JP2796430B2 - Pattern defect detection method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to a method of detecting defects by comparing patterns of, for example, LSI wafers and TFTs, and relates to a method of high-accuracy alignment and alignment marks. The present invention relates to a pattern defect detection method for detecting a defect by performing high-accuracy position detection and an apparatus therefor.

[Conventional technology]

As described in Japanese Patent Application Laid-Open No. 57-196377, a conventional pattern recognition method detects a target pattern, stores the detected pattern, and detects a pattern stored immediately before. A pattern defect is recognized by aligning a pattern with a pixel unit and extracting and comparing an error between the two aligned patterns.

The recognition target is a pattern of a semiconductor wafer such as a memory LSI, a pattern of a TFT (Thin Film Transister), a pattern of a printed wiring board, or the like, as illustrated in FIGS. 2 (a), 2 (b) and 2 (c). The pattern is a pattern of a ceramic substrate or a pattern of a mask or a reticle used in a process of manufacturing the pattern. Here, a pattern of a semiconductor wafer will be described as an example, but the same holds true for other patterns. The semiconductor wafer pattern has several tens of chips, which are finally separated into individual products, mounted on a single wafer, and they have the same pattern. The principle of recognizing such a pattern defect will be described with reference to FIGS. 2 (a) to 2 (c).

2 (a) to 2 (c) are diagrams for explaining the principle of a conventional general pattern comparison method. FIG. 2 (a) is a storage pattern, FIG. 2 (b) is a detection pattern, and FIG. ) Is a pattern difference. Paying attention to the fact that each chip has exactly the same pattern, the pattern of FIG. 2A is detected and stored, and another pattern which should be the same as that of FIG. Next, the two patterns are detected and aligned in pixel units, and the error between the two aligned patterns in FIG. 2C is extracted and compared. When there is no defect in any of the patterns, there is almost no difference between the patterns. However, when there is a defect in any of the patterns, for example, the detection pattern in FIG. 2B, FIG. Since there is a difference in the pattern at the defective portion as shown in (2), the pattern defect can be recognized by detecting a place where an error occurs by comparing the patterns. Note that if there is a difference, it can be said that any of the patterns has a defect, but it cannot be determined which pattern has a defect.

[Problems to be solved by the invention]

In the above prior art, since an image is digitized and input, only information on sampling points is obtained, and occurrence of a sampling error is inevitable. However, there is a problem that it is difficult to recognize a small defect due to the influence of the sampling error. This will be described with reference to FIGS. 3 (a) to 3 (c).

FIGS. 3 (a) to 3 (c) show X- in FIGS. 2 (a) to 2 (c).
FIG. 3 (a) is a waveform diagram of a detected signal of the storage pattern of FIG. 2 (a), and FIG. 3 (b) is a waveform diagram of a pattern on the X 'line.
Is a detection signal waveform diagram of the detection pattern of FIG.
FIG. 3 (c) is a detection signal waveform diagram when there is no sampling error in the detection pattern of FIG. 2 (b), and FIG.
3 (a) and 3 (b) show the difference signal waveforms when there is a sampling error, and FIG. 3 (e) shows the difference signal between FIGS. 3 (a) and 3 (c) when there is no sampling error. FIG. 3 is a waveform diagram, in which “・” indicates a detection signal at a sampling point. As shown in FIGS. 3 (b) and 3 (c), originally the same pattern cannot be set at the same point, so that the detected waveform at the sampling point becomes different, and the detected waveform and the detected pattern have an error. And this error is called a sampling error. If there is no sampling error in the detection pattern to be compared, the defect difference signal is sufficiently larger than the normal difference signal as shown in FIG.
If there is a sampling error in the detection pattern, the defect difference signal is almost the same as the normal difference signal as shown in FIG. 3 (d), making it difficult to recognize the defect. If the defect size to be recognized is sufficiently large with respect to the pixel size at the time of detection, the sampling error and the defect can be identified by using the difference in the area of the place with a large difference. Is about the same as that due to sampling errors, and it is difficult to identify defects.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a pattern defect detection method and apparatus capable of reducing the influence of sampling errors by setting the accuracy of pattern alignment to the detection pixel size or less.

[Means for solving the problem]

In order to achieve the above object, a pattern defect detection method according to the present invention detects a pattern formed on a substrate, generates image data, creates comparative image data using the image data, and generates image data. A method of correcting the deviation between the image data and the comparison image data based on the comparison image data with an accuracy equal to or less than the pixel, and comparing the image data with the corrected deviation with the comparison image data to detect a pattern defect. is there.

