JPH1145919A - Manufacture of semiconductor substrate - Google Patents

Abstract

PROBLEM TO BE SOLVED: To manufacture a semiconductor substrate of high quality at high yield, by early finding an abnormal situation from failure data obtained by inspecting with an inspection device, to significantly reduce defects in manufacturing. SOLUTION: Inspected failure position data 101, 102 and 103 on each semiconductor substrate are specified on the coordinate of image data comprising lattice pixels set on the semiconductor substrate, and the number of failures are added for each lattice pixel on the coordinate of specified image data for a plurality of semiconductor substrates, and failure distribution image data 111, 112 and 113 indicated with variable density values are prepared. Then, the prepared failure distribution image data is displayed on a display means, and a condition for generating failures on the semiconductor substrate is grasped based on the displayed failure distribution image data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for inspecting a semiconductor substrate such as a semiconductor wafer based on inspection data such as foreign matter inspection, appearance defect inspection, alignment inspection, film thickness inspection, electrical inspection and dimensional inspection. The present invention relates to a method for manufacturing a semiconductor substrate manufactured by performing analysis, quality control, and the like.

[0002]

2. Description of the Related Art As a prior art relating to a method of manufacturing a semiconductor substrate by performing a failure analysis, a quality control and the like, a semiconductor substrate is disclosed in Japanese Patent Application Laid-Open Nos. 5-44006 and 6-9.
The contents described in Japanese Patent Publication No. 915 are known. Also,
As a conventional technique as a support system for analyzing an abnormal situation in a method of manufacturing a semiconductor substrate, "Monthly Semiic"
producer World 1996.8 (p
p. 79-105)) are known. According to this conventional technique, a symbol, a color, and hatching are designated for each foreign substance and each defect, the distribution of the defects is displayed for each semiconductor wafer, and an approximate defect distribution can be analyzed.

[0003]

However, in the prior art in which a symbol, color, or hatching is designated for each foreign substance and each defect, the distribution of the defect is displayed for each semiconductor wafer, and the approximate defect distribution is analyzed, When multiple semiconductor wafers and many defects are displayed in a superimposed manner in order to analyze every day, daily trend, each manufacturing process, each manufacturing facility, etc., there is a defect in the entire semiconductor wafer. However, this method is not suitable for observing the tendency of defects due to color irregularities and the like, which makes analysis difficult. Further, in a conventional support system for analyzing an abnormal situation, since human intervention is always required, real-time abnormality detection cannot be performed, and a defect has to be created. Further, in the related art, there has not been sufficiently considered that an abnormal state is automatically monitored and detected, and the abnormal state is automatically transmitted to a related person.

[0004] An object of the present invention is to solve the above problems.
An abnormal situation can be found at an early stage from defect data such as foreign matter defects and appearance defects obtained by inspecting a semiconductor substrate manufactured on a production line with an inspection device, and the production of defects can be significantly reduced to improve the semiconductor substrate. It is an object of the present invention to provide a method of manufacturing a semiconductor substrate which can be manufactured with high yield and high quality. Another object of the present invention is to easily estimate a cause of a defect from defect data such as a foreign matter defect and an appearance defect obtained by inspecting a semiconductor substrate manufactured on a production line with an inspection device, It is an object of the present invention to provide a method of manufacturing a semiconductor substrate capable of manufacturing a semiconductor substrate with high yield and high quality by remarkably reducing the production of defects.

[0005]

In order to achieve the above object, the present invention provides a method of manufacturing a semiconductor substrate through a plurality of manufacturing steps, the method comprising the steps of: An inspection step of inspecting a position of a defect occurring on each semiconductor substrate with respect to the semiconductor substrate by an inspection device, and position data of the defect on each semiconductor substrate inspected in the inspection step are set on the semiconductor substrate. A failure distribution image data creating step of designating coordinates on image data composed of lattice pixels and adding the number of defects for each lattice pixel on a plurality of semiconductor substrates on the image data to create failure distribution image data And a failure occurrence state grasping step of grasping a failure occurrence state on the semiconductor substrate based on the failure distribution image data created in the failure distribution image data creation step. Is a semiconductor substrate manufacturing method characterized by. Further, according to the present invention, in a semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by a manufacturing line including a plurality of manufacturing apparatuses, a plurality of semiconductor substrates manufactured by the desired manufacturing apparatus are generated on each semiconductor substrate. Inspection step of inspecting the position of the defect with an inspection device, and position data of the defect on each semiconductor substrate inspected in the inspection step,
Coordinate designation is performed on image data composed of lattice-shaped pixels set on the semiconductor substrate, and the number of defects is added for each of the lattice-shaped pixels on the image data for a plurality of semiconductor substrates, and defect distribution image data And a failure occurrence state grasping step of grasping a failure occurrence state on the semiconductor substrate based on the failure distribution image data created in the failure distribution image data creation step. This is a method for manufacturing a semiconductor substrate.

Further, the present invention relates to a method of manufacturing a semiconductor substrate, which manufactures a semiconductor substrate through a plurality of manufacturing steps, wherein a plurality of semiconductor substrates manufactured in the desired manufacturing step are free of defects occurring on each semiconductor substrate. Inspection step of inspecting the position with an inspection apparatus, and specifying the position data of a defect on each semiconductor substrate inspected in the inspection step on image data composed of grid-like pixels set on the semiconductor substrate. A defect distribution image data generating step of adding the number of defects for each of the grid-like pixels on the plurality of semiconductor substrates on the image data to generate defect distribution image data represented by grayscale values; The defect distribution image data represented by the grayscale values created in the creation step is displayed on the display means, and the semiconductor substrate is determined based on the defect distribution image data represented by the displayed grayscale values. A method of manufacturing a semiconductor substrate and having a failure condition grasping step of grasping the state of occurrence of failure in. Further, according to the present invention, in a semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by a manufacturing line including a plurality of manufacturing apparatuses, a plurality of semiconductor substrates manufactured by the desired manufacturing apparatus are generated on each semiconductor substrate. An inspection step of inspecting the position of the defect with an inspection apparatus, and the position data of the defect on each semiconductor substrate inspected in the inspection step is placed on image data composed of grid-like pixels set on the semiconductor substrate. A defect distribution image data generating step of specifying coordinates and adding the number of defects for each of the lattice-shaped pixels on the coordinate-specified image data for a plurality of semiconductor substrates to generate defect distribution image data represented by grayscale values; Displaying the defect distribution image data indicated by the gray value created in the defect distribution image data creating step on the display means, and displaying the defect distribution image indicated by the displayed gray value. A semiconductor substrate manufacturing method characterized by having a failure state grasping step to grasp the occurrence of the failure in the semiconductor substrate on the basis of the data.

Further, the present invention provides a method of manufacturing a semiconductor substrate in which a semiconductor substrate is manufactured through a plurality of manufacturing steps, wherein a plurality of semiconductor substrates in lot units manufactured in the desired manufacturing step are formed on each semiconductor substrate. An inspection step of inspecting the position of the generated defect with an inspection apparatus, and the position data of the defect on each semiconductor substrate, which has been inspected in the inspection step, is converted into image data comprising grid-like pixels set on the semiconductor substrate. A defect distribution image data generating step of specifying the coordinates on the upper side and adding the number of defects for each lattice-shaped pixel on the image data for a plurality of semiconductor substrates in lot units to generate defect distribution image data in lot units And a defect occurrence for grasping a defect occurrence state on the semiconductor substrate based on the defect distribution image data in lot units created in the defect distribution image data creation step. Is a manufacturing method of a semiconductor substrate characterized by having a state grasping step.

The present invention also relates to a method of manufacturing a semiconductor substrate, wherein the semiconductor substrate is manufactured by a manufacturing line including a plurality of manufacturing apparatuses, wherein a plurality of semiconductor substrates in lot units manufactured by the desired manufacturing apparatus are provided. An inspection step of inspecting the position of a defect that has occurred on each semiconductor substrate by an inspection device; A defect in which coordinates are specified on image data composed of pixels, and the number of defects for each lattice-shaped pixel is added for a plurality of semiconductor substrates in lot units on the image data to create defect distribution image data in lot units A distribution image data generating step, and generation of a defect on the semiconductor substrate based on the defect distribution image data in lot units generated in the defect distribution image data generating step. A method of manufacturing a semiconductor substrate and having a failure condition grasping step of grasping the state.

Further, the present invention relates to a method of manufacturing a semiconductor substrate, wherein the semiconductor substrate is manufactured through a plurality of manufacturing steps, wherein the plurality of semiconductor substrates in lot units manufactured in the desired manufacturing step are formed on each semiconductor substrate. An inspection step of inspecting the position of the generated defect with an inspection apparatus, and the position data of the defect on each semiconductor substrate, which has been inspected in the inspection step, is converted into image data comprising grid-like pixels set on the semiconductor substrate. A defect in which the coordinates are designated above, and the number of defects for each pixel in a grid on the image data is added for a plurality of semiconductor substrates in lot units to create defect distribution image data indicated by shading values in lot units Distribution image data creation process,
The failure distribution image data represented by the gradation value in lot units created in the failure distribution image data creation step is displayed on the display means, and the defect distribution image data represented by the gradation values represented in the displayed lot units is displayed. And a failure occurrence state grasping step of grasping a failure occurrence state on the semiconductor substrate. Also,
The present invention is directed to a method of manufacturing a semiconductor substrate, wherein a semiconductor substrate is manufactured by a manufacturing line including a plurality of manufacturing apparatuses, wherein a plurality of semiconductor substrates in lot units manufactured by the desired manufacturing apparatus are provided on each semiconductor substrate. An inspection step of inspecting the position of the generated defect with an inspection apparatus, and the position data of the defect on each semiconductor substrate, which has been inspected in the inspection step, is converted into image data comprising grid-like pixels set on the semiconductor substrate. Defect distribution image data designated by the coordinates above, and adding the number of defects for each lattice-like pixel for a plurality of semiconductor substrates in lot units on the coordinate-designated image data and indicating the gradation value in lot units A failure distribution image data creating step of creating a failure distribution image data, and displaying the failure distribution image data represented by the grayscale values in lot units created in the failure distribution image data creation step. A failure occurrence state grasping step of grasping the occurrence state of a failure on the semiconductor substrate based on the failure distribution image data indicated by the displayed gradation value in lot units. It is a manufacturing method of.

Further, according to the present invention, there is provided a semiconductor substrate manufacturing method for manufacturing a semiconductor substrate through a plurality of manufacturing steps, wherein a plurality of semiconductor substrates manufactured in the desired manufacturing step are provided with a defect generated on each semiconductor substrate. Inspection step of inspecting the position with an inspection apparatus, and specifying the position data of a defect on each semiconductor substrate inspected in the inspection step on image data composed of grid-like pixels set on the semiconductor substrate. A defect distribution image data creating step of adding the number of defects for each of the lattice-shaped pixels for the plurality of semiconductor substrates on the image data to create defect distribution image data, and a defect distribution image data creation step. A defect distribution characteristic amount calculated from the defect distribution image data obtained, and a defect occurrence state on the semiconductor substrate is grasped based on the calculated defect distribution characteristic amount. Is a manufacturing method of a semiconductor substrate characterized by having a state grasping step. Further, according to the present invention, in a semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by a manufacturing line including a plurality of manufacturing apparatuses, a plurality of semiconductor substrates manufactured by the desired manufacturing apparatus are generated on each semiconductor substrate. An inspection step of inspecting the position of the defect with an inspection apparatus, and the position data of the defect on each semiconductor substrate inspected in the inspection step is placed on image data composed of grid-like pixels set on the semiconductor substrate. A defect distribution image data generating step of specifying coordinates and adding the number of defects for each of the lattice-shaped pixels on the plurality of semiconductor substrates on the image data to generate defect distribution image data; and The feature amount of the failure distribution is calculated from the created failure distribution image data, and the occurrence state of the failure on the semiconductor substrate is grasped based on the calculated feature amount of the failure distribution. A method of manufacturing a semiconductor substrate and having a failure status determination step of.

