An epitaxial pre-wafer surface treatment method and apparatus

Abstract

The present invention discloses a method and device for wafer surface treatment before epitaxy, relating to the technical field of wafers. The method includes: setting a preset incident angle, performing X-ray total reflection scanning on a target wafer, and a detector capturing and generating a three-dimensional mapping spectrum of the target wafer; preprocessing surface detection records and data-driven supervised training of a surface detection module; receiving the three-dimensional mapping spectrum, identifying and integrating defect elements based on the surface detection module to determine the surface defect distribution; coordinating dry processing methods and wet processing methods, traversing the surface defect distribution to make single-defect treatment decisions, performing global optimization with cross-infection as a constraint to determine a surface treatment strategy; and a device group responding to the surface treatment strategy to perform pre-positioning and defect treatment on the target wafer. The technical problem of low efficiency in wafer surface defect treatment in the prior art is solved, and the technical effect of improving the efficiency of wafer surface defect treatment is achieved by optimizing the surface treatment strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of wafers, and particularly relates to a method and device for wafer surface treatment before epitaxy. Background Art

[0002] In the process of semiconductor manufacturing, the quality of the wafer surface directly affects subsequent epitaxial growth and device performance. Therefore, the wafer surface treatment process plays a crucial role in production efficiency and product yield. Currently, conventional wafer surface treatment methods mainly rely on single detection techniques and simple surface cleaning means, and cannot effectively identify and process minute surface defects, especially those involving trace elements and surface micro-nano structures. In addition, existing treatment methods usually lack pertinence and are prone to cross-infection during the treatment process, affecting the overall surface quality of the wafer, thereby reducing production efficiency. Therefore, there is an urgent need for a more intelligent and precise wafer surface treatment method that can, without damaging the wafer, monitor, identify, and optimize the treatment of surface defects in real time, thereby improving the overall quality of the product and production efficiency. Summary of the Invention

[0003] The present application provides a method and device for wafer surface treatment before epitaxy, which solves the technical problem of low efficiency in treating wafer surface defects in the prior art.

[0004] In view of the above problems, the present application provides a method and device for wafer surface treatment before epitaxy.

[0005] In the first aspect of the present application, a method for wafer surface treatment before epitaxy is provided. The method includes:

[0006] Setting a preset incident angle, performing X-ray total reflection scanning on a target wafer based on a fluorescence spectrometer, and a detector capturing and generating a three-dimensional mapped spectrum of the target wafer; invoking surface detection records, preprocessing the surface detection records with trace elements, surface chemical states, and micro-nano structures as identification dimensions, and data-driven supervised training of a surface detection module, where the preprocessing results include a dimensional detection architecture and preprocessing samples; receiving the three-dimensional mapped spectrum, identifying and integrating defect elements based on the surface detection module to determine the surface defect distribution, where each distributed defect is marked with a relative spatial position; coordinating dry processing methods and wet processing methods, traversing the surface defect distribution to make single-defect treatment decisions, and performing global optimization with cross-infection as a constraint to determine a surface treatment strategy, where the surface treatment strategy includes multiple processing nodes; and an equipment group responding to the surface treatment strategy to perform pre-positioning and defect treatment on the target wafer.

[0007] In the second aspect of the present application, a device for wafer surface treatment before epitaxy is provided. The device includes:

[0008] Scanning component, which is used to set a preset incident angle, perform X-ray total reflection scanning on a target wafer based on a fluorescence spectrometer, and a detector captures and generates a three-dimensional mapped spectrum of the target wafer; Pretreatment component, which is used to call surface detection records, preprocess the surface detection records with trace elements, surface chemical states, and micro-nano structures as the identification dimensions, and data-drive supervised training of the surface detection module, where the pretreatment results include a dimensional detection architecture and pretreatment samples; Identification component, which is used to receive the three-dimensional mapped spectrum, identify and integrate defect elements based on the surface detection module, and determine the surface defect distribution, where each distributed defect is marked with a relative spatial position; Optimization component, which is used to coordinate dry processing methods and wet processing methods, traverse the surface defect distribution to make single-defect processing decisions, perform global optimization with cross-infection as a constraint, and determine a surface treatment strategy, where the surface treatment strategy includes multiple processing nodes; Surface treatment component, which is used for the equipment group to respond to the surface treatment strategy and perform pre-positioning and defect processing on the target wafer.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] First, set a preset incident angle, perform X-ray total reflection scanning on a target wafer based on a fluorescence spectrometer, and a detector captures and generates a three-dimensional mapped spectrum of the target wafer. Next, call surface detection records, preprocess the surface detection records with trace elements, surface chemical states, and micro-nano structures as the identification dimensions, and data-drive supervised training of the surface detection module, where the pretreatment results include a dimensional detection architecture and pretreatment samples. Further, receive the three-dimensional mapped spectrum, identify and integrate defect elements based on the surface detection module, and determine the surface defect distribution, where each distributed defect is marked with a relative spatial position. Then, coordinate dry processing methods and wet processing methods, traverse the surface defect distribution to make single-defect processing decisions, perform global optimization with cross-infection as a constraint, and determine a surface treatment strategy, where the surface treatment strategy includes multiple processing nodes. Finally, the equipment group responds to the surface treatment strategy and performs pre-positioning and defect processing on the target wafer. This solves the technical problem of low efficiency in processing wafer surface defects in the prior art, and achieves the technical effect of improving the efficiency of processing wafer surface defects by optimizing the surface treatment strategy. Description of the Drawings

