JP6623904B2 - Abnormality analysis apparatus and abnormality analysis method in semiconductor device manufacturing process - Google Patents

Description

  The present invention relates to an abnormality analysis apparatus and an abnormality analysis apparatus for automatically digitizing and analyzing an image of a non-defective product rate and a defect distribution in a wafer which change due to an abnormality of a manufacturing facility in a semiconductor device manufacturing process.

  In a semiconductor mass production line, a huge number of semiconductor products are manufactured using wafers made of semiconductor materials. Chips constituting a semiconductor product are two-dimensionally arranged in one wafer, and the number thereof is as large as several hundreds. The non-defective rate of a semiconductor product manufactured for each wafer is not always 100%, and some chips may be defective due to some reason in the manufacturing process. Particularly typical failure causes are adhesion of foreign substances and equipment abnormality generated during manufacturing.

  In the manufacture of semiconductor products, many wafers are processed on a wafer-by-wafer basis, and one wafer is processed under one processing condition. Therefore, the distribution within the wafer when processing is performed in each manufacturing facility is uniform. Hope that is. It is known that, when the equipment becomes abnormal, the uniformity and reproducibility are impaired, which causes a defect, and the distribution thereof matches the quality distribution of the final inspection.

  Therefore, a production line analysis device that analyzes a production line by using this fact is proposed in Patent Document 1. In the manufacturing line analysis apparatus disclosed in this document, foreign substances and defects are inspected during a wafer process, and a failure analysis result is notified to an operator so that the failure can be promptly dealt with.

  Specifically, in the manufacturing line analyzer disclosed in Patent Document 1, the number of particles present in a wafer and the like are stored in a failure analysis database in association with the location of the wafer. Whenever data such as the number of particles present in a wafer is obtained for a wafer processed by each process, the data is compared with past data stored in a failure analysis database to perform failure analysis. I do. As a result of the analysis, if a result indicating that the data is out of the standard is obtained, the operator is notified that the data is out of the standard, so that the operator can quickly detect the cause of the failure and deal with it.

JP 2010-173021 A

  However, in Patent Document 1 described above, newly obtained data is collated with past data that has been stored, and when they match, a failure similar to the failure shown in the data that has been stored in the past is detected. It can only be estimated that there is. In addition, since data is stored as past data without any data processing, and image data is only stored as it is, the data volume becomes enormous, and images of defective distribution are recognized by the data itself. And the tendency for abnormalities cannot be continuously monitored.

  In view of the above, the present invention provides an abnormality analysis apparatus and an abnormality analysis method in a semiconductor device manufacturing process capable of automatically quantifying a failure distribution and automatically analyzing and monitoring an abnormality of a manufacturing facility or the like. The purpose is to provide. Further, there is provided an abnormality analysis apparatus and an abnormality analysis method in a semiconductor device manufacturing process capable of automatically quantifying a failure distribution and automatically analyzing a causal relationship between an abnormality of a manufacturing facility and the failure distribution. That is another purpose.

  In order to achieve the above object, the abnormality analysis apparatus according to claim 1 includes a manufacturing facility (20a to 20c) and a monitor facility (20a to 20c) used when manufacturing semiconductor device chips using a wafer made of a semiconductor material. 30) An abnormality analysis device for analyzing an abnormality of a processing facility based on a processing result of at least one processing facility of a test facility (40), and a management system (110) for managing information on processing conditions in the processing facility. , An information obtaining and digitizing unit (131) for obtaining a processing result of the processing equipment, and digitizing an image of a failure distribution for each wafer based on the processing result; And a monitoring unit (133) for monitoring the abnormality of the processing equipment based on the information quantified in step (1). The information acquisition quantification unit divides the wafer surface into arbitrary physical positions, In terms of the position the goods chips are automatically digitized to fit the image a person feels the intensity of the defect distribution in the mutual position from the amount of defective chips at each location.

  As described above, the distribution image in which humans perceive the quality of product chips arranged two-dimensionally on a wafer is directly linked to the occurrence of equipment abnormality or foreign matter that causes the distribution. It is known that there is an opportunity to analyze the cause-and-effect relationship numerically and to quickly monitor the characteristic change points to quickly address the cause of the quality deterioration.

  Further, the abnormality analysis device according to claim 2 is based on the information storage unit (132) for normalizing and storing the information quantified by the information acquisition and quantification unit, and based on the quantified information stored in the information storage unit. And a causal relationship analysis unit (135) for analyzing the causal relationship with the abnormality of the processing equipment.

  In this way, the digitized information can be normalized and accumulated, and the causal relationship with the abnormality of the processing equipment can be analyzed based on the accumulated digitized information. For example, a semiconductor manufacturing process includes hundreds of manufacturing processes, and many manufacturing facilities are used. Even for the same product, the combination of manufacturing equipment differs depending on the manufacturing lot. If this is normalized by a product name, a processing step name, a manufacturing facility name, a processing condition name, a processing date and time, a change in a numerical value of the distribution in the range can be easily observed. By comparing this with the processing record of each product wafer and comparing each distribution, and comparing it with the distribution of the final test, it is possible to analyze processes and equipment that have a strong causal relationship with the distribution of the final test It becomes.

  In addition, the code | symbol in parenthesis of the said each means shows an example of the correspondence with the concrete means described in embodiment mentioned later.

FIG. 1 is a diagram illustrating a schematic configuration of a processing facility and an abnormality analysis device that perform a semiconductor device manufacturing process according to a first embodiment of the present invention. FIG. 3 is a diagram illustrating details of a control block configured in a computer. FIG. 3 is a diagram showing an example in which a wafer is divided into arbitrary physical positions. In order to match with the sense of position of the distribution image perceived by humans, here, the types of “concentric (cr: circle)”, “horizontal direction (ro: row)”, “vertical direction (co: column)” and Although an example of three divisions (x, y, z) is shown, various positions can be defined depending on the case. FIG. 4 is a diagram showing, as blocks, specifically subdivided positions generated from predetermined combinations of positions (ci, ro, co and x, y, z) shown in FIG. 3. It is a figure following FIG. 4A. FIG. 4B is a view following FIG. 4B. FIG. 5 is a diagram listing all blocks of the specific distribution shown in FIGS. 4A, 4B, and 4C. FIG. 6 is a diagram illustrating that all 17 types shown in FIG. 5 are covered on a wafer as a result of being arranged on the wafer. It is explanatory drawing of the distribution image which a person perceives, (a) is a figure which showed the difference of the intensity | strength of the distribution perceived by one defect in the area different for every block, and (b) with respect to (a). FIG. 9 is a diagram showing a state after area correction. FIG. 9 is a schematic diagram showing a relationship between chips formed in a wafer and respective positions, taking a wafer having a total number of effective chips of 968 as an example. 9 is a table showing a case where a correction coefficient for area correction is calculated based on an average value of the number of chips at all positions. A combination of three types (uniform, sparse, significant) of the strength of variation of defective chips at (x, y, z) for each position (ci, ro, co) is made into a matrix, and the combined strengths It is a figure roughly divided into ten types. FIG. 11 is a diagram illustrating combinations that actually exist in FIG. 5 (position) and FIG. 10 (strength of variation for each type of position), and is a diagram showing the same combinations of target combinations on a diagonal line. It is a figure following FIG. 11A. A list of all combinations that exist at 17 positions shown in FIG. 5 (positions) and 10 types of strengths shown in FIG. 10 (variation strength for each type of position), and corresponds to each digitization It is a figure showing a digit and a numerical value. It is a figure following FIG. 12A. It is a figure following FIG. 12B. It is a figure following FIG. 12C. FIG. 21 is a diagram illustrating a specific procedure of a process executed by a sixth processing unit.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following embodiments, portions that are the same or equivalent are denoted by the same reference numerals and described.

