JP2000183122A - Method and device for inspecting semiconductor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To inspect defects of a semiconductor by precisely aligning a reference image with an image to be inspected for comparison in a method and device for inspecting a semiconductor. SOLUTION: A reference image 2A of a semiconductor and an image 2B to be inspected of the semiconductor are taken in, a plurality of corresponding regions 7A and 7B are set in the reference image 2A and the inspected image 2B, the similarity between corresponding regions 7A and 7B of the reference image 2A and the inspected image 2B is determined, an alignment region is set according to similarity for each region obtained, and the reference image 2A is aligned with the inspected image 2B in the alignment region for comparison/inspection between the reference image 2A and inspected image 2B.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] The present invention relates to a semiconductor inspection apparatus.

[0002]

2. Description of the Related Art In recent years, as semiconductors have become larger and more highly integrated, semiconductor wiring patterns and the like have been miniaturized.
In addition, since a defect in a pattern or adhesion of a foreign substance on the pattern causes a connection failure, it is necessary to perform a strict defect inspection. When performing inspections, conventional visual inspections by workers have variations in the inspection standards of individual inspectors, and as the pattern complexity increases due to miniaturization, the inspection time increases and the inspectors become fatigued. This makes it difficult to perform inspections. Under such circumstances, attempts have been made to automate the visual inspection from the visual inspection of the inspector.

A semiconductor is formed on a wafer and is diced into chips. The semiconductor provided on each chip includes a memory, a logic part, a wiring part, and a pad part for bonding to an electrode of another component. If a foreign substance is attached to a semiconductor or a defect is found in a wiring portion or the like, a connection failure will be caused later, so that a defect inspection needs to be performed. Various methods have been used for this type of inspection, and one of them is a comparative inspection method using image information of a target semiconductor.

[0004]

In the comparative inspection, a difference image is formed by using images of a plurality of semiconductors of the same kind, and if there is a difference in the difference image, a defect is found in one of the two target semiconductors. It is said that there is. Conventionally, inspections have been performed by visual inspection by inspectors. However, inspections by inspectors have had problems such as variations in inspection standards of individual inspectors and fatigue of the inspectors.

Further, when comparing images of a plurality of semiconductors of the same type, if the corresponding portions of the two images are not precisely aligned so as to be at the same position, the result of the inspection may include a large error. And there was a problem that a correct judgment could not be made. An object of the present invention is to provide a semiconductor inspection method and apparatus capable of inspecting semiconductor defects by precisely aligning a reference image and an image to be inspected and comparing the images.

[0006]

According to a semiconductor inspection method of the present invention, a reference image of a semiconductor and an image to be inspected of a semiconductor are fetched, and a plurality of corresponding areas are defined in the reference image and the image to be inspected. Calculating the similarity between the corresponding areas of the reference image and the image to be inspected, determining the alignment area according to the obtained similarity of each area, and determining the reference image and the image to be inspected in the alignment area. And a comparative inspection of the reference image and the image to be inspected is performed.

In this method, a similarity is calculated for each region where a semiconductor defect is set on a reference image and an image to be inspected, and a region having a large similarity is determined as a positioning region. The reference image and the inspection image are aligned, and a comparison inspection of the reference image and the inspection image is performed. In this way, since the alignment between the reference image and the image to be inspected can be performed with high accuracy, highly accurate defect detection can be performed.

[0008] A semiconductor inspection apparatus capable of implementing this method is as follows.
Imaging means for capturing an image of a semiconductor, storage means for storing a reference image of the semiconductor and an image to be inspected of the semiconductor, area setting means for defining a plurality of areas respectively corresponding to the reference image and the image to be inspected, A positioning area setting means for determining a similarity between the corresponding areas of the reference image and the image to be inspected and determining a positioning area according to the obtained degree of similarity for each area; The image processing apparatus is characterized in that it comprises a positioning means for positioning the image and the image to be inspected, and an inspection means for comparing and inspecting the reference image and the image to be inspected.

[0009] These methods and apparatus may include the following features. The amount of shift between the reference image and the image to be inspected in the alignment region is calculated. When there are a plurality of alignment regions, the deviation amounts obtained in the respective regions are averaged to obtain the entire deviation amount. The plurality of alignment regions are set at positions separated from each other.
Preferably, the positioning area is set in a peripheral portion of each of the reference image and the inspection image.

