Optical element surface pattern structured light testing system

Abstract

The present application relates to the technical field of optical element surface physical property testing, and discloses a structured light testing system for optical element surface patterns, which comprises: a pair of phase conjugate structured light fields for alternately illuminating the optical element, synchronous acquisition of Fourier domain diffraction patterns and spatial domain in-situ light field intensity distribution, and calculation of a differential signal to determine a defect position after correction of the diffraction patterns by the in-situ light field intensity distribution. The present application presets a common mode signal cancellation mechanism for defect-free periodic patterns at the physical layer level through conjugate illumination and differential processing, thereby suppressing a large amount of redundant information and retaining and highlighting only the asymmetric scattering signal caused by defects, which makes the focus of detection shift from the accuracy of measuring geometric morphology to the uniformity of distinguishing the physical properties of material optical response, and realizes high signal-to-noise ratio rapid discrimination of functional defects.

Description

Technical Field

[0001] This invention relates to a structured light testing system for surface patterns of optical components, belonging to the technical field of testing the physical properties of optical component surfaces. Background Technology

[0002] Currently, optical elements based on surface micro / nano periodic patterns have been widely used in high-precision fields such as diffractive waveguides, security anti-counterfeiting, and photonic chips due to their unique light field manipulation capabilities. However, the optical performance of these elements is highly dependent on the periodicity of their structure, which itself brings a series of technical challenges. Among them, the strong diffraction effect generated by the periodic structure is particularly prominent. In some applications, this effect is even regarded as an interference signal that needs to be suppressed. For example, Chinese patent CN103229078B discloses a security element with an optically variable surface pattern. In order to obtain a colorless and interference-free visual effect, this technology introduces a non-periodic, pseudo-random, high degree of misalignment into periodically arranged reflective elements, such as sawtooth gratings or micromirror arrays, to actively destroy the conditions for coherent superposition of light waves at the physical level, thereby effectively suppressing the diffraction effect caused by the structure. The technology demonstrates, from a component design perspective, that the inherent optical signals (diffraction) generated by periodic structures are extremely significant and must be properly handled. However, in the field of quality control, this problem manifests in another form: when it is necessary to detect local functional defects in such components, the strong and uniform diffraction or reflection signals contributed by a massive number of defect-free periodic patterns will form structural background noise that is difficult to filter out, thus drowning out the weak defect signals. In the manufacturing process of high-precision optical components such as diffractive waveguides, photonic chips, and semiconductor masks, quality control of the millions of micro- and nano-functional patterns on their surfaces is the core link to ensure their final optical performance. Currently, the technical means used in the industry for such detection have different technical paths based on their working principles and application focuses, but all of them have their inherent limitations.

[0003] The first type of technology is high-resolution microscopic imaging based on point-by-point scanning, typically represented by scanning electron microscopy (SEM) or atomic force microscopy (AFM). This type of technology uses physical probes or electron beams to perform precise mechanical scanning on the sample surface, enabling accurate reconstruction of the geometry of a single pattern with high spatial resolution. Its technological value lies in providing detailed microscopic physical information for sample development or failure analysis. However, when this technology is applied to the specific scenario of online quality control in large-scale production, its inherent logic of reconstruction before judgment introduces constraints. In order to find a small number of abnormal defects, it is necessary to perform indiscriminate scanning of all patterns (including a large number of defect-free patterns), which has a high degree of information redundancy and is extremely time-consuming. This approach makes full-area coverage inspection of every product economically impractical. Manufacturers are therefore forced to accept a compromise of low-ratio sampling inspection, which undoubtedly puts the quality consistency of the final product at a probabilistic risk.

[0004] The second type of technical approach combines optical imaging with image processing, aiming to improve detection efficiency through parallel optical imaging. This type of method typically uses a high-magnification optical microscope in conjunction with a high-pixel industrial camera to acquire large-format images of the component surface. Then, a computer runs complex image processing algorithms, such as template comparison, edge detection, or machine learning models, to identify areas that have geometric differences from the standard pattern. Although this technology is far faster than scanning microscopes, its detection capability is highly dependent on the significant differences in the geometric morphology of the defects. When the root cause of the defect is not geometric deviations such as contour or size, but rather abnormal optical properties of the material itself, such as gradual changes in film thickness or local non-uniformity of the material's refractive index, the contrast of such defects in the optical image may be extremely low, or even unable to be effectively identified by conventional image processing algorithms, thus leading to missed detection.

[0005] The third type of technique utilizes scattered light for detection, such as dark-field microscopy or laser scattering measurement. Its principle is to receive only the scattered light generated by surface discontinuities in the sample, such as defects, particles, and scratches, while blocking specular reflections from flat, defect-free areas. This method is highly sensitive for detecting isolated defects with strong scattering characteristics (such as dust particles). However, it faces significant challenges when detecting large-scale periodic micro / nano patterns. The periodic patterns themselves generate strong diffraction, which, along with the weak scattered light from defects, enters the detector, forming a difficult-to-separate mixed signal. Strong structural background noise severely overwhelms the asymmetric scattering signal caused by functional defects, such as minor material contamination or etching residue, resulting in a low signal-to-noise ratio and difficulty in consistently detecting such defects.