(1st Embodiment)
An abnormality analysis apparatus and an abnormality analysis method in a manufacturing process of a semiconductor device according to the present embodiment will be described with reference to FIGS.

  In the manufacturing process of semiconductor products, after the product lot that was prepared in the preparation section was transported to multiple manufacturing facilities, various processes were performed at each manufacturing facility, and it was judged as a good product through the monitoring process and the testing process The product is a completed semiconductor product. The abnormality analysis apparatus described in the present embodiment performs data collection and monitoring of processing equipment such as manufacturing equipment, monitor equipment, and test equipment used in the manufacturing process of the semiconductor device, and performs processing in the manufacturing process of the semiconductor device. Perform an anomaly analysis. Specifically, the abnormality analyzer automatically digitizes and analyzes an image of a non-defective product rate and a defect distribution in a wafer which change due to an abnormality of the manufacturing equipment in a manufacturing process of the semiconductor device, thereby analyzing the manufacturing equipment. Automatically analyze the causal relationship between abnormalities and failure distribution. The image of the defect distribution is an image of the position and the strength of the defect distribution. In the following description, although there is a portion abbreviated as a distribution image, it means an image of a defect distribution here.

  As shown in FIG. 1, a semiconductor device manufacturing process is performed by an automated manufacturing line (hereinafter, referred to as FA (FAB Automation)). The FA includes processing equipment such as various manufacturing equipments 20a to 20c, a monitor equipment 30, and a test equipment 40 in addition to the preparation unit 10.

  The preparation unit 10 prepares a semiconductor device by preparing, for example, a product lot, that is, a set of a plurality of wafers made of a semiconductor material for a carrier.

  In each of the manufacturing facilities 20a to 20c, various processes performed as a semiconductor device manufacturing process, such as a thin film forming process, a patterning process, and an impurity ion implantation process are performed. Here, three manufacturing facilities 20a to 20c are shown as an example, but various other facilities corresponding to the manufacturing process are provided. Each of the manufacturing facilities 20a to 20c includes information on each of the manufacturing facilities 20a to 20c, for example, a result of actually measuring the performance in a process performed by the manufacturing facilities 20a to 20c for each product lot, a processing condition, a processing date and time, and a processing number. Computers 21a to 21c for storing processing histories related to the above are connected.

  Performance information and processing history information from each of the computers 21a to 21c are collected by an EDC (abbreviation of Engineering Data Collection) 22 and accumulated in a server 23 or the like. Then, the information is transmitted to an online information interface 120 provided in the abnormality analysis device 100 described later through an MES (abbreviation of Manufacturing Execute System) 24.

  The MES 24 is a central component of FA automation. Generally, automation consists of several components, the center of which is called MES24. Although only the MES 24 is shown here, it is needless to say that the MES 24 may be composed of a plurality of components. The MES 24 obtains performance information and processing history information from each of the computers 21a to 21c and transmits the information to the online information interface 120, and also obtains information on manufacturing conditions from the online information interface 120 and converts the manufacturing conditions to each of the computers 21a to 21c. Tell Based on this, each of the computers 21a to 21c executes the manufacturing process based on the transmitted manufacturing conditions. As described above, the MES 24 is a central component, the online information interface 120 and the computers 21a to 21c of the respective manufacturing facilities 20a to 20c collect information, and the management of each manufacturing process is centrally managed under desired manufacturing conditions. It has become.

  The monitor equipment 30 measures the performance check circuit (monitor) incorporated in the product chip, inputs the measurement result to the server 31, and transmits the measurement result to the online information interface 120 through the server 31. For example, chips on wafers used to manufacture semiconductor products are being formed with special circuits, etc. to confirm that the performance is being performed as designed during the manufacturing process. And a measurement is made as to whether the particular circuit operates properly. The measurement result is transmitted to the online information interface 120. In the monitor equipment 30, the distribution is measured for each measurement function (hereinafter referred to as a category).

  The test equipment 40 performs an electrical property test, that is, measures electrical properties of a semiconductor product that is actually manufactured as a final product, and determines whether the semiconductor device is a non-defective product or a non-defective product from the measurement result. At the same time, check the non-defective rate and the defective distribution. Then, semiconductor devices determined as non-defective products are placed in the distribution process as completed products, and defective products are excluded from the completed products. In addition, the test equipment 40 communicates the test result, that is, the non-defective product rate, the defect distribution, the content of the defect when a defect occurs, the coordinate information of the defective chip, and the like to the online information interface 120 through the server 41. In the test equipment 40 as well, the distribution of non-defective products is measured for each category, that is, for each test item.

  On the other hand, the abnormality analysis device 100 is configured to include an SCS (abbreviation for Specification Control System) 110, an online information interface 120, and a computer 130.

  The SCS 110 is a manufacturing condition management system, and is a master system that normalizes manufacturing conditions as electronic information for each product and manages the version number. Here, the SCS 110 is configured separately from the computer 130, but the SCS 110 may be built in the computer 130. The SCS 110 relates to various processing conditions such as manufacturing conditions for each manufacturing process, such as a material, a film thickness, a film forming temperature and a film forming atmosphere in a thin film forming process, and a material, a film thickness and a line width in a wiring manufacturing process. Data is managed. Various processing conditions managed by the SCS 110 are transmitted to the online information interface 120. Then, various processing conditions are transmitted to the MES 24 via the online information interface 120, and further transmitted to the computers 21a to 21c of the respective manufacturing facilities 20a to 20c and the servers 31 and 41 of the monitor facility 30 and the test facility 40. .

  The online information interface 120 exchanges various information. For example, the online information interface 120 plays a role of transmitting data on processing conditions transmitted from the SCS 110 to the MES 24, and transmitting, from the MES 24, performance information and processing history information on each of the manufacturing facilities 20 a to 20 c to the computer 130. The online information interface 120 also plays a role of acquiring the respective measurement results from the servers 31 and 41 of the monitor equipment 30 and the test equipment 40 and transmitting the measurement results to the computer 130.

