Electrical parameter screening method

Abstract

The invention provides an electrical parameter screening method, which comprises the following steps of: counting each electrical parameter of each wafer on a wafer, forming a parameter group by using the same electrical parameter, and identifying the distribution form of each parameter group; if the parameter group is in normal distribution, adopting a standard deviation method to screen abnormal data so as to identify an abnormal wafer; if the parameter group is in non-normal distribution, adopting a median absolute deviation method to primarily screen wafers with abnormal single electrical parameters; based on the primarily screened wafers, adopting a multivariate anomaly detection method to carry out secondary screening so as to identify wafers which are abnormal on a plurality of electrical parameters at the same time; counting each electrical parameter of each screened wafer, forming a parameter group by using the same electrical parameter, identifying the distribution form of each parameter group, and converting the data of the parameter groups into normal distribution by using a conversion model if the distribution form is abnormal distribution; and performing statistical index characterization on each electrical parameter based on the converted data. According to the method, efficient screening and reliable evaluation of the electrical parameters of the wafer can be realized.

Description

Technical Field

[0001] The present invention relates to the field of electrical parameter screening, and in particular to a method for screening electrical parameters. Background Art

[0002] In the field of wafer manufacturing, accurate testing and screening of electrical parameters are key to ensuring chip quality and performance. Disadvantages of the current electrical parameter screening methods:

[0003] 1. There is no complete set of screening methods for the reliability of electrical parameters

[0004] The data needs to obey the normal distribution, and the shape is a bell-shaped distribution that is symmetrical on both sides. However, in fact, electrical parameters do not necessarily obey the normal distribution. For example, IDSS (saturated drain-source current) has a numerical value characteristic of being greater than or equal to 0, and there is no negative value, that is, the data distribution of this parameter cannot be a normal distribution. If this method is used for card control, there will be inaccurate control positions, resulting in product risks, that is, products that are actually qualified products are mistakenly judged as unqualified products, or unqualified products that are actually qualified products are missed as qualified products.

[0005] In addition, the existing method for screening electrical parameters will ultimately be characterized by process capability index and defective rate, but the data distribution needs to be converted into a normal distribution before the index calculation can be performed, otherwise there will be deviations.

[0006] 2. The method of identifying abnormal wafers does not consider the correlation of electrical parameters

[0007] Current methods for screening and identifying electrical parameter anomalies mostly consider the performance of electrical parameters of a single chip, or consider the failure of multiple electrical parameters. However, these methods do not consider the correlation between electrical parameters and have limitations.

[0008] 3. Statistical indicators are not calculated intelligently

[0009] At present, the calculation of statistical indicators, especially the defect rate, is often obtained by table lookup. The premise of using the table lookup method is that the data obeys the normal distribution and there is no center shift. However, in the actual process, the data does not necessarily obey the distribution and may have deviations. Summary of the invention

[0010] In view of the above-mentioned shortcomings of the related art, an object of the present invention is to provide a method for screening electrical parameters to solve the above-mentioned problems.

[0011] To achieve the above-mentioned purpose and other related purposes, the present invention provides a method for screening electrical parameters, the steps of which include: counting the electrical parameters of each chip on the wafer and forming a group of parameter groups with the same electrical parameter, and identifying the distribution form of each parameter group; if the parameter group is normally distributed, using the standard deviation method to screen abnormal data to identify abnormal chips; if the parameter group is non-normally distributed, using the median absolute deviation method to preliminarily screen chips that show abnormalities in a single electrical parameter; based on the chips after the initial screening, using a multivariate anomaly detection method for secondary screening to identify chips that show abnormalities in multiple electrical parameters at the same time; counting the electrical parameters of each chip after screening, and forming a group of parameter groups with the same electrical parameter, identifying the distribution form of each parameter group, if it is non-normally distributed, using a conversion model to transform the data of the parameter group into a normal distribution; based on the converted data, performing statistical index characterization on each electrical parameter.

[0012] Optionally, for missing data or abnormal noise data that occurs during the process of counting the electrical parameters of each chip on the wafer, data processing is performed by eliminating or replacing the data.

