JP2886278B2 - Circuit pattern image detection method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention automatically detects defects such as shape defects and foreign matter on a pattern to be inspected (circuit pattern) such as an LSI wafer, a printed circuit board, and a thin film transistor (TFT) substrate. TECHNICAL FIELD The present invention relates to a circuit pattern image detection method and apparatus suitable for dynamic detection.

[Conventional technology]

Integrated circuits such as LSIs tend to be highly integrated and miniaturized. In the formation of such a fine wiring pattern, detection of a defect is important in determining the quality of the formation. Since such a defect is no longer visually detectable, automation of the defect detection is urgently required, not at the stage where a large number of personnel are arranged and visually performed.

Conventional techniques related to the automation of defect detection include information on the surface or inside of a semiconductor element obtained from an optical microscope, an infrared imaging device, an electron microscope, an X-ray imaging device, or the like, by using an imaging tube, an imaging device, or the like. 2. Description of the Related Art There are known methods and devices for converting a signal into a signal, performing predetermined signal processing (image processing) on an obtained image, and detecting a defect. This includes the Semiconductor World
d), June 1984, pages 112 to 119, and those described in JP-A-59-192943.

An example of the above-mentioned prior art components will be outlined with reference to FIG. In FIG. 9, the circuit pattern on the wafer 1 illuminated by the lamp 2 is enlarged and detected by a detector 4 ′ such as a one-dimensional image sensor via the objective lens 3. Then, the density image of the circuit pattern is compared with the image of the immediately preceding chip 7 a (adjacent chip) stored in the image memory 5, and the defect determination circuit 6 performs a defect determination process. The grayscale image of the detected circuit pattern is simultaneously stored in the image memory 5.
Is stored as a stored image and used for a comparative inspection of the next chip 7b. The defect determination is performed, for example, by comparing the gray levels of the corresponding pixels of the two images and outputting a region where the gray level difference is equal to or more than an allowable value as a defect.

[Problems to be solved by the invention]

However, as the miniaturization of LSI progresses, submicron LSI
Now, with the above-mentioned conventional detection technology, it is becoming difficult to detect a minute defect in a short time from the viewpoint of accuracy of a detected image. In the future, further miniaturization and area increase will be advanced, and if ultra-fine defects such as 0.1 to 0.3 μm in a complicated and fine multilayer pattern are to be detected with high reliability and extremely short time, ,
It is expected that conventional technology alone cannot cope. This is because, in order to detect a minute defect, it is necessary to detect an image having a small pixel size.However, if the pixel size is reduced, the number of pixels increases and detection takes time, and the pixel size is reduced. Even so, there is naturally a limit.

An object of the present invention is to provide a method and an apparatus for detecting an image of a pattern to-be-inspected with high accuracy so as to withstand the above-described defect detection. Furthermore, by applying this image detection method, it has become possible to detect ultra-small defects on the pattern to be inspected in a short time and with high accuracy in response to the miniaturization, large area, and multi-layering of LSI. An object of the present invention is to provide a circuit pattern defect detection method and apparatus capable of detecting the above-described ultra-fine defect such as 0.1 to 0.3 μm.

[Means for solving the problem]

In order to achieve the above object, according to the present invention, in detecting an image of a circuit pattern to be inspected, a configuration is adopted in which a detection pixel is allowed to have an area overlapping with another detection pixel adjacent to the detection pixel. This is to detect a grayscale image of a pattern to be inspected. In addition, the present invention detects a grayscale image in which pixels are arranged such that each detection pixel having an overlapping region is arranged with its center at a node of a triangular lattice, for example, a node of a right-angled triangular lattice. It was done. Further, in the detected image, in addition to the detection pixel, the detection pixel is interpolated (interpolated) by a pixel additionally formed using the detection pixel.

According to the present invention, a defect of a circuit pattern is determined using an image obtained by the image detecting means.

Further, in the present invention, for example, a pattern or the like is represented by an image obtained by the image detecting means.

