JP2005056907A - Method for managing image data and image data service system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for managing image data and an image data service system in which a semiconductor device manufacturer can sustain reliability of a client by prompt action. <P>SOLUTION: The image data service system comprises a storage device section 42 for accumulating and storing the image of a defect or a candidate of defect being transmitted at least from the review unit and the CD-SEM of a production division 20 through a communication means 30 and the feature amount (defect, or the like) of that image while imparting at least a semiconductor device ID, and a search engine section 43 for searching the defect, or the like, of a semiconductor device relative to the semiconductor device ID based on the defect, or the like, corresponding at least to the semiconductor device ID being inquired through a communication means. The image data service system is further provided with a data base center 40 for providing data of the defect, or the like, corresponding to a relevant semiconductor device searched by the search engine to a semiconductor device manufacturer, a design house or its client through a communication means. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention provides a large number of image databases obtained from a review apparatus, a length measurement SEM, etc. for defects or defect candidates selected from defects or defect candidates detected by an inspection apparatus in various manufacturing processes when manufacturing semiconductor devices. The present invention relates to a method for managing image data stored over a semiconductor device for a long period of time (for example, 5 to 10 years) and a service system for the image data.
[0002]
[Prior art]
For example, as described in Japanese Patent Application Laid-Open No. 2001-156135 (Patent Document 1), in a semiconductor manufacturing process, a foreign object / appearance inspection apparatus can be used to detect foreign matter / A defect inspection is performed, and based on the inspection data, an enlarged image pickup apparatus having a function of automatically acquiring an enlarged image of a foreign object / defect, that is, an ADR (Auto Defect Review) has been developed, and a large number of images are obtained in a short time Has been developed, and a function of classifying the collected large quantity of images into categories according to the characteristics of foreign matter / defects, that is, ADC (Auto Detect Classification) has been developed, and the ADC can automatically classify defect images into categories. Are known. That is, Patent Document 1 describes ADC.
[0003]
Japanese Patent Laid-Open No. 2003-59813 (Patent Document 2) describes a length measuring SEM (Scanning Electron Microscope) for monitoring fluctuations in exposure conditions.
[0004]
Also, WO97 / 35337 (Patent Document 3) describes a process management for defining a defect by a combination of elements characterizing a defect based on inspection information and image information collected from an inspection apparatus via a communication network. The system is described.
[0005]
[Patent Document 1]
JP 2001-156135 A
[Patent Document 2]
JP 2003-59813 A
[Patent Document 3]
WO97 / 35337
[0006]
[Problems to be solved by the invention]
In semiconductor device manufacturers, in order to prevent defects in semiconductor devices, many inspection apparatuses, review apparatuses, analysis apparatuses, and the like have been introduced into semiconductor manufacturing process lines.
[0007]
For example, in the case of a vertically integrated device manufacturer, it has been difficult to construct a consistent image database corresponding to the model update of the inspection / measurement apparatus over a long period of time (for example, 5 to 10 years). In particular, the model of the inspection / measurement apparatus is updated as the equipment manufacturer changes specifications (high functionality) due to miniaturization and introduction of a new process, and the image resolution / data output format changes. In addition, it is too expensive to maintain the system and operating organization on the device manufacturer side while absorbing the device manufacturer-driven changes in the image database. Furthermore, when the same semiconductor device is manufactured at a plurality of bases, the difficulty is further increased.
[0008]
SUMMARY OF THE INVENTION An object of the present invention is to provide an image data management method and an image data service system that can solve the above-mentioned problems and can maintain the trust of a semiconductor device manufacturer customer by quick response.
[0009]
Another object of the present invention is to provide an image data management method and an image data service system capable of realizing a minimum compensation expense for a customer by narrowing down a defective lot or a defective process in a semiconductor device manufacturer. It is in.
[0010]
Still another object of the present invention is to provide an image data management method and image data management method that realizes consistent management of an image database over a long period of time for a large number of semiconductor devices having a large number of manufacturing process steps. To provide a service system.
[0011]
[Means for Solving the Problems]
In order to achieve the above object, the present invention corresponds to the first embodiment, and is a defect or defect candidate transmitted from at least a review apparatus and a CD-SEM of a semiconductor device manufacturing department via a communication means. And a storage device that stores and stores at least a semiconductor device ID by assigning at least a semiconductor device ID and an image of a defect or defect candidate image corresponding to at least the semiconductor device ID inquired via the communication means A search engine that searches for an image group of defects or defect candidates corresponding to the semiconductor device related to the semiconductor device ID based on the feature amount, and the defect corresponding to the related semiconductor device searched by the search engine or Defect candidate image group data via communication means to semiconductor device manufacturer or design house or A service system of the image data, characterized in that customers provided with a database center to perform the service offered.
[0012]
Further, the present invention corresponds to the first embodiment, and an image of a defect or a defect candidate transmitted from at least a review device and a CD-SEM of a semiconductor device manufacturing department via a communication means and the image of the image. An image of a defect or a defect candidate corresponding to the semiconductor device ID based on at least the semiconductor device ID inquired through the communication means, and a storage device unit that accumulates and stores the feature amount with at least a semiconductor device ID. And at least an image ID is assigned and a primary answer is made using a communication means, and an image corresponding to the image ID is obtained based on at least the image ID obtained using the communication means as a selection result of the primary answer. A search engine for searching for a defect or a defect candidate image group corresponding to a related semiconductor device based on a feature amount; The image of the defect or defect candidate corresponding to the semiconductor device ID to which the image ID is assigned by searching is provided to the semiconductor device manufacturer or design house or its customer via the communication means as the primary answer, and the search engine A database center is provided for providing a service by providing image data of a defect or defect candidate corresponding to a related semiconductor device searched to a semiconductor device manufacturer or a design house or its customer via communication means. This is an image data service system.
