JP2001210690A - Macroscopic inspecting apparatus for semiconductor wafer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To exactly detect macroscopic irregularities appearing in the growth stage of a semiconductor wafer. SOLUTION: A photographed original image of the entire area or a part of a semiconductor wafer is collected, the die of the wafer is adopted as an inspection region unit, and the image densities of mutually adjacent dies are compared to extract a macroscopic irregularity appearing as a nonuniformity of the density over a wide range of the wafer. The normalized correlation for the density comparison of the dies one with the other is used to eliminate the influence of the illumination nonuniformity.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a macro inspection of a semiconductor wafer for automatically detecting a macro abnormality which cannot be detected by a so-called micro inspection such as an abnormal film thickness of a semiconductor ware or an exposure phenomenon defect. Related to the device.

[0002]

2. Description of the Related Art In the macro inspection of semiconductor ware, the loading of wafers from a cassette and the handling of wafers have been automated for the purpose of labor saving and maintaining a clean level. At present, there are few technologies related to automation.

[0003] To realize automation of macro inspection, as disclosed in Japanese Patent Application Laid-Open No. H10-144747, a reference region composed of one or a plurality of dies on a wafer is utilized by utilizing the repeatability of the dies of a semiconductor wafer. There is a macro inspection method in which is compared with another inspection area by using template matching as a reference template, and a place having a low degree of coincidence is output as a defect.

[0004]

However, the conventional macro inspection method has a problem that illumination unevenness and unevenness due to macro abnormality cannot be accurately separated.

[0005] For example, as shown in FIG.
If the profile of the density histogram of the inspection object is accompanied by a shape change due to macro abnormality (b), this can be detected as a defect. When shifting to the left or right with respect to (a) (c)
In this case, this is erroneously detected as a macro error instead of a non-defective product.

The present invention has been made in view of such circumstances, and has as its object to provide a semiconductor wafer macro inspection apparatus capable of always accurately detecting a macro abnormality of a semiconductor wafer.

[0007]

According to the present invention, there is provided a macro inspection apparatus for a semiconductor wafer, which collects an original image of an entire area or a part of the semiconductor wafer, and divides the die on the wafer into inspection area units. Means, an inspection area associating means for associating inspection areas adjacent to each other, setting an association of dies to be compared with each other, and a density comparing means using a normalized correlation in the density comparison between the mutually associated inspection areas. A determination unit that determines whether the normalized correlation value is below a predetermined threshold, and configured to output an area where the normalized correlation value is below a predetermined threshold as an abnormal point. Can be

In the semiconductor wafer macro inspection apparatus according to the present invention, the calculation of the normalized correlation value is performed not only on the given area but also on the area shifted by one pixel in the vicinity of the periphery 8.
The maximum value (the value in the region with the highest matching degree) may be selected from a total of nine normalized correlation values, and the determination may be made using this as a representative value.

According to the semiconductor wafer macro inspection apparatus of the present invention, the whole or a part of the original image of the semiconductor wafer is collected using an area sensor or a line sensor, and the dies on the wafer are used as inspection area units. Since the density comparison is performed between the dies adjacent to each other after the area is divided, it is possible to extract macro abnormalities that occur as density unevenness over a wide range on the wafer. Further, the use of the normalized correlation in the density comparison between the dies makes it possible to perform a macro abnormality inspection that is not easily affected by illumination unevenness.

Further, in the semiconductor wafer macro inspection apparatus according to the present invention, when calculating the normalized correlation, the calculation is performed not only on the given area but also on the area shifted by one pixel to the vicinity of the periphery 8. Then, the maximum value (the value in the region with the highest degree of coincidence) is selected from a total of nine normalized correlation values, and this is adopted as a representative value, so that the quantization error of the horizontal resolution of the image can be obtained. It is possible to perform more accurate macro abnormality inspection with less erroneous detection, which is hardly affected.

[0011]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a schematic configuration of an embodiment of the present invention.

The image input means 1 is an area sensor, a line sensor, or the like, and captures an image of the whole or a part of a semiconductor wafer to be inspected. The picked-up original image is input to the image processing device 2 and is an M × N 2 as shown in FIG.
It is stored in an image frame memory or the like as dimensional array density data.

The image processing apparatus 2 includes an inspection area dividing unit 2
1. It is composed of an inspection area associating unit 22, a gray scale comparing unit 23, a determining unit 24, and the like, and is configured to detect a macro abnormality of a semiconductor wafer by a process described later.

