JPH0636016A - Optical inspection method and device for fault of surface of body - Google Patents

Abstract

PURPOSE: To inspect a semiconductor wafer formed into a pattern at a comparatively high speed with a low false warning rate during the period of the manufacture of the wafer or immediately after the manufacture by scanning the whole surface of an object with an optical beams whose diameters are comparatively small and at a comparatively high speed in a first stage inspection. CONSTITUTION: The inspection device and a stand 2 to which a wafer W to be inspected is fitted are provided. The inspection of the first stage on the wafer W is executed by a laser 3. TheX laser 3 radiates the laser optical beams and it scans the whole surface of the wafer W. Plural condensers 4 are arranged in a circular array and they collect light beams scattered from the wafer W and send them to plural detectors 5. The outputs of the detects 5 are sent to a stage I picture processor 7 through a stage I preprocessor 6. The stage I picture processor 7 processes information under the control of a main controller 8. The stage I picture processor 7 processes the outputs of the detectors 5 and stores information which has the defect at a high probability on the wafer and shows a doubtful position in the storage device of the main controller 8.

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION The present invention relates to a method and apparatus for optically inspecting the surface of an object to detect defects. The present invention is particularly useful for optically inspecting patterned semiconductor wafers used in manufacturing integrated circuit dies or integrated circuit chips. Therefore, the present invention is described in detail below with respect to this application.

[0002]

2. Description of the Prior Art Inspecting an unpatterned semiconductor wafer for particles present on its surface is relatively straightforward and can be easily automated. In one known type of such device, the wafer is scanned with a laser light beam and the light scattered by the particles is collected by a photodetector to detect the presence of the particles. However, it is very difficult to inspect a semiconductor wafer formed in a pattern and detect defects in the pattern. It is because the light scattered by the pattern is predominantly larger than the light scattered by the particles or defects and will therefore give rise to many false alarms.

The inspection apparatus currently used for inspecting a wafer formed in a pattern is, as a whole,
On a pixel-by-pixel basis, it is based on analyzing a high resolution two-dimensional image of a patterned wafer using a photoelectric converter such as a CCD (charge coupled device). However, due to the extremely large number of pixels involved, the processing speed of such devices is very slow. Because of this, inspection of patterned wafers is currently used only for near statistical sampling purposes. as a result,
Until a large number of such wafers have been manufactured and the problems due to these defects begin to manifest, the microscopic defects in the patterned semiconductor wafer remain largely undetected. Therefore, the subsequent discovery of such defects would increase losses, reduce yields, and increase manufacturing downtime.

US Pat. No. 4,764, named Autumbu et al.
4,969 (Japanese Patent Application No. 61-14701)
Discloses an optical inspection technique having a first-stage inspection in which a macroscopic overall inspection is performed and a second-stage inspection in which a microscopic inspection is performed. However, macroscopic gross examinations have a much lower resolution than microscopic gist examinations, so macroscopic gross examinations will not detect very small defects. Therefore, such a small defect will not be noticed in the microscopic inspection. In the embodiment disclosed in the patent, the macroscopic gross inspection stage has a resolution of about 100 microns and the microscopic gist inspection stage has a resolution of about 1 micron.

Accordingly, comparisons are made because inspection can be performed during or shortly after the manufacture of the wafer so that the process step that causes the defect can be immediately identified, and thus corrective action can be taken immediately. High speed, relatively high resolution, and relatively low false alarm rate,
There is an urgent need to inspect patterned semiconductor wafers. This need has resulted in device integration in integrated circuits currently being manufactured from these wafers, and integrated circuits designed for future manufacturing.
Increasingly, due to increasing die size and number of layers. Such increased integration requires that the number of microdefects per wafer be significantly reduced in order to obtain reasonable die yields.

[0006]

According to one feature of the invention, in a first step an information is provided for optically inspecting the entire surface of the object and information indicating the position on said object suspected of having a defect. Electrically outputting, storing the suspicious position in a storage device, and in a second stage, determining the presence or absence of a defect at the suspicious position on the surface of the object. Optically inspecting only suspicious positions with high resolution, the first stage inspection being relatively small so that the first stage inspection can be performed with the same high resolution as the second stage inspection. A method for inspecting defects on the surface of an object is obtained, characterized in that it is performed by scanning the entire surface of the object with a light beam of diameter and at a relatively high velocity.

It will thus be seen that the novel method of the present invention is inherently capable of monitoring defect density while maintaining high throughput rates with relatively low false alarm rates. Therefore, the first stage inspection is performed at a relatively high speed so as to indicate only the suspicious position having the defect with a high probability, and the second stage inspection is performed with a high resolution only on the suspicious position having the defect with a high probability. . The first stage inspection can be performed with a relatively high resolution comparable to that of the second stage inspection. Therefore, this new method can take out very small defects in the first stage inspection, which could not be taken out by the microscopic-macroscopic inspection device. For example, the second stage inspection can be performed with a resolution of 1 micron or less, similar to that disclosed in the patent to Autumbu et al. However, by using a small diameter light beam in the first stage inspection, it is possible to achieve this stage of resolution to a high degree similar to that of the second stage. Therefore, the resolution can be as high as one tenth of the diameter of the light beam. For example, by using a laser light beam with a diameter of a few microns, the first stage inspection can also be performed with a high resolution of 1 micron or less, which is comparable to the second stage inspection. This is in contrast to the Automub et al. Device, where the first stage inspection is performed at a much lower resolution of about 100 microns.

The two-step sensitivity of this novel method can be adjusted according to the requirements of a particular application. Therefore, for applications involving a relatively small number of defects, the sensitivity of the first stage inspection can be increased at the expense of a higher false alarm rate of the first stage. However, since the second stage inspection only inspects a relatively small number of suspicious locations, the entire inspection can be performed relatively quickly if the false alarm rate is reduced and any processing step or device Defects caused by can be identified by manufacturing personnel and the cause of such defects can be immediately corrected.