  The computer 130 constitutes a control unit for executing various processes and accumulating data. The computer 130 transmits performance information and process history information transmitted from each of the manufacturing facilities 20a to 20c, and measurement results obtained by the monitor facility 30 and the test facility 40. It is stored and stored as data in a database. Then, based on the data stored in the database and the data obtained from time to time, the computer 130 determines a non-defective product rate that changes due to an abnormality in the manufacturing equipment 20a to 20c, the monitor equipment 30, and the test equipment 40 in the manufacturing process of the semiconductor device. Automatically digitize and analyze the image of the defect distribution in the wafer. As a result, the computer 130 automatically analyzes the causal relationship between the abnormalities of the manufacturing facilities 20a to 20c, the monitor facility 30, and the test facility 40 and the failure distribution.

  Specifically, when each process executed by the computer 130 is represented as a control block, the configuration shown in FIG. 2 is obtained.

  As shown in this figure, the computer 130 includes an information acquisition numerical unit 131, an information storage unit 132, a monitoring unit 133, a design know-how information storage unit 134, a causal relationship analysis unit 135, a causal relationship information storage unit 136, and a monitoring notification. A unit 137 and a feedback unit 138 are provided.

  The information obtaining and digitizing unit 131 obtains various information of each processing facility in the FA that executes the manufacturing process through the online information interface 120, and temporarily stores the information in a dedicated database (not shown). The information acquisition digitizing unit 131 digitizes the distribution image based on the acquired various information. More specifically, the information obtaining and digitizing unit 131 includes first to sixth processing units 131a to 131f. FIG. 13 shows a specific procedure of the process executed by the sixth processing unit 131f.

  The first processing unit 131a manufactures various information managed by the SCS 110, such as specification information, processing recipes, and condition change information for each type (ie, product) and process to be processed by the manufacturing equipment 20a to 20c. Get information about conditions. The second processing unit 131b obtains, for example, performance information for each type and lot to be processed in the manufacturing facilities 20a to 20c. The third processing unit 131c obtains a processing history for each type and lot to be processed in the manufacturing facilities 20a to 20c, for example, information on processing date and time, a result of workmanship, and the like. The fourth processing unit 131d obtains, as various kinds of information, measurement results for each category in the monitor equipment 30, that is, for each monitor item. The fifth processing unit 131e obtains the measurement results for each category of the non-defective rate in the test equipment 40, that is, for each test item. The sixth processing unit 131f obtains the determination result of good / bad in the test facility 40, that is, the measurement result of the good product ratio. Further, the fourth to sixth processing units 131d to 131f automatically quantify an image of a non-defective product rate and a defect distribution in a wafer, which change due to an abnormality of a manufacturing facility in a manufacturing process, based on various kinds of input information. I do.

  Here, various types of information are temporarily stored in a dedicated database, but if all the information is stored as it is, the data amount becomes enormous. Further, the tendency of the abnormality cannot be continuously monitored. In order to reduce the data amount and to continuously monitor the tendency of abnormalities, it is considered effective to replace the distribution image in the wafer surface with numerical values. For example, by continuously monitoring the numerical value even if there is no abnormality, it becomes possible to monitor a change point that is different from usual. Assuming that all the manufacturing facilities 20a to 20c reproduce the same result together, the distribution should always be constant, and the numerical value should be the same value. However, when the numerical value starts to change for some reason, it is possible to recognize the change simply by monitoring the numerical value, so that it is possible to prevent the failure before it becomes a major abnormality by investigating the cause. Become. Therefore, the information obtaining and digitizing unit 131 executes the numerical processing for performing the above-described numerical conversion based on the obtained various information. Then, the numerical value obtained by the digitization processing is transmitted to the information storage unit 132. It should be noted that the details of the numerical processing here will be described later.

  The information storage unit 132 receives the information quantified by the information acquisition and quantification unit 131 and normalizes the quantified information, that is, divides the information into the same keyword (type name, process name, equipment name, processing condition, etc.). Accumulate. Specifically, the first accumulation unit 132a accumulates information based on, for example, information such as a product name, a process name, an equipment name, and processing conditions obtained from the manufacturing equipments 20a to 20c. The second to fourth storage units 132b to 132d accumulate common information from the measurement results for each category in the monitor equipment 30, the measurement results for each category in the test equipment 40, and the chip coordinate information of the non-defective product result. To go. That is, when each process is executed by the information obtaining and digitizing unit 131, the same type is separated and accumulated from the result after each process.

  The monitoring unit 133 monitors a change “unusual” based on the numerical value of the information quantified by the information acquisition quantifying unit 131. Specifically, the monitoring unit 133 monitors a change in the characteristic of the non-defective product ratio due to the same product, for example, a change in the non-defective product ratio, a change in the position of the distribution of defective chips, and a change in the intensity of the defective distribution. The strength of the failure distribution is an index indicating the failure rate or the number of defective chips per unit, and becomes stronger as the failure rate or the number of defective chips per unit is higher. Details of the strength of the failure distribution will be described later. The monitoring unit 133 also monitors a change in the performance tendency for each manufacturing process, in other words, for each manufacturing facility. For example, for all manufacturing processes, processing recipes and workmanship specifications are normalized and monitored on the same scale. Further, at a semiconductor manufacturing site, there are a plurality of facilities that can be used in the same process, and the same result is expected. However, it is known that there is an individual difference of facilities (generally, a difference between units). Specifically, the timing of cleaning differs for each unit due to changes in the effect on the processing results within the cleaning cycle. The most important thing is that the number of units that become abnormal in the same equipment is limited, and it is possible to prevent defects by determining them early. It is known that many of them appear in the distribution of the processing results, and it is an effective means if the distribution change of each facility can be numerically monitored with respect to a certain standard. Further, when the manufacturing conditions are changed, the monitoring unit 133 monitors the manufacturing process and the manufacturing equipment before and after the change. The monitoring unit 133 has a function of performing the above-described defect determination using these as monitoring criteria and notifying a person in charge of the determination result.

  The design know-how information accumulating section 134 stores information on the causal relationship between the categories of the manufacturing equipment 20a to 20c, the monitor equipment 30, and the test equipment 40 and the manufacturing process that the designers and experienced manufactures of the semiconductor products have as knowledge (hereinafter, referred to as “manufacturing processes”). (Called causal relationship information). Specifically, when a user such as a designer or an experienced person inputs the causal relationship information through an input device (not shown) provided in the computer 130, for example, a keyboard, the causal relationship information is stored in the design know-how information storage unit 134. Can be accumulated.

  The causal relationship analysis unit 135 can narrow down the causal relationship by searching for the same keyword from the quantified information group stored in the information storage unit 132 and comparing it with the time axis. More specifically, a causal relationship analysis process is performed by comparing the change in the distribution of wafers having a reduced defect rate, the process record, and the workmanship distribution of the manufacturing process registered in advance in the causal relationship information.

