A method and system for defining an offline wafer defect inspection recipe

Abstract

The application discloses a method and system for defining wafer defect detection recipe offline. The method is executed in the following steps in sequence by judging recipe mode trigger parameter perfecting process in an offline state: corner point marking is performed through interaction at a target chip unit, wafer mapping recipe and alignment recipe are corrected and updated; alignment is automatically performed based on the updated alignment recipe to establish a detection coordinate system; multi-point automatic focusing is performed on the wafer and focus point data is saved; light intensity and camera gain tuning is automatically performed on a target area and the result is written into the detection recipe; microscope light source parameters are automatically corrected. The above steps are realized by event-driven decoupling of the main process and the interactive interface. After completion, the recipe mode is restored to the normal mode and all parameters are saved persistently. The application realizes efficient offline definition of the recipe, does not occupy production time, guarantees the reproducibility and consistency of the recipe, and is compatible with various hardware configurations.

Description

Technical Field

[0001] This invention relates to the fields of semiconductor manufacturing and automated optical inspection technology, and in particular to a method and system for offline definition of wafer defect detection formulations. Background Technology

[0002] In the semiconductor / microelectronics manufacturing field, automated optical inspection (AOI) is the core method for wafer defect detection. The wafer defect detection recipe is the core basis for AOI equipment to achieve accurate detection; the quality and efficiency of the recipe definition directly determine the accuracy of defect detection and the production line's efficiency. The recipe definition process requires key steps such as wafer map generation, alignment parameter calibration, lens focus calibration, light intensity and camera gain optimization, and microscope light source calibration. All steps must be based on good alignment and optical / focusing conditions to ensure the recipe's effectiveness.

[0003] As semiconductor manufacturing moves towards higher precision and higher capacity, production lines are increasingly demanding higher efficiency and stability in wafer defect detection. Traditional online recipe definition methods are no longer sufficient to meet production needs. Online definition requires AOI equipment's mass production scanning time, leading to a reduction in production line cycle time. Meanwhile, manual real-time debugging is susceptible to operational experience and environmental factors, resulting in fluctuations in recipe parameters that are difficult to reproduce, increasing the cost of repeated debugging.

[0004] Currently, the most commonly used method for defining wafer defect detection recipes in the industry is the "online real-time definition method," the core process of which is as follows: 1. Start the mass production scanning mode of the AOI equipment, place the wafer to be inspected at the inspection station, and complete the initial positioning of the wafer; 2. Manually input basic wafer parameters (such as diameter, die size, spacing, etc.) into the equipment operation interface to manually generate a preliminary wafer map; 3. Manually control the equipment to move to the target die position, manually mark the die corner points, adjust the alignment parameters, complete the manual alignment, and establish a coordinate system; 4. Manually control the lens to perform focusing operations at different positions on the wafer, record focus data, and manually enter the formula; 5. Manually adjust optical parameters such as light source intensity and camera gain, observe image clarity with the naked eye, determine the optimal parameters, and save them; 6. After completing all parameter adjustments, manually save the recipe for subsequent mass production scanning.

[0005] Some of the improved solutions support importing GoldenMap files to generate preliminary WaferMaps, but subsequent steps such as alignment, focus finding, and light intensity optimization still need to be completed online, and parameter debugging still relies on manual intervention, thus failing to achieve a complete offline automated process.

[0006] Disadvantages of existing solutions:

[0007] 1. Low efficiency and impact on production line cycle time: Existing technologies require online formulation definition, but in online mode, the equipment needs to occupy mass production scanning resources. The formulation debugging process (such as alignment, focus finding, and parameter optimization) directly occupies production line time, resulting in a reduction in production line cycle time. At the same time, manual debugging of each step (such as manual corner marking, manual focus finding, and manual light intensity optimization) is time-consuming and prone to operational errors, requiring repeated debugging, further reducing efficiency. The core reason is that existing technologies have not achieved offline formulation definition, cannot complete parameter debugging outside of mass production scanning mode, and lack automated debugging processes.

[0008] 2. Difficulty in reproducing formulas and poor consistency: In existing technologies, key parameters such as alignment parameters, focus data, and light intensity gain largely rely on manual recording and input. Differences in manual operation (such as deviations in marking corner points, differences in focus accuracy, and subjective judgments in light intensity testing) can lead to inconsistencies in formula parameters defined by different personnel at different times. At the same time, the parameters lack a standardized and persistent storage mechanism, making them prone to loss or corruption, resulting in formulas that cannot be reproduced. The core reason is that existing technologies lack automated parameter acquisition, calibration, and persistence processes, and parameter accuracy depends on manual operation without a unified parameter management mechanism.

