JPH0523063B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to a method and apparatus for detecting defects in fine patterns such as LSI patterns formed on a wafer.

[Background of the invention]

For example, LSI products are manufactured by forming layers of circuit patterns such as SiO ₂ on Si wafers, and during this process, inspecting the patterns of each layer for shape defects is essential to ensuring product yield.

An example of a conventional inspection device is shown in FIG. Sample 1
Objective lens 2 installed above the illumination optical system 3,
It is composed of a semi-transparent mirror 4 and a detector 5. The illumination optical system 3 and the objective lens 2 constitute an epi-illumination system, and as shown in FIG. 6 occurs. Here, the specularly reflected light 7 is the zero-order diffracted light (DC component), and the higher-order reflected light 6 involved in detection is the 1st to Nth higher-order diffracted lights (here, N is determined by the aperture of the objective lens 2). ).

The detector 5 uses a CCD or a photodiode array, and obtains a detection signal 8 from the sample shown in Fig. 1b (c). Signal 8 passes through a differentiating circuit (not shown) to obtain signal 9 (d). Signal 9 and set threshold 10,1
1 is compared to obtain a binary signal 9a indicating the edge position (e).

The detector 5 and the sample 1 are relatively scanned, and the signal 9a is two-dimensionally stored to obtain an edge binarized image 12 shown in FIG. 15a. Defects 13, 14, and 15 are detected by comparing the binarized image 12 with a standard pattern 16 stored in a memory circuit in advance.

The standard pattern 16 may be a pattern created from design data or a pattern in an adjacent chip detected in advance on the same wafer.

An example of a method for comparing pattern 12 and pattern 16 is shown in FIG. Figure a and b are each pattern 1.
2 and 16 are shown, and extraction patterns 17 and 18 are obtained. Similarly, extraction patterns 19 and 20 in the horizontal direction are shown in c and d. Similarly right
Extraction pattern edges 21 and 22 at 45° and 45° to the left are shown in e and d. Extraction results a and b, c and d, e and f
If the discrepancy is detected by comparing each, defect 13,
Get 14,15. A circuit that performs the above processing is shown in Document 1.

In addition to the above edge information, pattern defects include the first
Etched residue 23 of PolySi layer 23 shown in Figure 7
There is a flat part defect as in example a. Originally, the PolySi 23 was removed from this part, but due to a defect in the previous process, the PolySi remains, resulting in a defective cell circuit. This defect 23a cannot be detected using the edge information described above. This defect is detected, for example, by comparing the brightness information 7 of the reflected light of the adjacent good cell circuit 23b with the cell circuit 23a.

However, this brightness information 7 is affected by stray light 27 generated on the surface of the objective lens 2 and other photoelectric elements, and becomes a detection signal 8a shown in FIG. There is a problem that it cannot be done.

As mentioned above, there are two types of defects in LSI wafer patterns: edge information and brightness information.
With the method shown in the figure, only edge information can be obtained;
For the reasons shown in Figure 8, it was not possible to accurately detect the brightness and edge position.

[Purpose of the invention]

SUMMARY OF THE INVENTION In view of the above-mentioned conventional problems, an object of the present invention is to provide a method and apparatus for detecting defects in wafer patterns that accurately detects brightness and edge positions and enables stable automatic inspection of defects in LSI wafer patterns. It's about doing.

[Summary of the invention]

That is, in order to achieve the above object, the present invention provides separate optical systems for obtaining edge information and brightness information, and the former optically detects high-order reflected light 6 from the edge.
In the latter case, an optical isolator system is installed to remove stray light, and both pieces of information are processed separately to easily, stably, and automatically detect the wafer pattern. It is something.

[Embodiments of the invention]

The present invention will be explained below based on embodiments shown in the drawings. FIG. 1a shows an optical system that detects only high-order reflected light 6 from pattern edges and blocks specularly reflected light 7. On the Fourier transform surface of objective lens 2,
If a light shielding plate 27 is installed to block only the specularly reflected light 7 shown in FIG. 3, only the higher-order reflected light 6 shown in FIG. If the detector 5a and the sample 1 are relatively scanned, a detection signal 8b shown in FIG.
get. With this optical system, the edge shape (FIG. 15b) can be stably obtained without being affected by the specularly reflected light 7.

FIG. 2a shows an optical system that blocks stray light 27 from the objective lens 2. This is called an optical isolator system, and includes a polarizing beam splitter 4a and a quarter-wave plate 29.
It consists of When S-polarized light is used as the illumination light 3a, it is totally reflected by the polarization beam splitter 4a. Stray light 2 from the surface of objective lens 2 maintains S polarization and therefore does not reach detector 5a. Sample 1a
The illumination light that illuminates the 1/4 wavelength plate 29 passes through the 1/4 wavelength plate 29 twice, so it becomes P-polarized light and the polarized beam splitter 4
a and reaches the detector 5a. The detection signal 8c shown in b obtained thereby is not affected by the stray light 27.