[Action]

According to the above configuration, since the detected pixel detects an image having an area overlapping with another detected pixel adjacent to the detected pixel, the detected grayscale image is the same as the size of the detected pixel or its density value. In addition to the above parameters, the image has another parameter called the interval (pitch) of the detection pixels. In the conventional image, the size of the detection pixel is one processing unit. However, in the image of the present invention, the size of the detection pixel and the interval between the detection pixels can be independently set at the time of image detection without restricting each other. . Therefore, if a circuit pattern is represented by the image of the present invention, it is possible to perform a process suitable for an extremely fine defect on the circuit pattern. That is, according to the image of the present invention, it is possible to perform high-accuracy image registration in a smaller unit, which is determined by the interval of the detection pixels, which is the parameter, and to compare the shades. It is possible to detect even minute defects. Further, since the above processing can be performed without increasing the number of pixels to be detected, a defect can be detected at high speed. Further, the pixels for interpolating the detection pixels also have a small smoothing effect and are of high quality, so that they can contribute to detection of minute defects.

〔Example〕

Hereinafter, the present invention will be described specifically with reference to examples. The embodiment described below is an example in which the present invention is applied to a defect inspection of an LSI wafer pattern. However, it is needless to say that the embodiment can also be applied to a pattern inspection such as a TFT.

FIG. 1 is a configuration diagram of an LSI wafer pattern defect detection apparatus used in the embodiment. The apparatus includes an XY stage 8 for scanning the wafer 1, a lamp 2 for illuminating the wafer 1, an objective lens 3 for enlarging the illuminated wafer pattern by, for example, 40 times, and a half having a transmittance of 50%. mirror
The output signals of the one-dimensional image sensors 4a and 4b and the one-dimensional image sensors 4a and 4b, which detect the pattern enlarged by the objective lens 3 via the CCD 11, for example, are 8 to 1 respectively.
A / D converter for digitizing to 0-bit digital signal
9a, 9b, a mixer 10 for mixing these digital signals to form a grayscale image signal, an image memory 5 for storing the image signal, and comparing the image signal before and after being input to the image memory 5, A defect determination circuit 6 for performing defect determination.

The image signal is transmitted to the XY stage 8 using the one-dimensional image sensor 4
By moving and scanning at a constant speed in a direction orthogonal to the internal scanning directions of a and 4b, a two-dimensional image is obtained. The one-dimensional image sensors 4a and 4b are driven at, for example, 10 MHz. In the present embodiment, an example using a one-dimensional image sensor will be described, but a two-dimensional image sensor such as a TV camera may be used. In this case, the XY stage 8 performs a step-and-repeat movement to detect a two-dimensional image. In addition, the illumination may be illumination other than bright-field illumination as long as it is an illumination method suitable for detecting a defect, and the detection is not limited to regular reflection light detection, but may be any detection method suitable for detecting a defect. Anything may be used, for example, scattered light may be detected.

The two one-dimensional image sensors 4a and 4b are arranged at the in-focus position of the objective lens 3 so that the image sensors do not defocus. The detection position is shifted by 1/2 pixel on the wafer pattern in the longitudinal direction and the lateral direction of the one-dimensional image sensor. Now, assuming that the length of the one-dimensional image sensor is 1024 pixels and the detection pixel size on the wafer surface is 0.24 μm, the detection field of 246 μm × 0.24 μm on the wafer surface is 0.12 μm as shown in FIG. The positional relationship is shifted in the X and Y directions.
In addition, in order to eliminate the influence of illuminance fluctuation of the lighting lamp, the two one-dimensional image sensors 4a and 4b use the same clock,
It is driven by the same start timing signal. Also, XY
In order to eliminate the image distortion due to the speed fluctuation of the stage 8,
The one-dimensional image sensors 4a and 4b are driven in synchronization with the output of a position detection scale (not shown) mounted on the stage. Although a focusing mechanism is required for stable image detection, illustration and description are omitted here. Further, in FIG. 1, the half mirror 11 is used to split the detection light into the two image sensors. However, the invention is not limited to the half mirror, but may be anything as long as it splits the light. For example, a prism type beam splitter is used. it can.

