KR100696276B1 - Automatic defect classification system based on the measurement data from wafer defect inspection equipments - Google Patents

Abstract

An automatic defect classification system based on measurement data obtained from wafer defect inspection equipment is provided to apply an optimum automatic defect classification system to the wafer defect inspection equipment according to corresponding characteristics. An automatic defect classification system based on measurement data obtained from wafer defect inspection equipment includes a loader part, an automatic defect classification client part, and an automatic defect classification service part. The loader part(100) is used for collecting measurement data according to wafer inspection processes of a variety of inspection equipment(10) and ordering a track-out to the corresponding inspection equipment according to a track-out control signal. The automatic defect classification client part(300) is used for setting classification recipes for the operation of an automatic defect classification and evaluating a corresponding reference. The automatic defect classification service part(200) is used for performing the automatic defect classification by receiving upload data of the loader part according to an automatic classification reference supplied from the automatic defect classification client part and transferring the results to a production managing system server part(400).

Description

Automatic defect classification system based on the measurement data from wafer defect inspection equipments

1 is a block diagram of a conventional defect classification system.

2 is a diagram illustrating a conventional decision tree.

3 is a schematic diagram of an automatic defect classification system (ADC System) of the present invention.

4 is an overall configuration diagram of an automatic defect classification system according to the present invention.

5 is a detailed configuration diagram of a loader unit according to the present invention.

6 is a detailed configuration diagram of a service unit according to the present invention.

7 is a detailed configuration diagram of a client unit according to the present invention.

8 is a data flowchart of an automatic defect classification system according to the present invention.

9 is an exemplary configuration diagram of a decision tree and a discriminant function according to the present invention.

10 is an exemplary diagram for selecting a feature vector according to the present invention.

<Explanation of symbols for the main parts of the drawings>

10: inspection equipment 100: loader

110: inspection equipment interface (Interface)

120: data conversion unit 130: uploader

140: receiver 150: result database

200: automatic defect classification service (ADC Service)

210: Automatic defect classification master (ADC Master) unit

211 Loader Interface Module

212: Automatic defect classification client interface module

213: Manufacturing Execution System Interface Module

214: Automatic fault classification master control module

215: Automatic defect classification service control module

216: Automatic defect classification service monitoring module

220: automatic fault classification service module
230: Automatic Defect Classification Recipe Database
300: client
310: input / output interface unit 320: decision tree configuration unit
330: Feature Vector Selection Unit
340: Discriminant Function Module
350: classification rate evaluation module 360: user interface
370: Recipe / Sample Database
400: production management system server (MES Server)

The present invention relates to a wafer automatic detection system (Wafer Automatic Defect Classification System), in particular, to integrate the automatic classification system provided separately for each inspection equipment into one, but to apply the optimal automatic classification system according to the characteristics of each inspection equipment. The present invention relates to a system for constructing a decision tree based on measurement data obtained from each wafer defect measurement device, and automatically classifying the types of defects generated on a wafer.
A general conventional wafer defect measurement system is constructed as shown in FIG. 1 attached to the present invention. Referring to this description, in the conventional system, as shown by reference numeral 1, each inspection equipment manufacturing company for each inspection equipment 1A is represented. Although the automatic classification system (1B) of the equipment provided by the company is being introduced, in order to operate the automatic defect classification system effectively, the process operator must understand the inspection principle of each equipment and the characteristics and meaning of the data collected from the equipment. You must have enough prior knowledge, and every time you introduce a new piece of equipment, this knowledge must be learned anew.

delete

Moreover, poor data or confidentiality of the nature and meaning of the measurement data at the equipment manufacturer makes it more difficult for the user to actively identify automatic defects, which is a fundamental problem in yield management.

The decision tree adopted in the conventional system as shown in FIG. 1 is generally as shown in FIG. 2. In order to construct a decision tree, a user may use a feature vector for determining a defect class in the collected data. Vector) should be selected.

However, the number of feature vectors possessed by the measured data ranges from tens to hundreds. However, the number of feature vectors that a person can grasp at one time in order to construct a decision tree is difficult to exceed three or more.

Therefore, in reality, it is only possible to construct a decision tree based on a mathematical model using a computer to apply a decision tree to an N-dimensional feature vector. However, it is still impossible for the user to intuitively recognize this decision tree. This modifies the decision tree or renders the user unpredictable in the behavior of the decision tree.

On the other hand, classifiable feature vectors may be arbitrarily selected in order to effectively construct a decision tree and allow a user to intuitively recognize the decision tree.

