JP2005079274A - Pattern defect inspection method and apparatus thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for inspecting a pattern within a short period of time, which is taken by two times of imaging, namely the pattern which is taken only at a part of image through reduction of the number of times of imaging, by sythesizing images with the image process and then utilizing the synthesized image for the inspection. <P>SOLUTION: This method can reduce the time required for inspection through reduction in the number of times of imaging, by synthesizing the images taken by a plurality times of imaging and then using the same images for inspection. Moreover, this method conducts the inspection by comparing the images obtained by a plurality times of imaging and the image in the same position of the standard image. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a method for inspecting a pattern formed on a semiconductor substrate, a circuit board or the like, and an apparatus using the method.

  As a method for inspecting a pattern formed on a semiconductor substrate, a circuit board or the like, there is a method for inspecting the presence / absence of a defect by performing a density difference between an inspection image and a reference image. (For example, refer to Patent Document 1). This conventional technique will be described with reference to FIGS. 9, 10 and 11. FIG.

  FIG. 9 shows a configuration of a conventional pattern inspection apparatus. The image sensor 11 is provided above the inspection object 12 and images a pattern 22 (not shown) existing on the surface of the inspection object 12. The inspection object 12 is installed on a table 65, and can be moved in a horizontal plane by an XY stage 66 installed in the lower part thereof. The image sensor 11 relatively moves on the inspection object 12 and images all the surface patterns 22 on the inspection object 12. In order to inspect the pattern 22 with high accuracy, the moving speed of the XY stage 66 is set to be slow. Therefore, it takes a considerable time for the measurement. The captured image is subjected to image processing by the image processing unit 67 and then recorded in the inspection data recording unit 63 as an inspection image. On the other hand, an image of a non-defective pattern is recorded in advance in the reference data recording unit 64 as a reference image, and defect inspection is performed by sequentially comparing the recorded inspection image with reference pixels. The comparison is made by taking a density difference between the image data. The entire device is controlled by the control device 62. The reference image may be created from design data or a non-defective image may be used.

  FIG. 10 illustrates the positional relationship between the inspection object 12 of FIG. 9 and the imaging region 19 of the image sensor 11. FIG. 10 shows the positional relationship among the imaging region 19, the inspection object 12, the image sensor movement locus 13, the image sensor passage overlapping region 16, and the pattern 22. A pattern 22 of “H” exists on the inspection object 12. The imaging region 19 covers the inspection object 12 and captures an image of all patterns 22 on the inspection object 12. At this time, by setting the image sensor passing overlap area 16 so that the shooting areas 19 overlap, it is set so that all the “H” patterns 22 can be shot as one “H” pattern 22 without being divided. (The meaning of the division will be described with reference to FIG. 12).

  FIG. 11 shows the entire process flow. First, in the alignment step 31, the inspection object 12 is aligned with a predetermined position on the XY table 66. Next, in the imaging step 32, the image sensor 11 moves and images the pattern 22 on the inspection object 12. In the image processing step 33, the image data taken by the image sensor 11 is subjected to binarization processing, leveling, and smoothing processing in the image processing section 67, and finally one single “H” pattern 22 is extracted. Is done. Details will be described with reference to FIG. In the recording step 34, the one extracted “H” pattern 22 is recorded in the inspection data recording unit 63. The comparison step 35 compares (differs) the density difference between the image data recorded in the inspection data recording step 63 and the reference image data previously recorded in the reference data recording step 64. In the determination step 36, the compared data is compared with a certain threshold value to determine whether it is a defect. The display step 37 displays the determination result on a display (not shown) of the control device 62. Along with photographing, image processing is sequentially performed and defect inspection is performed.