[0006] In summary, existing technologies generally suffer from the following shortcomings when addressing the need for full-domain online inspection of large-scale micro-nano functional patterns: 1. The scanning microscope-based approach sacrifices efficiency for accuracy, and its redundant data acquisition mode is unsuitable for mass production requirements; 2. Conventional optical image processing techniques primarily focus on the accuracy of geometric morphology. When defects manifest as non-uniformity in the physical properties of the material's optical response, their detection sensitivity and reliability are limited; 3. When applied to the inspection of periodic patterns, techniques based on scattered light detection struggle to effectively separate weak defect signals from strong structural background noise. Therefore, the technical problem this invention aims to solve is to establish a new testing method that avoids redundant information acquisition from defect-free areas and shifts the focus of inspection from measuring the accuracy of geometric morphology to identifying the uniformity of the material's optical response physical properties, thereby achieving rapid identification of functional defects with a high signal-to-noise ratio. Summary of the Invention

[0007] This invention provides a structured light testing system for surface patterns of optical components. Its main purpose is to solve the problem that existing technologies, in pursuit of full-area detection, must redundantly collect information from a massive number of defect-free areas, resulting in a fundamental contradiction between detection efficiency and economy.

[0008] To achieve the above objectives, the present invention provides a structured light testing system for optical element surface patterns, the system comprising:

[0009] An illumination module is used to generate a first conjugate structured light field and a second conjugate structured light field with a phase difference of , and to alternately illuminate the first conjugate structured light field and the second conjugate structured light field onto the surface pattern of an optical element;

[0010] A Fourier optical module is used to perform Fourier transform on the first conjugate structured light field and the second conjugate structured light field after the surface pattern of the optical element is applied.

[0011] An image acquisition module has its photosensitive surface located at the back focal plane of a Fourier optical module. The image acquisition module acquires a first diffraction pattern corresponding to a first conjugate structure light field and a second diffraction pattern corresponding to a second conjugate structure light field in a first region of the photosensitive surface. The image acquisition module also acquires a first in-situ light field intensity distribution image corresponding to the first diffraction pattern and a second in-situ light field intensity distribution image corresponding to the second diffraction pattern in a second region of the photosensitive surface.

[0012] A processing module, connected to the image acquisition module, is used to: perform pixel-by-pixel intensity correction on the first diffraction pattern and the second diffraction pattern based on the first in-situ light field intensity distribution image and the second in-situ light field intensity distribution image, to generate the corrected first diffraction pattern and the corrected second diffraction pattern; calculate the difference image between the corrected first diffraction pattern and the corrected second diffraction pattern; and determine the pixel points in the difference image whose pixel intensity exceeds a threshold as the defect locations of the surface pattern of the optical element.

[0013] Preferably, the illumination module includes a Mach-Zehnder interferometer and a phase modulator; the Mach-Zehnder interferometer is used to split a coherent beam into a first beam and a second beam, and to superimpose the first beam and the second beam on the surface pattern of the optical element to form a structured light field; the phase modulator is placed in the optical path of the first beam, and the phase modulator is used to introduce a phase difference of 0 and 0 between the first beam and the second beam to generate a first conjugate structured light field and a second conjugate structured light field.

[0014] Preferably, the image acquisition module includes a beam splitter and an imaging lens; the beam splitter is placed between the Fourier optical module and the photosensitive surface of the image acquisition module, and is used to split the light field passing through the Fourier optical module into a main beam and a branch beam; the main beam propagates to a first region of the photosensitive surface to form a first diffraction pattern and a second diffraction pattern; the branch beam is imaged on a second region of the photosensitive surface by the imaging lens to form a first in-situ light field intensity distribution image and a second in-situ light field intensity distribution image.

[0015] Preferably, the phase modulator is a piezoelectric ceramic driven mirror, which is also used to introduce a phase difference between the first beam and the second beam to generate a third structured light field and a fourth structured light field; the image acquisition module also acquires a third diffraction pattern corresponding to the third structured light field and a fourth diffraction pattern corresponding to the fourth structured light field; the processing module is also used to: generate a phase defect map characterizing the phase anomaly of the surface pattern of the optical element by calculation, and generate an amplitude defect map characterizing the amplitude anomaly of the surface pattern of the optical element by calculation, wherein the intensity of the phase defect map is calculated by a formula, where is the third diffraction pattern and is the fourth diffraction pattern.

[0016] Preferably, the substrate of the optical element is made of a transparent material; the piezoelectric ceramic driven mirror is also used to introduce a time-dependent optical path difference between the first beam and the second beam; the image acquisition module is used to acquire a time-series diffraction pattern frame; the processing module is also used to: perform a Fourier transform on the time-series signal of each pixel in the time-series diffraction pattern frame, and separate the first frequency signal contributed only by the interface where the surface pattern of the optical element is located and the second frequency signal contributed by the interface of the back surface of the substrate of the optical element in the generated spectrum; and the processing module reconstructs a parasitic interference-free diffraction pattern based on the first frequency signal, and uses the parasitic interference-free diffraction pattern to determine the defect location.

[0017] Preferably, the processing module is further configured to: generate a defect distribution map containing spatial information of all identified defect locations; and perform spatial statistical analysis on the defect distribution map to identify whether multiple defect locations have feature signatures associated with process fault types in their spatial distribution, and output indication information associated with preset process problem sources based on the feature signatures.

[0018] Preferably, the processing module performs spatial statistical analysis on the defect distribution map by executing a two-dimensional Fourier transform to detect whether there are periodic arrangement feature signatures that appear as isolated characteristic frequency peaks in the frequency domain; and the processing module also performs spatial density clustering analysis on the defect distribution map to detect whether there are clustered feature signatures that appear as clustered regions of specific contours in the spatial domain.

[0019] Preferably, the illumination module further includes a Fabry-Perot etalon placed downstream of the optical path of the piezoelectric ceramic-driven mirror; after performing a Fourier transform on the time-series diffraction pattern frame, the processing module is also used to: identify a reference peak with a constant frequency generated by the Fabry-Perot etalon in the spectrum; and, based on the actual peak shape and center frequency position of the reference peak, to perform real-time correction of the frequency axis distortion introduced by the nonlinear motion of the piezoelectric ceramic-driven mirror and the optical path drift caused by changes in ambient temperature.