Further, according to the present invention, the pattern defect detection method detects a pattern formed on a substrate to generate image data,
Using this image data, comparative image data is created, an amount to be shifted based on the image data and the comparative image data is determined with an accuracy of a pixel or less, and the image data and the comparative image data are determined based on the determined amount to be shifted. Are compared with each other with a precision equal to or less than a pixel, and the image data shifted from each other is compared with comparative image data to detect a pattern defect.

Further, the pattern defect detection method according to the present invention includes an image data generating means for detecting a pattern formed on a substrate to generate image data, and comparing the comparison image data using the image data generated by the image data generating means. Comparison image data creation means to be created, shift amount calculation means for obtaining an amount to be shifted based on the image data and the comparison image data with an accuracy of a pixel or less, and image data based on the shift amount obtained by the shift amount calculation means. And a comparison image data, and a defect detection unit for comparing the image data shifted by the correction unit with the comparison image data to detect a pattern defect.

In order to perform positioning with an accuracy of this pixel unit or less, for example, the following least square method is used. F two patterns
When (x, y) and g (x, y), ε in equation (1)
2 (dx, dy) is aligned to the position (dx0, dy0) that minimizes the pixel unit, and the difference between the pattern of the detected image and the stored image is minimized between the values 0 and 1 for both x and y coordinates. Position (δx0,
δy0).

δ2 (dx, dy) = ε (dx, dy) + ε (dx + 1, dy) + ε (dx, dy + 1) + ε (dx + 1, dy + 1) (1) ε (dx, dy) = ΣΣ | f (x, y ) -G (x + dx, y + dy) |
(2) where x and y are the coordinates of the pattern in pixel units, dx and dy are the alignment amounts in pixel units of the two patterns, and dx0 and dy0 are ε.
Pixel-based registration amounts dx, dy, δx, δy that minimize 2.
Indicates the amount of alignment below the pixel, and δx0, δy0 indicate the amount of alignment δx, δy, ΣΣ below the pixel that minimizes the pattern difference, and indicates the sum of the x and y coordinates of the range to be aligned.

In order to perform pixel-by-pixel alignment, g1 obtained by shifting g shown in the following equation (3) is used.

g1 (x, y) = g (x + dx, y + dy) (3) An intermediate value between pixels is defined by equations (4) and (5).

 fd (x, δx, y, δy) = f (x, y) + δx * (f (x + 1, y) -f (x, y)) + δy * (f (x, y + 1) -f (x, y) (4) g1d (x, δx, y, δy) = g1 (x, y) + δx * (g1 (x−1, y) −g1 (x, y)) + δy * (g1 (x, y−1) ) −g1 (x, y)) (5) The square error can be defined by equation (6).

εd (δx, δy) = ΣΣ (fd (x, δx, y, δy) −g1d (x, δx, y, δy)) ** 2 (6) Equation (6) is partially differentiated by δx, δy. By setting this to 0, the equations (7) and (8) are obtained.

Here, Co = f (x, y) −g1 (x, y) Ci = (f (x + 1, y) −f (x, y)) − (g1 (x−1,
y) -g1 (x, y)) (9) Cj = (f (x, y + 1) -f (x, y))-(g1 (x, y-
1) -g1 (x, y)) Patterns f2 and g2 after alignment are calculated from the alignment amounts δx0 and δy0 below the pixel by the following equations (10) and (11).

f2 (x, y) = fd (x, δx0, y, δy0) (10) f2 (x, y) = g1d (x, δx0, y, δy0) (11) [Action] The action of this pattern recognition method FIG. 4 (a) to (d)
This will be described with reference to FIG.

4 (a) to 4 (d) are waveform diagrams of an operation example of sub-pixel positioning for positioning at a pitch equal to or less than the pixel of the pattern of FIG. 2 according to the present invention, and FIG. FIG. 4 (b) is a defect-free detected waveform, and FIG. 4 (c).
Fig. 4 (d) shows a difference waveform after sub-pixel positioning, in which a simple difference waveform is obtained by performing only pixel-by-pixel positioning. Fifth
The figures are the numerical table diagrams of FIGS. 4 (a) to 4 (d). The stored waveforms at the sampling positions 0 to 15, the detected waveforms, the simple difference waveforms, and the f2, g2, | f2-g2 | Numerical values are shown. The stored waveform and the detected waveform are, for example, as shown in FIG.
4B and FIG. 5, the detected waveform in FIG. 4B is a waveform obtained by averaging two pixels before and after the stored waveform in FIG. This is equivalent to the shifted waveform.