[0013]
Further, the present invention corresponds to the third embodiment, and the image ID of the defect or defect candidate transmitted from at least the review apparatus and the CD-SEM of the semiconductor device manufacturing department via the communication means and the image thereof. A storage unit that stores and stores at least a semiconductor device ID by assigning at least a semiconductor device ID and a feature amount of an image ID of a defect or defect candidate corresponding to at least the semiconductor device ID inquired through the communication unit And a search engine for searching for a defect or defect candidate image ID group corresponding to the semiconductor device related to the semiconductor device ID, and the defect or defect candidate corresponding to the related semiconductor device searched by the search engine The data of the image ID group of the semiconductor device manufacturer or the design house via the communication means A service system of the image data, characterized in that customers provided with a database center to perform the service offered.
[0014]
Further, the present invention corresponds to the third embodiment, and the image ID of the defect or defect candidate transmitted from at least the review apparatus and the CD-SEM of the semiconductor device manufacturing department via the communication means and the image thereof. A storage unit that stores and stores at least a semiconductor device ID with an ID feature amount, and a defect or a defect candidate corresponding to the semiconductor device ID based on at least the semiconductor device ID inquired through the communication unit To the associated semiconductor device based on at least the feature quantity of the image corresponding to the image ID obtained by using the communication means as the primary answer selection result. A search engine for searching for a corresponding defect or defect candidate image ID group is provided, and an image ID is assigned by searching the search engine. The image ID of the defect or defect candidate corresponding to the semiconductor device ID is provided to the semiconductor device manufacturer or design house or its customer via the communication means as the primary answer, and the related semiconductor device searched by the search engine is provided. An image data service system comprising a database center for providing a service by providing data of a corresponding defect or defect candidate image ID group to a semiconductor device manufacturer or a design house or its customer via communication means is there.
[0015]
In addition, the present invention corresponds to the second embodiment, and the storage device unit is composed of distributed sites.
[0016]
The present invention also corresponds to the first embodiment, and is a method for managing image data using a database center including a storage device unit and a search engine. The storage device unit includes a semiconductor device. A storage process for storing and storing at least a semiconductor device ID with at least a semiconductor device ID and an image of a defect or a defect candidate transmitted from at least a review apparatus and a CD-SEM of a manufacturing department via a communication unit; The defect or defect corresponding to the semiconductor device related to the semiconductor device ID based on the feature amount of the defect or defect candidate image corresponding to at least the semiconductor device ID inquired through the communication means using the search engine A search process for searching for candidate image groups, and corresponding to the related semiconductor device searched in the search process. It is a method of managing image data and providing a defect or a data image group of the defect candidate via the communication means to the semiconductor device manufacturer or design house, or the customer.
[0017]
The present invention also corresponds to the first embodiment, and is a method for managing image data using a database center including a storage device unit and a search engine. The storage device unit includes a semiconductor device. A storage process for storing and storing at least a semiconductor device ID with at least a semiconductor device ID and an image of a defect or a defect candidate transmitted from at least a review apparatus and a CD-SEM of a manufacturing department via a communication unit; The search engine is used to search for an image of a defect or a defect candidate corresponding to the semiconductor device ID based on at least the semiconductor device ID inquired through the communication means, and at least an image ID is assigned to the primary answer. And at least obtained using communication means as the selection result of the primary answer using the search engine. A search process for searching for an image group of defects or defect candidates corresponding to a semiconductor device based on a feature quantity of an image corresponding to an image ID, and the primary answer in the primary answer process is transmitted via communication means. The data of the defect or defect candidate image group corresponding to the related semiconductor device searched in the search process is provided to the semiconductor device manufacturer or design house or its customer via the communication means. Or it is the management method of the image data characterized by providing to the customer.
[0018]
Further, the present invention corresponds to the third embodiment, and is a method for managing image data using a database center including a storage device unit and a search engine, and the storage device unit includes a semiconductor device. Storage that stores and accumulates at least a semiconductor device ID and an image ID of a defect or defect candidate and a feature amount of the image ID transmitted from at least a review apparatus and a CD-SEM of a manufacturing department via a communication unit Corresponding to the semiconductor device related to the semiconductor device ID based on the process and the feature amount of the defect ID or defect candidate image ID corresponding to the semiconductor device ID inquired through the communication means using the search engine A search process for searching for a defect or defect candidate image ID group, and the related semiconductor searched in the search process It is a method of managing image data and providing data of an image ID group corresponding defects or defect candidates via the communication means to the semiconductor device manufacturer or design house or a customer device.
[0019]
Further, the present invention corresponds to the third embodiment, and is a method for managing image data using a database center including a storage device unit and a search engine, and the storage device unit includes a semiconductor device. Storage that stores and accumulates at least a semiconductor device ID and an image ID of a defect or defect candidate and a feature amount of the image ID transmitted from at least a review apparatus and a CD-SEM of a manufacturing department via a communication unit And a primary answer in which a search is made for a defect or a defect candidate image ID corresponding to the semiconductor device ID based on at least the semiconductor device ID inquired through the communication means using the search engine. A process and at least an image ID obtained using the communication means as a selection result of the primary answer using the search engine. A search process for searching for an image group of defects or defect candidates corresponding to a semiconductor device related based on the feature quantity of the image, and the primary answer in the primary answer process is sent to the semiconductor device manufacturer via the communication means or Provide the design house or its customer with the image ID group data of the defect or defect candidate corresponding to the related semiconductor device searched in the search process to the semiconductor device manufacturer or the design house or customer through the communication means. An image data management method is provided.
[0020]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments according to the present invention will be described with reference to the drawings.
[0021]
As shown in FIG. 1, the semiconductor device manufacturer A company 1a includes a sales / technical sales department 21, a development design department (including a quality assurance department) 22, and a manufacturing department 20. In addition, about manufacture of a semiconductor device, you may entrust FAB (foundry AC). In this case, the semiconductor device manufacturer is close to a design house having a sales / technical sales department and a development design department (including a quality assurance department).