[0020] Preferably, after separating the first frequency signal, the processing module is further configured to: extract the amplitude information of the first frequency signal to generate a diffraction pattern without parasitic interference; and extract the phase information of the first frequency signal in parallel, and generate a height distribution map characterizing the three-dimensional morphology of the surface pattern of the optical element by performing phase unwrapping processing on the phase information.

[0021] Preferably, the processing module performs intensity correction on the first diffraction pattern and the second diffraction pattern in the following manner: calculates the ratio of the total light intensity of the first in-situ light field intensity distribution image to the total light field intensity distribution image of the second in-situ light field, and uses this ratio of total light intensity as a scale factor to perform a multiplication operation on the intensity of all pixels of the second diffraction pattern to generate the corrected second diffraction pattern.

[0022] Compared with the prior art, the beneficial effects of the present invention are:

[0023] 1. This method alternately illuminates the surface of a component by generating a pair of spatially phase-conjugate structured light fields. By utilizing the interaction between the periodicity of the surface pattern and the periodicity of the structured light field, the signals contributed by a large number of defect-free uniform pattern units in the two acquired diffraction patterns present highly similar common-mode information. The subsequent differential processing is not a conventional background noise elimination, but a targeted common-mode signal cancellation process pre-set at the physical optics level. This process suppresses most of the information in the defect-free region in the final result, and only the asymmetric scattering signal caused by local functional defects that disrupts the symmetry of the optical response is retained as the differential result. Thus, this method transforms the problem of searching for weak signals in massive redundant information into a problem of directly interpreting high signal-to-noise ratio differential signals. Its focus also shifts from the accuracy of measuring geometric morphology to identifying the uniformity of the physical properties of the material's optical response.

[0024] 2. The step of generating the structured light field in this invention may further include introducing multiple preset orthogonal phase relationships. By acquiring multiple corresponding diffraction patterns, subsequent differential processing can separate two independent types of information, namely, those indicating phase anomalies and amplitude anomalies of the surface pattern, from the originally mixed defect signals. This separation of information is not based on subsequent algorithmic decomposition of a single defect signal, but rather stems from the fact that, in the initial physical stage of the interaction between the light field and the sample, the different physical properties of the defect are orthogonally modulated and encoded by structured light of different phases. Subsequent processing simply decodes and presents this information that has already been encoded at the optical level. Therefore, it can provide a basis for judgment that directly points to different physical causes such as drift in the photolithography etching process or residues in the surface cleaning process, providing direct technical guidance for the traceability and correction of the production process.

[0025] 3. The differential signal obtained by this method, after being thresholded to identify functional defects, forms a defect distribution map containing spatial location information of multiple defects. This map can be used as a new information source and input into the subsequent spatial statistical analysis process. By processing the distribution map, such as by two-dimensional Fourier transform or spatial density clustering, it is possible to reveal the periodic or clustered characteristics that individual defects do not possess but are exhibited by the spatial arrangement of defect groups. This method of extracting field information from point information transforms the endpoint data originally used for single-piece product qualification judgment into process data that can be used to understand the status of upstream production equipment or the stability of the process flow. This transforms the detection system from a product quality inspection tool into a production process monitoring and early warning tool. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the system testing and defect analysis process of the present invention;

[0027] Figure 2 This is a schematic diagram of the optical system for conjugate illumination differential detection of the present invention;

[0028] Figure 3 This is a timing diagram of the dual-frame acquisition and differential processing of the system of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] A structured light testing system for optical element surface patterns comprises an illumination module, a Fourier optics module, an image acquisition module, and a processing module. The illumination module generates a structured light field. The Fourier optics module and the image acquisition module work together to acquire information about the interaction between the light field and the element surface. The processing module performs signal interpretation and defect location determination. The entire system aims to convert the non-uniformity of the optical response properties of the optical element surface under test into a differential signal that can be analyzed. In specific applications, such as quality control of diffractive waveguides or semiconductor masks with large-scale periodic micro / nano functional patterns, existing technologies often require redundant information acquisition from a large number of defect-free areas to detect a few functional defects, leading to efficiency and cost constraints. To address this issue, this system employs a common-mode signal suppression method, shifting the focus of the detection task from measuring the geometric morphology of the pattern to identifying the uniformity of the material's optical response physical properties. As a result, the system can suppress redundant information from defect-free areas and focus processing resources on the abnormal physical property signals caused by defects.

[0031] The illumination module utilizes a Mach-Zehnder interferometer. A coherent beam, such as a 632.8nm helium-neon laser, is split into a first beam and a second beam by a beam splitter. The two beams are guided by mirrors in their respective optical paths and superimposed on the surface pattern of the optical element under test to form a structured light field. A phase modulator, specifically a piezoelectric ceramic-driven mirror, is placed in the optical path of the first beam. By applying a preset voltage to the piezoelectric ceramic-driven mirror, a normal displacement is generated, introducing a phase difference of 0 and π between the first and second beams. This allows for the alternating generation of a first conjugate structured light field and a second conjugate structured light field with a phase difference of π. The calibration of the operating voltage for the discrete phase steps generated by the piezoelectric ceramic-driven mirror in the illumination module is also discussed. Before undertaking batch testing tasks, the system performs a self-calibration procedure based on intensity response. This procedure first drives the piezoelectric ceramic to complete a continuous scan exceeding the range with no device under test in the optical path, using no less than 100 voltage steps. Simultaneously, it records the light intensity sequence acquired by specific pixels of the CMOS sensor, typically pixels in the center region of the zero-order diffraction spot. The processing module then performs sine fitting on the light intensity sequence, calculates the maximum and minimum light intensity values, and establishes a nonlinear response mapping relationship between voltage input and phase shift output based on the inverse cosine function. Finally, by retrieving the target value equal to this mapping relationship, the system automatically generates and stores a set of precise driving voltage values ​​corresponding to the four phases mentioned above, which serve as the calling parameters for all subsequent testing tasks.