The present invention also relates to a method of manufacturing a semiconductor substrate, wherein the semiconductor substrate is manufactured through a plurality of manufacturing steps. Inspection step of inspecting the position with an inspection apparatus, and specifying the position data of a defect on each semiconductor substrate inspected in the inspection step on image data composed of grid-like pixels set on the semiconductor substrate. A defect distribution image data creating step of adding the number of defects for each of the lattice-shaped pixels for the plurality of semiconductor substrates on the image data to create defect distribution image data, and a defect distribution image data creation step. A failure analysis step of collating and analyzing the failure distribution image data prepared with a plurality of prepared case databases capable of estimating a failure occurrence cause to determine a failure occurrence cause. A method for manufacturing a semiconductor substrate according to claim Rukoto. Further, according to the present invention, in a semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by a manufacturing line including a plurality of manufacturing apparatuses, a plurality of semiconductor substrates manufactured by the desired manufacturing apparatus are generated on each semiconductor substrate. An inspection step of inspecting the position of the defect with an inspection apparatus, and the position data of the defect on each semiconductor substrate inspected in the inspection step is placed on image data composed of grid-like pixels set on the semiconductor substrate. A defect distribution image data generating step of specifying coordinates and adding the number of defects for each of the lattice-shaped pixels on the plurality of semiconductor substrates on the image data to generate defect distribution image data; and A defect solution that collates and analyzes the created defect distribution image data with a prepared case database that can estimate the cause of the defect to determine the cause of the defect. Is a manufacturing method of a semiconductor substrate, characterized by a step.

Further, the present invention provides a method of manufacturing a semiconductor substrate, wherein the semiconductor substrate is manufactured through a plurality of manufacturing steps. Inspection step of inspecting the position with an inspection apparatus, and specifying the position data of a defect on each semiconductor substrate inspected in the inspection step on image data composed of grid-like pixels set on the semiconductor substrate. A defect distribution image data creating step of adding the number of defects for each of the lattice-shaped pixels for the plurality of semiconductor substrates on the image data to create defect distribution image data, and a defect distribution image data creation step. The characteristic amount of the defect distribution is calculated from the defect distribution image data, and the calculated characteristic amount of the defect distribution is displayed on the display means. A method of manufacturing a semiconductor substrate and having a failure condition grasping step of grasping the state of occurrence of defects on the body board. Also, the present invention
In a semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by a manufacturing line including a plurality of manufacturing apparatuses, an inspection apparatus is provided for inspecting a position of a defect occurring on each semiconductor substrate for a plurality of semiconductor substrates manufactured by the desired manufacturing apparatus. Inspection step of inspecting, and the position data of a defect on each semiconductor substrate inspected in the inspection step are designated by coordinates on image data composed of grid-like pixels set on the semiconductor substrate, A defect distribution image data creating step of adding the number of defects for each of the grid-shaped pixels for the plurality of semiconductor substrates on the data to create defect distribution image data; and a defect distribution image created in the defect distribution image data creating step. Calculating a feature amount of the failure distribution from the data; displaying the calculated feature amount of the failure distribution on a display unit; A method of manufacturing a semiconductor substrate and having a failure condition grasping step of grasping the state of occurrence of defects on the plate.

Further, according to the present invention, there is provided a semiconductor substrate manufacturing method for manufacturing a semiconductor substrate through a plurality of manufacturing steps, wherein a plurality of semiconductor substrates manufactured in the desired manufacturing step are provided with a defect generated on each semiconductor substrate. Inspection process of inspecting the position with an inspection device, and setting of the defect position data on each semiconductor substrate inspected in the inspection process by dividing and arranging a plurality of exposure shot units or chip units on the semiconductor substrate. The coordinates are specified on the shot-unit or chip-unit image data composed of the divided lattice-shaped pixels, and the number of defects for each lattice-shaped pixel is added for the plurality of semiconductor substrates on the shot-unit or chip-unit image data. A failure distribution image data creating step of creating failure distribution image data indicated by a gray scale value in a shot unit or a chip unit; The failure distribution image data represented by the gray value of the shot unit or the chip unit created in the distribution image data creation step is displayed on the display means, and based on the displayed failure distribution image data represented by the gray value of the shot unit A failure occurrence state grasping step of grasping a state of occurrence of a failure on the semiconductor substrate.

Further, the present invention provides a method of manufacturing a semiconductor substrate, wherein the semiconductor substrate is manufactured by a manufacturing line including a plurality of manufacturing apparatuses. Inspection step of inspecting the position of the defect occurred in the inspection device, and the position data of the defect on each semiconductor substrate inspected in the inspection step,
Coordinates are specified on shot-unit or chip-unit image data composed of grid-like pixels that are set by dividing into a plurality of exposure shot-units or chip-units. A defect distribution image data generating step of adding the number of defects for each of the pixels for a plurality of semiconductor substrates to generate defect distribution image data represented by a gray value in a shot unit or a chip unit; Displaying the defect distribution image data represented by the grayscale value in the shot unit or the chip unit created in the step (c) on the display means, and based on the displayed defect distribution image data represented by the grayscale value in the shot unit or the chip unit; A failure occurrence state grasping step of grasping a failure occurrence state on the above. It is a manufacturing method of the body substrate. Further, in the present invention, in the method of manufacturing a semiconductor substrate, a case database for analyzing and monitoring a defect distribution (regional defect) on the semiconductor substrate is an image database. Further, according to the present invention, in the method for manufacturing a semiconductor substrate, a failure distribution image created in the failure distribution image data creating step may include a case database for analyzing and monitoring a failure distribution (regional failure) on the semiconductor substrate. It is created based on data. The present invention also provides the semiconductor substrate manufacturing method, wherein each case database for analyzing and monitoring a defect distribution (regional defect) on the semiconductor substrate includes a defect countermeasure method, a defect manufacturing process, or a defect manufacturing facility. And the like are added.

As described above, according to the above-described configuration, in the method of manufacturing a semiconductor substrate, each desired manufacturing process, each desired manufacturing apparatus (manufacturing equipment), each production lot, etc.
It is possible to grasp the tendency of a defect distribution (defect area characteristic) such as a foreign substance defect or an appearance defect occurring on a semiconductor substrate on image data composed of lattice-like pixels set on the semiconductor substrate, As a result, an abnormal situation can be found at an early stage, the production of defects can be significantly reduced, and the semiconductor substrate can be manufactured with high yield and high quality. Further, according to the above configuration, in the method of manufacturing a semiconductor substrate, defects such as foreign matter defects and appearance defects occurring on the semiconductor substrate are generated for each desired manufacturing process, each desired manufacturing apparatus (manufacturing equipment), each production lot, and the like. By grasping the tendency of distribution (regional nature of defects) on image data composed of grid-like pixels set on the semiconductor substrate, regional characteristics such as concentration of failures and uniform scattering of failures Defects can be analyzed and monitored, leading to early detection of abnormal conditions and early countermeasures. Manufacturing of semiconductor substrates with a high yield and high quality by significantly reducing defects. Can be.

Further, according to the above configuration, in the method of manufacturing a semiconductor substrate, a foreign substance defect or an appearance defect generated on the semiconductor substrate for each desired manufacturing step, each desired manufacturing apparatus (manufacturing equipment), each production lot, or the like. By visually grasping the tendency of failure distribution (regional nature of failures), it is possible to analyze and monitor regionality failures such as concentration of failures and uniform scattering of failures. This can be linked to early detection of an abnormal state and early countermeasures, and the production of defects can be significantly reduced, so that semiconductor substrates can be manufactured with high yield and high quality. Further, according to the above configuration, in the method of manufacturing a semiconductor substrate, defects such as foreign matter defects and appearance defects occurring on the semiconductor substrate are generated for each desired manufacturing process, each desired manufacturing apparatus (manufacturing equipment), each production lot, and the like. The tendency of the distribution (regional nature of the defect) can be grasped on image data composed of lattice-like pixels set on the semiconductor substrate, and as a result, the cause of the defect can be easily estimated, and As a countermeasure can be taken, the production of defects can be significantly reduced, and the semiconductor substrate can be manufactured with high yield and high quality.

Further, according to the above configuration, in the method of manufacturing a semiconductor substrate, a foreign substance defect or an appearance defect generated on the semiconductor substrate may be generated for each desired manufacturing process, each desired manufacturing apparatus (manufacturing equipment), each production lot, or the like. The tendency of the defect distribution (defect area characteristic) such as can be grasped on the image data composed of the lattice-like pixels set on the semiconductor substrate,
As a result, defective manufacturing processes, defective manufacturing equipment, defective lot units, etc. can be specified, so that the cause of failure can be easily estimated, countermeasures can be taken at an early stage, and the production of defects can be significantly reduced. The substrate can be manufactured with high yield and high quality. Further, according to the configuration, in the method of manufacturing a semiconductor substrate, in addition to the regional defect, the abnormal state of various analysis items such as the standard value management of the inspection data, the time-series trend change analysis, and the machine-by-device difference analysis is automatically determined. By monitoring and notifying related parties in real time, it is possible to link to early detection of abnormal conditions and early countermeasures, remarkably reduce the production of defective products, and manufacture semiconductor substrates with high yield and high quality. be able to.
Further, according to the above configuration, in a method of manufacturing a semiconductor substrate, a failure distribution can be visually observed for each desired manufacturing process, each desired manufacturing facility, and each production lot, so that a person concerned can understand and understand the cause of the failure. It will be possible to help the investigation. Further, according to the configuration, in the method of manufacturing a semiconductor substrate, by registering a past case database on the distribution of defects, it is possible to easily estimate the cause of the failure,
Measures can be taken early.

[0018]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described with reference to the drawings. First, the principle of a semiconductor failure analysis method in a semiconductor manufacturing method according to the present invention will be described with reference to FIGS. A semiconductor wafer has a number of manufacturing steps (for example, a step of forming a device on a substrate made of Si or the like (mainly, an oxidation diffusion step, a photolithography step, and an electrode forming step)), and a multilayer wiring layer formed thereon. The process is roughly divided into a process and a process. FIG. 1 shows manufacturing processes A81 and B, which are representative of a number of manufacturing processes in which a foreign substance inspection or an appearance inspection is performed.
82 is a diagram for explaining a method of determining a manufacturing process in which a defect occurs based on the defect distribution image data or the defect distribution grayscale image data obtained for each of C82 and C83.

Reference numeral 101 denotes each semiconductor wafer extracted (sampled) in lot units, for example, or extracted (sampled) entirely for a specific lot from a large number of semiconductor wafers manufactured in the manufacturing process A81. The position of the center of gravity of each defect (foreign matter, pattern defect, etc.) inspected by the inspection device 91 (41, 42) is determined by specifying coordinates on grid image data as shown in FIG. a (showing a value corresponding to the number of defects existing in the same lattice-shaped pixel). That is, the foreign matter / appearance collection system 51 shown in FIG.
Is an inspection apparatus 91 for each semiconductor wafer sampled (sampled) in lot units or 100% sampled (sampled) for a specific lot from a large number of semiconductor wafers manufactured in the manufacturing process A81. The position of the center of gravity of each defect (foreign matter, pattern defect, etc.) obtained by inspection at (41, 42) is extracted, and the position of the center of gravity of each of the extracted defects is determined as shown in FIG. Defective position data a (representing a value corresponding to the number of defects existing in the same lattice-shaped pixel) 101 is determined by specifying coordinates on the upper side. Next, the foreign matter / appearance collection system 51 performs the defect position data a101 for each semiconductor wafer extracted in the manufacturing process A81.
4 and 5, a value (a value indicating the number of defects) is added for each pixel indicating the same coordinate position over a plurality of semiconductor wafers, so that the defect distribution image data in the manufacturing process A81 can be obtained. Is converted to defect distribution density image data 111 by converting a numerical value of each pixel in the defect distribution image data into a gray value, and the converted defect distribution gray image is generated. Data 111 is displayed on display means 51a such as a display.
Or it outputs to 61a.

