Semiconductor yield root cause determination method and device, equipment and storage medium

Abstract

The invention provides a semiconductor yield root cause determination method and device, equipment and a storage medium, and the method comprises the steps: obtaining original data corresponding to a target manufacturing process, and carrying out the preprocessing of the original data, and obtaining target data; for each initial feature value in the target data, determining suspicion degree data corresponding to the initial feature value according to a preset index calculation mode; deleting the initial feature values with the suspicion degree data meeting a preset exclusion condition in the target data to obtain alternative feature values; and determining a target feature value corresponding to the target manufacturing process from the alternative feature values according to the suspicion degree data of each alternative feature value. In this way, by calculating the doubt degree data based on the preset index calculation mode and performing feature elimination according to the preset elimination condition, the finally determined target feature value can better conform to the service logic of a semiconductor actual scene, and the accuracy of yield root cause determination can be improved.

Description

Technical Field

[0001] This application relates to the field of semiconductor manufacturing technology, and in particular to a method, apparatus, device and storage medium for determining the root causes of semiconductor yield. Background Technology

[0002] Semiconductor manufacturing is a highly precise engineering technology with a complex production process, typically involving hundreds or thousands of process steps. Each step involves different equipment, chambers, and recipes. Therefore, when semiconductor yield issues arise, it is often difficult to quickly pinpoint the root cause from a large amount of data manually. The root cause refers to the fundamental reason for the reduced yield in semiconductor manufacturing, specifically an abnormal processing step. This abnormal processing step usually manifests as an unusual change in indicators throughout the entire semiconductor manufacturing process, or an unusual change in indicators during intermediate manufacturing steps.

[0003] To quickly determine the root cause of yield issues, related technologies typically rely on statistical data processing methods, using statistical indicators to identify the root cause. However, this method ultimately determines the root cause of yield issues in a way that does not align with the business logic of actual semiconductor scenarios, resulting in low accuracy in determining the root cause of yield issues. Summary of the Invention

[0004] This application provides a method, apparatus, device, and storage medium for determining the root cause of semiconductor yield, which enables the finally determined root cause of yield to conform to the actual business logic of the semiconductor manufacturing scenario, thereby improving the accuracy of the root cause determination.

[0005] In a first aspect, embodiments of this application provide a method for determining the root causes of semiconductor yield, including:

[0006] Obtain the raw data corresponding to the target manufacturing process, and preprocess the raw data to obtain the target data;

[0007] For each initial feature value in the target data, the degree of suspicion corresponding to the initial feature value is determined according to a preset index calculation method;

[0008] The initial feature values ​​of the suspicion level data that meet the preset exclusion conditions are deleted from the target data to obtain alternative feature values;

[0009] Based on the degree of suspicion data for each candidate feature value, the target feature value corresponding to the target manufacturing process is determined from the candidate feature values.

[0010] In one possible implementation, the raw data includes yield index data and manufacturing process data;

[0011] The yield index data includes at least one of the wafer test data corresponding to the target manufacturing process and the defect quantity data in the target manufacturing process; the data type of the manufacturing process data includes at least one of the following: site identifier, machine identifier, chamber identifier, process formula identifier, and material identifier.

[0012] In one possible implementation, the preprocessing of the raw data to obtain the target data includes:

[0013] The original data is converted to obtain alternative data;

[0014] Based on the yield index data, determine the target status identifier corresponding to each wafer in the target manufacturing process;

[0015] The target status identifier is added to the alternative data to obtain the target data.

[0016] In one possible implementation, determining the degree of suspicion data corresponding to the initial feature value according to a preset index calculation method includes:

[0017] Based on a preset analysis of variance method, the various feature types in the target data are analyzed and processed to obtain significant difference parameters corresponding to each initial feature value in the feature type; and / or,

[0018] Determine the ratio of the number of abnormal wafers corresponding to the initial feature value to the total number of wafers corresponding to the initial feature value to obtain the first abnormality ratio corresponding to the initial feature value; and / or,

[0019] The ratio of the number of abnormal wafers corresponding to the initial feature value to the total number of abnormal wafers corresponding to the feature type to which the initial feature value belongs is determined to obtain the second abnormality ratio corresponding to the initial feature value; and / or,

[0020] The ratio of the target sum corresponding to the initial feature value to the total number of wafers corresponding to the feature type to which the initial feature value belongs is determined to obtain the common proportion of abnormal wafers corresponding to the initial feature value; the target sum is the sum of the number of abnormal wafers corresponding to the initial feature value and the number of normal wafers corresponding to other initial feature values ​​in the feature type to which the initial feature value belongs, excluding the initial feature value.

[0021] In one possible implementation, the method further includes:

[0022] If the significant difference parameter of the initial feature value is greater than a first preset threshold, then the initial feature value is determined to meet the preset exclusion condition; or,

[0023] For each feature type to which the initial feature value belongs, a first initial feature value with the largest number of abnormal wafers within that feature type is determined. If the first abnormality ratio corresponding to the first initial feature value is less than a second preset threshold, then it is determined that each initial feature value corresponding to the feature type satisfies the preset exclusion condition; or...

[0024] If the second anomaly ratio corresponding to the first initial feature value is greater than the third preset threshold, and the ratio of the number of wafers corresponding to the first initial feature value to the total number of wafers corresponding to the feature type is greater than the fourth preset threshold, then it is determined that each initial feature value in the feature type satisfies the preset exclusion condition; or,

[0025] If the proportion of abnormal wafers corresponding to the first initial feature value is less than the fifth preset threshold, then it is determined that each initial feature value corresponding to the feature type satisfies the preset exclusion condition.

[0026] In one possible implementation, determining the target feature value corresponding to the target manufacturing process from the candidate feature values ​​based on the suspicion level data of each candidate feature value includes:

[0027] For each candidate feature value, the suspicion level data of the candidate feature value is weighted and summed to obtain the target suspicion level data corresponding to the candidate feature value;

[0028] The target feature value is determined from the candidate feature values ​​based on the target suspicion level data.

[0029] Secondly, embodiments of this application provide a semiconductor yield root cause determination apparatus, comprising:

[0030] The acquisition module is used to acquire the raw data corresponding to the target manufacturing process and preprocess the raw data to obtain the target data;

[0031] The first determining module is used to determine the degree of suspicion data corresponding to each initial feature value in the target data according to a preset index calculation method.

