JPH0617778B2 - Pattern defect detection method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Field of Application] The present invention relates to an LSI wafer on which alignment errors, errors due to expansion and contraction, minute irregularities, etc. occur between two patterns to be inspected that have the same shape and are to be compared. The present invention relates to a pattern defect detection method and apparatus for detecting fine defects in a multilayer pattern or the like with high accuracy.

[Background of the Invention]

Integrated circuits such as LSI tend to be highly integrated and miniaturized. In the production of such a fine pattern, a defect is often generated in the pattern even if careful attention is paid in the production process, and a thorough inspection is required. The initial inspection is
It was performed by a large number of inspectors by visual inspection using a microscope, but there was a limit in terms of quality control due to the omission of defects due to eye fatigue. In addition, manual inspection results in obstructing the flow of production, which is a cause of lowering productivity. Therefore, automating the inspection is a very important issue from the viewpoint of quality assurance and productivity improvement.

Since the pattern on the semiconductor element targeted by the present invention has a complicated three-dimensional structure as shown in FIG. 13, for example, there are cases where the conventional inspection apparatus does not sufficiently function. In the conventional device (FIG. 14), the linear image sensors 5a and 5b are self-scanning and detect a pattern one-dimensionally. Further, the XY table 7 is configured to detect the pattern two-dimensionally by moving the LSI wafer 1 in the direction perpendicular to the scanning of the linear image sensor.

In the conventional wafer visual inspection apparatus (FIG. 14), since the determination circuit 13 is configured to handle binary signals, it is a general premise that the circuit patterns at two locations detected by the photoelectric converters 5a and 5b are the same. Met.

That is, it is required that there is no positional deviation between the two detected video signals, and if this condition is satisfied, the conventional binary comparison is effective.

Actually, it is unavoidable that a positional deviation occurs between the input patterns due to the accuracy of the XY table, the chip arrangement accuracy, the thermal deformation of the optical system, the mechanical system, and the like. It is necessary to correct it.

In the conventional example, a shift register for positional deviation correction was also provided, but the purpose is to compensate for the binarization error due to the binary image, and it is necessary to deal with the higher precision of inspection due to the miniaturization of the inspection object. I don't get it.

That is, as shown in FIGS. 15 (a) and 15 (b), when there is a positional deviation (alignment error) between the inspection target, for example, the first layer and the second layer of the semiconductor element, the dimension of the alignment error between the layers. Defects of the same order as or smaller than that cannot be detected. The inter-layer alignment error is a positional deviation that cannot be avoided when forming a pattern, and when the mismatch detection is performed after the alignment is performed by the conventional technique, it becomes as shown in FIG. 15 (c). It is impossible to detect only defects.

In addition, the pattern may have minute irregularities or deviations in width, which causes a problem that it is necessary to detect a defect that cannot be detected only by the conventional positional deviation correcting means.

As a known example of the invention relating to the alignment, Japanese Patent Laid-Open No. 57-
34402 and JP-A-57-196377.

[Object of the Invention]

The object of the present invention is to provide two patterns to be inspected even if there is an alignment error between the two patterns to be inspected having the same original shape to be inspected, an error due to expansion and contraction, and minute irregularities. To provide a pattern defect detection method and an apparatus thereof capable of detecting a true fine defect with high accuracy without erroneously detecting a normal part based on a grayscale image signal detected from a pattern. .

[Outline of Invention]