According to another feature of the invention, the first inspection step is performed by optically scanning the entire surface of the object with a laser beam, and the second inspection step is automatic immediately after the first inspection step. Performed by placing the image on the transducer only at the positions that are suspicious in question, and the converter converts the image into an electrical signal and the analysis of the electrical signal is performed.

According to further features in preferred embodiments of the invention described below, the surface of the object to be inspected has a pattern, eg, used to fabricate a plurality of integrated circuit dies or integrated circuit chips. A wafer formed in a patterned pattern. The first inspection step compares the pattern to be inspected with another pattern which acts as a reference pattern, and as a result identifies the position on the pattern to be inspected where there is a large difference with respect to the reference pattern. ,
It indicates that there is a high probability that defects are present at those locations on the pattern to be inspected. The second inspection step also performs a comparison between the inspected pattern and the reference pattern, inspects positions on the inspected pattern that have a large difference to the reference pattern, and determines suspicious positions on the inspected pattern Indicate that there is a defect.

The reference pattern can also be a pattern on another object (eg, die-to-die comparison), or another pattern on the same object (repeated pattern comparison), or a database. It can be the data stored in (die and database comparison).

According to another characteristic of the invention, the first inspection stage optically scans the surface of the object with a light beam, whereby N data pieces of data representing pixels of different images of the pattern to be inspected are examined. By producing a first stream of streams and a second stream of N streams of data representing pixels of images of different reference patterns, and comparing the data of the first stream with the data of the second stream. A first inspection stage is carried out, which makes it possible with a high probability to obtain an indication of the suspicious location of the defective pattern to be inspected.

According to yet another feature of the invention, the second inspection stage produces an image of the suspected location of the inspected pattern and the corresponding location of the reference pattern on the transducer,
As a result, two sets of electric signals corresponding to the respective pixels of the pattern to be inspected and the reference pattern are output, and the pixels of the pattern to be inspected and the pixels of the reference pattern are compared with each other, and predetermined at respective positions. This second inspection step is carried out by indicating a defect if there is a mismatch of a given magnitude. Each of the suspicious positions of the pattern to be inspected and the reference pattern is imaged at a plurality of different depths, and one set of electrical signals is applied to the other set to obtain image registration at each depth. Is shifted with respect to the electric signal of.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Overall Device The device shown in the drawings is for automatically inspecting patterned semiconductor wafers, each having a plurality of similar integrated circuit dies patterned in a similar pattern. Specially designed. This device inspects each of the patterns. This pattern is called the inspected pattern. This inspection is performed by comparing the pattern to be inspected with at least one other pattern on the wafer and detecting their difference. This at least one other pattern is called a reference pattern. When this difference is detected, it will indicate that the pattern being inspected is defective.

The inspection is carried out in two stages. In the first stage, the entire surface of the wafer is inspected at a relatively high speed, and information indicating a position on the wafer that is suspected to be defective is output. These positions are stored in storage.
In the second stage, only those suspicious locations stored in the storage device are examined with a relatively high spatial resolution.
Then, it is determined whether or not there is a defect. As a result, it is possible to easily identify and correct the processing step having the defect.

The inspection apparatus shown in FIGS. 1 to 3 has a base 2 for mounting a wafer W to be inspected. First wafer
The step inspection is performed by the laser 3. The laser 3 emits a laser light beam, which scans the entire surface of the wafer W. A plurality of collectors 4 are arranged in a circular array and collect the light scattered from the wafer and direct the scattered light to a plurality of detectors 5. The output of the detector 5 is sent to the stage I image processor 7 through the stage I preprocessor 6. The stage I image processor 7 processes the information under the control of the main controller 8. The stage I image processor 7 processes the output of the detector 5,
Then, information is generated that indicates a suspicious position having a high probability of having a defect on the wafer. These suspicious positions are stored in a storage device in the main controller 8.

Only suspicious locations with a high probability of being inspected are inspected by the stage II inspection system. This device has an optical device for producing an image of this suspect position on the opto-electrical converter. This photoelectric converter is, for example, CC
It can be a matrix 9 of D and converts the image on it to an electrical signal. These signals are Phase II
Through the preprocessor 10, the stage II image processor 11
Sent to. The stage II image processor 11, under the control of the main controller 8, outputs information indicating whether or not there is a defect at each suspicious position inspected in stage II.

In the block diagram of FIG. 2, the base 2 (FIG. 1) and associated elements included in the wafer handling apparatus are shown as a block 1.
2 as a whole. The stage 2 is controlled by the movement controller shown in block 13 to perform the correct positioning of the wafer on the stage 2 in each of the stage I inspection stage and the stage II inspection stage, and also in the stage I inspection of the wafer W. Perform a scan.

The photodetector 5 shown in FIG. 1 corresponds to the block 51 shown in FIG.
Is provided in the stage I image acquisition sensor indicated by. The opto-electrical converter 9 of FIG. 1 is included in the stage II image acquisition sensor shown in block 52 of FIG.

FIG. 2 also shows a post-processor 14 that processes the information from the Stage I processor 7, a main controller 8 that manages and synchronizes the data and controls the flow, and an operator controls the main controller 8. It has a keyboard 15 and a joystick with which information can be entered and a monitoring device 16 with which the operator can monitor the processing of the information.

The elements of the wafer handling and image acquisition sub-devices for both stages are provided in the dotted block generally shown in FIG. The elements of the image processing subsystem for both stages are provided in the dotted block indicated by B. The elements of the operator control console sub-device are provided in the dotted block indicated by C.
This operator control console sub-device not only has a main control device 8, a keyboard 15 and a monitoring device 16, but also
It also has the graph terminal device shown at 17 in FIG.

The other elements shown in FIG. 1 are described in detail below in connection with their respective subdevices.

Wafer Handling and Image Acquisition FIG. 3 illustrates wafer handling and image acquisition sub-device A (FIG. 2) in further detail. This sub-device has a large mass pedestal 2 (such as granite) mounted on a vibration isolation device 20. The vibration isolation device 20 attenuates high frequency vibrations from the outside world. This sub-device also has a movement control device 13 which is controlled by the main control device 8. The movement control device 13 controls the one-way scanning stage 21.
This stage moves the vacuum chuck 24. The vacuum chuck 24, when the laser light beam from the laser 3 is deflected in the other orthogonal direction to scan the entire surface of the wafer during the Stage I inspection, during its movement in one other direction, Hold the wafer flat with respect to the Stage I sensor 5.

