Planar optical element scatter budget gated acceptance method and system

Abstract

The application relates to the technical field of etching process control, and discloses a planar optical element scattering budget gated acceptance method and system, computer equipment and a storage medium. The method comprises the following steps: acquiring a microstructure sidewall roughness parameter and a surface roughness parameter of a target planar optical element; acquiring a multi-component identity of an etching cavity; inputting the roughness parameter and the multi-component identity into a multi-source heterogeneity scattering budget model to calculate an expected scattering loss value; calculating a preset scattering loss threshold value according to a microstructure aspect ratio and a material absorption coefficient; comparing the expected scattering loss value with the threshold value to output a qualified or unqualified determination result; when the determination result is unqualified, the expected scattering loss value is decomposed into a fixed effect component and a plurality of independent random effect components through variance decomposition, and a maintenance suggestion for the corresponding component is generated. The application realizes accurate prediction of scattering loss and component-level tracing of process abnormalities, and improves acceptance accuracy and abnormal troubleshooting efficiency.

Description

Technical Field

[0001] This application relates to the field of etching process control technology, specifically a method, system, computer equipment, and storage medium for gating and accepting scattering budgets of planar optical elements. Background Technology

[0002] With the rapid development of consumer electronics products such as augmented reality and smart glasses, fully planar, ultra-thin prescription optical components (such as Fresnel lenses and diffractive optical components) have become an important development direction. These components typically employ grayscale photolithography and inductively coupled plasma (ICP) etching processes to form micron- or submicron-scale microstructures on silicon substrates, which are then replicated onto transparent substrates using techniques such as nanoimprint lithography. During the etching process, the microscopic roughness of the sidewalls and surface of the microstructures directly causes scattering loss, leading to a decrease in the modulation transfer function (MTF) and an increase in stray light, ultimately affecting the optical imaging quality.

[0003] Currently, for quality acceptance of such microstructures, the industry typically uses offline roughness measurement (such as atomic force microscopy and white light interferometer) to obtain the root mean square roughness parameters of the sidewalls and surfaces, and then combines them with empirical thresholds to determine whether they are qualified or not. However, in actual production, it has been found that even if the roughness parameters meet the same specifications, the actual scattering loss of components processed by different etching cavities and different batches may still have significant differences. Moreover, when defective products occur, it is difficult to trace which specific component in the etching cavity (such as RF power supply, gas distribution plate, cavity bushing, or electrostatic chuck) has drifted based solely on the roughness value. This results in long process problem troubleshooting time, low maintenance efficiency, and restricts the stability and yield improvement of mass production. Summary of the Invention

[0004] Therefore, it is necessary to provide a planar optical element scattering budget gating acceptance method, system, computer equipment, and storage medium that can distinguish the physical sources of multi-source errors, perform scattering budget gating judgment, and output component-level maintenance suggestions to address the above-mentioned technical problems, so as to improve the consistency of mass production and the efficiency of anomaly detection.

[0005] Firstly, a method for gating and accepting scattering budgets of planar optical elements is provided, the method comprising: Obtain the microstructure sidewall roughness parameters and surface roughness parameters of the target planar optical element; Obtain the multi-component identification of the etching cavity used to process the target planar optical element; The microstructure sidewall roughness parameters, the surface roughness parameters, and the component identification are input into a multi-source heterogeneous scattering budget model to calculate the expected scattering loss value of the target planar optical element, wherein the multi-source heterogeneous scattering budget model includes independent random effect terms associated with each of the component identifications; Calculate the preset scattering loss threshold value based on the aspect ratio of the microstructure and the absorption coefficient of the material of the target planar optical element; When the expected scattering loss value is less than or equal to the preset scattering loss threshold, a pass / fail result is output; when the expected scattering loss value is greater than the preset scattering loss threshold, a fail / fail result is output, and the expected scattering loss value is decomposed into a fixed effect component and multiple independent random effect components through variance decomposition, wherein each independent random effect component is associated with a component identification identifier; maintenance suggestions for each corresponding component are generated based on the values ​​of each independent random effect component.

[0006] In one embodiment, the process of constructing the multi-source heterogeneous scattering budget model includes: The microstructure sample sidewall roughness parameters, sample surface roughness parameters, and measured scattering loss values ​​of calibration samples processed in multiple different etching cavity samples under multiple different processing batches were obtained, as well as the multi-sample component identification of each etching cavity sample. Using the sidewall roughness parameters and surface roughness parameters of the microstructure sample as fixed-effect input variables, independent random-effect terms are constructed for the identification of each sample component. The measured scattering loss value of the sample is used as the output variable. The fixed-effect coefficients and the variances of each random-effect term are obtained by regression fitting through a multi-factor linear mixed-effect model. The fixed effect coefficients and variances of each random effect term obtained from the fitting are stored in the model library and associated with the identity identifiers of each sample component.

[0007] In one embodiment, the step of decomposing the expected scattering loss value into a fixed effect component and multiple independent random effect components through variance decomposition includes: Substitute the microstructure sidewall roughness parameters and the surface roughness parameters into the fixed effect function composed of the fixed effect coefficients to calculate the fixed effect components; Subtract the fixed effect component from the expected scattering loss value to obtain the total random effect component; Based on the proportion of the variance of each random effect term to the total random effect variance, the total random effect component is decomposed into independent random effect components associated with the identity identifiers of each component through Bayesian inference.

[0008] In one embodiment, the step of decomposing the total random effects component into independent random effects components associated with the identity identifiers of each component through Bayesian inference includes: The proportion of the variance of each random effect term to the total random effect variance is used as the prior weight to construct the prior distribution of the contribution of each component; Using the total random effects component as the observed data, the posterior distribution is sampled using the Markov chain Monte Carlo method; The posterior expected value of the contribution of each component is extracted from the posterior distribution and used as the independent random effect component associated with the identity of each component.

[0009] In one embodiment, obtaining the microstructure sidewall roughness parameters and surface roughness parameters of the target planar optical element includes: The Mueller matrix spectrum of the microstructure region of the target planar optical element was acquired at multiple incident angles using a Mueller matrix ellipsometer, and the root mean square roughness, correlation length, skewness coefficient and kurtosis coefficient of the microstructure sidewall were extracted as the roughness parameters of the microstructure sidewall. The complex amplitude distribution of the surface morphology of the unstructured region of the target planar optical element is acquired using a digital holographic microscope. The filtered root mean square roughness, correlation length, skewness coefficient, and kurtosis coefficient are then extracted as the surface roughness parameters.

[0010] In one embodiment, extracting the root mean square roughness, correlation length, skewness coefficient, and kurtosis coefficient of the microstructure sidewalls as the roughness parameters of the microstructure sidewalls includes: The Mueller matrix spectra acquired at each incident angle are subjected to wavelength-separated inversion to obtain the root mean square roughness of the sidewalls, correlation length, skewness coefficient, and kurtosis coefficient corresponding to each wavelength. The average value of the inversion results for each wavelength is used as the sidewall roughness parameter of the microstructure.