Reference numeral 102 denotes each semiconductor wafer extracted (sampled) in lot units, for example, or extracted (sampled) entirely for a specific lot from a large number of semiconductor wafers manufactured in the manufacturing process B82. As shown in FIG. 3, the position of the center of gravity of each defect (foreign matter, pattern defect, etc.) inspected by the inspection device 92 is designated as defective position data a (the same lattice shape) obtained by specifying coordinates on lattice image data. A value corresponding to the number of defects existing in the pixel is indicated.). However, the defect position data a101 is excluded from the defect position data b102. That is, the foreign matter / appearance collection system 51 shown in FIG. 13 is sampled (sampled), for example, in lot units, or 100% in a specific lot from many semiconductor wafers manufactured in the manufacturing process A81 ( The centroid position of each defect (such as a foreign substance and a pattern defect) obtained by inspecting each sampled semiconductor wafer in the inspection device 92 (41, 42) is extracted, and the centroid position of each extracted defect is calculated as As shown in FIG. 3, defective position data a (indicating a value corresponding to the number of defects existing in the same grid-like pixel) 102 is obtained by specifying coordinates on the grid-like image data. Since the inspection apparatus 92 (41, 42) may detect the defective position data a101, it is necessary to delete the defective position data a101 from the defective position data b102. Next, the foreign matter / appearance collection system 51 compares the defect position data b102 for each semiconductor wafer extracted in the manufacturing process B82 over the plurality of semiconductor wafers with the same coordinate position as shown in FIGS. By adding a value (a value indicating the number of defects) for each pixel shown, failure distribution image data in the manufacturing process B82 is created, and the created failure distribution image data is used for each pixel in the failure distribution image data. By converting the numerical value into a gray value, it is converted into the defective distribution gray image data 112,
The converted defect distribution density image data 112 is output to the display means 51a or 61a such as a display.

Reference numeral 103 denotes each semiconductor wafer sampled (sampled) in lot units from a large number of semiconductor wafers manufactured in the manufacturing process C83, or all sampled (sampled) for a specific lot. As shown in FIG. 3, the position of the center of gravity of each of the inspected defects (foreign matter, pattern defect, etc.) inspected by the inspection device 93 (41, 42) can be obtained by designating the coordinates on the grid-like image data. Defect position data c
(A value corresponding to the number of defects existing in the same lattice-shaped pixel is shown.) However, the defect position data a101 and b102 are excluded from the defect position data c103. That is, the foreign matter / appearance collection system 51 shown in FIG.
Is an inspection device 93 for each semiconductor wafer sampled (sampled) in lot units, or 100% sampled (sampled) for a specific lot, from a large number of semiconductor wafers manufactured in the manufacturing process C83. The position of the center of gravity of each defect (foreign matter, pattern defect, etc.) obtained by inspection at (41, 42) is extracted, and the position of the center of gravity of each defect for each of the extracted semiconductor wafers is determined as shown in FIG. Defective position data c (indicating a value corresponding to the number of defects existing in the same lattice-like pixel) 103 is obtained by specifying coordinates on the lattice-like image data. Since the defect position data a101 and b102 may be detected from the inspection device 92 (41, 42), the defect position data a101 and b102 must be deleted from the defect position data c103. Next, the foreign matter / appearance collection system 51 compares the defective position data c103 for each semiconductor wafer extracted in the manufacturing process C83 with the same coordinate position over a plurality of semiconductor wafers as shown in FIGS. By adding a value (a value indicating the number of defects) for each pixel indicated, the manufacturing process C
The failure distribution image data in 83 is created, and the created failure distribution image data is converted into failure distribution gray image data 113 by converting a numerical value for each pixel in the failure distribution image data into a gray value. The obtained defect distribution density image data 113 is output to the display means 51a or 61a such as a display.

As described above, when the defect distribution image data or the defect distribution density image data 111, 112 and 113 created for each manufacturing process are compared, the defect distribution image data or the defect distribution density image data 111 includes: It can be seen that the failures are concentrated on the upper right and that the failure distribution image data or the failure distribution density image data 113 has a linear failure on the upper left. Each of the defect distribution image data or the defect distribution density image data 111 and 113 indicates a defect distribution in the semiconductor wafer, that is, a region where a defect occurs. Then, by finding a defect that is not detected in such a certain manufacturing process but is detected in a certain manufacturing process, the manufacturing process that causes the occurrence of the defect can be identified. That is, in the foreign matter / appearance collection system 51 shown in FIG. 13, each of the defect distribution image data created for each manufacturing process is converted into a numerical value for each pixel in each of the defect distribution image data into a gray scale value. The image data is converted into gray-scale image data 111, 112, and 113, and these converted data 111, 112, and 113 are directly converted into foreign matter /
By providing the display means 51a such as a display connected to the appearance collection system 51 or the abnormality monitoring system 61 and outputting the same to the display means 61a such as a display connected to the abnormality monitoring system 61, manufacturing which causes a defect may be generated. The process can be identified.

FIG. 2 shows a manufacturing apparatus A among a number of manufacturing steps.
In a certain manufacturing process in which a foreign substance inspection or an appearance inspection is performed on a semiconductor wafer manufactured using the semiconductor device 84 and the manufacturing apparatus B85, the defective distribution image data or the defective distribution grayscale image obtained for each of the manufacturing apparatuses A84 and B85. FIG. 7 is a diagram for explaining a method for determining a manufacturing apparatus in which a defect occurs based on data. Reference numeral 104 denotes, for example, a lot (sampled) or a whole lot of a specific lot (sampled) from a large number of semiconductor wafers manufactured by the manufacturing apparatus A84.
Inspection device 94 (41, 42) for each semiconductor wafer
As shown in FIG. 3, the position of the center of gravity of each defect (for example, a foreign matter and a pattern defect) inspected in the defect position data d (existing in the same lattice pixel The value corresponds to the number of defects to be formed.). Before entering manufacturing apparatus A84 and manufacturing apparatus A8
4, the inspection device 94 (41, 42) inspects the occurrence of a defect existing on each semiconductor wafer, and the production device A 8
4 from the defect position data obtained after exiting from step 4.
By erasing the defective position data obtained before entering the position 84, the defective position data generated by the manufacturing apparatus A84 can be obtained. That is, foreign matter / appearance collection system 5
1 is an inspection apparatus for each semiconductor wafer extracted (sampled) in lot units, for example, or extracted (sampled) in total for a specific lot from a large number of semiconductor wafers manufactured in the manufacturing apparatus A84. The position of the center of gravity of each defect (for example, foreign matter and pattern defect) obtained by inspection at 94 (41, 42) is extracted, and the position of the center of gravity of each of the extracted defects is extracted as shown in FIG. Defective position data d (indicating a value corresponding to the number of defects existing in the same lattice-shaped pixel) 104 is obtained by specifying coordinates on the data. Next, the foreign matter / appearance collection system 51 performs the defect position data d10 for each semiconductor wafer extracted in the manufacturing process A81.
4 over a plurality of semiconductor wafers, as shown in FIGS. 4 and 5, by adding a value (a value indicating the number of defects) for each pixel indicating the same coordinate position, thereby obtaining a defect distribution image in the manufacturing apparatus A84. Data is created, and the created defect distribution image data is converted into defect distribution density image data 114 by converting a numerical value for each pixel in the defect distribution image data into a gray value, and these converted defect distribution density data are converted. The image data 114 is displayed on a display unit 51 such as a display.
a or 61a.

Reference numeral 105 denotes each semiconductor wafer extracted (sampled) in lot units from a large number of semiconductor wafers manufactured by the manufacturing apparatus B85, or extracted (sampled) in total for a specific lot. The position of the center of gravity of each defect (foreign matter, pattern defect, etc.) inspected by the inspection apparatus 95 (41, 42) is determined by specifying coordinates on grid-like image data as shown in FIG. e (indicating a value corresponding to the number of defects existing in the same lattice-shaped pixel). Before entering the manufacturing apparatus B85 and after exiting the manufacturing apparatus B85, the inspection apparatus 95 checks the occurrence of defects existing on each semiconductor wafer.
Inspection is performed at (41, 42), and the defective position data obtained before entering the manufacturing apparatus B85 is deleted from the defective position data obtained after leaving the manufacturing apparatus B85, thereby obtaining the defective position data generated by the manufacturing apparatus B85. be able to. That is, the foreign matter / appearance collection system 51 is extracted (sampled), for example, in lot units from a large number of semiconductor wafers manufactured in the manufacturing apparatus B85.
Alternatively, the inspection apparatus 9 for each semiconductor wafer sampled (sampled) for a specific lot.
5 (41, 42), the center of gravity of each defect (foreign matter, pattern defect, etc.) obtained by inspection is extracted, and the center of gravity of each of the extracted defects is determined as shown in FIG.
Defective position data e by specifying coordinates on grid-like image data
(Representing a value corresponding to the number of defects existing in the same lattice-shaped pixel.) 105 is obtained. Next, the foreign matter / appearance collection system 51 compares the defect position data e105 for each semiconductor wafer extracted in the manufacturing process A81 over a plurality of semiconductor wafers with the same coordinate position as shown in FIGS. By adding a value (a value indicating the number of defects) for each pixel shown, failure distribution image data in the manufacturing apparatus B85 is created, and the created failure distribution image data is used for each pixel in the failure distribution image data. By converting a numerical value into a gray value, a defective distribution gray image data 1 is obtained.
And converting the defective distribution density image data 115 into a display means 51a or 6 such as a display.
1a.

As described above, when comparing the defect distribution image data or the defect distribution density image data 114 and 115 created for each manufacturing apparatus, it is found that the defect distribution image data or the defect distribution density image data 115 It can be seen that the defects are distributed in a ring shape. As described above, the defect distribution image data or the defect distribution grayscale image data 115 indicates the defect distribution in the semiconductor wafer, that is, the area in which the defect occurs. Then, by finding a defect that is not detected from such a certain manufacturing apparatus but is detected from a certain manufacturing apparatus, it is possible to determine a manufacturing apparatus that causes a defect. That is,
In the appearance collecting system 51, each of the defect distribution image data created for each manufacturing apparatus is converted into defect distribution gray image data 114, 115 by converting a numerical value for each pixel in each defect distribution image data into a gray value. The converted data 114 and 115 are provided to a display means 51a such as a display directly connected to the foreign matter / appearance collection system 51 or to the abnormality monitoring system 61, and are displayed on a display connected to the abnormality monitoring system 61. By outputting the data to the means 61a, it is possible to identify the manufacturing apparatus that causes the failure.

