JPS6129712A - Method and device for detecting defect of fine pattern - Google Patents

Abstract

PURPOSE:To inspect a complicate and fine pattern automatically by separating optical images of light visual field lighting and dark visual field lighting in the middle of an optical path of detection, and detecting and processing pattern shape information, density, and hue variation simultaneously. CONSTITUTION:The device consists of a dark visual field lighting system whose wavelength (alpha) by a narrow-band filter 3, light visual field system which is limited to light containing no alpha by a filter 9, image formation system, binary- coding circuit 16, memories 17 and 20, image comparing circuits 18 and 21, decision part 22, control mechanism for the whole, etc. Light visual field lighting and dark visual field lighting are performed at the same position and their respective images are formed on image sensors 14 and 15. In this case, a thin- film mirror 13 performs wavelength and optical path separation. Thus, the images are detected, and then a defect of a pattern shape is detected as abnormality of the plane shape of a pattern edge and abnormality of the thickness of a transparent thin film causes variation in brightness and hue, so they are inspected separately at the same time. Further, when a linear image sensor is used, the circuits 18 and 21 compare signals with signals which are one chip before and annunciate the results.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to an automatic pattern defect inspection method.
The present invention relates to a method and apparatus for detecting defects from complex and fine patterns such as circuit patterns formed on SI wafers.

[Background of the invention]

A fine pattern formed on an LSI wafer is formed of a plurality of thin film layers made of different materials. Therefore, the defects include pattern shape defects caused by poor planar shape of the layer,
There are hue defects or uneven tone defects due to abnormal layer thickness.

Humans detect and identify these defects by visually observing them with a microscope, but in order to do this automatically,
It is necessary to separately detect pattern shape information and hue or shade information.

Pattern shape information is originally based on pattern edges having steps, and dark field illumination is effective for efficiently detecting this. However, with dark field illumination, only the edge portions are detected, and the smooth hues and shades within the pattern cannot be detected. - Although bright-field illumination can detect hue and density unevenness, pattern edges are also detected at the same time, and attempting to separate this information requires extensive image processing, making it impractical.

[Purpose of the invention]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present invention provides a method and apparatus for automatically inspecting complex and fine patterns by separately detecting pattern shape information and hue/shade information.

[Summary of the invention]

That is, in order to achieve the above object, the present invention uses light having different optical properties such as wavelength and polarization direction as a light source for bright field illumination and dark field illumination, and simultaneously illuminates the same spot on the surface to be inspected. The area is detected with the same objective lens, and the optical image obtained with bright-field illumination and the optical image obtained with dark-field illumination are separated using the difference in wavelength in the middle of the detection optical path, and each image is detected using an image detector. It is characterized by detecting defects by detecting them and performing image processing.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to FIG. 1st
The figure shows a dark field illumination system with light source 1, condenser lens 2,
Narrow band filter for wavelength selection for dark field illumination (wavelength λ
1) 3, ring-shaped opening slit 4, ring-shaped opening slit 2
−5, a parabolic concave mirror 6, and a light source 7 as a bright field illumination system;
Condenser lens 8, wavelength selection filter 9, circular aperture slit 10, half mirror 11, objective lens 12 as an imaging system, wavelength separation mirror (e.g. dichroic mirror)
13, and a dark-field image detection image sensor 14, a bright-field image detection image sensor 15, a binarization circuit 16, and a memory 1.
7, image comparison circuit 18. binarization circuit 19, memory 20,
Image comparison circuit 21, determination unit 22, cave table 23, feed motor 24, table control circuit 25, overall control circuit 26
Consists of.

An object to be inspected, for example, an LSI wafer 27, is accurately positioned and fixed on the °C table 23 in advance.

Positioning and fixing mechanisms are not shown. The pattern on 27 is illuminated in dark field and bright field. The dark-field illumination is limited to a wavelength λ by a filter 5, and is illuminated onto the pattern from an oblique direction around the pattern by a parabolic concave mirror 6Vc. Bright field illumination is limited to light that does not include the wavelength λ by a filter 9, and is illuminated from above by an objective lens 12 via a half mirror 11. The pattern is magnified by the objective lens 12, and the dark field image at wavelength λ is captured by the image sensor 14.
Above, bright-field images at wavelengths other than λ are formed on the image sensor 15. Wavelength and optical path separation in the imaging system is performed by a thin film mirror 13.

