to a filter film thickness detection device and a detection method. Background Art

[0002] Traditional filter film thickness detection methods mainly rely on ellipsometers and single-wavelength interference techniques. Although they can meet the basic measurement requirements, when facing nanoscale accuracy requirements and filter films with complex structures, these methods have obvious limitations. Especially when detecting large-area, uneven or multi-layer composite filter films, single-wavelength interference is prone to phase ambiguity problems, resulting in inaccurate measurement results; while ellipsometry, although having high precision, has a slow measurement speed, high requirements for the surface roughness of samples, and is difficult to achieve large-area rapid scanning measurement, unable to meet the needs of efficient production in the modern optical thin film industry.

[0003] Spectral reflectance analysis method, as an accurate measurement tool for filter film thickness, has become the standard detection method in the optical thin film industry. However, due to the need to process spectral data at a large number of wavelength points, traditional reflectance analysis methods have low calculation efficiency and are easily affected by the selection of initial values. Especially for multi-layer composite filter films, the inversion calculation process is complex, the convergence is poor, and the measurement accuracy is limited. In addition, traditional methods do not adequately consider environmental factors such as temperature changes, incident angle fluctuations, and wavelength stability of the measurement system itself, and it is difficult to systematically compensate for various error sources, resulting in insufficient reliability of measurement results in the actual production environment. Summary of the Invention

[0004] The main object of the present invention is to provide a filter film thickness detection device and a detection method. The present invention solves the problem of uneven measurement of large-area filter film thickness and significantly improves the accuracy, efficiency, and reliability of filter film thickness detection.

[0005] To achieve the above object, the present invention provides a filter film thickness detection method, including the following steps: Processing a first-wavelength light source, a second-wavelength light source, and a third-wavelength light source through an optical collimation system and an angle control device to obtain a first parallel light beam, a second parallel light beam, and a third parallel light beam; Fixing a filter film sample to be measured on a sample stage and setting an incident angle, and measuring the filter film sample to be measured through a spectrophotometer to obtain a reflection spectrum and initial refractive index parameters; Irradiating the filter film sample to be measured with the first parallel light beam, the second parallel light beam, and the third parallel light beam, dividing the measurement area into multiple sub-regions according to the initial refractive index parameters, and obtaining the optimal thickness value and refractive index value of each sub-region through least squares fitting; Establish a mapping relationship between spectral reflectance and thickness based on the optimal thickness value, the refractive index value, and the reflection spectrum, and obtain the target thickness distribution map through iterative algorithm solving and error compensation calculation.

[0006] The present invention also provides a filter film thickness detection device, including: A processing module, configured to process a first wavelength light source, a second wavelength light source, and a third wavelength light source through an optical collimation system and an angle control device to obtain a first parallel light beam, a second parallel light beam, and a third parallel light beam; A measurement module, configured to fix a filter film sample to be measured on a sample stage and set an incident angle, and measure the filter film sample to be measured through a spectrophotometer to obtain a reflection spectrum and an initial refractive index parameter; A fitting module, configured to irradiate the filter film sample to be measured with the first parallel light beam, the second parallel light beam, and the third parallel light beam, divide a measurement area into multiple sub-areas according to the initial refractive index parameter, and obtain the optimal thickness value and refractive index value of each sub-area through least squares fitting; A calculation module, configured to establish a mapping relationship between spectral reflectance and thickness based on the optimal thickness value, the refractive index value, and the reflection spectrum, and obtain the target thickness distribution map through iterative algorithm solving and error compensation calculation.

[0007] In summary, the technical solution provided by the present invention effectively expands the measurement range and reduces the error caused by phase ambiguity through a cooperative measurement strategy using three different wavelength light sources. At the same time, combined with the adaptive region division technology, it solves the problem of uneven measurement of the thickness of large-area filter films. The method establishes a multi-factor error compensation mechanism for the system, including compensation for refractive index error, incident angle error, temperature error, wavelength error, and phase measurement error, realizing the improvement of thickness measurement accuracy. Through the GPU parallel computing technology and the adaptive grid refinement strategy, the data processing efficiency is improved. An accurate mapping relationship model between spectral reflectance and thickness is established for different types of filter films, enhancing the accuracy and applicability of the inversion algorithm. The relative error has been reduced through cross-validation and comparative measurement, which is applicable to on-line monitoring and quality control of industrial production lines, and significantly improves the accuracy, efficiency, and reliability of filter film thickness detection. Description of the Drawings

[0008] Figure 1 is a schematic diagram of the steps of the filter film thickness detection method in an embodiment of the present invention; Figure 2 is a structural block diagram of the filter film thickness detection device in an embodiment of the present invention.

[0009] The realization, functional characteristics, and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed implementation manners

[0010] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present 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 only used to explain the present invention and are not used to limit the present invention.

[0011] Refer to Figure 1 , this embodiment provides a method for detecting the thickness of a filter film, including the following steps: S1. Process the first wavelength light source, the second wavelength light source, and the third wavelength light source through an optical collimation system and an angle control device to obtain a first parallel light beam, a second parallel light beam, and a third parallel light beam; Among them, the wavelength of the selected laser light source is controlled to meet the wavelength requirements of high-precision interference measurement. Three lasers with different bands are configured, namely blue, green and red lasers. Among them, the first wavelength light source is a blue laser, and its central wavelength is accurately adjusted to 450±5nm through a wavelength control device. This band has the characteristics of a shorter wavelength and is used to improve the spatial resolution of interference fringes; the second wavelength light source is a green laser, and its central wavelength should be adjusted to 532±3nm, which belongs to the medium wavelength band in visible light, which helps to form synthetic interference with other wavelengths and expand the measurement range; and the third wavelength light source is a red laser, by controlling its emission band, the central wavelength is stabilized at around 650±5nm. The wavelength of this band is longer, which helps to enhance the sensitivity of the synthetic wavelength to large-scale film thickness changes. Through the above-mentioned regulation, the first central wavelength, the second central wavelength and the third central wavelength are obtained respectively. In order to ensure that the three laser beams can stably and consistently form spatially consistent parallel beams in subsequent interference measurements, the three beams are collimated in turn through the same or collaborative optical collimation system. The collimation system is equipped with a precisely adjustable lens group. The lens parameters include focal length, position and spacing, etc., which are optimized according to the three central wavelengths determined to ensure that lasers of different wavelengths can be converted into parallel beams with low divergence angle and smooth wavefront after passing through the lens system. Since blue light has a shorter wavelength, optical glass materials with small chromatic aberration are selected to avoid optical path distortion caused by dispersion; green light and red light need to be matched with a configuration with a smaller lens curvature to maintain their wavefront stability. After being processed by the collimation system, the three laser beams form initial parallel beams respectively and propagate along the optical axis. An angle control device is introduced to further deflect and adjust the three initial parallel beams to achieve angle calibration operation. The angle control device is a rotating stage or fine-tuning prism mechanism with sub-angle resolution. Its control accuracy is better than 0.01°, allowing the user to accurately set the incident angle of the laser within any range of 0° to 85° according to measurement requirements. Through this device, the three laser beams can be further precisely deflected on the basis of collimation to form angle-calibrated parallel beams with uniform incident angle parameters, ensuring that the beam incident on the surface of the filter film sample is completely consistent with the assumptions in the interference theory model, thereby eliminating the phase distortion caused by the angle error. After the three laser beams are collimated and angle-adjusted, they are subjected to path splitting processing to form a reference light path and a measurement light path for interference. This process is completed by a high-stability beam splitter, which divides each laser beam into two paths in equal proportion or according to a set ratio. One path is used as a reference light and does not interact with the sample. It is directly reflected and transmitted to the spectrum detector to form a reference phase in the interference signal; the other path is the measurement light, which is irradiated on the surface of the filter film sample to be measured at a specific incident angle after angle control. After being partially reflected by the upper surface of the sample, it interferes with the reference light in the interference detection module to form interference fringes containing film thickness information.The interference pattern is then collected in real time by a two-dimensional photoelectric detector array and digitally processed. Through the synergistic effects of the above-mentioned light source wavelength regulation, optical collimation, angle adjustment, and path beam splitting, high-quality interference measurements at three wavelengths are ultimately achieved on the filter film sample, thereby obtaining the first parallel beam, the second parallel beam, and the third parallel beam respectively.