[0013] Optionally, the step of identifying the distribution form of each parameter group is: first confirming whether the data distribution shape of each parameter group conforms to the normal distribution through descriptive statistics, and if it conforms to the normal distribution, then confirming it through inferential statistical methods.

[0014] Optionally, the step of first confirming the data distribution by descriptive statistics is: drawing a histogram of the electrical parameters based on all the data in the parameter group, and confirming whether the data is normally distributed by the shape of the histogram.

[0015] Optionally, the step of confirming by inferential statistical methods is: using the Shapiro-Wilk test method to determine whether the data is normally distributed by testing the significance of the results.

[0016] Optionally, the step of using the median absolute deviation method to initially screen chips that exhibit abnormalities in a single electrical parameter is as follows: each electrical parameter of each chip is tested one by one, and a detection range for judging each electrical parameter is determined based on the median absolute deviation method; if an electrical parameter is detected to be outside the prescribed detection range, the chip is abnormal and detection of other electrical parameters is stopped.

[0017] Optionally, each electrical parameter is divided into a key parameter and a non-key parameter, and different detection ranges are set for the key parameter and the non-key parameter for evaluation. If the detected electrical parameter exceeds the detection range, the chip is identified as abnormal.

[0018] Optionally, it is characterized in that: the step of using the multivariate anomaly detection method for secondary screening is: introducing a Mahalanobis distance model and improving the Mahalanobis distance model, and the improved Mahalanobis distance formula is as follows:

[0019]

[0020] x is the electrical parameter vector of the detected wafer;

[0021] S is the covariance matrix of all electrical parameters;

[0022] median(x) is the median vector of the electrical parameters of the chip after preliminary selection;

[0023] The Mahalanobis distance of each wafer is calculated based on the above improved formula, and then the detection range of the secondary screening is determined based on the number set composed of the Mahalanobis distances of each wafer and the median absolute deviation method to screen out abnormal wafers.

[0024] Optionally, the content represented by the statistical indicators includes the process capability index and defective rate of electrical parameters.

[0025] A readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the above-mentioned electrical parameter screening method.

[0026] As described above, the method for screening electrical parameters of the present invention has the following beneficial effects: the method includes the steps of identifying distribution, screening anomalies, converting distribution and characterizing statistical indicators, thereby realizing efficient screening and reliable evaluation of wafer electrical parameters. In semiconductor electrical parameters, there is often a certain correlation between different parameters (such as the relationship between voltage and current), and this correlation may affect the screening results. The multivariate anomaly detection method used in this method can eliminate this correlation interference, thereby accurately identifying wafers that exhibit abnormalities in multiple parameters at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Shown is a schematic flow chart of a method for screening electrical parameters in an embodiment of the present invention.

[0028] Figure 2 Displayed is a statistical histogram of the electrical parameter IDSS raw data in an embodiment of the present invention.

[0029] Figure 3 Shown is a statistical histogram of IDSS of each wafer after screening in an embodiment of the present invention.

[0030] Figure 4 Shown is the IDSS statistical histogram of each wafer after transformation in the embodiment of the present invention. DETAILED DESCRIPTION

[0031] The following describes the embodiments of the present invention through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention.

[0032] For example, when describing the embodiments of the present invention in detail, for the sake of convenience, the cross-sectional view showing the device structure will not be partially enlarged according to the general scale, and the schematic view is only an example, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional space dimensions of length, width and depth should be included.

[0033] For ease of description, spatial relational terms such as "under", "below", "below", "below", "above", "on", etc. may be used herein to describe the relationship of one element or feature shown in the drawings to other elements or features. It will be understood that these spatial relational terms are intended to include other directions of the device in use or operation in addition to the directions depicted in the drawings. In addition, when a layer is referred to as being "between" two layers, it can be the only layer between the two layers, or there can be one or more intervening layers. As used herein, "between..." means including the end point values.

[0034] In the context of the present application, a structure in which a first feature is described as being "above" a second feature may include embodiments in which the first and second features are formed in direct contact, and may also include embodiments in which additional features are formed between the first and second features, such that the first and second features may not be in direct contact.