Assume that there is a pattern shown in FIG. 9 on the LSI wafer 27. Figure 9(a) is a plan view of the pattern, Figure 9(b)
FIG. 3 is a sectional view taken along a dashed-dotted line in FIG. That is, a thick 8i0. There is a surface layer of 8
i0. The number of layers is extremely small. Here, the part indicated by A is a defect in which a part of the pattern shape is missing, that is, the outline shape is abnormal, and the part indicated by B is a defect of 8i0. This is a defect where the layer thickness is locally thick. A pattern shape defect like A is detected as an abnormality in the planar shape of the pattern edge. On the other hand, the defect of abnormal thickness of transparent thin film like B shows that the planar shape of the pattern edge is not abnormal, but
The brightness and hue of the flat image change. This occurs because the amount of transmitted light changes due to a change in the thickness of the thin film, or because the interference color due to the interference of reflected light from the surface and bottom surface of the thin film changes with the change in thickness. Although FIG. 9 shows an example of a negative layer pattern, similar defects occur in multilayer patterns as well. Now, when this Ueno A pattern is detected, a dark field illumination image as shown in FIG. 2(a) is formed on the imaging plane 14. In other words, the pattern edges are bright (detected), and the other almost smooth surfaces are dark (detected). Figure 2 (b) exemplifies the video signal on the dashed line shown in the center of (a). , the edges are bright (, clearly detected.

Therefore, when the binarization circuit 16 binarizes with the binarization threshold Vrn1 shown by the chain line in FIG. 2(b), the edge part becomes a stable value of 1 and the rest has a value of 0, as shown in FIG. 2(c). A value pattern is obtained. -For example, on the imaging plane of 15,000
A bright field illumination image like this is formed.

That is, the smooth surface of each pattern has shading determined by the properties of the pattern. In the example shown in FIG. 3, there is an abnormality in the pattern film thickness to the right of the bone-like pattern, and the shading changes. This part is defective. Figure 6 (bl is Figure 6 (
The video signal shown on the dashed line of al is shown, and the output is slightly reduced at the defective portion. This part is dark (value 0)
The threshold value VTH2 is set so that when the pattern is binarized, the pattern part with almost uniform brightness is set to a value of 1, and the part where discoloration or shading changes, edges, and their surroundings are set to a value of 0. A binary pattern is obtained.

FIG. 4 shows an LSI wafer as an example of the object to be inspected. On the wafer 27, LSI chips 28 are repeatedly arranged in alignment. Image sensor 1.4.15 When using a near image sensor, the image sensor has a field of view as shown by the solid line 29 in Figure 4, and ■Table 23
By scanning, the field of view of the linear image sensor moves like a zigzag arrow 30. Therefore memory 17.20
, the detection patterns are sequentially stored, and the pattern signal at the same position one chip before the top of the pattern being detected is read out. Therefore, the comparison circuit 18.21 compares the pattern being detected and the pattern at the same position of the pattern one chip before, and when there is a mismatch, the defect determination section 22
Notify that there is a discrepancy, its size, and location. When the defect determination unit is notified from 18 or 21 of the existence of a mismatch having a size greater than a preset value, the defect determination unit stores the position as a defect position. A circuit for outputting the defect position, defect size, etc. is not particularly shown. The overall control circuit 26 controls the sequence of the above-described operations. As described above, in this embodiment, the contour shape and shading information of a pattern can be simultaneously detected separately and automatically inspected using dark field illumination and bright field illumination.

Although a linear image sensor is used in this embodiment, a two-dimensional image sensor may be used and the M' table may be fed stepwise or continuously to emit pulsed light as the illumination light source. In addition, in this embodiment, the detected image is binarized,
Although the comparison was made using binary images, it is also possible to convert the images into multi-valued images using an A/D converter and compare them using multi-valued images. In addition, in this embodiment, a comparative inspection with the previous chip was illustrated, but this is also a comparative inspection between repeated patterns within one chip, a comparative inspection of two wafers,
It may also be a comparative inspection with non-defective pattern information that can be prepared in advance, such as a design pattern.

