JPH09232387A - Pattern fault specification method and pattern fault detection method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To objectively and precisely specify the cause of occurrence of a pattern fault by comparing occurring pattern fault contrast information with past pattern fault contrast information. SOLUTION: A data for storing past pattern fault contrast information constituted of the value of the contrast of at least one color in the picture of the past pattern fault being the pattern fault which occurs in the past and whose fault occurring cause is ascertained is generated. Then, the occurring pattern fault being the pattern fault occurring in the manufacture process of a semiconductor device is detected, and occurring pattern fault contrast information constituted of the value of the contrast of at least one color in the picture of the detected occurred pattern fault is calculated. Then, the occurring fault cause of the occurred pattern fault is specified by comparing occurred pattern fault contrast information with past pattern fault contrast information.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for inspecting a pattern defect occurring in a semiconductor device manufacturing process, and more particularly to a pattern defect detecting method for detecting a pattern defect, and a pattern defect generating step, generating device or cause. The present invention relates to a pattern defect identification method for identifying the cause of defect occurrence such as

[0002]

2. Description of the Related Art In order to maintain and improve the yield of semiconductor devices, it is important to manage and suppress the generation of pattern defects such as particles generated during the process. For this reason, the importance of the pattern defect inspection step of inspecting a wafer in the process by a pattern defect inspection apparatus and managing the occurrence status of pattern defects is increasing.

The conventional pattern defect detection method and pattern defect identification method will be described below.

As an example of a conventional pattern defect detection method, a plurality of wafers formed by using the same mask are compared with each other in the course of a semiconductor device manufacturing process, and a portion where the patterns of the plurality of wafers do not match is patterned. A method of detecting a defect is known.

Further, as an example of a conventional pattern defect identification method, the following procedure is known.

FIGS. 6 (a) and 6 (b) show the procedure of the conventional pattern defect identification method. In FIG.
Reference numeral 4 denotes a pattern defect map measured by the pattern defect inspection apparatus during the semiconductor device manufacturing process, 55 denotes a wafer displayed on the pattern defect map 54, and 56 denotes a pattern defect on the wafer 55.
Further, in FIG. 6B, 57 is an image of the pattern map introduced into the visual field of the optical microscope based on the coordinates of the pattern defect map 54, 58 is an image of the pattern of the semiconductor device, and 59 is an image of the pattern defect. Shows.

First, in a certain step in the process of manufacturing a semiconductor device, the pattern defect inspection apparatus is used to inspect the pattern defect 56 to display the position of the pattern defect 56 existing on the wafer 55. 54 is created.

Next, the pattern defects 5 are formed on the wafer 55 by inputting an optical microscope and coordinates on the wafer 55.
The operator observes the image 57 of the pattern map obtained by introducing 6 into the review station which can be put in the visual field. The operator specifies the generation process, generation device, generation cause, and the like of the pattern defect image 59 for all the pattern defects 59 in the pattern defect map 54 based on past experience.

[0009]

However, according to the above-described pattern defect detection method, since the patterns of a plurality of wafers are compared with each other and a portion where the patterns do not match is detected as a pattern defect, the pattern defect is detected. There is a problem that it takes time and the detection accuracy is not good.

Further, in the pattern defect identification method described above, it takes a lot of time to identify the cause of the defect occurrence of the pattern defect, and the throughput of the defect inspection process is poor, so the number of inspection processes and the number of inspection samplings are small. , The fact that it requires experience and skill to specify the defect occurrence reason of the pattern defect, the defect reason specified by the operator who works is different, the accuracy rate of the defect reason identification result is low, etc. Have a problem.

In view of the above, according to the present invention, a pattern defect detecting method capable of accurately detecting a pattern defect in a short time, and the throughput of the work for specifying the cause of the defect are improved, and the defect can be objectively specified without depending on the subjectivity of the operator. It is also an object of the present invention to provide a pattern defect identification method capable of obtaining an identification result with a high accuracy rate.

[0012]

In order to achieve the above object, the means for solving the problems according to the invention of claim 1 is at least one of a pattern defect generation step, a generation device and a generation cause of a pattern defect in a semiconductor device manufacturing process. At least one color contrast in an image of a past pattern defect, which is a pattern defect that has occurred in the past and whose cause of occurrence is known, is targeted for a pattern defect identification method that identifies one defect occurrence reason. A pattern forming step for forming a database storing past pattern defect contrast information having a value of, a pattern defect detecting step for detecting a generated pattern defect which is a pattern defect generated in a semiconductor device manufacturing process, A contrast of the at least one color in the image of the generated pattern defect A defect for identifying the cause of the defect of the generated pattern defect by comparing the generated pattern defect contrast information with the past pattern defect contrast information. It is configured to include a process for specifying the cause of occurrence.