Although FIG. 2 shows each apparatus, even if the same apparatus is used, similar processing is performed for each chamber.
The chamber that causes the failure can be determined. In addition, the inspection devices 91 to 95 shown in FIGS.
Has been described using separate inspection devices, but the same inspection device may be used. Inspection device 9
When 1 to 95 are configured by separate inspection devices, they can be incorporated in a semiconductor manufacturing line and used as an ON-line monitor. As described above, as shown in FIG. 1 and FIG. 2, the defect distribution density image data 11 obtained by the manufacturing process A81 for each defect category (foreign matter defect, appearance pattern defect, etc.)
1 and defective distribution density image data 1 by the manufacturing process C83
A different defect distribution pattern (indicating the area where a defect has occurred in a semiconductor wafer) is shown between 13 and the defect distribution density image data 115 by the manufacturing apparatus B85. Therefore, if these different failure distribution patterns are registered as the past case database (template database) 121, the abnormality monitoring system 61 can specify a manufacturing process and a manufacturing apparatus that cause a failure. By the way,
A semiconductor wafer is manufactured by a number of manufacturing processes, and in a certain manufacturing process, it is manufactured using a plurality of manufacturing apparatuses. The flow of semiconductor wafers, including lots for the production line, is managed by the semiconductor production management system 71 shown in FIG. 13, and the semiconductor production management system 71 stores a database of the flow of semiconductor wafers for the production line. Will be. Therefore, the abnormality monitoring system 61 performs the failure analysis for each manufacturing process shown in FIG. 1 and the failure analysis for each manufacturing apparatus shown in FIG. 2 based on the database of the flow of the semiconductor wafer for the manufacturing line obtained from the semiconductor manufacturing management system 71. Need to be mixed with the above.

Next, a method of creating defect distribution image data from the defect position data 101 to 105 shown in FIGS. 1 and 2 will be described with reference to FIG. Reference numeral 21 denotes coordinate data of a defect inspection result inspected by each of the inspection devices 91 to 95 (41, 42). The coordinate data 21 of the defect inspection result includes an X coordinate and a Y coordinate from a reference position indicating the center of gravity of a defect (defective) such as a foreign substance or a pattern defect, and a size indicated by the area of the defect. The foreign matter / appearance collection system 51 includes the inspection devices 91 to 91.
The X coordinate and the Y coordinate from the reference position indicating the center of gravity of the defect (defective) in the coordinate data 21 of the defect inspection result inspected and input from 95 (41, 42) are obtained by dividing the diameter of the semiconductor wafer by N. That is, quantization into N × N pixels (one pixel is 100 μm × 100 μm to 500 μm × 5
Divide to a size of 00 μm. ) Is converted to correspond to which pixel position in the image data 13 constituted by. First, in the image data of all pixel values 0, when one defect (defect) is present in one pixel, the pixel value 1 is added to the one pixel, and a plurality of defects are generated in the same pixel. When a (defect) occurs, the number of defects (the number of defects) is added to the pixel. As a result, the image data 13 indicates the number of defects (the number of defects) for each pixel. That is, image data 1 composed of N × N pixels
As 3, the numerical value shown for each pixel indicates the number of defects (the number of defects) detected at that pixel. Therefore, the image data 13 composed of the N × N pixels
Does not include information on the size indicated by the area of the defect or the like. However, the size of a foreign matter defect can be sufficiently included in one pixel, and it is sufficient that the number of generated foreign matter defects can be grasped. Also, as for pattern defects, most of the causes are caused by foreign matters, and therefore, as with the foreign matter defects, the size is sufficiently included in one pixel, and it is only necessary to know the number of occurrences. In this way, the foreign matter / appearance collection system 51 can create defect distribution image data based on the defect position data 101 to 105 as the image data 13 including N × N pixels. If the foreign substance / appearance collection system 51 displays the failure distribution image data on a display unit 51a such as a display connected to the foreign substance / appearance collection system 51 or a display unit 61a such as a display connected to the abnormality monitoring system 61. For this purpose, the image data 13 composed of N × N pixels is converted into N × N
By converting the numerical value of each pixel in the image data 13 composed of pixels into a gray value, the image data 13 is converted into defective distribution gray image data 31, which is displayed on the display means 51a or 61a.
The defect position data 101 to 10
5 can be displayed as a gray-scale image, and the tendency of the defect distribution can be visually grasped.

Next, as shown in FIGS. 1 and 2, the defect distribution image data 111 to
A specific example of conversion to 115 will be described with reference to FIG. In the foreign matter / appearance collection system 51, a plurality of semiconductors are obtained by adding the values of the group 14 of the defect distribution image data 13 of each semiconductor wafer shown in FIG. 3 for each corresponding pixel as shown in FIG. A defect distribution 15 having a value indicated by the number of occurrences of defects (defects) is obtained for each pixel over the wafer. Then, the foreign matter / appearance collection system 51 is N ×
The image data 15 composed of N pixels is converted into defect distribution density image data 32 by converting a numerical value for each pixel in the image data 15 composed of N × N pixels into a density value, and this is converted into display means such as a display. It is displayed on 51a or 61a. As a result, the distribution 15 of the number of defects (defects) occurring over a plurality of manufactured semiconductor wafers can be visually grasped as a grayscale image 32 for each manufacturing process and each manufacturing apparatus. Further, the pattern 22 based on the design data such as the chip configuration and the orientation flat on the semiconductor wafer is input to the foreign matter / appearance collection system 51 by using an input means 51b including a network connected to the CAD system 201 and a recording medium. The foreign matter / appearance collection system 51 displays the pattern 22 based on the input design data and the converted defect distribution density image data 32 on the display means 51a or 61a such as a display as indicated by 33. By doing so, it becomes possible to grasp the tendency of the chip configuration and the failure distribution. The design pattern is not shaded, but is displayed in color to distinguish it from the failure distribution, so that the display is easy to understand. As shown in FIG. 6, the chip configuration may not be displayed, and only the outer shape of the wafer may be displayed.

FIG. 5 is a diagram showing a case where the defect distribution image data 18 is created by adding the defect distribution image data 16 and 17 for two semiconductor wafers. That is, grid-like image data A (x (n), y (n)) 16 and B (x (n), y (n)) indicating the number (number) of occurrences of defects (defects) for each pixel
By adding the values of the pixels indicating the same coordinates (x (n), y (n)) of 17 based on the following (Equation 1), defects (defects) occurring over two semiconductor wafers ) Is generated as image data C (x (n), y (n)) 18. C (x (n), y (n)) = A (x (n), y (n)) + B (x (n), y (n)) (Equation 1) FIG. Display means 51 for displaying the data 32
a and 61a show examples shown in FIG. In FIG. 6, the defect distribution on the semiconductor wafer is treated as grid-like image data. The more defects in the same lattice, the brighter the display, and the fewer the defects, the darker the display. The defect distribution 11 has a rough grid and treats one chip as one pixel.
Has a fine lattice and handles one wafer with 128 × 128 pixels (about 1.6 mm per pixel). FIG. 6 shows an embodiment in which the number of defects is larger, the brightness is higher, and the number of defects is lower, the darker the display is. ,
It is valid. This selection may be adapted to the user's preference. In addition, the image size can be further reduced by making the grid thinner.
× 512 pixels (about 400 μm per pixel), 1024 × 10
It is also effective to use 24 pixels (one pixel is about 200 μm).

Next, as shown in FIGS. 3 and 4, defect distribution image data 13 detected from each semiconductor wafer manufactured by each manufacturing process or each manufacturing apparatus, and a plurality of defective distribution image data 13 manufactured by each manufacturing process or each manufacturing apparatus. FIGS. 7 and 8 show an embodiment of a method for automatically analyzing and monitoring an abnormality of a defect distribution in the foreign matter / appearance collection system 51 and the abnormality monitoring system 61 from the defect distribution image data 15 detected over the semiconductor wafer of FIG. It will be described using FIG. In FIG. 8, a solid line in a lattice formed on a semiconductor wafer indicates a chip. First, in step 131, the foreign matter / appearance collection system 51 detects the foreign matter from the foreign matter inspection device 41 and the appearance inspection device 42 for each semiconductor wafer or over a plurality of semiconductor wafers by the method shown in FIGS. The foreign matter map and the appearance map 141 (13 or 15) indicating the number of defects (defects such as foreign matter and pattern defects) for each pixel divided from the coordinate data are converted into image data, and the foreign matter and appearance map images are formed. 142 is obtained. That is, image data is generated for each defective category (foreign matter or appearance pattern defect). The foreign matter map and the appearance map 141 (13 or 15) are supplemented with information such as a number specifying the semiconductor wafer inspected by the foreign matter inspection device 41 and the appearance inspection device. Therefore, it is possible to identify from which semiconductor wafer the foreign matter map and the appearance map 141 are obtained. Next, in step 132, the abnormality monitoring system 61 binarizes the foreign matter and appearance map with a predetermined threshold (a threshold given by a predetermined number of defects (the number of defects) in one pixel) for the foreign matter and appearance map image 142. The binarized image of the converted foreign matter and appearance map is converted into an image, and a defect (defect) generated in an adjacent pixel is converted into one cluster (cluster) as shown in FIG. Then, the foreign matter and the appearance map dilated binary image 143 are created.

As the image expansion processing shown in FIG. 9, original image data 151 (14) represented by two-dimensional binary image data is used.
In contrast to 2), including its central pixel, its four neighboring pixels 154
In the above, the logical sum 155 is taken and the result is given to the central pixel. If at least one pixel “1” indicating a defect (defect) exists in the four neighboring pixels 154 including the central pixel, the central pixel As a result, "1" indicating a defect (defect) is added, and a signal "1" indicating a defect (defect) generated in a neighboring pixel is expanded and connected even if the pixel is separated by one pixel, thereby forming one lump (cluster). Thus, the image data 152 after the 4-neighbor expansion is obtained. Further, as the image expansion processing shown in FIG. 9, the OR operation is performed on the eight neighboring pixels 156 of the original image data 151 (142) represented by the two-dimensional binarized image data, including the center pixel.
By taking 7 and adding the result to the center pixel, including the center pixel, in its 8 neighboring pixels 156,
If at least one pixel indicating a defect (defect) exists, “1” indicating a defect (defect) is added as a central pixel, and “1” indicating a defect (defect) occurring in a neighboring pixel is assigned. Even if the signal “1” is separated by one pixel, the image data 153 is expanded and connected to form one lump (cluster), and the image data 153 is expanded after 8 neighborhoods. By repeating such image expansion processing a plurality of times, "1" indicating a defect (defect) is displayed.
Even if a plurality of signals are apart from each other by a plurality of pixels, they are expanded and connected, and image data in one lump (cluster) is obtained. In addition, uniform expansion can be performed by alternately performing near-four expansion and near-eight expansion.

As described above, in order to return to the original image while being connected to one cluster (cluster), image reduction processing may be performed. The four pixels 154 including the central pixel and the four neighboring pixels 154 or the central pixel are logically ANDed at the eight neighboring pixels 156 and the result is applied to the central pixel. The outermost one pixel in the clustered defective (defect) area is changed from a signal of "1" to a signal of "0" and reduced and connected to return in a state close to the original image data in a connected state. Become. In short, when a defect (defect) is detected in a nearby pixel,
The cause of the occurrence of the defect (defect) is considered to be the same, so that the image expansion processing is performed so as to form one block in order to facilitate the subsequent image processing. Further, in step 133, the foreign matter / appearance collection system 51 performs an image labeling process for applying labeling to each of the same clusters to create a labeling image 144. When performing this labeling process, a small block close to the noise component (for example, 3
Defective (defect) in 3x5 pixels or all 5x5 pixels
"1" indicating a defect (defect) in a small lump that is not determined to be defective, that is, in a state of being one pixel or two pixels in size before expansion, and in a state largely separated from other defects (defects).
(In the case of a small lump in which a certain signal is generated), a process of erasing is performed. Alternatively, the number of defects (the number of defects) in the same cluster may be counted, and the clusters of a predetermined value or less may be deleted. As a result, only the lump having a large defect (defect) is subjected to the labeling processing.