When the similarity of the same area is lower than the specified value for a certain number of consecutive times, or when the average value of the predetermined number of times is lower than the specified value, the similarity calculation of the area is performed at a certain number of times. Stop and speed up the alignment time. Further, when the similarity of the same region is better than a specified value for a certain number of consecutive times, or when the average value of a certain number of times is better than a specified value, the region is defined as a registration region for a certain number of times, and The calculation of the similarity other than the area is stopped, and the positioning time is shortened.

The optical magnification at the time of imaging the reference image is made larger than the optical magnification at the time of imaging the image to be inspected, to obtain an image with higher definition than the image to be inspected, and the reference image and the image to be inspected are compared. Sometimes, by reconstructing to the same size as the image to be inspected and calculating the similarity and the amount of deviation, the reference image and the image to be inspected have the same size to reduce the effect of blur due to shift of pixels or less, Achieve higher accuracy in calculating the amount of deviation.

When the reference image is reconstructed to the same size as the image to be inspected, the reference image is thinned and reconstructed, thereby shortening the positioning time. Further, when the reference image is reconstructed to the same size as the image to be inspected, by averaging and reconstructing the vicinity of the corresponding pixel of the reference image, the positioning time is shortened. N of the lens for capturing the reference image
A is made equal to the NA for the image to be inspected. In addition, the reference image is captured by adjusting the amount of illumination so that the brightness of the image obtained by capturing the reference image at high magnification is equal to the brightness of the image to be inspected. Also, to further improve the alignment accuracy,
After capturing the reference image at a high magnification, the brightness level of the reference image is corrected so as to match the brightness level of the image to be inspected. Then, the normalized correlation is used for the similarity calculation, so that the difference in the brightness level between the reference image and the image to be inspected can be ignored.

[0013]

FIG. 1 is a view showing a silicon wafer 1 on which a semiconductor integrated circuit is formed. The silicon wafer 1 has a plurality of unit semiconductors 2. FIG. 2 is a diagram showing a circled portion of the silicon wafer 1 of FIG. The silicon wafer 1 of FIG. 1 is diced and separated into respective semiconductors 2 as silicon chips. The present invention can be applied to the semiconductor 2 in the state of a silicon wafer 1 or can be applied to the semiconductor 2 in the state of a silicon chip after being diced.

Each semiconductor 2 includes, for example, a memory or logic section 3, a wiring section 4, and a pad section 5 for bonding to an electrode of another component (not shown). If a foreign substance is attached to the semiconductor 2 or a defect is present in the wiring portion 4 or the like, a connection failure may occur later, so that a defect inspection of the semiconductor 2 needs to be performed. As shown in FIG. 3, the defect inspection of the semiconductor 2 acquires the reference image 2A of the semiconductor 2 and the inspection image 2B of the semiconductor 2, and obtains the reference image 2A and the inspection image 2B.
This is done by comparing

FIG. 4 is a diagram showing a semiconductor inspection apparatus 10 according to an embodiment of the present invention. The semiconductor inspection apparatus 10 is configured to use a semiconductor 2 in a state of a silicon wafer 1 or a state of a silicon chip.
A movable stage 12 for supporting the CCD, a CCD camera 14 which is an imaging device for acquiring an image of the semiconductor 2 supported on the stage 12, a light source 16, and a half mirror 18. The illumination light of the light source 16 is the half mirror 1
The image light reflected by 8 and directed to the semiconductor 2, and the image light reflected by the semiconductor 2 is transmitted through the half mirror 18 and directed to the camera 14 to be imaged.

The semiconductor inspection device 10 further includes a video grabber 20, a first memory 22, a second memory 24, and a control device 26. The first memory 22 stores the inspection image 2B of the semiconductor 2 captured by the camera 14, and the second memory 24 stores the semiconductor reference image 2A captured by the camera 14. The reference image 2A of the semiconductor 2 is
For example, an image to be inspected of the semiconductor 2 used before in a series of inspections may be used, or an image obtained by imaging a good semiconductor 2 in advance may be used.