  The causal relationship information accumulation unit 136 is a unit that accumulates the causal relationship between the abnormality of the manufacturing equipment analyzed by the causal relationship analyzing unit 135 and the failure distribution. In the case of the present embodiment, the causal relationship information storage unit 136 corrects the causal relationship between the abnormality of the manufacturing equipment and the failure distribution in consideration of feedback information transmitted from the feedback unit 138 described later, and stores the corrected causal relationship. Is responsible for improving the accuracy of the causal relationship information.

  The monitoring notification unit 137 notifies the user of the result of the causal relationship analysis performed by the causal relationship analyzing unit 135 when the monitoring unit 133 determines that there is an abnormality, for example, when there is a characteristic change in the non-defective rate due to the same product. Perform relationship analysis result notification processing. For example, by displaying the analysis result on a display (not shown) provided in the computer 130, the user can be notified of the result.

  The feedback section 138 is a section where, when the user has taken action based on the result of the notification by the monitoring notification section 137, the user inputs the result of the action as feedback information. The feedback information can be input through an input device (not shown) provided in the computer 130, for example, a keyboard.

  As described above, the abnormality analysis device 100 according to the present embodiment is configured. Subsequently, the numerical analysis process performed by the abnormality analysis apparatus 100, that is, the process of automatically digitizing an image of a non-defective product ratio and a defect distribution in a wafer that change due to an abnormality of a manufacturing facility in a semiconductor product manufacturing process, includes In addition to the description of the abnormality analysis method, a typical procedure is shown in FIG.

  First, the concept of the digitization and how to realize the digitization will be described.

  A wafer to be quantified with respect to the defective distribution of the non-defective product rate or the defective rate is generally in a round shape. Chips constituting hundreds of semiconductor products arranged on the wafer are two-dimensionally arranged on the XY plane when the wafer surface is regarded as an XY plane.

  Many of the manufacturing equipments 20a to 20c, which influence the defective distribution of the non-defective rate or the defective rate, perform processing in a manufacturing process by mounting a wafer on a stage. For this reason, the in-plane distribution of the processing result in each of the manufacturing facilities 20a to 20c is managed. It is desirable that the in-plane distribution be uniform and reproducible. That is, assuming that all of the manufacturing facilities 20a to 20c that have performed the manufacturing process reproduce the same result together, it is desirable that the in-plane distribution is always constant. Regarding the in-plane distribution, there is no micro-distribution in the wafer temperature distribution, gas supply distribution, and the like, and a macro-distribution is generally used.

  With regard to such an in-plane distribution, a numerical value that reproduces a distribution image that a person feels by visually observing a wafer was aimed. The non-defective rate or the defective rate is obtained by calculating the number of non-defective chips or the number of defective chips with respect to the total number of chips, and has a known value by confirming the final product. Therefore, a known value is used for the non-defective rate or the defective rate.

  However, the non-defective product rate or the defective rate and the defect distribution on the wafer surface are not always the same even if the defect rate is the same. In order to make a numerical value, it is important that it is clear which position of the wafer such as a defective chip is to be identified on the wafer. For this reason, the "position" that can be recognized by a person is divided on the wafer, and, for example, concentrically or horizontally and vertically as described later, and further divided into blocks (the outermost periphery, the center, the center, the upper, the middle, (Lower, left, middle, right). Since the area of each part divided into 17 positions on the wafer is not the same, the numerical value is calculated after mathematically leveling in consideration of the area correction.

  In addition, the 17 positions are further grouped as blocks (the outermost periphery, the center, the center, and the upper, the middle, the lower, the left, the middle, the right, and the like) felt by a person. As a block felt by a person on the wafer, for example, an image of a block concentrically divided (hereinafter, referred to as a circle image) and an image of a block divided horizontally in one direction on the surface of the wafer (hereinafter, horizontal). Image) and an image of a block divided in a vertical direction orthogonal to one direction (hereinafter, referred to as a vertical image). In such a plurality of different blocks, 17 positions are respectively grouped. Of course, each position can be grouped as a block other than the blocks mentioned here.

  Further, according to human eyes, the strength of the defect distribution can be grasped as a sensation even at the same position. In other words, intensity can also be expressed as density, meaning the number of chips to be identified per unit. This can also be expressed as a non-defective rate per position per unit, but the resolution is only 100, and variations in the range of positions cannot be expressed.

  Therefore, the variation is obtained as a standard deviation, and the area ratio in the positional relationship is also taken into consideration so as to match the non-defective product rate. Specifically, an equation for matching is set. The value obtained by this equation can be provided as a numerical value having a resolution close to an image seen by a human, such as the defect rate, the position and strength of the distribution, and the like. More specifically, the distribution at 17 positions is obtained by the standard deviation. Further, the strength is obtained by multiplying the standard deviation of the area-corrected state by the non-defective rate. This equation realizes a numerical value having the same slope as the yield rate.

  The digitization of the distribution image is performed so that the computer can automatically recognize the distribution image of the defective chip on the wafer without directly observing the image. To do this, when digitizing the distribution image, it is necessary to digitize the "position" of the distribution image and to digitize the "strength" of the distribution image. For example, for convenience, when the user manually inputs the number of defective chips at each position, the strength of the distribution image is automatically quantified.

  Then, a distribution image of defective chips on the wafer is represented by using, for example, the “position” and “strength” quantified as described later, in addition to the known “defective rate”. For example, a composite value of “position” and “strength” of a numerically-distributed distribution as described below is represented by a numerical value of 4 to 5 digits. By managing the failure distribution quantified in this way, it is possible to automatically analyze the causal relationship between the failure distribution in the wafer and the abnormality of the manufacturing equipment. Hereinafter, a specific description will be given of how the distribution image is digitized.

[Numericalization of distribution image position]
Regarding the “position” of the distribution image, the surface of the wafer is regarded as an XY plane, and can be regarded as XY two-dimensional. Chips constituting a semiconductor product are two-dimensionally arranged on a wafer, and pass / fail judgment is performed for each chip of each semiconductor product. The coordinates of the defective chip exist as a distribution in the wafer. The distribution is for each wafer. The distribution for each wafer can be seen by human eyes as “one image” and can also be expressed as “number of defective chips”.

  However, even for a wafer having the same number of defective chips of 100 chips, there are various ways from the viewpoint of the "position" of the distribution image on the wafer. By setting the XY coordinate axes on the wafer plane and representing all the defective chip coordinate groups indicating the XY coordinates of all the defective chips, the "position" in the distribution image can be obtained. However, the “position” of the distribution image recognized by the human eye is “bundled”, which is a rough “around”, and does not recognize the XY coordinates of each defective chip. For this reason, the "position" of the distribution image recognized by the human eye is digitized so that it can be recognized as "grouped", in other words, "around here". That is, the “position” of the distribution image felt by a person on the wafer is quantified.