[0022]
A semiconductor device production line is constructed at each of the bases (Fab.A, Fab.B, Fab.C,...) Of the manufacturing department 20, and an inspection apparatus (optical (BF (Bright Field), DF (Dark Field)), SEM type) 11, review device (SEM type, optical type) 12, length measuring SEM (CD (Critical Dimension) -SEM) 13 and the like are installed. Each server 10 adds the image feature amount to the SEM review image or optical review image after ADC processing having high resolution from the SEM type or optical review device 12 and collects it via the network 15, and collects it via the CD-CEM 13. From the image of the gate and wiring on the semiconductor wafer imaged by the CD-SEM and the image feature amount that characterizes the shape calculated by the image processing (numerical values expressing the shape such as the width and inclination angle of the gate and wiring) Data) is added and collected via the network 15. A method for calculating the width, inclination angle, etc. of the cross-sectional shape as numerical information from the cross-sectional signal file of the wiring imaged by the CD-SEM will be referred to as MPPC (Multiple Parameter Profile Characterization) hereinafter. As the image feature amount of the defect or defect candidate, the size (area, length in the rotational symmetry axis direction, projection length in the X-axis and Y-axis directions), brightness (gradation value: gradation value), shape, There are textures. Of course, the real-time ADC image obtained from the inspection apparatus 11 is basically a rough image for selecting a defect or a defect candidate to be captured by the review apparatus and the CD-SEM, but at a high speed. Since the number of wafers to be sampled is very large because inspection is possible, an image of a defect or defect candidate that has not been captured by the review apparatus and the CD-SEM is obtained. Therefore, each server 10 can select a desired one from real-time ADC images obtained in large quantities from the inspection apparatus 11 and collect them in a form (hierarchical) attached to a review image. As described above, the real-time ADC image is coarse image data, but can be increased as an information amount.
[0023]
Then, each server 10 displays “image + image feature quantity” of the collected review image and length measurement image, as shown in FIG. 7A, device ID, date of manufacture, device product name, device type, inspection. Attaching image ID information such as measurement process and manufacturing FAB to the server 40a dedicated to the company A in the data center 40 that is consistently managed via the Internet 30, and storing it in the storage device 42a such as a disk array It will be. Note that the inspection device 11, the review device 12, the CD-SEM 13, the image ID information, and the like are managed by the PC 16. Security is ensured between each server 10 and the server 41a dedicated to the company A by encryption technology. Further, a large amount of “review image + its image feature amount” is acquired from the review apparatus 12 for each process, for each lot, for each device type, and from the CD-SEM 13 for each process, for each lot. Since a large amount of “measurement image + its feature value” is acquired for each product type, the server 10 performs image compression or the like and uses high-speed communication for exclusive use of the company A via the Internet (communication means) 30. The data is stored in a storage device 42a such as a disk array of the server 41a. As image compression, since similar images are acquired continuously, a compression method for taking a difference between images such as MPEG can be considered. Further, in the manufacturing department 20 of the company A 1a, it is necessary to standardize the connection IF between the review apparatus 12 and the CD-SEM 13 and the server 10 and the data structure of “image + image characteristic amount”.
[0024]
Further, the manufacturing department 20 of the semiconductor device manufacturer B company 1b transmits the disk to the server 41b dedicated to the company B via the Internet (communication means) 30 in the same manner as the company A 1a in a state where security is ensured. The data is accumulated in the storage device 42b such as an array.
[0025]
Further, the manufacturing department 20 of the semiconductor device maker C company 1c is transmitted to the server 41c dedicated to the company B via the Internet 30 and stored in the storage device 42c such as a disk array in a state where security is ensured. become.
[0026]
Naturally, in the data center 40, the data for each company can be kept secure by applying a number of keys similar to the encryption technology.
[0027]
Further, a method of transmitting “image + image feature quantity” from the manufacturing department of each semiconductor device manufacturer together with the image acquisition process information to the dedicated server 41 of each company of the image DB center 40 and storing it in the storage device 42 dedicated to each company. 5 and FIG. That is, as an apparatus for acquiring an image regarding one wafer, an inspection apparatus (optical (BF, DF), SEM type) 11, a review apparatus (SEM type, optical) 12, a CD-SEM 13, an analysis apparatus (mass spectrometry) , Spectroscopic analysis) and the like. In particular, the server 10 collects an SEM review image and an optical review image having a high resolution after ADC processing from the review device 12 and collects a length measurement image having a high resolution from the CD-SEM 13.
[0028]
Also, as an opportunity to acquire an image for one wafer by the above apparatus, [diffusion process (gate), capacitance process (capacitor, contact hole), wiring process (Al wiring, Cu wiring, low-k film),...] * [After film formation, after development, after etching, etc.] For example, the gate shape and gate dimensions in the diffusion process are after film formation and etching.
[0029]
Further, a plurality of [image + its image feature amount] are acquired for one wafer (one lot / one device).
[0030]
Furthermore, the image shown in FIG. 6 is acquired by the review SEM and the length measurement SEM (CD-SEM). In the case of the review SEM, (1) a fatal defect, (2) a non-fatal defect, and (3) a pseudo defect defect image and its image feature amount. For length measurement SEM, (4) dimensions, (5) cross-sectional shape information (MPPC, 3D technology), (6) actual circuit pattern image of exposure state (MPPC) and its image features, and (7) process diagnosis The TEG image and the image feature amount of the appearance such as TEG and reliability TEG. Therefore, the semiconductor device manufacturer can search for [image + image feature quantity] focused on the semiconductor process from the data center 40.
[0031]
These acquired [images of (1) to (7) + image feature quantity focused on the phenomenon of (1) to (7)] are stored in the storage device 42 of the data center 40 as a DB together with the device ID, the image acquisition process, and the like. Will be saved.