[0009] 3. Poor compatibility and limited adaptability: The existing technology's formula definition process is bound to specific AOI equipment hardware (such as microscope model, light source type, and alignment method). The manually adjusted parameters are only applicable to the current equipment configuration. After changing the equipment or adjusting the hardware configuration, the formula needs to be completely re-adjusted, resulting in high adaptation costs. The core reason is that the existing technology has not decoupled the process from the hardware, lacks a unified control interface and parameter mapping mechanism, and cannot be compatible with different alignment methods, microscopes, and light source configurations.

[0010] 4. High degree of manual intervention and error rate: The core steps of existing technologies (corner marking, alignment, focusing, and light intensity optimization) all require manual intervention. Manual operation is prone to deviations (such as corner marking offset and inaccurate focusing), resulting in abnormal formula parameters and affecting the accuracy of defect detection. At the same time, the high complexity of manual operation and the high professional requirements of operators increase labor costs and operational risks. The core reason is that existing technologies have not achieved an automated closed loop of the process and lack event-driven automatic interaction and parameter feedback mechanisms. Summary of the Invention

[0011] This invention aims to solve the technical problems of existing wafer defect detection recipe definition processes, which rely on online sampling or manual real-time operation at the workstation, resulting in repetitive recipe debugging, low definition efficiency, difficulty in parameter reproduction, and online debugging occupying production line mass production time and affecting production cycle. At the same time, it solves the problems of poor compatibility of existing recipe definition methods, difficulty in adapting to different alignment methods, microscope and light source configurations, and inability to effectively persist and reuse parameters.

[0012] To achieve the above objectives, this invention provides a method for defining wafer defect detection recipes offline. The core concept is to construct a control logic based on a core control module and event-driven and user interface interaction. In an offline state completely detached from the mass production scanning mode of the equipment, the method automatically and systematically completes the entire parameter definition process from wafer mapping generation, corner mark correction, alignment, multi-point focusing, light intensity gain optimization to microscope light source optimization by importing or inputting basic parameters. Finally, all parameters are persistently saved to generate a standard defect detection recipe that can be directly used for mass production scanning.

[0013] Specifically, the method includes the following steps: After receiving the recipe definition instruction, determine whether the current recipe mode is the preset fabless mode or gold mapping mode. If so, proceed to the offline parameter refinement process, which is executed sequentially: Move the detection device to the target chip cell, trigger and display the corner marking window, receive the corner marking results and offset of the target chip cell obtained through the window interaction, and update the wafer mapping recipe and alignment recipe accordingly. Based on the updated alignment formula, the alignment execution thread is started according to the preset alignment mode to establish a unified detection coordinate system; Calculate the coordinates of multiple focus points on the wafer, perform an automatic focus operation for each focus point, acquire focus data, and save it to the wafer mapping recipe; In the target chip unit area, an automatic light intensity and gain tuning window is triggered and displayed, automatically searching for and obtaining the optimal light source intensity and camera gain parameters, writing the tuning results back to the optical parameters of the detection formula and saving them; Trigger and display the microscope light source auto-adjustment window to complete the automatic correction of microscope light source parameters; Once the above steps are successfully executed, the current recipe mode is restored to normal mode, and all updated recipe parameters are saved to the storage medium through the data persistence interface.

[0014] Preferably, the triggering and displaying of the window, the execution of automated operations within the window, and the receiving of interaction results are all implemented through an event aggregator by publishing and subscribing to events, thereby decoupling the main process from the user interface window.

[0015] Preferably, the alignment mode includes a marker alignment mode based on alignment marker coordinates and an edge alignment mode based on wafer edge features.

[0016] Preferably, the automatic light intensity and gain optimization step specifically includes: constructing a detection area covering the target chip unit area, moving the device to the center of the area, automatically searching and obtaining the light intensity optimization results containing multiple channels, and updating the optimal light intensity value of each channel to the optical parameters of the detection formula according to the preset channel-light source color mapping relationship.

[0017] Preferably, the process further includes a fault tolerance mechanism: if a failure state is detected when executing any of the steps, the subsequent process is terminated, the current recipe mode is rolled back to the normal mode, and the currently completed parameter configuration is saved.