The transfer controller 13 further includes 2 in order to position the wafer at any desired position with respect to the Stage II detector 9 (CCD matrix) during the Stage II inspection.
The dimensional scanning stage 22 is effectively controlled. The movement control device 13 further controls the rotation / level / focus stage 23. The movement controller 13 rotates the wafer about its axis during the scan to angularly align it,
And keep it horizontal and keep the focus. The stage 23 also moves the vacuum chuck 24 and its wafer toward or away from the Stage II sensor 9, thereby allowing it to move to different depths during the Stage II inspection, as described in detail below. Allows to produce multiple images.

FIG. 3 also schematically illustrates the wafer handling apparatus 25. The wafer handling device 25 includes a vacuum chuck 24,
Wafer pre-alignment device 26, and cassettes 27 and 2
The wafer W is transferred between 8 times. Pre-wafer aligning device 26
First aligns the wafer angles and then the centers. An optical character recognition device 29 for reading the wafer identification code is also shown schematically in FIG.

The parts individually are generally well-known parts and will therefore not be described in detail here.

As shown in FIG. 4, a laser (eg an Argon laser) 3 emits a laser light beam. This laser light beam is incident on the polarizer light beam splitter 30. Polarizer light beam splitter 30
Are arranged so that the laser light beam is transmitted toward the wafer W, but the light reflected from the wafer is reflected toward the photodetector 31. The photodetector 31 outputs an electrical signal, which controls the stage I preprocessor 6. The laser light beam from the light beam splitter 30 passes through a light beam expander 32 and then a cylindrical lens 33a, a deflector 34 and another cylindrical lens 33b.
, A folding mirror 35, a multi-magnification telescope 36, a light beam splitter 37, a quarter-wave plate 38, and a microscope objective lens 39. The quarter-wave plate 38 converts linearly polarized light into circularly polarized light, and the microscope objective lens 39 focuses the laser light beam on the wafer W to focus it.

The light beam expander 32 expands the diameter of the laser light beam to illuminate the entire deflector 34 optical aperture. The cylindrical lens 33a focuses the laser light beam on the deflector 34 and focuses it. Deflector 34
Is an acoustoelectric deflector. The deflector 34 scans the laser light beam in one orthogonal direction in a sawtooth pattern in the time domain, while the movement controller moves the table (and the wafer above it) in the other orthogonal direction. This scans the entire surface of the wafer. The folding mirror 35 is
The laser light beam is reflected and advances toward the multi-magnification telescope 36. The multi-magnification telescope 36 matches the diameter of the laser light beam with the scanning aperture so that the entrance requirements of standard microscope optics are met. The slit 40 in the telescope 36 allows only the first-order diffracted light of the laser light beam to enter the wafer W.

Light beam splitter 37 directs a portion of the light beam toward the wafer and reflects another portion of the light beam toward autofocus device 41, as described above. In the autofocus device 41, the wafer is the microscope objective lens 3
It is determined whether or not the focus position is 9. The autofocus device can be a standard autofocus device, such as the autofocus device used in the Leitz Ergolux microscope.

The light incident on the wafer W to be inspected and reflected by the laser light beam is arranged in a circle around the objective lens 39, as shown in more detail in FIGS. 5 and 6a. It is collected by a plurality of light collectors 42.
The pattern on the wafer W is based on a grid of lines that are 45 ° apart from each other. The circular array of concentrators 42 concentrates the light in the middle region between the angularly spaced lines of the grating in order to minimize the amount of pattern reflected light collected by these concentrators. Will be placed. In the example shown in FIGS. 5 and 6a, such a collector 42 has
Are present, each of which is arranged with an intermediate spacing between adjacent grid lines. However, the device could have only four such concentrators, as described in more detail below with respect to Figures 7a, 7b, and 8b.

The baffle 43 (FIG. 6b) prevents the false laser light from reaching the wafer W. Another baffle 44 (FIG. 6a) between the concentrators 42 is provided to minimize the amount of spurious laser light collected by the concentrators 42.
To limit the field of view to a predetermined area of the wafer.

Each of the collectors 42 has an inlet end 42a.
(FIG. 6b) and an optical fiber with an outlet end 42b. The inlet end 42a is located near the point where the laser light beam is incident on the wafer W to collect the light scattered by the wafer, and the outlet end 42b focuses the light on the photodetector sensor 46. Placed near the lens for. The entrance end 42a of each of the optical fibers is shown in FIG.
Constrained to a curved region of constant shape, as shown in more detail at 47 of a. This end of each area
A pair of side surfaces 47 that converge from a bottom surface 47c that is arranged substantially parallel to the table 2 to which the wafer W to be inspected is attached.
a and 47b. The two side surfaces 47a and 47b converge toward a tip point 47d on the table on which the wafer is attached.

As shown in FIG. 8a, the optical fiber 42 defines a collection area α, and these collection areas are separated by a non-collection area β. In the illustrated embodiment, the width of each collection area α is 16 ° at the bottom surface 47c,
The height φ is 49 °. Such a structure minimizes the pattern reflected light and maximizes the defect reflected light collected by the collector.

Yet another example of collection optics that may be used is shown in Figures 7a, 7b and 8b, respectively, corresponding to Figures 6a, 6b and 8a, respectively. . In this example, we have only four collectors. These concentrators are shown at 42 ', and they are 45 °,
It is placed in angular positions of 135 °, 225 °, and 315 °. This construction is useful when the object to be inspected consists of lines in only two perpendicular directions (0 ° and 90 °). Yet another advantage of this arrangement is that the objective lens 39 'can have a large numerical aperture, thus further reducing the spot size that can be used for scanning. The collection area of this construct is shown at 47 'in Figure 8b. As an example,
The widths α of the collection areas can be 30 ° and their height can be 45 °.