[0032] The deletion module is used to delete the initial feature values ​​of the suspicion level data that meet the preset exclusion conditions in the target data, and obtain alternative feature values;

[0033] The second determining module is used to determine the target feature value corresponding to the target manufacturing process from the candidate feature values ​​based on the suspicion level data of each candidate feature value.

[0034] In one possible implementation, the raw data includes yield index data and manufacturing process data;

[0035] The yield index data includes at least one of the wafer test data corresponding to the target manufacturing process and the defect quantity data in the target manufacturing process; the data type of the manufacturing process data includes at least one of the following: site identifier, machine identifier, chamber identifier, process formula identifier, and material identifier.

[0036] In one possible implementation, the acquisition module is specifically used for:

[0037] The original data is converted to obtain alternative data;

[0038] Based on the yield index data, determine the target status identifier corresponding to each wafer in the target manufacturing process;

[0039] The target status identifier is added to the alternative data to obtain the target data.

[0040] In one possible implementation, the first determining module is specifically used for:

[0041] Based on a preset analysis of variance method, the various feature types in the target data are analyzed and processed to obtain significant difference parameters corresponding to each initial feature value in the feature type; and / or,

[0042] Determine the ratio of the number of abnormal wafers corresponding to the initial feature value to the total number of wafers corresponding to the initial feature value to obtain the first abnormality ratio corresponding to the initial feature value; and / or,

[0043] The ratio of the number of abnormal wafers corresponding to the initial feature value to the total number of abnormal wafers corresponding to the feature type to which the initial feature value belongs is determined to obtain the second abnormality ratio corresponding to the initial feature value; and / or,

[0044] The ratio of the target sum corresponding to the initial feature value to the total number of wafers corresponding to the feature type to which the initial feature value belongs is determined to obtain the common proportion of abnormal wafers corresponding to the initial feature value; the target sum is the sum of the number of abnormal wafers corresponding to the initial feature value and the number of normal wafers corresponding to other initial feature values ​​in the feature type to which the initial feature value belongs, excluding the initial feature value.

[0045] In one possible implementation, the device is further used for:

[0046] If the significant difference parameter of the initial feature value is greater than a first preset threshold, then the initial feature value is determined to meet the preset exclusion condition; or,

[0047] For each feature type to which the initial feature value belongs, a first initial feature value with the largest number of abnormal wafers within that feature type is determined. If the first abnormality ratio corresponding to the first initial feature value is less than a second preset threshold, then it is determined that each initial feature value corresponding to the feature type satisfies the preset exclusion condition; or...

[0048] If the second anomaly ratio corresponding to the first initial feature value is greater than the third preset threshold, and the ratio of the number of wafers corresponding to the first initial feature value to the total number of wafers corresponding to the feature type is greater than the fourth preset threshold, then it is determined that each initial feature value in the feature type satisfies the preset exclusion condition; or,

[0049] If the proportion of abnormal wafers corresponding to the first initial feature value is less than the fifth preset threshold, then it is determined that each initial feature value corresponding to the feature type satisfies the preset exclusion condition.

[0050] In one possible implementation, the second determining module is specifically used for:

[0051] For each candidate feature value, the suspicion level data of the candidate feature value is weighted and summed to obtain the target suspicion level data corresponding to the candidate feature value;

[0052] The target feature value is determined from the candidate feature values ​​based on the target suspicion level data.

[0053] Thirdly, embodiments of this application provide a semiconductor yield root cause determination device, including: a processor and a memory;

[0054] The memory stores computer-executed instructions;

[0055] The processor executes computer execution instructions stored in the memory to implement the semiconductor yield root cause determination method as described in any of the first aspects.

[0056] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed, are used to implement the semiconductor yield root cause determination method according to any one of the first aspects.

[0057] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed, implements the semiconductor yield root cause determination method as described in any of the first aspects.

[0058] The semiconductor yield root cause determination method, apparatus, device, and storage medium provided in this application acquire raw data corresponding to a target manufacturing process and preprocess the raw data to obtain target data. For each initial feature value in the target data, the suspicion level data corresponding to the initial feature value is determined according to a preset index calculation method. Initial feature values ​​whose suspicion level data meets preset exclusion conditions are deleted from the target data to obtain candidate feature values. Based on the suspicion level data of each candidate feature value, the target feature value corresponding to the target manufacturing process is determined from the candidate feature values. In this application, the electronic device acquires raw data in the target manufacturing process and preprocesses it to obtain target data. Then, it determines the suspicion level data corresponding to each initial feature value according to a preset index calculation method. Afterward, it first excludes initial feature values ​​that do not conform to business logic based on preset exclusion conditions to obtain candidate feature values. Then, based on the suspicion level data of each candidate feature value, it determines the target feature value corresponding to the target manufacturing process. This target feature value is the yield root cause corresponding to the target manufacturing process. In this way, by calculating the degree of suspicion data based on a preset index calculation method and excluding feature values ​​according to preset exclusion conditions, this application can make the final determined target feature value, the root cause of yield, more in line with the business logic of actual semiconductor scenarios, and improve the accuracy of yield root cause determination. Attached Figure Description

[0059] Figure 1 A schematic flowchart of a method for determining the root causes of semiconductor yield provided in an embodiment of this application;

[0060] Figure 2 A flowchart illustrating another method for determining the root causes of semiconductor yield provided in this application embodiment;

[0061] Figure 3 A logic diagram illustrating the determination of semiconductor yield root causes provided in an embodiment of this application;

[0062] Figure 4 This is a schematic diagram of a semiconductor yield root cause determination device provided in an embodiment of this application;

[0063] Figure 5 This is a schematic diagram of a semiconductor yield root cause determination device provided in an embodiment of this application. Detailed Implementation

[0064] To enable those skilled in the art to better understand the technical solutions of this application, the application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments and drawings described herein are merely for explaining this application and are not intended to limit this application. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with relevant laws, regulations and standards, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0065] When yield issues arise during semiconductor manufacturing, the sheer volume of data involved makes it difficult for manual methods to quickly pinpoint the problematic machine or chamber. To determine the root cause of yield problems, related technologies typically employ statistical methods, such as analysis of variance, to analyze the vast amounts of data generated during semiconductor manufacturing. These statistical indicators are used to identify the root cause. However, this method of determining the root cause of yield problems is only statistically significant and often produces results that do not align with business logic in real-world scenarios, leading to low accuracy in determining the root cause of yield issues.