In the optical systems shown in FIGS. 1 and 2 described above, if a laser beam is used as the illumination light 3a, high brightness can be easily obtained, and therefore a detection signal with a high S/N ratio can be obtained.

Next, a method for improving the resolution of the detection system is shown in FIGS. 3 to 5.

In FIG. 3, the semi-transmissive mirror 4 is omitted and the principle is shown expanded to the illumination system 3 and the detection system 34. The point light source 3b formed by the pinhole 33 is connected to the objective lens 2.
As a result, the light spreads over the sample to form a distribution 3c, and an infinitesimal pattern 30 on the sample is illuminated. The infinitesimal pattern 30 spreads to form a distribution 31 on the image plane of the objective lens 2. When the pinhole 28 is placed at this position, the spread distribution 31 becomes the distribution 32 after passing through the pinhole 28, resulting in improved resolution.

Here, the spread 31 is the radius M×R (M=×0.61λ/
NA), so it is effective to make the pinhole 28 smaller than this. (Here, M is the magnification of the objective lens 2, λ is the wavelength of the illumination light 3b, and NA is the aperture of the objective lens.) The resolution improvement in FIG. 3 will be explained in FIGS. 4 and 5. The upper layer of sample 1 is a transparent layer 35 of SiO ₂ etc.
In this case, the effect of the pinhole 28 allows the enlarged image i' of the upper layer pattern i to be separated from the enlarged image j' of the lower layer pattern j (FIG. 4).
This improves vertical resolution.

Similarly, the enlarged image h' of point h and the enlarged image l' of point l can be separated by the pinhole 28, and the resolution in the lateral direction is also improved (FIG. 5).

FIG. 6 shows an embodiment of the present invention. This is the first
2 and 3 are combined with an optical system. The detection system in FIGS. 1 and 2 is branched by a branch mirror 36, and the former has a pinhole 28b,
Detector 5b, in the latter pinhole 2 if necessary
8c and a detector 5c were installed. Since the detector 5c detects brightness, the pinhole 28c is not necessarily required. In this embodiment, the illumination light from the illumination optical system 3 is made into parallel light by the condenser lens 35, the objective lens 2 is made into an afocal system, and the imaging lens 37
A light shielding plate 27 was installed on the Fourier transform surface. At this time, the magnification M of the detection system 34 is the ratio (f ₁ /f) of the focal lengths of the objective lens 2 and the imaging lens 2a. Detection signals 8b and 8c of detectors 5b and 5c can be output simultaneously, and these signals 8b,
Defects (edge defects 13, 14, 15 and brightness defect 23a) can be detected by creating a binarized image 12 and a brightness image by 8c and comparing them with each normal image.

Fig. 7 shows the signal 8b of the detectors 5b and 5c in Fig. 6.
A method for creating a composite image of and 8c will be explained. Sample 1
are scanned by a scanning means, and two-dimensional images E and F of appropriate size are stored in memory circuits 37a and 37b. The amplifier circuits 38 and 39 multiply E and F by preset constants α and β, and create an image (αE+βF) in the storage circuit 40. Here, if α is set larger than β, an image with emphasized edges ((αE+βF)) can be created.If the created image and the image _G0 stored in advance in the memory circuit 50 are compared in the comparison circuit 51, , the aforementioned edge defect 13,1
4, 15, the brightness defect 23a can be detected. image
G ₀ may be an image calculated from design data or a detected image of the same position of adjacent chips within the same sample wafer.

FIG. 8 shows a method of detecting defects by comparing each of the two-dimensional images E and F with pre-stored images E ₀ and F ₀ . The image E stored in the memory circuit 37a and the image _E0 stored in the memory circuit 52 are compared in the comparison circuit 53, and the mismatched portions (defects 13, 14, 15) are compared with the image E0 stored in the memory circuit 52.
6 is stored.