As shown in FIG. 9, the wafer W to be inspected is made of a plurality of integrated circuit dies D ₁ -D _n having the same pattern. In the stage I inspection, the entire surface of the wafer is scanned with the laser light beam 3 and is inspected more carefully for the purpose of detecting defects or for at least the suspicious areas with a high probability of having defects, and thus during the stage II inspection. The scattered light from this scan is collected by the concentrator 42 in order to inspect the suspicious area to be examined. As also shown above, during the Stage I inspection (and also the Stage II inspection), the pattern of one die D, which serves as the pattern to be inspected, has at least one other pattern that serves as the reference pattern. It is compared to the pattern on the die and it is determined that defects may be present in the pattern being inspected.

9-11 show a method of performing a wafer scan in a Stage I inspection.

As shown in FIG. 9, the laser light beam forms a scan line indicated at 50 in FIG.
It is deflected in the x direction by an acoustoelectric deflector 34 (FIG. 4). At the same time, the scanning stage 2 of the table 2 holding the wafer W
1 is the wafer at a constant speed that keeps the wafer in the y direction.
Moving under the spot, which results in the raster scan indicated by 51 in FIG. In the illustrated embodiment,
The scan length of the scan line 50 is 1 mm (1,000 microns). The distance S _y between adjacent lines is 0.6 micron. The distance equal to the sampling distance S _x in the x direction is likewise 0.6 micron. The spot size of the laser light beam, shown at 52, is about 3.0 microns (ie, about 5 sample points are in range).

Therefore, the scanning stage 21 scans the wafer in the y direction between the points a and b, as shown in FIG. As a result, a region with a width w of about 1 mm and a length equal to the distance between point a and point b is covered.

The wafer is then moved in the x direction from point b to point c by the scanning stage 22 (FIG. 3), then the area between points c and d is scanned, and these It is done one after another.

The scanning is performed so that there is an overlap (t, FIG. 10) between adjacent strips scanned by the laser light beam 52. In the example shown in the drawing, this overlap t is 0.2 mm.

In this way, different dies on the same wafer are continuously scanned, and the scattered light generated thereby is collected by the collector 42.
(Or 42 ', FIG. 7a, FIG. 7b, FIG. 8b), such a die, referred to as the die assembled and inspected,
A die-by-die comparison with a die called the reference die is made to give an indication of the possible presence of defects in the die being inspected.

As indicated above, in order to inspect the wafer for defects, the Stage I inspection system uses eight photodetectors 46 (or the modified embodiment of FIGS. 7a, 7b, 8b). Can have four photodetectors). But to get additional information for the registration process,
It is also possible to have another detector (reflected light detector). Therefore, the mismatch between the rectangular pixels in the image being inspected and the rectangular pixel in the reference image is calculated for all possible mismatches to detect the mismatch from the reflected photodetector image. can do. This information can be used if the score matrix calculated by the match control circuit does not give a meaningful indication of the correct mismatch.

Stage I Image Processor The stage I inspection is (a) N streams of a first stream of data representing pixels of different images of the pattern being inspected, where N is the number of collectors 42 or 42 '. ) Occurs,
(B) producing N streams of a second stream of data representing pixels of the image with different reference patterns, and (c)
Comparing the data of the first stream with the data of the second stream to obtain an indication by comparison of suspicious positions of the pattern to be inspected having a high probability of having defects. This comparison corrects any mismatch between the data of the two streams and compares the data of each stream of the first stream with the data of the corresponding stream of the second stream to Obtaining a difference value or an alarm value indicating the importance of the presence of a suspicious pixel therein, and detecting a defect at a pixel location with N difference values or alarm values corresponding to N streams of data; Done by

FIG. 12 is a functional block diagram of the stage I image processing apparatus. The Stage I image processor includes eight sensors 46a-46b (each of which is the photodetector sensor 4 of FIG. 6b).
6) corresponding to their respective preprocessors 6
a-6g with inputs. These sensors convert the optical signals into analog electrical signals, and the preprocessor samples these analog electrical signals at pixel intervals and converts them into digital data. Therefore, the output of the preprocessor is in the form of a stream of pixel values and constitutes an image in digital form.

The preprocessor in each channel has a preamplifier 56, as shown in FIG. Preamplifier 56 converts the current received from each sensor 46 into a voltage and amplifies it to the appropriate magnitude as an input to A / D converter 57. The parameters of the amplification can be controlled according to the characteristics of the signal received from the inspected wafer. A / D converter 57 samples the analog voltage and converts it to a digital value. The sampling of this image is continuous
A two-dimensional image of the object is obtained.

Thus, two streams of eight streams of data occur. One stream represents the pixels of eight different images of the reference pattern previously stored in temporary storage, the other stream obtains an indication of the presence of defects in the pattern to be inspected. In order to represent the pixels of the different images of the inspected pattern to be compared with the image of the reference pattern. Defect detection is performed by the defect detector circuits 60a-60h for each of the eight streams.

The processing device shown in FIG. 12 further includes a second
It has a matching control circuit 62 which controls the register circuit for each of the defect detector circuits 60a-60h. Thus, the register circuits 64a, 64c, 64e and 64g.
Continuously monitors the registration between the reference image and the image to be inspected. These registration circuits produce a score matrix for each selected registration point and output a score matrix (ie, a matrix of values) for each possible shift position near the current registration point. The matching control circuit 62 analyzes the score matrix obtained from the four sensor channels (i.e., every other sensor channel). The values of the matching error signals D _x , D _y at which optimum matching occurs are calculated and a matching control signal is output to the defect detector circuits 60a-60h to correct the mismatch between the two streams of the data stream. .

Defect detector circuits 60a-60h provide their outputs to decision table 66. Judgment table 66
Is based on the alarm values obtained from all eight sensor channels, whether a global defect alarm (ie, a logic output indicating the presence of a defect at a given location) should occur or not, Make a decision. Therefore, decision table 66 receives as input the alarm values from all eight channels and outputs a defect flag.