[0035] It should be noted that the illustrations provided in this embodiment are only used to illustrate the basic concept of the present invention in a schematic manner, and therefore the illustrations only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0036] like Figure 1 As shown, this embodiment provides a method for screening electrical parameters, and the specific steps of the method include:

[0037] S1. Count the electrical parameters of each chip on the wafer and form a parameter group with the same electrical parameters, and identify the distribution form of each parameter group.

[0038] Specifically, each wafer has multiple chips. To ensure the quality of the chip, the electrical parameters of each chip need to be tested. This embodiment uses the DPAT scheme to test the electrical parameters of the chip. During the test, data missing or abnormal noise data may appear. At this time, the data needs to be preprocessed. The missing values ​​or abnormal noise data can be removed, or the mean or median of the electrical parameter can be used to replace them, so as to ensure the integrity of the data. After the data test and collection of each electrical parameter is completed, the data of the same electrical parameter is formed into a group of parameter groups. For example, the IDSS of each chip is tested one by one, and then the IDSS data of all chips are formed into a group of parameter groups.

[0039] Then, descriptive statistics are used to confirm whether the data distribution shape of each parameter group conforms to the normal distribution. If it conforms to the normal distribution, it is confirmed by inferential statistical methods.

[0040] Specifically, a histogram of electrical parameters is drawn based on all the data in the parameter group, and the shape of the histogram is used to determine whether the data is normally distributed. For example, when judging the distribution of the IDSS parameter group, the IDSS data is used as the x-axis and the frequency is used as the y-axis to draw a histogram, such as Figure 2 Then, based on the shape of the histogram, it is preliminarily determined whether the parameter group conforms to the normal distribution. If it is preliminarily determined to be a normal distribution, it is necessary to further determine it through inferential statistical methods. This embodiment adopts the Shapiro-Wilk test method to determine whether the data is normally distributed by the significance of the test results.

[0041] S2. If the data of the parameter group conforms to the normal distribution, the standard deviation method is used to screen abnormal data to identify abnormal chips. Specifically, the mean μ and standard deviation σ can be obtained according to the data in each parameter group, and then μ±3σ is determined as the detection range. Then, each data in each parameter group is compared with the corresponding detection range one by one. If the data exceeds the detection range, it indicates that the chip is abnormal. It should be noted that if the electrical parameters of the chip are screened as abnormal, the chip is directly identified as abnormal and there is no need to continue to detect whether other electrical parameters are abnormal. In this way, test time and resources can be saved.

[0042] S3. If the data of the parameter group does not conform to the normal distribution, the median absolute deviation method is used to preliminarily screen the wafers that show abnormal performance in a single electrical parameter. The steps of preliminarily screening the wafers that show abnormal performance in a single electrical parameter using the median absolute deviation method are: each electrical parameter of each wafer is tested one by one, and the detection range of each electrical parameter is determined based on the median absolute deviation method. If the electrical parameter is detected to be beyond the specified detection range, the wafer is abnormal and the detection of other electrical parameters is stopped.

[0043] Specifically, the median and median absolute deviation MAD can be obtained according to the data in each parameter group, where MAD = 1.4826 × median (|X i -median(X)|), X i The electrical parameter values ​​are measured. Taking the electrical parameter IDSS as an example, the difference between each measured IDSS value and the median in the parameter group is calculated, and then the median of the absolute values ​​of these differences is taken as the MAD.

[0044] Each electrical parameter is divided into key parameters and non-key parameters, and different detection ranges are set for the key parameters and non-key parameters for evaluation. If the detected electrical parameter exceeds the detection range, the chip is identified as abnormal.

[0045] Specifically, for key parameters, the detection range for judging electrical parameters is set to median±3MAD. If the detected electrical parameters exceed median±3MAD, the chip is abnormal; for non-key parameters, the detection range for judging electrical parameters is set to median±4MAD. If the detected electrical parameters exceed median±4MAD, the chip is abnormal. Similarly, if the electrical parameters of a chip are screened as abnormal, the chip is directly identified as abnormal and there is no need to continue to detect whether other electrical parameters are abnormal.

