JPH0644362A - Pattern matching method - Google Patents

Abstract

PURPOSE:To accelerate pattern matching by performing sampling at prescribed interval to an image smoothed from the prescribed direction of a wiring pattern image. CONSTITUTION:An LSI is designed based on prescribed pattern rules. Namely, at the time of pattern matching in an X direction, among information in a Y direction, the information to be changed at shorter intervals in comparison with the wiring width of the LSI can be omitted. Before performing pattern matching processing, smoothing processing is performed for erasing fine dot noise to an SME image to be obtained with the equal span width to the wiring width in the Y direction beforehand. Therefore, the S/N of the image can be improved without damaging information required for the pattern matching. Further, the information of a processing object can be reduced by compressing the SEM image by sampling the image at intervals shorter than the span width and constituting the image again.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern matching method, and more specifically, for example, CAD (Computer Aided).
Design) CAD such as mask drawing, which is suitable for use in fields such as an integrated electron beam tester (hereinafter referred to as EB tester)
The present invention relates to a pattern matching method in which data is used to specify measurement points on a mask diagram to specify detailed measurement points on an image display of an SEM image linked with the mask diagram and to associate the mask diagram with the SEM image.

BACKGROUND OF THE INVENTION In recent years, LSI (Large Scal)
Semiconductor integrated circuits represented by e integrated circuits) are becoming highly complex due to high density and high integration. For the design development and failure analysis of LSI, the voltage state of wiring inside the LSI ( For example, evaluation of waveform timing, amplitude, etc. has become very important.

As an internal operation analysis means of LSI, LS is used.
There are a probe method in which a metal probe is brought into direct contact with the wiring in I and a method in which a non-contact probe such as an EB probe is used. In the method using the optical microscope image and the EB probe, the probe position is determined on the image while observing the SEM image.
There is a demand for a method of automatically aligning a probe with a measurement position on an LSI by associating it with an M image and designating a measurement point in a wiring pattern obtained from CAD data. In order to realize such probe alignment, a wiring pattern and SE obtained from CAD data are used.
It is necessary to accurately associate the M image.

However, in reality, as described above, the LSI
Since the inside is extremely densified, the acquired SEM
It is difficult to align the image with the wiring data obtained from the CAD data. Therefore, it is necessary to perform a matching process for accurately aligning the distribution data obtained from the CAD data and the SEM image.

[0005]

2. Description of the Related Art As a conventional pattern matching method of this type, for example, there is a method shown in FIG. Figure 5
3 is a flow chart for explaining the operation of the conventional example. First, a CAD data matching area is determined (step 11), and an edge portion is extracted from the CAD data in this area (step 12).

FIG. 6A shows an example of edges extracted from CAD data. Since the CAD data is originally composed of vector data, the process of step 12 is easy. Next, the CA determined in the process of step 11
An SEM image of the area corresponding to the D data matching area is acquired (step 13), and as shown in FIG. 6B, an image in which an edge portion is extracted from the acquired SEM image is obtained.

Then, the degree of correlation between the edges of the edge image of the CAD data (FIG. 6A) and the edge image of the SEM image (FIG. 6B) is used as an evaluation amount, that is, the mutual correlation. The pattern matching processing is performed by obtaining the magnification and the shift amount so that the correlation becomes maximum.

[0008]

However, in such a conventional pattern matching method, the CAD
In order to perform pattern matching between the data (mask diagram) and the SEM image, first, the mask diagram and the SEM in the X direction are shown.
The image and the image were compared to obtain the magnification and the shift amount, and then the magnification and the shift amount were obtained in the Y direction.

That is, in order to perform pattern matching, it is necessary to set the magnification and the shift amount in the X direction and the Y direction, and four parameters (two directions of X and Y × 2).
Parameter) matching was required. And
In order to speed up the processing, the X-direction and the Y-direction are independently matched, but the information amount of the SEM image accompanying the high density and high integration of the semiconductor integrated circuit represented by LSI However, there is a problem in that the matching processing time increases due to the increase of

[Object] Therefore, an object of the present invention is to provide a high-speed pattern matching method.