As described above, two one-dimensional image sensors 4a and 4 respectively.
FIG. 3 shows images detected by the b and A / D converters 9a and 9b and the mixer 10. In FIG. 3, an example of an image of the detected circuit pattern is shown in the center, and this image is an image A detected by the one-dimensional image sensor 4a and an image A detected by the one-dimensional image sensor 4b. An image B, which is an image, is synthesized. The upper part of FIG. 3 shows the relationship between the pixels constituting the image A and the pixels constituting the image B. As is clear from the figure, both are vertically and horizontally shifted by 1/2 pixel. . As described above, in the present embodiment, an image in which the image A and the image B are combined is detected. When the combined image is referred to as an image C, the features of the image C are images A and B. The center position of each pixel (the position of the node of the lattice) forms a square (square lattice), whereas the center position of each pixel of the image C (the position of the node of the lattice) forms a right triangle. (A right-angled triangular lattice). In order to more clearly show this state, FIG. 4 (a) schematically shows a conventional image expressed by a square lattice, and FIG. 4 (b) schematically shows an image expressed by a right-angled triangular lattice in this embodiment. Shown in However, in FIG. 4 (b), each pixel is
It should be noted that the pixels overlap by 1/2 pixel.

In the mixer 10, as shown in FIG. 5, offsets and gains of an output amplifier (not shown) of the one-dimensional image sensor are stored in a look-up table (RAM or the like) to store correction data measured in advance. It converts the input data and outputs it.) Compensation by 10-1a, 10-1b and correction of the mismatch due to the difference in sensitivity between the two one-dimensional image sensors, etc.
An extremely accurate grayscale image signal can be obtained. The obtained grayscale image signals are alternately output as 8-bit image signals or simultaneously output as 8-bit × 2 channel image signals by the switcher 10-2, for example. Then, the output is transferred to the image memory 5 at the same time,
The image is input to the defect determination circuit 6 and compared with the image of the immediately preceding adjacent chip stored in the image memory 5, and a difference in shading or a difference in shape is detected as a defect.

Here, the defect detection performance of the present embodiment will be described.
Where the inventors of the present application have searched, the defect detection performance is not uniquely determined by the pixel size when detecting an image, apart from the defect determination algorithm,
It is considered that it is determined by the pixel pitch at the time of processing an image (interval between grids when the grid is considered as described above). This is the IEICE Transactions shown in Figure 6.
D-II, Vol. J72-D-II, No. 12 (1989), pp. 2041-2050, a comparison of the defect detection ability
Image detected with a pixel size of μm (shown by a black circle in the figure)
And the newly formed image (indicated by white circles in the figure) after detecting with a pixel size of 0.24 μm and interpolating a new pixel with a pixel pitch of 0.12 μm (in the known example,
0.3 μm defect detection performance). That is,
For an image with a pixel size of 0.24 μm, 0.12
Although a pixel having a pitch of μm is obtained, the interpolated pixel is a pixel having an average and smoothed density value whose density value is formed from surrounding pixels, and the pixel size itself is substantially the same as the original pixel size. It can be said that it is twice as large or three times as large (in terms of area, four times or nine times). Therefore, although the quality of the interpolated pixel is low, the defect detection performance of an image having a pixel size of 0.24 μm does not deteriorate despite the comparison with such an averaged density value is 0.12 μm. It is nothing but processing the image at the pitch. In other words, if the interval between the nearest neighboring pixels (interval between the grids) is defined as “distance”, it can be understood that both of the above images have the same “distance” of 0.12 μm, and this “distance” indicates the defect. That is, the detection performance is determined. However, in the method of interpolating pixels and processing an image at this pitch, the amount of information at the time of image detection cannot be further increased by interpolation.
It is expected that it is impossible in principle to detect an ultra-fine defect such as an m defect. On the other hand, in general, if the pixel size at the time of detection is reduced, the possibility of detecting even a minute defect increases, but at the same time, the detection time increases in inverse proportion to the square of the pixel size. Therefore, the pixel size is restricted by the inspection time and the like and cannot be reduced unnecessarily.

Therefore, in this embodiment, as shown in FIGS. 3 and 4 (b), the image A having the conventional pixel size is used in the x and y directions by 1 using the method described with reference to FIG. / 2
A composite image C on which an image B shifted by pixels is superimposed at the same time is detected. In this case, assuming that the pixel size of each of the images A and B before synthesis is α (for example, α = 0.24 μm), the interval between adjacent pixels in each image before synthesis is α in the x and y directions. In the diagonal direction Becomes On the other hand, the interval between adjacent pixels in the synthesized image C is α in the X and Y directions, but is α in the oblique direction. And the above “distance” representing the interval between the nearest neighboring pixels.
In the concept that the former is α, the latter is And in the latter This means that the lattice spacing has become smaller. Therefore, if a defect is determined using the synthesized image C, a finer defect, that is, almost It is expected that a defect having a size of up to 10 mm can be detected. In other words, the image in which the line connecting the center position (grid position) of each pixel forms a right triangle is more defective than the conventional image in which the center of each pixel is located at a node of a square grid. It is more advantageous.

