JPH0798281A - Method and apparatus for detecting defect in pattern - Google Patents

Abstract

PURPOSE:To detect the defect in a pattern, which imparts a great deal of adverse effects at high reliability and at a high speed even if there is dispersion in height of the pattern on a material under inspection and there are noise irregularities at the pattern edge. CONSTITUTION:With respect to whole light from a pattern, which is accurately formed with space filters 12a and 12b as designed, the light cannot be transmitted through the space filters 12a and 12b. With respect to the light from a pattern, wherein abnormality has occurred, the light is transmitted through the space filters 12a and 12b, and the images are focused on image sensors 3a and 3b. Since there are noise irregularities at the pattern edge and there is dispersion in height of the pattern, however, the irregular images close to noises appear on the image sensors 3a and 3b. Therefore, both images of image sensors 3a and 3b are compared in an image processor 14, and the defective part is extracted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern defect detection method and apparatus, and more particularly to a method and apparatus for detecting regular and fine pattern defects formed on the surface of a semiconductor wafer, liquid crystal or the like.

[0002]

2. Description of the Related Art First, a typical method in the conventional pattern defect detection method will be described below. Firstly, the spatial filter method is used as reported by Yoko Miyazaki and two other members of the Institute of Industrial Science, Mitsubishi Electric Corp. in P707 to P708 of the academic conference presentations of the Precision Engineering Society Spring Conference in 1988. There is a pattern defect detection method. Hereinafter, this method is referred to as a first conventional example.

In this method, diffracted light from a regular pattern of a semiconductor integrated circuit or the like is Fourier transformed by a lens,
This Fourier transform pattern is recorded on a photograph, and by using this as a spatial filter, scattered light from a pattern defect can be selectively transmitted to be applied to pattern defect detection.

Actually, the detection is performed using a device as shown in FIG. Laser light from He-Cd laser 32 is mirrored 3
3, the wafer to be inspected 37 is irradiated through the mirror 34, the collimator 35, and the half mirror 36, and this image is projected onto the lens 38.
And is detected by the camera 39 placed at the image forming position. Then, the detected image is recognized by the image processing device 40. A spatial filter 41 made of a holographic photographic dry plate is installed on the focal plane of the lens 38. If the reflecting surface of the wafer to be inspected 37, which is a reflecting object, shines, the Fourier transform pattern deviates from the recording pattern on the spatial filter 41. Therefore, in order to remove the pattern on the measured wafer 37 and selectively detect only the defects, the spatial filter 4
1 has a structure in which the recording pattern on 1 and the Fourier transform pattern are aligned.

In this apparatus, the optical path is divided by the half mirror 42, and the position of the diffracted light and the position of the wafer 37 to be inspected are obtained. The position of the diffracted light is divided by the half mirror 43, the position of the 0th order diffracted light in the diffraction pattern of the light reaching the spatial filter 41 is detected by the camera 44 from the reflected diffraction pattern, and the position of the 1st order diffracted light is detected by the camera 45. The position is detected from the reflection diffraction pattern by.

Then, the deviation amount of the tilt angle and the deviation amount of the rotation angle with respect to the optical axis are calculated from the positional deviation amount between the positions of the respective diffracted light beams of the spatial filter 41 stored in advance and the actually detected respective diffracted light beams. This is corrected using the six-axis control stage 46.

Further, the actual position of the wafer 37 to be inspected is detected by the camera 48 by dividing the optical path by the half mirror 47. Second, as reported by Mr. Sadao Sasaki of Toshiba ULSI Laboratories, Inc. in P181 to P188 of the "Electronic Materials" magazine in 1984, this wafer to be inspected was inspected. There is a method in which the difference signal is recognized as a defect by enlarging and comparing the same portion of two specific chips among the chips existing in one mask by utilizing the periodicity of the mask pattern used for manufacturing. As typical examples, "5MD-25" manufactured by Nippon Automatic Control Co., Ltd. and "KLA-101" manufactured by KLA Co., Ltd. have been put into practical use. Hereinafter, this method is referred to as a second conventional example.