However, even in such a case, the conventional system has important problems such as the method of selecting a feature vector and the accuracy of a defect classification rate when a decision tree is constructed from the selected feature vectors. In other words, the user must evaluate whether it is a feature vector that can improve the classification rate, but it is realistic to construct a decision tree by calculating a combination of n feature vectors from the N-dimensional feature vectors, and calculate and evaluate the classification rate. impossible.

Therefore, the decision tree construction system needs a method that selects the feature vector with the highest classification performance and recommends it to the user. It is obvious that the data having N-dimensional feature vectors collected in the defect measuring equipment of the wafer should have a decision tree based on the N-dimensional feature vectors in order to have the maximum classification rate.

On the contrary, as described above, in order to make a decision tree that can be understood by the user, it is appropriate to select two or three feature vectors. In the conventional system, since the decision tree is constructed using only the selected feature vector, there is no limit compared with the classification of the feature vectors of the N-dimensional feature. In addition, in a conventional decision tree, a decision rule can be set only vertically and horizontally. There is a limit in expressing various types of crystal regions, which in turn lowers the automatic classification rate of wafer defects.

delete

An object of the present invention for solving the above problems relates to a wafer defect measurement system, in particular, to integrate the automatic defect classification system provided separately for each inspection equipment into one, but the optimal automatic classification system according to the characteristics of each inspection equipment It is to provide a system that constructs a decision tree based on the measurement data obtained from each wafer defect measuring equipment so that the method can be applied, and automatically classifies the types of defects generated on the wafer.

Automatic defect classification system using the measurement data obtained from the wafer defect inspection equipment according to the present invention for achieving the above object, which manages all the inspection equipment and various parts of the production process for wafer (wafer) defect inspection A system for manufacturing a system for automatically classifying a kind of defects generated in a semiconductor wafer manufacturing process, comprising a manufacturing execution system server part (MES), and measuring according to wafer inspection from the various inspection equipments 10. A loader part 100 configured to collect the collected data, convert the converted data into an integrated data integrated format, output the received data, and instruct track out to the corresponding inspection equipment according to the received track-out control signal; An automatic defect classification client part 300 for setting a classification recipe for performing an automatic defect classification (ADC) operation and evaluating corresponding criteria; Receive the data uploaded from the loader unit 100 according to the automatic classification criteria provided from the automatic defect classification client unit 300 and performs automatic defect classification, and the result is the production management system server unit ( It characterized in that it comprises an automatic defect classification service unit 200 to transmit to 400.

The loader unit 100 includes a test equipment interface unit 110 for collecting a map data file and a test result file for each device generated by a plurality of different types of test equipment; A data converter 120 for converting all data collected through the test equipment interface 110 into a designated integrated data format; The data conversion unit 120 may be configured to include an upload unit 130 that transmits the data converted by the automatic defect classification service unit 200.

The inspection equipment interface unit 110 may convert the map data file for each device into a designated standard map data file.

The automatic defect classification service unit 200 includes a loader interface module 211 for checking and parsing a format of data transmitted from the loader unit 100; An automatic defect classification client interface module (212) for collecting a preset automatic classification recipe (Recipe) in the automatic defect classification client unit (300); An automatic defect classification service module 220 for automatically classifying input data based on a preset automatic defect classification recipe introduced through the automatic defect classification client interface module 212; ; An Automatic Defect Classification Service Control Module (215) for monitoring the task status of each of the modules and assigning tasks; And a recipe database module 230 that stores classification criteria and procedures for performing automatic defect classification.

The apparatus may further include an automatic defect classification master control module 214 for controlling each module including the automatic defect classification service control module 215.

The automatic defect classification client part 300 includes an input / output interface 310 for performing data communication with the automatic defect classification service part 200 and a decision tree of the decision tree. A discriminant function module 340 for setting free lines, curves, and complex forms thereof as discriminant functions at each node; the most classifiable in a multidimensional feature vector to improve classification performance; A feature vector selector 330 for selecting the high feature vector, and a decision tree capable of constructing a final tree by applying a feature vector combination having a high classification rate at each decision node. (Decision Tree) unit 320, a classification rate evaluation module 350 for evaluating and monitoring the classification performance of the set decision tree; and a recipe / sample database module for storing the classification rate evaluation module 350 (Recipe / Sampl) e Database Module) 370.

The automatic defect classification client unit 300 may further include a user interface 360 that allows a user to comprehensively interpret a recipe setting and classification result.