According to the first aspect of the present invention, the generated pattern defect contrast information including the contrast value of at least one color in the detected image of the generated pattern defect and the cause of the defect that has occurred in the past are identified. In order to identify the defect occurrence reason of the generated pattern defect by comparing with the past pattern defect contrast information consisting of the contrast value of at least one color in the image of the past pattern defect existing, that is, comparing the color contrast values. By doing so, the cause of the defect occurrence of the pattern defect is specified, so that the cause of the defect occurrence can be specified by the computer without relying on human hands.

According to a second aspect of the present invention, in the structure according to the first aspect, in the database manufacturing step, after the past pattern defects generated in a semiconductor device manufacturing process are subjected to a process such as film peeling, A configuration including a step of obtaining a contrast value of the at least one color in the image of the pattern defect and storing the obtained contrast value as the past pattern defect contrast information in the database is added.

According to a third aspect of the present invention, in the configuration of the first aspect, the past pattern defect contrast information is an average value of numerical values of gray scales of contrast for each RGB in each pixel of the image of the past pattern defect. Therefore, the generated pattern defect contrast information is added with a configuration that is an average value of numerical values of gradations of contrast shades of RGB for each pixel of the image of the generated pattern defect.

According to a fourth aspect of the present invention, in the structure according to the first aspect, the defect occurrence reason specifying step is an average value of numerical values of gradations of contrast of each RGB in each pixel of the image of the past pattern defect. , The average value of the square of the difference of each RGB from the average value of the gray scale of the contrast of each RGB in each pixel of the image of the generated pattern defect is determined, and the calculated average value of the square is the most A configuration including a step of identifying the cause of the defect of the past pattern defect that becomes smaller as the cause of the defect of the generated pattern defect is added.

According to a fifth aspect of the present invention, there is provided a means for solving the problem of pattern defect detection, which comprises dividing a semiconductor wafer into a plurality of pattern blocks having a repeating pattern, and for each pixel of an image of the plurality of pattern blocks. A step of obtaining a pattern block contrast value which is a numerical value of gradation of light and shade of color contrast, the pattern block contrast value, and the color of each pixel of the image of the reference block having the repeating pattern and having no pattern defect A step of obtaining a difference between a reference block contrast value, which is a numerical value of gradation of contrast, and an area in which the difference between the pattern block contrast value and the reference block contrast value is within a predetermined value is a pattern defect occurrence area. And a step of specifying that there is.

According to the configuration of claim 5, the pattern block contrast value, which is a numerical value of the gradation of the contrast of the color of each pixel of the image of the pattern block in which the semiconductor wafer is divided, and the reference block having no pattern defect. In order to specify the area where the difference from the reference block contrast value, which is the numerical value of the gradation of the color contrast of each pixel of the image, is within the predetermined value, it is specified as the pattern defect occurrence area, that is, the color contrast value. Since the pattern defect is detected by comparing them, the pattern defect can be detected by the computer.

[0019]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of a pattern defect detecting method and a pattern defect specifying method according to the present invention will be described below with reference to the drawings.

FIG. 1 shows an example of an image A of a pattern map displayed on the monitor of the review station. In FIG. 1, 1 is an element of a memory element after silicon nitride etching in an element isolation process of the memory device. The image of the pattern of the separation area is shown, and 2 is the image of the pattern defect made of particles or the like.
The image 2 of the pattern defect is composed of R, G, and B contrast values, and the R, G, and B contrast values are displayed in 256 gradations of 0 to 255 for each pixel.

An algorithm for detecting a pattern defect by extracting only the pattern defect image 2 from the pattern map image A shown in FIG. 1 will be described below with reference to FIGS. 2A to 2C. To do. FIG. 2 (a)
2 shows the image A of the pattern map shown in FIG.
2B shows a reference pattern block (hereinafter, referred to as a reference block) 3 having a normal pattern, and FIG.
(C) shows the image 2 of the extracted pattern defect.