The control device 26 includes at least a function of positioning the reference image 2A of the semiconductor 2 and the image 2B to be inspected, and a function of comparing and inspecting the aligned reference image 2A and the image 2B to be inspected. . That is, the control device 26 obtains the similarity between the corresponding areas of the reference image 2A and the image to be inspected 2A and the area setting means for determining a plurality of areas corresponding to the respective areas in the reference image 2A and the image to be inspected 2B. Positioning area setting means for determining a positioning area in accordance with the similarity of each set area, positioning means for positioning the reference image 2A and the inspection image 2B in the positioning area, and a reference image 2A. Inspection means for performing an inspection for comparison with the image to be inspected 2B.

In FIG. 3, the reference image 2A and the image to be inspected 2B are images of the same kind of semiconductor 2, and
Basically, the same memory or logic unit 3 and wiring unit 4
And an image of the pad unit 5. However, each semiconductor 2 may have a different defect. In FIG. 3, for example, the reference image 2A has a defect 6A, and the inspection image 2B has a defect 6B.

The position of the reference image 2A and the image to be inspected 2B are aligned, and if there is a difference between the two, it is determined that there is a defect. However, the reference image 2A and the inspection image 2
If a misalignment occurs between the two images when the position B is aligned, noise enters during the comparison, making it impossible to make an accurate defect determination. Therefore, precise alignment is required.

FIG. 5 is a diagram for explaining an example in which misalignment occurs between the reference image 2A and the image to be inspected 2B. The reference image 2A and the inspected image 2B have a characteristic portion 2p, and usually, the characteristic portion 2p
Is selected as the alignment criterion. But, for example,
The inspection image 2B has a defect 6B near the characteristic portion 2p.
And the reference image 2A based on the characteristic portion 2p of the reference image 2A and the defect 6B of the inspected image 2B.
And the image to be inspected 2B may be aligned. In such a case, misalignment occurs between the reference image 2A and the image to be inspected 2B.

Therefore, in the present invention, alignment is not performed with reference to a portion where there is a possibility of the defects 6A and 6B. For this purpose, as shown in FIG. 3, a plurality of minute regions 7A and 7B respectively corresponding to the reference image 2A and the inspection image 2B are determined. The area 7A of the reference image 2A is divided into the same number as the area 7B of the inspection image 2B and exists at the same position.

Then, the images are compared for each of the corresponding areas 7A and 7B, and the similarity is calculated. The similarity calculation may be a sum of absolute values of differences between images, a sum of squares of differences, or a normalized correlation value. In the case of the sum of the absolute values of the differences or the sum of the squares of the differences, the value is 0 when the similarity is maximum, and the value increases as the similarity worsens. In the case of the normalized correlation value, the value of the similarity becomes 1 when the similarity is maximum, and the value approaches 0 as the similarity becomes worse.

FIG. 6 shows an example of the similarity calculated for each of the corresponding areas 7A and 7B. In the embodiment, the similarity uses a normalized correlation value. Among these areas 7A and 7B, the similarity of area 7C is large. Therefore, the area 7 having a large similarity
C is defined as a positioning area. In the alignment area 7C having a high degree of similarity, it is unlikely that one of the reference image 2A and the inspection image 2B contains a defect. Therefore, if the alignment is performed with reference to a specific point (for example, the characteristic portion 2p in FIG. 5) in the alignment area 7C, and the comparison inspection of the reference image 2A and the inspection image 2B is performed, the position between the two images is obtained. There is no misalignment, no noise enters during the comparative inspection, and accurate defect determination can be performed. The comparison inspection of the reference image 2A and the inspection image 2B can be performed according to a known method.

Furthermore, the corresponding regions 7A and 7B are provided at corresponding positions, but need not necessarily have the same area. FIG. 7 is a diagram showing an example in which one area 7A of the reference image 2A and the inspection image 2B includes the other area 7B. Thus, the included area 7B
Performs the similarity calculation at each position while shifting over the other area 7A as indicated by the arrow. FIG. 6 shows the maximum similarity among the similarities calculated at each position. The positioning area 7C is an area where the maximum similarity is large.