In order to achieve the above-mentioned object, the present invention detects a grayscale image signal for each of two patterns to be inspected having originally the same shape, and relatively moves the detected two grayscale image signals. The misregistration correction is performed on the two patterns, and the gradient of the shading of the grayscale image signal is detected in at least the area where the pattern misregistration remains for each of the two grayscale image signals that have been misaligned for the two patterns. Then, the detected gradation gradients are compared with each other and the mutual gradation gradients are judged as defects based on a desired difference. Further, according to the present invention, a grayscale image signal is detected for each of two patterns to be inspected having the same original shape, and the detected two grayscale image signals are relatively moved to shift the positions of the two patterns. After the correction, the positional deviation of the two patterns is corrected to 2
An area in which the positional deviation of the pattern is left is detected for each of the two grayscale image signals, the gradient of the grayscale of each of the grayscale image signals is detected in the detected area, and the gradient of each of the detected grayscales is detected. The pattern defect detection method is characterized in that the comparison is performed to determine a defect based on a desired difference between the gradients of the mutual shading. Further, the present invention is characterized in that, in the pattern defect detection method, the area where the positional deviation is left is an area in which a gray level difference between two gray level image signals is a predetermined value or more. Further, the present invention is characterized in that, in the pattern defect detecting method, the pattern is a multilayer pattern. Further, the present invention comprises a grayscale image signal detecting means for detecting a grayscale image signal for each of two patterns to be inspected having the same original shape,
A positional shift correction unit that relatively shifts the two grayscale image signals detected by the grayscale image signal detection unit to correct the positional shift between the two patterns; and a positional shift correction between the two patterns by the positional shift correction unit. For each of the two generated grayscale image signals, at least a grayscale gradient detecting means for detecting a grayscale gradient of the grayscale image signal in a region where a positional deviation of the pattern is left, and each of the grayscale gradient detecting means A pattern defect detecting apparatus comprising: a defect determining unit (comparing unit) which compares the gradation gradients and judges the mutual gradation gradients as a defect based on a desired difference. Further, according to the present invention, the grayscale image signal detecting means for detecting the grayscale image signal for each of the two patterns to be inspected having the originally same shape, and the two grayscale image signals detected by the grayscale image signal detecting means are relatively arranged. Position shift correction means for moving the two patterns to correct the position shift, and the position shift of the pattern is left for each of the two grayscale image signals corrected by the position shift correction means for the two patterns. Displacement area detecting means for detecting an area, an intensity gradient detecting means for detecting an intensity gradient of each of the intensity image signals in the area detected by the displacement area detecting means, and an intensity gradient detecting means. Defect determining means for comparing the respective gradations of the light and shade, and judging as a defect based on a desired difference between the gradations of the light and shade of each other. A pattern defect detecting apparatus characterized by comprising.

Example of Invention

 Hereinafter, examples will be described after the principle of the present invention is outlined.

<Principle> Alignment error between layers or minute pattern irregularities must be detected as defects if they are larger than a certain reference value, but must be regarded as normal if they are smaller than a certain reference value and must be allowed. I have to. When comparing two adjacent chips, after aligning the corresponding patterns f and g on the two chips as shown in FIG.
If the difference in brightness of g is taken and the alignment error between layers is small (FIG. 8), the difference in brightness 1f−g1 is larger than a certain binarization threshold th, and the original pattern f, g
The brightness gradients of have the same value or tend to be similar even if they do not have the same value. However, when the alignment error between layers is large (FIG. 9), the brightness gradients of the patterns f and g have completely different values.

On the other hand, when a defect such as an arc defect is present in the pattern f, the brightness difference 1f−g1 ≧ th. In the region, the brightness gradient has different values between the two patterns f and g.

Therefore, from the difference in the waveform of the brightness difference 1f-g1 (FIGS. 8 and 9), the positional deviation caused by the alignment error between layers and the minute unevenness of the pattern is By detecting areas with a large amount of mismatch as defect candidates and comparing the shapes of the brightness waveforms of these areas to determine the size, it is possible to detect an unbalanced value due to a defect that is smaller than a large mismatch caused by misalignment. Is possible.

For example, when the patterns f and g are aligned as shown in FIG. 10, the alignment is not completed due to the alignment error between the layers (the underlying pattern is not shown). (A) The alignment error of one picture element , (B) when there is an alignment error between 2 picture elements and (c) when there is an alignment error between 3 picture elements, consider applying a 3 × 3 window to an area with a large amount of mismatch.
This window detects a mismatched defect having a size of 3 × 3. In the case of (a), there is no problem because the amount of pattern mismatch is small. In the case of (b), the mismatch amount of the picture elements shown in the figure becomes large and the pixel becomes a defect candidate. When the 3 × 3 window is re-applied to this defect candidate and the brightness gradients of the two patterns in the window are obtained, they are almost the same, and thus it can be seen that there is a small local displacement.
In the case of (c), the mismatch amount of the picture elements shown in the figure becomes large, and the pixel becomes a defect candidate. It can be seen that the gradients in the 3 × 3 window have slightly different values, and the positional deviation is larger than in the case of (b). Depending on the value of this gradient, it can be considered as a defect or a normal part.

When the edges are thin and thin, the mismatch amount of the picture elements shown in the figure becomes large as shown in FIG. 11, but the gradients within the 3 × 3 window are the same, and it is considered that the positional deviation is allowable in the figure. Can be interpreted.

In the case of a defect such as a disconnection, as shown in FIG. 12, it can be seen that the gradients in the 3 × 3 window of the picture elements having a large amount of mismatch differ greatly and the pattern is not normal.

As described above, in the example of the 3 × 3 window described above, a defect having a size of 3 picture elements can be detected while allowing the positional deviation of 2 picture elements or 3 picture elements. That is, although the displacement of the two picture elements is uniformly detected as a defect of the size mismatch of three picture elements or more according to the conventional technique, the size deviation of the three picture elements is detected by the above method. Defects can be detected.

<Example> Next, an example of the present invention will be described.

As the photoelectric converter, any device such as a linear image sensor or a TV camera can be used, but in this embodiment, a linear image sensor will be used for description.