[0011]

In order to achieve the above object, a pattern matching method according to the present invention has wiring pattern data in a predetermined rectangular area as design data and wiring in an XY plane on an actual device corresponding to the rectangular area. A pattern matching method for performing matching with a pattern image, wherein the wiring pattern image is sampled with a predetermined width in a predetermined direction, and the wiring pattern data is obtained in a direction perpendicular to a sampling direction of an image obtained by sampling. It is configured to collate, compare, and associate with.

In this case, the wiring pattern image is subjected to smoothing processing for smoothing with a predetermined span width from a predetermined one of the X and Y directions, and the smoothed span width is applied to the smoothed image. It is preferable to perform sampling with a shorter interval width, and it is effective that the span width of the smoothing is smaller than the minimum image width in the wiring pattern image in the smoothing direction in the smoothing process.

[0013]

According to the present invention, the wiring pattern image to be matched with the graphic data is sampled at an interval shorter than the smoothing span width with respect to the smoothed image from the predetermined direction of the wiring pattern image. Information is compressed.

That is, for example, when the wiring pattern data and the wiring pattern image are compared in the X direction,
When the magnification and shift amount in the X direction are obtained, Y
Since the magnification and the shift amount in the direction have values according to the magnification and the shift amount in the X direction, the comparison accuracy is maintained even if there is some image information to be compared in the Y direction.

Therefore, in this case, by sampling the wiring pattern image in the Y direction at predetermined intervals, the amount of information to be processed is reduced by compression, and pattern matching is performed at high speed between the graphic data and the gray scale image. .

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings. FIG. 1 shows the principle of the present invention and is a diagram for explaining the efficiency of processing by compressing an SEM image. The LSI is designed based on a predetermined pattern rule.

That is, in the pattern matching in the X direction, as described above, it is possible to omit the information in the Y direction, which information changes at intervals shorter than the wiring width of the LSI. Therefore, before performing pattern matching processing, smoothing processing for removing fine dot noise is performed on the obtained SEM image in advance with a span width similar to the wiring width in the Y direction, which is necessary for pattern matching. S / N of images without damaging important information
The ratio can be improved.

Further, by sampling the image at intervals equal to or less than the span width and reconstructing the image, the SEM
The image can be compressed and the information to be processed can be reduced. This embodiment will be described based on the above principle. FIG. 2 is a block diagram showing a schematic configuration of an apparatus for carrying out the present invention.

This apparatus is roughly classified into an E as a computer.
A WS (Engineering Work Station) 1, an SEM device 2 controlled by the EWS 1, and a data storage device 3 for storing CAD data. The SEM device 2 is an electron gun for irradiating an electron beam (hereinafter referred to as EB). 4, an XY stage 5 on which an LSI as a sample is mounted, and a detector 6
, Scan generator 7, frame memory 8,
It is composed of a stage controller 9.

Next, the operation will be described. FIG. 3 is a flowchart showing the preprocessing in the pattern matching of this embodiment. First, the operator gives CAD to EWS1.
When the information of the display area is specified, LS is set by EWS1.
The position of the XY stage 5 on which I is mounted and the magnification of the SEM image are manipulated to match the matching area of CAD data (hereinafter, C
The AD display area) is determined (step 1), and CA
An SEM image (see FIG. 1A) corresponding to the D display area is acquired (step 2).

Next, for example, when pattern matching is performed in the X direction, the lengths of "1" and "0" as binary images from the mask diagram (hereinafter, referred to as run length).
Is detected (step 3), and smoothing processing is applied to the Y direction of the SEM image with 50% of the run length as the span width (step 4). Then, the SEM image is sampled and compressed at an interval approximately the same as the span width.
An EM image (see FIG. 1B) is obtained (step 5).

The preprocessing in pattern matching is completed by the above processing. Here, with this pattern matching operation, E with an accuracy of ± 0.1 μm can be applied to a wiring with a width of 1 μm.
Consider B positioning. By the way, even when using an inexpensive XY stage 5 that does not perform feedback control, it is easy to control the position error of the stage to about ± 2 μm, and it is also possible to suppress the magnification error of the SEM image to about ± 2%. It's easy.

