Adaptive air pressure control method for semiconductor polishing

Abstract

This application provides an adaptive air pressure control method for semiconductor polishing, relating to the field of polishing control technology. The method includes: acquiring a cascaded air pressure control module for a chemical mechanical polishing (CMP) device; constructing an original three-dimensional model of the semiconductor wafer to be processed and a preset three-dimensional model of the polishing target; performing spatial alignment fitting to generate a model to be polished; performing neighborhood clustering for polishing requirements at multiple points; optimizing the polishing air pressure based on multiple clustered regions and polishing requirement identification information, combined with preset polishing fluid information, to obtain multiple optimal polishing air pressure parameters for multiple clustered regions; generating air pressure control signals according to the multiple optimal polishing air pressure parameters using a first air pressure regulator for polishing control, and performing feedback optimization control through a second air pressure regulator. This application can solve the technical problem of insufficient precision in polishing parameter settings in existing technologies, achieving the technical effect of improving the accuracy and efficiency of semiconductor polishing.

Description

Technical Field

[0001] This application relates to the field of polishing control technology, and more particularly to an adaptive air pressure control method for semiconductor polishing. Background Technology

[0002] Adaptive air pressure control is designed to automatically adjust the air pressure parameters during the polishing process according to polishing needs and conditions in order to achieve the best polishing results.

[0003] Currently, existing semiconductor polishing processes do not group points with similar polishing requirements, leading to inaccurate polishing parameter settings for different areas and affecting the overall surface polishing consistency. Furthermore, setting polishing gas pressure parameters may not be optimized for specific areas, resulting in over-polishing or under-polishing in certain areas. This increases polishing time and cost, reduces overall polishing efficiency, and impacts product quality and performance. Therefore, a method is needed to address these issues.

[0004] In summary, existing technologies suffer from the technical problem that, since most of the points with similar polishing requirements are not grouped, the polishing parameters for different areas are not set precisely enough, resulting in low polishing efficiency and further affecting the quality and performance of the product. Summary of the Invention

[0005] The purpose of this application is to provide an adaptive air pressure control method for semiconductor polishing, in order to solve the technical problem in the prior art that, due to the lack of grouping of multiple points with similar polishing needs, the polishing parameters of different areas are not set accurately, resulting in low polishing efficiency and further affecting the quality and performance of the product.

[0006] In view of the above problems, this application provides an adaptive air pressure control method for semiconductor polishing.

[0007] This application provides an adaptive air pressure control method for semiconductor polishing, wherein the method includes: acquiring a cascaded air pressure control module for a chemical mechanical polishing (CMP) device, wherein the cascaded air pressure control module includes a first air pressure regulator and a second air pressure regulator; constructing an original three-dimensional model of the semiconductor wafer to be processed and a preset three-dimensional model of the polishing target; performing spatial alignment fitting on the original three-dimensional model and the preset three-dimensional model of the polishing target to generate a model to be polished; performing neighborhood clustering on the model to be polished for polishing requirements at multiple points to obtain multiple clustering regions, wherein the multiple clustering regions carry polishing requirement identification information; based on the multiple clustering regions and the polishing requirement identification information, combined with preset polishing fluid information, performing polishing air pressure optimization to obtain multiple optimal polishing air pressure parameters for the multiple clustering regions; generating an air pressure control signal based on the first air pressure regulator according to the multiple optimal polishing air pressure parameters for polishing control, and performing feedback optimization control through the second air pressure regulator.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] By acquiring a cascaded air pressure control module for a chemical mechanical polishing (CMP) device, wherein the cascaded air pressure control module includes a first air pressure regulator and a second air pressure regulator; constructing an original 3D model of the semiconductor wafer to be processed and a preset 3D model of the polishing target; performing spatial alignment fitting on the original 3D model and the preset 3D model of the polishing target to generate a model to be polished; performing neighborhood clustering on the model to be polished based on the polishing requirements at multiple points to obtain multiple cluster regions, each carrying polishing requirement identification information; based on the multiple cluster regions and the polishing requirement identification information, combined with preset polishing fluid information, optimizing the polishing air pressure to obtain multiple optimal polishing air pressure parameters for the multiple cluster regions; generating air pressure control signals based on the first air pressure regulator according to the multiple optimal polishing air pressure parameters for polishing control, and performing feedback optimization control through the second air pressure regulator, that is, by performing neighborhood clustering on the model to be polished based on the polishing requirements at multiple points and optimizing the polishing air pressure based on the clustering results, the technical goal of improving the accuracy and efficiency of semiconductor polishing is ultimately achieved, resulting in improved product quality.