[0011] In one embodiment, calculating a preset scattering loss threshold based on the microstructure aspect ratio and material absorption coefficient of the target planar optical element includes: The preset scattering loss threshold value is calculated according to the following formula: in, To preset the scattering loss threshold, As the baseline threshold value, The aspect ratio of the microstructure, Based on the aspect ratio, The aspect ratio sensitivity index, The absorption coefficient and sensitivity coefficient are, The absorption coefficient of the material.

[0012] In one embodiment, before inputting the microstructure sidewall roughness parameters, the surface roughness parameters, and the multi-component identification into the multi-source heterogeneous scattering budget model, the method further includes: Query the component lifespan database to obtain the cumulative operating parameters of each component; When the cumulative operating parameters of any component exceed the corresponding lifespan threshold, the variance of the random effects term corresponding to that component is increased.

[0013] In one embodiment, generating maintenance recommendations for each corresponding component based on the values ​​of each independent random effect component includes: When the RF power supply contribution exceeds the preset RF contribution threshold, an RF power supply impedance matching network calibration suggestion is generated and the suggested calibration offset is output. When the contribution of the gas distribution plate exceeds the preset gas plate contribution threshold, a gas flow uniformity detection suggestion is generated for the gas distribution plate, and the gas pipeline number to be detected is output. When the contribution of the cavity bushing exceeds the preset bushing contribution threshold, a cavity bushing replacement suggestion is generated, and the remaining service life of the bushing is predicted based on the time-series change rate of the cavity bushing contribution. When the contribution of the electrostatic chuck exceeds the preset chuck contribution threshold, an electrostatic chuck temperature control calibration suggestion is generated, and the suggested temperature control parameter adjustment value is output.

[0014] Secondly, a scattering budget gating acceptance system for planar optical elements is provided, the system comprising: The parameter acquisition module is used to acquire the microstructure sidewall roughness parameters and surface roughness parameters of the target planar optical element; The identity acquisition module is used to acquire the identity identifiers of multiple components of the etching cavity used to process the target planar optical element; The model calculation module is used to input the microstructure sidewall roughness parameters, the surface roughness parameters, and the multi-component identification into the multi-source heterogeneous scattering budget model to calculate the expected scattering loss value of the target planar optical element, wherein the multi-source heterogeneous scattering budget model includes independent random effect terms associated with each of the component identifications; The threshold calculation module is used to calculate a preset scattering loss threshold value based on the aspect ratio of the microstructure and the absorption coefficient of the material of the target planar optical element. The judgment output module is used to output a qualified judgment result when the expected scattering loss value is less than or equal to the preset scattering loss threshold value; otherwise, it outputs a unqualified judgment result. The suggestion generation module is used to decompose the expected scattering loss value into a fixed effect component and multiple independent random effect components through variance decomposition when the expected scattering loss value is greater than the preset scattering loss threshold value. Each of the independent random effect components is associated with a component identification identifier. Maintenance suggestions are generated for each corresponding component based on the values ​​of each independent random effect component.

[0015] Thirdly, a computer device is provided, including a memory and a processor, the memory being communicatively connected to the processor, and the memory storing a computer program executable on the processor, wherein the processor, when executing the computer program, implements a planar optical element scattering budget gating acceptance method as described above.

[0016] Fourthly, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described planar optical element scattering budget gating acceptance method.

[0017] The aforementioned method, system, computer equipment, and storage medium for gating and accepting planar optical elements based on scattering budgets achieve accurate prediction of scattering loss by acquiring the microstructure sidewall and surface roughness parameters of the target planar optical element and combining them with the multi-component identification of the etching cavity. A multi-source heterogeneous scattering budget model containing independent random effect terms is used to calculate the expected scattering loss value, and the prediction results are compared with a threshold value dynamically calculated based on the microstructure aspect ratio and material absorption coefficient to automatically determine product qualification. When the predicted value exceeds the limit, variance decomposition separates the total deviation into independent random effect components associated with each component, thereby accurately locating the component source causing the scattering exceedance and generating targeted maintenance recommendations. Therefore, this application elevates process acceptance from a single roughness judgment to refined gating based on scattering budgets, achieving component-level traceability and precise maintenance of process deviations, improving the acceptance accuracy of planar optical elements and the efficiency of process anomaly troubleshooting, and effectively ensuring the consistency and yield of mass production. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a planar optical element scattering budget-gated acceptance method in one embodiment; Figure 2 This is a block diagram of a planar optical element scattering budget gated acceptance system in one embodiment. Detailed Implementation

[0019] To facilitate understanding of the technical solutions provided in the embodiments of this application, the background technology involved in the embodiments of this application will be described below.

[0020] With the rapid development of consumer electronics products such as augmented reality and smart glasses, fully planar, ultra-thin prescription optical components (such as Fresnel lenses and diffractive optical components) have become an important development direction. These components typically employ grayscale photolithography to form a three-dimensional photoresist outline on a silicon substrate, then use inductively coupled plasma (ICP) etching to transfer the photoresist outline onto the silicon substrate to form micron- or submicron-scale microstructures. Finally, they are replicated onto a transparent substrate using techniques such as nanoimprint lithography. In this process, the quality of the etching step directly determines the performance of the final component.

[0021] During ICP etching, the micro-roughness of the sidewalls and surface of microstructures directly causes scattering loss, leading to a decrease in modulation transfer function (MTF) and an increase in stray light, ultimately affecting optical imaging quality. Because the etching process involves multiple physical field couplings (plasma density, ion energy, gas distribution, temperature field, etc.), and different etching cavities and component conditions (such as impedance matching of the RF power supply, flow uniformity of the gas distribution disk, deposition state of the cavity bushing, and temperature control accuracy of the electrostatic chuck) all affect the etching results, even with the same process formulation, the actual scattering loss of components processed in different cavities and batches may still vary significantly.

[0022] Currently, for quality acceptance of such microstructures, the industry typically uses offline roughness measurement (such as atomic force microscopy and white light interferometer) to obtain the root mean square roughness parameters of the sidewalls and surfaces, and then combines them with empirical thresholds to determine whether they are qualified or not. However, this acceptance method has two problems: First, the roughness parameters can only reflect the amplitude characteristics of the surface morphology and are difficult to comprehensively characterize scattering loss. Moreover, the existing acceptance thresholds are usually fixed values ​​and do not consider the influence of different aspect ratios of microstructures and different absorption coefficients of materials on scattering sensitivity, resulting in a weak correlation between the qualification judgment and the actual optical performance of the product. Second, when non-conforming products are found, the roughness values ​​alone cannot distinguish whether the excessive scattering is caused by the sidewall roughness itself or by the drift of a component in the etching cavity (such as the RF power supply, gas distribution plate, cavity bushing, or electrostatic chuck). Troubleshooting process anomalies often requires testing each component one by one, which is time-consuming and seriously affects mass production efficiency and yield stability.

[0023] Therefore, in order to achieve accurate prediction of scattering loss during etching process acceptance and to quickly locate specific drifting components when anomalies occur, this application provides a scattering budget gating acceptance method, system, computer equipment, and storage medium for planar optical elements.