Each of the eight alarm values has one of three values (0, 1, or 2) indicating no alarm, minor alarm, and major alarm, respectively. The determination table includes (a) at least one alarm value is "2", and (b) at least two adjacent alarm values are "2" or "1" (channel "a" and "g" alarm values). Are adjacent to each other), and only then, is set to output a defect flag “1” indicating the presence of a defect.

The output of the judgment table 66 is sent to the parameter buffer circuit 68. The parameter buffer circuit 68 records the exact coordinates of the pixels in the immediate vicinity of the defect in both the inspected image and the reference image, and parameters describing each defect, such as type and intensity. .
The parameter buffer circuit 68 receives as input an alarm flag trigger (“0” indicates no defects and “1” indicates defects) and all parameters to be recorded. These parameters are received from the temporary storage associated with each of the eight channels. The parameter buffer 68 stores a list of defects associated with those parameters in the post-processing device 14
Output to.

The post-processor 14 receives a list of suspicious defects, together with their associated parameters,
And a decision is made before sending them to the main controller for processing by the stage II image processor. The post-processor outputs a list of suspicious points, including their parameters, to send to the Stage II inspector, and also outputs a list of defects that would not be sent to the Stage II inspector.

FIG. 14 shows in more detail the defect detector (eg 60a) and its associated register 64a in one channel of the image processing apparatus of FIG.

Detecting defects by the defect detector in each channel is based on comparing each pixel in the inspected stream with the corresponding pixel in the corresponding reference stream. These pixels are compared against an adaptive threshold that determines the detection sensitivity according to the pixel type. The type of each pixel is determined by the pixel characteristics, such as signal strength and shape in the 3x3 neighborhood.

Therefore, the digital images from the preprocessors (6a-6h) in each stream are sent to the threshold processor 70 and also the delay buffer 71.
Sent to. Threshold processing device 70 and delay buffer 71
The output of is sent to pixel characterizers 72 and 74. The pixel characterizer 72 is in the register circuit 64a (FIG. 12). The registration circuit 64a is the matching control circuit 62 (FIG. 12).
It outputs a signal to a score calculator 73 (FIG. 14) which controls (with three other streams as described above). The pixel characterizer 74 is used for comparison. The pixel characterizer 74 is connected to the reference die store 75. Reference die store 75 also receives a signal from delay buffer 71. The reference die store 75 outputs a signal to the score calculator 73 and a pixel matcher 76. The pixel matching unit 76 outputs a signal to the comparator 77.

A comparator 77 included in the defect detector 60 for each channel performs a comparison between the inspected image in the vicinity of the current pixel and the reference image in the vicinity of the corresponding pixel. This comparison is carried out with respect to the threshold level depending on the pixel type of the current pixel in the reference image and the image to be examined.

Therefore, the comparator 77 has four inputs, namely (1) the reference pixel input a corresponding to the intensity of the pixel in the reference image and (2) the type of pixel in the reference image. A reference type input b, (3) an inspected pixel input c corresponding to the type of pixel in the inspected image, and (4)
An inspected pixel input d corresponding to the intensity of a pixel in the inspected image. As a result of the comparison performed by the comparator 77, the comparator 77, through its alarm output (e), has three possible outcomes of the comparison:
(A) High threshold is exceeded, (b) Low threshold is exceeded,
(C) An alarm value equal to or less than the threshold value is output. As shown in FIG. 12, the output of comparator 77 in all eight streams is sent to decision table 66.

Thresholding device 70 calculates a threshold for pixel classification as the pixels are scanned. This calculation is
It is based on a histogram of characteristic parameters. Three thresholds for each parameter: (a) a threshold for the decision about the registration point, (b) a threshold for the classification of pixels in the reference image, (c) a threshold for the classification of pixels in the image to be examined, There is.

The threshold processor 70 receives the pixel stream from the object to be scanned through its preprocessor (eg 6a (FIG. 12)) and its threshold level to the pixel characterizers 72 and 74. Output for registration and one for comparison.

The delay buffer 71 delays the processing in each defect detector (eg 60a) and register (eg 64a) until the threshold is calculated. This ensures that the threshold is set by the parameters in the scanned area. Thus, the delay buffer receives the stream of pixel pixels from the object to be scanned through its respective preprocessor and, after appropriate delay, to the two pixel characterizers 72, 74 and to the reference die recorder. To 75.

The pixel characterizer 74 calculates the current pixel type. Thus, during the scan of the reference pattern, the pixel characterizer calculates each type of pixel imaged therein for storage in the reference die store 75, and during the scan of the pattern being examined. , Continuously calculates the new pixel type sent to the comparator 77.

The pixel characterization unit 72 selects a registration point based on the pixel type determined from the result of pixel parameter calculation and the result of comparison with those. Therefore, its input is the examined image from the delay buffer 71 and the thresholds for all pixel parameters from the thresholding device 70. Pixel characterizer 72
Is a score calculator 73 for the points selected as registration points.
The registration point flag is output to.

The score calculator 73 calculates a score matrix of the correlation between the examined image and the reference image at all possible shifts up to the maximum amount allowed around the current pixel. The score calculator 73 has three inputs,
Ie, (a) an image to be inspected to define the region around which the correlation is inspected, (b) a reference image to define the region of possible alignment within the maximum range of horizontal and vertical shifts, (C) a control input from the pixel characterizer 72, which allows selection of registration points based on pixel type,
To receive.

In order to calculate the correct registration, 4 (8
The output of each stream is the matching control circuit 62 (FIG. 12).
Sent to.

The pixel characterizer 74 calculates the current pixel type. Thus, during the scan of the reference pattern, the pixel characterizer calculates each type of pixel that has an image therein for storage in the reference die store 75, and during the scan of the pattern to be examined. , Continuously calculates the new pixel type sent to the comparator 77. The pixel characterizer 74 has two inputs, namely (a) a digital image output from the delay buffer 71, to enable it to make a determination as to the type of pixel.
(B) The threshold value from the threshold value processing device 70 for the related parameter.