Next, the present embodiment will be described more specifically.

When an image of a semiconductor wafer is taken, as shown in FIG. 3, macro abnormalities E may be observed as shading unevenness along with dies D,... D regularly arranged in the horizontal and vertical directions. The macro abnormality described here is a defect caused by a film thickness abnormality in a wafer manufacturing process, an exposure and development defect, and the like, and is a type of defect that cannot be found by a so-called micro inspection using a microscope. It is assumed that macro abnormality E is detected. The main processing flow of the macro abnormality detection is as shown in the processing flow of FIG.

First, an original image taken from the image input means 1 is fetched (S1), and a die on a semiconductor wafer is divided into inspection areas in units of inspection areas (S2). Next, the inspection areas adjacent to each other are associated with each other, and the association of the dies to be compared with each other is set (S3). Then, the grayscale comparison between the inspection areas mutually associated, that is, the normalized correlation value between the associated inspection areas is performed. Is calculated (S4), and it is determined whether or not the calculated value of the normalized correlation value is below the threshold value (S5). If the calculated value of the normalized correlation value is below the threshold value, the area is regarded as an abnormal point. Output (S6). Then, each processing of the above steps S4 and S5 (including the processing of step S6)
Is performed for each inspection area of the semiconductor wafer (S7),
When the inspection of all the inspection areas is completed, the process ends.

Next, a description will be given of a method of setting an inspection area for a captured image as shown in FIG.

First, the inspection target is targeted only to the dies which completely enter the image, and the association of the inspection areas to be compared with each other is set. It is desirable that the association rules are adjacent to each other as much as possible in order to minimize the influence of illumination unevenness. As an example, a case of a main body in the horizontal direction as shown in FIG. 4A and a case of a main body in the vertical direction as shown in FIG. In FIGS. 4A and 4B, the arrows indicate the association, and the inspection areas at both ends of the arrows are to be compared with each other.

Next, a description will be given of a method for comparing gray levels between inspection areas associated with each other, that is, a normalized correlation.

First, as shown in FIG. 5, the inspection regions Rn, R
An area narrowed by one pixel from n + 1 is set as a basic normalized correlation operation target area. An array vector of the pixel grayscale values included in the calculation target area in the inspection area Rn is Pn, and similarly, an array vector for the inspection area Rn + 1 is Pn + 1.
Assuming that the normalized correlation value between these arrays is NR, the normalized correlation is calculated by the following equation (1).

[0022]

(Equation 1)

The left side of this equation (1): NR means the cosine between two array vectors, and the larger the degree of coincidence, the larger the value (1 at maximum).

Hereinafter, the normalized correlation calculation will be described in more detail.

First, when the two inspection areas Rn and Rn + 1 have a pixel array of 8 × 5 = 40 pixels as shown in FIG. 6, for example, the normalized correlation with respect to the area surrounded by the thick line in FIG. The operation is specifically as follows.

First, the gray value of the 40 pixels constituting Rn is represented by Ii (i = 1, 2, 3,..., 40). Similarly, the gray value of the 40 pixels constituting Rn + 1 is represented by Ji (i = 1, 2). , 3... 40), then 6 × 3 =
Array vectors Pn, Pn + using 18 pixel gray values
1 is

[0027]

(Equation 2)

## EQU1 ##

In the case of this example, the left side NR of the equation (1)
Is the cosine (COS θ) of the angle θ between two array vectors Pn and Pn + 1 in an 18-dimensional space. Therefore, when the grayscale patterns of the two normalized correlation calculation target regions are very similar, θ ≒ 0, that is, NR ≒ 1. Further, when the density pattern of the entire area is multiplied by a constant due to illumination unevenness or the like, since the size of the array vector is simply multiplied by a constant, θ is unchanged, that is, is not affected by the illumination unevenness.

Here, in the case where the normalized correlation calculation target area is automatically generated based on the die size and the pixel size, the size and the position of the target area are determined if the die size is not a constant multiple of the pixel size. It may be shifted as shown in FIG. In this state, if the normalized correlation value is calculated as it is, if the contrast of the light and shade pattern in the target area is large, the degree of coincidence will be low, and there is a possibility that the pattern will be erroneously inspected as defective.