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0012] Figure 1 Schematic flow chart of a method for processing the surface of a pre-epitaxial wafer provided by an embodiment of the present application;

[0013] Figure 2 Schematic structural diagram of a device for processing the surface of a pre-epitaxial wafer provided by an embodiment of the present application.

[0014] Explanation of reference numerals: Scanning component 11, Pretreatment component 12, Identification component 13, Optimization component 14, Surface treatment component 15. Detailed implementation manners

[0015] By providing a method and a device for processing the surface of a pre-epitaxial wafer, the present application solves the technical problem of low efficiency in processing surface defects of wafers in the prior art.

[0016] Next, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0017] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product or server that includes a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0018] Embodiment 1, as Figure 1 shown, the present application provides a method for processing the surface of a pre-epitaxial wafer, where the method includes:

[0019] Set a preset incident angle, and perform X-ray total reflection scanning on the target wafer based on a fluorescence spectrometer, and a detector captures and generates a three-dimensional mapped spectrum of the target wafer.

[0020] By setting a preset incident angle and using a fluorescence spectrometer to perform X-ray total reflection scanning on the target wafer, the fluorescence spectrometer controls the incident angle to make the X-ray incident on the wafer surface at a specific angle, ensuring that the X-ray can fully cover and reflect the microscopic features on the wafer surface. The detector is responsible for capturing the X-ray signal reflected from the wafer surface during this process and generating a three-dimensional mapped spectrum of the target wafer. The three-dimensional mapped spectrum contains detailed information about the wafer surface features, providing basic data support for subsequent surface detection and defect identification.

[0021] Furthermore, the setting of the preset incident angle includes:

[0022] According to the wafer characteristics and detector performance, determine the incident angle range, wherein the preset incident angle is of a small magnitude; based on the incident angle range, randomly determine any incident angle as the preset incident angle, perform X-ray total reflection scanning based on the preset incident angle, and receive fluorescence signals: verify the fluorescence signals and perform compensation adjustment on the preset incident angle; wherein, the determination of the incident angle range includes: combining the incident wavelength to determine the critical incident angle, wherein the critical incident angle is positively correlated with the incident wavelength; the incident angle range is less than the critical incident angle.

[0023] Specifically, according to the characteristics of the wafer (such as material composition, surface roughness, etc.) and the performance of the detector (such as sensitivity, detection range), determine a suitable incident angle range. The setting of this angle range is aimed at ensuring that the reflection of X-rays can effectively capture the microscopic features of the wafer surface and making the intensity of the fluorescence spectrum signal within the measurable range. Among them, the preset incident angle is of a small magnitude to meet the fine detection requirements of X-ray total reflection; within the determined incident angle range, randomly select any incident angle as the preset incident angle for this scan. Use the selected preset incident angle to perform X-ray total reflection scanning on the target wafer. The detector receives the fluorescence signals reflected from the wafer surface. The fluorescence signals record the optical characteristics and structural information of the wafer surface; after receiving the fluorescence signals, verify the signals, evaluate their signal-to-noise ratio and effectiveness. If the signal intensity or quality does not meet the expected standard, perform compensation adjustment on the preset incident angle and fine-tune the incident angle to optimize the signal capture effect; the incident wavelength is positively correlated with the critical incident angle. The longer the wavelength, the larger the critical incident angle. The critical incident angle refers to the minimum angle at which X-rays can be completely reflected under the condition of total reflection; to ensure the accuracy of total reflection scanning, the incident angle range is set to be less than the critical incident angle range to ensure that X-rays can be totally reflected and effectively obtain surface information during the scanning process.