The relationship between the lower layer 26 and the upper layer 25 of the sample 1 is as shown in FIG. 9, and there is an allowable overlay error (alignment error) between these two layers. For example, in the example of images E ₀ 52 and E 37a shown in FIG. 9, the upper layer 25 of E is located 5 pixels higher in the y direction than the upper layer 25 of E ₀ . If the unmatched portion shown at this time is extracted, a defect will be extracted even though there is no defect. To prevent this, the comparison circuit 56 performs the following operation. That is, image E is moved one pixel at a time in the y direction within a pixel range (5 pixels in FIG. ₉ corresponding to the above-mentioned allowed overlay error) and compared with image E Only the mismatched parts are extracted and treated as defects. In Fig. 9, for the sake of simplicity, we have explained the overlay error only in the y direction, but in general there is an overlay error in the x and y directions.For example, if the tolerance is for 5 pixels in both the x and y directions, it is 25 (5 x 5). Compare times. As a result, if the image E is in Figure 15a and the image _E0 is in Figure 9a, even if an overlay error occurs, defects 13, 14, 1
5 can be detected. Naturally, images E (Fig. 9b) and E ₀
In the comparison shown in FIG. 9a, no defects are detected. The circuit for detecting a mismatched portion while shifting the above-mentioned image in the x and y directions is similar to that in Japanese Patent Laid-Open No. 133022/1983. As shown in Fig. 9, the area around the edge is affected by the pattern step, and the brightness information based on the film thickness shown in Fig. ₁₇ cannot be clearly detected. It is desirable to remove the area around the edges. FIG. 18 shows a procedure for removing comparisons around edges.

Since the brightness image F37b and the image E37a are stored simultaneously, the edge enlargement circuit 57 obtains the image H (FIG. 18c). This is a process of expanding the edge image E by appropriate pixels in the x and y directions. Similarly, the edge enlarging circuit 61 obtains the image H ₀ from the image F ₀ 54 and the image E ₀ 52. Here, images H and H ₀ are “0” and “1” images (binarized images)
"0" indicates the area around the edge, and "1" indicates the other area. If continuous areas of "1" are extracted from images H and _H0 , closed areas '~' and ~ are obtained. The closed regions of images F and F ₀ with the same number originally have the same brightness. The brightness in the corresponding regions of the closed regions ' to ' and ~ of the multivalued images F and F ₀ is extracted by brightness extraction circuits 55 and 62 . This is the average value of the brightness within the area. For example, in the area ', the sum of brightness within m' × n' pixels is averaged by m' × n', and in the corresponding area, the sum of brightness within m × n pixels is averaged by m' × n'. The total brightness is averaged by m×n. The average brightness within the corresponding areas is compared by a comparison circuit 58, and if there is a difference in brightness, the defect 23a is stored in the memory circuit 59.

Edge defect memory 56 and brightness defect memory 59
is connected to the OR circuit 60, so the OR circuit 60 outputs all defects.

A method of removing only the area around the pattern edge from the above image F will be described. When image F is stored in the circuit 55 for masking with H, only the edges of image F become "0". Similarly, F ₀ is stored in a circuit 62 that masks H ₀ . The circuits 55 and 62 respectively store images in which only the edges of the multi-valued images F and F ₀ are "0". Taking into account the above-mentioned overlay tolerance, a circuit 58 calculates the difference between images 55 and 62 to create a difference image, and compares this with a set threshold to identify the brightness defect 23a.
get. When performing the above calculation, if the target is "0" (near the edge), the comparison result can be ignored and no defect will be detected.

An example of a circuit for realizing this is shown in FIG. “0”
The level determination circuits 65 and 66 are circuits that determine whether the information f ₀ (f) of each pixel F ₀ and F is “0”. The output of the OR circuit 67 becomes "1" when f or f ₀ is "0", and the inverting circuit 68 further obtains an output "0". The differential amplifier 58a is f-
The computation of f ₀ is performed, and the difference signal and the threshold value TH are compared in the comparator circuit 58b. When the AND circuit 58c calculates the logical sum of the output of the circuit 58b and the output of the inversion circuit 68, the output is only the defect 23a, and information around the edges can be ignored.

The standard images E ₀ and F ₀ mentioned above may be calculated from design data, or images detected at the same position of adjacent chips in the same sample may be used.

FIG. 12 shows an embodiment of a detection optical system using a laser 63 as a light source. The laser beam from the pinhole 33 placed between the lenses 64 and 35 forms a spot with a diameter a on the sample 1 by the objective lens 2.

As the rotating polygon mirror 62 rotates, the spot scans the sample in the x direction. If the sample 1 is sent in the y direction by a stage (not shown) in synchronization with this, images E and F are stored in the memory circuit 37a (and 37b).
are respectively memorized. In the figure, for simplicity, light shielding plate 2
7, the quarter-wave plate 29, etc. are omitted from illustration.

FIG. 13 shows an embodiment in which one-dimensional solid-state imaging devices are used for collective illumination and detectors 5c and 5b. It is important that the size of one pixel of the solid-state image sensing device is smaller than the pinhole diameter 28 (M×R) explained in FIG. In FIGS. 12 and 13, the entire surface of the sample 1 can be inspected by feeding the sample 1 in a zigzag manner in the y direction.