[0032]
As described above, the data center 30 has a variety of semiconductor devices manufactured by semiconductor device manufacturers for a long period of time (for example, 5 to 10 years) and sold to customers. [Images of (1) to (7) + image feature amounts focused on the phenomena of (1) to (7)] obtained at least in units of lots are accumulated. If the storage period is about 5 to 6 years, the storage capacity of the storage device 42 can be stored at about 100 TB per site without image compression.
If image compression is performed, the storage capacity per site can be reduced.
[0033]
Next, analysis and analysis of inquiries or requests by semiconductor device manufacturers (including design houses) about the quality of semiconductor devices (including claims such as failures) after a long period of time from customers who sold manufactured semiconductor devices The result report (including compensation) will be described with reference to FIGS.
[0034]
[Workflow 1]
First, the first embodiment shown in FIG. 2 will be described.
[0035]
When a sales / technical sales department 21 of a semiconductor device manufacturer makes an inquiry or a request about the quality (including complaints such as failure) about a semiconductor device sold after a long period of time, the development design department (quality assurance department) 22) confirms a location (for example, a failure location) to be inspected for quality, and checks the quality (for example, failure, defect) as shown in FIGS. 2 and 3 for the device ID (chip ID (product type), wafer) ID, lot ID, date of manufacture, etc.) are extracted.
[0036]
(1-1) The semiconductor device manufacturer makes an inquiry (S1) to the dedicated server 41 of the data center (DB center) 40 via the Internet 30 with the device ID from the development design department 22 or the like. For the management in the image DB stored in the storage device 42, image ID information corresponding to the device ID is given to [image + its image feature quantity].
[0037]
(1-2) The DB center 40 retrieves the image data for the inquiry device ID from the database stored in the storage device 42 and returns it to the device manufacturer 1 as a primary answer.
[0038]
That is, the dedicated server 41 uses the search engine 43 capable of high-efficiency and high-speed search, as shown in FIG. 7B, when there is data with the same device ID as the device ID of the inquiry (S1), A series of images as shown in FIG. 8 corresponding to the IDs and their image feature amounts are searched (S2), and a primary answer is sent to the dedicated PC 17 of the development design department 22 or the manufacturing department 20 via the Internet 30 (S3). . FIG. 8 shows a review image / length measurement SEM image that is inspected and measured for each process when the device ID is manufactured. Naturally, the image feature amount is attached to the image. In FIG. 9, the defect image searched by the search key and the reference image which is a comparison image at that time are shown side by side. The reference image is a non-defective image (that is, there is no defect at all) in the same part as the defect image used for specifying the position of the defect in the defect image.
By subtracting the reference image and the defect image, a portion having a difference can be specified as the defect position. In the manufacturing process of a semiconductor wafer, it is common to manufacture by regularly arranging chips on the wafer. The difference in obtaining a reference image for a certain defective part is that other positions in the chip are the same. It is usual to image a chip (for example, an adjacent chip).
[0039]
If there is no data with the same device ID as the device ID of the inquiry (S1) by the search engine 43, as shown in FIG. Items close to the day are sent to the development design department 22 or the dedicated PC 17 of the manufacturing department 20 via the Internet 30 as alternative primary answers.
[0040]
(1-3) Next, the device manufacturer browses the returned image and designates a manufacturing process or the like to which attention is paid. That is, the dedicated PC 17 selects the problem image (the length measurement SEM process C in FIG. 8), and limits the acquisition process to the selected image ID (the image feature amount is also grasped). In addition, a similar image search inquiry (S4) based on the selected image ID is made to the DB center 40.
[0041]
(1-4) The DB center 40 searches for similar images and device IDs using the search engine 43 only for the manufacturing process designated by the device manufacturer. That is, the dedicated server 41 of the DB center 40 is a search engine capable of high-efficiency and high-speed search based on the similar image search inquiry (S4) based on the image ID (including the image acquisition process) selected by the device manufacturer. 43, for the same product and the same image acquisition process, image ID information having an image feature amount similar to the image feature amount of the image ID and its device ID are searched (S5).
[0042]
(1-5) The retrieved result is sent from the DB center 40 to the device manufacturer 1. That is, the dedicated server 41 of the DB center 40 acquires the searched image ID information having the similar image feature amount and its device ID (S6), and the development design department 22 or the manufacturing department 20 via the Internet 30. To the dedicated PC 17 (S7).
[0043]
As a high-speed image search method by the search engine 43, there is a method of using a defect image feature amount space. FIG. 13 shows the image search using the defect type and the defect feature amount space, and the search accuracy rate at that time. As the type (category) of the defect, large foreign matter, potential contrast change, small foreign matter, lower layer bridge, and the like are shown. In this figure, the feature space created as a result of expressing the defect data captured in each image with dozens of feature values consisting of brightness, size, defect texture, etc., is analyzed using a statistical method such as principal component analysis. A three-dimensional space that has been dimensionally compressed by using is shown. One point in the figure corresponds to one image. Here, if the similarity of the image feature amount is the distance of the defect in the feature amount space, when an input image is given, the image search is performed by searching for an image that is close to the image in the feature amount space. Yes. By using such a dimension-compressed space, image ID information having a similar image feature amount and its device ID can be searched at high speed.
[0044]
FIG. 10 shows an example of a related image list (basic form) and visualization data display. The basic form of the related image list includes an image, device ID, date of manufacture, similarity of image feature amounts, and the like. Examples of data display include the relationship between the similarity of the image feature and the date of manufacture, the relationship between the similarity of the image feature and the device ID, and the relationship between the similarity of the image feature and the image. Specifically, the manufacturing period is higher than the similarity XX, the device ID is higher than the similarity XX, the image list is displayed with the similarity XX, and the image list is displayed in the similarity [XX, YY] section.
[0045]
As a result, it becomes possible for the device manufacturer to acquire the device ID and its image that are highly relevant to the device that investigates the quality requested by the customer (for example, a failure), and quickly reports the investigation report to the customer ( It is possible to submit a form with an image attached.
