JPH05129397A - Foreign object detection method and apparatus - Google Patents

Abstract

(57) [Abstract] [Purpose] To illuminate a patterned sample with laser or white light from diagonally above, detect the images of two adjacent patterns from above, and provide a multi-step threshold for the difference image between the two. Thus, it is possible to realize a foreign matter detection method and an apparatus therefor, which change the detected foreign matter size on the low step pattern on the sample and on the high step pattern to enable highly accurate foreign matter detection. [Structure] On a wafer sample 1 on which a large number of LSI patterns are formed, a laser 8 is obliquely illuminated and diffracted light is converted into an objective lens 9.
The image is picked up by the image sensor 11, and the current chip image is focused on 22.
The adjacent chip image is stored in 21, and the difference image between the two is stored in 23. The detected image signal is binarized by the threshold values l ₁ and l ₂ , further expanded, logical product (AND) is taken, the coefficient is multiplied by 32 and 33 and added by 35, and the threshold value of 3 stages k ₀ , k ₁ , k
Create a threshold signal with ₂ and store in 36. At 37,
The difference between the difference image signal and the three-step threshold signal is taken and binarized,
The foreign matter 4 on the sample is detected.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for detecting minute foreign matter on a sample, and more particularly to a method and apparatus for detecting foreign matter suitable for detecting minute foreign matter on a patterned wafer which is a product. It is about.

[0002]

2. Description of the Related Art Taking a foreign matter inspection on a patterned wafer as an example, the conventional technique is, for example, as shown in Japanese Unexamined Patent Publication No. 1-117024, in which a laser beam or a laser beam is slanted from above the wafer 1 in FIG. Illuminate with white light 8, detect two adjacent chip images 21 and 22, create a difference image 23 between the two,
Among them, the one that outputs a certain value or more is determined as a foreign matter. This is because the circuit patterns of all the chips on the same wafer are the same, so that when the difference image is taken, the circuit pattern signal is erased and only the foreign matter signal remains. This makes it possible to detect foreign matter on the patterned wafer. Although FIG. 1 shows an overall configuration diagram of the present invention, which will be described later, since a part of the configuration includes the conventional technique, it has been used in common.

[0003]

FIG. 10 shows an image of the local multi-valued image memories 21 and 22 (shown in FIG. 1) detected by the prior art.
21c and 22c. It is assumed that two foreign substances 4 are present in 22c. The image of 21c is represented by g ', the image of 22c is represented by f', and the image obtained by the equation (3) is represented by e'in the local multi-valued image memory 23c.

[0004]

## EQU1 ## e '= | f'-g' | (3) If the images f'and g 'are exactly the same, e'will be zero, but in reality, a very small position between them. There is a shift and an output signal difference, and e ′ does not become zero.

Scan line 5 in local multi-valued image memory 23c
The output signal of 0 is indicated by 51 in FIG. When this is binarized with the fixed threshold value 52, a binarized signal 53 is obtained as shown in FIG. However, this makes it impossible to distinguish between the output signal difference based on a minute positional deviation and the signal based on the foreign matter 4, and the foreign matter 4 cannot be detected.

Therefore, an object of the present invention is to solve the above-mentioned problems of the prior art and to provide a foreign matter detection method and apparatus capable of surely detecting minute foreign matter even on a wafer having a complicated pattern. Especially.

[0007]

The above-mentioned object is shown in FIG.
As shown in (c), the local multi-valued image memory 21c storing the original image is not fixed to the fixed threshold value applied to the local multi-valued image memory 23c storing the difference image as in the prior art, 22c, the multi-step threshold value 54
Is generated and applied to the local multi-valued image memory 23c, the foreign matter signal 56 is reliably detected and achieved as shown in FIG. 11 (d). Although a wafer for manufacturing an LSI will be described as a representative example as a foreign matter detection sample here, the present invention is not limited to this, and various circuit boards in which a plurality of identical patterns are repeatedly formed, such as a liquid crystal board, are also used. It goes without saying that electronic parts on which various patterns are formed can be inspected.