Next, in step 134, the foreign matter / appearance collection system 51 determines the representative position (representative coordinate) of each label as the feature extraction 145 of the data in each label processed by the same labeling, that is, the same cluster. ), The area indicating the size of each label, the x direction, y direction, r direction (radial direction) of each label,
the length in the θ direction, the perimeter of each label, the circularity of each label,
Calculate the number of defects in each label (number of defects) (shading value), moment of defect in each label, density of defects in each label (number of defects / area), dispersion value of defects in each label, It is transmitted to the monitoring system 61. That is, in step 134, the foreign matter / appearance collection system 51 determines in step 134 the feature amount (for example, the representative position (representative position) of each label) in each label for each defect category and for each semiconductor wafer or a plurality of semiconductor wafers. Coordinates) (center of gravity coordinates)
145, which indicates the size of each label, the number of defects, the number of defects, the density of defects, the variance of defects, etc.) 145, and the number identifying the semiconductor wafer inspected by the foreign matter inspection device 41 and the appearance inspection device 42, etc. The information is transmitted to the abnormality monitoring system 61 in a state where the information is supplemented.

The abnormality monitoring system 61 proceeds to step 135
In (2), a characteristic amount in each label (for example, a two-dimensional area or a three-dimensional area indicating the size of each label) for each semiconductor wafer or a plurality of semiconductor wafers extracted by the foreign matter / appearance collection system 51 Defects (number of defects)
It is determined whether or not the density of the defect (the number of defects / area) or the variance value from the position of the center of gravity of the defect 145 exceeds an allowable value. The information is output and an abnormality notification 62 is issued to the user or the administrator. Abnormal notification 6 to this user or administrator
2 is a display means 61 connected to the abnormality monitoring system 61
This may be performed by using an output means such as a, or by outputting to a terminal connected to the abnormality monitoring system 61 via a network. In step 135, the abnormality monitoring system 61 determines the center of gravity position of each label among the feature amounts in each label for each semiconductor wafer extracted by the foreign matter / appearance collection system 51 or for a plurality of semiconductor wafers ( The area where one block of defects (defects) has occurred on the semiconductor wafer can be recognized by the coordinates of the center of gravity, and the area, length, and number of defects (defects) indicating the size of each label among the feature amounts in each label. The two-dimensional size of a block in which a defect (defect) has occurred or the three-dimensional size including the number of defects (the number of defects) can be recognized based on the number or the like. In addition, the abnormality monitoring system 61
In step 135, the foreign matter / appearance collection system 51
From the center of gravity of the defect, the density of the defect (number of defects / area), and the variance 145 from the centroid of the defect, which are the characteristic amounts in each label for each semiconductor wafer or a plurality of semiconductor wafers extracted by As a result, it is possible to grasp the occurrence status of defects, and as a result, empirically identify the manufacturing process in which the defect (defect) has occurred, the manufacturing apparatus in which the defect (defect) has occurred, and estimate the cause of the defect (defect). It is also possible to do.

The processing shown in FIG. 7 described above is basically performed by the abnormality monitoring system 61 for each defect category (foreign matter defect or appearance pattern defect) extracted by the foreign matter / appearance collection system 51 and for each semiconductor. This is simply to determine an abnormality based on the feature amount in each label for each wafer or for a plurality of semiconductor wafers. Next, FIG.
As shown in FIG. 4 and FIG. 4, the defect distribution image data 13 detected from each semiconductor wafer manufactured by each manufacturing process or each manufacturing apparatus, and the detection is performed over a plurality of semiconductor wafers manufactured by each manufacturing process or each manufacturing apparatus. Another embodiment of a method of automatically analyzing and monitoring a defect distribution in the foreign matter / appearance collection system 51 and the abnormality monitoring system 61 from the defect distribution image data 15 based on past cases will be described with reference to FIG. As shown in FIG. 10, first, in step 131, the foreign matter / appearance collection system 51 sends each semiconductor wafer or a plurality of semiconductors from the foreign object inspection device 41 and the appearance inspection device 42 by the method shown in Foreign matter map and appearance map 14 indicating the number of defects (defects such as foreign matter and pattern defects) generated for each pixel divided from the coordinate data detected over the wafer
1 (13 or 15) is converted into image data as a foreign matter and appearance map image 142. The foreign matter map and the appearance map 141 (13 or 15) are supplemented with information such as a number specifying the semiconductor wafer inspected by the foreign matter inspection device 41 and the appearance inspection device. Therefore, it is possible to identify from which semiconductor wafer the foreign matter map and the appearance map 141 are obtained.

Next, in step 132, the foreign matter / appearance collection system 51
Is converted into a binarized image of the foreign matter and the appearance map by a predetermined threshold value (threshold given by a predetermined number of defects (the number of defects) in one pixel), and the converted foreign matter and the binarized image of the appearance map are converted. In contrast, as shown in FIG. 9, defects (defects) generated in adjacent pixels are combined into one cluster (cluster).
A foreign matter and appearance map dilated binary image 143 is created by performing image expansion processing to make the foreign matter and appearance map expanded binary information 143, and information such as a number specifying a semiconductor wafer inspected by the foreign matter inspection device 41 and the appearance inspection device 42 is supplemented. It is transmitted to the abnormality monitoring system 61. Next, the abnormality monitoring system 61 proceeds to step 13
In FIG. 7, for each defect category (foreign matter defect or appearance pattern defect), the created defect (defect) as one lump (cluster) and the appearance map expanded binarized image 143 as shown in FIG. In addition, a past case database prepared for each failure category prepared in advance, including supplementary information on a manufacturing apparatus causing a failure (defect), a manufacturing process causing a failure, a countermeasure method for removing the cause of the failure, and the like. , And a matching rate is calculated by performing matching with a plurality of types of failure distribution templates (represented by a symbol). Then, in step 138, the abnormality monitoring system 61 identifies and detects, for each defect category, the type of the template of the defect distribution having the highest matching rate and the occurrence position (coordinate) of the defect distribution. Next, in step 139, the abnormality monitoring system 61 determines whether or not the matching rate with the template of the specified failure distribution has exceeded an allowable value.
At step 36b, an abnormality is output to warn the user or the administrator of the abnormality to give a warning 62. Therefore, by specifying the type of the failure distribution template having the highest matching rate for which the abnormality monitoring system 61 has issued the abnormality notification 62, the information attached to the failure distribution template is output to the output unit 61a or the like. By outputting the information to the user, the user or the administrator can specify a manufacturing device that causes a defect (defect), a manufacturing process that causes the defect, a countermeasure method for removing the cause of the defect, and the like. Thus, it is possible to easily take a countermeasure and remove the cause of the failure.

Next, a description will be given of a method of creating a plurality of representative failure distribution templates. As shown in FIG. 14, the abnormality monitoring system 61 includes a plurality of types of defect distribution image data 122 serving as templates acquired from the foreign matter / appearance collection system 51 in a number of past cases in which the cause of occurrence of a defect is estimated and countermeasures are taken. Is selected using the input means 61b, and a manufacturing apparatus which causes a defect (defect)
It is stored in the storage device as the template database 121 with additional information such as a manufacturing process that causes a defect and a countermeasure method for removing the defect. As the image data 122 of a plurality of types of defect distributions, a large number of foreign substances and appearance map expansions 2 obtained for each semiconductor wafer obtained from the foreign substance / appearance collection system 51 in the past case or over a plurality of semiconductor wafers The value image 143 may be used. Naturally, the abnormality monitoring system 61 uses a plurality of types of failure distribution image data (failure distribution template images) as representatives from many past cases.
When selecting 122, based on the result of estimating the cause of the occurrence of the defect (defect) and taking a countermeasure, the manufacturing apparatus is estimated to be the cause of the occurrence of the defect (defect), and the cause of the occurrence of the defect (defect) is estimated. The supplementary information such as the manufacturing process and a countermeasure method for eliminating the occurrence of the defect (defect) is added to the storage device as the template database 121.

When there is no past case, the abnormality monitoring system 61 needs to create a new case database (template database). That is, when the abnormality monitoring system 61 determines that there is an abnormality in the processing flow shown in FIG. 7, based on the processing flow shown in FIG. 10, the past case database (template database) 1
21 and if a past case database having a matching ratio equal to or higher than a certain reference is not found, a new case database (template database) needs to be constructed. When the abnormality monitoring system 61 determines that an abnormality has occurred in the processing flow illustrated in FIG. 7 and issues an abnormality notification, the user or the administrator determines whether or not the abnormality notification has been performed on the semiconductor wafer determined to be abnormal or over a plurality of semiconductor wafers. Based on the results, a detailed analysis is performed by observing an electron microscope, searching for past cases using the abnormality monitoring system 61, and the like, to find out a manufacturing process and a manufacturing apparatus that may cause a failure (step 63), and manufacturing equipment (a manufacturing apparatus) (6) Takes measures such as maintenance including maintenance, adjustment of process parameters, and feedback to design (step 64), and the cause of the failure identified in step 63 together with the image data 122 of the failure distribution determined to be abnormal. In the past case, additional information such as the manufacturing process and the manufacturing apparatus to be It is registered in the database 121.
That is, if a new area-based defect is not registered as the template image data 122 in the past case database 121, the image data 122 of the defect distribution is generated.
The cause device, cause process,
Registered in the past case database (template database) 121 as supplementary information 123 such as countermeasures,
After that, use it as a template.

Further, a representative failure distribution template may be created interactively as shown in FIG. 11 and stored in the storage device as a past case database (template database) 121. That is, the abnormality monitoring system 6
In step 161, the CAD system 201
Then, the pattern of the chip arrangement of the design data is read and displayed on the display means 61a. Then, in step 162, the input means 61b is used as an area for each chip as a clickable area based on the search results and experience of past cases. Clicking on these chips based on past search results and experience will change the color and brightness of the display of that chip,
A template shape of a representative failure distribution template is created.
It is also effective to draw a template figure using commercially available photo retouching software or the like and create the template by registering it as an image file. that's all,
When the creation of the template shape of the defect distribution is completed in step 163, the storage device stores the template image data of the shape created on the screen as a past case database (template database) 121 in step 164. This makes it possible to use the template of the representative failure distribution by reading it from the template database 121. In this case, too, based on past case search results and experience,
When the representative failure distribution template image 124 is created, the manufacturing apparatus estimated to be the cause of the defect (defect), the manufacturing process estimated to be the cause of the defect (defect), and the defect (defect) Auxiliary information such as a countermeasure for eliminating occurrence is added and stored in the storage device as the past case database 121.

Next, as shown in FIG. 3 and FIG. 4, defect distribution image data 13 detected from each semiconductor wafer manufactured by each manufacturing process or each manufacturing apparatus, and a plurality of defective distribution image data 13 manufactured by each manufacturing process or each manufacturing apparatus. FIG. 11 shows another embodiment of a method of automatically analyzing and monitoring the area defect of the defect distribution in the foreign matter / appearance collection system 51 and the abnormality monitoring system 61 from the defect distribution image data 15 detected over the semiconductor wafer of FIG. It will be described using FIG. The difference from the embodiment shown in FIG. 7 is that in step 140, the abnormality monitoring system 61 uses a neural network or a statistical network for the feature amount extraction result for each data in the label obtained from the foreign matter / appearance collection system 51. Is performed with feature amount data 121 'which is a past case database acquired from a past case, and a high-matching rate is output as abnormal in step 136c, and an abnormal notification 62 is issued to a user or an administrator. . In this case, similar to the template database 121, the feature data 121 'obtained from a neural network or a statistical past case is converted into a manufacturing device estimated to be the cause of the defect (defect) and the defect (defect). It is necessary to add a number of supplementary information such as a manufacturing process that is presumed to be a cause of occurrence and a countermeasure method for eliminating the occurrence of defects (defects), and prepare a large number of them.