[0046]
In addition, the device manufacturer can acquire the device ID and its image that are highly relevant to the device that investigates the quality requested by the customer (for example, failure or defect), and the defect lot or the defect process. As a result, it is possible to realize a minimum compensation expense for the customer.
[0047]
[Workflow 2]
Next, a second embodiment shown in FIG. 3 will be described.
[0048]
(2-1) The device manufacturer 1 inquires the DB center 40 about the ID of the defective or failed device.
[0049]
(2-2) The DB center 40 searches the database for image data for the device ID for which the inquiry has been made.
[0050]
(2-3) The DB center 40 searches for an image similar to the searched image data and its device ID.
[0051]
(2-4) The search results of (2-2) and (2-3) are sent from the DB center 40 to the device manufacturer 1.
[0052]
In the second embodiment, the difference from the first embodiment shown in FIG. 2 is that the dedicated server 41 of the DB center 40 searches the database for image data for the inquired device ID. Next, the dedicated server 41 of the DB center 40 searches for an image similar to the searched image data and its device ID. In other words, the dedicated server 41 uses the search engine 43 that can perform a high-efficiency and high-speed search based on the image feature quantity of the device ID that has been inquired, and has similar image feature quantities for the same product. Data (highly relevant data) is searched (S8), and the searched image ID information having the similar image feature quantity and its device ID are acquired (S6). The primary answer (S3) is only presented to the device manufacturer.
[0053]
As described above, according to the first and second embodiments, the device manufacturer responds to a request for a quality (including failure) survey for a desired semiconductor device from a customer after a long period of time. It becomes possible to quickly acquire a device ID and its image that are highly relevant to the device, and as a result, it is possible to quickly submit a survey report (a form with an image attached) to the customer.
[0054]
In addition, it is possible to access the DB center 40 from the review device 12 or the CD-SEM 13 via the server 10 to search for images.
[0055]
Further, since the [image + image feature value] accumulated in the DB center 40 is obtained from the review device 12, the CD-SEM 13, or the like, the [manufacturer] manages the [image + image feature value]. Is convenient.
[0056]
Further, a company that stores and stores image data in the DB center 40 for a long period of time and a search service can charge the device manufacturer or its customers for the cost required for these services.
[0057]
[Workflow 3]
Next, a third embodiment shown in FIG. 4 will be described.
[0058]
(3-1) The device manufacturer 1 inquires the DB center 40 about the ID of the defective device (including the failed device) (S1).
[0059]
(3-2) The DB center 40 retrieves image data for a plurality of defective devices from the database based on the device ID inquired from the database (S2), and makes a primary answer to the device manufacturer 1 (S3).
[0060]
(3-3) The device manufacturer 1 makes a request for data necessary for analyzing the image data obtained in S3 (here, it is called a 1.5th order request) (S9).
[0061]
For example, the device manufacturer 1 tries to know the problem process or the problem defect mode by examining what is different from the non-defective device in the image group of the defective device that is the search result of S3. Specify the time when a defect does not occur within the period during which the device was manufactured (the time when the user of the product in which the semiconductor device is mounted has no complaint about the defect), and the image data of the device manufactured at that time This 1.5th order request is made to the DB center 40. The production time of the non-defective product can be acquired from the customer manufacturer, that is, the supply of the device of the model number is started. From then on, it can be considered that good devices were manufactured during the period until many claims were made.
[0062]
(3-4) The DB center 40 collects predetermined image data according to the 1.5th order request from the device manufacturer 1 and returns it to the device manufacturer 1 (S10). At that time, the similarity evaluation value result with the defective device image is also sent to the device manufacturer 1.
[0063]
In accordance with the 1.5th order request, the DB center 40 searches for image data of a device at a certain time and returns it to the device manufacturer 1. At this time, instead of returning only the collected database, it is calculated by calculating how much data similar to the search result of S3 (that is, the image data of the defective device) is included in the collected image group. The results of sex evaluation will also be sent.
[0064]
For example, it is assumed that there are 10 steps of image data in the search result of S3. A 1.5-order request provides an image data set of a plurality of devices for devices in good production time, and there are also 10 steps of image data. Then, for each process (that is, every 10 processes), the number of images included in the image data group of the non-defective device similar to (related to) the image included in the image data group of the defective product device is checked. . For the evaluation of similarity, the distance in the feature amount space (FIG. 13) is used. When the distance in the feature amount space is smaller than a predetermined threshold value, the two images are similar. As a result, for each image data of the defective device, it can be understood how much the image similar to that is included in the image data of the non-defective device.
[0065]
FIG. 14 shows an image of data to be sent. Here, the data from step 1 to step 10 indicates how much data similar to each defect is included. Here, a category is assigned to each defect. Thereby, it can obtain | require, for example by clustering in the feature-value space. Since similar defects tend to be dense (form clusters) in the feature space, if the distance between each point in the feature space is smaller than a certain threshold, the cluster is detected by considering it as the same cluster. Can do. For each defect belonging to each category, the ID and the number of non-defective images similar to the defect are recorded. From these data, the lower bar graph compares the sum of the defects in the same category in the defective device and the number of non-defective device images similar to the defect for each category. In this example, for category 1, the number of non-defective devices is almost the same as the number of defective devices, and there is no particular difference, whereas category 2 shows that the defects are hardly seen in non-defective devices. means. From this result, it is predicted that the efficiency will be improved if the subsequent search is performed focusing on the category 2 defect in this process. Of course, in this display, if there is no difference between a defective device and a non-defective device in any category, it means that the process does not need attention.
[0066]
(3-5) In the device manufacturer 1, using the return result (good image and / or defect / non-defective image similarity evaluation data) for the 1.5th request, the process to be focused on (problem process / search range instruction) Is selected and answered to the DB center 40 (S11).