  Specifically, first, "which part" of the wafer formed in a circular shape is defined based on the sense of position of a person. The wafer has a circular shape that is known in advance, but there are different wafers such as a 6-inch wafer, an 8-inch wafer, and a 10-inch wafer. However, regardless of the size of the wafer, a person has a certain sense of "what part" of the wafer, and therefore defines it.

  In addition, the resolution of the distribution should be set to a resolution suitable for the usage phase. Here, the resolution is set so as to indicate the distribution affected by the failure of the manufacturing facilities 20a to 20c.

(1) Resolution of Position of Distribution Image As shown in FIG. 3A, a wafer is divided into three concentric circles as a circular image. That is, the resolution is divided into three parts, the outermost circumference, the middle, and the center. The outermost circumference is defined as x, the middle is defined as y, and the center is defined as z.

  Similarly, as shown in FIG. 3B, the wafer is divided into three along the horizontal direction as a horizontal image. That is, upper, middle, and lower resolutions are defined as three divisions, where x is defined as upper, y is defined as middle, and z is defined as lower.

  Further, as shown in FIG. 3C, the wafer is divided into three along the vertical direction as a vertical image. That is, the resolution is divided into three divisions of left, middle, and right, and left is defined as x, middle is defined as y, and right is defined as z.

  In this specification, a circle image is represented by ci which means circle, a horizontal image is represented by ro which means row, and a vertical image is represented by co which means column.

  As described above, in three types of the circle image, the horizontal image, and the vertical image, the position of the distribution image of the wafer is defined by three elements each. Since each of the three types has a resolution of three points, there are 27 possible combinations, but there are 17 actual combinations of positions as described later. Therefore, the position of the distribution image is divided into 17 positions, and the resolution can be defined as “17”.

  Here, the case where the distribution image is divided into 17 positions has been described, but the distribution image can be appropriately changed according to the purpose of use. For example, a circle image, a horizontal image, and a vertical image are each divided into three, but the number of divisions may be smaller or larger, and the number of positions of the distribution image, that is, the resolution, must be defined according to the division number. Can be.

(2) Details of the position of the distribution image When the distribution image is divided into the circle image, the horizontal image, and the vertical image as described above, the respective positions are defined as the positions shown in FIGS. 4A to 4C. In the figure, the position of the distribution image is shown by a matrix of ci, ro, and co. That is, for each of the ci, ro, and co, which of the three divided x, y, and z correspond to is indicated by indicating a “*” mark. 4A to 4C are not cross-sectional views, but the corresponding positions are indicated by hatching to make them easier to see.

  As described above, the positions of the distribution image on the wafer are defined in order, such as the uppermost outermost position, the uppermost outermost position, the uppermost outermost position,..., And finally the center position is defined. This defines 17 positions. Regarding these 17 positions, as shown in FIGS. 1 to No. 17 is assigned an identification number.

  As described above, the position of the distribution image can be defined by 17 positions listed in FIG. 5, that is, 17 patterns. When all of these 17 patterns are combined, as shown in FIG. 6, the entire surface of the wafer is covered. Although FIGS. 5 and 6 are not cross-sectional views, the corresponding positions are indicated by hatching to make them easy to see.

[Numericalization of distribution image strength]
The “strength” of the distribution image indicates the number of defective chips, that is, the number of defective chips, and if the “position” can be regarded as two-dimensional XY, the “strength” is the Z-axis. Is equivalent to

  As described above, the chips constituting the semiconductor product are two-dimensionally arranged on the wafer, and pass / fail judgment is performed for each chip of each semiconductor product. The coordinates of the defective chip exist as a distribution in the wafer. The coordinates of the defective chip exist as a distribution in the wafer. The distribution is for each wafer. The distribution for each wafer can be seen by human eyes as “one image” and can also be expressed as “number of defective chips”.

  However, even for a wafer having the same number of defective chips of 100, there are various ways from the viewpoint of the "strength" of the distribution image on the wafer. However, the “strength” of the distribution image is unique by “density of defective chip at position”. Therefore, here, the "strength" of the distribution image perceived by a human is replaced with a number representing "the density of defective chips whose position is the denominator". That is, the “strength” of the distribution image felt by a person on the wafer is quantified.

  Specifically, the area at each position, that is, the area correction value corresponding to the number of chips is obtained by calculation, and the area correction is performed on the known position information. Then, the coordinates of the defective chip on the wafer are matched with the position information, the number of defective chips is calculated, and correction is performed using the area correction value. The sum of the number of defective chips at each position after such area correction is equivalent to the density of defective chips on the wafer. Therefore, the position having the maximum value is determined. This makes it possible to determine a place where the number of defective chips after area correction is the largest, that is, a place where the “strength” of the distribution image is strong.

  As shown in FIG. 5, the area of each position on the wafer defined by the 17 patterns is different. For this reason, the number of chips included in each position is also different, and it is not possible to simply define the “strength” of the distribution image using the absolute value of the number of defective chips as the density. That is, since the density of a defective chip that a person sees and feels is different from the density indicated by the absolute value of the number of defective chips calculated, it is necessary to correct the density indicated by the absolute value of the number of defective chips.

  Then, people finally feel the image of one strength with one wafer. It is necessary to reproduce this logically. This is done by considering the “high-ranking position” and the “whole variation” separately.

(1) Method of Area Correction In order to determine the density of defective chips at each position and rank the strongest positions, it is necessary to correct each position having various areas to the same area. Therefore, here, the density of the defective chip is determined after performing the area correction based on the number of defective chips at positions having different areas on the wafer.

  For example, as shown in FIG. 7A, on the wafer, the position is divided into three types: left, middle, and right. The left area is 9, the middle area is 4, and the right area is 1. Suppose there was. In this case, as shown in the figure, assuming that the number of defective chips is 1, when viewed by a person, the density of the defective chips is felt to be in a magnitude relationship of left <middle <right. That is, the density of the defective chip is interpreted as 1/9, 1/4, 1/1 in the head for each of the left, middle, and right.

  In this case, since the physical position information cannot be corrected, the number of defective chips is corrected. At each position shown in FIG. 7A, the number of defective chips is actually one at all positions. However, as shown in FIG. In the position, one is corrected to 2.25, in the middle position one remains one, and in the right position one is corrected to 0.25. This is an example in which correction is performed so as to be an average value. Since the result is obtained by mathematically correcting the actual number of defective chips, a decimal point may be added. By doing so, correction is performed so that the entire area is represented by the same density at all positions.

(2) How to Obtain Correction Value As described above, 17 positions are defined for the “position” of the distribution image. The area of this position can be obtained by calculation on a desk. In this method, an X-Y array covering the wafer is created with a chip size that is the smallest in operation, and then a difference in the number of chips positioned at a position is replaced with a difference in area to determine a correction coefficient. In this case, it is desirable to calculate the correction coefficient so as to match the average area. This correction coefficient is a coefficient that can be commonly used for the same inch wafer regardless of the chip size. For example, when a wafer having a total number of effective chips of 968 as shown in FIG. 8 is taken as an example, for example, the correction coefficient is set as shown in the table of FIG.

