Radiant energy – Inspection of solids or liquids by charged particles – Electron probe type
Utility Patent
1998-10-29
2001-01-02
Berman, Jack (Department: 2878)
Radiant energy
Inspection of solids or liquids by charged particles
Electron probe type
C250S307000, C356S394000, C382S145000, C382S149000
Utility Patent
active
06169282
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to visual inspection for detecting a defect of a pattern to be inspected and more particularly to a defect inspection method and apparatus therefor for a pattern to be inspected in a semiconductor wafer or liquid crystal display.
2. Description of the Prior Art
Conventionally, this kind of inspection apparatus, like the art described in Japanese Patent Laid-Open Application No. 55-74409, detects an image of a pattern to be inspected by an image sensor such as a line sensor by moving the pattern to be inspected, compares the aforementioned detected image signal with an image signal delayed by the predetermined time in the gray level, and recognizes a mismatch between them as a defect.
Furthermore, as described in Japanese Patent Publication 8-10463, the comparison is made by lining up two images.
The aforementioned conventional defect recognition method will be explained in detail with reference to
FIGS. 14
to
17
.
FIG. 14
is a schematic illustration of a memory mat and peripheral circuit in a memory chip of a pattern to be inspected by the prior art, and
FIG. 15
shows a brightness histogram in the memory mat and peripheral circuit in the memory chip shown in
FIG. 14
, and
FIG. 16
shows drawings indicating the cutline of the process of flattening the surface of multi-layered film on a semiconductor wafer by the CMP (chemical mechanical polishing) process.
As shown in
FIG. 14
, many memory chips
20
are arranged and formed on a semiconductor wafer. Each memory chip
20
is broadly divided into a memory mat
21
and a peripheral circuit
22
. The memory mat
21
is a collection of small repetitive patterns (cells) and the peripheral circuit
22
is basically a collection of random patterns. However, in many cases, as viewed minutely, it can be regarded as a collection of repetitive patterns having a plurality of different cell pitches.
FIG. 15
shows a brightness distribution in the memory mat
21
and the peripheral circuit
22
in the memory chip
21
shown in
FIG. 14
, that is, the frequency (histogram) for the brightness in the memory chip as a maximum of 1024 gray levels in a 10-bit constitution. However, the memory mat
21
has a high pattern density and is generally dark. On the other hand, the peripheral circuit
22
has a low pattern density and is generally bright.
With respect to the brightness distribution on the semiconductor wafer subjected to the CMP process as shown in
FIG. 16
, the circuit pattern in the memory mat
21
is different in brightness depending on the pattern film thickness as indicated by the histogram shown in FIG.
17
. In the drawing, the wiring layer is deposited and then flattened by the CMP process. In such a pattern, the film thickness varies locally and gray scale differences are easily generated. In such a pattern, the patterns shown in
FIGS. 15 and 17
are compared in the pattern brightness and when the threshold is set so as to prevent a maldetection of a difference in brightness, the defect detection sensitivity will be extremely lowered. Such a difference in brightness may be offset to a certain extent if illumination light with a wide wave length band is used. However, in a pattern subjected to the CMP process, it is limited because the brightness variation is great. Therefore, it is desirable to detect a minute defect from a pattern with different brightness.
Furthermore, conventionally, the sum of squares of differences in two images is calculated and the image dislocation is detected by fitting a parabolic sphere surface to it. However, the method does not guarantee a match between the two images to be compared and optimum matching is desired for comparison.
FIG. 20
shows experimental results when the sum of squares of differences in each pixel of two images (f(x,y) shown in
FIG. 6
described later) is obtained by shifting one of the images within a range of ±1 pixel in the x and y directions. The transverse axis indicates the x direction and the ordinate axis indicates the y direction. Each value shown in the drawing indicates the sum of squares of differences. In this case, the same image (f(x,y) shown in
FIG. 6
) is used. Namely, as a sum of squares of differences, &Sgr;(f(x,y)−f(x±1, y±1))
2
is calculated. As shown in
FIG. 20
, even in the same image, the sum of squares of differences is not symmetrical about (0,0) and has an asymmetry of about 0.6%. Since the same image is used, the sum of squares of differences is 0 at (0,0). Therefore, even if a parabola is fitted to this data and a position where the sum of squares of differences is minimized with a resolution less than the pixel dimension is obtained, the accurate dislocation, (0,0) in this case cannot be detected. Furthermore, in a wafer of the CMP process, the brightness is different. The effect of the difference in brightness is shown. In this case, an image and another image having a brightness 1.1 times of that of the image are used. The value of 1.1 times is normal or smaller as a variation of the brightness of a CMP wafer. Experimental results are shown in FIG.
21
. The sum of absolute values of differences is shown. The minimum position is (0,1). There are great errors not only with a resolution less than the pixel dimension but also on the pixel level. The sum of squares of differences has the same tendency. The drawing shows that the image dislocation cannot be accurately obtained from these data. Needless to say, the case of 1.05 times also has the same tendency. The method for fitting a parabolic sphere surface to the sum of squares of differences and obtaining the minimum position from it like this may be called a method having an extremely large error.
SUMMARY OF THE INVENTION
An object of the present invention is to eliminate the difficulties of the prior arts mentioned above and to provide a defect inspection method for a pattern to be inspected for making capable of inspecting by comparison of patterns different in brightness and always inspecting a defect in high sensitivity and high reliability.
Another object of the present invention is to provide a defect inspection method for a pattern to be inspected using a highly accurate image matching method.
Still another object of the present invention is to provide a highly sensitive defect inspection method even for a wafer pattern subjected to the CMP process.
To accomplish these objects, the defect inspection method and apparatus therefor of the present invention images a sample having a plurality of patterns formed so as to be identical, detects an image of the first pattern arranged on the sample and an image of the second pattern, converts the gray level of at least one of the detected image of the first pattern and the detected image of the second pattern, and detects a defect of at least the first or second pattern using the one image whose gray level is converted and the image of the other aforementioned first or second pattern different from the one image.
More concretely, to accomplish the aforementioned objects, in the defect inspection method and apparatus therefor for a pattern to be inspected of the present invention which images a substrate having a plurality of patterns formed so as to be identical, detects an image of the first pattern arranged on this substrate, stores the detected image of the first pattern, images the substrate, detects an image of the second pattern formed so as to be identical with the first pattern, corrects the dislocation between the stored image of the first pattern and the detected second image in pixel units, and detects a pattern defect using the stored image of the first pattern and the detected second image in which the dislocation is corrected in pixel units, before detecting a pattern defect, the gray level of at least one of the stored image of the first pattern and the detected image of the second pattern is converted.
The conversion of gray level is to convert the brightness of each of two image signals to be compared so as to
Maeda Shunji
Makihira Hiroshi
Nakayama Yasuhiko
Oka Kenji
Shibata Yukihiro
Antonelli Terry Stout & Kraus LLP
Berman Jack
Hitachi , Ltd.
LandOfFree
Defect inspection method and apparatus therefor does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Defect inspection method and apparatus therefor, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Defect inspection method and apparatus therefor will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2523623