Next, image processing will be described with reference to FIG. FIG. 12 shows an actually captured image. This image is extracted from the image photographed by the image sensor 11 so that the “H” pattern 22 is in the center in the horizontal direction of the image. There are a pattern 22 and a pixel 21 without a pattern. In FIG. 12, there are two “H” patterns 22. These two patterns 22 are extracted and separated into one “H” pattern 22, which are shown in FIGS. 13 (a) and 13 (b). FIG. 13A and FIG. 13B are extracted from FIG. 12 so that the “H” pattern 22 is in the center in the vertical direction of the image. A reference “H” pattern 22 is shown in FIG. The density difference between the inspection pattern image (FIGS. 13A and 13B) and the reference pattern “H” image (FIG. 13C) is determined to determine the presence or absence of a defect. Obviously, the defective portion 70 in FIG. 13B can be detected. At this time, a part of the “H” pattern 22 is photographed in the lower part of FIG. 12. This partially photographed “H” pattern 22 is “H” during the subsequent photographing. The entire pattern 22 is photographed and inspected in the same manner as described above.
JP 2002-022421 A (Page 2, [0014])

  However, in the conventional method, when the image is not photographed as one “H” pattern 22 (the remaining pattern in the lower part of FIG. 12), the pattern 22 is photographed again, so that it takes an extra photographing time. Had problems. In particular, high-accuracy pattern inspection requires high-accuracy imaging and requires a lot of imaging time.

  The present invention solves this problem, and an image obtained by synthesizing an image by image processing for an image obtained by dividing one pattern into two images, that is, an image obtained by shooting only one part. It is an object of the present invention to provide a method of reducing the number of times of photographing and inspecting in a short time by using for inspection.

  In order to achieve the object described above, the following inventive method is used. The method of the first invention is a method of detecting a defect by comparing an inspection pattern obtained by photographing a pattern on an inspection object and a reference pattern, and was photographed separately in two different images. A step of estimating a common part of a pattern, a step of averaging the common part or deleting one of the common parts, and a step of extracting a defect by comparing the inspection pattern with a reference pattern A pattern defect inspection method consisting of:

  Furthermore, it is preferable to estimate the common part of the images by combining the feature points of the two images. Furthermore, it is preferable to extract a sequence of images from the mating surfaces of the two images, determine the matching of the images, and estimate the common part of the images.

  The method of the second invention is a method for detecting a defect by comparing an inspection image obtained by photographing a pattern on an inspection object with a reference image, and for an inspection image divided by photographing. A pattern defect inspection method is used which includes a synthesis step of creating a corresponding reference image and an extraction step of comparing the inspection image with the corresponding reference image to extract a defect.

  Further, as a compositing step, a compositing that creates a corresponding reference image by estimating the common part of the image from the feature points of the two images of the inspection image and the reference image and aligning the position of the common part in the image. It may be a process. Also, as a compositing step, each image row is extracted from the joining surface of the two images of the inspection image and the reference pixel, the matching of the images is judged, the common part of the image is estimated, and the image of the common part It is good to set it as the process of producing a corresponding reference image by matching the position in the inside.

  The method of the third invention is a method for detecting a defect by comparing an inspection image obtained by photographing a pattern on an inspection object with a reference image, and for an inspection image divided by photographing. Then, a pattern defect inspection method including a synthesis step for creating a corresponding reference image and an extraction step for comparing the inspection image with the corresponding reference image to extract a defect is used.

  Further, the synthesis step estimates the common part of the image from the feature points of the two images of the inspection image and the reference image, and creates a corresponding reference image by matching the position of the common part in the image. It may be a synthesis process.

Further, the synthesis step extracts a sequence of each image from the joining surface of the two images of the inspection image and the reference pixel, judges the coincidence of the images, estimates the common part of the image, It may be a step of creating a corresponding reference image by matching the positions in the image.
A method for detecting defects by comparing an inspection image obtained by photographing a pattern on an inspection object with a reference image,
An estimation process for estimating a common part between the inspection image and the reference image;
An extraction step of extracting the common part from the inspection image and the reference image, respectively.
The extracted common part is compared, and a pattern defect inspection method comprising an extraction process for extracting defects is used.