The reference die storage device 75 stores the image of the reference parameter. This image contains both the intensities of the pixels and their classification. The reference die store 75 has a pixel input a which receives the density value for each pixel from the delay buffer 71 and a type input b which receives the pixel classification from the pixel characterizer 74. These inputs are only active when the fiducial pattern is scanned, and the fiducial image is regenerated when needed for comparison with the image being examined.
The reference die store 75 has a pixel output b connected to the score calculator 73 and also connected to a pixel matcher 76.
And a type output connected to the pixel matcher 76.

The pixel matcher 76 has a reference die storage device 75.
In order to obtain a match between the pixels output by the and the pixels in the image to be inspected, advance or delay of the pixels output by the reference die store 75 is performed before these pixels reach the comparison stage. To do. Its inputs are the pixel intensity output and type output from the reference die store 75, and also the match control input from the match computer 62 (FIG. 12). Pixel matcher 76 outputs an image pixel stream with advance or delay.

The comparator 77 performs a comparison between the inspected image in the vicinity of the current pixel and the reference image in the vicinity of the corresponding pixel. This comparison is made with respect to a variable threshold level depending on the pixel type of the current pixel of the reference image and the image to be examined. Therefore, its inputs a-d are
It has the pixel intensity and type of the reference image from the pixel matcher 76 and the pixel intensity and type of the inspected image from each of the delay buffer 71 and the pixel characterizer 74.

Therefore, the Stage I post-processor 14 (FIG. 12) receives the list of suspect defects along with their associated parameters and makes a determination before sending them to the Stage II inspection system.

Overall Stage II Device As briefly described above, the Stage II inspection is performed automatically after the completion of the Stage I inspection while the wafer is still on the pedestal 2. Performed only for those locations on the wafer W that were shown to have a high probability of defects in between.
Therefore, although the Stage I inspection is performed at a relatively high rate, the Stage II inspection is performed to indicate whether there is actually a defect at a location that is suspected of having a defect during the Stage I inspection.
The examination is performed at a very slow speed and with high resolution.

Briefly, the Stage II test is performed as follows. That is, a converter 9 such as a CCD
An image of each of the suspicious positions of the pattern to be inspected and the corresponding position of the reference pattern is formed on (FIG. 1 and FIG. 26), and two sets corresponding to each pixel of the inspected pattern and the reference pattern are formed. It is found that a predetermined amount of misalignment exists at each position by comparing the step of outputting the electrical signal of the above and the pixel of the pattern to be inspected with the corresponding pixel of the reference pattern. Indicating a defect whenever
The inspection is performed. Each of the suspicious positions of the pattern to be inspected to accommodate variations in thickness of the wafer and / or pattern, and / or the multilayer pattern,
And the reference pattern is imaged at multiple different depths, and to align the images at each depth,
Shifting one set of electrical signals with respect to another set of electrical signals.

Stage II Optics The Stage II optics are shown generally in FIG. 1 and in detail in FIG. The stage II optics has a microscope objective 100. Microscope objective lens 1
00 is capable of moving the selected objective lens into the optical path between the wafer W and the image converter 9.
It is mounted on a rotating turret 101 with different objective lenses. The wafer W is a flash lamp device 1
02 illuminates through an optical device 103 having a beam splitter 104 and a second beam splitter 105. The device 102 also has a continuous light source, such as a standard tungsten lamp. This continuous light source is used with a standard TV camera and / or viewing device III, as described below.

The beam splitter 104 reflects the infrared portion of the light from the wafer to the autofocus device 106, while the beam splitter 105 reflects the flash light through the selected objective lens 100 and into the vacuum chuck 24. Light is directed to the wafer W above. The beam splitter 105 also transmits the light reflected by the wafer W and advances it through the imaging lens 107, and another beam splitter 108 advances the light to the image converter 9. Beamsplitter 108 reflects a portion of the image and through another beam splitter 109, a standard T
Light is directed to a viewing device 111 having a V-camera 110 and / or a binocular eyepiece. By the binocular observation device 111,
The observer can visually observe the wafer while the TV
The camera 110 can view the wafer through the TV.

Stage II Image Processor FIG. 16 shows a stage II image preprocessor 10 and stage I.
The I image processing apparatus 11 is shown.

The information detected by the image converter 9 is
It is sent to the preamplifier 120 of the preprocessor 10 and then to the digitizer 121 of the image processor 110 and then to the storage buffer 122. The image processing device 11 further includes a digital signal processing device. This digital signal processor is shown in FIG. 16 under software control (block 124) from the main controller (8, FIG. 2).
The following operations are performed as shown in. That is, the matching operation 125, the registration operation 126, the comparison operation 127, and the classification operation 128 are executed. The output from the digital signal processor 123 then returns to the main controller 8.

FIG. 16 further illustrates that the stage II image processing device 11 has a hardware accelerator 129 for accelerating registration and comparison operations, among others.

The above operation will be described in particular detail in the following description with reference to FIGS.

As explained above, the input to the Stage II image processor comprises two sets of images taken from each of the pattern to be inspected and the reference pattern.
Each of these sets has 5 images taken to focus on different depths to accommodate changes in wafer or pattern thickness, or to accommodate multilayer patterns.

As shown in more detail in FIG. 17, a depth matching operation 125 is performed to match the two depth sets to the reference image and the image to be examined, and a registration operation 126 is also performed. Done. In the registration operation, a misalignment between the reference image and the inspected image is detected at each depth.
The list of mismatches is sent to the comparison circuit 127. Circuit 12
Reference numeral 7 compares the density images using the adaptive threshold value obtained from the dynamic range equalization circuit 129 and the surrounding pixels. This equalization circuit compensates for processing, lighting, and other variations. The output of the comparison circuit 127 indicates the suspected defect, location and score. This output is sent to the defect classification circuit 128. The circuit 128 not only outputs the data of the comparison circuit 127, but also the data previously collected.
The data stored in the base 130 is also used to characterize the data defect. The output of the defect classification circuit 128 is sent to the main controller (8, FIG. 1, FIG. 2) for display, printing, or the like.