Then, the shift amount between the two images can be calculated from the maximum similarity value in the corresponding areas 7A and 7B according to a calculation formula experimentally determined. Therefore, precise alignment is performed with reference to a specific point in the alignment area 7C having a high degree of similarity, and by correcting the positional shift amount of the specific point, and comparing the reference image 2A and the image to be inspected 2B. Inspection can be performed, and accurate defect determination can be performed. When there is one alignment area 7C, the shift amount is determined by the shift amount of the area 7C. Multiple alignment areas 7C
Is selected, the position shift amount is obtained by averaging the shift amounts obtained in the respective areas to obtain the total shift amount.

The calculation of the similarity need not be performed in all the areas 7A and 7B as shown in FIG. The calculation of the similarity may be performed in some predetermined areas 7A and 7B, or the calculation of the similarity may be performed in accordance with the mode selected in the calculation process. 8 and 9 show an area 7A to be an alignment area 7C,
7B in advance, the peripheral portion 7 of the reference image 2A and the inspection image 2B.
Set to D. For example, when the reference image 2A and the image to be inspected 2B are rectangular images, the regions 7A and 7B to be the alignment regions 7C are limited to three or four corners of the rectangle. Alternatively, the alignment area 7C
The areas 7A and 7B that should become
Limited to the area.

The peripheral portion 7D of the reference image 2A and the inspected image 2B includes a plurality of areas 7A and 7B.
In A and 7B, the similarity is calculated in the same manner as described above, the positioning area 7C is determined, and the shift amount is calculated. Since the regions 7A and 7B between the central portion and the peripheral portion 7D of the reference image 2A and the image to be inspected 2B are not initially set as the positioning regions 7C, the calculation of similarity is not performed. Accordingly, it is possible to speed up the processing of calculating the similarity.

The reference image 2A and the image to be inspected 2B are obtained by using the shift amount of the alignment area 7C selected in this manner.
By performing the position alignment, it is possible to reduce the influence of an error generated in the comparative inspection of the images including the defects 6A and 6B, and to perform the defect determination. FIG. 10 is a diagram illustrating an example of similarity calculation in a plurality of tests. In the process of performing the inspection a plurality of times, regions 7A and 7B having a similar degree of similarity may appear continuously. For example, FIG. 10 shows an example in which similarity calculation is performed in three tests in the order of upper left, center, and lower right. Areas with successively poor similarities are indicated by x points in the lower right figure, and in these areas the calculation of similarity is stopped in subsequent inspections. That is, these areas are excluded from the positioning area in the subsequent inspection.

As described above, the calculation of the similarity of the regions 7A and 7B having the poor similarity is continuously stopped in the plurality of inspections, or the similarity of the regions 7A and 7B other than the regions 7A and 7B having the good similarity is continuously calculated. By stopping the calculation, the positioning time can be shortened. Also, the similarity of the same area is
If a certain number of consecutive times or the average value of a certain number of times is better than the specified value, the area is defined as a registration area for a certain number of times, calculation of similarity other than the above area is stopped, and the positioning time is shortened. Can also be planned.

FIG. 11 shows an example in which the calculation of the degree of similarity at the time of positioning is performed with an accuracy of a pixel or less. In FIG. 11, regions 7A and 7B are divided into a plurality of pixels 2x and 2y, and each pixel 2x and 2y is divided into sub-pixels 2'x and 2'y. The calculation of the similarity is performed based on the sub-pixels 2'x, 2'y. By doing so, the alignment accuracy can be further improved.

However, when each pixel of the image is divided with an accuracy of less than a pixel as shown in the figure, it is necessary to divide not only the reference image 2A but also the image 2B to be inspected with an accuracy of less than a pixel for each inspection. Therefore, the calculation takes time. It may be preferable to perform imaging with a higher resolution of the camera 14 than to perform sub-pixel conversion by image processing in order to reduce calculation time.