[0024] Call the surface detection record, use trace elements, surface chemical state and micro-nano structure as the identification dimensions, preprocess the surface detection record, and perform data-driven supervised training on the surface detection module, wherein the preprocessing result includes a dimensional detection architecture and preprocessing samples.

[0025] Call the surface inspection records from the database, using trace elements, surface chemical states, and micro-nano structures as identification dimensions. Trace elements refer to whether there are harmful trace elements on the wafer surface. Surface chemical states refer to chemical corrosion on the wafer surface, oxide layers formed by reactions, etc. Micro-nano structures refer to the microscopic structures on the wafer surface, such as micro-nano structure changes on the surface with tiny cracks, defect pits, etc. Based on the above identification dimensions, preprocess the surface inspection records to obtain a dimensional inspection framework and preprocessing samples. Through the preprocessed surface inspection records, perform supervised training in a data-driven manner to train the surface inspection module so that the surface inspection module can accurately identify various defects and features on the wafer surface according to the input data.

[0026] Furthermore, preprocessing the surface inspection records includes:

[0027] Traverse the surface inspection records, perform a first clustering based on the identification dimensions to determine dimensional clustering clusters; identify the dimensional clustering clusters, perform hierarchical clustering processing with multi-level subdivision to determine a record classification tree, and the record classification tree corresponds one-to-one with the identification dimensions; distinguish between characteristic features and general features, split the record classification tree to determine a discriminative classification tree; based on the discriminative classification tree, determine the dimensional inspection framework and the preprocessing samples.

[0028] Specifically, traverse all surface detection records and perform a clustering analysis on the data based on predefined recognition dimensions (such as trace elements, surface chemical state, micro-nano structure, etc.). By clustering records with similar characteristics together, preliminary dimension clustering clusters are determined. After completing one clustering, the internal structure of each dimension clustering cluster will be further identified. Through multi-level subdivision and hierarchical clustering processing, these clustering clusters are divided more finely. Hierarchical clustering can further split the records in each clustering cluster according to fine-grained characteristics and form a record classification tree. The record classification tree corresponds one-to-one with the recognition dimensions and reflects the hierarchical structure of the records under each recognition dimension. After constructing the record classification tree, the characteristics in the record classification tree will be further analyzed. By distinguishing between discriminative features and universal features, the classification tree is segmented. Discriminative features refer to features with significant differences in specific detection records, while universal features are features common to various types of records. By segmenting the record classification tree, a discriminative classification tree can be generated. The discriminative classification tree highlights the discriminative features between different categories. For example, for the oxide layer under the surface chemical state dimension, the recognition features of the oxide layers of different underlying elements are different, and further down, the recognition features of different elements also vary in strength, and it is constructed based on the discriminative features. After obtaining the discriminative classification tree, based on the structure and node characteristics of the discriminative classification tree, the final dimension detection architecture is further determined. The dimension detection architecture represents the key feature dimensions that need to be concerned during the detection process and provides an organized detection framework. At the same time, the node information in the discriminative classification tree will be used as preprocessing samples for training the detection module.

[0029] Furthermore, determining the dimension detection architecture and the preprocessing samples includes:

[0030] Taking the discriminative classification tree as the preprocessing sample, traversing the discriminative classification tree and mining node characteristics to determine a feature classification tree as the dimension detection architecture; traversing the discriminative classification tree, based on the sample record span corresponding to the tree nodes, taking the maximum value of the span as the feature recognition intensity; and identifying the dimension detection architecture based on the feature recognition intensity.

[0031] Specifically, the discrimination classification tree is a classification structure based on multi-dimensional recognition features (such as trace elements, surface chemical state, micro-nano structure, etc.). Each node in the discrimination classification tree represents a different feature classification, and the node information extracted therefrom will be used as a preprocessed sample. Traverse the discrimination classification tree, analyze each node, and mine its feature information, such as the specific defect type represented by the node, detection conditions, etc., in order to construct a feature classification tree. The feature classification tree is a refinement and structuring of the discrimination classification tree, containing detailed feature information under each recognition dimension. The feature classification tree serves as a dimension detection framework to guide subsequent defect detection. When traversing the discrimination classification tree, analyze according to the sample record span corresponding to each node. The sample record span refers to the distribution range of different sample data in the relevant dimension under a specific node. By extracting the maximum value of these spans, the recognition intensity of the node features can be determined. The larger the maximum span value, the greater the variation of the feature in different samples, indicating that the feature has higher discriminative value in classification. According to the feature recognition intensity of each node, label the feature classification tree (i.e., the dimension detection framework). Features with higher recognition intensity will be given higher priority in the detection framework to ensure that these key features are preferentially recognized in the subsequent defect detection process, thereby improving the accuracy and efficiency of detection.