  Further, as the estimation step, a step of estimating the common part of the image by combining the feature points of the two images of the inspection image and the reference image is preferable. In addition, it is preferable that the estimation step is a step of extracting a sequence of images from the joining surface of two images of the inspection image and the reference image, judging the coincidence of the images, and estimating the common part of the images.

  In order to achieve the object described above, the following apparatus of the present invention is used.

  In an apparatus for detecting a defect by comparing an inspection image obtained by photographing a pattern on the inspection object and a reference image, an imaging unit that is located above the inspection object and images the inspection object A pattern comprising: a moving mechanism unit that relatively moves the object to be inspected and the imaging unit; an image device unit that performs image processing on image data captured by the imaging unit; and a control unit that controls the moving mechanism unit A defect inspection apparatus, wherein the image apparatus unit uses a pattern defect inspection apparatus including an inspection data recording unit, a reference data recording unit, and an image composition unit.

  As described above, according to the inspection method and apparatus of the present invention, the number of times of photographing can be reduced by estimating a common part between images photographed separately and synthesizing the inspection images and comparing them with the reference image. As a result, the inspection time can be shortened.

  In addition, by newly creating a reference image for the divided inspection images and comparing and inspecting both images, the number of times of photographing can be reduced, and as a result, the inspection time can be shortened.

Further, by extracting the common part 20 of the inspection image and the reference image that are separately photographed and comparing them, the number of times of photographing can be reduced, and as a result, the inspection time can be shortened.
As a result, the inspection time required for inspecting the pattern formed on the surface of the semiconductor substrate, the circuit board, or the like can be shortened, and the productivity can be improved.

(Embodiment 1)
Hereinafter, Embodiment 1 of the present invention will be described with reference to FIGS. The apparatus used in Embodiment 1 is shown in FIG. Details of the conventional apparatus shown in FIG. 9 are omitted. The apparatus of this embodiment is different from FIGS. 1 and 9 in that an image composition unit 69 is installed. The image composition unit 69 synthesizes one inspection image from two inspection images. Details will be described with reference to FIGS. In the apparatus of FIG. 1, the positional relationship between the inspection object 12 and the image sensor 11 of the present embodiment is shown in FIG. The image sensor movement locus 13 indicates a locus on which the image sensor 11 moves for the first time, and the image sensor movement locus 14 indicates a path on which the image sensor 11 moves on the inspection object 12 for the second time. The difference from FIG. 10 showing the movement of the conventional image sensor 11 is that the number of movements is reduced because the area of the image sensor passing overlap region 16 is small. FIG. 3A shows an image captured and extracted by the first image sensor movement locus 13 of FIG. FIG. 3B is an image taken and extracted along the path of the second image sensor movement locus 14 of FIG. In FIG. 3A and FIG. 3B, one “H” pattern 22 as a whole is extracted as one “H” pattern 22 with respect to what is photographed, and FIG. Let b) extract. These are compared with the reference pattern of FIG. On the other hand, the pattern is not a single “H” pattern 22, but is divided into the lower part and the upper part of FIG. 2), the common portion 20 photographed in both figures is corrected to synthesize a pattern, and an inspection image as shown in FIGS. 3A and 3B is synthesized. The correction method will be described below.

  Next, the process flow of this embodiment is demonstrated using FIG. Only the difference from the conventional flowchart of FIG. 11 will be described. When the pattern is extracted in the pattern extraction step 33, the pattern extracted as one “H” pattern 22 proceeds to the recording step 34 and the comparison step 35 as in the conventional case. On the other hand, in the pattern extraction step 33, the one obtained by dividing one “H” pattern 22 (lower part of FIG. 3A) is recorded in the temporary recording step 38. After that, when an image is acquired by the next shooting, the pattern is divided and shot (upper part of FIG. 3B) is recorded in the temporary recording step 38 as described above. When these two images are recorded in the temporary recording step 38, they are transferred to the pattern synthesis step 39, where the common portion 20 between the two images is estimated and corrected, and one pattern is synthesized. This step is processed by the image composition unit 69 in FIG. Thereafter, the pattern is recorded in the recording step 34, and is inspected by going through the same inspection step as before.