From the above relationship, the pixel size of each of the images A and B is , The “distance” in the synthesized image C
Is the same as the “distance” of the original images A and B, and is 0.24 μm.
Therefore, if the synthesized image C is used, the defect detection performance is almost the same as that of the original image, and in this case, the detection time is only 1/2 times the original, and the inspection speed is dramatically increased. Improvement is possible.

Expressing the pattern to be inspected with an image in which the grid positions of the pixels form a right-angled triangle has more dense information than a conventional image in which the pixels are arranged at the nodes of a square grid. , And also for image compression. In addition, expressing an image with a right-angled triangular grid in this way is based on a conventional square grid or hexagonal grid (see "Introduction to Computer Image Processing", Tamura, Soken Publishing,
Has higher expressive ability than that of textbooks.

In the above description of the embodiment, the detection positions of the two one-dimensional image sensors 4a and 4b
It is assumed that only one pixel is shifted in the longitudinal direction and the short direction of the one-dimensional image sensor. However, only one pixel is shifted only in the longitudinal direction, and a half cycle (here, about 50 μs, that is, half of one scanning time 100 μs) is shifted. Even if each one-dimensional image sensor is driven by different start timing signals,
The same effect as the above embodiment can be obtained. In addition, the detection by the two one-dimensional image sensors is shifted by half a pixel in the longitudinal direction on the wafer pattern, and is shifted at a position on the order of several tens of μm in the short direction. To output the output of one image sensor to the distance between the image sensors (several tens
Even if m) is delayed by the time during which the stage moves, substantially the same effects as in the above embodiment can be obtained. In these cases, the two one-dimensional image sensors can be integrated.

Also, in the above embodiment, an image was detected at a position shifted by 1/2 pixel using two one-dimensional image sensors.
An arbitrary amount of shift (shift of 1 / N pixel) is set by using a plurality of N one-dimensional image sensors, such as detecting an image at a position shifted by 1/3 pixel using three one-dimensional image sensors. be able to. However, when three or more one-dimensional image sensors are used, the lattice position is not always a right triangle as in the above embodiment.

Next, a specific example of defect determination using a composite image in the above embodiment will be described. FIG. 7 shows a configuration example of the defect determination circuit 6. The defect determination circuit 6 includes a pixel interpolation circuit 12 for pixel-interpolating one of the input composite images (in the figure, a composite image from the mixer 10), an alignment circuit 13 for aligning the image,
Mismatch detection circuit for detecting misalignment of aligned images
Consists of fourteen. As shown in FIG. 8 (a), the pixel interpolation circuit 12 determines the center position (grid position) of each pixel forming a right triangle by the input composite image as shown in FIG. 8 (b).
As shown in (1), a new pixel is interpolated between each pixel of the composite image so that it forms a square. In FIG. 8, an image is represented by arranging pixels at positions of nodes of the lattice. For example, pixel interpolation is performed. It is given by one of the following equations.

e = (a + c) / 2 (1) e = (b + d) / 2 (2) e = (a + b + c + d) / 4 (3) where a to e are shown in FIG. As an example, e is the density value of the interpolated pixel, a and c are the density values of the detection pixels by the one-dimensional image sensor 4a, and b and d are the density values of the detection pixels by the one-dimensional image sensor 4b. . Note that interpolation need not necessarily be linear interpolation as in the above equation,
Any method such as a conventionally proposed nonlinear method may be applied. By this interpolation, a high-precision gray-scale image of a lattice position having an interval of α / 2 (0.12 μm) is obtained. In the above interpolation, a pixel is newly formed at the center (half position) of two pixels. However, a new pixel may be formed at an arbitrary position such as the 1/3 and 2/3 positions. it can.