[0032] Receive the three-dimensional mapped spectrum, identify and integrate defect elements based on the surface detection module, and determine the surface defect distribution, where each distributed defect is marked with a relative spatial position.

[0033] By receiving the three-dimensional mapped spectrum and using the surface detection module for analysis, that is, identifying and integrating defect elements for each recognition dimension (such as trace elements, surface chemical state, and micro-nano structure), determine the defect distribution on the wafer surface, and mark the relative spatial position of each defect to ensure that the precise position of each defect is clearly located on the wafer surface.

[0034] Coordinate the dry processing method and the wet processing method, traverse the surface defect distribution to make a single-defect processing decision, perform global optimization with cross-infection as a constraint, and determine the surface treatment strategy, where the surface treatment strategy includes multiple processing nodes.

[0035] Adopt a dry processing method and a wet processing method in combination, and optimize the processing for different types of defects. The dry processing method usually includes non-contact processing methods such as laser cleaning and plasma processing, while the wet processing method includes processing methods that require a liquid medium such as chemical solution cleaning and soaking. By combining these two processing methods, various defects on the wafer surface can be processed more comprehensively, thus ensuring a higher processing effect. Traverse the distribution of surface defects, and make a single-item processing decision for each specific defect; according to the nature, location of the defect and the corresponding surface detection records, determine whether to use the dry processing method, the wet processing method, or a combination of both; at the same time, according to the processing effect and the potential impact on other areas, determine the best processing method to avoid unnecessary cross-infection, that is, prevent the defect processing in one area from having a negative impact on other areas during the processing process. After traversing all the defects and completing the single-item processing decision, with cross-infection as the main constraint, perform global optimization. The purpose of global optimization is to ensure that the processing scheme has the optimal effect as a whole, that is, not only to solve each defect, but also to avoid causing additional impacts on other unprocessed areas; through global optimization, the rationality and effectiveness of each processing step can be ensured, and finally the best surface processing strategy can be determined. The surface processing strategy contains multiple processing nodes, and each node represents a specific processing operation point. It not only includes the processing method for defects in a specific area, but also contains information such as the processing order, the equipment used, and the parameter configuration; the purpose of the multi-processing node strategy is to ensure that the wafer surface processing can effectively cover all defect areas, and the various processing steps are coordinated with each other to avoid processing conflicts.

[0036] Furthermore, traversing the distribution of the surface defects to make a single-defect processing decision and performing global optimization with cross-infection as the constraint includes:

[0037] Construct a decision database, and the decision database includes a dry processing strategy sub-library and a wet processing strategy sub-library; traverse the distribution of the surface defects, take each surface defect as the first-level greedy target, perform greedy search based on the decision database to determine an alternative strategy library; take the optimal alternative as the second-level greedy target, traverse the alternative strategy library to perform greedy search, determine and adjust the initial greedy strategy to determine a single-item strategy set, and there is a corresponding relationship between the alternative strategy library, the single-item strategy set and the distribution of the surface defects; traverse the single-item strategy set, and perform global iterative optimization based on strategy cross-contamination to determine the surface processing strategy.