[0018] The advantages of the present invention through the above technical solution are as follows: 1. Enable offline definition of wafer defect detection formulas, eliminating the need for mass production scanning mode, avoiding production line time occupation, and improving production line cycle time; 2. Build an automated recipe definition process to automate corner marking, alignment, focusing, light intensity / gain optimization, and microscope light intensity optimization, thereby reducing manual intervention, lowering the error rate, and improving recipe definition efficiency; 3. Establish a standardized parameter persistence mechanism to achieve complete storage and one-click reuse of formula parameters, ensuring the reproducibility and consistency of the formula; 4. Decouple the process from the hardware, and be compatible with different alignment methods (marker alignment, edge alignment), microscope models and multi-channel light source configurations to improve the adaptability and versatility of the formula; 5. Simplify the formula definition process, reduce the professional requirements for operators, and lower labor costs and operational risks. Attached Figure Description

[0019] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0020] Figure 1 This is a schematic diagram of the overall process of the offline wafer defect detection formulation method of the present invention.

[0021] Figure 2 This is a schematic diagram of the initial parameter input / import in this invention.

[0022] Figure 3 This is a schematic diagram of the target die corner marking and parameter correction in this invention.

[0023] Figure 4 This is a schematic diagram of the corner point marking interaction in this invention.

[0024] Figure 5 This is a schematic diagram illustrating the initial adjustment of feeding error in this invention.

[0025] Figure 6 This is a schematic diagram of automatic light intensity / camera gain optimization in this invention.

[0026] Figure 7 This is a schematic diagram illustrating the event-driven and UI interaction principles in this invention. Detailed Implementation

[0027] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0028] I. Overall Overview of the Technical Solution

[0029] This invention provides a method for defining wafer defect detection recipes offline. Using RecipeManager as the core control module and event-driven and UI interaction as control methods, basic parameters are generated by offline input or import of GoldenMap files. The process sequentially completes corner marking and parameter correction, alignment, five-point focusing, automatic light intensity / gain adjustment, and microscope light intensity optimization. Finally, all parameters are persistently saved, generating a defect detection recipe that can be directly used for mass production scanning. The entire process is completed in offline scanning mode, without occupying mass production line resources, and achieves full automation, compatible with various hardware configurations.

[0030] II. Overall Flowchart (Flow Logic Description)

[0031] The technical process of this invention consists of 8 core steps, which are executed sequentially to form a closed loop. The flowchart is as follows: Figure 1 As shown.

[0032] Figure 1This invention demonstrates the complete process of defining wafer defect detection recipes offline, with the "offline recipe parameter improvement process" being the core component. All steps are completed in offline scanning mode. Each step is executed sequentially. If any step fails, the process can be stopped and rolled back to Normal mode, saving the currently completed parameter configurations. After the process is completed, the improved recipe can be loaded to enter the normal defect detection process.

[0033] 1. Initial parameter input / import: Input parameters offline via RecipeManager (Method A) or import GoldenMap file (Method B) to generate WaferMapRecipe and initial alignment parameters; 2. Recipe Mode Judgment: When loading the offline-defined Recipe for the first time, check whether MainRecipe.RecipeMode is Waferless or GoldenMap. If so, proceed to the offline recipe parameter improvement process; otherwise, proceed directly to the normal testing process. 3. Target Die Corner Marking and Parameter Correction: Move to the specified target die position, a marking window pops up, obtain the corner marking results and offset, and update WaferMapRecipe and AlignmentRecipe (alignment recipe). 4. Alignment process execution: Set alignment parameters according to the preset alignment mode (mark alignment or edge alignment), start the alignment thread, complete the alignment and establish a unified coordinate system; 5. Five-point focusing and focus saving: Calculate the five-point focusing coordinates of the wafer, pop up the focusing window to perform the focusing operation, and save the focus data to WaferMapRecipe after confirmation; 6. Automatic tuning of light intensity / camera gain: A tuning window pops up in the target die area, performs automatic search for light intensity and camera gain, writes the tuning results back to the optical parameters of TestsRecipe and saves them; 7. Automatic Microscope Light Source Optimization: A microscope light intensity optimization window pops up at the center of the target die to automatically correct microscope-level light source parameters; 8. Recipe Persistence and Mode Restoration: Restore MainRecipe.RecipeMode to Normal, and persistently save all updated recipe parameters (WaferMapRecipe, AlignmentRecipe, TestsRecipe, etc.) through RecipeAssist to complete the offline recipe definition; subsequently, loading this recipe will allow the normal defect detection process to begin.