In order to further improve the defect detection performance in the defect determination circuit 6 shown in FIG.
The method proposed in 3587 can be used. When the combined image C of A and B is used as in the above-described embodiment, high-accuracy alignment with an interval of α / 2 (0.12 μm) can be performed. This alignment may use only the original image A or the image B, or may use an image before pixel interpolation. In the mismatch detection, the aligned and interpolated images are compared with each other between chips, and differences in shading and shapes are output as mismatches. At that time, for the pixel of interest of one image and the corresponding pixel of the other image corresponding to the pixel of interest, the respective neighboring pixels are located closer to the neighboring pixels of the conventional image, and therefore, also be compared with the neighboring pixels. As a result, compared with the conventional method, it is possible to perform detailed analysis in a finer unit, and it is possible to perform mismatch detection with high accuracy, thereby detecting an ultrafine defect.

In the above embodiment, the detection pixel size is described as 0.24 μm. However, this size may be an arbitrary value, and is determined in consideration of an inspection target, an inspection time, and the like.

In the above embodiments, the image detection method and the defect determination using the detected image have been described. However, according to the above configuration, the detection pixel interval corresponding to the arrangement interval of the two one-dimensional image sensors is smaller. It is possible to perform high-accuracy image registration in units, comparison of shading, and the like, and therefore, it is possible to detect even finer defects.

Further, in the above-described embodiment, an example in which the means for obtaining an image of an object to be inspected is an optical microscope has been described. However, the concept of the image described above is widely applicable to a general public, and is extremely versatile. . That is, information on the surface or inside of a semiconductor element or the like obtained from an electron microscope or an X-ray imaging device is converted into an electric signal by an image pickup tube or an image display, and the obtained image is subjected to predetermined signal processing ( Image processing). Further, it can also be used for purposes other than inspection, for example, pattern recognition, position shift detection using the pattern, or positioning.

〔The invention's effect〕

According to the present invention, in detecting an image of a circuit pattern,
An image having a region where the detection pixel overlaps with the adjacent detection pixel is detected. This image has a size of the detection pixel,
The image has another parameter called the interval (pitch) of the detection pixels in addition to the parameter called the density value, and is suitable for defect detection. In other words, in this image, the size of the detection pixel and the interval between the detection pixels can be set independently at the time of image detection without being constrained to each other, so that the height in a smaller unit determined by the latter parameter, the detection pixel interval, is used. Accurate image registration and comparison of light and shade can be performed. Therefore, it is possible to detect minute defects of a pattern to be inspected with high accuracy and in a short time, sufficiently responding to miniaturization, large area, and multilayer of LSI, and as a result, 0.1 to 0.3 μm
Such a minute defect can be detected. Also, the detected image has a new concept of an image, unlike a conventional image, and can be widely applied to purposes other than inspection.

[Brief description of the drawings]

FIG. 1 is a block diagram of an apparatus used in an embodiment of a circuit pattern image detection method and apparatus according to the present invention, FIG. 2 is an explanatory view of a detection field in the embodiment, and FIG. FIGS. 4 (a) and 4 (b) are explanatory diagrams showing the configuration of the detected image of FIG.
Is an explanatory diagram of an image formed by a conventional square grid and an image formed by a right-angled triangular grid in this embodiment. FIG. 5 is a configuration diagram of a mixer that combines outputs of two image sensors in the embodiment. Is a comparison diagram of the defect detection performance at different detection pixel dimensions (known example), FIG. 7 is a configuration diagram of a defect determination circuit in the embodiment, and FIG. 8 shows additional pixel formation by pixel interpolation in the embodiment. FIG. 9 is an explanatory diagram of a conventional circuit pattern defect determination. Description of reference numerals 1 ... wafer, 2 ... lamp 3 ... objective lens 4a, 4b ... one-dimensional image sensor 5 ... image memory, 6 ... defect determination circuit 8 ... XY stage 9a, 9b ... A / D converter 10 Mixer 10-1a, 10-1b Lookup table 10-2 Switcher 11, Half mirror 12 Pixel interpolator 13, Positioning circuit 14 Mismatch detector

──────────────────────────────────────────────────の Continuation of the front page (51) Int.Cl. ⁶ Identification code FI G06T 7/00 G06F 15/62 405A (72) Inventor Takashi Hiroi 292 Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture Hitachi, Ltd. Production Technology (56) References JP-A-61-120572 (JP, A) JP-A-63-208964 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) H01L 21/66 G06T 9/00 G01N 21/188