[0038] By constructing a decision database to store and manage treatment strategies for different types of surface defects, the decision database includes a dry process strategy sub-library and a wet process strategy sub-library. The dry process strategy sub-library contains all strategies applicable to dry processing, such as non-contact processing methods like laser cleaning and plasma treatment. The wet process strategy sub-library contains all strategies applicable to wet processing, such as chemical solution cleaning and immersion treatment. The strategies in each sub-library correspond to different treatment parameters, including treatment type, conditions, effect prediction, etc. Traverse the identified defect distribution on the wafer surface, and each defect is regarded as a first-level greedy target; for each defect, from the dry process strategy sub-library and the wet process strategy sub-library in the decision database, based on the greedy search algorithm, quickly select a set of possible alternative strategies to form an alternative strategy library; the greedy search algorithm selects the local optimal solution and preferentially selects those strategies with the best treatment effects as alternative solutions. After obtaining the alternative strategy library, take the optimal alternative strategy as the second-level greedy target and further traverse the alternative strategy library; through secondary greedy search, select the optimal treatment plan for each defect to form an initial greedy strategy; then, according to the characteristics of each defect and its relevance to other defects, adjust the initial greedy strategy to finally form a single strategy set. There is a one-to-one correspondence between the single strategy set and the surface defect distribution, that is, each defect has its own exclusive treatment strategy. After completing the construction of the single strategy set, traverse all the single strategy sets to check for possible cross-contamination problems between different strategies. For example, the treatment strategy for a certain defect may affect the defects in the adjacent area or cause cross-infection; based on the cross-contamination effect of the strategies, perform global iterative optimization. Through repeated iterative adjustment, modify and optimize each treatment strategy each time to ensure that all treatment steps can be executed under the optimal conditions and avoid the negative impacts between strategies. After global iterative optimization, finally determine a globally optimal surface treatment strategy. The surface treatment strategy includes multi-node operation steps and treatment sequences to ensure that all defects can be effectively eliminated, cross-contamination can be avoided, and the overall treatment efficiency and the quality of the wafer surface can be improved.

[0039] Furthermore, global iterative optimization based on strategy cross-contamination includes:

[0040] Based on the alternative strategy library, establish a global optimization space; construct a first strategy group and a second strategy group, store the single strategy set in the second strategy group, and initialize the first strategy group to be empty; traverse the second strategy group, identify and mark the cross-contamination characteristics of the strategies, and based on the global optimization space, perform loop judgment and strategy transfer to determine the surface treatment strategy.

[0041] Specifically, based on the alternative strategy library, a global optimization space is constructed. The global optimization space is a multi-dimensional space that contains all possible strategy combinations and their related processing parameters. Each dimension corresponds to a specific processing parameter or strategy choice, and all alternative strategies are mapped into this global optimization space. The first strategy group and the second strategy group are constructed. The first strategy group is used to store the optimized strategies and is empty in the initial stage. The second strategy group is used to store all the determined single strategy sets, and each strategy set corresponds to a preliminary processing plan for surface defects. Traverse each strategy in the second strategy group, analyze its relationship with other surrounding strategies, and identify possible cross-contamination characteristics. Cross-contamination characteristics refer to the situation where the execution of a certain strategy may have a negative impact on adjacent areas or other strategies, such as the spread of contamination during processing, mechanical interference, etc. After marking the cross-contamination characteristics, perform a loop judgment within the global optimization space. By evaluating the mutual influence between strategies, determine which strategies need to be adjusted or replaced. The result of each judgment may cause the transfer of strategies from the second strategy group to the first strategy group. The goal of strategy transfer is to store those strategies that no longer cause cross-contamination after adjustment and optimization into the first strategy group, while the strategies with unresolved cross-contamination problems remain in the second strategy group. Through repeated loop judgments and strategy transfers, continuously reduce the cross-contamination between strategies, and gradually transfer the optimized strategies from the second strategy group to the first strategy group. When all strategies have been optimized and cross-contamination has been eliminated, the strategy set in the first strategy group becomes the final surface treatment strategy.

[0042] Furthermore, the performing of the loop judgment and strategy transfer includes:

[0043] Based on the second strategy group, identify the first-order strategy and determine whether there is a mark. If there is a mark, combine with the global optimization space to determine the first alternative strategy, and based on the avoidance of the cross-contamination characteristics, determine the optimized first-order strategy and transfer it to the first strategy group. If there is no mark, transfer the first-order strategy to the first strategy group. Determine the second-order strategy based on the second strategy group, perform loop judgment and transfer until the second strategy group is empty, and take the first strategy group as the surface treatment strategy, where the first-order strategy has a predecessor node mark based on the second-order strategy.