[0012] S2, Fix the filter film sample to be measured on the sample stage and set the incident angle. Measure the filter film sample to be measured with a spectrophotometer to obtain the reflection spectrum and the initial refractive index parameters. Specifically, install the filter film sample to be measured on the sample stage with a high-precision three-dimensional fine-tuning mechanism. This sample stage is equipped with a precise displacement control device to adjust the spatial position of the sample within the nanometer range to ensure that the attitude of the filter film meets the preset requirements in the XYZ three-dimensional directions. Fine-tune the sample through the displacement control device to ensure the flatness of the filter film surface, and at the same time maintain its stable fixed state at the set position to avoid systematic errors caused by displacement or vibration during subsequent optical measurements. After the sample is successfully fixed, identify the film layer type of the filter film. The identification is achieved through pre-scanning the reflectivity or user input of technical parameters, so as to determine whether the sample belongs to a dielectric multilayer film, a metal film, or a hybrid thin film. Based on the identified film layer type, automatically match the preset standard incident angle parameters. For example, set the incident angle to about five degrees for a dielectric multilayer film, fifteen degrees for a metal film, and ten degrees for a hybrid thin film, and combine the allowed small adjustment range to obtain an accurate angle setting value, which is used as the basis for the incident angle of the incident light in subsequent measurements. According to the set incident angle, use the high-resolution spectrophotometer configured in the system to scan the spectrum of the sample. The spectrophotometer continuously scans in the visible light range of 380 nm to 780 nm and records the reflectivity corresponding to each wavelength with a wavelength resolution of up to 0.5 nm to obtain a reflection spectrum dataset. The reflection spectrum contains the interference oscillation information caused by the thickness and optical properties of the filter film layer and reflects the intrinsic characteristics of the refractive index of the film layer material changing with the wavelength. Based on the reflection spectrum, construct the commonly used Cauchy model in optics. This model describes the relationship between the refractive index and the wavelength in the form of a mathematical function. In order to obtain more accurate model parameters, use the least squares method to fit the coefficients of the Cauchy model. By minimizing the square of the error between the reflectivity curve and the model calculation result, finally obtain the initial refractive index parameters of the filter film in the measured wavelength range.

[0013] S3, Use the first parallel beam, the second parallel beam, and the third parallel beam to irradiate the filter film sample to be measured, divide the measurement area into multiple sub-regions according to the initial refractive index parameters, and obtain the optimal thickness value and refractive index value of each sub-region through least squares fitting. It should be noted that the configured electronic shutter controller activates light sources of three different wavelengths in sequence. This shutter controller has a high response speed and precise timing control capabilities, ensuring that the three beams of light irradiate the surface of the filter film sample in sequence within their respective independent time windows, avoiding optical crosstalk caused by the simultaneous interference of multi-wavelength light. During the process of activating the light sources, the blue laser emitted by the first-wavelength light source forms a first parallel beam and irradiates the filter film. Subsequently, the first light source is turned off and the green laser of the second wavelength is turned on to irradiate and form a second parallel beam. Finally, the red laser of the third wavelength is activated in sequence to complete the irradiation process of the third parallel beam. Each laser irradiation will form clear interference fringe patterns on the surface of the filter film, and these fringes record information on the optical path difference of the film layer, which is directly related to the film thickness distribution. The high-resolution two-dimensional photodetector array in the system acquires the original interference fringe images generated during these three irradiation processes with high precision, forming a digital interference image dataset. These original images are subjected to dark field correction processing, that is, the background dark field image is obtained without laser irradiation, and pixel-level difference is performed with the actually acquired image to remove the background response deviation of the detector system itself, obtaining the corrected interference image. The background noise elimination operation is performed on the corrected interference image. By using an image denoising algorithm based on wavelet transform, such as the Daubechies wavelet function for multi-layer decomposition, the high-frequency noise components are removed, thereby retaining the true spatial structure characteristics of the interference fringes and obtaining the interference image. The improved five-step phase-shifting method is used to perform phase extraction processing on each group of interference images. This method extracts the optical phase information related to the fringe structure by setting the phase shift step and sequentially obtaining the interference images in multiple phase shift states, respectively obtaining the wrapped phase diagrams corresponding to the three wavelengths. Due to the 2π periodic jump characteristic of the wrapped phase diagram, it is converted into a continuous phase distribution diagram through the corresponding phase unwrapping algorithm. The quality-guided phase unwrapping technology is adopted, which weights and sorts each pixel based on the image quality factor, so as to preferentially unwrap the regions with stable phase changes and gradually expand to the boundary and mutation regions to eliminate the jumps, obtaining the continuous first phase distribution diagram, second phase distribution diagram, and third phase distribution diagram. Combining the initial refractive index parameters obtained by the spectrophotometer in the early stage, the entire interference measurement region is spatially divided, and the regions with similar phase characteristics are divided into multiple sub-regions. The boundary of the region is dynamically determined based on the rough thickness gradient and the phase change trend, ensuring that the optical characteristics within each sub-region are as consistent as possible. In each sub-region, a fitting model between the light intensity, film thickness, and refractive index is established, and the objective function is optimized by the least squares method to minimize the error between the theoretical light intensity and the measured light intensity, thereby respectively solving the optimal thickness value and the optimal refractive index value corresponding to each sub-region.

[0014] Based on the first phase distribution map and the second phase distribution map, perform a synthesis operation to calculate the first synthesized wavelength. Its physical meaning is to utilize the interference beat effect between two neighboring wavelengths to expand the effective measurement depth. Subsequently, use the first phase distribution map and the third phase distribution map to construct the second synthesized wavelength to obtain an optical modulation response with a longer scale. Through the design of the synthesized wavelength, the 2π ambiguity problem existing in traditional single-wavelength interference is effectively avoided, and the non-jump range of film thickness measurement is significantly broadened. After the calculation of the synthesized wavelength is completed, based on these two synthesized wavelengths, a short synthesized wavelength phase map and a long synthesized wavelength phase map are respectively constructed. The short wavelength phase map has a higher phase sensitivity, while the long wavelength phase map has a stronger thickness penetration ability. When these two types of information are combined, multi-scale fusion from local details to the overall contour can be achieved. According to the long synthesized wavelength phase map and the initial refractive index parameters obtained by fitting with the Cauchy model in the early stage, calculate the rough film thickness value at each pixel point to generate an original thickness distribution map, which approximately reflects the thickness distribution trend of the filter film surface in space. Since the long synthesized wavelength has strong anti-jump ability, this original thickness map can not only maintain global continuity but also truly present the film layer undulation within a large range. Perform a dynamic threshold segmentation operation on the original thickness distribution map to achieve an adaptive division of the measurement area. In this segmentation process, the threshold step is automatically calculated based on the minimum and maximum values in the thickness distribution map, and the number and boundaries of the segmented sub-regions are determined according to the change trend of the film thickness gradient within the region, so that the thickness change within each sub-region tends to be gentle, which is conducive to subsequent local modeling and fitting. After the sub-region division is completed, perform local least squares fitting optimization on each sub-region respectively. The optimization objective function is to minimize the sum of the squares of the residuals between the measured light intensity and the theoretically calculated light intensity. The variables include the film layer thickness and the refractive index. In the fitting process, an initial value selection strategy guided by gradient information is introduced, and the Levenberg-Marquardt algorithm with high numerical stability is used for non-linear optimization, so that each sub-region can finally obtain a set of optimal thickness values and refractive index values with good convergence and high precision.