[0008]

In general, in a semiconductor wafer, the output signals of 21c and 22c shown in FIG. 10 are small and the difference between f'and g'is small in the part where the pattern step is small. Therefore, in this part, the absolute value e'of the image difference is e '. Is small. On the other hand, the output signal is large at the end of the memory mat or a part where the pattern step is large, such as a part of the peripheral circuit, and the difference between f ′ and g ′ is also large. Therefore, the absolute value e ′ of the image difference is also large at this part. Becomes Therefore, paying attention to f ′ and g ′, the threshold of e ′ is set to be large in a portion where these output signals are large, and the threshold of e ′ is set to be small in a portion where the output signals of f ′ and g ′ are small. For example, minute foreign matter can be detected in a portion where the output signals of f ′ and g ′ are small. Further, since the portion where f ′ and g ′ are large is 0.1% or less of the whole wafer, even if the threshold value of e ′ is increased in this portion, there is no great influence.

In the present invention, as shown in FIGS. 11C and 11D, the scanning line signal 51 is multiplied by the multi-stage threshold signal 54,
By outputting the binarized signal 55, a means for correctly detecting a foreign substance is provided.

[0010]

Embodiments of the present invention will be described below with reference to FIGS. First, the overall configuration of the present invention will be described with reference to FIG. A large number of memory mats 2 each having a fine pattern, a memory control circuit 3, and other peripheral circuits (not shown) are regularly formed on the wafer 1 as shown in a partially enlarged view. A foreign matter detection sample is one in which the foreign matter 4 is present. The wafer 1 is vacuum-sucked by a wafer chuck 5, and is mounted on an x stage 6 that reciprocates in the x direction and a y stage 7 that reciprocates in the y direction.

As shown in FIG. 2, the entire wafer moves in the x direction at a constant distance (for example, 200 mm) at a constant speed, and then moves in the y direction at a constant distance (for example, 1.5 mm) in steps, and then -x.
Repeat the movement to move in the direction. Fixed area on wafer
(For example, 0.2 mm x 2 mm) is a semiconductor laser 8a, 8b, 8c,
Illuminated from 4 directions by 8d (not shown). The reflected light on the wafer surface is condensed by the objective lens 9, reflected by the mirror 10, and converted into an electric signal by the image pickup device 11 (usually a one-dimensional CCD linear image sensor is used). The detection signal is stored in the local multi-valued image memory 21 via the delay memory 20 for one chip. (This image is g). At the same time, this signal is also stored in another local multi-valued image memory 22 (this image is f), and the image obtained by the following equation (1) (this image is e) is stored in the memory 23.

[0012]

[Equation 2] e = | f−g | (1) Next, the signals of the local multi-valued image memories 21 and 22 are thresholded by l
₁ and binarized by (El ₁ at will) local binary image memory 24
After being stored in a and 24b and expanded by 2 pixels respectively, they are stored in the local binary image memories 25a and 25b, and the logical product (AN
D) The result is passed through the element 26 to the local binary image memory 27
Store at.

Similarly, local multi-valued image memories 21 and 22 are provided.
Signal is binarized with a threshold value l ₂ (meaning L ₂ ), and the result is passed through the same electric circuit as above, and the result is stored in the local binary image memory 3
Store at 1.

[0014] On the other hand, the potentiometer 32 is set by a coefficient k ₁ -k _0, and the potentiometer 33 is set by a coefficient k ₂ -k ₁ -k _0, respectively local binary image memory 2
The potentiometers 32,
The output of 33 is led to the adder circuit 35. The output voltage is set to k ₀
The output of the constant voltage device 34 set to is applied to the adder circuit 35, and the output is guided to the local multi-valued image memory 36 and stored.
(This image is h). By binarizing the image e with the image h in the binarizing circuit 37 according to the following equation (2), the target final output signal 38 (binary signal of j = 1 or j = 0) is obtained.

[0015]

[Equation 3]

Next, the operation of the local multi-valued image memory having the above configuration will be described in detail with reference to FIG. Since many same chips are formed on the wafer 1, if the chip pitch is p, two images 21 separated by p in FIG.
By detecting a and 22a, a difference image between them is obtained. The delay memory 20 of FIG. 1 is a memory for delaying the image signal by p, and the image 21a of FIG. 3 is stored in the local multi-valued image memory 21 by passing it. On the other hand, 22
The currently detected image 22a is stored in. Figure 1
In the local multi-valued image memory, for example, as shown in FIG.
An image of 56 × 256 pixels is multivalued (for example, 8 bits, 25
It is stored in 6 gradations.