  FIG. 9 shows a case where the average number of chips at 17 positions is normalized. As shown in FIG. 9, the positions 1, 3, 6, and 8 have 60 chips, the positions 2, 4, 5, and 7 have 70 chips, the positions 9, 11, 14, and 16 have 10 chips, and 10, 12, 13, and 15 have 66 chips, and position 17 has 144 chips. As for the average value, since the total number of effective chips is 968 and 17 positions, the average number of chips per position is 56.9. Therefore, the correction coefficient at each position is (average number of chips / number of chips at each position). For example, since the number of chips at position 1 is 60, the average number of chips / the number of chips at position 1 is 56.9 / 60, and the correction coefficient is 0.95. Therefore, the value obtained by multiplying the number of chips at each position by the correction coefficient is uniformly 56.94, indicating that the number of chips can be normalized for all positions.

  In this way, a correction coefficient that serves as an area correction value can be set, and the number of chips at each position can be normalized by multiplying the number of chips at each position by the correction coefficient.

(3) Correction by Correction Value and Rank of Strong Position Subsequently, the coordinates of defective chips on the wafer are matched with the position information, the number of defective chips is calculated, and correction is performed using the area correction value.

  First, with respect to a wafer to be used in advance, an origin that matches information measured by the inspection apparatus in the inspection process, for example, an upper left point is determined as the origin, and the surface of the wafer is regarded as an XY plane, and defined by the “position” of the distribution image. The maximum value of the XY coordinates is obtained from the XY coordinate information indicating each position. Thereby, the chip arrangement in each of the X direction and the Y direction on the XY plane can be determined, and the number of chips at each position and the total number of chips on the wafer can be determined.

  For example, the XY coordinates are determined by dividing XY coordinate information indicating each position defined by the “position” of the distribution image, that is, ci, ro, and co into x, y, and z, respectively. For example, if the maximum value of the X coordinate and the Y coordinate in the XY coordinate is 36, x, y, z of the horizontal image and x, y, z of the vertical image are allocated to each of 12 chips obtained by dividing the X and Y coordinates into three. become.

  Finally, the total number of defective chips at each position after the alignment is obtained. Then, the number of defective chips at the analysis position is corrected by multiplying the correction coefficient obtained as described above. The area correction is performed in this manner, and the area becomes equivalent at all positions. Therefore, the number of defective chips corrected at each position becomes equivalent to the density at each position.

  Then, using the corrected number of defective chips, the total sum of the number of defective chips for each position is calculated, and the position of the largest numerical value is ranked.

(4) Method of expressing strength The degree of significance of analysis is determined by the strength of the failure distribution felt by humans. Specifically, a person often feels strength in three degrees, "strong, normal, and weak."

  Considering the degree required for the analysis, the best situation is that the strength of the defect distribution is “uniform”, and when the strength deviates from the uniformity, it is determined by the computer 130 where “how much” and “how much” deviated. It is preferable to be able to automatically determine at.

  In view of these, the strength of the failure distribution is expressed by dividing it into a plurality of degrees, and the degree of the strength is given a meaning so that the distribution can be automatically recognized. Here, the degree of the strength is divided based on the standard deviation, and the degree of the divided strength is “three types” of “0”, “5”, and “9”. It is determined that “0 is uniform”, “5 is sparse”, and “9 is significant”, respectively. What is left to the standard deviation here is the variation of the unit divided into blocks of x, y, z for each of the circle image, the horizontal image, and the vertical image. That is, the population is a circle, a horizontal, and a vertical, respectively, and by determining the standard deviation of the values identified as x, y, and z, the bias of x, y, and z of each image can be quantified. Become. At this stage, it merely shows how the distribution seen by a person appears in a circle, a horizontal direction, and a vertical direction. In other words, this is the strength of the distribution position, and care must be taken so as not to be confused with the overall strength of the entire wafer described later. Note that the degree of the strength of the failure distribution shown here is an example, and the numerical value and its meaning may be determined according to the purpose of operation.

  When the above-described area correction is performed, the strength is simply determined by the total number of defective chips after the correction, because the strength at the position is only determined here. In consideration of the strength at all positions, the degree of density that images the entire wafer is obtained. That is, it is possible to express human sensitivity.

  Up to this point, since the positions of the distribution image are divided into three types (ci, ro, and co), each is corrected independently. Here, the sum of the three types (ci, ro, co) of strength is referred to as “the strength of the entire wafer”.

(5) Summation Method of Strength The degree of strength “0”, “5”, “9” divided as described above indicates that the larger the value, the stronger the strength. Therefore, the summation is performed simply by calculating the sum of the strengths of the three types (ci, ro, and co). Three (ci, ro, co) strength degree of intensity each of "0", since the "5", there three "9", as shown in FIG. 10, as the combination, ^{3 3} 27 types. However, since the sum value may be the same even in different combinations, there are ten types of sums of 0, 5, 9, 10, 14, 15, 18, 19, 23, and 27. .

  The sum referred to here is the final strength and is a continuous value. For example, in FIG. 10, when the sum = 0, the initial value of the continuous value is 10, and the values of the continuous values 10 to 19 are added up to the sum 27. Here, the initial value of the continuous value is 10 and the continuous values 10 to 19 are given. However, this value is arbitrary and the initial value does not necessarily have to be 10.

[Composition of distribution image position and intensity]
As described above, the position and intensity of the distribution image can be obtained. The position and intensity of the distribution image obtained in this way are combined and digitized. 11A and 11B show, as a matrix, all patterns obtained when the position and intensity of the distribution image are combined. In this drawing, the shaded image is shown by hatching, but the intensity is divided into shades at 17 positions and arranged in a matrix. No. 2-No. Regarding No. 15, two positions are represented in the same matrix. However, these are positions that are symmetrical with respect to a diagonal line passing from the upper left corner to the lower right corner. Because they are equal, they are shown as common figures. The pattern shown here is the resolution at which the “position” and “strength” of the distribution image can be expressed by the anomaly analyzer.

(1) Numericalization of the position of the distribution image For all the combinations in the matrix shown in FIGS. 11A and 11B, the position and the intensity of the distribution image are quantified according to a predetermined rule. Here, the predetermined rule is defined as follows.

  First, the position of the distribution image, specifically, 17 positions is set as “classification”, and an arbitrary digit is assigned as a digit representing each classification. Here, as an example, digits of 0100000 to 1700000 are assigned to “classification”.

  Next, an arbitrary digit is assigned as a digit representing each of ci, ro, and co indicating the “position” of the distribution image. Here, a digit different from the “classification” is assigned as a digit representing each of ci, ro, and co. Ci is assigned a digit of 10,000, ro is assigned a digit of 1,000, and co is assigned a digit of 100.