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. Furthermore, the data collection, storage, and use operations involved in the embodiments of this application are all performed in compliance with relevant laws and regulations and with appropriate authorization, ensuring that user privacy and device data security are not infringed.

[0025] Firstly, in existing technologies, when conducting quality acceptance testing of planar optical components, only roughness parameters are typically measured and compared with a fixed threshold. This fails to establish an accurate mapping relationship between roughness parameters and scattering loss, and when defective products are found, the specific reasons for excessive scattering cannot be distinguished, leading to low efficiency in troubleshooting process anomalies. Therefore, in one embodiment, a scattering budget gating acceptance method for planar optical components is provided, such as... Figure 1 As shown, the method includes: Step S1: Obtain the microstructure sidewall roughness parameters and surface roughness parameters of the target planar optical element.

[0026] Specifically, the target planar optical element refers to a planar optical element that has been processed by inductively coupled plasma etching and is awaiting quality acceptance. The microstructure sidewall roughness parameter is used to characterize the microscopic morphology of the microstructure sidewall of the element, while the surface roughness parameter is used to characterize the surface microscopic morphology of the unstructured region of the element (i.e., the flat region where no microstructure has been formed).

[0027] Step S2: Obtain the multi-component identification of the etching cavity used to process the target planar optical element.

[0028] Specifically, the etching cavity refers to an inductively coupled plasma etching (ICP) device used to perform etching of target planar optical components. Multi-component identification includes unique identification information for several key components associated with the etching cavity. These key components may include, but are not limited to, the RF power supply, gas distribution disk, cavity bushing, and electrostatic chuck. The RF power supply provides the RF energy required to generate the plasma, and its serial number serves as the component's identification. The gas distribution disk distributes the etching gas evenly within the cavity, and its serial number serves as the component's identification. The cavity bushing protects the inner wall of the cavity and influences plasma distribution; its serial number serves as the component's identification. The electrostatic chuck is used to fix the wafer and control its temperature, and its serial number serves as the component's identification. The purpose of obtaining these component identifications is to differentiate the contributions of different components to scattering loss in subsequent steps.

[0029] Step S3: Input the microstructure sidewall roughness parameters, surface roughness parameters, and component identification into the multi-source heterogeneous scattering budget model to calculate the expected scattering loss value of the target planar optical element. The multi-source heterogeneous scattering budget model includes independent random effect terms associated with each component identification.

[0030] Specifically, the multi-source heterogeneous scattering budget model is a pre-constructed mathematical model that uses microstructure sidewall roughness parameters and surface roughness parameters as fixed-effect inputs (i.e., the influence of these parameters on scattering loss is deterministic and quantifiable), and uses the identity of each component as an independent random-effect factor (i.e., the influence of each component on scattering loss contains a random component, reflecting the differences between individual components and the uncertainty caused by component state drift). By substituting the measured roughness parameters and component identity into the model, the expected scattering loss value of the target planar optical element can be calculated. This value reflects the level of scattering loss that the element may produce under given roughness parameters and current component state.

[0031] Step S4: Calculate the preset scattering loss threshold value based on the aspect ratio of the microstructure of the target planar optical element and the material absorption coefficient.

[0032] Specifically, the aspect ratio of a microstructure refers to the ratio of its height to its width, and the material absorption coefficient refers to the absorption capacity of the material constituting the target planar optical element for incident light. Since microstructures with different aspect ratios are less sensitive to scattering, and materials with different absorption coefficients have different tolerances to scattering loss, this embodiment dynamically calculates the threshold value based on these two parameters, rather than using a fixed threshold. This threshold value serves as the criterion for determining whether the structure is qualified or not.

[0033] Step S5: When the expected scattering loss value is less than or equal to the preset scattering loss threshold, output a qualified judgment result; when the expected scattering loss value is greater than the preset scattering loss threshold, output a unqualified judgment result, and decompose the expected scattering loss value into a fixed effect component and multiple independent random effect components through variance decomposition, wherein each independent random effect component is associated with a component identification; generate maintenance suggestions for each corresponding component based on the value of each independent random effect component.

[0034] Specifically, the fixed effect component represents the scattering loss contributed by the microstructure sidewall roughness parameters and surface roughness parameters, reflecting the influence of the component's morphology on scattering. The independent random effect component represents the scattering loss contributed by the differences in the states of each component, such as the impedance matching state of the RF power supply, the flow uniformity of the gas distribution plate, the deposition level of the cavity bushing, and the temperature control accuracy of the electrostatic chuck. Through variance decomposition, the portion of the total expected scattering loss exceeding the fixed effect component can be allocated to each component according to the variance weight of each random effect term, thus obtaining the contribution of each component to the scattering exceedance. Based on these contribution values, targeted maintenance recommendations can be generated; for example, when the RF power supply contribution is high, it is recommended to calibrate the impedance matching network; when the gas distribution plate contribution is high, it is recommended to check the flow uniformity of each gas pipeline.

[0035] Based on the above, this embodiment incorporates roughness parameters and component identification into scattering prediction using a multi-source heterogeneous scattering budget model, achieving accurate prediction of scattering loss. Through dynamic threshold calculation based on microstructure aspect ratio and material absorption coefficient, it closely links acceptance criteria with the actual optical performance of the product. Furthermore, by using variance decomposition to isolate the causes of scattering exceedances to specific components, it enables precise tracing of process anomalies and targeted maintenance. This solves the technical problems of low acceptance accuracy and difficulty in anomaly investigation in existing technologies.

[0036] Accurately obtaining the roughness parameters of the microstructure sidewalls and surfaces is fundamental for subsequent scattering prediction. Existing technologies typically employ a single measurement method to obtain roughness parameters, often measuring only the root mean square roughness, neglecting topographic features such as correlation length, skewness coefficient, and kurtosis coefficient, which significantly influence scattering. Therefore, in one embodiment, obtaining the microstructure sidewall roughness parameters and surface roughness parameters of the target planar optical element includes: The Mueller matrix spectrum of the microstructure region of the target planar optical element was acquired at multiple incident angles using a Mueller matrix ellipsometer. The root mean square roughness, correlation length, skewness coefficient and kurtosis coefficient of the microstructure sidewall were extracted as microstructure sidewall roughness parameters. The complex amplitude distribution of the surface morphology of the unstructured region of the target planar optical element was acquired using a digital holographic microscope. The filtered root mean square roughness, correlation length, skewness coefficient, and kurtosis coefficient were then extracted as surface roughness parameters.

[0037] Specifically, a Mueller matrix ellipsometer is an optical measurement device capable of measuring the Mueller matrix of a sample. The Mueller matrix contains 16 elements and comprehensively describes the sample's transformation characteristics for polarized light. Since the roughness of the microstructure sidewalls affects the scattering characteristics of polarized light, by measuring the Mueller matrix spectra at different incident angles and matching or inverting them with theoretical models, the root mean square roughness (characterizing amplitude), correlation length (characterizing lateral correlation), skewness coefficient (characterizing the asymmetry of the morphology distribution), and kurtosis coefficient (characterizing the sharpness of the morphology distribution) of the sidewalls can be extracted. These four parameters collectively describe the statistical characteristics of the sidewall morphology, providing a more comprehensive reflection of the sidewall's influence on scattering compared to using only the root mean square roughness. Specific models of Mueller matrix ellipsometers can be the Woollam RC2 series or equivalent devices.