The above object and various advantages of the present invention will become more apparent from the preferred embodiments of the present invention described below with reference to the accompanying drawings by those skilled in the art.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

3 is a schematic diagram of an automatic defect classification system (ADC System) according to the present invention, which is connected to various inspection equipment 10 for defect inspection of a semiconductor wafer, and the inspection equipment 10 through a network. Loader part (100) for collecting the results of the wafer inspection from the inspection equipment (10), and automatic defect classification service for performing the automatic defect classification by receiving the test results collected through the loader 100 There is an automatic defect classification service part (200), and is largely composed of an automatic defect classification client (ADC Client Part; 300) for setting the classification criteria required by the automatic defect classification service unit (200).

The Manufacturing Execution System Server, referred to by reference 400, is a system that manages all parts of the production process.

Looking at the above-described configuration in more detail, the loader part (100) is connected to a variety of inspection equipment 10 by a network (at this time using a variety of protocols (Protocol)) generated in each inspection equipment 10 Map data and inspection result data are collected, converted to a designated standard format, and then transmitted to the automatic defect classification service unit 200.

In addition, the automatic defect classification service unit 200 receives the data transmitted from the loader unit 100, and uses the classification criteria set in advance in the automatic defect classification client (ADC Client) 300, automatic defect The classification is performed, and the automatic classification result is aggregated and transmitted to the production management system server unit 400.

The automatic defect classification client unit 300 sets an automatic classification criterion (ADC Recipe) so that the automatic defect classification service unit 200 may perform automatic defect classification.

In this case, the user can set the automatic / manual classification function, configure the recipe by recommending the appropriate combination of feature vectors from the automatic defect classification client, and transmit or manage the recipe to the automatic defect classification service. In addition, classification performance can be assessed using previously collected data.

In addition, the production management system server unit 400 is a system that manages all parts of the production process, process-related messages (eg, track-out or interlock) based on the classification result collected from the automatic defect classification service unit. (Inter-Lock)) to the automatic defect classification service. The delivered message is delivered to the inspection equipment 10 through the data loader unit 100.

FIG. 3 is a schematic diagram of a system according to the present invention, with reference to FIGS. 4 to 7 showing the detailed configuration of each component thereof.

4 is an overall configuration diagram of an automatic defect classification system according to the present invention, FIG. 5 is a detailed configuration diagram of a loader unit according to the present invention, FIG. 6 is a detailed configuration diagram of a service unit according to the present invention, and FIG. The detailed configuration diagram of the client unit by FIG.

Looking at the detailed configuration and operation of the automatic defect classification system according to the present invention with reference to the above-described drawings, as shown in the accompanying Figure 5, and map data files generated by a plurality of different types of inspection equipment 10 and The test result file is collected through the tester interface unit 110 of the loader unit 100, converted into a standard format designated by the data converter unit 120, and then uploaded by the upload unit 130. Is configured to transmit).

In this case, the inspection equipment interface unit 110 converts the map data file into a designated standard map data file, and the data conversion unit 120 integrates the standard map data file and the inspection result data and uploads the upload unit 130. ) Transmits the file to the automatic defect classification service unit 200 to perform automatic defect classification.

In addition, the automatic defect classification service unit 200 may classify the automatic defect classification master 210, the automatic defect classification service module 220, and the classification criteria for performing the automatic defect classification as illustrated in FIG. 6. And a recipe database 230 for storing the procedure.

Therefore, the automatic defect classification master 210 receives the data collected by the loader 100 through the loader interface module 211.

The loader interface module 211 transfers this data file to the automatic defect classification master control module 214. The automatic defect classification master control module 214 manages / operates all modules belonging to the automatic defect classification master 210.

Meanwhile, the automatic defect classification client unit 300 transmits an automatic classification recipe set by the user through the automatic defect classification client interface module 212. If using an automatic classification recipe set up in the past, the automatic defect classification master control module 214 pulls it from the recipe database 230.

The automatic defect classification master control module 214 transfers the input map / test result integrated data and the automatic classification recipe to the automatic defect classification service control module 215. The automatic defect classification service control module is a part that manages and supervises a plurality of automatic defect classification service modules 220. The automatic defect classification service control module monitors a workload or idle state of the automatic defect classification service module 220 and performs automatic defect classification service. It is in charge of control of start, end, execution, etc.