Claims (20)

(57) [Claims] In a method for detecting an image of a circuit pattern to be inspected by detecting pixels of the circuit pattern to be inspected, the detection pixel has an area overlapping with another detection pixel adjacent to the detection pixel. A method for detecting an image of a circuit pattern. 2. The circuit pattern image detection method according to claim 1, wherein the detection pixels are arranged so as to have their centers at the positions of the nodes of the triangular lattice. Method. 3. The circuit pattern image detecting method according to claim 2, wherein the triangles forming the triangular grid are right triangles. 4. The circuit pattern image detecting method according to claim 1, wherein a pixel for interpolating the detected pixel is additionally formed by using the detected pixel. Image detection method. 5. The method for detecting an image of a circuit pattern according to claim 1, wherein a detection pixel having an area overlapping with another adjacent detection pixel is obtained. A circuit pattern image detection method, comprising using two or more types of detectors that are displaced in units of pixels or less on the same imaging plane. 6. The circuit pattern image detecting method according to claim 1, wherein a detection pixel having an area overlapping with another adjacent detection pixel is obtained. A circuit pattern image detection method comprising using two or more types of detectors arranged on one and the same image plane in units of one pixel or less and displaced in one direction, and driving the detectors with different timing signals. Method. 7. For each circuit pattern to be inspected, which is originally formed to be the same, an image in which a detection pixel includes a pixel having an area overlapping with another detection pixel adjacent to the detection pixel is detected. And detecting a defect in the circuit pattern by comparing corresponding portions. 8. The circuit pattern defect detection method according to claim 7, wherein the image to be detected is an image composed of pixels arranged such that a detected pixel has a center at a position of a node of a triangular lattice. A method for detecting a defect in a circuit pattern, comprising: 9. The method according to claim 8, wherein the triangles forming the triangular lattice are right triangles. 10. An image expression method characterized by expressing an image by a group of pixels which are arranged so as to have a center at a node position of a triangular lattice and have an area overlapping with another adjacent pixel. 11. An apparatus for detecting an image of a circuit pattern to be inspected by detecting pixels of the circuit pattern to be inspected, wherein the detected pixel has an area overlapping with another detected pixel adjacent to the detected pixel. And a pixel detecting means for detecting a pixel. 12. The circuit pattern image detecting device according to claim 11, wherein the pixel detecting means detects the detected pixel such that the detected pixel is arranged at a node position of the triangular lattice with its center. A circuit pattern image detecting device, characterized in that: 13. The circuit pattern image detecting apparatus according to claim 12, wherein the triangles forming the triangular grid are right triangles. 14. The circuit pattern image detecting apparatus according to claim 11, further comprising means for using a detected pixel to additionally form a pixel for interpolating the detected pixel. Circuit pattern image detection device. 15. The circuit pattern image detecting apparatus according to claim 11, wherein the pixel detecting means shifts the position in units of pixels or less on the same image forming plane on the optical axis. An image detecting apparatus for a circuit pattern, comprising two or more detectors arranged. 16. The circuit pattern image detecting apparatus according to claim 11, wherein the pixel detecting means is located on the same image plane on the optical axis in one direction in units of pixels or less. A circuit pattern image detecting apparatus, comprising: two or more types of detectors which are displaced from each other; and means for driving the detectors with different timing signals. 17. For each of the circuit patterns to be inspected formed so as to be originally identical, an image in which a detection pixel is formed of a pixel having an area overlapping with another detection pixel adjacent to the detection pixel is detected. A circuit pattern defect detecting device for detecting a defect in a circuit pattern by comparing a corresponding portion of the image with a corresponding portion of the image. 18. The circuit pattern defect detecting device according to claim 17, wherein the image detecting means performs the detection such that the detected pixels are arranged at the positions of the nodes of the triangular lattice with the centers thereof. A defect detecting apparatus for a circuit pattern, comprising: 19. The defect detecting apparatus for a circuit pattern according to claim 18, wherein the triangles forming the triangular lattice are right-angled triangles. 20. A detection device for realizing the circuit pattern image detection device according to claim 15 or 16, wherein a plurality of detectors are integrated.