[0066] To address the aforementioned issues, this application provides a method, apparatus, device, and storage medium for determining the root cause of semiconductor yield problems. The electronic device acquires raw data from the target manufacturing process and preprocesses it to obtain target data. Then, based on a preset index calculation method, it determines the degree of suspicion data corresponding to each initial feature value. Next, it first eliminates initial feature values ​​that do not conform to business logic based on preset exclusion conditions, obtaining candidate feature values. Finally, based on the degree of suspicion data of each candidate feature value, it determines the target feature value corresponding to the target manufacturing process. This target feature value is the root cause of yield problems in the target manufacturing process. Thus, by calculating the degree of suspicion data based on a preset index calculation method and eliminating features according to preset exclusion conditions, this application ensures that the finally determined target feature value, the root cause of yield problems, better aligns with the business logic of the actual scenario, thereby improving the accuracy of yield root cause determination.

[0067] The following detailed description of the solution presented in this application is provided through specific embodiments. It should be noted that the following embodiments may exist independently or in combination with each other; identical or similar content will not be repeated in different embodiments.

[0068] Figure 1 This is a flowchart illustrating a method for determining the root causes of semiconductor yield, provided in an embodiment of this application.

[0069] Please see Figure 1The semiconductor yield root cause determination method may include:

[0070] S101. Obtain the raw data corresponding to the target manufacturing process, and preprocess the raw data to obtain the target data.

[0071] The execution subject of this application embodiment can be an electronic device or a semiconductor yield root cause determination device installed in an electronic device. The semiconductor yield root cause determination device can be implemented by software or by a combination of software and hardware. For ease of understanding, the following description uses an electronic device as the execution subject. The electronic device can specifically refer to a mobile phone, computer, or wearable device, etc. The specific type of electronic device is not limited in this application embodiment.

[0072] In this embodiment, the target manufacturing process can refer to a semiconductor manufacturing process, specifically the processing of wafers. The raw data can refer to the initial data of the target manufacturing process acquired by the electronic device. This raw data can be discrete data, specifically including yield index data and manufacturing process data. The yield index data can refer to various index data corresponding to the target manufacturing process, such as yield data reflecting wafer quality and the number of defects in the target manufacturing process. The manufacturing process data can refer to the process data corresponding to each wafer processing step in the target manufacturing process. The specific data type of this manufacturing process data can include machine identifiers, chamber identifiers, process recipe identifiers, and material identifiers passed through a certain station (Step). It should be noted that the raw data can also include other types of data, which can be set based on actual needs; this embodiment does not limit this.

[0073] The target data can refer to the data obtained after preprocessing the original data. The target data can be in a target format such as a table, where each row in the table corresponds to a wafer in the target manufacturing process, and each column corresponds to a feature type (or data type). Of course, the target data can also be in other formats, and this application embodiment does not limit it.

[0074] In this step, when yield root cause analysis is required, the electronic equipment can first acquire raw data, specifically from the database of the Manufacturing Execution System (MES). MES is a shop floor-level production information management system that improves the efficiency of the entire production process, coordinates various production stages, monitors production activities in real time, and provides detailed production data. To facilitate subsequent data analysis and calculation, the electronic equipment can preprocess the raw data after acquisition. This preprocessing may include format conversion, data sorting, and adding feature types (i.e., feature columns) such as target status identifiers for each wafer, ultimately obtaining the target data corresponding to the target manufacturing process.

[0075] S102. For each initial feature value in the target data, determine the degree of suspicion data corresponding to the initial feature value according to the preset index calculation method.

[0076] In this embodiment, the initial feature value can refer to the specific numerical value of each feature in the target data. The preset index calculation method can refer to a pre-set calculation method or rule for the degree of suspicion index. The degree of suspicion data can refer to the specific numerical value of the degree of suspicion index.

[0077] In this step, after preprocessing the raw data to obtain the target data, the electronic device can determine the suspicion level data corresponding to each suspicion level indicator type according to a preset indicator calculation method based on business logic settings. For example, this suspicion level data may specifically include a significant difference parameter, a first anomaly ratio, a second anomaly ratio, and a common ratio of abnormal wafers. The significant difference parameter can be used to characterize the significant difference level of different initial feature values; the first anomaly ratio can be used to characterize the wafer defect rate corresponding to each initial feature value; the second anomaly ratio can be used to characterize the proportion of abnormal wafers corresponding to each initial feature value to the total number of abnormal wafers corresponding to that feature; the common ratio of abnormal wafers can be the proportion of the sum of the number of abnormal wafers that have undergone the initial feature value and the number of normal wafers that have not undergone the initial feature value to the total number of wafers processed by that feature. Based on the different suspicion level data, the electronic device can calculate the corresponding suspicion level data based on the preset indicator calculation method.

[0078] It should be noted that the preset index calculation method may also include other calculation methods based on the business logic design of the semiconductor target manufacturing process, which can calculate other types of doubt level data. This application embodiment does not limit this.

[0079] S103. Delete the initial feature values ​​of the suspected data that meet the preset exclusion conditions in the target data to obtain alternative feature values.

[0080] In this embodiment, the preset exclusion condition can refer to the screening and exclusion conditions determined based on the business logic of the semiconductor target manufacturing process. Specifically, it can refer to the threshold conditions corresponding to different types of doubt data. This embodiment does not limit the specific type of the preset exclusion condition. The alternative feature value can refer to the feature value remaining after deleting the initial feature value of the doubt data that meets the preset exclusion condition from the target data.

[0081] In this step, after calculating the suspicion level data corresponding to each initial feature value, the electronic device can filter the initial feature values ​​according to preset exclusion conditions. When the suspicion level data meets the preset exclusion conditions, the electronic device can determine that the initial feature value corresponding to that suspicion level data has a low suspicion level in the business logic of the semiconductor target manufacturing process and is unlikely to be a root cause of yield problems. The electronic device can then delete the initial feature value from the target data, ultimately obtaining candidate feature values. In this way, the electronic device can filter and select initial feature values ​​based on preset exclusion conditions that conform to the business logic of the actual scenario, ensuring the rationality and accuracy of subsequent determination of semiconductor yield root causes, while also reducing the amount of computation and saving computing resources.

[0082] S104. Based on the doubt level data of each candidate feature value, determine the target feature value corresponding to the target manufacturing process from the candidate feature values.