[0038] Secondly, a digital holographic microscope is a three-dimensional topography measurement device based on the principle of interference. It can acquire the complex amplitude distribution of a sample surface and then reconstruct its three-dimensional topography. Unstructured regions refer to flat areas on the target planar optical element where no microstructures have formed, typically located at the edges of a microstructure array or specific test points. The acquired surface topography complex amplitude distribution is processed by wavelet transform or Gaussian filtering to separate the roughness component (high-frequency component) and waviness component (mid-frequency component). Then, statistical calculations are performed on the filtered roughness data to obtain the root mean square roughness, correlation length, skewness coefficient, and kurtosis coefficient. Specific models of digital holographic microscopes can be the Lyncée Tec DHM series or equivalent devices.

[0039] Based on the above, this embodiment uses two complementary measurement techniques, Mueller matrix ellipsometer and digital holographic microscope, to obtain multi-dimensional roughness parameters of the microstructure sidewalls and surface, providing more comprehensive and accurate input data for subsequent scattering prediction.

[0040] Furthermore, the Mueller matrix spectrum with multiple incident angles can be denoised and the instrument response corrected first, and then the sidewall roughness statistical parameters can be obtained by combining multi-angle joint inversion. After removing the plane tilt term, spatial bandpass filtering and outlier removal from the surface complex amplitude distribution obtained by digital holographic microscopy, the root mean square roughness, correlation length, skewness coefficient and kurtosis coefficient can be calculated, thereby improving the feasibility and repeatability of parameter extraction.

[0041] When using a Mueller matrix ellipsometer for measurement, the selection of measurement conditions directly affects the accuracy and stability of the measurement results. Existing technologies often use a single incident angle or a few incident angles for measurement, making them susceptible to measurement noise and model errors. Therefore, in one embodiment, the root mean square roughness, correlation length, skewness coefficient, and kurtosis coefficient of the microstructure sidewalls are extracted as microstructure sidewall roughness parameters, including: Wavelength inversion was performed on the Mueller matrix spectra collected at each incident angle to obtain the root mean square roughness of the sidewalls, correlation length, skewness coefficient and kurtosis coefficient corresponding to each wavelength; The average value of the inversion results for each wavelength is used as the roughness parameter of the microstructure sidewall.

[0042] Specifically, the incident angle refers to the angle between the incident ray and the normal to the sample surface. This embodiment selects an incident angle within the range of 45° to 75°. This range is chosen because within this angle range, the microstructure is more sensitive to the scattering effect of polarized light, which is beneficial for extracting roughness parameters. Selecting at least five incident angles provides redundant information and reduces errors that may be introduced by measuring a single angle. For example, at least five angles from 45°, 50°, 55°, 60°, 65°, 70°, and 75° can be selected for measurement.

[0043] Secondly, wavelength-specific inversion refers to performing model matching or optimization separately for each wavelength to extract roughness parameters from the Mueller matrix data. Mueller matrix ellipsometers typically measure wavelengths from ultraviolet to near-infrared (approximately 200 nm to 1700 nm), with each wavelength corresponding to a set of Mueller matrix data. For each incident angle and each wavelength, by matching the measured Mueller matrix with a theoretical model established based on Rigorous Coupled Wave Analysis (RCWA) or the finite element method, the sidewall roughness parameters corresponding to that wavelength can be inverted. The theoretical model uses the microstructure sidewall roughness as input parameters, calculates the corresponding Mueller matrix, and iteratively optimizes to minimize the difference between the theoretical and measured Mueller matrices, thereby obtaining the inverted values ​​of the roughness parameters.

[0044] Furthermore, for each incident angle, the average value of the inversion results at each wavelength under that incident angle is first calculated; then, the overall average value of the average values ​​of each incident angle is calculated as the final microstructure sidewall roughness parameter. This multi-incident-angle, multi-wavelength averaging method can effectively reduce the influence of measurement noise and model error, and improve the stability and repeatability of the measurement results.

[0045] Based on the above, this embodiment improves the accuracy and stability of sidewall roughness parameter measurement by using multiple incident angles and multiple wavelengths for measurement and averaging.

[0046] Before applying a multi-source heterogeneous scattering budget model, the model needs to be built in advance. Existing technologies typically use calibration data from a single cavity and a single batch to build the model, which fails to reflect the differences between different cavities and components, resulting in insufficient model prediction accuracy. Therefore, in one embodiment, the process of building a multi-source heterogeneous scattering budget model includes: The microstructure sample sidewall roughness parameters, sample surface roughness parameters, and measured scattering loss values ​​of calibration samples processed in multiple different etching cavity samples under multiple different processing batches were obtained, as well as the multi-sample component identification of each etching cavity sample. Using the sidewall roughness parameters and surface roughness parameters of the microstructure samples as fixed effect input variables, independent random effect terms are constructed for the identification of each sample component. The measured scattering loss value of the samples is used as the output variable. The fixed effect coefficients and the variances of each random effect term are obtained by regression fitting through a multi-factor linear mixed effect model. The fixed effect coefficients and variances of each random effect term obtained from the fitting are stored in the model library and associated with the identity identifiers of each sample component.

[0047] Specifically, etched cavity samples refer to the etching equipment used to collect calibration data. These devices should cover different cavity types and states that may be used in actual production. Processing batches refer to processing batches performed at different times, with each batch corresponding to a set of process parameters and cavity states. Calibration samples refer to test samples specifically used for model calibration, whose materials and microstructure design are the same as or similar to the target planar optical element. The microstructure sample sidewall roughness parameters and sample surface roughness parameters are the roughness parameters obtained by measuring the calibration sample. The measured scattering loss value of the sample refers to the actual scattering loss value of the calibration sample obtained by standard methods such as the integrating sphere method and angle-resolved scattering method. The multi-sample component identification refers to the RF power supply serial number, gas distribution disk serial number, cavity bushing number, and electrostatic chuck serial number of each etched cavity sample.

[0048] Secondly, the multi-factor linear mixed-effects model can be expressed as: in, The measured scattering loss value of the calibration sample obtained by processing the i-th RF power supply, j-th gas distribution disk, k-th cavity bushing, and l-th electrostatic chuck in the m-th batch; , , , These are the root mean square roughness, correlation length, skewness coefficient, and kurtosis coefficient of the microstructure sidewalls, respectively. , , , These are the root mean square roughness, correlation length, skewness coefficient, and kurtosis coefficient of the surface, respectively. to This is the fixed effects coefficient; For the random effects term associated with the i-th RF power supply sequence number; For the random effects term associated with the j-th gas distribution disk sequence number; For the random effects term associated with the number of the k-th cavity bushing; For the random effects term associated with the sequence number of the l-th electrostatic chuck; These are the residual terms. Each random effect term follows a normal distribution with a mean of zero, and their variances are respectively... , , , These variances reflect the degree of fluctuation in the contribution of each component to scattering loss.