In addition, the automatic defect classification service control module 215 uses the service monitoring module 216 to monitor the status of the automatic defect classification service modules 220 which are being executed first, and to perform a new automatic defect classification service. Select modules to deliver standard map / inspection integration data and automatic defect classification recipes and perform automatic classification.

Therefore, the automatic defect classification service module 220 classifies the data by the automatic defect classification recipe and transmits the result to the automatic defect classification master control module 214 through the automatic defect classification service control module 215.

The analyzed result is transmitted to the production management system server unit 400 through the production management system interface module 213.

In addition, the automatic defect classification client unit 300, as shown in the accompanying Figure 7, the decision tree (Decision Tree) configuration unit 320 for establishing the classification configuration criteria, and the inspection results collected from the inspection equipment 10 The main components are a part 330 for selecting a feature vector for which automatic classification is effective from the data, and a discriminant function module 340 constituting a function capable of effectively classifying the selected feature vector.

In addition, there is a user interface 360 for allowing a user to effectively operate a client, a portion 370 for storing test result data and classification recipes, and a classification rate evaluation module 350 for evaluating performance of a set recipe.

The standard map / test result integrated data input to the automatic defect classification client unit 300 is slightly different from the data input to the automatic defect classification service module 220 of the automatic defect classification service unit 200. In other words, it becomes learning data for constructing an automatic defect classification recipe.

Therefore, the class of the defect must be already coded for all the checked defects, and the operation must be set by the user through the user interface 360 of the automatic defect classification client unit 300. do. This is a criterion for discriminating positive / false recognition when evaluating the classification performance of the automatic defect classification recipe in the classification rate evaluation module 350 later.

In addition, the production management system server unit 400 manages wafer information, statistical process analysis information (SPC: Statistical Process Control), and wafer messages transmitted from an equipment control system (ECS). As for the invention, the external system is the basic equipment for wafer measurement and process control.

The signal flow between the above-described components is illustrated as shown in FIG.

As shown in the drawing, the inspection equipment 10 stores map data and inspection results as a first step.

As a second step, the inspection equipment interface 110 converts the map data file into a standard map file (where the standard map file uses any standard format for analysis in the system of the present invention and the result processing thereof. do).

As a third step, after integrating the standard map file and the wafer inspection result data, the fourth step is performed by transferring the data to the upload unit 130. The upload unit 130 transmits the integrated file to the automatic defect classification service unit 200. The automatic defect classification service unit analyzes an input file based on an automatic defect classification recipe set by a user, that is, a decision tree, and generates automatic defect classification result data.

In the fifth step, the production management system interface module 213 transmits the result to the production management system server unit 400.

9 is an example of the role of the decision tree constructing unit 320 and the feature vector selecting unit 330 according to the present invention.

In comparison with FIG. 2 showing the conventional decision tree, the conventional decision tree is mostly classified by setting only horizontal / vertical regions for the selected feature vector.

However, in the present invention, as shown in FIG. 9, the feature vector is set differently for each decision node of the decision tree.

In other words, the classification ambiguity of the data that is obtained when only n (<< N) feature vectors are selected from the N-dimensional data and the difficulty of decision making from the user's point of view regarding all the N-dimensional data are considered. It can effectively solve these problems.

Therefore, by separately selecting feature vectors for each node, a decision tree can be created by considering all data in N-dimensional through a decision tree, which is a collection of nodes. N-analyses are typically 1, 2, 3 feature vectors to choose from.

In addition, as shown in the example of the classification function configuration of FIG. 9, it is possible to freely set the classification area such as a straight line, a curve, or a complex type so that the area setting is more accurate and easier to access than the vertical / horizontal only. In particular, in the case of complex region setting, a problem in which the decision tree grows rapidly can be solved.

10 is an exemplary view of the function of the feature vector selecting unit 330 according to the present invention.

The feature vectors generated by the inspection equipment 10 may range from several to as many as hundreds. Among them, it is far from a practical application for the user to select the feature vectors for the selection of the feature vectors having good classification performance.

Therefore, in the present invention, after automatically constructing an optimal discriminant function for each feature vector combination, a table shown in Table 1 below is formed.

Therefore, as shown in Table 1, the accuracy of classification and the purity of classification are evaluated to recommend the combination of the two highest feature vectors so that the user can build a decision tree based on this. To help.

Here, the accuracy and purity of the classification are explained by using the table in Table 1.

Assuming that each defect belongs to two classes (Class 1, Class 2), the relationship between each defect is summarized in Tables 2 to 5 below.