Next, the operation of FIG. 1 will be described with reference to the image of FIG. 4 showing the contents of the local multi-valued image memory. The signal on the scanning line 40 of the image 21b in the local multi-valued image memory 21 is shown by the signal 41 in FIG. Two foreign matters 4 are present on the scanning line 40. Similarly, the corresponding scanning line signal of 22 images (not shown, with no foreign substance 4 present) is shown at 42. This is multiplied by a threshold value l ₂ (meaning L ₂ ) to be binarized, and a logical product (AND) of the two values is shown as a binarized signal 31 a, and a threshold value l ₁ (meaning L ₁ ) A binarized signal 27a is obtained by performing the same operation by multiplying. Further, a scanning line signal corresponding to the scanning line 40 (shown in FIG. 4) of the local multi-valued image memory 36 (shown in FIG. 1) is shown at 36a in FIG. This signal is k ₀ , k
_It has a three-step threshold value having values of ₁ and k ₂ , and is k ₀ when the output of the original signals 41 and 42 is low, k ₂ when the output is high, and has a medium output. Then k ₁
Has become. The scanning line signal corresponding to the scanning line 40 (shown in FIG. 4) of the local multi-valued image memory 23 (shown in FIG. 1) is sent to 23a.
Shown in. Although the scanning line signal 43 in which 36a and 23a are overlapped is shown for the sake of explanation, it can be seen from this figure that the output 44 of the binarization circuit 37 (shown in FIG. 1) is obtained. A foreign substance signal 46 based on the foreign substance 4 is correctly obtained as the output signal 44.

Next, the threshold values l ₁ , l ₂ , k ₀ , k ₁ , k ₂
An example of a method of obtaining the above will be described with reference to FIGS.
These threshold values are determined for each process of the wafer manufacturing process and need not be set for each wafer. Therefore, when a wafer of a certain process is completed, it is necessary to obtain these threshold values and register them as work conditions before the subsequent series of works.

In FIG. 1, the completed wafer 1 is mounted on the wafer chuck 5, and the wafer is moved as shown in FIG. 2 to detect signals on the entire surface of the wafer. At this time, it is necessary to take out the signals from the local multi-valued image memories 22 and 23 in FIG. 1 and create a histogram with a histogram measuring device. As shown in FIG. 6, the histogram means a diagram in which the horizontal axis 50 represents the gradation of an image and the vertical axis 51 represents the number (frequency) of the pixels of each gradation.

_First , how to obtain the original image threshold values l ₁ and l ₂ will be described with reference to FIG.
When the output (shown in FIG. 1) is connected to the histogram measuring device, the result as shown in FIG. 6 is obtained, so that the gradations at which the frequency is at the bottom are l ₁ and l ₂ . Here, a region 52 is a signal of a peripheral circuit part having a small pattern step on the wafer, a region 53 is a signal of a circuit part having a complicated structure such as a memory mat part, and a region 54 is like a memory mat end. This is a signal in a portion where the pattern step is extremely high.

Next, how to obtain the difference signal threshold value k ₀ will be described with reference to FIG.
When the signal of ₁ or less is output to the local multi-valued image memory 23 for storage and the output is connected to the histogram measuring instrument, the result of FIG. 7 is obtained. Although k ₀ ′ is the maximum value of the difference signal, k ₀ ′ changes in the case of different wafers even in the same process. It is assumed that twice the _{value of} ′, that is, 2k ₀ ′ is k ₀ .

Similarly, the method of obtaining the difference signal thresholds k ₁ and k ₂ will be described with reference to FIGS. 8 and 9. The signals of l ₂ or less at the outputs of 21 and 22 are led to 23 and stored, and the histogram is stored. Since the result shown in Fig. 8 is obtained when it is created, k ₁ = 2k ₁
´ Further, when all the outputs of 21 and 22 are led to 23 and stored and a histogram is created, the result of FIG. 9 is obtained, so k ₂ = 2k ₂ ′. By the above operation, l ₁ ,
It is possible to obtain l ₂ , k ₀ , k ₁ and k ₂ .

In the above embodiment, the local multi-valued image memory 2
Two cases have been described in which the threshold values applied to ₁ and 22 are l ₁ and l ₂ , but the number may be one or more, and may be changed according to the purpose. Further, the number of voltages applied to the adder circuit 35 is three in the present embodiment, but the number may be one or more, and may be changed according to the purpose. Further, in this embodiment, the logical product (AND) of 29a and 29b is taken, but this may be a logical sum (OR) and may be changed according to the purpose.