  Subsequently, a value is assigned to each of x, y, and z, which are divided into three, for each of ci, ro, and co. Here, x is 1, y is 2, and z is 3. The number of digits is changed depending on whether x, y, or z is ci, ro, or co. In other words, x, y, and z are represented by changing the number of digits for each of ci, ro, and co. For example, ci is a digit of 10,000, x in ci is 10,000, and y in ci is 20,000. Also, ro is set to the digit of 1000, and x in ro is set to 1000. Also, co is a digit of 100, and x in co is 100.

(2) Composition of “Position” and “Strength” As described above, the “strength” of the distribution image is a value obtained by quantifying the “strength” in the final strength.

  Then, the quantified values of the position of the distribution image, that is, “classification”, “ci, ro, co”, and “x, y, z” are assigned to specific digits, and “strength” is further quantified. The sum is obtained by adding the values. For example, no. In the position of 2, when ci is x, ro is x, co is y, and the final strength is 17, the digit of 0100000 is No. 0200000 representing 2; 10,000 representing ci represents 10000 representing x; ro representing 1000 represents x representing 1000; and 100 representing co represents 200 representing y. Then, a value obtained by adding 17 of the final strength representing “strength” to these, that is, 0200000 + 10000 + 1000 + 200 + 17 = 0212112, is a value obtained by combining the “position” and “strength” of the distribution image and quantifying it.

  By allocating to a specific digit in this way, a numerical value of "classification", "position" and "strength" can be obtained. The resulting tables are FIGS. 12A to 12D. As shown in these figures, the sum results all have different values.

  In this way, by continuing to monitor at least one of the summation result and the serial number, it becomes possible to monitor the tendency of an abnormality such as the following (a) to (d).

  (A) When the distribution changes, the unit of ** 00000 digits changes. (B) Even if the distribution is the same, if the intensity changes, the unit of 10 digits changes. (C) It becomes easy to notice abnormalities from the serial number plots. (D) By mutually monitoring different types of graphs, it becomes easy to notice common abnormalities.

  Regarding (a), as described above, since the ** 00000 digit indicates the "position" of the distribution image, it can be seen that the position of the failure distribution has changed by changing the ** 00000 digit. Similarly, as for (b), as described above, the final strength is added as a numerical value as it is and is expressed in units of 10 digits. Therefore, when the intensity changes in the same distribution, 10 digits are used. It will change in units. As a result, it is possible to continuously monitor changes in the “position” and “strength” of the distribution for one product.

  On the other hand, since the same manufacturing apparatus is responsible for processing various products and different processes, it is also used to monitor the tendency of abnormalities when a plurality of products are processed by the manufacturing apparatus. That is, as for (c), the same result should be obtained if the same processing is performed when a plurality of the same products are manufactured. In the case of (d), when different products are manufactured, the distribution of each product is monitored, and when a result indicating a similar change in serial number is obtained, it is noticed that there is a common abnormality tendency. be able to.

  As described above, by automatically quantifying the “position” and “strength” of the distribution image, it is possible to monitor the tendency of abnormality by simply monitoring the numerical values. Here, in addition to defining “classification” at 17 positions, it is also possible to sense how a person, such as a circle image, a horizontal image, and a vertical image, becomes stronger when looking at the distribution of wafers. It is divided into easy blocks. Then, since "strength" is quantified for each of x, y, and z obtained by dividing each block into three parts, and the final strength is obtained by the total value, the "what" is felt when a person views the distribution of wafers. With such a feeling, it is possible to express whether the defect distribution is strengthened.

  In addition, a standard deviation is used to quantify that “strength” is “uniform”, “sparse”, or “significant”. The number of chips manufactured for a wafer can be changed according to the product, the wafer inch, and the like, and is not fixed. Therefore, “strength” cannot be simply expressed by the number of defective chips. On the other hand, by using the standard deviation, parameters within the wafer surface can be determined with the standard deviation. For this reason, the “strength” is determined based on the variation of the standard deviation, and when the “strength” is quantified, a function that allows a user to input a coefficient as a threshold value as a variable from outside is provided so that an appropriate coefficient can be obtained. Can be set.

  In the case of the present embodiment, a first threshold (for example, 13) and a second threshold (for example, 30) are used as coefficients for determining whether “strength” is “uniform”, “sparse”, or “significant”. You have set. If the standard deviation is less than the first threshold value, it is “uniform”; if it is equal to or more than the first threshold value and less than the second threshold value, it is “sparse”; At this time, the user can appropriately set the threshold value serving as the coefficient. Further, by using the standard deviation, it is possible to set appropriate values for the number of chips in the plane, the distribution thereof, and the like. That is, the number of chips changes according to the size of the chip even if the area is the same. For example, if the size per chip is reduced, the number of chips is increased even if the area is the same. Therefore, the number of chips included in each position changes according to the size of the chips. However, the failure rate, that is, the number of defective chips with respect to the total number of chips is represented by the variation of the standard deviation based on three parameters of x, y, and z, and when it is represented in a graph, the slope is obtained. Therefore, if the threshold value is determined based on the variation of the standard deviation of the number of defective chips in the divided positions, even if the number of chips in the plane changes, the “strength” can be determined appropriately in accordance with the change. it can.

  Further, in this way, by digitizing and combining the “position” and “strength” of the distribution image, the computer 130 can grasp the distribution image based on the numerical values. By being able to grasp the “position” and “strength” of the distribution image, the computer 130 can use the computer 130 to display the “state” of the distribution image as to how the failure distribution looks when a person looks at the wafer. It is also possible to grasp.

  As described above, the distribution image is digitized. That is, the wafer is divided into a plurality of positions in the plane, the plurality of positions are quantified, and the variation in strength at each of the plurality of positions is determined as a standard deviation, and the strength is quantified based on the standard deviation. ing. Such digitization is performed for each main manufacturing facility in the semiconductor device manufacturing process, for example, for the manufacturing facilities 20a to 20c, the monitor facility 30, and the test facility 40. Thereby, it becomes possible to analyze the abnormality of each of the manufacturing facilities 20a to 20c, the monitor facility 30, and the test facility 40 and to analyze the causal relationship between the abnormality and the failure distribution.

  For example, data is stored one by one for each manufacturing facility, and the data is replaced with numerical data according to the same algorithm and stored in the information storage unit 132. Then, after the final process is completed, based on the non-defective product rate and the defective distribution obtained as a result of the test process by the test equipment 40, the accumulated data of the same lot is reversely searched, and the place where the similar distribution is present is determined. Investigation makes it possible to analyze the causal relationship. By monitoring the distribution of defects, it is possible to anticipate the point of change and to deal with the cause of the quality deterioration at an early stage.

  Further, as described above, since the data is digitized and stored for each manufacturing facility, the data amount can be reduced, and the tendency of the abnormality can be reduced by the computer 130. It is possible to continuously monitor.