[0008]

However, the conventional pattern defect detecting method having the above-mentioned structure has the following problems. First, in the first conventional example,
Due to the imperfections in the manufacturing process of the wafer 37 to be inspected, the fine pattern usually has a pattern edge that does not form a complete straight line and has noise-like unevenness. Further, around the unit area 49 as shown in FIG. 6, the corners are rounded due to the imperfect manufacturing process of the wafer 37 to be inspected. Further, due to optical distortion, the optical image at the corner becomes round. The height of the pattern on the inspected wafer 37 also varies. For this reason, in the first conventional example, the reliability of defect detection is low, and it is not possible to detect the defect of the pattern edge, which has a great adverse effect on the characteristics of the wafer to be inspected 37, or the characteristics of the wafer to be inspected 37 are adversely affected. The roundness of the corners of the pattern or the variation in height on the uninspected wafer 37 may be detected as a defect.

In addition, there are many patterns with low regularity in the peripheral portion of the chip constituting the wafer to be inspected 37. In the method of the first conventional example, irregular diffracted light from this pattern is reflected by the lens. It becomes difficult to collect light and perform Fourier transform. Therefore, the recording pattern on the spatial filter 41 is also irregular, and the detection sensitivity is significantly reduced.

Further, in the second conventional example, it is necessary to compare two positions on a fine pattern to detect a pattern defect, and therefore it is necessary to position the two positions on the pattern with high precision. In addition, the pixels of the captured image for detecting the pattern defect are made as fine as the size of the pattern, and complicated calculation is required to detect only the defective portion other than the regular pattern. Therefore, the inspection requires a lot of time.

[0011]

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned technical problems. First, it detects defects in a regular pattern formed on a plate-like object. In the pattern defect detection method, the result of applying a spatial filter to the Fourier transform image of the regular pattern and the result of comparing optical images of the same pattern at a plurality of points on the regular pattern are used. The present invention provides a pattern defect detection method.

Secondly, in a pattern defect detecting method for detecting defects of a regular pattern formed on a plate-like object, optical images at odd places and a plurality of places on the regular pattern are compared, and this comparison is performed. A method for detecting a pattern defect, characterized in that the regular pattern having the same result and occupying a large number is identified as the accurate regular pattern, and the regular pattern having a defect is identified from the identified result. is there.

Thirdly, in a pattern defect detecting apparatus for detecting defects of a regular pattern formed on a plate-like object, a transforming means for Fourier transforming an optical image of the regular pattern, a spatial filter, and the regular rule. A pattern defect using a comparison result of the Fourier transform image obtained by the transforming means and the comparison result of the comparing means. The present invention provides a pattern defect detection device characterized by detecting

Fourthly, in a pattern defect detecting apparatus for detecting defects of a regular pattern formed on a plate-like object, a comparison means for comparing optical images at odd-numbered locations and a plurality of locations on the regular pattern, and this comparison. Recognizing means for making the regular pattern, which is equal and majority in comparison result by means, the accurate regular pattern, and identifies the defective regular pattern from the qualifying result of the recognizing means. And a pattern defect detecting device.

[0015]

According to the pattern defect detecting method of the present invention, even if the height of the pattern on the object to be inspected varies and the pattern edge has noise-like unevenness,
It is a highly reliable and high-speed detection of a pattern defect that has a great adverse effect.

[0016]

FIG. 1 shows a first embodiment of a pattern detecting apparatus using the pattern defect detecting method of the present invention. In FIG. 1, the inspection object 1 is a semiconductor wafer, a liquid crystal panel, or a regular pattern such as a mask for manufacturing these. The inspection object 1 is placed on the drive table 2.

The regular pattern formed on the inspection object 1 is designed so that the patterns separated by a distance l are the same pattern. The optical systems A and B are the inspection object 1
Sensor 3a corresponding to each of the above patterns projected
And an image is formed on the sensor 3b (dotted line in the figure).

The optical axis 4a of the optical system A and the optical axis 4b of the optical system B are separated by a distance 1 (dashed line in the figure).
The optical system A and the optical system B are provided with an adjusting mechanism (not shown) for matching the distance between the optical axes 4a and 4b with the distance l.