〔Effect of the invention〕

As explained above, according to the present invention, it is possible to stably detect minute pattern defects.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing an external appearance of a detection device according to the present invention, a cross section of a sample, and its signal waveform, and FIG. 2 is a diagram showing a cross section of a detection device according to the present invention and its signal waveform. 3 is a principle diagram showing the optical path of the optical type according to the present invention, FIGS. 4 and 5 are diagrams showing the optical path according to the present invention, and FIG. 6 is a diagram showing the optical path of an embodiment of the present invention. 7 and 8 are block diagrams showing electrical processing according to an embodiment of the present invention, and FIG. 9 is an image F ₀
(or E ₀ ) and the image F (or E), FIG. 10 shows the image F, the binarized image of the image E, and the image H
FIG. 11 is a diagram showing an electric circuit for not performing comparison processing of brightness near edges, and FIG. 12 is a diagram showing a binary image of image H ₀ according to the present invention. External view showing the laser scanning detection optical system, 1st
Figure 3 is an external view showing the collective illumination detection optical system according to the present invention, and Figure 14 is an external view of a conventional detection device, a cross section of sample 1, a signal waveform, a differential waveform of this signal, and its binary signal. FIG. 15 is a diagram showing a wafer pattern, FIG. 16 is a diagram showing an edge extraction image of a wafer pattern, FIG. 17 is a cross-sectional view of a sample, and FIG. 18 is an external view of a conventional detection device. It is a figure which shows the signal waveform. 1, 1a...sample, 2...objective lens, 2a...
...Imaging lens, 3, 3a, 3b...Light source, 4a...
...beam splitter, 5, 5a, 5b, 5c...
Detector, 6...high-order reflected light, 7... specular reflected light,
8, 8a, 8b...detection signal, 9...differential signal,
10, 11...Threshold value, 9a...Binarized signal, 12
...Inspection target pattern, 16...Standard pattern,
27... Light shielding plate, 28, 28a, 28b... Pinhole, 29... 1/4 wavelength plate, 30... Infinitely small pattern, 33... Pinhole, 34... Detection system,
35...Transparent layer ( _SiO2 etc.), 36...Semi-transparent mirror,
37a, 37b...memory circuit, 38, 39...
Amplifier circuit, 40... Memory circuit, 50... Memory circuit, 51... Comparison circuit, 52, 54... Memory circuit, 53... Comparison circuit, 55, 62... Brightness extraction circuit (masking circuit), 57 , 61...edge expansion circuit, 56...edge defect memory, 59...
...Brightness defect memory, 58...Brightness comparison circuit,
60...OR circuit, 63...Laser, 64...Lens, 62...Rotating polygon mirror, 65, 66...Level judgment circuit, 67...OR circuit, 68...Inverting circuit, 58a...Differential amplifier , 58b...comparison circuit, 58c...AND circuit, 69...mirror.

Claims (1)

[Claims] 1. Of the light obtained from the pattern on the wafer, the diffraction component excluding the zero-order component and the component including the zero-order component are detected separately by light receiving elements, and the light obtained from each light receiving element is detected. 1. A method for detecting defects in patterns on a wafer, the method comprising the step of detecting edge defects and brightness defects in a pattern by comparing a normal pattern signal with a video signal. 2. A laser oscillator that outputs a laser beam, a polarizing beam splitter that reflects the laser beam output from the laser oscillator and passes a predetermined polarized light, and a focusing optical system that focuses the laser beam onto a pattern on the wafer. a polarizing optical system for converting the laser beam into predetermined polarized light, a scanning means for relatively scanning the pattern on the wafer and the laser spot, and splitting the detection light that has passed through the polarizing beam splitter. A defect detection device for a pattern on a wafer, comprising a branching mirror, a spatial filter that blocks zero-order components of detection light on one side branched by the branching mirror, and a laser spot. 3. The defect detection apparatus for a pattern on a wafer according to claim 2, wherein the first and second photoelectric elements are each constructed from a one-dimensional solid-state image sensor. 4. A laser oscillator that outputs a laser beam, a polarizing beam splitter that reflects the laser beam output from the laser oscillator and passes a predetermined polarized beam that returns from the pattern on the wafer, and a polarizing beam splitter that outputs the laser beam onto the wafer. an objective lens that focuses light on the pattern;
a polarizing optical system that converts the laser beam into predetermined polarized light; a scanning device that relatively scans the pattern on the wafer and the irradiated laser spot; and splitting the detection light that has passed through the polarizing beam splitter. A branching optical system, a spatial filter that blocks the zero-order component of the detected light on one side branched by the branching optical system, a pinhole arranged at a conjugate/enlarged position of the laser spot, and the above-mentioned spatial filter and pinhole. A first detector having a first photoelectric element that receives the component obtained by passing the light, and a second detector having a second photoelectric element that receives the detection light branched to the other side by the branching optics. What is claimed is: 1. A defect detection device for patterns on a wafer, comprising: