JPS61176807A - Inspection of pattern - Google Patents

Abstract

PURPOSE:To detect a micro-defect with high reliability without generating a pseudo defect by cutting out a pattern to be inspected by using a small window, matching the position of the partial image within a model pattern and detecting the rate of dissidence. CONSTITUTION:The pattern 5 to be inspected is cut out as the partial image 7 by using the window consisting of, for example, 3X3 picture elements centering at the certain picture element 7A and the model pattern 6 is cut out as the partial image 8 by using the window of, for example, 7X7 picture elements centering at the picture element 7B corresponding to the picture element 7A. The partial image 7 is moved in the partial image 8 and both partial patterns are so matched in position as to minimize the number of the dissidence picture elements. The dissidence of the pattern 5 to be inspected and the part A of the model pattern 6 is thereby absorbed and the generation of the pseudo defect is prevented. On the other hand, the dissident picture elements remain even if the partial image 7' is matched in postion in whatever way within the partial image 8'. A defect 9 is thus detected.

Description

[Detailed Description of the Invention] (Field of Application of the Invention) The present invention relates to a pattern inspection method, and is particularly used for defect inspection of circuit patterns of printed circuit boards and semiconductor integrated circuits Ueno-1 or masks for manufacturing the same. This invention relates to a pattern inspection method suitable for.

[Background of the invention]

Figures 5 (4) and (b) show one of the masks for manufacturing integrated circuits.
In the enlarged view of the part, (4) is a complete pattern 1 with no defects.
, (A) is the protrusion 2a of the pattern, cut tI! A2b shows defect pattern 3 including isolated defect 2G. Conventionally, the method of inspecting defects in such mask patterns is to simultaneously image two chip patterns having the same pattern on the mask to be inspected, and use one pattern as a model pattern and the other pattern as the pattern to be inspected. A method has been implemented in which the two are compared and a portion where there is a difference between the two is determined to be defective.

However, even if two chip patterns having the same pattern on a mask to be inspected are simultaneously imaged, each is used as a model pattern and a pattern to be inspected, and the positional deviation between the two is completely removed and filtered, (G) As shown in the figure, the sample pattern (turn 1 (shown in the second part) and the pattern to be inspected (shown in the second part) do not match completely except for the defective parts 2cL to 20. Possible causes include the following reasons.

(1) Two patterns are formed with subtle differences to the extent that they are not defects during mask production such as original, current, etching, etc. (2) Two patterns that should be compared are photographed. The shape distortion increases due to the unbalance of the optical system and the running error of the stage on which the mask is placed.

Therefore, as shown in (d), in addition to the defects 2a to 2G, the mismatched portion between the model pattern 1 and the pattern to be inspected 3 is enriched with pseudo defects 4 caused by mismatch between the pattern and the contour. become.

As a method for removing this pseudo defect 4, there is a conventional method described in, for example, Japanese Patent Publication No. 54-37475. In this method, in order to remove false defects, a mismatched part between the model pattern and the pattern to be inspected is detected as a defect candidate, and at the same time, the contour parts of the pattern are extracted from both the model pattern and the pattern to be inspected. .

Then, a common contour portion is extracted from the contour portions of each of the model pattern and the pattern to be inspected, and among the 8U defect candidates, those belonging to the common contour portion are removed as pseudo defects.

However, with this method, although the cut f@24 and isolated defect 2G of the pattern shown in #145 (6) can be detected, the minute protrusion 24 that occurs at the scar of pattern 4111 lies across the common contour part and cannot be detected. becomes 1-. Furthermore, in order to detect such a minute defect 1If1, if the pattern detection lll1element size is made small, the inspection speed will be slowed down and the throughput of the apparatus will be reduced.

(Object of the Invention) An object of the present invention is to provide a pattern inspection method that solves the above-mentioned problems of the prior art and can detect small defects with high reliability without generating pseudo-defects.

[Summary of the invention]

In order to achieve the above objectives, the present invention cuts out one turn of the test f/'' using a small window of several pixels x several pixels, positions the partial image with the same model pattern, and calculates the amount of discrepancy. The special feature is to detect the ridges.This allows the slight differences in the tape pattern and the effects of positional deviations to be removed and the occurrence of false defects to be suppressed.

[Example of invention]

An embodiment of the present invention will be described below based on the drawings.