In the optical system A and the optical system B, when the inspection object 1 having a regular pattern has only an opaque portion such as a semiconductor wafer, only the monochromatic light sources 5a and 5b which emit monochromatic light are used. To use. Then, those lights are reflected by the half mirrors 7a and 7b, respectively, and are incident on the object 1 to be inspected along the optical axes 4a and 4b through the lenses 6a and 6b (the monochromatic light sources 5a and 5b are used as starting points). solid line).

The object to be inspected 1 having a regular pattern
When a liquid crystal panel, a semiconductor mask, a shadow mask or the like has a transparent portion and an opaque portion, only the monochromatic light sources 8a and 8b which are monochromatic light sources are used. Then, these lights are made into parallel lights by the lenses 9a and 9b, respectively, and the parallel light illuminates the DUT 1 along the optical axes 4a and 4b, respectively (the solid lines in the figure starting from the monochromatic light sources 8a and 8b). ).

Here, in the case of handling only the inspection object 1 having only the opaque portion, it is sufficient to provide only the illumination system by the monochromatic light sources 5a and 5b, and the inspection object having the pattern having the transparent portion and the opaque portion is provided. In the case of handling only one, it is sufficient to provide only the illumination system by the monochromatic light sources 8a and 8b.

Further, the monochromatic light sources 5a, 5b, 8a and 8b are a combination of a laser light source or a white light source such as a mercury lamp and an interference filter. Now, the inspection object 1
The light 10a, 10b reflected, transmitted, scattered, or diffracted by the inspection object 1 is reflected by the lens 6a both when the inspection object 1 has only an opaque portion and when the inspection object 1 has a pattern having a transparent portion and an opaque portion.・ It is condensed at 6b. Spatial filters 12a and 12b are provided at the focal positions of the lenses 6a and 6b. On the other hand, the lights 11a and 11b, which are images of the inspection object 1, are reflected by the spatial filters 12a and 1a.
The pattern on the inspection object 1 is imaged on the image sensors 3a and 3b by passing through the lenses 13a and 13b after passing through 2b.

The image data from both of the image sensors 3a and 3b are amplified and both are input to the image processing device 14. The output from the image processing device 14 is input to the inspection result output device 15.

The process of defect detection in this embodiment will be described below. First, when the inspection object 1 is illuminated with monochromatic light, the inspection object 1
Diffracted light is generated from each of the above regular patterns. These diffracted lights, including the 0th order light, are condensed into regular point images on the focal planes of the lenses 6a and 6b. This phenomenon is optically called the Fourier transform of the pattern.

Spatial filter 12a placed on this focal plane
As shown by 12 in FIG. 2, 12b is opaque only at the position where the above-mentioned regular point image is generated and the central portion of the light (the cross-shaped portion in FIG. 2), and the other portion. Is transparent. The spatial filters 12a and 12b are manufactured by using a non-defective item 1 to be inspected having a regular pattern accurately formed as designed in advance. Specifically, the spatial filters 12a and 12b are exposed to the diffracted light from the regular pattern on the object to be inspected 1 or the portion where the regular point image is focused is made opaque by using a computer.

All the light from the pattern accurately formed as designed by the spatial filters 12a and 12b cannot pass through the spatial filters 12a and 12b. However, the light from the abnormal pattern is transmitted through the spatial filters 12a and 12b and the image sensor 3a.
An image is formed on 3b.

However, even in a regular fine pattern on the object 1 to be inspected, which is allowed to have no abnormality, the pattern edge is not normally formed as a complete straight line due to the imperfection of the manufacturing process, and the unevenness like noise is generated. Is occurring. Also, the height of the pattern varies. Therefore, an irregular image close to noise appears on the image sensors 3a and 3b. Therefore, the images of both the image sensors 3a and 3b are compared in the image processing device 14 by the following method to extract the defective portion.

(1) Accurately align both images of the image sensors 3a and 3b, calculate the difference in light intensity between the two, and if the difference is equal to or greater than a predetermined value C ₁ , mark that part as defective. to decide. However, the predetermined value C ₁ is set by a person in advance, or is set by automatically calculating γ ₁ times the standard deviation σ ₁ of the intensity distribution of the image, that is, C ₁ = γ ₁ σ _1. To do so. However, here too, γ ₁ will be set by humans in advance.