By applying the least squares method to these waveforms, δx0 is obtained, and when it is actually calculated, δx0 = 0.2. From this value, (10),
The patterns f2 and g2 after the alignment are obtained by using the equation (11). At this time, the difference signal waveforms with and without sub-pixel alignment are shown in FIGS.
As shown in the figure, the residual is reduced by half. As a result, the value of the pattern difference due to the sampling error becomes sufficiently smaller than the value of the defect, so that the defect can be easily identified.

〔Example〕

Hereinafter, a first embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a configuration diagram of an LSI wafer pattern defect detection apparatus. This pattern defect detection device comprises a XY stage 2 for scanning the wafer 1, a light source 3 for illuminating the wafer, an illumination optical system 4, an objective lens 5 for detecting an optical image of the illuminated wafer, and a one-dimensional image sensor 6. An image input unit 9 including an A / D converter 7 and an image memory unit 8 for digitizing and storing a signal of the one-dimensional image sensor 6, and a detected image input to the image input unit 9.
An image extracting unit 12 for extracting the reference image 10 and the comparative image 11 from the image memory unit 8, and calculating the difference between the detected image 10 and the comparative image 11 by using the expression (2), and moving the comparative image as in the expression (3). Image unit matching unit 13
And a sub-pixel matching unit that calculates an alignment amount δx0, δy0 of a pixel or less represented by the equations (7) and (8) from the image 14 after the pixel-unit position correction by the pixel-unit matching unit 13 and the detected image 10 Based on the position adjustment amount from the sub-pixel matching unit 15 and the position adjustment unit 16 that performs the position correction expressed by the equations (10) and (11), the difference image 17 and the difference image 17 of the image 17 after the alignment are obtained. An image processing unit 20 including a defect determination unit 19 that binarizes and extracts various feature amounts at a location where a difference exists to determine a defect, and control of the XY stage 2 and defect information output from the image processing unit 20 Control unit composed of computers that store and display data and manage the entire sequence
Consists of 21.

The operation of detecting a pattern defect with this configuration will now be described. First, after initialization of each unit by a command from the overall control unit 21, the pattern of the wafer 1 illuminated by the light source 3 and the illumination optical system 4 is synchronized with the scanning of the XY stage by the one-dimensional image sensor via the objective lens 5. A two-dimensional pattern is detected by photoelectric conversion at 6 and an A / D converter 7
The obtained detection image is stored in the image memory unit 8 as the two-dimensional detection image 10 digitized in the step (1).

The image extracting section 12 extracts the comparative image 11 by referring to a fixed address in the image memory section 8. here,
The operation of the pixel unit matching unit 13 will be described with reference to FIG. FIG. 6 is an explanatory view of the operation principle of the matching unit 13 in FIG. Based on the detected image 10 and the comparative image 11, the comparative image
± δ pixels of the permissible displacement in the X and ΔY directions (this embodiment will be described with δ = 1, but the value is determined by the dimensional accuracy of the detection target and the positioning accuracy of the defect detection device, and the necessary values are set appropriately. The difference between the image of the detected image 10 and the image of the comparative image 11 when the image is shifted by only
Dx0, dy0 (ΔX in FIG. 6) that minimizes 2 (S1 in FIG. 6)
= -1, ΔY = 0), the position of the comparative image 11 is corrected using the equation (3), and the image 14 after the image unit position correction is output.

Sub-pixel matching unit 15 is a pixel unit matching unit
Image 14 and detected image after pixel unit position correction from 13
From 10, the amounts of alignment δx0 and δyp below the pixel represented by the equations (7) and (8) are calculated. Based on the amount of alignment from the sub-pixel matching unit 15 (10),
The position correction represented by the equation (11) is performed. The difference image extracting unit 18 extracts the difference image 17 from the image after the position correction by the following equation (12).

S (i, j) = | f2 (i, j) −g2 (i, j) | (12) The defect determination unit 19 binarizes the difference image 17 with a threshold value Vth for defect determination, and determines a location where the difference exists. Defects are determined by extracting various features such as the area, width, and projection length of the image.