FIG. 2A shows pattern blocks 11 to 24 in which the image A of the pattern map is divided into regions corresponding to the reference block 3, and each pattern block 11 to 24 is used for reference. An area in which the contrast values of the pattern blocks 11 to 24 and the contrast value of the reference block 3 are different is recognized as a pattern defect by sequentially comparing with the block 3, and is extracted as an image 2 of the pattern defect shown in FIG. 2C. . As a method for extracting the image 2 of the pattern defect, instead of extracting the areas having different contrast values, the area having a difference in contrast values smaller than a predetermined value may be extracted.

FIG. 3 shows the image of each of the pattern blocks 11 to 24 shown in FIG. 2 as the value of the R contrast for each pixel, and an area other than the area of the image 2 of the pattern defect, that is, a pattern defect occurs. The value of 125 is displayed in the area where no pattern is formed, but the pattern defect image 2
In the area of, a value larger than 125 is displayed. Although not shown, the R contrast value for each pixel of the image of the reference block 3 is all displayed as 125. By comparing the R contrast value of each pixel of each pattern block 11 to 24 with the R contrast value of each pixel of the reference block 3, and recognizing and extracting regions having different contrast values as pattern defect regions, Only pattern defects can be extracted.

Next, although the display of the G and B contrast values is omitted, the G and B contrast values of each pixel of each pattern block 11 to 24 are used as the R and B contrast values of each pixel of the reference block 3. Regions having different R and G contrast values are also extracted by comparison with the respective values.

FIGS. 4A, 4B and 4C show the average of the contrast values of the image 2 of the pattern defect extracted by the above-mentioned algorithm, with the horizontal axis representing contrast and the vertical axis representing the number of pixels. The histogram of the values is R, G,
It is shown for each B. Each of the R, G, and B histograms is unique to the type of pattern defect.
Next, the average value of the contrast values is calculated for each of the R, G, and B histograms, and from the database of pattern defects that have occurred in the past and whose generation process, generation device, generation cause, etc. are known, By selecting a pattern defect having an average value that is closest to the average value of the R, G, and B contrasts, the generation process, generation device, generation cause, etc. of the detected pattern defect are specified and output.

FIG. 5 shows a method for retrieving the image closest to the image 2 of the pattern defect having the average value of the R, G, and B contrast values obtained by the above algorithm from the database. In FIG. 5, X _r is the average value of the R contrast values of the image information in the failure analysis case in the database, X _r1 is the average value of the R contrast values of the image information of the detected pattern defect, and X _g is the database. Average value of the R contrast value of the image information in the defect analysis example, X _g1 is the average value of the R contrast value of the image information of the detected pattern defect, and X _b is the R value of the image information in the defect analysis example in the database. , X _b1 represents the average value of the R contrast values of the image information of the detected pattern defect.

First, the average value of the R, G, and B contrast values of the image information of the detected pattern defect (X _r1 ,
X _g1 , X _b1 ) and the average value of the contrast (X _r , X _g , between the image information of the failure analysis cases in the database)
_Xb ) is calculated by calculating the average value of the squares of the differences between R, G, and B: ΔX, and searching the database for a failure analysis case with the minimum average value: ΔX. The generation process, generation device, generation cause, etc. of the generated pattern defect are specified.

As described above, according to this embodiment, the contrast value of each pixel of the monitor image is 2 for each of R, G, and B.
Expressed in 56 gradations, the G, G, B contrast values of each pixel of each pattern block 11-24 are compared with the G, G, B contrast values of each pixel of the reference block 3 consisting of the normal pattern, A step of detecting a portion having a different contrast value as a pattern defect, an average value of the contrast values of the detected pattern defects for each of R, G, and B, and a contrast value of the R, G, and B of the pattern defect of the past database. Since it has a step of comparing the average value, it becomes possible to perform the identification work of the defect occurrence reason of the pattern defect, which was conventionally done manually, by the computer. The reason for the defect occurrence can be specified in a short time.

[0029]

According to the pattern defect specifying method of the first aspect of the present invention, the reason for the defect occurrence of the pattern defect is specified by comparing the color contrast values.
Since the cause of the defect occurrence can be specified by the computer without relying on human hands, the cause of the defect occurrence of the pattern defect can be specified objectively and accurately.

According to the pattern defect identifying method of the second aspect of the present invention, since the contrast value of at least one color in the image of the past pattern defect generated in the manufacturing process of the semiconductor device is stored in the database, the actual manufacturing process is performed. It is possible to create a database according to.

According to the pattern defect identifying method of the third aspect of the present invention, the average value of the gradation values of the gradation of RGB for each pixel of the image of the past pattern defect and each pixel of the image of the generated pattern defect. Since the cause of the defect occurrence of the pattern defect is specified by comparing with the average value of the gradation values of the light and shade of the contrast in each of RGB, the cause of the defect occurrence of the pattern defect can be specified more accurately.