  Therefore, the causal relationship between the abnormalities of the manufacturing facilities 20a to 20c, the monitor facility 30, and the test facility 40 and the failure distribution can be automatically analyzed without accumulating a huge amount of data. This makes it possible to provide an abnormality analysis device.

(Other embodiments)
The present invention is not limited to the embodiments described above, and can be appropriately modified within the scope described in the claims.

  For example, in the first embodiment, when a distribution image of a wafer used for manufacturing a semiconductor device is grouped as a block felt by a person, the distribution image is grasped as a circle image, a horizontal image, and a vertical image. However, this is only an example of the case of grouping as a block, and all of these images may not be used, or a distribution image may be added by adding other images. That is, the entire surface of the wafer is classified into a plurality of positions (17 positions in the first embodiment), and a plurality of blocks (circles ci, horizontal ro, vertical co in the first embodiment) are divided according to human senses. And the failure distribution is analyzed for each block. Then, when the analysis results of the respective blocks are superimposed, it is possible to determine which block has the most scattered defect and has occurred. In order to superimpose the analysis results in this manner, the entire surface of the wafer may be divided into a plurality of types of blocks, and the number of types is arbitrary.

  Further, in the first embodiment, an example in which the “position” and “strength” of the distribution image are quantified has been described, but the quantification method may be another method. For example, in the first embodiment, the entire surface of the wafer is classified into 17 positions by dividing three types of blocks (circle ci, horizontal ro, vertical co) into three. The positions obtained by dividing the three types of blocks into three are represented by x, y, and z, respectively, and when digitizing, the digit representing the classification and the digit representing the block are set to different digits. Assign different values. As described above, if the classification (that is, the position) and the block are set to different digits, and the value assigned to each position when the block is divided is set to a different value, the “position” of the distribution image is obtained regardless of the number of digits or the numerical value. ”And“ strength ”. Therefore, the same effect can be obtained even if a different digit or a different value from the description of the first embodiment is used.

  Further, in the first embodiment, in addition to a part for quantifying a distribution image and automatically analyzing and monitoring an abnormality, a case where a causal relationship between an abnormality such as a manufacturing facility and a failure distribution is also analyzed. This is described using an example. However, at least, the distribution image is digitized, and based on the result, it is only necessary to analyze the tendency of some abnormality, and it is not always possible to analyze the causal relationship between the abnormality of the manufacturing equipment and the failure distribution. May be.

  In the first embodiment, both the acquisition of information such as the processing result and the quantification of the information are performed by the information acquisition and quantification unit 131. However, these need not be performed in the same block, and are separately performed. It can be done in blocks.

20a-20c Manufacturing equipment 21a-21c Computer 30 Monitor equipment 40 Test equipment 100 Abnormality analyzer 120 Online information interface 130 Computer

Claims (10)

A processing result of at least one of a manufacturing facility (20a to 20c), a monitor facility (30), and a test facility (40) used when manufacturing semiconductor device chips using a wafer made of a semiconductor material. An abnormality analysis device for analyzing an abnormality of the processing equipment based on the
A management system (110) for managing information relating to processing conditions in the processing equipment;
An information obtaining and digitizing unit (131) for obtaining the processing conditions, obtaining a processing result of the processing equipment, and digitizing an image of a failure distribution for each wafer based on the processing result;
A monitoring unit (133) for monitoring an abnormality of the processing equipment based on the information quantified by the information acquisition quantification unit;
The information obtaining and digitizing unit divides the wafer into a plurality of positions in the plane, automatically quantifies the strength of the defect distribution at the mutual positions from the amount of defective chips at the plurality of positions, An abnormality analysis device that obtains a variation in the strength of the defect distribution for each of a plurality of positions as a standard deviation, and quantifies the strength based on the standard deviation. An information accumulation unit (132) for normalizing and accumulating the information digitized by the information acquisition digitization unit;
The abnormality analysis according to claim 1, further comprising: a causal relationship analysis unit (135) configured to analyze a causal relationship with the abnormality of the processing facility based on the digitized information stored in the information storage unit. apparatus.   A monitoring notifying unit (137) for notifying a user of information on the causal relationship analyzed by the causal relationship analyzing unit when the monitoring unit outputs a monitoring result indicating that the processing equipment is abnormal; The abnormality analysis device according to claim 2.   A causal relationship information storage unit (136) for storing information on the causal relationship analyzed by the causal relationship analyzing unit when the monitoring unit outputs a monitoring result indicating that there is an abnormality in the processing equipment. Item 3. The abnormality analysis device according to item 3.   After the monitoring result is issued by the monitoring notifying unit, the result of the action taken by the user based on the causal relationship is input as feedback information, and the feedback information is transmitted to the causal relationship information accumulating unit, and the causal relationship is transmitted. The abnormality analysis device according to claim 4, further comprising a feedback unit configured to correct the causal relationship stored in the information storage unit based on the feedback information.   The information obtaining and digitizing unit divides the plurality of positions into a plurality of blocks that are grouped differently, quantifies the strength for each of the plurality of blocks, and converts the plurality of blocks into blocks. The abnormality analysis apparatus according to any one of claims 1 to 5, wherein a final strength is obtained by summing the strengths quantified in each case.   As the plurality of blocks, a horizontal image block that divides the position into a plurality in the horizontal direction with one direction of the surface of the wafer being a horizontal direction, and a vertical direction orthogonal to the one direction on the surface of the wafer. The abnormality analysis apparatus according to claim 6, wherein a plurality of the vertical image blocks that divide the position into a plurality of parts and a plurality of circular image blocks that divide the wafer into a plurality of concentric circles are included.   In the information obtaining and digitizing unit, the digit representing the plurality of positions and the digit representing each of the plurality of blocks are respectively different digits, and the digitization of the plurality of positions and the digitization of the strength are performed. The abnormality analysis device according to claim 6.   The information obtaining and digitizing unit, by adding a numerical value representing the final strength in addition to a numerical value representing each of the plurality of positions and a numerical value representing each of the plurality of blocks as determined digits, the The abnormality analysis device according to claim 8, wherein the position is quantified and the strength is quantified. A processing result of at least one of processing equipment (20a to 20c), a monitor equipment (30), and a test equipment (40) used when manufacturing semiconductor device chips using a wafer made of a semiconductor material. An abnormality analysis method for analyzing an abnormality of the processing equipment based on,
Obtaining information on processing conditions in the processing equipment managed by the manufacturing condition management system (110), and obtaining processing results of the processing equipment;
Digitizing an image of a defect distribution for each wafer based on the processing result;
Monitoring the abnormality of the processing equipment based on the quantified information,
In the quantification, the wafer is divided into a plurality of positions in a plane, and the strength of the defect distribution at the mutual positions is automatically quantified from the amount of defective chips at the plurality of positions, An abnormality analysis method in which a variation in the strength of the defect distribution at each of a plurality of positions is obtained as a standard deviation, and the strength is quantified based on the standard deviation.

Images