FIG. 1 shows an embodiment of the defect detection circuit.

The outputs of the linear image sensors 5a and 5b are aligned by the misregistration correction circuit 22. The signals 19 and 20 having no positional deviation are then sent to the difference image detection circuit 23,
The absolute value of the difference between the signals 19, 20 is detected. Next, the binarization circuit 24 binarizes the difference image to detect defect candidates. From the detected defect candidates, the 3 × 3 window processing circuit 25
Those with less than 3 picture elements are removed. The output of the window processing circuit 25 has a large amount of mismatch between the signals 19 and 20, and has a size of 3 picture elements or more. Next, the gradient detection circuits 26 and 27 obtain the in-window gradients of the signals 19 and 20 for these defect candidates. By comparing the gradients thus obtained with each other by the comparison circuit 28, when the gradients are significantly different, they are detected as defects.

Next, the configuration of each unit will be described in more detail. In FIG. 2 and FIG. 3, on the average, the aligned signals 19 and 20 (here, 8 bits) are input to the subtracter 29 (see FIG. (Fig. 2), EXR
The circuit 30 detects the absolute value of the difference between 19 and 20. The absolute value of the difference is calculated by the comparator 31 as the threshold value th.
Is binarized with. The output of the comparator 31 is a shift register 32 for delaying one scan of the linear image sensor,
33, and a 3 × 3 pixel segmentation circuit composed of a serial-in / parallel-out shift register 34, and A
If the brightness of all the picture elements in the 3 × 3 window is th or more by the ND circuit 36 (FIG. 3), 41 is enab.
Set to le, otherwise set to disable. on the other hand,
An AND circuit output 4 from the signals 19 and 20 is output by a 3 × 3 pixel cutout circuit composed of the shift registers 37 and 38 and the serial-in / parallel-out shift register 39.
Cut out 3 × 3 picture elements in synchronization with 1. AND circuit output 41
Becomes enable if the absolute value of the difference between the signals 19 and 20 is th or more over the 3 × 3 picture element, and the clipping circuit for the 3 × 3 picture element is enabled. The brightness is extracted from the 3 × 3 picture element clipping circuit and input to the subtractor 40 to detect (brightness gradient). The detected gradient is compared by the comparator circuit 28 with the gradients corresponding to the signals 19 and 20, and if the gradients are significantly different, the gradient is detected as a defect. The inside of the comparison circuit 28 is composed of a subtractor and a comparator.

An example of the gradient is shown in FIG. If the respective picture elements of the 3 × 3 picture element cutting circuit are A, B, C, ..., A gradient table as shown in FIG. 4 can be created. If the values in the gradient tables are compared and even one value is significantly different between the signals 19 and 20, then there is a defect that has caused the signals 19 and 20 to differ.

In the example described above, the absolute value of the difference between the two signals 19 and 20 is binarized by the binarization circuits 24 and 31, but as shown in FIG. 5, the absolute value of the difference between 19 and 20 is shifted. Register 42, serial-in parallel-out shift register 4
There is also a method in which the sum of the brightness of the 3 × 3 picture elements is obtained by the 3 × 3 picture element cutout circuit configured by 3 and the addition circuit 44, and this is binarized by the comparator 31.

Another embodiment is shown in FIG. FIG. 6 shows the comparison of the brightness gradients, but the brightness gradients of the 3 × 3 picture elements are compared within the range in which the two picture elements around the corresponding point on the corresponding pattern are enlarged. In the same figure, for example, it is shown that the brightness gradients in the shaded areas are compared.
A portion having the closest brightness gradient is searched for within the range of the picture element, and local alignment is performed at that time. Therefore, it is judged whether or not there is a defect by comparing the values of the gradient.

In the special case of FIG. 6, the windows 46 and 48 are 1 × 1 picture elements and 3 × 3 picture elements, respectively, and the brightness of the picture elements in the window 46 is within the range of 3 × 3 picture elements in the window 48. There is also a method of determining whether or not there is a defect by comparing the brightnesses, since a local alignment is performed at that time by searching for a location close to the brightness. This corresponds to an example in which the brightness gradient defined in the 3 × 3 picture element window is expanded to the 1 × 1 picture element window. In other words, 1 if the defect is small
It is desirable that the brightness gradients can be compared even in the case of a × 1 picture element window, and in this case, the present invention can be applied.

According to the examples of FIGS. 5 and 6, wafers having various pattern edge shapes can be dealt with, and the accuracy of defect detection can be improved.

〔The invention's effect〕

According to the present invention, such as a multi-layer pattern of an LSI wafer, a portion that cannot be aligned due to alignment error, error due to expansion and contraction, minute unevenness, etc. between two patterns to be inspected that have the same original shape to be compared. Even if occurs, there is an effect that a true fine defect can be detected with high accuracy without erroneously detecting a normal part based on a grayscale image signal detected from two patterns to be inspected.