[0032] The photosensitive surface of the image acquisition module, such as a 2048×2048 pixel CMOS sensor, is placed on the back focal plane of the Fourier optical module, i.e., a Fourier transform lens, to acquire the diffraction pattern of the light field after the illumination light field is applied to the component surface. To simultaneously acquire diffraction information and spatial information of the illumination light field, a beam splitter is placed between the photosensitive surface of the Fourier optical module and the image acquisition module. This beam splitter divides the light field into a main beam and a branch beam. The main beam propagates to the first region of the photosensitive surface to form a diffraction pattern, while the branch beam passes through an imaging lens to form the surface pattern of the component under test in the second region of the photosensitive surface. The in-situ light field intensity distribution image is obtained, thus enabling a single exposure to simultaneously acquire the first diffraction pattern in the Fourier domain and the first in-situ light field intensity distribution image in the spatial domain, or the second diffraction pattern and the second in-situ light field intensity distribution image, on different sensor regions. Upon receiving the image data, the processing module, to eliminate measurement errors introduced by fluctuations in the total energy of the illumination light field, first performs intensity correction. This procedure sums the full-frame pixel intensities of the first and second in-situ light field intensity distribution images to obtain the total light intensity and its ratio as a scaling factor. Subsequently, the second diffraction pattern is... The intensity values ​​of all pixels in the diffraction pattern are multiplied by this scale factor to generate the corrected second diffraction pattern. As a numerical example, if the value is 5.80 × 10⁷ or 5.75 × 10⁷, the scale factor is approximately 1.0087. The processing module then multiplies each pixel value in the second diffraction pattern by 1.0087 to align the total light intensity with the first diffraction pattern. Furthermore, for the intensity correction stage, to eliminate the fixed-mode noise introduced by the uneven distribution of illumination light field energy and the inconsistency of sensor pixel response, rather than just the fluctuation of total energy between frames, the system adopts a method of generating a two-dimensional correction matrix. The specific procedure for replacing the calculation of a single scale factor is as follows: Before formal testing, a standard mirror with a surface flatness better than the standard mirror is placed at the test position. At least 16 frames of in-situ light field images generated by the first conjugate structured light field are acquired and averaged to obtain a reference image with a high signal-to-noise ratio. This process is repeated to obtain a reference image corresponding to the second conjugate structured light field. The processing module calculates a correction matrix based on this. In subsequent batch testing, this matrix is ​​used to perform pixel-by-pixel multiplication correction on each acquired second diffraction pattern. That is, while correcting the total light intensity difference, all fixed spatial non-uniformities in the optical path are compensated.

[0033] After intensity correction, the processing module calculates the difference image between the first diffraction pattern and the corrected second diffraction pattern. For a defect-free periodic pattern, the diffraction pattern produced under illumination by two conjugate structured light fields presents common-mode information, and the signal intensity is significantly weakened after differential calculation. However, a local functional defect, such as a material contamination, will generate an asymmetric scattering signal because it disrupts the local optical response symmetry. This signal is retained as a residual after differential processing. Finally, the processing module performs a threshold judgment on the difference image to determine the defect location. The threshold setting process is as follows: first, calculate the average value and standard deviation of the pixel intensity in the background region of the difference image, and then determine the defect location. The threshold is set to , where is a coefficient set according to the detection sensitivity requirements, for example, a value of 6. Pixels in the differential image whose pixel intensity exceeds this threshold are identified as defect locations. To further distinguish the physical source of the defect, the system can also perform information decoupling operations. By controlling the piezoelectric ceramic driven mirror, an additional phase difference between and is introduced between the first beam and the second beam to generate a third structured light field and a fourth structured light field, and the corresponding third diffraction pattern and fourth diffraction pattern are acquired. The processing module calculates and generates a phase defect map characterizing the phase anomaly of the surface pattern of the optical element. This calculation method is sensitive to changes in optical path, thus highlighting phase defects caused by geometric shape deviations.

[0034] When the substrate of the optical element under test is made of a transparent material, to avoid parasitic interference caused by reflections from the back surface of the substrate, the system can be configured to perform coherent gated signal separation. In this case, the piezoelectric ceramic-driven mirror is driven by a linearly varying voltage signal and performs continuous scanning, introducing a time-varying optical path difference between the two arms of the interferometer. The image acquisition module simultaneously acquires a time-series diffraction pattern frame. The processing module performs a Fourier transform on the time-series signal of each pixel in the time-series diffraction pattern frame. In the generated spectrum, the signals contributed by the front surface pattern and the signals contributed by the back surface will be at the first and second frequencies respectively due to the different optical path differences. A bandpass filter separates the first frequency signal, which is contributed only by the interface where the surface pattern of the optical element is located, and reconstructs a parasitic interference-free diffraction pattern based on this signal for subsequent defect location determination. Simultaneously, the processing module can also extract the phase information of the first frequency signal in parallel and generate a height distribution map characterizing the three-dimensional morphology of the optical element's surface pattern through phase unwrapping processing. To ensure the accuracy of the above time-series analysis, the system can also integrate a self-calibration mechanism to address [variable issues]. The nonlinear motion of the piezoelectric ceramic-driven mirror and the optical path drift caused by changes in ambient temperature are addressed by placing a Fabry-Perot etalon downstream of the piezoelectric ceramic-driven mirror's optical path. This etalon generates a reference peak with a constant frequency in the spectrum. The processing module identifies the actual peak shape and center frequency position of this reference peak and performs real-time correction of frequency axis distortion and drift. After determining all defect locations, the processing module can also perform defect distribution feature signature analysis. This procedure first generates a defect distribution map containing spatial information of all determined defect locations. Then, spatial statistical analysis is performed on the distribution map to identify whether there are feature signatures associated with process fault types in the spatial distribution of defects. Specific analysis methods include: performing a two-dimensional Fourier transform to detect the existence of periodic arrangement feature signatures that appear as isolated characteristic frequency peaks in the frequency domain, and performing spatial density clustering analysis to detect the existence of clustered feature signatures that appear as specific contour clusters in the spatial domain. By comparing the identified feature signatures with a preset process problem feature library, a basis for tracing and correcting upstream production processes can be provided.