FIG. 12 is a diagram showing a method of imaging the reference image 2A with a higher resolution. When only one of the images is shifted with an accuracy of less than the pixel, the time can be reduced, but there is a problem that blur occurs after the shift. Therefore, as a method of calculating the similarity with an accuracy equal to or less than the pixel, for example, the reference image 2A is imaged at a higher resolution than the image to be inspected 2B. In FIG. 12, the upper reference image 2A is an image captured at a high resolution having the sub-pixel 2′x, and when this is used for calculating the similarity,
The area 7A is reconstructed to have the pixel 2x as shown in the lower left diagram. In FIG. 12, one pixel 2x is composed of four sub-pixels 2'x. The image 2B to be inspected is picked up by the pixel 2y at the same level as the reconstructed pixel 2x, and the similarity is calculated by this. As a result, it is possible to reduce the effect of blurring of the image due to subpixel conversion,
In addition, since it is sufficient to perform the operation of subpixel conversion only for the reference image, it is possible to reduce the similarity calculation time.

As described above, the optical magnification at the time of imaging the reference image is made larger than the optical magnification at the time of imaging the image to be inspected, thereby obtaining an image with higher definition than the image to be inspected. When comparing images, reconstruct the image to the same size as the image to be inspected and calculate the similarity and the amount of shift, thereby reducing the effect of blurring due to shifts of pixels or less when the reference image and the image to be inspected have the same size. Can be reduced, and the accuracy of calculating the amount of deviation can be improved. Then, when the reference image is reconstructed to the same size as the image to be inspected, the reference image is thinned and reconstructed, whereby the positioning time can be shortened. Further, when reconstructing the reference image to the same size as the image to be inspected, by averaging and reconstructing the vicinity of the corresponding pixel of the reference image, the positioning time can be shortened.

When obtaining such a high-resolution reference image, the imaging optical system for obtaining the reference image 2A is different from the imaging optical system for obtaining the inspection image 2B. A device is needed to make it smaller. FIG. 13A shows a high-resolution imaging optical system for the reference image 2A, and FIG. 13B shows an imaging optical system for the inspection image 2B.
16 is an illumination light source, 18 is a half mirror, 14 is a camera,
30 is an imaging lens, 32 is a high-resolution objective lens for the reference image 2A, 34 is a circular slit, and 36 is the image 2 to be inspected.
This is an objective lens for B.

Here, in order to make the appearance of the reference image 2A and the image to be inspected 2B equal, the NA of the objective lens 36 for the image 2B to be inspected is changed by using the slit 34 so that the high-resolution objective lens for the reference image 2A is used. Adjust to be equal to 32 NA. Thereby, the appearance of both images becomes equal.
In addition, the average of the brightness of the image captured by the imaging optical system for obtaining the image 2B to be inspected (B) is stored,
The output of the illumination light source 16 of (A) may be adjusted so that the brightness of the image captured by the high-resolution imaging optical system for the reference image 2A of (A) is equivalent to that of the image. As described above, by performing the positioning, the positioning can be performed more precisely and at a higher speed, and the accuracy and speed of the inspection can be improved.

As described above, the reference image can be captured by adjusting the amount of illumination so that the brightness of the image obtained by capturing the reference image at a high magnification is equal to the brightness of the image to be inspected. Further, in order to further improve the positioning accuracy, after the reference image is captured at a high magnification, the brightness level of the reference image is corrected so as to match the brightness level of the image to be inspected. If necessary, a normalized correlation is used for similarity calculation, so that a difference in brightness level between the reference image and the image to be inspected can be ignored.

[0037]

As described above, according to the present invention,
In an inspection device that detects a defect in a semiconductor, a similarity is calculated for each region set on the reference image and the image to be inspected, alignment is performed using a region having a large similarity, and the defect is compared and detected. Highly accurate defect detection can be performed, which greatly contributes to automation and high efficiency of semiconductor inspection. In addition, it is preferable that high-accuracy defect detection be performed by calculating a similarity with an accuracy equal to or smaller than a pixel, calculating a shift using a region having a maximum maximum similarity, performing alignment, and comparing and detecting defects.

[Brief description of the drawings]

FIG. 1 is a view showing a silicon wafer on which a semiconductor integrated circuit is formed.

FIG. 2 is an enlarged view showing a circle portion of the silicon wafer of FIG. 1;

FIG. 3 is a view showing a reference image and an image to be inspected of a semiconductor for explaining a semiconductor inspection method according to the present invention.