[0010] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0012] Figure 1 This is a schematic flowchart of the adaptive air pressure control method for semiconductor polishing according to this application;

[0013] Figure 2 This is a schematic diagram of the process for obtaining multiple clustered regions in the adaptive air pressure control method for semiconductor polishing of this application. Detailed Implementation

[0014] This application provides an adaptive air pressure control method for semiconductor polishing, solving the technical problem in existing technologies where the lack of grouping for multiple points with similar polishing needs leads to inaccurate polishing parameter settings in different areas, resulting in low polishing efficiency and further affecting product quality and performance. It achieves the technical goal of improving the accuracy and efficiency of semiconductor polishing, thereby enhancing product quality.

[0015] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0016] Example

[0017] Please see the appendix Figure 1 This application provides an adaptive air pressure control method for semiconductor polishing, wherein the method specifically includes the following steps:

[0018] Step 1: Obtain the cascaded air pressure control module of the chemical mechanical polishing equipment, wherein the cascaded air pressure control module includes a first air pressure regulator and a second air pressure regulator;

[0019] Specifically, chemical mechanical polishing (CMP) equipment is used in semiconductor manufacturing processes to remove minute bumps and unevenness from the surface of silicon wafers, achieving a highly smooth surface. During manufacturing, a cascaded pneumatic control module precisely regulates the air pressure during polishing, ensuring polishing quality and efficiency. This cascaded pneumatic control module includes a first air pressure regulator and a second air pressure regulator. The first air pressure regulator is an electro-proportional valve. The second air pressure regulator includes a pressure sensor and a PID controller.

[0020] Step 2: Construct the original 3D model of the semiconductor wafer to be processed and the preset 3D model of the polishing target;

[0021] Specifically, a sample of the finished semiconductor wafer to be processed is obtained, representing the final state of the wafer after a complete polishing process. Laser scanning technology is used to scan the sample. A laser beam is emitted and the reflection time is measured to obtain the three-dimensional coordinate data of the sample surface. The obtained sample point cloud data is processed to generate a three-dimensional model, converting the point cloud data into a continuous surface in three-dimensional space, thus constructing a three-dimensional model of the sample that accurately reflects the shape and details of the sample surface. Based on the processed three-dimensional model, a preset polishing target three-dimensional model is generated as a reference during the polishing process, guiding the movement and parameter adjustments of the polishing equipment. By comparing the original three-dimensional model of the semiconductor wafer to be processed with the preset polishing target three-dimensional model, the amount of material to be removed and the polishing path are determined, thereby achieving precise polishing control.

[0022] Step 3: Spatial alignment and fitting are performed on the original 3D model and the preset polishing target 3D model to generate the model to be polished;

[0023] Specifically, an alignment reference is selected, and the point cloud data of the original 3D model and the preset polishing target 3D model are aligned. The differences between the original 3D model and the preset polishing target 3D model are compared. The differences are subtracted from the original 3D model to obtain the model to be polished.

[0024] Step 4: Perform neighborhood clustering on the polishing requirements of the multi-point model to be polished to obtain multiple cluster regions, each carrying polishing requirement identification information;

[0025] Specifically, multiple points are selected on the model to be polished, and the polishing thickness information at these points is extracted. The deviation between the polishing thickness information of these multiple points is calculated, i.e., the neighborhood thickness deviation. The calculated neighborhood thickness deviation is compared with a preset neighborhood deviation threshold. If the neighborhood thickness deviation is less than or equal to the preset threshold, it indicates that the polishing requirements of adjacent points are similar, and they are merged into clustered regions, thus obtaining multiple clustered regions, where the points within each region have similar polishing requirements.