[0035] Example 1: In the production scenario of a diffractive waveguide element for augmented reality display, online quality control faces a technical challenge: some elements, after being confirmed to meet design standards through sampling inspection based on scanning electron microscopy (SEM), are still deemed unqualified due to stray light issues during end-use. Subsequent analysis shows that this problem stems from a large-area, low-contrast, gradually varying etching depth. This defect manifests physically as non-uniformity in optical response, with a weak spatial gradient that is difficult to effectively identify using traditional local morphology measurement methods. In this production process, a structured light testing system is deployed after the plasma etching process for full-area inspection. When a diffractive waveguide element enters the inspection area, the illumination module in the system illuminates the element surface with a first conjugate structured light field, and the image acquisition module acquires a first diffraction pattern in a first region of its photosensitive surface and simultaneously acquires a first in-situ light field intensity distribution image in a second region. Subsequently, the illumination module switches to a second conjugate structured light field, and the image acquisition module acquires a second diffraction pattern and a second in-situ light field intensity distribution image.

[0036] When analyzing the data, the processing module faces the challenge that the low-frequency differential signal generated by the gradual change in etching depth has characteristics similar to the background signal formed by the Gaussian energy distribution of the illumination beam itself. If differential calculations are performed directly, the former will be masked by the latter. Therefore, two technical features in the system work synergistically in a specific order. First, the processing module uses the simultaneously acquired first and second in-situ light field intensity distribution images to normalize the total light intensity of the two frames and generate a correction matrix to compensate for the non-uniformity of the illumination light field. The output of this step is two corrected first and second diffraction patterns, with the systematic background signal removed, providing clean data input for subsequent processing. Based on this, the processing module then performs differential calculations on the two corrected diffraction patterns. Since the signals from millions of defect-free grating units are canceled out as common-mode information, the original... The masked asymmetric scattering signal caused by the gradual change in etching depth over a large area appears as a low-intensity region with a recognizable spatial contour in the differential image. This detection method transforms the problem of measuring the geometric accuracy of a single pattern into discerning the uniformity of the optical response of the pattern array. It integrates the speed advantage of parallel optical acquisition and the sensitivity advantage of differential signal processing within a single architecture, thus enabling it to meet the online detection requirements of such large-area low-contrast defects. After identifying the low-intensity region as a defect based on a preset threshold, the system immediately performs defect distribution feature signature analysis, parametrically extracting the spatial features such as the contour and gradient of the region. The analysis results are not only used to determine the current optical waveguide component as unqualified, but the output feature signature is also associated with the gas flow control unit of a specific upstream plasma etching equipment, providing data support for the real-time correction of process parameters.

[0037] Example 2: To quantitatively verify the effectiveness of the phase-locked differential information decoupling mechanism in the aforementioned technical solution for identifying defects from different physical sources, this experiment was set up. Its purpose was to objectively evaluate whether the system could separate mixed defect signals into independent information indicating phase and amplitude anomalies, respectively. The experiment was conducted on a structured light testing system built on an optical vibration isolation platform. The system used a 632.8nm helium-neon laser and was equipped with a piezoelectric ceramic-driven mirror with a displacement resolution of 2nm and a 16-bit dynamic range CMOS industrial camera. To establish a verification benchmark, a specially prepared silicon substrate sample was used. The sample surface was etched with a grating pattern with a period of 1μm, and two types of defects with known physical properties were introduced onto it: in region A of the sample, a phase-type defect was formed by deep etching of the grating within a 10μm × 10μm area using a focused ion beam at a depth of 20nm; in region B of the sample, a 10μm × 10μm area and a 5nm thick semi-transparent chromium film was deposited by electron beam evaporation, forming an amplitude-type defect.

[0038] The experiment included a control group and an experimental group. In the control group, the sample was illuminated with first and second conjugate structured light fields with a phase difference of 0, and diffraction patterns were collected, with their difference images calculated. In the experimental group, in addition to the control group's procedure, diffraction patterns generated by third and fourth structured light fields with phase differences of π / 2 and 3π / 2 were additionally collected, and difference images representing amplitude anomalies and phase anomalies were calculated respectively. After processing and normalizing the collected data, in the difference images of the control group, both phase-type defects in region A and amplitude-type defects in region B exhibited high... The intensity signals, with normalized intensities of 0.85 and 0.89 respectively, are so close that they cannot be distinguished. However, in the analysis results of the experimental group, in the differential image representing amplitude anomalies, the signal intensity of amplitude-type defects in region B is 0.93, while the signal of phase-type defects in region A is suppressed to 0.09. Correspondingly, in the differential image representing phase anomalies, the signal intensity of phase-type defects in region A is 0.91, while the signal of amplitude-type defects in region B is suppressed to 0.07. Meanwhile, in the defect-free background areas of all differential images, the signal intensity is no higher than 0.04.