[0044] Specifically, select a strategy with the highest priority from the second strategy group as the first-priority strategy. The priority of the first-priority strategy is determined according to the order of defect handling. Check the first-priority strategy to determine whether there is a mark of cross-contamination characteristics. If it is detected that the first-priority strategy has a cross-contamination mark, analyze the strategy conflict in combination with the global optimization space and optimize it based on the cross-contamination characteristics. Through global optimization, determine a first alternative strategy that can effectively avoid cross-contamination characteristics, thereby minimizing interference to other strategies. The optimized first-priority strategy will be corrected to the version after avoidance to ensure that cross-contamination will not occur. This optimized first-priority strategy is then transferred to the first strategy group and marked as an optimized strategy. After processing the first-priority strategy, continue to select the next strategy with the highest priority from the second strategy group, that is, the second-priority strategy, and repeat the above loop judgment and transfer process. The second-priority strategy will also be checked for the presence of cross-contamination marks and optimized or directly transferred according to the situation. Continuously perform this loop judgment and strategy transfer process to gradually process all strategies in the second strategy group. After each judgment and transfer, the number of strategies in the second strategy group will decrease, while the number of strategies in the first strategy group will gradually increase. When all strategies have been judged, optimized, and transferred, the second strategy group will finally be empty. When the second strategy group is empty, all strategies in the first strategy group have been optimized and processed, and there will be no cross-contamination or interference. At this time, the set of strategies in the first strategy group is used as the final surface treatment strategy.

[0045] The equipment group responds to the surface treatment strategy and pre-positions and handles defects on the target wafer.

[0046] The equipment group receives the surface treatment strategy through the communication interface. According to the surface treatment strategy, the equipment group will start in sequence to ensure that all processing steps are strictly executed according to the plan. Each device will be configured according to the processing nodes and parameters specified in the strategy, such as the processing order, the choice of dry or wet processing, etc. By the equipment group responding to the surface treatment strategy and performing pre-positioning and defect handling, the defects on the surface of the target wafer can be effectively and accurately repaired, improving the wafer quality and optimizing the overall production efficiency.

[0047] In summary, the embodiments of the present application have at least the following technical effects:

[0048] First, set a preset incident angle, perform X-ray total reflection scanning on the target wafer based on a fluorescence spectrometer, and the detector captures and generates a three-dimensional mapped spectrum of the target wafer. Next, call the surface detection record, preprocess the surface detection record with trace elements, surface chemical state, and micro-nano structure as the identification dimensions, and data-drive supervised training of the surface detection module. Among them, the preprocessing result includes a dimensional detection architecture and preprocessing samples. Further, receive the three-dimensional mapped spectrum, identify and integrate defect elements based on the surface detection module, and determine the surface defect distribution. Among them, each distributed defect is marked with a relative spatial position. Then, coordinate the dry processing method and the wet processing method, traverse the surface defect distribution to make a single defect processing decision, and perform global optimization with cross-infection as a constraint to determine the surface treatment strategy. The surface treatment strategy includes multiple processing nodes. Finally, the equipment group responds to the surface treatment strategy to perform pre-positioning and defect processing on the target wafer. This solves the technical problem of low efficiency in processing wafer surface defects in the prior art, and achieves the technical effect of improving the efficiency of processing wafer surface defects by optimizing the surface treatment strategy.

[0049] Embodiment 2, based on the same inventive concept as the method for processing the surface of a pre-epitaxial wafer in the foregoing embodiment, as Figure 2 shown, the present application provides a device for processing the surface of a pre-epitaxial wafer, wherein the device includes:

[0050] A scanning component 11 for setting a preset incident angle, performing X-ray total reflection scanning on the target wafer based on a fluorescence spectrometer, and the detector captures and generates a three-dimensional mapped spectrum of the target wafer; a preprocessing component 12 for calling the surface detection record, preprocessing the surface detection record with trace elements, surface chemical state, and micro-nano structure as the identification dimensions, and data-drive supervised training of the surface detection module. Among them, the preprocessing result includes a dimensional detection architecture and preprocessing samples; an identification component 13 for receiving the three-dimensional mapped spectrum, identifying and integrating defect elements based on the surface detection module, and determining the surface defect distribution. Among them, each distributed defect is marked with a relative spatial position; an optimization component 14 for coordinating the dry processing method and the wet processing method, traversing the surface defect distribution to make a single defect processing decision, and performing global optimization with cross-infection as a constraint to determine the surface treatment strategy. The surface treatment strategy includes multiple processing nodes; a surface treatment component 15 for the equipment group to respond to the surface treatment strategy to perform pre-positioning and defect processing on the target wafer.

[0051] Further, the scanning component 11 is used to execute the following method:

[0052] According to the wafer characteristics and detector performance, determine the incident angle range, where the preset incident angle is of a small magnitude; based on the incident angle range, randomly determine any incident angle as the preset incident angle, perform X-ray total reflection scanning based on the preset incident angle, and receive fluorescence signals: verify the fluorescence signals and perform compensation adjustment on the preset incident angle; where the determination of the incident angle range includes: combining the incident wavelength to determine the critical incident angle, where the critical incident angle is positively correlated with the incident wavelength; the incident angle range is less than the critical incident angle.