[0083] In this embodiment, the target feature value can refer to the root cause of yield corresponding to the target semiconductor manufacturing process. After filtering the initial feature values ​​based on preset exclusion conditions to obtain candidate feature values, the electronic device can determine the target feature value corresponding to the target manufacturing process from the candidate feature values ​​based on the suspicion level data of each candidate feature value. For example, the electronic device can perform weighted summation of different types of suspicion level data corresponding to the candidate feature values ​​to obtain the target suspicion level data of the candidate feature value, and then determine the target feature value based on the numerical relationship of the target suspicion level data. Of course, the electronic device can also use other methods to determine the target feature value, such as comparing and determining it based on the average value of the suspicion level data, which is not limited in this embodiment.

[0084] The semiconductor yield root cause determination method provided in this application involves an electronic device acquiring raw data from the target manufacturing process and preprocessing it to obtain target data. Then, based on a preset index calculation method, the suspicion level data corresponding to each initial feature value is determined. Next, initial feature values ​​that do not conform to business logic are first excluded based on preset exclusion conditions to obtain candidate feature values. Finally, based on the suspicion level data of each candidate feature value, the target feature value corresponding to the target manufacturing process is determined. This target feature value is the yield root cause in the target manufacturing process. Thus, by calculating the suspicion level data based on a preset index calculation method and excluding features according to preset exclusion conditions, this application ensures that the final determined target feature value, the target root cause, better aligns with the business logic of actual semiconductor scenarios, thereby improving the accuracy of yield root cause determination.

[0085] Based on the above embodiments, Figure 2 This is a flowchart illustrating another method for determining the root causes of semiconductor yield provided in an embodiment of this application. Please refer to... Figure 2 The semiconductor yield root cause determination method may include:

[0086] S201. Obtain the raw data corresponding to the target manufacturing process.

[0087] In one possible implementation, the raw data includes yield index data and manufacturing process data; the yield index data includes at least one of wafer test data corresponding to the target manufacturing process and defect quantity data in the target manufacturing process; the data type of the manufacturing process data includes at least one of site identifier, machine identifier, chamber identifier, process recipe identifier and material identifier.

[0088] In this embodiment, the electronic device can obtain raw data of the target manufacturing process from the manufacturing execution system database. This raw data may include yield index data and manufacturing process data. Yield index data can refer to test data reflecting whether the wafer is normal or abnormal (good or bad), specifically including yield data after the target manufacturing process is completed and the number of defects in the target manufacturing process. Manufacturing process data can refer to discrete feature data automatically collected from each processing step of the wafer during the target manufacturing process. The data type of manufacturing process data can include site identifiers (e.g., site name), machine identifiers (e.g., machine name), chamber identifiers (e.g., chamber name), process recipe identifiers (e.g., process recipe name), and material identifiers (e.g., material name). Of course, the raw data may also include other types of data, which are not limited in this embodiment.

[0089] In the embodiments of this application, the raw data acquired by the electronic device may include various types of initial data from the semiconductor manufacturing process, which can ensure the comprehensiveness of the data and thus improve the accuracy of subsequent yield root cause determination.

[0090] S202. Convert the format of the original data to obtain alternative data.

[0091] In this embodiment, the alternative data can refer to the target format data obtained after converting the original data. Specifically, the original data acquired by the electronic device is quite complex and requires preprocessing. The electronic device can first convert the original data into alternative data in the target format to facilitate subsequent analysis and calculation. Of course, preprocessing processes such as data sorting can also be performed during the format conversion process, which is not limited in this embodiment. For example, Table 1 is a schematic table of alternative data in a target format provided by an exemplary embodiment of this application.

[0092] Table 1

[0093]

[0094]

[0095] As shown in Table 1, each row in the candidate data represents the data for one wafer, each column represents a feature type that needs to be traced, and each cell represents the specific value of that wafer under that feature type. For example, the feature type value for wafer 1 at process station 1 is "machine 1". The yield index data in the candidate data is the indicator data reflecting the quality of the wafer.

[0096] S203. Based on the yield index data, determine the target status identifier corresponding to each wafer in the target manufacturing process; add the target status identifier to the candidate data to obtain the target data.

[0097] In this embodiment, the target status identifier can refer to the status data or status label corresponding to each wafer in the target manufacturing process, specifically, it can be normal (Good) or abnormal (Bad), which can be represented by the values ​​0 and 1 respectively. Specifically, since yield index data is a numerical indicator reflecting the quality of wafers, the electronic device can further determine the status label reflecting the quality of the wafers based on the yield index data, thus obtaining the target status identifier corresponding to each wafer. The specific method for determining this target status identifier can be set by the user based on actual needs, or it can be determined and marked based on a preset threshold; this embodiment does not limit this.

[0098] For example, an electronic device can determine the target state identifier of a wafer according to the following formula (1):

[0099]

[0100] In formula (1), L represents the target status identifier corresponding to the wafer. When L is 0, it indicates that the wafer is normal, and when L is 1, it indicates that the wafer is abnormal. The values ​​Q1 and Q3 are the 1 / 4 and 3 / 4 quantiles of the yield index data in the original data, respectively. Among them, the 1 / 4 quantile Q1, also known as the lower quartile, refers to the data value located at the 25th percentile after sorting the yield index data from smallest to largest. That is, 25% of the yield index data is less than Q1, and 75% of the data is greater than Q1. The 3 / 4 quantile Q3, also known as the upper quartile, refers to the data value located at the 75th percentile after sorting the yield index data from smallest to largest. That is, 75% of the yield index data is less than Q3, and 25% of the data is greater than Q3.

[0101] In formula (1), IQR is the difference between the 3 / 4 quantile and the 1 / 4 quantile of the yield index data, i.e., Q3-Q1; k is a preset constant, which can generally be taken as 1.5. In formula (1), when the yield index data corresponding to the wafer is within the range of (Q1-1.5*IQR, Q3+1.5*IQR), the target status of the wafer is marked as 0, and the wafer is in a normal state; when the yield index data corresponding to the wafer is outside the above range, the target status of the wafer is marked as 1, and the wafer is abnormal.

[0102] S204. For each initial feature value in the target data, determine the degree of suspicion data corresponding to the initial feature value according to the preset index calculation method.

[0103] In one possible implementation, the suspicion level data may include types such as significant difference parameters, first anomaly ratio, second anomaly ratio, and common ratio of anomaly wafers. The suspicion level data in step 204 can be achieved through the following steps (1) to (4):

[0104] (1) Based on the preset variance analysis method, analyze and process each feature type in the target data to obtain the significant difference parameters corresponding to each initial feature value in the feature type.

[0105] In this embodiment, the preset analysis of variance method can refer to the analysis of variance (ANOVA) method, etc. The significance difference parameter can be the significance level parameter (P value) corresponding to each initial feature value in the target data.