[0039] Experimental data shows that by acquiring four diffraction patterns with orthogonal phase relationships and performing specific differential operations, defect signals from different sources can be separated. This separation effect stems from the selective response of different differential calculations to different physical property changes. Differential operations are sensitive to amplitude changes during the interaction between the light field and the sample, while differential operations are sensitive to phase changes. The experimental results confirm that this technical solution can provide detection information with clear distinction for defects with different physical properties, providing data support for the root cause analysis of production processes.

[0040] Example 3: This example combines Figures 1 to 3 A structured light testing system for surface patterns of an optical element is described, such as... Figure 1 As shown, the process begins with the preparation of the optical element under test, which contains micro-nano functional patterns. Subsequently, the illumination module generates conjugate structured light to alternately illuminate the sample. The light field after being applied to the sample is subjected to Fourier transform by the Fourier optics module and simultaneously acquired by the image acquisition module, along with the diffraction pattern and the in-situ light field. The acquired image data first undergoes an intensity correction step to eliminate energy fluctuation errors, and then enters the differential processing stage. This stage aims to suppress the common-mode signal contributed by the defect-free region and highlight the asymmetric signal caused by defects. The processing module then performs threshold judgment to accurately determine the defect location. After the defect is located, the system can selectively perform phase-locked differential information decoupling to separate the phase and amplitude information of the defect, thereby distinguishing its physical origin. Finally, the system performs defect distribution feature signature analysis on all identified defect locations. By identifying spatial features such as periodicity or clustering, it provides direct evidence for tracing the source of upstream process problems.

[0041] like Figure 2 As shown, a coherent beam emitted by a laser 1 first passes through a beam splitter 2 and is split into two coherent beams. One beam has a piezoelectric ceramic phase modulator 4 in its optical path and can selectively incorporate a Fabry-Perot etalon 5. The other beam propagates through a mirror 3. The two beams eventually interfere and superimpose at the surface pattern 6 of the optical element under test, forming a structured light field that interacts with the surface pattern. The light field after passing through the optical element under test enters a Fourier lens 7 for Fourier transform. The transformed light field is then split into two beams by a beam splitter 9. The main beam is directly projected onto the first region of the CMOS sensor 10 to obtain the diffraction pattern, while the branch beam passes through an imaging lens 8 and is imaged onto the second region of the CMOS sensor 10 to obtain the in-situ light field distribution. The image signal acquired by the CMOS sensor 10 is transmitted to a processing module, which sequentially performs a series of data processing operations such as intensity correction, differential calculation, and defect identification to ultimately determine the defect location of the surface pattern of the optical element.

[0042] like Figure 3As shown, after system initialization, the illumination module first emits a first conjugate structured light field with a phase of 0. This light field interacts with the optical elements to form a diffraction light field. After Fourier transformation, the image acquisition module acquires the first diffraction pattern in the first region and the first in-situ light field distribution in the second region. Then, this first set of data is transmitted to the processing module. Immediately afterwards, the illumination module performs a phase switching operation, emits a second conjugate structured light field with a phase of 0, and repeats the above acquisition and transmission process to obtain the second diffraction pattern and the second in-situ light field distribution. After receiving the two sets of data, the processing module sequentially performs the steps of calculating the light intensity ratio for intensity correction, calculating the differential image, and determining the defect location by threshold judgment. Finally, it outputs a defect distribution map containing all defect spatial information.

[0043] Example 4: In a specific system calibration application, to enable the aforementioned defect distribution feature signature analysis function to identify specific process problems, a feature signature knowledge base needs to be established. This process aims to associate foreseeable process failures with different physical causes with the spatial distribution patterns of defects detected by the system. To this end, a standardized system calibration and knowledge base construction procedure is executed on a semiconductor photoresist process verification line equipped with the structured light testing system of this invention. To determine the basic threshold for defect detection, a standard silicon wafer confirmed by atomic force microscopy to be free of functional defects is first used as a reference sample. The system performs 10 consecutive repeated measurements on the reference sample and statistically analyzes the background area of ​​the differential image generated by each measurement, calculating the average pixel intensity and standard deviation of all measurement results. Furthermore, the value of the coefficient in the defect judgment threshold is set to a value that can control the misjudgment rate of the reference sample to below 0.001%, which is satisfied when determined by calculation in this procedure.

[0044] After the detection threshold is determined, two types of process faults are actively introduced and their defect distribution patterns are collected by controlling the bias of the process equipment parameters. The first type is the collection of periodic characteristics. A sinusoidal electrical noise with an amplitude of 5mV and a frequency of 50Hz is introduced into the Y-axis driver of the lithography machine scanning mirror to simulate the periodic interference of the equipment's electronic components. A test wafer is then processed under this condition. The second type is the collection of aggregation characteristics. The back suction parameter in the center dispensing program of the spin coater is set to 0 to simulate the adhesive residue caused by valve failure. A test wafer is then prepared under this condition. The system then processes these two wafers with pre-existing defects. For test wafers with known process fault sources, defect distribution feature signature analysis is performed. For the first wafer prepared under periodic interference, the system generates a defect distribution map showing defect points arranged periodically along the Y-axis. The processing module performs a two-dimensional Fourier transform on the defect distribution map. In the generated spectrum, two isolated characteristic frequency peaks appear at positions on the positive and negative half-axis of the Y-axis. Based on this, a record is created in the feature signature knowledge base, with the following content: Fault type - periodic, Fault source - 50Hz interference on the Y-axis of the lithography machine scanning galvanometer, Feature signature type - two-dimensional Fourier transform spectrum characteristic peak, Feature parameter -.