Figure 1 (4) - (- is a map of the original castle of Ming Dynasty, (&) is B
The pattern 5 to be inspected, which has 9kt of defects due to protrusions of the pattern, and the model pattern 6 are stacked on top of each other! At'', (
b) The diagram shows the square portion 2s xsII & Ii of the pattern to be inspected 5
The part cut out using the 787i1ii window = m7 (r-) The figure shows a partial image 8 of the same part of the hand pattern 6 cut out using the window of the 787i1ii plan, and the figure (d) shows the part to be inspected. Part-1 result 7 of pattern 5 and partial image 8 of model pattern 6 are aligned so that the number of mismatched images is minimized. (e) - (S figure is (1 to ld)) Similar to the figure, the partial image i17, partial image 8, and the number of unmatched images of the partial image cut out from part B of the pattern to be inspected 5 and the model pattern 6 This shows the state in which the positions are aligned so that the distance is minimized.

(&) As shown in the figure, when the pattern 5 to be inspected and the pattern 6 are superimposed and the mismatched part is detected as a defect, the pattern Cs is displayed at the same time as the defect 9 in part B. However, as described above, the pattern deviation of Cs is a pseudo defect that should not be detected as a defect. In order to prevent this pseudo defect from occurring, (
b) As shown in the figure, the pattern to be inspected 5 is cut out as a part iiijw7 using a window of 5 x 3iai elements, for example, centered on a certain pixel 7A, and (C) the sample pattern 6 is cut out as a part iiijw7 as shown in the figure. For example, use a window of 7 parts and 7 pixels to create partial image 1 centered on &X7B corresponding to
Cut out as 1!8. The section sentence partial image 7 is moved within the partial image 8, and alignment is performed so that the number of mismatched lines between the two partial images is minimized as shown in FIG. By detecting the mismatched portion after alignment using this window as a defect, the mismatch in the gap portion of the pattern to be inspected 5 and the model pattern 6 shown in the figure (a) can be absorbed, and similar defects can be detected. Occurrence can be prevented.

Furthermore, by performing the same alignment over the entire surface of the pattern to be inspected, the mismatch due to the pattern shift of the 0 part shown in Figure 3 can be absorbed and the occurrence of similar defects can be prevented. On the other hand, the pattern of part B shown in (4) Figure (e) no matter how much the partial image gR7 shown in Figure (e) is aligned with the partial image lj image 8 circle shown in (f) - is (
t) Defect 9 can be detected because a mismatched fraction remains as shown in the figure. For example, if the window f)-r size of 5 pixels and 5 pixels to cut out the pattern 5 to be inspected is set to 5 pixels and 5 pixels, the alignment correction m range will be ±1-element, for example, 9 pixels and 9 parts. In this case, the alignment correction range is ±3 pixels. By changing the wine toe size at which the model pattern 6 is cut out in this manner, the alignment correction range between the pattern to be inspected and the model pattern can be changed.

FIG. 2 is a block diagram of an embodiment of the butter/inspection method according to the present invention described above. In the diagram, 10 is 5X3
A partial image of 5 parts and 5 pixels is cut out from the pattern to be inspected 5 at the cutting exit path. 11 is a 7X7 cutout circuit, model pattern 6
A partial image of 7 parts and 7 strokes is cut out from the image. 12-1 to 12
-25 is an SXS cutting circuit which further cuts out 25 partial image sources of 3 parts and 3 strokes from the partial image of 7 parts and 7 strokes cut out from the model pattern 6 by the 7X7 cutting circuit 11. 13
-1 to 13-25 are exclusive ORs, which are exclusive ORs for each Tanigagyo for the output from the 3X5 cutout circuit 10 and the output from the 3X3 cutout circuits 12-1 to 12-25. Calculate.

- The logical sum circuits 14-1 to 14-25 calculate the logical sum for each output from the exclusive logical sums 13-1 to 15-25, and - In the filling circuit 15, the logical sum circuits 14-1 to 14 −
The logical AND of the outputs of 25 is calculated.

The output of this -filling becomes the defect signal 16.

Figure 3 shows 3x
3 is an explanatory diagram of the operation of cutting out 25 3-part, 3-art work partial images from 3-cutting circuits 12-1 to 12-25; FIG. (') The figure shows a partial image cut out by the 7X7 cutout circuit 11.
', (J) shows 2 images further cut out from partial image 8.
Five 3-part, 3-pixel sub-pixels 8-1 to 8-25 are shown.

As shown in the figure, the 5 part 3 painting partial images 8-1 to 8-25 of 25-a are cut out so that each 5x5 painting purple part of the 7 part 7 painting tree partial image 8 is centered.

In this way, 25 3's were cut out from the model pattern 6.
Comparing the part 3 pixel partial images 8-1 to 8-25 with the SXS pixel partial image 7 cut out from the inspected pattern 5 shown in FIG. 2 and calculating the exclusive OR for each mg. Yes, partial images 7 can be aligned within partial images 8. If at least one of the outputs of the exclusive ORs 13-1 to 13-25 is all "01", the alignment is performed, and the test lid pattern 5 and the model pattern 6 are
The defect signal 16, which is the output of the Fa buried layer 15, also becomes "0'", indicating that the values are completely matched. On the other hand, if a mismatched picture kite occurs no matter how the partial image 7 is aligned within the partial image 8, the defect signal 1 which is the output of the logical product 15
6 is ``11'' and defects can be detected.