As described above, it is effective to perform the processing shown in FIG. 7, the processing shown in FIG. 10, and the processing shown in FIG. 12 individually. However, each processing is performed in parallel, and any one is detected. By notifying the user of the result, it is possible to detect an abnormality with few omissions. Next, one embodiment of a failure analysis system in semiconductor manufacturing according to the present invention will be described with reference to FIG. FIG. 13 is an overall schematic configuration diagram showing one embodiment of a failure analysis system in semiconductor manufacturing according to the present invention. The manufacturing process A81 includes an oxidation process 81a, a film formation process 81b, a photolithography process 81c, and an etching process 8
1d, and one or more manufacturing apparatuses are used in each of the processes 81a to 81d.
The manufacturing process B82 includes a film forming process 82a, a photolithography process 82b, and an etching process 82c, and one or a plurality of manufacturing devices are used in each of the processes 82a to 82c. The manufacturing process C83 includes a sputtering process 83a, a film forming process 83b, a photolithography process 83c, and an etching process 83d, and one or more manufacturing devices are used in each of the processes 83a to 83d. These manufacturing steps A81, B8
2, C83 schematically shows a representative process in manufacturing a semiconductor wafer. Film forming steps 81b, 82
The same film forming apparatus may be used in a and 83b, the same photolithography apparatus may be used in photolithography steps 81c, 82b and 83c, and the same etching apparatus may be used in etching steps 81d, 82c and 83d. It may be done. In short, a manufacturing line for manufacturing a semiconductor wafer is defined as a manufacturing process A81,
This is schematically illustrated by a manufacturing process B82 and a manufacturing process C83. The semiconductor manufacturing management system 71 manages semiconductor wafers flowing through the manufacturing line, and a database of the flow of semiconductor wafers is constructed. Therefore, by looking at the database of the flow of semiconductor wafers constructed in the semiconductor production management system 71, it is possible to ascertain at what point in time which production process was performed using which production equipment.

Reference numeral 72 denotes a tester for performing an operation test of each semiconductor chip, that is, an electrical test, in a state of the completed semiconductor wafer. The foreign substance inspection device 41 is used as an online monitor or an offline monitor, and is used for each manufacturing process A8.
This is an apparatus for inspecting minute foreign matter defects adhered on sampled semiconductor wafers manufactured in desired manufacturing steps of 1, B82 and C83. Naturally, the foreign matter inspection apparatus 41 can obtain at least information on which semiconductor wafer has been inspected for minute foreign matter defects and information on foreign matter defect position data shown in FIGS. 1 and 2. The visual inspection device 42 is usually used as an off-line monitor, and is a device for inspecting a minute defect of a pattern formed on a sampled semiconductor wafer manufactured in a desired manufacturing process of each of the manufacturing processes A81, B82, and C83. is there.
Naturally, the appearance inspection device 42 can also obtain at least information on which semiconductor wafer has been inspected for minute pattern defects and information on the appearance pattern defect position data shown in FIGS. 1 and 2.

Therefore, the foreign substance / appearance collection system 51 mainly constituted by the CPU relates to the position data of the foreign substance defect inspected by the foreign substance inspection apparatus 41 on the semiconductor wafer manufactured by the desired manufacturing apparatus in the desired manufacturing process. Information and information on position data of appearance pattern defects inspected by the appearance inspection device 42 on the semiconductor wafer manufactured by the desired manufacturing apparatus in the desired manufacturing process are obtained. Then, the foreign matter / appearance collection system 51 converts foreign matter defect position image data inspected by the foreign matter inspection apparatus 41 into foreign matter defect distribution image data indicated by the number of foreign matter defects for each divided pixel over a plurality of semiconductor wafers. Add to create foreign matter defect distribution image data,
The appearance defect distribution image data is created by adding the appearance defect distribution image data indicated by the number of pattern defects for each of the divided pixels to the pattern defect position data inspected by the appearance inspection device 42 over a plurality of semiconductor wafers. I do. Further, the foreign substance / appearance collection system 51 can obtain the foreign substance defect distribution density image data by converting the numerical value of each pixel in the generated foreign substance defect distribution image data into a gray level value. By converting the numerical value of each pixel in the image data into a gray value, it is possible to obtain the external defect distribution density image data. , 61a. In the foreign matter / appearance collection system 51, the foreign matter inspection device 41 and the appearance inspection device 4
When collecting the inspected data from the semiconductor wafer 2, by referring to the data of the flow of the semiconductor wafer stored in the semiconductor manufacturing management system 71, regarding what kind of manufacturing path and when the semiconductor wafer is manufactured, It becomes possible to grasp.

The electric tester 43 obtains the operation test result of each completed semiconductor chip in the state of a semiconductor wafer from the tester 72. When each semiconductor chip is, for example, a memory, a defective memory cell is specified as an operation test result,
This defective memory cell is classified into a memory cell that can be repaired by switching to a memory cell for repair and a memory cell that cannot be repaired.
Therefore, from the electrical inspection device 43, all the semiconductor wafers are classified into those that can be repaired in units of cells in each semiconductor chip and those that cannot be repaired, and the position data of the defective cell (defective element) and the entire chip cannot be repaired. Information about the chip failure as a result is obtained. Mainly C
The electrical test collection system 52 constituted by the PU includes:
Information on the position data of the malfunctioning cell (malfunctioning element) inspected by the electrical inspection device 43 on the completed semiconductor wafer and information on the chip failure can be obtained. And an electrical inspection and collection system 5 mainly constituted by a CPU.
Reference numeral 2 denotes a defective element distribution image obtained by adding position data of defective elements inspected by the electrical inspection device 43 to defective element distribution image data indicated by the number of defective elements for each divided pixel over a plurality of semiconductor wafers. By creating data and further converting the numerical value of each pixel in the created defective element distribution image data into a gray value, it is possible to obtain defective element distribution gray image data, and display this defective element distribution gray image data. It can be displayed on the means 51a and 61a. When collecting data inspected from the electrical inspection device 43 in the electrical inspection collection system 52, the semiconductor manufacturing management system 71 stores information on what kind of manufacturing path the semiconductor wafer has been manufactured and when. It can be grasped by referring to the data of the flow of the semiconductor wafer.

The alignment inspection apparatus 44 inspects an alignment error in an exposure apparatus used in the photolithography process by measuring a test pattern obtained by actual exposure. Therefore, the alignment inspection device 44 measures and inspects the alignment error in the exposure apparatus used. The dimension inspection device 45 inspects the dimensions of a pattern formed on a semiconductor wafer by etching or the like. The film thickness inspection device 46 inspects the film thickness formed on the semiconductor wafer. The quality collection system 53 calculates the semiconductor alignment error from the alignment error in the exposure device obtained from the alignment inspection device 44, the pattern size obtained from the dimension inspection device 45, and the film thickness dimension obtained from the film thickness inspection device 46. Information on the quality of the circuit pattern formed on the wafer is collected. In the quality collection system 53,
When collecting data measured from the alignment inspection device 44, the dimension inspection device 45, and the film thickness inspection device 46, the semiconductor manufacturing management system 71 determines which manufacturing path has passed and when the semiconductor wafer has been manufactured. It can be understood by referring to the data on the flow of the semiconductor wafer stored in the memory.

Mainly CPU and past case database 1
An abnormality monitoring system 61 including a storage device for storing the information 21 (121 ′) is a foreign matter / appearance collection system 5.
1. Automatic monitoring is performed based on data collected by the electrical inspection collection system 52, the quality collection system 53, and the like. 7 and FIG.
In addition to the failure distribution analysis (analysis based on failure occurrence area) shown in FIG. Automatically monitor and display the monitoring result
a and output from the terminal device or the semiconductor manufacturing management system 71 via the network to notify the user or the administrator. In the automatic abnormality monitoring system 61, the interrelationship between data collected by the foreign substance / appearance collection system 51, the electrical inspection collection system 52, the quality collection system 53, etc. The correlation between the collected data and the data on the wafer) can be grasped by referring to the data on the flow of the semiconductor wafer accumulated in the semiconductor manufacturing management system 71. Therefore, the automatic abnormality monitoring system 61 can perform not only the time series analysis in the specified manufacturing process but also the machine difference analysis including the time series in the specified manufacturing apparatus.

Incidentally, the automatic abnormality monitoring system 61 and the collection systems 51, 52, 53 may be constituted by the same CPU. Further, the functions of the automatic abnormality monitoring system 61 and the collection systems 51, 52, 53 may be incorporated into the semiconductor manufacturing management system 72. Next, an embodiment of a method in which the automatic abnormality monitoring system 61 automatically notifies a user or an administrator will be described with reference to FIG. The automatic abnormality monitoring system 61 can be operated by using a desired terminal device or a semiconductor manufacturing management system 71 by using e-mail or the like via a network.
Automatically send to. FIG. 15 shows contents 171 automatically transmitted to a desired terminal device or semiconductor manufacturing management system 71. TO: Related person, FROM: Abnormality monitoring system, SUB:
Foreign matter frequently occurs above the wafer where the foreign matter area defect has occurred (shows the state of occurrence of foreign matter defects). Product name: 64MDRAM, Manufacturing process name: Third G
ate, inspection content: foreign substance inspection # 1, number of foreign substances: 203 /
Wafer, detection date and time: 1996/12/06, etc. The manufacturing apparatus name may be entered here.

In the automatically transmitted content 171, as an attached file, a time-series transition diagram 172 until the abnormal state of the inspection data is obtained, or converted into defect distribution image data or a grayscale image when the abnormal state occurs. The defect distribution density image data 173 and the like are sent. FIG. 172 shows a time-series transition until the inspection data becomes abnormal.
The time or lot unit for sampling a semiconductor wafer from Third Gate is shown. ) And the number of foreign particles, which is one of the characteristic amounts of the defect distribution detected from one semiconductor wafer. The number of foreign particles, which is one of the characteristic amounts of the defect distribution, may be the number of foreign particles detected from one cluster (lump) in one semiconductor wafer. When a plurality of clusters (lumps) are generated from one semiconductor wafer, this is the maximum number of foreign substances among them. The vertical axis may be a value of a two-dimensional area in which a foreign matter from which a noise component, which is one of the feature amounts of the defect distribution, is detected. The defect distribution image data 173 converted into the defect distribution image data or the gray image when the abnormal state occurs is indicated by 32 or 33 in FIG.

Although the content 171 has been described as a defect in the case of a foreign matter defect, it is possible to similarly configure an external pattern defect, a malfunctioning element, and the like as a defect. Then, in the desired terminal device, by outputting the contents 171 to an output unit (for example, a display unit), not only the result of the abnormal state but also information effective for investigating the cause of the abnormality is promptly transmitted to the concerned person. It becomes possible to notify. As the defect distribution image data 173, a density image shown in FIG. 6 or a defect distribution image 33 superimposed with the design data shown in FIG. 4 is a good method by visually indicating the area characteristics of the defect distribution.