[0045] For the second wafer affected by adhesive residue, its defect distribution map shows a cluster of defects highly concentrated in the central region of the wafer. The processing module performs spatial density clustering analysis on this map, identifying a cluster with an approximately circular outline and an area of ​​12.3 mm², whose centroid coordinates deviate from the wafer's geometric center by less than 0.5 mm. Based on this, another record is created in the knowledge base, with the following content: Fault Type - Clustering, Fault Source - Adhesive Residue in the Center of the Spinneret, Feature Signature Type - Spatial Density Clustering Feature, Feature Parameter - Number of Clusters = 1, Centroid Radial Position <, Area >. Through the execution of this procedure, the content of the feature signature knowledge base can be constructed and expanded through a standardized experimental process.

[0046] Example 5: When the system of the present invention is applied to the detection of transparent optical elements with thin substrates, the optical path difference formed by the reflected light from the front and rear surfaces is small. During coherent gated differential analysis, the center frequencies of the first frequency signal and the second frequency signal separated by the processing module in the frequency domain are close, causing signal peak aliasing, which in turn interferes with the subsequent extraction of diffraction patterns without parasitic interference. To address this situation, the system performs a pre-calibration procedure for the built-in reference channel before deployment. This procedure first places a standard plane mirror with a highly reflective dielectric film on its surface and no transparent substrate at the test position, and drives a piezoelectric ceramic driven mirror to perform continuous scanning under a temperature-controlled environment, synchronously acquiring time-series diffraction pattern frames. The processing module performs a Fourier transform on the acquired signal. At this time, the spectrum only contains the main signal peak from the standard plane mirror and the reference peak generated by the Fabry-Perot etalon. The system stores the center frequency and peak shape profile of the acquired reference peak as a set of reference parameters to characterize the instrument response function of the system under this deployment environment.

[0047] After completing the pre-calibration, the system proceeds to inspect the thin-substrate transparent optical element. The processing module performs a Fourier transform on the acquired time-series signal and first uses the real-time acquired reference peak to correct for optical path drift caused by changes in ambient temperature. Subsequently, using the reference parameter set stored in the calibration procedure as the kernel function for deconvolution, deconvolution processing is performed on the spectrum with signal peak aliasing. This process improves the effective resolution of the spectrum, separating the previously overlapping first and second frequency signals into two independent signal peaks. The processing module can then apply bandpass filtering to extract the first frequency signal contributed only by the interface where the element surface pattern is located, and use it for subsequent defect location determination. It should be noted that in thin-substrate... During the inspection of the substrate transparent optical element, the processing module performs bandpass filtering on the spectrum after deconvolution to extract the amplitude information of the first frequency signal used for defect analysis, and simultaneously extracts the phase information of the signal. By performing phase unwrapping processing on the phase information, the system synchronously generates a height distribution map characterizing the three-dimensional morphology of the surface pattern of the element without adding any extra acquisition time. Through analysis of the height distribution map, a gradually changing concave morphology with a radius of curvature of 5.2m was found on the surface of the element. Although this morphology is within the geometric tolerance range of the individual element, it provides preliminary quantitative input on stress distribution and focus compensation for the subsequent integration and calibration of the optical module.

[0048] Example 6: In a specific system application scenario, when a new specification of optical element is introduced into the production line, a standardized pre-calibration procedure needs to be executed to achieve optimal and repeatable performance of the detection system. This procedure generates a dedicated working parameter matrix for the element, aiming to transform the system's general detection capabilities into dedicated detection capabilities for the specific physical characteristics of the element. The procedure first calibrates the spatial frequency of the structured light field in the illumination module. The processing module takes the nominal pattern period of the new element as input and calculates the target period of the structured light field based on the relevant formula. Then, it calculates the target angle between the two beams of the interferometer, where λ is the laser wavelength. The system then controls... The piezoelectric ceramic-driven reflector adjusts the beam angle until the spatial frequency of the interference fringes measured by the calibration grating matches the target value within a 0.1% tolerance, and this angle setting value is stored in the working parameter matrix. After determining the illumination parameters, the procedure performs signal-to-noise ratio verification and adaptive adjustment of the acquisition parameters. The system acquires a frame of diffraction pattern with an initial integration time of 10ms, calculates the peak intensity of the first-order diffraction peak and the root mean square value of the noise in the background region, and obtains the initial signal-to-noise ratio. If this value is lower than the preset threshold of 25dB, the system increases the integration time in 2ms increments and repeats the above process until the signal-to-noise ratio is greater than or equal to 25dB, and the finally determined integration time value is stored in the working parameter matrix.

[0049] For transparent components requiring coherent gated differential analysis, the procedure also includes calibrating the scanning parameters of the piezoelectric ceramic-driven mirror. To achieve a balance between avoiding signal aliasing and suppressing low-frequency noise, the system performs a scan speed optimization. It drives the piezoelectric ceramic at an initial low speed and performs a Fourier transform on the acquired time-series signal to obtain the peak frequency of the main signal. The system then gradually increases the scan speed until the center frequency stabilizes at 50% of the camera's Nyquist frequency. That is, the corresponding scan speed and the preset scan range to ensure spectral resolution are stored together in the working parameter matrix. The completion of this procedure signifies that a complete set of validated working parameters has been generated for this new specification component, which can be used for subsequent mass production testing.