[0106] Specifically, for any feature type and corresponding index column (such as yield index data column or target status identifier column) in the target data of electronic devices, ANOVA analysis is performed to obtain the P value corresponding to each initial feature value in that feature type. This is the significant difference parameter corresponding to each initial feature value. The smaller the significant difference parameter, the greater the difference between the initial feature value and other initial feature values, the higher the degree of suspicion, and the higher the probability that the initial feature value is the root cause of the yield of the semiconductor target manufacturing process.

[0107] (2) Determine the ratio of the number of abnormal wafers corresponding to the initial feature value to the number of wafers corresponding to the initial feature value, and obtain the first abnormality ratio corresponding to the initial feature value.

[0108] In this embodiment, the first anomalous ratio (Bad Ratio, BR) can refer to the proportion of the number of anomalous wafers corresponding to the initial feature value to the total number of wafers corresponding to that initial feature value. In this step, for each feature type, the electronic device can first calculate the number of wafers processed for each initial feature value in that feature type (ProcessWafer Count, PWC), and then calculate the number of anomalous wafers corresponding to each initial feature value (Bad Wafer Count, BWC). At this time, the first anomalous ratio corresponding to the initial feature value can be calculated by the following formula (2):

[0109]

[0110] In the above formula (2), the first anomaly ratio BR is equal to the ratio of the number of abnormal wafers corresponding to the initial feature value to the number of wafers corresponding to the initial feature value. The larger the value of the first anomaly ratio, the higher the anomaly ratio of the wafers processed using the initial feature value, the higher the suspicion level of the initial feature value, and the higher the possibility that the initial feature value is the root cause of the yield of the semiconductor target manufacturing process.

[0111] (3) Determine the ratio of the number of abnormal wafers corresponding to the initial feature value to the total number of abnormal wafers corresponding to the feature type to which the initial feature value belongs, and obtain the second abnormality ratio corresponding to the initial feature value.

[0112] In this embodiment, the second bad proportion (BP) can refer to the proportion of the number of bad wafers corresponding to the initial feature value to the total number of bad wafers corresponding to the feature type to which the initial feature value belongs. Specifically, for each feature type, the electronic device can first calculate the number of bad wafers corresponding to each initial feature value, and then calculate the total number of bad wafers corresponding to all initial feature values ​​in that feature type. At this time, the second bad proportion corresponding to the initial feature value can be calculated by the following formula (3):

[0113]

[0114] In the above formula (3), j represents a certain initial characteristic value, and J represents the set of all initial characteristic values ​​under that characteristic. The larger the second anomaly ratio of the initial characteristic value, the higher the proportion of abnormal wafers of that initial characteristic value, the greater the suspicion of that initial characteristic value, and the higher the possibility that the initial characteristic value is the root cause of the yield of the semiconductor target manufacturing process.

[0115] (4) Determine the ratio of the target sum corresponding to the initial feature value to the total number of wafers corresponding to the feature type to which the initial feature value belongs, and obtain the common proportion of abnormal wafers corresponding to the initial feature value; the target sum is the sum of the number of abnormal wafers corresponding to the initial feature value and the number of normal wafers corresponding to other initial feature values ​​in the feature type to which the initial feature value belongs.

[0116] In this embodiment of the application, the Common Score (CS) of abnormal wafers can refer to the target sum of the number of abnormal wafers that have passed a certain initial characteristic value and the number of normal wafers that have not passed the initial characteristic value, in the total number of wafers corresponding to the characteristic type to which the initial characteristic value belongs.

[0117] Specifically, for each feature type, the electronic device can first calculate the number of abnormal wafers (BWC) with a certain initial feature value in that feature type, and then calculate the number of wafers processed with that initial feature value. The difference between the number of wafers and the number of abnormal wafers is the number of normal wafers processed with that initial feature value. After that, the electronic device can calculate the total number of wafers (Total Wafer Count, TWC) corresponding to that feature type, and at the same time, it can calculate the total number of normal wafers (Total Good Wafer Count, TGWC) corresponding to that feature type. Then, the electronic device can calculate the target sum of the number of abnormal wafers with a certain initial feature value and the number of normal wafers that have not undergone that initial feature value, and calculate the proportion of the target sum to the total number of wafers corresponding to the feature type to which the initial feature value belongs, so as to obtain the common proportion of abnormal wafers. For example, the common proportion of abnormal wafers can be calculated by the following formula (4):

[0118]

[0119] In formula (4) above, (PWC-BWC) represents the number of normal wafers processed for a certain initial characteristic value, and (TGWC-(PWC-BWC)) represents the number of normal wafers that did not undergo the initial characteristic value under that characteristic type. Since the total number of wafers processed for a characteristic type is fixed, the more abnormal wafers that have undergone a certain initial characteristic value and the more normal wafers that have not undergone the initial characteristic value, the worse the performance of that initial characteristic value is, and the higher the possibility that it may be the root cause of the semiconductor target manufacturing process yield. That is, the larger the common proportion of abnormal wafers of the initial characteristic value, the greater the suspicion of that initial characteristic value.

[0120] For example, Table 2 is a schematic table of suspicion level data according to an embodiment of this application, as follows:

[0121] Table 2

[0122]

[0123] As shown in Table 2, the feature type (feature column) is the chamber of site X, and the initial feature values ​​included in this feature type are chamber 1, chamber 2, and chamber 3. For example, for the initial feature value of chamber 1, the electronic device can determine the number of abnormal wafers corresponding to this initial feature value and the number of wafers processed using this initial feature value. Subsequently, it can calculate the suspicion level data, such as the first abnormality ratio, the second abnormality ratio, the common ratio of abnormal wafers, and significant difference parameters (not shown in Table 2), corresponding to this initial feature value. In this way, the electronic device determines the suspicion level data for each initial feature value according to a preset index calculation method. It can calculate the suspicion level data based on the business logic of the semiconductor target manufacturing process, which can improve the rationality of the suspicion level data calculation and thus improve the accuracy of subsequent yield root cause determination.

[0124] It should be noted that the calculation methods of each suspicion level data in steps (1) to (4) above are AND / OR relationships. The electronic device can calculate at least one of them. Of course, it can also set other preset indicator calculation methods based on business logic to calculate other types of suspicion level data. This application embodiment does not limit this.

[0125] S205. Delete the initial feature values ​​of the suspected data that meet the preset exclusion conditions in the target data to obtain alternative feature values.