[Brief description of drawings]

FIG. 1 is a diagram showing an embodiment of a defect detection circuit according to the present invention, FIGS. 2 and 3 are diagrams for explaining the details of FIG. 1, and FIG. 4 is an embodiment of a brightness gradient. FIG. 5, FIG. 5 is a diagram showing an embodiment of a binarization circuit for obtaining defect candidates, FIG. 6 is a diagram showing another embodiment of the brightness gradient, and FIG. FIGS. 8 and 9 are diagrams showing alignment of patterns, FIGS. 8 and 9 are diagrams showing mismatched waveforms and brightness gradients after alignment, and FIGS. 10, 11 and 12 are 3 of mismatched waveforms.
FIG. 13 is an enlarged perspective view showing details of a chip having a pattern, FIG. 14 is a view showing a conventional example of an LSI wafer appearance inspection apparatus, and FIG. 15 is an example of pattern mismatch. It is a figure. 1 ... LSI wafer, 2 ... Chip 5 ... Photoelectric converter, 7 ... XY table 14-17 ... Shift register 22 ... Positional deviation correction circuit, 23 ... Difference image detection circuit 24 ... Binarization Circuit, 25 ... Window processing circuit 26, 27 ... Gradient detection circuit, 28 ... Comparison circuit

Claims (10)

[Claims] 1. A misalignment correction is performed on two patterns by detecting a grayscale image signal for each of two patterns to be inspected having the same original shape and relatively moving the two detected grayscale image signals. Then, the gradient of the grayscale of the grayscale image signal is detected in at least the area where the positional shift of the pattern is left for each of the two grayscale image signals whose positional shift has been corrected for the two patterns, and each of the detected grayscales is detected. The pattern defect detection method, characterized in that the gradation gradients of are compared and the mutual gradation gradients are judged as defects based on a desired difference. 2. The pattern defect detection according to claim 1, wherein the area in which the positional deviation is left is an area in which the gray level difference between two gray level image signals is equal to or more than a predetermined value. Method. 3. The pattern defect detection method according to claim 1, wherein the pattern is a multilayer pattern. 4. A grayscale image signal is detected for each of two patterns to be inspected having the same original shape, and the detected two grayscale image signals are relatively moved to correct the positional deviation between the two patterns. Then, an area in which the positional deviation of the pattern is left is detected for each of the two grayscale image signals whose positional deviation has been corrected for the two patterns, and the gradient of the grayscale of each of the grayscale image signals is detected in the detected area. Is detected, and the detected gradation gradients are compared with each other, and the mutual gradation gradients are judged as defects based on a desired difference. 5. The pattern defect detecting method according to claim 4, wherein the detection of the area is performed as an area in which a difference in gray level between two gray level image signals is a predetermined value or more. 6. The pattern defect detecting method according to claim 4, wherein the pattern is a multilayer pattern. 7. A grayscale image signal detecting means for detecting a grayscale image signal for each of two patterns to be inspected having the same original shape, and two grayscale image signals detected by the grayscale image signal detecting means are relative to each other. Position deviation of at least the pattern for each of the two grayscale image signals for which the positional deviation is corrected for the two patterns by the positional deviation correcting means for correcting the positional deviation for the two patterns. In the region where the grayscale image signal is shaded, the grayscale gradient detection means for detecting the grayscale gradient of the grayscale image signal is compared with the respective grayscale gradients detected by the grayscale gradient detection means to obtain a desired difference in the grayscale gradient between them. A pattern defect detection apparatus comprising: a defect determination unit that determines a defect based on 8. The gradation gradient detecting means is configured such that the area where the positional deviation is left is an area in which the gradation difference between two image signals is a predetermined value or more. 5. A pattern defect detection device according to claim 7. 9. A grayscale image signal detecting means for detecting a grayscale image signal for each of the two patterns to be inspected having the same original shape, and two grayscale image signals detected by the grayscale image signal detecting means are relative to each other. Position shift correction means for moving the two patterns to correct the position shift, and the position shift of the pattern is left for each of the two grayscale image signals corrected by the position shift correction means for the two patterns. Displacement area detecting means for detecting an area, an intensity gradient detecting means for detecting an intensity gradient of each of the intensity image signals in the area detected by the displacement area detecting means, and an intensity gradient detecting means. Defect determination means for comparing the respective gradations of the light and shade that have been determined and judging as a defect based on the desired difference in the gradient of the light and shade of each other. Pattern defect detecting apparatus characterized by was e. 10. The position shift area detecting means is configured to detect the area as an area in which a difference in gray level between two image signals is a predetermined value or more. Item 9. The pattern defect detection device according to item 9.