[0053] Further, the preprocessing component 12 is used to execute the following method:

[0054] Traverse the surface detection records, perform a first clustering based on the recognition dimension to determine dimension clustering clusters; identify the dimension clustering clusters, perform hierarchical clustering processing with multi-level subdivision to determine a record classification tree, and the record classification tree corresponds one-to-one with the recognition dimension; segment the record classification tree with distinguishing features and general features to determine a distinguishing classification tree; based on the distinguishing classification tree, determine the dimension detection architecture and the preprocessing samples.

[0055] Further, the preprocessing component 12 is used to execute the following method:

[0056] Use the distinguishing classification tree as the preprocessing sample, traverse the distinguishing classification tree and mine node features to determine a feature classification tree as the dimension detection architecture; traverse the distinguishing classification tree, based on the sample record span corresponding to the tree node, take the maximum span value as the feature recognition intensity; based on the feature recognition intensity, identify the dimension detection architecture.

[0057] Further, the optimization component 14 is used to execute the following method:

[0058] Construct a decision database, which includes a dry process strategy sub-library and a wet process strategy sub-library; traverse the surface defect distribution, take each surface defect as the first-level greedy target, perform greedy search based on the decision database to determine an alternative strategy library; take the optimal alternative as the second-level greedy target, traverse the alternative strategy library to perform greedy search, determine and adjust the initialized greedy strategy to determine a single strategy set, and there is a corresponding relationship between the alternative strategy library, the single strategy set and the surface defect distribution; traverse the single strategy set, perform global iterative optimization based on strategy cross-contamination to determine the surface treatment strategy.

[0059] Further, the optimization component 14 is used to execute the following method:

[0060] Based on the alternative strategy library, a global optimization space is established; a first strategy group and a second strategy group are constructed, the single-item strategy set is stored in the second strategy group, and the first strategy group is initialized to be empty; the second strategy group is traversed, the cross-contamination characteristics of the strategies are identified and marked, and based on the global optimization space, cyclic judgment and strategy transfer are performed to determine the surface treatment strategy.

[0061] Further, the optimization component 14 is used to execute the following method:

[0062] Based on the second strategy group, the first-rank strategy is identified and it is judged whether there is a mark; if there is a mark, the first alternative strategy is determined in combination with the global optimization space, so as to determine the optimized first-rank strategy based on the avoidance of the cross-contamination characteristics and transfer it to the first strategy group; if there is no mark, the first-rank strategy is transferred to the first strategy group; the second-rank strategy based on the second strategy group is determined, and cyclic judgment and transfer are performed until the second strategy group is empty, and the first strategy group is used as the surface treatment strategy, wherein the first-rank strategy has a predecessor node mark based on the second-rank strategy.

[0063] It should be noted that the above sequence of embodiments of the present application is only for description and does not represent the advantages or disadvantages of the embodiments. And the above description of specific embodiments of this specification has been made. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0064] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0065] This specification and the drawings are only exemplary descriptions of the present application and are considered to have covered any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A method for treating a wafer surface before epitaxy, characterized in that: The method comprises: A preset incident angle is set, and a total reflection X-ray scanning is performed on the target wafer based on a fluorescence spectrometer, and a detector captures and generates a three-dimensional mapping spectrum of the target wafer; Calling the surface detection record, taking trace elements, surface chemical state and micro-nano structure as identification dimensions, preprocessing the surface detection record, and data-driven supervised training of the surface detection module, wherein the preprocessing result includes the dimension detection architecture and the preprocessing sample; receiving the three-dimensional mapping spectrum, identifying and integrating defect elements based on the surface detection module, and determining the surface defect distribution, wherein each distributed defect mark has a relative spatial position; Cooperating with the dry treatment method and the wet treatment method, traversing the surface defect distribution to make a single defect treatment decision, performing global optimization with cross infection as a constraint, and determining a surface treatment strategy, wherein the surface treatment strategy includes multiple processing nodes; The equipment group performs pre-positioning and defect treatment on the target wafer in response to the surface treatment strategy; Traversing the surface defect distribution to make a single defect treatment decision, and performing global optimization with cross infection as a constraint, including: Constructing a decision database, wherein the decision database includes a dry strategy sub-library and a wet strategy sub-library; Traversing the surface defect distribution, taking each surface defect as a first-level greedy target, performing greedy search based on the decision database, and determining an alternative strategy library; Taking the best candidate as the secondary greedy target, traversing the candidate strategy library for greedy search, determining and adjusting the initial greedy strategy, and determining a single strategy set, wherein the candidate strategy library, the single strategy set and the surface defect distribution have a corresponding relationship; Traversing the single strategy set, performing global iterative optimization based on strategy cross contamination, and determining the surface treatment strategy; Global iterative optimization based on cross-contamination of strategies, including: Based on the candidate strategy library, a global optimization space is established; Constructing a first policy group and a second policy group, storing the single policy set into the second policy group, and initializing the first policy group to be empty; The second strategy group is traversed, the cross-contamination characteristics of the strategies are identified and marked, and cyclic judgment and strategy transfer are performed based on the global optimization space to determine the surface treatment strategy.