[0034] III. Detailed Explanation of Each Step (Including Implementation Details)

[0035] 3.1 Step 0: As Figure 2 As shown, the initial parameter input / import (WaferMap and alignment parameter generation)

[0036] This step forms the basis for offline recipe definition, providing two parameter input methods, both completed offline and without consuming production line scanning resources:

[0037] Method A: Generate WaferMap offline using RecipeManager input parameters Users can directly set key fields of RecipeAssist.Instance.CurrentRecipeData.WaferMapRecipe in the recipe editing interface or script, including DiePitchMark1X / DiePitchMark1Y (coordinates of the first die spacing marker), DiePitchMark2X / DiePitchMark2Y (coordinates of the second die spacing marker), DiePitchMark3X / DiePitchMark3Y (coordinates of the third die spacing marker), Diameter (wafer diameter), StreetSizeX / StreetSizeY (street size), DieOriginX / DieOriginY (coordinates of the die origin), etc. The system automatically generates a complete WaferMap (containing the coordinates, indexes, and status of all dies) by calculating the coordinate relationship of the initial three points DiePitchMark1 / 2 / 3.

[0038] Method B: Import GoldenMap to generate WaferMap and alignment parameters The GoldenMap file stores parameters in a Key:Value format. Core fields include: basic wafer parameters (Diameter, Flat Type), Die parameters (Die Width, Die Height, Die Count X / Y, Die Array Center X / Y, Shot Row / Column Die Count, Center Die Shot Row / Column), and DiePosition List (recording the index, coordinates, and status of each die in X, Y, and Status rows). After reading this file, the system calculates the coordinates and status of all dies using the parameters, generates a WaferMap, and extracts relevant parameters as initial alignment parameters, storing them in AlignmentRecipe.

[0039] 3.2 Step 1: Formula Mode Determination

[0040] The entry point for the process is WaferlessRecipeFlow.Setup(IEventAggregator ea). This method first checks the value of MainRecipe.RecipeMode: if it is Waferless or GoldenMap, it indicates that the current mode is offline recipe definition, and the process proceeds to the subsequent parameter improvement stage; if it is Normal, it indicates that the recipe is already complete, and the recipe is directly loaded to enter the normal defect detection process. MainRecipe is the core recipe entity, containing sub-recipes such as AlignmentRecipe, WaferMapRecipe, and TestsRecipe. Its RecipeMode field is used to control the process branches.

[0041] 3.3 Step 2: Target Die Corner Marking and Parameter Correction

[0042] like Figure 3 As shown, step 2 includes the following process:

[0043] 1. Target Die Selection: If RecipeMode is Waferless, the system automatically selects the die with XIndex==0 and YIndex==0 in WaferMapRecipe.DieList as the target die; if it is GoldenMap, the base die specified in GoldenMap is selected.

[0044] 2. Location movement: Move the device to the specified location of the target die using the RecipeStageOperate.Instance.MoveToCustomXY(x,y) method.

[0045] 3. Corner marker interaction: such as Figure 4 As shown, the OpenWindowCommandEvent event is published through IEventAggregator, which calls MarkDieCornerWindow (corner marking window). The user completes the corner marking of the target die in the window. The window returns the marking result (_markCorner) and offset (_markOffset), and returns it through WindowSetVariableCommandEvent event. It is then processed by SetVariableEventHandler to update the marking state (_isMarkOK).

[0046] 4. Parameter Correction: After successful marking, two core correction methods are called: (1) CorrectWaferMapParam: Recalculate the coordinates of WaferMapRecipe.DiePitchMark1 / 2 / 3 based on the marked corner point (_markCorner), call the RecipeAssist.Instance.CreateWaferMapBy3Pt(...) method, regenerate the Die layout, update and save WaferMapRecipe; (2) CorrectAlignParam: Using the left and right Die coordinates and the mark offset (_markOffset), update AlignmentRecipe.AlignParamList[0].Mark1X / Mark1Y / Mark2X / Mark2Y (alignment mark coordinates), complete the alignment parameter correction and save.

[0047] 3.4 Step 3: Alignment Process Execution

[0048] 1. Alignment Mode Selection: Supports two alignment modes, which can be configured according to actual needs, such as... Figure 5 As shown, preliminary adjustments were made to the feeding error: (1) AlignMode.ByMark (Mark Alignment): Aligns the markers based on the corrected Mark1 / Mark2 coordinates in AlignmentRecipe; (2) AlignMode.ByEdge: Creates EdgeAlignParam (edge ​​alignment parameter) using the CreateEdgeAlignMark method, and performs alignment based on the wafer edge features.