Even when the difference in light intensity becomes C ₁ or more, it is not possible to determine which of the image sensors 3a and 3b has the defective portion. Therefore, at the position when the difference in light intensity becomes C ₁ or more, it is considered that there is a defective portion in both patterns, and the position on both patterns is displayed by the inspection result output device 15. Become.

It is also conceivable to move the drive table 2 further by 1 and compare the third image on the pattern with the inspected image. This third image and the image sensor 3a
The difference C ₁ in the light intensity between the two images measured in 3b is measured again, and the defective portion is limited to the position of the inspected image portion where the difference in the light intensity is C ₁ or more. You can

(2) In the method of (1) above, the image sensor 3a
Although both images of 3b are accurately aligned and the difference in light intensity between them is calculated, the difference between the differential values of both light intensity distributions may be calculated. When the difference between the differential values is a predetermined value C ₂ or more, the portion is determined as a defective portion in the same manner as the method (1).

Here, the predetermined value C ₂ is determined by a human being in advance or by the standard deviation σ of the intensity distribution of the image.
₂ gamma ₂ fold that is, to be set automatically computed and C ₂ = _{γ 2} σ _2. However, here too, γ ₂ will be set by humans in advance.

In practice, either (1) or (2) or (1) and (2) may be calculated at the same time. With the above-described configuration, it is possible to greatly improve the low reliability, which is the greatest drawback of the pattern defect detection method using the spatial filter method as in the first conventional example.

Further, by utilizing the periodicity of the mask pattern used for manufacturing the wafer to be inspected as in the second conventional example, the same portion of two specific chips among the chips existing in one mask is enlarged and compared. In the case of recognizing the difference signal as a defect, very accurate positioning is required, but according to the above-mentioned method, everything from the pattern accurately formed as designed at the time of passing through the spatial filters 12a and 12b. Since there is almost no such light, it is not necessary to make the pixels of the captured image as fine as the size of the pattern in order to detect pattern defects.
Since this pixel is large, it is not necessary to make the positioning accuracy as high as in the conventional case.

Here, when the peripheral portion of the chip forming the wafer to be inspected 21 is the object, 1 can be set to a large value and the inspection can be performed by comparing with the same portion of other chips. Then, when a highly regular portion such as a memory in a chip constituting the wafer to be inspected 21 is targeted, the inspection can be performed by setting l to be small and comparing with the same pattern portion in the same chip.

Further, since the pixels are large, the calculation time required for the inspection is shortened, so that the inspection speed itself is increased. Since most regular pattern images are not detected, the defect detection calculation method itself becomes simple.

In this embodiment, two optical systems are used, but by adding an image storage unit, the second optical system
It is also possible to inspect using one optical system as described in the above embodiment. First, the image data is fetched and the image data is stored in the image storage unit, and then the drive table 2 is compared with the image data when the drive table 2 is moved by l or a multiple of l. Alternatively, the defective portion may be extracted and the same processing as that in the above-described embodiment may be performed.

Further, the image sensors 3a and 3b may be ordinary television cameras, or line sensors may be used to collect image data in synchronization with the movement of the drive table 2.

Next, a second modification of the first embodiment will be described.
Examples of are described below. This embodiment has the same basic configuration as that of the first embodiment. However, as in the case of the first embodiment, a comparison is made by enlarging and comparing the same portions of specific two chips among the chips existing in one mask. Instead, the same portion of a plurality of odd-numbered specific chips in the chip is enlarged and compared, and the defective portion is detected by using the “majority decision algorithm” described later.

FIG. 3 shows a second embodiment of the pattern defect detecting apparatus using the pattern defect detecting method of the present invention. The essence of the present embodiment is a method of comparing image data captured to detect a defect. The process and configuration of the defect detection in this embodiment other than this are the same as those in the above-described first embodiment and will not be described here.

In FIG. 3, the inspection object 16 is a semiconductor wafer, a liquid crystal panel, or a regular pattern such as a mask for manufacturing these. The inspection object 16 is mounted on the drive table 17.