FIG. 4 is a diagram showing a semiconductor inspection device according to the present invention.

FIG. 5 is a diagram illustrating an example in which misalignment occurs between a reference image and an image to be inspected.

FIG. 6 is a diagram illustrating an example of a similarity calculated for each corresponding region.

FIG. 7 is a diagram showing an example in which one of a reference image and a test image includes the other in a corresponding area.

FIG. 8 is a diagram illustrating an example in which a region to be a positioning region is set in advance in a peripheral portion of a reference image and a test image;

FIG. 9 is a diagram illustrating an example of a similarity calculated at a portion corresponding to FIG. 8;

FIG. 10 is a diagram showing an example of similarity in a plurality of tests.

FIG. 11 is a diagram illustrating an example in which the calculation of the similarity at the time of alignment is performed with an accuracy equal to or less than a pixel.

FIG. 12 is a diagram illustrating a method of imaging a reference image with a higher resolution.

FIG. 13 is a diagram showing a high-resolution imaging optical system for a reference image and an imaging optical system for an image to be inspected.

[Explanation of symbols]

 2 Semiconductor 2A Reference image 2B Inspection image 7A, 7B Area 7C Positioning area 10 Semiconductor inspection apparatus 12 Stage 14 Camera 16 Light source 18 Half mirror 22 First memory 24 Second memory Memory 26 ... Control device

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Fumiyuki Takahashi 4-1, 1-1 Uedanaka, Nakahara-ku, Kawasaki-shi, Kanagawa Prefecture Inside Fujitsu Limited (72) Inventor Mitaka Oshima 4 Kamikadanaka, Nakahara-ku, Kawasaki-shi, Kanagawa Prefecture 1-1 1-1 Fujitsu Limited (72) Inventor Hiroyuki Tsukahara 4-1-1 1-1 Kamidadanaka, Nakahara-ku, Kawasaki City, Kanagawa Prefecture F-term within Fujitsu Limited (reference) 2G051 AA51 AB01 AB02 BA00 CA03 CA04 CB01 DA07 EA08 EA14 EB01 EB02 EC03 EC06 ED07 4M106 AA01 AA02 BA20 CA38 CA40 CA41 DA15 DB04 DJ17 DJ18 5B057 AA03 DA03 DA07 DB02 DC02 DC32

Claims (5)

[Claims] 1. A reference image of a semiconductor and an image to be inspected of a semiconductor are fetched, and a plurality of areas corresponding to the reference image and the image to be inspected are defined, and a plurality of areas corresponding to the reference image and the image to be inspected are defined. The similarity between the regions is obtained, a positioning region is determined in accordance with the obtained similarity for each region, the reference image and the image to be inspected are aligned in the positioning region, and the reference image and the image are compared. A semiconductor inspection method, wherein a comparative inspection of inspection images is performed. 2. The semiconductor inspection method according to claim 1, wherein a shift amount between the reference image and the image to be inspected is calculated in the alignment area. 3. The semiconductor inspection method according to claim 1, wherein an alignment region is set in a peripheral portion of each of the reference image and the inspection image. 4. An optical magnification when capturing a reference image,
Obtain a higher definition image than the inspected image by making it larger than the optical magnification at the time of imaging the inspected image, and reconstruct to the same size as the inspected image when comparing the reference image and the inspected image. By calculating the similarity and the shift amount, it is possible to reduce the influence of blur due to shift of pixels or less when the reference image and the image to be inspected have the same size, and to improve the accuracy of the shift amount calculation. The semiconductor inspection method according to claim 1. 5. An image pickup means for picking up an image of a semiconductor, a storage means for storing a reference image of a semiconductor and an image to be inspected of a semiconductor, and a plurality of areas respectively corresponding to the reference image and the image to be inspected are determined. Area setting means, positioning area setting means for determining a similarity between corresponding areas of the reference image and the inspection image, and determining a positioning area according to the obtained degree of similarity for each area; A semiconductor inspection apparatus comprising: a positioning unit that positions the reference image and the image to be inspected in an area for use; and an inspection unit that performs a comparative inspection between the reference image and the image to be inspected.