[0067]
The device manufacturer 1 searches for a process having a defect image that can be found in a defective device and that is not found when manufacturing a non-defective device, and pays attention to the defect mode (category).
[0068]
(3-6) The DB center 40 searches for an image and a device ID similar to the image of the target defect mode (category) only for the process specified in S11 by the device manufacturer side (S12). At that time, device IDs manufactured before and after the defective device is manufactured are targeted.
[0069]
Since the steps to be noted are narrowed down in order to see the difference between the non-defective device and the defective device by the above S11, what time do you want to know next when did the difference occur? When did it go away (if that happened)? It turns out that. Therefore, a data search is performed using data manufactured before and after the defective device is manufactured as a search target. The search period can be designated as, for example, one month before or after, or the first half of the year in the instruction (S11) from the device manufacturer 1.
[0070]
(3-7) The search result of S6 and the time series similarity evaluation data are sent from the DB center 40 to the device manufacturer 1 (S13).
[0071]
In accordance with the secondary request, the DB center 40 searches for device image data for data within a specified period, and returns it to the device manufacturer 1. At this time, rather than returning only the collected database, the collected image group includes how much data is similar (related) to the search results of S2 and S3 (that is, the image data of the defective device). Also send the similarity evaluation result calculated. The similarity evaluation index is calculated in the same manner as in step S10. As a result, the number of observed defect modes (categories) observed in the defective device can be viewed in time series in the target process within the designated period.
[0072]
FIG. 15 shows an image of this data. In this example, one of the bar graphs means one week, and the transition of defects similar to the defect mode existing in the defective product device in the device that has passed the target process is shown in the one week. In this example, it is shown that the occurrence of defects is remarkably seen from 12 weeks before and 5 weeks after the defective device (failed device) is manufactured.
[0073]
FIG. 16 shows an embodiment in which the time when a defective device is generated is estimated based on the CD (Critical Dimension) value of the gate wiring measured by the CD-SEM 13.
It is known that the gate length of the semiconductor device measured in the CD-SEM 13 affects the performance of the transistor. The CD-SEM 13 periodically observes certain points of a plurality of chips in the wafer, and from the image. By determining the gate length (CD value), it is possible to see the trend. In the previous example of the result image, the image search is performed using the data obtained by digitizing the brightness of the defect as the image feature amount. However, in the case of the CD-SEM, the same is performed from the gate profile calculated from the captured image. The image feature amount can be calculated, and data search can be performed using the image feature amount. FIG. 16 shows the average values (for example, for one week) of the gate length (CD value) of the device before and after the gate creation process of the defective device in time series. Thus, it can be seen that the CD value is lower than the values in the other sections in the 9 weeks before and 2 weeks after the manufacture of the defective device (failed device). Such an abnormal device section can be specified by comparison with a threshold value for a predetermined CD value. Since the CD-SEM is used for regular fixed-point observation unlike the review device, it is not always necessary to perform an image search in step S6 in order to display the trend as shown in FIG.
[0074]
Thereafter, the device manufacturer 1 examines a countermeasure based on the result obtained in S13. Then, by tracing the index value of similarity evaluation included in the reply result of S13 in the time direction, at which point did the problem causing the current defect occur? And where did it last? If it can be confirmed, it is possible to define a device that seems to contain the problem as a risky device and collect them. As a result of S13, if there are a plurality of failure modes that are considered to have such a problem, it is also possible to regard devices related to all of them as risky devices. If the device in question is currently in production and the problem still appears, it can be fed back into the current process.
[0075]
Next, variations of the embodiment will be described. That is, if the workflow 3 and the workflows 1 and 2 are used as a base, variations of various embodiments can be considered.
[0076]
[If there are multiple defective devices]
There may be a case where there are a plurality of defective devices that trigger the first inquiry (primary inquiry) from the device manufacturer 1 to the DB center 40. This is because the number of devices in which defects occur after shipment is not limited to one. In this case, the determination of the problem process in the device manufacturer 1 (S12 in the workflow 3 or S5 in the workflow 1) is made after comparing data obtained from a plurality of defective devices. I don't think there will be any difference in the above workflow.
[0077]
Further, the input (inquiry) to the DB center 40 can be considered as a module product. That is, the semiconductor device according to the present invention naturally includes a module product. This is because a semiconductor device shipped from a device manufacturer to a customer is considered to be a combination of a plurality of semiconductor components (a plurality of devices mounted on one substrate), and thus there are a plurality of defective devices. Like the case.
[0078]
[Further refine defect information]
In the workflow 3 described above, an example is described in which the search is performed using the feature amount of the image data after the search target is narrowed down by the manufacturing process instructed by the device manufacturer 1 or the device manufacturing time. It is not limited to only. In other words, the device manufacturer can further narrow down the data if there is any other information. For example, it may be possible to search only for data in which a characteristic that is an electrical characteristic inspection at the time of shipment of a certain device is seen (eg, the threshold voltage of a transistor is within a specified range but is slightly lower). The conditions used for such narrowing down include the material used in the process (the film material in the case of a film forming process) and the manufacturing equipment (if a plurality of manufacturing equipment are used in the same process, the machine number). Conceivable.
[0079]
[Application to post-wafer process]
In the description so far, only the pre-process of the wafer (process for forming a circuit pattern on the wafer) has been considered, but the same idea can be applied to the post-process of the wafer (chip cutting, packaging, etc.). The image data in this case is assumed to be image data in an appearance inspection (packaging or lead wire portion appearance inspection) performed in a later process. In wire bonding (a step of attaching a wire to a terminal portion of a chip), since positioning is performed using an image, image data used for the positioning can be stored.
[0080]
Next, a second embodiment of an image database integrated management system inspected in the manufacturing department of a semiconductor device manufacturer according to the present invention will be described with reference to FIG. The configuration shown in FIG. 11 shows a case where the DB sites 45 a to 45 d are distributed as the virtual DB 45 in the DB center 40.