[0049] 2. Preprocessing: Set the microscope lens parameters and the initial intensity of the light source, and reset the coordinate system using the CoordinateManager.Instance.CreateCoordinate() method to prepare for alignment.

[0050] 3. Alignment Execution: Assign AlignmentRecipe to AlignmentFlow.AlignmentRecipe, initialize calibration using AlignmentFlow.evInitCalibration.Set(), and start the AlignmentFlow.AlignmentStartThread() thread to execute the alignment process. The system waits for AlignmentFlow.IsAlignmentEnd (alignment end flag) and checks AlignmentFlow.IsAlignmentOK (alignment success flag). If alignment fails, the process is aborted and rolled back.

[0051] 4. Coordinate system establishment: After successful alignment, the CoordinateManager.Instance.CreateCoordinate() method establishes a unified detection coordinate system to ensure the consistency of coordinates for subsequent focusing, optimization, and detection.

[0052] 3.5 Step 4: Five-point focus finding and focus saving

[0053] 1. Focusing coordinate calculation: The system automatically calculates the coordinates of the five focusing points relative to the wafer center (usually one point at the wafer center and four points around the perimeter), converts them into device stage coordinates, and stores them in WaferMapRecipe.PointCoordinates.

[0054] 2. Focusing operation: Publish the ExecuteCommandEvent event through IEventAggregator to pop up the FivePointCalFocusWindow (five-point focus window). The window automatically executes the focus process and detects the lens focus data of each focus point.

[0055] 3. Data saving: After successful focus acquisition, the window returns LensFocusList (focus data list), which the system saves to MainRecipe.WaferMapRecipe.LensFocusList. At the same time, WaferMapRecipe is updated and saved to provide stable focus parameters for subsequent detection.

[0056] 3.6 Step 5: Automatic adjustment of light intensity / camera gain, such as... Figure 6 As shown

[0057] 1. Detection region construction: Construct a DieRegion (Die detection region) with the same width and height as the DiePitch (Die spacing), and include a multi-pitch list to ensure that the optimization range covers the entire target Die region.

[0058] 2. Location Relocation: Move the device to the center of the target die to ensure the accuracy of the tuning area.

[0059] 3. Automatic tuning: The LightIntensityAutoTuneWindow pops up and automatically performs a search for light intensity and camera gain to obtain the IntensityTuneResult.

[0060] 4. Parameter write-back and saving: Fill IntensityTuneResult into TestsRecipe.TestList[0].OpticsParam.IntensityTuneResult, call the AutoTuneResultSetIntensity(opticParam) method to update IntensityByPercent (light intensity percentage), set CameraGain (camera gain) and GainTuneResult (gain tuning result), and finally save TestsRecipe to ensure the accuracy of optical parameters.

[0061] 3.7 Step 6: Automatic adjustment of microscope light source

[0062] At the center of the target die, the MicroscopeLightIntensityAutoTuneWindow is called via IEventAggregator. The window automatically corrects the light source parameters at the microscope level and returns the tuning result (_isMicroscopeLightTuneOK) after the correction is completed, ensuring that the light source parameters are compatible with the microscope and improving the clarity of the detection image.

[0063] 3.8 Steps 7-8: Formula persistence and pattern recovery, and start of the testing process

[0064] 1. Mode restoration: MainRecipe.RecipeMode is restored from Waferless / GoldenMap to Normal, marking the recipe as complete.

[0065] 2. Persistent storage: The RecipeAssist.Instance.Save(XXX, name, nameof(XXX),false) method saves all sub-recipes such as WaferMapRecipe, AlignmentRecipe, and TestsRecipe, as well as the parameters of MainRecipe, to persistent media (such as local device storage or server) to ensure that recipe parameters are not lost and can be directly reused in subsequent mass production scans.

[0066] 3. Process completion: After loading the improved formula, the equipment officially enters the defect detection process in Normal mode, without the need for further parameter adjustments.

[0067] IV. Event-Driven and UI Interaction Principles

[0068] like Figure 7As shown, this invention uses an event-driven mechanism to implement process control and UI interaction, with its core relying on IEventAggregator (event aggregator), mainly involving three types of events: 1. OpenWindowCommandEvent: Used to publish commands to open interactive windows, such as opening MarkDieCornerWindow, FivePointCalFocusWindow, etc.; 2. ExecuteCommandEvent: Used to publish commands to execute in the window, triggering automated operations within the window (such as focus finding and optimization). 3. WindowSetVariableCommandEvent: Used for the window to send interaction results (such as marking results, focus data, and tuning parameters) back to the main process. It is received and processed by SetVariableEventHandler to update the process status and parameters.