  Next, an image composition method (correction method) will be described. FIG. 5A and FIG. 5B show image portions taken by dividing patterns. These figures are extracted from the lower part of FIG. 3 (a) and the upper part of (b).

  First, as a first method, a method of setting feature points in a pattern and synthesizing an image will be described with reference to FIGS. 5 (a) and 5 (b). In FIG. 5A, feature points 41 and 42 are selected as characteristic points in the image. These points are intersections that are the maximum density values from the intensity distribution of the image density shown in the right and lower diagrams of FIG. 5A, and are features that are easy to detect. Similarly, feature points 41 and 42 are found in FIG. A pattern can be synthesized by correcting so that the coordinates of these feature points 41 and 42 are matched. Of course, if the pattern changes, it is necessary to reset the feature points 41 and 42. Further, the feature points 41 and 42 need to be located in the common part 20 of the two images. It is necessary to take an image in advance and set the feature points 41 and 42. The setting method of the feature points 41 and 42 is required to be easy to detect the location and not to be mistaken for other points. That is, as in this embodiment, the intersection where the patterns are concentrated is the best.

  When the feature points 41 and 42 cannot be set in the common unit 20, another method described below can be used. The correction of the common unit 20 is performed by deleting or averaging one of the common units 20.

  Next, as a second method, a method of aligning images from the edge of the image will be described with reference to FIGS. FIGS. 6 (a) and 6 (b) are digitized images of FIGS. 5 (a) and 5 (b). A method for estimating the common unit 20 will be described. First, images shown in FIGS. 7A to 7D are separately taken out from the boundary 55 between FIGS. 6A and 6B. FIG. 7A shows one row, FIG. 7B shows two rows, FIG. 7C shows three rows, and FIG. 7D shows four rows. These figures are compared to find the number of matching pixels. Table 1 shows the ratio of the number of columns and the number of matching pixels. Although five columns or more are not shown in FIG. 7, they all match. As the number of columns increases, the point where the coincidence ratio becomes 90% or more is defined as a common part. In the first column, all the pixels match, and the matching ratio is high and is 90% or more, but the matching ratio in the second column is not 90% or more, so it is not the common part 20. From Table 1, four rows of figures (FIG. 7D) can be determined as the common part 20.

  Next, the third method will be described with reference to FIG. This method is a method in which the number of columns to be compared is fixed from the beginning and compared in the second method. In this case, it was set to compare every three columns. The portion of the extraction region 51 in FIG. 6A and the extraction region 52 in FIG. 6B are compared, and if they do not match, the extraction region 53 and the extraction region 54 in FIG. Compare. The reference for comparison can be made by using the number of coincident pixels and the coincidence ratio as in the second method. Table 2 shows the number of matching pixels and the matching ratio. By setting the determination criterion to be 90% of the coincidence ratio, the result is that the extraction areas 51 and 53 coincide with each other, and the common unit 20 determines that the extraction area 53 and corrects the extraction areas 51 and 53 together. Now you can synthesize the pattern. In selecting the extraction regions to be compared, the optimum number of columns must be selected depending on the type of pattern. In the case of a simple pattern, a large number of matches is set, and in a complicated case, a small value is set. In addition, if the number of columns to be set is large, it takes time to determine whether the images match, and inspection time is required. If the number of columns to be set is small, there is a high possibility that the common unit 20 is erroneously estimated. Here, the extraction area 51 is divided into three columns. In the case of the pattern “H”, there is a feature in the horizontal bar portion, and the preceding and following columns including the column of this portion are set.

  In addition, from the beginning, it is possible to determine whether or not FIG. 6A and FIG. 6B match with each other while shifting the respective drawings, but whether or not the images are matched is wide. This image processing takes a considerable amount of time.