[0050] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A structured light testing system for optical element surface patterns, characterized in that The system includes: An illumination module is used to generate a first conjugate structured light field and a second conjugate structured light field with a phase difference of π, and to alternately illuminate the first conjugate structured light field and the second conjugate structured light field onto the surface pattern of an optical element; A Fourier optical module is used to perform Fourier transform on the first conjugate structured light field and the second conjugate structured light field after the surface pattern of the optical element is applied. An image acquisition module has its photosensitive surface located at the back focal plane of a Fourier optical module. The image acquisition module acquires a first diffraction pattern corresponding to a first conjugate structure light field and a second diffraction pattern corresponding to a second conjugate structure light field in a first region of the photosensitive surface. The image acquisition module also acquires a first in-situ light field intensity distribution image corresponding to the first diffraction pattern and a second in-situ light field intensity distribution image corresponding to the second diffraction pattern in a second region of the photosensitive surface. A processing module, connected to the image acquisition module, is used to: perform pixel-by-pixel intensity correction on the first diffraction pattern and the second diffraction pattern based on the first in-situ light field intensity distribution image and the second in-situ light field intensity distribution image, to generate the corrected first diffraction pattern and the corrected second diffraction pattern; calculate the difference image between the corrected first diffraction pattern and the corrected second diffraction pattern; and determine the pixel points in the difference image whose pixel intensity exceeds a threshold as the defect locations of the surface pattern of the optical element.

2. The structured light testing system for optical element surface patterns according to claim 1, characterized in that, The illumination module includes a Mach-Zehnder interferometer and a phase modulator. The Mach-Zehnder interferometer is used to split a coherent beam into a first beam and a second beam, and to superimpose the first beam and the second beam on the surface pattern of the optical element to form a structured light field. The phase modulator is placed in the optical path of the first beam and is used to introduce a phase difference of 0 and π between the first beam and the second beam to generate a first conjugate structured light field and a second conjugate structured light field.

3. The structured light testing system for optical element surface patterns according to claim 1, characterized in that, The image acquisition module includes a beam splitter and an imaging lens; the beam splitter is placed between the Fourier optical module and the photosensitive surface of the image acquisition module, and is used to split the light field passing through the Fourier optical module into a main beam and a branch beam; the main beam propagates to a first region of the photosensitive surface to form a first diffraction pattern and a second diffraction pattern. The branched light beam is imaged onto the second region of the photosensitive surface through the imaging lens to form a first in-situ light field intensity distribution image and a second in-situ light field intensity distribution image.

4. The structured light testing system for optical element surface patterns according to claim 2, characterized in that, The phase modulator is a piezoelectric ceramic driven mirror, which is also used to introduce a phase modulator between the first beam and the second beam. and The phase difference is used to generate a third structured light field and a fourth structured light field; the image acquisition module also acquires the third diffraction pattern corresponding to the third structured light field and the fourth diffraction pattern corresponding to the fourth structured light field; the processing module is also used to: generate a phase defect map characterizing the phase anomaly of the surface pattern of the optical element by calculation, and generate an amplitude defect map characterizing the amplitude anomaly of the surface pattern of the optical element by calculation, wherein the intensity of the phase defect map is... Through formula The calculation yielded that, This is the third diffraction pattern. This is the fourth diffraction pattern.

5. The structured light testing system for optical element surface patterns according to claim 4, characterized in that, The substrate of the optical element is made of a transparent material; the piezoelectric ceramic driven mirror is also used to introduce a time-dependent optical path difference between the first beam and the second beam; the image acquisition module is used to acquire a time-series diffraction pattern frame; the processing module is also used to: perform a Fourier transform on the time-series signal of each pixel in the time-series diffraction pattern frame, and separate the first frequency signal contributed only by the interface where the surface pattern of the optical element is located and the second frequency signal contributed by the interface of the back surface of the substrate of the optical element in the generated spectrum; Furthermore, the processing module reconstructs a parasitic interference-free diffraction pattern based on the first frequency signal, and uses the parasitic interference-free diffraction pattern to determine the defect location.

6. The structured light testing system for optical element surface patterns according to claim 1, characterized in that, The processing module is also used to: generate a defect distribution map containing spatial information of all identified defect locations; and perform spatial statistical analysis on the defect distribution map to identify whether multiple defect locations have feature signatures associated with process fault types in their spatial distribution, and output indication information associated with preset process problem sources based on the feature signatures.

7. The structured light testing system for optical element surface patterns according to claim 6, characterized in that, The processing module performs spatial statistical analysis on the defect distribution map by executing a two-dimensional Fourier transform to detect whether there are periodic arrangement feature signatures that appear as isolated characteristic frequency peaks in the frequency domain; and the processing module also performs spatial statistical analysis on the defect distribution map by executing spatial density clustering analysis.

8. The structured light testing system for optical element surface patterns according to claim 5, characterized in that, The illumination module also includes a Fabry-Perot etalon placed downstream of the optical path of the piezoelectric ceramic-driven mirror; after performing a Fourier transform on the time-series diffraction pattern frame, the processing module is also used to: identify a reference peak with a constant frequency generated in the spectrum by the Fabry-Perot etalon; and, based on the actual peak shape and center frequency position of the reference peak, to perform real-time correction of the frequency axis distortion introduced by the nonlinear motion of the piezoelectric ceramic-driven mirror and the optical path drift caused by changes in ambient temperature.

9. The structured light testing system for optical element surface patterns according to claim 5, characterized in that, After separating the first frequency signal, the processing module is also used to: extract the amplitude information of the first frequency signal to generate a diffraction pattern without parasitic interference; and extract the phase information of the first frequency signal in parallel, and generate a height distribution map characterizing the three-dimensional morphology of the surface pattern of the optical element by performing phase unwrapping processing on the phase information.

10. The structured light testing system for optical element surface patterns according to claim 1, characterized in that, The processing module performs intensity correction on the first and second diffraction patterns in the following way: it calculates the ratio of the total light intensity of the first in-situ light field intensity distribution image to the total light field intensity distribution image of the second in-situ light field intensity distribution image, and uses this ratio as a scale factor to multiply the intensity of all pixels in the second diffraction pattern to generate the corrected second diffraction pattern.

Images