[0026] Step 5: Based on the multiple clustered regions and the polishing requirement identification information, combined with the preset polishing fluid information, optimize the polishing air pressure to obtain multiple optimal polishing air pressure parameters for the multiple clustered regions;

[0027] Specifically, in the historical polishing control record database, the polishing demand identification information of multiple clustered regions is retrieved to find historical records similar to the current polishing demand. For each clustered region, the historical records similar to the current polishing demand are used to construct the polishing consistency fitness optimization. The candidate parameter set is generated by random combination and cross-expansion of the historical polishing control record set to find the optimal polishing air pressure parameter.

[0028] Step 6: Based on the first air pressure regulator, generate an air pressure control signal according to the multiple optimal polishing air pressure parameters to perform polishing control, and perform feedback optimization control through the second air pressure regulator.

[0029] Specifically, based on multiple optimal polishing air pressure parameters, a first air pressure regulator generates an air pressure control signal for polishing control. The second air pressure regulator collects the air pressure data of the cylinder in real time to reflect the actual air pressure situation during the polishing process. The collected real-time air pressure data of the cylinder is sent to the first air pressure regulator for pressure control loss compensation.

[0030] The adaptive air pressure control method for semiconductor polishing can achieve the technical goal of improving the accuracy and efficiency of semiconductor polishing, thereby improving product quality.

[0031] Furthermore, this application also includes the following steps:

[0032] The first air pressure regulator is an electro-proportional valve, and the second air pressure regulator includes an air pressure sensor and a PID controller located at the cylinder position in the chemical mechanical polishing equipment. One end of the air pressure sensor and one end of the PID controller are connected, and the other end of the PID controller is connected to the first air pressure regulator.

[0033] The second pressure regulator is used to optimize the pressure control signal of the first pressure regulator by feedback.

[0034] Specifically, such as Figure 2As shown, the first pressure regulator in the cascaded air pressure control module uses an electro-proportional valve. The electro-proportional valve controls the position of the valve core via an electrical signal, thereby regulating flow or pressure. In the chemical mechanical polishing (CMP) equipment, the electro-proportional valve acts as the main regulator, connected to a high-pressure air source, responsible for adjusting the input air pressure to a preset initial value. Based on the signal from the control loop, it controls the opening and closing of the supply and exhaust solenoid valves, causing controlled changes in the gas pressure in the pilot chamber. A pressure sensor detects the signal and feeds it back to the control loop, causing the valve core to move under electromagnetic force, ultimately outputting the required air pressure. The accuracy and stability of the electro-proportional valve ensure basic air pressure control during the polishing process. Further, the second pressure regulator consists of a pressure sensor located at the cylinder position in the CMP equipment and a PID controller. The pressure sensor measures the air pressure changes within the cylinder. Using physical carriers such as resistors, capacitors, thermoelectrics, and semiconductors, the gas pressure changes are converted into electrical signals, which are further converted into voltage, current, or frequency for transmission, processing, and control. By monitoring the air pressure at the cylinder position in real time, the air pressure sensor provides real-time feedback data to the PID controller. The PID controller receives the feedback signal from the air pressure sensor and, based on the deviation between the setpoint and the measured value, performs proportional, integral, and derivative calculations to derive a control signal, which is then transmitted to the electro-proportional valve in the first air pressure regulator for fine-tuning. The goal of the PID controller is to make the actual air pressure as close as possible to the setpoint, thereby achieving precise control of the air pressure during the polishing process.

[0035] By working in concert with an electro-proportional valve, a pressure sensor, and a PID controller, precise and stable control of the air pressure in the chemical mechanical polishing equipment is achieved, providing a strong guarantee for high-quality polishing operations.

[0036] Furthermore, this application also includes the following steps:

[0037] Based on the model to be polished, extract multiple polishing thickness information at multiple points;

[0038] Extract the first and second polishing thickness information of adjacent first and second points from the plurality of polishing thickness information;

[0039] Calculate the neighborhood thickness deviation between the first polishing thickness information and the second polishing thickness information;

[0040] Determine whether the neighborhood thickness deviation meets the preset neighborhood deviation threshold. If so, merge the first point and the second point.

[0041] Repeat the neighborhood point merging analysis until all the points have been traversed, and obtain the multiple clustering regions based on the merging results;

[0042] Calculate the average polishing thickness information within the multiple clustered regions to generate the polishing requirement identification information.