FIG. 16 shows another embodiment in which an abnormality is monitored and the result is automatically notified to a user or an administrator. In this embodiment, a foreign matter inspection device 41 and a visual inspection device 42
Inside, the functions of the foreign matter / appearance collection system 51 and the abnormality monitoring system 61 are provided. In this embodiment, the foreign matter / appearance collection system 5 is installed in the electrical inspection device 43.
1 and the function of the abnormality monitoring system 61. That is, as an inspection device 181 (41, 42, 43, etc.), an inspection unit 182 that performs inspection, an abnormality monitoring unit 183 that collects failure (defect) data inspected by the inspection unit 182 and monitors abnormality, A transmission unit 18 for outputting information on an abnormal result monitored by the abnormality monitoring unit 183 to another terminal device or the semiconductor manufacturing management system 71 via a network.
4, the abnormality monitoring unit 183 manages the standard value,
Various analyzes such as time-series trend change analysis, machine difference analysis for each manufacturing device, and regionality analysis are performed, and the results are notified to a user or a manager. In this case, in addition to transmitting the content 171 shown in FIG. 15 to another terminal device or the semiconductor manufacturing management system 71 via the network, an operation screen provided on the inspection device 181 for directly notifying the operator of the inspection device. It is more effective to display the monitoring result on the monitor.

Next, FIG. 16 shows an embodiment in which the defect distribution is displayed in a map not in a wafer unit but in a shot unit (exposure unit) upon exposure of a semiconductor wafer. Basically, it is similar to the concept shown in FIG. Each shot unit is treated as rectangular image data of K × T pixels, and image data is formed for each shot unit depending on where the coordinate data 21 of the defect inspection result exists in each shot for each defect. In this image data, a pixel having one or more defects is binarized to 1 and a pixel having no defect is binarized to 0 (step 191). By adding all the binary image data groups of the shots (step 192) and creating grayscale image data of a defect distribution in shot units, it is possible to visually grasp the tendency of defects in shot units. If defects are concentrated at a certain position of the added image data, it can be easily found that there is a defect in the reticle or the like at the time of exposure. This is an embodiment in which the image data is converted in units of exposure shots. However, it is obvious that the image data may be converted in units of chips. If the image data is converted into the chip unit, it is possible to grasp the occurrence state of the defect or the defect distribution in the chip unit.

[0052]

According to the present invention, in a method of manufacturing a semiconductor substrate, foreign matter defects or the like occurring on a semiconductor substrate are determined for each desired manufacturing process, each desired manufacturing apparatus (manufacturing equipment), each production lot, and the like. By grasping the tendency of defect distribution (defect area characteristics) such as appearance defects on image data composed of lattice-like pixels set on the semiconductor substrate, the concentration of defects and the uniformity of defects It is possible to analyze and monitor regional defects such as scattering, which can lead to early detection of abnormal conditions and early countermeasures. An effect that can be manufactured with quality is exhibited. Further, according to the present invention, in a method of manufacturing a semiconductor substrate, a foreign substance defect, an appearance defect, and the like generated on a semiconductor substrate for each desired manufacturing process, each desired manufacturing apparatus (manufacturing equipment), each production lot, and the like. By visually grasping the tendency of the defect distribution (regional nature of defects), it is possible to analyze and monitor regionality defects such as the concentration of defects and the uniform scattering of defects. This can be linked to early detection of the state and early countermeasures, so that the production of defects can be significantly reduced and the semiconductor substrate can be manufactured with high yield and high quality.

Further, according to the present invention, a defect distribution such as a foreign matter defect or an appearance defect occurring on a semiconductor substrate is determined for each desired manufacturing process, each desired manufacturing apparatus (manufacturing equipment), or each production lot. The tendency of defect area) can be grasped on image data composed of grid-like pixels set on the semiconductor substrate, and as a result, the cause of the defect can be easily estimated and countermeasures taken at an early stage. As a result, there is an effect that semiconductor devices can be manufactured with high yield and high quality by significantly reducing the formation of defects. Further, according to the present invention, in a method of manufacturing a semiconductor substrate, a foreign substance defect, an appearance defect, and the like generated on a semiconductor substrate for each desired manufacturing process, each desired manufacturing apparatus (manufacturing equipment), each production lot, and the like. The tendency of the defect distribution (defect area characteristic) can be grasped on the image data composed of the grid-like pixels set on the semiconductor substrate, and as a result, the defective manufacturing process, the defective manufacturing equipment, and the defective lot By being able to specify the unit and the like, it is possible to easily estimate the cause of the occurrence of a defect and to take a countermeasure at an early stage, thereby significantly reducing the production of the defect and manufacturing the semiconductor substrate with high yield and high quality. The effect that can be achieved.

Further, according to the present invention, in the method of manufacturing a semiconductor substrate, in addition to the regional defect, various analysis items such as standard value management of inspection data, time-series trend change analysis, and machine-by-device difference analysis. By automatically monitoring abnormal situations and notifying related parties in real time, it is possible to detect abnormal situations early and take measures as early as possible, significantly reducing the production of defective products and achieving a high yield of semiconductor substrates. It has the effect that it can be manufactured with high quality. Further, according to the present invention, in a method of manufacturing a semiconductor substrate, a failure distribution can be visually observed for each desired manufacturing process, each desired manufacturing facility, and each production lot, so that a person concerned can understand the failure. , Helping to determine the cause. Further, according to the present invention, in a method of manufacturing a semiconductor substrate, by registering a past case database regarding a defect distribution,
It is possible to easily estimate the cause of the failure and take an early action.

[Brief description of the drawings]

FIG. 1 shows a method for manufacturing a semiconductor substrate according to the present invention.
FIG. 4 is a diagram for explaining a basic embodiment for finding a defective manufacturing process.

FIG. 2 shows a method for manufacturing a semiconductor substrate according to the present invention.
It is a figure for explaining one basic embodiment which finds a defective manufacturing device (manufacturing equipment).

FIG. 3 is a diagram for explaining an embodiment in which defect distribution image data according to the present invention is created and displayed on a screen.

FIG. 4 is a diagram for describing an embodiment of the present invention in which defect distribution image data is created from a plurality of semiconductor wafers, and design data is displayed on a screen with overlapping design data.

FIG. 5 is an explanatory diagram for creating defect distribution image data by adding the number of defects for each of the grid-like pixels according to the present invention for a plurality of semiconductor wafers.

FIG. 6 is a diagram showing an embodiment in which a defect distribution is displayed on a screen by changing the size of a grid-like pixel set on a semiconductor wafer according to the present invention.

FIG. 7 is a processing flowchart showing a first embodiment for analyzing a failure distribution (defect area property) according to the present invention.

FIG. 8 is an explanatory diagram showing a failure distribution (failure area characteristics) in the processing flow shown in FIG. 7;

FIG. 9 is a diagram for explaining an expansion process for making a defect (defect) that has occurred in the vicinity into one lump (cluster);

FIG. 10 is a process flowchart showing a second embodiment for analyzing a failure distribution (defect area property) according to the present invention.

FIG. 11 is a diagram for explaining an embodiment for creating a new case database (template) according to the present invention.

FIG. 12 is a process flowchart showing a third embodiment for analyzing a failure distribution (defect area property) according to the present invention.

FIG. 13 is a view showing an overall schematic configuration of a failure analysis system used in the method of manufacturing a semiconductor substrate according to the present invention.

14 is a diagram showing a specific configuration of the abnormality monitoring system shown in FIG.

FIG. 15 is a view showing a related person (user or administrator) when an abnormal situation occurs in the method of manufacturing a semiconductor substrate according to the present invention.
FIG. 6 is a diagram showing an embodiment of the content reported in FIG.

FIG. 16 is a diagram showing an embodiment in which an inspection unit, an abnormality monitoring unit, and a transmission unit are provided in the inspection device.

FIG. 17 is a diagram for describing an embodiment of the present invention in which defect distribution image data is created for each exposure shot and displayed on a screen.

[Explanation of symbols]

11, 12: Display of density of defective distribution image data, 13, 1
6, 17, 122 ... defect distribution image data, 14, 10
1, 102, 103, 104, 105 ... defect distribution image data group, 15, 18, 111, 112, 113, 11
4, 115: defective distribution added image data, 21: coordinate data (defective position data) of a defect inspection result, 22: pattern by design data, 31, 32: gray scale display of defective distribution image data, 33: defective distribution and design data Simultaneous display of patterns by 41, foreign matter inspection device, 42 appearance inspection device, 43 electric inspection device, 44 alignment inspection device, 45
Dimension inspection device, 46: film thickness inspection device, 47: inspection device,
51: foreign matter / appearance collection system, 51a: display means, 5
1b ... input means, 52 ... electrical inspection collection system, 53 ...
Quality collection system, 61 ... Abnormality monitoring system, 61a ...
Display means, 61b ... input means, 62 ... abnormality monitoring unit, 63
... Transmission unit, 72 ... Tester, 71 ... Semiconductor manufacturing management system, 81 ... Manufacturing process A, 82 ... Manufacturing process B, 83 ... Manufacturing process C, 84 ... Manufacturing device A, 85 ... Manufacturing device B, 91
92, 93, 94, 95 ... inspection device, 121, 121 '
… Case database (template database), 1
71: Contents to be reported to the related parties, 172: Bad transition, 20
1. CAD system

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Masaki Korai 5-2-1, Josuihonmachi, Kodaira-shi, Tokyo Inside the Semiconductor Division, Hitachi, Ltd. (72) Inventor Naoki Go 292, Yoshidacho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture House Co., Ltd., Hitachi, Ltd.

Claims (16)