According to the pattern defect specifying method of the fourth aspect of the present invention, the average value of the gradation values of the gradation of RGB for each pixel in the image of the past pattern defect and each pixel of the image of the generated pattern defect. RGB with the average value of the gradation values of contrast for each RGB in
Since the defect occurrence reason of the past pattern defect in which the average value of the square of the difference for each is the smallest is the defect occurrence reason of the pattern defect, it is possible to more accurately specify the defect occurrence reason of the pattern defect. it can.

According to the pattern defect detecting method of the fifth aspect of the present invention, since the pattern defect is detected by comparing the color contrast values with each other, the pattern defect can be detected by the computer. In addition, the pattern defect can be detected accurately.

[Brief description of drawings]

FIG. 1 is a diagram showing an example of a pattern image displayed on a monitor of a review station according to an embodiment of the present invention.

2A is a diagram showing a pattern block in which a pattern image displayed on a monitor of the review station is partitioned, and FIG. 2B is a diagram showing a reference block having no pattern defect; (C) is a figure which shows the image of the detected pattern defect.

FIG. 3 is a diagram showing an R contrast value for each pixel of the image of the pattern block.

4A and 4B show histograms of the average value of contrast values of pattern defect images, where FIG. 4A is an R histogram, FIG. 4B is a G histogram, and FIG. 4C is a B histogram.

FIG. 5 is a diagram showing a process of identifying a defect occurrence reason of a pattern defect in the one embodiment.

6A and 6B are diagrams illustrating a conventional pattern defect identification method.

[Explanation of symbols]

 A pattern map image 1 element isolation region pattern image 2 pattern defect image 3 reference pattern block 11-24 pattern block

Claims (5)

[Claims] 1. A pattern defect identification method for identifying a defect occurrence event comprising at least one of a pattern defect occurrence step, a generation device and an occurrence cause in a semiconductor device manufacturing process, which has occurred in the past. And a database preparation step of preparing a database that stores past pattern defect contrast information including the contrast value of at least one color in the image of the past pattern defect, which is a pattern defect whose cause of defect is known. A pattern defect detection step of detecting a generated pattern defect, which is a pattern defect generated in the manufacturing process of the device, and calculation of generated pattern defect contrast information consisting of the contrast value of the at least one color in the image of the detected generated pattern defect. Contrast calculation process And a defect occurrence reason specifying step of specifying the defect occurrence reason of the generated pattern defect by comparing the generated pattern defect contrast information and the past pattern defect contrast information. Defect identification method. 2. The database manufacturing step comprises performing a process such as film peeling on the past pattern defects generated in a semiconductor device manufacturing process, and then performing a contrast of the at least one color in an image of the past pattern defects. 2. The pattern defect identification method according to claim 1, further comprising the step of: determining the value of ## EQU1 ## and storing the obtained contrast value in the database as the past pattern defect contrast information. 3. The past pattern defect contrast information is RG in each pixel of the image of the past pattern defect.
The generated pattern defect contrast information is an average value of the numerical values of the gradation of contrast of each RGB in each pixel of the image of the generated pattern defect. The pattern defect identification method according to claim 1, wherein the pattern defect identification method is provided. 4. The defect occurrence reason specifying step comprises: an average value of gradation values of contrast gradation for each RGB in each pixel of the image of the past pattern defect, and RGB of each pixel of the image of the generated pattern defect. The average value of the squares of the differences for each RGB from the average value of the gradation values of each contrast is calculated, and the defect occurrence reason of the past pattern defect having the smallest average value of the calculated squares is determined. The pattern defect identifying method according to claim 1, further comprising a step of identifying the cause of the defect of the generated pattern defect as a cause of the defect. 5. A step of dividing a semiconductor wafer into a plurality of pattern blocks having a repetitive pattern, and a pattern block contrast value which is a numerical value of gradation of shade of color contrast for each pixel of an image of the plurality of pattern blocks. And a step of determining the pattern block contrast value and a reference block contrast value that is a numerical value of a gray scale of contrast of color of each pixel of a reference block having the repeating pattern and having no pattern defect A pattern comprising: a step of obtaining a difference; and a step of specifying an area in which a difference between the pattern block contrast value and the reference block contrast value is within a predetermined value as a pattern defect occurrence area. Defect detection method.