[0015] S4. Based on the optimal thickness values, refractive index values, and reflection spectrum, establish the mapping relationship between the spectral reflectance and the thickness, and through iterative algorithm solution and error compensation calculation, obtain the target thickness distribution map.

[0016] Specifically, a reflectivity calculation model is established by selecting a suitable optical modeling method based on the specific structural characteristics of the filter film. For single-layer homogeneous film samples, a spectral reflectivity calculation model based on the principle of interference optics is constructed. This model accurately simulates the reflectivity change trend under different wavelengths by considering the relationship between parameters such as film thickness, refractive index and incident angle of light, and obtains a set of mapping relationships, which converts any given film thickness and refractive index into the expected reflectivity curve. For multi-layer composite filter film samples with more complex structures, a characteristic matrix calculation model is constructed. This model compiles the optical characteristic parameters of each layer of film material, such as refractive index, thickness and optical admittance, into a matrix form, and obtains the overall optical behavior of the entire film system by layer-by-layer matrix product method, and obtains the reflectivity mapping relationship of the multilayer film. The optimal thickness value and refractive index value of each sub-region obtained by phase inversion and local fitting in the previous stage are substituted into the corresponding reflectivity mapping model as the initial estimate to participate in the thickness inversion calculation. In order to improve the inversion accuracy, the Newton-Raphson iteration method is introduced. By calculating the derivative relationship between reflectivity and thickness, the true thickness value and refractive index value are gradually approached, so that the error between the theoretically calculated reflectivity and the actually measured reflectivity is minimized, and a set of preliminary thickness inversion results are obtained. Since the calculation amount required for high-resolution inversion calculation tasks on large-area filter film samples is huge, in order to avoid the problem of low efficiency of traditional serial algorithm processing, a parallel computing strategy is implemented to divide the entire filter film surface into multiple calculation blocks. Each block is independently assigned tasks according to a fixed pixel dimension, and the reflectivity of all blocks is solved and iteratively updated in parallel through GPU acceleration technology, thereby significantly improving the calculation speed while maintaining the inversion accuracy. After completing the parallel processing, the local mutation area in the thickness data is refined and analyzed. By performing boundary scanning and gradient analysis on the accelerated thickness map, the location of the thickness mutation area is identified, and an adaptive grid refinement strategy is used for these areas to divide the calculation units in the mutation area more densely to improve the local fitting and inversion accuracy. After refinement, the initial thickness distribution map and refractive index distribution map with higher resolution are reconstructed, and the error compensation process is performed based on this data. The compensation calculations of five typical error sources are implemented in sequence on the initial thickness map, including refractive index error, incident angle error, temperature error, wavelength error and phase measurement error. Among them, the refractive index error is compensated by introducing a dispersion function correction term, the incident angle error is fitted and corrected based on the thickness-to-angle sensitivity function, the temperature error is jointly corrected by thermal expansion and thermo-optical effect coefficients, the wavelength error is updated by the correction value after calibration of the standard sample, and the phase error is smoothed and adjusted by statistical weighting of multiple measurements. After compensating all the above error sources one by one, the target thickness distribution map is finally generated.

[0017] Identify and correct the refractive index errors introduced in the initial thickness distribution map due to inaccurate estimation of material intrinsic parameters. By introducing multiple auxiliary measurement wavelength points, the measured reflectivity data at these wavelengths are obtained on the filter film sample, and a more accurate dispersion function is constructed based on these data. The initial refractive index values are dynamically adjusted using the corrected dispersion relation to obtain a new set of refractive index error compensation data. Based on the compensated refractive index data, differential sensitivity analysis is carried out. By establishing the response function of thickness to the change of incident angle, that is, the partial derivative model of thickness with respect to incident angle, the change trend of film thickness under different angle deviations is evaluated, and the measured incident angle deviation value is used to finely correct the thickness data to obtain a data set including incident angle error compensation. Based on the existing incident angle compensation data, temperature error compensation is carried out according to the thermal expansion coefficient and thermo-optic coefficient of the filter film material. The system measures the deviation between the current ambient temperature and the reference temperature, and introduces temperature response correction factors to the thickness value and refractive index value respectively to obtain temperature error compensation data, so that the film thickness information measured under different experimental environments has good thermal stability and comparability. On this basis, to eliminate the systematic error caused by the wavelength shift of the light source, a single-crystal silicon wafer with known thickness and optical properties is used as the standard reference material. By analyzing the deviation between its interference pattern and the theoretical model, the actual working wavelengths of the current laser light sources are deduced, and the wavelength parameters used in the inversion model are updated accordingly to correct the wavelength error of the temperature error compensation data and obtain wavelength error compensation data. After completing the compensation for the above four error sources, the inevitable phase measurement error in the measurement is processed. By repeatedly measuring the same area multiple times, the average value and standard deviation of the phase extraction results of each measurement are calculated, and the weighted average method is used to combine the confidence coefficients of each measurement to construct the final phase estimation map. Then, the thickness data corresponding to this map is applied to the previously compensated wavelength data to generate statistically robust phase error compensation data. The data results after integrating all compensation steps are fused to generate the target thickness distribution map.

[0018] In one example, the first wavelength light source, the second wavelength light source, and the third wavelength light source are processed by an optical collimation system and an angle control device to obtain a first parallel light beam, a second parallel light beam, and a third parallel light beam, including: Perform wavelength regulation on the first wavelength light source, and adjust the central wavelength of the first wavelength light source to a blue laser with a central wavelength of 450 ± 5 nm to obtain a first central wavelength; Perform wavelength regulation on the second wavelength light source, and adjust the central wavelength of the second wavelength light source to a green laser with a central wavelength of 532 ± 3 nm to obtain a second central wavelength; Perform wavelength regulation on the third wavelength light source, and adjust the central wavelength of the third wavelength light source to a red laser with a central wavelength of 650 ± 5 nm to obtain a third central wavelength; Set the lens group parameters of the optical collimation system based on the first central wavelength, the second central wavelength, and the third central wavelength, and collimate the three laser beams through the optical collimation system to obtain an initial parallel light beam; Use the angle control device to deflect and adjust the initial parallel light beam to obtain an angle-calibrated parallel light beam; Split the angle-calibrated parallel light beam through a beam splitter, and divide each beam of light into a reference light and a measurement light. The reference light directly enters the spectral detector, and the measurement light irradiates on the surface of the filter film to be measured, obtaining a first parallel light beam, a second parallel light beam, and a third parallel light beam.