[0043] Specifically, multiple points are selected on the model to be polished, and the polishing thickness information at these points is extracted. These multiple points can be evenly distributed across the entire model, or they can be selected selectively based on the complexity of the model surface and the polishing requirements.

[0044] Then, from the extracted points, randomly select adjacent first and second points, and obtain first and second polishing thickness information to analyze the polishing thickness difference between adjacent points.

[0045] Next, the deviation between the polishing thickness information of the first and second points is calculated, namely the neighborhood thickness deviation, which is used to reflect the degree of difference in polishing requirements between adjacent points.

[0046] Next, the calculated neighborhood thickness deviation is compared with a preset neighborhood deviation threshold. If the neighborhood thickness deviation is less than or equal to the preset threshold, it indicates that the polishing requirements of adjacent points are similar, and they can be merged into a single cluster region.

[0047] Then, the polishing thickness information of adjacent points is repeatedly extracted, the neighborhood thickness deviation is calculated, and the points are judged and merged until all points are traversed, thereby gradually forming multiple clustered regions, with points in each region having similar polishing requirements.

[0048] Next, for each formed cluster region, the mean value of the polishing thickness information of all points in the cluster region is calculated, which represents the overall polishing requirement of the cluster region, and corresponding polishing requirement identification information is generated.

[0049] By clustering multiple points on the model to be polished according to polishing requirements and generating clear polishing requirement identification information for each cluster region, it is helpful to optimize the polishing path, reduce errors in the polishing process, and improve the overall polishing efficiency and quality.

[0050] Furthermore, this application also includes the following steps:

[0051] Perform 3D structural alignment between the original 3D model and the preset polishing target 3D model;

[0052] Based on the alignment results, the original 3D model is subjected to position extraction to remove the position, and the model to be polished is generated from the position extraction results.

[0053] Specifically, an alignment reference is selected. For example, the alignment reference is a known point, line, or surface existing in both the original 3D model and the preset polishing target 3D model, possessing the same geometric properties. A point cloud registration algorithm, such as the iterative nearest-point algorithm, is used to align the point cloud data of the original 3D model and the preset polishing target 3D model. During the alignment process, the optimal rotation and translation between the two point clouds are iteratively calculated to minimize the distance between corresponding points.

[0054] Then, after the 3D structure alignment is completed, the differences between the original 3D model and the preset polishing target 3D model are compared. By calculating the difference between the two models, the areas to be removed are extracted. Boolean operations are performed on the point cloud data to obtain point cloud data representing the removal areas. Based on the removal location map, one or more removal layers are constructed on the original 3D model, representing the material portions to be gradually removed. Subtracting the removal layers from the original 3D model yields the model to be polished, which reflects the intermediate states to be achieved during the polishing process, providing precise guidance for the polishing equipment.

[0055] By ensuring the precision of the polishing operation, optimizing the polishing path, and reducing unnecessary material removal, polishing efficiency and quality can be improved.

[0056] Furthermore, this application also includes the following steps:

[0057] Based on multiple clustering regions and the polishing requirement identification information, a search is performed in the historical polishing control records to obtain multiple sets of historical polishing control records for multiple clustering regions. Each historical polishing control record includes polishing control parameters and the surface roughness and material removal rate of the semiconductor wafer.

[0058] Construct the polishing consistency fitness function for the multiple clustered regions:

[0059] The polishing consistency fitness function is used to perform consistency optimization on multiple historical polishing control record sets to generate multiple optimal polishing air pressure parameters.

[0060] Specifically, the historical polishing control record database is searched based on polishing demand identifiers from multiple clustered regions to find historical records similar to the current polishing demand. These historical records contain polishing control parameters, as well as corresponding information such as semiconductor wafer surface roughness and material removal rate. The material removal rate refers to the waste material that needs to be removed; waste material includes materials that have been over-polished or damaged.

[0061] Then, for each cluster region, the consistency of polishing results under different polishing control parameters is evaluated using the polishing consistency fitness function.

[0062] Next, the polishing consistency fitness function is used to optimize the consistency of multiple historical polishing control record sets. Candidate parameter sets are generated by random combination and cross-expansion of historical polishing control record sets. The polishing consistency fitness function is then used to find the optimal polishing air pressure parameters to meet the polishing consistency and quality requirements. The polishing control parameters are continuously adjusted to find the optimal parameter combination that makes the polishing results most consistent.