The regular pattern formed on the inspection object 16 is designed so that the patterns separated by the distance 1 in the x direction and the y direction are the same pattern. The optical system C projects the pattern on the inspection object 16 and forms an image on the sensor 18 corresponding to this pattern (dotted line in the figure).

The optical axis 19 of the optical system C (one-dot chain line in the figure) can be moved on the object 16 to be inspected in the x and y directions by the drive table 17 by a distance l or a multiple of l.

In the optical system C, the monochromatic light source 2 which emits monochromatic light when the inspection object 16 having a regular pattern has only an opaque portion such as a semiconductor wafer.
Only 0 is used. Then, each of those lights is reflected by the half mirror 21, and is transmitted through the lens 22 to the optical axis 1.
The inspection object 16 is illuminated by epi-illumination along the line 9 (solid line in the figure starting from the monochromatic light source 20).

The object to be inspected 1 having a regular pattern
When 6 has a transparent part and an opaque part like a liquid crystal panel, a semiconductor mask, a shadow mask, etc., only the monochromatic light source 23 which is a light source which emits monochromatic light is used. And
The light is made parallel by the lens 24,
The parallel light respectively illuminates the inspection object 16 along the optical axis 19 (a solid line in the figure starting from the monochromatic light source 23).

Here, in the case of handling only the inspection object 16 having only the opaque portion, it is sufficient to provide only the illumination system by the monochromatic light source 20, and only the inspection object 16 having the pattern having the transparent portion and the opaque portion is provided. When dealing with, it is sufficient to have only the illumination system by the monochromatic light source 23.

Further, the monochromatic light sources 20 and 23 are a combination of a laser light source or a white light source such as a mercury lamp and an interference filter. Now, in both the case where the inspection object 16 has only an opaque portion and the inspection object 16 has a pattern having a transparent portion and an opaque portion,
The light 25 reflected, transmitted, scattered, or diffracted by the inspection object 16 is condensed by the lens 22. A spatial filter 27 is provided at the focal position of the lens 22. On the other hand, the light 26 representing the image of the inspection object 16 passes through the spatial filter 27 and the lens 28, so that the pattern on the inspection object 1 is imaged on the image sensor 18. Then, the image data from the image sensor 18 is stored in the image storage unit 29.

Specifically, first, the image data from the image sensor 18 is fetched and stored in the image storage section 29. Next, the drive table 2 is moved from this storage location in the x-direction or the y-direction by l or a multiple (however, odd number) times of l times to obtain plural (however, odd number) image data from the image sensor 18. . Then, these image data are compared in the same manner as in the first embodiment, the defective portion is extracted by the "majority decision algorithm" described later, and the same processing as in the first embodiment is performed. At this time, each image data is amplified and input to the image processing device 30. The output from the image processing device 30 is input to the inspection result output device 31.

Here, the "majority voting algorithm" will be described by taking the case of comparing three patterns as an example. According to this algorithm, it is easy to identify the defective portion without using design data that is the basis of pattern fabrication or by visual inspection.

FIG. 4 is a block diagram showing the flow of processing of the pattern defect detecting apparatus of this embodiment using the "majority decision algorithm" and will be described in detail below. First, the following three cases can be considered in detecting the difference between the image data of the three patterns. From these cases, a pattern having a defect is specified in the image processing apparatus 30 by taking a majority decision.

(1) If no difference is detected between the three patterns, it is considered that all three patterns have no defect.

(2) When a difference is detected between one of the three patterns and the other two patterns, one pattern in which the difference is detected between the other two patterns Is considered to be defective.

(3) When all the differences are detected between the three patterns, it is considered that all three patterns have defects.

That is, according to this method, a pattern having the same pattern comparison result and a large number of pattern comparison results is recognized as an accurate pattern, and a pattern having a defect is specified by this result. On the contrary, it is also possible to identify a pattern having different pattern comparison results and occupying a small number as a pattern having a defect, and to identify the pattern having the defect based on this result.

In addition to the above, in the case of (2) above, in the case of identifying a defective portion from a large number of patterns, it is appropriate to classify as defective products when there are too many patterns for which differences are detected.