[0019] In this example, in the laser light source section, three lasers with different central wavelengths are selected as the first, second, and third wavelength light sources, corresponding to the blue, green, and red light bands respectively. To ensure that these three laser beams meet the predetermined spectral specifications, wavelength regulation is performed on each laser. This process relies on the coordinated control of built-in temperature control elements and current modulation mechanisms to achieve fine adjustment of the central wavelength. The first wavelength light source stabilizes and regulates its output wavelength within the blue light range of 450 ± 5 nm by adjusting the operating temperature and electrical drive current of the laser, ensuring its characteristics of short wavelength, high spatial resolution, and strong interference sensitivity, and obtaining the first central wavelength that meets the design specifications. The second wavelength light source is regulated to the green light range of 532 ± 3 nm. This wavelength band is located in the middle of the visible spectrum, has good system compatibility and moderate penetration ability, and can provide stable signals in different film thickness regions, thus obtaining the second central wavelength. The third wavelength light source is set in the red light region of 650 ± 5 nm through the same thermoelectric modulation and drive current adjustment. The red light has a longer wavelength and better interference modulation depth, which helps to maintain the clarity of interference fringes in the thick film region and obtain the required third central wavelength. After the central wavelengths of the three laser beams reach the above-mentioned regulation targets respectively, these three laser beams are respectively guided into the optical collimation system. This system is composed of multiple groups of precisely designed lens assemblies, capable of adjusting the focal length, radius of curvature, and lens spacing. Its design principle is to configure the optimal lens group combination for different wavelengths based on the relationship between the laser wavelength and the divergence angle. Since light of different wavelengths will produce different degrees of chromatic aberration and optical axis deviation when propagating in the lens, the propagation paths of blue, green, and red lasers in the lens system are calculated respectively during the design process, and the arrangement order, optical axis concentricity, and focal length matching of each lens group are adjusted to ensure that the three beams of light can be converted into high-quality initial parallel beams with low divergence angle, flat wavefront, and uniform beam waist after collimation processing. To ensure the collimation effect, an automatic alignment mechanism and an on-line beam quality monitoring device are equipped. By adjusting the lens position in real-time feedback, the three laser beams are accurately converted from the point light source form into collimated parallel light with consistent quality. Although the three collimated beams of light already have consistent parallelism in the spatial direction, they need to be deflected and adjusted in direction through an angle control device. This angle control device is composed of a high-precision electric rotating platform, a fine-tuning prism, or a beam deflection module. Its main function is to apply a small angular offset to the propagation direction of the light beam, thereby achieving precise setting of the incident angle. The control accuracy of this device is as high as 0.01°, allowing the system to flexibly adjust the incident angle according to the film type structure of the filter film. For example, it is set to 5° for the dielectric film, 15° for the metal film, and 10° for the hybrid film. The angle adjustment ensures the direction controllability of the three beams of light and ensures that they have the same incident conditions as the theoretical model in the subsequent interference, eliminating the phase error caused by the incident angle deviation.After angle adjustment, the three laser beams respectively form angle-calibrated parallel beams with the target incident angles, providing precise light source conditions for interference imaging. These three angle-calibrated parallel beams are sequentially introduced into a high-stability beam splitter for path separation. The beam splitter is designed with an optical film layer structure that partially transmits and partially reflects, and can reasonably distribute each laser beam into two optical paths, one is the reference light and the other is the measurement light. The reference light directly enters the spectral detector after beam splitting to record the standard optical path information in the interference signal that is not affected by the sample, while the measurement light continues to propagate at the original incident angle and irradiates on the surface of the filter film fixed on the sample stage. After being reflected by the surface of the filter film, it is superimposed with the reference light in the interference system to form interference fringes. The phase change carried by the measurement light after passing through the filter film is exactly caused by the optical path difference caused by the change in film thickness and refractive index. By measuring the phase difference change between the measurement light and the reference light, a mapping relationship between film thickness and interference phase is established. The first parallel beam, the second parallel beam, and the third parallel beam are obtained through the above process.

[0020] In one example, the filter film sample to be measured is fixed on the sample stage and the incident angle is set. The filter film sample to be measured is measured by a spectrophotometer to obtain the reflection spectrum and the initial refractive index parameters, including: The filter film sample to be measured is installed on the sample stage with a three-dimensional fine adjustment mechanism, and the position of the sample stage is adjusted by a displacement control device to obtain the filter film sample to be measured in a fixed state; The film layer type of the filter film sample to be measured in a fixed state is identified to obtain the film layer type, and a specific incident angle is set according to the film layer type to obtain the angle setting value; According to the angle setting value, the filter film sample to be measured in a fixed state is scanned in the wavelength range of 380 nm to 780 nm by a spectrophotometer to obtain the reflection spectrum; A Cauchy model is constructed based on the reflection spectrum, and the parameters of the Cauchy model are fitted by the least squares method to obtain the initial refractive index parameters.

[0021] In this example, the filter film sample to be tested is installed on a sample stage with a high-resolution three-dimensional fine-tuning mechanism, which supports nanometer-level micro-displacement adjustment along the X, Y, and Z directions. During the installation process, the filter film sample is firmly placed in the central area of ​​the sample stage by a mechanical limit device or a vacuum adsorption structure, and then it is precisely aligned by a displacement control device matched with the sample stage. The displacement control system is equipped with a closed-loop feedback system and a high-resolution electric platform, which can achieve adjustment steps in units of 10 nanometers, and monitor the position change of the sample in real time through a high-precision encoder, so that the angle between the normal direction and the incident light direction can reach an ideal state while ensuring the flatness of the sample surface. When the system determines that the sample has reached a stable and accurately positioned state, it can be regarded as being in a fixed state. After the sample is fixed, the system calls its built-in film type recognition algorithm to automatically identify the filter film. The identification process is carried out in combination with a variety of methods such as material database, historical spectral data matching and user input information. By analyzing the basic parameters of the sample such as color, initial value of surface reflectance, preparation process information, etc., it is determined that the filter film belongs to a dielectric multilayer film, a metal film or a hybrid film structure. The identified film type affects the selection method of the film model and directly determines the setting basis of the subsequent light source incident angle. For example, if the system identifies the sample as a dielectric multilayer film, the incident angle is set to a smaller value, such as 5 degrees, to avoid multiple interference at too large an angle; if it is identified as a metal film, it is suitable to be set to 15 degrees because its reflectivity characteristics are relatively stable to the angle; and if it is identified as a composite hybrid film, an intermediate value, such as 10 degrees, is taken to retain the reflection sensitivity while taking into account phase stability. By matching the identification result with the incident angle database, the angle setting value of the current sample is calculated, and this value is automatically transmitted to the angle control module to complete the calibration preset of the subsequent light source incident direction. When the sample is fixed and the incident angle is set, the spectrophotometer is started to perform a comprehensive scan of the reflection spectrum of the filter film sample. The spectrophotometer has high wavelength resolution and low stray light characteristics. The scanning range covers the entire visible spectrum from 380 nanometers to 780 nanometers, and samples at equal intervals with a step size of 0.5 nanometers, so that the tiny reflectivity fluctuations caused by film interference can be accurately captured during the entire measurement process. During the scanning process, the incident light irradiates the surface of the filter film at a preset angle, and its reflected light is merged into the detection light path through the optical sampler, and is finally received by the high-sensitivity detector and converted into a digital signal. The system stores, normalizes and suppresses noise for the collected full-spectrum reflectivity data to obtain a set of high signal-to-noise ratio reflection spectrum curves, which contain information about the film structure and optical path difference, and reflect the refractive characteristics of the film material at different wavelengths. The Cauchy model is constructed based on the acquired reflection spectrum data. The Cauchy model is a classic empirical function that describes the relationship between the refractive index of optically transparent materials and the wavelength, and is used in thin film optics.The model is presented in a three-parameter form, that is, the refractive index is expressed as a function of wavelength. By fitting the reflectance data over the entire visible light band, the refractive characteristics of the material at each wavelength are inversely deduced. To improve the fitting accuracy and robustness, the least squares method is used to optimize the parameters of the Cauchy model. The error between the measured reflectance curve and the theoretical curve calculated by the model is compared, and the sum of squared errors is used as the objective function. An iterative solution is carried out relying on a nonlinear fitting algorithm to minimize the error between the model output result and the measured data. During the fitting process, to avoid the local optimal trap, multiple groups of initial value strategies and boundary conditions are introduced, and the convergence threshold and the maximum number of iterations are set to ensure the convergence and efficiency of the fitting process. Through this least squares fitting process, the initial refractive index parameter group of the filter film within the measured spectral range is obtained.