18-20 show in detail how the depth matching operation is performed. Therefore, the sequence of images taken from the inspected pattern is aligned with the sequence of images taken from the reference pattern. The goal is to align each image of the pattern being inspected with the image of the reference pattern taken at the corresponding depth of focus. Two assumptions are made. That is, (1)
The images are taken in order of increasing depth, the difference between the two subsequent images being constant, (2) the error in the depth of the first image of the two sequences being at most the two subsequent images. Is assumed to be equal to the difference between.

Therefore, if I _i (1 ≦ i ≦ 5) and R _i (1 ≦ i ≦ 5) are the image to be inspected and the reference image, respectively, then x is −1,0,1. 1 in
And i = 1, ..., 5 (I _i , R _{i + x} )
The matching procedure is to detect x such that is a pair of comparable images (see Figure 18). The depth-of-focus correlation of the two images is measured by calculating the similarity of the variances of the density values of the two images. The measure of correlation used is the difference between the intensity value histograms of the image.
The shift x is calculated as giving the best correlation for all images in the sequence.

FIG. 19 shows the matching procedure in more detail. This procedure has the following stages: (1) Calculate density histograms for all images (blocks 131, 132). The density value histogram of the image includes a distribution of density values. Image Histogram H
Contains H (j), in the jth cell, the number of pixels in the image having a density value equal to j. (2) Calculate the distance between the histograms (block 133). The distance is taken as the sum of absolute differences between corresponding cells in the histogram. The distance will be calculated by the following formula.

[Equation 1] Here, H _RK and H _IL are histograms of R _K and I _L , respectively. (3) Create a distance table (block 134). The distance table contains the correlation measure calculated for each pair of images.

[Number 2] _{d (R 1 -I 1) d} (R 1 -I 2) d (R 1 - I 3) ... d (R 2 -I 1) d (R 2 -I 2) d (R 2 - I ₃ ) ... (4) By calculating the average of the three main diagonals (block 135) and choosing the minimum average (block 136) to produce the depth shift, the distance table giving the minimum average is calculated. Find the diagonal line inside (see Figure 20).
The shift x corresponds to the diagonal that gives the minimum mean, thus minimizing the overall distance between the two sets.

Stage II Inspection Improvements FIGS. 21 and 22 show a number of improvements made to the Stage II inspection apparatus. FIG. 21 shows FIG.
However, certain changes described below have been made. FIG. 22 is a diagram to help explain these improvements. For ease of understanding and ease of explanation, only the changes made in FIG. 21 compared to FIG. 15 are described in detail below. Furthermore, globally equivalent elements are given the same reference numbers, but with the addition of "300" and new elements with reference numbers starting with "400".

The main difference between the optical device shown in FIG. 21 and the optical device shown in FIG. 15 is that the optical device shown in FIG. 21 uses a dark field image instead of a bright field image of the object. Is to use. Darkfield images have been found to be more sensitive to small defects than brightfield images. The use of dark field images in Stage II exams is excellent for confirming or dismissing the alarms detected in Stage I exams. Therefore, the probability of detection is higher and the probability of false alarms is lower.

The Stage II optics shown in FIG. 21 has a dark field microscope objective 300 mounted on a rotating turret 301. The turret 301 is further equipped with another objective lens, and by rotating the turret the selected objective lens can be placed in the optical path between the wafer W and the image converter 309. .
The wafer W is illuminated by an illumination device 400 through an optical device 303 having beam splitters 304 and 305. The lighting device 400 can be a standard device using a mercury lamp, such as is commercially available from Leitz. It is a 200 watt mercury lamp 402
, A reflector 404, and a condenser lens 406.

The beam splitter 304 reflects the infrared portion of the light reflected from the wafer W toward the autofocus device 306. The beam splitter 305, on the other hand, reflects the light from the device 400 and selects the selected objective lens 30.
0 through the wafer W on the vacuum chuck 324. The beam splitter 305 also includes the wafer W
Light reflected by is directed through an imaging lens 307 and another beam splitter 308 toward an image converter 309. The beam splitter 308 reflects a part of the image and the observation device 31 equipped with a binocular eyepiece.
Proceed toward 1. The observer can visually observe the wafer with this observation device. Image converter 309
Can be a CCD camera with exposure control, such as the Parnix TM64.

FIG. 21 further has a dark field shutter 408. The dark field shutter 408 is an optical device that produces a dark field image by blocking the central region of the illuminating beam IB. The optical device shown in FIG.
It has a D filter 410 and a color filter 412. The ND filter 410 is used to adjust the illumination intensity above the object, and the color filter 412 is used to enhance the contrast of the image.

FIG. 22 illustrates the creation of multiple depth images at one location. In the illustrated embodiment, there are three such depth images, but in practice any number of depth images can be produced, as will be explained below.

As a result of the stage I inspection, image formation of a position determined to have a defect with a high probability is performed as follows. That is, first, the xy stage (22, FIG. 3) is set so that the position having a high possibility of the defect detected by the stage I inspection is below the stage II objective lens 300 (FIG. 21).
Thus, the wafer is moved. The autofocus device 306
The lens is focused to a predetermined depth with respect to the surface of the object by moving the rotation / level / focus stage 323 in the correct z direction.

The rotation / level / focus stage is accelerated to a constant predetermined velocity equal to the separation distance (h) between depth images divided by the time between frames. When passing the set distance, three (or any number) of images are recorded at equal intervals.

The separation distance (h) between depth images is approximately equal to the depth of focus. This creates an image of the defect focused on the location of at least one depth image.

Another feature of the image production technique shown in FIG. 22 is that the exposure time used for each image is significantly shorter than the frame time. This prevents blurring of the image due to continuous movement of the stage 323 in the z direction at the time the image is recorded. As one example, the frame time can be about 16 milliseconds,
On the other hand, the exposure time can be 0.5 ms. This short exposure time can be obtained by the built-in exposure control of the CCD camera 309.

In the above apparatus, the reference pattern is yet another similar object pattern (eg, die-to-die comparison), but the method uses another similar pattern on the same object with the same reference pattern. Either (repeating pattern comparison) or data stored in the database (compare die to database)
Can be executed.

[Brief description of drawings]

FIG. 1 is a schematic diagram of one type of device constructed in accordance with the present invention.