[0069] This mechanism decouples the main process from the UI window, ensuring the smooth execution of automated processes while supporting manual intervention (such as manually adjusting parameters within the window), thus improving flexibility.

[0070] V. Description of Core Data Structures

[0071] The core data structure is MainRecipe, which contains multiple sub-recipes. The key fields and relationships of each sub-recipe are as follows: 1. MainRecipe (Core Recipe Entity): Contains sub-recipes such as AlignmentRecipe, WaferMapRecipe, DieRegionRecipe, and TestsRecipe. The RecipeMode field (values: Normal / Waferless / GoldenMap) controls the flow branches. 2. WaferMapRecipe: Includes DieList (position and index information of all dies), DiePitchX / Y (die spacing), Diameter (wafer diameter), PointCoordinates (five-point focus coordinates), DiePitchMark1 / 2 / 3 (three-point marker coordinates), and LensFocusList (five-point focus results); 3. AlignmentRecipe: Includes AlignParamList (alignment parameter list), AlignMode (alignment mode), and Mark1X / Mark1Y / Mark2X / Mark2Y (alignment marker coordinates); 4. TestsRecipe / OpticsParam (Detection / Optical Parameters): Includes IntensityTuneResult (Light Intensity Optimization Result), CameraGain (Camera Gain), IntensityByPercent (Light Intensity Percentage), etc.

[0072] The advantages of this invention can be directly derived from the disadvantages of existing technologies and the technical solution of this invention: Because this invention enables offline definition of the formula, all parameter debugging processes are separated from the online mass production scanning mode, eliminating the need to occupy production line resources. This effectively solves the problem of existing technologies' online debugging affecting production line cycle time, significantly improving production line efficiency. Simultaneously, this invention automates steps such as corner marking, alignment, focusing, and light intensity optimization through event-driven processes, reducing manual intervention, avoiding the differences and errors inherent in manual operations, and solving the problems of poor formula reproducibility and consistency in existing technologies. It also reduces labor costs and operational risks. Furthermore, by decoupling the process from hardware, this invention is compatible with different alignment methods, microscopes, and multi-channel light sources, solving the problem of poor compatibility in existing technologies and improving the universality and adaptability of formulas. The standardized parameter persistence mechanism enables one-click reuse of formulas, further reducing the time and cost of repeated debugging, and overall improving the efficiency and quality of wafer defect detection formula definition.

[0073] Regarding the technical solution of this invention, the following alternative solutions can achieve the same inventive objective, and all fall within the protection scope of this invention: 1. Alternative to offline parameter input method: In addition to method A (offline parameter input) and method B (import GoldenMap), an "import historical recipe parameters" method can be added. By loading existing offline recipe parameters, modifying some key parameters to complete the definition of a new recipe, offline initialization can also be achieved, reducing the workload of parameter input. 2. Alternatives to focus finding methods: Five-point focus finding can be replaced by three-point focus finding, seven-point focus finding, or other focus finding methods, as long as stable lens focus data can be obtained through multi-point focus finding to ensure detection sharpness, the same purpose can be achieved; the calculation method of focus finding coordinates can be replaced by other algorithms based on Die array distribution without affecting the focus finding effect; 3. Event-driven approach alternative: Message queues can be used instead of IEventAggregator to implement the interaction between the main process and the UI window. The core logic (window invocation, result return) remains the same, and automated process control can also be achieved. 4. Alternatives for light intensity optimization: Automatic optimization of light intensity and camera gain can be achieved using different search algorithms (such as genetic algorithms and gradient descent algorithms). As long as the optimal matching of optical parameters can be achieved and image clarity can be improved, the same effect can be achieved. 5. Alternative to persistent storage: Formula parameters can be saved to a cloud server to replace local storage, enabling formula sharing across multiple devices. This also achieves parameter persistence and reuse without affecting the achievement of the invention's purpose.