According to this embodiment, since the two images of the detected image and the stored image are moved by the same amount in the opposite direction to create an alignment image of pixels or less, an image smoothing effect (for example, δx0 = F2 (x, y) = 0.5, δy0 = 0
(Equivalent to applying an average filter at f (x + 1, y) + f (x, y)) / 2), the same effect is obtained for the two patterns, and the error of the resulting difference image can be minimized. There is.

A first modification of the present embodiment is that the equations (10) and (11) are
Equations 3) and (14) are used.

f2 (x, y) = f (x, y) (13) g2 (x, y) = f (x, y) + δx * (fx−1, y) −f (x, y)) + 2 * δy * (F (x, y-1) -f (x, y)) if0.0 ≦ δx <0.25,0.0 ≦ δy <0.25 f (x + 1, y) + (1-δx) * (f (x, y) −f (x−1, y) + 2 * δy * (f (x−1, y−1) −f (x−1,
y)) if0.25 ≦ δx <0.5,0,0 ≦ δy <0.25 f (x, y + 1) + δx * (f (x−1, y−1) −f (x, y−1)) + (1 −2 * δy) * (f (x, y) −f (x, y−
1)) if = 0.0 ≦ δx <0.25, 0.25 ≦ δy <0.5 f (x + 1, y + 1) + (1−δx) * (f (x, y−
1) −f (x−1, y−1)) + (1-2 * δy) * (f (x−1, y) −f (x−1, y−1)) if0.25 ≦ δx < 0.5, 0.25 ≦ δy <0.5 (14) According to this modification, when δx and δy take values close to 0.5, the values of f2 and g2 can be made more continuous. That is, for the data example of g shown in FIG. 7, the case of dx = 0, δx = 0.49 and the case of dx
FIG. 7 shows the value of g2 in the case where this modification is performed when = 1 and δx = 0.01 and when this deformation is not performed. When the main deformation is not performed, the values are greatly different, but when the main deformation is performed, the values are substantially the same.

Further, a second modification of the present embodiment is that the XY stage 10 is step-moved instead of detecting a two-dimensional pattern by photoelectrically converting the one-dimensional image sensor 14 in synchronization with scanning of the XY stage 10. A two-dimensional pattern is detected by photoelectric conversion with a TV camera. Alternatively, instead of the one-dimensional image sensor 14, various types of sensors such as using a point type sensor such as a photomultiplier and a scanning mechanism can be used. Further, the image processing unit may not be entirely hardware, but may be configured by hardware + software.

A third modification of the present embodiment is to calculate the difference between the detected image and the stored image by equation (2) and output the difference between the images corresponding to the respective shift amounts as the matching value, instead of the detected image and the comparison image. Then, the edges are extracted by applying filters to the filter images, and the difference between the images is calculated by the equation (2) with respect to the filtered image, and f2 and g2 are calculated with respect to the image before being filtered. Alternatively, the detected image and the comparison image are each subjected to filtering to be binarized, and the difference between the filtered binary images is calculated by equation (2), and f2 and g2 are calculated for the image before being filtered. According to this modification, since the filter is used, there is an effect that it is hardly affected by a difference between unnecessary information of the detected image and the comparative image.

The fourth modification of this embodiment is (1), (4),
Equations (5) and (9) are replaced by (15-1), (15-4), (15-
5) Replace with (15-9).

ε2 (dx, dy) = ΣΣε (dx + nx, dy + ny) (15-1) fd (x, δx, y, δy) = f (x, y) + δx * (f (x + n, y) −f (x, y )) + Δy * (f (x, y + n) −f (x, y)) (15-4) g1d (x, δx, y, δy) = g1 (x, y) + δx * (g1 (x−n, y) −g1 (x, y) + δy * (g1 (x, y−n) g1 (x, y)) (15-5) Co = f (x, y) −g1 (x, y) Ci = ( f (x + n, y) -f (x, y))-(g1 (x-n, y) -g1 (x, y)) (15-9) Cj = (f (x, y + n) -f (x , y))-(g1 (x, yn)-g1 (x, y)) where n is the pitch of the calculation of the intermediate value between pixels, and n-1, 2, 3,. ΣΣ in equation (1) means the sum of nx and ny from 0 to n, ΣΣ in equations (6) and (7) indicates x, ny in the range to be aligned.
Shows the sum of all y coordinates or the sum of every n.

(1) The pixel is aligned to (dx, dy) which minimizes ε2 (dx, dy) in equation (1), and the difference between the two patterns between 0 and n in both x and y coordinates is minimized. The position ((δx, δy)) may be used.