2 is a block diagram of the apparatus of FIG.

FIG. 3 is a schematic diagram of a wafer handling and image acquisition apparatus in the apparatus of FIGS. 1 and 2.

4 is a schematic diagram of an optical device in a first inspection stage of the device of FIG.

5 is a plan view showing the arrangement of light collectors in the optical device of FIG. 4. FIG.

6 a is a plan view showing the arrangement of the collectors of FIG. 5 in more detail, and b is a view of one collector in the arrangement of FIG. 6 a.

7a is a view of the modified embodiment of FIG. 6a, b is a view of one collector in the arrangement of FIG. 7a.

FIG. 8 is a more detailed view of the light collection region in each of the structures of FIGS. 6a and 7a.

FIG. 9 is a diagram showing a wafer scanning method in the stage I inspection.

FIG. 10 is a diagram showing a wafer scanning method in the stage I inspection.

FIG. 11 is a diagram showing a wafer scanning method in the stage I inspection.

FIG. 12 is a block diagram illustrating a Stage I processor.

FIG. 13 is a block diagram of the major components of the one-channel pre-processor of the processor of FIG.

FIG. 14 is a block diagram showing one channel of the processor of FIG. 12 after the preprocessor.

FIG. 15 shows the main elements of a Stage II optical device.

FIG. 16 is a diagram showing the configuration and operation of a stage II inspection device.

FIG. 17 is a diagram showing the configuration and operation of a stage II inspection device.

FIG. 18 is a diagram showing the configuration and operation of a stage II inspection device.

FIG. 19 is a diagram showing the configuration and operation of a stage II inspection device.

FIG. 20 is a diagram showing the configuration and operation of a stage II inspection device.

FIG. 21 is a view corresponding to FIG. 15 but with changes made to the optical device in the stage II inspection.

FIG. 22 is an illustration of a modified example of a Stage II inspection.

[Explanation of symbols]

 3 Laser 4 Concentrator 5 Detector 6 Stage I Preprocessor 7 Stage I Image Processor 8 Main Controller 9 Photoelectric Converter 10 Stage II Preprocessor 11 Stage II Image Processor 14 Postprocessor

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Ribka Sherman Ramat Hashar, Nahal Solek Street, Israel 10 (72) Inventor Ehud Tirosh Mosheskor Street, Jerusalem, Israel 1

Claims (11)

[Claims] 1. A step of optically inspecting the entire surface of an object in a first step, the step of electrically outputting information indicating a position on the object suspected of having a defect and the suspicious position being stored. Storing in the device, second
Optically inspecting only the suspicious location on the surface of the object with high resolution to determine whether a defect is present or absent in the suspicious location in step 1; The first stage inspection is performed by scanning the entire surface of the object with a light beam of a relatively small diameter and at a relatively high velocity so that it can be performed with a high resolution similar to the second stage inspection. A method for inspecting defects on the surface of an object, characterized by: 2. The inspection method according to claim 1, wherein
The first inspection step is performed by optically scanning the entire surface of the object with a laser beam, and the second
An inspection step is carried out immediately after the first inspection step by automatically image-forming only the suspicious positions on the transducer, and the transducer converting the image into an electrical signal and then the electrical signal. The above-mentioned inspection method for analyzing. 3. The inspection method according to claim 1, wherein the surface of the object has a pattern to be inspected, and the first inspecting step includes the inspected pattern and a reference pattern. The step of comparing with another pattern which plays a role of and a position which shows that there is a large difference between the reference pattern as a result of the comparison and there is a high probability that a defect exists in the inspected pattern. The inspection method performed by the steps of: 4. The inspection method according to claim 3,
The second inspecting step also includes comparing the inspected pattern and the reference pattern, and the comparison results in a large difference with respect to the reference pattern, which results in the suspicious position of the inspected pattern. Identifying a location on the inspected pattern that indicates the presence of a defect. 5. Optically scanning the surface of the object with a light beam to produce a first stream of N streams of data representing pixels of different images of the object to be examined, and corresponding reference images. Producing a second stream of N streams of data representative of pixels of the first stream of data and the first stream of data to obtain an indication of a suspicious location of the inspected object having a high probability of being defective. A method of inspecting for defects on the surface of an object, comprising the steps of: comparing two flow data. 6. The inspection method according to claim 5, wherein the step of correcting a mismatch between the data of the two streams and an alarm indicating the significance of the presence of a suspicious position in the stream. Comparing the data of each stream of the first stream with the data of the corresponding stream of the second stream to obtain the pixel position by N alarms corresponding to the N streams of data. The step of detecting defects, wherein the comparing operation is performed. 7. The inspection method according to claim 6, wherein a corresponding registration point is selected in each stream of streams, and a mismatch between the registration points of the two streams is detected. And the step of shifting one stream with respect to another stream to correct the mismatch between the two streams, wherein the correction of the mismatch is performed. 8. An image of each position of the inspected object at a plurality of different depths and an image of the corresponding position of the reference object are created on the opto-electrical converter and are inspected at that depth. Outputting two sets of electrical signals for each depth corresponding to the pixels of the object and the pixels of the reference object, and one set of said electrical signals for the other to match the images of each depth. A step of shifting with respect to said set of electrical signals; a predetermined magnitude of mismatch of the object being inspected to indicate a defect if found to be present at each position of the object being inspected. A method of inspecting defects on the surface of an object, comprising the steps of: comparing the pixel with a corresponding pixel of a reference object. 9. The inspection method according to claim 8, wherein:
The inspection method, wherein the imaging step at a plurality of different depths is performed by moving an object to be inspected and a reference object in a direction toward and away from the transducer. 10. The inspection method according to claim 8 or 9, wherein a lamp emits a flash of light at periodic time intervals during the imaging of a plurality of different depths, while the inspection is performed. The inspection method, wherein an object to be subjected and a reference object are moved with respect to the optoelectric converter. 11. Claim 2-Claim 4 or Claim 8-
The inspection method according to claim 10, wherein the image forming step is performed by a dark-field image forming apparatus.