[0126] In one possible implementation, whether the suspicion level data meets the preset exclusion criteria can be determined based on the following steps (5) to (8):

[0127] (5) If the significant difference parameter of the initial feature value is greater than the first preset threshold, then the initial feature value is determined to meet the preset exclusion condition.

[0128] In this embodiment, the first preset threshold may refer to a pre-set threshold for a significant difference parameter. For example, the first preset threshold may be 0.05. Specifically, if the significant difference parameter of the initial feature value is greater than the first preset threshold, as determined by a preset variance analysis method, then the initial feature value meets the preset exclusion condition. In this case, there is no significant difference between the initial feature value and other initial feature values, the suspicion level of the initial feature value is low, and the initial feature value can be excluded from the target data.

[0129] (6) For the feature type to which the initial feature value belongs, determine the first initial feature value with the largest number of abnormal wafers in the feature type. If the first abnormality ratio corresponding to the first initial feature value is less than the second preset threshold, then determine that each initial feature value corresponding to the feature type meets the preset exclusion condition.

[0130] In this embodiment, the first initial feature value can refer to the initial feature value with the largest number of abnormal wafers in a feature type. The second preset threshold can refer to a pre-set critical threshold for the first abnormality ratio, for example, the second preset threshold can be 0.4, etc.

[0131] Specifically, for each feature type, the electronic device can determine the initial feature value with the largest number of abnormal wafers in that feature type as the first initial feature value. If the first abnormality ratio corresponding to the first initial feature value is less than the second preset threshold, the electronic device can determine that all initial feature values ​​under that feature type meet the preset exclusion conditions. That is, when the proportion of abnormal wafers in the initial feature value with the largest number of abnormal wafers in a feature type is lower than the second preset threshold, the probability of abnormal wafers appearing in that initial feature value is low. By analogy, the electronic device can determine that each initial feature value in that feature type meets the preset exclusion conditions, and the electronic device can exclude each initial feature value in that feature type from the target data.

[0132] (7) If the second abnormality ratio corresponding to the first initial feature value is greater than the third preset threshold, and the ratio of the number of wafers corresponding to the first initial feature value to the total number of wafers corresponding to the feature type is greater than the fourth preset threshold, then it is determined that each initial feature value in the feature type meets the preset exclusion conditions.

[0133] In this embodiment, the third preset threshold may refer to a pre-set critical threshold for the second abnormal ratio, such as 0.8. The fourth preset threshold may refer to a pre-set ratio threshold, such as 0.7.

[0134] Specifically, for each feature type, the electronic device can determine the initial feature value with the largest number of abnormal wafers in that feature type as the first initial feature value. If the second abnormality ratio of the first initial feature value is greater than the third preset threshold, and the ratio of the number of wafers of the first initial feature value to the total number of wafers corresponding to the feature type (PWC / TWC) is greater than the fourth preset threshold, then the electronic device can determine that all initial feature values ​​of that feature type meet the preset exclusion conditions, and can delete the initial feature value corresponding to that feature type from the target data.

[0135] That is, when the proportion of abnormal wafers with the first initial feature value that has the most abnormal wafers is high, but the proportion of wafers using the first initial feature value is also high, it means that abnormal wafers often appear with the first initial feature value because most wafers use the first initial feature value, rather than because the first initial feature value itself is performing poorly. The first initial feature value meets the preset exclusion condition. Similarly, each initial feature value in the feature type meets the preset exclusion condition, and the electronic device can delete each initial feature value corresponding to the feature type from the target data.

[0136] (8) If the proportion of abnormal wafers corresponding to the first initial feature value is less than the fifth preset threshold, then each initial feature value corresponding to the feature type is determined to meet the preset exclusion conditions.

[0137] In this embodiment, the fifth preset threshold can refer to a pre-set threshold for the common proportion of abnormal wafers, specifically 0.4, etc. Specifically, for each feature type, the electronic device can determine the initial feature value with the highest number of abnormal wafers in that feature type as the first initial feature value. If the common proportion of abnormal wafers of the first initial feature value is less than the fifth preset threshold, the electronic device can determine that each initial feature value in that feature type meets the preset exclusion conditions, and the electronic device can delete each initial feature value under that feature type from the target data. That is, if the target sum of the number of abnormal wafers that have passed the first initial feature value and the number of normal wafers that have not passed the first initial feature value accounts for a low proportion of the total number of wafers processed in the feature type, it indicates that the performance of the first initial feature value is good, and the first initial feature value meets the preset exclusion conditions. Similarly, each initial feature value in that feature type meets the preset exclusion conditions.

[0138] It should be noted that the above steps (5) to (8) are in an OR relationship. As long as one of them is satisfied, the electronic device can determine that the initial feature value meets the preset exclusion conditions. Of course, the electronic device can also use other preset exclusion conditions based on business logic and domain knowledge for exclusion, and this application embodiment does not limit this.

[0139] In this embodiment, after the electronic device obtains the suspicion level data of each initial feature value in the target data, it can exclude the initial feature values ​​that do not conform to the business logic according to the preset exclusion conditions based on business logic and domain knowledge. This can improve the accuracy of subsequent yield root cause determination, reduce the amount of data calculation, and save computing resources.

[0140] S206. For each candidate feature value, perform a weighted summation of the suspicion level data of the candidate feature values ​​to obtain the target suspicion level data corresponding to the candidate feature value; determine the target feature value from the candidate feature values ​​based on the target suspicion level data.

[0141] In this embodiment of the application, the target suspicion level data can refer to a comprehensive index of the suspicion level data of each candidate feature value, specifically, it can refer to the weighted sum of the suspicion level data, etc.

[0142] Specifically, after filtering the initial feature values ​​according to preset exclusion conditions to obtain candidate feature values, the electronic device can determine the target suspicion level data corresponding to each candidate feature value based on the suspicion level data of that candidate feature value. This can be calculated using a weighted summation method. The weights of each suspicion level data can be user-defined preset weights or weights determined based on preset weight assignment methods such as entropy methods; this embodiment does not limit this. After obtaining the target suspicion level data for each candidate feature value, the electronic device can sort the candidate feature values ​​according to the target suspicion level data from high to low, and then select the top N candidate feature values ​​as the target feature values, where N is a positive integer. For example, Table 3 is a schematic table of target feature values ​​provided in this embodiment. Details are as follows:

[0143] Table 3

[0144]

[0145] As shown in Table 3, the electronic device sorts the candidate feature values ​​according to the target suspicion level data from high to low, and then selects the one with the highest ranking, indicating a higher suspicion level. The electronic device can determine the target feature value based on this ranking; for example, the candidate feature value with the highest target suspicion level data can be used as the target feature value. In this embodiment, the electronic device comprehensively calculates the target suspicion level data of the candidate feature values ​​through weighted summation and other methods, and then determines the target feature value according to the ranking of the target suspicion level data, which can improve the rationality and accuracy of determining the root cause of yield in semiconductor manufacturing processes.