[0046] S4. Based on the wafers after the initial screening, a multivariate anomaly detection method is used for secondary screening to identify wafers that show abnormalities in multiple electrical parameters at the same time. Specifically, the Mahalanobis distance model is introduced and improved, that is, the median is used to replace the mean in the original formula. The improved Mahalanobis distance formula is as follows:

[0047]

[0048] x is a vector composed of various electrical parameters of the chip;

[0049] S is the covariance matrix of the electrical parameters of the chip;

[0050] median(x) is the median vector of the electrical parameters of the chip after preliminary selection;

[0051] Based on the above improved formula, the Mahalanobis distance of each chip can be calculated, and then the detection range of the secondary screening is determined based on the number set composed of the Mahalanobis distances of each chip and the median absolute deviation method to screen out abnormal chips.

[0052] Specifically, the wafers after the initial screening are taken as the detection objects, and then the Mahalanobis distance of each wafer is calculated one by one. When calculating a certain wafer, the electrical parameters of the wafer are represented in the form of vectors, which is x in the formula; at the same time, the covariance matrix of the electrical parameters of the wafer is calculated, which is S; then the Mahalanobis distance of the wafer is calculated by the improved Mahalanobis distance formula. It can be understood that each wafer can calculate a Mahalanobis distance through the above formula. At this time, the Mahalanobis distances of all wafers constitute a number set. The median median and median absolute deviation MAD of the number set are obtained using the data in the number set, and the detection range is determined to be median ± 3MAD. Finally, the Mahalanobis distance of each wafer is compared with the detection range. If it exceeds the detection range, the wafer is identified as abnormal. In this embodiment, the improved Mahalanobis distance transforms the original data through the inverse matrix of the covariance matrix, and the median is used instead of the mean to represent the data center position, so as to accurately identify wafers that are abnormal in multiple parameters at the same time.

[0053] S5, statistically analyzing the electrical parameters of each wafer after screening, and forming a parameter group with the same electrical parameters, and identifying the distribution form of each parameter group, Figure 3 is the IDSS statistical histogram of each wafer after screening. If it is a non-normal distribution, the conversion model is used to transform the data of the parameter group into a normal distribution. Figure 4 is the IDSS statistical histogram of each wafer after transformation. It can be understood that the step of identifying the normal distribution in this step is the same as the step of identifying the normal distribution mentioned above.

[0054] Specifically, this embodiment uses the Box-cox power transformation model to transform non-normally distributed data into a form that satisfies normal distribution. The formula of the Box-cox power transformation model is as follows:

[0055]

[0056] The Box-Cox power distribution depends on the use of a single parameter λ, whose optimal value can be found by maximum likelihood estimation.

[0057] Among them, λ is the transformation parameter. The core of Box-Cox transformation is to find an optimal transformation that makes the data close to normal distribution by adjusting the value of λ. When λ=1, Box-Cox transformation is linear (that is, it does not change the data). When λ=0, Box-Cox transformation degenerates into logarithmic transformation. Different λ values ​​will exert different degrees of stretching or compression on the data, thereby adjusting the distribution form of the data. Box-Cox is used to optimize the assumptions of the model and improve the predictive ability of the model.

[0058] S6. Based on the converted data, statistical indicators are used to characterize each electrical parameter.

[0059] Specifically, the statistical indicators include the process capability index C of the electrical parameters. pk and defect rate P pm Among them, C pk The higher the value, the higher the process capability. pm The smaller the value, the higher the product reliability. Taking the VDSON (on-state drain-source voltage) parameter as an example, Table 1 compares the process capability index C before and after the screening method of this embodiment is used. pk and defect rate P pm .

[0060] Table 1

[0061] stage <![CDATA[C pk ]]> <![CDATA[P pm ]]> Raw data 1.60 0.75 Filtered data 2.15 0.000060

[0062] This embodiment also discloses a readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the above-mentioned electrical parameter screening method. The computer instructions can also enable the computer to generate charts and reports to help users intuitively understand the screening results. After the screening is completed, a screening report is automatically generated to record in detail the screening process, parameter settings, screening results and statistical analysis results.