[0022] In one example, the filter film sample to be measured is irradiated with a first parallel beam, a second parallel beam, and a third parallel beam, and the measurement area is divided into multiple sub-areas according to the initial refractive index parameters, and the optimal thickness value and refractive index value of each sub-area are obtained by least squares fitting, including: The first wavelength light source, the second wavelength light source, and the third wavelength light source are sequentially activated by an electronic shutter controller, so that the first parallel beam, the second parallel beam, and the third parallel beam sequentially irradiate the filter film sample to be measured, and three groups of original interference fringe images are obtained; A two-dimensional photodetector array is used to collect the three groups of original interference fringe images to obtain digital interference fringe image data, and the digital interference fringe image data is subjected to dark field correction processing to obtain a corrected interference fringe image; Background noise is eliminated from the corrected interference fringe image to obtain a noise-reduced interference fringe image, and the five-step phase-shifting method is used to extract the phase of the noise-reduced interference fringe image, and wrapped phase diagrams corresponding to three wavelengths are obtained respectively; The 2π jump of the wrapped phase diagram is eliminated by a quality-guided phase unwrapping algorithm to obtain a first phase distribution diagram, a second phase distribution diagram, and a third phase distribution diagram; The measurement area is divided into multiple sub-areas according to the first phase distribution diagram, the second phase distribution diagram, the third phase distribution diagram, and the initial refractive index parameters and least squares fitting is carried out to obtain the optimal thickness value and refractive index value of each sub-area.

[0023] In this example, a shutter controller sequentially activates three lasers with different central wavelengths. This shutter controller has high-precision timing management capabilities and independently schedules the opening and closing sequences of the three laser beams with a control accuracy at the millisecond level, so as to ensure that the blue laser corresponding to the first wavelength, the green laser corresponding to the second wavelength, and the red laser corresponding to the third wavelength are strictly staggered in time, each forming an independent illumination window. Whenever a laser with a certain wavelength is activated, the corresponding collimated parallel beam irradiates the surface of the filter film sample with a pre-adjusted angle, and forms an interference fringe image with the same-wavelength beam returning from the reference optical path in the interference detection system. Through this step, the original interference fringe images at three wavelengths of blue, green, and red are obtained respectively. The built-in two-dimensional photodetector array in the system acquires the above three groups of interference images with high resolution. This detector array has a pixel structure of 2048×2048, the size of each pixel is controlled at the micron level, and it is equipped with 16-bit gray-scale imaging capabilities with a high dynamic range, ensuring high-fidelity data recording under different light intensity distributions. The detector operates in the synchronous sampling mode and, through signal linkage with the shutter controller, realizes independent acquisition of the image corresponding to each laser irradiation. After the acquisition is completed, the three groups of image data are formatted and stored, and the dark-field correction process is started. During the dark-field correction process, a dark-field image is obtained under the condition of no light illumination to record the electrical noise of the detector itself and the system background signal. Subsequently, pixel-level difference operation is performed on the dark-field image from the actually acquired images, so as to effectively eliminate the errors caused by non-optical components and obtain the corrected interference image. The background noise elimination operation is performed on the corrected image. To ensure that the fringe information in the interference image is not weakened during the processing, a multi-scale wavelet transform method is used for noise reduction. The Daubechies wavelet basis is selected for five-level decomposition, a noise threshold is set in each wavelet coefficient layer and soft threshold processing is performed, and then the interference image that is smooth and retains edge features is reconstructed through inverse wavelet transform. After the noise reduction is completed, the five-step phase-shifting method is used to extract the phase information for each group of images. By sequentially acquiring five interference images with a known phase difference (π / 2) and using the formula to calculate the corresponding wrapped phase diagram, the optical phase information in the interference fringes can be accurately restored. This method has strong anti-noise performance and high resolution and is suitable for high-precision optical interference systems. In this step, each wavelength image group will obtain a corresponding wrapped phase diagram, and the phase values in these images are still in a 2π ambiguity state and cannot be directly used for film thickness calculation, so phase unwrapping processing is performed. In order to restore the wrapped phase diagram to a real and continuous phase distribution diagram, a quality-guided phase unwrapping algorithm is used to process it. This algorithm evaluates the phase quality factor of each pixel in the entire image, judges its local phase gradient, noise distribution, and edge continuity, and constructs an unwrapping path priority diagram based on this.The algorithm preferentially performs phase unwrapping in high-quality regions and gradually expands to the edges and complex regions, thereby effectively avoiding the propagation of jump errors, and finally obtaining the first phase distribution map, the second phase distribution map, and the third phase distribution map in the blue, green, and red wavelength bands respectively. These three phase maps are all continuous phase information, with high spatial resolution and high phase accuracy. After obtaining the phase maps, combined with the initial refractive index parameters obtained by fitting with the Cauchy model in the early stage, the measurement region is intelligently divided. The division method is based on the local gradient change, adaptive thickness distribution estimation, and spatial structure characteristics in the phase distribution map, and divides the entire interference field of view into multiple sub-regions, making the optical properties within each sub-region relatively uniform, which is conducive to improving the accuracy of local inversion. For each divided sub-region, a non-linear relationship model between reflectivity, phase, thickness, and refractive index is established, and the least squares method is used to fit the parameters of this model. The least squares method iteratively solves for the optimal film thickness and refractive index values by minimizing the sum of the squares of the differences between the actual measured light intensity and the light intensity calculated by the theoretical model. During the solution process, the Levenberg-Marquardt optimization algorithm is introduced as the core of numerical solution to improve the fitting accuracy and convergence speed. The optimal film thickness value and refractive index value within each spatial sub-region are obtained.

[0024] In one example, according to the first phase distribution map, the second phase distribution map, the third phase distribution map, and the initial refractive index parameters, the measurement region is divided into multiple sub-regions and least squares fitting is performed to obtain the optimal thickness value and refractive index value of each sub-region, including: Based on the first phase distribution map and the second phase distribution map, wavelength synthesis calculation is performed to calculate the first synthesized wavelength, and based on the first phase distribution map and the third phase distribution map, the second synthesized wavelength is calculated; According to the first synthesized wavelength and the second synthesized wavelength, and simultaneously using the first phase distribution map, the second phase distribution map, and the third phase distribution map to construct a synthesized phase, the short synthesized wavelength phase map and the long synthesized wavelength phase map are respectively obtained; According to the long synthesized wavelength phase map and the initial refractive index parameters, the rough thickness distribution is calculated to obtain the original thickness distribution map; The original thickness distribution map is subjected to dynamic threshold segmentation to obtain multiple sub-regions, and local optimization is performed on each sub-region to obtain the optimal thickness value and refractive index value of each sub-region.