It goes without saying that this "majority decision algorithm" can be used not only in the pattern defect detection method of the present invention but also in various comparison objects. Although one optical system is used by adding the image storage unit 29 in this embodiment, it is also possible to simultaneously inspect a plurality of patterns by using a plurality of optical systems as in the first embodiment. it can.
Further, the image sensor 18 may be an ordinary television camera, or a line sensor may be used to collect image data in synchronization with the movement of the drive table 17.

[0057]

By combining the pattern defect detecting method using the optical filter and the detecting method of calculating the difference between the same patterns at a plurality of locations, the normal pattern edge forms a complete straight line due to the imperfection of the manufacturing process. Not even, it has unevenness like noise, and the height of the pattern also varies, even for semiconductor wafers, liquid crystal panels, or those with regular patterns such as masks for manufacturing these , Pattern defects can be detected with high reliability at high speed and with simple calculation.

Further, it becomes possible to detect a defect of a pattern having a low regularity such as a peripheral portion of a semiconductor chip. Then, it became easy to identify the defective portion without using the design data that is the basis of the pattern fabrication or by visual inspection.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a first embodiment of the present invention.

FIG. 2 is a plan view of a spatial filter used in the present invention.

FIG. 3 is a schematic configuration diagram of a second embodiment of the present invention.

FIG. 4 is a second example of the present invention using a “majority voting algorithm”.
Block diagram showing the processing in the embodiment of FIG.

FIG. 5 is a schematic configuration diagram of a first conventional example.

FIG. 6 is a schematic view of an example of a unit area of an inspection pattern.

[Explanation of symbols]

1.16 ... Object to be inspected 2.17 ... Drive table 3a, 3b, 18 ... Image sensor 4a, 4b, 19 ... Optical axis 5a, 5b, 8a, 8b, 20, 23 ... Monochromatic light source 6a, 6b, 9a, 9b・ 13a ・ 13b ・ 22 ・ 28
.38 lens 7a, 7b, 21, 36, 42, 43, 47 ... half mirror 10a, 10b, 11a, 11b, 25, 26 ... light 12, 12a, 12b, 27, 41 ... spatial filter 14, 30, 40 Image processing device 15/31 Inspection result output device 29 Image storage unit 32 He-Cd laser 33/34 Mirror 35 Collimator 37 Inspected wafer 39/44/45/48 Camera 46 / Six-axis control Stage 47 ... Unit area

Claims (6)

[Claims] 1. A pattern defect detection method for detecting defects of a regular pattern formed on a plate-like object,
A pattern defect detection method characterized by using a result of applying a spatial filter to a Fourier transform image of the regular pattern and a result of comparing optical images of the same pattern at a plurality of locations on the regular pattern. 2. The result of comparing the optical images of the same pattern at a plurality of points on the regular pattern uses the difference in the light intensity of the optical images of the same pattern at a plurality of points on the regular pattern. The pattern defect detection method according to claim 1. 3. The result of comparing the optical images of the same pattern at a plurality of points on the regular pattern is the difference of the differential values of the light intensity distribution of the optical images of the same pattern at a plurality of points on the regular pattern. The pattern defect detection method according to claim 1, which is used. 4. A pattern defect detection method for detecting defects of a regular pattern formed on a plate-like object,
Optical images at odd and multiple points on the regular pattern are compared with each other, and the regular pattern whose comparison result is equal and occupies a large number is identified as the accurate regular pattern. A pattern defect detection method characterized by specifying the regular pattern. 5. A pattern defect detection device for detecting defects of a regular pattern formed on a plate-like object,
Transforming means for Fourier transforming the optical image of the regular pattern, spatial filter, and means for comparing optical images of the same pattern at a plurality of points on the regular pattern,
A pattern defect detecting apparatus, which detects a pattern defect using a result of applying a spatial filter to a Fourier transform image by the converting unit and a comparison result of the comparing unit. 6. A pattern defect detection device for detecting defects of a regular pattern formed on a plate-shaped object,
The image forming apparatus further comprises: comparing means for comparing optical images at odd-numbered locations and a plurality of locations on the regular pattern; and recognizing means for making the regular pattern, which has the same comparison result and a large number of comparison results, to be the accurate regular pattern. A pattern defect detection device characterized in that the regular pattern having a defect is specified from the result of the recognition by the recognition means.