FIG. 5 shows another embodiment of the invention.

The dark field illumination system is composed of a light source 1, a condenser lens 2, a filter 3 with a wavelength λ, a polarizing plate/ring slit 31, a ring mirror 5, and a parabolic concave pin 6. Perform dark field polarized illumination. Polarization is S polarized light (
vibration is parallel to the surface to be inspected). Bright field illumination system has wavelength λ
Two light sources 52 with different wavelength components λ′, λ″
, 35, condenser lens 34, narrow band filter 35 with wavelength λ', narrow band filter 36 with wavelength λ'', λ' and λ''
half mirror 37, circular opening slit 10,
It is composed of a half mirror 11. The detection system is objective lens 1
Human bright field and dark field illumination light separation mirror 38, which transmits only the wavelength λ, reflects λ' and λ'', and transmits the bright field illumination light to λ',
Separation mirror 39 for separating λ"K, narrow band filter 40 for λ', narrow band filter 41 for λ", image sensor 42 for bright field λ', image sensor 43 for bright field λ"
It consists of The determination circuit section includes an A/D converter 44, a memory 45, a hue extraction circuit 46, an image comparison circuit 47.4B,
It consists of a determination section 22, along with a parallel table 23, a feed motor 24, a table control circuit 25, and an overall control circuit 2.
Consists of 6.

Here, the polarizing plate and ring-shaped slit 31 is K as shown in FIG.
, by dividing and combining polarizing plates, the surrounding area becomes uniform S
Polarized darkfield illumination becomes possible.

The reason why S-polarized light is used is that it is effective for detecting an internal upper layer pattern of a multilayer complex pattern with good contrast. FIG. 10 shows the structure of a dynamic RAM as an example of a wafer multilayer pattern. i.e. data@
49 is A, word line 50 is formed of polycrystalline silicon, and a storage area consisting of an electrode 51 also made of polycrystalline silicon and an inversion layer region of P substrate 52 is laid out directly below data a 49 in the data line direction. This is what is being done.

Note that an insulating film is formed between the word line 50 and the data line 49. In order to inspect the pattern of each layer of an LSI wafer consisting of many types of layers, it is necessary to inspect whether the top layer has been formed correctly each time the top layer is formed. Detection meets this purpose. The principle is shown in Figure 7. 5. S-polarized light has a higher reflectance than P-polarized light. In other words, most of the light illuminated on a multilayer pattern is reflected from the surface, making the pattern edges of the top layer obvious. It is something. The dark-field illumination image is detected by the image sensor 14 and compared with the dark-field illumination image at the same position of the pattern one chip before by 45° 47. This point is the same as the first embodiment.

On the other hand, in bright-field illumination, two wavelengths λ' and λ'' are set in advance that make it easy to detect color changes in the target pattern, and the illumination and detection optical images are separated by the wavelengths λ' and λ''.
The images are detected by image sensors 42 and 45, respectively. The hue extraction circuit 46 does not need to accurately extract the hue, but only needs to be able to detect subtle changes in tone of the multilayer thin film. Generally, a thin film produces an interference color as illustrated in FIG. 8 under bright field illumination, and its spectral reflectance differs from the film thickness tVC. Therefore, if the two wavelengths λ' and λ'' where the characteristics are noticeable are selected, the output at each wavelength is V(λ) and ■(λ"), then ■(λ')/V(λ") By determining the color tone change, it is possible to make the color tone change obvious.Here, V(λ")<V
By not performing division and setting the output to zero when TI[5], it is possible to prevent false detection in dark areas such as pattern edges, and to achieve the same effect as the first embodiment in terms of density changes. can bring. At step 46, the above processing is performed. 45 and 46 have the same function as in the first embodiment, detecting discrepancies with hue and shade changes,
22 can be recognized as a defect.