That is, the SE for which the positional error is ± 2 μm and the magnification error is ± 2% with respect to the designated CAD display area.
The acquisition of the M image is sufficiently possible with the existing technology. Therefore, for example, assuming that the SEM image is represented by 512 × 512 pixels and the wiring width is 1 μm, the pixel interval is 0.1 in order to perform EB positioning with an accuracy of ± 0.1 μm.
Must be μm, in this case 1 μm wiring ± 0.1
The field size of the SEM image for EB positioning with an accuracy of μm is approximately 50 μm □ (≈ (512 × 0.1) ² ).
Becomes

In this case, it is considered that a necessary amount of information can be maintained by thinning out about half of the wiring width as a compression standard. Therefore, 0.5 μm, which is half the wiring width in the direction of compression, is considered. If pixels are thinned out at intervals, the image data size can be compressed to 1/5 or less. By the way, as shown in FIG. 1A, 512 × 5
As shown in FIG. 1B, when the SEM image represented by 12 pixels is sampled at 50% of the wiring width, the image can be compressed to about 20 pixels in the Y direction in the figure.

FIG. 4 shows an example in which the image compression rate is further increased. In the above-described embodiment, the standard is to compress half of the wiring width, but in this embodiment, the maximum compression that can maintain the required amount of information is tried. That is, in the case of the SEM image as shown in FIG. 4A, the feature amount indicating a significant change in the image in the Y direction is the length of the minimum gap G, and therefore the minimum gap G serves as a reference for image compression. Become.

Therefore, in this embodiment, as shown in FIG. 4 (b), SE is used up to the limit where the characteristics of the gap G are preserved.
When the M image is compressed, the image can be compressed to about 11 pixels in the Y direction in the figure. As described above, in each of the embodiments described above, the SEM image is subjected to smoothing processing in a predetermined span width from a predetermined direction, and the SEM image subjected to the smoothing processing is more smoothed than the smoothing span width. By performing sampling at short intervals, the SEM image can be compressed without losing necessary information.

Therefore, since the amount of information of the SEM image to be processed is reduced, the processing speed can be increased.

[0028]

According to the present invention, the wiring pattern image to be matched with the graphic data is sampled by sampling the image subjected to the smoothing process from the predetermined direction of the wiring pattern image at intervals shorter than the smoothing span width. Information can be compressed.

Therefore, the amount of information to be processed can be reduced by the compression, and the pattern matching can be performed at high speed between the graphic data and the gray scale image.

[Brief description of drawings]

FIG. 1 is a principle diagram of a pattern matching method of the present invention.

FIG. 2 is a block diagram showing a schematic configuration of an apparatus for carrying out the present invention.

FIG. 3 is a flowchart for explaining the pattern matching method of the present embodiment.

FIG. 4 is a diagram for explaining a processing method of another embodiment.

FIG. 5 is a flowchart for explaining the operation of the conventional example.

FIG. 6 is a diagram showing an edge image extracted by a method of a conventional example.

[Explanation of symbols]

 1 EWS (computer) 2 SEM device 3 data storage device 4 electron gun 5 XY stage 6 detector 7 scan generator 8 frame memory 9 stage controller

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification code Office reference number FI technical display location H01L 21/66 A 8406-4M 21/68 G 8418-4M

Claims (3)

[Claims] 1. A pattern matching method for performing matching between wiring pattern data in a predetermined rectangular area as design data and a wiring pattern image on an XY plane on an actual device corresponding to the rectangular area, the wiring pattern comprising: A pattern matching method characterized in that sampling is performed in a predetermined width of an image with a predetermined width, and an image obtained by sampling is collated / compared with the wiring pattern data in a direction perpendicular to a direction in which the image is sampled and associated. 2. The wiring pattern image is subjected to smoothing processing for smoothing with a predetermined span width from a predetermined one of X and Y directions, and the smoothed image is subjected to a smoothing span width. The pattern matching method according to claim 1, wherein the sampling is performed with a short interval width. 3. The pattern matching method according to claim 2, wherein the span width of the smoothing is smaller than the minimum image width in the wiring pattern image in the smoothing direction in the smoothing process.