[0049] By collecting data from multiple cavities, multiple batches, and multiple calibration samples, and substituting them into the above model for regression fitting, the fixed effects coefficients can be obtained. to The estimated values, and the variances of each random effects term. , , , The estimated value.

[0050] Furthermore, the model library can be a database or configuration file used to store model parameters for later use. Fixed effects coefficients are used to construct the fixed effects function, and the variance of the random effects term is used for subsequent variance decomposition. Associating model parameters with component identifiers allows for querying the corresponding random effects term variance based on the component identifier of the current cavity during subsequent applications.

[0051] Based on the above, this embodiment collects calibration data from multiple different cavities and multiple different batches, and uses a multi-factor linear mixed effect model for fitting, so that the model can effectively distinguish the independent contributions of different components to scattering loss, providing a foundation for subsequent variance decomposition and component-level traceability.

[0052] After obtaining the expected scattering loss value, it is necessary to separate the portion contributed by the roughness parameter from the portion contributed by each component in order to determine the specific source of the excessive scattering. Existing technologies typically cannot perform this separation, making it impossible to pinpoint the specific problematic component. Therefore, in one embodiment, variance decomposition is used to decompose the expected scattering loss value into a fixed effect component and multiple independent random effect components, including: Substitute the microstructure sidewall roughness parameters and surface roughness parameters into the fixed effect function composed of fixed effect coefficients to calculate the fixed effect components; Subtract the fixed effect component from the expected scattering loss value to obtain the total random effect component; Based on the proportion of the variance of each random effect term to the total random effect variance, Bayesian inference decomposes the total random effect component into independent random effect components associated with the identity identifiers of each component.

[0053] Specifically, the form of the fixed effects function is determined by the fixed effects coefficients obtained from the aforementioned fitting, as follows: in, For fixed effect components, it represents the scattering loss contributed by the roughness parameter. , , , These are the root mean square roughness of the microstructure sidewalls, correlation length, skewness coefficient, and kurtosis coefficient of the target planar optical element, respectively. , , , These represent the root mean square surface roughness, correlation length, skewness coefficient, and kurtosis coefficient of the target planar optical element, respectively.

[0054] Secondly, the total random effects component represents the scattering loss contributed by the differences in the states of each component and unobservable factors. The specific calculation formula is as follows: in, The expected scattering loss value obtained from the aforementioned calculation, This represents the total random effects component.

[0055] Furthermore, the variance of the total random effects component is equal to the sum of the variances of each random effects term, i.e.: The proportions of the variance of each random effect term to the total random effect variance are as follows: , , , These proportions reflect the expected percentage contribution of each component to the total random effect. Through Bayesian inference, these proportions are used as prior information, and the total random effect components are... As observational data, the independent random effects components of the actual contributions of each component can be inferred. (Corresponding RF power supply) (Corresponding gas distribution plate) (Corresponding cavity bushing) The posterior distribution (corresponding to the electrostatic chuck) is used to obtain the estimated contribution of each component.

[0056] Based on the above, this embodiment calculates the scattering component of roughness contribution using a fixed effect function, obtains the total random effect through subtraction, and then decomposes the total random effect into the contribution of each component through Bayesian inference, thus achieving accurate attribution of the cause of scattering exceeding the standard.

[0057] Bayesian inference is a statistical inference method based on Bayes' theorem, which can combine prior information and observed data to obtain the posterior distribution. In this embodiment, the prior information comes from the proportion of the variance of each random effect term to the total random effect variance, and the observed data are the components of the total random effect. Existing technologies typically only use maximum likelihood estimation to obtain point estimates, which cannot quantify the uncertainty of the contribution of each component. Therefore, in one embodiment, Bayesian inference decomposes the total random effect component into independent random effect components associated with the identity identifiers of each component, including: The prior distribution of the contribution of each component is constructed by using the proportion of the variance of each random effect term to the total variance of random effects as the prior weight. Using the total random effects component as the observed data, the posterior distribution was sampled using the Markov chain Monte Carlo method. The posterior expected value of the contribution of each component is extracted from the posterior distribution and used as an independent random effect component associated with the identity of each component.

[0058] Specifically, assuming the contribution of each component , , , It follows a mean of zero and variances of . , , , The components of the total random effects follow a normal distribution and are independent. Follows a pattern with a mean of zero and a variance of . The normal distribution. Given... Under the given conditions, the conditional distribution of the contribution of each component is a normal distribution, and its mean and variance can be calculated using Bayes' theorem.

[0059] Secondly, the Markov chain Monte Carlo method is a numerical method for sampling from complex distributions. In this embodiment, Gibbs sampling or Hamiltonian Monte Carlo methods are used to perform multiple samplings from the posterior distribution of the contribution of each component, generating a sampling sequence. The number of samplings can be set to 2000 to 10000, with the first 500 samples discarded as a warm-up phase, and the remaining samples used for subsequent statistics.

[0060] Next, the sample mean of the contribution of each component in the sampling sequence is calculated as the posterior expected value of the contribution of that component. This estimated value is the aforementioned independent random effect component, which reflects the scattering loss value most likely to be contributed by each component under the given total random effect component and the prior variance of each component.

[0061] Based on the above, this embodiment decomposes the total random effect component into the contribution of each component through Bayesian inference, which not only provides point estimates but also implies quantitative information on uncertainty, providing a more reliable basis for the generation of subsequent maintenance suggestions.

[0062] A preset scattering loss threshold is used to determine whether a target planar optical element is qualified. Existing technologies typically use a fixed threshold, which cannot accommodate the influence of different microstructure aspect ratios and different material absorption coefficients on scattering sensitivity, resulting in a threshold that is either too strict or too lenient. Therefore, in one embodiment, a preset scattering loss threshold is calculated based on the microstructure aspect ratio and material absorption coefficient of the target planar optical element, including: The preset scattering loss threshold value is calculated using the following formula: in, To preset the scattering loss threshold, As the baseline threshold value, The aspect ratio of the microstructure, Based on the aspect ratio, The aspect ratio sensitivity index, The absorption coefficient and sensitivity coefficient are, The absorption coefficient of the material.

[0063] Specifically, the preset scattering loss threshold represents the maximum allowable scattering loss value under the current microstructure aspect ratio and material absorption coefficient. The reference threshold is based on a reference aspect ratio. The threshold value determined under the condition that the absorption coefficient is zero can be obtained through experimental calibration. The aspect ratio of the microstructure, i.e., the ratio of the height to the width of the microstructure, can be measured using a scanning electron microscope or atomic force microscope. Typical reference aspect ratios are usually 1:1 or 2:1. The aspect ratio sensitivity index reflects the sensitivity of scattering loss to changes in aspect ratio and can be obtained through fitting experimental data, with a typical range of 0.5 to 2. The absorption coefficient sensitivity index reflects the sensitivity of scattering loss to changes in absorption coefficient and can be obtained through fitting experimental data. The material absorption coefficient can be measured using an ellipsometry or spectrophotometer.