  Further, as a method for determining the coincidence of the images, a pattern matching algorithm such as a correlation method can be used in addition to the difference method described above. In the correlation method, since the coincidence between images is determined using a correlation function, an accurate determination can be made. In the difference method, the difference between the two images is taken to determine the coincidence based on the magnitude of the difference. Which one is adopted is selected depending on the object. In the case of a semiconductor, since it is necessary to inspect a pattern with high accuracy, a correlation method is used. On the other hand, in the case of a circuit pattern, since the pattern accuracy is not high, a difference method may be used.

  In the case of the correlation method, a value of 1.0 is obtained with perfect match. Since there is a case where perfect matching may not occur due to the presence of defects, dust, etc., the matching condition can be set to 0.8 to 1.0, for example. The range of this standard is determined for each object.

  Actually, Table 3 shows the ratio of the time when the transistor pattern on the surface of the 6-inch wafer was inspected by the above-mentioned three methods, the conventional method, and the method for judging the coincidence of all images. [1] The conventional method is a method of inspecting by increasing the number of photographing without synthesizing patterns, [2] to [4] are the methods described above, [2] a method using feature points, [3] A method for determining matching by changing the number of columns, [4] A method for determining image matching by a certain number of columns, [5] A comparative example shows two images (for example, FIG. 6A). This is a method in which the consistency of the entire image is judged, the patterns are synthesized, and the inspection is performed while shifting FIG. The difference in time between [1] and [2], [3], [4] is the difference in the number of times of shooting. That is, in the conventional [1], as shown in FIG. 10, shooting is performed 6 times, whereas as shown in FIG. 2, shooting is performed 4 times, and the difference in shooting time appears in Table 3. ing. The process for estimating the common unit 20 is not shown because the control processing capability of the control device 62 is high. However, in the case of checking the coincidence of the entire images in [5], the calculation processing time is long, so the shooting time is long.

  From Table 3, the inspection time can be shortened in the methods [2] to [4] of the present embodiment as compared with the conventional method. As a result, the number of inspections can be increased even at the same time, and the yield rate can be improved.

(Embodiment 2)
Hereinafter, Embodiment 2 of the present invention will be described with reference to FIGS. FIG. 5A shows what is shown in the lower part of FIG. 2, in which the pattern “H” was not photographed completely. In the first embodiment, one complete pattern “H” is synthesized from FIG. 5A and FIG. 5B, which is an image combined with the image, and used for inspection. In the second embodiment, a method for inspecting by comparing FIG. 5A with a newly synthesized reference image, or extracting a common portion from FIG. 5A and the reference image, and comparing them. Is a method of inspection.

  For example, in FIG. 8A showing the inspection image, the coordinates of the feature points 41 and 42 are set, and the coordinates of the feature points 41 and 42 of the pattern of FIG. Data conversion is performed so that a new reference image shown in FIG. 8C is synthesized. A defect can be inspected by comparing FIG. 8C of this new reference image with FIG. Alternatively, a method may be used in which FIG. 8A is compared with the reference image FIG. 8B, the common part is estimated, and a new reference image FIG. 8C is synthesized and inspected.

  As another method, the common part of FIGS. 8A and 8B can be estimated, and the common part figure 8D can be extracted and compared. The method for estimating the common part can use the three methods shown in the first embodiment. That is, a method of obtaining by feature points, a method of obtaining by comparing each column, and a method of obtaining by comparing each determined column can be used.

  Since the flowchart of this embodiment is the same as that of FIG. 3 of the first embodiment, only the differences will be described. In the temporary recording step 38, the pattern extraction step 33 records the one “H” pattern 22 divided (lower part of FIG. 3A). In the first embodiment, processing is performed after the next divided image is recorded, but in the second embodiment, the process immediately proceeds to the pattern synthesis step 39. In the pattern synthesis step 39, the common portion 20 between the two images to be inspected in the first embodiment is estimated and corrected, and one pattern is synthesized. In this second embodiment, the reference image and the inspection image are compared. Extract common parts. Alternatively, a new reference image is synthesized in accordance with the inspection image. Thereafter, these patterns are recorded by the recording unit 34, and are inspected by proceeding with the inspection process as in the prior art.

  In the second embodiment, the same effect as in the first embodiment can be obtained.