[0081]
Next, a third embodiment of the image database integrated management system inspected in the manufacturing department of the semiconductor device manufacturer according to the present invention will be described with reference to FIG. The configuration shown in FIG. 12 shows a case in which only the image feature amount with the defect or defect candidate image ID attached is stored in the DB center 40 and stored for a long period of time. For this reason, images obtained by the review device 12 or the CD-SEM 13 are accumulated in the image DB 18 at each base of the manufacturing department and stored for a long time. With this configuration, the storage capacity in the DB center 40 can be reduced. On the other hand, it becomes necessary for the device manufacturer to manage the image DB 18 for a long period of time.
[0082]
【The invention's effect】
According to the present invention, image data obtained from a review apparatus or a CD-SEM installed in a semiconductor production line is stored and stored for a long period of time, and an image of a semiconductor device highly relevant to a required semiconductor device. By making it possible to search for information, it is possible to quickly provide semiconductor device manufacturer information about the quality of semiconductor devices to customers even after a long period of time, thereby maintaining the trust of customers. Play.
[0083]
Further, according to the present invention, it is possible to realize a minimum compensation expense for a customer by narrowing down a defective lot or a defective process in a semiconductor device manufacturer.
[0084]
In addition, according to the present invention, it is possible to realize consistent management of an image database over a long period of a year for a large number of semiconductor devices having a large number of manufacturing process steps.
[Brief description of the drawings]
FIG. 1 is a schematic configuration diagram showing a first embodiment of an image database integrated management system inspected in a manufacturing department of a semiconductor device manufacturer according to the present invention.
FIG. 2 is a first method of acquiring a semiconductor device information and an inspection image thereof that are highly related to a semiconductor device that requires quality information and an inquiry flow and an inquiry flow of an image search in the integrated image database management system shown in FIG. It is a figure for demonstrating the Example of.
3 is a second example of a method for acquiring a semiconductor device information and an inspection image thereof that are highly related to a semiconductor device that requires quality information and an inquiry flow and an inquiry flow of an image search in the integrated image database management system shown in FIG. It is a figure for demonstrating the Example of.
4 is a third example of a method for acquiring semiconductor device information and an inspection image thereof that are highly related to a semiconductor device that requires quality information and an inquiry flow and an inquiry flow of an image search in the image database integrated management system shown in FIG. It is a figure for demonstrating the Example of.
FIG. 5 is a diagram for explaining the contents of data related to an apparatus for acquiring an image installed in a semiconductor manufacturing process and a process for acquiring an image, which are stored in a DB center according to the present invention.
FIG. 6 is a diagram for explaining the contents of image data acquired from a review device, a CD-SEM, or the like, stored in a DB center according to the present invention.
FIG. 7A is a diagram showing the contents of data stored in the DB center according to the present invention, and FIG. 7B is a diagram for explaining the difference in primary answering method depending on presence / absence of data of the same device ID. It is.
FIG. 8 is a view showing an embodiment of a primary answer image for an inquiry according to the present invention.
FIG. 9 is a diagram illustrating a search result image searched and displayed by a search key and a reference image compared at that time.
FIG. 10 is a diagram for explaining an embodiment of a method for presenting related images according to the present invention.
FIG. 11 is a schematic configuration diagram showing a second embodiment of an image database integrated management system inspected in a manufacturing department of a semiconductor device manufacturer according to the present invention.
FIG. 12 is a schematic configuration diagram showing a third embodiment of an image database integrated management system inspected in a manufacturing department of a semiconductor device manufacturer according to the present invention.
FIG. 13 is a diagram for explaining a pattern recognition type which is an embodiment of a search engine according to the present invention, and a diagram showing an image search using a defect type and a defect feature amount space, and a search accuracy rate at that time. It is.
FIG. 14 is a diagram showing an image showing how much data similar to each defect with a category is included in the data from step 1 to step 10 according to the present invention.
FIG. 15 shows an image of data in which the number of observed defect modes (categories) observed in a defective device can be viewed in time series in a target process within a specified period according to the present invention. It is.
FIG. 16 is a diagram showing an average value (for example, for one week) of gate lengths (CD values) of devices before and after the gate creation process of a defective device according to the present invention arranged in time series.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Device maker, 10 ... Server, 11 ... Inspection apparatus, 12 ... Review apparatus, 13 ... CD-SEM (measurement SEM), 15 ... Network, 16 ... PC, 17 ... Dedicated PC, 18 ... Image DB, 20 ... Manufacturing department, 21 ... Sales / Technology sales department, 22 ... Development design department (including quality assurance department), 30 ... Internet (communication means), 40 ... Data center (DB center), 41a-41c ... Dedicated server, 42a- 42d ... Storage device, 43 ... Search engine, 45 ... Virtual DB, 45a-45d ... Site.