[0025] In this example, the synthetic wavelength is calculated using the first phase distribution map and the second phase distribution map. These two phase maps respectively correspond to the interference measurement results of the first wavelength and the second wavelength. Therefore, the phase difference reflected by them is the interference period difference between the two-wavelength lights under the same optical path. Furthermore, the first synthetic wavelength is calculated through the wavelength synthesis formula. The physical meaning of this wavelength value is that a synthetic wavelength is constructed by the difference between different wavelengths, and its period is much larger than the original wavelength, thereby effectively expanding the unambiguous measurement range of the interference measurement system and reducing the influence of phase jump errors. In the same way, based on the first phase distribution map and the third phase distribution map, the second synthetic wavelength is calculated, enabling the system to obtain two synthetic wavelengths. Among them, the first synthetic wavelength is shorter, and the second synthetic wavelength is longer, which are respectively used to construct phase distribution maps under different measurement scales subsequently. The aforementioned first phase distribution map, second phase distribution map, and third phase distribution map are combined and used to generate a synthetic phase map through point-by-point phase difference operations. The specific method is to calculate the differences between the first phase distribution map minus the second phase distribution map and the first phase distribution map minus the third phase distribution map respectively, which respectively correspond to the first synthetic wavelength and the second synthetic wavelength, to obtain two synthetic phase maps. Among them, the first synthetic phase map has a shorter wavelength scale, so it has a higher spatial resolution and is suitable for capturing small changes in local film thickness. The second synthetic phase map, due to its longer wavelength and larger interference fringe spacing, has good anti-2π jump ability and is suitable for extracting the thickness structure contour on a large scale. The two synthetic phase maps are respectively named the short synthetic wavelength phase map and the long synthetic wavelength phase map, and the thickness calculation is carried out based on this. The long synthetic wavelength phase map is selected as the basis for the preliminary thickness estimation because its large wavelength characteristic can effectively avoid phase jumps in the thick film region, thereby enabling the rough thickness estimation to have strong global continuity. During the thickness calculation process, combined with the initial refractive index parameters previously obtained by fitting with the Cauchy model, using this refractive index as the effective refractive index input, and combining the synthetic wavelength and phase map data, the thickness-phase inverse formula is used to calculate the film layer thickness value at each pixel point, obtaining a complete original thickness distribution map, which macroscopically reflects the thickness change trend of the filter film sample within the entire field of view. Dynamic threshold segmentation is performed on the original thickness distribution map to divide the measurement area into several sub-regions. The dynamic threshold segmentation is based on the difference range between the maximum and minimum values in the thickness map. By setting an adaptive threshold stepping strategy, the thickness space is divided into multiple intervals, and the number of partitions and boundaries are determined according to the gradient trend of the thickness change. This division method ensures that the thickness change within each sub-region is relatively gentle, thereby improving the stability and accuracy of the subsequent fitting process. After the division, each sub-region is used as an independent modeling unit for fine optimization calculation of local thickness and refractive index.During the local optimization process, an objective function containing the square of the difference between the theoretical reflectivity and the measured reflectivity is constructed, and the least squares method is introduced as an optimization means. An iterative-based solution strategy is adopted to continuously adjust the thickness and refractive index parameters to minimize the error term. To improve the calculation efficiency and convergence performance, the Levenberg-Marquardt algorithm is introduced in the optimization process, which combines the advantages of the gradient descent method and the Newton method and can quickly converge under the condition that the initial value is relatively close to the solution. At the same time, at the boundary position of each sub-region, a bicubic interpolation strategy is introduced to smooth the fitting result, ensuring that the finally obtained thickness values are continuous and spatially consistent between sub-regions and avoiding local discontinuous thickness jumps. The optimal thickness values and refractive index values corresponding to each sub-region are obtained.

[0026] In one example, based on the optimal thickness values, refractive index values, and reflection spectra, a mapping relationship between spectral reflectivity and thickness is established, and through an iterative algorithm for solution and error compensation calculation, a target thickness distribution map is obtained, including: A spectral reflectivity calculation model is established based on a single-layer homogeneous film, and reflectivity calculations are performed based on the spectral reflectivity calculation model to obtain a reflectivity mapping relationship for the single-layer film; A characteristic matrix calculation model is established based on a multi-layer composite filter film, and reflectivity calculations are performed based on the characteristic matrix calculation model to obtain a reflectivity mapping relationship for the multi-layer film; The optimal thickness values and refractive index values are substituted into the reflectivity mapping relationship for the single-layer film or the reflectivity mapping relationship for the multi-layer film for thickness inversion, and the Newton-Raphson iterative method is used for solution to obtain a preliminary thickness inversion result; A parallel computing strategy is implemented for the preliminary thickness inversion result. The surface of the filter film is divided into multiple computing blocks, and the computing tasks are processed simultaneously through GPU acceleration technology to obtain the thickness data after acceleration processing; Adaptive grid refinement processing is performed on the thickness mutation regions in the thickness data after acceleration processing to obtain the thickness data after refinement processing, and an initial thickness distribution map and a refractive index distribution map are generated based on the thickness data after refinement processing; Compensation calculations for refractive index error, incident angle error, temperature error, wavelength error, and phase measurement error are performed on the initial thickness distribution map to obtain the target thickness distribution map.

[0027] In this example, for a single-layer homogeneous film sample, a spectral reflectance calculation model is established based on the thin-film interference theory. This model takes the incident light wavelength, film thickness, film refractive index, substrate refractive index, and incident angle as input parameters, and combines electromagnetic boundary conditions to derive the expression form of the reflectance formed by the interference of light at multiple air-film-substrate interfaces. This model reveals the periodic variation law of reflectance with film thickness and wavelength, and is used to construct the mapping relationship between film thickness and reflectance. Through a large number of parametric simulation calculations, a reflectance mapping diagram covering different thicknesses and refractive index conditions is generated. For a more complex multi-layer composite filter film sample, the characteristic matrix method is used for modeling. In this method, each layer of film is represented as a 2×2 matrix with specific optical admittance and phase delay, and these characteristic matrices are multiplied in sequence according to the layer order to form the equivalent propagation matrix of the entire film system. Through this total matrix, the complex reflection coefficient and reflectance at any wavelength are calculated. Since the interference behavior of light in the multi-layer film system has strong inter-layer coupling characteristics, it is difficult to accurately express with a simple analytical model. Therefore, this matrix model provides a calculation method with strong generality and high precision, making the system applicable to filter film structures with any number of layers and any material combinations. After the modeling is completed and the reflectance mapping relationship is generated, the optimal thickness values and refractive index values obtained in each sub-region in the early stage are substituted into the corresponding models respectively, and then thickness inversion is carried out. For the known reflectance measurement values and the initially estimated optical parameters, the Newton-Raphson iterative method is used for inversion solution. This method takes the reflectance model as the objective function, and by obtaining the first-order derivative (i.e., partial derivative information) of the function at the current parameter point, the film thickness and refractive index are gradually corrected, and gradually approach the true physical solution as the error decreases. The convergence threshold and the maximum number of iterations are set to ensure that the calculation complexity is controlled while ensuring the accuracy, and the preliminary thickness inversion results of each calculation point are obtained. Considering the large-area and high-resolution spatial characteristics of the entire filter film sample, a parallel computing strategy is implemented. The entire filter film surface is divided into multiple calculation blocks, and each block contains a fixed number of pixel units, such as a pixel array of 128×128. These blocks are assigned to different calculation cores of the GPU, and the reflectance calculation, model iteration, and thickness solution operations are carried out simultaneously. With the help of the CUDA parallel computing platform, the system schedules a large number of concurrent threads to maximize the computing power of the GPU, thereby greatly improving the overall computing speed, making the originally time-consuming inversion calculation efficiently completed within an acceptable time. To address the problem of sudden or sharp changes in the thickness of the filter film surface in local areas, edge detection and difference analysis are performed on the thickness data after acceleration processing. By judging whether the thickness difference between adjacent pixels exceeds the set threshold, the thickness mutation areas are identified.Within these regions, to improve the model resolution and fitting accuracy, an adaptive mesh refinement strategy is initiated. The original mesh division scale is reduced to one-fourth or less of the original, generating denser computational cells, and local optimization calculations are re-executed in the refined mesh. In this way, while maintaining the overall computational efficiency, the detail accuracy is taken into account, significantly enhancing the film thickness modeling effect in complex structure regions. After completing the data processing for all regions, the thickness results are re-stitched and integrated to generate the initial thickness distribution map and the corresponding refractive index distribution map. These images are consistent with the initially acquired interference images in terms of spatial resolution, achieving a high-precision mapping from phase data to physical parameters. Considering that the system is affected by various error sources during the measurement process, such as refractive index model error, incident angle deviation, environmental temperature fluctuation, laser wavelength drift, and phase extraction uncertainty, multi-factor compensation is performed on the initial thickness results to improve the accuracy of thickness measurement. For refractive index error, a dispersion relation correction strategy is introduced. The reflectivity of the filter film is measured at multiple additional wavelength points, and the refractive index curve is corrected across the entire wavelength band in combination with the Sellmeier equation. For incident angle error, a response function of thickness to angle change is established through differential sensitivity analysis, and the thickness data is corrected in combination with the actual incident angle deviation. For temperature error, based on the thermal expansion coefficient and thermo-optic coefficient of the film material, a compensation formula is used to jointly correct the thickness and refractive index. For wavelength error, the interference spectrum of a single-crystal silicon wafer standard sample is used for wavelength calibration to reverse-correct the light source offset. For phase measurement error, by performing multiple measurements and statistical averaging on the same region, a weighted phase estimation method is used to improve the reliability of the phase data. After all error compensations are completed, the finally generated target thickness distribution map is comprehensively corrected.

