JPH04154110A - Mark for alignment and its detection - Google Patents

Abstract

PURPOSE:To accurately detect an alignment mark as a whole even when individual stripes are detected inaccurately by a method wherein edges of the stripes are detected from a dark and bright image, the arrangement of the edges are recognized and the mark is detected. CONSTITUTION:A grating shaped mark 1 is scanned; parts where a density is changed relatively highly are detected; the number of increasing and decreasing densities and their interval are investigated within a range of the width of the mark. Thereby, even when one stripe 3 is missed due to a noise or the like, the existence of the mark is found out because many stripes exist. When the number of stripes is counted and their interval is investigated, it is possible to discriminate whether it is a mark or not. When the number and the interval are correct, it is judged that the position of the mark has been grasped accurately. Its arrangement is recognized and compared; the mark can accurately be detected. Regarding the direction of gratings and the horizontal direction, a mark which has been turned by 90 deg. is arranged near the mark, and the same operation is executed. By this constitution, the recognition rate of the mark can be enhanced.

Description

DETAILED DESCRIPTION OF THE INVENTION <Industrial Application Field> The present invention relates to an alignment mark and a method for detecting the same. It is used to align and connect to the electrode portion of a printed circuit pattern.

<Conventional technology> Conventionally, alignment between an LSI chip and a glass substrate is performed using an image of the density of reflected light obtained by upwardly illuminating a square or round mark pattern for alignment provided on the chip and glass substrate. can be directly observed by humans or by CCD.
After capturing the image into image memory using a camera, a computer examines the density sound of each pixel to determine the presence or absence of a mark.

<Problems to be Solved by the Invention> By the way, in the method of detecting the mark and determining the alignment position by using the shading of the light projected at 70 degrees from the mark pattern for alignment, ζ Catfish buttock small embroidery cam tr fence 1 lacquer 8 lf + ki is necessary, but if the ultraviolet ray shielding film is completely passed through, the density difference will decrease and the mark pattern will be difficult to read with a computer. Therefore, when using such a filter, it is necessary to use the human eye to find the mark, or to first determine the position of the chip or glass substrate separately under favorable conditions, and then calculate the amount of movement of the stage. By managing it, you can align it. However, in this case, the accuracy of positioning is determined by the positioning accuracy of the stage and the deviation of the chip. There is also a method of mechanically retracting the filter only when detecting a mark, but this method complicates the apparatus, takes time to insert and remove the filter, and there is a risk that the filter may become misaligned.

Therefore, the present invention provides a noise-resistant alignment mark shape and a mark detection method that allows a computer to accurately read the mark position even from a grayscale image in which the density difference has been reduced due to the use of an ultraviolet light shielding filter, etc. The purpose is to

<Means for Solving All the Problems> The alignment mark of the present invention is an alignment mark formed on a chip component, and is characterized by being composed of a plurality of stripes.

Furthermore, the alignment mark detection method of the present invention is a method of irradiating light onto a chip component and detecting the alignment mark formed on a border component by a total of one machine from a gray scale image of the emitted light. Then, the edges of the stripes are detected by scanning the entire grayscale image in the direction across the stripes, the arrangement of the detected edges is recognized, and the alignment mark to be obtained from the recognized arrangement of edges and the alignment mark to be detected is It is characterized in that an alignment mark to be pressure-tested is detected by comparing it with an arrangement of edges.

<Operation> The alignment mark of the present invention is detected by scanning the mark in a direction across the stripe and detecting the edge of the stripe. At this time, since there are a plurality of stripes, a large number of edges can be detected compared to conventional marks in which only two edges can be detected. Then, the mark is detected by recognizing the arrangement of a large number of detected edges.

The alignment mark detection method of the present invention detects the edge of a stripe from a grayscale image, and thereby recognizes the edge arrangement and detects the mark. Even if each stripe is inaccurately detected when detecting a mark, the mark can be detected accurately as a whole. Also,
Since the image is scanned across the entire stripe, no stripes will be overlooked.

<Example> Next, examples of the present invention will be shown.

FIG. 2 is a configuration diagram of a synutem for recognizing alignment marks. On top of the LSI chip 21 is a galanus plate 22, on which a mark pattern for alignment is printed. In order to observe these with a COD camera 25, light from a light source 26 is illuminated using a half mirror 24.

A filter 23 for blocking ultraviolet rays is placed on the glass substrate. The image from the camera is input to the image processing device 27, where mark detection and position calculation are performed. L
The SI chip 21 and the glass substrate 22 are first subjected to rough mechanical alignment so that the pixel arrangement of the image memory has some degree of orthogonality to the sides of the mark pattern.

How the image captured in this manner is processed to detect marks will be explained in comparison with a conventional method.