[Claims] 1. A method of manufacturing a semiconductor substrate, comprising manufacturing a semiconductor substrate through a plurality of manufacturing steps, wherein an inspection device is provided for detecting a position of a defect occurring on each of the plurality of semiconductor substrates manufactured in the desired manufacturing step. Inspection step of inspecting in the inspection step, the defect position data on each semiconductor substrate inspected in the inspection step, the coordinates are specified on image data composed of lattice-like pixels set on the semiconductor substrate, A defect distribution image data creating step of adding the number of defects for each of the grid-shaped pixels for the plurality of semiconductor substrates on the data to create defect distribution image data; and a defect distribution image created in the defect distribution image data creating step. A failure occurrence state grasping step of grasping a failure occurrence state on the semiconductor substrate based on the data. 2. A semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by a manufacturing line including a plurality of manufacturing apparatuses, wherein a defect generated on each of the plurality of semiconductor substrates manufactured by the desired manufacturing apparatus. An inspection process of inspecting the position of the semiconductor device with an inspection device, and the position data of a defect on each semiconductor substrate inspected in the inspection process is coordinated on image data composed of grid-like pixels set on the semiconductor substrate. A defect distribution image data creating step of creating the defect distribution image data by adding the number of defects for each of the lattice-shaped pixels on the plurality of semiconductor substrates on the image data;
A failure occurrence state grasping step of grasping a failure occurrence state on the semiconductor substrate based on the failure distribution image data created in the failure distribution image data creation step. 3. A method for manufacturing a semiconductor substrate, wherein the semiconductor substrate is manufactured through a plurality of manufacturing steps, wherein the inspection apparatus detects a position of a defect generated on each semiconductor substrate in the plurality of semiconductor substrates manufactured in the desired manufacturing step. Inspection step of inspecting in the inspection step, the defect position data on each semiconductor substrate inspected in the inspection step, the coordinates are specified on image data composed of lattice-like pixels set on the semiconductor substrate, A defect distribution image data generating step of adding the number of defects for each of the grid-shaped pixels on the plurality of semiconductor substrates on the data to generate defect distribution image data represented by a grayscale value; Display the defect distribution image data indicated by the displayed grayscale value on the display means, and determine the defect distribution on the semiconductor substrate based on the defective distribution image data indicated by the displayed grayscale value. Method of manufacturing a semiconductor substrate and having a failure condition grasping step of grasping the green state. 4. A semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by using a manufacturing line including a plurality of manufacturing apparatuses, wherein a plurality of semiconductor substrates manufactured by the desired manufacturing apparatus have defects generated on each semiconductor substrate. An inspection process of inspecting the position of the semiconductor device with an inspection device, and the position data of a defect on each semiconductor substrate inspected in the inspection process is coordinated on image data composed of grid-like pixels set on the semiconductor substrate. A defect distribution image data creating step of adding the number of defects for each of the grid-like pixels for the plurality of semiconductor substrates on the coordinate-designated image data to create defect distribution image data represented by a grayscale value; The failure distribution image data represented by the gray value created in the failure distribution image data creation step is displayed on the display means, and the failure distribution image data represented by the displayed gray value is displayed. Method of manufacturing a semiconductor substrate and having a failure condition grasping step of grasping the state of occurrence of defects on a semiconductor substrate based on the data. 5. A method of manufacturing a semiconductor substrate, comprising manufacturing a semiconductor substrate through a plurality of manufacturing steps, wherein a plurality of semiconductor substrates in lots manufactured in the desired manufacturing step are provided with a defect generated on each semiconductor substrate. Inspection step of inspecting the position with an inspection apparatus, and coordinate designation of defect position data on each semiconductor substrate inspected in the inspection step on image data composed of lattice-like pixels set on the semiconductor substrate. And
A defect distribution image data creating step of adding the number of defects for each lattice-shaped pixel for the plurality of semiconductor substrates in lot units on the image data to create defect distribution image data in lot units; A failure occurrence state grasping step of grasping a failure occurrence state on the semiconductor substrate based on defect distribution image data in lot units created in the data creation step. 6. A semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by a manufacturing line including a plurality of manufacturing apparatuses, wherein a plurality of semiconductor substrates in lot units manufactured by the desired manufacturing apparatus are provided on each semiconductor substrate. An inspection step of inspecting the position of a defect that has occurred in the inspection apparatus by using an inspection apparatus, and an image composed of lattice-like pixels set on the semiconductor substrate by using the defect position data on each semiconductor substrate inspected in the inspection step. Defect distribution image data creation for designating coordinates on data and adding the number of defects for each of the lattice-shaped pixels on the image data for a plurality of semiconductor substrates in lot units to create defect distribution image data in lot units Process and a failure occurrence state on the semiconductor substrate is grasped based on the failure distribution image data in lot units created in the failure distribution image data creation process. Method of manufacturing a semiconductor substrate and having a good generation state grasping step. 7. A semiconductor substrate manufacturing method for manufacturing a semiconductor substrate through a plurality of manufacturing steps, wherein a plurality of semiconductor substrates in lot units manufactured in the desired manufacturing step are provided with a defect generated on each semiconductor substrate. Inspection step of inspecting the position with an inspection apparatus, and coordinate designation of defect position data on each semiconductor substrate inspected in the inspection step on image data composed of lattice-like pixels set on the semiconductor substrate. And
A defect distribution image data creating step of adding the number of defects for each lattice-like pixel on the image data for a plurality of semiconductor substrates in lot units to create defect distribution image data represented by shading values in lot units; Displaying, on the display means, the defect distribution image data represented by the grayscale value in the lot unit created in the defect distribution image data creating step, and displaying the defect distribution image data represented by the grayscale value in the lot unit. And a failure occurrence state grasping step of grasping a state of occurrence of a failure on the semiconductor substrate based on the defect occurrence state. 8. A semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by a manufacturing line including a plurality of manufacturing apparatuses, wherein a plurality of semiconductor substrates in lot units manufactured by the desired manufacturing apparatus are provided on each semiconductor substrate. An inspection step of inspecting the position of a defect that has occurred in the inspection apparatus by using an inspection apparatus, and an image composed of lattice-like pixels set on the semiconductor substrate by using the defect position data on each semiconductor substrate inspected in the inspection step. A defect distribution image in which coordinates are designated on data, and the number of defects for each grid-like pixel is added for a plurality of semiconductor substrates in lot units on the coordinate-designated image data, and the defect distribution image is indicated by a grayscale value in lot units A defect distribution image data creating process for creating data, and defect distribution image data represented by shading values in lot units created in the defect distribution image data creating process are displayed. A failure occurrence state grasping step of grasping a failure occurrence state on a semiconductor substrate on the basis of the failure distribution image data indicated by a gray scale value in the displayed lot unit and displayed on a row. Substrate manufacturing method. 9. A method of manufacturing a semiconductor substrate, wherein the semiconductor substrate is manufactured through a plurality of manufacturing steps, wherein a plurality of semiconductor substrates manufactured in the desired manufacturing step are inspected for a position of a defect occurring on each semiconductor substrate. Inspection step of inspecting, and the position data of a defect on each semiconductor substrate inspected in the inspection step are designated by coordinates on image data composed of grid-like pixels set on the semiconductor substrate, A defect distribution image data creating step of adding the number of defects for each of the grid-shaped pixels for the plurality of semiconductor substrates on the data to create defect distribution image data; and a defect distribution image created in the defect distribution image data creating step. A failure occurrence state grasping step of calculating a failure distribution feature amount from the data and grasping a failure occurrence state on the semiconductor substrate based on the calculated failure distribution feature amount; Method of manufacturing a semiconductor substrate and having. 10. A semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by a manufacturing line including a plurality of manufacturing apparatuses, wherein a defect generated on each of the plurality of semiconductor substrates manufactured by the desired manufacturing apparatus is generated. An inspection process of inspecting the position of the semiconductor device with an inspection device, and the position data of a defect on each semiconductor substrate inspected in the inspection process is coordinated on image data composed of grid-like pixels set on the semiconductor substrate. A failure distribution image data creating step of creating a failure distribution image data by designating and adding the number of failures for each of the lattice-shaped pixels on the plurality of semiconductor substrates on the image data; A feature amount of the failure distribution is calculated from the obtained failure distribution image data, and an occurrence state of the failure on the semiconductor substrate is grasped based on the calculated feature amount of the failure distribution. Method of manufacturing a semiconductor substrate and having a failure condition grasping step that. 11. A method for manufacturing a semiconductor substrate, wherein the semiconductor substrate is manufactured through a plurality of manufacturing steps, wherein a position of a defect generated on each of the plurality of semiconductor substrates manufactured in the desired manufacturing step is inspected. Inspection step of inspecting in the inspection step, the defect position data on each semiconductor substrate inspected in the inspection step, the coordinates are specified on image data composed of lattice-like pixels set on the semiconductor substrate, A defect distribution image data creating step of adding the number of defects for each of the lattice-shaped pixels on the plurality of semiconductor substrates on the data to create defect distribution image data; and a defect distribution image created in the defect distribution image data creating step. A failure analysis step of collating and analyzing data with a plurality of prepared case databases from which the cause of failure can be estimated to determine the cause of failure. A method of manufacturing a semiconductor substrate. 12. A semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by a manufacturing line including a plurality of manufacturing apparatuses, wherein a defect generated on each semiconductor substrate for the plurality of semiconductor substrates manufactured by the desired manufacturing apparatus. An inspection process of inspecting the position of the semiconductor device with an inspection device, and the position data of a defect on each semiconductor substrate inspected in the inspection process is coordinated on image data composed of grid-like pixels set on the semiconductor substrate. A failure distribution image data creating step of creating a failure distribution image data by designating and adding the number of failures for each of the lattice-shaped pixels on the plurality of semiconductor substrates on the image data; Failure analysis to find out the cause of failure by collating and analyzing the provided failure distribution image data with a prepared case database that can estimate the cause of failure Method of manufacturing a semiconductor substrate and having a degree. 13. A method of manufacturing a semiconductor substrate, wherein the semiconductor substrate is manufactured through a plurality of manufacturing steps, wherein a plurality of semiconductor substrates manufactured in the desired manufacturing step are inspected for a position of a defect occurring on each semiconductor substrate. Inspection step of inspecting, and the position data of a defect on each semiconductor substrate inspected in the inspection step are designated by coordinates on image data composed of grid-like pixels set on the semiconductor substrate, A defect distribution image data creating step of adding the number of defects for each of the grid-shaped pixels for the plurality of semiconductor substrates on the data to create defect distribution image data; and a defect distribution image created in the defect distribution image data creating step. The feature amount of the failure distribution is calculated from the data, the calculated feature amount of the failure distribution is displayed on the display means, and the defect amount is displayed on the semiconductor substrate based on the displayed feature amount of the failure distribution. Method of manufacturing a semiconductor substrate and having a failure condition grasping step of grasping the state of occurrence of that failure. 14. A semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by a manufacturing line including a plurality of manufacturing apparatuses, wherein a defect generated on each semiconductor substrate for the plurality of semiconductor substrates manufactured by the desired manufacturing apparatus. An inspection process of inspecting the position of the semiconductor device with an inspection device, and the position data of a defect on each semiconductor substrate inspected in the inspection process is coordinated on image data composed of grid-like pixels set on the semiconductor substrate. A failure distribution image data creating step of creating a failure distribution image data by designating and adding the number of failures for each of the lattice-shaped pixels on the plurality of semiconductor substrates on the image data; Calculating a feature amount of the failure distribution from the obtained failure distribution image data, displaying the calculated feature amount of the failure distribution on a display unit, and displaying the feature amount of the displayed failure distribution. Based method of manufacturing a semiconductor substrate and having a failure condition grasping step of grasping the state of occurrence of defects on a semiconductor substrate. 15. A method of manufacturing a semiconductor substrate, wherein the semiconductor substrate is manufactured through a plurality of manufacturing steps, wherein a plurality of semiconductor substrates manufactured in the desired manufacturing step are inspected for a position of a defect occurring on each semiconductor substrate. Inspection step of inspecting, and a grid-like pattern set by dividing the defect position data on each semiconductor substrate inspected in the inspection step into a plurality of exposure shot units or chip units on the semiconductor substrate. The coordinates are specified on the image data of the shot unit or chip unit composed of pixels of the above, and the number of defects is added for each of the lattice-shaped pixels for the plurality of semiconductor substrates on the image data of the shot unit or chip unit for the plurality of semiconductor substrates. A failure distribution image data creating step of creating failure distribution image data represented by a gray value in a chip unit; The failure distribution image data represented by the gray scale value of the shot unit or the chip unit created in the creation step is displayed on the display means, and the defect distribution image data represented by the displayed gray scale value of the shot unit is displayed on the semiconductor substrate. A failure occurrence state grasping step of grasping a state of occurrence of a failure in the semiconductor substrate. 16. A semiconductor substrate manufacturing method for manufacturing a semiconductor substrate by using a manufacturing line including a plurality of manufacturing apparatuses, wherein a plurality of semiconductor substrates manufactured by the desired manufacturing apparatus have defects generated on each semiconductor substrate. An inspection step of inspecting the position of the semiconductor device with an inspection apparatus, and a grid-like pattern set by dividing and positioning defect position data on each semiconductor substrate inspected in the inspection step into a plurality of exposure shot units or chip units. Coordinates are specified on image data of a shot unit or a chip unit composed of pixels, and the number of defects is added for a plurality of semiconductor substrates on a plurality of semiconductor substrates on a grid-like pixel basis on the coordinate-designated image data of a shot unit. A failure distribution image data creating step of creating failure distribution image data represented by a gray value in a chip unit; The failure distribution image data represented by the gray scale value of the shot unit or the chip unit created in the creation process is displayed on the display means, and based on the displayed failure distribution image data represented by the gray value of the shot unit or the chip unit. A failure occurrence state grasping step of grasping a failure occurrence state on the semiconductor substrate.