[0146] Based on the above embodiments, Figure 3This is a schematic diagram illustrating the practical application of a semiconductor yield root cause determination method provided in this application embodiment. For example... Figure 3 As shown, the electronic device first acquires raw data, which may include yield index data and manufacturing process data. Then, the electronic device can preprocess the raw data, specifically by converting the raw data into a format to obtain candidate data in the target format. Then, the electronic device can calculate the target status identifier of each wafer in the target manufacturing process based on the yield index data, and add the target status identifier to the candidate data to obtain the target data corresponding to the target manufacturing process.

[0147] Subsequently, the electronic device can calculate the degree of suspicion data corresponding to each initial feature value in the target data according to the preset index calculation method. Specifically, it can include significant difference parameters, first anomaly ratio, second anomaly ratio, and common ratio of abnormal wafers, etc. Then, the electronic device can exclude the initial feature values ​​that meet the preset exclusion conditions from the target data according to the business logic and domain knowledge set based on the actual semiconductor scenario, and obtain candidate feature values. Finally, the yield root cause is sorted according to the target degree of suspicion data of the candidate feature values ​​to obtain the target feature value, which is the yield root cause of the target manufacturing process.

[0148] In this embodiment, the electronic device is configured with a preset index calculation method based on semiconductor business logic. This preset index calculation method is used to analyze the root causes of semiconductor yield, reflecting the degree of suspicion of initial characteristic values ​​such as machine identifiers, chamber identifiers, process formula identifiers, and material identifiers from multiple different dimensions. Finally, through comprehensive calculation and sorting, the root cause determination process is made more accurate and reasonable. Furthermore, this embodiment configures preset exclusion conditions based on business logic and domain knowledge to exclude initial characteristic values ​​that meet the preset exclusion conditions, further improving the accuracy and rationality of root cause determination.

[0149] Figure 4 This is a schematic diagram of a semiconductor yield root cause determination device provided in an embodiment of this application.

[0150] Please see Figure 4 The semiconductor yield root cause determination device 40 may include:

[0151] The acquisition module 41 is used to acquire the raw data corresponding to the target manufacturing process and preprocess the raw data to obtain the target data;

[0152] The first determining module 42 is used to determine the degree of suspicion data corresponding to each initial feature value in the target data according to a preset index calculation method.

[0153] The deletion module 43 is used to delete the initial feature values ​​of the doubt level data that meet the preset exclusion conditions in the target data to obtain alternative feature values;

[0154] The second determining module 44 is used to determine the target feature value corresponding to the target manufacturing process from the candidate feature values ​​based on the suspicion level data of each candidate feature value.

[0155] In one possible implementation, the raw data includes yield index data and manufacturing process data;

[0156] Yield metrics data include at least one of the wafer testing data corresponding to the target manufacturing process and the defect quantity data in the target manufacturing process; the data types of manufacturing process data include at least one of the following: site identifier, machine identifier, chamber identifier, process formula identifier, and material identifier.

[0157] In one possible implementation, the acquisition module 41 is specifically used for:

[0158] The original data is converted to a different format to obtain alternative data.

[0159] Based on yield index data, determine the target status identifier corresponding to each wafer in the target manufacturing process;

[0160] Add the target status identifier to the candidate data to obtain the target data.

[0161] In one possible implementation, the first determining module 42 is specifically used for:

[0162] Based on a pre-defined analysis of variance method, the various feature types in the target data are analyzed and processed to obtain the significant difference parameters corresponding to each initial feature value in each feature type; and / or,

[0163] Determine the ratio of the number of abnormal wafers corresponding to the initial feature value to the total number of wafers corresponding to the initial feature value, thus obtaining the first abnormality ratio corresponding to the initial feature value; and / or,

[0164] Determine the ratio of the number of abnormal wafers corresponding to the initial feature value to the total number of abnormal wafers corresponding to the feature type to which the initial feature value belongs, to obtain the second abnormality ratio corresponding to the initial feature value; and / or,

[0165] The ratio of the target sum corresponding to the initial feature value to the total number of wafers corresponding to the feature type to which the initial feature value belongs is determined to obtain the common proportion of abnormal wafers corresponding to the initial feature value; the target sum is the sum of the number of abnormal wafers corresponding to the initial feature value and the number of normal wafers corresponding to other initial feature values ​​in the feature type to which the initial feature value belongs.

[0166] In one possible implementation, the device 40 is further used for:

[0167] If the significant difference parameter of the initial feature value is greater than the first preset threshold, then the initial feature value is determined to meet the preset exclusion condition; or,

[0168] For each initial feature value, determine the first initial feature value that has the largest number of abnormal wafers within that feature type. If the first abnormality ratio corresponding to the first initial feature value is less than a second preset threshold, then determine that all initial feature values ​​corresponding to the feature type meet the preset exclusion conditions; or...

[0169] If the second anomaly ratio corresponding to the first initial feature value is greater than the third preset threshold, and the ratio of the number of wafers corresponding to the first initial feature value to the total number of wafers corresponding to the feature type is greater than the fourth preset threshold, then it is determined that each initial feature value in the feature type meets the preset exclusion conditions; or,

[0170] If the proportion of abnormal wafers corresponding to the first initial feature value is less than the fifth preset threshold, then each initial feature value corresponding to the feature type is determined to meet the preset exclusion conditions.

[0171] In one possible implementation, the second determining module 44 is specifically used for:

[0172] For each candidate feature value, the doubt level data of the candidate feature value is weighted and summed to obtain the target doubt level data corresponding to the candidate feature value;

[0173] Based on the target suspicion level data, the target feature value is determined from the candidate feature values.

[0174] The semiconductor yield root cause determination device 40 provided in this application embodiment can execute the technical solution shown in the above method embodiment. Its implementation principle and beneficial effects are similar, and will not be repeated here.

[0175] Figure 5 This is a schematic diagram of a semiconductor yield root cause determination device provided in an embodiment of this application. Please refer to... Figure 5 The semiconductor yield root cause determination device 50 may include a memory 51 and a processor 52. Exemplarily, the memory 51 and the processor 52 are interconnected via a bus 53.