[0064] The physical meaning of the formula for calculating the preset scattering loss threshold is as follows: As the aspect ratio increases, the sidewall area of ​​the microstructure increases, and the scattering effect intensifies. Therefore, the preset scattering loss threshold should be reduced accordingly, hence the use of a power-law attenuation factor. Conversely, as the material absorption coefficient increases, the effective transmitted signal weakens, and the system becomes more sensitive to additional scattering. To ensure imaging quality, the preset scattering loss threshold should also be reduced accordingly, hence the use of an exponential attenuation factor. The combined effect of these two factors allows the preset scattering loss threshold to adaptively change with the microstructure characteristics and material properties.

[0065] In one specific implementation, the above threshold calculation results are only used to perform gating judgment and release / interception decisions on the expected scattering loss value, and serve as the triggering basis for subsequent component-level maintenance recommendations, rather than being directly used to generate real-time compensation amounts.

[0066] Based on the above, this embodiment uses a dynamic threshold function based on the power-law decay of the aspect ratio and the exponential decay of the absorption coefficient to enable the acceptance criteria to adaptively adapt to different microstructure designs and material properties, thereby improving the accuracy and rationality of the acceptance process.

[0067] Before inputting the measured roughness parameters into the model, the variance of the random effects term can be corrected using information from the component lifetime database, enabling the model to reflect the impact of component aging on scattering loss. Existing technologies typically do not consider component lifetime information, causing the model to be unable to adapt to performance drift caused by component aging. Therefore, in one embodiment, before inputting the microstructure sidewall roughness parameters, surface roughness parameters, and multi-component identification into the multi-source heterogeneous scattering budget model, the following method is also included: Query the component lifespan database to obtain the cumulative operating parameters of each component; When the cumulative operating parameters of any component exceed the corresponding lifespan threshold, the variance of the random effects term corresponding to that component is increased.

[0068] Specifically, the component lifespan database is a database used to record the operating history and lifespan information of each etching chamber component. Cumulative operating parameters include: cumulative RF power supply runtime (in hours), cumulative gas throughput of the gas distribution plate (in standard liters), cumulative etching time of the chamber bushing (in hours), and cumulative number of times the electrostatic chuck has been used. These parameters can be obtained from the equipment control system or maintenance management system.

[0069] Secondly, the lifespan threshold is a critical value determined based on the component's design lifespan and actual usage experience. For example, the lifespan threshold for an RF power supply can be set to 5000 hours, for a gas distribution plate to 10000 standard liters, for a cavity bushing to 2000 hours, and for an electrostatic chuck to 5000 cycles. When the cumulative operating parameters exceed the lifespan threshold, it indicates that the component may have entered the aging stage, and its performance fluctuations may increase. Therefore, it is necessary to increase the variance of the random effects term corresponding to the component to reflect this uncertainty. Specifically, the variance can be multiplied by a coefficient greater than 1, such as 1.5 or 2.0. The magnitude of the coefficient can be determined based on the correlation between the component's aging degree and performance fluctuations.

[0070] Based on the above, this embodiment uses component lifespan database information to dynamically adjust the variance of the random effects term before model input, enabling the model to adapt to performance changes caused by component aging and improving the model's prediction accuracy.

[0071] After obtaining the values ​​of each independent random effect component through variance decomposition, these values ​​need to be transformed into specific and actionable maintenance recommendations. Existing technologies can only provide general process anomaly alerts and cannot guide specific maintenance operations. Therefore, in one embodiment, maintenance recommendations for each corresponding component are generated based on the values ​​of each independent random effect component, including: When the RF power supply contribution exceeds the preset RF contribution threshold, an RF power supply impedance matching network calibration suggestion is generated and the suggested calibration offset is output. When the contribution of the gas distribution plate exceeds the preset gas plate contribution threshold, a gas flow uniformity detection suggestion is generated for the gas distribution plate, and the gas pipeline number to be detected is output. When the contribution of the cavity bushing exceeds the preset bushing contribution threshold, a cavity bushing replacement suggestion is generated, and the remaining service life of the bushing is predicted based on the time-series change rate of the cavity bushing contribution. When the contribution of the electrostatic chuck exceeds the preset chuck contribution threshold, an electrostatic chuck temperature control calibration suggestion is generated, and the suggested temperature control parameter adjustment value is output.

[0072] Specifically, the contribution of the radio frequency power supply is an independent random effect component. This reflects the contribution of the RF power supply status to the excessive scattering. A preset RF contribution threshold can be set empirically, for example, to 0.1 (relative to scattering loss units). The calibration offset refers to the adjustment value required for adjustable components (such as variable capacitors) in the impedance matching network. There is a mapping relationship between this offset and the RF power supply contribution value. This mapping relationship can be pre-established through regression analysis of historical data showing the RF power supply contribution and the corresponding calibration offset. For example, for every 0.01 increase in contribution, the calibration offset increases by 0.5 degrees.

[0073] The contribution of the gas distribution disk is an independent random effect component. This reflects the contribution of the gas distribution disk status to excessive scattering. The preset gas distribution disk contribution threshold can be set to 0.08. Gas distribution disks typically have multiple gas lines (e.g., central line, edge line, auxiliary line), and the flow uniformity of each line affects the etching uniformity. The recommended gas line number to be tested can be determined by correlation analysis between the gas distribution disk contribution and the flow fluctuations of each line; that is, the line number whose flow fluctuation has the highest correlation with the gas distribution disk contribution in historical data is output.

[0074] The contribution of the cavity bushing is an independent random effect component. This reflects the contribution of the cavity bushing condition to excessive scattering. The preset bushing contribution threshold can be set to 0.12. The time-series change rate refers to the slope of the cavity bushing contribution over time, which can be obtained by linear regression calculation on historical contribution data. Based on the Weibull distribution model or linear degradation model, using the current contribution value, the time-series change rate, and the bushing failure threshold, the remaining number of processing batches or the remaining service time required for the bushing to reach the failure threshold can be predicted.

[0075] The contribution of the electrostatic chuck is an independent random effect component. This reflects the contribution of the electrostatic chuck's condition to excessive scattering. The preset chuck contribution threshold can be set to 0.1. The temperature control parameter adjustment value refers to the value that needs to be adjusted for the temperature setpoint of each temperature control zone (e.g., the center zone, the edge zone) of the electrostatic chuck. There is a mapping relationship between this adjustment value and the value of the electrostatic chuck's contribution. This mapping relationship can be pre-established through regression analysis of the electrostatic chuck's contribution and the corresponding temperature control adjustment value in historical data.

[0076] Based on the above, this embodiment transforms each independent random effect component into specific component maintenance suggestions and outputs quantified adjustment parameters, providing operators with accurate and actionable maintenance guidance and improving the efficiency of process anomaly troubleshooting and maintenance.

[0077] Secondly, to implement the aforementioned planar optical element scattering budget gating acceptance method, this embodiment provides a planar optical element scattering budget gating acceptance system. For example... Figure 2 As shown, the system includes a parameter acquisition module, an identity acquisition module, a model calculation module, a threshold calculation module, a decision output module, and a suggestion generation module. These modules can be deployed on the same industrial control computer or distributed across different computing devices or embedded systems. Data exchange between modules occurs via industrial Ethernet, serial bus, or internal process communication.