Claims (10)

A memory for storing and storing at least a semiconductor device ID with at least a defect or defect candidate image transmitted from a semiconductor device manufacturing department via a communication means from a review device and a CD-SEM. A device section;
Retrieval for searching for a defect or defect candidate image group corresponding to a semiconductor device related to the semiconductor device ID based on at least the feature quantity of the defect or defect candidate image corresponding to the semiconductor device ID inquired through the communication means With an engine,
A database center is provided that provides services by providing data of defect or defect candidate image groups corresponding to related semiconductor devices searched by the search engine to a semiconductor device manufacturer or a design house or its customers via communication means. An image data service system characterized by that. A memory for storing and storing at least a semiconductor device ID with at least a defect or defect candidate image transmitted from a semiconductor device manufacturing department via a communication means from a review device and a CD-SEM. A device section;
Search for an image of a defect or defect candidate corresponding to the semiconductor device ID based on at least the semiconductor device ID inquired through the communication means, assign at least an image ID, and make a primary answer using the communication means, Based on at least the image ID obtained using the communication means as the selection result of the primary answer, the image group of the defect or defect candidate corresponding to the related semiconductor device is searched based on the feature amount of the image corresponding to the image ID. With a search engine,
Providing the image of the defect or defect candidate corresponding to the semiconductor device ID to which the image ID is given by searching with the search engine to the semiconductor device manufacturer or the design house or its customer as the primary answer via the communication means; A database center is provided that provides services by providing data of defect or defect candidate image groups corresponding to related semiconductor devices searched by the search engine to a semiconductor device manufacturer or design house or its customers via communication means. An image data service system characterized by that. At least a semiconductor device ID is assigned to the image ID of the defect or defect candidate transmitted from the semiconductor device manufacturing department via the communication means from the review apparatus and the CD-SEM, and the feature amount of the image ID is accumulated and stored. A storage device unit,
A defect or defect candidate image ID group corresponding to the semiconductor device related to the semiconductor device ID is searched based on at least the feature quantity of the defect or defect candidate image ID corresponding to the semiconductor device ID inquired through the communication means. And a search engine
A database center for providing a service by providing data of a defect or a defect candidate image ID group corresponding to a related semiconductor device searched by the search engine to a semiconductor device maker or a design house or its customer via communication means An image data service system characterized by being provided. At least a semiconductor device ID is assigned to the image ID of the defect or defect candidate transmitted from the semiconductor device manufacturing department via the communication means from the review apparatus and the CD-SEM, and the feature amount of the image ID is accumulated and stored. A storage device unit,
Based on at least the semiconductor device ID inquired via the communication means, the image ID of the defect or defect candidate corresponding to the semiconductor device ID is searched, the primary answer is made using the communication means, and the selection result of the primary answer A search engine for searching for a defect or defect candidate image ID group corresponding to a semiconductor device based on at least the feature amount of the image corresponding to the image ID obtained using the communication means,
The image ID of the defect or defect candidate corresponding to the semiconductor device ID to which the image ID is assigned by searching with the search engine is provided as the primary answer to the semiconductor device manufacturer or the design house or its customer via the communication means. A database center that provides services to semiconductor device manufacturers or design houses or their customers via communication means with image ID group data of defects or defect candidates corresponding to related semiconductor devices searched by the search engine An image data service system characterized by comprising: 5. The image data service system according to claim 1, wherein the storage device unit is configured by distributed sites. A method for managing image data using a database center including a storage device and a search engine,
At least a semiconductor device ID is assigned to the storage device unit with at least a review device of a semiconductor device manufacturing department and an image of a defect or defect candidate transmitted from a CD-SEM via a communication means and a feature amount of the image. Storage process to accumulate and save,
Using the search engine, the defect or defect candidate corresponding to the semiconductor device related to the semiconductor device ID based on the feature quantity of the image of the defect or defect candidate corresponding to at least the semiconductor device ID inquired through the communication means A search process for searching a group of images,
Image data management characterized by providing image data of defects or defect candidates corresponding to related semiconductor devices searched in the search process to a semiconductor device manufacturer or a design house or its customer via communication means Method. A method for managing image data using a database center including a storage device and a search engine,
At least a semiconductor device ID is assigned to the storage device unit with at least a review device of a semiconductor device manufacturing department and an image of a defect or defect candidate transmitted from a CD-SEM via a communication means and a feature amount of the image. Storage process to accumulate and save,
Using the search engine, a defect or a defect candidate image corresponding to the semiconductor device ID is searched based on at least the semiconductor device ID inquired through the communication means, and at least an image ID is assigned and a primary answer is given. A defect corresponding to a semiconductor device associated with a primary answer process and a search result using the search engine based on at least a feature quantity of an image corresponding to an image ID obtained using a communication unit as a selection result of the primary answer A search process for searching for image groups of defect candidates,
A primary answer in the primary answering process is provided to a semiconductor device manufacturer or a design house or its customer via communication means, and a defect or defect candidate image group corresponding to a related semiconductor device searched in the searching process is provided. A method for managing image data, characterized in that the above data is provided to a semiconductor device manufacturer, a design house, or a customer thereof via communication means. A method for managing image data using a database center including a storage device and a search engine,
At least a semiconductor device ID is assigned to the storage device unit with an image ID of a defect or defect candidate transmitted from a semiconductor device manufacturing department via at least a review device and a CD-SEM via communication means, and a feature amount of the image ID. Storage process that accumulates and stores,
Using the search engine, the defect or defect corresponding to the semiconductor device related to the semiconductor device ID based on the feature quantity of the image ID of the defect or defect candidate corresponding to at least the semiconductor device ID inquired through the communication means A search process for searching for candidate image ID groups,
Data of image ID groups of defects or defect candidates corresponding to related semiconductor devices searched in the search process is provided to a semiconductor device manufacturer or a design house or its customer via communication means. Management method. A method for managing image data using a database center including a storage device and a search engine,
At least a semiconductor device ID is assigned to the storage device unit with an image ID of a defect or defect candidate transmitted from a semiconductor device manufacturing department via at least a review device and a CD-SEM via communication means, and a feature amount of the image ID. Storage process that accumulates and stores,
Using the search engine, a primary answering process for searching for a primary image by searching for an image ID of a defect or a defect candidate corresponding to the semiconductor device ID based on at least the semiconductor device ID inquired through the communication means;
Using the search engine, an image group of defects or defect candidates corresponding to a semiconductor device related to at least an image feature amount corresponding to an image ID obtained using a communication unit as a selection result of the primary answer. A search process for searching,
An image ID of a defect or a defect candidate corresponding to a related semiconductor device searched in the search process is provided to a semiconductor device manufacturer or a design house or its customer via the communication means through the primary response in the primary answer process. A method for managing image data, comprising providing group data to a semiconductor device manufacturer, a design house, or a customer thereof via communication means. The image data management method according to claim 6, 7, 8, or 9, wherein the storage unit is configured by distributed sites.

Images