[0063] Specifically, the operation of computer instructions is realized through software. For example, to realize automatic calculation through Visual Basic, select a suitable development environment and method, create a new Windows Form application in Visual Studio, and add necessary controls, including TextBox (for inputting or displaying data), Button (triggering calculation), DataGridView (for displaying result list), and Label (for displaying status information or parameters). The application is packaged as an executable file and provided to testers or production lines. Based on user feedback and production needs, the algorithm and user interface are further optimized to enhance the performance and usability of the software. Through reasonable deployment and optimization, the software can be used for electrical parameter testing and screening in the field of wafer manufacturing.

[0064] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A method for screening electrical parameters, characterized in that: The steps of the method include: Count the electrical parameters of each chip on the wafer and form a parameter group with the same electrical parameters, and identify the distribution form of each parameter group; If the parameter group is normally distributed, the standard deviation method is used to screen abnormal data to identify abnormal wafers; If the parameter group is non-normally distributed, the median absolute deviation method is used to preliminarily screen out wafers that show abnormal performance in a single electrical parameter; Based on the wafers after the initial screening, a multivariate anomaly detection method is used for secondary screening to identify wafers that exhibit abnormalities in multiple electrical parameters at the same time; The electrical parameters of each wafer after screening are counted, and the same electrical parameters are grouped into a parameter group, and the distribution form of each parameter group is identified. If it is a non-normal distribution, the data of the parameter group is transformed into a normal distribution using a conversion model; Based on the converted data, statistical indicators are used to characterize each electrical parameter.

2. The method for screening electrical parameters according to claim 1, characterized in that: In the process of counting the electrical parameters of each chip on the wafer, missing data or abnormal noise data is processed by elimination or replacement.

3. The method for screening electrical parameters according to claim 1, characterized in that: The step of identifying the distribution form of each parameter group is: first confirm whether the data distribution shape of each parameter group conforms to the normal distribution through descriptive statistics, and if it conforms to the normal distribution, then confirm it through inferential statistical methods.

4. The method for screening electrical parameters according to claim 3, characterized in that: The step of first confirming the data distribution through descriptive statistics is: drawing a histogram of the electrical parameters based on all the data in the parameter group, and confirming whether the data is normally distributed through the shape of the histogram.

5. The method for screening electrical parameters according to claim 3, characterized in that: The step of confirming by inferential statistical methods is: using the Shapiro-Wilk test method to determine whether the data is normally distributed by checking the significance of the test results.

6. The method for screening electrical parameters according to claim 1, characterized in that: The steps of using the median absolute deviation method to initially screen chips that show abnormalities in a single electrical parameter are: each electrical parameter of each chip is tested one by one, and the detection range of each electrical parameter is determined based on the median absolute deviation method. If the electrical parameter is detected to be beyond the specified detection range, the chip is abnormal and the detection of other electrical parameters is stopped.

7. The method for screening electrical parameters according to claim 6, characterized in that: Each electrical parameter is divided into key parameters and non-key parameters, and different detection ranges are set for the key parameters and non-key parameters for evaluation. If the detected electrical parameter exceeds the detection range, the chip is identified as abnormal.

8. The method for screening electrical parameters according to claim 1, characterized in that: The steps of using the multivariate anomaly detection method for secondary screening are: introducing the Mahalanobis distance model and improving the Mahalanobis distance model. The improved Mahalanobis distance formula is as follows: x is the electrical parameter vector of the detected wafer; S is the covariance matrix of all electrical parameters; median(x) is the median vector of the electrical parameters of the chip after preliminary selection; The Mahalanobis distance of each wafer is calculated based on the above improved formula, and then the detection range of the secondary screening is determined based on the number set composed of the Mahalanobis distances of each wafer and the median absolute deviation method to screen out abnormal wafers.

9. The method for screening electrical parameters according to claim 1, characterized in that: The statistical indicators characterize the content including the process capability index and defective rate of electrical parameters.

10. A readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable the computer to execute the electrical parameter screening method according to any one of claims 1 to 9.

Images