In this case, the accuracy means the ratio of classifying the recognized class 1 when the automatic defect classification recipe is applied among the class 1 defects. Purity means the recognized classification of the defects classified as Class 1 with the result of applying the automatic defect classification recipe to all the defects.

While the invention has been shown and described in connection with specific embodiments thereof, it is well known in the art that various modifications and changes can be made without departing from the spirit and scope of the invention as indicated by the claims. Anyone who owns it can easily find out.

When providing an automatic defect classification system using the measurement data obtained from the wafer defect inspection equipment according to the present invention as described above, the user can grasp the measurement principle of the equipment and the meaning of each data calculated therefrom Solve the need to cultivate prior knowledge that is less relevant to the manufacturing process, and users can build decision trees more familiar and faster with a single tool, regardless of the type of equipment.

In addition, the cost of operating the system can reduce the additional cost of purchasing expensive analysis tools provided by the manufacturer for automatic defect classification for each analysis equipment.

In addition, it is possible to have a high user convenience in management, such as eliminating the need to set and manage the decision tree constructed from the measurement data for each device.

Claims (7)

Various inspection equipments for wafer defect inspection and a manufacturing execution system server part (MES) that manage all parts of the production process and automatically classify the types of defects that occur in the semiconductor wafer manufacturing process In the system for Collect the measured data according to the wafer inspection from the various inspection equipment 10, converts it into an integrated data integrated format, outputs it, and commands the track equipment to the corresponding inspection equipment according to the received track-out control signal. A loader part (Loader Part) 100; An automatic defect classification client part 300 for setting a classification recipe for performing an automatic defect classification (ADC) operation and evaluating corresponding criteria; Receive the data uploaded from the loader unit 100 according to the automatic classification criteria provided from the automatic defect classification client unit 300 and performs automatic defect classification, and the result is the production management system server unit ( Automatic defect classification system using the measurement data obtained from the wafer defect inspection equipment, characterized in that it comprises an automatic defect classification service unit (200) for transmitting to. The method of claim 1, The loader unit 100 includes a test equipment interface unit 110 for collecting a map data file and a test result file for each device generated by a plurality of different types of test equipment; A data converter 120 for converting all data collected through the test equipment interface 110 into a designated integrated data format; Measurement obtained from the wafer defect inspection equipment, characterized in that it comprises an upload unit (130) for transmitting the data converted through the data conversion unit 120 to the automatic defect classification service unit 200 Automatic defect classification system using data. The method of claim 2, The inspection equipment interface unit 110, the automatic defect classification system using the measurement data obtained from the wafer defect inspection equipment, characterized in that for converting the map data file for each device into a designated standard map data file. The method of claim 1, The automatic defect classification service unit 200 includes a loader interface module 211 for checking and parsing a format of data transmitted from the loader unit 100; An automatic defect classification client interface module (212) for collecting a preset automatic classification recipe (Recipe) in the automatic defect classification client unit (300); An automatic defect classification service module 220 for automatically classifying input data based on a preset automatic defect classification recipe introduced through the automatic defect classification client interface module 212; An Automatic Defect Classification Service Control Module (215) for monitoring the task status of each of the modules and assigning tasks; And An automatic defect classification system using measurement data obtained from wafer defect inspection equipment, comprising a recipe database module 230 for storing classification criteria and procedures for performing automatic defect classification. The method of claim 4, wherein Obtained from wafer defect inspection equipment, characterized in that it further comprises an Automatic Defect Classification Master Control Module (214) for controlling each module including the automatic defect classification service control module (215). Defect classification system using the measured measurement data. The method of claim 1, The automatic defect classification client part 300 includes an input / output interface 310 for performing data communication with the automatic defect classification service part 200; A discriminant function module 340 for setting free lines, curves, and complex forms thereof as discriminant functions at each node of the decision tree; A feature vector selector 330 that selects a feature vector having the most classifiability from a multidimensional feature vector to increase classification performance; A decision tree constructing unit 320 capable of constructing a final tree by applying a feature vector combination having a high classification rate in each decision node; A classification ratio evaluation module 350 for evaluating and monitoring classification performance of the set decision tree; And Automatic defect classification system using the measurement data obtained from the wafer defect inspection equipment, characterized in that it comprises a Recipe / Sample Database Module (370) for storing it. The method of claim 6, The automatic defect classification client unit 300 further includes a user interface 360 that allows a user to comprehensively interpret the recipe setting and classification results. Automatic defect classification system.

Images