According to this modification, there is a feature that the range to be aligned and placed in advance by the pixel unit matching unit may be rough.

A fifth modification of the present embodiment is a pixel unit matching unit.
Do not use 13. Depending on the target condition, the position (δx, δy) that minimizes ε2 (dx, dy) in equation (1) is between 0 and 1 for both x and y coordinates without performing pixel unit matching. In this case, it is not necessary to perform positioning in pixel units. In particular, when the fourth modification of the present embodiment is performed, there is a high possibility that it becomes unnecessary. According to this modification, there is a feature that the configuration is simple.

In a sixth modification of the present embodiment, the same method as the sub-pixel matching unit 15 is used for the matching unit 13 for each pixel. At this time, a third modification and a fourth modification are generally added to the matching unit 13 for each pixel. In other words, after the stored image g is shifted by n pixels according to the equation (16),
The calculation is performed at the pitch n of the pixel calculation with the average value filter of equation (17) applied. Equations (10) and (11) are obtained from equation (1).
Use 8) and (19). According to this modification, there are only two stages of sub-pixel matching, so that there are two sets of circuits having the same configuration when implemented in hardware, which is an efficient feature.

 gs (x, y) = g (x + n, y + n) (16) fo (x, y) = {f (x + i, y + j) go (x, y) = {gs (x + i, y + j) (17)} is i, j = Sum of -m to m.

f2 (x, y) = f (x, δx0, y, δy0) (18) g2 (x, y) = gs (x + 1, δx0, y + 1, δy0) (19) A seventh modification of the present embodiment If the equations (4) and (5) can be determined analytically or numerically by using arbitrary equations instead of linear equations, these equations can be used. According to this modification, an arbitrary expression can be used, and therefore, there is a feature of high versatility.

Next, a second embodiment of the present invention will be described with reference to FIG. FIG. 8 is a block diagram of a pattern recognition device for detecting a position error of a pattern. This pattern recognition device is a wafer 1
XY stage 2 for positioning wafer and light source 3 for illuminating wafer
A detection unit including an illumination optical system 4, an objective lens 5 for detecting an optical image of the illuminated wafer, and a TV camera 22, and a TV camera
A / D converter 7 for digitizing and storing 22 signals
And an image input unit 9 composed of an image memory unit 8, and a detected image 10 input to the image input unit and a stored image 11 in the image memory, and an alignment amount of a pixel or less represented by the equations (7) and (8). an image processing unit 20 including a sub-pixel matching unit 15 for calculating δx0 and δy0, and a control and image processing unit for the XY stage 2
It comprises an overall control unit 21 composed of a computer for storing and displaying information output from 20 and managing the overall sequence.

The operation of detecting a pattern error with this configuration will now be described. First, after initializing each unit according to a command from the overall control unit 21, the XY stage 2 is driven and positioned, and the pattern of the wafer 1 illuminated by the light source 3 and the illumination optical system 4 is passed through the objective lens 5 to the TV camera 22. Then, a two-dimensional pattern is detected by photoelectric conversion in step (2), and is converted into a two-dimensional detection image 10 digitized by the A / D converter 7. The obtained detection image is stored in the image memory unit 8. After the stored image 11 in the image memory 8 detected and stored immediately before is shifted by one pixel by the equation (20), the position errors δx0 and δy are calculated by the equations (7) and (8).
Calculate 0. The position error is passed to the overall control unit.

g1 (x, y) = g (x + 1, y + 1) (20) According to the present invention, a pattern position error is output using pattern alignment information as it is, so that a highly accurate position error detection is possible. There is.

In the first modification of the present invention, there is a method of masking an image when calculating δx0 and δy0 from the detected image 10 and the stored image 11 by the equations (7) and (8). According to this modification, the position errors Δx0 and Δy0 can be calculated from only the desired information, which is a feature of high accuracy.

In a second modification of the present invention, alignment is performed by feeding back positional error information used for detecting an alignment mark. According to this modification, there is a feature that highly accurate alignment can be performed.

A third variant of the invention is used for angle detection. The angle is calculated based on error information of two or more positions. According to this modification, there is a feature that highly accurate angle detection can be performed.

〔The invention's effect〕

According to the present invention, it is possible to reduce the influence of sampling errors by setting the pattern alignment accuracy to be equal to or smaller than the detection pixel size, and to facilitate detection of a defect having a size similar to the pixel size.