[0063] By using information from historical polishing control records, constructing a polishing consistency fitness function and performing consistency optimization, optimal polishing gas pressure parameters can be generated for multiple clustered regions. This will help improve the stability and efficiency of the polishing process, while ensuring that the surface quality of semiconductor wafers meets requirements.

[0064] Furthermore, this application also includes the following steps:

[0065] The polishing consistency fitness function is:

[0066]

[0067] FIT = σ 2 [|FQ i -FQ i+1 |];

[0068] Among them, FQ i SR represents the independent fitness of the i-th cluster region. i Let MRR be the surface roughness of the i-th cluster region; i S represents the material removal rate of the i-th cluster region; MRR ω1 and ω2 are the standard material removal rate thresholds; i∈N, where N is the number of regions in the multiple clusters; ω1 and ω2 are weighting coefficients; FIT is the polishing consistency fitness of the multiple clusters; σ 2 [|FQ i -FQ i+1 |] represents the variance of the independent fitness difference between the i-th cluster region and the (i+1)-th cluster region.

[0069] Specifically, independent fitness (FQ) i It is the fitness value calculated separately for the i-th cluster region, based on the surface roughness SR of the i-th cluster region. i Material removal rate (MRR) of the i-th cluster region i and standard material removal rate threshold S MRR To perform a comprehensive evaluation. i∈N, where N is the number of regions in the multiple clusters. ω1 and ω2 are weighting coefficients used for weighted calculation. The surface roughness SR of the i-th cluster is... i The smaller the value, the higher the independent fitness FQ of the i-th cluster region. iThe larger the value, the smaller the value; conversely, the smaller the value, the smaller the value. The Material Removal Rate (MRR) of the i-th cluster region. i and standard material removal rate threshold S MRR The smaller the difference, the higher the independent fitness FQ of the i-th cluster region. i The larger the value, the smaller the value. Furthermore, the polishing consistency fitness (FIT) of multiple cluster regions is based on the variance σ of the independent fitness differences between the i-th and (i+1)-th cluster regions. 2 [|FQ i -FQ i+1 |] Obtained. The variance of the independent fitness difference is used to measure the consistency of polishing results among different cluster regions. The smaller the variance, the closer the polishing results of each cluster region are, the higher the consistency, and the smoother the polished surface; conversely, the larger the variance, the lower the consistency.

[0070] The polishing results of multiple clustered regions are evaluated using a polishing consistency fitness function, and the polishing control parameters that achieve optimal polishing consistency are found for adjustment and optimization.

[0071] Furthermore, this application also includes the following steps:

[0072] Based on multiple historical polishing control record sets of the multiple clustering regions, the control parameters of multiple regions are randomly combined to obtain the parameter combination results of multiple regions.

[0073] The combined results of the regional parameters are cross-expanded to obtain a set of candidate regional parameters;

[0074] The polishing consistency fitness function is used to calculate and obtain multiple polishing consistency fitness values ​​for multiple region candidate parameters in the region candidate parameter set.

[0075] Based on the multiple polishing consistency fitness values, the candidate parameters of the region corresponding to the minimum polishing consistency fitness value are extracted to obtain the multiple optimal polishing air pressure parameters.

[0076] Specifically, control parameters are randomly selected and combined from multiple historical polishing control record sets across multiple clusters. Each cluster has a set of historical polishing control records, including different polishing gas pressure parameters and other control parameters. By randomly combining these parameters, multiple regional parameter combination results are generated.

[0077] Then, the results of the regional parameter combinations are cross-expanded by exchanging or mixing parameters in different regional parameter combinations to generate new regional candidate parameter sets, thereby expanding the search range and increasing the possibility of finding the optimal parameters.

[0078] Next, for each region candidate parameter, a polishing consistency fitness function is used to calculate its polishing consistency fitness. By calculating the polishing consistency fitness, the performance of each candidate parameter in achieving polishing consistency and quality requirements is evaluated.

[0079] Next, the polishing consistency fitness of all candidate parameters in all regions is compared, and the parameter with the smallest fitness is selected as the optimal polishing air pressure parameter, indicating that the parameter with the smallest fitness meets the polishing quality requirements while maintaining polishing consistency.