In the above embodiment, a semiconductor wafer for manufacturing an LSI has been described as an inspection sample, but the same applies to a substrate having a large number of liquid crystal patterns formed on a glass substrate, and the present invention is extremely important in quality control. Played a role.

[0025]

As described above, according to the present invention,
The threshold for detecting foreign matter can be changed in multiple steps, and since foreign matter was detected with only one threshold so far, a small foreign matter could not be detected. Can detect a minute foreign substance, and can detect a relatively large foreign substance in a region having an extremely high pattern level difference such as the end of the memory mat, which greatly improves the detection performance.

[Brief description of drawings]

1 is an overall configuration diagram of the present invention having an image signal processing function by the multi-step threshold method. 2 is a perspective view showing the movement of the wafer. 3 is an explanatory diagram of the operation of the local multi-valued image memory. 4 is a diagram showing the contents of the local multi-valued image memory. 5 is an explanatory diagram of the multi-step threshold method according to the present invention. 6 is an explanatory diagram of how to obtain the original signal thresholds l ₁ and l ₂ . 7 is an explanatory diagram of how to obtain the difference signal threshold value k ₀ . 8 is an explanatory diagram of how to determine the difference signal threshold value k ₁ . 9 is an explanatory diagram of how to determine the difference signal threshold value k ₂ . 10 is an explanatory view of a detected image obtained by the conventional technique. 11 is an explanatory view of the binarization method according to the prior art.

[Explanation of symbols]

1 ... Wafer, 4 ... Foreign matter, 8a, 8b, 8c ... Semiconductor laser, 9
… Objective lens, 11… Image sensor, 20…
Delay memory, 21, 22, 23, 36 ... Local multi-valued image memory, 24a, 24b ... Binarization circuit (threshold l ₁ ), 2
8a, 28b ... Binarization circuit (threshold l ₂ ), 25, ₂
7, 31 ... Local binary image memory, 32, 33, 34 ... Potentiometer, 35 ... Addition circuit, 37 ... Binarization circuit, 38 ... Final output signal (j), 46, 56 ... Foreign object signal.

Claims (6)

[Claims] 1. A sample on which a plurality of identical patterns are formed is illuminated obliquely from above with a laser or white light to detect original images of two patterns adjacent or close to each other on the sample, and a signal of a difference image between them. A method for detecting foreign matter on a sample based on the above, wherein the original images of the two patterns are binarized by applying one or more threshold values, and the two binarized images obtained here are Logical product or logical sum is taken to make one binary image,
By multiplying this by a certain constant to create a multi-stage threshold signal, and setting this as the threshold of the difference image signal, the threshold of the difference image signal is increased when the original image signal is large. And a foreign object detection method in which the threshold value of the difference image signal is reduced when the original image signal is small. 2. The foreign matter detection method according to claim 1, wherein the binary images are obtained by applying one or more threshold values to the original images of the two patterns, and the binary image. And a logical sum of two binarized images obtained by subjecting the signal to an expansion process of one pixel or more to form one binarized image. 3. The foreign matter detecting method according to claim 1 or 2, wherein the sample is a semiconductor wafer having a plurality of LSI patterns, and a chip configured for each pattern is an inspection target. 4. The foreign matter detection method according to claim 1 or 2, wherein the sample is a liquid crystal substrate having a plurality of liquid crystal element patterns, and liquid crystal elements configured for each pattern are to be inspected. 5. A sample on which at least a plurality of identical patterns are formed is illuminated obliquely from above with a laser or white light,
A foreign matter detection apparatus comprising: means for detecting two original patterns of adjacent or close patterns on a sample; and means for detecting a foreign matter on the sample based on a signal of the difference image. A method for binarizing one original image of one pattern by applying one or more threshold values and a logical product or a logical sum of the two binarized images obtained here to create one binarized image. , A means for creating a multistage threshold signal by multiplying this by a constant, and by using the multistage threshold signal as the threshold of the difference image signal, when the original image signal is large And a means for increasing the threshold value of the difference image signal and decreasing the threshold value of the difference image signal when the original image signal is small. 6. The foreign matter detecting device according to claim 5, wherein the original images of the two patterns are multiplied by one or more threshold values to obtain a binarized image signal, and the binarized image. A foreign matter detecting apparatus comprising: means for taking a logical product or a logical sum of two binarized images obtained by subjecting a signal to an expansion process of one pixel or more to form one binarized image.