[Brief description of the drawings]

FIG. 1 is a block diagram of one embodiment of the present invention, and FIG.
3 (a) to 3 (c) are explanatory diagrams of the principle of a pattern defect detection method using a conventional general pattern comparison method, FIGS. 3 (a) to 3 (e) are pattern waveform diagrams of FIGS. 2 (a) to 2 (c), and FIG. Figure (a)
4 (d) are waveform diagrams of an example of the sub-pixel matching operation of the pattern of FIG. 2 according to the present invention, and FIG.
FIG. 6 is an explanatory diagram of the numerical value of (d), FIG. 6 is an explanatory diagram of the operation of the pixel unit matching unit of FIG. 1, FIG. 7 is an explanatory diagram of a modification of the first embodiment, and FIG. FIG. 3 is a block diagram of an apparatus showing an embodiment of the present invention. 1 ... wafer, 2 ... XY stage, 3 ... light source, 4 ...
Illumination optical system, 5 ... objective lens, 6 ... one-dimensional image sensor, 7 ... A / D converter, 8 ... image memory unit, 9 ...
… Image input unit, 10… Detected image, 11… Comparative image, 12…
… Image extraction unit, 13… matching unit, 14… image after pixel unit position correction, 15… sub-pixel matching unit, 16… alignment unit, 17… difference image, 18… difference image extraction Unit, 19 ... Defect determination unit, 20 ... Image processing unit, 21
…… Overall control unit, 22 …… Sub-pixel operation unit, 23 …… TV
camera.

──────────────────────────────────────────────────の Continuation of the front page (51) Int.Cl. ⁶ Identification code FI H01L 21/66 G06F 15/62 405C (72) Inventor Takaki Makihira 292 Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture Production technology of Hitachi, Ltd. (56) References JP-A-57-196377 (JP, A) JP-A-61-176807 (JP, A) JP-A-62-140009 (JP, A) JP-A-59-157505 (JP, A) JP-A-63-88682 (JP, A) JP-A-3-177040 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) G01B 11/00-11/30 G01N 21 / 84-21/91 G06T 1/00-9/20 H01L 21/64-21/66

Claims (9)

(57) [Claims] An image data is generated by detecting a pattern formed on a substrate, comparative image data is created using the image data, and the image data is generated based on the image data and the comparative image data. Pattern defect detection, wherein a shift between the image data and the comparison image data is corrected with an accuracy of a pixel or less, and the image data corrected for the shift is compared with the comparison image data to detect a pattern defect. Method. 2. A method for detecting a pattern formed on a substrate to generate image data, generating comparative image data using the image data, and determining an amount to be shifted based on the image data and the comparative image data. Is calculated with an accuracy of a pixel or less, and the image data and the comparative image data are mutually shifted with a precision of a pixel or less based on the obtained amount to be shifted, and the image data and the comparative image data shifted from each other are compared. And detecting a defect in the pattern. 3. The pattern according to claim 1, wherein the shifting of the image data and the comparison image data is performed by shifting the image data and the comparison image data in opposite directions. Defect detection method. 4. The method according to claim 1, wherein the shifting of the image data and the comparison image data in opposite directions is performed by shifting the image data and the comparison image data in the opposite directions by the same amount. 3. The pattern defect detection method according to 3. 5. The pattern defect detecting method according to claim 1, wherein the defect of the pattern to be detected is a defect having a size substantially equal to the size of the pixel. 6. The method of claim 1, wherein the alignment of the plurality of pixels or less is performed by a plurality of pixels.
3. The pattern defect detection method according to claim 2, wherein the position of the unit pixel is aligned as a unit pixel. 7. Image data generating means for detecting a pattern formed on a substrate and generating image data, and comparing image data for generating comparative image data using said image data generated by said image data generating means. Creating means, a shift amount calculating means for obtaining an amount to be shifted based on the image data and the comparison image data with an accuracy of less than a pixel, and the image data and the image data based on the shift amount obtained by the shift amount calculating means. Correction means for shifting the comparison image data from each other;
A defect detection unit for comparing the image data shifted by the correction unit with the comparison image data to detect a defect in the pattern. 8. The pattern defect detecting device according to claim 7, wherein said comparison image data creating means includes a storage unit for storing said image data. 9. The apparatus according to claim 7, wherein said correction means shifts said image data and said comparison image data by the same amount in opposite directions based on the shift amount obtained by said shift amount calculation means. Pattern defect detection device.