[0080] Candidate parameter sets are generated by randomly combining and cross-expanding the historical polishing control record set, and the optimal polishing gas pressure parameters are found by using the polishing consistency fitness function to meet the polishing consistency and quality requirements.

[0081] Furthermore, this application also includes the following steps:

[0082] The first cluster region is polished using the first optimal polishing air pressure parameter among the plurality of optimal polishing air pressure parameters, and the real-time air pressure of the cylinder is collected by the air pressure sensor in the second air pressure regulator.

[0083] The real-time air pressure of the cylinder is sent to the PID controller for pressure control loss compensation, and the feedback control signal is generated and sent to the first air pressure regulator.

[0084] The first pressure regulator optimizes the control of the electro-proportional valve according to the feedback control signal.

[0085] Specifically, the first optimal polishing air pressure parameter is randomly selected from multiple optimal polishing air pressure parameters and used to control the polishing of the first clustered region. During the polishing process, the air pressure data of the cylinder is collected in real time by the air pressure sensor in the second air pressure regulator to reflect the actual air pressure situation during the polishing process.

[0086] Then, the collected real-time cylinder air pressure data is sent to the PID controller for pressure control loss compensation. By calculating and processing the deviation between the real-time air pressure and the set air pressure, a corresponding feedback control signal is generated to compensate for possible pressure losses during polishing. The PID controller calculates the feedback control signal based on the real-time air pressure data and the set optimal polishing air pressure parameters. The feedback control signal contains instructions to adjust the air pressure and is sent to the first air pressure regulator to bring the actual air pressure closer to the set optimal value.

[0087] Next, after receiving the feedback control signal, the first air pressure regulator optimizes the control of the electro-proportional valve according to the instructions in the feedback control signal, adjusts the opening of the electro-proportional valve, and thus achieves precise control of the polishing air pressure.

[0088] By enabling real-time monitoring and precise control of polishing air pressure, the air pressure is kept stable during the polishing process, achieving optimal polishing results, improving the quality and consistency of polishing, as well as the efficiency and stability of the polishing process.

[0089] In summary, the adaptive air pressure control method for semiconductor polishing provided in this application has the following technical advantages:

[0090] By acquiring a cascaded air pressure control module for a chemical mechanical polishing (CMP) device, wherein the cascaded air pressure control module includes a first air pressure regulator and a second air pressure regulator; constructing an original 3D model of the semiconductor wafer to be processed and a preset 3D model of the polishing target; performing spatial alignment fitting on the original 3D model and the preset 3D model of the polishing target to generate a model to be polished; performing neighborhood clustering on the model to be polished based on the polishing requirements at multiple points to obtain multiple cluster regions, each carrying polishing requirement identification information; based on the multiple cluster regions and the polishing requirement identification information, combined with preset polishing fluid information, optimizing the polishing air pressure to obtain multiple optimal polishing air pressure parameters for the multiple cluster regions; generating air pressure control signals based on the first air pressure regulator according to the multiple optimal polishing air pressure parameters for polishing control, and performing feedback optimization control through the second air pressure regulator, that is, by performing neighborhood clustering on the model to be polished based on the polishing requirements at multiple points and optimizing the polishing air pressure based on the clustering results, the technical goal of improving the accuracy and efficiency of semiconductor polishing is ultimately achieved, resulting in improved product quality.