  As described above, the pattern defect inspection method and apparatus according to the present invention can use a pattern formed on a semiconductor substrate, a circuit board or the like for inspection.

The figure which shows the apparatus structure of embodiment of this invention. The figure which shows the direction of the relative movement of the image capture image sensor and test article of embodiment of this invention The figure which shows the image image | photographed by embodiment of this invention. Flowchart of an embodiment of the present invention The figure explaining embodiment of this invention The figure explaining embodiment of this invention The figure explaining embodiment of this invention The figure explaining embodiment of this invention The figure which shows the conventional pattern inspection apparatus constitution The figure which shows the relative movement direction of the conventional image capturing image sensor and the inspection object Conventional Embodiment Flowchart Figure showing a conventional photographed image Figure of conventional extracted image

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Image sensor 12 To-be-inspected object 13 Image sensor moving area 15 Image sensor moving locus 16 Image sensor passage overlap area 19 Imaging area 20 Common part 22 Pattern 41, 42 Feature point 51-54 Extraction area 55 Matching surface 61 Imaging means 62 Control apparatus 63 Inspection Data Recording Unit 64 Reference Data Recording Unit 65 Table 66 XY Table 67 Image Processing Unit 69 Image Processing Device

Claims (10)

A method for detecting defects by comparing an inspection image obtained by photographing a pattern on an inspection object with a reference image,
An estimation process for estimating the common part between inspection images taken separately,
A synthesizing step of synthesizing an inspection image by averaging the common part or deleting one,
A pattern defect inspection method comprising an extraction step of comparing the inspection image with a reference image and extracting a defect. The pattern defect inspection method according to claim 1, wherein the estimating step is a step of estimating a common portion of the images by combining the feature points of the two images. The pattern defect inspection method according to claim 1, wherein the estimating step is a step of extracting a sequence of images from the mating surfaces of the two images, judging the coincidence of the images, and estimating a common part of the images. A method for detecting defects by comparing an inspection image obtained by photographing a pattern on an inspection object with a reference image,
A synthesis step for creating a corresponding reference image for an inspection image divided by photographing,
An extraction step of comparing the inspection image with the corresponding reference image and extracting defects;
A pattern defect inspection method comprising: A compositing step is a compositing step of estimating a common part of an image from each feature point of two images of an inspection image and a reference image, and creating a corresponding reference image by aligning the positions of the common part in the image. The pattern defect inspection method according to claim 4. The compositing step extracts a sequence of each image from the joining surface of the two images of the inspection image and the reference pixel, determines the coincidence of the images, estimates the common part of the image, and within the image of the common part 5. The pattern defect inspection method according to claim 4, which is a step of creating a corresponding reference image by aligning the positions of. A method for detecting defects by comparing an inspection image obtained by photographing a pattern on an inspection object with a reference image,
An estimation process for estimating a common part between the inspection image and the reference image;
An extraction step of extracting the common part from the inspection image and the reference image, respectively.
A pattern defect inspection method comprising an extraction step of comparing the extracted common parts and extracting defects. The pattern defect inspection method according to claim 7, wherein the estimating step is a step of estimating a common portion of the images by combining the feature points of the two images of the inspection image and the reference image. 9. The pattern according to claim 8, wherein the estimating step is a step of extracting a sequence of images from the joining surface of the two images of the inspection image and the reference image, determining the coincidence of the images, and estimating a common part of the images. Defect inspection method. In the apparatus for detecting defects by comparing the inspection image obtained by photographing the pattern on the inspection object and the reference image,
An imaging unit that is located above the inspection object and images the inspection object;
A moving mechanism for relatively moving the object to be inspected and the imaging unit;
An image device unit that performs image processing on image data captured by the imaging unit;
A pattern defect inspection apparatus comprising a control unit for controlling the moving mechanism unit,
A pattern defect characterized in that the image device unit comprises an inspection data recording unit for recording a photographed image, a reference data recording unit for recording a reference image, and an image combining unit for combining two images Inspection device.

Images