The above explanation takes as an example the case where both the pattern to be inspected 5 and the model pattern 6 are binary images, but the same effect can be obtained even if the pattern to be inspected 5 and the model pattern 6 are both multivalued images. be able to.

FIG. 4 is a block diagram of another embodiment of the pattern inspection method according to the present invention, showing a case where both the pattern to be inspected and the model pattern are multivalued images. As in FIG. 2, first, from the pattern to be inspected 5', p5 is
At the same time as cutting out the 85-stroke part picture 7, model pattern 6
A 7 x 7 pixel partial image 8'' is cut out from the 7 x 7 cut-out circuit 11'. From the 7 x 7 pixel partial direct image 8, 925 pieces are cut out by the 3 x 3 cut-out circuits 12-1 to 12-25. After cutting out a 6×3 pixel partial image, the partial image and the partial image 7
Difference circuits 17-1 to 17-25 calculate the difference between each fraction.
Calculate accordingly. Next, cough differential circuits 17-1 to 17-2
Absolute value circuits 18-1 to 1B-
25, the output is sent to integration circuits 19-1 to 19.
Integrate at −25. These integrating circuits 19-1 to 19-
The output 25 represents the amount of image mismatch at each position when the partial image 7 of the distortion pattern 5 to be tested is aligned within the partial image rule B of the model pattern 60. Therefore, if partial image 7 has a mismatch amount of zero at a certain position within partial image 8,
When the alignment is complete, the output of the integrating circuit 19 corresponding to the position becomes zero. However, when aligning multivalued images, the amount of mismatch is unlikely to be completely zero, and if the amount of mismatch is less than a certain value, it is necessary to consider that alignment has been achieved. Therefore, the integration circuits 19-1 to 19-25
The output of ! is compared with a certain constant 21 by the comparison circuits 20-1 to 20-25, and ! If the output of the integrator circuit written in IJ is greater than the constant 21, it outputs "1"; otherwise, it outputs "0"'.

Then, by calculating the logical product of the outputs of the comparison circuit in the logical product circuit 15, the partial image 7'' of the test target I (turn 5') is changed to the partial image 8'' of the sample turn 6'. A defect signal 16 can be obtained which is "0" if alignment is achieved within the range, and "fail" if the amount of mismatch exceeds a certain value no matter how the alignment is performed.

As explained in detail above, according to this embodiment, even if the size of the pattern to be inspected is slightly different from the pattern I (turn), it is possible to detect minute defects with high reliability without generating false defects. In addition, according to this embodiment, it is possible to detect minute defects with a size comparable to the detection pixel size, so it is possible to image the turn at high speed. This is effective in significantly shortening inspection time.

〔Effect of the invention〕

As described in detail above, according to the present invention, it is possible to realize a high-speed and highly reliable pattern defect inspection device, thereby improving the effectiveness of, for example, the mask pattern inspection process for integrated circuit manufacturing, Improving the reliability of products such as C, etc.
Remarkable effects on cost reduction etc.0

[Brief explanation of drawings]

Figure 881 is a diagram showing the principle of the pattern inspection method according to the present invention, Figure 2 is a block diagram of an embodiment of the turn inspection method based on the principle of Figure 1, and Figure 5 is the operation of the 3x3 cutting circuit. FIG. 4 is a block diagram of another embodiment of the pattern inspection method according to the present invention, and FIG. 5 is an explanatory diagram of the conventional I-turn inspection method. 5.5... Turn to be inspected 6.6... Model pattern io, io, 12-1 to 12-25, 12
-1 to 12-25...3x5 cutout circuit 11.11...7x7 cutout circuit 13-1 to 15-25...exclusive OR 14-1 to 14-25...logic Sum 15...Logic product 16...Defect

Claims (1)

[Claims] 1. The pattern to be inspected is compared with a model pattern corresponding to the pattern or a pattern having the same shape as the pattern to be inspected, and the pattern to be inspected is determined. In a pattern inspection method for detecting defects,
The pattern to be inspected is cut out as a two-dimensional image using a minute window, the cut out partial image is aligned with the pattern to be compared, and the inspection is performed based on the amount of mismatch between the patterns after alignment. A pattern inspection method characterized by detecting defects in a target pattern. 2. In the pattern inspection method according to claim 1, a difference is obtained without binarizing the signal cut out as the two-dimensional image, and a signal obtained by integrating the absolute value of the difference is compared. , a pattern inspection method for calculating the mismatch amount.