[0091] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0092] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. An adaptive air pressure control method for semiconductor polishing, characterized in that, The method includes: A cascaded air pressure control module for a chemical mechanical polishing (CMP) device is obtained, wherein the cascaded air pressure control module includes a first air pressure regulator and a second air pressure regulator. Construct the original 3D model of the semiconductor wafer to be processed and the preset 3D model of the polishing target; Spatial alignment and fitting are performed on the original 3D model and the preset polishing target 3D model to generate a model to be polished; The polishing requirements of the model to be polished are clustered in neighborhood to obtain multiple cluster regions, and the multiple cluster regions carry polishing requirement identification information. Based on the multiple clustering regions and the polishing requirement identification information, combined with the preset polishing fluid information, the polishing air pressure is optimized to obtain multiple optimal polishing air pressure parameters for the multiple clustering regions; Based on the first air pressure regulator, an air pressure control signal is generated according to the multiple optimal polishing air pressure parameters to perform polishing control, and feedback optimization control is performed through the second air pressure regulator. Specifically, the polishing requirements at multiple points on the model to be polished are clustered to obtain multiple cluster regions. These multiple cluster regions carry polishing requirement identification information, including: Based on the model to be polished, extract multiple polishing thickness information at multiple points; Extract the first and second polishing thickness information of adjacent first and second points from the plurality of polishing thickness information; Calculate the neighborhood thickness deviation between the first polishing thickness information and the second polishing thickness information; Determine whether the neighborhood thickness deviation meets the preset neighborhood deviation threshold. If so, merge the first point and the second point. Repeat the neighborhood point merging analysis until all the points have been traversed, and obtain the multiple clustering regions based on the merging results; Calculate the average polishing thickness information within the multiple clustering regions to generate the polishing requirement identification information; Specifically, based on the multiple clustering regions and the polishing requirement identification information, combined with preset polishing fluid information, polishing air pressure optimization is performed to obtain multiple optimal polishing air pressure parameters for the multiple clustering regions, including: Based on multiple clustering regions and the polishing requirement identification information, a search is performed in the historical polishing control records to obtain multiple sets of historical polishing control records for multiple clustering regions. Each historical polishing control record includes polishing control parameters and the surface roughness and material removal rate of the semiconductor wafer. Construct the polishing consistency fitness function for the multiple clustered regions: The polishing consistency fitness function is used to perform consistency optimization on multiple historical polishing control record sets to generate the multiple optimal polishing air pressure parameters. The polishing consistency fitness function is as follows: ； ； in, For the first Independent fitness of each cluster region; For the first Surface roughness of each cluster region; For the first Material removal rate of each cluster region; The standard material removal rate threshold; , The number of regions in multiple clusters; and These are the weighting coefficients; For polishing consistency fitness across multiple clustered regions; For the first The cluster regions and the first The variance of the independent fitness differences between clustered regions.

2. The method as described in claim 1, characterized in that, The first air pressure regulator is an electro-proportional valve, and the second air pressure regulator includes an air pressure sensor and a PID controller located at the cylinder position in the chemical mechanical polishing equipment. One end of the air pressure sensor and one end of the PID controller are connected, and the other end of the PID controller is connected to the first air pressure regulator. The second pressure regulator is used to optimize the pressure control signal of the first pressure regulator by feedback.

3. The method as described in claim 1, characterized in that, Spatial alignment and fitting are performed on the original 3D model and the preset polishing target 3D model to generate a model to be polished, including: Perform 3D structural alignment between the original 3D model and the preset polishing target 3D model; Based on the alignment results, the original 3D model is subjected to position extraction to remove the position, and the model to be polished is generated from the position extraction results.

4. The method as described in claim 1, characterized in that, The polishing consistency fitness function is used to perform consistency optimization on multiple historical polishing control record sets to generate the multiple optimal polishing gas pressure parameters, including: Based on multiple historical polishing control record sets of the multiple clustering regions, the control parameters of multiple regions are randomly combined to obtain the parameter combination results of multiple regions. The combined results of the regional parameters are cross-expanded to obtain a set of candidate regional parameters; The polishing consistency fitness function is used to calculate and obtain multiple polishing consistency fitness values ​​for multiple region candidate parameters in the region candidate parameter set. Based on the multiple polishing consistency fitness values, the candidate parameters of the region corresponding to the minimum polishing consistency fitness value are extracted to obtain the multiple optimal polishing air pressure parameters.

5. The method as described in claim 2, characterized in that, Based on the first air pressure regulator, an air pressure control signal is generated according to the multiple optimal polishing air pressure parameters for polishing control, and feedback optimization control is performed through the second air pressure regulator, including: The first cluster region is polished using the first optimal polishing air pressure parameter among the plurality of optimal polishing air pressure parameters, and the real-time air pressure of the cylinder is collected by the air pressure sensor in the second air pressure regulator. The real-time air pressure of the cylinder is sent to the PID controller for pressure control loss compensation, and a feedback control signal is generated and sent to the first air pressure regulator. The first pressure regulator optimizes the control of the electro-proportional valve according to the feedback control signal.

Images