FIG. 1 is a diagram illustrating alignment marks. Figure 1 (a) shows a lattice-shaped mark 1 that generates a lattice-shaped shading pattern according to the present invention, and Figure 1 (b) shows a square mark 4 that is commonly used in the past. The size is about 10 μm. Note that a density section obtained by importing these into the image memory and examining the shading of the pixels in a direction perpendicular to the grid is drawn below.

In conventional mark inspection algorithms, this density cross-section is examined from the edge, and first, the area where the density of neighboring pixels relatively changes higher than a certain threshold value is searched as the starting edge of the mark. Once it is found, it is searched for as the end edge of the partial mark whose density changes relatively low in the next nearby pixel. Then, it is determined that the mark is a mark by comparing the width between them with the width of the mark, which is known in advance. Then, the center point of the starting edge and ending edge of the mark is set as the center position of the mark in that direction. However, with this method, when the difference in density between the mark and the background decreases due to the insertion of a filter, the threshold value must be reduced, and the noise component cannot be ignored, causing areas inside the mark or in the background to be mistakenly detected as edges. Many pressure points appear, and some parts are judged to be marks, and real edges cannot be detected properly.

Therefore, it is necessary to perform more complicated image processing to select the correct mark and determine its position.

On the other hand, in the detection method according to the present invention when the mark according to the present invention is fully opened, it first similarly searches for a portion where the density changes relatively high, but once it is found, it is possible to determine the width of the mark from there. Find out how many times the concentration rises and falls within the range, and whether the intervals are equal.

By doing this, even if there is noise! Even if you miss 11 stripes 3, there are many stripes 3, so you can tell whether there is a mark from the remaining stripes 3, and by counting the number of stripes 3 or checking the interval tone, you can determine if it is a mark or another pattern. Can be distinguished. Furthermore, if the number and spacing are correct, it is assumed that the position of the mark has been accurately grasped, and the position in that direction can be determined.

That is, by recognizing and comparing the entire array, the pressure of the mark can be accurately and precisely detected. Incidentally, in the above embodiment, since there are linear stripes at equal intervals all around the circumference, it is easy to realize eight patterns, and it is also easy to form them on a chip.

By the way, with the above method, it is not possible to accurately determine the position of the mark in the grid direction and horizontal direction. This can be investigated by taking the concentration distribution in the direction perpendicular to the grid.

An example is shown in FIG.

The outer frame in FIG. 3 represents an image memory, into which an image of mark 2 consisting of a pair of grid patterns for alignment is input. The marks are arranged so that the orientation of the grid is 90 degrees. In order to find all marks 2, the computer scans the image memory in the order of ■■■■. Here, the direction of scanning is from left to right and top to bottom.First, nothing is found by scanning ■, so we move downwards and scan ■.As again, nothing is found, so we move further downwards. te ■ scan. Here, the vertical striped lattice is pressure-tested to determine the horizontal position, and based on this information, the vertical position can also be determined by scanning (■).

When performing the above scanning, since the mark 2 is composed of stripes 3, it is possible to reliably scan over the necessary stripes 3. For example, if a stripe section is constructed by arranging square blocks at regular intervals instead of stripes, the number of edges will increase, but when scanning, it may be possible to accidentally scan between blocks, making it difficult to accurately Arrays cannot be recognized.

The alignment mark is formed on the chip at the same time as the circuit pattern. The mark 2 consisting of a pair of lattice-like patterns rotated by 90 degrees with respect to each other is formed at two corners of the upper side of the chip, respectively. By forming such a marker, the inclination of the chip can be controlled more accurately.

<Effects of the Invention> According to the alignment mark or detection method of the present invention, it is easy to recognize the mark pattern using a computer even for a fake image with a small difference in shading, and the mark recognition rate can be increased with a conventional device configuration. I can do it.

[Brief explanation of drawings]

Fig. 1 is a diagram for explaining alignment marks, Fig. 2 is a system configuration diagram for recognizing alignment marks, and Fig. 3 is a diagram for explaining the entire mark consisting of a pair of grid patterns according to an embodiment of the present invention. be. 1... Grid-like mark 2... Mark 3 consisting of a pair of grid-like patterns...
stripe

Claims (1)

[Scope of Claims] 1. An alignment mark formed on a chip component, characterized in that it is composed of a plurality of stripes. 2. A method of irradiating light onto a chip component and detecting the alignment mark formed on the chip component using a computer from a grayscale image of the reflected light, the method comprising scanning the grayscale image in a direction across the stripe. an alignment mark to be detected by detecting the edges of the stripe, recognizing the detected edge arrangement, and comparing the recognized edge arrangement with the edge arrangement to be obtained from the alignment mark to be detected. A method for detecting an alignment mark, the method comprising: detecting an alignment mark.