[0074] VI. Key Information in Code Implementation

[0075] 1. Main entry point: WaferlessRecipeFlow.Setup(IEventAggregator ea), located in the WaferlessRecipeFlow.cs file; 2. Core auxiliary methods: MarkDieCorner(MainRecipe), Alignment(AlignMode), FivePtsSearchFocus(), AutoTuning(), MicroscopeAutoTuning(), CorrectWaferMapParam, CorrectAlignParam, AutoTuneResultSetIntensity, etc., are all located in WaferlessRecipeFlow.cs; 3. Location of core data structures: The fields and storage logic of MainRecipe and WaferMapRecipe are located in Subsystems / RecipeManager / RecipeEntity / MainRecipe.cs and WaferMapRecipe.cs, respectively; 4. Alignment Implementation Details: Handled by AlignmentFlow (Subsystems / RecipeManager / RecipeEntity / RecipeUtility / AlignmentFlow.cs), including multi-lens / moving or fixed camera alignment, nine-grid search, image matching, etc. 5. Event interaction details: IEventAggregator is used to publish / subscribe to OpenWindowCommandEvent, ExecuteCommandEvent, and WindowSetVariableCommandEvent. All interaction results are handled uniformly by SetVariableEventHandler. 6. Persistence Implementation: The parameters of each sub-recipe are saved to persistent media through the RecipeAssist.Instance.Save(...) method, supporting both local storage and server storage.

[0076] VII. Hardware Compatibility Instructions

[0077] 1. Compatible microscope types: Supports mainstream industrial microscopes, and decouples from different microscope models through MicroscopeController, without requiring modification of the core process; 2. Light source adaptation: Supports multi-channel light sources. The channel type is mapped to IlluminationColors through _dict_LightTypeConversion, and the corresponding light source intensity is automatically set. 3. Alignment devices: Supports both moving and fixed cameras for alignment. AlignmentFlow can automatically adjust the alignment logic according to the device type. 4. Stage: Supports stages with different travel distances, and adapts to the coordinate system of different devices through the RecipeStageOperate.Instance.MoveToCustomXY(...) method.

[0078] VIII. Process Fault Tolerance and Exception Handling

[0079] 1. Exception handling mechanism: Each step is equipped with a failure judgment (such as marking failure, alignment failure, focus failure). After failure, the process can be stopped and rolled back to Normal mode, saving the currently completed parameter configuration to avoid parameter loss. 2. Manual intervention interface: All automated steps have a reserved manual intervention interface. If the automated operation fails, parameters can be manually entered through the UI window to continue to complete the recipe definition; 3. Parameter validation: The system automatically validates the input parameters (such as GoldenMap file format and Die coordinate range). If the parameters are abnormal, a prompt will pop up and the process will be stopped to avoid generating invalid recipes.

[0080] IX. Description of Practical Application Scenarios

[0081] This invention is mainly applied to the recipe management of AOI inspection equipment for semiconductor wafers. It is suitable for defect detection of mainstream wafer sizes such as 12-inch and 8-inch wafers and can be widely used in the manufacturing process of various semiconductor products such as logic chips, memory chips, and power chips. It is especially suitable for scenarios with tight production line capacity, high requirements for recipe reusability, and diverse equipment configurations. It can significantly reduce recipe debugging time and improve production line efficiency and inspection consistency.

[0082] 10. Glossary of relevant technical terms

[0083] 1. Wafer Defect Detection Recipe: This refers to a set of parameters used by AOI equipment to detect wafer defects, including wafer mapping parameters, alignment parameters, optical parameters (light intensity, camera gain), focus parameters, etc. It is the core basis for the equipment to achieve accurate detection. 2. Offline definition: refers to the mass production scanning mode without starting the equipment, completing the parameter input, debugging, calibration and saving of the formula in a non-production state, without occupying production line resources; 3. GoldenMap: A standard file containing information such as basic wafer parameters, die layout parameters, and die status. It is used to quickly generate a wafer map and ensure the accuracy of wafer mapping. 4. Wafer Map: Used to describe the mapping relationship of the location, index, and status (such as pass, fail) of all dies on the wafer. It is the basis for recipe definition and defect detection. 5. Die (chip die): An independent chip unit separated from the wafer, which is the basic unit for defect detection; 6. Alignment: By using markers or edge features, the relative position of the wafer and the inspection equipment is adjusted to establish a unified coordinate system and ensure the accuracy of the inspection position; 7. Search Focus: Adjust the lens focal length to make the image of the inspection area (Die) as clear as possible, ensuring the accuracy of defect detection; 8. Automatic Light Intensity / Gain Tuning: The algorithm automatically searches for the optimal light source intensity and camera gain parameters to achieve the best contrast and clarity of the detected image, thereby improving the defect recognition rate. 9. Event-driven: Centered on events, it enables interaction between the main process, UI window, and hardware modules. By publishing / subscribing to events, it triggers relevant operations and receives results, thereby automating the process. 10. Persistent storage: Save the formula parameters to storage media (local or server) to ensure that the parameters are not lost and can be directly loaded and reused in subsequent tests; 11. IEventAggregator: Used to manage the publishing and subscription of events, enabling decoupled communication between different modules; 12. Stage: Used to place wafers and enable precise movement of wafers, it is one of the core hardware components of AOI equipment.