[0176] Memory 51 is used to store program instructions;

[0177] The processor 52 is used to execute the program instructions stored in the memory to implement the semiconductor yield root cause determination method shown in the above embodiment.

[0178] Figure 5The semiconductor yield root cause determination device 50 shown can execute the technical solution shown in the above method embodiment. Its implementation principle and beneficial effects are similar, and will not be described again here.

[0179] This application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the above-described semiconductor yield root cause determination method.

[0180] This application embodiment may also provide a computer program product, including a computer program that, when executed by a processor, can implement the above-described semiconductor yield root cause determination method.

[0181] This application provides a chip that stores a computer program. When the computer program is executed by the chip, the above-mentioned semiconductor yield root cause determination method is implemented.

[0182] This application provides a chip module that stores a computer program. When the computer program is executed by the chip module, the above-mentioned semiconductor yield root cause determination method is implemented.

[0183] It should be noted that the processor mentioned in the embodiments of this application may be a central processing unit (CPU), etc., and the memory mentioned in the embodiments of this application may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The embodiments of this application do not limit the specific types of processors and memory.

[0184] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0185] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processing unit of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processing unit of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0186] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0187] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0188] Regarding the modules / units included in the various devices and products described in the above embodiments, they can be software modules / units, hardware modules / units, or a combination of both. Each device and product can be applied to or integrated into a chip, chip module, or terminal device. For example, for devices and products applied to or integrated into a chip, each included module / chip can be implemented entirely using hardware methods such as circuits, or at least some modules / units can be implemented using software programs running on a processor integrated within the chip, while the remaining modules / units can be implemented using hardware methods such as circuits.

[0189] In this application, the term "comprising" and its variations can refer to non-limiting inclusion; the term "or" and its variations can refer to "and / or". The terms "first", "second", etc., in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. In this application, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0190] The above are only some embodiments of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for determining the root causes of semiconductor yield, characterized in that, include: Obtain the raw data corresponding to the target manufacturing process, and preprocess the raw data to obtain the target data; For each initial feature value in the target data, the degree of suspicion corresponding to the initial feature value is determined according to a preset index calculation method; The initial feature values ​​of the suspicion level data that meet the preset exclusion conditions are deleted from the target data to obtain alternative feature values; Based on the degree of suspicion data for each candidate feature value, the target feature value corresponding to the target manufacturing process is determined from the candidate feature values.

2. The method according to claim 1, characterized in that, The raw data includes yield index data and manufacturing process data; The yield index data includes at least one of the wafer test data corresponding to the target manufacturing process and the defect quantity data in the target manufacturing process; The data types of the manufacturing process data include at least one of the following: site identifier, machine identifier, chamber identifier, process formula identifier, and material identifier.

3. The method according to claim 2, characterized in that, The preprocessing of the original data to obtain the target data includes: The original data is converted to obtain alternative data; Based on the yield index data, determine the target status identifier corresponding to each wafer in the target manufacturing process; The target status identifier is added to the alternative data to obtain the target data.

4. The method according to claim 1, characterized in that, The step of determining the degree of suspicion data corresponding to the initial feature value according to the preset index calculation method includes: Based on a preset analysis of variance method, the various feature types in the target data are analyzed and processed to obtain significant difference parameters corresponding to each initial feature value in the feature type; and / or, Determine the ratio of the number of abnormal wafers corresponding to the initial feature value to the total number of wafers corresponding to the initial feature value to obtain the first abnormality ratio corresponding to the initial feature value; and / or, The ratio of the number of abnormal wafers corresponding to the initial feature value to the total number of abnormal wafers corresponding to the feature type to which the initial feature value belongs is determined to obtain the second abnormality ratio corresponding to the initial feature value; and / or, The ratio of the target sum corresponding to the initial feature value to the total number of wafers corresponding to the feature type to which the initial feature value belongs is determined to obtain the common proportion of abnormal wafers corresponding to the initial feature value; the target sum is the sum of the number of abnormal wafers corresponding to the initial feature value and the number of normal wafers corresponding to other initial feature values ​​in the feature type to which the initial feature value belongs, excluding the initial feature value.

5. The method according to claim 4, characterized in that, The method further includes: If the significant difference parameter of the initial feature value is greater than a first preset threshold, then the initial feature value is determined to meet the preset exclusion condition; or, For each feature type to which the initial feature value belongs, a first initial feature value with the largest number of abnormal wafers within that feature type is determined. If the first abnormality ratio corresponding to the first initial feature value is less than a second preset threshold, then it is determined that each initial feature value corresponding to the feature type satisfies the preset exclusion condition; or... If the second anomaly ratio corresponding to the first initial feature value is greater than the third preset threshold, and the ratio of the number of wafers corresponding to the first initial feature value to the total number of wafers corresponding to the feature type is greater than the fourth preset threshold, then it is determined that each initial feature value in the feature type satisfies the preset exclusion condition; or, If the proportion of abnormal wafers corresponding to the first initial feature value is less than the fifth preset threshold, then it is determined that each initial feature value corresponding to the feature type satisfies the preset exclusion condition.

6. The method according to any one of claims 1 to 5, characterized in that, The step of determining the target feature value corresponding to the target manufacturing process from the candidate feature values ​​based on the suspicion level data of each candidate feature value includes: For each candidate feature value, the suspicion level data of the candidate feature value is weighted and summed to obtain the target suspicion level data corresponding to the candidate feature value; The target feature value is determined from the candidate feature values ​​based on the target suspicion level data.

7. A semiconductor yield root cause determination device, characterized in that, include: The acquisition module is used to acquire the raw data corresponding to the target manufacturing process and preprocess the raw data to obtain the target data; The first determining module is used to determine the degree of suspicion data corresponding to each initial feature value in the target data according to a preset index calculation method. The deletion module is used to delete the initial feature values ​​of the suspicion level data that meet the preset exclusion conditions in the target data, and obtain alternative feature values; The second determining module is used to determine the target feature value corresponding to the target manufacturing process from the candidate feature values ​​based on the suspicion level data of each candidate feature value.

8. A semiconductor yield root cause determination device, characterized in that, include: Processor, memory; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the semiconductor yield root cause determination method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed, are used to implement the semiconductor yield root cause determination method according to any one of claims 1 to 6.

10. A computer program product, characterized in that, Includes a computer program that, when executed, implements the semiconductor yield root cause determination method according to any one of claims 1 to 6.

Images