To prevent this, for the target area on the Rn + 1 side, a normalized correlation value is calculated for eight neighboring areas, and the largest (highest degree of coincidence) is selected from a total of nine normalized correlation values. ) The method of using the case as a representative value for the judgment is adopted. Specifically, in the case of the example shown in FIG. 7, as shown in FIG.
) Has the highest normalized correlation value, which is used as a representative value.

When the above-described quantization error prevention processing is performed, the vicinity of the periphery 8 of Rn + 1 is taken as shown in FIG.

After performing the above-described normalized correlation, as shown in FIG. 2, a preset threshold value is compared with the normalized correlation value, and if the normalized correlation value falls below the threshold value,
The two inspection areas Rn and Rn + 1 are output as areas containing macro abnormalities.

By performing the above-described processing for all the associations in the image, it is possible to detect a macro error. For example, when the association shown in FIG. 4A is used for an image as shown in FIG. 3, a macro abnormality is detected in a form as shown in FIG.

[0035]

As described above, according to the semiconductor wafer macro inspection apparatus of the present invention, the whole or a part of the original image of the semiconductor wafer is sampled using the area sensor, the line sensor, etc. Since the density comparison is performed between the dies adjacent to each other after the area division is performed with the die as a unit of the inspection area, it is possible to extract the macro abnormality that occurs as the density unevenness over a wide range on the wafer. . Moreover, since the normalized correlation is used in the density comparison between the dies, it is possible to perform an automatic macro abnormality inspection that is not easily affected by illumination unevenness.

Further, in the semiconductor wafer macro inspection apparatus of the present invention, when calculating the normalized correlation, it is not necessary to calculate only for a given area, but to calculate 1 in the vicinity of the periphery 8.
The calculation is also performed for the pixel-shifted region, and the maximum value (the value in the region with the highest degree of coincidence) is selected from a total of nine normalized correlation values, and this is used as a representative value for determination. A more accurate macro abnormality inspection with less erroneous detection, which is less affected by the quantization error of the horizontal resolution of the image, can be performed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of an embodiment of the present invention.

FIG. 2 is a diagram showing a processing flow of macro abnormality detection.

FIG. 3 is a diagram illustrating an example of a macro target image.

FIG. 4 is a diagram illustrating an example of an inspection area association process.

FIG. 5 is a diagram illustrating a calculation target region of a normalization region.

FIG. 6 is a diagram illustrating a specific example of a calculation target area of a normalized area.

FIG. 7 is an explanatory diagram of a quantization error generated at the time of a normalized correlation operation.

FIG. 8 is an explanatory diagram of a process for preventing a quantization error.

FIG. 9 is a diagram showing the outline of how to take the vicinity of the periphery 8;

FIG. 10 is a diagram illustrating an example of macro abnormality detection.

FIG. 11 is an explanatory diagram of a problem of a conventional macro inspection method.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image input means 2 Image processing apparatus 21 Inspection area division means 22 Inspection area association means 23 Shade comparison means 24 Judgment means D Die E Macro abnormality

 ──────────────────────────────────────────────────続 き Continuing on the front page F term (reference) 2F065 AA49 BB02 CC19 FF42 JJ02 JJ03 JJ25 JJ26 QQ00 QQ24 QQ25 QQ26 QQ41 QQ42 RR01 2G051 AA51 AB20 CA03 CA04 EA14 EA20 EB01 EB02 EC03 EC06 DJ01 A14 DJ01

Claims (2)

[Claims] 1. An inspection area dividing means for collecting an entire or a part of an original image of a semiconductor wafer and dividing a die on the wafer into inspection area units, and associating adjacent inspection areas with each other and comparing them with each other. Inspection area association means for setting association of mutually associated dies, density comparison means using normalized correlation in density comparison between mutually associated inspection areas, and whether or not the normalized correlation value falls below a predetermined threshold value A macro-inspecting apparatus for a semiconductor wafer, characterized in that it is provided with a judging means for judging a value, and is configured to output a region where the normalized correlation value is smaller than a predetermined threshold value as an abnormal point. 2. A normalization correlation operation is performed on a region shifted by one pixel in the vicinity of the periphery 8 in addition to a given region, and a maximum value is selected from a total of nine normalized correlation values.
2. The semiconductor wafer macro inspection apparatus according to claim 1, wherein the apparatus is configured to make a determination using the value as a representative value.