[0084] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for defining wafer defect detection formulations offline, characterized in that, include: In the offline state where mass production scanning mode is not started, determine whether the formulation mode of the formulation to be improved is fabless mode or gold mapping mode. If so, then the offline parameter improvement process is executed, which includes the following steps performed in sequence: a) Control the detection equipment to move to the target chip unit, trigger corner marking interaction, receive the corner marking results and offset obtained through the interaction, and correct and update the wafer mapping recipe and alignment recipe based on the corner marking results and offset; b) Based on the updated alignment formula, and according to the preset alignment mode, start the alignment execution thread to establish a unified detection coordinate system; c) Calculate the coordinates of multiple focus points on the wafer, perform a focus operation at each focus point, and save the obtained focus data to the wafer mapping recipe; d) In the target chip unit area, trigger light intensity and gain tuning interaction to search for and determine the optimal light source intensity parameters and camera gain parameters, and write the tuning results into the optical parameters of the detection formula and save them; e) Trigger the microscope light source tuning interaction to complete the calibration of the microscope light source parameters; Once all steps in the offline parameter improvement process have been successfully executed, the recipe mode of the recipe to be improved is restored to the normal mode, and all updated recipe parameters are saved to the storage medium through the data persistence interface.

2. The method for defining an offline wafer defect detection formula according to claim 1, characterized in that, The triggering of corner marker interactions, the execution of operations, and the receiving of interaction results are all implemented through an event aggregator by publishing and subscribing to events. The events include a first event for opening the interaction window, a second event for triggering operations within the window, and a third event for sending the interaction results back to the main process.

3. The method for offline definition of wafer defect detection formulation according to claim 1, characterized in that, The preset alignment modes include a marker alignment mode based on the corrected alignment marker coordinates in the alignment recipe, or an edge alignment mode based on wafer edge features.

4. The method for offline definition of wafer defect detection formulation according to claim 1, characterized in that, The correction and updating of the wafer mapping recipe and alignment recipe in step a specifically includes: Using the received corner marker results, the coordinates of the multi-point markers in the wafer mapping recipe are recalculated and updated to regenerate the chip cell layout; The alignment mark coordinates in the alignment recipe are updated using the received corner mark results and offsets.

5. The method for offline definition of wafer defect detection formulation according to claim 1, characterized in that, Step c, which involves calculating the coordinates of multiple focus points on the wafer and performing focus finding, specifically includes: calculating the coordinates of five focus points relative to the center of the wafer, and detecting lens focus data at each coordinate point.

6. The method for offline definition of wafer defect detection formulation according to claim 1, characterized in that, Step d involves triggering an interaction between light intensity and gain tuning to search for and determine the optimal light source intensity parameters and camera gain parameters, and then writing the tuning results into the optical parameters of the detection formula. Specifically, this includes: Construct a detection area covering the target chip unit area, and move the device to the center of that area; Trigger the light intensity optimization window, perform light intensity and camera gain search, and obtain optimization results containing multiple channels; Based on the preset channel-light source color mapping relationship, the optimal light intensity value corresponding to each channel is filled into the optical parameters of the detection formula and saved.

7. The method for defining an offline wafer defect detection formula according to claim 1, characterized in that, Also includes: If an operation failure is detected during any step of the offline parameter improvement process, the subsequent process is terminated, the recipe mode of the recipe to be improved is rolled back to the normal mode, and the currently completed parameter configuration is saved.

8. The method for defining an offline wafer defect detection formula according to claim 1, characterized in that, Before determining the formulation pattern of the formulation to be improved, an initialization step is also included, which includes: Initial wafer mapping recipes and / or alignment parameters can be generated by inputting parameters offline or by importing a gold mapping file containing basic wafer parameters, chip cell layout parameters, and a list of chip cell locations.

9. A system for defining wafer defect detection formulas offline, characterized in that, include: processor; And a memory having a computer program stored thereon, which, when executed by the processor, implements the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 8.

Images