The additional items in the first embodiment (variations of the two-dimensional image sensor and comparison target) also hold true in the second embodiment. In addition, in the second embodiment, 2 of λ′ and λ″
Although the case in which wavelengths are used has been exemplified, a method of detecting the three primary colors of blue, red, and green and determining the phase angle of the colors may also be used. Alternatively, other methods using multiple wavelengths may be used.

Alternatively, a method may be used in which a wavelength light source is used, only the wavelength λ is cut, and images of multiple wavelengths are detected using multiple image sensors.

〔Effect of the invention〕

As explained above, according to the present invention, it is possible to simultaneously detect pattern outline information at the same position of the same pattern and pattern shading and hue changes, so that it is possible to simultaneously detect pattern contour information at the same position of the same pattern, and to detect complex multilayer thin films such as LSI patterns formed on Si wafers. Patterns can be inspected efficiently (i.e., with a simple structure and at high speed. In addition, since contour information and shading and hue changes are processed separately, it is possible to identify whether a pattern defect is an abnormal contour shape or not.
Defects caused by changes in thin film thickness can be easily identified.

[Brief explanation of the drawing]

Figure 1 is an overall configuration diagram for explaining one embodiment of the present invention, Figure 2 is a diagram for explaining pattern detection status and binarization in dark-field illumination, and Figure 3 is a diagram for explaining pattern detection status in bright-field illumination. 4 is a diagram illustrating the detection field of view and the detection sequence in this embodiment. FIG. 5 is an overall configuration diagram illustrating another embodiment of the present invention.
The figure is a diagram explaining a specific example of the ring-shaped polarizing plate in the embodiment of Figure 5, Figure 7 is a diagram explaining the difference in the reflection characteristics of S-polarized light and P-polarized light, and Figure 8 is a diagram explaining the interference color of the thin film. Figure 9(a) is a plan view showing the pattern on an LSI wafer, and Figure 9(b) is a diagram showing the spectral reflectance characteristics for
FIG. 10 is a diagram showing a dynamic structure as an example of a wafer multilayer pattern. 1... Light 1i 2...
Condenser lens 3...Narrowband filter
4... Ring-shaped opening slit 5... Ring-shaped mirror 6... Parabolic concave mirror 7...
... Light source 8 ... Curved condenser lens 9 ... Wavelength selection filter 10 ... Circular aperture slit 11
...Half mirror 12...Objective lens 13...Wavelength separation mirror 14...
Dark field image detection image sensor 15...Bright field image detection image 16...2
Value conversion circuit - sensor 17, 20... Memo! J 18,21・
... Image comparison circuit 22 ... Judgment section
27... LSI wafer rack 1 Figure 2 Figure 3 (α) (d) (b)
' (b) Figure 10 Figure 7 Figure 7 Notebook 8 x N '/Genryo Atsushi Figure 9 (and λ-n (b) Figure 1O

Claims (1)

[Claims] 1. Dark-field illumination and bright-field illumination with different wavelengths are simultaneously illuminated on the same spot on the surface to be inspected, and this is detected with an objective lens;
In the middle of the detection optical path, the optical image obtained with bright field illumination and the optical image obtained with dark field illumination are separated using the difference in wavelength, and each is detected by an image detector, and changes in pattern density are detected from the bright field illumination image. A method for detecting defects in fine patterns, characterized by detecting defects due to color changes, hue changes, and pattern contour defects from dark-field illumination images. 2. The method for detecting defects in fine patterns according to claim 1, characterized in that S-polarized light is irradiated from the surroundings as the dark-field illumination to make the uppermost layer pattern visible. 3. A dark field illumination device and a bright field illumination device with different wavelengths,
A means for separating the optical path of a dark-field illumination image and a bright-field illumination image according to wavelength in the microscope detection optical system, two or more image detectors for detecting the dark-field illumination image and the bright-field illumination image, and images obtained from these. A fine pattern defect characterized by having an image processing/judgment circuit that processes signals, and detecting defects due to changes in pattern shading or hue from bright-field illumination images, and defects in pattern contours from dark-field illumination images. Detection device. 4. The micro pattern defect detection device according to claim 3, wherein the dark field illumination device has means for irradiating S-polarized light from the surroundings.