[0078] The parameter acquisition module is used to acquire the microstructure sidewall roughness parameters and surface roughness parameters of the target planar optical element. The implementation of this module includes, but is not limited to, the following two methods: First, the parameter acquisition module can be one or more measuring devices and their accompanying data acquisition software, such as a Mueller matrix ellipsometer and a digital holographic microscope. These measuring devices output the measured roughness parameters in a structured data format (such as CSV, JSON, or OPCUA protocol) through their own software interfaces, and the parameter acquisition module receives this data through a data acquisition card or network communication protocol. Second, the parameter acquisition module can also be a data processing unit stored in a computer-readable medium, which reads the measured and stored roughness parameter files from a database. In practical applications, the parameter acquisition module can also be a dedicated hardware device integrating the above-mentioned measurement functions; this embodiment does not limit this.

[0079] The identification module is used to acquire the multi-component identification identifiers of the etching cavity used to process the target planar optical element. This module can be implemented in ways including, but not limited to: The identification module connects to the database interface of the Manufacturing Execution System (MES) and, based on the batch number or wafer number of the target planar optical element, queries and obtains the RF power supply serial number, gas distribution disk serial number, cavity bushing number, and electrostatic chuck serial number of the etching cavity through which the element passes. Alternatively, the identification module can communicate directly with the etching equipment control system, acquiring the aforementioned identification identifiers by reading the equipment's hardware configuration registers or scanning the equipment's QR code. The output of the identification module is structured component identification data, which can be used by subsequent modules.

[0080] The model calculation module is used to input microstructure sidewall roughness parameters, surface roughness parameters, and multi-component identifiers into a multi-source heterogeneous scattering budget model to calculate the expected scattering loss value of the target planar optical element. The multi-source heterogeneous scattering budget model includes independent random effect terms associated with each component identifier. The implementation of the model calculation module includes, but is not limited to: the model calculation module can be a software program running on a central processing unit (CPU) or graphics processing unit (GPU), which loads pre-built and stored fixed effect coefficients and variances of each random effect term in a model library. The model calculation module first queries and loads the corresponding random effect term variance from the model library based on the received multi-component identifiers; then, it substitutes the microstructure sidewall roughness parameters and surface roughness parameters into the fixed effect function to calculate the fixed effect prediction value; finally, it generates random effect prediction values ​​based on the variances of the random effect terms, and adds the fixed effect prediction value to the random effect prediction value to obtain the expected scattering loss value. The output of this module is the expected scattering loss value, which can be a floating-point number, expressed as a percentage or an absolute scattering intensity value.

[0081] The threshold calculation module is used to calculate a preset scattering loss threshold value based on the aspect ratio of the microstructure and the material absorption coefficient of the target planar optical element. The implementation of this module includes, but is not limited to: the threshold calculation module is a computing unit running on an embedded microcontroller or central processing unit, which internally stores parameters such as the reference threshold value, reference aspect ratio, aspect ratio sensitivity index, and absorption coefficient sensitivity coefficient. The threshold calculation module receives the microstructure aspect ratio and material absorption coefficient from external inputs (such as process design documents or measurement equipment) and calculates the preset scattering loss threshold value according to the formula. The output of this module is the preset scattering loss threshold value, which has the same dimensions as the expected scattering loss value.

[0082] The judgment output module outputs a pass / fail judgment result when the expected scattering loss value is less than or equal to a preset scattering loss threshold value; otherwise, it outputs a fail / fail judgment result. The implementation of this module includes, but is not limited to, a logic comparison unit, which can be implemented by a comparator circuit or by software logic running on a central processing unit. This module receives the expected scattering loss value output by the model calculation module and the preset scattering loss threshold value output by the threshold calculation module, and performs a numerical comparison operation. When the expected scattering loss value is less than or equal to the preset scattering loss threshold value, the judgment output module generates a pass signal (such as a logic high level or a "PASS" status code); when the expected scattering loss value is greater than the preset scattering loss threshold value, the judgment output module generates a fail / fail signal (such as a logic low level or a "FAIL" status code). The output of this module can be displayed through a human-machine interface, sent to the manufacturing execution system via an industrial communication protocol, or used to trigger subsequent sorting actions.

[0083] The suggestion generation module is used to decompose the expected scattering loss value into a fixed-effect component and multiple independent random-effect components through variance decomposition when the expected scattering loss value exceeds a preset scattering loss threshold. Each independent random-effect component is associated with a component identifier. Maintenance suggestions are generated for each corresponding component based on the values ​​of each independent random-effect component. The implementation of this module includes, but is not limited to: the suggestion generation module is a software program running on a central processing unit or digital signal processor, which is triggered upon receiving a non-compliance signal from the judgment output module. The suggestion generation module first calls the variance decomposition method to decompose the expected scattering loss value into a fixed-effect component and contributions from the RF power supply, gas distribution plate, cavity bushing, and electrostatic chuck. Then, based on the comparison results of each contribution value with a preset threshold, the suggestion generation module calls the maintenance suggestion generation logic to generate maintenance suggestion text or structured data containing specific quantitative parameters (such as calibration offset, gas pipeline number to be inspected, predicted remaining service life of the bushing, and temperature control parameter adjustment values). The output of this module can be presented to process engineers or maintenance personnel via a monitor, printer, or network interface, or it can be directly pushed to the maintenance work order module of the manufacturing execution system.

[0084] Based on the above, this embodiment forms a complete planar optical element scattering budget gating acceptance system. This system can automatically complete roughness parameter acquisition, scattering loss prediction, dynamic threshold calculation, qualification judgment, and component-level traceability and maintenance suggestions for non-conforming products. It realizes closed-loop automation from measurement to decision-making, improving acceptance efficiency and the accuracy of process anomaly investigation.

[0085] Additionally, it should be noted that the specific limitations of a planar optical element scattering budget gating acceptance system can be found in the limitations of a planar optical element scattering budget gating acceptance method mentioned above, and will not be repeated here.

[0086] Thirdly, a computer device is provided, including a memory and a processor. The memory is communicatively connected to the processor, and the memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the planar optical element scattering budget gating acceptance method as described above. Furthermore, the specific limitations of the computer device in implementing the planar optical element scattering budget gating acceptance method can be found in the above-described limitations of the planar optical element scattering budget gating acceptance method, and will not be repeated here.

[0087] Fourthly, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the planar optical element scattering budget gating acceptance method as described above. Furthermore, the specific limitations of the computer-readable storage medium in implementing the planar optical element scattering budget gating acceptance method can be found in the above-described limitations of the planar optical element scattering budget gating acceptance method, and will not be repeated here.

[0088] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0089] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0090] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A gating acceptance method for scattering budget of planar optical elements, characterized in that, The method includes: Obtain the microstructure sidewall roughness parameters and surface roughness parameters of the target planar optical element; Obtain the multi-component identification of the etching cavity used to process the target planar optical element; The microstructure sidewall roughness parameters, the surface roughness parameters, and the component identification are input into a multi-source heterogeneous scattering budget model to calculate the expected scattering loss value of the target planar optical element, wherein the multi-source heterogeneous scattering budget model includes independent random effect terms associated with each of the component identifications; Calculate the preset scattering loss threshold value based on the aspect ratio of the microstructure and the absorption coefficient of the material of the target planar optical element; When the expected scattering loss value is less than or equal to the preset scattering loss threshold, a pass / fail result is output; when the expected scattering loss value is greater than the preset scattering loss threshold, a fail / fail result is output, and the expected scattering loss value is decomposed into a fixed effect component and multiple independent random effect components through variance decomposition, wherein each independent random effect component is associated with a component identification identifier; maintenance suggestions for each corresponding component are generated based on the values ​​of each independent random effect component.

2. The method according to claim 1, characterized in that, The construction process of the multi-source heterogeneous scattering budget model includes: The microstructure sample sidewall roughness parameters, sample surface roughness parameters, and measured scattering loss values ​​of calibration samples processed in multiple different etching cavity samples under multiple different processing batches were obtained, as well as the multi-sample component identification of each etching cavity sample. Using the sidewall roughness parameters and surface roughness parameters of the microstructure sample as fixed-effect input variables, independent random-effect terms are constructed for the identification of each sample component. The measured scattering loss value of the sample is used as the output variable. The fixed-effect coefficients and the variances of each random-effect term are obtained by regression fitting through a multi-factor linear mixed-effect model. The fixed effect coefficients and variances of each random effect term obtained from the fitting are stored in the model library and associated with the identity identifiers of each sample component.

3. The method according to claim 2, characterized in that, The step of decomposing the expected scattering loss value into a fixed effect component and multiple independent random effect components through variance decomposition includes: Substitute the microstructure sidewall roughness parameters and the surface roughness parameters into the fixed effect function composed of the fixed effect coefficients to calculate the fixed effect components; Subtract the fixed effect component from the expected scattering loss value to obtain the total random effect component; Based on the proportion of the variance of each random effect term to the total random effect variance, the total random effect component is decomposed into independent random effect components associated with the identity identifiers of each component through Bayesian inference.

4. The method according to claim 3, characterized in that, The process of decomposing the total random effects component into independent random effects components associated with the identity identifiers of each component through Bayesian inference includes: The proportion of the variance of each random effect term to the total random effect variance is used as the prior weight to construct the prior distribution of the contribution of each component; Using the total random effects component as the observed data, the posterior distribution is sampled using the Markov chain Monte Carlo method; The posterior expected value of the contribution of each component is extracted from the posterior distribution and used as the independent random effect component associated with the identity of each component.

5. The method according to claim 1, characterized in that, The acquisition of the microstructure sidewall roughness parameters and surface roughness parameters of the target planar optical element includes: The Mueller matrix spectrum of the microstructure region of the target planar optical element was acquired at multiple incident angles using a Mueller matrix ellipsometer, and the root mean square roughness, correlation length, skewness coefficient and kurtosis coefficient of the microstructure sidewall were extracted as the roughness parameters of the microstructure sidewall. The complex amplitude distribution of the surface morphology of the unstructured region of the target planar optical element is acquired using a digital holographic microscope. The filtered root mean square roughness, correlation length, skewness coefficient, and kurtosis coefficient are then extracted as the surface roughness parameters.

6. The method according to claim 5, characterized in that, The extraction of the root mean square roughness, correlation length, skewness coefficient, and kurtosis coefficient of the microstructure sidewalls as the roughness parameters of the microstructure sidewalls includes: The Mueller matrix spectra acquired at each incident angle are subjected to wavelength-separated inversion to obtain the root mean square roughness of the sidewalls, correlation length, skewness coefficient, and kurtosis coefficient corresponding to each wavelength. The average value of the inversion results for each wavelength is used as the sidewall roughness parameter of the microstructure.

7. The method according to claim 1, characterized in that, The step of calculating a preset scattering loss threshold based on the aspect ratio of the microstructure and the material absorption coefficient of the target planar optical element includes: The preset scattering loss threshold value is calculated according to the following formula: in, To preset the scattering loss threshold, As the baseline threshold value, The aspect ratio of the microstructure, Based on the aspect ratio, The aspect ratio sensitivity index, The absorption coefficient and sensitivity coefficient are, The absorption coefficient of the material.

8. The method according to claim 1, characterized in that, Before inputting the microstructure sidewall roughness parameters, the surface roughness parameters, and the multi-component identification into the multi-source heterogeneous scattering budget model, the method further includes: Query the component lifespan database to obtain the cumulative operating parameters of each component; When the cumulative operating parameters of any component exceed the corresponding lifespan threshold, the variance of the random effects term corresponding to that component is increased.

9. The method according to any one of claims 1 to 8, characterized in that, The process of generating maintenance recommendations for each corresponding component based on the values ​​of each independent random effect component includes: When the RF power supply contribution exceeds the preset RF contribution threshold, an RF power supply impedance matching network calibration suggestion is generated and the suggested calibration offset is output. When the contribution of the gas distribution plate exceeds the preset gas plate contribution threshold, a gas flow uniformity detection suggestion is generated for the gas distribution plate, and the gas pipeline number to be detected is output. When the contribution of the cavity bushing exceeds the preset bushing contribution threshold, a cavity bushing replacement suggestion is generated, and the remaining service life of the bushing is predicted based on the time-series change rate of the cavity bushing contribution. When the contribution of the electrostatic chuck exceeds the preset chuck contribution threshold, an electrostatic chuck temperature control calibration suggestion is generated, and the suggested temperature control parameter adjustment value is output.

10. A scattering budget gating acceptance system for planar optical elements, characterized in that, The system includes: The parameter acquisition module is used to acquire the microstructure sidewall roughness parameters and surface roughness parameters of the target planar optical element; The identity acquisition module is used to acquire the identity identifiers of multiple components of the etching cavity used to process the target planar optical element; The model calculation module is used to input the microstructure sidewall roughness parameters, the surface roughness parameters, and the multi-component identification into the multi-source heterogeneous scattering budget model to calculate the expected scattering loss value of the target planar optical element, wherein the multi-source heterogeneous scattering budget model includes independent random effect terms associated with each of the component identifications; The threshold calculation module is used to calculate a preset scattering loss threshold value based on the aspect ratio of the microstructure and the absorption coefficient of the material of the target planar optical element. The judgment output module is used to output a qualified judgment result when the expected scattering loss value is less than or equal to the preset scattering loss threshold value; otherwise, it outputs a unqualified judgment result. The suggestion generation module is used to decompose the expected scattering loss value into a fixed effect component and multiple independent random effect components through variance decomposition when the expected scattering loss value is greater than the preset scattering loss threshold value. Each of the independent random effect components is associated with a component identification identifier. Maintenance suggestions are generated for each corresponding component based on the values ​​of each independent random effect component.

11. A computer device comprising a memory and a processor, the memory being communicatively connected to the processor, and the memory storing a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements a planar optical element scattering budget gating acceptance method as described in any one of claims 1 to 9.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a planar optical element scattering budget gating acceptance method as described in any one of claims 1 to 9.

Images