[0028] In one example, compensation calculations for refractive index error, incident angle error, temperature error, wavelength error, and phase measurement error are performed on the initial thickness distribution map to obtain the target thickness distribution map, including: Compensation calculations for refractive index error in the initial thickness distribution map are performed to obtain refractive index error compensation data; Based on differential sensitivity analysis, incident angle error compensation is performed on the refractive index error compensation data to obtain incident angle error compensation data; According to the incident angle error compensation data, temperature error compensation is performed using the thermal expansion coefficient and thermo-optic coefficient of the material to obtain temperature error compensation data; Wavelength error compensation is performed on the temperature error compensation data using the single-crystal silicon wafer standard material to obtain wavelength error compensation data; Phase measurement error compensation is performed on the wavelength error compensation data to obtain phase error compensation data, and the target thickness distribution map is generated based on the phase error compensation data.

[0029] In this example, compensation calculations are performed on the refractive index errors in the initial thickness distribution map. Since the refractive index was obtained based on the Cauchy model or empirical fitting in the early-stage modeling, but the dispersion characteristics of the actual material deviate due to process fluctuations, material impurities, or interface structure differences, multi-band spectral reflectance measurement data is introduced to correct the refractive index. By adding five auxiliary wavelength points, such as 480 nm, 500 nm, 550 nm, 600 nm, and 630 nm, the corresponding reflectance values are obtained, and these data are compared with the results calculated by the theoretical model. Then, combined with the Sellmeier equation, a corrected dispersion model is constructed to obtain the refractive index deviation at each wavelength. These deviation values are weighted back into the original refractive index distribution data in the form of a function to obtain the corrected refractive index error compensation data, and from this, the film thickness values under conditions closer to the true refractive index are derived. After completing the refractive index error compensation, the measurement errors caused by the incident angle deviation are identified. Since in actual measurements, the laser incident angle deviates slightly from the set value due to the mechanical tolerances of the optical system, platform drift, or thermal stress deformation, this small angle error will significantly affect the phase distribution of the interference fringes, and thus affect the thickness inversion result. The differential sensitivity analysis method is used to establish a partial derivative model of the thickness with respect to the change in the incident angle, that is, to construct a response function of the thickness with respect to the angle change. By fitting the change trend of the film thickness under different angle conditions, the sensitivity function is obtained. The film thickness is measured at multiple deviation angles such as θ±0.5°, θ±1°, θ±1.5°, etc., and substituted into the sensitivity function for reverse correction to obtain the incident angle error compensation data. The errors caused by temperature are processed. Since the physical dimensions and optical properties of the filter film material change under different temperature conditions, the thickness and refractive index are compensated bidirectionally by combining the thermal expansion coefficient and thermo-optic coefficient of the material. The system detects the difference between the ambient temperature during measurement and the reference temperature, and then corrects the physical thickness of the film layer for thermal expansion according to the thickness compensation formula, and modifies the optical refractive index value for thermo-optic modulation using the refractive index compensation formula. Temperature correction processing is performed on all pixel points in sequence to generate temperature error compensation data, so that the measurement results under different ambient temperature conditions have good consistency and stability. After temperature compensation is completed, wavelength error compensation is performed. The deviation between the actual operating wavelength of the laser and the nominal value is identified. Since the center wavelength of the laser drifts after long-term operation, and the ambient conditions during the measurement also affect the laser output wavelength, a single-crystalline silicon wafer with a known thickness and stable material properties is used as a standard reference material. Interference measurement is performed on this material to obtain the phase data corresponding to the interference fringes, and it is compared with the theoretical calculated value to inversely deduce the current actual operating wavelength. After obtaining the true center wavelength of each light source, the wavelength parameter in the thickness calculation is updated, and the reflectance model is recalculated to obtain the film thickness data after wavelength compensation processing, that is, the wavelength error compensation data.After completing the above four types of systematic error compensation, the measurement errors caused by phase extraction uncertainty are processed. Especially in multi-wavelength interference, phase offsets or jitters occur in the phase images of different bands due to uneven light source intensity, non-linear response of the detector, or poor interference fringe quality. The statistical averaging method is used to correct the phase. Multiple repeated measurements are performed on each region to obtain multiple sets of phase diagrams, and the average phase value and standard deviation of each pixel are calculated. Then, using the weighted average algorithm, a greater weight is given to the phase with higher stability, thereby constructing a weighted average phase diagram. This diagram is used for the final thickness inversion calculation, and the corresponding results are combined with the wavelength error compensation data to obtain the final thickness result after phase measurement error compensation.

[0030] Referring to Figure 2 , this embodiment provides a filter film thickness detection device, including: A processing module 1, configured to process a first-wavelength light source, a second-wavelength light source, and a third-wavelength light source through an optical collimation system and an angle control device to obtain a first parallel light beam, a second parallel light beam, and a third parallel light beam; A measurement module 2, configured to fix a filter film sample to be measured on a sample stage and set an incident angle, and measure the filter film sample to be measured through a spectrophotometer to obtain a reflection spectrum and initial refractive index parameters; A fitting module 3, configured to irradiate the filter film sample to be measured with the first parallel light beam, the second parallel light beam, and the third parallel light beam, divide the measurement region into multiple sub-regions according to the initial refractive index parameters, and obtain the optimal thickness value and refractive index value of each sub-region by least squares fitting; A calculation module 4, configured to establish a mapping relationship between spectral reflectance and thickness based on the optimal thickness value, refractive index value, and reflection spectrum, and obtain a target thickness distribution map through iterative algorithm solving and error compensation calculation.

[0031] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to that described in the above method embodiment, and details will not be elaborated here.

[0032] It should be noted that in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, device, article or method. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, device, article or method including the element.

[0033] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be similarly included in the patent protection scope of the present invention.

Claims

1. A method for detecting the thickness of a filter film, characterized in that: The following steps are involved: Processing the first wavelength light source, the second wavelength light source and the third wavelength light source through an optical collimation system and an angle control device to obtain a first parallel light beam, a second parallel light beam and a third parallel light beam; Fixing the filter film sample to be tested on a sample stage and setting the incident angle, measuring the filter film sample to be tested by a spectrophotometer to obtain a reflection spectrum and an initial refractive index parameter; irradiating the filter film sample to be measured with the first parallel light beam, the second parallel light beam and the third parallel light beam, dividing the measurement area into a plurality of sub-areas according to the initial refractive index parameter, and obtaining the optimal thickness value and refractive index value of each sub-area by least squares fitting; A mapping relationship between spectral reflectance and thickness is established based on the optimal thickness value, the refractive index value and the reflection spectrum, and a target thickness distribution map is obtained through iterative algorithm solution and error compensation calculation.