2. A method for treating a wafer surface before epitaxy as claimed in claim 1, characterized in that: The setting of the preset incident angle includes: Determine the incident angle interval according to wafer characteristics and detector performance, wherein the preset incident angle is of small magnitude; Based on the incident angle interval, any incident angle is randomly determined as the preset incident angle, and X-ray total reflection scanning is performed based on the preset incident angle, and a fluorescence signal is received: Verifying the fluorescence signal and making compensation adjustments to the preset incident angle; Wherein, the determination of the incident angle interval includes: Determine a critical incident angle in combination with the incident wavelength, wherein the critical incident angle is positively correlated with the incident wavelength; The incident angle interval is smaller than the critical incident angle.

3. A wafer surface treatment method before epitaxy as claimed in claim 1, characterized in that: Preprocessing the surface detection record includes: Traversing the surface detection records, performing a clustering based on the identified dimension, and determining a dimensional clustering cluster; Identify the dimensional clusters, perform hierarchical clustering with multiple levels of subdivision, and determine a record classification tree, wherein the record classification tree corresponds one-to-one to the identified dimensions; Segmenting the record classification tree based on the distinguishing features and the universal features to determine a distinguishing classification tree; Based on the distinguishing classification tree, the dimension detection framework and the pre-processing samples are determined.

4. A method for treating a wafer surface before epitaxy as claimed in claim 3, characterized in that: Determining the dimension detection architecture and the preprocessing sample includes: Taking the distinguishing classification tree as the preprocessing sample, traversing the distinguishing classification tree and mining node features, determining a feature classification tree as the dimension detection architecture; Traversing the distinguishing classification tree, taking the maximum span value as the feature recognition strength based on the span of the sample records corresponding to the tree nodes; Based on the feature recognition strength, the dimension detection architecture is identified.

5. The method for treating the surface of a wafer before epitaxy as claimed in claim 1, characterized in that: The loop judgment and strategy transfer include: Based on the second strategy group, identifying the first priority strategy and determining whether there is a mark; If there is a mark, determine the first candidate strategy in combination with the global optimization space, determine the optimized first priority strategy based on the avoidance of the cross-contamination feature, and transfer to the first strategy group; If there is no mark, transferring the first priority strategy to the first strategy group; Determine a second priority strategy based on the second priority strategy group, perform cyclic judgment and transfer until the second strategy group is empty, and use the first strategy group as the surface treatment strategy, wherein the first priority strategy carries a predecessor node mark based on the second priority strategy.

6. A wafer surface treatment device before epitaxy, characterized in that: The device is used to implement a method for treating a wafer surface before epitaxy according to any one of claims 1 to 5, comprising: A scanning component, wherein the scanning component is used to set a preset incident angle, perform X-ray total reflection scanning on the target wafer based on a fluorescence spectrometer, and the detector captures and generates a three-dimensional mapping spectrum of the target wafer; A preprocessing component, the preprocessing component is used to call the surface detection record, preprocess the surface detection record with trace elements, surface chemical state and micro-nano structure as identification dimensions, and data-driven supervised training of the surface detection module, wherein the preprocessing result includes the dimension detection architecture and the preprocessing sample; An identification component, the identification component is used to receive the three-dimensional mapping spectrum, identify and integrate defect elements based on the surface detection module, and determine the distribution of surface defects, wherein each distributed defect mark has a relative spatial position; An optimization component, the optimization component is used to coordinate the dry processing method and the wet processing method, traverse the surface defect distribution to make a single defect processing decision, perform global optimization with cross-infection as a constraint, and determine a surface processing strategy, wherein the surface processing strategy includes multiple processing nodes; A surface treatment component is used for the equipment group to respond to the surface treatment strategy and perform pre-positioning and defect treatment on the target wafer.

Images