2. The method for detecting the thickness of the filter film according to claim 1, characterized in that: The method of processing the first wavelength light source, the second wavelength light source and the third wavelength light source through the optical collimation system and the angle control device to obtain the first parallel light beam, the second parallel light beam and the third parallel light beam comprises: The first wavelength light source is wavelength-controlled to adjust the central wavelength of the first wavelength light source to a blue laser of 450±5 nm to obtain a first central wavelength; The wavelength of the second wavelength light source is controlled to adjust the central wavelength of the second wavelength light source to a green laser of 532±3 nm to obtain a second central wavelength; The third wavelength light source is wavelength-controlled, and the central wavelength of the third wavelength light source is adjusted to a red laser of 650±5 nm to obtain a third central wavelength; Setting the lens group parameters of the optical collimation system based on the first central wavelength, the second central wavelength and the third central wavelength, and performing collimation processing on the three laser beams through the optical collimation system to obtain initial parallel light beams; Using an angle control device to deflect and adjust the initial parallel light beam to obtain an angle-calibrated parallel light beam; The angle-calibrated parallel light beams are split by a beam splitter to divide each light beam into a reference light and a measurement light, wherein the reference light directly enters a spectrum detector and the measurement light is irradiated on the surface of the filter film to be measured to obtain a first parallel light beam, a second parallel light beam and a third parallel light beam.

3. The method for detecting the thickness of the filter film according to claim 1, characterized in that: The method of fixing the filter film sample to be tested on a sample stage and setting an incident angle, and measuring the filter film sample to be tested by a spectrophotometer to obtain a reflection spectrum and an initial refractive index parameter includes: The filter film sample to be tested is mounted on a sample stage with a three-dimensional fine-tuning mechanism, and the position of the sample stage is adjusted by a displacement control device to obtain the filter film sample to be tested in a fixed state; Identifying the film type of the filter film sample to be tested in the fixed state to obtain the film type, and setting a specific incident angle according to the film type to obtain an angle setting value; According to the angle setting value, the fixed state filter film sample to be tested is scanned in the wavelength range of 380nm to 780nm by using a spectrophotometer to obtain a reflection spectrum; A Cauchy model is constructed based on the reflection spectrum, and the parameters of the Cauchy model are fitted by the least square method to obtain initial refractive index parameters.

4. The method for detecting the thickness of the filter film according to claim 1, characterized in that: The method of irradiating the filter film sample to be measured with the first parallel light beam, the second parallel light beam and the third parallel light beam, dividing the measurement area into a plurality of sub-areas according to the initial refractive index parameter, and obtaining the optimal thickness value and refractive index value of each sub-area by least squares fitting includes: The first wavelength light source, the second wavelength light source and the third wavelength light source are sequentially activated by an electronic shutter controller, so that the first parallel light beam, the second parallel light beam and the third parallel light beam sequentially irradiate the filter film sample to be tested, and three groups of original interference fringe images are obtained; The three groups of original interference fringe images are collected by using a two-dimensional photodetector array to obtain digital interference fringe image data, and the digital interference fringe image data are subjected to dark field correction processing to obtain a corrected interference fringe image; Eliminating background noise from the corrected interference fringe image to obtain a noise-reduced interference fringe image, and performing phase extraction on the noise-reduced interference fringe image using a five-step phase shift method to obtain wrapped phase images corresponding to three wavelengths respectively; Eliminate 2π jumps on the wrapped phase diagram using a quality-guided phase unwrapping algorithm to obtain a first phase distribution diagram, a second phase distribution diagram, and a third phase distribution diagram; The measurement area is divided into a plurality of sub-areas according to the first phase distribution diagram, the second phase distribution diagram, the third phase distribution diagram and the initial refractive index parameter, and a least squares fitting is performed to obtain an optimal thickness value and a refractive index value of each sub-area.

5. The method for detecting the thickness of the filter film according to claim 4, characterized in that: The method of dividing the measurement area into a plurality of sub-areas according to the first phase distribution diagram, the second phase distribution diagram, the third phase distribution diagram and the initial refractive index parameter and performing least squares fitting to obtain an optimal thickness value and a refractive index value of each sub-area includes: Performing wavelength synthesis calculation based on the first phase distribution diagram and the second phase distribution diagram to calculate a first synthesized wavelength, and calculating a second synthesized wavelength based on the first phase distribution diagram and the third phase distribution diagram; According to the first synthetic wavelength and the second synthetic wavelength, a synthetic phase is constructed by using the first phase distribution diagram, the second phase distribution diagram and the third phase distribution diagram at the same time, and a short synthetic wavelength phase diagram and a long synthetic wavelength phase diagram are obtained respectively; Calculate a rough thickness distribution according to the long synthetic wavelength phase map and the initial refractive index parameter to obtain an original thickness distribution map; The original thickness distribution map is segmented by dynamic threshold value to obtain multiple sub-regions, and each sub-region is locally optimized to obtain the optimal thickness value and refractive index value of each sub-region.

6. The method for detecting the thickness of the filter film according to claim 1, characterized in that: The mapping relationship between spectral reflectance and thickness is established based on the optimal thickness value, the refractive index value and the reflection spectrum, and a target thickness distribution map is obtained by solving the problem through an iterative algorithm and calculating error compensation, including: A spectral reflectance calculation model is established according to a single-layer homogeneous film, and reflectance calculation is performed based on the spectral reflectance calculation model to obtain a single-layer film reflectance mapping relationship; Establishing a characteristic matrix calculation model according to the multi-layer composite filter film, and performing reflectivity calculation based on the characteristic matrix calculation model to obtain a multi-layer film reflectivity mapping relationship; Substituting the optimal thickness value and the refractive index value into the single-layer film reflectivity mapping relationship or the multi-layer film reflectivity mapping relationship to perform thickness inversion, and using the Newton-Raphson iteration method to solve, to obtain a preliminary thickness inversion result; Implementing a parallel computing strategy on the preliminary thickness inversion result, dividing the filter film surface into multiple computing blocks, and processing computing tasks simultaneously through GPU acceleration technology to obtain accelerated thickness data; Performing adaptive grid refinement processing on a thickness mutation region in the thickness data after the acceleration processing to obtain the thickness data after the refinement processing, and generating an initial thickness distribution map and a refractive index distribution map based on the thickness data after the refinement processing; Compensation calculations are performed on the initial thickness distribution map for refractive index error, incident angle error, temperature error, wavelength error and phase measurement error to obtain a target thickness distribution map.

7. The method for detecting the thickness of the filter film according to claim 6, characterized in that: The method of performing compensation calculation for refractive index error, incident angle error, temperature error, wavelength error and phase measurement error on the initial thickness distribution map to obtain a target thickness distribution map includes: Performing compensation calculation on the refractive index error in the initial thickness distribution map to obtain refractive index error compensation data; Performing incident angle error compensation on the refractive index error compensation data based on differential sensitivity analysis to obtain incident angle error compensation data; Performing temperature error compensation using the thermal expansion coefficient and the thermo-optic coefficient of the material according to the incident angle error compensation data to obtain temperature error compensation data; Performing wavelength error compensation on the temperature error compensation data using a single crystal silicon wafer standard material to obtain wavelength error compensation data; Phase measurement error compensation is performed on the wavelength error compensation data to obtain phase error compensation data, and a target thickness distribution map is generated based on the phase error compensation data.

8. A filter film thickness detection device, characterized in that: The method for detecting the thickness of a filter film according to any one of claims 1 to 7 is used to implement the steps of the method, wherein the filter film thickness detection device comprises: A processing module, used for processing the first wavelength light source, the second wavelength light source and the third wavelength light source through an optical collimation system and an angle control device to obtain a first parallel light beam, a second parallel light beam and a third parallel light beam; A measuring module is used to fix the filter film sample to be tested on a sample stage and set an incident angle, and measure the filter film sample to be tested by a spectrophotometer to obtain a reflection spectrum and an initial refractive index parameter; A fitting module, used for irradiating the filter film sample to be measured with the first parallel light beam, the second parallel light beam and the third parallel light beam, dividing the measurement area into a plurality of sub-areas according to the initial refractive index parameter, and obtaining the optimal thickness value and refractive index value of each sub-area by least squares fitting; The calculation module is used to establish a mapping relationship between spectral reflectance and thickness based on the optimal thickness value, the refractive index value and the reflection spectrum, and obtain a target thickness distribution map through iterative algorithm solution and error compensation calculation.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images